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
26 This is the lower layer of the Perl parser, managing characters and tokens.
28 =for apidoc AmU|yy_parser *|PL_parser
30 Pointer to a structure encapsulating the state of the parsing operation
31 currently in progress. The pointer can be locally changed to perform
32 a nested parse without interfering with the state of an outer parse.
33 Individual members of C<PL_parser> have their own documentation.
39 #define PERL_IN_TOKE_C
41 #include "dquote_static.c"
43 #define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46 #define pl_yylval (PL_parser->yylval)
48 /* XXX temporary backwards compatibility */
49 #define PL_lex_brackets (PL_parser->lex_brackets)
50 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
51 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
52 #define PL_lex_brackstack (PL_parser->lex_brackstack)
53 #define PL_lex_casemods (PL_parser->lex_casemods)
54 #define PL_lex_casestack (PL_parser->lex_casestack)
55 #define PL_lex_defer (PL_parser->lex_defer)
56 #define PL_lex_dojoin (PL_parser->lex_dojoin)
57 #define PL_lex_formbrack (PL_parser->lex_formbrack)
58 #define PL_lex_inpat (PL_parser->lex_inpat)
59 #define PL_lex_inwhat (PL_parser->lex_inwhat)
60 #define PL_lex_op (PL_parser->lex_op)
61 #define PL_lex_repl (PL_parser->lex_repl)
62 #define PL_lex_starts (PL_parser->lex_starts)
63 #define PL_lex_stuff (PL_parser->lex_stuff)
64 #define PL_multi_start (PL_parser->multi_start)
65 #define PL_multi_open (PL_parser->multi_open)
66 #define PL_multi_close (PL_parser->multi_close)
67 #define PL_preambled (PL_parser->preambled)
68 #define PL_sublex_info (PL_parser->sublex_info)
69 #define PL_linestr (PL_parser->linestr)
70 #define PL_expect (PL_parser->expect)
71 #define PL_copline (PL_parser->copline)
72 #define PL_bufptr (PL_parser->bufptr)
73 #define PL_oldbufptr (PL_parser->oldbufptr)
74 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
75 #define PL_linestart (PL_parser->linestart)
76 #define PL_bufend (PL_parser->bufend)
77 #define PL_last_uni (PL_parser->last_uni)
78 #define PL_last_lop (PL_parser->last_lop)
79 #define PL_last_lop_op (PL_parser->last_lop_op)
80 #define PL_lex_state (PL_parser->lex_state)
81 #define PL_rsfp (PL_parser->rsfp)
82 #define PL_rsfp_filters (PL_parser->rsfp_filters)
83 #define PL_in_my (PL_parser->in_my)
84 #define PL_in_my_stash (PL_parser->in_my_stash)
85 #define PL_tokenbuf (PL_parser->tokenbuf)
86 #define PL_multi_end (PL_parser->multi_end)
87 #define PL_error_count (PL_parser->error_count)
89 # define PL_nexttoke (PL_parser->nexttoke)
90 # define PL_nexttype (PL_parser->nexttype)
91 # define PL_nextval (PL_parser->nextval)
93 static const char* const ident_too_long = "Identifier too long";
95 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
97 #define XENUMMASK 0x3f
99 #define XFAKEBRACK 0x80
101 #ifdef USE_UTF8_SCRIPTS
102 # define UTF (!IN_BYTES)
104 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
107 /* The maximum number of characters preceding the unrecognized one to display */
108 #define UNRECOGNIZED_PRECEDE_COUNT 10
110 /* In variables named $^X, these are the legal values for X.
111 * 1999-02-27 mjd-perl-patch@plover.com */
112 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
114 #define SPACE_OR_TAB(c) isBLANK_A(c)
116 #define HEXFP_PEEK(s) \
118 (isXDIGIT(s[1]) || isALPHA_FOLD_EQ(s[1], 'p'))) || \
119 isALPHA_FOLD_EQ(s[0], 'p'))
121 /* LEX_* are values for PL_lex_state, the state of the lexer.
122 * They are arranged oddly so that the guard on the switch statement
123 * can get by with a single comparison (if the compiler is smart enough).
125 * These values refer to the various states within a sublex parse,
126 * i.e. within a double quotish string
129 /* #define LEX_NOTPARSING 11 is done in perl.h. */
131 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
132 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
133 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
134 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
135 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
137 /* at end of code, eg "$x" followed by: */
138 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
139 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
141 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
142 string or after \E, $foo, etc */
143 #define LEX_INTERPCONST 2 /* NOT USED */
144 #define LEX_FORMLINE 1 /* expecting a format line */
145 #define LEX_KNOWNEXT 0 /* next token known; just return it */
149 static const char* const lex_state_names[] = {
164 #include "keywords.h"
166 /* CLINE is a macro that ensures PL_copline has a sane value */
168 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
170 # define SKIPSPACE0(s) skipspace(s)
171 # define SKIPSPACE1(s) skipspace(s)
172 # define SKIPSPACE2(s,tsv) skipspace(s)
173 # define PEEKSPACE(s) skipspace(s)
176 * Convenience functions to return different tokens and prime the
177 * lexer for the next token. They all take an argument.
179 * TOKEN : generic token (used for '(', DOLSHARP, etc)
180 * OPERATOR : generic operator
181 * AOPERATOR : assignment operator
182 * PREBLOCK : beginning the block after an if, while, foreach, ...
183 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
184 * PREREF : *EXPR where EXPR is not a simple identifier
185 * TERM : expression term
186 * POSTDEREF : postfix dereference (->$* ->@[...] etc.)
187 * LOOPX : loop exiting command (goto, last, dump, etc)
188 * FTST : file test operator
189 * FUN0 : zero-argument function
190 * FUN0OP : zero-argument function, with its op created in this file
191 * FUN1 : not used, except for not, which isn't a UNIOP
192 * BOop : bitwise or or xor
194 * SHop : shift operator
195 * PWop : power operator
196 * PMop : pattern-matching operator
197 * Aop : addition-level operator
198 * AopNOASSIGN : addition-level operator that is never part of .=
199 * Mop : multiplication-level operator
200 * Eop : equality-testing operator
201 * Rop : relational operator <= != gt
203 * Also see LOP and lop() below.
206 #ifdef DEBUGGING /* Serve -DT. */
207 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
209 # define REPORT(retval) (retval)
212 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
213 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
214 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
215 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
216 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
217 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
218 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
219 #define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
220 #define LOOPX(f) return (PL_bufptr = force_word(s,WORD,TRUE,FALSE), \
222 PL_expect = PL_nexttoke ? XOPERATOR : XTERM, \
224 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
225 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
226 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
227 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
228 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
229 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
230 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
231 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
232 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
233 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
234 #define AopNOASSIGN(f) return (pl_yylval.ival=f, PL_bufptr=s, REPORT((int)ADDOP))
235 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
236 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
237 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
239 /* This bit of chicanery makes a unary function followed by
240 * a parenthesis into a function with one argument, highest precedence.
241 * The UNIDOR macro is for unary functions that can be followed by the //
242 * operator (such as C<shift // 0>).
244 #define UNI3(f,x,have_x) { \
245 pl_yylval.ival = f; \
246 if (have_x) PL_expect = x; \
248 PL_last_uni = PL_oldbufptr; \
249 PL_last_lop_op = f; \
251 return REPORT( (int)FUNC1 ); \
253 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
255 #define UNI(f) UNI3(f,XTERM,1)
256 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
257 #define UNIPROTO(f,optional) { \
258 if (optional) PL_last_uni = PL_oldbufptr; \
262 #define UNIBRACK(f) UNI3(f,0,0)
264 /* grandfather return to old style */
267 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
268 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
269 pl_yylval.ival = (f); \
275 #define COPLINE_INC_WITH_HERELINES \
277 CopLINE_inc(PL_curcop); \
278 if (PL_parser->herelines) \
279 CopLINE(PL_curcop) += PL_parser->herelines, \
280 PL_parser->herelines = 0; \
282 /* Called after scan_str to update CopLINE(PL_curcop), but only when there
283 * is no sublex_push to follow. */
284 #define COPLINE_SET_FROM_MULTI_END \
286 CopLINE_set(PL_curcop, PL_multi_end); \
287 if (PL_multi_end != PL_multi_start) \
288 PL_parser->herelines = 0; \
294 /* how to interpret the pl_yylval associated with the token */
298 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
303 static struct debug_tokens {
305 enum token_type type;
307 } const debug_tokens[] =
309 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
310 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
311 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
312 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
313 { ARROW, TOKENTYPE_NONE, "ARROW" },
314 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
315 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
316 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
317 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
318 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
319 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
320 { DO, TOKENTYPE_NONE, "DO" },
321 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
322 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
323 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
324 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
325 { ELSE, TOKENTYPE_NONE, "ELSE" },
326 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
327 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
328 { FOR, TOKENTYPE_IVAL, "FOR" },
329 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
330 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
331 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
332 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
333 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
334 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
335 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
336 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
337 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
338 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
339 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
340 { IF, TOKENTYPE_IVAL, "IF" },
341 { LABEL, TOKENTYPE_PVAL, "LABEL" },
342 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
343 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
344 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
345 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
346 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
347 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
348 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
349 { MY, TOKENTYPE_IVAL, "MY" },
350 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
351 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
352 { OROP, TOKENTYPE_IVAL, "OROP" },
353 { OROR, TOKENTYPE_NONE, "OROR" },
354 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
355 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
356 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
357 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
358 { POSTJOIN, TOKENTYPE_NONE, "POSTJOIN" },
359 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
360 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
361 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
362 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
363 { PREINC, TOKENTYPE_NONE, "PREINC" },
364 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
365 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
366 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
367 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
368 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
369 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
370 { SUB, TOKENTYPE_NONE, "SUB" },
371 { THING, TOKENTYPE_OPVAL, "THING" },
372 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
373 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
374 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
375 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
376 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
377 { USE, TOKENTYPE_IVAL, "USE" },
378 { WHEN, TOKENTYPE_IVAL, "WHEN" },
379 { WHILE, TOKENTYPE_IVAL, "WHILE" },
380 { WORD, TOKENTYPE_OPVAL, "WORD" },
381 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
382 { 0, TOKENTYPE_NONE, NULL }
385 /* dump the returned token in rv, plus any optional arg in pl_yylval */
388 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
390 PERL_ARGS_ASSERT_TOKEREPORT;
393 const char *name = NULL;
394 enum token_type type = TOKENTYPE_NONE;
395 const struct debug_tokens *p;
396 SV* const report = newSVpvs("<== ");
398 for (p = debug_tokens; p->token; p++) {
399 if (p->token == (int)rv) {
406 Perl_sv_catpv(aTHX_ report, name);
407 else if ((char)rv > ' ' && (char)rv <= '~')
409 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
411 sv_catpvs(report, " (pending identifier)");
414 sv_catpvs(report, "EOF");
416 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
421 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
423 case TOKENTYPE_OPNUM:
424 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
425 PL_op_name[lvalp->ival]);
428 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
430 case TOKENTYPE_OPVAL:
432 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
433 PL_op_name[lvalp->opval->op_type]);
434 if (lvalp->opval->op_type == OP_CONST) {
435 Perl_sv_catpvf(aTHX_ report, " %s",
436 SvPEEK(cSVOPx_sv(lvalp->opval)));
441 sv_catpvs(report, "(opval=null)");
444 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
450 /* print the buffer with suitable escapes */
453 S_printbuf(pTHX_ const char *const fmt, const char *const s)
455 SV* const tmp = newSVpvs("");
457 PERL_ARGS_ASSERT_PRINTBUF;
459 GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
460 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
468 S_deprecate_commaless_var_list(pTHX) {
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
477 * This subroutine looks for an '=' next to the operator that has just been
478 * parsed and turns it into an ASSIGNOP if it finds one.
482 S_ao(pTHX_ int toketype)
484 if (*PL_bufptr == '=') {
486 if (toketype == ANDAND)
487 pl_yylval.ival = OP_ANDASSIGN;
488 else if (toketype == OROR)
489 pl_yylval.ival = OP_ORASSIGN;
490 else if (toketype == DORDOR)
491 pl_yylval.ival = OP_DORASSIGN;
499 * When Perl expects an operator and finds something else, no_op
500 * prints the warning. It always prints "<something> found where
501 * operator expected. It prints "Missing semicolon on previous line?"
502 * if the surprise occurs at the start of the line. "do you need to
503 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
504 * where the compiler doesn't know if foo is a method call or a function.
505 * It prints "Missing operator before end of line" if there's nothing
506 * after the missing operator, or "... before <...>" if there is something
507 * after the missing operator.
511 S_no_op(pTHX_ const char *const what, char *s)
513 char * const oldbp = PL_bufptr;
514 const bool is_first = (PL_oldbufptr == PL_linestart);
516 PERL_ARGS_ASSERT_NO_OP;
522 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
523 if (ckWARN_d(WARN_SYNTAX)) {
525 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
526 "\t(Missing semicolon on previous line?)\n");
527 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
529 for (t = PL_oldoldbufptr; (isWORDCHAR_lazy_if(t,UTF) || *t == ':');
530 t += UTF ? UTF8SKIP(t) : 1)
532 if (t < PL_bufptr && isSPACE(*t))
533 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
534 "\t(Do you need to predeclare %"UTF8f"?)\n",
535 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
539 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
540 "\t(Missing operator before %"UTF8f"?)\n",
541 UTF8fARG(UTF, s - oldbp, oldbp));
549 * Complain about missing quote/regexp/heredoc terminator.
550 * If it's called with NULL then it cauterizes the line buffer.
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
557 S_missingterm(pTHX_ char *s)
562 char * const nl = strrchr(s,'\n');
566 else if ((U8) PL_multi_close < 32) {
568 tmpbuf[1] = (char)toCTRL(PL_multi_close);
573 *tmpbuf = (char)PL_multi_close;
577 q = strchr(s,'"') ? '\'' : '"';
578 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
584 * Check whether the named feature is enabled.
587 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
589 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
591 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
593 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
595 if (namelen > MAX_FEATURE_LEN)
597 memcpy(&he_name[8], name, namelen);
599 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
600 REFCOUNTED_HE_EXISTS));
604 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
605 * utf16-to-utf8-reversed.
608 #ifdef PERL_CR_FILTER
612 const char *s = SvPVX_const(sv);
613 const char * const e = s + SvCUR(sv);
615 PERL_ARGS_ASSERT_STRIP_RETURN;
617 /* outer loop optimized to do nothing if there are no CR-LFs */
619 if (*s++ == '\r' && *s == '\n') {
620 /* hit a CR-LF, need to copy the rest */
624 if (*s == '\r' && s[1] == '\n')
635 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
637 const I32 count = FILTER_READ(idx+1, sv, maxlen);
638 if (count > 0 && !maxlen)
645 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
647 Creates and initialises a new lexer/parser state object, supplying
648 a context in which to lex and parse from a new source of Perl code.
649 A pointer to the new state object is placed in L</PL_parser>. An entry
650 is made on the save stack so that upon unwinding the new state object
651 will be destroyed and the former value of L</PL_parser> will be restored.
652 Nothing else need be done to clean up the parsing context.
654 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
655 non-null, provides a string (in SV form) containing code to be parsed.
656 A copy of the string is made, so subsequent modification of I<line>
657 does not affect parsing. I<rsfp>, if non-null, provides an input stream
658 from which code will be read to be parsed. If both are non-null, the
659 code in I<line> comes first and must consist of complete lines of input,
660 and I<rsfp> supplies the remainder of the source.
662 The I<flags> parameter is reserved for future use. Currently it is only
663 used by perl internally, so extensions should always pass zero.
668 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
669 can share filters with the current parser.
670 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
671 caller, hence isn't owned by the parser, so shouldn't be closed on parser
672 destruction. This is used to handle the case of defaulting to reading the
673 script from the standard input because no filename was given on the command
674 line (without getting confused by situation where STDIN has been closed, so
675 the script handle is opened on fd 0) */
678 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
680 const char *s = NULL;
681 yy_parser *parser, *oparser;
682 if (flags && flags & ~LEX_START_FLAGS)
683 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
685 /* create and initialise a parser */
687 Newxz(parser, 1, yy_parser);
688 parser->old_parser = oparser = PL_parser;
691 parser->stack = NULL;
693 parser->stack_size = 0;
695 /* on scope exit, free this parser and restore any outer one */
697 parser->saved_curcop = PL_curcop;
699 /* initialise lexer state */
701 parser->nexttoke = 0;
702 parser->error_count = oparser ? oparser->error_count : 0;
703 parser->copline = parser->preambling = NOLINE;
704 parser->lex_state = LEX_NORMAL;
705 parser->expect = XSTATE;
707 parser->rsfp_filters =
708 !(flags & LEX_START_SAME_FILTER) || !oparser
710 : MUTABLE_AV(SvREFCNT_inc(
711 oparser->rsfp_filters
712 ? oparser->rsfp_filters
713 : (oparser->rsfp_filters = newAV())
716 Newx(parser->lex_brackstack, 120, char);
717 Newx(parser->lex_casestack, 12, char);
718 *parser->lex_casestack = '\0';
719 Newxz(parser->lex_shared, 1, LEXSHARED);
723 s = SvPV_const(line, len);
724 parser->linestr = flags & LEX_START_COPIED
725 ? SvREFCNT_inc_simple_NN(line)
726 : newSVpvn_flags(s, len, SvUTF8(line));
727 sv_catpvn(parser->linestr, "\n;", rsfp ? 1 : 2);
729 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
731 parser->oldoldbufptr =
734 parser->linestart = SvPVX(parser->linestr);
735 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
736 parser->last_lop = parser->last_uni = NULL;
738 assert(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
739 |LEX_DONT_CLOSE_RSFP));
740 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
741 |LEX_DONT_CLOSE_RSFP));
743 parser->in_pod = parser->filtered = 0;
747 /* delete a parser object */
750 Perl_parser_free(pTHX_ const yy_parser *parser)
752 PERL_ARGS_ASSERT_PARSER_FREE;
754 PL_curcop = parser->saved_curcop;
755 SvREFCNT_dec(parser->linestr);
757 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
758 PerlIO_clearerr(parser->rsfp);
759 else if (parser->rsfp && (!parser->old_parser ||
760 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
761 PerlIO_close(parser->rsfp);
762 SvREFCNT_dec(parser->rsfp_filters);
763 SvREFCNT_dec(parser->lex_stuff);
764 SvREFCNT_dec(parser->sublex_info.repl);
766 Safefree(parser->lex_brackstack);
767 Safefree(parser->lex_casestack);
768 Safefree(parser->lex_shared);
769 PL_parser = parser->old_parser;
774 Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
776 I32 nexttoke = parser->nexttoke;
777 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
779 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
780 && parser->nextval[nexttoke].opval
781 && parser->nextval[nexttoke].opval->op_slabbed
782 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
783 op_free(parser->nextval[nexttoke].opval);
784 parser->nextval[nexttoke].opval = NULL;
791 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
793 Buffer scalar containing the chunk currently under consideration of the
794 text currently being lexed. This is always a plain string scalar (for
795 which C<SvPOK> is true). It is not intended to be used as a scalar by
796 normal scalar means; instead refer to the buffer directly by the pointer
797 variables described below.
799 The lexer maintains various C<char*> pointers to things in the
800 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
801 reallocated, all of these pointers must be updated. Don't attempt to
802 do this manually, but rather use L</lex_grow_linestr> if you need to
803 reallocate the buffer.
805 The content of the text chunk in the buffer is commonly exactly one
806 complete line of input, up to and including a newline terminator,
807 but there are situations where it is otherwise. The octets of the
808 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
809 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
810 flag on this scalar, which may disagree with it.
812 For direct examination of the buffer, the variable
813 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
814 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
815 of these pointers is usually preferable to examination of the scalar
816 through normal scalar means.
818 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
820 Direct pointer to the end of the chunk of text currently being lexed, the
821 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
822 + SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
823 always located at the end of the buffer, and does not count as part of
824 the buffer's contents.
826 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
828 Points to the current position of lexing inside the lexer buffer.
829 Characters around this point may be freely examined, within
830 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
831 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
832 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
834 Lexing code (whether in the Perl core or not) moves this pointer past
835 the characters that it consumes. It is also expected to perform some
836 bookkeeping whenever a newline character is consumed. This movement
837 can be more conveniently performed by the function L</lex_read_to>,
838 which handles newlines appropriately.
840 Interpretation of the buffer's octets can be abstracted out by
841 using the slightly higher-level functions L</lex_peek_unichar> and
842 L</lex_read_unichar>.
844 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
846 Points to the start of the current line inside the lexer buffer.
847 This is useful for indicating at which column an error occurred, and
848 not much else. This must be updated by any lexing code that consumes
849 a newline; the function L</lex_read_to> handles this detail.
855 =for apidoc Amx|bool|lex_bufutf8
857 Indicates whether the octets in the lexer buffer
858 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
859 of Unicode characters. If not, they should be interpreted as Latin-1
860 characters. This is analogous to the C<SvUTF8> flag for scalars.
862 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
863 contains valid UTF-8. Lexing code must be robust in the face of invalid
866 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
867 is significant, but not the whole story regarding the input character
868 encoding. Normally, when a file is being read, the scalar contains octets
869 and its C<SvUTF8> flag is off, but the octets should be interpreted as
870 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
871 however, the scalar may have the C<SvUTF8> flag on, and in this case its
872 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
873 is in effect. This logic may change in the future; use this function
874 instead of implementing the logic yourself.
880 Perl_lex_bufutf8(pTHX)
886 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
888 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
889 at least I<len> octets (including terminating C<NUL>). Returns a
890 pointer to the reallocated buffer. This is necessary before making
891 any direct modification of the buffer that would increase its length.
892 L</lex_stuff_pvn> provides a more convenient way to insert text into
895 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
896 this function updates all of the lexer's variables that point directly
903 Perl_lex_grow_linestr(pTHX_ STRLEN len)
907 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
908 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
909 linestr = PL_parser->linestr;
910 buf = SvPVX(linestr);
911 if (len <= SvLEN(linestr))
913 bufend_pos = PL_parser->bufend - buf;
914 bufptr_pos = PL_parser->bufptr - buf;
915 oldbufptr_pos = PL_parser->oldbufptr - buf;
916 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
917 linestart_pos = PL_parser->linestart - buf;
918 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
919 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
920 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
921 PL_parser->lex_shared->re_eval_start - buf : 0;
923 buf = sv_grow(linestr, len);
925 PL_parser->bufend = buf + bufend_pos;
926 PL_parser->bufptr = buf + bufptr_pos;
927 PL_parser->oldbufptr = buf + oldbufptr_pos;
928 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
929 PL_parser->linestart = buf + linestart_pos;
930 if (PL_parser->last_uni)
931 PL_parser->last_uni = buf + last_uni_pos;
932 if (PL_parser->last_lop)
933 PL_parser->last_lop = buf + last_lop_pos;
934 if (PL_parser->lex_shared->re_eval_start)
935 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
940 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
942 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
943 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
944 reallocating the buffer if necessary. This means that lexing code that
945 runs later will see the characters as if they had appeared in the input.
946 It is not recommended to do this as part of normal parsing, and most
947 uses of this facility run the risk of the inserted characters being
948 interpreted in an unintended manner.
950 The string to be inserted is represented by I<len> octets starting
951 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
952 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
953 The characters are recoded for the lexer buffer, according to how the
954 buffer is currently being interpreted (L</lex_bufutf8>). If a string
955 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
956 function is more convenient.
962 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
966 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
967 if (flags & ~(LEX_STUFF_UTF8))
968 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
970 if (flags & LEX_STUFF_UTF8) {
973 STRLEN highhalf = 0; /* Count of variants */
974 const char *p, *e = pv+len;
975 for (p = pv; p != e; p++) {
976 if (! UTF8_IS_INVARIANT(*p)) {
982 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
983 bufptr = PL_parser->bufptr;
984 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
985 SvCUR_set(PL_parser->linestr,
986 SvCUR(PL_parser->linestr) + len+highhalf);
987 PL_parser->bufend += len+highhalf;
988 for (p = pv; p != e; p++) {
990 if (! UTF8_IS_INVARIANT(c)) {
991 *bufptr++ = UTF8_TWO_BYTE_HI(c);
992 *bufptr++ = UTF8_TWO_BYTE_LO(c);
999 if (flags & LEX_STUFF_UTF8) {
1000 STRLEN highhalf = 0;
1001 const char *p, *e = pv+len;
1002 for (p = pv; p != e; p++) {
1004 if (UTF8_IS_ABOVE_LATIN1(c)) {
1005 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1006 "non-Latin-1 character into Latin-1 input");
1007 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
1010 } else if (! UTF8_IS_INVARIANT(c)) {
1011 /* malformed UTF-8 */
1013 SAVESPTR(PL_warnhook);
1014 PL_warnhook = PERL_WARNHOOK_FATAL;
1015 utf8n_to_uvchr((U8*)p, e-p, NULL, 0);
1021 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1022 bufptr = PL_parser->bufptr;
1023 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1024 SvCUR_set(PL_parser->linestr,
1025 SvCUR(PL_parser->linestr) + len-highhalf);
1026 PL_parser->bufend += len-highhalf;
1029 if (UTF8_IS_INVARIANT(*p)) {
1035 *bufptr++ = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
1041 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1042 bufptr = PL_parser->bufptr;
1043 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1044 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1045 PL_parser->bufend += len;
1046 Copy(pv, bufptr, len, char);
1052 =for apidoc Amx|void|lex_stuff_pv|const char *pv|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 represented by octets starting at I<pv>
1063 and continuing to the first nul. These octets are interpreted as either
1064 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1065 in I<flags>. The characters are recoded for the lexer buffer, according
1066 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1067 If it is not convenient to nul-terminate a string to be inserted, the
1068 L</lex_stuff_pvn> function is more appropriate.
1074 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1076 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1077 lex_stuff_pvn(pv, strlen(pv), flags);
1081 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1083 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1084 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1085 reallocating the buffer if necessary. This means that lexing code that
1086 runs later will see the characters as if they had appeared in the input.
1087 It is not recommended to do this as part of normal parsing, and most
1088 uses of this facility run the risk of the inserted characters being
1089 interpreted in an unintended manner.
1091 The string to be inserted is the string value of I<sv>. The characters
1092 are recoded for the lexer buffer, according to how the buffer is currently
1093 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1094 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1095 need to construct a scalar.
1101 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1105 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1107 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1109 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1113 =for apidoc Amx|void|lex_unstuff|char *ptr
1115 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1116 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1117 This hides the discarded text from any lexing code that runs later,
1118 as if the text had never appeared.
1120 This is not the normal way to consume lexed text. For that, use
1127 Perl_lex_unstuff(pTHX_ char *ptr)
1131 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1132 buf = PL_parser->bufptr;
1134 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1137 bufend = PL_parser->bufend;
1139 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1140 unstuff_len = ptr - buf;
1141 Move(ptr, buf, bufend+1-ptr, char);
1142 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1143 PL_parser->bufend = bufend - unstuff_len;
1147 =for apidoc Amx|void|lex_read_to|char *ptr
1149 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1150 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1151 performing the correct bookkeeping whenever a newline character is passed.
1152 This is the normal way to consume lexed text.
1154 Interpretation of the buffer's octets can be abstracted out by
1155 using the slightly higher-level functions L</lex_peek_unichar> and
1156 L</lex_read_unichar>.
1162 Perl_lex_read_to(pTHX_ char *ptr)
1165 PERL_ARGS_ASSERT_LEX_READ_TO;
1166 s = PL_parser->bufptr;
1167 if (ptr < s || ptr > PL_parser->bufend)
1168 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1169 for (; s != ptr; s++)
1171 COPLINE_INC_WITH_HERELINES;
1172 PL_parser->linestart = s+1;
1174 PL_parser->bufptr = ptr;
1178 =for apidoc Amx|void|lex_discard_to|char *ptr
1180 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1181 up to I<ptr>. The remaining content of the buffer will be moved, and
1182 all pointers into the buffer updated appropriately. I<ptr> must not
1183 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1184 it is not permitted to discard text that has yet to be lexed.
1186 Normally it is not necessarily to do this directly, because it suffices to
1187 use the implicit discarding behaviour of L</lex_next_chunk> and things
1188 based on it. However, if a token stretches across multiple lines,
1189 and the lexing code has kept multiple lines of text in the buffer for
1190 that purpose, then after completion of the token it would be wise to
1191 explicitly discard the now-unneeded earlier lines, to avoid future
1192 multi-line tokens growing the buffer without bound.
1198 Perl_lex_discard_to(pTHX_ char *ptr)
1202 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1203 buf = SvPVX(PL_parser->linestr);
1205 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1208 if (ptr > PL_parser->bufptr)
1209 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1210 discard_len = ptr - buf;
1211 if (PL_parser->oldbufptr < ptr)
1212 PL_parser->oldbufptr = ptr;
1213 if (PL_parser->oldoldbufptr < ptr)
1214 PL_parser->oldoldbufptr = ptr;
1215 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1216 PL_parser->last_uni = NULL;
1217 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1218 PL_parser->last_lop = NULL;
1219 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1220 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1221 PL_parser->bufend -= discard_len;
1222 PL_parser->bufptr -= discard_len;
1223 PL_parser->oldbufptr -= discard_len;
1224 PL_parser->oldoldbufptr -= discard_len;
1225 if (PL_parser->last_uni)
1226 PL_parser->last_uni -= discard_len;
1227 if (PL_parser->last_lop)
1228 PL_parser->last_lop -= discard_len;
1232 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1234 Reads in the next chunk of text to be lexed, appending it to
1235 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1236 looked to the end of the current chunk and wants to know more. It is
1237 usual, but not necessary, for lexing to have consumed the entirety of
1238 the current chunk at this time.
1240 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1241 chunk (i.e., the current chunk has been entirely consumed), normally the
1242 current chunk will be discarded at the same time that the new chunk is
1243 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1244 will not be discarded. If the current chunk has not been entirely
1245 consumed, then it will not be discarded regardless of the flag.
1247 Returns true if some new text was added to the buffer, or false if the
1248 buffer has reached the end of the input text.
1253 #define LEX_FAKE_EOF 0x80000000
1254 #define LEX_NO_TERM 0x40000000
1257 Perl_lex_next_chunk(pTHX_ U32 flags)
1261 STRLEN old_bufend_pos, new_bufend_pos;
1262 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1263 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1264 bool got_some_for_debugger = 0;
1266 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1267 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1268 linestr = PL_parser->linestr;
1269 buf = SvPVX(linestr);
1270 if (!(flags & LEX_KEEP_PREVIOUS) &&
1271 PL_parser->bufptr == PL_parser->bufend) {
1272 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1274 if (PL_parser->last_uni != PL_parser->bufend)
1275 PL_parser->last_uni = NULL;
1276 if (PL_parser->last_lop != PL_parser->bufend)
1277 PL_parser->last_lop = NULL;
1278 last_uni_pos = last_lop_pos = 0;
1282 old_bufend_pos = PL_parser->bufend - buf;
1283 bufptr_pos = PL_parser->bufptr - buf;
1284 oldbufptr_pos = PL_parser->oldbufptr - buf;
1285 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1286 linestart_pos = PL_parser->linestart - buf;
1287 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1288 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1290 if (flags & LEX_FAKE_EOF) {
1292 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1294 } else if (filter_gets(linestr, old_bufend_pos)) {
1296 got_some_for_debugger = 1;
1297 } else if (flags & LEX_NO_TERM) {
1300 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1301 sv_setpvs(linestr, "");
1303 /* End of real input. Close filehandle (unless it was STDIN),
1304 * then add implicit termination.
1306 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1307 PerlIO_clearerr(PL_parser->rsfp);
1308 else if (PL_parser->rsfp)
1309 (void)PerlIO_close(PL_parser->rsfp);
1310 PL_parser->rsfp = NULL;
1311 PL_parser->in_pod = PL_parser->filtered = 0;
1312 if (!PL_in_eval && PL_minus_p) {
1314 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1315 PL_minus_n = PL_minus_p = 0;
1316 } else if (!PL_in_eval && PL_minus_n) {
1317 sv_catpvs(linestr, /*{*/";}");
1320 sv_catpvs(linestr, ";");
1323 buf = SvPVX(linestr);
1324 new_bufend_pos = SvCUR(linestr);
1325 PL_parser->bufend = buf + new_bufend_pos;
1326 PL_parser->bufptr = buf + bufptr_pos;
1327 PL_parser->oldbufptr = buf + oldbufptr_pos;
1328 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1329 PL_parser->linestart = buf + linestart_pos;
1330 if (PL_parser->last_uni)
1331 PL_parser->last_uni = buf + last_uni_pos;
1332 if (PL_parser->last_lop)
1333 PL_parser->last_lop = buf + last_lop_pos;
1334 if (PL_parser->preambling != NOLINE) {
1335 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1336 PL_parser->preambling = NOLINE;
1338 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1339 PL_curstash != PL_debstash) {
1340 /* debugger active and we're not compiling the debugger code,
1341 * so store the line into the debugger's array of lines
1343 update_debugger_info(NULL, buf+old_bufend_pos,
1344 new_bufend_pos-old_bufend_pos);
1350 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1352 Looks ahead one (Unicode) character in the text currently being lexed.
1353 Returns the codepoint (unsigned integer value) of the next character,
1354 or -1 if lexing has reached the end of the input text. To consume the
1355 peeked character, use L</lex_read_unichar>.
1357 If the next character is in (or extends into) the next chunk of input
1358 text, the next chunk will be read in. Normally the current chunk will be
1359 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1360 then the current chunk will not be discarded.
1362 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1363 is encountered, an exception is generated.
1369 Perl_lex_peek_unichar(pTHX_ U32 flags)
1373 if (flags & ~(LEX_KEEP_PREVIOUS))
1374 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1375 s = PL_parser->bufptr;
1376 bufend = PL_parser->bufend;
1382 if (!lex_next_chunk(flags))
1384 s = PL_parser->bufptr;
1385 bufend = PL_parser->bufend;
1388 if (UTF8_IS_INVARIANT(head))
1390 if (UTF8_IS_START(head)) {
1391 len = UTF8SKIP(&head);
1392 while ((STRLEN)(bufend-s) < len) {
1393 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1395 s = PL_parser->bufptr;
1396 bufend = PL_parser->bufend;
1399 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1400 if (retlen == (STRLEN)-1) {
1401 /* malformed UTF-8 */
1403 SAVESPTR(PL_warnhook);
1404 PL_warnhook = PERL_WARNHOOK_FATAL;
1405 utf8n_to_uvchr((U8*)s, bufend-s, NULL, 0);
1411 if (!lex_next_chunk(flags))
1413 s = PL_parser->bufptr;
1420 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1422 Reads the next (Unicode) character in the text currently being lexed.
1423 Returns the codepoint (unsigned integer value) of the character read,
1424 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1425 if lexing has reached the end of the input text. To non-destructively
1426 examine the next character, use L</lex_peek_unichar> instead.
1428 If the next character is in (or extends into) the next chunk of input
1429 text, the next chunk will be read in. Normally the current chunk will be
1430 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1431 then the current chunk will not be discarded.
1433 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1434 is encountered, an exception is generated.
1440 Perl_lex_read_unichar(pTHX_ U32 flags)
1443 if (flags & ~(LEX_KEEP_PREVIOUS))
1444 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1445 c = lex_peek_unichar(flags);
1448 COPLINE_INC_WITH_HERELINES;
1450 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1452 ++(PL_parser->bufptr);
1458 =for apidoc Amx|void|lex_read_space|U32 flags
1460 Reads optional spaces, in Perl style, in the text currently being
1461 lexed. The spaces may include ordinary whitespace characters and
1462 Perl-style comments. C<#line> directives are processed if encountered.
1463 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1464 at a non-space character (or the end of the input text).
1466 If spaces extend into the next chunk of input text, the next chunk will
1467 be read in. Normally the current chunk will be discarded at the same
1468 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1469 chunk will not be discarded.
1474 #define LEX_NO_INCLINE 0x40000000
1475 #define LEX_NO_NEXT_CHUNK 0x80000000
1478 Perl_lex_read_space(pTHX_ U32 flags)
1481 const bool can_incline = !(flags & LEX_NO_INCLINE);
1482 bool need_incline = 0;
1483 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
1484 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1485 s = PL_parser->bufptr;
1486 bufend = PL_parser->bufend;
1492 } while (!(c == '\n' || (c == 0 && s == bufend)));
1493 } else if (c == '\n') {
1496 PL_parser->linestart = s;
1502 } else if (isSPACE(c)) {
1504 } else if (c == 0 && s == bufend) {
1507 if (flags & LEX_NO_NEXT_CHUNK)
1509 PL_parser->bufptr = s;
1510 l = CopLINE(PL_curcop);
1511 CopLINE(PL_curcop) += PL_parser->herelines + 1;
1512 got_more = lex_next_chunk(flags);
1513 CopLINE_set(PL_curcop, l);
1514 s = PL_parser->bufptr;
1515 bufend = PL_parser->bufend;
1518 if (can_incline && need_incline && PL_parser->rsfp) {
1526 PL_parser->bufptr = s;
1531 =for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1533 This function performs syntax checking on a prototype, C<proto>.
1534 If C<warn> is true, any illegal characters or mismatched brackets
1535 will trigger illegalproto warnings, declaring that they were
1536 detected in the prototype for C<name>.
1538 The return value is C<true> if this is a valid prototype, and
1539 C<false> if it is not, regardless of whether C<warn> was C<true> or
1542 Note that C<NULL> is a valid C<proto> and will always return C<true>.
1549 Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1551 STRLEN len, origlen;
1552 char *p = proto ? SvPV(proto, len) : NULL;
1553 bool bad_proto = FALSE;
1554 bool in_brackets = FALSE;
1555 bool after_slash = FALSE;
1556 char greedy_proto = ' ';
1557 bool proto_after_greedy_proto = FALSE;
1558 bool must_be_last = FALSE;
1559 bool underscore = FALSE;
1560 bool bad_proto_after_underscore = FALSE;
1562 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1568 for (; len--; p++) {
1571 proto_after_greedy_proto = TRUE;
1573 if (!strchr(";@%", *p))
1574 bad_proto_after_underscore = TRUE;
1577 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1584 in_brackets = FALSE;
1585 else if ((*p == '@' || *p == '%') &&
1588 must_be_last = TRUE;
1597 after_slash = FALSE;
1602 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
1605 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1606 origlen, UNI_DISPLAY_ISPRINT)
1607 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1609 if (proto_after_greedy_proto)
1610 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1611 "Prototype after '%c' for %"SVf" : %s",
1612 greedy_proto, SVfARG(name), p);
1614 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1615 "Missing ']' in prototype for %"SVf" : %s",
1618 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1619 "Illegal character in prototype for %"SVf" : %s",
1621 if (bad_proto_after_underscore)
1622 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1623 "Illegal character after '_' in prototype for %"SVf" : %s",
1627 return (! (proto_after_greedy_proto || bad_proto) );
1632 * This subroutine has nothing to do with tilting, whether at windmills
1633 * or pinball tables. Its name is short for "increment line". It
1634 * increments the current line number in CopLINE(PL_curcop) and checks
1635 * to see whether the line starts with a comment of the form
1636 * # line 500 "foo.pm"
1637 * If so, it sets the current line number and file to the values in the comment.
1641 S_incline(pTHX_ const char *s)
1648 PERL_ARGS_ASSERT_INCLINE;
1650 COPLINE_INC_WITH_HERELINES;
1651 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1652 && s+1 == PL_bufend && *s == ';') {
1653 /* fake newline in string eval */
1654 CopLINE_dec(PL_curcop);
1659 while (SPACE_OR_TAB(*s))
1661 if (strnEQ(s, "line", 4))
1665 if (SPACE_OR_TAB(*s))
1669 while (SPACE_OR_TAB(*s))
1677 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1679 while (SPACE_OR_TAB(*s))
1681 if (*s == '"' && (t = strchr(s+1, '"'))) {
1687 while (!isSPACE(*t))
1691 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1693 if (*e != '\n' && *e != '\0')
1694 return; /* false alarm */
1696 line_num = grok_atou(n, &e) - 1;
1699 const STRLEN len = t - s;
1701 if (!PL_rsfp && !PL_parser->filtered) {
1702 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1703 * to *{"::_<newfilename"} */
1704 /* However, the long form of evals is only turned on by the
1705 debugger - usually they're "(eval %lu)" */
1706 GV * const cfgv = CopFILEGV(PL_curcop);
1709 STRLEN tmplen2 = len;
1713 if (tmplen2 + 2 <= sizeof smallbuf)
1716 Newx(tmpbuf2, tmplen2 + 2, char);
1721 memcpy(tmpbuf2 + 2, s, tmplen2);
1724 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1726 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1727 /* adjust ${"::_<newfilename"} to store the new file name */
1728 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1729 /* The line number may differ. If that is the case,
1730 alias the saved lines that are in the array.
1731 Otherwise alias the whole array. */
1732 if (CopLINE(PL_curcop) == line_num) {
1733 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1734 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
1736 else if (GvAV(cfgv)) {
1737 AV * const av = GvAV(cfgv);
1738 const I32 start = CopLINE(PL_curcop)+1;
1739 I32 items = AvFILLp(av) - start;
1741 AV * const av2 = GvAVn(gv2);
1742 SV **svp = AvARRAY(av) + start;
1743 I32 l = (I32)line_num+1;
1745 av_store(av2, l++, SvREFCNT_inc(*svp++));
1750 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1753 CopFILE_free(PL_curcop);
1754 CopFILE_setn(PL_curcop, s, len);
1756 CopLINE_set(PL_curcop, line_num);
1759 #define skipspace(s) skipspace_flags(s, 0)
1763 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1765 AV *av = CopFILEAVx(PL_curcop);
1768 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1770 sv = *av_fetch(av, 0, 1);
1771 SvUPGRADE(sv, SVt_PVMG);
1773 if (!SvPOK(sv)) sv_setpvs(sv,"");
1775 sv_catsv(sv, orig_sv);
1777 sv_catpvn(sv, buf, len);
1782 if (PL_parser->preambling == NOLINE)
1783 av_store(av, CopLINE(PL_curcop), sv);
1789 * Called to gobble the appropriate amount and type of whitespace.
1790 * Skips comments as well.
1794 S_skipspace_flags(pTHX_ char *s, U32 flags)
1796 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
1797 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1798 while (s < PL_bufend && SPACE_OR_TAB(*s))
1801 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1803 lex_read_space(flags | LEX_KEEP_PREVIOUS |
1804 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1805 LEX_NO_NEXT_CHUNK : 0));
1807 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1808 if (PL_linestart > PL_bufptr)
1809 PL_bufptr = PL_linestart;
1817 * Check the unary operators to ensure there's no ambiguity in how they're
1818 * used. An ambiguous piece of code would be:
1820 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1821 * the +5 is its argument.
1830 if (PL_oldoldbufptr != PL_last_uni)
1832 while (isSPACE(*PL_last_uni))
1835 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1837 if ((t = strchr(s, '(')) && t < PL_bufptr)
1840 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1841 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1842 (int)(s - PL_last_uni), PL_last_uni);
1846 * LOP : macro to build a list operator. Its behaviour has been replaced
1847 * with a subroutine, S_lop() for which LOP is just another name.
1850 #define LOP(f,x) return lop(f,x,s)
1854 * Build a list operator (or something that might be one). The rules:
1855 * - if we have a next token, then it's a list operator (no parens) for
1856 * which the next token has already been parsed; e.g.,
1859 * - if the next thing is an opening paren, then it's a function
1860 * - else it's a list operator
1864 S_lop(pTHX_ I32 f, int x, char *s)
1866 PERL_ARGS_ASSERT_LOP;
1871 PL_last_lop = PL_oldbufptr;
1872 PL_last_lop_op = (OPCODE)f;
1877 return REPORT(FUNC);
1880 return REPORT(FUNC);
1883 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1884 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1885 return REPORT(LSTOP);
1891 * When the lexer realizes it knows the next token (for instance,
1892 * it is reordering tokens for the parser) then it can call S_force_next
1893 * to know what token to return the next time the lexer is called. Caller
1894 * will need to set PL_nextval[] and possibly PL_expect to ensure
1895 * the lexer handles the token correctly.
1899 S_force_next(pTHX_ I32 type)
1903 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1904 tokereport(type, &NEXTVAL_NEXTTOKE);
1907 PL_nexttype[PL_nexttoke] = type;
1909 if (PL_lex_state != LEX_KNOWNEXT) {
1910 PL_lex_defer = PL_lex_state;
1911 PL_lex_state = LEX_KNOWNEXT;
1918 * This subroutine handles postfix deref syntax after the arrow has already
1919 * been emitted. @* $* etc. are emitted as two separate token right here.
1920 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1921 * only the first, leaving yylex to find the next.
1925 S_postderef(pTHX_ int const funny, char const next)
1927 assert(funny == DOLSHARP || strchr("$@%&*", funny));
1928 assert(strchr("*[{", next));
1930 PL_expect = XOPERATOR;
1931 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
1932 assert('@' == funny || '$' == funny || DOLSHARP == funny);
1933 PL_lex_state = LEX_INTERPEND;
1934 force_next(POSTJOIN);
1940 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1941 && !PL_lex_brackets)
1943 PL_expect = XOPERATOR;
1952 int yyc = PL_parser->yychar;
1953 if (yyc != YYEMPTY) {
1955 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1956 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1957 PL_lex_allbrackets--;
1959 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1960 } else if (yyc == '('/*)*/) {
1961 PL_lex_allbrackets--;
1966 PL_parser->yychar = YYEMPTY;
1971 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1973 SV * const sv = newSVpvn_utf8(start, len,
1976 && !is_ascii_string((const U8*)start, len)
1977 && is_utf8_string((const U8*)start, len));
1983 * When the lexer knows the next thing is a word (for instance, it has
1984 * just seen -> and it knows that the next char is a word char, then
1985 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1989 * char *start : buffer position (must be within PL_linestr)
1990 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1991 * int check_keyword : if true, Perl checks to make sure the word isn't
1992 * a keyword (do this if the word is a label, e.g. goto FOO)
1993 * int allow_pack : if true, : characters will also be allowed (require,
1994 * use, etc. do this)
1995 * int allow_initial_tick : used by the "sub" lexer only.
1999 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
2004 PERL_ARGS_ASSERT_FORCE_WORD;
2006 start = SKIPSPACE1(start);
2008 if (isIDFIRST_lazy_if(s,UTF) ||
2009 (allow_pack && *s == ':') )
2011 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2012 if (check_keyword) {
2013 char *s2 = PL_tokenbuf;
2014 if (allow_pack && len > 6 && strnEQ(s2, "CORE::", 6))
2016 if (keyword(s2, len, 0))
2019 if (token == METHOD) {
2024 PL_expect = XOPERATOR;
2027 NEXTVAL_NEXTTOKE.opval
2028 = (OP*)newSVOP(OP_CONST,0,
2029 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2030 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2038 * Called when the lexer wants $foo *foo &foo etc, but the program
2039 * text only contains the "foo" portion. The first argument is a pointer
2040 * to the "foo", and the second argument is the type symbol to prefix.
2041 * Forces the next token to be a "WORD".
2042 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2046 S_force_ident(pTHX_ const char *s, int kind)
2048 PERL_ARGS_ASSERT_FORCE_IDENT;
2051 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2052 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2053 UTF ? SVf_UTF8 : 0));
2054 NEXTVAL_NEXTTOKE.opval = o;
2057 o->op_private = OPpCONST_ENTERED;
2058 /* XXX see note in pp_entereval() for why we forgo typo
2059 warnings if the symbol must be introduced in an eval.
2061 gv_fetchpvn_flags(s, len,
2062 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2063 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2064 kind == '$' ? SVt_PV :
2065 kind == '@' ? SVt_PVAV :
2066 kind == '%' ? SVt_PVHV :
2074 S_force_ident_maybe_lex(pTHX_ char pit)
2076 NEXTVAL_NEXTTOKE.ival = pit;
2081 Perl_str_to_version(pTHX_ SV *sv)
2086 const char *start = SvPV_const(sv,len);
2087 const char * const end = start + len;
2088 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2090 PERL_ARGS_ASSERT_STR_TO_VERSION;
2092 while (start < end) {
2096 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2101 retval += ((NV)n)/nshift;
2110 * Forces the next token to be a version number.
2111 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2112 * and if "guessing" is TRUE, then no new token is created (and the caller
2113 * must use an alternative parsing method).
2117 S_force_version(pTHX_ char *s, int guessing)
2122 PERL_ARGS_ASSERT_FORCE_VERSION;
2130 while (isDIGIT(*d) || *d == '_' || *d == '.')
2132 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2134 s = scan_num(s, &pl_yylval);
2135 version = pl_yylval.opval;
2136 ver = cSVOPx(version)->op_sv;
2137 if (SvPOK(ver) && !SvNIOK(ver)) {
2138 SvUPGRADE(ver, SVt_PVNV);
2139 SvNV_set(ver, str_to_version(ver));
2140 SvNOK_on(ver); /* hint that it is a version */
2143 else if (guessing) {
2148 /* NOTE: The parser sees the package name and the VERSION swapped */
2149 NEXTVAL_NEXTTOKE.opval = version;
2156 * S_force_strict_version
2157 * Forces the next token to be a version number using strict syntax rules.
2161 S_force_strict_version(pTHX_ char *s)
2164 const char *errstr = NULL;
2166 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2168 while (isSPACE(*s)) /* leading whitespace */
2171 if (is_STRICT_VERSION(s,&errstr)) {
2173 s = (char *)scan_version(s, ver, 0);
2174 version = newSVOP(OP_CONST, 0, ver);
2176 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2177 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2181 yyerror(errstr); /* version required */
2185 /* NOTE: The parser sees the package name and the VERSION swapped */
2186 NEXTVAL_NEXTTOKE.opval = version;
2194 * Tokenize a quoted string passed in as an SV. It finds the next
2195 * chunk, up to end of string or a backslash. It may make a new
2196 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2201 S_tokeq(pTHX_ SV *sv)
2208 PERL_ARGS_ASSERT_TOKEQ;
2212 assert (!SvIsCOW(sv));
2213 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
2217 /* This is relying on the SV being "well formed" with a trailing '\0' */
2218 while (s < send && !(*s == '\\' && s[1] == '\\'))
2223 if ( PL_hints & HINT_NEW_STRING ) {
2224 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2225 SVs_TEMP | SvUTF8(sv));
2229 if (s + 1 < send && (s[1] == '\\'))
2230 s++; /* all that, just for this */
2235 SvCUR_set(sv, d - SvPVX_const(sv));
2237 if ( PL_hints & HINT_NEW_STRING )
2238 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2243 * Now come three functions related to double-quote context,
2244 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2245 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2246 * interact with PL_lex_state, and create fake ( ... ) argument lists
2247 * to handle functions and concatenation.
2251 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2256 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2258 * Pattern matching will set PL_lex_op to the pattern-matching op to
2259 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2261 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2263 * Everything else becomes a FUNC.
2265 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2266 * had an OP_CONST or OP_READLINE). This just sets us up for a
2267 * call to S_sublex_push().
2271 S_sublex_start(pTHX)
2273 const I32 op_type = pl_yylval.ival;
2275 if (op_type == OP_NULL) {
2276 pl_yylval.opval = PL_lex_op;
2280 if (op_type == OP_CONST) {
2281 SV *sv = tokeq(PL_lex_stuff);
2283 if (SvTYPE(sv) == SVt_PVIV) {
2284 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2286 const char * const p = SvPV_const(sv, len);
2287 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2291 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2292 PL_lex_stuff = NULL;
2296 PL_sublex_info.super_state = PL_lex_state;
2297 PL_sublex_info.sub_inwhat = (U16)op_type;
2298 PL_sublex_info.sub_op = PL_lex_op;
2299 PL_lex_state = LEX_INTERPPUSH;
2303 pl_yylval.opval = PL_lex_op;
2313 * Create a new scope to save the lexing state. The scope will be
2314 * ended in S_sublex_done. Returns a '(', starting the function arguments
2315 * to the uc, lc, etc. found before.
2316 * Sets PL_lex_state to LEX_INTERPCONCAT.
2323 const bool is_heredoc = PL_multi_close == '<';
2326 PL_lex_state = PL_sublex_info.super_state;
2327 SAVEI8(PL_lex_dojoin);
2328 SAVEI32(PL_lex_brackets);
2329 SAVEI32(PL_lex_allbrackets);
2330 SAVEI32(PL_lex_formbrack);
2331 SAVEI8(PL_lex_fakeeof);
2332 SAVEI32(PL_lex_casemods);
2333 SAVEI32(PL_lex_starts);
2334 SAVEI8(PL_lex_state);
2335 SAVESPTR(PL_lex_repl);
2336 SAVEVPTR(PL_lex_inpat);
2337 SAVEI16(PL_lex_inwhat);
2340 SAVECOPLINE(PL_curcop);
2341 SAVEI32(PL_multi_end);
2342 SAVEI32(PL_parser->herelines);
2343 PL_parser->herelines = 0;
2345 SAVEI8(PL_multi_close);
2346 SAVEPPTR(PL_bufptr);
2347 SAVEPPTR(PL_bufend);
2348 SAVEPPTR(PL_oldbufptr);
2349 SAVEPPTR(PL_oldoldbufptr);
2350 SAVEPPTR(PL_last_lop);
2351 SAVEPPTR(PL_last_uni);
2352 SAVEPPTR(PL_linestart);
2353 SAVESPTR(PL_linestr);
2354 SAVEGENERICPV(PL_lex_brackstack);
2355 SAVEGENERICPV(PL_lex_casestack);
2356 SAVEGENERICPV(PL_parser->lex_shared);
2357 SAVEBOOL(PL_parser->lex_re_reparsing);
2358 SAVEI32(PL_copline);
2360 /* The here-doc parser needs to be able to peek into outer lexing
2361 scopes to find the body of the here-doc. So we put PL_linestr and
2362 PL_bufptr into lex_shared, to ‘share’ those values.
2364 PL_parser->lex_shared->ls_linestr = PL_linestr;
2365 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2367 PL_linestr = PL_lex_stuff;
2368 PL_lex_repl = PL_sublex_info.repl;
2369 PL_lex_stuff = NULL;
2370 PL_sublex_info.repl = NULL;
2372 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2373 = SvPVX(PL_linestr);
2374 PL_bufend += SvCUR(PL_linestr);
2375 PL_last_lop = PL_last_uni = NULL;
2376 SAVEFREESV(PL_linestr);
2377 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2379 PL_lex_dojoin = FALSE;
2380 PL_lex_brackets = PL_lex_formbrack = 0;
2381 PL_lex_allbrackets = 0;
2382 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2383 Newx(PL_lex_brackstack, 120, char);
2384 Newx(PL_lex_casestack, 12, char);
2385 PL_lex_casemods = 0;
2386 *PL_lex_casestack = '\0';
2388 PL_lex_state = LEX_INTERPCONCAT;
2390 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2391 PL_copline = NOLINE;
2393 Newxz(shared, 1, LEXSHARED);
2394 shared->ls_prev = PL_parser->lex_shared;
2395 PL_parser->lex_shared = shared;
2397 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2398 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2399 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2400 PL_lex_inpat = PL_sublex_info.sub_op;
2402 PL_lex_inpat = NULL;
2404 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2405 PL_in_eval &= ~EVAL_RE_REPARSING;
2412 * Restores lexer state after a S_sublex_push.
2418 if (!PL_lex_starts++) {
2419 SV * const sv = newSVpvs("");
2420 if (SvUTF8(PL_linestr))
2422 PL_expect = XOPERATOR;
2423 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2427 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2428 PL_lex_state = LEX_INTERPCASEMOD;
2432 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2433 assert(PL_lex_inwhat != OP_TRANSR);
2435 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
2436 PL_linestr = PL_lex_repl;
2438 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2439 PL_bufend += SvCUR(PL_linestr);
2440 PL_last_lop = PL_last_uni = NULL;
2441 PL_lex_dojoin = FALSE;
2442 PL_lex_brackets = 0;
2443 PL_lex_allbrackets = 0;
2444 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2445 PL_lex_casemods = 0;
2446 *PL_lex_casestack = '\0';
2448 if (SvEVALED(PL_lex_repl)) {
2449 PL_lex_state = LEX_INTERPNORMAL;
2451 /* we don't clear PL_lex_repl here, so that we can check later
2452 whether this is an evalled subst; that means we rely on the
2453 logic to ensure sublex_done() is called again only via the
2454 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2457 PL_lex_state = LEX_INTERPCONCAT;
2460 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2461 CopLINE(PL_curcop) +=
2462 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xpad_cop_seq.xlow
2463 + PL_parser->herelines;
2464 PL_parser->herelines = 0;
2469 const line_t l = CopLINE(PL_curcop);
2471 if (PL_multi_close == '<')
2472 PL_parser->herelines += l - PL_multi_end;
2473 PL_bufend = SvPVX(PL_linestr);
2474 PL_bufend += SvCUR(PL_linestr);
2475 PL_expect = XOPERATOR;
2476 PL_sublex_info.sub_inwhat = 0;
2481 PERL_STATIC_INLINE SV*
2482 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2484 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2485 * interior, hence to the "}". Finds what the name resolves to, returning
2486 * an SV* containing it; NULL if no valid one found */
2488 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2495 const U8* first_bad_char_loc;
2496 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2498 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2500 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2502 &first_bad_char_loc))
2504 /* If warnings are on, this will print a more detailed analysis of what
2505 * is wrong than the error message below */
2506 utf8n_to_uvchr(first_bad_char_loc,
2507 e - ((char *) first_bad_char_loc),
2510 /* We deliberately don't try to print the malformed character, which
2511 * might not print very well; it also may be just the first of many
2512 * malformations, so don't print what comes after it */
2513 yyerror(Perl_form(aTHX_
2514 "Malformed UTF-8 character immediately after '%.*s'",
2515 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr));
2519 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2520 /* include the <}> */
2521 e - backslash_ptr + 1);
2523 SvREFCNT_dec_NN(res);
2527 /* See if the charnames handler is the Perl core's, and if so, we can skip
2528 * the validation needed for a user-supplied one, as Perl's does its own
2530 table = GvHV(PL_hintgv); /* ^H */
2531 cvp = hv_fetchs(table, "charnames", FALSE);
2532 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2533 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
2535 const char * const name = HvNAME(stash);
2536 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2537 && strEQ(name, "_charnames")) {
2542 /* Here, it isn't Perl's charname handler. We can't rely on a
2543 * user-supplied handler to validate the input name. For non-ut8 input,
2544 * look to see that the first character is legal. Then loop through the
2545 * rest checking that each is a continuation */
2547 /* This code makes the reasonable assumption that the only Latin1-range
2548 * characters that begin a character name alias are alphabetic, otherwise
2549 * would have to create a isCHARNAME_BEGIN macro */
2552 if (! isALPHAU(*s)) {
2557 if (! isCHARNAME_CONT(*s)) {
2560 if (*s == ' ' && *(s-1) == ' ') {
2563 if ((U8) *s == NBSP_NATIVE && ckWARN_d(WARN_DEPRECATED)) {
2564 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2565 "NO-BREAK SPACE in a charnames "
2566 "alias definition is deprecated");
2572 /* Similarly for utf8. For invariants can check directly; for other
2573 * Latin1, can calculate their code point and check; otherwise use a
2575 if (UTF8_IS_INVARIANT(*s)) {
2576 if (! isALPHAU(*s)) {
2580 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2581 if (! isALPHAU(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1)))) {
2587 if (! PL_utf8_charname_begin) {
2588 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2589 PL_utf8_charname_begin = _core_swash_init("utf8",
2590 "_Perl_Charname_Begin",
2592 1, 0, NULL, &flags);
2594 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2601 if (UTF8_IS_INVARIANT(*s)) {
2602 if (! isCHARNAME_CONT(*s)) {
2605 if (*s == ' ' && *(s-1) == ' ') {
2610 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2611 if (! isCHARNAME_CONT(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1))))
2615 if (*s == *NBSP_UTF8
2616 && *(s+1) == *(NBSP_UTF8+1)
2617 && ckWARN_d(WARN_DEPRECATED))
2619 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2620 "NO-BREAK SPACE in a charnames "
2621 "alias definition is deprecated");
2626 if (! PL_utf8_charname_continue) {
2627 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2628 PL_utf8_charname_continue = _core_swash_init("utf8",
2629 "_Perl_Charname_Continue",
2631 1, 0, NULL, &flags);
2633 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2640 if (*(s-1) == ' ') {
2643 "charnames alias definitions may not contain trailing "
2644 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2645 (int)(s - backslash_ptr + 1), backslash_ptr,
2646 (int)(e - s + 1), s + 1
2648 UTF ? SVf_UTF8 : 0);
2652 if (SvUTF8(res)) { /* Don't accept malformed input */
2653 const U8* first_bad_char_loc;
2655 const char* const str = SvPV_const(res, len);
2656 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2657 /* If warnings are on, this will print a more detailed analysis of
2658 * what is wrong than the error message below */
2659 utf8n_to_uvchr(first_bad_char_loc,
2660 (char *) first_bad_char_loc - str,
2663 /* We deliberately don't try to print the malformed character,
2664 * which might not print very well; it also may be just the first
2665 * of many malformations, so don't print what comes after it */
2668 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2669 (int) (e - backslash_ptr + 1), backslash_ptr,
2670 (int) ((char *) first_bad_char_loc - str), str
2681 /* The final %.*s makes sure that should the trailing NUL be missing
2682 * that this print won't run off the end of the string */
2685 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2686 (int)(s - backslash_ptr + 1), backslash_ptr,
2687 (int)(e - s + 1), s + 1
2689 UTF ? SVf_UTF8 : 0);
2696 "charnames alias definitions may not contain a sequence of "
2697 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2698 (int)(s - backslash_ptr + 1), backslash_ptr,
2699 (int)(e - s + 1), s + 1
2701 UTF ? SVf_UTF8 : 0);
2708 Extracts the next constant part of a pattern, double-quoted string,
2709 or transliteration. This is terrifying code.
2711 For example, in parsing the double-quoted string "ab\x63$d", it would
2712 stop at the '$' and return an OP_CONST containing 'abc'.
2714 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2715 processing a pattern (PL_lex_inpat is true), a transliteration
2716 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2718 Returns a pointer to the character scanned up to. If this is
2719 advanced from the start pointer supplied (i.e. if anything was
2720 successfully parsed), will leave an OP_CONST for the substring scanned
2721 in pl_yylval. Caller must intuit reason for not parsing further
2722 by looking at the next characters herself.
2726 \N{FOO} => \N{U+hex_for_character_FOO}
2727 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2730 all other \-char, including \N and \N{ apart from \N{ABC}
2733 @ and $ where it appears to be a var, but not for $ as tail anchor
2738 In transliterations:
2739 characters are VERY literal, except for - not at the start or end
2740 of the string, which indicates a range. If the range is in bytes,
2741 scan_const expands the range to the full set of intermediate
2742 characters. If the range is in utf8, the hyphen is replaced with
2743 a certain range mark which will be handled by pmtrans() in op.c.
2745 In double-quoted strings:
2747 double-quoted style: \r and \n
2748 constants: \x31, etc.
2749 deprecated backrefs: \1 (in substitution replacements)
2750 case and quoting: \U \Q \E
2753 scan_const does *not* construct ops to handle interpolated strings.
2754 It stops processing as soon as it finds an embedded $ or @ variable
2755 and leaves it to the caller to work out what's going on.
2757 embedded arrays (whether in pattern or not) could be:
2758 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2760 $ in double-quoted strings must be the symbol of an embedded scalar.
2762 $ in pattern could be $foo or could be tail anchor. Assumption:
2763 it's a tail anchor if $ is the last thing in the string, or if it's
2764 followed by one of "()| \r\n\t"
2766 \1 (backreferences) are turned into $1 in substitutions
2768 The structure of the code is
2769 while (there's a character to process) {
2770 handle transliteration ranges
2771 skip regexp comments /(?#comment)/ and codes /(?{code})/
2772 skip #-initiated comments in //x patterns
2773 check for embedded arrays
2774 check for embedded scalars
2776 deprecate \1 in substitution replacements
2777 handle string-changing backslashes \l \U \Q \E, etc.
2778 switch (what was escaped) {
2779 handle \- in a transliteration (becomes a literal -)
2780 if a pattern and not \N{, go treat as regular character
2781 handle \132 (octal characters)
2782 handle \x15 and \x{1234} (hex characters)
2783 handle \N{name} (named characters, also \N{3,5} in a pattern)
2784 handle \cV (control characters)
2785 handle printf-style backslashes (\f, \r, \n, etc)
2788 } (end if backslash)
2789 handle regular character
2790 } (end while character to read)
2795 S_scan_const(pTHX_ char *start)
2797 char *send = PL_bufend; /* end of the constant */
2798 SV *sv = newSV(send - start); /* sv for the constant. See note below
2800 char *s = start; /* start of the constant */
2801 char *d = SvPVX(sv); /* destination for copies */
2802 bool dorange = FALSE; /* are we in a translit range? */
2803 bool didrange = FALSE; /* did we just finish a range? */
2804 bool in_charclass = FALSE; /* within /[...]/ */
2805 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2806 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2807 UTF8? But, this can show as true
2808 when the source isn't utf8, as for
2809 example when it is entirely composed
2811 SV *res; /* result from charnames */
2813 /* Note on sizing: The scanned constant is placed into sv, which is
2814 * initialized by newSV() assuming one byte of output for every byte of
2815 * input. This routine expects newSV() to allocate an extra byte for a
2816 * trailing NUL, which this routine will append if it gets to the end of
2817 * the input. There may be more bytes of input than output (eg., \N{LATIN
2818 * CAPITAL LETTER A}), or more output than input if the constant ends up
2819 * recoded to utf8, but each time a construct is found that might increase
2820 * the needed size, SvGROW() is called. Its size parameter each time is
2821 * based on the best guess estimate at the time, namely the length used so
2822 * far, plus the length the current construct will occupy, plus room for
2823 * the trailing NUL, plus one byte for every input byte still unscanned */
2825 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2828 UV literal_endpoint = 0;
2829 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2832 PERL_ARGS_ASSERT_SCAN_CONST;
2834 assert(PL_lex_inwhat != OP_TRANSR);
2835 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2836 /* If we are doing a trans and we know we want UTF8 set expectation */
2837 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2838 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2841 /* Protect sv from errors and fatal warnings. */
2842 ENTER_with_name("scan_const");
2845 while (s < send || dorange) {
2847 /* get transliterations out of the way (they're most literal) */
2848 if (PL_lex_inwhat == OP_TRANS) {
2849 /* expand a range A-Z to the full set of characters. AIE! */
2851 I32 i; /* current expanded character */
2852 I32 min; /* first character in range */
2853 I32 max; /* last character in range */
2864 char * const c = (char*)utf8_hop((U8*)d, -1);
2868 *c = (char) ILLEGAL_UTF8_BYTE;
2869 /* mark the range as done, and continue */
2875 i = d - SvPVX_const(sv); /* remember current offset */
2878 SvLEN(sv) + ((has_utf8)
2879 ? (512 - UTF_CONTINUATION_MARK
2882 /* How many two-byte within 0..255: 128 in UTF-8,
2883 * 96 in UTF-8-mod. */
2885 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2887 d = SvPVX(sv) + i; /* refresh d after realloc */
2891 for (j = 0; j <= 1; j++) {
2892 char * const c = (char*)utf8_hop((U8*)d, -1);
2893 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2899 max = (U8)0xff; /* only to \xff */
2900 uvmax = uv; /* \x{100} to uvmax */
2902 d = c; /* eat endpoint chars */
2907 d -= 2; /* eat the first char and the - */
2908 min = (U8)*d; /* first char in range */
2909 max = (U8)d[1]; /* last char in range */
2916 "Invalid range \"%c-%c\" in transliteration operator",
2917 (char)min, (char)max);
2921 /* Because of the discontinuities in EBCDIC A-Z and a-z, expand
2922 * any subsets of these ranges into individual characters */
2923 if (literal_endpoint == 2 &&
2924 ((isLOWER_A(min) && isLOWER_A(max)) ||
2925 (isUPPER_A(min) && isUPPER_A(max))))
2927 for (i = min; i <= max; i++) {
2934 for (i = min; i <= max; i++)
2937 append_utf8_from_native_byte(i, &d);
2945 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2947 *d++ = (char) ILLEGAL_UTF8_BYTE;
2949 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2953 /* mark the range as done, and continue */
2957 literal_endpoint = 0;
2962 /* range begins (ignore - as first or last char) */
2963 else if (*s == '-' && s+1 < send && s != start) {
2965 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2972 *d++ = (char) ILLEGAL_UTF8_BYTE; /* use illegal utf8 byte--see pmtrans */
2982 literal_endpoint = 0;
2983 native_range = TRUE;
2988 /* if we get here, we're not doing a transliteration */
2990 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
2993 while (s1 >= start && *s1-- == '\\')
2996 in_charclass = TRUE;
2999 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3002 while (s1 >= start && *s1-- == '\\')
3005 in_charclass = FALSE;
3008 /* skip for regexp comments /(?#comment)/, except for the last
3009 * char, which will be done separately.
3010 * Stop on (?{..}) and friends */
3012 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
3014 while (s+1 < send && *s != ')')
3017 else if (!PL_lex_casemods &&
3018 ( s[2] == '{' /* This should match regcomp.c */
3019 || (s[2] == '?' && s[3] == '{')))
3025 /* likewise skip #-initiated comments in //x patterns */
3026 else if (*s == '#' && PL_lex_inpat && !in_charclass &&
3027 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3028 while (s+1 < send && *s != '\n')
3032 /* no further processing of single-quoted regex */
3033 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3034 goto default_action;
3036 /* check for embedded arrays
3037 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3039 else if (*s == '@' && s[1]) {
3040 if (isWORDCHAR_lazy_if(s+1,UTF))
3042 if (strchr(":'{$", s[1]))
3044 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3045 break; /* in regexp, neither @+ nor @- are interpolated */
3048 /* check for embedded scalars. only stop if we're sure it's a
3051 else if (*s == '$') {
3052 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3054 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3056 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3057 "Possible unintended interpolation of $\\ in regex");
3059 break; /* in regexp, $ might be tail anchor */
3063 /* End of else if chain - OP_TRANS rejoin rest */
3066 if (*s == '\\' && s+1 < send) {
3067 char* e; /* Can be used for ending '}', etc. */
3071 /* warn on \1 - \9 in substitution replacements, but note that \11
3072 * is an octal; and \19 is \1 followed by '9' */
3073 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3074 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3076 /* diag_listed_as: \%d better written as $%d */
3077 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3082 /* string-change backslash escapes */
3083 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3087 /* In a pattern, process \N, but skip any other backslash escapes.
3088 * This is because we don't want to translate an escape sequence
3089 * into a meta symbol and have the regex compiler use the meta
3090 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3091 * in spite of this, we do have to process \N here while the proper
3092 * charnames handler is in scope. See bugs #56444 and #62056.
3093 * There is a complication because \N in a pattern may also stand
3094 * for 'match a non-nl', and not mean a charname, in which case its
3095 * processing should be deferred to the regex compiler. To be a
3096 * charname it must be followed immediately by a '{', and not look
3097 * like \N followed by a curly quantifier, i.e., not something like
3098 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3100 else if (PL_lex_inpat
3103 || regcurly(s + 1)))
3106 goto default_action;
3111 /* quoted - in transliterations */
3113 if (PL_lex_inwhat == OP_TRANS) {
3120 if ((isALPHANUMERIC(*s)))
3121 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3122 "Unrecognized escape \\%c passed through",
3124 /* default action is to copy the quoted character */
3125 goto default_action;
3128 /* eg. \132 indicates the octal constant 0132 */
3129 case '0': case '1': case '2': case '3':
3130 case '4': case '5': case '6': case '7':
3132 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3134 uv = grok_oct(s, &len, &flags, NULL);
3136 if (len < 3 && s < send && isDIGIT(*s)
3137 && ckWARN(WARN_MISC))
3139 Perl_warner(aTHX_ packWARN(WARN_MISC),
3140 "%s", form_short_octal_warning(s, len));
3143 goto NUM_ESCAPE_INSERT;
3145 /* eg. \o{24} indicates the octal constant \024 */
3150 bool valid = grok_bslash_o(&s, &uv, &error,
3151 TRUE, /* Output warning */
3152 FALSE, /* Not strict */
3153 TRUE, /* Output warnings for
3160 goto NUM_ESCAPE_INSERT;
3163 /* eg. \x24 indicates the hex constant 0x24 */
3168 bool valid = grok_bslash_x(&s, &uv, &error,
3169 TRUE, /* Output warning */
3170 FALSE, /* Not strict */
3171 TRUE, /* Output warnings for
3181 /* Insert oct or hex escaped character. There will always be
3182 * enough room in sv since such escapes will be longer than any
3183 * UTF-8 sequence they can end up as, except if they force us
3184 * to recode the rest of the string into utf8 */
3186 /* Here uv is the ordinal of the next character being added */
3187 if (!UVCHR_IS_INVARIANT(uv)) {
3188 if (!has_utf8 && uv > 255) {
3189 /* Might need to recode whatever we have accumulated so
3190 * far if it contains any chars variant in utf8 or
3193 SvCUR_set(sv, d - SvPVX_const(sv));
3196 /* See Note on sizing above. */
3197 sv_utf8_upgrade_flags_grow(sv,
3198 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3199 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3200 d = SvPVX(sv) + SvCUR(sv);
3205 d = (char*)uvchr_to_utf8((U8*)d, uv);
3206 if (PL_lex_inwhat == OP_TRANS &&
3207 PL_sublex_info.sub_op) {
3208 PL_sublex_info.sub_op->op_private |=
3209 (PL_lex_repl ? OPpTRANS_FROM_UTF
3213 if (uv > 255 && !dorange)
3214 native_range = FALSE;
3227 /* In a non-pattern \N must be a named character, like \N{LATIN
3228 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3229 * mean to match a non-newline. For non-patterns, named
3230 * characters are converted to their string equivalents. In
3231 * patterns, named characters are not converted to their
3232 * ultimate forms for the same reasons that other escapes
3233 * aren't. Instead, they are converted to the \N{U+...} form
3234 * to get the value from the charnames that is in effect right
3235 * now, while preserving the fact that it was a named character
3236 * so that the regex compiler knows this */
3238 /* The structure of this section of code (besides checking for
3239 * errors and upgrading to utf8) is:
3240 * Further disambiguate between the two meanings of \N, and if
3241 * not a charname, go process it elsewhere
3242 * If of form \N{U+...}, pass it through if a pattern;
3243 * otherwise convert to utf8
3244 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3245 * pattern; otherwise convert to utf8 */
3247 /* Here, s points to the 'N'; the test below is guaranteed to
3248 * succeed if we are being called on a pattern as we already
3249 * know from a test above that the next character is a '{'.
3250 * On a non-pattern \N must mean 'named sequence, which
3251 * requires braces */
3254 yyerror("Missing braces on \\N{}");
3259 /* If there is no matching '}', it is an error. */
3260 if (! (e = strchr(s, '}'))) {
3261 if (! PL_lex_inpat) {
3262 yyerror("Missing right brace on \\N{}");
3264 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
3269 /* Here it looks like a named character */
3271 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3272 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3273 | PERL_SCAN_DISALLOW_PREFIX;
3276 /* For \N{U+...}, the '...' is a unicode value even on
3277 * EBCDIC machines */
3278 s += 2; /* Skip to next char after the 'U+' */
3280 uv = grok_hex(s, &len, &flags, NULL);
3281 if (len == 0 || len != (STRLEN)(e - s)) {
3282 yyerror("Invalid hexadecimal number in \\N{U+...}");
3289 /* On non-EBCDIC platforms, pass through to the regex
3290 * compiler unchanged. The reason we evaluated the
3291 * number above is to make sure there wasn't a syntax
3292 * error. But on EBCDIC we convert to native so
3293 * downstream code can continue to assume it's native
3295 s -= 5; /* Include the '\N{U+' */
3297 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3300 (unsigned int) UNI_TO_NATIVE(uv));
3302 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3306 else { /* Not a pattern: convert the hex to string */
3308 /* If destination is not in utf8, unconditionally
3309 * recode it to be so. This is because \N{} implies
3310 * Unicode semantics, and scalars have to be in utf8
3311 * to guarantee those semantics */
3313 SvCUR_set(sv, d - SvPVX_const(sv));
3316 /* See Note on sizing above. */
3317 sv_utf8_upgrade_flags_grow(
3319 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3320 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3321 d = SvPVX(sv) + SvCUR(sv);
3325 /* Add the (Unicode) code point to the output. */
3326 if (UNI_IS_INVARIANT(uv)) {
3327 *d++ = (char) LATIN1_TO_NATIVE(uv);
3330 d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
3334 else /* Here is \N{NAME} but not \N{U+...}. */
3335 if ((res = get_and_check_backslash_N_name(s, e)))
3338 const char *str = SvPV_const(res, len);
3341 if (! len) { /* The name resolved to an empty string */
3342 Copy("\\N{}", d, 4, char);
3346 /* In order to not lose information for the regex
3347 * compiler, pass the result in the specially made
3348 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3349 * the code points in hex of each character
3350 * returned by charnames */
3352 const char *str_end = str + len;
3353 const STRLEN off = d - SvPVX_const(sv);
3355 if (! SvUTF8(res)) {
3356 /* For the non-UTF-8 case, we can determine the
3357 * exact length needed without having to parse
3358 * through the string. Each character takes up
3359 * 2 hex digits plus either a trailing dot or
3361 d = off + SvGROW(sv, off
3363 + 6 /* For the "\N{U+", and
3365 + (STRLEN)(send - e));
3366 Copy("\\N{U+", d, 5, char);
3368 while (str < str_end) {
3371 my_snprintf(hex_string,
3373 "%02X.", (U8) *str);
3374 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(hex_string));
3375 Copy(hex_string, d, 3, char);
3379 d--; /* We will overwrite below the final
3380 dot with a right brace */
3383 STRLEN char_length; /* cur char's byte length */
3385 /* and the number of bytes after this is
3386 * translated into hex digits */
3387 STRLEN output_length;
3389 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3390 * for max('U+', '.'); and 1 for NUL */
3391 char hex_string[2 * UTF8_MAXBYTES + 5];
3393 /* Get the first character of the result. */
3394 U32 uv = utf8n_to_uvchr((U8 *) str,
3398 /* Convert first code point to hex, including
3399 * the boiler plate before it. */
3401 my_snprintf(hex_string, sizeof(hex_string),
3405 /* Make sure there is enough space to hold it */
3406 d = off + SvGROW(sv, off
3408 + (STRLEN)(send - e)
3409 + 2); /* '}' + NUL */
3411 Copy(hex_string, d, output_length, char);
3414 /* For each subsequent character, append dot and
3415 * its ordinal in hex */
3416 while ((str += char_length) < str_end) {
3417 const STRLEN off = d - SvPVX_const(sv);
3418 U32 uv = utf8n_to_uvchr((U8 *) str,
3423 my_snprintf(hex_string,
3428 d = off + SvGROW(sv, off
3430 + (STRLEN)(send - e)
3431 + 2); /* '}' + NUL */
3432 Copy(hex_string, d, output_length, char);
3437 *d++ = '}'; /* Done. Add the trailing brace */
3440 else { /* Here, not in a pattern. Convert the name to a
3443 /* If destination is not in utf8, unconditionally
3444 * recode it to be so. This is because \N{} implies
3445 * Unicode semantics, and scalars have to be in utf8
3446 * to guarantee those semantics */
3448 SvCUR_set(sv, d - SvPVX_const(sv));
3451 /* See Note on sizing above. */
3452 sv_utf8_upgrade_flags_grow(sv,
3453 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3454 len + (STRLEN)(send - s) + 1);
3455 d = SvPVX(sv) + SvCUR(sv);
3457 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3459 /* See Note on sizing above. (NOTE: SvCUR() is not
3460 * set correctly here). */
3461 const STRLEN off = d - SvPVX_const(sv);
3462 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3464 if (! SvUTF8(res)) { /* Make sure is \N{} return is UTF-8 */
3465 sv_utf8_upgrade(res);
3466 str = SvPV_const(res, len);
3468 Copy(str, d, len, char);
3474 } /* End \N{NAME} */
3477 native_range = FALSE; /* \N{} is defined to be Unicode */
3479 s = e + 1; /* Point to just after the '}' */
3482 /* \c is a control character */
3486 *d++ = grok_bslash_c(*s++, 1);
3489 yyerror("Missing control char name in \\c");
3493 /* printf-style backslashes, formfeeds, newlines, etc */
3519 } /* end if (backslash) */
3526 /* If we started with encoded form, or already know we want it,
3527 then encode the next character */
3528 if (! NATIVE_BYTE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3532 /* One might think that it is wasted effort in the case of the
3533 * source being utf8 (this_utf8 == TRUE) to take the next character
3534 * in the source, convert it to an unsigned value, and then convert
3535 * it back again. But the source has not been validated here. The
3536 * routine that does the conversion checks for errors like
3539 const UV nextuv = (this_utf8)
3540 ? utf8n_to_uvchr((U8*)s, send - s, &len, 0)
3542 const STRLEN need = UNISKIP(nextuv);
3544 SvCUR_set(sv, d - SvPVX_const(sv));
3547 /* See Note on sizing above. */
3548 sv_utf8_upgrade_flags_grow(sv,
3549 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3550 need + (STRLEN)(send - s) + 1);
3551 d = SvPVX(sv) + SvCUR(sv);
3553 } else if (need > len) {
3554 /* encoded value larger than old, may need extra space (NOTE:
3555 * SvCUR() is not set correctly here). See Note on sizing
3557 const STRLEN off = d - SvPVX_const(sv);
3558 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3562 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3564 if (uv > 255 && !dorange)
3565 native_range = FALSE;
3571 } /* while loop to process each character */
3573 /* terminate the string and set up the sv */
3575 SvCUR_set(sv, d - SvPVX_const(sv));
3576 if (SvCUR(sv) >= SvLEN(sv))
3577 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3578 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3581 if (PL_encoding && !has_utf8) {
3582 sv_recode_to_utf8(sv, PL_encoding);
3588 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3589 PL_sublex_info.sub_op->op_private |=
3590 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3594 /* shrink the sv if we allocated more than we used */
3595 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3596 SvPV_shrink_to_cur(sv);
3599 /* return the substring (via pl_yylval) only if we parsed anything */
3602 for (; s2 < s; s2++) {
3604 COPLINE_INC_WITH_HERELINES;
3606 SvREFCNT_inc_simple_void_NN(sv);
3607 if ( (PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ))
3608 && ! PL_parser->lex_re_reparsing)
3610 const char *const key = PL_lex_inpat ? "qr" : "q";
3611 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3615 if (PL_lex_inwhat == OP_TRANS) {
3618 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3621 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3629 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3632 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3634 LEAVE_with_name("scan_const");
3639 * Returns TRUE if there's more to the expression (e.g., a subscript),
3642 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3644 * ->[ and ->{ return TRUE
3645 * ->$* ->$#* ->@* ->@[ ->@{ return TRUE if postderef_qq is enabled
3646 * { and [ outside a pattern are always subscripts, so return TRUE
3647 * if we're outside a pattern and it's not { or [, then return FALSE
3648 * if we're in a pattern and the first char is a {
3649 * {4,5} (any digits around the comma) returns FALSE
3650 * if we're in a pattern and the first char is a [
3652 * [SOMETHING] has a funky algorithm to decide whether it's a
3653 * character class or not. It has to deal with things like
3654 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3655 * anything else returns TRUE
3658 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3661 S_intuit_more(pTHX_ char *s)
3663 PERL_ARGS_ASSERT_INTUIT_MORE;
3665 if (PL_lex_brackets)
3667 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3669 if (*s == '-' && s[1] == '>'
3670 && FEATURE_POSTDEREF_QQ_IS_ENABLED
3671 && ( (s[2] == '$' && (s[3] == '*' || (s[3] == '#' && s[4] == '*')))
3672 ||(s[2] == '@' && strchr("*[{",s[3])) ))
3674 if (*s != '{' && *s != '[')
3679 /* In a pattern, so maybe we have {n,m}. */
3687 /* On the other hand, maybe we have a character class */
3690 if (*s == ']' || *s == '^')
3693 /* this is terrifying, and it works */
3696 const char * const send = strchr(s,']');
3697 unsigned char un_char, last_un_char;
3698 char tmpbuf[sizeof PL_tokenbuf * 4];
3700 if (!send) /* has to be an expression */
3702 weight = 2; /* let's weigh the evidence */
3706 else if (isDIGIT(*s)) {
3708 if (isDIGIT(s[1]) && s[2] == ']')
3714 Zero(seen,256,char);
3716 for (; s < send; s++) {
3717 last_un_char = un_char;
3718 un_char = (unsigned char)*s;
3723 weight -= seen[un_char] * 10;
3724 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3726 char *tmp = PL_bufend;
3727 PL_bufend = (char*)send;
3728 scan_ident(s, tmpbuf, sizeof tmpbuf, FALSE);
3730 len = (int)strlen(tmpbuf);
3731 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3732 UTF ? SVf_UTF8 : 0, SVt_PV))
3737 else if (*s == '$' && s[1] &&
3738 strchr("[#!%*<>()-=",s[1])) {
3739 if (/*{*/ strchr("])} =",s[2]))
3748 if (strchr("wds]",s[1]))
3750 else if (seen[(U8)'\''] || seen[(U8)'"'])
3752 else if (strchr("rnftbxcav",s[1]))
3754 else if (isDIGIT(s[1])) {
3756 while (s[1] && isDIGIT(s[1]))
3766 if (strchr("aA01! ",last_un_char))
3768 if (strchr("zZ79~",s[1]))
3770 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3771 weight -= 5; /* cope with negative subscript */
3774 if (!isWORDCHAR(last_un_char)
3775 && !(last_un_char == '$' || last_un_char == '@'
3776 || last_un_char == '&')
3777 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3782 if (keyword(tmpbuf, d - tmpbuf, 0))
3785 if (un_char == last_un_char + 1)
3787 weight -= seen[un_char];
3792 if (weight >= 0) /* probably a character class */
3802 * Does all the checking to disambiguate
3804 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3805 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3807 * First argument is the stuff after the first token, e.g. "bar".
3809 * Not a method if foo is a filehandle.
3810 * Not a method if foo is a subroutine prototyped to take a filehandle.
3811 * Not a method if it's really "Foo $bar"
3812 * Method if it's "foo $bar"
3813 * Not a method if it's really "print foo $bar"
3814 * Method if it's really "foo package::" (interpreted as package->foo)
3815 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3816 * Not a method if bar is a filehandle or package, but is quoted with
3821 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3823 char *s = start + (*start == '$');
3824 char tmpbuf[sizeof PL_tokenbuf];
3828 PERL_ARGS_ASSERT_INTUIT_METHOD;
3830 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3832 if (cv && SvPOK(cv)) {
3833 const char *proto = CvPROTO(cv);
3835 while (*proto && (isSPACE(*proto) || *proto == ';'))
3842 if (*start == '$') {
3843 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3844 isUPPER(*PL_tokenbuf))
3849 return *s == '(' ? FUNCMETH : METHOD;
3852 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3853 /* start is the beginning of the possible filehandle/object,
3854 * and s is the end of it
3855 * tmpbuf is a copy of it (but with single quotes as double colons)
3858 if (!keyword(tmpbuf, len, 0)) {
3859 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3864 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3865 if (indirgv && GvCVu(indirgv))
3867 /* filehandle or package name makes it a method */
3868 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3870 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3871 return 0; /* no assumptions -- "=>" quotes bareword */
3873 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3874 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3875 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3879 return *s == '(' ? FUNCMETH : METHOD;
3885 /* Encoded script support. filter_add() effectively inserts a
3886 * 'pre-processing' function into the current source input stream.
3887 * Note that the filter function only applies to the current source file
3888 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3890 * The datasv parameter (which may be NULL) can be used to pass
3891 * private data to this instance of the filter. The filter function
3892 * can recover the SV using the FILTER_DATA macro and use it to
3893 * store private buffers and state information.
3895 * The supplied datasv parameter is upgraded to a PVIO type
3896 * and the IoDIRP/IoANY field is used to store the function pointer,
3897 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3898 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3899 * private use must be set using malloc'd pointers.
3903 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3911 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3912 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3914 if (!PL_rsfp_filters)
3915 PL_rsfp_filters = newAV();
3918 SvUPGRADE(datasv, SVt_PVIO);
3919 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3920 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3921 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3922 FPTR2DPTR(void *, IoANY(datasv)),
3923 SvPV_nolen(datasv)));
3924 av_unshift(PL_rsfp_filters, 1);
3925 av_store(PL_rsfp_filters, 0, datasv) ;
3927 !PL_parser->filtered
3928 && PL_parser->lex_flags & LEX_EVALBYTES
3929 && PL_bufptr < PL_bufend
3931 const char *s = PL_bufptr;
3932 while (s < PL_bufend) {
3934 SV *linestr = PL_parser->linestr;
3935 char *buf = SvPVX(linestr);
3936 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3937 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3938 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3939 STRLEN const linestart_pos = PL_parser->linestart - buf;
3940 STRLEN const last_uni_pos =
3941 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3942 STRLEN const last_lop_pos =
3943 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3944 av_push(PL_rsfp_filters, linestr);
3945 PL_parser->linestr =
3946 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3947 buf = SvPVX(PL_parser->linestr);
3948 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3949 PL_parser->bufptr = buf + bufptr_pos;
3950 PL_parser->oldbufptr = buf + oldbufptr_pos;
3951 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3952 PL_parser->linestart = buf + linestart_pos;
3953 if (PL_parser->last_uni)
3954 PL_parser->last_uni = buf + last_uni_pos;
3955 if (PL_parser->last_lop)
3956 PL_parser->last_lop = buf + last_lop_pos;
3957 SvLEN(linestr) = SvCUR(linestr);
3958 SvCUR(linestr) = s-SvPVX(linestr);
3959 PL_parser->filtered = 1;
3969 /* Delete most recently added instance of this filter function. */
3971 Perl_filter_del(pTHX_ filter_t funcp)
3975 PERL_ARGS_ASSERT_FILTER_DEL;
3978 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3979 FPTR2DPTR(void*, funcp)));
3981 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3983 /* if filter is on top of stack (usual case) just pop it off */
3984 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3985 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3986 sv_free(av_pop(PL_rsfp_filters));
3990 /* we need to search for the correct entry and clear it */
3991 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3995 /* Invoke the idxth filter function for the current rsfp. */
3996 /* maxlen 0 = read one text line */
3998 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4002 /* This API is bad. It should have been using unsigned int for maxlen.
4003 Not sure if we want to change the API, but if not we should sanity
4004 check the value here. */
4005 unsigned int correct_length = maxlen < 0 ? PERL_INT_MAX : maxlen;
4007 PERL_ARGS_ASSERT_FILTER_READ;
4009 if (!PL_parser || !PL_rsfp_filters)
4011 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4012 /* Provide a default input filter to make life easy. */
4013 /* Note that we append to the line. This is handy. */
4014 DEBUG_P(PerlIO_printf(Perl_debug_log,
4015 "filter_read %d: from rsfp\n", idx));
4016 if (correct_length) {
4019 const int old_len = SvCUR(buf_sv);
4021 /* ensure buf_sv is large enough */
4022 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4023 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4024 correct_length)) <= 0) {
4025 if (PerlIO_error(PL_rsfp))
4026 return -1; /* error */
4028 return 0 ; /* end of file */
4030 SvCUR_set(buf_sv, old_len + len) ;
4031 SvPVX(buf_sv)[old_len + len] = '\0';
4034 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4035 if (PerlIO_error(PL_rsfp))
4036 return -1; /* error */
4038 return 0 ; /* end of file */
4041 return SvCUR(buf_sv);
4043 /* Skip this filter slot if filter has been deleted */
4044 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4045 DEBUG_P(PerlIO_printf(Perl_debug_log,
4046 "filter_read %d: skipped (filter deleted)\n",
4048 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4050 if (SvTYPE(datasv) != SVt_PVIO) {
4051 if (correct_length) {
4053 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4054 if (!remainder) return 0; /* eof */
4055 if (correct_length > remainder) correct_length = remainder;
4056 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4057 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4060 const char *s = SvEND(datasv);
4061 const char *send = SvPVX(datasv) + SvLEN(datasv);
4069 if (s == send) return 0; /* eof */
4070 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4071 SvCUR_set(datasv, s-SvPVX(datasv));
4073 return SvCUR(buf_sv);
4075 /* Get function pointer hidden within datasv */
4076 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4077 DEBUG_P(PerlIO_printf(Perl_debug_log,
4078 "filter_read %d: via function %p (%s)\n",
4079 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4080 /* Call function. The function is expected to */
4081 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4082 /* Return: <0:error, =0:eof, >0:not eof */
4083 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4087 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4089 PERL_ARGS_ASSERT_FILTER_GETS;
4091 #ifdef PERL_CR_FILTER
4092 if (!PL_rsfp_filters) {
4093 filter_add(S_cr_textfilter,NULL);
4096 if (PL_rsfp_filters) {
4098 SvCUR_set(sv, 0); /* start with empty line */
4099 if (FILTER_READ(0, sv, 0) > 0)
4100 return ( SvPVX(sv) ) ;
4105 return (sv_gets(sv, PL_rsfp, append));
4109 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4113 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4115 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4119 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4120 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4122 return GvHV(gv); /* Foo:: */
4125 /* use constant CLASS => 'MyClass' */
4126 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4127 if (gv && GvCV(gv)) {
4128 SV * const sv = cv_const_sv(GvCV(gv));
4130 pkgname = SvPV_const(sv, len);
4133 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4138 S_tokenize_use(pTHX_ int is_use, char *s) {
4139 PERL_ARGS_ASSERT_TOKENIZE_USE;
4141 if (PL_expect != XSTATE)
4142 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4143 is_use ? "use" : "no"));
4146 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4147 s = force_version(s, TRUE);
4148 if (*s == ';' || *s == '}'
4149 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4150 NEXTVAL_NEXTTOKE.opval = NULL;
4153 else if (*s == 'v') {
4154 s = force_word(s,WORD,FALSE,TRUE);
4155 s = force_version(s, FALSE);
4159 s = force_word(s,WORD,FALSE,TRUE);
4160 s = force_version(s, FALSE);
4162 pl_yylval.ival = is_use;
4166 static const char* const exp_name[] =
4167 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4168 "ATTRTERM", "TERMBLOCK", "XBLOCKTERM", "POSTDEREF",
4173 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4175 S_word_takes_any_delimeter(char *p, STRLEN len)
4177 return (len == 1 && strchr("msyq", p[0])) ||
4179 (p[0] == 't' && p[1] == 'r') ||
4180 (p[0] == 'q' && strchr("qwxr", p[1]))));
4184 S_check_scalar_slice(pTHX_ char *s)
4187 while (*s == ' ' || *s == '\t') s++;
4188 if (*s == 'q' && s[1] == 'w'
4189 && !isWORDCHAR_lazy_if(s+2,UTF))
4191 while (*s && (isWORDCHAR_lazy_if(s,UTF) || strchr(" \t$#+-'\"", *s)))
4192 s += UTF ? UTF8SKIP(s) : 1;
4193 if (*s == '}' || *s == ']')
4194 pl_yylval.ival = OPpSLICEWARNING;
4200 Works out what to call the token just pulled out of the input
4201 stream. The yacc parser takes care of taking the ops we return and
4202 stitching them into a tree.
4205 The type of the next token
4208 Switch based on the current state:
4209 - if we already built the token before, use it
4210 - if we have a case modifier in a string, deal with that
4211 - handle other cases of interpolation inside a string
4212 - scan the next line if we are inside a format
4213 In the normal state switch on the next character:
4215 if alphabetic, go to key lookup
4216 unrecoginized character - croak
4217 - 0/4/26: handle end-of-line or EOF
4218 - cases for whitespace
4219 - \n and #: handle comments and line numbers
4220 - various operators, brackets and sigils
4223 - 'v': vstrings (or go to key lookup)
4224 - 'x' repetition operator (or go to key lookup)
4225 - other ASCII alphanumerics (key lookup begins here):
4228 scan built-in keyword (but do nothing with it yet)
4229 check for statement label
4230 check for lexical subs
4231 goto just_a_word if there is one
4232 see whether built-in keyword is overridden
4233 switch on keyword number:
4234 - default: just_a_word:
4235 not a built-in keyword; handle bareword lookup
4236 disambiguate between method and sub call
4237 fall back to bareword
4238 - cases for built-in keywords
4246 char *s = PL_bufptr;
4250 const bool saw_infix_sigil = cBOOL(PL_parser->saw_infix_sigil);
4254 /* orig_keyword, gvp, and gv are initialized here because
4255 * jump to the label just_a_word_zero can bypass their
4256 * initialization later. */
4257 I32 orig_keyword = 0;
4262 SV* tmp = newSVpvs("");
4263 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4264 (IV)CopLINE(PL_curcop),
4265 lex_state_names[PL_lex_state],
4266 exp_name[PL_expect],
4267 pv_display(tmp, s, strlen(s), 0, 60));
4271 switch (PL_lex_state) {
4273 case LEX_INTERPNORMAL:
4276 /* when we've already built the next token, just pull it out of the queue */
4279 pl_yylval = PL_nextval[PL_nexttoke];
4281 PL_lex_state = PL_lex_defer;
4282 PL_lex_defer = LEX_NORMAL;
4286 next_type = PL_nexttype[PL_nexttoke];
4287 if (next_type & (7<<24)) {
4288 if (next_type & (1<<24)) {
4289 if (PL_lex_brackets > 100)
4290 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4291 PL_lex_brackstack[PL_lex_brackets++] =
4292 (char) ((next_type >> 16) & 0xff);
4294 if (next_type & (2<<24))
4295 PL_lex_allbrackets++;
4296 if (next_type & (4<<24))
4297 PL_lex_allbrackets--;
4298 next_type &= 0xffff;
4300 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4303 /* interpolated case modifiers like \L \U, including \Q and \E.
4304 when we get here, PL_bufptr is at the \
4306 case LEX_INTERPCASEMOD:
4308 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4310 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4311 PL_bufptr, PL_bufend, *PL_bufptr);
4313 /* handle \E or end of string */
4314 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4316 if (PL_lex_casemods) {
4317 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4318 PL_lex_casestack[PL_lex_casemods] = '\0';
4320 if (PL_bufptr != PL_bufend
4321 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4322 || oldmod == 'F')) {
4324 PL_lex_state = LEX_INTERPCONCAT;
4326 PL_lex_allbrackets--;
4329 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4330 /* Got an unpaired \E */
4331 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4332 "Useless use of \\E");
4334 if (PL_bufptr != PL_bufend)
4336 PL_lex_state = LEX_INTERPCONCAT;
4340 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4341 "### Saw case modifier\n"); });
4343 if (s[1] == '\\' && s[2] == 'E') {
4345 PL_lex_state = LEX_INTERPCONCAT;
4350 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4351 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4352 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4353 (strchr(PL_lex_casestack, 'L')
4354 || strchr(PL_lex_casestack, 'U')
4355 || strchr(PL_lex_casestack, 'F'))) {
4356 PL_lex_casestack[--PL_lex_casemods] = '\0';
4357 PL_lex_allbrackets--;
4360 if (PL_lex_casemods > 10)
4361 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4362 PL_lex_casestack[PL_lex_casemods++] = *s;
4363 PL_lex_casestack[PL_lex_casemods] = '\0';
4364 PL_lex_state = LEX_INTERPCONCAT;
4365 NEXTVAL_NEXTTOKE.ival = 0;
4366 force_next((2<<24)|'(');
4368 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4370 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4372 NEXTVAL_NEXTTOKE.ival = OP_LC;
4374 NEXTVAL_NEXTTOKE.ival = OP_UC;
4376 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4378 NEXTVAL_NEXTTOKE.ival = OP_FC;
4380 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4384 if (PL_lex_starts) {
4387 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4388 if (PL_lex_casemods == 1 && PL_lex_inpat)
4391 AopNOASSIGN(OP_CONCAT);
4397 case LEX_INTERPPUSH:
4398 return REPORT(sublex_push());
4400 case LEX_INTERPSTART:
4401 if (PL_bufptr == PL_bufend)
4402 return REPORT(sublex_done());
4403 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4404 "### Interpolated variable\n"); });
4406 /* for /@a/, we leave the joining for the regex engine to do
4407 * (unless we're within \Q etc) */
4408 PL_lex_dojoin = (*PL_bufptr == '@'
4409 && (!PL_lex_inpat || PL_lex_casemods));
4410 PL_lex_state = LEX_INTERPNORMAL;
4411 if (PL_lex_dojoin) {
4412 NEXTVAL_NEXTTOKE.ival = 0;
4414 force_ident("\"", '$');
4415 NEXTVAL_NEXTTOKE.ival = 0;
4417 NEXTVAL_NEXTTOKE.ival = 0;
4418 force_next((2<<24)|'(');
4419 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4422 /* Convert (?{...}) and friends to 'do {...}' */
4423 if (PL_lex_inpat && *PL_bufptr == '(') {
4424 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4426 if (*PL_bufptr != '{')
4428 PL_expect = XTERMBLOCK;
4432 if (PL_lex_starts++) {
4434 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4435 if (!PL_lex_casemods && PL_lex_inpat)
4438 AopNOASSIGN(OP_CONCAT);
4442 case LEX_INTERPENDMAYBE:
4443 if (intuit_more(PL_bufptr)) {
4444 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4450 if (PL_lex_dojoin) {
4451 const U8 dojoin_was = PL_lex_dojoin;
4452 PL_lex_dojoin = FALSE;
4453 PL_lex_state = LEX_INTERPCONCAT;
4454 PL_lex_allbrackets--;
4455 return REPORT(dojoin_was == 1 ? ')' : POSTJOIN);
4457 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4458 && SvEVALED(PL_lex_repl))
4460 if (PL_bufptr != PL_bufend)
4461 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4464 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4465 re_eval_str. If the here-doc body’s length equals the previous
4466 value of re_eval_start, re_eval_start will now be null. So
4467 check re_eval_str as well. */
4468 if (PL_parser->lex_shared->re_eval_start
4469 || PL_parser->lex_shared->re_eval_str) {
4471 if (*PL_bufptr != ')')
4472 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4474 /* having compiled a (?{..}) expression, return the original
4475 * text too, as a const */
4476 if (PL_parser->lex_shared->re_eval_str) {
4477 sv = PL_parser->lex_shared->re_eval_str;
4478 PL_parser->lex_shared->re_eval_str = NULL;
4480 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4481 SvPV_shrink_to_cur(sv);
4483 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4484 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4485 NEXTVAL_NEXTTOKE.opval =
4486 (OP*)newSVOP(OP_CONST, 0,
4489 PL_parser->lex_shared->re_eval_start = NULL;
4495 case LEX_INTERPCONCAT:
4497 if (PL_lex_brackets)
4498 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4499 (long) PL_lex_brackets);
4501 if (PL_bufptr == PL_bufend)
4502 return REPORT(sublex_done());
4504 /* m'foo' still needs to be parsed for possible (?{...}) */
4505 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4506 SV *sv = newSVsv(PL_linestr);
4508 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4512 s = scan_const(PL_bufptr);
4514 PL_lex_state = LEX_INTERPCASEMOD;
4516 PL_lex_state = LEX_INTERPSTART;
4519 if (s != PL_bufptr) {
4520 NEXTVAL_NEXTTOKE = pl_yylval;
4523 if (PL_lex_starts++) {
4524 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4525 if (!PL_lex_casemods && PL_lex_inpat)
4528 AopNOASSIGN(OP_CONCAT);
4538 s = scan_formline(PL_bufptr);
4539 if (!PL_lex_formbrack)
4548 /* We really do *not* want PL_linestr ever becoming a COW. */
4549 assert (!SvIsCOW(PL_linestr));
4551 PL_oldoldbufptr = PL_oldbufptr;
4553 PL_parser->saw_infix_sigil = 0;
4558 if (UTF ? isIDFIRST_utf8((U8*)s) : isALNUMC(*s))
4561 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4562 const char *c = UTF ? sv_uni_display(dsv, newSVpvn_flags(s,
4564 SVs_TEMP | SVf_UTF8),
4565 10, UNI_DISPLAY_ISPRINT)
4566 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4567 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4568 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4569 d = UTF ? (char *) utf8_hop((U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4573 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"UTF8f"<-- HERE near column %d", c,
4574 UTF8fARG(UTF, (s - d), d),
4579 goto fake_eof; /* emulate EOF on ^D or ^Z */
4581 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4584 if (PL_lex_brackets &&
4585 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4586 yyerror((const char *)
4588 ? "Format not terminated"
4589 : "Missing right curly or square bracket"));
4591 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4592 "### Tokener got EOF\n");
4596 if (s++ < PL_bufend)
4597 goto retry; /* ignore stray nulls */
4600 if (!PL_in_eval && !PL_preambled) {
4601 PL_preambled = TRUE;
4603 /* Generate a string of Perl code to load the debugger.
4604 * If PERL5DB is set, it will return the contents of that,
4605 * otherwise a compile-time require of perl5db.pl. */
4607 const char * const pdb = PerlEnv_getenv("PERL5DB");
4610 sv_setpv(PL_linestr, pdb);
4611 sv_catpvs(PL_linestr,";");
4613 SETERRNO(0,SS_NORMAL);
4614 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4616 PL_parser->preambling = CopLINE(PL_curcop);
4618 sv_setpvs(PL_linestr,"");
4619 if (PL_preambleav) {
4620 SV **svp = AvARRAY(PL_preambleav);
4621 SV **const end = svp + AvFILLp(PL_preambleav);
4623 sv_catsv(PL_linestr, *svp);
4625 sv_catpvs(PL_linestr, ";");
4627 sv_free(MUTABLE_SV(PL_preambleav));
4628 PL_preambleav = NULL;
4631 sv_catpvs(PL_linestr,
4632 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4633 if (PL_minus_n || PL_minus_p) {
4634 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4636 sv_catpvs(PL_linestr,"chomp;");
4639 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4640 || *PL_splitstr == '"')
4641 && strchr(PL_splitstr + 1, *PL_splitstr))
4642 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4644 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4645 bytes can be used as quoting characters. :-) */
4646 const char *splits = PL_splitstr;
4647 sv_catpvs(PL_linestr, "our @F=split(q\0");
4650 if (*splits == '\\')
4651 sv_catpvn(PL_linestr, splits, 1);
4652 sv_catpvn(PL_linestr, splits, 1);
4653 } while (*splits++);
4654 /* This loop will embed the trailing NUL of
4655 PL_linestr as the last thing it does before
4657 sv_catpvs(PL_linestr, ");");
4661 sv_catpvs(PL_linestr,"our @F=split(' ');");
4664 sv_catpvs(PL_linestr, "\n");
4665 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4666 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4667 PL_last_lop = PL_last_uni = NULL;
4668 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4669 update_debugger_info(PL_linestr, NULL, 0);
4674 bof = PL_rsfp ? TRUE : FALSE;
4677 fake_eof = LEX_FAKE_EOF;
4679 PL_bufptr = PL_bufend;
4680 COPLINE_INC_WITH_HERELINES;
4681 if (!lex_next_chunk(fake_eof)) {
4682 CopLINE_dec(PL_curcop);
4684 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4686 CopLINE_dec(PL_curcop);
4688 /* If it looks like the start of a BOM or raw UTF-16,
4689 * check if it in fact is. */
4690 if (bof && PL_rsfp &&
4692 *(U8*)s == BOM_UTF8_FIRST_BYTE ||
4695 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4696 bof = (offset == (Off_t)SvCUR(PL_linestr));
4697 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4698 /* offset may include swallowed CR */
4700 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4703 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4704 s = swallow_bom((U8*)s);
4707 if (PL_parser->in_pod) {
4708 /* Incest with pod. */
4709 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4710 sv_setpvs(PL_linestr, "");
4711 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4712 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4713 PL_last_lop = PL_last_uni = NULL;
4714 PL_parser->in_pod = 0;
4717 if (PL_rsfp || PL_parser->filtered)
4719 } while (PL_parser->in_pod);
4720 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4721 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4722 PL_last_lop = PL_last_uni = NULL;
4723 if (CopLINE(PL_curcop) == 1) {
4724 while (s < PL_bufend && isSPACE(*s))
4726 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4730 if (*s == '#' && *(s+1) == '!')
4732 #ifdef ALTERNATE_SHEBANG
4734 static char const as[] = ALTERNATE_SHEBANG;
4735 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4736 d = s + (sizeof(as) - 1);
4738 #endif /* ALTERNATE_SHEBANG */
4747 while (*d && !isSPACE(*d))
4751 #ifdef ARG_ZERO_IS_SCRIPT
4752 if (ipathend > ipath) {
4754 * HP-UX (at least) sets argv[0] to the script name,
4755 * which makes $^X incorrect. And Digital UNIX and Linux,
4756 * at least, set argv[0] to the basename of the Perl
4757 * interpreter. So, having found "#!", we'll set it right.
4759 SV* copfilesv = CopFILESV(PL_curcop);
4762 GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4764 assert(SvPOK(x) || SvGMAGICAL(x));
4765 if (sv_eq(x, copfilesv)) {
4766 sv_setpvn(x, ipath, ipathend - ipath);
4772 const char *bstart = SvPV_const(copfilesv, blen);
4773 const char * const lstart = SvPV_const(x, llen);
4775 bstart += blen - llen;
4776 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4777 sv_setpvn(x, ipath, ipathend - ipath);
4784 /* Anything to do if no copfilesv? */
4786 TAINT_NOT; /* $^X is always tainted, but that's OK */
4788 #endif /* ARG_ZERO_IS_SCRIPT */
4793 d = instr(s,"perl -");
4795 d = instr(s,"perl");
4797 /* avoid getting into infinite loops when shebang
4798 * line contains "Perl" rather than "perl" */
4800 for (d = ipathend-4; d >= ipath; --d) {
4801 if (isALPHA_FOLD_EQ(*d, 'p')
4802 && !ibcmp(d, "perl", 4))
4812 #ifdef ALTERNATE_SHEBANG
4814 * If the ALTERNATE_SHEBANG on this system starts with a
4815 * character that can be part of a Perl expression, then if
4816 * we see it but not "perl", we're probably looking at the
4817 * start of Perl code, not a request to hand off to some
4818 * other interpreter. Similarly, if "perl" is there, but
4819 * not in the first 'word' of the line, we assume the line
4820 * contains the start of the Perl program.
4822 if (d && *s != '#') {
4823 const char *c = ipath;
4824 while (*c && !strchr("; \t\r\n\f\v#", *c))
4827 d = NULL; /* "perl" not in first word; ignore */
4829 *s = '#'; /* Don't try to parse shebang line */
4831 #endif /* ALTERNATE_SHEBANG */
4836 !instr(s,"indir") &&
4837 instr(PL_origargv[0],"perl"))
4844 while (s < PL_bufend && isSPACE(*s))
4846 if (s < PL_bufend) {
4847 Newx(newargv,PL_origargc+3,char*);
4849 while (s < PL_bufend && !isSPACE(*s))
4852 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4855 newargv = PL_origargv;
4858 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4860 Perl_croak(aTHX_ "Can't exec %s", ipath);
4863 while (*d && !isSPACE(*d))
4865 while (SPACE_OR_TAB(*d))
4869 const bool switches_done = PL_doswitches;
4870 const U32 oldpdb = PL_perldb;
4871 const bool oldn = PL_minus_n;
4872 const bool oldp = PL_minus_p;
4876 bool baduni = FALSE;
4878 const char *d2 = d1 + 1;
4879 if (parse_unicode_opts((const char **)&d2)
4883 if (baduni || isALPHA_FOLD_EQ(*d1, 'M')) {
4884 const char * const m = d1;
4885 while (*d1 && !isSPACE(*d1))
4887 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4890 d1 = moreswitches(d1);
4892 if (PL_doswitches && !switches_done) {
4893 int argc = PL_origargc;
4894 char **argv = PL_origargv;
4897 } while (argc && argv[0][0] == '-' && argv[0][1]);
4898 init_argv_symbols(argc,argv);
4900 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4901 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4902 /* if we have already added "LINE: while (<>) {",
4903 we must not do it again */
4905 sv_setpvs(PL_linestr, "");
4906 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4907 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4908 PL_last_lop = PL_last_uni = NULL;
4909 PL_preambled = FALSE;
4910 if (PERLDB_LINE || PERLDB_SAVESRC)
4911 (void)gv_fetchfile(PL_origfilename);
4918 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4919 PL_lex_state = LEX_FORMLINE;
4920 NEXTVAL_NEXTTOKE.ival = 0;
4921 force_next(FORMRBRACK);
4926 #ifdef PERL_STRICT_CR
4927 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4929 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4931 case ' ': case '\t': case '\f': case 013:
4936 if (PL_lex_state != LEX_NORMAL ||
4937 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
4938 const bool in_comment = *s == '#';
4939 if (*s == '#' && s == PL_linestart && PL_in_eval
4940 && !PL_rsfp && !PL_parser->filtered) {
4941 /* handle eval qq[#line 1 "foo"\n ...] */
4942 CopLINE_dec(PL_curcop);
4946 while (d < PL_bufend && *d != '\n')
4950 else if (d > PL_bufend)
4951 /* Found by Ilya: feed random input to Perl. */
4952 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
4955 if (in_comment && d == PL_bufend
4956 && PL_lex_state == LEX_INTERPNORMAL
4957 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
4958 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
4961 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4962 PL_lex_state = LEX_FORMLINE;
4963 NEXTVAL_NEXTTOKE.ival = 0;
4964 force_next(FORMRBRACK);
4969 while (s < PL_bufend && *s != '\n')
4977 else if (s > PL_bufend)
4978 /* Found by Ilya: feed random input to Perl. */
4979 Perl_croak(aTHX_ "panic: input overflow");
4983 if (s[1] && isALPHA(s[1]) && !isWORDCHAR(s[2])) {
4991 while (s < PL_bufend && SPACE_OR_TAB(*s))
4994 if (strnEQ(s,"=>",2)) {
4995 s = force_word(PL_bufptr,WORD,FALSE,FALSE);
4996 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4997 OPERATOR('-'); /* unary minus */
5000 case 'r': ftst = OP_FTEREAD; break;
5001 case 'w': ftst = OP_FTEWRITE; break;
5002 case 'x': ftst = OP_FTEEXEC; break;
5003 case 'o': ftst = OP_FTEOWNED; break;
5004 case 'R': ftst = OP_FTRREAD; break;
5005 case 'W': ftst = OP_FTRWRITE; break;
5006 case 'X': ftst = OP_FTREXEC; break;
5007 case 'O': ftst = OP_FTROWNED; break;
5008 case 'e': ftst = OP_FTIS; break;
5009 case 'z': ftst = OP_FTZERO; break;
5010 case 's': ftst = OP_FTSIZE; break;
5011 case 'f': ftst = OP_FTFILE; break;
5012 case 'd': ftst = OP_FTDIR; break;
5013 case 'l': ftst = OP_FTLINK; break;
5014 case 'p': ftst = OP_FTPIPE; break;
5015 case 'S': ftst = OP_FTSOCK; break;
5016 case 'u': ftst = OP_FTSUID; break;
5017 case 'g': ftst = OP_FTSGID; break;
5018 case 'k': ftst = OP_FTSVTX; break;
5019 case 'b': ftst = OP_FTBLK; break;
5020 case 'c': ftst = OP_FTCHR; break;
5021 case 't': ftst = OP_FTTTY; break;
5022 case 'T': ftst = OP_FTTEXT; break;
5023 case 'B': ftst = OP_FTBINARY; break;
5024 case 'M': case 'A': case 'C':
5025 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5027 case 'M': ftst = OP_FTMTIME; break;
5028 case 'A': ftst = OP_FTATIME; break;
5029 case 'C': ftst = OP_FTCTIME; break;
5037 PL_last_uni = PL_oldbufptr;
5038 PL_last_lop_op = (OPCODE)ftst;
5039 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5040 "### Saw file test %c\n", (int)tmp);
5045 /* Assume it was a minus followed by a one-letter named
5046 * subroutine call (or a -bareword), then. */
5047 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5048 "### '-%c' looked like a file test but was not\n",
5055 const char tmp = *s++;
5058 if (PL_expect == XOPERATOR)
5063 else if (*s == '>') {
5066 if (FEATURE_POSTDEREF_IS_ENABLED && (
5067 ((*s == '$' || *s == '&') && s[1] == '*')
5068 ||(*s == '$' && s[1] == '#' && s[2] == '*')
5069 ||((*s == '@' || *s == '%') && strchr("*[{", s[1]))
5070 ||(*s == '*' && (s[1] == '*' || s[1] == '{'))
5073 Perl_ck_warner_d(aTHX_
5074 packWARN(WARN_EXPERIMENTAL__POSTDEREF),
5075 "Postfix dereference is experimental"
5077 PL_expect = XPOSTDEREF;
5080 if (isIDFIRST_lazy_if(s,UTF)) {
5081 s = force_word(s,METHOD,FALSE,TRUE);
5089 if (PL_expect == XOPERATOR) {
5090 if (*s == '=' && !PL_lex_allbrackets &&
5091 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5098 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5100 OPERATOR('-'); /* unary minus */
5106 const char tmp = *s++;
5109 if (PL_expect == XOPERATOR)
5114 if (PL_expect == XOPERATOR) {
5115 if (*s == '=' && !PL_lex_allbrackets &&
5116 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5123 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5130 if (PL_expect == XPOSTDEREF) POSTDEREF('*');
5131 if (PL_expect != XOPERATOR) {
5132 s = scan_ident(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5133 PL_expect = XOPERATOR;
5134 force_ident(PL_tokenbuf, '*');
5142 if (*s == '=' && !PL_lex_allbrackets &&
5143 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5149 if (*s == '=' && !PL_lex_allbrackets &&
5150 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5154 PL_parser->saw_infix_sigil = 1;
5159 if (PL_expect == XOPERATOR) {
5160 if (s[1] == '=' && !PL_lex_allbrackets &&
5161 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5164 PL_parser->saw_infix_sigil = 1;
5167 else if (PL_expect == XPOSTDEREF) POSTDEREF('%');
5168 PL_tokenbuf[0] = '%';
5169 s = scan_ident(s, PL_tokenbuf + 1,
5170 sizeof PL_tokenbuf - 1, FALSE);
5172 if (!PL_tokenbuf[1]) {
5175 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5177 PL_tokenbuf[0] = '@';
5179 PL_expect = XOPERATOR;
5180 force_ident_maybe_lex('%');
5184 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5185 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5190 if (PL_lex_brackets > 100)
5191 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5192 PL_lex_brackstack[PL_lex_brackets++] = 0;
5193 PL_lex_allbrackets++;
5195 const char tmp = *s++;
5200 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5202 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5205 Perl_ck_warner_d(aTHX_
5206 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
5207 "Smartmatch is experimental");
5213 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5220 goto just_a_word_zero_gv;
5226 switch (PL_expect) {
5228 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5230 PL_bufptr = s; /* update in case we back off */
5233 "Use of := for an empty attribute list is not allowed");
5240 PL_expect = XTERMBLOCK;
5244 while (isIDFIRST_lazy_if(s,UTF)) {
5247 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5248 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5249 if (tmp < 0) tmp = -tmp;
5264 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5266 d = scan_str(d,TRUE,TRUE,FALSE,NULL);
5267 COPLINE_SET_FROM_MULTI_END;
5269 /* MUST advance bufptr here to avoid bogus
5270 "at end of line" context messages from yyerror().
5272 PL_bufptr = s + len;
5273 yyerror("Unterminated attribute parameter in attribute list");
5277 return REPORT(0); /* EOF indicator */
5281 sv_catsv(sv, PL_lex_stuff);
5282 attrs = op_append_elem(OP_LIST, attrs,
5283 newSVOP(OP_CONST, 0, sv));
5284 SvREFCNT_dec(PL_lex_stuff);
5285 PL_lex_stuff = NULL;
5288 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5290 if (PL_in_my == KEY_our) {
5291 deprecate(":unique");
5294 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5297 /* NOTE: any CV attrs applied here need to be part of
5298 the CVf_BUILTIN_ATTRS define in cv.h! */
5299 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5301 CvLVALUE_on(PL_compcv);
5303 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5305 deprecate(":locked");
5307 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5309 CvMETHOD_on(PL_compcv);
5311 /* After we've set the flags, it could be argued that
5312 we don't need to do the attributes.pm-based setting
5313 process, and shouldn't bother appending recognized
5314 flags. To experiment with that, uncomment the
5315 following "else". (Note that's already been
5316 uncommented. That keeps the above-applied built-in
5317 attributes from being intercepted (and possibly
5318 rejected) by a package's attribute routines, but is
5319 justified by the performance win for the common case
5320 of applying only built-in attributes.) */
5322 attrs = op_append_elem(OP_LIST, attrs,
5323 newSVOP(OP_CONST, 0,
5327 if (*s == ':' && s[1] != ':')
5330 break; /* require real whitespace or :'s */
5331 /* XXX losing whitespace on sequential attributes here */
5334 if (*s != ';' && *s != '}' &&
5335 !(PL_expect == XOPERATOR
5336 ? (*s == '=' || *s == ')')
5337 : (*s == '{' || *s == '('))) {
5338 const char q = ((*s == '\'') ? '"' : '\'');
5339 /* If here for an expression, and parsed no attrs, back
5341 if (PL_expect == XOPERATOR && !attrs) {
5345 /* MUST advance bufptr here to avoid bogus "at end of line"
5346 context messages from yyerror().
5349 yyerror( (const char *)
5351 ? Perl_form(aTHX_ "Invalid separator character "
5352 "%c%c%c in attribute list", q, *s, q)
5353 : "Unterminated attribute list" ) );
5361 NEXTVAL_NEXTTOKE.opval = attrs;
5367 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5371 PL_lex_allbrackets--;
5375 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5376 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5380 PL_lex_allbrackets++;
5383 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5390 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5393 PL_lex_allbrackets--;
5399 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5402 if (PL_lex_brackets <= 0)
5403 /* diag_listed_as: Unmatched right %s bracket */
5404 yyerror("Unmatched right square bracket");
5407 PL_lex_allbrackets--;
5408 if (PL_lex_state == LEX_INTERPNORMAL) {
5409 if (PL_lex_brackets == 0) {
5410 if (*s == '-' && s[1] == '>')
5411 PL_lex_state = LEX_INTERPENDMAYBE;
5412 else if (*s != '[' && *s != '{')
5413 PL_lex_state = LEX_INTERPEND;
5420 if (PL_lex_brackets > 100) {
5421 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5423 switch (PL_expect) {
5425 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5426 PL_lex_allbrackets++;
5427 OPERATOR(HASHBRACK);
5429 while (s < PL_bufend && SPACE_OR_TAB(*s))
5432 PL_tokenbuf[0] = '\0';
5433 if (d < PL_bufend && *d == '-') {
5434 PL_tokenbuf[0] = '-';
5436 while (d < PL_bufend && SPACE_OR_TAB(*d))
5439 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5440 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5442 while (d < PL_bufend && SPACE_OR_TAB(*d))
5445 const char minus = (PL_tokenbuf[0] == '-');
5446 s = force_word(s + minus, WORD, FALSE, TRUE);
5454 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5455 PL_lex_allbrackets++;
5460 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5461 PL_lex_allbrackets++;
5465 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5466 PL_lex_allbrackets++;
5471 if (PL_oldoldbufptr == PL_last_lop)
5472 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5474 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5475 PL_lex_allbrackets++;
5478 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5480 /* This hack is to get the ${} in the message. */
5482 yyerror("syntax error");
5485 OPERATOR(HASHBRACK);
5487 if (PL_expect == XREF && PL_oldoldbufptr != PL_last_lop) {
5488 /* ${...} or @{...} etc., but not print {...} */
5492 /* This hack serves to disambiguate a pair of curlies
5493 * as being a block or an anon hash. Normally, expectation
5494 * determines that, but in cases where we're not in a
5495 * position to expect anything in particular (like inside
5496 * eval"") we have to resolve the ambiguity. This code
5497 * covers the case where the first term in the curlies is a
5498 * quoted string. Most other cases need to be explicitly
5499 * disambiguated by prepending a "+" before the opening
5500 * curly in order to force resolution as an anon hash.
5502 * XXX should probably propagate the outer expectation
5503 * into eval"" to rely less on this hack, but that could
5504 * potentially break current behavior of eval"".
5508 if (*s == '\'' || *s == '"' || *s == '`') {
5509 /* common case: get past first string, handling escapes */
5510 for (t++; t < PL_bufend && *t != *s;)
5515 else if (*s == 'q') {
5518 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5519 && !isWORDCHAR(*t))))
5521 /* skip q//-like construct */
5523 char open, close, term;
5526 while (t < PL_bufend && isSPACE(*t))
5528 /* check for q => */
5529 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5530 OPERATOR(HASHBRACK);
5534 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5538 for (t++; t < PL_bufend; t++) {
5539 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5541 else if (*t == open)
5545 for (t++; t < PL_bufend; t++) {
5546 if (*t == '\\' && t+1 < PL_bufend)
5548 else if (*t == close && --brackets <= 0)
5550 else if (*t == open)
5557 /* skip plain q word */
5558 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5561 else if (isWORDCHAR_lazy_if(t,UTF)) {
5563 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5566 while (t < PL_bufend && isSPACE(*t))
5568 /* if comma follows first term, call it an anon hash */
5569 /* XXX it could be a comma expression with loop modifiers */
5570 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5571 || (*t == '=' && t[1] == '>')))
5572 OPERATOR(HASHBRACK);
5573 if (PL_expect == XREF)
5576 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5582 pl_yylval.ival = CopLINE(PL_curcop);
5583 if (isSPACE(*s) || *s == '#')
5584 PL_copline = NOLINE; /* invalidate current command line number */
5585 TOKEN(formbrack ? '=' : '{');
5587 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5591 if (PL_lex_brackets <= 0)
5592 /* diag_listed_as: Unmatched right %s bracket */
5593 yyerror("Unmatched right curly bracket");
5595 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5596 PL_lex_allbrackets--;
5597 if (PL_lex_state == LEX_INTERPNORMAL) {
5598 if (PL_lex_brackets == 0) {
5599 if (PL_expect & XFAKEBRACK) {
5600 PL_expect &= XENUMMASK;
5601 PL_lex_state = LEX_INTERPEND;
5603 return yylex(); /* ignore fake brackets */
5605 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5606 && SvEVALED(PL_lex_repl))
5607 PL_lex_state = LEX_INTERPEND;
5608 else if (*s == '-' && s[1] == '>')
5609 PL_lex_state = LEX_INTERPENDMAYBE;
5610 else if (*s != '[' && *s != '{')
5611 PL_lex_state = LEX_INTERPEND;
5614 if (PL_expect & XFAKEBRACK) {
5615 PL_expect &= XENUMMASK;
5617 return yylex(); /* ignore fake brackets */
5619 force_next(formbrack ? '.' : '}');
5620 if (formbrack) LEAVE;
5621 if (formbrack == 2) { /* means . where arguments were expected */
5627 if (PL_expect == XPOSTDEREF) POSTDEREF('&');
5630 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5631 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5638 if (PL_expect == XOPERATOR) {
5639 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5640 && isIDFIRST_lazy_if(s,UTF))
5642 CopLINE_dec(PL_curcop);
5643 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5644 CopLINE_inc(PL_curcop);
5646 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5647 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5651 PL_parser->saw_infix_sigil = 1;
5655 PL_tokenbuf[0] = '&';
5656 s = scan_ident(s - 1, PL_tokenbuf + 1,
5657 sizeof PL_tokenbuf - 1, TRUE);
5658 if (PL_tokenbuf[1]) {
5659 PL_expect = XOPERATOR;
5660 force_ident_maybe_lex('&');
5664 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5670 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5671 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5678 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5679 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5687 const char tmp = *s++;
5689 if (!PL_lex_allbrackets &&
5690 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5697 if (!PL_lex_allbrackets &&
5698 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5706 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5707 && strchr("+-*/%.^&|<",tmp))
5708 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5709 "Reversed %c= operator",(int)tmp);
5711 if (PL_expect == XSTATE && isALPHA(tmp) &&
5712 (s == PL_linestart+1 || s[-2] == '\n') )
5714 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
5715 || PL_lex_state != LEX_NORMAL) {
5720 if (strnEQ(s,"=cut",4)) {
5734 PL_parser->in_pod = 1;
5738 if (PL_expect == XBLOCK) {
5740 #ifdef PERL_STRICT_CR
5741 while (SPACE_OR_TAB(*t))
5743 while (SPACE_OR_TAB(*t) || *t == '\r')
5746 if (*t == '\n' || *t == '#') {
5749 SAVEI8(PL_parser->form_lex_state);
5750 SAVEI32(PL_lex_formbrack);
5751 PL_parser->form_lex_state = PL_lex_state;
5752 PL_lex_formbrack = PL_lex_brackets + 1;
5756 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5765 const char tmp = *s++;
5767 /* was this !=~ where !~ was meant?
5768 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5770 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5771 const char *t = s+1;
5773 while (t < PL_bufend && isSPACE(*t))
5776 if (*t == '/' || *t == '?' ||
5777 ((*t == 'm' || *t == 's' || *t == 'y')
5778 && !isWORDCHAR(t[1])) ||
5779 (*t == 't' && t[1] == 'r' && !isWORDCHAR(t[2])))
5780 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5781 "!=~ should be !~");
5783 if (!PL_lex_allbrackets &&
5784 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5796 if (PL_expect != XOPERATOR) {
5797 if (s[1] != '<' && !strchr(s,'>'))
5800 s = scan_heredoc(s);
5802 s = scan_inputsymbol(s);
5803 PL_expect = XOPERATOR;
5804 TOKEN(sublex_start());
5810 if (*s == '=' && !PL_lex_allbrackets &&
5811 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5815 SHop(OP_LEFT_SHIFT);
5820 if (!PL_lex_allbrackets &&
5821 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5828 if (!PL_lex_allbrackets &&
5829 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5837 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5845 const char tmp = *s++;
5847 if (*s == '=' && !PL_lex_allbrackets &&
5848 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5852 SHop(OP_RIGHT_SHIFT);
5854 else if (tmp == '=') {
5855 if (!PL_lex_allbrackets &&
5856 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5864 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5873 if (PL_expect == XOPERATOR) {
5874 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5875 return deprecate_commaless_var_list();
5878 else if (PL_expect == XPOSTDEREF) {
5881 POSTDEREF(DOLSHARP);
5886 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
5887 PL_tokenbuf[0] = '@';
5888 s = scan_ident(s + 1, PL_tokenbuf + 1,
5889 sizeof PL_tokenbuf - 1, FALSE);
5890 if (PL_expect == XOPERATOR)
5891 no_op("Array length", s);
5892 if (!PL_tokenbuf[1])
5894 PL_expect = XOPERATOR;
5895 force_ident_maybe_lex('#');
5899 PL_tokenbuf[0] = '$';
5900 s = scan_ident(s, PL_tokenbuf + 1,
5901 sizeof PL_tokenbuf - 1, FALSE);
5902 if (PL_expect == XOPERATOR)
5904 if (!PL_tokenbuf[1]) {
5906 yyerror("Final $ should be \\$ or $name");
5912 const char tmp = *s;
5913 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5916 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5917 && intuit_more(s)) {
5919 PL_tokenbuf[0] = '@';
5920 if (ckWARN(WARN_SYNTAX)) {
5923 while (isSPACE(*t) || isWORDCHAR_lazy_if(t,UTF) || *t == '$')
5926 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5927 while (t < PL_bufend && *t != ']')
5929 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5930 "Multidimensional syntax %.*s not supported",
5931 (int)((t - PL_bufptr) + 1), PL_bufptr);
5935 else if (*s == '{') {
5937 PL_tokenbuf[0] = '%';
5938 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5939 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5941 char tmpbuf[sizeof PL_tokenbuf];
5944 } while (isSPACE(*t));
5945 if (isIDFIRST_lazy_if(t,UTF)) {
5947 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5952 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
5953 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5954 "You need to quote \"%"UTF8f"\"",
5955 UTF8fARG(UTF, len, tmpbuf));
5961 PL_expect = XOPERATOR;
5962 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5963 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5964 if (!islop || PL_last_lop_op == OP_GREPSTART)
5965 PL_expect = XOPERATOR;
5966 else if (strchr("$@\"'`q", *s))
5967 PL_expect = XTERM; /* e.g. print $fh "foo" */
5968 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5969 PL_expect = XTERM; /* e.g. print $fh &sub */
5970 else if (isIDFIRST_lazy_if(s,UTF)) {
5971 char tmpbuf[sizeof PL_tokenbuf];
5973 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5974 if ((t2 = keyword(tmpbuf, len, 0))) {
5975 /* binary operators exclude handle interpretations */
5987 PL_expect = XTERM; /* e.g. print $fh length() */
5992 PL_expect = XTERM; /* e.g. print $fh subr() */
5995 else if (isDIGIT(*s))
5996 PL_expect = XTERM; /* e.g. print $fh 3 */
5997 else if (*s == '.' && isDIGIT(s[1]))
5998 PL_expect = XTERM; /* e.g. print $fh .3 */
5999 else if ((*s == '?' || *s == '-' || *s == '+')
6000 && !isSPACE(s[1]) && s[1] != '=')
6001 PL_expect = XTERM; /* e.g. print $fh -1 */
6002 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6004 PL_expect = XTERM; /* e.g. print $fh /.../
6005 XXX except DORDOR operator
6007 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6009 PL_expect = XTERM; /* print $fh <<"EOF" */
6012 force_ident_maybe_lex('$');
6016 if (PL_expect == XOPERATOR)
6018 else if (PL_expect == XPOSTDEREF) POSTDEREF('@');
6019 PL_tokenbuf[0] = '@';
6020 s = scan_ident(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6022 if (!PL_tokenbuf[1]) {
6025 if (PL_lex_state == LEX_NORMAL)
6027 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6029 PL_tokenbuf[0] = '%';
6031 /* Warn about @ where they meant $. */
6032 if (*s == '[' || *s == '{') {
6033 if (ckWARN(WARN_SYNTAX)) {
6034 S_check_scalar_slice(aTHX_ s);
6038 PL_expect = XOPERATOR;
6039 force_ident_maybe_lex('@');
6042 case '/': /* may be division, defined-or, or pattern */
6043 if ((PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR) && s[1] == '/') {
6044 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6045 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6050 else if (PL_expect == XOPERATOR) {
6052 if (*s == '=' && !PL_lex_allbrackets &&
6053 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6060 /* Disable warning on "study /blah/" */
6061 if (PL_oldoldbufptr == PL_last_uni
6062 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6063 || memNE(PL_last_uni, "study", 5)
6064 || isWORDCHAR_lazy_if(PL_last_uni+5,UTF)
6067 s = scan_pat(s,OP_MATCH);
6068 TERM(sublex_start());
6071 case '?': /* conditional */
6073 if (!PL_lex_allbrackets &&
6074 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6078 PL_lex_allbrackets++;
6082 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6083 #ifdef PERL_STRICT_CR
6086 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6088 && (s == PL_linestart || s[-1] == '\n') )
6091 formbrack = 2; /* dot seen where arguments expected */
6094 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6098 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6101 if (!PL_lex_allbrackets &&
6102 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6109 pl_yylval.ival = OPf_SPECIAL;
6115 if (*s == '=' && !PL_lex_allbrackets &&
6116 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6123 case '0': case '1': case '2': case '3': case '4':
6124 case '5': case '6': case '7': case '8': case '9':
6125 s = scan_num(s, &pl_yylval);
6126 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6127 if (PL_expect == XOPERATOR)
6132 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6135 COPLINE_SET_FROM_MULTI_END;
6136 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6137 if (PL_expect == XOPERATOR) {
6138 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6139 return deprecate_commaless_var_list();
6144 pl_yylval.ival = OP_CONST;
6145 TERM(sublex_start());
6148 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6151 printbuf("### Saw string before %s\n", s);
6153 PerlIO_printf(Perl_debug_log,
6154 "### Saw unterminated string\n");
6156 if (PL_expect == XOPERATOR) {
6157 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6158 return deprecate_commaless_var_list();
6165 pl_yylval.ival = OP_CONST;
6166 /* FIXME. I think that this can be const if char *d is replaced by
6167 more localised variables. */
6168 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6169 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6170 pl_yylval.ival = OP_STRINGIFY;
6174 if (pl_yylval.ival == OP_CONST)
6175 COPLINE_SET_FROM_MULTI_END;
6176 TERM(sublex_start());
6179 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6180 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6181 if (PL_expect == XOPERATOR)
6182 no_op("Backticks",s);
6185 pl_yylval.ival = OP_BACKTICK;
6186 TERM(sublex_start());
6190 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6192 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6194 if (PL_expect == XOPERATOR)
6195 no_op("Backslash",s);
6199 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6200 char *start = s + 2;
6201 while (isDIGIT(*start) || *start == '_')
6203 if (*start == '.' && isDIGIT(start[1])) {
6204 s = scan_num(s, &pl_yylval);
6207 else if ((*start == ':' && start[1] == ':')
6208 || (PL_expect == XSTATE && *start == ':'))
6210 else if (PL_expect == XSTATE) {
6212 while (d < PL_bufend && isSPACE(*d)) d++;
6213 if (*d == ':') goto keylookup;
6215 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6216 if (!isALPHA(*start) && (PL_expect == XTERM
6217 || PL_expect == XREF || PL_expect == XSTATE
6218 || PL_expect == XTERMORDORDOR)) {
6219 GV *const gv = gv_fetchpvn_flags(s, start - s,
6220 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6222 s = scan_num(s, &pl_yylval);
6229 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6282 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6284 /* Some keywords can be followed by any delimiter, including ':' */
6285 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6287 /* x::* is just a word, unless x is "CORE" */
6288 if (!anydelim && *s == ':' && s[1] == ':') {
6289 if (strEQ(PL_tokenbuf, "CORE")) goto case_KEY_CORE;
6294 while (d < PL_bufend && isSPACE(*d))
6295 d++; /* no comments skipped here, or s### is misparsed */
6297 /* Is this a word before a => operator? */
6298 if (*d == '=' && d[1] == '>') {
6302 = (OP*)newSVOP(OP_CONST, 0,
6303 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6304 pl_yylval.opval->op_private = OPpCONST_BARE;
6308 /* Check for plugged-in keyword */
6312 char *saved_bufptr = PL_bufptr;
6314 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6316 if (result == KEYWORD_PLUGIN_DECLINE) {
6317 /* not a plugged-in keyword */
6318 PL_bufptr = saved_bufptr;
6319 } else if (result == KEYWORD_PLUGIN_STMT) {
6320 pl_yylval.opval = o;
6322 if (!PL_nexttoke) PL_expect = XSTATE;
6323 return REPORT(PLUGSTMT);
6324 } else if (result == KEYWORD_PLUGIN_EXPR) {
6325 pl_yylval.opval = o;
6327 if (!PL_nexttoke) PL_expect = XOPERATOR;
6328 return REPORT(PLUGEXPR);
6330 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6335 /* Check for built-in keyword */
6336 tmp = keyword(PL_tokenbuf, len, 0);
6338 /* Is this a label? */
6339 if (!anydelim && PL_expect == XSTATE
6340 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6342 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6343 pl_yylval.pval[len] = '\0';
6344 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6349 /* Check for lexical sub */
6350 if (PL_expect != XOPERATOR) {
6351 char tmpbuf[sizeof PL_tokenbuf + 1];
6353 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6354 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6355 if (off != NOT_IN_PAD) {
6356 assert(off); /* we assume this is boolean-true below */
6357 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6358 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6359 HEK * const stashname = HvNAME_HEK(stash);
6360 sv = newSVhek(stashname);
6361 sv_catpvs(sv, "::");
6362 sv_catpvn_flags(sv, PL_tokenbuf, len,
6363 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6364 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6374 rv2cv_op = newOP(OP_PADANY, 0);
6375 rv2cv_op->op_targ = off;
6376 cv = find_lexical_cv(off);
6384 if (tmp < 0) { /* second-class keyword? */
6385 GV *ogv = NULL; /* override (winner) */
6386 GV *hgv = NULL; /* hidden (loser) */
6387 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6389 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6390 (UTF ? SVf_UTF8 : 0)|GV_NOTQUAL,
6394 if (GvIMPORTED_CV(gv))
6396 else if (! CvMETHOD(cv))
6400 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6404 ? GvCVu(gv) && GvIMPORTED_CV(gv)
6405 : SvPCS_IMPORTED(gv)
6406 && (gv_init(gv, PL_globalstash, PL_tokenbuf,
6415 tmp = 0; /* overridden by import or by GLOBAL */
6418 && -tmp==KEY_lock /* XXX generalizable kludge */
6421 tmp = 0; /* any sub overrides "weak" keyword */
6423 else { /* no override */
6425 if (tmp == KEY_dump) {
6426 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6427 "dump() better written as CORE::dump()");
6431 if (hgv && tmp != KEY_x) /* never ambiguous */
6432 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6433 "Ambiguous call resolved as CORE::%s(), "
6434 "qualify as such or use &",
6439 if (tmp && tmp != KEY___DATA__ && tmp != KEY___END__
6440 && (!anydelim || *s != '#')) {
6441 /* no override, and not s### either; skipspace is safe here
6442 * check for => on following line */
6444 STRLEN bufoff = PL_bufptr - SvPVX(PL_linestr);
6445 STRLEN soff = s - SvPVX(PL_linestr);
6446 s = skipspace_flags(s, LEX_NO_INCLINE);
6447 arrow = *s == '=' && s[1] == '>';
6448 PL_bufptr = SvPVX(PL_linestr) + bufoff;
6449 s = SvPVX(PL_linestr) + soff;
6457 default: /* not a keyword */
6458 /* Trade off - by using this evil construction we can pull the
6459 variable gv into the block labelled keylookup. If not, then
6460 we have to give it function scope so that the goto from the
6461 earlier ':' case doesn't bypass the initialisation. */
6463 just_a_word_zero_gv:
6475 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6476 const char penultchar =
6477 lastchar && PL_bufptr - 2 >= PL_linestart
6482 /* Get the rest if it looks like a package qualifier */
6484 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6486 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6489 Perl_croak(aTHX_ "Bad name after %"UTF8f"%s",
6490 UTF8fARG(UTF, len, PL_tokenbuf),
6491 *s == '\'' ? "'" : "::");
6496 if (PL_expect == XOPERATOR) {
6497 if (PL_bufptr == PL_linestart) {
6498 CopLINE_dec(PL_curcop);
6499 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6500 CopLINE_inc(PL_curcop);
6503 no_op("Bareword",s);
6506 /* Look for a subroutine with this name in current package,
6507 unless this is a lexical sub, or name is "Foo::",
6508 in which case Foo is a bareword
6509 (and a package name). */
6512 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6514 if (ckWARN(WARN_BAREWORD)
6515 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6516 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6517 "Bareword \"%"UTF8f"\" refers to nonexistent package",
6518 UTF8fARG(UTF, len, PL_tokenbuf));
6520 PL_tokenbuf[len] = '\0';
6526 /* Mustn't actually add anything to a symbol table.
6527 But also don't want to "initialise" any placeholder
6528 constants that might already be there into full
6529 blown PVGVs with attached PVCV. */
6530 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6531 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6537 /* if we saw a global override before, get the right name */
6540 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6541 len ? len : strlen(PL_tokenbuf));
6543 SV * const tmp_sv = sv;
6544 sv = newSVpvs("CORE::GLOBAL::");
6545 sv_catsv(sv, tmp_sv);
6546 SvREFCNT_dec(tmp_sv);
6550 /* Presume this is going to be a bareword of some sort. */
6552 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6553 pl_yylval.opval->op_private = OPpCONST_BARE;
6555 /* And if "Foo::", then that's what it certainly is. */
6561 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6562 const_op->op_private = OPpCONST_BARE;
6564 newCVREF(OPpMAY_RETURN_CONSTANT<<8, const_op);
6566 ? isGV(gv) ? GvCV(gv) : (CV *)gv
6567 : rv2cv_op_cv(rv2cv_op, RV2CVOPCV_RETURN_STUB);
6570 /* See if it's the indirect object for a list operator. */
6572 if (PL_oldoldbufptr &&
6573 PL_oldoldbufptr < PL_bufptr &&
6574 (PL_oldoldbufptr == PL_last_lop
6575 || PL_oldoldbufptr == PL_last_uni) &&
6576 /* NO SKIPSPACE BEFORE HERE! */
6577 (PL_expect == XREF ||
6578 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6580 bool immediate_paren = *s == '(';
6582 /* (Now we can afford to cross potential line boundary.) */
6583 s = SKIPSPACE2(s,nextPL_nextwhite);
6585 /* Two barewords in a row may indicate method call. */
6587 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6588 (tmp = intuit_method(s, gv, cv))) {
6590 if (tmp == METHOD && !PL_lex_allbrackets &&
6591 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6592 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6596 /* If not a declared subroutine, it's an indirect object. */
6597 /* (But it's an indir obj regardless for sort.) */
6598 /* Also, if "_" follows a filetest operator, it's a bareword */
6601 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6603 (PL_last_lop_op != OP_MAPSTART &&
6604 PL_last_lop_op != OP_GREPSTART))))
6605 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6606 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6609 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6614 PL_expect = XOPERATOR;
6617 /* Is this a word before a => operator? */
6618 if (*s == '=' && s[1] == '>' && !pkgname) {
6621 /* This is our own scalar, created a few lines above,
6623 SvREADONLY_off(cSVOPx(pl_yylval.opval)->op_sv);
6624 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6625 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6626 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6627 SvREADONLY_on(cSVOPx(pl_yylval.opval)->op_sv);
6631 /* If followed by a paren, it's certainly a subroutine. */
6636 while (SPACE_OR_TAB(*d))
6638 if (*d == ')' && (sv = cv_const_sv_or_av(cv))) {
6643 NEXTVAL_NEXTTOKE.opval =
6644 off ? rv2cv_op : pl_yylval.opval;
6646 op_free(pl_yylval.opval), force_next(PRIVATEREF);
6647 else op_free(rv2cv_op), force_next(WORD);
6652 /* If followed by var or block, call it a method (unless sub) */
6654 if ((*s == '$' || *s == '{') && !cv) {
6656 PL_last_lop = PL_oldbufptr;
6657 PL_last_lop_op = OP_METHOD;
6658 if (!PL_lex_allbrackets &&
6659 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6660 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6661 PL_expect = XBLOCKTERM;
6663 return REPORT(METHOD);
6666 /* If followed by a bareword, see if it looks like indir obj. */
6669 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6670 && (tmp = intuit_method(s, gv, cv))) {
6672 if (tmp == METHOD && !PL_lex_allbrackets &&
6673 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6674 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6678 /* Not a method, so call it a subroutine (if defined) */
6682 if (lastchar == '-' && penultchar != '-') {
6683 const STRLEN l = len ? len : strlen(PL_tokenbuf);
6684 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6685 "Ambiguous use of -%"UTF8f" resolved as -&%"UTF8f"()",
6686 UTF8fARG(UTF, l, PL_tokenbuf),
6687 UTF8fARG(UTF, l, PL_tokenbuf));
6689 /* Check for a constant sub */
6690 if ((sv = cv_const_sv_or_av(cv))) {
6693 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6694 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6695 if (SvTYPE(sv) == SVt_PVAV)
6696 pl_yylval.opval = newUNOP(OP_RV2AV, OPf_PARENS,
6699 pl_yylval.opval->op_private = 0;
6700 pl_yylval.opval->op_folded = 1;
6701 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6706 /* Resolve to GV now if this is a placeholder. */
6707 if (!off && (gvop = cUNOPx(rv2cv_op)->op_first)
6708 && gvop->op_type == OP_GV) {
6709 GV *gv2 = cGVOPx_gv(gvop);
6710 if (gv2 && !isGV(gv2)) {
6711 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
6712 assert (SvTYPE(gv) == SVt_PVGV);
6713 /* cv must have been some sort of placeholder,
6714 so now needs replacing with a real code
6720 op_free(pl_yylval.opval);
6722 off ? (OP *)newCVREF(0, rv2cv_op) : rv2cv_op;
6723 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6724 PL_last_lop = PL_oldbufptr;
6725 PL_last_lop_op = OP_ENTERSUB;
6726 /* Is there a prototype? */
6730 STRLEN protolen = CvPROTOLEN(cv);
6731 const char *proto = CvPROTO(cv);
6733 proto = S_strip_spaces(aTHX_ proto, &protolen);
6736 if ((optional = *proto == ';'))
6739 while (*proto == ';');
6743 *proto == '$' || *proto == '_'
6744 || *proto == '*' || *proto == '+'
6749 *proto == '\\' && proto[1] && proto[2] == '\0'
6752 UNIPROTO(UNIOPSUB,optional);
6753 if (*proto == '\\' && proto[1] == '[') {
6754 const char *p = proto + 2;
6755 while(*p && *p != ']')
6757 if(*p == ']' && !p[1])
6758 UNIPROTO(UNIOPSUB,optional);
6760 if (*proto == '&' && *s == '{') {
6762 sv_setpvs(PL_subname, "__ANON__");
6764 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6765 if (!PL_lex_allbrackets &&
6766 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6767 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6771 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6773 force_next(off ? PRIVATEREF : WORD);
6774 if (!PL_lex_allbrackets &&
6775 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6776 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6780 /* Call it a bare word */
6782 if (PL_hints & HINT_STRICT_SUBS)
6783 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6786 /* after "print" and similar functions (corresponding to
6787 * "F? L" in opcode.pl), whatever wasn't already parsed as
6788 * a filehandle should be subject to "strict subs".
6789 * Likewise for the optional indirect-object argument to system
6790 * or exec, which can't be a bareword */
6791 if ((PL_last_lop_op == OP_PRINT
6792 || PL_last_lop_op == OP_PRTF
6793 || PL_last_lop_op == OP_SAY
6794 || PL_last_lop_op == OP_SYSTEM
6795 || PL_last_lop_op == OP_EXEC)
6796 && (PL_hints & HINT_STRICT_SUBS))
6797 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6798 if (lastchar != '-') {
6799 if (ckWARN(WARN_RESERVED)) {
6803 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
6805 /* PL_warn_reserved is constant */
6806 GCC_DIAG_IGNORE(-Wformat-nonliteral);
6807 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6817 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
6818 && saw_infix_sigil) {
6819 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6820 "Operator or semicolon missing before %c%"UTF8f,
6822 UTF8fARG(UTF, strlen(PL_tokenbuf),
6824 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6825 "Ambiguous use of %c resolved as operator %c",
6826 lastchar, lastchar);
6833 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
6838 (OP*)newSVOP(OP_CONST, 0,
6839 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
6842 case KEY___PACKAGE__:
6844 (OP*)newSVOP(OP_CONST, 0,
6846 ? newSVhek(HvNAME_HEK(PL_curstash))
6853 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6854 HV * const stash = PL_tokenbuf[2] == 'D' && PL_curstash
6857 gv = (GV *)*hv_fetchs(stash, "DATA", 1);
6859 gv_init(gv,stash,"DATA",4,0);
6862 GvIOp(gv) = newIO();
6863 IoIFP(GvIOp(gv)) = PL_rsfp;
6864 #if defined(HAS_FCNTL) && defined(F_SETFD)
6866 const int fd = PerlIO_fileno(PL_rsfp);
6867 fcntl(fd,F_SETFD,fd >= 3);
6870 /* Mark this internal pseudo-handle as clean */
6871 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6872 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6873 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6875 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6876 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6877 /* if the script was opened in binmode, we need to revert
6878 * it to text mode for compatibility; but only iff it has CRs
6879 * XXX this is a questionable hack at best. */
6880 if (PL_bufend-PL_bufptr > 2
6881 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6884 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6885 loc = PerlIO_tell(PL_rsfp);
6886 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6889 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6891 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6892 #endif /* NETWARE */
6894 PerlIO_seek(PL_rsfp, loc, 0);
6898 #ifdef PERLIO_LAYERS
6901 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6902 else if (PL_encoding) {
6908 XPUSHs(PL_encoding);
6910 call_method("name", G_SCALAR);
6914 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6915 Perl_form(aTHX_ ":encoding(%"SVf")",
6928 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
6937 if (PL_expect == XSTATE) {
6948 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6949 if ((*s == ':' && s[1] == ':')
6950 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
6954 Copy(PL_bufptr, PL_tokenbuf, olen, char);
6958 Perl_croak(aTHX_ "CORE::%"UTF8f" is not a keyword",
6959 UTF8fARG(UTF, len, PL_tokenbuf));
6962 else if (tmp == KEY_require || tmp == KEY_do
6964 /* that's a way to remember we saw "CORE::" */
6976 LOP(OP_ACCEPT,XTERM);
6979 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
6984 LOP(OP_ATAN2,XTERM);
6990 LOP(OP_BINMODE,XTERM);
6993 LOP(OP_BLESS,XTERM);
7002 /* We have to disambiguate the two senses of
7003 "continue". If the next token is a '{' then
7004 treat it as the start of a continue block;
7005 otherwise treat it as a control operator.
7015 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7025 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7034 if (!PL_cryptseen) {
7035 PL_cryptseen = TRUE;
7039 LOP(OP_CRYPT,XTERM);
7042 LOP(OP_CHMOD,XTERM);
7045 LOP(OP_CHOWN,XTERM);
7048 LOP(OP_CONNECT,XTERM);
7068 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7070 if (len && (len != 4 || strNE(PL_tokenbuf+1, "CORE"))
7071 && !keyword(PL_tokenbuf + 1, len, 0)) {
7074 force_ident_maybe_lex('&');
7079 if (orig_keyword == KEY_do) {
7088 PL_hints |= HINT_BLOCK_SCOPE;
7098 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7099 STR_WITH_LEN("NDBM_File::"),
7100 STR_WITH_LEN("DB_File::"),
7101 STR_WITH_LEN("GDBM_File::"),
7102 STR_WITH_LEN("SDBM_File::"),
7103 STR_WITH_LEN("ODBM_File::"),
7105 LOP(OP_DBMOPEN,XTERM);
7117 pl_yylval.ival = CopLINE(PL_curcop);
7121 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7133 if (*s == '{') { /* block eval */
7134 PL_expect = XTERMBLOCK;
7135 UNIBRACK(OP_ENTERTRY);
7137 else { /* string eval */
7139 UNIBRACK(OP_ENTEREVAL);
7144 UNIBRACK(-OP_ENTEREVAL);
7158 case KEY_endhostent:
7164 case KEY_endservent:
7167 case KEY_endprotoent:
7178 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7180 pl_yylval.ival = CopLINE(PL_curcop);
7182 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7185 if ((PL_bufend - p) >= 3 &&
7186 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7188 else if ((PL_bufend - p) >= 4 &&
7189 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7192 /* skip optional package name, as in "for my abc $x (..)" */
7193 if (isIDFIRST_lazy_if(p,UTF)) {
7194 p = scan_word(p, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7198 Perl_croak(aTHX_ "Missing $ on loop variable");
7203 LOP(OP_FORMLINE,XTERM);
7212 LOP(OP_FCNTL,XTERM);
7218 LOP(OP_FLOCK,XTERM);
7221 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7226 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7231 LOP(OP_GREPSTART, XREF);
7248 case KEY_getpriority:
7249 LOP(OP_GETPRIORITY,XTERM);
7251 case KEY_getprotobyname:
7254 case KEY_getprotobynumber:
7255 LOP(OP_GPBYNUMBER,XTERM);
7257 case KEY_getprotoent:
7269 case KEY_getpeername:
7270 UNI(OP_GETPEERNAME);
7272 case KEY_gethostbyname:
7275 case KEY_gethostbyaddr:
7276 LOP(OP_GHBYADDR,XTERM);
7278 case KEY_gethostent:
7281 case KEY_getnetbyname:
7284 case KEY_getnetbyaddr:
7285 LOP(OP_GNBYADDR,XTERM);
7290 case KEY_getservbyname:
7291 LOP(OP_GSBYNAME,XTERM);
7293 case KEY_getservbyport:
7294 LOP(OP_GSBYPORT,XTERM);
7296 case KEY_getservent:
7299 case KEY_getsockname:
7300 UNI(OP_GETSOCKNAME);
7302 case KEY_getsockopt:
7303 LOP(OP_GSOCKOPT,XTERM);
7318 pl_yylval.ival = CopLINE(PL_curcop);
7319 Perl_ck_warner_d(aTHX_
7320 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
7321 "given is experimental");
7326 orig_keyword==KEY_glob ? -OP_GLOB : OP_GLOB,
7334 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7336 pl_yylval.ival = CopLINE(PL_curcop);
7340 LOP(OP_INDEX,XTERM);
7346 LOP(OP_IOCTL,XTERM);
7374 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7379 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7393 LOP(OP_LISTEN,XTERM);
7402 s = scan_pat(s,OP_MATCH);
7403 TERM(sublex_start());
7406 LOP(OP_MAPSTART, XREF);
7409 LOP(OP_MKDIR,XTERM);
7412 LOP(OP_MSGCTL,XTERM);
7415 LOP(OP_MSGGET,XTERM);
7418 LOP(OP_MSGRCV,XTERM);
7421 LOP(OP_MSGSND,XTERM);
7426 PL_in_my = (U16)tmp;
7428 if (isIDFIRST_lazy_if(s,UTF)) {
7429 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7430 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7432 if (!FEATURE_LEXSUBS_IS_ENABLED)
7434 "Experimental \"%s\" subs not enabled",
7435 tmp == KEY_my ? "my" :
7436 tmp == KEY_state ? "state" : "our");
7437 Perl_ck_warner_d(aTHX_
7438 packWARN(WARN_EXPERIMENTAL__LEXICAL_SUBS),
7439 "The lexical_subs feature is experimental");
7442 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7443 if (!PL_in_my_stash) {
7447 len = my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7448 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(tmpbuf));
7449 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7459 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7464 s = tokenize_use(0, s);
7468 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7471 if (!PL_lex_allbrackets &&
7472 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7473 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7479 if (isIDFIRST_lazy_if(s,UTF)) {
7481 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE,
7483 for (t=d; isSPACE(*t);)
7485 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7487 && !(t[0] == '=' && t[1] == '>')
7488 && !(t[0] == ':' && t[1] == ':')
7489 && !keyword(s, d-s, 0)
7491 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7492 "Precedence problem: open %"UTF8f" should be open(%"UTF8f")",
7493 UTF8fARG(UTF, d-s, s), UTF8fARG(UTF, d-s, s));
7499 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7501 pl_yylval.ival = OP_OR;
7511 LOP(OP_OPEN_DIR,XTERM);
7514 checkcomma(s,PL_tokenbuf,"filehandle");
7518 checkcomma(s,PL_tokenbuf,"filehandle");
7537 s = force_word(s,WORD,FALSE,TRUE);
7539 s = force_strict_version(s);
7543 LOP(OP_PIPE_OP,XTERM);
7546 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7549 COPLINE_SET_FROM_MULTI_END;
7550 pl_yylval.ival = OP_CONST;
7551 TERM(sublex_start());
7558 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7561 COPLINE_SET_FROM_MULTI_END;
7562 PL_expect = XOPERATOR;
7563 if (SvCUR(PL_lex_stuff)) {
7564 int warned_comma = !ckWARN(WARN_QW);
7565 int warned_comment = warned_comma;
7566 d = SvPV_force(PL_lex_stuff, len);
7568 for (; isSPACE(*d) && len; --len, ++d)
7573 if (!warned_comma || !warned_comment) {
7574 for (; !isSPACE(*d) && len; --len, ++d) {
7575 if (!warned_comma && *d == ',') {
7576 Perl_warner(aTHX_ packWARN(WARN_QW),
7577 "Possible attempt to separate words with commas");
7580 else if (!warned_comment && *d == '#') {
7581 Perl_warner(aTHX_ packWARN(WARN_QW),
7582 "Possible attempt to put comments in qw() list");
7588 for (; !isSPACE(*d) && len; --len, ++d)
7591 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7592 words = op_append_elem(OP_LIST, words,
7593 newSVOP(OP_CONST, 0, tokeq(sv)));
7598 words = newNULLLIST();
7600 SvREFCNT_dec(PL_lex_stuff);
7601 PL_lex_stuff = NULL;
7603 PL_expect = XOPERATOR;
7604 pl_yylval.opval = sawparens(words);
7609 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7612 pl_yylval.ival = OP_STRINGIFY;
7613 if (SvIVX(PL_lex_stuff) == '\'')
7614 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7615 TERM(sublex_start());
7618 s = scan_pat(s,OP_QR);
7619 TERM(sublex_start());
7622 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7625 pl_yylval.ival = OP_BACKTICK;
7626 TERM(sublex_start());
7634 s = force_version(s, FALSE);
7636 else if (*s != 'v' || !isDIGIT(s[1])
7637 || (s = force_version(s, TRUE), *s == 'v'))
7639 *PL_tokenbuf = '\0';
7640 s = force_word(s,WORD,TRUE,TRUE);
7641 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7642 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
7643 GV_ADD | (UTF ? SVf_UTF8 : 0));
7645 yyerror("<> at require-statement should be quotes");
7647 if (orig_keyword == KEY_require) {
7653 PL_expect = PL_nexttoke ? XOPERATOR : XTERM;
7655 PL_last_uni = PL_oldbufptr;
7656 PL_last_lop_op = OP_REQUIRE;
7658 return REPORT( (int)REQUIRE );
7667 LOP(OP_RENAME,XTERM);
7676 LOP(OP_RINDEX,XTERM);
7685 UNIDOR(OP_READLINE);
7688 UNIDOR(OP_BACKTICK);
7697 LOP(OP_REVERSE,XTERM);
7700 UNIDOR(OP_READLINK);
7707 if (pl_yylval.opval)
7708 TERM(sublex_start());
7710 TOKEN(1); /* force error */
7713 checkcomma(s,PL_tokenbuf,"filehandle");
7723 LOP(OP_SELECT,XTERM);
7729 LOP(OP_SEMCTL,XTERM);
7732 LOP(OP_SEMGET,XTERM);
7735 LOP(OP_SEMOP,XTERM);
7741 LOP(OP_SETPGRP,XTERM);
7743 case KEY_setpriority:
7744 LOP(OP_SETPRIORITY,XTERM);
7746 case KEY_sethostent:
7752 case KEY_setservent:
7755 case KEY_setprotoent:
7765 LOP(OP_SEEKDIR,XTERM);
7767 case KEY_setsockopt:
7768 LOP(OP_SSOCKOPT,XTERM);
7774 LOP(OP_SHMCTL,XTERM);
7777 LOP(OP_SHMGET,XTERM);
7780 LOP(OP_SHMREAD,XTERM);
7783 LOP(OP_SHMWRITE,XTERM);
7786 LOP(OP_SHUTDOWN,XTERM);
7795 LOP(OP_SOCKET,XTERM);
7797 case KEY_socketpair:
7798 LOP(OP_SOCKPAIR,XTERM);
7801 checkcomma(s,PL_tokenbuf,"subroutine name");
7804 s = force_word(s,WORD,TRUE,TRUE);
7808 LOP(OP_SPLIT,XTERM);
7811 LOP(OP_SPRINTF,XTERM);
7814 LOP(OP_SPLICE,XTERM);
7829 LOP(OP_SUBSTR,XTERM);
7835 char * const tmpbuf = PL_tokenbuf + 1;
7836 expectation attrful;
7837 bool have_name, have_proto;
7838 const int key = tmp;
7839 SV *format_name = NULL;
7844 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7845 (*s == ':' && s[1] == ':'))
7849 attrful = XATTRBLOCK;
7850 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
7852 if (key == KEY_format)
7853 format_name = S_newSV_maybe_utf8(aTHX_ s, d - s);
7855 if (memchr(tmpbuf, ':', len) || key != KEY_sub
7857 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
7859 sv_setpvn(PL_subname, tmpbuf, len);
7861 sv_setsv(PL_subname,PL_curstname);
7862 sv_catpvs(PL_subname,"::");
7863 sv_catpvn(PL_subname,tmpbuf,len);
7865 if (SvUTF8(PL_linestr))
7866 SvUTF8_on(PL_subname);
7873 if (key == KEY_my || key == KEY_our || key==KEY_state)
7876 /* diag_listed_as: Missing name in "%s sub" */
7878 "Missing name in \"%s\"", PL_bufptr);
7880 PL_expect = XTERMBLOCK;
7881 attrful = XATTRTERM;
7882 sv_setpvs(PL_subname,"?");
7886 if (key == KEY_format) {
7888 NEXTVAL_NEXTTOKE.opval
7889 = (OP*)newSVOP(OP_CONST,0, format_name);
7890 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
7896 /* Look for a prototype */
7897 if (*s == '(' && !FEATURE_SIGNATURES_IS_ENABLED) {
7898 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7899 COPLINE_SET_FROM_MULTI_END;
7901 Perl_croak(aTHX_ "Prototype not terminated");
7902 (void)validate_proto(PL_subname, PL_lex_stuff, ckWARN(WARN_ILLEGALPROTO));
7910 if (*s == ':' && s[1] != ':')
7911 PL_expect = attrful;
7912 else if ((*s != '{' && *s != '(') && key == KEY_sub) {
7914 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7915 else if (*s != ';' && *s != '}')
7916 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7920 NEXTVAL_NEXTTOKE.opval =
7921 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7922 PL_lex_stuff = NULL;
7927 sv_setpvs(PL_subname, "__ANON__");
7929 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7932 force_ident_maybe_lex('&');
7937 LOP(OP_SYSTEM,XREF);
7940 LOP(OP_SYMLINK,XTERM);
7943 LOP(OP_SYSCALL,XTERM);
7946 LOP(OP_SYSOPEN,XTERM);
7949 LOP(OP_SYSSEEK,XTERM);
7952 LOP(OP_SYSREAD,XTERM);
7955 LOP(OP_SYSWRITE,XTERM);
7960 TERM(sublex_start());
7981 LOP(OP_TRUNCATE,XTERM);
7993 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7995 pl_yylval.ival = CopLINE(PL_curcop);
7999 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8001 pl_yylval.ival = CopLINE(PL_curcop);
8005 LOP(OP_UNLINK,XTERM);
8011 LOP(OP_UNPACK,XTERM);
8014 LOP(OP_UTIME,XTERM);
8020 LOP(OP_UNSHIFT,XTERM);
8023 s = tokenize_use(1, s);
8033 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8035 pl_yylval.ival = CopLINE(PL_curcop);
8036 Perl_ck_warner_d(aTHX_
8037 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
8038 "when is experimental");
8042 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8044 pl_yylval.ival = CopLINE(PL_curcop);
8048 PL_hints |= HINT_BLOCK_SCOPE;
8055 LOP(OP_WAITPID,XTERM);
8061 /* Make sure $^L is defined. 0x0C is CTRL-L on ASCII platforms, and
8062 * we use the same number on EBCDIC */
8063 gv_fetchpvs("\x0C", GV_ADD|GV_NOTQUAL, SVt_PV);
8067 if (PL_expect == XOPERATOR) {
8068 if (*s == '=' && !PL_lex_allbrackets &&
8069 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8077 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8079 pl_yylval.ival = OP_XOR;
8088 Looks up an identifier in the pad or in a package
8091 PRIVATEREF if this is a lexical name.
8092 WORD if this belongs to a package.
8095 if we're in a my declaration
8096 croak if they tried to say my($foo::bar)
8097 build the ops for a my() declaration
8098 if it's an access to a my() variable
8099 build ops for access to a my() variable
8100 if in a dq string, and they've said @foo and we can't find @foo
8102 build ops for a bareword
8106 S_pending_ident(pTHX)
8109 const char pit = (char)pl_yylval.ival;
8110 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8111 /* All routes through this function want to know if there is a colon. */
8112 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8114 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8115 "### Pending identifier '%s'\n", PL_tokenbuf); });
8117 /* if we're in a my(), we can't allow dynamics here.
8118 $foo'bar has already been turned into $foo::bar, so
8119 just check for colons.
8121 if it's a legal name, the OP is a PADANY.
8124 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8126 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8127 "variable %s in \"our\"",
8128 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8129 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8133 /* PL_no_myglob is constant */
8134 GCC_DIAG_IGNORE(-Wformat-nonliteral);
8135 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8136 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8137 UTF ? SVf_UTF8 : 0);
8141 pl_yylval.opval = newOP(OP_PADANY, 0);
8142 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8143 UTF ? SVf_UTF8 : 0);
8149 build the ops for accesses to a my() variable.
8154 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8155 UTF ? SVf_UTF8 : 0);
8156 if (tmp != NOT_IN_PAD) {
8157 /* might be an "our" variable" */
8158 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8159 /* build ops for a bareword */
8160 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8161 HEK * const stashname = HvNAME_HEK(stash);
8162 SV * const sym = newSVhek(stashname);
8163 sv_catpvs(sym, "::");
8164 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8165 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8166 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8170 ? (GV_ADDMULTI | GV_ADDINEVAL)
8173 ((PL_tokenbuf[0] == '$') ? SVt_PV
8174 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8179 pl_yylval.opval = newOP(OP_PADANY, 0);
8180 pl_yylval.opval->op_targ = tmp;
8186 Whine if they've said @foo in a doublequoted string,
8187 and @foo isn't a variable we can find in the symbol
8190 if (ckWARN(WARN_AMBIGUOUS) &&
8191 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8192 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8193 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8194 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8195 /* DO NOT warn for @- and @+ */
8196 && !( PL_tokenbuf[2] == '\0' &&
8197 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8200 /* Downgraded from fatal to warning 20000522 mjd */
8201 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8202 "Possible unintended interpolation of %"UTF8f
8204 UTF8fARG(UTF, tokenbuf_len, PL_tokenbuf));
8208 /* build ops for a bareword */
8209 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8210 newSVpvn_flags(PL_tokenbuf + 1,
8212 UTF ? SVf_UTF8 : 0 ));
8213 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8215 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8216 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8217 | ( UTF ? SVf_UTF8 : 0 ),
8218 ((PL_tokenbuf[0] == '$') ? SVt_PV
8219 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8225 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8227 PERL_ARGS_ASSERT_CHECKCOMMA;
8229 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8230 if (ckWARN(WARN_SYNTAX)) {
8233 for (w = s+2; *w && level; w++) {
8241 /* the list of chars below is for end of statements or
8242 * block / parens, boolean operators (&&, ||, //) and branch
8243 * constructs (or, and, if, until, unless, while, err, for).
8244 * Not a very solid hack... */
8245 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8246 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8247 "%s (...) interpreted as function",name);
8250 while (s < PL_bufend && isSPACE(*s))
8254 while (s < PL_bufend && isSPACE(*s))
8256 if (isIDFIRST_lazy_if(s,UTF)) {
8257 const char * const w = s;
8258 s += UTF ? UTF8SKIP(s) : 1;
8259 while (isWORDCHAR_lazy_if(s,UTF))
8260 s += UTF ? UTF8SKIP(s) : 1;
8261 while (s < PL_bufend && isSPACE(*s))
8265 if (keyword(w, s - w, 0))
8268 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8269 if (gv && GvCVu(gv))
8271 Perl_croak(aTHX_ "No comma allowed after %s", what);
8276 /* S_new_constant(): do any overload::constant lookup.
8278 Either returns sv, or mortalizes/frees sv and returns a new SV*.
8279 Best used as sv=new_constant(..., sv, ...).
8280 If s, pv are NULL, calls subroutine with one argument,
8281 and <type> is used with error messages only.
8282 <type> is assumed to be well formed UTF-8 */
8285 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8286 SV *sv, SV *pv, const char *type, STRLEN typelen)
8289 HV * table = GvHV(PL_hintgv); /* ^H */
8294 const char *why1 = "", *why2 = "", *why3 = "";
8296 PERL_ARGS_ASSERT_NEW_CONSTANT;
8297 /* We assume that this is true: */
8298 if (*key == 'c') { assert (strEQ(key, "charnames")); }
8301 /* charnames doesn't work well if there have been errors found */
8302 if (PL_error_count > 0 && *key == 'c')
8304 SvREFCNT_dec_NN(sv);
8305 return &PL_sv_undef;
8308 sv_2mortal(sv); /* Parent created it permanently */
8310 || ! (PL_hints & HINT_LOCALIZE_HH)
8311 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8316 /* Here haven't found what we're looking for. If it is charnames,
8317 * perhaps it needs to be loaded. Try doing that before giving up */
8319 Perl_load_module(aTHX_
8321 newSVpvs("_charnames"),
8322 /* version parameter; no need to specify it, as if
8323 * we get too early a version, will fail anyway,
8324 * not being able to find '_charnames' */
8329 assert(sp == PL_stack_sp);
8330 table = GvHV(PL_hintgv);
8332 && (PL_hints & HINT_LOCALIZE_HH)
8333 && (cvp = hv_fetch(table, key, keylen, FALSE))
8339 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8340 msg = Perl_form(aTHX_
8341 "Constant(%.*s) unknown",
8342 (int)(type ? typelen : len),
8348 why3 = "} is not defined";
8351 msg = Perl_form(aTHX_
8352 /* The +3 is for '\N{'; -4 for that, plus '}' */
8353 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
8357 msg = Perl_form(aTHX_ "Constant(%.*s): %s%s%s",
8358 (int)(type ? typelen : len),
8359 (type ? type: s), why1, why2, why3);
8362 yyerror_pv(msg, UTF ? SVf_UTF8 : 0);
8363 return SvREFCNT_inc_simple_NN(sv);
8368 pv = newSVpvn_flags(s, len, SVs_TEMP);
8370 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8372 typesv = &PL_sv_undef;
8374 PUSHSTACKi(PERLSI_OVERLOAD);
8386 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8390 /* Check the eval first */
8391 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
8393 const char * errstr;
8394 sv_catpvs(errsv, "Propagated");
8395 errstr = SvPV_const(errsv, errlen);
8396 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
8398 res = SvREFCNT_inc_simple_NN(sv);
8402 SvREFCNT_inc_simple_void_NN(res);
8411 why1 = "Call to &{$^H{";
8413 why3 = "}} did not return a defined value";
8415 (void)sv_2mortal(sv);
8422 PERL_STATIC_INLINE void
8423 S_parse_ident(pTHX_ char **s, char **d, char * const e, int allow_package, bool is_utf8) {
8424 PERL_ARGS_ASSERT_PARSE_IDENT;
8428 Perl_croak(aTHX_ "%s", ident_too_long);
8429 if (is_utf8 && isIDFIRST_utf8((U8*)*s)) {
8430 /* The UTF-8 case must come first, otherwise things
8431 * like c\N{COMBINING TILDE} would start failing, as the
8432 * isWORDCHAR_A case below would gobble the 'c' up.
8435 char *t = *s + UTF8SKIP(*s);
8436 while (isIDCONT_utf8((U8*)t))
8438 if (*d + (t - *s) > e)
8439 Perl_croak(aTHX_ "%s", ident_too_long);
8440 Copy(*s, *d, t - *s, char);
8444 else if ( isWORDCHAR_A(**s) ) {
8447 } while (isWORDCHAR_A(**s) && *d < e);
8449 else if (allow_package && **s == '\'' && isIDFIRST_lazy_if(*s+1,is_utf8)) {
8454 else if (allow_package && **s == ':' && (*s)[1] == ':'
8455 /* Disallow things like Foo::$bar. For the curious, this is
8456 * the code path that triggers the "Bad name after" warning
8457 * when looking for barewords.
8459 && (*s)[2] != '$') {
8469 /* Returns a NUL terminated string, with the length of the string written to
8473 S_scan_word(pTHX_ char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8476 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8477 bool is_utf8 = cBOOL(UTF);
8479 PERL_ARGS_ASSERT_SCAN_WORD;
8481 parse_ident(&s, &d, e, allow_package, is_utf8);
8488 S_scan_ident(pTHX_ char *s, char *dest, STRLEN destlen, I32 ck_uni)
8490 I32 herelines = PL_parser->herelines;
8491 SSize_t bracket = -1;
8494 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8495 bool is_utf8 = cBOOL(UTF);
8496 I32 orig_copline = 0, tmp_copline = 0;
8498 PERL_ARGS_ASSERT_SCAN_IDENT;
8503 while (isDIGIT(*s)) {
8505 Perl_croak(aTHX_ "%s", ident_too_long);
8510 parse_ident(&s, &d, e, 1, is_utf8);
8515 /* Either a digit variable, or parse_ident() found an identifier
8516 (anything valid as a bareword), so job done and return. */
8517 if (PL_lex_state != LEX_NORMAL)
8518 PL_lex_state = LEX_INTERPENDMAYBE;
8521 if (*s == '$' && s[1] &&
8522 (isIDFIRST_lazy_if(s+1,is_utf8)
8523 || isDIGIT_A((U8)s[1])
8526 || strnEQ(s+1,"::",2)) )
8528 /* Dereferencing a value in a scalar variable.
8529 The alternatives are different syntaxes for a scalar variable.
8530 Using ' as a leading package separator isn't allowed. :: is. */
8533 /* Handle the opening { of @{...}, &{...}, *{...}, %{...}, ${...} */
8535 bracket = s - SvPVX(PL_linestr);
8537 orig_copline = CopLINE(PL_curcop);
8538 if (s < PL_bufend && isSPACE(*s)) {
8543 /* Is the byte 'd' a legal single character identifier name? 'u' is true
8544 * iff Unicode semantics are to be used. The legal ones are any of:
8546 * b) ASCII punctuation
8547 * c) When not under Unicode rules, any upper Latin1 character
8548 * d) \c?, \c\, \c^, \c_, and \cA..\cZ, minus the ones that have traditionally
8549 * been matched by \s on ASCII platforms. That is: \c?, plus 1-32, minus
8551 #define VALID_LEN_ONE_IDENT(d, u) (isPUNCT_A((U8)(d)) \
8552 || isDIGIT_A((U8)(d)) \
8553 || (!(u) && !isASCII((U8)(d))) \
8554 || ((((U8)(d)) < 32) \
8555 && (((((U8)(d)) >= 14) \
8556 || (((U8)(d)) <= 8 && (d) != 0) \
8557 || (((U8)(d)) == 13)))) \
8558 || (((U8)(d)) == toCTRL('?')))
8560 && (isIDFIRST_lazy_if(s, is_utf8) || VALID_LEN_ONE_IDENT(*s, is_utf8)))
8562 if ( isCNTRL_A((U8)*s) ) {
8563 deprecate("literal control characters in variable names");
8567 const STRLEN skip = UTF8SKIP(s);
8570 for ( i = 0; i < skip; i++ )
8578 /* Convert $^F, ${^F} and the ^F of ${^FOO} to control characters */
8579 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8583 /* Warn about ambiguous code after unary operators if {...} notation isn't
8584 used. There's no difference in ambiguity; it's merely a heuristic
8585 about when not to warn. */
8586 else if (ck_uni && bracket == -1)
8588 if (bracket != -1) {
8589 /* If we were processing {...} notation then... */
8590 if (isIDFIRST_lazy_if(d,is_utf8)) {
8591 /* if it starts as a valid identifier, assume that it is one.
8592 (the later check for } being at the expected point will trap
8593 cases where this doesn't pan out.) */
8594 d += is_utf8 ? UTF8SKIP(d) : 1;
8595 parse_ident(&s, &d, e, 1, is_utf8);
8597 tmp_copline = CopLINE(PL_curcop);
8598 if (s < PL_bufend && isSPACE(*s)) {
8601 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8602 /* ${foo[0]} and ${foo{bar}} notation. */
8603 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8604 const char * const brack =
8606 ((*s == '[') ? "[...]" : "{...}");
8607 orig_copline = CopLINE(PL_curcop);
8608 CopLINE_set(PL_curcop, tmp_copline);
8609 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8610 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8611 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8612 funny, dest, brack, funny, dest, brack);
8613 CopLINE_set(PL_curcop, orig_copline);
8616 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8617 PL_lex_allbrackets++;
8621 /* Handle extended ${^Foo} variables
8622 * 1999-02-27 mjd-perl-patch@plover.com */
8623 else if (! isPRINT(*d) /* isCNTRL(d), plus all non-ASCII */
8627 while (isWORDCHAR(*s) && d < e) {
8631 Perl_croak(aTHX_ "%s", ident_too_long);
8636 tmp_copline = CopLINE(PL_curcop);
8637 if (s < PL_bufend && isSPACE(*s)) {
8641 /* Expect to find a closing } after consuming any trailing whitespace.
8645 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8646 PL_lex_state = LEX_INTERPEND;
8649 if (PL_lex_state == LEX_NORMAL) {
8650 if (ckWARN(WARN_AMBIGUOUS) &&
8651 (keyword(dest, d - dest, 0)
8652 || get_cvn_flags(dest, d - dest, is_utf8 ? SVf_UTF8 : 0)))
8654 SV *tmp = newSVpvn_flags( dest, d - dest,
8655 SVs_TEMP | (is_utf8 ? SVf_UTF8 : 0) );
8658 orig_copline = CopLINE(PL_curcop);
8659 CopLINE_set(PL_curcop, tmp_copline);
8660 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8661 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
8662 funny, SVfARG(tmp), funny, SVfARG(tmp));
8663 CopLINE_set(PL_curcop, orig_copline);
8668 /* Didn't find the closing } at the point we expected, so restore
8669 state such that the next thing to process is the opening { and */
8670 s = SvPVX(PL_linestr) + bracket; /* let the parser handle it */
8671 CopLINE_set(PL_curcop, orig_copline);
8672 PL_parser->herelines = herelines;
8676 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8677 PL_lex_state = LEX_INTERPEND;
8682 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8684 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8685 * the parse starting at 's', based on the subset that are valid in this
8686 * context input to this routine in 'valid_flags'. Advances s. Returns
8687 * TRUE if the input should be treated as a valid flag, so the next char
8688 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
8689 * first call on the current regex. This routine will set it to any
8690 * charset modifier found. The caller shouldn't change it. This way,
8691 * another charset modifier encountered in the parse can be detected as an
8692 * error, as we have decided to allow only one */
8695 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
8697 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
8698 if (isWORDCHAR_lazy_if(*s, UTF)) {
8699 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
8700 UTF ? SVf_UTF8 : 0);
8702 /* Pretend that it worked, so will continue processing before
8711 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8712 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8713 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8714 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8715 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8716 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8717 case LOCALE_PAT_MOD:
8719 goto multiple_charsets;
8721 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8724 case UNICODE_PAT_MOD:
8726 goto multiple_charsets;
8728 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8731 case ASCII_RESTRICT_PAT_MOD:
8733 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8737 /* Error if previous modifier wasn't an 'a', but if it was, see
8738 * if, and accept, a second occurrence (only) */
8740 || get_regex_charset(*pmfl)
8741 != REGEX_ASCII_RESTRICTED_CHARSET)
8743 goto multiple_charsets;
8745 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8749 case DEPENDS_PAT_MOD:
8751 goto multiple_charsets;
8753 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8762 if (*charset != c) {
8763 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8765 else if (c == 'a') {
8766 /* diag_listed_as: Regexp modifier "/%c" may appear a maximum of twice */
8767 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
8770 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8773 /* Pretend that it worked, so will continue processing before dieing */
8779 S_scan_pat(pTHX_ char *start, I32 type)
8783 const char * const valid_flags =
8784 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8785 char charset = '\0'; /* character set modifier */
8787 PERL_ARGS_ASSERT_SCAN_PAT;
8789 s = scan_str(start,TRUE,FALSE, (PL_in_eval & EVAL_RE_REPARSING), NULL);
8791 Perl_croak(aTHX_ "Search pattern not terminated");
8793 pm = (PMOP*)newPMOP(type, 0);
8794 if (PL_multi_open == '?') {
8795 /* This is the only point in the code that sets PMf_ONCE: */
8796 pm->op_pmflags |= PMf_ONCE;
8798 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8799 allows us to restrict the list needed by reset to just the ??
8801 assert(type != OP_TRANS);
8803 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8806 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8809 elements = mg->mg_len / sizeof(PMOP**);
8810 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8811 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8812 mg->mg_len = elements * sizeof(PMOP**);
8813 PmopSTASH_set(pm,PL_curstash);
8817 /* if qr/...(?{..}).../, then need to parse the pattern within a new
8818 * anon CV. False positives like qr/[(?{]/ are harmless */
8820 if (type == OP_QR) {
8822 char *e, *p = SvPV(PL_lex_stuff, len);
8824 for (; p < e; p++) {
8825 if (p[0] == '(' && p[1] == '?'
8826 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
8828 pm->op_pmflags |= PMf_HAS_CV;
8832 pm->op_pmflags |= PMf_IS_QR;
8835 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
8836 /* issue a warning if /c is specified,but /g is not */
8837 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8839 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8840 "Use of /c modifier is meaningless without /g" );
8843 PL_lex_op = (OP*)pm;
8844 pl_yylval.ival = OP_MATCH;
8849 S_scan_subst(pTHX_ char *start)
8856 char charset = '\0'; /* character set modifier */
8859 PERL_ARGS_ASSERT_SCAN_SUBST;
8861 pl_yylval.ival = OP_NULL;
8863 s = scan_str(start, TRUE, FALSE, FALSE, &t);
8866 Perl_croak(aTHX_ "Substitution pattern not terminated");
8870 first_start = PL_multi_start;
8871 first_line = CopLINE(PL_curcop);
8872 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8875 SvREFCNT_dec(PL_lex_stuff);
8876 PL_lex_stuff = NULL;
8878 Perl_croak(aTHX_ "Substitution replacement not terminated");
8880 PL_multi_start = first_start; /* so whole substitution is taken together */
8882 pm = (PMOP*)newPMOP(OP_SUBST, 0);
8886 if (*s == EXEC_PAT_MOD) {
8890 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
8896 if ((pm->op_pmflags & PMf_CONTINUE)) {
8897 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
8901 SV * const repl = newSVpvs("");
8904 pm->op_pmflags |= PMf_EVAL;
8907 sv_catpvs(repl, "eval ");
8909 sv_catpvs(repl, "do ");
8911 sv_catpvs(repl, "{");
8912 sv_catsv(repl, PL_sublex_info.repl);
8913 sv_catpvs(repl, "}");
8915 SvREFCNT_dec(PL_sublex_info.repl);
8916 PL_sublex_info.repl = repl;
8918 if (CopLINE(PL_curcop) != first_line) {
8919 sv_upgrade(PL_sublex_info.repl, SVt_PVNV);
8920 ((XPVNV*)SvANY(PL_sublex_info.repl))->xnv_u.xpad_cop_seq.xlow =
8921 CopLINE(PL_curcop) - first_line;
8922 CopLINE_set(PL_curcop, first_line);
8925 PL_lex_op = (OP*)pm;
8926 pl_yylval.ival = OP_SUBST;
8931 S_scan_trans(pTHX_ char *start)
8938 bool nondestruct = 0;
8941 PERL_ARGS_ASSERT_SCAN_TRANS;
8943 pl_yylval.ival = OP_NULL;
8945 s = scan_str(start,FALSE,FALSE,FALSE,&t);
8947 Perl_croak(aTHX_ "Transliteration pattern not terminated");
8951 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8954 SvREFCNT_dec(PL_lex_stuff);
8955 PL_lex_stuff = NULL;
8957 Perl_croak(aTHX_ "Transliteration replacement not terminated");
8960 complement = del = squash = 0;
8964 complement = OPpTRANS_COMPLEMENT;
8967 del = OPpTRANS_DELETE;
8970 squash = OPpTRANS_SQUASH;
8982 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
8983 o->op_private &= ~OPpTRANS_ALL;
8984 o->op_private |= del|squash|complement|
8985 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
8986 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
8989 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
8996 Takes a pointer to the first < in <<FOO.
8997 Returns a pointer to the byte following <<FOO.
8999 This function scans a heredoc, which involves different methods
9000 depending on whether we are in a string eval, quoted construct, etc.
9001 This is because PL_linestr could containing a single line of input, or
9002 a whole string being evalled, or the contents of the current quote-
9005 The two basic methods are:
9006 - Steal lines from the input stream
9007 - Scan the heredoc in PL_linestr and remove it therefrom
9009 In a file scope or filtered eval, the first method is used; in a
9010 string eval, the second.
9012 In a quote-like operator, we have to choose between the two,
9013 depending on where we can find a newline. We peek into outer lex-
9014 ing scopes until we find one with a newline in it. If we reach the
9015 outermost lexing scope and it is a file, we use the stream method.
9016 Otherwise it is treated as an eval.
9020 S_scan_heredoc(pTHX_ char *s)
9022 I32 op_type = OP_SCALAR;
9029 const bool infile = PL_rsfp || PL_parser->filtered;
9030 const line_t origline = CopLINE(PL_curcop);
9031 LEXSHARED *shared = PL_parser->lex_shared;
9033 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9036 d = PL_tokenbuf + 1;
9037 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9038 *PL_tokenbuf = '\n';
9040 while (SPACE_OR_TAB(*peek))
9042 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9045 s = delimcpy(d, e, s, PL_bufend, term, &len);
9047 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9053 /* <<\FOO is equivalent to <<'FOO' */
9057 if (!isWORDCHAR_lazy_if(s,UTF))
9058 deprecate("bare << to mean <<\"\"");
9059 for (; isWORDCHAR_lazy_if(s,UTF); s++) {
9064 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9065 Perl_croak(aTHX_ "Delimiter for here document is too long");
9068 len = d - PL_tokenbuf;
9070 #ifndef PERL_STRICT_CR
9071 d = strchr(s, '\r');
9073 char * const olds = s;
9075 while (s < PL_bufend) {
9081 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9090 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9095 tmpstr = newSV_type(SVt_PVIV);
9099 SvIV_set(tmpstr, -1);
9101 else if (term == '`') {
9102 op_type = OP_BACKTICK;
9103 SvIV_set(tmpstr, '\\');
9106 PL_multi_start = origline + 1 + PL_parser->herelines;
9107 PL_multi_open = PL_multi_close = '<';
9108 /* inside a string eval or quote-like operator */
9109 if (!infile || PL_lex_inwhat) {
9112 char * const olds = s;
9113 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9114 /* These two fields are not set until an inner lexing scope is
9115 entered. But we need them set here. */
9116 shared->ls_bufptr = s;
9117 shared->ls_linestr = PL_linestr;
9119 /* Look for a newline. If the current buffer does not have one,
9120 peek into the line buffer of the parent lexing scope, going
9121 up as many levels as necessary to find one with a newline
9124 while (!(s = (char *)memchr(
9125 (void *)shared->ls_bufptr, '\n',
9126 SvEND(shared->ls_linestr)-shared->ls_bufptr
9128 shared = shared->ls_prev;
9129 /* shared is only null if we have gone beyond the outermost
9130 lexing scope. In a file, we will have broken out of the
9131 loop in the previous iteration. In an eval, the string buf-
9132 fer ends with "\n;", so the while condition above will have
9133 evaluated to false. So shared can never be null. */
9135 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9136 most lexing scope. In a file, shared->ls_linestr at that
9137 level is just one line, so there is no body to steal. */
9138 if (infile && !shared->ls_prev) {
9144 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9147 linestr = shared->ls_linestr;
9148 bufend = SvEND(linestr);
9150 while (s < bufend - len + 1 &&
9151 memNE(s,PL_tokenbuf,len) ) {
9153 ++PL_parser->herelines;
9155 if (s >= bufend - len + 1) {
9158 sv_setpvn(tmpstr,d+1,s-d);
9160 /* the preceding stmt passes a newline */
9161 PL_parser->herelines++;
9163 /* s now points to the newline after the heredoc terminator.
9164 d points to the newline before the body of the heredoc.
9167 /* We are going to modify linestr in place here, so set
9168 aside copies of the string if necessary for re-evals or
9170 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9171 check shared->re_eval_str. */
9172 if (shared->re_eval_start || shared->re_eval_str) {
9173 /* Set aside the rest of the regexp */
9174 if (!shared->re_eval_str)
9175 shared->re_eval_str =
9176 newSVpvn(shared->re_eval_start,
9177 bufend - shared->re_eval_start);
9178 shared->re_eval_start -= s-d;
9180 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9181 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9182 cx->blk_eval.cur_text == linestr)
9184 cx->blk_eval.cur_text = newSVsv(linestr);
9185 SvSCREAM_on(cx->blk_eval.cur_text);
9187 /* Copy everything from s onwards back to d. */
9188 Move(s,d,bufend-s + 1,char);
9189 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9190 /* Setting PL_bufend only applies when we have not dug deeper
9191 into other scopes, because sublex_done sets PL_bufend to
9192 SvEND(PL_linestr). */
9193 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9200 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9201 term = PL_tokenbuf[1];
9203 linestr_save = PL_linestr; /* must restore this afterwards */
9204 d = s; /* and this */
9205 PL_linestr = newSVpvs("");
9206 PL_bufend = SvPVX(PL_linestr);
9208 PL_bufptr = PL_bufend;
9209 CopLINE_set(PL_curcop,
9210 origline + 1 + PL_parser->herelines);
9211 if (!lex_next_chunk(LEX_NO_TERM)
9212 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9213 SvREFCNT_dec(linestr_save);
9216 CopLINE_set(PL_curcop, origline);
9217 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9218 s = lex_grow_linestr(SvLEN(PL_linestr) + 3);
9219 /* ^That should be enough to avoid this needing to grow: */
9220 sv_catpvs(PL_linestr, "\n\0");
9221 assert(s == SvPVX(PL_linestr));
9222 PL_bufend = SvEND(PL_linestr);
9225 PL_parser->herelines++;
9226 PL_last_lop = PL_last_uni = NULL;
9227 #ifndef PERL_STRICT_CR
9228 if (PL_bufend - PL_linestart >= 2) {
9229 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9230 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9232 PL_bufend[-2] = '\n';
9234 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9236 else if (PL_bufend[-1] == '\r')
9237 PL_bufend[-1] = '\n';
9239 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9240 PL_bufend[-1] = '\n';
9242 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
9243 SvREFCNT_dec(PL_linestr);
9244 PL_linestr = linestr_save;
9245 PL_linestart = SvPVX(linestr_save);
9246 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9251 sv_catsv(tmpstr,PL_linestr);
9255 PL_multi_end = origline + PL_parser->herelines;
9256 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9257 SvPV_shrink_to_cur(tmpstr);
9260 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9262 else if (PL_encoding)
9263 sv_recode_to_utf8(tmpstr, PL_encoding);
9265 PL_lex_stuff = tmpstr;
9266 pl_yylval.ival = op_type;
9270 SvREFCNT_dec(tmpstr);
9271 CopLINE_set(PL_curcop, origline);
9272 missingterm(PL_tokenbuf + 1);
9276 takes: current position in input buffer
9277 returns: new position in input buffer
9278 side-effects: pl_yylval and lex_op are set.
9283 <FH> read from filehandle
9284 <pkg::FH> read from package qualified filehandle
9285 <pkg'FH> read from package qualified filehandle
9286 <$fh> read from filehandle in $fh
9292 S_scan_inputsymbol(pTHX_ char *start)
9294 char *s = start; /* current position in buffer */
9297 char *d = PL_tokenbuf; /* start of temp holding space */
9298 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9300 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9302 end = strchr(s, '\n');
9305 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9307 /* die if we didn't have space for the contents of the <>,
9308 or if it didn't end, or if we see a newline
9311 if (len >= (I32)sizeof PL_tokenbuf)
9312 Perl_croak(aTHX_ "Excessively long <> operator");
9314 Perl_croak(aTHX_ "Unterminated <> operator");
9319 Remember, only scalar variables are interpreted as filehandles by
9320 this code. Anything more complex (e.g., <$fh{$num}>) will be
9321 treated as a glob() call.
9322 This code makes use of the fact that except for the $ at the front,
9323 a scalar variable and a filehandle look the same.
9325 if (*d == '$' && d[1]) d++;
9327 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9328 while (*d && (isWORDCHAR_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9329 d += UTF ? UTF8SKIP(d) : 1;
9331 /* If we've tried to read what we allow filehandles to look like, and
9332 there's still text left, then it must be a glob() and not a getline.
9333 Use scan_str to pull out the stuff between the <> and treat it
9334 as nothing more than a string.
9337 if (d - PL_tokenbuf != len) {
9338 pl_yylval.ival = OP_GLOB;
9339 s = scan_str(start,FALSE,FALSE,FALSE,NULL);
9341 Perl_croak(aTHX_ "Glob not terminated");
9345 bool readline_overriden = FALSE;
9347 /* we're in a filehandle read situation */
9350 /* turn <> into <ARGV> */
9352 Copy("ARGV",d,5,char);
9354 /* Check whether readline() is overriden */
9355 if ((gv_readline = gv_override("readline",8)))
9356 readline_overriden = TRUE;
9358 /* if <$fh>, create the ops to turn the variable into a
9362 /* try to find it in the pad for this block, otherwise find
9363 add symbol table ops
9365 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9366 if (tmp != NOT_IN_PAD) {
9367 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9368 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9369 HEK * const stashname = HvNAME_HEK(stash);
9370 SV * const sym = sv_2mortal(newSVhek(stashname));
9371 sv_catpvs(sym, "::");
9377 OP * const o = newOP(OP_PADSV, 0);
9379 PL_lex_op = readline_overriden
9380 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9381 op_append_elem(OP_LIST, o,
9382 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9383 : (OP*)newUNOP(OP_READLINE, 0, o);
9392 ? (GV_ADDMULTI | GV_ADDINEVAL)
9393 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
9395 PL_lex_op = readline_overriden
9396 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9397 op_append_elem(OP_LIST,
9398 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9399 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9400 : (OP*)newUNOP(OP_READLINE, 0,
9401 newUNOP(OP_RV2SV, 0,
9402 newGVOP(OP_GV, 0, gv)));
9404 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9405 pl_yylval.ival = OP_NULL;
9408 /* If it's none of the above, it must be a literal filehandle
9409 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9411 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9412 PL_lex_op = readline_overriden
9413 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9414 op_append_elem(OP_LIST,
9415 newGVOP(OP_GV, 0, gv),
9416 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9417 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9418 pl_yylval.ival = OP_NULL;
9428 start position in buffer
9429 keep_bracketed_quoted preserve \ quoting of embedded delimiters, but
9430 only if they are of the open/close form
9431 keep_delims preserve the delimiters around the string
9432 re_reparse compiling a run-time /(?{})/:
9433 collapse // to /, and skip encoding src
9434 delimp if non-null, this is set to the position of
9435 the closing delimiter, or just after it if
9436 the closing and opening delimiters differ
9437 (i.e., the opening delimiter of a substitu-
9439 returns: position to continue reading from buffer
9440 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9441 updates the read buffer.
9443 This subroutine pulls a string out of the input. It is called for:
9444 q single quotes q(literal text)
9445 ' single quotes 'literal text'
9446 qq double quotes qq(interpolate $here please)
9447 " double quotes "interpolate $here please"
9448 qx backticks qx(/bin/ls -l)
9449 ` backticks `/bin/ls -l`
9450 qw quote words @EXPORT_OK = qw( func() $spam )
9451 m// regexp match m/this/
9452 s/// regexp substitute s/this/that/
9453 tr/// string transliterate tr/this/that/
9454 y/// string transliterate y/this/that/
9455 ($*@) sub prototypes sub foo ($)
9456 (stuff) sub attr parameters sub foo : attr(stuff)
9457 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9459 In most of these cases (all but <>, patterns and transliterate)
9460 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9461 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9462 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9465 It skips whitespace before the string starts, and treats the first
9466 character as the delimiter. If the delimiter is one of ([{< then
9467 the corresponding "close" character )]}> is used as the closing
9468 delimiter. It allows quoting of delimiters, and if the string has
9469 balanced delimiters ([{<>}]) it allows nesting.
9471 On success, the SV with the resulting string is put into lex_stuff or,
9472 if that is already non-NULL, into lex_repl. The second case occurs only
9473 when parsing the RHS of the special constructs s/// and tr/// (y///).
9474 For convenience, the terminating delimiter character is stuffed into
9479 S_scan_str(pTHX_ char *start, int keep_bracketed_quoted, int keep_delims, int re_reparse,
9483 SV *sv; /* scalar value: string */
9484 const char *tmps; /* temp string, used for delimiter matching */
9485 char *s = start; /* current position in the buffer */
9486 char term; /* terminating character */
9487 char *to; /* current position in the sv's data */
9488 I32 brackets = 1; /* bracket nesting level */
9489 bool has_utf8 = FALSE; /* is there any utf8 content? */
9490 I32 termcode; /* terminating char. code */
9491 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9492 STRLEN termlen; /* length of terminating string */
9493 int last_off = 0; /* last position for nesting bracket */
9496 PERL_ARGS_ASSERT_SCAN_STR;
9498 /* skip space before the delimiter */
9503 /* mark where we are, in case we need to report errors */
9506 /* after skipping whitespace, the next character is the terminator */
9509 termcode = termstr[0] = term;
9513 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
9514 Copy(s, termstr, termlen, U8);
9515 if (!UTF8_IS_INVARIANT(term))
9519 /* mark where we are */
9520 PL_multi_start = CopLINE(PL_curcop);
9521 PL_multi_open = term;
9522 herelines = PL_parser->herelines;
9524 /* find corresponding closing delimiter */
9525 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9526 termcode = termstr[0] = term = tmps[5];
9528 PL_multi_close = term;
9530 if (PL_multi_open == PL_multi_close) {
9531 keep_bracketed_quoted = FALSE;
9534 /* create a new SV to hold the contents. 79 is the SV's initial length.
9535 What a random number. */
9536 sv = newSV_type(SVt_PVIV);
9538 SvIV_set(sv, termcode);
9539 (void)SvPOK_only(sv); /* validate pointer */
9541 /* move past delimiter and try to read a complete string */
9543 sv_catpvn(sv, s, termlen);
9546 if (PL_encoding && !UTF && !re_reparse) {
9550 int offset = s - SvPVX_const(PL_linestr);
9551 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9552 &offset, (char*)termstr, termlen);
9556 if (SvIsCOW(PL_linestr)) {
9557 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos;
9558 STRLEN oldoldbufptr_pos, linestart_pos, last_uni_pos;
9559 STRLEN last_lop_pos, re_eval_start_pos, s_pos;
9560 char *buf = SvPVX(PL_linestr);
9561 bufend_pos = PL_parser->bufend - buf;
9562 bufptr_pos = PL_parser->bufptr - buf;
9563 oldbufptr_pos = PL_parser->oldbufptr - buf;
9564 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
9565 linestart_pos = PL_parser->linestart - buf;
9566 last_uni_pos = PL_parser->last_uni
9567 ? PL_parser->last_uni - buf
9569 last_lop_pos = PL_parser->last_lop
9570 ? PL_parser->last_lop - buf
9573 PL_parser->lex_shared->re_eval_start ?
9574 PL_parser->lex_shared->re_eval_start - buf : 0;
9577 sv_force_normal(PL_linestr);
9579 buf = SvPVX(PL_linestr);
9580 PL_parser->bufend = buf + bufend_pos;
9581 PL_parser->bufptr = buf + bufptr_pos;
9582 PL_parser->oldbufptr = buf + oldbufptr_pos;
9583 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
9584 PL_parser->linestart = buf + linestart_pos;
9585 if (PL_parser->last_uni)
9586 PL_parser->last_uni = buf + last_uni_pos;
9587 if (PL_parser->last_lop)
9588 PL_parser->last_lop = buf + last_lop_pos;
9589 if (PL_parser->lex_shared->re_eval_start)
9590 PL_parser->lex_shared->re_eval_start =
9591 buf + re_eval_start_pos;
9594 ns = SvPVX_const(PL_linestr) + offset;
9595 svlast = SvEND(sv) - 1;
9597 for (; s < ns; s++) {
9598 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9599 COPLINE_INC_WITH_HERELINES;
9602 goto read_more_line;
9604 /* handle quoted delimiters */
9605 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9607 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9609 if ((svlast-1 - t) % 2) {
9610 if (!keep_bracketed_quoted) {
9613 SvCUR_set(sv, SvCUR(sv) - 1);
9618 if (PL_multi_open == PL_multi_close) {
9624 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9625 /* At here, all closes are "was quoted" one,
9626 so we don't check PL_multi_close. */
9628 if (!keep_bracketed_quoted && *(t+1) == PL_multi_open)
9633 else if (*t == PL_multi_open)
9641 SvCUR_set(sv, w - SvPVX_const(sv));
9643 last_off = w - SvPVX(sv);
9644 if (--brackets <= 0)
9650 SvCUR_set(sv, SvCUR(sv) - 1);
9656 /* extend sv if need be */
9657 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9658 /* set 'to' to the next character in the sv's string */
9659 to = SvPVX(sv)+SvCUR(sv);
9661 /* if open delimiter is the close delimiter read unbridle */
9662 if (PL_multi_open == PL_multi_close) {
9663 for (; s < PL_bufend; s++,to++) {
9664 /* embedded newlines increment the current line number */
9665 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9666 COPLINE_INC_WITH_HERELINES;
9667 /* handle quoted delimiters */
9668 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9669 if (!keep_bracketed_quoted
9671 || (re_reparse && s[1] == '\\'))
9674 else /* any other quotes are simply copied straight through */
9677 /* terminate when run out of buffer (the for() condition), or
9678 have found the terminator */
9679 else if (*s == term) {
9682 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9685 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9691 /* if the terminator isn't the same as the start character (e.g.,
9692 matched brackets), we have to allow more in the quoting, and
9693 be prepared for nested brackets.
9696 /* read until we run out of string, or we find the terminator */
9697 for (; s < PL_bufend; s++,to++) {
9698 /* embedded newlines increment the line count */
9699 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9700 COPLINE_INC_WITH_HERELINES;
9701 /* backslashes can escape the open or closing characters */
9702 if (*s == '\\' && s+1 < PL_bufend) {
9703 if (!keep_bracketed_quoted &&
9704 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9711 /* allow nested opens and closes */
9712 else if (*s == PL_multi_close && --brackets <= 0)
9714 else if (*s == PL_multi_open)
9716 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9721 /* terminate the copied string and update the sv's end-of-string */
9723 SvCUR_set(sv, to - SvPVX_const(sv));
9726 * this next chunk reads more into the buffer if we're not done yet
9730 break; /* handle case where we are done yet :-) */
9732 #ifndef PERL_STRICT_CR
9733 if (to - SvPVX_const(sv) >= 2) {
9734 if ((to[-2] == '\r' && to[-1] == '\n') ||
9735 (to[-2] == '\n' && to[-1] == '\r'))
9739 SvCUR_set(sv, to - SvPVX_const(sv));
9741 else if (to[-1] == '\r')
9744 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9749 /* if we're out of file, or a read fails, bail and reset the current
9750 line marker so we can report where the unterminated string began
9752 COPLINE_INC_WITH_HERELINES;
9753 PL_bufptr = PL_bufend;
9754 if (!lex_next_chunk(0)) {
9756 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9762 /* at this point, we have successfully read the delimited string */
9764 if (!PL_encoding || UTF || re_reparse) {
9767 sv_catpvn(sv, s, termlen);
9770 if (has_utf8 || (PL_encoding && !re_reparse))
9773 PL_multi_end = CopLINE(PL_curcop);
9774 CopLINE_set(PL_curcop, PL_multi_start);
9775 PL_parser->herelines = herelines;
9777 /* if we allocated too much space, give some back */
9778 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9779 SvLEN_set(sv, SvCUR(sv) + 1);
9780 SvPV_renew(sv, SvLEN(sv));
9783 /* decide whether this is the first or second quoted string we've read
9788 PL_sublex_info.repl = sv;
9791 if (delimp) *delimp = PL_multi_open == PL_multi_close ? s-termlen : s;
9797 takes: pointer to position in buffer
9798 returns: pointer to new position in buffer
9799 side-effects: builds ops for the constant in pl_yylval.op
9801 Read a number in any of the formats that Perl accepts:
9803 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
9804 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
9805 0b[01](_?[01])* binary integers
9806 0[0-7](_?[0-7])* octal integers
9807 0x[0-9A-Fa-f](_?[0-9A-Fa-f])* hexadecimal integers
9808 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*(?:\.\d*)?p[+-]?[0-9]+ hexadecimal floats
9810 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
9813 If it reads a number without a decimal point or an exponent, it will
9814 try converting the number to an integer and see if it can do so
9815 without loss of precision.
9819 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
9821 const char *s = start; /* current position in buffer */
9822 char *d; /* destination in temp buffer */
9823 char *e; /* end of temp buffer */
9824 NV nv; /* number read, as a double */
9825 SV *sv = NULL; /* place to put the converted number */
9826 bool floatit; /* boolean: int or float? */
9827 const char *lastub = NULL; /* position of last underbar */
9828 static const char* const number_too_long = "Number too long";
9829 /* Hexadecimal floating point.
9831 * In many places (where we have quads and NV is IEEE 754 double)
9832 * we can fit the mantissa bits of a NV into an unsigned quad.
9833 * (Note that UVs might not be quads even when we have quads.)
9834 * This will not work everywhere, though (either no quads, or
9835 * using long doubles), in which case we have to resort to NV,
9836 * which will probably mean horrible loss of precision due to
9837 * multiple fp operations. */
9840 #if NVSIZE == 8 && defined(HAS_QUAD) && defined(Uquad_t)
9841 # define HEXFP_UQUAD
9842 Uquad_t hexfp_uquad = 0;
9843 int hexfp_frac_bits = 0;
9848 NV hexfp_mult = 1.0;
9849 UV high_non_zero = 0; /* highest digit */
9851 PERL_ARGS_ASSERT_SCAN_NUM;
9853 /* We use the first character to decide what type of number this is */
9857 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
9859 /* if it starts with a 0, it could be an octal number, a decimal in
9860 0.13 disguise, or a hexadecimal number, or a binary number. */
9864 u holds the "number so far"
9865 shift the power of 2 of the base
9866 (hex == 4, octal == 3, binary == 1)
9867 overflowed was the number more than we can hold?
9869 Shift is used when we add a digit. It also serves as an "are
9870 we in octal/hex/binary?" indicator to disallow hex characters
9876 bool overflowed = FALSE;
9877 bool just_zero = TRUE; /* just plain 0 or binary number? */
9878 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
9879 static const char* const bases[5] =
9880 { "", "binary", "", "octal", "hexadecimal" };
9881 static const char* const Bases[5] =
9882 { "", "Binary", "", "Octal", "Hexadecimal" };
9883 static const char* const maxima[5] =
9885 "0b11111111111111111111111111111111",
9889 const char *base, *Base, *max;
9892 if (isALPHA_FOLD_EQ(s[1], 'x')) {
9896 } else if (isALPHA_FOLD_EQ(s[1], 'b')) {
9901 /* check for a decimal in disguise */
9902 else if (s[1] == '.' || isALPHA_FOLD_EQ(s[1], 'e'))
9904 /* so it must be octal */
9911 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
9912 "Misplaced _ in number");
9916 base = bases[shift];
9917 Base = Bases[shift];
9918 max = maxima[shift];
9920 /* read the rest of the number */
9922 /* x is used in the overflow test,
9923 b is the digit we're adding on. */
9928 /* if we don't mention it, we're done */
9932 /* _ are ignored -- but warned about if consecutive */
9934 if (lastub && s == lastub + 1)
9935 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
9936 "Misplaced _ in number");
9940 /* 8 and 9 are not octal */
9943 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
9947 case '2': case '3': case '4':
9948 case '5': case '6': case '7':
9950 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
9954 b = *s++ & 15; /* ASCII digit -> value of digit */
9958 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
9959 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
9960 /* make sure they said 0x */
9965 /* Prepare to put the digit we have onto the end
9966 of the number so far. We check for overflows.
9972 x = u << shift; /* make room for the digit */
9974 total_bits += shift;
9976 if ((x >> shift) != u
9977 && !(PL_hints & HINT_NEW_BINARY)) {
9980 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
9981 "Integer overflow in %s number",
9984 u = x | b; /* add the digit to the end */
9987 n *= nvshift[shift];
9988 /* If an NV has not enough bits in its
9989 * mantissa to represent an UV this summing of
9990 * small low-order numbers is a waste of time
9991 * (because the NV cannot preserve the
9992 * low-order bits anyway): we could just
9993 * remember when did we overflow and in the
9994 * end just multiply n by the right
9999 if (high_non_zero == 0 && b > 0)
10002 /* this could be hexfp, but peek ahead
10003 * to avoid matching ".." */
10004 if (UNLIKELY(HEXFP_PEEK(s))) {
10012 /* if we get here, we had success: make a scalar value from
10017 /* final misplaced underbar check */
10018 if (s[-1] == '_') {
10019 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10022 if (UNLIKELY(HEXFP_PEEK(s))) {
10023 /* Do sloppy (on the underbars) but quick detection
10024 * (and value construction) for hexfp, the decimal
10025 * detection will shortly be more thorough with the
10026 * underbar checks. */
10030 #else /* HEXFP_NV */
10035 NV mult = 1 / 16.0;
10038 while (isXDIGIT(*h) || *h == '_') {
10039 if (isXDIGIT(*h)) {
10040 U8 b = XDIGIT_VALUE(*h);
10041 total_bits += shift;
10043 hexfp_uquad <<= shift;
10045 hexfp_frac_bits += shift;
10046 #else /* HEXFP_NV */
10047 hexfp_nv += b * mult;
10055 if (total_bits >= 4) {
10056 if (high_non_zero < 0x8)
10058 if (high_non_zero < 0x4)
10060 if (high_non_zero < 0x2)
10064 if (total_bits > 0 && (isALPHA_FOLD_EQ(*h, 'p'))) {
10065 bool negexp = FALSE;
10069 else if (*h == '-') {
10075 while (isDIGIT(*h) || *h == '_') {
10078 hexfp_exp += *h - '0';
10081 -hexfp_exp < NV_MIN_EXP - 1) {
10082 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10083 "Hexadecimal float: exponent underflow");
10090 hexfp_exp > NV_MAX_EXP - 1) {
10091 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10092 "Hexadecimal float: exponent overflow");
10101 hexfp_exp = -hexfp_exp;
10103 hexfp_exp -= hexfp_frac_bits;
10105 hexfp_mult = pow(2.0, hexfp_exp);
10113 if (n > 4294967295.0)
10114 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10115 "%s number > %s non-portable",
10121 if (u > 0xffffffff)
10122 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10123 "%s number > %s non-portable",
10128 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10129 sv = new_constant(start, s - start, "integer",
10130 sv, NULL, NULL, 0);
10131 else if (PL_hints & HINT_NEW_BINARY)
10132 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10137 handle decimal numbers.
10138 we're also sent here when we read a 0 as the first digit
10140 case '1': case '2': case '3': case '4': case '5':
10141 case '6': case '7': case '8': case '9': case '.':
10144 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10153 /* read next group of digits and _ and copy into d */
10154 while (isDIGIT(*s) || *s == '_' ||
10155 UNLIKELY(hexfp && isXDIGIT(*s))) {
10156 /* skip underscores, checking for misplaced ones
10160 if (lastub && s == lastub + 1)
10161 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10162 "Misplaced _ in number");
10166 /* check for end of fixed-length buffer */
10168 Perl_croak(aTHX_ "%s", number_too_long);
10169 /* if we're ok, copy the character */
10174 /* final misplaced underbar check */
10175 if (lastub && s == lastub + 1) {
10176 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10179 /* read a decimal portion if there is one. avoid
10180 3..5 being interpreted as the number 3. followed
10183 if (*s == '.' && s[1] != '.') {
10188 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10189 "Misplaced _ in number");
10193 /* copy, ignoring underbars, until we run out of digits.
10195 for (; isDIGIT(*s) || *s == '_' ||
10196 UNLIKELY(hexfp && isXDIGIT(*s));
10198 /* fixed length buffer check */
10200 Perl_croak(aTHX_ "%s", number_too_long);
10202 if (lastub && s == lastub + 1)
10203 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10204 "Misplaced _ in number");
10210 /* fractional part ending in underbar? */
10211 if (s[-1] == '_') {
10212 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10213 "Misplaced _ in number");
10215 if (*s == '.' && isDIGIT(s[1])) {
10216 /* oops, it's really a v-string, but without the "v" */
10222 /* read exponent part, if present */
10223 if ((isALPHA_FOLD_EQ(*s, 'e')
10224 || UNLIKELY(hexfp && isALPHA_FOLD_EQ(*s, 'p')))
10225 && strchr("+-0123456789_", s[1]))
10229 /* regardless of whether user said 3E5 or 3e5, use lower 'e',
10230 ditto for p (hexfloats) */
10231 if ((isALPHA_FOLD_EQ(*s, 'e'))) {
10232 /* At least some Mach atof()s don't grok 'E' */
10235 else if (UNLIKELY(hexfp && (isALPHA_FOLD_EQ(*s, 'p')))) {
10242 /* stray preinitial _ */
10244 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10245 "Misplaced _ in number");
10249 /* allow positive or negative exponent */
10250 if (*s == '+' || *s == '-')
10253 /* stray initial _ */
10255 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10256 "Misplaced _ in number");
10260 /* read digits of exponent */
10261 while (isDIGIT(*s) || *s == '_') {
10264 Perl_croak(aTHX_ "%s", number_too_long);
10268 if (((lastub && s == lastub + 1) ||
10269 (!isDIGIT(s[1]) && s[1] != '_')))
10270 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10271 "Misplaced _ in number");
10279 We try to do an integer conversion first if no characters
10280 indicating "float" have been found.
10285 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10287 if (flags == IS_NUMBER_IN_UV) {
10289 sv = newSViv(uv); /* Prefer IVs over UVs. */
10292 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10293 if (uv <= (UV) IV_MIN)
10294 sv = newSViv(-(IV)uv);
10301 STORE_NUMERIC_LOCAL_SET_STANDARD();
10302 /* terminate the string */
10304 if (UNLIKELY(hexfp)) {
10305 # ifdef NV_MANT_DIG
10306 if (total_bits > NV_MANT_DIG)
10307 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10308 "Hexadecimal float: mantissa overflow");
10311 nv = hexfp_uquad * hexfp_mult;
10312 #else /* HEXFP_NV */
10313 nv = hexfp_nv * hexfp_mult;
10316 nv = Atof(PL_tokenbuf);
10318 RESTORE_NUMERIC_LOCAL();
10323 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10324 const char *const key = floatit ? "float" : "integer";
10325 const STRLEN keylen = floatit ? 5 : 7;
10326 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10327 key, keylen, sv, NULL, NULL, 0);
10331 /* if it starts with a v, it could be a v-string */
10334 sv = newSV(5); /* preallocate storage space */
10335 ENTER_with_name("scan_vstring");
10337 s = scan_vstring(s, PL_bufend, sv);
10338 SvREFCNT_inc_simple_void_NN(sv);
10339 LEAVE_with_name("scan_vstring");
10343 /* make the op for the constant and return */
10346 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10348 lvalp->opval = NULL;
10354 S_scan_formline(pTHX_ char *s)
10358 SV * const stuff = newSVpvs("");
10359 bool needargs = FALSE;
10360 bool eofmt = FALSE;
10362 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10364 while (!needargs) {
10367 #ifdef PERL_STRICT_CR
10368 while (SPACE_OR_TAB(*t))
10371 while (SPACE_OR_TAB(*t) || *t == '\r')
10374 if (*t == '\n' || t == PL_bufend) {
10379 eol = (char *) memchr(s,'\n',PL_bufend-s);
10383 for (t = s; t < eol; t++) {
10384 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10386 goto enough; /* ~~ must be first line in formline */
10388 if (*t == '@' || *t == '^')
10392 sv_catpvn(stuff, s, eol-s);
10393 #ifndef PERL_STRICT_CR
10394 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10395 char *end = SvPVX(stuff) + SvCUR(stuff);
10398 SvCUR_set(stuff, SvCUR(stuff) - 1);
10406 if ((PL_rsfp || PL_parser->filtered)
10407 && PL_parser->form_lex_state == LEX_NORMAL) {
10409 PL_bufptr = PL_bufend;
10410 COPLINE_INC_WITH_HERELINES;
10411 got_some = lex_next_chunk(0);
10412 CopLINE_dec(PL_curcop);
10420 if (!SvCUR(stuff) || needargs)
10421 PL_lex_state = PL_parser->form_lex_state;
10422 if (SvCUR(stuff)) {
10423 PL_expect = XSTATE;
10425 const char *s2 = s;
10426 while (*s2 == '\r' || *s2 == ' ' || *s2 == '\t' || *s2 == '\f'
10430 PL_expect = XTERMBLOCK;
10431 NEXTVAL_NEXTTOKE.ival = 0;
10434 NEXTVAL_NEXTTOKE.ival = 0;
10435 force_next(FORMLBRACK);
10438 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10440 else if (PL_encoding)
10441 sv_recode_to_utf8(stuff, PL_encoding);
10443 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10447 SvREFCNT_dec(stuff);
10449 PL_lex_formbrack = 0;
10455 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10457 const I32 oldsavestack_ix = PL_savestack_ix;
10458 CV* const outsidecv = PL_compcv;
10460 SAVEI32(PL_subline);
10461 save_item(PL_subname);
10462 SAVESPTR(PL_compcv);
10464 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10465 CvFLAGS(PL_compcv) |= flags;
10467 PL_subline = CopLINE(PL_curcop);
10468 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10469 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10470 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10471 if (outsidecv && CvPADLIST(outsidecv))
10472 CvPADLIST(PL_compcv)->xpadl_outid =
10473 PadlistNAMES(CvPADLIST(outsidecv));
10475 return oldsavestack_ix;
10479 S_yywarn(pTHX_ const char *const s, U32 flags)
10481 PERL_ARGS_ASSERT_YYWARN;
10483 PL_in_eval |= EVAL_WARNONLY;
10484 yyerror_pv(s, flags);
10485 PL_in_eval &= ~EVAL_WARNONLY;
10490 Perl_yyerror(pTHX_ const char *const s)
10492 PERL_ARGS_ASSERT_YYERROR;
10493 return yyerror_pvn(s, strlen(s), 0);
10497 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
10499 PERL_ARGS_ASSERT_YYERROR_PV;
10500 return yyerror_pvn(s, strlen(s), flags);
10504 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
10506 const char *context = NULL;
10509 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
10510 int yychar = PL_parser->yychar;
10512 PERL_ARGS_ASSERT_YYERROR_PVN;
10514 if (!yychar || (yychar == ';' && !PL_rsfp))
10515 sv_catpvs(where_sv, "at EOF");
10516 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10517 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10518 PL_oldbufptr != PL_bufptr) {
10521 The code below is removed for NetWare because it abends/crashes on NetWare
10522 when the script has error such as not having the closing quotes like:
10523 if ($var eq "value)
10524 Checking of white spaces is anyway done in NetWare code.
10527 while (isSPACE(*PL_oldoldbufptr))
10530 context = PL_oldoldbufptr;
10531 contlen = PL_bufptr - PL_oldoldbufptr;
10533 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10534 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10537 The code below is removed for NetWare because it abends/crashes on NetWare
10538 when the script has error such as not having the closing quotes like:
10539 if ($var eq "value)
10540 Checking of white spaces is anyway done in NetWare code.
10543 while (isSPACE(*PL_oldbufptr))
10546 context = PL_oldbufptr;
10547 contlen = PL_bufptr - PL_oldbufptr;
10549 else if (yychar > 255)
10550 sv_catpvs(where_sv, "next token ???");
10551 else if (yychar == -2) { /* YYEMPTY */
10552 if (PL_lex_state == LEX_NORMAL ||
10553 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10554 sv_catpvs(where_sv, "at end of line");
10555 else if (PL_lex_inpat)
10556 sv_catpvs(where_sv, "within pattern");
10558 sv_catpvs(where_sv, "within string");
10561 sv_catpvs(where_sv, "next char ");
10563 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10564 else if (isPRINT_LC(yychar)) {
10565 const char string = yychar;
10566 sv_catpvn(where_sv, &string, 1);
10569 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10571 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
10572 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10573 OutCopFILE(PL_curcop),
10574 (IV)(PL_parser->preambling == NOLINE
10575 ? CopLINE(PL_curcop)
10576 : PL_parser->preambling));
10578 Perl_sv_catpvf(aTHX_ msg, "near \"%"UTF8f"\"\n",
10579 UTF8fARG(UTF, contlen, context));
10581 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
10582 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10583 Perl_sv_catpvf(aTHX_ msg,
10584 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10585 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10588 if (PL_in_eval & EVAL_WARNONLY) {
10589 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10593 if (PL_error_count >= 10) {
10595 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
10596 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10597 SVfARG(errsv), OutCopFILE(PL_curcop));
10599 Perl_croak(aTHX_ "%s has too many errors.\n",
10600 OutCopFILE(PL_curcop));
10603 PL_in_my_stash = NULL;
10608 S_swallow_bom(pTHX_ U8 *s)
10610 const STRLEN slen = SvCUR(PL_linestr);
10612 PERL_ARGS_ASSERT_SWALLOW_BOM;
10616 if (s[1] == 0xFE) {
10617 /* UTF-16 little-endian? (or UTF-32LE?) */
10618 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10619 /* diag_listed_as: Unsupported script encoding %s */
10620 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10621 #ifndef PERL_NO_UTF16_FILTER
10622 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10624 if (PL_bufend > (char*)s) {
10625 s = add_utf16_textfilter(s, TRUE);
10628 /* diag_listed_as: Unsupported script encoding %s */
10629 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10634 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10635 #ifndef PERL_NO_UTF16_FILTER
10636 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10638 if (PL_bufend > (char *)s) {
10639 s = add_utf16_textfilter(s, FALSE);
10642 /* diag_listed_as: Unsupported script encoding %s */
10643 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10647 case BOM_UTF8_FIRST_BYTE: {
10648 const STRLEN len = sizeof(BOM_UTF8_TAIL) - 1; /* Exclude trailing NUL */
10649 if (slen > len && memEQ(s+1, BOM_UTF8_TAIL, len)) {
10650 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10651 s += len + 1; /* UTF-8 */
10658 if (s[2] == 0xFE && s[3] == 0xFF) {
10659 /* UTF-32 big-endian */
10660 /* diag_listed_as: Unsupported script encoding %s */
10661 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10664 else if (s[2] == 0 && s[3] != 0) {
10667 * are a good indicator of UTF-16BE. */
10668 #ifndef PERL_NO_UTF16_FILTER
10669 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10670 s = add_utf16_textfilter(s, FALSE);
10672 /* diag_listed_as: Unsupported script encoding %s */
10673 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10680 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10683 * are a good indicator of UTF-16LE. */
10684 #ifndef PERL_NO_UTF16_FILTER
10685 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10686 s = add_utf16_textfilter(s, TRUE);
10688 /* diag_listed_as: Unsupported script encoding %s */
10689 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10697 #ifndef PERL_NO_UTF16_FILTER
10699 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10701 SV *const filter = FILTER_DATA(idx);
10702 /* We re-use this each time round, throwing the contents away before we
10704 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10705 SV *const utf8_buffer = filter;
10706 IV status = IoPAGE(filter);
10707 const bool reverse = cBOOL(IoLINES(filter));
10710 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10712 /* As we're automatically added, at the lowest level, and hence only called
10713 from this file, we can be sure that we're not called in block mode. Hence
10714 don't bother writing code to deal with block mode. */
10716 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10719 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10721 DEBUG_P(PerlIO_printf(Perl_debug_log,
10722 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10723 FPTR2DPTR(void *, S_utf16_textfilter),
10724 reverse ? 'l' : 'b', idx, maxlen, status,
10725 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10732 /* First, look in our buffer of existing UTF-8 data: */
10733 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10737 } else if (status == 0) {
10739 IoPAGE(filter) = 0;
10740 nl = SvEND(utf8_buffer);
10743 STRLEN got = nl - SvPVX(utf8_buffer);
10744 /* Did we have anything to append? */
10746 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10747 /* Everything else in this code works just fine if SVp_POK isn't
10748 set. This, however, needs it, and we need it to work, else
10749 we loop infinitely because the buffer is never consumed. */
10750 sv_chop(utf8_buffer, nl);
10754 /* OK, not a complete line there, so need to read some more UTF-16.
10755 Read an extra octect if the buffer currently has an odd number. */
10759 if (SvCUR(utf16_buffer) >= 2) {
10760 /* Location of the high octet of the last complete code point.
10761 Gosh, UTF-16 is a pain. All the benefits of variable length,
10762 *coupled* with all the benefits of partial reads and
10764 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10765 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10767 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10771 /* We have the first half of a surrogate. Read more. */
10772 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10775 status = FILTER_READ(idx + 1, utf16_buffer,
10776 160 + (SvCUR(utf16_buffer) & 1));
10777 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10778 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10781 IoPAGE(filter) = status;
10786 chars = SvCUR(utf16_buffer) >> 1;
10787 have = SvCUR(utf8_buffer);
10788 SvGROW(utf8_buffer, have + chars * 3 + 1);
10791 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10792 (U8*)SvPVX_const(utf8_buffer) + have,
10793 chars * 2, &newlen);
10795 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10796 (U8*)SvPVX_const(utf8_buffer) + have,
10797 chars * 2, &newlen);
10799 SvCUR_set(utf8_buffer, have + newlen);
10802 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10803 it's private to us, and utf16_to_utf8{,reversed} take a
10804 (pointer,length) pair, rather than a NUL-terminated string. */
10805 if(SvCUR(utf16_buffer) & 1) {
10806 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10807 SvCUR_set(utf16_buffer, 1);
10809 SvCUR_set(utf16_buffer, 0);
10812 DEBUG_P(PerlIO_printf(Perl_debug_log,
10813 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10815 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10816 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
10821 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
10823 SV *filter = filter_add(S_utf16_textfilter, NULL);
10825 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
10827 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
10828 sv_setpvs(filter, "");
10829 IoLINES(filter) = reversed;
10830 IoPAGE(filter) = 1; /* Not EOF */
10832 /* Sadly, we have to return a valid pointer, come what may, so we have to
10833 ignore any error return from this. */
10834 SvCUR_set(PL_linestr, 0);
10835 if (FILTER_READ(0, PL_linestr, 0)) {
10836 SvUTF8_on(PL_linestr);
10838 SvUTF8_on(PL_linestr);
10840 PL_bufend = SvEND(PL_linestr);
10841 return (U8*)SvPVX(PL_linestr);
10846 Returns a pointer to the next character after the parsed
10847 vstring, as well as updating the passed in sv.
10849 Function must be called like
10851 sv = sv_2mortal(newSV(5));
10852 s = scan_vstring(s,e,sv);
10854 where s and e are the start and end of the string.
10855 The sv should already be large enough to store the vstring
10856 passed in, for performance reasons.
10858 This function may croak if fatal warnings are enabled in the
10859 calling scope, hence the sv_2mortal in the example (to prevent
10860 a leak). Make sure to do SvREFCNT_inc afterwards if you use
10866 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
10868 const char *pos = s;
10869 const char *start = s;
10871 PERL_ARGS_ASSERT_SCAN_VSTRING;
10873 if (*pos == 'v') pos++; /* get past 'v' */
10874 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10876 if ( *pos != '.') {
10877 /* this may not be a v-string if followed by => */
10878 const char *next = pos;
10879 while (next < e && isSPACE(*next))
10881 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
10882 /* return string not v-string */
10883 sv_setpvn(sv,(char *)s,pos-s);
10884 return (char *)pos;
10888 if (!isALPHA(*pos)) {
10889 U8 tmpbuf[UTF8_MAXBYTES+1];
10892 s++; /* get past 'v' */
10897 /* this is atoi() that tolerates underscores */
10900 const char *end = pos;
10902 while (--end >= s) {
10904 const UV orev = rev;
10905 rev += (*end - '0') * mult;
10908 /* diag_listed_as: Integer overflow in %s number */
10909 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10910 "Integer overflow in decimal number");
10914 if (rev > 0x7FFFFFFF)
10915 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10917 /* Append native character for the rev point */
10918 tmpend = uvchr_to_utf8(tmpbuf, rev);
10919 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10920 if (!UVCHR_IS_INVARIANT(rev))
10922 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
10928 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10932 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
10939 Perl_keyword_plugin_standard(pTHX_
10940 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
10942 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
10943 PERL_UNUSED_CONTEXT;
10944 PERL_UNUSED_ARG(keyword_ptr);
10945 PERL_UNUSED_ARG(keyword_len);
10946 PERL_UNUSED_ARG(op_ptr);
10947 return KEYWORD_PLUGIN_DECLINE;
10950 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
10952 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
10954 SAVEI32(PL_lex_brackets);
10955 if (PL_lex_brackets > 100)
10956 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
10957 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
10958 SAVEI32(PL_lex_allbrackets);
10959 PL_lex_allbrackets = 0;
10960 SAVEI8(PL_lex_fakeeof);
10961 PL_lex_fakeeof = (U8)fakeeof;
10962 if(yyparse(gramtype) && !PL_parser->error_count)
10963 qerror(Perl_mess(aTHX_ "Parse error"));
10966 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
10968 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
10972 SAVEVPTR(PL_eval_root);
10973 PL_eval_root = NULL;
10974 parse_recdescent(gramtype, fakeeof);
10980 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
10982 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
10985 if (flags & ~PARSE_OPTIONAL)
10986 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
10987 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
10988 if (!exprop && !(flags & PARSE_OPTIONAL)) {
10989 if (!PL_parser->error_count)
10990 qerror(Perl_mess(aTHX_ "Parse error"));
10991 exprop = newOP(OP_NULL, 0);
10997 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
10999 Parse a Perl arithmetic expression. This may contain operators of precedence
11000 down to the bit shift operators. The expression must be followed (and thus
11001 terminated) either by a comparison or lower-precedence operator or by
11002 something that would normally terminate an expression such as semicolon.
11003 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11004 otherwise it is mandatory. It is up to the caller to ensure that the
11005 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11006 the source of the code to be parsed and the lexical context for the
11009 The op tree representing the expression is returned. If an optional
11010 expression is absent, a null pointer is returned, otherwise the pointer
11013 If an error occurs in parsing or compilation, in most cases a valid op
11014 tree is returned anyway. The error is reflected in the parser state,
11015 normally resulting in a single exception at the top level of parsing
11016 which covers all the compilation errors that occurred. Some compilation
11017 errors, however, will throw an exception immediately.
11023 Perl_parse_arithexpr(pTHX_ U32 flags)
11025 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11029 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11031 Parse a Perl term expression. This may contain operators of precedence
11032 down to the assignment operators. The expression must be followed (and thus
11033 terminated) either by a comma or lower-precedence operator or by
11034 something that would normally terminate an expression such as semicolon.
11035 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11036 otherwise it is mandatory. It is up to the caller to ensure that the
11037 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11038 the source of the code to be parsed and the lexical context for the
11041 The op tree representing the expression is returned. If an optional
11042 expression is absent, a null pointer is returned, otherwise the pointer
11045 If an error occurs in parsing or compilation, in most cases a valid op
11046 tree is returned anyway. The error is reflected in the parser state,
11047 normally resulting in a single exception at the top level of parsing
11048 which covers all the compilation errors that occurred. Some compilation
11049 errors, however, will throw an exception immediately.
11055 Perl_parse_termexpr(pTHX_ U32 flags)
11057 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11061 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11063 Parse a Perl list expression. This may contain operators of precedence
11064 down to the comma operator. The expression must be followed (and thus
11065 terminated) either by a low-precedence logic operator such as C<or> or by
11066 something that would normally terminate an expression such as semicolon.
11067 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11068 otherwise it is mandatory. It is up to the caller to ensure that the
11069 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11070 the source of the code to be parsed and the lexical context for the
11073 The op tree representing the expression is returned. If an optional
11074 expression is absent, a null pointer is returned, otherwise the pointer
11077 If an error occurs in parsing or compilation, in most cases a valid op
11078 tree is returned anyway. The error is reflected in the parser state,
11079 normally resulting in a single exception at the top level of parsing
11080 which covers all the compilation errors that occurred. Some compilation
11081 errors, however, will throw an exception immediately.
11087 Perl_parse_listexpr(pTHX_ U32 flags)
11089 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11093 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11095 Parse a single complete Perl expression. This allows the full
11096 expression grammar, including the lowest-precedence operators such
11097 as C<or>. The expression must be followed (and thus terminated) by a
11098 token that an expression would normally be terminated by: end-of-file,
11099 closing bracketing punctuation, semicolon, or one of the keywords that
11100 signals a postfix expression-statement modifier. If I<flags> includes
11101 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11102 mandatory. It is up to the caller to ensure that the dynamic parser
11103 state (L</PL_parser> et al) is correctly set to reflect the source of
11104 the code to be parsed and the lexical context for the expression.
11106 The op tree representing the expression is returned. If an optional
11107 expression is absent, a null pointer is returned, otherwise the pointer
11110 If an error occurs in parsing or compilation, in most cases a valid op
11111 tree is returned anyway. The error is reflected in the parser state,
11112 normally resulting in a single exception at the top level of parsing
11113 which covers all the compilation errors that occurred. Some compilation
11114 errors, however, will throw an exception immediately.
11120 Perl_parse_fullexpr(pTHX_ U32 flags)
11122 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11126 =for apidoc Amx|OP *|parse_block|U32 flags
11128 Parse a single complete Perl code block. This consists of an opening
11129 brace, a sequence of statements, and a closing brace. The block
11130 constitutes a lexical scope, so C<my> variables and various compile-time
11131 effects can be contained within it. It is up to the caller to ensure
11132 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11133 reflect the source of the code to be parsed and the lexical context for
11136 The op tree representing the code block is returned. This is always a
11137 real op, never a null pointer. It will normally be a C<lineseq> list,
11138 including C<nextstate> or equivalent ops. No ops to construct any kind
11139 of runtime scope are included by virtue of it being a block.
11141 If an error occurs in parsing or compilation, in most cases a valid op
11142 tree (most likely null) is returned anyway. The error is reflected in
11143 the parser state, normally resulting in a single exception at the top
11144 level of parsing which covers all the compilation errors that occurred.
11145 Some compilation errors, however, will throw an exception immediately.
11147 The I<flags> parameter is reserved for future use, and must always
11154 Perl_parse_block(pTHX_ U32 flags)
11157 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11158 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11162 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11164 Parse a single unadorned Perl statement. This may be a normal imperative
11165 statement or a declaration that has compile-time effect. It does not
11166 include any label or other affixture. It is up to the caller to ensure
11167 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11168 reflect the source of the code to be parsed and the lexical context for
11171 The op tree representing the statement is returned. This may be a
11172 null pointer if the statement is null, for example if it was actually
11173 a subroutine definition (which has compile-time side effects). If not
11174 null, it will be ops directly implementing the statement, suitable to
11175 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11176 equivalent op (except for those embedded in a scope contained entirely
11177 within the statement).
11179 If an error occurs in parsing or compilation, in most cases a valid op
11180 tree (most likely null) is returned anyway. The error is reflected in
11181 the parser state, normally resulting in a single exception at the top
11182 level of parsing which covers all the compilation errors that occurred.
11183 Some compilation errors, however, will throw an exception immediately.
11185 The I<flags> parameter is reserved for future use, and must always
11192 Perl_parse_barestmt(pTHX_ U32 flags)
11195 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11196 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11200 =for apidoc Amx|SV *|parse_label|U32 flags
11202 Parse a single label, possibly optional, of the type that may prefix a
11203 Perl statement. It is up to the caller to ensure that the dynamic parser
11204 state (L</PL_parser> et al) is correctly set to reflect the source of
11205 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11206 label is optional, otherwise it is mandatory.
11208 The name of the label is returned in the form of a fresh scalar. If an
11209 optional label is absent, a null pointer is returned.
11211 If an error occurs in parsing, which can only occur if the label is
11212 mandatory, a valid label is returned anyway. The error is reflected in
11213 the parser state, normally resulting in a single exception at the top
11214 level of parsing which covers all the compilation errors that occurred.
11220 Perl_parse_label(pTHX_ U32 flags)
11222 if (flags & ~PARSE_OPTIONAL)
11223 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11224 if (PL_lex_state == LEX_KNOWNEXT) {
11225 PL_parser->yychar = yylex();
11226 if (PL_parser->yychar == LABEL) {
11227 char * const lpv = pl_yylval.pval;
11228 STRLEN llen = strlen(lpv);
11229 PL_parser->yychar = YYEMPTY;
11230 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11237 STRLEN wlen, bufptr_pos;
11240 if (!isIDFIRST_lazy_if(s, UTF))
11242 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11243 if (word_takes_any_delimeter(s, wlen))
11245 bufptr_pos = s - SvPVX(PL_linestr);
11247 lex_read_space(LEX_KEEP_PREVIOUS);
11249 s = SvPVX(PL_linestr) + bufptr_pos;
11250 if (t[0] == ':' && t[1] != ':') {
11251 PL_oldoldbufptr = PL_oldbufptr;
11254 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11258 if (flags & PARSE_OPTIONAL) {
11261 qerror(Perl_mess(aTHX_ "Parse error"));
11262 return newSVpvs("x");
11269 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11271 Parse a single complete Perl statement. This may be a normal imperative
11272 statement or a declaration that has compile-time effect, and may include
11273 optional labels. It is up to the caller to ensure that the dynamic
11274 parser state (L</PL_parser> et al) is correctly set to reflect the source
11275 of the code to be parsed and the lexical context for the statement.
11277 The op tree representing the statement is returned. This may be a
11278 null pointer if the statement is null, for example if it was actually
11279 a subroutine definition (which has compile-time side effects). If not
11280 null, it will be the result of a L</newSTATEOP> call, normally including
11281 a C<nextstate> or equivalent op.
11283 If an error occurs in parsing or compilation, in most cases a valid op
11284 tree (most likely null) is returned anyway. The error is reflected in
11285 the parser state, normally resulting in a single exception at the top
11286 level of parsing which covers all the compilation errors that occurred.
11287 Some compilation errors, however, will throw an exception immediately.
11289 The I<flags> parameter is reserved for future use, and must always
11296 Perl_parse_fullstmt(pTHX_ U32 flags)
11299 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11300 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11304 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11306 Parse a sequence of zero or more Perl statements. These may be normal
11307 imperative statements, including optional labels, or declarations
11308 that have compile-time effect, or any mixture thereof. The statement
11309 sequence ends when a closing brace or end-of-file is encountered in a
11310 place where a new statement could have validly started. It is up to
11311 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11312 is correctly set to reflect the source of the code to be parsed and the
11313 lexical context for the statements.
11315 The op tree representing the statement sequence is returned. This may
11316 be a null pointer if the statements were all null, for example if there
11317 were no statements or if there were only subroutine definitions (which
11318 have compile-time side effects). If not null, it will be a C<lineseq>
11319 list, normally including C<nextstate> or equivalent ops.
11321 If an error occurs in parsing or compilation, in most cases a valid op
11322 tree is returned anyway. The error is reflected in the parser state,
11323 normally resulting in a single exception at the top level of parsing
11324 which covers all the compilation errors that occurred. Some compilation
11325 errors, however, will throw an exception immediately.
11327 The I<flags> parameter is reserved for future use, and must always
11334 Perl_parse_stmtseq(pTHX_ U32 flags)
11339 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11340 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11341 c = lex_peek_unichar(0);
11342 if (c != -1 && c != /*{*/'}')
11343 qerror(Perl_mess(aTHX_ "Parse error"));
11347 #define lex_token_boundary() S_lex_token_boundary(aTHX)
11349 S_lex_token_boundary(pTHX)
11351 PL_oldoldbufptr = PL_oldbufptr;
11352 PL_oldbufptr = PL_bufptr;
11355 #define parse_opt_lexvar() S_parse_opt_lexvar(aTHX)
11357 S_parse_opt_lexvar(pTHX)
11362 lex_token_boundary();
11363 sigil = lex_read_unichar(0);
11364 if (lex_peek_unichar(0) == '#') {
11365 qerror(Perl_mess(aTHX_ "Parse error"));
11369 c = lex_peek_unichar(0);
11370 if (c == -1 || !(UTF ? isIDFIRST_uni(c) : isIDFIRST_A(c)))
11373 d = PL_tokenbuf + 1;
11374 PL_tokenbuf[0] = (char)sigil;
11375 parse_ident(&s, &d, PL_tokenbuf + sizeof(PL_tokenbuf) - 1, 0, cBOOL(UTF));
11377 if (d == PL_tokenbuf+1)
11380 var = newOP(sigil == '$' ? OP_PADSV : sigil == '@' ? OP_PADAV : OP_PADHV,
11381 OPf_MOD | (OPpLVAL_INTRO<<8));
11382 var->op_targ = allocmy(PL_tokenbuf, d - PL_tokenbuf, UTF ? SVf_UTF8 : 0);
11387 Perl_parse_subsignature(pTHX)
11390 int prev_type = 0, pos = 0, min_arity = 0, max_arity = 0;
11391 OP *initops = NULL;
11393 c = lex_peek_unichar(0);
11394 while (c != /*(*/')') {
11398 if (prev_type == 2)
11399 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11400 var = parse_opt_lexvar();
11402 newBINOP(OP_AELEM, 0,
11403 ref(newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv)),
11405 newSVOP(OP_CONST, 0, newSViv(pos))) :
11408 c = lex_peek_unichar(0);
11410 lex_token_boundary();
11411 lex_read_unichar(0);
11413 c = lex_peek_unichar(0);
11414 if (c == ',' || c == /*(*/')') {
11416 qerror(Perl_mess(aTHX_ "Optional parameter "
11417 "lacks default expression"));
11419 OP *defexpr = parse_termexpr(0);
11420 if (defexpr->op_type == OP_UNDEF &&
11421 !(defexpr->op_flags & OPf_KIDS)) {
11426 scalar(newUNOP(OP_RV2AV, 0,
11427 newGVOP(OP_GV, 0, PL_defgv))),
11428 newSVOP(OP_CONST, 0, newSViv(pos+1)));
11430 newCONDOP(0, ifop, expr, defexpr) :
11431 newLOGOP(OP_OR, 0, ifop, defexpr);
11436 if (prev_type == 1)
11437 qerror(Perl_mess(aTHX_ "Mandatory parameter "
11438 "follows optional parameter"));
11440 min_arity = pos + 1;
11442 if (var) expr = newASSIGNOP(OPf_STACKED, var, 0, expr);
11444 initops = op_append_list(OP_LINESEQ, initops,
11445 newSTATEOP(0, NULL, expr));
11451 if (prev_type == 2)
11452 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11453 var = parse_opt_lexvar();
11455 OP *chkop = newLOGOP((pos & 1) ? OP_OR : OP_AND, 0,
11456 newBINOP(OP_BIT_AND, 0,
11457 scalar(newUNOP(OP_RV2AV, 0,
11458 newGVOP(OP_GV, 0, PL_defgv))),
11459 newSVOP(OP_CONST, 0, newSViv(1))),
11460 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11461 newSVOP(OP_CONST, 0,
11462 newSVpvs("Odd name/value argument "
11463 "for subroutine"))));
11464 if (pos != min_arity)
11465 chkop = newLOGOP(OP_AND, 0,
11467 scalar(newUNOP(OP_RV2AV, 0,
11468 newGVOP(OP_GV, 0, PL_defgv))),
11469 newSVOP(OP_CONST, 0, newSViv(pos))),
11471 initops = op_append_list(OP_LINESEQ,
11472 newSTATEOP(0, NULL, chkop),
11477 op_prepend_elem(OP_ASLICE,
11478 newOP(OP_PUSHMARK, 0),
11479 newLISTOP(OP_ASLICE, 0,
11481 newSVOP(OP_CONST, 0, newSViv(pos)),
11482 newUNOP(OP_AV2ARYLEN, 0,
11483 ref(newUNOP(OP_RV2AV, 0,
11484 newGVOP(OP_GV, 0, PL_defgv)),
11486 ref(newUNOP(OP_RV2AV, 0,
11487 newGVOP(OP_GV, 0, PL_defgv)),
11489 newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv));
11490 initops = op_append_list(OP_LINESEQ, initops,
11491 newSTATEOP(0, NULL,
11492 newASSIGNOP(OPf_STACKED, var, 0, slice)));
11499 qerror(Perl_mess(aTHX_ "Parse error"));
11503 c = lex_peek_unichar(0);
11505 case /*(*/')': break;
11508 lex_token_boundary();
11509 lex_read_unichar(0);
11511 c = lex_peek_unichar(0);
11512 } while (c == ',');
11518 if (min_arity != 0) {
11519 initops = op_append_list(OP_LINESEQ,
11520 newSTATEOP(0, NULL,
11523 scalar(newUNOP(OP_RV2AV, 0,
11524 newGVOP(OP_GV, 0, PL_defgv))),
11525 newSVOP(OP_CONST, 0, newSViv(min_arity))),
11526 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11527 newSVOP(OP_CONST, 0,
11528 newSVpvs("Too few arguments for subroutine"))))),
11531 if (max_arity != -1) {
11532 initops = op_append_list(OP_LINESEQ,
11533 newSTATEOP(0, NULL,
11536 scalar(newUNOP(OP_RV2AV, 0,
11537 newGVOP(OP_GV, 0, PL_defgv))),
11538 newSVOP(OP_CONST, 0, newSViv(max_arity))),
11539 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11540 newSVOP(OP_CONST, 0,
11541 newSVpvs("Too many arguments for subroutine"))))),
11549 * c-indentation-style: bsd
11550 * c-basic-offset: 4
11551 * indent-tabs-mode: nil
11554 * ex: set ts=8 sts=4 sw=4 et: