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_expect (PL_parser->lex_expect)
58 #define PL_lex_formbrack (PL_parser->lex_formbrack)
59 #define PL_lex_inpat (PL_parser->lex_inpat)
60 #define PL_lex_inwhat (PL_parser->lex_inwhat)
61 #define PL_lex_op (PL_parser->lex_op)
62 #define PL_lex_repl (PL_parser->lex_repl)
63 #define PL_lex_starts (PL_parser->lex_starts)
64 #define PL_lex_stuff (PL_parser->lex_stuff)
65 #define PL_multi_start (PL_parser->multi_start)
66 #define PL_multi_open (PL_parser->multi_open)
67 #define PL_multi_close (PL_parser->multi_close)
68 #define PL_preambled (PL_parser->preambled)
69 #define PL_sublex_info (PL_parser->sublex_info)
70 #define PL_linestr (PL_parser->linestr)
71 #define PL_expect (PL_parser->expect)
72 #define PL_copline (PL_parser->copline)
73 #define PL_bufptr (PL_parser->bufptr)
74 #define PL_oldbufptr (PL_parser->oldbufptr)
75 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
76 #define PL_linestart (PL_parser->linestart)
77 #define PL_bufend (PL_parser->bufend)
78 #define PL_last_uni (PL_parser->last_uni)
79 #define PL_last_lop (PL_parser->last_lop)
80 #define PL_last_lop_op (PL_parser->last_lop_op)
81 #define PL_lex_state (PL_parser->lex_state)
82 #define PL_rsfp (PL_parser->rsfp)
83 #define PL_rsfp_filters (PL_parser->rsfp_filters)
84 #define PL_in_my (PL_parser->in_my)
85 #define PL_in_my_stash (PL_parser->in_my_stash)
86 #define PL_tokenbuf (PL_parser->tokenbuf)
87 #define PL_multi_end (PL_parser->multi_end)
88 #define PL_error_count (PL_parser->error_count)
90 # define PL_nexttoke (PL_parser->nexttoke)
91 # define PL_nexttype (PL_parser->nexttype)
92 # define PL_nextval (PL_parser->nextval)
94 static const char* const ident_too_long = "Identifier too long";
96 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
98 #define XENUMMASK 0x3f
100 #define XFAKEBRACK 0x80
102 #ifdef USE_UTF8_SCRIPTS
103 # define UTF (!IN_BYTES)
105 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
108 /* The maximum number of characters preceding the unrecognized one to display */
109 #define UNRECOGNIZED_PRECEDE_COUNT 10
111 /* In variables named $^X, these are the legal values for X.
112 * 1999-02-27 mjd-perl-patch@plover.com */
113 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
115 #define SPACE_OR_TAB(c) isBLANK_A(c)
117 /* LEX_* are values for PL_lex_state, the state of the lexer.
118 * They are arranged oddly so that the guard on the switch statement
119 * can get by with a single comparison (if the compiler is smart enough).
121 * These values refer to the various states within a sublex parse,
122 * i.e. within a double quotish string
125 /* #define LEX_NOTPARSING 11 is done in perl.h. */
127 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
128 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
129 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
130 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
131 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
133 /* at end of code, eg "$x" followed by: */
134 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
135 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
137 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
138 string or after \E, $foo, etc */
139 #define LEX_INTERPCONST 2 /* NOT USED */
140 #define LEX_FORMLINE 1 /* expecting a format line */
141 #define LEX_KNOWNEXT 0 /* next token known; just return it */
145 static const char* const lex_state_names[] = {
160 #include "keywords.h"
162 /* CLINE is a macro that ensures PL_copline has a sane value */
164 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
166 # define SKIPSPACE0(s) skipspace(s)
167 # define SKIPSPACE1(s) skipspace(s)
168 # define SKIPSPACE2(s,tsv) skipspace(s)
169 # define PEEKSPACE(s) skipspace(s)
172 * Convenience functions to return different tokens and prime the
173 * lexer for the next token. They all take an argument.
175 * TOKEN : generic token (used for '(', DOLSHARP, etc)
176 * OPERATOR : generic operator
177 * AOPERATOR : assignment operator
178 * PREBLOCK : beginning the block after an if, while, foreach, ...
179 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
180 * PREREF : *EXPR where EXPR is not a simple identifier
181 * TERM : expression term
182 * POSTDEREF : postfix dereference (->$* ->@[...] etc.)
183 * LOOPX : loop exiting command (goto, last, dump, etc)
184 * FTST : file test operator
185 * FUN0 : zero-argument function
186 * FUN0OP : zero-argument function, with its op created in this file
187 * FUN1 : not used, except for not, which isn't a UNIOP
188 * BOop : bitwise or or xor
190 * SHop : shift operator
191 * PWop : power operator
192 * PMop : pattern-matching operator
193 * Aop : addition-level operator
194 * Mop : multiplication-level operator
195 * Eop : equality-testing operator
196 * Rop : relational operator <= != gt
198 * Also see LOP and lop() below.
201 #ifdef DEBUGGING /* Serve -DT. */
202 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
204 # define REPORT(retval) (retval)
207 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
208 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
209 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
210 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
211 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
212 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
213 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
214 #define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
215 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
216 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
217 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
218 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
219 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
220 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
221 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
222 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
223 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
224 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
225 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
226 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
227 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
228 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
230 /* This bit of chicanery makes a unary function followed by
231 * a parenthesis into a function with one argument, highest precedence.
232 * The UNIDOR macro is for unary functions that can be followed by the //
233 * operator (such as C<shift // 0>).
235 #define UNI3(f,x,have_x) { \
236 pl_yylval.ival = f; \
237 if (have_x) PL_expect = x; \
239 PL_last_uni = PL_oldbufptr; \
240 PL_last_lop_op = f; \
242 return REPORT( (int)FUNC1 ); \
244 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
246 #define UNI(f) UNI3(f,XTERM,1)
247 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
248 #define UNIPROTO(f,optional) { \
249 if (optional) PL_last_uni = PL_oldbufptr; \
253 #define UNIBRACK(f) UNI3(f,0,0)
255 /* grandfather return to old style */
258 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
259 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
260 pl_yylval.ival = (f); \
266 #define COPLINE_INC_WITH_HERELINES \
268 CopLINE_inc(PL_curcop); \
269 if (PL_parser->herelines) \
270 CopLINE(PL_curcop) += PL_parser->herelines, \
271 PL_parser->herelines = 0; \
273 /* Called after scan_str to update CopLINE(PL_curcop), but only when there
274 * is no sublex_push to follow. */
275 #define COPLINE_SET_FROM_MULTI_END \
277 CopLINE_set(PL_curcop, PL_multi_end); \
278 if (PL_multi_end != PL_multi_start) \
279 PL_parser->herelines = 0; \
285 /* how to interpret the pl_yylval associated with the token */
289 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
294 static struct debug_tokens {
296 enum token_type type;
298 } const debug_tokens[] =
300 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
301 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
302 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
303 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
304 { ARROW, TOKENTYPE_NONE, "ARROW" },
305 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
306 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
307 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
308 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
309 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
310 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
311 { DO, TOKENTYPE_NONE, "DO" },
312 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
313 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
314 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
315 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
316 { ELSE, TOKENTYPE_NONE, "ELSE" },
317 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
318 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
319 { FOR, TOKENTYPE_IVAL, "FOR" },
320 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
321 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
322 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
323 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
324 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
325 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
326 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
327 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
328 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
329 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
330 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
331 { IF, TOKENTYPE_IVAL, "IF" },
332 { LABEL, TOKENTYPE_PVAL, "LABEL" },
333 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
334 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
335 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
336 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
337 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
338 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
339 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
340 { MY, TOKENTYPE_IVAL, "MY" },
341 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
342 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
343 { OROP, TOKENTYPE_IVAL, "OROP" },
344 { OROR, TOKENTYPE_NONE, "OROR" },
345 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
346 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
347 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
348 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
349 { POSTJOIN, TOKENTYPE_NONE, "POSTJOIN" },
350 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
351 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
352 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
353 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
354 { PREINC, TOKENTYPE_NONE, "PREINC" },
355 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
356 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
357 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
358 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
359 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
360 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
361 { SUB, TOKENTYPE_NONE, "SUB" },
362 { THING, TOKENTYPE_OPVAL, "THING" },
363 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
364 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
365 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
366 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
367 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
368 { USE, TOKENTYPE_IVAL, "USE" },
369 { WHEN, TOKENTYPE_IVAL, "WHEN" },
370 { WHILE, TOKENTYPE_IVAL, "WHILE" },
371 { WORD, TOKENTYPE_OPVAL, "WORD" },
372 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
373 { 0, TOKENTYPE_NONE, NULL }
376 /* dump the returned token in rv, plus any optional arg in pl_yylval */
379 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
383 PERL_ARGS_ASSERT_TOKEREPORT;
386 const char *name = NULL;
387 enum token_type type = TOKENTYPE_NONE;
388 const struct debug_tokens *p;
389 SV* const report = newSVpvs("<== ");
391 for (p = debug_tokens; p->token; p++) {
392 if (p->token == (int)rv) {
399 Perl_sv_catpv(aTHX_ report, name);
400 else if ((char)rv > ' ' && (char)rv <= '~')
402 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
404 sv_catpvs(report, " (pending identifier)");
407 sv_catpvs(report, "EOF");
409 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
414 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
416 case TOKENTYPE_OPNUM:
417 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
418 PL_op_name[lvalp->ival]);
421 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
423 case TOKENTYPE_OPVAL:
425 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
426 PL_op_name[lvalp->opval->op_type]);
427 if (lvalp->opval->op_type == OP_CONST) {
428 Perl_sv_catpvf(aTHX_ report, " %s",
429 SvPEEK(cSVOPx_sv(lvalp->opval)));
434 sv_catpvs(report, "(opval=null)");
437 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
443 /* print the buffer with suitable escapes */
446 S_printbuf(pTHX_ const char *const fmt, const char *const s)
448 SV* const tmp = newSVpvs("");
450 PERL_ARGS_ASSERT_PRINTBUF;
452 GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
453 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
461 S_deprecate_commaless_var_list(pTHX) {
463 deprecate("comma-less variable list");
464 return REPORT(','); /* grandfather non-comma-format format */
470 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
471 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
475 S_ao(pTHX_ int toketype)
477 if (*PL_bufptr == '=') {
479 if (toketype == ANDAND)
480 pl_yylval.ival = OP_ANDASSIGN;
481 else if (toketype == OROR)
482 pl_yylval.ival = OP_ORASSIGN;
483 else if (toketype == DORDOR)
484 pl_yylval.ival = OP_DORASSIGN;
492 * When Perl expects an operator and finds something else, no_op
493 * prints the warning. It always prints "<something> found where
494 * operator expected. It prints "Missing semicolon on previous line?"
495 * if the surprise occurs at the start of the line. "do you need to
496 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
497 * where the compiler doesn't know if foo is a method call or a function.
498 * It prints "Missing operator before end of line" if there's nothing
499 * after the missing operator, or "... before <...>" if there is something
500 * after the missing operator.
504 S_no_op(pTHX_ const char *const what, char *s)
506 char * const oldbp = PL_bufptr;
507 const bool is_first = (PL_oldbufptr == PL_linestart);
509 PERL_ARGS_ASSERT_NO_OP;
515 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
516 if (ckWARN_d(WARN_SYNTAX)) {
518 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
519 "\t(Missing semicolon on previous line?)\n");
520 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
522 for (t = PL_oldoldbufptr; (isWORDCHAR_lazy_if(t,UTF) || *t == ':');
523 t += UTF ? UTF8SKIP(t) : 1)
525 if (t < PL_bufptr && isSPACE(*t))
526 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
527 "\t(Do you need to predeclare %"UTF8f"?)\n",
528 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
532 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
533 "\t(Missing operator before %"UTF8f"?)\n",
534 UTF8fARG(UTF, s - oldbp, oldbp));
542 * Complain about missing quote/regexp/heredoc terminator.
543 * If it's called with NULL then it cauterizes the line buffer.
544 * If we're in a delimited string and the delimiter is a control
545 * character, it's reformatted into a two-char sequence like ^C.
550 S_missingterm(pTHX_ char *s)
555 char * const nl = strrchr(s,'\n');
559 else if ((U8) PL_multi_close < 32) {
561 tmpbuf[1] = (char)toCTRL(PL_multi_close);
566 *tmpbuf = (char)PL_multi_close;
570 q = strchr(s,'"') ? '\'' : '"';
571 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
577 * Check whether the named feature is enabled.
580 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
582 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
584 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
586 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
588 if (namelen > MAX_FEATURE_LEN)
590 memcpy(&he_name[8], name, namelen);
592 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
593 REFCOUNTED_HE_EXISTS));
597 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
598 * utf16-to-utf8-reversed.
601 #ifdef PERL_CR_FILTER
605 const char *s = SvPVX_const(sv);
606 const char * const e = s + SvCUR(sv);
608 PERL_ARGS_ASSERT_STRIP_RETURN;
610 /* outer loop optimized to do nothing if there are no CR-LFs */
612 if (*s++ == '\r' && *s == '\n') {
613 /* hit a CR-LF, need to copy the rest */
617 if (*s == '\r' && s[1] == '\n')
628 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
630 const I32 count = FILTER_READ(idx+1, sv, maxlen);
631 if (count > 0 && !maxlen)
638 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
640 Creates and initialises a new lexer/parser state object, supplying
641 a context in which to lex and parse from a new source of Perl code.
642 A pointer to the new state object is placed in L</PL_parser>. An entry
643 is made on the save stack so that upon unwinding the new state object
644 will be destroyed and the former value of L</PL_parser> will be restored.
645 Nothing else need be done to clean up the parsing context.
647 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
648 non-null, provides a string (in SV form) containing code to be parsed.
649 A copy of the string is made, so subsequent modification of I<line>
650 does not affect parsing. I<rsfp>, if non-null, provides an input stream
651 from which code will be read to be parsed. If both are non-null, the
652 code in I<line> comes first and must consist of complete lines of input,
653 and I<rsfp> supplies the remainder of the source.
655 The I<flags> parameter is reserved for future use. Currently it is only
656 used by perl internally, so extensions should always pass zero.
661 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
662 can share filters with the current parser.
663 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
664 caller, hence isn't owned by the parser, so shouldn't be closed on parser
665 destruction. This is used to handle the case of defaulting to reading the
666 script from the standard input because no filename was given on the command
667 line (without getting confused by situation where STDIN has been closed, so
668 the script handle is opened on fd 0) */
671 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
673 const char *s = NULL;
674 yy_parser *parser, *oparser;
675 if (flags && flags & ~LEX_START_FLAGS)
676 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
678 /* create and initialise a parser */
680 Newxz(parser, 1, yy_parser);
681 parser->old_parser = oparser = PL_parser;
684 parser->stack = NULL;
686 parser->stack_size = 0;
688 /* on scope exit, free this parser and restore any outer one */
690 parser->saved_curcop = PL_curcop;
692 /* initialise lexer state */
694 parser->nexttoke = 0;
695 parser->error_count = oparser ? oparser->error_count : 0;
696 parser->copline = parser->preambling = NOLINE;
697 parser->lex_state = LEX_NORMAL;
698 parser->expect = XSTATE;
700 parser->rsfp_filters =
701 !(flags & LEX_START_SAME_FILTER) || !oparser
703 : MUTABLE_AV(SvREFCNT_inc(
704 oparser->rsfp_filters
705 ? oparser->rsfp_filters
706 : (oparser->rsfp_filters = newAV())
709 Newx(parser->lex_brackstack, 120, char);
710 Newx(parser->lex_casestack, 12, char);
711 *parser->lex_casestack = '\0';
712 Newxz(parser->lex_shared, 1, LEXSHARED);
716 s = SvPV_const(line, len);
717 parser->linestr = flags & LEX_START_COPIED
718 ? SvREFCNT_inc_simple_NN(line)
719 : newSVpvn_flags(s, len, SvUTF8(line));
720 sv_catpvn(parser->linestr, "\n;", rsfp ? 1 : 2);
722 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
724 parser->oldoldbufptr =
727 parser->linestart = SvPVX(parser->linestr);
728 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
729 parser->last_lop = parser->last_uni = NULL;
731 assert(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
732 |LEX_DONT_CLOSE_RSFP));
733 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
734 |LEX_DONT_CLOSE_RSFP));
736 parser->in_pod = parser->filtered = 0;
740 /* delete a parser object */
743 Perl_parser_free(pTHX_ const yy_parser *parser)
745 PERL_ARGS_ASSERT_PARSER_FREE;
747 PL_curcop = parser->saved_curcop;
748 SvREFCNT_dec(parser->linestr);
750 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
751 PerlIO_clearerr(parser->rsfp);
752 else if (parser->rsfp && (!parser->old_parser ||
753 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
754 PerlIO_close(parser->rsfp);
755 SvREFCNT_dec(parser->rsfp_filters);
756 SvREFCNT_dec(parser->lex_stuff);
757 SvREFCNT_dec(parser->sublex_info.repl);
759 Safefree(parser->lex_brackstack);
760 Safefree(parser->lex_casestack);
761 Safefree(parser->lex_shared);
762 PL_parser = parser->old_parser;
767 Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
769 I32 nexttoke = parser->nexttoke;
770 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
772 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
773 && parser->nextval[nexttoke].opval
774 && parser->nextval[nexttoke].opval->op_slabbed
775 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
776 op_free(parser->nextval[nexttoke].opval);
777 parser->nextval[nexttoke].opval = NULL;
784 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
786 Buffer scalar containing the chunk currently under consideration of the
787 text currently being lexed. This is always a plain string scalar (for
788 which C<SvPOK> is true). It is not intended to be used as a scalar by
789 normal scalar means; instead refer to the buffer directly by the pointer
790 variables described below.
792 The lexer maintains various C<char*> pointers to things in the
793 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
794 reallocated, all of these pointers must be updated. Don't attempt to
795 do this manually, but rather use L</lex_grow_linestr> if you need to
796 reallocate the buffer.
798 The content of the text chunk in the buffer is commonly exactly one
799 complete line of input, up to and including a newline terminator,
800 but there are situations where it is otherwise. The octets of the
801 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
802 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
803 flag on this scalar, which may disagree with it.
805 For direct examination of the buffer, the variable
806 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
807 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
808 of these pointers is usually preferable to examination of the scalar
809 through normal scalar means.
811 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
813 Direct pointer to the end of the chunk of text currently being lexed, the
814 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
815 + SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
816 always located at the end of the buffer, and does not count as part of
817 the buffer's contents.
819 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
821 Points to the current position of lexing inside the lexer buffer.
822 Characters around this point may be freely examined, within
823 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
824 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
825 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
827 Lexing code (whether in the Perl core or not) moves this pointer past
828 the characters that it consumes. It is also expected to perform some
829 bookkeeping whenever a newline character is consumed. This movement
830 can be more conveniently performed by the function L</lex_read_to>,
831 which handles newlines appropriately.
833 Interpretation of the buffer's octets can be abstracted out by
834 using the slightly higher-level functions L</lex_peek_unichar> and
835 L</lex_read_unichar>.
837 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
839 Points to the start of the current line inside the lexer buffer.
840 This is useful for indicating at which column an error occurred, and
841 not much else. This must be updated by any lexing code that consumes
842 a newline; the function L</lex_read_to> handles this detail.
848 =for apidoc Amx|bool|lex_bufutf8
850 Indicates whether the octets in the lexer buffer
851 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
852 of Unicode characters. If not, they should be interpreted as Latin-1
853 characters. This is analogous to the C<SvUTF8> flag for scalars.
855 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
856 contains valid UTF-8. Lexing code must be robust in the face of invalid
859 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
860 is significant, but not the whole story regarding the input character
861 encoding. Normally, when a file is being read, the scalar contains octets
862 and its C<SvUTF8> flag is off, but the octets should be interpreted as
863 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
864 however, the scalar may have the C<SvUTF8> flag on, and in this case its
865 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
866 is in effect. This logic may change in the future; use this function
867 instead of implementing the logic yourself.
873 Perl_lex_bufutf8(pTHX)
879 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
881 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
882 at least I<len> octets (including terminating C<NUL>). Returns a
883 pointer to the reallocated buffer. This is necessary before making
884 any direct modification of the buffer that would increase its length.
885 L</lex_stuff_pvn> provides a more convenient way to insert text into
888 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
889 this function updates all of the lexer's variables that point directly
896 Perl_lex_grow_linestr(pTHX_ STRLEN len)
900 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
901 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
902 linestr = PL_parser->linestr;
903 buf = SvPVX(linestr);
904 if (len <= SvLEN(linestr))
906 bufend_pos = PL_parser->bufend - buf;
907 bufptr_pos = PL_parser->bufptr - buf;
908 oldbufptr_pos = PL_parser->oldbufptr - buf;
909 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
910 linestart_pos = PL_parser->linestart - buf;
911 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
912 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
913 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
914 PL_parser->lex_shared->re_eval_start - buf : 0;
916 buf = sv_grow(linestr, len);
918 PL_parser->bufend = buf + bufend_pos;
919 PL_parser->bufptr = buf + bufptr_pos;
920 PL_parser->oldbufptr = buf + oldbufptr_pos;
921 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
922 PL_parser->linestart = buf + linestart_pos;
923 if (PL_parser->last_uni)
924 PL_parser->last_uni = buf + last_uni_pos;
925 if (PL_parser->last_lop)
926 PL_parser->last_lop = buf + last_lop_pos;
927 if (PL_parser->lex_shared->re_eval_start)
928 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
933 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
935 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
936 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
937 reallocating the buffer if necessary. This means that lexing code that
938 runs later will see the characters as if they had appeared in the input.
939 It is not recommended to do this as part of normal parsing, and most
940 uses of this facility run the risk of the inserted characters being
941 interpreted in an unintended manner.
943 The string to be inserted is represented by I<len> octets starting
944 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
945 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
946 The characters are recoded for the lexer buffer, according to how the
947 buffer is currently being interpreted (L</lex_bufutf8>). If a string
948 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
949 function is more convenient.
955 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
959 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
960 if (flags & ~(LEX_STUFF_UTF8))
961 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
963 if (flags & LEX_STUFF_UTF8) {
966 STRLEN highhalf = 0; /* Count of variants */
967 const char *p, *e = pv+len;
968 for (p = pv; p != e; p++) {
969 if (! UTF8_IS_INVARIANT(*p)) {
975 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
976 bufptr = PL_parser->bufptr;
977 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
978 SvCUR_set(PL_parser->linestr,
979 SvCUR(PL_parser->linestr) + len+highhalf);
980 PL_parser->bufend += len+highhalf;
981 for (p = pv; p != e; p++) {
983 if (! UTF8_IS_INVARIANT(c)) {
984 *bufptr++ = UTF8_TWO_BYTE_HI(c);
985 *bufptr++ = UTF8_TWO_BYTE_LO(c);
992 if (flags & LEX_STUFF_UTF8) {
994 const char *p, *e = pv+len;
995 for (p = pv; p != e; p++) {
997 if (UTF8_IS_ABOVE_LATIN1(c)) {
998 Perl_croak(aTHX_ "Lexing code attempted to stuff "
999 "non-Latin-1 character into Latin-1 input");
1000 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
1003 } else if (! UTF8_IS_INVARIANT(c)) {
1004 /* malformed UTF-8 */
1006 SAVESPTR(PL_warnhook);
1007 PL_warnhook = PERL_WARNHOOK_FATAL;
1008 utf8n_to_uvchr((U8*)p, e-p, NULL, 0);
1014 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1015 bufptr = PL_parser->bufptr;
1016 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1017 SvCUR_set(PL_parser->linestr,
1018 SvCUR(PL_parser->linestr) + len-highhalf);
1019 PL_parser->bufend += len-highhalf;
1022 if (UTF8_IS_INVARIANT(*p)) {
1028 *bufptr++ = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
1034 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1035 bufptr = PL_parser->bufptr;
1036 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1037 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1038 PL_parser->bufend += len;
1039 Copy(pv, bufptr, len, char);
1045 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1047 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1048 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1049 reallocating the buffer if necessary. This means that lexing code that
1050 runs later will see the characters as if they had appeared in the input.
1051 It is not recommended to do this as part of normal parsing, and most
1052 uses of this facility run the risk of the inserted characters being
1053 interpreted in an unintended manner.
1055 The string to be inserted is represented by octets starting at I<pv>
1056 and continuing to the first nul. These octets are interpreted as either
1057 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1058 in I<flags>. The characters are recoded for the lexer buffer, according
1059 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1060 If it is not convenient to nul-terminate a string to be inserted, the
1061 L</lex_stuff_pvn> function is more appropriate.
1067 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1069 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1070 lex_stuff_pvn(pv, strlen(pv), flags);
1074 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1076 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1077 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1078 reallocating the buffer if necessary. This means that lexing code that
1079 runs later will see the characters as if they had appeared in the input.
1080 It is not recommended to do this as part of normal parsing, and most
1081 uses of this facility run the risk of the inserted characters being
1082 interpreted in an unintended manner.
1084 The string to be inserted is the string value of I<sv>. The characters
1085 are recoded for the lexer buffer, according to how the buffer is currently
1086 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1087 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1088 need to construct a scalar.
1094 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1098 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1100 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1102 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1106 =for apidoc Amx|void|lex_unstuff|char *ptr
1108 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1109 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1110 This hides the discarded text from any lexing code that runs later,
1111 as if the text had never appeared.
1113 This is not the normal way to consume lexed text. For that, use
1120 Perl_lex_unstuff(pTHX_ char *ptr)
1124 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1125 buf = PL_parser->bufptr;
1127 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1130 bufend = PL_parser->bufend;
1132 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1133 unstuff_len = ptr - buf;
1134 Move(ptr, buf, bufend+1-ptr, char);
1135 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1136 PL_parser->bufend = bufend - unstuff_len;
1140 =for apidoc Amx|void|lex_read_to|char *ptr
1142 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1143 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1144 performing the correct bookkeeping whenever a newline character is passed.
1145 This is the normal way to consume lexed text.
1147 Interpretation of the buffer's octets can be abstracted out by
1148 using the slightly higher-level functions L</lex_peek_unichar> and
1149 L</lex_read_unichar>.
1155 Perl_lex_read_to(pTHX_ char *ptr)
1158 PERL_ARGS_ASSERT_LEX_READ_TO;
1159 s = PL_parser->bufptr;
1160 if (ptr < s || ptr > PL_parser->bufend)
1161 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1162 for (; s != ptr; s++)
1164 COPLINE_INC_WITH_HERELINES;
1165 PL_parser->linestart = s+1;
1167 PL_parser->bufptr = ptr;
1171 =for apidoc Amx|void|lex_discard_to|char *ptr
1173 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1174 up to I<ptr>. The remaining content of the buffer will be moved, and
1175 all pointers into the buffer updated appropriately. I<ptr> must not
1176 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1177 it is not permitted to discard text that has yet to be lexed.
1179 Normally it is not necessarily to do this directly, because it suffices to
1180 use the implicit discarding behaviour of L</lex_next_chunk> and things
1181 based on it. However, if a token stretches across multiple lines,
1182 and the lexing code has kept multiple lines of text in the buffer for
1183 that purpose, then after completion of the token it would be wise to
1184 explicitly discard the now-unneeded earlier lines, to avoid future
1185 multi-line tokens growing the buffer without bound.
1191 Perl_lex_discard_to(pTHX_ char *ptr)
1195 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1196 buf = SvPVX(PL_parser->linestr);
1198 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1201 if (ptr > PL_parser->bufptr)
1202 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1203 discard_len = ptr - buf;
1204 if (PL_parser->oldbufptr < ptr)
1205 PL_parser->oldbufptr = ptr;
1206 if (PL_parser->oldoldbufptr < ptr)
1207 PL_parser->oldoldbufptr = ptr;
1208 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1209 PL_parser->last_uni = NULL;
1210 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1211 PL_parser->last_lop = NULL;
1212 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1213 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1214 PL_parser->bufend -= discard_len;
1215 PL_parser->bufptr -= discard_len;
1216 PL_parser->oldbufptr -= discard_len;
1217 PL_parser->oldoldbufptr -= discard_len;
1218 if (PL_parser->last_uni)
1219 PL_parser->last_uni -= discard_len;
1220 if (PL_parser->last_lop)
1221 PL_parser->last_lop -= discard_len;
1225 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1227 Reads in the next chunk of text to be lexed, appending it to
1228 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1229 looked to the end of the current chunk and wants to know more. It is
1230 usual, but not necessary, for lexing to have consumed the entirety of
1231 the current chunk at this time.
1233 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1234 chunk (i.e., the current chunk has been entirely consumed), normally the
1235 current chunk will be discarded at the same time that the new chunk is
1236 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1237 will not be discarded. If the current chunk has not been entirely
1238 consumed, then it will not be discarded regardless of the flag.
1240 Returns true if some new text was added to the buffer, or false if the
1241 buffer has reached the end of the input text.
1246 #define LEX_FAKE_EOF 0x80000000
1247 #define LEX_NO_TERM 0x40000000
1250 Perl_lex_next_chunk(pTHX_ U32 flags)
1254 STRLEN old_bufend_pos, new_bufend_pos;
1255 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1256 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1257 bool got_some_for_debugger = 0;
1259 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1260 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1261 linestr = PL_parser->linestr;
1262 buf = SvPVX(linestr);
1263 if (!(flags & LEX_KEEP_PREVIOUS) &&
1264 PL_parser->bufptr == PL_parser->bufend) {
1265 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1267 if (PL_parser->last_uni != PL_parser->bufend)
1268 PL_parser->last_uni = NULL;
1269 if (PL_parser->last_lop != PL_parser->bufend)
1270 PL_parser->last_lop = NULL;
1271 last_uni_pos = last_lop_pos = 0;
1275 old_bufend_pos = PL_parser->bufend - buf;
1276 bufptr_pos = PL_parser->bufptr - buf;
1277 oldbufptr_pos = PL_parser->oldbufptr - buf;
1278 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1279 linestart_pos = PL_parser->linestart - buf;
1280 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1281 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1283 if (flags & LEX_FAKE_EOF) {
1285 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1287 } else if (filter_gets(linestr, old_bufend_pos)) {
1289 got_some_for_debugger = 1;
1290 } else if (flags & LEX_NO_TERM) {
1293 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1294 sv_setpvs(linestr, "");
1296 /* End of real input. Close filehandle (unless it was STDIN),
1297 * then add implicit termination.
1299 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1300 PerlIO_clearerr(PL_parser->rsfp);
1301 else if (PL_parser->rsfp)
1302 (void)PerlIO_close(PL_parser->rsfp);
1303 PL_parser->rsfp = NULL;
1304 PL_parser->in_pod = PL_parser->filtered = 0;
1305 if (!PL_in_eval && PL_minus_p) {
1307 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1308 PL_minus_n = PL_minus_p = 0;
1309 } else if (!PL_in_eval && PL_minus_n) {
1310 sv_catpvs(linestr, /*{*/";}");
1313 sv_catpvs(linestr, ";");
1316 buf = SvPVX(linestr);
1317 new_bufend_pos = SvCUR(linestr);
1318 PL_parser->bufend = buf + new_bufend_pos;
1319 PL_parser->bufptr = buf + bufptr_pos;
1320 PL_parser->oldbufptr = buf + oldbufptr_pos;
1321 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1322 PL_parser->linestart = buf + linestart_pos;
1323 if (PL_parser->last_uni)
1324 PL_parser->last_uni = buf + last_uni_pos;
1325 if (PL_parser->last_lop)
1326 PL_parser->last_lop = buf + last_lop_pos;
1327 if (PL_parser->preambling != NOLINE) {
1328 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1329 PL_parser->preambling = NOLINE;
1331 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1332 PL_curstash != PL_debstash) {
1333 /* debugger active and we're not compiling the debugger code,
1334 * so store the line into the debugger's array of lines
1336 update_debugger_info(NULL, buf+old_bufend_pos,
1337 new_bufend_pos-old_bufend_pos);
1343 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1345 Looks ahead one (Unicode) character in the text currently being lexed.
1346 Returns the codepoint (unsigned integer value) of the next character,
1347 or -1 if lexing has reached the end of the input text. To consume the
1348 peeked character, use L</lex_read_unichar>.
1350 If the next character is in (or extends into) the next chunk of input
1351 text, the next chunk will be read in. Normally the current chunk will be
1352 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1353 then the current chunk will not be discarded.
1355 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1356 is encountered, an exception is generated.
1362 Perl_lex_peek_unichar(pTHX_ U32 flags)
1366 if (flags & ~(LEX_KEEP_PREVIOUS))
1367 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1368 s = PL_parser->bufptr;
1369 bufend = PL_parser->bufend;
1375 if (!lex_next_chunk(flags))
1377 s = PL_parser->bufptr;
1378 bufend = PL_parser->bufend;
1381 if (UTF8_IS_INVARIANT(head))
1383 if (UTF8_IS_START(head)) {
1384 len = UTF8SKIP(&head);
1385 while ((STRLEN)(bufend-s) < len) {
1386 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1388 s = PL_parser->bufptr;
1389 bufend = PL_parser->bufend;
1392 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1393 if (retlen == (STRLEN)-1) {
1394 /* malformed UTF-8 */
1396 SAVESPTR(PL_warnhook);
1397 PL_warnhook = PERL_WARNHOOK_FATAL;
1398 utf8n_to_uvchr((U8*)s, bufend-s, NULL, 0);
1404 if (!lex_next_chunk(flags))
1406 s = PL_parser->bufptr;
1413 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1415 Reads the next (Unicode) character in the text currently being lexed.
1416 Returns the codepoint (unsigned integer value) of the character read,
1417 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1418 if lexing has reached the end of the input text. To non-destructively
1419 examine the next character, use L</lex_peek_unichar> instead.
1421 If the next character is in (or extends into) the next chunk of input
1422 text, the next chunk will be read in. Normally the current chunk will be
1423 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1424 then the current chunk will not be discarded.
1426 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1427 is encountered, an exception is generated.
1433 Perl_lex_read_unichar(pTHX_ U32 flags)
1436 if (flags & ~(LEX_KEEP_PREVIOUS))
1437 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1438 c = lex_peek_unichar(flags);
1441 COPLINE_INC_WITH_HERELINES;
1443 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1445 ++(PL_parser->bufptr);
1451 =for apidoc Amx|void|lex_read_space|U32 flags
1453 Reads optional spaces, in Perl style, in the text currently being
1454 lexed. The spaces may include ordinary whitespace characters and
1455 Perl-style comments. C<#line> directives are processed if encountered.
1456 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1457 at a non-space character (or the end of the input text).
1459 If spaces extend into the next chunk of input text, the next chunk will
1460 be read in. Normally the current chunk will be discarded at the same
1461 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1462 chunk will not be discarded.
1467 #define LEX_NO_INCLINE 0x40000000
1468 #define LEX_NO_NEXT_CHUNK 0x80000000
1471 Perl_lex_read_space(pTHX_ U32 flags)
1474 const bool can_incline = !(flags & LEX_NO_INCLINE);
1475 bool need_incline = 0;
1476 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
1477 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1478 s = PL_parser->bufptr;
1479 bufend = PL_parser->bufend;
1485 } while (!(c == '\n' || (c == 0 && s == bufend)));
1486 } else if (c == '\n') {
1489 PL_parser->linestart = s;
1495 } else if (isSPACE(c)) {
1497 } else if (c == 0 && s == bufend) {
1500 if (flags & LEX_NO_NEXT_CHUNK)
1502 PL_parser->bufptr = s;
1503 l = CopLINE(PL_curcop);
1504 CopLINE(PL_curcop) += PL_parser->herelines + 1;
1505 got_more = lex_next_chunk(flags);
1506 CopLINE_set(PL_curcop, l);
1507 s = PL_parser->bufptr;
1508 bufend = PL_parser->bufend;
1511 if (can_incline && need_incline && PL_parser->rsfp) {
1519 PL_parser->bufptr = s;
1524 =for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1526 This function performs syntax checking on a prototype, C<proto>.
1527 If C<warn> is true, any illegal characters or mismatched brackets
1528 will trigger illegalproto warnings, declaring that they were
1529 detected in the prototype for C<name>.
1531 The return value is C<true> if this is a valid prototype, and
1532 C<false> if it is not, regardless of whether C<warn> was C<true> or
1535 Note that C<NULL> is a valid C<proto> and will always return C<true>.
1542 Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1544 STRLEN len, origlen;
1545 char *p = proto ? SvPV(proto, len) : NULL;
1546 bool bad_proto = FALSE;
1547 bool in_brackets = FALSE;
1548 bool after_slash = FALSE;
1549 char greedy_proto = ' ';
1550 bool proto_after_greedy_proto = FALSE;
1551 bool must_be_last = FALSE;
1552 bool underscore = FALSE;
1553 bool bad_proto_after_underscore = FALSE;
1555 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1561 for (; len--; p++) {
1564 proto_after_greedy_proto = TRUE;
1566 if (!strchr(";@%", *p))
1567 bad_proto_after_underscore = TRUE;
1570 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1577 in_brackets = FALSE;
1578 else if ((*p == '@' || *p == '%') &&
1581 must_be_last = TRUE;
1590 after_slash = FALSE;
1595 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
1598 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1599 origlen, UNI_DISPLAY_ISPRINT)
1600 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1602 if (proto_after_greedy_proto)
1603 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1604 "Prototype after '%c' for %"SVf" : %s",
1605 greedy_proto, SVfARG(name), p);
1607 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1608 "Missing ']' in prototype for %"SVf" : %s",
1611 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1612 "Illegal character in prototype for %"SVf" : %s",
1614 if (bad_proto_after_underscore)
1615 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1616 "Illegal character after '_' in prototype for %"SVf" : %s",
1620 return (! (proto_after_greedy_proto || bad_proto) );
1625 * This subroutine has nothing to do with tilting, whether at windmills
1626 * or pinball tables. Its name is short for "increment line". It
1627 * increments the current line number in CopLINE(PL_curcop) and checks
1628 * to see whether the line starts with a comment of the form
1629 * # line 500 "foo.pm"
1630 * If so, it sets the current line number and file to the values in the comment.
1634 S_incline(pTHX_ const char *s)
1641 PERL_ARGS_ASSERT_INCLINE;
1643 COPLINE_INC_WITH_HERELINES;
1644 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1645 && s+1 == PL_bufend && *s == ';') {
1646 /* fake newline in string eval */
1647 CopLINE_dec(PL_curcop);
1652 while (SPACE_OR_TAB(*s))
1654 if (strnEQ(s, "line", 4))
1658 if (SPACE_OR_TAB(*s))
1662 while (SPACE_OR_TAB(*s))
1670 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1672 while (SPACE_OR_TAB(*s))
1674 if (*s == '"' && (t = strchr(s+1, '"'))) {
1680 while (!isSPACE(*t))
1684 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1686 if (*e != '\n' && *e != '\0')
1687 return; /* false alarm */
1689 line_num = atoi(n)-1;
1692 const STRLEN len = t - s;
1694 if (!PL_rsfp && !PL_parser->filtered) {
1695 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1696 * to *{"::_<newfilename"} */
1697 /* However, the long form of evals is only turned on by the
1698 debugger - usually they're "(eval %lu)" */
1699 GV * const cfgv = CopFILEGV(PL_curcop);
1702 STRLEN tmplen2 = len;
1706 if (tmplen2 + 2 <= sizeof smallbuf)
1709 Newx(tmpbuf2, tmplen2 + 2, char);
1714 memcpy(tmpbuf2 + 2, s, tmplen2);
1717 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1719 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1720 /* adjust ${"::_<newfilename"} to store the new file name */
1721 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1722 /* The line number may differ. If that is the case,
1723 alias the saved lines that are in the array.
1724 Otherwise alias the whole array. */
1725 if (CopLINE(PL_curcop) == line_num) {
1726 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1727 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
1729 else if (GvAV(cfgv)) {
1730 AV * const av = GvAV(cfgv);
1731 const I32 start = CopLINE(PL_curcop)+1;
1732 I32 items = AvFILLp(av) - start;
1734 AV * const av2 = GvAVn(gv2);
1735 SV **svp = AvARRAY(av) + start;
1736 I32 l = (I32)line_num+1;
1738 av_store(av2, l++, SvREFCNT_inc(*svp++));
1743 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1746 CopFILE_free(PL_curcop);
1747 CopFILE_setn(PL_curcop, s, len);
1749 CopLINE_set(PL_curcop, line_num);
1752 #define skipspace(s) skipspace_flags(s, 0)
1756 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1758 AV *av = CopFILEAVx(PL_curcop);
1761 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1763 sv = *av_fetch(av, 0, 1);
1764 SvUPGRADE(sv, SVt_PVMG);
1766 if (!SvPOK(sv)) sv_setpvs(sv,"");
1768 sv_catsv(sv, orig_sv);
1770 sv_catpvn(sv, buf, len);
1775 if (PL_parser->preambling == NOLINE)
1776 av_store(av, CopLINE(PL_curcop), sv);
1782 * Called to gobble the appropriate amount and type of whitespace.
1783 * Skips comments as well.
1787 S_skipspace_flags(pTHX_ char *s, U32 flags)
1789 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
1790 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1791 while (s < PL_bufend && SPACE_OR_TAB(*s))
1794 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1796 lex_read_space(flags | LEX_KEEP_PREVIOUS |
1797 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1798 LEX_NO_NEXT_CHUNK : 0));
1800 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1801 if (PL_linestart > PL_bufptr)
1802 PL_bufptr = PL_linestart;
1810 * Check the unary operators to ensure there's no ambiguity in how they're
1811 * used. An ambiguous piece of code would be:
1813 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1814 * the +5 is its argument.
1823 if (PL_oldoldbufptr != PL_last_uni)
1825 while (isSPACE(*PL_last_uni))
1828 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1830 if ((t = strchr(s, '(')) && t < PL_bufptr)
1833 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1834 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1835 (int)(s - PL_last_uni), PL_last_uni);
1839 * LOP : macro to build a list operator. Its behaviour has been replaced
1840 * with a subroutine, S_lop() for which LOP is just another name.
1843 #define LOP(f,x) return lop(f,x,s)
1847 * Build a list operator (or something that might be one). The rules:
1848 * - if we have a next token, then it's a list operator [why?]
1849 * - if the next thing is an opening paren, then it's a function
1850 * - else it's a list operator
1854 S_lop(pTHX_ I32 f, int x, char *s)
1856 PERL_ARGS_ASSERT_LOP;
1862 PL_last_lop = PL_oldbufptr;
1863 PL_last_lop_op = (OPCODE)f;
1867 return REPORT(FUNC);
1870 return REPORT(FUNC);
1873 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1874 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1875 return REPORT(LSTOP);
1881 * When the lexer realizes it knows the next token (for instance,
1882 * it is reordering tokens for the parser) then it can call S_force_next
1883 * to know what token to return the next time the lexer is called. Caller
1884 * will need to set PL_nextval[] and possibly PL_expect to ensure
1885 * the lexer handles the token correctly.
1889 S_force_next(pTHX_ I32 type)
1893 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1894 tokereport(type, &NEXTVAL_NEXTTOKE);
1897 PL_nexttype[PL_nexttoke] = type;
1899 if (PL_lex_state != LEX_KNOWNEXT) {
1900 PL_lex_defer = PL_lex_state;
1901 PL_lex_expect = PL_expect;
1902 PL_lex_state = LEX_KNOWNEXT;
1909 * This subroutine handles postfix deref syntax after the arrow has already
1910 * been emitted. @* $* etc. are emitted as two separate token right here.
1911 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1912 * only the first, leaving yylex to find the next.
1916 S_postderef(pTHX_ int const funny, char const next)
1918 assert(funny == DOLSHARP || strchr("$@%&*", funny));
1919 assert(strchr("*[{", next));
1921 PL_expect = XOPERATOR;
1922 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
1923 assert('@' == funny || '$' == funny || DOLSHARP == funny);
1924 PL_lex_state = LEX_INTERPEND;
1925 force_next(POSTJOIN);
1931 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1932 && !PL_lex_brackets)
1934 PL_expect = XOPERATOR;
1943 int yyc = PL_parser->yychar;
1944 if (yyc != YYEMPTY) {
1946 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1947 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1948 PL_lex_allbrackets--;
1950 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1951 } else if (yyc == '('/*)*/) {
1952 PL_lex_allbrackets--;
1957 PL_parser->yychar = YYEMPTY;
1962 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1964 SV * const sv = newSVpvn_utf8(start, len,
1967 && !is_ascii_string((const U8*)start, len)
1968 && is_utf8_string((const U8*)start, len));
1974 * When the lexer knows the next thing is a word (for instance, it has
1975 * just seen -> and it knows that the next char is a word char, then
1976 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1980 * char *start : buffer position (must be within PL_linestr)
1981 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1982 * int check_keyword : if true, Perl checks to make sure the word isn't
1983 * a keyword (do this if the word is a label, e.g. goto FOO)
1984 * int allow_pack : if true, : characters will also be allowed (require,
1985 * use, etc. do this)
1986 * int allow_initial_tick : used by the "sub" lexer only.
1990 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
1995 PERL_ARGS_ASSERT_FORCE_WORD;
1997 start = SKIPSPACE1(start);
1999 if (isIDFIRST_lazy_if(s,UTF) ||
2000 (allow_pack && *s == ':') )
2002 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2003 if (check_keyword) {
2004 char *s2 = PL_tokenbuf;
2005 if (allow_pack && len > 6 && strnEQ(s2, "CORE::", 6))
2007 if (keyword(s2, len, 0))
2010 if (token == METHOD) {
2015 PL_expect = XOPERATOR;
2018 NEXTVAL_NEXTTOKE.opval
2019 = (OP*)newSVOP(OP_CONST,0,
2020 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2021 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2029 * Called when the lexer wants $foo *foo &foo etc, but the program
2030 * text only contains the "foo" portion. The first argument is a pointer
2031 * to the "foo", and the second argument is the type symbol to prefix.
2032 * Forces the next token to be a "WORD".
2033 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2037 S_force_ident(pTHX_ const char *s, int kind)
2039 PERL_ARGS_ASSERT_FORCE_IDENT;
2042 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2043 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2044 UTF ? SVf_UTF8 : 0));
2045 NEXTVAL_NEXTTOKE.opval = o;
2048 o->op_private = OPpCONST_ENTERED;
2049 /* XXX see note in pp_entereval() for why we forgo typo
2050 warnings if the symbol must be introduced in an eval.
2052 gv_fetchpvn_flags(s, len,
2053 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2054 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2055 kind == '$' ? SVt_PV :
2056 kind == '@' ? SVt_PVAV :
2057 kind == '%' ? SVt_PVHV :
2065 S_force_ident_maybe_lex(pTHX_ char pit)
2067 NEXTVAL_NEXTTOKE.ival = pit;
2072 Perl_str_to_version(pTHX_ SV *sv)
2077 const char *start = SvPV_const(sv,len);
2078 const char * const end = start + len;
2079 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2081 PERL_ARGS_ASSERT_STR_TO_VERSION;
2083 while (start < end) {
2087 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2092 retval += ((NV)n)/nshift;
2101 * Forces the next token to be a version number.
2102 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2103 * and if "guessing" is TRUE, then no new token is created (and the caller
2104 * must use an alternative parsing method).
2108 S_force_version(pTHX_ char *s, int guessing)
2113 PERL_ARGS_ASSERT_FORCE_VERSION;
2121 while (isDIGIT(*d) || *d == '_' || *d == '.')
2123 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2125 s = scan_num(s, &pl_yylval);
2126 version = pl_yylval.opval;
2127 ver = cSVOPx(version)->op_sv;
2128 if (SvPOK(ver) && !SvNIOK(ver)) {
2129 SvUPGRADE(ver, SVt_PVNV);
2130 SvNV_set(ver, str_to_version(ver));
2131 SvNOK_on(ver); /* hint that it is a version */
2134 else if (guessing) {
2139 /* NOTE: The parser sees the package name and the VERSION swapped */
2140 NEXTVAL_NEXTTOKE.opval = version;
2147 * S_force_strict_version
2148 * Forces the next token to be a version number using strict syntax rules.
2152 S_force_strict_version(pTHX_ char *s)
2155 const char *errstr = NULL;
2157 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2159 while (isSPACE(*s)) /* leading whitespace */
2162 if (is_STRICT_VERSION(s,&errstr)) {
2164 s = (char *)scan_version(s, ver, 0);
2165 version = newSVOP(OP_CONST, 0, ver);
2167 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2168 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2172 yyerror(errstr); /* version required */
2176 /* NOTE: The parser sees the package name and the VERSION swapped */
2177 NEXTVAL_NEXTTOKE.opval = version;
2185 * Tokenize a quoted string passed in as an SV. It finds the next
2186 * chunk, up to end of string or a backslash. It may make a new
2187 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2192 S_tokeq(pTHX_ SV *sv)
2199 PERL_ARGS_ASSERT_TOKEQ;
2203 assert (!SvIsCOW(sv));
2204 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
2208 /* This is relying on the SV being "well formed" with a trailing '\0' */
2209 while (s < send && !(*s == '\\' && s[1] == '\\'))
2214 if ( PL_hints & HINT_NEW_STRING ) {
2215 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2216 SVs_TEMP | SvUTF8(sv));
2220 if (s + 1 < send && (s[1] == '\\'))
2221 s++; /* all that, just for this */
2226 SvCUR_set(sv, d - SvPVX_const(sv));
2228 if ( PL_hints & HINT_NEW_STRING )
2229 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2234 * Now come three functions related to double-quote context,
2235 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2236 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2237 * interact with PL_lex_state, and create fake ( ... ) argument lists
2238 * to handle functions and concatenation.
2242 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2247 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2249 * Pattern matching will set PL_lex_op to the pattern-matching op to
2250 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2252 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2254 * Everything else becomes a FUNC.
2256 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2257 * had an OP_CONST or OP_READLINE). This just sets us up for a
2258 * call to S_sublex_push().
2262 S_sublex_start(pTHX)
2264 const I32 op_type = pl_yylval.ival;
2266 if (op_type == OP_NULL) {
2267 pl_yylval.opval = PL_lex_op;
2271 if (op_type == OP_CONST) {
2272 SV *sv = tokeq(PL_lex_stuff);
2274 if (SvTYPE(sv) == SVt_PVIV) {
2275 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2277 const char * const p = SvPV_const(sv, len);
2278 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2282 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2283 PL_lex_stuff = NULL;
2287 PL_sublex_info.super_state = PL_lex_state;
2288 PL_sublex_info.sub_inwhat = (U16)op_type;
2289 PL_sublex_info.sub_op = PL_lex_op;
2290 PL_lex_state = LEX_INTERPPUSH;
2294 pl_yylval.opval = PL_lex_op;
2304 * Create a new scope to save the lexing state. The scope will be
2305 * ended in S_sublex_done. Returns a '(', starting the function arguments
2306 * to the uc, lc, etc. found before.
2307 * Sets PL_lex_state to LEX_INTERPCONCAT.
2314 const bool is_heredoc = PL_multi_close == '<';
2317 PL_lex_state = PL_sublex_info.super_state;
2318 SAVEI8(PL_lex_dojoin);
2319 SAVEI32(PL_lex_brackets);
2320 SAVEI32(PL_lex_allbrackets);
2321 SAVEI32(PL_lex_formbrack);
2322 SAVEI8(PL_lex_fakeeof);
2323 SAVEI32(PL_lex_casemods);
2324 SAVEI32(PL_lex_starts);
2325 SAVEI8(PL_lex_state);
2326 SAVESPTR(PL_lex_repl);
2327 SAVEVPTR(PL_lex_inpat);
2328 SAVEI16(PL_lex_inwhat);
2331 SAVECOPLINE(PL_curcop);
2332 SAVEI32(PL_multi_end);
2333 SAVEI32(PL_parser->herelines);
2334 PL_parser->herelines = 0;
2336 SAVEI8(PL_multi_close);
2337 SAVEPPTR(PL_bufptr);
2338 SAVEPPTR(PL_bufend);
2339 SAVEPPTR(PL_oldbufptr);
2340 SAVEPPTR(PL_oldoldbufptr);
2341 SAVEPPTR(PL_last_lop);
2342 SAVEPPTR(PL_last_uni);
2343 SAVEPPTR(PL_linestart);
2344 SAVESPTR(PL_linestr);
2345 SAVEGENERICPV(PL_lex_brackstack);
2346 SAVEGENERICPV(PL_lex_casestack);
2347 SAVEGENERICPV(PL_parser->lex_shared);
2348 SAVEBOOL(PL_parser->lex_re_reparsing);
2349 SAVEI32(PL_copline);
2351 /* The here-doc parser needs to be able to peek into outer lexing
2352 scopes to find the body of the here-doc. So we put PL_linestr and
2353 PL_bufptr into lex_shared, to ‘share’ those values.
2355 PL_parser->lex_shared->ls_linestr = PL_linestr;
2356 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2358 PL_linestr = PL_lex_stuff;
2359 PL_lex_repl = PL_sublex_info.repl;
2360 PL_lex_stuff = NULL;
2361 PL_sublex_info.repl = NULL;
2363 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2364 = SvPVX(PL_linestr);
2365 PL_bufend += SvCUR(PL_linestr);
2366 PL_last_lop = PL_last_uni = NULL;
2367 SAVEFREESV(PL_linestr);
2368 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2370 PL_lex_dojoin = FALSE;
2371 PL_lex_brackets = PL_lex_formbrack = 0;
2372 PL_lex_allbrackets = 0;
2373 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2374 Newx(PL_lex_brackstack, 120, char);
2375 Newx(PL_lex_casestack, 12, char);
2376 PL_lex_casemods = 0;
2377 *PL_lex_casestack = '\0';
2379 PL_lex_state = LEX_INTERPCONCAT;
2381 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2382 PL_copline = NOLINE;
2384 Newxz(shared, 1, LEXSHARED);
2385 shared->ls_prev = PL_parser->lex_shared;
2386 PL_parser->lex_shared = shared;
2388 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2389 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2390 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2391 PL_lex_inpat = PL_sublex_info.sub_op;
2393 PL_lex_inpat = NULL;
2395 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2396 PL_in_eval &= ~EVAL_RE_REPARSING;
2403 * Restores lexer state after a S_sublex_push.
2409 if (!PL_lex_starts++) {
2410 SV * const sv = newSVpvs("");
2411 if (SvUTF8(PL_linestr))
2413 PL_expect = XOPERATOR;
2414 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2418 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2419 PL_lex_state = LEX_INTERPCASEMOD;
2423 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2424 assert(PL_lex_inwhat != OP_TRANSR);
2426 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
2427 PL_linestr = PL_lex_repl;
2429 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2430 PL_bufend += SvCUR(PL_linestr);
2431 PL_last_lop = PL_last_uni = NULL;
2432 PL_lex_dojoin = FALSE;
2433 PL_lex_brackets = 0;
2434 PL_lex_allbrackets = 0;
2435 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2436 PL_lex_casemods = 0;
2437 *PL_lex_casestack = '\0';
2439 if (SvEVALED(PL_lex_repl)) {
2440 PL_lex_state = LEX_INTERPNORMAL;
2442 /* we don't clear PL_lex_repl here, so that we can check later
2443 whether this is an evalled subst; that means we rely on the
2444 logic to ensure sublex_done() is called again only via the
2445 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2448 PL_lex_state = LEX_INTERPCONCAT;
2451 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2452 CopLINE(PL_curcop) +=
2453 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xpad_cop_seq.xlow
2454 + PL_parser->herelines;
2455 PL_parser->herelines = 0;
2460 const line_t l = CopLINE(PL_curcop);
2462 if (PL_multi_close == '<')
2463 PL_parser->herelines += l - PL_multi_end;
2464 PL_bufend = SvPVX(PL_linestr);
2465 PL_bufend += SvCUR(PL_linestr);
2466 PL_expect = XOPERATOR;
2467 PL_sublex_info.sub_inwhat = 0;
2472 PERL_STATIC_INLINE SV*
2473 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2475 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2476 * interior, hence to the "}". Finds what the name resolves to, returning
2477 * an SV* containing it; NULL if no valid one found */
2479 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2486 const U8* first_bad_char_loc;
2487 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2489 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2491 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2493 &first_bad_char_loc))
2495 /* If warnings are on, this will print a more detailed analysis of what
2496 * is wrong than the error message below */
2497 utf8n_to_uvchr(first_bad_char_loc,
2498 e - ((char *) first_bad_char_loc),
2501 /* We deliberately don't try to print the malformed character, which
2502 * might not print very well; it also may be just the first of many
2503 * malformations, so don't print what comes after it */
2504 yyerror(Perl_form(aTHX_
2505 "Malformed UTF-8 character immediately after '%.*s'",
2506 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr));
2510 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2511 /* include the <}> */
2512 e - backslash_ptr + 1);
2514 SvREFCNT_dec_NN(res);
2518 /* See if the charnames handler is the Perl core's, and if so, we can skip
2519 * the validation needed for a user-supplied one, as Perl's does its own
2521 table = GvHV(PL_hintgv); /* ^H */
2522 cvp = hv_fetchs(table, "charnames", FALSE);
2523 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2524 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
2526 const char * const name = HvNAME(stash);
2527 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2528 && strEQ(name, "_charnames")) {
2533 /* Here, it isn't Perl's charname handler. We can't rely on a
2534 * user-supplied handler to validate the input name. For non-ut8 input,
2535 * look to see that the first character is legal. Then loop through the
2536 * rest checking that each is a continuation */
2538 /* This code makes the reasonable assumption that the only Latin1-range
2539 * characters that begin a character name alias are alphabetic, otherwise
2540 * would have to create a isCHARNAME_BEGIN macro */
2543 if (! isALPHAU(*s)) {
2548 if (! isCHARNAME_CONT(*s)) {
2551 if (*s == ' ' && *(s-1) == ' ') {
2554 if ((U8) *s == NBSP_NATIVE && ckWARN_d(WARN_DEPRECATED)) {
2555 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2556 "NO-BREAK SPACE in a charnames "
2557 "alias definition is deprecated");
2563 /* Similarly for utf8. For invariants can check directly; for other
2564 * Latin1, can calculate their code point and check; otherwise use a
2566 if (UTF8_IS_INVARIANT(*s)) {
2567 if (! isALPHAU(*s)) {
2571 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2572 if (! isALPHAU(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1)))) {
2578 if (! PL_utf8_charname_begin) {
2579 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2580 PL_utf8_charname_begin = _core_swash_init("utf8",
2581 "_Perl_Charname_Begin",
2583 1, 0, NULL, &flags);
2585 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2592 if (UTF8_IS_INVARIANT(*s)) {
2593 if (! isCHARNAME_CONT(*s)) {
2596 if (*s == ' ' && *(s-1) == ' ') {
2601 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2602 if (! isCHARNAME_CONT(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1))))
2606 if (*s == *NBSP_UTF8
2607 && *(s+1) == *(NBSP_UTF8+1)
2608 && ckWARN_d(WARN_DEPRECATED))
2610 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2611 "NO-BREAK SPACE in a charnames "
2612 "alias definition is deprecated");
2617 if (! PL_utf8_charname_continue) {
2618 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2619 PL_utf8_charname_continue = _core_swash_init("utf8",
2620 "_Perl_Charname_Continue",
2622 1, 0, NULL, &flags);
2624 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2631 if (*(s-1) == ' ') {
2634 "charnames alias definitions may not contain trailing "
2635 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2636 (int)(s - backslash_ptr + 1), backslash_ptr,
2637 (int)(e - s + 1), s + 1
2639 UTF ? SVf_UTF8 : 0);
2643 if (SvUTF8(res)) { /* Don't accept malformed input */
2644 const U8* first_bad_char_loc;
2646 const char* const str = SvPV_const(res, len);
2647 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2648 /* If warnings are on, this will print a more detailed analysis of
2649 * what is wrong than the error message below */
2650 utf8n_to_uvchr(first_bad_char_loc,
2651 (char *) first_bad_char_loc - str,
2654 /* We deliberately don't try to print the malformed character,
2655 * which might not print very well; it also may be just the first
2656 * of many malformations, so don't print what comes after it */
2659 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2660 (int) (e - backslash_ptr + 1), backslash_ptr,
2661 (int) ((char *) first_bad_char_loc - str), str
2672 /* The final %.*s makes sure that should the trailing NUL be missing
2673 * that this print won't run off the end of the string */
2676 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2677 (int)(s - backslash_ptr + 1), backslash_ptr,
2678 (int)(e - s + 1), s + 1
2680 UTF ? SVf_UTF8 : 0);
2687 "charnames alias definitions may not contain a sequence of "
2688 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2689 (int)(s - backslash_ptr + 1), backslash_ptr,
2690 (int)(e - s + 1), s + 1
2692 UTF ? SVf_UTF8 : 0);
2699 Extracts the next constant part of a pattern, double-quoted string,
2700 or transliteration. This is terrifying code.
2702 For example, in parsing the double-quoted string "ab\x63$d", it would
2703 stop at the '$' and return an OP_CONST containing 'abc'.
2705 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2706 processing a pattern (PL_lex_inpat is true), a transliteration
2707 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2709 Returns a pointer to the character scanned up to. If this is
2710 advanced from the start pointer supplied (i.e. if anything was
2711 successfully parsed), will leave an OP_CONST for the substring scanned
2712 in pl_yylval. Caller must intuit reason for not parsing further
2713 by looking at the next characters herself.
2717 \N{FOO} => \N{U+hex_for_character_FOO}
2718 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2721 all other \-char, including \N and \N{ apart from \N{ABC}
2724 @ and $ where it appears to be a var, but not for $ as tail anchor
2729 In transliterations:
2730 characters are VERY literal, except for - not at the start or end
2731 of the string, which indicates a range. If the range is in bytes,
2732 scan_const expands the range to the full set of intermediate
2733 characters. If the range is in utf8, the hyphen is replaced with
2734 a certain range mark which will be handled by pmtrans() in op.c.
2736 In double-quoted strings:
2738 double-quoted style: \r and \n
2739 constants: \x31, etc.
2740 deprecated backrefs: \1 (in substitution replacements)
2741 case and quoting: \U \Q \E
2744 scan_const does *not* construct ops to handle interpolated strings.
2745 It stops processing as soon as it finds an embedded $ or @ variable
2746 and leaves it to the caller to work out what's going on.
2748 embedded arrays (whether in pattern or not) could be:
2749 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2751 $ in double-quoted strings must be the symbol of an embedded scalar.
2753 $ in pattern could be $foo or could be tail anchor. Assumption:
2754 it's a tail anchor if $ is the last thing in the string, or if it's
2755 followed by one of "()| \r\n\t"
2757 \1 (backreferences) are turned into $1 in substitutions
2759 The structure of the code is
2760 while (there's a character to process) {
2761 handle transliteration ranges
2762 skip regexp comments /(?#comment)/ and codes /(?{code})/
2763 skip #-initiated comments in //x patterns
2764 check for embedded arrays
2765 check for embedded scalars
2767 deprecate \1 in substitution replacements
2768 handle string-changing backslashes \l \U \Q \E, etc.
2769 switch (what was escaped) {
2770 handle \- in a transliteration (becomes a literal -)
2771 if a pattern and not \N{, go treat as regular character
2772 handle \132 (octal characters)
2773 handle \x15 and \x{1234} (hex characters)
2774 handle \N{name} (named characters, also \N{3,5} in a pattern)
2775 handle \cV (control characters)
2776 handle printf-style backslashes (\f, \r, \n, etc)
2779 } (end if backslash)
2780 handle regular character
2781 } (end while character to read)
2786 S_scan_const(pTHX_ char *start)
2788 char *send = PL_bufend; /* end of the constant */
2789 SV *sv = newSV(send - start); /* sv for the constant. See note below
2791 char *s = start; /* start of the constant */
2792 char *d = SvPVX(sv); /* destination for copies */
2793 bool dorange = FALSE; /* are we in a translit range? */
2794 bool didrange = FALSE; /* did we just finish a range? */
2795 bool in_charclass = FALSE; /* within /[...]/ */
2796 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2797 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2798 UTF8? But, this can show as true
2799 when the source isn't utf8, as for
2800 example when it is entirely composed
2802 SV *res; /* result from charnames */
2804 /* Note on sizing: The scanned constant is placed into sv, which is
2805 * initialized by newSV() assuming one byte of output for every byte of
2806 * input. This routine expects newSV() to allocate an extra byte for a
2807 * trailing NUL, which this routine will append if it gets to the end of
2808 * the input. There may be more bytes of input than output (eg., \N{LATIN
2809 * CAPITAL LETTER A}), or more output than input if the constant ends up
2810 * recoded to utf8, but each time a construct is found that might increase
2811 * the needed size, SvGROW() is called. Its size parameter each time is
2812 * based on the best guess estimate at the time, namely the length used so
2813 * far, plus the length the current construct will occupy, plus room for
2814 * the trailing NUL, plus one byte for every input byte still unscanned */
2816 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2819 UV literal_endpoint = 0;
2820 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2823 PERL_ARGS_ASSERT_SCAN_CONST;
2825 assert(PL_lex_inwhat != OP_TRANSR);
2826 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2827 /* If we are doing a trans and we know we want UTF8 set expectation */
2828 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2829 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2832 /* Protect sv from errors and fatal warnings. */
2833 ENTER_with_name("scan_const");
2836 while (s < send || dorange) {
2838 /* get transliterations out of the way (they're most literal) */
2839 if (PL_lex_inwhat == OP_TRANS) {
2840 /* expand a range A-Z to the full set of characters. AIE! */
2842 I32 i; /* current expanded character */
2843 I32 min; /* first character in range */
2844 I32 max; /* last character in range */
2855 char * const c = (char*)utf8_hop((U8*)d, -1);
2859 *c = (char) ILLEGAL_UTF8_BYTE;
2860 /* mark the range as done, and continue */
2866 i = d - SvPVX_const(sv); /* remember current offset */
2869 SvLEN(sv) + ((has_utf8)
2870 ? (512 - UTF_CONTINUATION_MARK
2873 /* How many two-byte within 0..255: 128 in UTF-8,
2874 * 96 in UTF-8-mod. */
2876 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2878 d = SvPVX(sv) + i; /* refresh d after realloc */
2882 for (j = 0; j <= 1; j++) {
2883 char * const c = (char*)utf8_hop((U8*)d, -1);
2884 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2890 max = (U8)0xff; /* only to \xff */
2891 uvmax = uv; /* \x{100} to uvmax */
2893 d = c; /* eat endpoint chars */
2898 d -= 2; /* eat the first char and the - */
2899 min = (U8)*d; /* first char in range */
2900 max = (U8)d[1]; /* last char in range */
2907 "Invalid range \"%c-%c\" in transliteration operator",
2908 (char)min, (char)max);
2912 /* Because of the discontinuities in EBCDIC A-Z and a-z, expand
2913 * any subsets of these ranges into individual characters */
2914 if (literal_endpoint == 2 &&
2915 ((isLOWER_A(min) && isLOWER_A(max)) ||
2916 (isUPPER_A(min) && isUPPER_A(max))))
2918 for (i = min; i <= max; i++) {
2925 for (i = min; i <= max; i++)
2928 append_utf8_from_native_byte(i, &d);
2936 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2938 *d++ = (char) ILLEGAL_UTF8_BYTE;
2940 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2944 /* mark the range as done, and continue */
2948 literal_endpoint = 0;
2953 /* range begins (ignore - as first or last char) */
2954 else if (*s == '-' && s+1 < send && s != start) {
2956 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2963 *d++ = (char) ILLEGAL_UTF8_BYTE; /* use illegal utf8 byte--see pmtrans */
2973 literal_endpoint = 0;
2974 native_range = TRUE;
2979 /* if we get here, we're not doing a transliteration */
2981 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
2984 while (s1 >= start && *s1-- == '\\')
2987 in_charclass = TRUE;
2990 else if (*s == ']' && PL_lex_inpat && in_charclass) {
2993 while (s1 >= start && *s1-- == '\\')
2996 in_charclass = FALSE;
2999 /* skip for regexp comments /(?#comment)/, except for the last
3000 * char, which will be done separately.
3001 * Stop on (?{..}) and friends */
3003 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
3005 while (s+1 < send && *s != ')')
3008 else if (!PL_lex_casemods &&
3009 ( s[2] == '{' /* This should match regcomp.c */
3010 || (s[2] == '?' && s[3] == '{')))
3016 /* likewise skip #-initiated comments in //x patterns */
3017 else if (*s == '#' && PL_lex_inpat && !in_charclass &&
3018 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3019 while (s+1 < send && *s != '\n')
3023 /* no further processing of single-quoted regex */
3024 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3025 goto default_action;
3027 /* check for embedded arrays
3028 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3030 else if (*s == '@' && s[1]) {
3031 if (isWORDCHAR_lazy_if(s+1,UTF))
3033 if (strchr(":'{$", s[1]))
3035 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3036 break; /* in regexp, neither @+ nor @- are interpolated */
3039 /* check for embedded scalars. only stop if we're sure it's a
3042 else if (*s == '$') {
3043 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3045 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3047 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3048 "Possible unintended interpolation of $\\ in regex");
3050 break; /* in regexp, $ might be tail anchor */
3054 /* End of else if chain - OP_TRANS rejoin rest */
3057 if (*s == '\\' && s+1 < send) {
3058 char* e; /* Can be used for ending '}', etc. */
3062 /* warn on \1 - \9 in substitution replacements, but note that \11
3063 * is an octal; and \19 is \1 followed by '9' */
3064 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3065 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3067 /* diag_listed_as: \%d better written as $%d */
3068 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3073 /* string-change backslash escapes */
3074 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3078 /* In a pattern, process \N, but skip any other backslash escapes.
3079 * This is because we don't want to translate an escape sequence
3080 * into a meta symbol and have the regex compiler use the meta
3081 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3082 * in spite of this, we do have to process \N here while the proper
3083 * charnames handler is in scope. See bugs #56444 and #62056.
3084 * There is a complication because \N in a pattern may also stand
3085 * for 'match a non-nl', and not mean a charname, in which case its
3086 * processing should be deferred to the regex compiler. To be a
3087 * charname it must be followed immediately by a '{', and not look
3088 * like \N followed by a curly quantifier, i.e., not something like
3089 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3091 else if (PL_lex_inpat
3094 || regcurly(s + 1)))
3097 goto default_action;
3102 /* quoted - in transliterations */
3104 if (PL_lex_inwhat == OP_TRANS) {
3111 if ((isALPHANUMERIC(*s)))
3112 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3113 "Unrecognized escape \\%c passed through",
3115 /* default action is to copy the quoted character */
3116 goto default_action;
3119 /* eg. \132 indicates the octal constant 0132 */
3120 case '0': case '1': case '2': case '3':
3121 case '4': case '5': case '6': case '7':
3123 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3125 uv = grok_oct(s, &len, &flags, NULL);
3127 if (len < 3 && s < send && isDIGIT(*s)
3128 && ckWARN(WARN_MISC))
3130 Perl_warner(aTHX_ packWARN(WARN_MISC),
3131 "%s", form_short_octal_warning(s, len));
3134 goto NUM_ESCAPE_INSERT;
3136 /* eg. \o{24} indicates the octal constant \024 */
3141 bool valid = grok_bslash_o(&s, &uv, &error,
3142 TRUE, /* Output warning */
3143 FALSE, /* Not strict */
3144 TRUE, /* Output warnings for
3151 goto NUM_ESCAPE_INSERT;
3154 /* eg. \x24 indicates the hex constant 0x24 */
3159 bool valid = grok_bslash_x(&s, &uv, &error,
3160 TRUE, /* Output warning */
3161 FALSE, /* Not strict */
3162 TRUE, /* Output warnings for
3172 /* Insert oct or hex escaped character. There will always be
3173 * enough room in sv since such escapes will be longer than any
3174 * UTF-8 sequence they can end up as, except if they force us
3175 * to recode the rest of the string into utf8 */
3177 /* Here uv is the ordinal of the next character being added */
3178 if (!UVCHR_IS_INVARIANT(uv)) {
3179 if (!has_utf8 && uv > 255) {
3180 /* Might need to recode whatever we have accumulated so
3181 * far if it contains any chars variant in utf8 or
3184 SvCUR_set(sv, d - SvPVX_const(sv));
3187 /* See Note on sizing above. */
3188 sv_utf8_upgrade_flags_grow(sv,
3189 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3190 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3191 d = SvPVX(sv) + SvCUR(sv);
3196 d = (char*)uvchr_to_utf8((U8*)d, uv);
3197 if (PL_lex_inwhat == OP_TRANS &&
3198 PL_sublex_info.sub_op) {
3199 PL_sublex_info.sub_op->op_private |=
3200 (PL_lex_repl ? OPpTRANS_FROM_UTF
3204 if (uv > 255 && !dorange)
3205 native_range = FALSE;
3218 /* In a non-pattern \N must be a named character, like \N{LATIN
3219 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3220 * mean to match a non-newline. For non-patterns, named
3221 * characters are converted to their string equivalents. In
3222 * patterns, named characters are not converted to their
3223 * ultimate forms for the same reasons that other escapes
3224 * aren't. Instead, they are converted to the \N{U+...} form
3225 * to get the value from the charnames that is in effect right
3226 * now, while preserving the fact that it was a named character
3227 * so that the regex compiler knows this */
3229 /* The structure of this section of code (besides checking for
3230 * errors and upgrading to utf8) is:
3231 * Further disambiguate between the two meanings of \N, and if
3232 * not a charname, go process it elsewhere
3233 * If of form \N{U+...}, pass it through if a pattern;
3234 * otherwise convert to utf8
3235 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3236 * pattern; otherwise convert to utf8 */
3238 /* Here, s points to the 'N'; the test below is guaranteed to
3239 * succeed if we are being called on a pattern as we already
3240 * know from a test above that the next character is a '{'.
3241 * On a non-pattern \N must mean 'named sequence, which
3242 * requires braces */
3245 yyerror("Missing braces on \\N{}");
3250 /* If there is no matching '}', it is an error. */
3251 if (! (e = strchr(s, '}'))) {
3252 if (! PL_lex_inpat) {
3253 yyerror("Missing right brace on \\N{}");
3255 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
3260 /* Here it looks like a named character */
3262 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3263 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3264 | PERL_SCAN_DISALLOW_PREFIX;
3267 /* For \N{U+...}, the '...' is a unicode value even on
3268 * EBCDIC machines */
3269 s += 2; /* Skip to next char after the 'U+' */
3271 uv = grok_hex(s, &len, &flags, NULL);
3272 if (len == 0 || len != (STRLEN)(e - s)) {
3273 yyerror("Invalid hexadecimal number in \\N{U+...}");
3280 /* On non-EBCDIC platforms, pass through to the regex
3281 * compiler unchanged. The reason we evaluated the
3282 * number above is to make sure there wasn't a syntax
3283 * error. But on EBCDIC we convert to native so
3284 * downstream code can continue to assume it's native
3286 s -= 5; /* Include the '\N{U+' */
3288 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3291 (unsigned int) UNI_TO_NATIVE(uv));
3293 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3297 else { /* Not a pattern: convert the hex to string */
3299 /* If destination is not in utf8, unconditionally
3300 * recode it to be so. This is because \N{} implies
3301 * Unicode semantics, and scalars have to be in utf8
3302 * to guarantee those semantics */
3304 SvCUR_set(sv, d - SvPVX_const(sv));
3307 /* See Note on sizing above. */
3308 sv_utf8_upgrade_flags_grow(
3310 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3311 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3312 d = SvPVX(sv) + SvCUR(sv);
3316 /* Add the (Unicode) code point to the output. */
3317 if (UNI_IS_INVARIANT(uv)) {
3318 *d++ = (char) LATIN1_TO_NATIVE(uv);
3321 d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
3325 else /* Here is \N{NAME} but not \N{U+...}. */
3326 if ((res = get_and_check_backslash_N_name(s, e)))
3329 const char *str = SvPV_const(res, len);
3332 if (! len) { /* The name resolved to an empty string */
3333 Copy("\\N{}", d, 4, char);
3337 /* In order to not lose information for the regex
3338 * compiler, pass the result in the specially made
3339 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3340 * the code points in hex of each character
3341 * returned by charnames */
3343 const char *str_end = str + len;
3344 const STRLEN off = d - SvPVX_const(sv);
3346 if (! SvUTF8(res)) {
3347 /* For the non-UTF-8 case, we can determine the
3348 * exact length needed without having to parse
3349 * through the string. Each character takes up
3350 * 2 hex digits plus either a trailing dot or
3352 d = off + SvGROW(sv, off
3354 + 6 /* For the "\N{U+", and
3356 + (STRLEN)(send - e));
3357 Copy("\\N{U+", d, 5, char);
3359 while (str < str_end) {
3362 my_snprintf(hex_string,
3364 "%02X.", (U8) *str);
3365 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(hex_string));
3366 Copy(hex_string, d, 3, char);
3370 d--; /* We will overwrite below the final
3371 dot with a right brace */
3374 STRLEN char_length; /* cur char's byte length */
3376 /* and the number of bytes after this is
3377 * translated into hex digits */
3378 STRLEN output_length;
3380 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3381 * for max('U+', '.'); and 1 for NUL */
3382 char hex_string[2 * UTF8_MAXBYTES + 5];
3384 /* Get the first character of the result. */
3385 U32 uv = utf8n_to_uvchr((U8 *) str,
3389 /* Convert first code point to hex, including
3390 * the boiler plate before it. */
3392 my_snprintf(hex_string, sizeof(hex_string),
3396 /* Make sure there is enough space to hold it */
3397 d = off + SvGROW(sv, off
3399 + (STRLEN)(send - e)
3400 + 2); /* '}' + NUL */
3402 Copy(hex_string, d, output_length, char);
3405 /* For each subsequent character, append dot and
3406 * its ordinal in hex */
3407 while ((str += char_length) < str_end) {
3408 const STRLEN off = d - SvPVX_const(sv);
3409 U32 uv = utf8n_to_uvchr((U8 *) str,
3414 my_snprintf(hex_string,
3419 d = off + SvGROW(sv, off
3421 + (STRLEN)(send - e)
3422 + 2); /* '}' + NUL */
3423 Copy(hex_string, d, output_length, char);
3428 *d++ = '}'; /* Done. Add the trailing brace */
3431 else { /* Here, not in a pattern. Convert the name to a
3434 /* If destination is not in utf8, unconditionally
3435 * recode it to be so. This is because \N{} implies
3436 * Unicode semantics, and scalars have to be in utf8
3437 * to guarantee those semantics */
3439 SvCUR_set(sv, d - SvPVX_const(sv));
3442 /* See Note on sizing above. */
3443 sv_utf8_upgrade_flags_grow(sv,
3444 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3445 len + (STRLEN)(send - s) + 1);
3446 d = SvPVX(sv) + SvCUR(sv);
3448 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3450 /* See Note on sizing above. (NOTE: SvCUR() is not
3451 * set correctly here). */
3452 const STRLEN off = d - SvPVX_const(sv);
3453 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3455 if (! SvUTF8(res)) { /* Make sure is \N{} return is UTF-8 */
3456 sv_utf8_upgrade(res);
3457 str = SvPV_const(res, len);
3459 Copy(str, d, len, char);
3465 } /* End \N{NAME} */
3468 native_range = FALSE; /* \N{} is defined to be Unicode */
3470 s = e + 1; /* Point to just after the '}' */
3473 /* \c is a control character */
3477 *d++ = grok_bslash_c(*s++, 1);
3480 yyerror("Missing control char name in \\c");
3484 /* printf-style backslashes, formfeeds, newlines, etc */
3501 *d++ = ASCII_TO_NATIVE('\033');
3510 } /* end if (backslash) */
3517 /* If we started with encoded form, or already know we want it,
3518 then encode the next character */
3519 if (! NATIVE_BYTE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3523 /* One might think that it is wasted effort in the case of the
3524 * source being utf8 (this_utf8 == TRUE) to take the next character
3525 * in the source, convert it to an unsigned value, and then convert
3526 * it back again. But the source has not been validated here. The
3527 * routine that does the conversion checks for errors like
3530 const UV nextuv = (this_utf8)
3531 ? utf8n_to_uvchr((U8*)s, send - s, &len, 0)
3533 const STRLEN need = UNISKIP(nextuv);
3535 SvCUR_set(sv, d - SvPVX_const(sv));
3538 /* See Note on sizing above. */
3539 sv_utf8_upgrade_flags_grow(sv,
3540 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3541 need + (STRLEN)(send - s) + 1);
3542 d = SvPVX(sv) + SvCUR(sv);
3544 } else if (need > len) {
3545 /* encoded value larger than old, may need extra space (NOTE:
3546 * SvCUR() is not set correctly here). See Note on sizing
3548 const STRLEN off = d - SvPVX_const(sv);
3549 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3553 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3555 if (uv > 255 && !dorange)
3556 native_range = FALSE;
3562 } /* while loop to process each character */
3564 /* terminate the string and set up the sv */
3566 SvCUR_set(sv, d - SvPVX_const(sv));
3567 if (SvCUR(sv) >= SvLEN(sv))
3568 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3569 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3572 if (PL_encoding && !has_utf8) {
3573 sv_recode_to_utf8(sv, PL_encoding);
3579 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3580 PL_sublex_info.sub_op->op_private |=
3581 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3585 /* shrink the sv if we allocated more than we used */
3586 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3587 SvPV_shrink_to_cur(sv);
3590 /* return the substring (via pl_yylval) only if we parsed anything */
3593 for (; s2 < s; s2++) {
3595 COPLINE_INC_WITH_HERELINES;
3597 SvREFCNT_inc_simple_void_NN(sv);
3598 if ( (PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ))
3599 && ! PL_parser->lex_re_reparsing)
3601 const char *const key = PL_lex_inpat ? "qr" : "q";
3602 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3606 if (PL_lex_inwhat == OP_TRANS) {
3609 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3612 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3620 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3623 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3625 LEAVE_with_name("scan_const");
3630 * Returns TRUE if there's more to the expression (e.g., a subscript),
3633 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3635 * ->[ and ->{ return TRUE
3636 * ->$* ->$#* ->@* ->@[ ->@{ return TRUE if postderef_qq is enabled
3637 * { and [ outside a pattern are always subscripts, so return TRUE
3638 * if we're outside a pattern and it's not { or [, then return FALSE
3639 * if we're in a pattern and the first char is a {
3640 * {4,5} (any digits around the comma) returns FALSE
3641 * if we're in a pattern and the first char is a [
3643 * [SOMETHING] has a funky algorithm to decide whether it's a
3644 * character class or not. It has to deal with things like
3645 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3646 * anything else returns TRUE
3649 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3652 S_intuit_more(pTHX_ char *s)
3654 PERL_ARGS_ASSERT_INTUIT_MORE;
3656 if (PL_lex_brackets)
3658 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3660 if (*s == '-' && s[1] == '>'
3661 && FEATURE_POSTDEREF_QQ_IS_ENABLED
3662 && ( (s[2] == '$' && (s[3] == '*' || (s[3] == '#' && s[4] == '*')))
3663 ||(s[2] == '@' && strchr("*[{",s[3])) ))
3665 if (*s != '{' && *s != '[')
3670 /* In a pattern, so maybe we have {n,m}. */
3678 /* On the other hand, maybe we have a character class */
3681 if (*s == ']' || *s == '^')
3684 /* this is terrifying, and it works */
3687 const char * const send = strchr(s,']');
3688 unsigned char un_char, last_un_char;
3689 char tmpbuf[sizeof PL_tokenbuf * 4];
3691 if (!send) /* has to be an expression */
3693 weight = 2; /* let's weigh the evidence */
3697 else if (isDIGIT(*s)) {
3699 if (isDIGIT(s[1]) && s[2] == ']')
3705 Zero(seen,256,char);
3707 for (; s < send; s++) {
3708 last_un_char = un_char;
3709 un_char = (unsigned char)*s;
3714 weight -= seen[un_char] * 10;
3715 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3717 char *tmp = PL_bufend;
3718 PL_bufend = (char*)send;
3719 scan_ident(s, tmpbuf, sizeof tmpbuf, FALSE);
3721 len = (int)strlen(tmpbuf);
3722 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3723 UTF ? SVf_UTF8 : 0, SVt_PV))
3728 else if (*s == '$' && s[1] &&
3729 strchr("[#!%*<>()-=",s[1])) {
3730 if (/*{*/ strchr("])} =",s[2]))
3739 if (strchr("wds]",s[1]))
3741 else if (seen[(U8)'\''] || seen[(U8)'"'])
3743 else if (strchr("rnftbxcav",s[1]))
3745 else if (isDIGIT(s[1])) {
3747 while (s[1] && isDIGIT(s[1]))
3757 if (strchr("aA01! ",last_un_char))
3759 if (strchr("zZ79~",s[1]))
3761 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3762 weight -= 5; /* cope with negative subscript */
3765 if (!isWORDCHAR(last_un_char)
3766 && !(last_un_char == '$' || last_un_char == '@'
3767 || last_un_char == '&')
3768 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3773 if (keyword(tmpbuf, d - tmpbuf, 0))
3776 if (un_char == last_un_char + 1)
3778 weight -= seen[un_char];
3783 if (weight >= 0) /* probably a character class */
3793 * Does all the checking to disambiguate
3795 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3796 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3798 * First argument is the stuff after the first token, e.g. "bar".
3800 * Not a method if foo is a filehandle.
3801 * Not a method if foo is a subroutine prototyped to take a filehandle.
3802 * Not a method if it's really "Foo $bar"
3803 * Method if it's "foo $bar"
3804 * Not a method if it's really "print foo $bar"
3805 * Method if it's really "foo package::" (interpreted as package->foo)
3806 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3807 * Not a method if bar is a filehandle or package, but is quoted with
3812 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3814 char *s = start + (*start == '$');
3815 char tmpbuf[sizeof PL_tokenbuf];
3819 PERL_ARGS_ASSERT_INTUIT_METHOD;
3821 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3823 if (cv && SvPOK(cv)) {
3824 const char *proto = CvPROTO(cv);
3826 while (*proto && (isSPACE(*proto) || *proto == ';'))
3833 if (*start == '$') {
3834 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3835 isUPPER(*PL_tokenbuf))
3840 return *s == '(' ? FUNCMETH : METHOD;
3843 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3844 /* start is the beginning of the possible filehandle/object,
3845 * and s is the end of it
3846 * tmpbuf is a copy of it (but with single quotes as double colons)
3849 if (!keyword(tmpbuf, len, 0)) {
3850 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3855 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3856 if (indirgv && GvCVu(indirgv))
3858 /* filehandle or package name makes it a method */
3859 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3861 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3862 return 0; /* no assumptions -- "=>" quotes bareword */
3864 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3865 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3866 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3870 return *s == '(' ? FUNCMETH : METHOD;
3876 /* Encoded script support. filter_add() effectively inserts a
3877 * 'pre-processing' function into the current source input stream.
3878 * Note that the filter function only applies to the current source file
3879 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3881 * The datasv parameter (which may be NULL) can be used to pass
3882 * private data to this instance of the filter. The filter function
3883 * can recover the SV using the FILTER_DATA macro and use it to
3884 * store private buffers and state information.
3886 * The supplied datasv parameter is upgraded to a PVIO type
3887 * and the IoDIRP/IoANY field is used to store the function pointer,
3888 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3889 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3890 * private use must be set using malloc'd pointers.
3894 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3902 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3903 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3905 if (!PL_rsfp_filters)
3906 PL_rsfp_filters = newAV();
3909 SvUPGRADE(datasv, SVt_PVIO);
3910 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3911 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3912 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3913 FPTR2DPTR(void *, IoANY(datasv)),
3914 SvPV_nolen(datasv)));
3915 av_unshift(PL_rsfp_filters, 1);
3916 av_store(PL_rsfp_filters, 0, datasv) ;
3918 !PL_parser->filtered
3919 && PL_parser->lex_flags & LEX_EVALBYTES
3920 && PL_bufptr < PL_bufend
3922 const char *s = PL_bufptr;
3923 while (s < PL_bufend) {
3925 SV *linestr = PL_parser->linestr;
3926 char *buf = SvPVX(linestr);
3927 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3928 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3929 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3930 STRLEN const linestart_pos = PL_parser->linestart - buf;
3931 STRLEN const last_uni_pos =
3932 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3933 STRLEN const last_lop_pos =
3934 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3935 av_push(PL_rsfp_filters, linestr);
3936 PL_parser->linestr =
3937 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3938 buf = SvPVX(PL_parser->linestr);
3939 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3940 PL_parser->bufptr = buf + bufptr_pos;
3941 PL_parser->oldbufptr = buf + oldbufptr_pos;
3942 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3943 PL_parser->linestart = buf + linestart_pos;
3944 if (PL_parser->last_uni)
3945 PL_parser->last_uni = buf + last_uni_pos;
3946 if (PL_parser->last_lop)
3947 PL_parser->last_lop = buf + last_lop_pos;
3948 SvLEN(linestr) = SvCUR(linestr);
3949 SvCUR(linestr) = s-SvPVX(linestr);
3950 PL_parser->filtered = 1;
3960 /* Delete most recently added instance of this filter function. */
3962 Perl_filter_del(pTHX_ filter_t funcp)
3966 PERL_ARGS_ASSERT_FILTER_DEL;
3969 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3970 FPTR2DPTR(void*, funcp)));
3972 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3974 /* if filter is on top of stack (usual case) just pop it off */
3975 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3976 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3977 sv_free(av_pop(PL_rsfp_filters));
3981 /* we need to search for the correct entry and clear it */
3982 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3986 /* Invoke the idxth filter function for the current rsfp. */
3987 /* maxlen 0 = read one text line */
3989 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3993 /* This API is bad. It should have been using unsigned int for maxlen.
3994 Not sure if we want to change the API, but if not we should sanity
3995 check the value here. */
3996 unsigned int correct_length = maxlen < 0 ? PERL_INT_MAX : maxlen;
3998 PERL_ARGS_ASSERT_FILTER_READ;
4000 if (!PL_parser || !PL_rsfp_filters)
4002 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4003 /* Provide a default input filter to make life easy. */
4004 /* Note that we append to the line. This is handy. */
4005 DEBUG_P(PerlIO_printf(Perl_debug_log,
4006 "filter_read %d: from rsfp\n", idx));
4007 if (correct_length) {
4010 const int old_len = SvCUR(buf_sv);
4012 /* ensure buf_sv is large enough */
4013 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4014 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4015 correct_length)) <= 0) {
4016 if (PerlIO_error(PL_rsfp))
4017 return -1; /* error */
4019 return 0 ; /* end of file */
4021 SvCUR_set(buf_sv, old_len + len) ;
4022 SvPVX(buf_sv)[old_len + len] = '\0';
4025 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4026 if (PerlIO_error(PL_rsfp))
4027 return -1; /* error */
4029 return 0 ; /* end of file */
4032 return SvCUR(buf_sv);
4034 /* Skip this filter slot if filter has been deleted */
4035 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4036 DEBUG_P(PerlIO_printf(Perl_debug_log,
4037 "filter_read %d: skipped (filter deleted)\n",
4039 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4041 if (SvTYPE(datasv) != SVt_PVIO) {
4042 if (correct_length) {
4044 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4045 if (!remainder) return 0; /* eof */
4046 if (correct_length > remainder) correct_length = remainder;
4047 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4048 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4051 const char *s = SvEND(datasv);
4052 const char *send = SvPVX(datasv) + SvLEN(datasv);
4060 if (s == send) return 0; /* eof */
4061 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4062 SvCUR_set(datasv, s-SvPVX(datasv));
4064 return SvCUR(buf_sv);
4066 /* Get function pointer hidden within datasv */
4067 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4068 DEBUG_P(PerlIO_printf(Perl_debug_log,
4069 "filter_read %d: via function %p (%s)\n",
4070 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4071 /* Call function. The function is expected to */
4072 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4073 /* Return: <0:error, =0:eof, >0:not eof */
4074 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4078 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4080 PERL_ARGS_ASSERT_FILTER_GETS;
4082 #ifdef PERL_CR_FILTER
4083 if (!PL_rsfp_filters) {
4084 filter_add(S_cr_textfilter,NULL);
4087 if (PL_rsfp_filters) {
4089 SvCUR_set(sv, 0); /* start with empty line */
4090 if (FILTER_READ(0, sv, 0) > 0)
4091 return ( SvPVX(sv) ) ;
4096 return (sv_gets(sv, PL_rsfp, append));
4100 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4104 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4106 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4110 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4111 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4113 return GvHV(gv); /* Foo:: */
4116 /* use constant CLASS => 'MyClass' */
4117 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4118 if (gv && GvCV(gv)) {
4119 SV * const sv = cv_const_sv(GvCV(gv));
4121 pkgname = SvPV_const(sv, len);
4124 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4129 S_tokenize_use(pTHX_ int is_use, char *s) {
4130 PERL_ARGS_ASSERT_TOKENIZE_USE;
4132 if (PL_expect != XSTATE)
4133 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4134 is_use ? "use" : "no"));
4137 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4138 s = force_version(s, TRUE);
4139 if (*s == ';' || *s == '}'
4140 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4141 NEXTVAL_NEXTTOKE.opval = NULL;
4144 else if (*s == 'v') {
4145 s = force_word(s,WORD,FALSE,TRUE);
4146 s = force_version(s, FALSE);
4150 s = force_word(s,WORD,FALSE,TRUE);
4151 s = force_version(s, FALSE);
4153 pl_yylval.ival = is_use;
4157 static const char* const exp_name[] =
4158 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4159 "ATTRTERM", "TERMBLOCK", "POSTDEREF", "TERMORDORDOR"
4163 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4165 S_word_takes_any_delimeter(char *p, STRLEN len)
4167 return (len == 1 && strchr("msyq", p[0])) ||
4169 (p[0] == 't' && p[1] == 'r') ||
4170 (p[0] == 'q' && strchr("qwxr", p[1]))));
4174 S_check_scalar_slice(pTHX_ char *s)
4177 while (*s == ' ' || *s == '\t') s++;
4178 if (*s == 'q' && s[1] == 'w'
4179 && !isWORDCHAR_lazy_if(s+2,UTF))
4181 while (*s && (isWORDCHAR_lazy_if(s,UTF) || strchr(" \t$#+-'\"", *s)))
4182 s += UTF ? UTF8SKIP(s) : 1;
4183 if (*s == '}' || *s == ']')
4184 pl_yylval.ival = OPpSLICEWARNING;
4190 Works out what to call the token just pulled out of the input
4191 stream. The yacc parser takes care of taking the ops we return and
4192 stitching them into a tree.
4195 The type of the next token
4198 Switch based on the current state:
4199 - if we already built the token before, use it
4200 - if we have a case modifier in a string, deal with that
4201 - handle other cases of interpolation inside a string
4202 - scan the next line if we are inside a format
4203 In the normal state switch on the next character:
4205 if alphabetic, go to key lookup
4206 unrecoginized character - croak
4207 - 0/4/26: handle end-of-line or EOF
4208 - cases for whitespace
4209 - \n and #: handle comments and line numbers
4210 - various operators, brackets and sigils
4213 - 'v': vstrings (or go to key lookup)
4214 - 'x' repetition operator (or go to key lookup)
4215 - other ASCII alphanumerics (key lookup begins here):
4218 scan built-in keyword (but do nothing with it yet)
4219 check for statement label
4220 check for lexical subs
4221 goto just_a_word if there is one
4222 see whether built-in keyword is overridden
4223 switch on keyword number:
4224 - default: just_a_word:
4225 not a built-in keyword; handle bareword lookup
4226 disambiguate between method and sub call
4227 fall back to bareword
4228 - cases for built-in keywords
4236 char *s = PL_bufptr;
4240 const bool saw_infix_sigil = cBOOL(PL_parser->saw_infix_sigil);
4244 /* orig_keyword, gvp, and gv are initialized here because
4245 * jump to the label just_a_word_zero can bypass their
4246 * initialization later. */
4247 I32 orig_keyword = 0;
4252 SV* tmp = newSVpvs("");
4253 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4254 (IV)CopLINE(PL_curcop),
4255 lex_state_names[PL_lex_state],
4256 exp_name[PL_expect],
4257 pv_display(tmp, s, strlen(s), 0, 60));
4261 switch (PL_lex_state) {
4263 case LEX_INTERPNORMAL:
4266 /* when we've already built the next token, just pull it out of the queue */
4269 pl_yylval = PL_nextval[PL_nexttoke];
4271 PL_lex_state = PL_lex_defer;
4272 PL_expect = PL_lex_expect;
4273 PL_lex_defer = LEX_NORMAL;
4277 next_type = PL_nexttype[PL_nexttoke];
4278 if (next_type & (7<<24)) {
4279 if (next_type & (1<<24)) {
4280 if (PL_lex_brackets > 100)
4281 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4282 PL_lex_brackstack[PL_lex_brackets++] =
4283 (char) ((next_type >> 16) & 0xff);
4285 if (next_type & (2<<24))
4286 PL_lex_allbrackets++;
4287 if (next_type & (4<<24))
4288 PL_lex_allbrackets--;
4289 next_type &= 0xffff;
4291 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4294 /* interpolated case modifiers like \L \U, including \Q and \E.
4295 when we get here, PL_bufptr is at the \
4297 case LEX_INTERPCASEMOD:
4299 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4301 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4302 PL_bufptr, PL_bufend, *PL_bufptr);
4304 /* handle \E or end of string */
4305 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4307 if (PL_lex_casemods) {
4308 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4309 PL_lex_casestack[PL_lex_casemods] = '\0';
4311 if (PL_bufptr != PL_bufend
4312 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4313 || oldmod == 'F')) {
4315 PL_lex_state = LEX_INTERPCONCAT;
4317 PL_lex_allbrackets--;
4320 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4321 /* Got an unpaired \E */
4322 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4323 "Useless use of \\E");
4325 if (PL_bufptr != PL_bufend)
4327 PL_lex_state = LEX_INTERPCONCAT;
4331 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4332 "### Saw case modifier\n"); });
4334 if (s[1] == '\\' && s[2] == 'E') {
4336 PL_lex_state = LEX_INTERPCONCAT;
4341 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4342 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4343 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4344 (strchr(PL_lex_casestack, 'L')
4345 || strchr(PL_lex_casestack, 'U')
4346 || strchr(PL_lex_casestack, 'F'))) {
4347 PL_lex_casestack[--PL_lex_casemods] = '\0';
4348 PL_lex_allbrackets--;
4351 if (PL_lex_casemods > 10)
4352 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4353 PL_lex_casestack[PL_lex_casemods++] = *s;
4354 PL_lex_casestack[PL_lex_casemods] = '\0';
4355 PL_lex_state = LEX_INTERPCONCAT;
4356 NEXTVAL_NEXTTOKE.ival = 0;
4357 force_next((2<<24)|'(');
4359 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4361 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4363 NEXTVAL_NEXTTOKE.ival = OP_LC;
4365 NEXTVAL_NEXTTOKE.ival = OP_UC;
4367 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4369 NEXTVAL_NEXTTOKE.ival = OP_FC;
4371 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4375 if (PL_lex_starts) {
4378 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4379 if (PL_lex_casemods == 1 && PL_lex_inpat)
4388 case LEX_INTERPPUSH:
4389 return REPORT(sublex_push());
4391 case LEX_INTERPSTART:
4392 if (PL_bufptr == PL_bufend)
4393 return REPORT(sublex_done());
4394 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4395 "### Interpolated variable\n"); });
4397 /* for /@a/, we leave the joining for the regex engine to do
4398 * (unless we're within \Q etc) */
4399 PL_lex_dojoin = (*PL_bufptr == '@'
4400 && (!PL_lex_inpat || PL_lex_casemods));
4401 PL_lex_state = LEX_INTERPNORMAL;
4402 if (PL_lex_dojoin) {
4403 NEXTVAL_NEXTTOKE.ival = 0;
4405 force_ident("\"", '$');
4406 NEXTVAL_NEXTTOKE.ival = 0;
4408 NEXTVAL_NEXTTOKE.ival = 0;
4409 force_next((2<<24)|'(');
4410 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4413 /* Convert (?{...}) and friends to 'do {...}' */
4414 if (PL_lex_inpat && *PL_bufptr == '(') {
4415 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4417 if (*PL_bufptr != '{')
4419 PL_expect = XTERMBLOCK;
4423 if (PL_lex_starts++) {
4425 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4426 if (!PL_lex_casemods && PL_lex_inpat)
4433 case LEX_INTERPENDMAYBE:
4434 if (intuit_more(PL_bufptr)) {
4435 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4441 if (PL_lex_dojoin) {
4442 const U8 dojoin_was = PL_lex_dojoin;
4443 PL_lex_dojoin = FALSE;
4444 PL_lex_state = LEX_INTERPCONCAT;
4445 PL_lex_allbrackets--;
4446 return REPORT(dojoin_was == 1 ? ')' : POSTJOIN);
4448 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4449 && SvEVALED(PL_lex_repl))
4451 if (PL_bufptr != PL_bufend)
4452 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4455 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4456 re_eval_str. If the here-doc body’s length equals the previous
4457 value of re_eval_start, re_eval_start will now be null. So
4458 check re_eval_str as well. */
4459 if (PL_parser->lex_shared->re_eval_start
4460 || PL_parser->lex_shared->re_eval_str) {
4462 if (*PL_bufptr != ')')
4463 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4465 /* having compiled a (?{..}) expression, return the original
4466 * text too, as a const */
4467 if (PL_parser->lex_shared->re_eval_str) {
4468 sv = PL_parser->lex_shared->re_eval_str;
4469 PL_parser->lex_shared->re_eval_str = NULL;
4471 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4472 SvPV_shrink_to_cur(sv);
4474 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4475 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4476 NEXTVAL_NEXTTOKE.opval =
4477 (OP*)newSVOP(OP_CONST, 0,
4480 PL_parser->lex_shared->re_eval_start = NULL;
4486 case LEX_INTERPCONCAT:
4488 if (PL_lex_brackets)
4489 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4490 (long) PL_lex_brackets);
4492 if (PL_bufptr == PL_bufend)
4493 return REPORT(sublex_done());
4495 /* m'foo' still needs to be parsed for possible (?{...}) */
4496 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4497 SV *sv = newSVsv(PL_linestr);
4499 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4503 s = scan_const(PL_bufptr);
4505 PL_lex_state = LEX_INTERPCASEMOD;
4507 PL_lex_state = LEX_INTERPSTART;
4510 if (s != PL_bufptr) {
4511 NEXTVAL_NEXTTOKE = pl_yylval;
4514 if (PL_lex_starts++) {
4515 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4516 if (!PL_lex_casemods && PL_lex_inpat)
4529 s = scan_formline(PL_bufptr);
4530 if (!PL_lex_formbrack)
4539 /* We really do *not* want PL_linestr ever becoming a COW. */
4540 assert (!SvIsCOW(PL_linestr));
4542 PL_oldoldbufptr = PL_oldbufptr;
4544 PL_parser->saw_infix_sigil = 0;
4549 if (UTF ? isIDFIRST_utf8((U8*)s) : isALNUMC(*s))
4552 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4553 const char *c = UTF ? sv_uni_display(dsv, newSVpvn_flags(s,
4555 SVs_TEMP | SVf_UTF8),
4556 10, UNI_DISPLAY_ISPRINT)
4557 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4558 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4559 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4560 d = UTF ? (char *) utf8_hop((U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4564 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"UTF8f"<-- HERE near column %d", c,
4565 UTF8fARG(UTF, (s - d), d),
4570 goto fake_eof; /* emulate EOF on ^D or ^Z */
4572 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4575 if (PL_lex_brackets &&
4576 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4577 yyerror((const char *)
4579 ? "Format not terminated"
4580 : "Missing right curly or square bracket"));
4582 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4583 "### Tokener got EOF\n");
4587 if (s++ < PL_bufend)
4588 goto retry; /* ignore stray nulls */
4591 if (!PL_in_eval && !PL_preambled) {
4592 PL_preambled = TRUE;
4594 /* Generate a string of Perl code to load the debugger.
4595 * If PERL5DB is set, it will return the contents of that,
4596 * otherwise a compile-time require of perl5db.pl. */
4598 const char * const pdb = PerlEnv_getenv("PERL5DB");
4601 sv_setpv(PL_linestr, pdb);
4602 sv_catpvs(PL_linestr,";");
4604 SETERRNO(0,SS_NORMAL);
4605 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4607 PL_parser->preambling = CopLINE(PL_curcop);
4609 sv_setpvs(PL_linestr,"");
4610 if (PL_preambleav) {
4611 SV **svp = AvARRAY(PL_preambleav);
4612 SV **const end = svp + AvFILLp(PL_preambleav);
4614 sv_catsv(PL_linestr, *svp);
4616 sv_catpvs(PL_linestr, ";");
4618 sv_free(MUTABLE_SV(PL_preambleav));
4619 PL_preambleav = NULL;
4622 sv_catpvs(PL_linestr,
4623 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4624 if (PL_minus_n || PL_minus_p) {
4625 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4627 sv_catpvs(PL_linestr,"chomp;");
4630 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4631 || *PL_splitstr == '"')
4632 && strchr(PL_splitstr + 1, *PL_splitstr))
4633 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4635 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4636 bytes can be used as quoting characters. :-) */
4637 const char *splits = PL_splitstr;
4638 sv_catpvs(PL_linestr, "our @F=split(q\0");
4641 if (*splits == '\\')
4642 sv_catpvn(PL_linestr, splits, 1);
4643 sv_catpvn(PL_linestr, splits, 1);
4644 } while (*splits++);
4645 /* This loop will embed the trailing NUL of
4646 PL_linestr as the last thing it does before
4648 sv_catpvs(PL_linestr, ");");
4652 sv_catpvs(PL_linestr,"our @F=split(' ');");
4655 sv_catpvs(PL_linestr, "\n");
4656 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4657 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4658 PL_last_lop = PL_last_uni = NULL;
4659 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4660 update_debugger_info(PL_linestr, NULL, 0);
4665 bof = PL_rsfp ? TRUE : FALSE;
4668 fake_eof = LEX_FAKE_EOF;
4670 PL_bufptr = PL_bufend;
4671 COPLINE_INC_WITH_HERELINES;
4672 if (!lex_next_chunk(fake_eof)) {
4673 CopLINE_dec(PL_curcop);
4675 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4677 CopLINE_dec(PL_curcop);
4679 /* If it looks like the start of a BOM or raw UTF-16,
4680 * check if it in fact is. */
4681 if (bof && PL_rsfp &&
4683 *(U8*)s == BOM_UTF8_FIRST_BYTE ||
4686 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4687 bof = (offset == (Off_t)SvCUR(PL_linestr));
4688 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4689 /* offset may include swallowed CR */
4691 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4694 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4695 s = swallow_bom((U8*)s);
4698 if (PL_parser->in_pod) {
4699 /* Incest with pod. */
4700 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4701 sv_setpvs(PL_linestr, "");
4702 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4703 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4704 PL_last_lop = PL_last_uni = NULL;
4705 PL_parser->in_pod = 0;
4708 if (PL_rsfp || PL_parser->filtered)
4710 } while (PL_parser->in_pod);
4711 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4712 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4713 PL_last_lop = PL_last_uni = NULL;
4714 if (CopLINE(PL_curcop) == 1) {
4715 while (s < PL_bufend && isSPACE(*s))
4717 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4721 if (*s == '#' && *(s+1) == '!')
4723 #ifdef ALTERNATE_SHEBANG
4725 static char const as[] = ALTERNATE_SHEBANG;
4726 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4727 d = s + (sizeof(as) - 1);
4729 #endif /* ALTERNATE_SHEBANG */
4738 while (*d && !isSPACE(*d))
4742 #ifdef ARG_ZERO_IS_SCRIPT
4743 if (ipathend > ipath) {
4745 * HP-UX (at least) sets argv[0] to the script name,
4746 * which makes $^X incorrect. And Digital UNIX and Linux,
4747 * at least, set argv[0] to the basename of the Perl
4748 * interpreter. So, having found "#!", we'll set it right.
4750 SV* copfilesv = CopFILESV(PL_curcop);
4753 GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4755 assert(SvPOK(x) || SvGMAGICAL(x));
4756 if (sv_eq(x, copfilesv)) {
4757 sv_setpvn(x, ipath, ipathend - ipath);
4763 const char *bstart = SvPV_const(copfilesv, blen);
4764 const char * const lstart = SvPV_const(x, llen);
4766 bstart += blen - llen;
4767 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4768 sv_setpvn(x, ipath, ipathend - ipath);
4775 /* Anything to do if no copfilesv? */
4777 TAINT_NOT; /* $^X is always tainted, but that's OK */
4779 #endif /* ARG_ZERO_IS_SCRIPT */
4784 d = instr(s,"perl -");
4786 d = instr(s,"perl");
4788 /* avoid getting into infinite loops when shebang
4789 * line contains "Perl" rather than "perl" */
4791 for (d = ipathend-4; d >= ipath; --d) {
4792 if ((*d == 'p' || *d == 'P')
4793 && !ibcmp(d, "perl", 4))
4803 #ifdef ALTERNATE_SHEBANG
4805 * If the ALTERNATE_SHEBANG on this system starts with a
4806 * character that can be part of a Perl expression, then if
4807 * we see it but not "perl", we're probably looking at the
4808 * start of Perl code, not a request to hand off to some
4809 * other interpreter. Similarly, if "perl" is there, but
4810 * not in the first 'word' of the line, we assume the line
4811 * contains the start of the Perl program.
4813 if (d && *s != '#') {
4814 const char *c = ipath;
4815 while (*c && !strchr("; \t\r\n\f\v#", *c))
4818 d = NULL; /* "perl" not in first word; ignore */
4820 *s = '#'; /* Don't try to parse shebang line */
4822 #endif /* ALTERNATE_SHEBANG */
4827 !instr(s,"indir") &&
4828 instr(PL_origargv[0],"perl"))
4835 while (s < PL_bufend && isSPACE(*s))
4837 if (s < PL_bufend) {
4838 Newx(newargv,PL_origargc+3,char*);
4840 while (s < PL_bufend && !isSPACE(*s))
4843 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4846 newargv = PL_origargv;
4849 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4851 Perl_croak(aTHX_ "Can't exec %s", ipath);
4854 while (*d && !isSPACE(*d))
4856 while (SPACE_OR_TAB(*d))
4860 const bool switches_done = PL_doswitches;
4861 const U32 oldpdb = PL_perldb;
4862 const bool oldn = PL_minus_n;
4863 const bool oldp = PL_minus_p;
4867 bool baduni = FALSE;
4869 const char *d2 = d1 + 1;
4870 if (parse_unicode_opts((const char **)&d2)
4874 if (baduni || *d1 == 'M' || *d1 == 'm') {
4875 const char * const m = d1;
4876 while (*d1 && !isSPACE(*d1))
4878 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4881 d1 = moreswitches(d1);
4883 if (PL_doswitches && !switches_done) {
4884 int argc = PL_origargc;
4885 char **argv = PL_origargv;
4888 } while (argc && argv[0][0] == '-' && argv[0][1]);
4889 init_argv_symbols(argc,argv);
4891 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4892 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4893 /* if we have already added "LINE: while (<>) {",
4894 we must not do it again */
4896 sv_setpvs(PL_linestr, "");
4897 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4898 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4899 PL_last_lop = PL_last_uni = NULL;
4900 PL_preambled = FALSE;
4901 if (PERLDB_LINE || PERLDB_SAVESRC)
4902 (void)gv_fetchfile(PL_origfilename);
4909 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4910 PL_lex_state = LEX_FORMLINE;
4911 NEXTVAL_NEXTTOKE.ival = 0;
4912 force_next(FORMRBRACK);
4917 #ifdef PERL_STRICT_CR
4918 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4920 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4922 case ' ': case '\t': case '\f': case 013:
4927 if (PL_lex_state != LEX_NORMAL ||
4928 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
4929 const bool in_comment = *s == '#';
4930 if (*s == '#' && s == PL_linestart && PL_in_eval
4931 && !PL_rsfp && !PL_parser->filtered) {
4932 /* handle eval qq[#line 1 "foo"\n ...] */
4933 CopLINE_dec(PL_curcop);
4937 while (d < PL_bufend && *d != '\n')
4941 else if (d > PL_bufend)
4942 /* Found by Ilya: feed random input to Perl. */
4943 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
4946 if (in_comment && d == PL_bufend
4947 && PL_lex_state == LEX_INTERPNORMAL
4948 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
4949 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
4952 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4953 PL_lex_state = LEX_FORMLINE;
4954 NEXTVAL_NEXTTOKE.ival = 0;
4955 force_next(FORMRBRACK);
4960 while (s < PL_bufend && *s != '\n')
4968 else if (s > PL_bufend)
4969 /* Found by Ilya: feed random input to Perl. */
4970 Perl_croak(aTHX_ "panic: input overflow");
4974 if (s[1] && isALPHA(s[1]) && !isWORDCHAR(s[2])) {
4982 while (s < PL_bufend && SPACE_OR_TAB(*s))
4985 if (strnEQ(s,"=>",2)) {
4986 s = force_word(PL_bufptr,WORD,FALSE,FALSE);
4987 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4988 OPERATOR('-'); /* unary minus */
4991 case 'r': ftst = OP_FTEREAD; break;
4992 case 'w': ftst = OP_FTEWRITE; break;
4993 case 'x': ftst = OP_FTEEXEC; break;
4994 case 'o': ftst = OP_FTEOWNED; break;
4995 case 'R': ftst = OP_FTRREAD; break;
4996 case 'W': ftst = OP_FTRWRITE; break;
4997 case 'X': ftst = OP_FTREXEC; break;
4998 case 'O': ftst = OP_FTROWNED; break;
4999 case 'e': ftst = OP_FTIS; break;
5000 case 'z': ftst = OP_FTZERO; break;
5001 case 's': ftst = OP_FTSIZE; break;
5002 case 'f': ftst = OP_FTFILE; break;
5003 case 'd': ftst = OP_FTDIR; break;
5004 case 'l': ftst = OP_FTLINK; break;
5005 case 'p': ftst = OP_FTPIPE; break;
5006 case 'S': ftst = OP_FTSOCK; break;
5007 case 'u': ftst = OP_FTSUID; break;
5008 case 'g': ftst = OP_FTSGID; break;
5009 case 'k': ftst = OP_FTSVTX; break;
5010 case 'b': ftst = OP_FTBLK; break;
5011 case 'c': ftst = OP_FTCHR; break;
5012 case 't': ftst = OP_FTTTY; break;
5013 case 'T': ftst = OP_FTTEXT; break;
5014 case 'B': ftst = OP_FTBINARY; break;
5015 case 'M': case 'A': case 'C':
5016 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5018 case 'M': ftst = OP_FTMTIME; break;
5019 case 'A': ftst = OP_FTATIME; break;
5020 case 'C': ftst = OP_FTCTIME; break;
5028 PL_last_uni = PL_oldbufptr;
5029 PL_last_lop_op = (OPCODE)ftst;
5030 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5031 "### Saw file test %c\n", (int)tmp);
5036 /* Assume it was a minus followed by a one-letter named
5037 * subroutine call (or a -bareword), then. */
5038 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5039 "### '-%c' looked like a file test but was not\n",
5046 const char tmp = *s++;
5049 if (PL_expect == XOPERATOR)
5054 else if (*s == '>') {
5057 if (FEATURE_POSTDEREF_IS_ENABLED && (
5058 ((*s == '$' || *s == '&') && s[1] == '*')
5059 ||(*s == '$' && s[1] == '#' && s[2] == '*')
5060 ||((*s == '@' || *s == '%') && strchr("*[{", s[1]))
5061 ||(*s == '*' && (s[1] == '*' || s[1] == '{'))
5064 Perl_ck_warner_d(aTHX_
5065 packWARN(WARN_EXPERIMENTAL__POSTDEREF),
5066 "Postfix dereference is experimental"
5068 PL_expect = XPOSTDEREF;
5071 if (isIDFIRST_lazy_if(s,UTF)) {
5072 s = force_word(s,METHOD,FALSE,TRUE);
5080 if (PL_expect == XOPERATOR) {
5081 if (*s == '=' && !PL_lex_allbrackets &&
5082 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5089 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5091 OPERATOR('-'); /* unary minus */
5097 const char tmp = *s++;
5100 if (PL_expect == XOPERATOR)
5105 if (PL_expect == XOPERATOR) {
5106 if (*s == '=' && !PL_lex_allbrackets &&
5107 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5114 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5121 if (PL_expect == XPOSTDEREF) POSTDEREF('*');
5122 if (PL_expect != XOPERATOR) {
5123 s = scan_ident(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5124 PL_expect = XOPERATOR;
5125 force_ident(PL_tokenbuf, '*');
5133 if (*s == '=' && !PL_lex_allbrackets &&
5134 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5140 if (*s == '=' && !PL_lex_allbrackets &&
5141 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5145 PL_parser->saw_infix_sigil = 1;
5150 if (PL_expect == XOPERATOR) {
5151 if (s[1] == '=' && !PL_lex_allbrackets &&
5152 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5155 PL_parser->saw_infix_sigil = 1;
5158 else if (PL_expect == XPOSTDEREF) POSTDEREF('%');
5159 PL_tokenbuf[0] = '%';
5160 s = scan_ident(s, PL_tokenbuf + 1,
5161 sizeof PL_tokenbuf - 1, FALSE);
5163 if (!PL_tokenbuf[1]) {
5166 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5168 PL_tokenbuf[0] = '@';
5170 PL_expect = XOPERATOR;
5171 force_ident_maybe_lex('%');
5175 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5176 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5181 if (PL_lex_brackets > 100)
5182 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5183 PL_lex_brackstack[PL_lex_brackets++] = 0;
5184 PL_lex_allbrackets++;
5186 const char tmp = *s++;
5191 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5193 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5196 Perl_ck_warner_d(aTHX_
5197 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
5198 "Smartmatch is experimental");
5204 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5211 goto just_a_word_zero_gv;
5217 switch (PL_expect) {
5219 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5221 PL_bufptr = s; /* update in case we back off */
5224 "Use of := for an empty attribute list is not allowed");
5231 PL_expect = XTERMBLOCK;
5235 while (isIDFIRST_lazy_if(s,UTF)) {
5238 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5239 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5240 if (tmp < 0) tmp = -tmp;
5255 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5257 d = scan_str(d,TRUE,TRUE,FALSE,NULL);
5258 COPLINE_SET_FROM_MULTI_END;
5260 /* MUST advance bufptr here to avoid bogus
5261 "at end of line" context messages from yyerror().
5263 PL_bufptr = s + len;
5264 yyerror("Unterminated attribute parameter in attribute list");
5268 return REPORT(0); /* EOF indicator */
5272 sv_catsv(sv, PL_lex_stuff);
5273 attrs = op_append_elem(OP_LIST, attrs,
5274 newSVOP(OP_CONST, 0, sv));
5275 SvREFCNT_dec(PL_lex_stuff);
5276 PL_lex_stuff = NULL;
5279 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5281 if (PL_in_my == KEY_our) {
5282 deprecate(":unique");
5285 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5288 /* NOTE: any CV attrs applied here need to be part of
5289 the CVf_BUILTIN_ATTRS define in cv.h! */
5290 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5292 CvLVALUE_on(PL_compcv);
5294 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5296 deprecate(":locked");
5298 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5300 CvMETHOD_on(PL_compcv);
5302 /* After we've set the flags, it could be argued that
5303 we don't need to do the attributes.pm-based setting
5304 process, and shouldn't bother appending recognized
5305 flags. To experiment with that, uncomment the
5306 following "else". (Note that's already been
5307 uncommented. That keeps the above-applied built-in
5308 attributes from being intercepted (and possibly
5309 rejected) by a package's attribute routines, but is
5310 justified by the performance win for the common case
5311 of applying only built-in attributes.) */
5313 attrs = op_append_elem(OP_LIST, attrs,
5314 newSVOP(OP_CONST, 0,
5318 if (*s == ':' && s[1] != ':')
5321 break; /* require real whitespace or :'s */
5322 /* XXX losing whitespace on sequential attributes here */
5325 if (*s != ';' && *s != '}' &&
5326 !(PL_expect == XOPERATOR
5327 ? (*s == '=' || *s == ')')
5328 : (*s == '{' || *s == '('))) {
5329 const char q = ((*s == '\'') ? '"' : '\'');
5330 /* If here for an expression, and parsed no attrs, back
5332 if (PL_expect == XOPERATOR && !attrs) {
5336 /* MUST advance bufptr here to avoid bogus "at end of line"
5337 context messages from yyerror().
5340 yyerror( (const char *)
5342 ? Perl_form(aTHX_ "Invalid separator character "
5343 "%c%c%c in attribute list", q, *s, q)
5344 : "Unterminated attribute list" ) );
5352 NEXTVAL_NEXTTOKE.opval = attrs;
5358 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5362 PL_lex_allbrackets--;
5366 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5367 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5371 PL_lex_allbrackets++;
5374 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5380 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5383 PL_lex_allbrackets--;
5389 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5392 if (PL_lex_brackets <= 0)
5393 /* diag_listed_as: Unmatched right %s bracket */
5394 yyerror("Unmatched right square bracket");
5397 PL_lex_allbrackets--;
5398 if (PL_lex_state == LEX_INTERPNORMAL) {
5399 if (PL_lex_brackets == 0) {
5400 if (*s == '-' && s[1] == '>')
5401 PL_lex_state = LEX_INTERPENDMAYBE;
5402 else if (*s != '[' && *s != '{')
5403 PL_lex_state = LEX_INTERPEND;
5410 if (PL_lex_brackets > 100) {
5411 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5413 switch (PL_expect) {
5415 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5416 PL_lex_allbrackets++;
5417 OPERATOR(HASHBRACK);
5419 while (s < PL_bufend && SPACE_OR_TAB(*s))
5422 PL_tokenbuf[0] = '\0';
5423 if (d < PL_bufend && *d == '-') {
5424 PL_tokenbuf[0] = '-';
5426 while (d < PL_bufend && SPACE_OR_TAB(*d))
5429 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5430 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5432 while (d < PL_bufend && SPACE_OR_TAB(*d))
5435 const char minus = (PL_tokenbuf[0] == '-');
5436 s = force_word(s + minus, WORD, FALSE, TRUE);
5444 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5445 PL_lex_allbrackets++;
5450 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5451 PL_lex_allbrackets++;
5456 if (PL_oldoldbufptr == PL_last_lop)
5457 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5459 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5460 PL_lex_allbrackets++;
5463 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5465 /* This hack is to get the ${} in the message. */
5467 yyerror("syntax error");
5470 OPERATOR(HASHBRACK);
5472 /* This hack serves to disambiguate a pair of curlies
5473 * as being a block or an anon hash. Normally, expectation
5474 * determines that, but in cases where we're not in a
5475 * position to expect anything in particular (like inside
5476 * eval"") we have to resolve the ambiguity. This code
5477 * covers the case where the first term in the curlies is a
5478 * quoted string. Most other cases need to be explicitly
5479 * disambiguated by prepending a "+" before the opening
5480 * curly in order to force resolution as an anon hash.
5482 * XXX should probably propagate the outer expectation
5483 * into eval"" to rely less on this hack, but that could
5484 * potentially break current behavior of eval"".
5488 if (*s == '\'' || *s == '"' || *s == '`') {
5489 /* common case: get past first string, handling escapes */
5490 for (t++; t < PL_bufend && *t != *s;)
5491 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5495 else if (*s == 'q') {
5498 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5499 && !isWORDCHAR(*t))))
5501 /* skip q//-like construct */
5503 char open, close, term;
5506 while (t < PL_bufend && isSPACE(*t))
5508 /* check for q => */
5509 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5510 OPERATOR(HASHBRACK);
5514 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5518 for (t++; t < PL_bufend; t++) {
5519 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5521 else if (*t == open)
5525 for (t++; t < PL_bufend; t++) {
5526 if (*t == '\\' && t+1 < PL_bufend)
5528 else if (*t == close && --brackets <= 0)
5530 else if (*t == open)
5537 /* skip plain q word */
5538 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5541 else if (isWORDCHAR_lazy_if(t,UTF)) {
5543 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5546 while (t < PL_bufend && isSPACE(*t))
5548 /* if comma follows first term, call it an anon hash */
5549 /* XXX it could be a comma expression with loop modifiers */
5550 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5551 || (*t == '=' && t[1] == '>')))
5552 OPERATOR(HASHBRACK);
5553 if (PL_expect == XREF)
5556 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5562 pl_yylval.ival = CopLINE(PL_curcop);
5563 if (isSPACE(*s) || *s == '#')
5564 PL_copline = NOLINE; /* invalidate current command line number */
5565 TOKEN(formbrack ? '=' : '{');
5567 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5571 if (PL_lex_brackets <= 0)
5572 /* diag_listed_as: Unmatched right %s bracket */
5573 yyerror("Unmatched right curly bracket");
5575 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5576 PL_lex_allbrackets--;
5577 if (PL_lex_state == LEX_INTERPNORMAL) {
5578 if (PL_lex_brackets == 0) {
5579 if (PL_expect & XFAKEBRACK) {
5580 PL_expect &= XENUMMASK;
5581 PL_lex_state = LEX_INTERPEND;
5583 return yylex(); /* ignore fake brackets */
5585 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5586 && SvEVALED(PL_lex_repl))
5587 PL_lex_state = LEX_INTERPEND;
5588 else if (*s == '-' && s[1] == '>')
5589 PL_lex_state = LEX_INTERPENDMAYBE;
5590 else if (*s != '[' && *s != '{')
5591 PL_lex_state = LEX_INTERPEND;
5594 if (PL_expect & XFAKEBRACK) {
5595 PL_expect &= XENUMMASK;
5597 return yylex(); /* ignore fake brackets */
5599 force_next(formbrack ? '.' : '}');
5600 if (formbrack) LEAVE;
5601 if (formbrack == 2) { /* means . where arguments were expected */
5607 if (PL_expect == XPOSTDEREF) POSTDEREF('&');
5610 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5611 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5618 if (PL_expect == XOPERATOR) {
5619 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5620 && isIDFIRST_lazy_if(s,UTF))
5622 CopLINE_dec(PL_curcop);
5623 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5624 CopLINE_inc(PL_curcop);
5626 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5627 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5631 PL_parser->saw_infix_sigil = 1;
5635 PL_tokenbuf[0] = '&';
5636 s = scan_ident(s - 1, PL_tokenbuf + 1,
5637 sizeof PL_tokenbuf - 1, TRUE);
5638 if (PL_tokenbuf[1]) {
5639 PL_expect = XOPERATOR;
5640 force_ident_maybe_lex('&');
5644 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5650 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5651 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5658 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5659 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5667 const char tmp = *s++;
5669 if (!PL_lex_allbrackets &&
5670 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5677 if (!PL_lex_allbrackets &&
5678 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5686 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5687 && strchr("+-*/%.^&|<",tmp))
5688 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5689 "Reversed %c= operator",(int)tmp);
5691 if (PL_expect == XSTATE && isALPHA(tmp) &&
5692 (s == PL_linestart+1 || s[-2] == '\n') )
5694 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
5695 || PL_lex_state != LEX_NORMAL) {
5700 if (strnEQ(s,"=cut",4)) {
5714 PL_parser->in_pod = 1;
5718 if (PL_expect == XBLOCK) {
5720 #ifdef PERL_STRICT_CR
5721 while (SPACE_OR_TAB(*t))
5723 while (SPACE_OR_TAB(*t) || *t == '\r')
5726 if (*t == '\n' || *t == '#') {
5729 SAVEI8(PL_parser->form_lex_state);
5730 SAVEI32(PL_lex_formbrack);
5731 PL_parser->form_lex_state = PL_lex_state;
5732 PL_lex_formbrack = PL_lex_brackets + 1;
5736 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5745 const char tmp = *s++;
5747 /* was this !=~ where !~ was meant?
5748 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5750 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5751 const char *t = s+1;
5753 while (t < PL_bufend && isSPACE(*t))
5756 if (*t == '/' || *t == '?' ||
5757 ((*t == 'm' || *t == 's' || *t == 'y')
5758 && !isWORDCHAR(t[1])) ||
5759 (*t == 't' && t[1] == 'r' && !isWORDCHAR(t[2])))
5760 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5761 "!=~ should be !~");
5763 if (!PL_lex_allbrackets &&
5764 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5776 if (PL_expect != XOPERATOR) {
5777 if (s[1] != '<' && !strchr(s,'>'))
5780 s = scan_heredoc(s);
5782 s = scan_inputsymbol(s);
5783 PL_expect = XOPERATOR;
5784 TOKEN(sublex_start());
5790 if (*s == '=' && !PL_lex_allbrackets &&
5791 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5795 SHop(OP_LEFT_SHIFT);
5800 if (!PL_lex_allbrackets &&
5801 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5808 if (!PL_lex_allbrackets &&
5809 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5817 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5825 const char tmp = *s++;
5827 if (*s == '=' && !PL_lex_allbrackets &&
5828 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5832 SHop(OP_RIGHT_SHIFT);
5834 else if (tmp == '=') {
5835 if (!PL_lex_allbrackets &&
5836 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5844 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5853 if (PL_expect == XOPERATOR) {
5854 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5855 return deprecate_commaless_var_list();
5858 else if (PL_expect == XPOSTDEREF) {
5861 POSTDEREF(DOLSHARP);
5866 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
5867 PL_tokenbuf[0] = '@';
5868 s = scan_ident(s + 1, PL_tokenbuf + 1,
5869 sizeof PL_tokenbuf - 1, FALSE);
5870 if (PL_expect == XOPERATOR)
5871 no_op("Array length", s);
5872 if (!PL_tokenbuf[1])
5874 PL_expect = XOPERATOR;
5875 force_ident_maybe_lex('#');
5879 PL_tokenbuf[0] = '$';
5880 s = scan_ident(s, PL_tokenbuf + 1,
5881 sizeof PL_tokenbuf - 1, FALSE);
5882 if (PL_expect == XOPERATOR)
5884 if (!PL_tokenbuf[1]) {
5886 yyerror("Final $ should be \\$ or $name");
5892 const char tmp = *s;
5893 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5896 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5897 && intuit_more(s)) {
5899 PL_tokenbuf[0] = '@';
5900 if (ckWARN(WARN_SYNTAX)) {
5903 while (isSPACE(*t) || isWORDCHAR_lazy_if(t,UTF) || *t == '$')
5906 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5907 while (t < PL_bufend && *t != ']')
5909 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5910 "Multidimensional syntax %.*s not supported",
5911 (int)((t - PL_bufptr) + 1), PL_bufptr);
5915 else if (*s == '{') {
5917 PL_tokenbuf[0] = '%';
5918 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5919 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5921 char tmpbuf[sizeof PL_tokenbuf];
5924 } while (isSPACE(*t));
5925 if (isIDFIRST_lazy_if(t,UTF)) {
5927 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5932 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
5933 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5934 "You need to quote \"%"UTF8f"\"",
5935 UTF8fARG(UTF, len, tmpbuf));
5941 PL_expect = XOPERATOR;
5942 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5943 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5944 if (!islop || PL_last_lop_op == OP_GREPSTART)
5945 PL_expect = XOPERATOR;
5946 else if (strchr("$@\"'`q", *s))
5947 PL_expect = XTERM; /* e.g. print $fh "foo" */
5948 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5949 PL_expect = XTERM; /* e.g. print $fh &sub */
5950 else if (isIDFIRST_lazy_if(s,UTF)) {
5951 char tmpbuf[sizeof PL_tokenbuf];
5953 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5954 if ((t2 = keyword(tmpbuf, len, 0))) {
5955 /* binary operators exclude handle interpretations */
5967 PL_expect = XTERM; /* e.g. print $fh length() */
5972 PL_expect = XTERM; /* e.g. print $fh subr() */
5975 else if (isDIGIT(*s))
5976 PL_expect = XTERM; /* e.g. print $fh 3 */
5977 else if (*s == '.' && isDIGIT(s[1]))
5978 PL_expect = XTERM; /* e.g. print $fh .3 */
5979 else if ((*s == '?' || *s == '-' || *s == '+')
5980 && !isSPACE(s[1]) && s[1] != '=')
5981 PL_expect = XTERM; /* e.g. print $fh -1 */
5982 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5984 PL_expect = XTERM; /* e.g. print $fh /.../
5985 XXX except DORDOR operator
5987 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5989 PL_expect = XTERM; /* print $fh <<"EOF" */
5992 force_ident_maybe_lex('$');
5996 if (PL_expect == XOPERATOR)
5998 else if (PL_expect == XPOSTDEREF) POSTDEREF('@');
5999 PL_tokenbuf[0] = '@';
6000 s = scan_ident(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6002 if (!PL_tokenbuf[1]) {
6005 if (PL_lex_state == LEX_NORMAL)
6007 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6009 PL_tokenbuf[0] = '%';
6011 /* Warn about @ where they meant $. */
6012 if (*s == '[' || *s == '{') {
6013 if (ckWARN(WARN_SYNTAX)) {
6014 S_check_scalar_slice(aTHX_ s);
6018 PL_expect = XOPERATOR;
6019 force_ident_maybe_lex('@');
6022 case '/': /* may be division, defined-or, or pattern */
6023 if ((PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR) && s[1] == '/') {
6024 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6025 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6030 else if (PL_expect == XOPERATOR) {
6032 if (*s == '=' && !PL_lex_allbrackets &&
6033 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6040 /* Disable warning on "study /blah/" */
6041 if (PL_oldoldbufptr == PL_last_uni
6042 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6043 || memNE(PL_last_uni, "study", 5)
6044 || isWORDCHAR_lazy_if(PL_last_uni+5,UTF)
6047 s = scan_pat(s,OP_MATCH);
6048 TERM(sublex_start());
6051 case '?': /* conditional */
6053 if (!PL_lex_allbrackets &&
6054 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6058 PL_lex_allbrackets++;
6062 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6063 #ifdef PERL_STRICT_CR
6066 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6068 && (s == PL_linestart || s[-1] == '\n') )
6071 formbrack = 2; /* dot seen where arguments expected */
6074 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6078 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6081 if (!PL_lex_allbrackets &&
6082 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6089 pl_yylval.ival = OPf_SPECIAL;
6095 if (*s == '=' && !PL_lex_allbrackets &&
6096 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6103 case '0': case '1': case '2': case '3': case '4':
6104 case '5': case '6': case '7': case '8': case '9':
6105 s = scan_num(s, &pl_yylval);
6106 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6107 if (PL_expect == XOPERATOR)
6112 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6115 COPLINE_SET_FROM_MULTI_END;
6116 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6117 if (PL_expect == XOPERATOR) {
6118 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6119 return deprecate_commaless_var_list();
6124 pl_yylval.ival = OP_CONST;
6125 TERM(sublex_start());
6128 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6131 printbuf("### Saw string before %s\n", s);
6133 PerlIO_printf(Perl_debug_log,
6134 "### Saw unterminated string\n");
6136 if (PL_expect == XOPERATOR) {
6137 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6138 return deprecate_commaless_var_list();
6145 pl_yylval.ival = OP_CONST;
6146 /* FIXME. I think that this can be const if char *d is replaced by
6147 more localised variables. */
6148 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6149 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6150 pl_yylval.ival = OP_STRINGIFY;
6154 if (pl_yylval.ival == OP_CONST)
6155 COPLINE_SET_FROM_MULTI_END;
6156 TERM(sublex_start());
6159 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6160 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6161 if (PL_expect == XOPERATOR)
6162 no_op("Backticks",s);
6165 pl_yylval.ival = OP_BACKTICK;
6166 TERM(sublex_start());
6170 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6172 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6174 if (PL_expect == XOPERATOR)
6175 no_op("Backslash",s);
6179 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6180 char *start = s + 2;
6181 while (isDIGIT(*start) || *start == '_')
6183 if (*start == '.' && isDIGIT(start[1])) {
6184 s = scan_num(s, &pl_yylval);
6187 else if ((*start == ':' && start[1] == ':')
6188 || (PL_expect == XSTATE && *start == ':'))
6190 else if (PL_expect == XSTATE) {
6192 while (d < PL_bufend && isSPACE(*d)) d++;
6193 if (*d == ':') goto keylookup;
6195 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6196 if (!isALPHA(*start) && (PL_expect == XTERM
6197 || PL_expect == XREF || PL_expect == XSTATE
6198 || PL_expect == XTERMORDORDOR)) {
6199 GV *const gv = gv_fetchpvn_flags(s, start - s,
6200 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6202 s = scan_num(s, &pl_yylval);
6209 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6262 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6264 /* Some keywords can be followed by any delimiter, including ':' */
6265 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6267 /* x::* is just a word, unless x is "CORE" */
6268 if (!anydelim && *s == ':' && s[1] == ':') {
6269 if (strEQ(PL_tokenbuf, "CORE")) goto case_KEY_CORE;
6274 while (d < PL_bufend && isSPACE(*d))
6275 d++; /* no comments skipped here, or s### is misparsed */
6277 /* Is this a word before a => operator? */
6278 if (*d == '=' && d[1] == '>') {
6282 = (OP*)newSVOP(OP_CONST, 0,
6283 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6284 pl_yylval.opval->op_private = OPpCONST_BARE;
6288 /* Check for plugged-in keyword */
6292 char *saved_bufptr = PL_bufptr;
6294 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6296 if (result == KEYWORD_PLUGIN_DECLINE) {
6297 /* not a plugged-in keyword */
6298 PL_bufptr = saved_bufptr;
6299 } else if (result == KEYWORD_PLUGIN_STMT) {
6300 pl_yylval.opval = o;
6303 return REPORT(PLUGSTMT);
6304 } else if (result == KEYWORD_PLUGIN_EXPR) {
6305 pl_yylval.opval = o;
6307 PL_expect = XOPERATOR;
6308 return REPORT(PLUGEXPR);
6310 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6315 /* Check for built-in keyword */
6316 tmp = keyword(PL_tokenbuf, len, 0);
6318 /* Is this a label? */
6319 if (!anydelim && PL_expect == XSTATE
6320 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6322 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6323 pl_yylval.pval[len] = '\0';
6324 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6329 /* Check for lexical sub */
6330 if (PL_expect != XOPERATOR) {
6331 char tmpbuf[sizeof PL_tokenbuf + 1];
6333 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6334 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6335 if (off != NOT_IN_PAD) {
6336 assert(off); /* we assume this is boolean-true below */
6337 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6338 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6339 HEK * const stashname = HvNAME_HEK(stash);
6340 sv = newSVhek(stashname);
6341 sv_catpvs(sv, "::");
6342 sv_catpvn_flags(sv, PL_tokenbuf, len,
6343 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6344 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6354 rv2cv_op = newOP(OP_PADANY, 0);
6355 rv2cv_op->op_targ = off;
6356 cv = find_lexical_cv(off);
6364 if (tmp < 0) { /* second-class keyword? */
6365 GV *ogv = NULL; /* override (winner) */
6366 GV *hgv = NULL; /* hidden (loser) */
6367 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6369 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6370 (UTF ? SVf_UTF8 : 0)|GV_NOTQUAL,
6374 if (GvIMPORTED_CV(gv))
6376 else if (! CvMETHOD(cv))
6380 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6384 ? GvCVu(gv) && GvIMPORTED_CV(gv)
6385 : SvPCS_IMPORTED(gv)
6386 && (gv_init(gv, PL_globalstash, PL_tokenbuf,
6395 tmp = 0; /* overridden by import or by GLOBAL */
6398 && -tmp==KEY_lock /* XXX generalizable kludge */
6401 tmp = 0; /* any sub overrides "weak" keyword */
6403 else { /* no override */
6405 if (tmp == KEY_dump) {
6406 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6407 "dump() better written as CORE::dump()");
6411 if (hgv && tmp != KEY_x) /* never ambiguous */
6412 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6413 "Ambiguous call resolved as CORE::%s(), "
6414 "qualify as such or use &",
6419 if (tmp && tmp != KEY___DATA__ && tmp != KEY___END__
6420 && (!anydelim || *s != '#')) {
6421 /* no override, and not s### either; skipspace is safe here
6422 * check for => on following line */
6424 STRLEN bufoff = PL_bufptr - SvPVX(PL_linestr);
6425 STRLEN soff = s - SvPVX(PL_linestr);
6426 s = skipspace_flags(s, LEX_NO_INCLINE);
6427 arrow = *s == '=' && s[1] == '>';
6428 PL_bufptr = SvPVX(PL_linestr) + bufoff;
6429 s = SvPVX(PL_linestr) + soff;
6437 default: /* not a keyword */
6438 /* Trade off - by using this evil construction we can pull the
6439 variable gv into the block labelled keylookup. If not, then
6440 we have to give it function scope so that the goto from the
6441 earlier ':' case doesn't bypass the initialisation. */
6443 just_a_word_zero_gv:
6455 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6456 const char penultchar =
6457 lastchar && PL_bufptr - 2 >= PL_linestart
6462 /* Get the rest if it looks like a package qualifier */
6464 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6466 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6469 Perl_croak(aTHX_ "Bad name after %"UTF8f"%s",
6470 UTF8fARG(UTF, len, PL_tokenbuf),
6471 *s == '\'' ? "'" : "::");
6476 if (PL_expect == XOPERATOR) {
6477 if (PL_bufptr == PL_linestart) {
6478 CopLINE_dec(PL_curcop);
6479 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6480 CopLINE_inc(PL_curcop);
6483 no_op("Bareword",s);
6486 /* Look for a subroutine with this name in current package,
6487 unless this is a lexical sub, or name is "Foo::",
6488 in which case Foo is a bareword
6489 (and a package name). */
6492 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6494 if (ckWARN(WARN_BAREWORD)
6495 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6496 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6497 "Bareword \"%"UTF8f"\" refers to nonexistent package",
6498 UTF8fARG(UTF, len, PL_tokenbuf));
6500 PL_tokenbuf[len] = '\0';
6506 /* Mustn't actually add anything to a symbol table.
6507 But also don't want to "initialise" any placeholder
6508 constants that might already be there into full
6509 blown PVGVs with attached PVCV. */
6510 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6511 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6517 /* if we saw a global override before, get the right name */
6520 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6521 len ? len : strlen(PL_tokenbuf));
6523 SV * const tmp_sv = sv;
6524 sv = newSVpvs("CORE::GLOBAL::");
6525 sv_catsv(sv, tmp_sv);
6526 SvREFCNT_dec(tmp_sv);
6530 /* Presume this is going to be a bareword of some sort. */
6532 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6533 pl_yylval.opval->op_private = OPpCONST_BARE;
6535 /* And if "Foo::", then that's what it certainly is. */
6541 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6542 const_op->op_private = OPpCONST_BARE;
6543 rv2cv_op = newCVREF(0, const_op);
6544 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
6547 /* See if it's the indirect object for a list operator. */
6549 if (PL_oldoldbufptr &&
6550 PL_oldoldbufptr < PL_bufptr &&
6551 (PL_oldoldbufptr == PL_last_lop
6552 || PL_oldoldbufptr == PL_last_uni) &&
6553 /* NO SKIPSPACE BEFORE HERE! */
6554 (PL_expect == XREF ||
6555 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6557 bool immediate_paren = *s == '(';
6559 /* (Now we can afford to cross potential line boundary.) */
6560 s = SKIPSPACE2(s,nextPL_nextwhite);
6562 /* Two barewords in a row may indicate method call. */
6564 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6565 (tmp = intuit_method(s, gv, cv))) {
6567 if (tmp == METHOD && !PL_lex_allbrackets &&
6568 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6569 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6573 /* If not a declared subroutine, it's an indirect object. */
6574 /* (But it's an indir obj regardless for sort.) */
6575 /* Also, if "_" follows a filetest operator, it's a bareword */
6578 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6580 (PL_last_lop_op != OP_MAPSTART &&
6581 PL_last_lop_op != OP_GREPSTART))))
6582 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6583 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6586 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6591 PL_expect = XOPERATOR;
6594 /* Is this a word before a => operator? */
6595 if (*s == '=' && s[1] == '>' && !pkgname) {
6598 /* This is our own scalar, created a few lines above,
6600 SvREADONLY_off(cSVOPx(pl_yylval.opval)->op_sv);
6601 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6602 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6603 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6604 SvREADONLY_on(cSVOPx(pl_yylval.opval)->op_sv);
6608 /* If followed by a paren, it's certainly a subroutine. */
6613 while (SPACE_OR_TAB(*d))
6615 if (*d == ')' && (sv = cv_const_sv_or_av(cv))) {
6620 NEXTVAL_NEXTTOKE.opval =
6621 off ? rv2cv_op : pl_yylval.opval;
6622 PL_expect = XOPERATOR;
6624 op_free(pl_yylval.opval), force_next(PRIVATEREF);
6625 else op_free(rv2cv_op), force_next(WORD);
6630 /* If followed by var or block, call it a method (unless sub) */
6632 if ((*s == '$' || *s == '{') && !cv) {
6634 PL_last_lop = PL_oldbufptr;
6635 PL_last_lop_op = OP_METHOD;
6636 if (!PL_lex_allbrackets &&
6637 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6638 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6642 /* If followed by a bareword, see if it looks like indir obj. */
6645 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6646 && (tmp = intuit_method(s, gv, cv))) {
6648 if (tmp == METHOD && !PL_lex_allbrackets &&
6649 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6650 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6654 /* Not a method, so call it a subroutine (if defined) */
6657 if (lastchar == '-' && penultchar != '-') {
6658 const STRLEN l = len ? len : strlen(PL_tokenbuf);
6659 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6660 "Ambiguous use of -%"UTF8f" resolved as -&%"UTF8f"()",
6661 UTF8fARG(UTF, l, PL_tokenbuf),
6662 UTF8fARG(UTF, l, PL_tokenbuf));
6664 /* Check for a constant sub */
6665 if ((sv = cv_const_sv_or_av(cv))) {
6668 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6669 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6670 if (SvTYPE(sv) == SVt_PVAV)
6671 pl_yylval.opval = newUNOP(OP_RV2AV, OPf_PARENS,
6674 pl_yylval.opval->op_private = 0;
6675 pl_yylval.opval->op_folded = 1;
6676 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6681 op_free(pl_yylval.opval);
6683 off ? (OP *)newCVREF(0, rv2cv_op) : rv2cv_op;
6684 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6685 PL_last_lop = PL_oldbufptr;
6686 PL_last_lop_op = OP_ENTERSUB;
6687 /* Is there a prototype? */
6691 STRLEN protolen = CvPROTOLEN(cv);
6692 const char *proto = CvPROTO(cv);
6694 proto = S_strip_spaces(aTHX_ proto, &protolen);
6697 if ((optional = *proto == ';'))
6700 while (*proto == ';');
6704 *proto == '$' || *proto == '_'
6705 || *proto == '*' || *proto == '+'
6710 *proto == '\\' && proto[1] && proto[2] == '\0'
6713 UNIPROTO(UNIOPSUB,optional);
6714 if (*proto == '\\' && proto[1] == '[') {
6715 const char *p = proto + 2;
6716 while(*p && *p != ']')
6718 if(*p == ']' && !p[1])
6719 UNIPROTO(UNIOPSUB,optional);
6721 if (*proto == '&' && *s == '{') {
6723 sv_setpvs(PL_subname, "__ANON__");
6725 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6726 if (!PL_lex_allbrackets &&
6727 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6728 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6732 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6734 force_next(off ? PRIVATEREF : WORD);
6735 if (!PL_lex_allbrackets &&
6736 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6737 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6741 /* Call it a bare word */
6743 if (PL_hints & HINT_STRICT_SUBS)
6744 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6747 /* after "print" and similar functions (corresponding to
6748 * "F? L" in opcode.pl), whatever wasn't already parsed as
6749 * a filehandle should be subject to "strict subs".
6750 * Likewise for the optional indirect-object argument to system
6751 * or exec, which can't be a bareword */
6752 if ((PL_last_lop_op == OP_PRINT
6753 || PL_last_lop_op == OP_PRTF
6754 || PL_last_lop_op == OP_SAY
6755 || PL_last_lop_op == OP_SYSTEM
6756 || PL_last_lop_op == OP_EXEC)
6757 && (PL_hints & HINT_STRICT_SUBS))
6758 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6759 if (lastchar != '-') {
6760 if (ckWARN(WARN_RESERVED)) {
6764 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
6766 /* PL_warn_reserved is constant */
6767 GCC_DIAG_IGNORE(-Wformat-nonliteral);
6768 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6778 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
6779 && saw_infix_sigil) {
6780 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6781 "Operator or semicolon missing before %c%"UTF8f,
6783 UTF8fARG(UTF, strlen(PL_tokenbuf),
6785 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6786 "Ambiguous use of %c resolved as operator %c",
6787 lastchar, lastchar);
6794 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
6799 (OP*)newSVOP(OP_CONST, 0,
6800 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
6803 case KEY___PACKAGE__:
6805 (OP*)newSVOP(OP_CONST, 0,
6807 ? newSVhek(HvNAME_HEK(PL_curstash))
6814 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6815 HV * const stash = PL_tokenbuf[2] == 'D' && PL_curstash
6818 gv = (GV *)*hv_fetchs(stash, "DATA", 1);
6820 gv_init(gv,stash,"DATA",4,0);
6823 GvIOp(gv) = newIO();
6824 IoIFP(GvIOp(gv)) = PL_rsfp;
6825 #if defined(HAS_FCNTL) && defined(F_SETFD)
6827 const int fd = PerlIO_fileno(PL_rsfp);
6828 fcntl(fd,F_SETFD,fd >= 3);
6831 /* Mark this internal pseudo-handle as clean */
6832 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6833 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6834 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6836 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6837 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6838 /* if the script was opened in binmode, we need to revert
6839 * it to text mode for compatibility; but only iff it has CRs
6840 * XXX this is a questionable hack at best. */
6841 if (PL_bufend-PL_bufptr > 2
6842 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6845 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6846 loc = PerlIO_tell(PL_rsfp);
6847 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6850 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6852 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6853 #endif /* NETWARE */
6855 PerlIO_seek(PL_rsfp, loc, 0);
6859 #ifdef PERLIO_LAYERS
6862 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6863 else if (PL_encoding) {
6869 XPUSHs(PL_encoding);
6871 call_method("name", G_SCALAR);
6875 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6876 Perl_form(aTHX_ ":encoding(%"SVf")",
6889 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
6898 if (PL_expect == XSTATE) {
6909 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6910 if ((*s == ':' && s[1] == ':')
6911 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
6915 Copy(PL_bufptr, PL_tokenbuf, olen, char);
6919 Perl_croak(aTHX_ "CORE::%"UTF8f" is not a keyword",
6920 UTF8fARG(UTF, len, PL_tokenbuf));
6923 else if (tmp == KEY_require || tmp == KEY_do
6925 /* that's a way to remember we saw "CORE::" */
6937 LOP(OP_ACCEPT,XTERM);
6940 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
6945 LOP(OP_ATAN2,XTERM);
6951 LOP(OP_BINMODE,XTERM);
6954 LOP(OP_BLESS,XTERM);
6963 /* We have to disambiguate the two senses of
6964 "continue". If the next token is a '{' then
6965 treat it as the start of a continue block;
6966 otherwise treat it as a control operator.
6976 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6986 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6995 if (!PL_cryptseen) {
6996 PL_cryptseen = TRUE;
7000 LOP(OP_CRYPT,XTERM);
7003 LOP(OP_CHMOD,XTERM);
7006 LOP(OP_CHOWN,XTERM);
7009 LOP(OP_CONNECT,XTERM);
7029 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7031 if (len && (len != 4 || strNE(PL_tokenbuf+1, "CORE"))
7032 && !keyword(PL_tokenbuf + 1, len, 0)) {
7035 force_ident_maybe_lex('&');
7040 if (orig_keyword == KEY_do) {
7049 PL_hints |= HINT_BLOCK_SCOPE;
7059 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7060 STR_WITH_LEN("NDBM_File::"),
7061 STR_WITH_LEN("DB_File::"),
7062 STR_WITH_LEN("GDBM_File::"),
7063 STR_WITH_LEN("SDBM_File::"),
7064 STR_WITH_LEN("ODBM_File::"),
7066 LOP(OP_DBMOPEN,XTERM);
7072 PL_expect = XOPERATOR;
7073 s = force_word(s,WORD,TRUE,FALSE);
7080 pl_yylval.ival = CopLINE(PL_curcop);
7084 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7096 if (*s == '{') { /* block eval */
7097 PL_expect = XTERMBLOCK;
7098 UNIBRACK(OP_ENTERTRY);
7100 else { /* string eval */
7102 UNIBRACK(OP_ENTEREVAL);
7107 UNIBRACK(-OP_ENTEREVAL);
7121 case KEY_endhostent:
7127 case KEY_endservent:
7130 case KEY_endprotoent:
7141 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7143 pl_yylval.ival = CopLINE(PL_curcop);
7145 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7148 if ((PL_bufend - p) >= 3 &&
7149 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7151 else if ((PL_bufend - p) >= 4 &&
7152 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7155 /* skip optional package name, as in "for my abc $x (..)" */
7156 if (isIDFIRST_lazy_if(p,UTF)) {
7157 p = scan_word(p, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7161 Perl_croak(aTHX_ "Missing $ on loop variable");
7166 LOP(OP_FORMLINE,XTERM);
7175 LOP(OP_FCNTL,XTERM);
7181 LOP(OP_FLOCK,XTERM);
7184 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7189 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7194 LOP(OP_GREPSTART, XREF);
7197 PL_expect = XOPERATOR;
7198 s = force_word(s,WORD,TRUE,FALSE);
7213 case KEY_getpriority:
7214 LOP(OP_GETPRIORITY,XTERM);
7216 case KEY_getprotobyname:
7219 case KEY_getprotobynumber:
7220 LOP(OP_GPBYNUMBER,XTERM);
7222 case KEY_getprotoent:
7234 case KEY_getpeername:
7235 UNI(OP_GETPEERNAME);
7237 case KEY_gethostbyname:
7240 case KEY_gethostbyaddr:
7241 LOP(OP_GHBYADDR,XTERM);
7243 case KEY_gethostent:
7246 case KEY_getnetbyname:
7249 case KEY_getnetbyaddr:
7250 LOP(OP_GNBYADDR,XTERM);
7255 case KEY_getservbyname:
7256 LOP(OP_GSBYNAME,XTERM);
7258 case KEY_getservbyport:
7259 LOP(OP_GSBYPORT,XTERM);
7261 case KEY_getservent:
7264 case KEY_getsockname:
7265 UNI(OP_GETSOCKNAME);
7267 case KEY_getsockopt:
7268 LOP(OP_GSOCKOPT,XTERM);
7283 pl_yylval.ival = CopLINE(PL_curcop);
7284 Perl_ck_warner_d(aTHX_
7285 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
7286 "given is experimental");
7291 orig_keyword==KEY_glob ? -OP_GLOB : OP_GLOB,
7299 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7301 pl_yylval.ival = CopLINE(PL_curcop);
7305 LOP(OP_INDEX,XTERM);
7311 LOP(OP_IOCTL,XTERM);
7323 PL_expect = XOPERATOR;
7324 s = force_word(s,WORD,TRUE,FALSE);
7341 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7346 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7360 LOP(OP_LISTEN,XTERM);
7369 s = scan_pat(s,OP_MATCH);
7370 TERM(sublex_start());
7373 LOP(OP_MAPSTART, XREF);
7376 LOP(OP_MKDIR,XTERM);
7379 LOP(OP_MSGCTL,XTERM);
7382 LOP(OP_MSGGET,XTERM);
7385 LOP(OP_MSGRCV,XTERM);
7388 LOP(OP_MSGSND,XTERM);
7393 PL_in_my = (U16)tmp;
7395 if (isIDFIRST_lazy_if(s,UTF)) {
7396 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7397 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7399 if (!FEATURE_LEXSUBS_IS_ENABLED)
7401 "Experimental \"%s\" subs not enabled",
7402 tmp == KEY_my ? "my" :
7403 tmp == KEY_state ? "state" : "our");
7404 Perl_ck_warner_d(aTHX_
7405 packWARN(WARN_EXPERIMENTAL__LEXICAL_SUBS),
7406 "The lexical_subs feature is experimental");
7409 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7410 if (!PL_in_my_stash) {
7414 len = my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7415 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(tmpbuf));
7416 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7423 PL_expect = XOPERATOR;
7424 s = force_word(s,WORD,TRUE,FALSE);
7428 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7433 s = tokenize_use(0, s);
7437 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7440 if (!PL_lex_allbrackets &&
7441 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7442 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7448 if (isIDFIRST_lazy_if(s,UTF)) {
7450 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE,
7452 for (t=d; isSPACE(*t);)
7454 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7456 && !(t[0] == '=' && t[1] == '>')
7457 && !(t[0] == ':' && t[1] == ':')
7458 && !keyword(s, d-s, 0)
7460 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7461 "Precedence problem: open %"UTF8f" should be open(%"UTF8f")",
7462 UTF8fARG(UTF, d-s, s), UTF8fARG(UTF, d-s, s));
7468 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7470 pl_yylval.ival = OP_OR;
7480 LOP(OP_OPEN_DIR,XTERM);
7483 checkcomma(s,PL_tokenbuf,"filehandle");
7487 checkcomma(s,PL_tokenbuf,"filehandle");
7506 s = force_word(s,WORD,FALSE,TRUE);
7508 s = force_strict_version(s);
7509 PL_lex_expect = XBLOCK;
7513 LOP(OP_PIPE_OP,XTERM);
7516 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7519 COPLINE_SET_FROM_MULTI_END;
7520 pl_yylval.ival = OP_CONST;
7521 TERM(sublex_start());
7528 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7531 COPLINE_SET_FROM_MULTI_END;
7532 PL_expect = XOPERATOR;
7533 if (SvCUR(PL_lex_stuff)) {
7534 int warned_comma = !ckWARN(WARN_QW);
7535 int warned_comment = warned_comma;
7536 d = SvPV_force(PL_lex_stuff, len);
7538 for (; isSPACE(*d) && len; --len, ++d)
7543 if (!warned_comma || !warned_comment) {
7544 for (; !isSPACE(*d) && len; --len, ++d) {
7545 if (!warned_comma && *d == ',') {
7546 Perl_warner(aTHX_ packWARN(WARN_QW),
7547 "Possible attempt to separate words with commas");
7550 else if (!warned_comment && *d == '#') {
7551 Perl_warner(aTHX_ packWARN(WARN_QW),
7552 "Possible attempt to put comments in qw() list");
7558 for (; !isSPACE(*d) && len; --len, ++d)
7561 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7562 words = op_append_elem(OP_LIST, words,
7563 newSVOP(OP_CONST, 0, tokeq(sv)));
7568 words = newNULLLIST();
7570 SvREFCNT_dec(PL_lex_stuff);
7571 PL_lex_stuff = NULL;
7573 PL_expect = XOPERATOR;
7574 pl_yylval.opval = sawparens(words);
7579 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7582 pl_yylval.ival = OP_STRINGIFY;
7583 if (SvIVX(PL_lex_stuff) == '\'')
7584 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7585 TERM(sublex_start());
7588 s = scan_pat(s,OP_QR);
7589 TERM(sublex_start());
7592 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7595 pl_yylval.ival = OP_BACKTICK;
7596 TERM(sublex_start());
7603 PL_expect = XOPERATOR;
7605 s = force_version(s, FALSE);
7607 else if (*s != 'v' || !isDIGIT(s[1])
7608 || (s = force_version(s, TRUE), *s == 'v'))
7610 *PL_tokenbuf = '\0';
7611 s = force_word(s,WORD,TRUE,TRUE);
7612 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7613 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
7614 GV_ADD | (UTF ? SVf_UTF8 : 0));
7616 yyerror("<> at require-statement should be quotes");
7618 if (orig_keyword == KEY_require) {
7626 PL_last_uni = PL_oldbufptr;
7627 PL_last_lop_op = OP_REQUIRE;
7629 return REPORT( (int)REQUIRE );
7635 PL_expect = XOPERATOR;
7636 s = force_word(s,WORD,TRUE,FALSE);
7640 LOP(OP_RENAME,XTERM);
7649 LOP(OP_RINDEX,XTERM);
7658 UNIDOR(OP_READLINE);
7661 UNIDOR(OP_BACKTICK);
7670 LOP(OP_REVERSE,XTERM);
7673 UNIDOR(OP_READLINK);
7680 if (pl_yylval.opval)
7681 TERM(sublex_start());
7683 TOKEN(1); /* force error */
7686 checkcomma(s,PL_tokenbuf,"filehandle");
7696 LOP(OP_SELECT,XTERM);
7702 LOP(OP_SEMCTL,XTERM);
7705 LOP(OP_SEMGET,XTERM);
7708 LOP(OP_SEMOP,XTERM);
7714 LOP(OP_SETPGRP,XTERM);
7716 case KEY_setpriority:
7717 LOP(OP_SETPRIORITY,XTERM);
7719 case KEY_sethostent:
7725 case KEY_setservent:
7728 case KEY_setprotoent:
7738 LOP(OP_SEEKDIR,XTERM);
7740 case KEY_setsockopt:
7741 LOP(OP_SSOCKOPT,XTERM);
7747 LOP(OP_SHMCTL,XTERM);
7750 LOP(OP_SHMGET,XTERM);
7753 LOP(OP_SHMREAD,XTERM);
7756 LOP(OP_SHMWRITE,XTERM);
7759 LOP(OP_SHUTDOWN,XTERM);
7768 LOP(OP_SOCKET,XTERM);
7770 case KEY_socketpair:
7771 LOP(OP_SOCKPAIR,XTERM);
7774 checkcomma(s,PL_tokenbuf,"subroutine name");
7777 s = force_word(s,WORD,TRUE,TRUE);
7781 LOP(OP_SPLIT,XTERM);
7784 LOP(OP_SPRINTF,XTERM);
7787 LOP(OP_SPLICE,XTERM);
7802 LOP(OP_SUBSTR,XTERM);
7808 char * const tmpbuf = PL_tokenbuf + 1;
7809 expectation attrful;
7810 bool have_name, have_proto;
7811 const int key = tmp;
7812 SV *format_name = NULL;
7817 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7818 (*s == ':' && s[1] == ':'))
7822 attrful = XATTRBLOCK;
7823 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
7825 if (key == KEY_format)
7826 format_name = S_newSV_maybe_utf8(aTHX_ s, d - s);
7828 if (memchr(tmpbuf, ':', len) || key != KEY_sub
7830 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
7832 sv_setpvn(PL_subname, tmpbuf, len);
7834 sv_setsv(PL_subname,PL_curstname);
7835 sv_catpvs(PL_subname,"::");
7836 sv_catpvn(PL_subname,tmpbuf,len);
7838 if (SvUTF8(PL_linestr))
7839 SvUTF8_on(PL_subname);
7846 if (key == KEY_my || key == KEY_our || key==KEY_state)
7849 /* diag_listed_as: Missing name in "%s sub" */
7851 "Missing name in \"%s\"", PL_bufptr);
7853 PL_expect = XTERMBLOCK;
7854 attrful = XATTRTERM;
7855 sv_setpvs(PL_subname,"?");
7859 if (key == KEY_format) {
7861 NEXTVAL_NEXTTOKE.opval
7862 = (OP*)newSVOP(OP_CONST,0, format_name);
7863 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
7869 /* Look for a prototype */
7870 if (*s == '(' && !FEATURE_SIGNATURES_IS_ENABLED) {
7871 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7872 COPLINE_SET_FROM_MULTI_END;
7874 Perl_croak(aTHX_ "Prototype not terminated");
7875 (void)validate_proto(PL_subname, PL_lex_stuff, ckWARN(WARN_ILLEGALPROTO));
7883 if (*s == ':' && s[1] != ':')
7884 PL_expect = attrful;
7885 else if ((*s != '{' && *s != '(') && key == KEY_sub) {
7887 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7888 else if (*s != ';' && *s != '}')
7889 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7893 NEXTVAL_NEXTTOKE.opval =
7894 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7895 PL_lex_stuff = NULL;
7900 sv_setpvs(PL_subname, "__ANON__");
7902 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7905 force_ident_maybe_lex('&');
7910 LOP(OP_SYSTEM,XREF);
7913 LOP(OP_SYMLINK,XTERM);
7916 LOP(OP_SYSCALL,XTERM);
7919 LOP(OP_SYSOPEN,XTERM);
7922 LOP(OP_SYSSEEK,XTERM);
7925 LOP(OP_SYSREAD,XTERM);
7928 LOP(OP_SYSWRITE,XTERM);
7933 TERM(sublex_start());
7954 LOP(OP_TRUNCATE,XTERM);
7966 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7968 pl_yylval.ival = CopLINE(PL_curcop);
7972 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7974 pl_yylval.ival = CopLINE(PL_curcop);
7978 LOP(OP_UNLINK,XTERM);
7984 LOP(OP_UNPACK,XTERM);
7987 LOP(OP_UTIME,XTERM);
7993 LOP(OP_UNSHIFT,XTERM);
7996 s = tokenize_use(1, s);
8006 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8008 pl_yylval.ival = CopLINE(PL_curcop);
8009 Perl_ck_warner_d(aTHX_
8010 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
8011 "when is experimental");
8015 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8017 pl_yylval.ival = CopLINE(PL_curcop);
8021 PL_hints |= HINT_BLOCK_SCOPE;
8028 LOP(OP_WAITPID,XTERM);
8034 /* Make sure $^L is defined. 0x0C is CTRL-L on ASCII platforms, and
8035 * we use the same number on EBCDIC */
8036 gv_fetchpvs("\x0C", GV_ADD|GV_NOTQUAL, SVt_PV);
8040 if (PL_expect == XOPERATOR) {
8041 if (*s == '=' && !PL_lex_allbrackets &&
8042 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8050 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8052 pl_yylval.ival = OP_XOR;
8061 Looks up an identifier in the pad or in a package
8064 PRIVATEREF if this is a lexical name.
8065 WORD if this belongs to a package.
8068 if we're in a my declaration
8069 croak if they tried to say my($foo::bar)
8070 build the ops for a my() declaration
8071 if it's an access to a my() variable
8072 build ops for access to a my() variable
8073 if in a dq string, and they've said @foo and we can't find @foo
8075 build ops for a bareword
8079 S_pending_ident(pTHX)
8082 const char pit = (char)pl_yylval.ival;
8083 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8084 /* All routes through this function want to know if there is a colon. */
8085 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8087 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8088 "### Pending identifier '%s'\n", PL_tokenbuf); });
8090 /* if we're in a my(), we can't allow dynamics here.
8091 $foo'bar has already been turned into $foo::bar, so
8092 just check for colons.
8094 if it's a legal name, the OP is a PADANY.
8097 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8099 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8100 "variable %s in \"our\"",
8101 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8102 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8106 /* PL_no_myglob is constant */
8107 GCC_DIAG_IGNORE(-Wformat-nonliteral);
8108 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8109 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8110 UTF ? SVf_UTF8 : 0);
8114 pl_yylval.opval = newOP(OP_PADANY, 0);
8115 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8116 UTF ? SVf_UTF8 : 0);
8122 build the ops for accesses to a my() variable.
8127 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8128 UTF ? SVf_UTF8 : 0);
8129 if (tmp != NOT_IN_PAD) {
8130 /* might be an "our" variable" */
8131 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8132 /* build ops for a bareword */
8133 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8134 HEK * const stashname = HvNAME_HEK(stash);
8135 SV * const sym = newSVhek(stashname);
8136 sv_catpvs(sym, "::");
8137 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8138 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8139 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8143 ? (GV_ADDMULTI | GV_ADDINEVAL)
8146 ((PL_tokenbuf[0] == '$') ? SVt_PV
8147 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8152 pl_yylval.opval = newOP(OP_PADANY, 0);
8153 pl_yylval.opval->op_targ = tmp;
8159 Whine if they've said @foo in a doublequoted string,
8160 and @foo isn't a variable we can find in the symbol
8163 if (ckWARN(WARN_AMBIGUOUS) &&
8164 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8165 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8166 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8167 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8168 /* DO NOT warn for @- and @+ */
8169 && !( PL_tokenbuf[2] == '\0' &&
8170 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8173 /* Downgraded from fatal to warning 20000522 mjd */
8174 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8175 "Possible unintended interpolation of %"UTF8f
8177 UTF8fARG(UTF, tokenbuf_len, PL_tokenbuf));
8181 /* build ops for a bareword */
8182 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8183 newSVpvn_flags(PL_tokenbuf + 1,
8185 UTF ? SVf_UTF8 : 0 ));
8186 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8188 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8189 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8190 | ( UTF ? SVf_UTF8 : 0 ),
8191 ((PL_tokenbuf[0] == '$') ? SVt_PV
8192 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8198 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8200 PERL_ARGS_ASSERT_CHECKCOMMA;
8202 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8203 if (ckWARN(WARN_SYNTAX)) {
8206 for (w = s+2; *w && level; w++) {
8214 /* the list of chars below is for end of statements or
8215 * block / parens, boolean operators (&&, ||, //) and branch
8216 * constructs (or, and, if, until, unless, while, err, for).
8217 * Not a very solid hack... */
8218 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8219 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8220 "%s (...) interpreted as function",name);
8223 while (s < PL_bufend && isSPACE(*s))
8227 while (s < PL_bufend && isSPACE(*s))
8229 if (isIDFIRST_lazy_if(s,UTF)) {
8230 const char * const w = s;
8231 s += UTF ? UTF8SKIP(s) : 1;
8232 while (isWORDCHAR_lazy_if(s,UTF))
8233 s += UTF ? UTF8SKIP(s) : 1;
8234 while (s < PL_bufend && isSPACE(*s))
8238 if (keyword(w, s - w, 0))
8241 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8242 if (gv && GvCVu(gv))
8244 Perl_croak(aTHX_ "No comma allowed after %s", what);
8249 /* S_new_constant(): do any overload::constant lookup.
8251 Either returns sv, or mortalizes/frees sv and returns a new SV*.
8252 Best used as sv=new_constant(..., sv, ...).
8253 If s, pv are NULL, calls subroutine with one argument,
8254 and <type> is used with error messages only.
8255 <type> is assumed to be well formed UTF-8 */
8258 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8259 SV *sv, SV *pv, const char *type, STRLEN typelen)
8262 HV * table = GvHV(PL_hintgv); /* ^H */
8267 const char *why1 = "", *why2 = "", *why3 = "";
8269 PERL_ARGS_ASSERT_NEW_CONSTANT;
8270 /* We assume that this is true: */
8271 if (*key == 'c') { assert (strEQ(key, "charnames")); }
8274 /* charnames doesn't work well if there have been errors found */
8275 if (PL_error_count > 0 && *key == 'c')
8277 SvREFCNT_dec_NN(sv);
8278 return &PL_sv_undef;
8281 sv_2mortal(sv); /* Parent created it permanently */
8283 || ! (PL_hints & HINT_LOCALIZE_HH)
8284 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8289 /* Here haven't found what we're looking for. If it is charnames,
8290 * perhaps it needs to be loaded. Try doing that before giving up */
8292 Perl_load_module(aTHX_
8294 newSVpvs("_charnames"),
8295 /* version parameter; no need to specify it, as if
8296 * we get too early a version, will fail anyway,
8297 * not being able to find '_charnames' */
8302 assert(sp == PL_stack_sp);
8303 table = GvHV(PL_hintgv);
8305 && (PL_hints & HINT_LOCALIZE_HH)
8306 && (cvp = hv_fetch(table, key, keylen, FALSE))
8312 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8313 msg = Perl_form(aTHX_
8314 "Constant(%.*s) unknown",
8315 (int)(type ? typelen : len),
8321 why3 = "} is not defined";
8324 msg = Perl_form(aTHX_
8325 /* The +3 is for '\N{'; -4 for that, plus '}' */
8326 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
8330 msg = Perl_form(aTHX_ "Constant(%.*s): %s%s%s",
8331 (int)(type ? typelen : len),
8332 (type ? type: s), why1, why2, why3);
8335 yyerror_pv(msg, UTF ? SVf_UTF8 : 0);
8336 return SvREFCNT_inc_simple_NN(sv);
8341 pv = newSVpvn_flags(s, len, SVs_TEMP);
8343 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8345 typesv = &PL_sv_undef;
8347 PUSHSTACKi(PERLSI_OVERLOAD);
8359 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8363 /* Check the eval first */
8364 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
8366 const char * errstr;
8367 sv_catpvs(errsv, "Propagated");
8368 errstr = SvPV_const(errsv, errlen);
8369 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
8371 res = SvREFCNT_inc_simple_NN(sv);
8375 SvREFCNT_inc_simple_void_NN(res);
8384 why1 = "Call to &{$^H{";
8386 why3 = "}} did not return a defined value";
8388 (void)sv_2mortal(sv);
8395 PERL_STATIC_INLINE void
8396 S_parse_ident(pTHX_ char **s, char **d, char * const e, int allow_package, bool is_utf8) {
8397 PERL_ARGS_ASSERT_PARSE_IDENT;
8401 Perl_croak(aTHX_ "%s", ident_too_long);
8402 if (is_utf8 && isIDFIRST_utf8((U8*)*s)) {
8403 /* The UTF-8 case must come first, otherwise things
8404 * like c\N{COMBINING TILDE} would start failing, as the
8405 * isWORDCHAR_A case below would gobble the 'c' up.
8408 char *t = *s + UTF8SKIP(*s);
8409 while (isIDCONT_utf8((U8*)t))
8411 if (*d + (t - *s) > e)
8412 Perl_croak(aTHX_ "%s", ident_too_long);
8413 Copy(*s, *d, t - *s, char);
8417 else if ( isWORDCHAR_A(**s) ) {
8420 } while (isWORDCHAR_A(**s) && *d < e);
8422 else if (allow_package && **s == '\'' && isIDFIRST_lazy_if(*s+1,is_utf8)) {
8427 else if (allow_package && **s == ':' && (*s)[1] == ':'
8428 /* Disallow things like Foo::$bar. For the curious, this is
8429 * the code path that triggers the "Bad name after" warning
8430 * when looking for barewords.
8432 && (*s)[2] != '$') {
8442 /* Returns a NUL terminated string, with the length of the string written to
8446 S_scan_word(pTHX_ char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8449 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8450 bool is_utf8 = cBOOL(UTF);
8452 PERL_ARGS_ASSERT_SCAN_WORD;
8454 parse_ident(&s, &d, e, allow_package, is_utf8);
8461 S_scan_ident(pTHX_ char *s, char *dest, STRLEN destlen, I32 ck_uni)
8463 I32 herelines = PL_parser->herelines;
8464 SSize_t bracket = -1;
8467 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8468 bool is_utf8 = cBOOL(UTF);
8469 I32 orig_copline = 0, tmp_copline = 0;
8471 PERL_ARGS_ASSERT_SCAN_IDENT;
8476 while (isDIGIT(*s)) {
8478 Perl_croak(aTHX_ "%s", ident_too_long);
8483 parse_ident(&s, &d, e, 1, is_utf8);
8488 /* Either a digit variable, or parse_ident() found an identifier
8489 (anything valid as a bareword), so job done and return. */
8490 if (PL_lex_state != LEX_NORMAL)
8491 PL_lex_state = LEX_INTERPENDMAYBE;
8494 if (*s == '$' && s[1] &&
8495 (isIDFIRST_lazy_if(s+1,is_utf8)
8496 || isDIGIT_A((U8)s[1])
8499 || strnEQ(s+1,"::",2)) )
8501 /* Dereferencing a value in a scalar variable.
8502 The alternatives are different syntaxes for a scalar variable.
8503 Using ' as a leading package separator isn't allowed. :: is. */
8506 /* Handle the opening { of @{...}, &{...}, *{...}, %{...}, ${...} */
8508 bracket = s - SvPVX(PL_linestr);
8510 orig_copline = CopLINE(PL_curcop);
8511 if (s < PL_bufend && isSPACE(*s)) {
8516 /* Is the byte 'd' a legal single character identifier name? 'u' is true
8517 * iff Unicode semantics are to be used. The legal ones are any of:
8519 * b) ASCII punctuation
8520 * c) When not under Unicode rules, any upper Latin1 character
8521 * d) \c?, \c\, \c^, \c_, and \cA..\cZ, minus the ones that have traditionally
8522 * been matched by \s on ASCII platforms. That is: \c?, plus 1-32, minus
8524 #define VALID_LEN_ONE_IDENT(d, u) (isPUNCT_A((U8)(d)) \
8525 || isDIGIT_A((U8)(d)) \
8526 || (!(u) && !isASCII((U8)(d))) \
8527 || ((((U8)(d)) < 32) \
8528 && (((((U8)(d)) >= 14) \
8529 || (((U8)(d)) <= 8 && (d) != 0) \
8530 || (((U8)(d)) == 13)))) \
8531 || (((U8)(d)) == toCTRL('?')))
8533 && (isIDFIRST_lazy_if(s, is_utf8) || VALID_LEN_ONE_IDENT(*s, is_utf8)))
8535 if ( isCNTRL_A((U8)*s) ) {
8536 deprecate("literal control characters in variable names");
8540 const STRLEN skip = UTF8SKIP(s);
8543 for ( i = 0; i < skip; i++ )
8551 /* Convert $^F, ${^F} and the ^F of ${^FOO} to control characters */
8552 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8556 /* Warn about ambiguous code after unary operators if {...} notation isn't
8557 used. There's no difference in ambiguity; it's merely a heuristic
8558 about when not to warn. */
8559 else if (ck_uni && bracket == -1)
8561 if (bracket != -1) {
8562 /* If we were processing {...} notation then... */
8563 if (isIDFIRST_lazy_if(d,is_utf8)) {
8564 /* if it starts as a valid identifier, assume that it is one.
8565 (the later check for } being at the expected point will trap
8566 cases where this doesn't pan out.) */
8567 d += is_utf8 ? UTF8SKIP(d) : 1;
8568 parse_ident(&s, &d, e, 1, is_utf8);
8570 tmp_copline = CopLINE(PL_curcop);
8571 if (s < PL_bufend && isSPACE(*s)) {
8574 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8575 /* ${foo[0]} and ${foo{bar}} notation. */
8576 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8577 const char * const brack =
8579 ((*s == '[') ? "[...]" : "{...}");
8580 orig_copline = CopLINE(PL_curcop);
8581 CopLINE_set(PL_curcop, tmp_copline);
8582 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8583 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8584 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8585 funny, dest, brack, funny, dest, brack);
8586 CopLINE_set(PL_curcop, orig_copline);
8589 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8590 PL_lex_allbrackets++;
8594 /* Handle extended ${^Foo} variables
8595 * 1999-02-27 mjd-perl-patch@plover.com */
8596 else if (! isPRINT(*d) /* isCNTRL(d), plus all non-ASCII */
8600 while (isWORDCHAR(*s) && d < e) {
8604 Perl_croak(aTHX_ "%s", ident_too_long);
8609 tmp_copline = CopLINE(PL_curcop);
8610 if (s < PL_bufend && isSPACE(*s)) {
8614 /* Expect to find a closing } after consuming any trailing whitespace.
8618 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8619 PL_lex_state = LEX_INTERPEND;
8622 if (PL_lex_state == LEX_NORMAL) {
8623 if (ckWARN(WARN_AMBIGUOUS) &&
8624 (keyword(dest, d - dest, 0)
8625 || get_cvn_flags(dest, d - dest, is_utf8 ? SVf_UTF8 : 0)))
8627 SV *tmp = newSVpvn_flags( dest, d - dest,
8628 SVs_TEMP | (is_utf8 ? SVf_UTF8 : 0) );
8631 orig_copline = CopLINE(PL_curcop);
8632 CopLINE_set(PL_curcop, tmp_copline);
8633 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8634 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
8635 funny, SVfARG(tmp), funny, SVfARG(tmp));
8636 CopLINE_set(PL_curcop, orig_copline);
8641 /* Didn't find the closing } at the point we expected, so restore
8642 state such that the next thing to process is the opening { and */
8643 s = SvPVX(PL_linestr) + bracket; /* let the parser handle it */
8644 CopLINE_set(PL_curcop, orig_copline);
8645 PL_parser->herelines = herelines;
8649 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8650 PL_lex_state = LEX_INTERPEND;
8655 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8657 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8658 * the parse starting at 's', based on the subset that are valid in this
8659 * context input to this routine in 'valid_flags'. Advances s. Returns
8660 * TRUE if the input should be treated as a valid flag, so the next char
8661 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
8662 * first call on the current regex. This routine will set it to any
8663 * charset modifier found. The caller shouldn't change it. This way,
8664 * another charset modifier encountered in the parse can be detected as an
8665 * error, as we have decided to allow only one */
8668 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
8670 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
8671 if (isWORDCHAR_lazy_if(*s, UTF)) {
8672 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
8673 UTF ? SVf_UTF8 : 0);
8675 /* Pretend that it worked, so will continue processing before
8684 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8685 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8686 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8687 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8688 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8689 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8690 case LOCALE_PAT_MOD:
8692 goto multiple_charsets;
8694 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8697 case UNICODE_PAT_MOD:
8699 goto multiple_charsets;
8701 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8704 case ASCII_RESTRICT_PAT_MOD:
8706 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8710 /* Error if previous modifier wasn't an 'a', but if it was, see
8711 * if, and accept, a second occurrence (only) */
8713 || get_regex_charset(*pmfl)
8714 != REGEX_ASCII_RESTRICTED_CHARSET)
8716 goto multiple_charsets;
8718 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8722 case DEPENDS_PAT_MOD:
8724 goto multiple_charsets;
8726 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8735 if (*charset != c) {
8736 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8738 else if (c == 'a') {
8739 /* diag_listed_as: Regexp modifier "/%c" may appear a maximum of twice */
8740 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
8743 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8746 /* Pretend that it worked, so will continue processing before dieing */
8752 S_scan_pat(pTHX_ char *start, I32 type)
8756 const char * const valid_flags =
8757 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8758 char charset = '\0'; /* character set modifier */
8760 PERL_ARGS_ASSERT_SCAN_PAT;
8762 s = scan_str(start,TRUE,FALSE, (PL_in_eval & EVAL_RE_REPARSING), NULL);
8764 Perl_croak(aTHX_ "Search pattern not terminated");
8766 pm = (PMOP*)newPMOP(type, 0);
8767 if (PL_multi_open == '?') {
8768 /* This is the only point in the code that sets PMf_ONCE: */
8769 pm->op_pmflags |= PMf_ONCE;
8771 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8772 allows us to restrict the list needed by reset to just the ??
8774 assert(type != OP_TRANS);
8776 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8779 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8782 elements = mg->mg_len / sizeof(PMOP**);
8783 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8784 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8785 mg->mg_len = elements * sizeof(PMOP**);
8786 PmopSTASH_set(pm,PL_curstash);
8790 /* if qr/...(?{..}).../, then need to parse the pattern within a new
8791 * anon CV. False positives like qr/[(?{]/ are harmless */
8793 if (type == OP_QR) {
8795 char *e, *p = SvPV(PL_lex_stuff, len);
8797 for (; p < e; p++) {
8798 if (p[0] == '(' && p[1] == '?'
8799 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
8801 pm->op_pmflags |= PMf_HAS_CV;
8805 pm->op_pmflags |= PMf_IS_QR;
8808 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
8809 /* issue a warning if /c is specified,but /g is not */
8810 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8812 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8813 "Use of /c modifier is meaningless without /g" );
8816 PL_lex_op = (OP*)pm;
8817 pl_yylval.ival = OP_MATCH;
8822 S_scan_subst(pTHX_ char *start)
8829 char charset = '\0'; /* character set modifier */
8832 PERL_ARGS_ASSERT_SCAN_SUBST;
8834 pl_yylval.ival = OP_NULL;
8836 s = scan_str(start, TRUE, FALSE, FALSE, &t);
8839 Perl_croak(aTHX_ "Substitution pattern not terminated");
8843 first_start = PL_multi_start;
8844 first_line = CopLINE(PL_curcop);
8845 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8848 SvREFCNT_dec(PL_lex_stuff);
8849 PL_lex_stuff = NULL;
8851 Perl_croak(aTHX_ "Substitution replacement not terminated");
8853 PL_multi_start = first_start; /* so whole substitution is taken together */
8855 pm = (PMOP*)newPMOP(OP_SUBST, 0);
8859 if (*s == EXEC_PAT_MOD) {
8863 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
8869 if ((pm->op_pmflags & PMf_CONTINUE)) {
8870 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
8874 SV * const repl = newSVpvs("");
8877 pm->op_pmflags |= PMf_EVAL;
8880 sv_catpvs(repl, "eval ");
8882 sv_catpvs(repl, "do ");
8884 sv_catpvs(repl, "{");
8885 sv_catsv(repl, PL_sublex_info.repl);
8886 sv_catpvs(repl, "}");
8888 SvREFCNT_dec(PL_sublex_info.repl);
8889 PL_sublex_info.repl = repl;
8891 if (CopLINE(PL_curcop) != first_line) {
8892 sv_upgrade(PL_sublex_info.repl, SVt_PVNV);
8893 ((XPVNV*)SvANY(PL_sublex_info.repl))->xnv_u.xpad_cop_seq.xlow =
8894 CopLINE(PL_curcop) - first_line;
8895 CopLINE_set(PL_curcop, first_line);
8898 PL_lex_op = (OP*)pm;
8899 pl_yylval.ival = OP_SUBST;
8904 S_scan_trans(pTHX_ char *start)
8911 bool nondestruct = 0;
8914 PERL_ARGS_ASSERT_SCAN_TRANS;
8916 pl_yylval.ival = OP_NULL;
8918 s = scan_str(start,FALSE,FALSE,FALSE,&t);
8920 Perl_croak(aTHX_ "Transliteration pattern not terminated");
8924 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8927 SvREFCNT_dec(PL_lex_stuff);
8928 PL_lex_stuff = NULL;
8930 Perl_croak(aTHX_ "Transliteration replacement not terminated");
8933 complement = del = squash = 0;
8937 complement = OPpTRANS_COMPLEMENT;
8940 del = OPpTRANS_DELETE;
8943 squash = OPpTRANS_SQUASH;
8955 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
8956 o->op_private &= ~OPpTRANS_ALL;
8957 o->op_private |= del|squash|complement|
8958 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
8959 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
8962 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
8969 Takes a pointer to the first < in <<FOO.
8970 Returns a pointer to the byte following <<FOO.
8972 This function scans a heredoc, which involves different methods
8973 depending on whether we are in a string eval, quoted construct, etc.
8974 This is because PL_linestr could containing a single line of input, or
8975 a whole string being evalled, or the contents of the current quote-
8978 The two basic methods are:
8979 - Steal lines from the input stream
8980 - Scan the heredoc in PL_linestr and remove it therefrom
8982 In a file scope or filtered eval, the first method is used; in a
8983 string eval, the second.
8985 In a quote-like operator, we have to choose between the two,
8986 depending on where we can find a newline. We peek into outer lex-
8987 ing scopes until we find one with a newline in it. If we reach the
8988 outermost lexing scope and it is a file, we use the stream method.
8989 Otherwise it is treated as an eval.
8993 S_scan_heredoc(pTHX_ char *s)
8995 I32 op_type = OP_SCALAR;
9002 const bool infile = PL_rsfp || PL_parser->filtered;
9003 const line_t origline = CopLINE(PL_curcop);
9004 LEXSHARED *shared = PL_parser->lex_shared;
9006 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9009 d = PL_tokenbuf + 1;
9010 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9011 *PL_tokenbuf = '\n';
9013 while (SPACE_OR_TAB(*peek))
9015 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9018 s = delimcpy(d, e, s, PL_bufend, term, &len);
9020 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9026 /* <<\FOO is equivalent to <<'FOO' */
9030 if (!isWORDCHAR_lazy_if(s,UTF))
9031 deprecate("bare << to mean <<\"\"");
9032 for (; isWORDCHAR_lazy_if(s,UTF); s++) {
9037 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9038 Perl_croak(aTHX_ "Delimiter for here document is too long");
9041 len = d - PL_tokenbuf;
9043 #ifndef PERL_STRICT_CR
9044 d = strchr(s, '\r');
9046 char * const olds = s;
9048 while (s < PL_bufend) {
9054 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9063 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9068 tmpstr = newSV_type(SVt_PVIV);
9072 SvIV_set(tmpstr, -1);
9074 else if (term == '`') {
9075 op_type = OP_BACKTICK;
9076 SvIV_set(tmpstr, '\\');
9079 PL_multi_start = origline + 1 + PL_parser->herelines;
9080 PL_multi_open = PL_multi_close = '<';
9081 /* inside a string eval or quote-like operator */
9082 if (!infile || PL_lex_inwhat) {
9085 char * const olds = s;
9086 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9087 /* These two fields are not set until an inner lexing scope is
9088 entered. But we need them set here. */
9089 shared->ls_bufptr = s;
9090 shared->ls_linestr = PL_linestr;
9092 /* Look for a newline. If the current buffer does not have one,
9093 peek into the line buffer of the parent lexing scope, going
9094 up as many levels as necessary to find one with a newline
9097 while (!(s = (char *)memchr(
9098 (void *)shared->ls_bufptr, '\n',
9099 SvEND(shared->ls_linestr)-shared->ls_bufptr
9101 shared = shared->ls_prev;
9102 /* shared is only null if we have gone beyond the outermost
9103 lexing scope. In a file, we will have broken out of the
9104 loop in the previous iteration. In an eval, the string buf-
9105 fer ends with "\n;", so the while condition above will have
9106 evaluated to false. So shared can never be null. */
9108 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9109 most lexing scope. In a file, shared->ls_linestr at that
9110 level is just one line, so there is no body to steal. */
9111 if (infile && !shared->ls_prev) {
9117 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9120 linestr = shared->ls_linestr;
9121 bufend = SvEND(linestr);
9123 while (s < bufend - len + 1 &&
9124 memNE(s,PL_tokenbuf,len) ) {
9126 ++PL_parser->herelines;
9128 if (s >= bufend - len + 1) {
9131 sv_setpvn(tmpstr,d+1,s-d);
9133 /* the preceding stmt passes a newline */
9134 PL_parser->herelines++;
9136 /* s now points to the newline after the heredoc terminator.
9137 d points to the newline before the body of the heredoc.
9140 /* We are going to modify linestr in place here, so set
9141 aside copies of the string if necessary for re-evals or
9143 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9144 check shared->re_eval_str. */
9145 if (shared->re_eval_start || shared->re_eval_str) {
9146 /* Set aside the rest of the regexp */
9147 if (!shared->re_eval_str)
9148 shared->re_eval_str =
9149 newSVpvn(shared->re_eval_start,
9150 bufend - shared->re_eval_start);
9151 shared->re_eval_start -= s-d;
9153 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9154 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9155 cx->blk_eval.cur_text == linestr)
9157 cx->blk_eval.cur_text = newSVsv(linestr);
9158 SvSCREAM_on(cx->blk_eval.cur_text);
9160 /* Copy everything from s onwards back to d. */
9161 Move(s,d,bufend-s + 1,char);
9162 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9163 /* Setting PL_bufend only applies when we have not dug deeper
9164 into other scopes, because sublex_done sets PL_bufend to
9165 SvEND(PL_linestr). */
9166 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9173 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9174 term = PL_tokenbuf[1];
9176 linestr_save = PL_linestr; /* must restore this afterwards */
9177 d = s; /* and this */
9178 PL_linestr = newSVpvs("");
9179 PL_bufend = SvPVX(PL_linestr);
9181 PL_bufptr = PL_bufend;
9182 CopLINE_set(PL_curcop,
9183 origline + 1 + PL_parser->herelines);
9184 if (!lex_next_chunk(LEX_NO_TERM)
9185 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9186 SvREFCNT_dec(linestr_save);
9189 CopLINE_set(PL_curcop, origline);
9190 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9191 s = lex_grow_linestr(SvLEN(PL_linestr) + 3);
9192 /* ^That should be enough to avoid this needing to grow: */
9193 sv_catpvs(PL_linestr, "\n\0");
9194 assert(s == SvPVX(PL_linestr));
9195 PL_bufend = SvEND(PL_linestr);
9198 PL_parser->herelines++;
9199 PL_last_lop = PL_last_uni = NULL;
9200 #ifndef PERL_STRICT_CR
9201 if (PL_bufend - PL_linestart >= 2) {
9202 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9203 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9205 PL_bufend[-2] = '\n';
9207 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9209 else if (PL_bufend[-1] == '\r')
9210 PL_bufend[-1] = '\n';
9212 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9213 PL_bufend[-1] = '\n';
9215 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
9216 SvREFCNT_dec(PL_linestr);
9217 PL_linestr = linestr_save;
9218 PL_linestart = SvPVX(linestr_save);
9219 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9224 sv_catsv(tmpstr,PL_linestr);
9228 PL_multi_end = origline + PL_parser->herelines;
9229 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9230 SvPV_shrink_to_cur(tmpstr);
9233 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9235 else if (PL_encoding)
9236 sv_recode_to_utf8(tmpstr, PL_encoding);
9238 PL_lex_stuff = tmpstr;
9239 pl_yylval.ival = op_type;
9243 SvREFCNT_dec(tmpstr);
9244 CopLINE_set(PL_curcop, origline);
9245 missingterm(PL_tokenbuf + 1);
9249 takes: current position in input buffer
9250 returns: new position in input buffer
9251 side-effects: pl_yylval and lex_op are set.
9256 <FH> read from filehandle
9257 <pkg::FH> read from package qualified filehandle
9258 <pkg'FH> read from package qualified filehandle
9259 <$fh> read from filehandle in $fh
9265 S_scan_inputsymbol(pTHX_ char *start)
9267 char *s = start; /* current position in buffer */
9270 char *d = PL_tokenbuf; /* start of temp holding space */
9271 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9273 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9275 end = strchr(s, '\n');
9278 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9280 /* die if we didn't have space for the contents of the <>,
9281 or if it didn't end, or if we see a newline
9284 if (len >= (I32)sizeof PL_tokenbuf)
9285 Perl_croak(aTHX_ "Excessively long <> operator");
9287 Perl_croak(aTHX_ "Unterminated <> operator");
9292 Remember, only scalar variables are interpreted as filehandles by
9293 this code. Anything more complex (e.g., <$fh{$num}>) will be
9294 treated as a glob() call.
9295 This code makes use of the fact that except for the $ at the front,
9296 a scalar variable and a filehandle look the same.
9298 if (*d == '$' && d[1]) d++;
9300 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9301 while (*d && (isWORDCHAR_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9302 d += UTF ? UTF8SKIP(d) : 1;
9304 /* If we've tried to read what we allow filehandles to look like, and
9305 there's still text left, then it must be a glob() and not a getline.
9306 Use scan_str to pull out the stuff between the <> and treat it
9307 as nothing more than a string.
9310 if (d - PL_tokenbuf != len) {
9311 pl_yylval.ival = OP_GLOB;
9312 s = scan_str(start,FALSE,FALSE,FALSE,NULL);
9314 Perl_croak(aTHX_ "Glob not terminated");
9318 bool readline_overriden = FALSE;
9320 /* we're in a filehandle read situation */
9323 /* turn <> into <ARGV> */
9325 Copy("ARGV",d,5,char);
9327 /* Check whether readline() is overriden */
9328 if ((gv_readline = gv_override("readline",8)))
9329 readline_overriden = TRUE;
9331 /* if <$fh>, create the ops to turn the variable into a
9335 /* try to find it in the pad for this block, otherwise find
9336 add symbol table ops
9338 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9339 if (tmp != NOT_IN_PAD) {
9340 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9341 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9342 HEK * const stashname = HvNAME_HEK(stash);
9343 SV * const sym = sv_2mortal(newSVhek(stashname));
9344 sv_catpvs(sym, "::");
9350 OP * const o = newOP(OP_PADSV, 0);
9352 PL_lex_op = readline_overriden
9353 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9354 op_append_elem(OP_LIST, o,
9355 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9356 : (OP*)newUNOP(OP_READLINE, 0, o);
9365 ? (GV_ADDMULTI | GV_ADDINEVAL)
9366 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
9368 PL_lex_op = readline_overriden
9369 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9370 op_append_elem(OP_LIST,
9371 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9372 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9373 : (OP*)newUNOP(OP_READLINE, 0,
9374 newUNOP(OP_RV2SV, 0,
9375 newGVOP(OP_GV, 0, gv)));
9377 if (!readline_overriden)
9378 PL_lex_op->op_flags |= OPf_SPECIAL;
9379 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9380 pl_yylval.ival = OP_NULL;
9383 /* If it's none of the above, it must be a literal filehandle
9384 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9386 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9387 PL_lex_op = readline_overriden
9388 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9389 op_append_elem(OP_LIST,
9390 newGVOP(OP_GV, 0, gv),
9391 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9392 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9393 pl_yylval.ival = OP_NULL;
9403 start position in buffer
9404 keep_bracketed_quoted preserve \ quoting of embedded delimiters, but
9405 only if they are of the open/close form
9406 keep_delims preserve the delimiters around the string
9407 re_reparse compiling a run-time /(?{})/:
9408 collapse // to /, and skip encoding src
9409 delimp if non-null, this is set to the position of
9410 the closing delimiter, or just after it if
9411 the closing and opening delimiters differ
9412 (i.e., the opening delimiter of a substitu-
9414 returns: position to continue reading from buffer
9415 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9416 updates the read buffer.
9418 This subroutine pulls a string out of the input. It is called for:
9419 q single quotes q(literal text)
9420 ' single quotes 'literal text'
9421 qq double quotes qq(interpolate $here please)
9422 " double quotes "interpolate $here please"
9423 qx backticks qx(/bin/ls -l)
9424 ` backticks `/bin/ls -l`
9425 qw quote words @EXPORT_OK = qw( func() $spam )
9426 m// regexp match m/this/
9427 s/// regexp substitute s/this/that/
9428 tr/// string transliterate tr/this/that/
9429 y/// string transliterate y/this/that/
9430 ($*@) sub prototypes sub foo ($)
9431 (stuff) sub attr parameters sub foo : attr(stuff)
9432 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9434 In most of these cases (all but <>, patterns and transliterate)
9435 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9436 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9437 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9440 It skips whitespace before the string starts, and treats the first
9441 character as the delimiter. If the delimiter is one of ([{< then
9442 the corresponding "close" character )]}> is used as the closing
9443 delimiter. It allows quoting of delimiters, and if the string has
9444 balanced delimiters ([{<>}]) it allows nesting.
9446 On success, the SV with the resulting string is put into lex_stuff or,
9447 if that is already non-NULL, into lex_repl. The second case occurs only
9448 when parsing the RHS of the special constructs s/// and tr/// (y///).
9449 For convenience, the terminating delimiter character is stuffed into
9454 S_scan_str(pTHX_ char *start, int keep_bracketed_quoted, int keep_delims, int re_reparse,
9458 SV *sv; /* scalar value: string */
9459 const char *tmps; /* temp string, used for delimiter matching */
9460 char *s = start; /* current position in the buffer */
9461 char term; /* terminating character */
9462 char *to; /* current position in the sv's data */
9463 I32 brackets = 1; /* bracket nesting level */
9464 bool has_utf8 = FALSE; /* is there any utf8 content? */
9465 I32 termcode; /* terminating char. code */
9466 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9467 STRLEN termlen; /* length of terminating string */
9468 int last_off = 0; /* last position for nesting bracket */
9471 PERL_ARGS_ASSERT_SCAN_STR;
9473 /* skip space before the delimiter */
9478 /* mark where we are, in case we need to report errors */
9481 /* after skipping whitespace, the next character is the terminator */
9484 termcode = termstr[0] = term;
9488 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
9489 Copy(s, termstr, termlen, U8);
9490 if (!UTF8_IS_INVARIANT(term))
9494 /* mark where we are */
9495 PL_multi_start = CopLINE(PL_curcop);
9496 PL_multi_open = term;
9497 herelines = PL_parser->herelines;
9499 /* find corresponding closing delimiter */
9500 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9501 termcode = termstr[0] = term = tmps[5];
9503 PL_multi_close = term;
9505 if (PL_multi_open == PL_multi_close) {
9506 keep_bracketed_quoted = FALSE;
9509 /* create a new SV to hold the contents. 79 is the SV's initial length.
9510 What a random number. */
9511 sv = newSV_type(SVt_PVIV);
9513 SvIV_set(sv, termcode);
9514 (void)SvPOK_only(sv); /* validate pointer */
9516 /* move past delimiter and try to read a complete string */
9518 sv_catpvn(sv, s, termlen);
9521 if (PL_encoding && !UTF && !re_reparse) {
9525 int offset = s - SvPVX_const(PL_linestr);
9526 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9527 &offset, (char*)termstr, termlen);
9531 if (SvIsCOW(PL_linestr)) {
9532 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos;
9533 STRLEN oldoldbufptr_pos, linestart_pos, last_uni_pos;
9534 STRLEN last_lop_pos, re_eval_start_pos, s_pos;
9535 char *buf = SvPVX(PL_linestr);
9536 bufend_pos = PL_parser->bufend - buf;
9537 bufptr_pos = PL_parser->bufptr - buf;
9538 oldbufptr_pos = PL_parser->oldbufptr - buf;
9539 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
9540 linestart_pos = PL_parser->linestart - buf;
9541 last_uni_pos = PL_parser->last_uni
9542 ? PL_parser->last_uni - buf
9544 last_lop_pos = PL_parser->last_lop
9545 ? PL_parser->last_lop - buf
9548 PL_parser->lex_shared->re_eval_start ?
9549 PL_parser->lex_shared->re_eval_start - buf : 0;
9552 sv_force_normal(PL_linestr);
9554 buf = SvPVX(PL_linestr);
9555 PL_parser->bufend = buf + bufend_pos;
9556 PL_parser->bufptr = buf + bufptr_pos;
9557 PL_parser->oldbufptr = buf + oldbufptr_pos;
9558 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
9559 PL_parser->linestart = buf + linestart_pos;
9560 if (PL_parser->last_uni)
9561 PL_parser->last_uni = buf + last_uni_pos;
9562 if (PL_parser->last_lop)
9563 PL_parser->last_lop = buf + last_lop_pos;
9564 if (PL_parser->lex_shared->re_eval_start)
9565 PL_parser->lex_shared->re_eval_start =
9566 buf + re_eval_start_pos;
9569 ns = SvPVX_const(PL_linestr) + offset;
9570 svlast = SvEND(sv) - 1;
9572 for (; s < ns; s++) {
9573 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9574 COPLINE_INC_WITH_HERELINES;
9577 goto read_more_line;
9579 /* handle quoted delimiters */
9580 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9582 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9584 if ((svlast-1 - t) % 2) {
9585 if (!keep_bracketed_quoted) {
9588 SvCUR_set(sv, SvCUR(sv) - 1);
9593 if (PL_multi_open == PL_multi_close) {
9599 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9600 /* At here, all closes are "was quoted" one,
9601 so we don't check PL_multi_close. */
9603 if (!keep_bracketed_quoted && *(t+1) == PL_multi_open)
9608 else if (*t == PL_multi_open)
9616 SvCUR_set(sv, w - SvPVX_const(sv));
9618 last_off = w - SvPVX(sv);
9619 if (--brackets <= 0)
9625 SvCUR_set(sv, SvCUR(sv) - 1);
9631 /* extend sv if need be */
9632 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9633 /* set 'to' to the next character in the sv's string */
9634 to = SvPVX(sv)+SvCUR(sv);
9636 /* if open delimiter is the close delimiter read unbridle */
9637 if (PL_multi_open == PL_multi_close) {
9638 for (; s < PL_bufend; s++,to++) {
9639 /* embedded newlines increment the current line number */
9640 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9641 COPLINE_INC_WITH_HERELINES;
9642 /* handle quoted delimiters */
9643 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9644 if (!keep_bracketed_quoted
9646 || (re_reparse && s[1] == '\\'))
9649 else /* any other quotes are simply copied straight through */
9652 /* terminate when run out of buffer (the for() condition), or
9653 have found the terminator */
9654 else if (*s == term) {
9657 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9660 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9666 /* if the terminator isn't the same as the start character (e.g.,
9667 matched brackets), we have to allow more in the quoting, and
9668 be prepared for nested brackets.
9671 /* read until we run out of string, or we find the terminator */
9672 for (; s < PL_bufend; s++,to++) {
9673 /* embedded newlines increment the line count */
9674 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9675 COPLINE_INC_WITH_HERELINES;
9676 /* backslashes can escape the open or closing characters */
9677 if (*s == '\\' && s+1 < PL_bufend) {
9678 if (!keep_bracketed_quoted &&
9679 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9686 /* allow nested opens and closes */
9687 else if (*s == PL_multi_close && --brackets <= 0)
9689 else if (*s == PL_multi_open)
9691 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9696 /* terminate the copied string and update the sv's end-of-string */
9698 SvCUR_set(sv, to - SvPVX_const(sv));
9701 * this next chunk reads more into the buffer if we're not done yet
9705 break; /* handle case where we are done yet :-) */
9707 #ifndef PERL_STRICT_CR
9708 if (to - SvPVX_const(sv) >= 2) {
9709 if ((to[-2] == '\r' && to[-1] == '\n') ||
9710 (to[-2] == '\n' && to[-1] == '\r'))
9714 SvCUR_set(sv, to - SvPVX_const(sv));
9716 else if (to[-1] == '\r')
9719 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9724 /* if we're out of file, or a read fails, bail and reset the current
9725 line marker so we can report where the unterminated string began
9727 COPLINE_INC_WITH_HERELINES;
9728 PL_bufptr = PL_bufend;
9729 if (!lex_next_chunk(0)) {
9731 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9737 /* at this point, we have successfully read the delimited string */
9739 if (!PL_encoding || UTF || re_reparse) {
9742 sv_catpvn(sv, s, termlen);
9745 if (has_utf8 || (PL_encoding && !re_reparse))
9748 PL_multi_end = CopLINE(PL_curcop);
9749 CopLINE_set(PL_curcop, PL_multi_start);
9750 PL_parser->herelines = herelines;
9752 /* if we allocated too much space, give some back */
9753 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9754 SvLEN_set(sv, SvCUR(sv) + 1);
9755 SvPV_renew(sv, SvLEN(sv));
9758 /* decide whether this is the first or second quoted string we've read
9763 PL_sublex_info.repl = sv;
9766 if (delimp) *delimp = PL_multi_open == PL_multi_close ? s-termlen : s;
9772 takes: pointer to position in buffer
9773 returns: pointer to new position in buffer
9774 side-effects: builds ops for the constant in pl_yylval.op
9776 Read a number in any of the formats that Perl accepts:
9778 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
9779 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
9782 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
9784 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
9787 If it reads a number without a decimal point or an exponent, it will
9788 try converting the number to an integer and see if it can do so
9789 without loss of precision.
9793 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
9795 const char *s = start; /* current position in buffer */
9796 char *d; /* destination in temp buffer */
9797 char *e; /* end of temp buffer */
9798 NV nv; /* number read, as a double */
9799 SV *sv = NULL; /* place to put the converted number */
9800 bool floatit; /* boolean: int or float? */
9801 const char *lastub = NULL; /* position of last underbar */
9802 static const char* const number_too_long = "Number too long";
9804 PERL_ARGS_ASSERT_SCAN_NUM;
9806 /* We use the first character to decide what type of number this is */
9810 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
9812 /* if it starts with a 0, it could be an octal number, a decimal in
9813 0.13 disguise, or a hexadecimal number, or a binary number. */
9817 u holds the "number so far"
9818 shift the power of 2 of the base
9819 (hex == 4, octal == 3, binary == 1)
9820 overflowed was the number more than we can hold?
9822 Shift is used when we add a digit. It also serves as an "are
9823 we in octal/hex/binary?" indicator to disallow hex characters
9829 bool overflowed = FALSE;
9830 bool just_zero = TRUE; /* just plain 0 or binary number? */
9831 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
9832 static const char* const bases[5] =
9833 { "", "binary", "", "octal", "hexadecimal" };
9834 static const char* const Bases[5] =
9835 { "", "Binary", "", "Octal", "Hexadecimal" };
9836 static const char* const maxima[5] =
9838 "0b11111111111111111111111111111111",
9842 const char *base, *Base, *max;
9845 if (s[1] == 'x' || s[1] == 'X') {
9849 } else if (s[1] == 'b' || s[1] == 'B') {
9854 /* check for a decimal in disguise */
9855 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
9857 /* so it must be octal */
9864 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
9865 "Misplaced _ in number");
9869 base = bases[shift];
9870 Base = Bases[shift];
9871 max = maxima[shift];
9873 /* read the rest of the number */
9875 /* x is used in the overflow test,
9876 b is the digit we're adding on. */
9881 /* if we don't mention it, we're done */
9885 /* _ are ignored -- but warned about if consecutive */
9887 if (lastub && s == lastub + 1)
9888 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
9889 "Misplaced _ in number");
9893 /* 8 and 9 are not octal */
9896 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
9900 case '2': case '3': case '4':
9901 case '5': case '6': case '7':
9903 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
9907 b = *s++ & 15; /* ASCII digit -> value of digit */
9911 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
9912 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
9913 /* make sure they said 0x */
9918 /* Prepare to put the digit we have onto the end
9919 of the number so far. We check for overflows.
9925 x = u << shift; /* make room for the digit */
9927 if ((x >> shift) != u
9928 && !(PL_hints & HINT_NEW_BINARY)) {
9931 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
9932 "Integer overflow in %s number",
9935 u = x | b; /* add the digit to the end */
9938 n *= nvshift[shift];
9939 /* If an NV has not enough bits in its
9940 * mantissa to represent an UV this summing of
9941 * small low-order numbers is a waste of time
9942 * (because the NV cannot preserve the
9943 * low-order bits anyway): we could just
9944 * remember when did we overflow and in the
9945 * end just multiply n by the right
9953 /* if we get here, we had success: make a scalar value from
9958 /* final misplaced underbar check */
9960 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
9964 if (n > 4294967295.0)
9965 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
9966 "%s number > %s non-portable",
9973 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
9974 "%s number > %s non-portable",
9979 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
9980 sv = new_constant(start, s - start, "integer",
9982 else if (PL_hints & HINT_NEW_BINARY)
9983 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
9988 handle decimal numbers.
9989 we're also sent here when we read a 0 as the first digit
9991 case '1': case '2': case '3': case '4': case '5':
9992 case '6': case '7': case '8': case '9': case '.':
9995 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
9998 /* read next group of digits and _ and copy into d */
9999 while (isDIGIT(*s) || *s == '_') {
10000 /* skip underscores, checking for misplaced ones
10004 if (lastub && s == lastub + 1)
10005 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10006 "Misplaced _ in number");
10010 /* check for end of fixed-length buffer */
10012 Perl_croak(aTHX_ "%s", number_too_long);
10013 /* if we're ok, copy the character */
10018 /* final misplaced underbar check */
10019 if (lastub && s == lastub + 1) {
10020 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10023 /* read a decimal portion if there is one. avoid
10024 3..5 being interpreted as the number 3. followed
10027 if (*s == '.' && s[1] != '.') {
10032 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10033 "Misplaced _ in number");
10037 /* copy, ignoring underbars, until we run out of digits.
10039 for (; isDIGIT(*s) || *s == '_'; s++) {
10040 /* fixed length buffer check */
10042 Perl_croak(aTHX_ "%s", number_too_long);
10044 if (lastub && s == lastub + 1)
10045 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10046 "Misplaced _ in number");
10052 /* fractional part ending in underbar? */
10053 if (s[-1] == '_') {
10054 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10055 "Misplaced _ in number");
10057 if (*s == '.' && isDIGIT(s[1])) {
10058 /* oops, it's really a v-string, but without the "v" */
10064 /* read exponent part, if present */
10065 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10069 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10070 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10072 /* stray preinitial _ */
10074 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10075 "Misplaced _ in number");
10079 /* allow positive or negative exponent */
10080 if (*s == '+' || *s == '-')
10083 /* stray initial _ */
10085 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10086 "Misplaced _ in number");
10090 /* read digits of exponent */
10091 while (isDIGIT(*s) || *s == '_') {
10094 Perl_croak(aTHX_ "%s", number_too_long);
10098 if (((lastub && s == lastub + 1) ||
10099 (!isDIGIT(s[1]) && s[1] != '_')))
10100 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10101 "Misplaced _ in number");
10109 We try to do an integer conversion first if no characters
10110 indicating "float" have been found.
10115 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10117 if (flags == IS_NUMBER_IN_UV) {
10119 sv = newSViv(uv); /* Prefer IVs over UVs. */
10122 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10123 if (uv <= (UV) IV_MIN)
10124 sv = newSViv(-(IV)uv);
10131 STORE_NUMERIC_LOCAL_SET_STANDARD();
10132 /* terminate the string */
10134 nv = Atof(PL_tokenbuf);
10135 RESTORE_NUMERIC_LOCAL();
10140 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10141 const char *const key = floatit ? "float" : "integer";
10142 const STRLEN keylen = floatit ? 5 : 7;
10143 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10144 key, keylen, sv, NULL, NULL, 0);
10148 /* if it starts with a v, it could be a v-string */
10151 sv = newSV(5); /* preallocate storage space */
10152 ENTER_with_name("scan_vstring");
10154 s = scan_vstring(s, PL_bufend, sv);
10155 SvREFCNT_inc_simple_void_NN(sv);
10156 LEAVE_with_name("scan_vstring");
10160 /* make the op for the constant and return */
10163 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10165 lvalp->opval = NULL;
10171 S_scan_formline(pTHX_ char *s)
10175 SV * const stuff = newSVpvs("");
10176 bool needargs = FALSE;
10177 bool eofmt = FALSE;
10179 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10181 while (!needargs) {
10184 #ifdef PERL_STRICT_CR
10185 while (SPACE_OR_TAB(*t))
10188 while (SPACE_OR_TAB(*t) || *t == '\r')
10191 if (*t == '\n' || t == PL_bufend) {
10196 eol = (char *) memchr(s,'\n',PL_bufend-s);
10200 for (t = s; t < eol; t++) {
10201 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10203 goto enough; /* ~~ must be first line in formline */
10205 if (*t == '@' || *t == '^')
10209 sv_catpvn(stuff, s, eol-s);
10210 #ifndef PERL_STRICT_CR
10211 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10212 char *end = SvPVX(stuff) + SvCUR(stuff);
10215 SvCUR_set(stuff, SvCUR(stuff) - 1);
10223 if ((PL_rsfp || PL_parser->filtered)
10224 && PL_parser->form_lex_state == LEX_NORMAL) {
10226 PL_bufptr = PL_bufend;
10227 COPLINE_INC_WITH_HERELINES;
10228 got_some = lex_next_chunk(0);
10229 CopLINE_dec(PL_curcop);
10237 if (!SvCUR(stuff) || needargs)
10238 PL_lex_state = PL_parser->form_lex_state;
10239 if (SvCUR(stuff)) {
10240 PL_expect = XSTATE;
10242 const char *s2 = s;
10243 while (*s2 == '\r' || *s2 == ' ' || *s2 == '\t' || *s2 == '\f'
10247 PL_expect = XTERMBLOCK;
10248 NEXTVAL_NEXTTOKE.ival = 0;
10251 NEXTVAL_NEXTTOKE.ival = 0;
10252 force_next(FORMLBRACK);
10255 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10257 else if (PL_encoding)
10258 sv_recode_to_utf8(stuff, PL_encoding);
10260 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10264 SvREFCNT_dec(stuff);
10266 PL_lex_formbrack = 0;
10272 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10274 const I32 oldsavestack_ix = PL_savestack_ix;
10275 CV* const outsidecv = PL_compcv;
10277 SAVEI32(PL_subline);
10278 save_item(PL_subname);
10279 SAVESPTR(PL_compcv);
10281 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10282 CvFLAGS(PL_compcv) |= flags;
10284 PL_subline = CopLINE(PL_curcop);
10285 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10286 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10287 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10288 if (outsidecv && CvPADLIST(outsidecv))
10289 CvPADLIST(PL_compcv)->xpadl_outid =
10290 PadlistNAMES(CvPADLIST(outsidecv));
10292 return oldsavestack_ix;
10296 S_yywarn(pTHX_ const char *const s, U32 flags)
10298 PERL_ARGS_ASSERT_YYWARN;
10300 PL_in_eval |= EVAL_WARNONLY;
10301 yyerror_pv(s, flags);
10302 PL_in_eval &= ~EVAL_WARNONLY;
10307 Perl_yyerror(pTHX_ const char *const s)
10309 PERL_ARGS_ASSERT_YYERROR;
10310 return yyerror_pvn(s, strlen(s), 0);
10314 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
10316 PERL_ARGS_ASSERT_YYERROR_PV;
10317 return yyerror_pvn(s, strlen(s), flags);
10321 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
10323 const char *context = NULL;
10326 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
10327 int yychar = PL_parser->yychar;
10329 PERL_ARGS_ASSERT_YYERROR_PVN;
10331 if (!yychar || (yychar == ';' && !PL_rsfp))
10332 sv_catpvs(where_sv, "at EOF");
10333 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10334 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10335 PL_oldbufptr != PL_bufptr) {
10338 The code below is removed for NetWare because it abends/crashes on NetWare
10339 when the script has error such as not having the closing quotes like:
10340 if ($var eq "value)
10341 Checking of white spaces is anyway done in NetWare code.
10344 while (isSPACE(*PL_oldoldbufptr))
10347 context = PL_oldoldbufptr;
10348 contlen = PL_bufptr - PL_oldoldbufptr;
10350 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10351 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10354 The code below is removed for NetWare because it abends/crashes on NetWare
10355 when the script has error such as not having the closing quotes like:
10356 if ($var eq "value)
10357 Checking of white spaces is anyway done in NetWare code.
10360 while (isSPACE(*PL_oldbufptr))
10363 context = PL_oldbufptr;
10364 contlen = PL_bufptr - PL_oldbufptr;
10366 else if (yychar > 255)
10367 sv_catpvs(where_sv, "next token ???");
10368 else if (yychar == -2) { /* YYEMPTY */
10369 if (PL_lex_state == LEX_NORMAL ||
10370 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10371 sv_catpvs(where_sv, "at end of line");
10372 else if (PL_lex_inpat)
10373 sv_catpvs(where_sv, "within pattern");
10375 sv_catpvs(where_sv, "within string");
10378 sv_catpvs(where_sv, "next char ");
10380 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10381 else if (isPRINT_LC(yychar)) {
10382 const char string = yychar;
10383 sv_catpvn(where_sv, &string, 1);
10386 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10388 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
10389 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10390 OutCopFILE(PL_curcop),
10391 (IV)(PL_parser->preambling == NOLINE
10392 ? CopLINE(PL_curcop)
10393 : PL_parser->preambling));
10395 Perl_sv_catpvf(aTHX_ msg, "near \"%"UTF8f"\"\n",
10396 UTF8fARG(UTF, contlen, context));
10398 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
10399 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10400 Perl_sv_catpvf(aTHX_ msg,
10401 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10402 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10405 if (PL_in_eval & EVAL_WARNONLY) {
10406 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10410 if (PL_error_count >= 10) {
10412 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
10413 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10414 SVfARG(errsv), OutCopFILE(PL_curcop));
10416 Perl_croak(aTHX_ "%s has too many errors.\n",
10417 OutCopFILE(PL_curcop));
10420 PL_in_my_stash = NULL;
10425 S_swallow_bom(pTHX_ U8 *s)
10427 const STRLEN slen = SvCUR(PL_linestr);
10429 PERL_ARGS_ASSERT_SWALLOW_BOM;
10433 if (s[1] == 0xFE) {
10434 /* UTF-16 little-endian? (or UTF-32LE?) */
10435 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10436 /* diag_listed_as: Unsupported script encoding %s */
10437 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10438 #ifndef PERL_NO_UTF16_FILTER
10439 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10441 if (PL_bufend > (char*)s) {
10442 s = add_utf16_textfilter(s, TRUE);
10445 /* diag_listed_as: Unsupported script encoding %s */
10446 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10451 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10452 #ifndef PERL_NO_UTF16_FILTER
10453 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10455 if (PL_bufend > (char *)s) {
10456 s = add_utf16_textfilter(s, FALSE);
10459 /* diag_listed_as: Unsupported script encoding %s */
10460 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10464 case BOM_UTF8_FIRST_BYTE: {
10465 const STRLEN len = sizeof(BOM_UTF8_TAIL) - 1; /* Exclude trailing NUL */
10466 if (slen > len && memEQ(s+1, BOM_UTF8_TAIL, len)) {
10467 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10468 s += len + 1; /* UTF-8 */
10475 if (s[2] == 0xFE && s[3] == 0xFF) {
10476 /* UTF-32 big-endian */
10477 /* diag_listed_as: Unsupported script encoding %s */
10478 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10481 else if (s[2] == 0 && s[3] != 0) {
10484 * are a good indicator of UTF-16BE. */
10485 #ifndef PERL_NO_UTF16_FILTER
10486 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10487 s = add_utf16_textfilter(s, FALSE);
10489 /* diag_listed_as: Unsupported script encoding %s */
10490 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10497 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10500 * are a good indicator of UTF-16LE. */
10501 #ifndef PERL_NO_UTF16_FILTER
10502 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10503 s = add_utf16_textfilter(s, TRUE);
10505 /* diag_listed_as: Unsupported script encoding %s */
10506 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10514 #ifndef PERL_NO_UTF16_FILTER
10516 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10518 SV *const filter = FILTER_DATA(idx);
10519 /* We re-use this each time round, throwing the contents away before we
10521 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10522 SV *const utf8_buffer = filter;
10523 IV status = IoPAGE(filter);
10524 const bool reverse = cBOOL(IoLINES(filter));
10527 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10529 /* As we're automatically added, at the lowest level, and hence only called
10530 from this file, we can be sure that we're not called in block mode. Hence
10531 don't bother writing code to deal with block mode. */
10533 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10536 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10538 DEBUG_P(PerlIO_printf(Perl_debug_log,
10539 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10540 FPTR2DPTR(void *, S_utf16_textfilter),
10541 reverse ? 'l' : 'b', idx, maxlen, status,
10542 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10549 /* First, look in our buffer of existing UTF-8 data: */
10550 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10554 } else if (status == 0) {
10556 IoPAGE(filter) = 0;
10557 nl = SvEND(utf8_buffer);
10560 STRLEN got = nl - SvPVX(utf8_buffer);
10561 /* Did we have anything to append? */
10563 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10564 /* Everything else in this code works just fine if SVp_POK isn't
10565 set. This, however, needs it, and we need it to work, else
10566 we loop infinitely because the buffer is never consumed. */
10567 sv_chop(utf8_buffer, nl);
10571 /* OK, not a complete line there, so need to read some more UTF-16.
10572 Read an extra octect if the buffer currently has an odd number. */
10576 if (SvCUR(utf16_buffer) >= 2) {
10577 /* Location of the high octet of the last complete code point.
10578 Gosh, UTF-16 is a pain. All the benefits of variable length,
10579 *coupled* with all the benefits of partial reads and
10581 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10582 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10584 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10588 /* We have the first half of a surrogate. Read more. */
10589 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10592 status = FILTER_READ(idx + 1, utf16_buffer,
10593 160 + (SvCUR(utf16_buffer) & 1));
10594 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10595 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10598 IoPAGE(filter) = status;
10603 chars = SvCUR(utf16_buffer) >> 1;
10604 have = SvCUR(utf8_buffer);
10605 SvGROW(utf8_buffer, have + chars * 3 + 1);
10608 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10609 (U8*)SvPVX_const(utf8_buffer) + have,
10610 chars * 2, &newlen);
10612 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10613 (U8*)SvPVX_const(utf8_buffer) + have,
10614 chars * 2, &newlen);
10616 SvCUR_set(utf8_buffer, have + newlen);
10619 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10620 it's private to us, and utf16_to_utf8{,reversed} take a
10621 (pointer,length) pair, rather than a NUL-terminated string. */
10622 if(SvCUR(utf16_buffer) & 1) {
10623 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10624 SvCUR_set(utf16_buffer, 1);
10626 SvCUR_set(utf16_buffer, 0);
10629 DEBUG_P(PerlIO_printf(Perl_debug_log,
10630 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10632 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10633 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
10638 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
10640 SV *filter = filter_add(S_utf16_textfilter, NULL);
10642 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
10644 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
10645 sv_setpvs(filter, "");
10646 IoLINES(filter) = reversed;
10647 IoPAGE(filter) = 1; /* Not EOF */
10649 /* Sadly, we have to return a valid pointer, come what may, so we have to
10650 ignore any error return from this. */
10651 SvCUR_set(PL_linestr, 0);
10652 if (FILTER_READ(0, PL_linestr, 0)) {
10653 SvUTF8_on(PL_linestr);
10655 SvUTF8_on(PL_linestr);
10657 PL_bufend = SvEND(PL_linestr);
10658 return (U8*)SvPVX(PL_linestr);
10663 Returns a pointer to the next character after the parsed
10664 vstring, as well as updating the passed in sv.
10666 Function must be called like
10668 sv = sv_2mortal(newSV(5));
10669 s = scan_vstring(s,e,sv);
10671 where s and e are the start and end of the string.
10672 The sv should already be large enough to store the vstring
10673 passed in, for performance reasons.
10675 This function may croak if fatal warnings are enabled in the
10676 calling scope, hence the sv_2mortal in the example (to prevent
10677 a leak). Make sure to do SvREFCNT_inc afterwards if you use
10683 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
10685 const char *pos = s;
10686 const char *start = s;
10688 PERL_ARGS_ASSERT_SCAN_VSTRING;
10690 if (*pos == 'v') pos++; /* get past 'v' */
10691 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10693 if ( *pos != '.') {
10694 /* this may not be a v-string if followed by => */
10695 const char *next = pos;
10696 while (next < e && isSPACE(*next))
10698 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
10699 /* return string not v-string */
10700 sv_setpvn(sv,(char *)s,pos-s);
10701 return (char *)pos;
10705 if (!isALPHA(*pos)) {
10706 U8 tmpbuf[UTF8_MAXBYTES+1];
10709 s++; /* get past 'v' */
10714 /* this is atoi() that tolerates underscores */
10717 const char *end = pos;
10719 while (--end >= s) {
10721 const UV orev = rev;
10722 rev += (*end - '0') * mult;
10725 /* diag_listed_as: Integer overflow in %s number */
10726 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10727 "Integer overflow in decimal number");
10731 if (rev > 0x7FFFFFFF)
10732 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10734 /* Append native character for the rev point */
10735 tmpend = uvchr_to_utf8(tmpbuf, rev);
10736 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10737 if (!UVCHR_IS_INVARIANT(rev))
10739 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
10745 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10749 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
10756 Perl_keyword_plugin_standard(pTHX_
10757 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
10759 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
10760 PERL_UNUSED_CONTEXT;
10761 PERL_UNUSED_ARG(keyword_ptr);
10762 PERL_UNUSED_ARG(keyword_len);
10763 PERL_UNUSED_ARG(op_ptr);
10764 return KEYWORD_PLUGIN_DECLINE;
10767 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
10769 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
10771 SAVEI32(PL_lex_brackets);
10772 if (PL_lex_brackets > 100)
10773 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
10774 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
10775 SAVEI32(PL_lex_allbrackets);
10776 PL_lex_allbrackets = 0;
10777 SAVEI8(PL_lex_fakeeof);
10778 PL_lex_fakeeof = (U8)fakeeof;
10779 if(yyparse(gramtype) && !PL_parser->error_count)
10780 qerror(Perl_mess(aTHX_ "Parse error"));
10783 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
10785 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
10789 SAVEVPTR(PL_eval_root);
10790 PL_eval_root = NULL;
10791 parse_recdescent(gramtype, fakeeof);
10797 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
10799 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
10802 if (flags & ~PARSE_OPTIONAL)
10803 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
10804 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
10805 if (!exprop && !(flags & PARSE_OPTIONAL)) {
10806 if (!PL_parser->error_count)
10807 qerror(Perl_mess(aTHX_ "Parse error"));
10808 exprop = newOP(OP_NULL, 0);
10814 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
10816 Parse a Perl arithmetic expression. This may contain operators of precedence
10817 down to the bit shift operators. The expression must be followed (and thus
10818 terminated) either by a comparison or lower-precedence operator or by
10819 something that would normally terminate an expression such as semicolon.
10820 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
10821 otherwise it is mandatory. It is up to the caller to ensure that the
10822 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
10823 the source of the code to be parsed and the lexical context for the
10826 The op tree representing the expression is returned. If an optional
10827 expression is absent, a null pointer is returned, otherwise the pointer
10830 If an error occurs in parsing or compilation, in most cases a valid op
10831 tree is returned anyway. The error is reflected in the parser state,
10832 normally resulting in a single exception at the top level of parsing
10833 which covers all the compilation errors that occurred. Some compilation
10834 errors, however, will throw an exception immediately.
10840 Perl_parse_arithexpr(pTHX_ U32 flags)
10842 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
10846 =for apidoc Amx|OP *|parse_termexpr|U32 flags
10848 Parse a Perl term expression. This may contain operators of precedence
10849 down to the assignment operators. The expression must be followed (and thus
10850 terminated) either by a comma or lower-precedence operator or by
10851 something that would normally terminate an expression such as semicolon.
10852 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
10853 otherwise it is mandatory. It is up to the caller to ensure that the
10854 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
10855 the source of the code to be parsed and the lexical context for the
10858 The op tree representing the expression is returned. If an optional
10859 expression is absent, a null pointer is returned, otherwise the pointer
10862 If an error occurs in parsing or compilation, in most cases a valid op
10863 tree is returned anyway. The error is reflected in the parser state,
10864 normally resulting in a single exception at the top level of parsing
10865 which covers all the compilation errors that occurred. Some compilation
10866 errors, however, will throw an exception immediately.
10872 Perl_parse_termexpr(pTHX_ U32 flags)
10874 return parse_expr(LEX_FAKEEOF_COMMA, flags);
10878 =for apidoc Amx|OP *|parse_listexpr|U32 flags
10880 Parse a Perl list expression. This may contain operators of precedence
10881 down to the comma operator. The expression must be followed (and thus
10882 terminated) either by a low-precedence logic operator such as C<or> or by
10883 something that would normally terminate an expression such as semicolon.
10884 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
10885 otherwise it is mandatory. It is up to the caller to ensure that the
10886 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
10887 the source of the code to be parsed and the lexical context for the
10890 The op tree representing the expression is returned. If an optional
10891 expression is absent, a null pointer is returned, otherwise the pointer
10894 If an error occurs in parsing or compilation, in most cases a valid op
10895 tree is returned anyway. The error is reflected in the parser state,
10896 normally resulting in a single exception at the top level of parsing
10897 which covers all the compilation errors that occurred. Some compilation
10898 errors, however, will throw an exception immediately.
10904 Perl_parse_listexpr(pTHX_ U32 flags)
10906 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
10910 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
10912 Parse a single complete Perl expression. This allows the full
10913 expression grammar, including the lowest-precedence operators such
10914 as C<or>. The expression must be followed (and thus terminated) by a
10915 token that an expression would normally be terminated by: end-of-file,
10916 closing bracketing punctuation, semicolon, or one of the keywords that
10917 signals a postfix expression-statement modifier. If I<flags> includes
10918 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
10919 mandatory. It is up to the caller to ensure that the dynamic parser
10920 state (L</PL_parser> et al) is correctly set to reflect the source of
10921 the code to be parsed and the lexical context for the expression.
10923 The op tree representing the expression is returned. If an optional
10924 expression is absent, a null pointer is returned, otherwise the pointer
10927 If an error occurs in parsing or compilation, in most cases a valid op
10928 tree is returned anyway. The error is reflected in the parser state,
10929 normally resulting in a single exception at the top level of parsing
10930 which covers all the compilation errors that occurred. Some compilation
10931 errors, however, will throw an exception immediately.
10937 Perl_parse_fullexpr(pTHX_ U32 flags)
10939 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
10943 =for apidoc Amx|OP *|parse_block|U32 flags
10945 Parse a single complete Perl code block. This consists of an opening
10946 brace, a sequence of statements, and a closing brace. The block
10947 constitutes a lexical scope, so C<my> variables and various compile-time
10948 effects can be contained within it. It is up to the caller to ensure
10949 that the dynamic parser state (L</PL_parser> et al) is correctly set to
10950 reflect the source of the code to be parsed and the lexical context for
10953 The op tree representing the code block is returned. This is always a
10954 real op, never a null pointer. It will normally be a C<lineseq> list,
10955 including C<nextstate> or equivalent ops. No ops to construct any kind
10956 of runtime scope are included by virtue of it being a block.
10958 If an error occurs in parsing or compilation, in most cases a valid op
10959 tree (most likely null) is returned anyway. The error is reflected in
10960 the parser state, normally resulting in a single exception at the top
10961 level of parsing which covers all the compilation errors that occurred.
10962 Some compilation errors, however, will throw an exception immediately.
10964 The I<flags> parameter is reserved for future use, and must always
10971 Perl_parse_block(pTHX_ U32 flags)
10974 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
10975 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
10979 =for apidoc Amx|OP *|parse_barestmt|U32 flags
10981 Parse a single unadorned Perl statement. This may be a normal imperative
10982 statement or a declaration that has compile-time effect. It does not
10983 include any label or other affixture. It is up to the caller to ensure
10984 that the dynamic parser state (L</PL_parser> et al) is correctly set to
10985 reflect the source of the code to be parsed and the lexical context for
10988 The op tree representing the statement is returned. This may be a
10989 null pointer if the statement is null, for example if it was actually
10990 a subroutine definition (which has compile-time side effects). If not
10991 null, it will be ops directly implementing the statement, suitable to
10992 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
10993 equivalent op (except for those embedded in a scope contained entirely
10994 within the statement).
10996 If an error occurs in parsing or compilation, in most cases a valid op
10997 tree (most likely null) is returned anyway. The error is reflected in
10998 the parser state, normally resulting in a single exception at the top
10999 level of parsing which covers all the compilation errors that occurred.
11000 Some compilation errors, however, will throw an exception immediately.
11002 The I<flags> parameter is reserved for future use, and must always
11009 Perl_parse_barestmt(pTHX_ U32 flags)
11012 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11013 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11017 =for apidoc Amx|SV *|parse_label|U32 flags
11019 Parse a single label, possibly optional, of the type that may prefix a
11020 Perl statement. It is up to the caller to ensure that the dynamic parser
11021 state (L</PL_parser> et al) is correctly set to reflect the source of
11022 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11023 label is optional, otherwise it is mandatory.
11025 The name of the label is returned in the form of a fresh scalar. If an
11026 optional label is absent, a null pointer is returned.
11028 If an error occurs in parsing, which can only occur if the label is
11029 mandatory, a valid label is returned anyway. The error is reflected in
11030 the parser state, normally resulting in a single exception at the top
11031 level of parsing which covers all the compilation errors that occurred.
11037 Perl_parse_label(pTHX_ U32 flags)
11039 if (flags & ~PARSE_OPTIONAL)
11040 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11041 if (PL_lex_state == LEX_KNOWNEXT) {
11042 PL_parser->yychar = yylex();
11043 if (PL_parser->yychar == LABEL) {
11044 char * const lpv = pl_yylval.pval;
11045 STRLEN llen = strlen(lpv);
11046 PL_parser->yychar = YYEMPTY;
11047 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11054 STRLEN wlen, bufptr_pos;
11057 if (!isIDFIRST_lazy_if(s, UTF))
11059 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11060 if (word_takes_any_delimeter(s, wlen))
11062 bufptr_pos = s - SvPVX(PL_linestr);
11064 lex_read_space(LEX_KEEP_PREVIOUS);
11066 s = SvPVX(PL_linestr) + bufptr_pos;
11067 if (t[0] == ':' && t[1] != ':') {
11068 PL_oldoldbufptr = PL_oldbufptr;
11071 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11075 if (flags & PARSE_OPTIONAL) {
11078 qerror(Perl_mess(aTHX_ "Parse error"));
11079 return newSVpvs("x");
11086 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11088 Parse a single complete Perl statement. This may be a normal imperative
11089 statement or a declaration that has compile-time effect, and may include
11090 optional labels. It is up to the caller to ensure that the dynamic
11091 parser state (L</PL_parser> et al) is correctly set to reflect the source
11092 of the code to be parsed and the lexical context for the statement.
11094 The op tree representing the statement is returned. This may be a
11095 null pointer if the statement is null, for example if it was actually
11096 a subroutine definition (which has compile-time side effects). If not
11097 null, it will be the result of a L</newSTATEOP> call, normally including
11098 a C<nextstate> or equivalent op.
11100 If an error occurs in parsing or compilation, in most cases a valid op
11101 tree (most likely null) is returned anyway. The error is reflected in
11102 the parser state, normally resulting in a single exception at the top
11103 level of parsing which covers all the compilation errors that occurred.
11104 Some compilation errors, however, will throw an exception immediately.
11106 The I<flags> parameter is reserved for future use, and must always
11113 Perl_parse_fullstmt(pTHX_ U32 flags)
11116 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11117 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11121 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11123 Parse a sequence of zero or more Perl statements. These may be normal
11124 imperative statements, including optional labels, or declarations
11125 that have compile-time effect, or any mixture thereof. The statement
11126 sequence ends when a closing brace or end-of-file is encountered in a
11127 place where a new statement could have validly started. It is up to
11128 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11129 is correctly set to reflect the source of the code to be parsed and the
11130 lexical context for the statements.
11132 The op tree representing the statement sequence is returned. This may
11133 be a null pointer if the statements were all null, for example if there
11134 were no statements or if there were only subroutine definitions (which
11135 have compile-time side effects). If not null, it will be a C<lineseq>
11136 list, normally including C<nextstate> or equivalent ops.
11138 If an error occurs in parsing or compilation, in most cases a valid op
11139 tree is returned anyway. The error is reflected in the parser state,
11140 normally resulting in a single exception at the top level of parsing
11141 which covers all the compilation errors that occurred. Some compilation
11142 errors, however, will throw an exception immediately.
11144 The I<flags> parameter is reserved for future use, and must always
11151 Perl_parse_stmtseq(pTHX_ U32 flags)
11156 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11157 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11158 c = lex_peek_unichar(0);
11159 if (c != -1 && c != /*{*/'}')
11160 qerror(Perl_mess(aTHX_ "Parse error"));
11164 #define lex_token_boundary() S_lex_token_boundary(aTHX)
11166 S_lex_token_boundary(pTHX)
11168 PL_oldoldbufptr = PL_oldbufptr;
11169 PL_oldbufptr = PL_bufptr;
11172 #define parse_opt_lexvar() S_parse_opt_lexvar(aTHX)
11174 S_parse_opt_lexvar(pTHX)
11179 lex_token_boundary();
11180 sigil = lex_read_unichar(0);
11181 if (lex_peek_unichar(0) == '#') {
11182 qerror(Perl_mess(aTHX_ "Parse error"));
11186 c = lex_peek_unichar(0);
11187 if (c == -1 || !(UTF ? isIDFIRST_uni(c) : isIDFIRST_A(c)))
11190 d = PL_tokenbuf + 1;
11191 PL_tokenbuf[0] = (char)sigil;
11192 parse_ident(&s, &d, PL_tokenbuf + sizeof(PL_tokenbuf) - 1, 0, cBOOL(UTF));
11194 if (d == PL_tokenbuf+1)
11197 var = newOP(sigil == '$' ? OP_PADSV : sigil == '@' ? OP_PADAV : OP_PADHV,
11198 OPf_MOD | (OPpLVAL_INTRO<<8));
11199 var->op_targ = allocmy(PL_tokenbuf, d - PL_tokenbuf, UTF ? SVf_UTF8 : 0);
11204 Perl_parse_subsignature(pTHX)
11207 int prev_type = 0, pos = 0, min_arity = 0, max_arity = 0;
11208 OP *initops = NULL;
11210 c = lex_peek_unichar(0);
11211 while (c != /*(*/')') {
11215 if (prev_type == 2)
11216 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11217 var = parse_opt_lexvar();
11219 newBINOP(OP_AELEM, 0,
11220 ref(newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv)),
11222 newSVOP(OP_CONST, 0, newSViv(pos))) :
11225 c = lex_peek_unichar(0);
11227 lex_token_boundary();
11228 lex_read_unichar(0);
11230 c = lex_peek_unichar(0);
11231 if (c == ',' || c == /*(*/')') {
11233 qerror(Perl_mess(aTHX_ "Optional parameter "
11234 "lacks default expression"));
11236 OP *defexpr = parse_termexpr(0);
11237 if (defexpr->op_type == OP_UNDEF &&
11238 !(defexpr->op_flags & OPf_KIDS)) {
11243 scalar(newUNOP(OP_RV2AV, 0,
11244 newGVOP(OP_GV, 0, PL_defgv))),
11245 newSVOP(OP_CONST, 0, newSViv(pos+1)));
11247 newCONDOP(0, ifop, expr, defexpr) :
11248 newLOGOP(OP_OR, 0, ifop, defexpr);
11253 if (prev_type == 1)
11254 qerror(Perl_mess(aTHX_ "Mandatory parameter "
11255 "follows optional parameter"));
11257 min_arity = pos + 1;
11259 if (var) expr = newASSIGNOP(OPf_STACKED, var, 0, expr);
11261 initops = op_append_list(OP_LINESEQ, initops,
11262 newSTATEOP(0, NULL, expr));
11268 if (prev_type == 2)
11269 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11270 var = parse_opt_lexvar();
11272 OP *chkop = newLOGOP((pos & 1) ? OP_OR : OP_AND, 0,
11273 newBINOP(OP_BIT_AND, 0,
11274 scalar(newUNOP(OP_RV2AV, 0,
11275 newGVOP(OP_GV, 0, PL_defgv))),
11276 newSVOP(OP_CONST, 0, newSViv(1))),
11277 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11278 newSVOP(OP_CONST, 0,
11279 newSVpvs("Odd name/value argument "
11280 "for subroutine"))));
11281 if (pos != min_arity)
11282 chkop = newLOGOP(OP_AND, 0,
11284 scalar(newUNOP(OP_RV2AV, 0,
11285 newGVOP(OP_GV, 0, PL_defgv))),
11286 newSVOP(OP_CONST, 0, newSViv(pos))),
11288 initops = op_append_list(OP_LINESEQ,
11289 newSTATEOP(0, NULL, chkop),
11294 op_prepend_elem(OP_ASLICE,
11295 newOP(OP_PUSHMARK, 0),
11296 newLISTOP(OP_ASLICE, 0,
11298 newSVOP(OP_CONST, 0, newSViv(pos)),
11299 newUNOP(OP_AV2ARYLEN, 0,
11300 ref(newUNOP(OP_RV2AV, 0,
11301 newGVOP(OP_GV, 0, PL_defgv)),
11303 ref(newUNOP(OP_RV2AV, 0,
11304 newGVOP(OP_GV, 0, PL_defgv)),
11306 newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv));
11307 initops = op_append_list(OP_LINESEQ, initops,
11308 newSTATEOP(0, NULL,
11309 newASSIGNOP(OPf_STACKED, var, 0, slice)));
11316 qerror(Perl_mess(aTHX_ "Parse error"));
11320 c = lex_peek_unichar(0);
11322 case /*(*/')': break;
11325 lex_token_boundary();
11326 lex_read_unichar(0);
11328 c = lex_peek_unichar(0);
11329 } while (c == ',');
11335 if (min_arity != 0) {
11336 initops = op_append_list(OP_LINESEQ,
11337 newSTATEOP(0, NULL,
11340 scalar(newUNOP(OP_RV2AV, 0,
11341 newGVOP(OP_GV, 0, PL_defgv))),
11342 newSVOP(OP_CONST, 0, newSViv(min_arity))),
11343 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11344 newSVOP(OP_CONST, 0,
11345 newSVpvs("Too few arguments for subroutine"))))),
11348 if (max_arity != -1) {
11349 initops = op_append_list(OP_LINESEQ,
11350 newSTATEOP(0, NULL,
11353 scalar(newUNOP(OP_RV2AV, 0,
11354 newGVOP(OP_GV, 0, PL_defgv))),
11355 newSVOP(OP_CONST, 0, newSViv(max_arity))),
11356 newLISTOP(OP_DIE, 0, newOP(OP_PUSHMARK, 0),
11357 newSVOP(OP_CONST, 0,
11358 newSVpvs("Too many arguments for subroutine"))))),
11366 * c-indentation-style: bsd
11367 * c-basic-offset: 4
11368 * indent-tabs-mode: nil
11371 * ex: set ts=8 sts=4 sw=4 et: