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_inline.h"
43 #define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46 #define pl_yylval (PL_parser->yylval)
48 /* XXX temporary backwards compatibility */
49 #define PL_lex_brackets (PL_parser->lex_brackets)
50 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
51 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
52 #define PL_lex_brackstack (PL_parser->lex_brackstack)
53 #define PL_lex_casemods (PL_parser->lex_casemods)
54 #define PL_lex_casestack (PL_parser->lex_casestack)
55 #define PL_lex_defer (PL_parser->lex_defer)
56 #define PL_lex_dojoin (PL_parser->lex_dojoin)
57 #define PL_lex_formbrack (PL_parser->lex_formbrack)
58 #define PL_lex_inpat (PL_parser->lex_inpat)
59 #define PL_lex_inwhat (PL_parser->lex_inwhat)
60 #define PL_lex_op (PL_parser->lex_op)
61 #define PL_lex_repl (PL_parser->lex_repl)
62 #define PL_lex_starts (PL_parser->lex_starts)
63 #define PL_lex_stuff (PL_parser->lex_stuff)
64 #define PL_multi_start (PL_parser->multi_start)
65 #define PL_multi_open (PL_parser->multi_open)
66 #define PL_multi_close (PL_parser->multi_close)
67 #define PL_preambled (PL_parser->preambled)
68 #define PL_sublex_info (PL_parser->sublex_info)
69 #define PL_linestr (PL_parser->linestr)
70 #define PL_expect (PL_parser->expect)
71 #define PL_copline (PL_parser->copline)
72 #define PL_bufptr (PL_parser->bufptr)
73 #define PL_oldbufptr (PL_parser->oldbufptr)
74 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
75 #define PL_linestart (PL_parser->linestart)
76 #define PL_bufend (PL_parser->bufend)
77 #define PL_last_uni (PL_parser->last_uni)
78 #define PL_last_lop (PL_parser->last_lop)
79 #define PL_last_lop_op (PL_parser->last_lop_op)
80 #define PL_lex_state (PL_parser->lex_state)
81 #define PL_rsfp (PL_parser->rsfp)
82 #define PL_rsfp_filters (PL_parser->rsfp_filters)
83 #define PL_in_my (PL_parser->in_my)
84 #define PL_in_my_stash (PL_parser->in_my_stash)
85 #define PL_tokenbuf (PL_parser->tokenbuf)
86 #define PL_multi_end (PL_parser->multi_end)
87 #define PL_error_count (PL_parser->error_count)
89 # define PL_nexttoke (PL_parser->nexttoke)
90 # define PL_nexttype (PL_parser->nexttype)
91 # define PL_nextval (PL_parser->nextval)
93 static const char* const ident_too_long = "Identifier too long";
95 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
97 #define XENUMMASK 0x3f
99 #define XFAKEBRACK 0x80
101 #ifdef USE_UTF8_SCRIPTS
102 # define UTF cBOOL(!IN_BYTES)
104 # define UTF cBOOL((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
107 /* The maximum number of characters preceding the unrecognized one to display */
108 #define UNRECOGNIZED_PRECEDE_COUNT 10
110 /* In variables named $^X, these are the legal values for X.
111 * 1999-02-27 mjd-perl-patch@plover.com */
112 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
114 #define SPACE_OR_TAB(c) isBLANK_A(c)
116 #define HEXFP_PEEK(s) \
118 (isXDIGIT(s[1]) || isALPHA_FOLD_EQ(s[1], 'p'))) || \
119 isALPHA_FOLD_EQ(s[0], 'p'))
121 /* LEX_* are values for PL_lex_state, the state of the lexer.
122 * They are arranged oddly so that the guard on the switch statement
123 * can get by with a single comparison (if the compiler is smart enough).
125 * These values refer to the various states within a sublex parse,
126 * i.e. within a double quotish string
129 /* #define LEX_NOTPARSING 11 is done in perl.h. */
131 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
132 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
133 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
134 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
135 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
137 /* at end of code, eg "$x" followed by: */
138 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
139 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
141 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
142 string or after \E, $foo, etc */
143 #define LEX_INTERPCONST 2 /* NOT USED */
144 #define LEX_FORMLINE 1 /* expecting a format line */
145 #define LEX_KNOWNEXT 0 /* next token known; just return it */
149 static const char* const lex_state_names[] = {
164 #include "keywords.h"
166 /* CLINE is a macro that ensures PL_copline has a sane value */
168 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
171 * Convenience functions to return different tokens and prime the
172 * lexer for the next token. They all take an argument.
174 * TOKEN : generic token (used for '(', DOLSHARP, etc)
175 * OPERATOR : generic operator
176 * AOPERATOR : assignment operator
177 * PREBLOCK : beginning the block after an if, while, foreach, ...
178 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
179 * PREREF : *EXPR where EXPR is not a simple identifier
180 * TERM : expression term
181 * POSTDEREF : postfix dereference (->$* ->@[...] etc.)
182 * LOOPX : loop exiting command (goto, last, dump, etc)
183 * FTST : file test operator
184 * FUN0 : zero-argument function
185 * FUN0OP : zero-argument function, with its op created in this file
186 * FUN1 : not used, except for not, which isn't a UNIOP
187 * BOop : bitwise or or xor
189 * BCop : bitwise complement
190 * SHop : shift operator
191 * PWop : power operator
192 * PMop : pattern-matching operator
193 * Aop : addition-level operator
194 * AopNOASSIGN : addition-level operator that is never part of .=
195 * Mop : multiplication-level operator
196 * Eop : equality-testing operator
197 * Rop : relational operator <= != gt
199 * Also see LOP and lop() below.
202 #ifdef DEBUGGING /* Serve -DT. */
203 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
205 # define REPORT(retval) (retval)
208 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
209 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
210 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, retval))
211 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
212 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
213 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
214 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
215 #define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
216 #define LOOPX(f) return (PL_bufptr = force_word(s,WORD,TRUE,FALSE), \
218 PL_expect = PL_nexttoke ? XOPERATOR : XTERM, \
220 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
221 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
222 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
223 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
224 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITOROP))
225 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITANDOP))
226 #define BCop(f) return pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr = s, \
228 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)SHIFTOP))
229 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)POWOP))
230 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
231 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)ADDOP))
232 #define AopNOASSIGN(f) return (pl_yylval.ival=f, PL_bufptr=s, REPORT((int)ADDOP))
233 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)MULOP))
234 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
235 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
237 /* This bit of chicanery makes a unary function followed by
238 * a parenthesis into a function with one argument, highest precedence.
239 * The UNIDOR macro is for unary functions that can be followed by the //
240 * operator (such as C<shift // 0>).
242 #define UNI3(f,x,have_x) { \
243 pl_yylval.ival = f; \
244 if (have_x) PL_expect = x; \
246 PL_last_uni = PL_oldbufptr; \
247 PL_last_lop_op = f; \
249 return REPORT( (int)FUNC1 ); \
251 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
253 #define UNI(f) UNI3(f,XTERM,1)
254 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
255 #define UNIPROTO(f,optional) { \
256 if (optional) PL_last_uni = PL_oldbufptr; \
260 #define UNIBRACK(f) UNI3(f,0,0)
262 /* grandfather return to old style */
265 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
266 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
267 pl_yylval.ival = (f); \
273 #define COPLINE_INC_WITH_HERELINES \
275 CopLINE_inc(PL_curcop); \
276 if (PL_parser->herelines) \
277 CopLINE(PL_curcop) += PL_parser->herelines, \
278 PL_parser->herelines = 0; \
280 /* Called after scan_str to update CopLINE(PL_curcop), but only when there
281 * is no sublex_push to follow. */
282 #define COPLINE_SET_FROM_MULTI_END \
284 CopLINE_set(PL_curcop, PL_multi_end); \
285 if (PL_multi_end != PL_multi_start) \
286 PL_parser->herelines = 0; \
292 /* how to interpret the pl_yylval associated with the token */
296 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
301 static struct debug_tokens {
303 enum token_type type;
305 } const debug_tokens[] =
307 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
308 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
309 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
310 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
311 { ARROW, TOKENTYPE_NONE, "ARROW" },
312 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
313 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
314 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
315 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
316 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
317 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
318 { DO, TOKENTYPE_NONE, "DO" },
319 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
320 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
321 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
322 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
323 { ELSE, TOKENTYPE_NONE, "ELSE" },
324 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
325 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
326 { FOR, TOKENTYPE_IVAL, "FOR" },
327 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
328 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
329 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
330 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
331 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
332 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
333 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
334 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
335 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
336 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
337 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
338 { IF, TOKENTYPE_IVAL, "IF" },
339 { LABEL, TOKENTYPE_PVAL, "LABEL" },
340 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
341 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
342 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
343 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
344 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
345 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
346 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
347 { MY, TOKENTYPE_IVAL, "MY" },
348 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
349 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
350 { OROP, TOKENTYPE_IVAL, "OROP" },
351 { OROR, TOKENTYPE_NONE, "OROR" },
352 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
353 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
354 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
355 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
356 { POSTJOIN, TOKENTYPE_NONE, "POSTJOIN" },
357 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
358 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
359 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
360 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
361 { PREINC, TOKENTYPE_NONE, "PREINC" },
362 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
363 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
364 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
365 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
366 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
367 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
368 { SUB, TOKENTYPE_NONE, "SUB" },
369 { THING, TOKENTYPE_OPVAL, "THING" },
370 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
371 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
372 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
373 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
374 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
375 { USE, TOKENTYPE_IVAL, "USE" },
376 { WHEN, TOKENTYPE_IVAL, "WHEN" },
377 { WHILE, TOKENTYPE_IVAL, "WHILE" },
378 { WORD, TOKENTYPE_OPVAL, "WORD" },
379 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
380 { 0, TOKENTYPE_NONE, NULL }
383 /* dump the returned token in rv, plus any optional arg in pl_yylval */
386 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
388 PERL_ARGS_ASSERT_TOKEREPORT;
391 const char *name = NULL;
392 enum token_type type = TOKENTYPE_NONE;
393 const struct debug_tokens *p;
394 SV* const report = newSVpvs("<== ");
396 for (p = debug_tokens; p->token; p++) {
397 if (p->token == (int)rv) {
404 Perl_sv_catpv(aTHX_ report, name);
405 else if (isGRAPH(rv))
407 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
409 sv_catpvs(report, " (pending identifier)");
412 sv_catpvs(report, "EOF");
414 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
419 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
421 case TOKENTYPE_OPNUM:
422 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
423 PL_op_name[lvalp->ival]);
426 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
428 case TOKENTYPE_OPVAL:
430 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
431 PL_op_name[lvalp->opval->op_type]);
432 if (lvalp->opval->op_type == OP_CONST) {
433 Perl_sv_catpvf(aTHX_ report, " %s",
434 SvPEEK(cSVOPx_sv(lvalp->opval)));
439 sv_catpvs(report, "(opval=null)");
442 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
448 /* print the buffer with suitable escapes */
451 S_printbuf(pTHX_ const char *const fmt, const char *const s)
453 SV* const tmp = newSVpvs("");
455 PERL_ARGS_ASSERT_PRINTBUF;
457 GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
458 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
466 S_deprecate_commaless_var_list(pTHX) {
468 deprecate("comma-less variable list");
469 return REPORT(','); /* grandfather non-comma-format format */
475 * This subroutine looks for an '=' next to the operator that has just been
476 * parsed and turns it into an ASSIGNOP if it finds one.
480 S_ao(pTHX_ int toketype)
482 if (*PL_bufptr == '=') {
484 if (toketype == ANDAND)
485 pl_yylval.ival = OP_ANDASSIGN;
486 else if (toketype == OROR)
487 pl_yylval.ival = OP_ORASSIGN;
488 else if (toketype == DORDOR)
489 pl_yylval.ival = OP_DORASSIGN;
492 return REPORT(toketype);
497 * When Perl expects an operator and finds something else, no_op
498 * prints the warning. It always prints "<something> found where
499 * operator expected. It prints "Missing semicolon on previous line?"
500 * if the surprise occurs at the start of the line. "do you need to
501 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
502 * where the compiler doesn't know if foo is a method call or a function.
503 * It prints "Missing operator before end of line" if there's nothing
504 * after the missing operator, or "... before <...>" if there is something
505 * after the missing operator.
507 * PL_bufptr is expected to point to the start of the thing that was found,
508 * and s after the next token or partial token.
512 S_no_op(pTHX_ const char *const what, char *s)
514 char * const oldbp = PL_bufptr;
515 const bool is_first = (PL_oldbufptr == PL_linestart);
517 PERL_ARGS_ASSERT_NO_OP;
523 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
524 if (ckWARN_d(WARN_SYNTAX)) {
526 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
527 "\t(Missing semicolon on previous line?)\n");
528 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
530 for (t = PL_oldoldbufptr; (isWORDCHAR_lazy_if(t,UTF) || *t == ':');
531 t += UTF ? UTF8SKIP(t) : 1)
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %"UTF8f"?)\n",
536 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
541 "\t(Missing operator before %"UTF8f"?)\n",
542 UTF8fARG(UTF, s - oldbp, oldbp));
550 * Complain about missing quote/regexp/heredoc terminator.
551 * If it's called with NULL then it cauterizes the line buffer.
552 * If we're in a delimited string and the delimiter is a control
553 * character, it's reformatted into a two-char sequence like ^C.
558 S_missingterm(pTHX_ char *s)
563 char * const nl = strrchr(s,'\n');
567 else if ((U8) PL_multi_close < 32) {
569 tmpbuf[1] = (char)toCTRL(PL_multi_close);
574 *tmpbuf = (char)PL_multi_close;
578 q = strchr(s,'"') ? '\'' : '"';
579 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
585 * Check whether the named feature is enabled.
588 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
590 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
592 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
594 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
596 if (namelen > MAX_FEATURE_LEN)
598 memcpy(&he_name[8], name, namelen);
600 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
601 REFCOUNTED_HE_EXISTS));
605 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
606 * utf16-to-utf8-reversed.
609 #ifdef PERL_CR_FILTER
613 const char *s = SvPVX_const(sv);
614 const char * const e = s + SvCUR(sv);
616 PERL_ARGS_ASSERT_STRIP_RETURN;
618 /* outer loop optimized to do nothing if there are no CR-LFs */
620 if (*s++ == '\r' && *s == '\n') {
621 /* hit a CR-LF, need to copy the rest */
625 if (*s == '\r' && s[1] == '\n')
636 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
638 const I32 count = FILTER_READ(idx+1, sv, maxlen);
639 if (count > 0 && !maxlen)
646 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
648 Creates and initialises a new lexer/parser state object, supplying
649 a context in which to lex and parse from a new source of Perl code.
650 A pointer to the new state object is placed in L</PL_parser>. An entry
651 is made on the save stack so that upon unwinding the new state object
652 will be destroyed and the former value of L</PL_parser> will be restored.
653 Nothing else need be done to clean up the parsing context.
655 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
656 non-null, provides a string (in SV form) containing code to be parsed.
657 A copy of the string is made, so subsequent modification of I<line>
658 does not affect parsing. I<rsfp>, if non-null, provides an input stream
659 from which code will be read to be parsed. If both are non-null, the
660 code in I<line> comes first and must consist of complete lines of input,
661 and I<rsfp> supplies the remainder of the source.
663 The I<flags> parameter is reserved for future use. Currently it is only
664 used by perl internally, so extensions should always pass zero.
669 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
670 can share filters with the current parser.
671 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
672 caller, hence isn't owned by the parser, so shouldn't be closed on parser
673 destruction. This is used to handle the case of defaulting to reading the
674 script from the standard input because no filename was given on the command
675 line (without getting confused by situation where STDIN has been closed, so
676 the script handle is opened on fd 0) */
679 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
681 const char *s = NULL;
682 yy_parser *parser, *oparser;
683 if (flags && flags & ~LEX_START_FLAGS)
684 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
686 /* create and initialise a parser */
688 Newxz(parser, 1, yy_parser);
689 parser->old_parser = oparser = PL_parser;
692 parser->stack = NULL;
694 parser->stack_size = 0;
696 /* on scope exit, free this parser and restore any outer one */
698 parser->saved_curcop = PL_curcop;
700 /* initialise lexer state */
702 parser->nexttoke = 0;
703 parser->error_count = oparser ? oparser->error_count : 0;
704 parser->copline = parser->preambling = NOLINE;
705 parser->lex_state = LEX_NORMAL;
706 parser->expect = XSTATE;
708 parser->rsfp_filters =
709 !(flags & LEX_START_SAME_FILTER) || !oparser
711 : MUTABLE_AV(SvREFCNT_inc(
712 oparser->rsfp_filters
713 ? oparser->rsfp_filters
714 : (oparser->rsfp_filters = newAV())
717 Newx(parser->lex_brackstack, 120, char);
718 Newx(parser->lex_casestack, 12, char);
719 *parser->lex_casestack = '\0';
720 Newxz(parser->lex_shared, 1, LEXSHARED);
724 s = SvPV_const(line, len);
725 parser->linestr = flags & LEX_START_COPIED
726 ? SvREFCNT_inc_simple_NN(line)
727 : newSVpvn_flags(s, len, SvUTF8(line));
728 sv_catpvn(parser->linestr, "\n;", rsfp ? 1 : 2);
730 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
732 parser->oldoldbufptr =
735 parser->linestart = SvPVX(parser->linestr);
736 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
737 parser->last_lop = parser->last_uni = NULL;
739 STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
740 |LEX_DONT_CLOSE_RSFP));
741 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
742 |LEX_DONT_CLOSE_RSFP));
744 parser->in_pod = parser->filtered = 0;
748 /* delete a parser object */
751 Perl_parser_free(pTHX_ const yy_parser *parser)
753 PERL_ARGS_ASSERT_PARSER_FREE;
755 PL_curcop = parser->saved_curcop;
756 SvREFCNT_dec(parser->linestr);
758 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
759 PerlIO_clearerr(parser->rsfp);
760 else if (parser->rsfp && (!parser->old_parser ||
761 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
762 PerlIO_close(parser->rsfp);
763 SvREFCNT_dec(parser->rsfp_filters);
764 SvREFCNT_dec(parser->lex_stuff);
765 SvREFCNT_dec(parser->sublex_info.repl);
767 Safefree(parser->lex_brackstack);
768 Safefree(parser->lex_casestack);
769 Safefree(parser->lex_shared);
770 PL_parser = parser->old_parser;
775 Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
777 I32 nexttoke = parser->nexttoke;
778 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
780 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
781 && parser->nextval[nexttoke].opval
782 && parser->nextval[nexttoke].opval->op_slabbed
783 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
784 op_free(parser->nextval[nexttoke].opval);
785 parser->nextval[nexttoke].opval = NULL;
792 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
794 Buffer scalar containing the chunk currently under consideration of the
795 text currently being lexed. This is always a plain string scalar (for
796 which C<SvPOK> is true). It is not intended to be used as a scalar by
797 normal scalar means; instead refer to the buffer directly by the pointer
798 variables described below.
800 The lexer maintains various C<char*> pointers to things in the
801 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
802 reallocated, all of these pointers must be updated. Don't attempt to
803 do this manually, but rather use L</lex_grow_linestr> if you need to
804 reallocate the buffer.
806 The content of the text chunk in the buffer is commonly exactly one
807 complete line of input, up to and including a newline terminator,
808 but there are situations where it is otherwise. The octets of the
809 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
810 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
811 flag on this scalar, which may disagree with it.
813 For direct examination of the buffer, the variable
814 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
815 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
816 of these pointers is usually preferable to examination of the scalar
817 through normal scalar means.
819 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
821 Direct pointer to the end of the chunk of text currently being lexed, the
822 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
823 + SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
824 always located at the end of the buffer, and does not count as part of
825 the buffer's contents.
827 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
829 Points to the current position of lexing inside the lexer buffer.
830 Characters around this point may be freely examined, within
831 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
832 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
833 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
835 Lexing code (whether in the Perl core or not) moves this pointer past
836 the characters that it consumes. It is also expected to perform some
837 bookkeeping whenever a newline character is consumed. This movement
838 can be more conveniently performed by the function L</lex_read_to>,
839 which handles newlines appropriately.
841 Interpretation of the buffer's octets can be abstracted out by
842 using the slightly higher-level functions L</lex_peek_unichar> and
843 L</lex_read_unichar>.
845 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
847 Points to the start of the current line inside the lexer buffer.
848 This is useful for indicating at which column an error occurred, and
849 not much else. This must be updated by any lexing code that consumes
850 a newline; the function L</lex_read_to> handles this detail.
856 =for apidoc Amx|bool|lex_bufutf8
858 Indicates whether the octets in the lexer buffer
859 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
860 of Unicode characters. If not, they should be interpreted as Latin-1
861 characters. This is analogous to the C<SvUTF8> flag for scalars.
863 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
864 contains valid UTF-8. Lexing code must be robust in the face of invalid
867 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
868 is significant, but not the whole story regarding the input character
869 encoding. Normally, when a file is being read, the scalar contains octets
870 and its C<SvUTF8> flag is off, but the octets should be interpreted as
871 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
872 however, the scalar may have the C<SvUTF8> flag on, and in this case its
873 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
874 is in effect. This logic may change in the future; use this function
875 instead of implementing the logic yourself.
881 Perl_lex_bufutf8(pTHX)
887 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
889 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
890 at least I<len> octets (including terminating C<NUL>). Returns a
891 pointer to the reallocated buffer. This is necessary before making
892 any direct modification of the buffer that would increase its length.
893 L</lex_stuff_pvn> provides a more convenient way to insert text into
896 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
897 this function updates all of the lexer's variables that point directly
904 Perl_lex_grow_linestr(pTHX_ STRLEN len)
908 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
909 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
910 linestr = PL_parser->linestr;
911 buf = SvPVX(linestr);
912 if (len <= SvLEN(linestr))
914 bufend_pos = PL_parser->bufend - buf;
915 bufptr_pos = PL_parser->bufptr - buf;
916 oldbufptr_pos = PL_parser->oldbufptr - buf;
917 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
918 linestart_pos = PL_parser->linestart - buf;
919 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
920 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
921 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
922 PL_parser->lex_shared->re_eval_start - buf : 0;
924 buf = sv_grow(linestr, len);
926 PL_parser->bufend = buf + bufend_pos;
927 PL_parser->bufptr = buf + bufptr_pos;
928 PL_parser->oldbufptr = buf + oldbufptr_pos;
929 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
930 PL_parser->linestart = buf + linestart_pos;
931 if (PL_parser->last_uni)
932 PL_parser->last_uni = buf + last_uni_pos;
933 if (PL_parser->last_lop)
934 PL_parser->last_lop = buf + last_lop_pos;
935 if (PL_parser->lex_shared->re_eval_start)
936 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
941 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
943 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
944 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
945 reallocating the buffer if necessary. This means that lexing code that
946 runs later will see the characters as if they had appeared in the input.
947 It is not recommended to do this as part of normal parsing, and most
948 uses of this facility run the risk of the inserted characters being
949 interpreted in an unintended manner.
951 The string to be inserted is represented by I<len> octets starting
952 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
953 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
954 The characters are recoded for the lexer buffer, according to how the
955 buffer is currently being interpreted (L</lex_bufutf8>). If a string
956 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
957 function is more convenient.
963 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
967 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
968 if (flags & ~(LEX_STUFF_UTF8))
969 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
971 if (flags & LEX_STUFF_UTF8) {
974 STRLEN highhalf = 0; /* Count of variants */
975 const char *p, *e = pv+len;
976 for (p = pv; p != e; p++) {
977 if (! UTF8_IS_INVARIANT(*p)) {
983 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
984 bufptr = PL_parser->bufptr;
985 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
986 SvCUR_set(PL_parser->linestr,
987 SvCUR(PL_parser->linestr) + len+highhalf);
988 PL_parser->bufend += len+highhalf;
989 for (p = pv; p != e; p++) {
991 if (! UTF8_IS_INVARIANT(c)) {
992 *bufptr++ = UTF8_TWO_BYTE_HI(c);
993 *bufptr++ = UTF8_TWO_BYTE_LO(c);
1000 if (flags & LEX_STUFF_UTF8) {
1001 STRLEN highhalf = 0;
1002 const char *p, *e = pv+len;
1003 for (p = pv; p != e; p++) {
1005 if (UTF8_IS_ABOVE_LATIN1(c)) {
1006 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1007 "non-Latin-1 character into Latin-1 input");
1008 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
1011 } else if (! UTF8_IS_INVARIANT(c)) {
1012 /* malformed UTF-8 */
1014 SAVESPTR(PL_warnhook);
1015 PL_warnhook = PERL_WARNHOOK_FATAL;
1016 utf8n_to_uvchr((U8*)p, e-p, NULL, 0);
1022 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1023 bufptr = PL_parser->bufptr;
1024 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1025 SvCUR_set(PL_parser->linestr,
1026 SvCUR(PL_parser->linestr) + len-highhalf);
1027 PL_parser->bufend += len-highhalf;
1030 if (UTF8_IS_INVARIANT(*p)) {
1036 *bufptr++ = TWO_BYTE_UTF8_TO_NATIVE(*p, *(p+1));
1042 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1043 bufptr = PL_parser->bufptr;
1044 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1045 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1046 PL_parser->bufend += len;
1047 Copy(pv, bufptr, len, char);
1053 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1055 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1056 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1057 reallocating the buffer if necessary. This means that lexing code that
1058 runs later will see the characters as if they had appeared in the input.
1059 It is not recommended to do this as part of normal parsing, and most
1060 uses of this facility run the risk of the inserted characters being
1061 interpreted in an unintended manner.
1063 The string to be inserted is represented by octets starting at I<pv>
1064 and continuing to the first nul. These octets are interpreted as either
1065 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1066 in I<flags>. The characters are recoded for the lexer buffer, according
1067 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1068 If it is not convenient to nul-terminate a string to be inserted, the
1069 L</lex_stuff_pvn> function is more appropriate.
1075 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1077 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1078 lex_stuff_pvn(pv, strlen(pv), flags);
1082 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1084 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1085 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1086 reallocating the buffer if necessary. This means that lexing code that
1087 runs later will see the characters as if they had appeared in the input.
1088 It is not recommended to do this as part of normal parsing, and most
1089 uses of this facility run the risk of the inserted characters being
1090 interpreted in an unintended manner.
1092 The string to be inserted is the string value of I<sv>. The characters
1093 are recoded for the lexer buffer, according to how the buffer is currently
1094 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1095 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1096 need to construct a scalar.
1102 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1106 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1108 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1110 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1114 =for apidoc Amx|void|lex_unstuff|char *ptr
1116 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1117 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1118 This hides the discarded text from any lexing code that runs later,
1119 as if the text had never appeared.
1121 This is not the normal way to consume lexed text. For that, use
1128 Perl_lex_unstuff(pTHX_ char *ptr)
1132 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1133 buf = PL_parser->bufptr;
1135 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1138 bufend = PL_parser->bufend;
1140 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1141 unstuff_len = ptr - buf;
1142 Move(ptr, buf, bufend+1-ptr, char);
1143 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1144 PL_parser->bufend = bufend - unstuff_len;
1148 =for apidoc Amx|void|lex_read_to|char *ptr
1150 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1151 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1152 performing the correct bookkeeping whenever a newline character is passed.
1153 This is the normal way to consume lexed text.
1155 Interpretation of the buffer's octets can be abstracted out by
1156 using the slightly higher-level functions L</lex_peek_unichar> and
1157 L</lex_read_unichar>.
1163 Perl_lex_read_to(pTHX_ char *ptr)
1166 PERL_ARGS_ASSERT_LEX_READ_TO;
1167 s = PL_parser->bufptr;
1168 if (ptr < s || ptr > PL_parser->bufend)
1169 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1170 for (; s != ptr; s++)
1172 COPLINE_INC_WITH_HERELINES;
1173 PL_parser->linestart = s+1;
1175 PL_parser->bufptr = ptr;
1179 =for apidoc Amx|void|lex_discard_to|char *ptr
1181 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1182 up to I<ptr>. The remaining content of the buffer will be moved, and
1183 all pointers into the buffer updated appropriately. I<ptr> must not
1184 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1185 it is not permitted to discard text that has yet to be lexed.
1187 Normally it is not necessarily to do this directly, because it suffices to
1188 use the implicit discarding behaviour of L</lex_next_chunk> and things
1189 based on it. However, if a token stretches across multiple lines,
1190 and the lexing code has kept multiple lines of text in the buffer for
1191 that purpose, then after completion of the token it would be wise to
1192 explicitly discard the now-unneeded earlier lines, to avoid future
1193 multi-line tokens growing the buffer without bound.
1199 Perl_lex_discard_to(pTHX_ char *ptr)
1203 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1204 buf = SvPVX(PL_parser->linestr);
1206 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1209 if (ptr > PL_parser->bufptr)
1210 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1211 discard_len = ptr - buf;
1212 if (PL_parser->oldbufptr < ptr)
1213 PL_parser->oldbufptr = ptr;
1214 if (PL_parser->oldoldbufptr < ptr)
1215 PL_parser->oldoldbufptr = ptr;
1216 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1217 PL_parser->last_uni = NULL;
1218 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1219 PL_parser->last_lop = NULL;
1220 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1221 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1222 PL_parser->bufend -= discard_len;
1223 PL_parser->bufptr -= discard_len;
1224 PL_parser->oldbufptr -= discard_len;
1225 PL_parser->oldoldbufptr -= discard_len;
1226 if (PL_parser->last_uni)
1227 PL_parser->last_uni -= discard_len;
1228 if (PL_parser->last_lop)
1229 PL_parser->last_lop -= discard_len;
1233 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1235 Reads in the next chunk of text to be lexed, appending it to
1236 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1237 looked to the end of the current chunk and wants to know more. It is
1238 usual, but not necessary, for lexing to have consumed the entirety of
1239 the current chunk at this time.
1241 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1242 chunk (i.e., the current chunk has been entirely consumed), normally the
1243 current chunk will be discarded at the same time that the new chunk is
1244 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1245 will not be discarded. If the current chunk has not been entirely
1246 consumed, then it will not be discarded regardless of the flag.
1248 Returns true if some new text was added to the buffer, or false if the
1249 buffer has reached the end of the input text.
1254 #define LEX_FAKE_EOF 0x80000000
1255 #define LEX_NO_TERM 0x40000000 /* here-doc */
1258 Perl_lex_next_chunk(pTHX_ U32 flags)
1262 STRLEN old_bufend_pos, new_bufend_pos;
1263 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1264 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1265 bool got_some_for_debugger = 0;
1267 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1268 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1269 if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
1271 linestr = PL_parser->linestr;
1272 buf = SvPVX(linestr);
1273 if (!(flags & LEX_KEEP_PREVIOUS) &&
1274 PL_parser->bufptr == PL_parser->bufend) {
1275 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1277 if (PL_parser->last_uni != PL_parser->bufend)
1278 PL_parser->last_uni = NULL;
1279 if (PL_parser->last_lop != PL_parser->bufend)
1280 PL_parser->last_lop = NULL;
1281 last_uni_pos = last_lop_pos = 0;
1285 old_bufend_pos = PL_parser->bufend - buf;
1286 bufptr_pos = PL_parser->bufptr - buf;
1287 oldbufptr_pos = PL_parser->oldbufptr - buf;
1288 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1289 linestart_pos = PL_parser->linestart - buf;
1290 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1291 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1293 if (flags & LEX_FAKE_EOF) {
1295 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1297 } else if (filter_gets(linestr, old_bufend_pos)) {
1299 got_some_for_debugger = 1;
1300 } else if (flags & LEX_NO_TERM) {
1303 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1304 sv_setpvs(linestr, "");
1306 /* End of real input. Close filehandle (unless it was STDIN),
1307 * then add implicit termination.
1309 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1310 PerlIO_clearerr(PL_parser->rsfp);
1311 else if (PL_parser->rsfp)
1312 (void)PerlIO_close(PL_parser->rsfp);
1313 PL_parser->rsfp = NULL;
1314 PL_parser->in_pod = PL_parser->filtered = 0;
1315 if (!PL_in_eval && PL_minus_p) {
1317 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1318 PL_minus_n = PL_minus_p = 0;
1319 } else if (!PL_in_eval && PL_minus_n) {
1320 sv_catpvs(linestr, /*{*/";}");
1323 sv_catpvs(linestr, ";");
1326 buf = SvPVX(linestr);
1327 new_bufend_pos = SvCUR(linestr);
1328 PL_parser->bufend = buf + new_bufend_pos;
1329 PL_parser->bufptr = buf + bufptr_pos;
1330 PL_parser->oldbufptr = buf + oldbufptr_pos;
1331 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1332 PL_parser->linestart = buf + linestart_pos;
1333 if (PL_parser->last_uni)
1334 PL_parser->last_uni = buf + last_uni_pos;
1335 if (PL_parser->last_lop)
1336 PL_parser->last_lop = buf + last_lop_pos;
1337 if (PL_parser->preambling != NOLINE) {
1338 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1339 PL_parser->preambling = NOLINE;
1341 if (got_some_for_debugger && PERLDB_LINE_OR_SAVESRC &&
1342 PL_curstash != PL_debstash) {
1343 /* debugger active and we're not compiling the debugger code,
1344 * so store the line into the debugger's array of lines
1346 update_debugger_info(NULL, buf+old_bufend_pos,
1347 new_bufend_pos-old_bufend_pos);
1353 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1355 Looks ahead one (Unicode) character in the text currently being lexed.
1356 Returns the codepoint (unsigned integer value) of the next character,
1357 or -1 if lexing has reached the end of the input text. To consume the
1358 peeked character, use L</lex_read_unichar>.
1360 If the next character is in (or extends into) the next chunk of input
1361 text, the next chunk will be read in. Normally the current chunk will be
1362 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1363 then the current chunk will not be discarded.
1365 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1366 is encountered, an exception is generated.
1372 Perl_lex_peek_unichar(pTHX_ U32 flags)
1376 if (flags & ~(LEX_KEEP_PREVIOUS))
1377 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1378 s = PL_parser->bufptr;
1379 bufend = PL_parser->bufend;
1385 if (!lex_next_chunk(flags))
1387 s = PL_parser->bufptr;
1388 bufend = PL_parser->bufend;
1391 if (UTF8_IS_INVARIANT(head))
1393 if (UTF8_IS_START(head)) {
1394 len = UTF8SKIP(&head);
1395 while ((STRLEN)(bufend-s) < len) {
1396 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1398 s = PL_parser->bufptr;
1399 bufend = PL_parser->bufend;
1402 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1403 if (retlen == (STRLEN)-1) {
1404 /* malformed UTF-8 */
1406 SAVESPTR(PL_warnhook);
1407 PL_warnhook = PERL_WARNHOOK_FATAL;
1408 utf8n_to_uvchr((U8*)s, bufend-s, NULL, 0);
1414 if (!lex_next_chunk(flags))
1416 s = PL_parser->bufptr;
1423 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1425 Reads the next (Unicode) character in the text currently being lexed.
1426 Returns the codepoint (unsigned integer value) of the character read,
1427 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1428 if lexing has reached the end of the input text. To non-destructively
1429 examine the next character, use L</lex_peek_unichar> instead.
1431 If the next character is in (or extends into) the next chunk of input
1432 text, the next chunk will be read in. Normally the current chunk will be
1433 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1434 then the current chunk will not be discarded.
1436 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1437 is encountered, an exception is generated.
1443 Perl_lex_read_unichar(pTHX_ U32 flags)
1446 if (flags & ~(LEX_KEEP_PREVIOUS))
1447 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1448 c = lex_peek_unichar(flags);
1451 COPLINE_INC_WITH_HERELINES;
1453 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1455 ++(PL_parser->bufptr);
1461 =for apidoc Amx|void|lex_read_space|U32 flags
1463 Reads optional spaces, in Perl style, in the text currently being
1464 lexed. The spaces may include ordinary whitespace characters and
1465 Perl-style comments. C<#line> directives are processed if encountered.
1466 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1467 at a non-space character (or the end of the input text).
1469 If spaces extend into the next chunk of input text, the next chunk will
1470 be read in. Normally the current chunk will be discarded at the same
1471 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1472 chunk will not be discarded.
1477 #define LEX_NO_INCLINE 0x40000000
1478 #define LEX_NO_NEXT_CHUNK 0x80000000
1481 Perl_lex_read_space(pTHX_ U32 flags)
1484 const bool can_incline = !(flags & LEX_NO_INCLINE);
1485 bool need_incline = 0;
1486 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
1487 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1488 s = PL_parser->bufptr;
1489 bufend = PL_parser->bufend;
1495 } while (!(c == '\n' || (c == 0 && s == bufend)));
1496 } else if (c == '\n') {
1499 PL_parser->linestart = s;
1505 } else if (isSPACE(c)) {
1507 } else if (c == 0 && s == bufend) {
1510 if (flags & LEX_NO_NEXT_CHUNK)
1512 PL_parser->bufptr = s;
1513 l = CopLINE(PL_curcop);
1514 CopLINE(PL_curcop) += PL_parser->herelines + 1;
1515 got_more = lex_next_chunk(flags);
1516 CopLINE_set(PL_curcop, l);
1517 s = PL_parser->bufptr;
1518 bufend = PL_parser->bufend;
1521 if (can_incline && need_incline && PL_parser->rsfp) {
1531 PL_parser->bufptr = s;
1536 =for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1538 This function performs syntax checking on a prototype, C<proto>.
1539 If C<warn> is true, any illegal characters or mismatched brackets
1540 will trigger illegalproto warnings, declaring that they were
1541 detected in the prototype for C<name>.
1543 The return value is C<true> if this is a valid prototype, and
1544 C<false> if it is not, regardless of whether C<warn> was C<true> or
1547 Note that C<NULL> is a valid C<proto> and will always return C<true>.
1554 Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1556 STRLEN len, origlen;
1557 char *p = proto ? SvPV(proto, len) : NULL;
1558 bool bad_proto = FALSE;
1559 bool in_brackets = FALSE;
1560 bool after_slash = FALSE;
1561 char greedy_proto = ' ';
1562 bool proto_after_greedy_proto = FALSE;
1563 bool must_be_last = FALSE;
1564 bool underscore = FALSE;
1565 bool bad_proto_after_underscore = FALSE;
1567 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1573 for (; len--; p++) {
1576 proto_after_greedy_proto = TRUE;
1578 if (!strchr(";@%", *p))
1579 bad_proto_after_underscore = TRUE;
1582 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1589 in_brackets = FALSE;
1590 else if ((*p == '@' || *p == '%') &&
1593 must_be_last = TRUE;
1602 after_slash = FALSE;
1607 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
1610 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1611 origlen, UNI_DISPLAY_ISPRINT)
1612 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1614 if (proto_after_greedy_proto)
1615 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1616 "Prototype after '%c' for %"SVf" : %s",
1617 greedy_proto, SVfARG(name), p);
1619 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1620 "Missing ']' in prototype for %"SVf" : %s",
1623 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1624 "Illegal character in prototype for %"SVf" : %s",
1626 if (bad_proto_after_underscore)
1627 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1628 "Illegal character after '_' in prototype for %"SVf" : %s",
1632 return (! (proto_after_greedy_proto || bad_proto) );
1637 * This subroutine has nothing to do with tilting, whether at windmills
1638 * or pinball tables. Its name is short for "increment line". It
1639 * increments the current line number in CopLINE(PL_curcop) and checks
1640 * to see whether the line starts with a comment of the form
1641 * # line 500 "foo.pm"
1642 * If so, it sets the current line number and file to the values in the comment.
1646 S_incline(pTHX_ const char *s)
1654 PERL_ARGS_ASSERT_INCLINE;
1656 COPLINE_INC_WITH_HERELINES;
1657 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1658 && s+1 == PL_bufend && *s == ';') {
1659 /* fake newline in string eval */
1660 CopLINE_dec(PL_curcop);
1665 while (SPACE_OR_TAB(*s))
1667 if (strnEQ(s, "line", 4))
1671 if (SPACE_OR_TAB(*s))
1675 while (SPACE_OR_TAB(*s))
1683 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1685 while (SPACE_OR_TAB(*s))
1687 if (*s == '"' && (t = strchr(s+1, '"'))) {
1693 while (!isSPACE(*t))
1697 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1699 if (*e != '\n' && *e != '\0')
1700 return; /* false alarm */
1702 if (!grok_atoUV(n, &uv, &e))
1704 line_num = ((line_t)uv) - 1;
1707 const STRLEN len = t - s;
1709 if (!PL_rsfp && !PL_parser->filtered) {
1710 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1711 * to *{"::_<newfilename"} */
1712 /* However, the long form of evals is only turned on by the
1713 debugger - usually they're "(eval %lu)" */
1714 GV * const cfgv = CopFILEGV(PL_curcop);
1717 STRLEN tmplen2 = len;
1721 if (tmplen2 + 2 <= sizeof smallbuf)
1724 Newx(tmpbuf2, tmplen2 + 2, char);
1729 memcpy(tmpbuf2 + 2, s, tmplen2);
1732 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1734 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1735 /* adjust ${"::_<newfilename"} to store the new file name */
1736 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1737 /* The line number may differ. If that is the case,
1738 alias the saved lines that are in the array.
1739 Otherwise alias the whole array. */
1740 if (CopLINE(PL_curcop) == line_num) {
1741 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1742 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
1744 else if (GvAV(cfgv)) {
1745 AV * const av = GvAV(cfgv);
1746 const I32 start = CopLINE(PL_curcop)+1;
1747 I32 items = AvFILLp(av) - start;
1749 AV * const av2 = GvAVn(gv2);
1750 SV **svp = AvARRAY(av) + start;
1751 I32 l = (I32)line_num+1;
1753 av_store(av2, l++, SvREFCNT_inc(*svp++));
1758 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1761 CopFILE_free(PL_curcop);
1762 CopFILE_setn(PL_curcop, s, len);
1764 CopLINE_set(PL_curcop, line_num);
1767 #define skipspace(s) skipspace_flags(s, 0)
1771 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1773 AV *av = CopFILEAVx(PL_curcop);
1776 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1778 sv = *av_fetch(av, 0, 1);
1779 SvUPGRADE(sv, SVt_PVMG);
1781 if (!SvPOK(sv)) sv_setpvs(sv,"");
1783 sv_catsv(sv, orig_sv);
1785 sv_catpvn(sv, buf, len);
1790 if (PL_parser->preambling == NOLINE)
1791 av_store(av, CopLINE(PL_curcop), sv);
1797 * Called to gobble the appropriate amount and type of whitespace.
1798 * Skips comments as well.
1802 S_skipspace_flags(pTHX_ char *s, U32 flags)
1804 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
1805 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1806 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
1809 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1811 lex_read_space(flags | LEX_KEEP_PREVIOUS |
1812 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
1813 LEX_NO_NEXT_CHUNK : 0));
1815 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1816 if (PL_linestart > PL_bufptr)
1817 PL_bufptr = PL_linestart;
1825 * Check the unary operators to ensure there's no ambiguity in how they're
1826 * used. An ambiguous piece of code would be:
1828 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1829 * the +5 is its argument.
1838 if (PL_oldoldbufptr != PL_last_uni)
1840 while (isSPACE(*PL_last_uni))
1843 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1844 s += UTF ? UTF8SKIP(s) : 1;
1845 if ((t = strchr(s, '(')) && t < PL_bufptr)
1848 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1849 "Warning: Use of \"%"UTF8f"\" without parentheses is ambiguous",
1850 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
1854 * LOP : macro to build a list operator. Its behaviour has been replaced
1855 * with a subroutine, S_lop() for which LOP is just another name.
1858 #define LOP(f,x) return lop(f,x,s)
1862 * Build a list operator (or something that might be one). The rules:
1863 * - if we have a next token, then it's a list operator (no parens) for
1864 * which the next token has already been parsed; e.g.,
1867 * - if the next thing is an opening paren, then it's a function
1868 * - else it's a list operator
1872 S_lop(pTHX_ I32 f, int x, char *s)
1874 PERL_ARGS_ASSERT_LOP;
1879 PL_last_lop = PL_oldbufptr;
1880 PL_last_lop_op = (OPCODE)f;
1885 return REPORT(FUNC);
1888 return REPORT(FUNC);
1891 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1892 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1893 return REPORT(LSTOP);
1899 * When the lexer realizes it knows the next token (for instance,
1900 * it is reordering tokens for the parser) then it can call S_force_next
1901 * to know what token to return the next time the lexer is called. Caller
1902 * will need to set PL_nextval[] and possibly PL_expect to ensure
1903 * the lexer handles the token correctly.
1907 S_force_next(pTHX_ I32 type)
1911 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1912 tokereport(type, &NEXTVAL_NEXTTOKE);
1915 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
1916 PL_nexttype[PL_nexttoke] = type;
1918 if (PL_lex_state != LEX_KNOWNEXT) {
1919 PL_lex_defer = PL_lex_state;
1920 PL_lex_state = LEX_KNOWNEXT;
1927 * This subroutine handles postfix deref syntax after the arrow has already
1928 * been emitted. @* $* etc. are emitted as two separate token right here.
1929 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1930 * only the first, leaving yylex to find the next.
1934 S_postderef(pTHX_ int const funny, char const next)
1936 assert(funny == DOLSHARP || strchr("$@%&*", funny));
1937 assert(strchr("*[{", next));
1939 PL_expect = XOPERATOR;
1940 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
1941 assert('@' == funny || '$' == funny || DOLSHARP == funny);
1942 PL_lex_state = LEX_INTERPEND;
1943 force_next(POSTJOIN);
1949 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1950 && !PL_lex_brackets)
1952 PL_expect = XOPERATOR;
1961 int yyc = PL_parser->yychar;
1962 if (yyc != YYEMPTY) {
1964 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1965 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1966 PL_lex_allbrackets--;
1968 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1969 } else if (yyc == '('/*)*/) {
1970 PL_lex_allbrackets--;
1975 PL_parser->yychar = YYEMPTY;
1980 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1982 SV * const sv = newSVpvn_utf8(start, len,
1985 && !is_invariant_string((const U8*)start, len)
1986 && is_utf8_string((const U8*)start, len));
1992 * When the lexer knows the next thing is a word (for instance, it has
1993 * just seen -> and it knows that the next char is a word char, then
1994 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1998 * char *start : buffer position (must be within PL_linestr)
1999 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2000 * int check_keyword : if true, Perl checks to make sure the word isn't
2001 * a keyword (do this if the word is a label, e.g. goto FOO)
2002 * int allow_pack : if true, : characters will also be allowed (require,
2003 * use, etc. do this)
2007 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
2012 PERL_ARGS_ASSERT_FORCE_WORD;
2014 start = skipspace(start);
2016 if (isIDFIRST_lazy_if(s,UTF) ||
2017 (allow_pack && *s == ':') )
2019 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2020 if (check_keyword) {
2021 char *s2 = PL_tokenbuf;
2023 if (allow_pack && len > 6 && strnEQ(s2, "CORE::", 6))
2025 if (keyword(s2, len2, 0))
2028 if (token == METHOD) {
2033 PL_expect = XOPERATOR;
2036 NEXTVAL_NEXTTOKE.opval
2037 = (OP*)newSVOP(OP_CONST,0,
2038 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2039 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2047 * Called when the lexer wants $foo *foo &foo etc, but the program
2048 * text only contains the "foo" portion. The first argument is a pointer
2049 * to the "foo", and the second argument is the type symbol to prefix.
2050 * Forces the next token to be a "WORD".
2051 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2055 S_force_ident(pTHX_ const char *s, int kind)
2057 PERL_ARGS_ASSERT_FORCE_IDENT;
2060 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2061 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2062 UTF ? SVf_UTF8 : 0));
2063 NEXTVAL_NEXTTOKE.opval = o;
2066 o->op_private = OPpCONST_ENTERED;
2067 /* XXX see note in pp_entereval() for why we forgo typo
2068 warnings if the symbol must be introduced in an eval.
2070 gv_fetchpvn_flags(s, len,
2071 (PL_in_eval ? GV_ADDMULTI
2072 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2073 kind == '$' ? SVt_PV :
2074 kind == '@' ? SVt_PVAV :
2075 kind == '%' ? SVt_PVHV :
2083 S_force_ident_maybe_lex(pTHX_ char pit)
2085 NEXTVAL_NEXTTOKE.ival = pit;
2090 Perl_str_to_version(pTHX_ SV *sv)
2095 const char *start = SvPV_const(sv,len);
2096 const char * const end = start + len;
2097 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2099 PERL_ARGS_ASSERT_STR_TO_VERSION;
2101 while (start < end) {
2105 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2110 retval += ((NV)n)/nshift;
2119 * Forces the next token to be a version number.
2120 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2121 * and if "guessing" is TRUE, then no new token is created (and the caller
2122 * must use an alternative parsing method).
2126 S_force_version(pTHX_ char *s, int guessing)
2131 PERL_ARGS_ASSERT_FORCE_VERSION;
2139 while (isDIGIT(*d) || *d == '_' || *d == '.')
2141 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2143 s = scan_num(s, &pl_yylval);
2144 version = pl_yylval.opval;
2145 ver = cSVOPx(version)->op_sv;
2146 if (SvPOK(ver) && !SvNIOK(ver)) {
2147 SvUPGRADE(ver, SVt_PVNV);
2148 SvNV_set(ver, str_to_version(ver));
2149 SvNOK_on(ver); /* hint that it is a version */
2152 else if (guessing) {
2157 /* NOTE: The parser sees the package name and the VERSION swapped */
2158 NEXTVAL_NEXTTOKE.opval = version;
2165 * S_force_strict_version
2166 * Forces the next token to be a version number using strict syntax rules.
2170 S_force_strict_version(pTHX_ char *s)
2173 const char *errstr = NULL;
2175 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2177 while (isSPACE(*s)) /* leading whitespace */
2180 if (is_STRICT_VERSION(s,&errstr)) {
2182 s = (char *)scan_version(s, ver, 0);
2183 version = newSVOP(OP_CONST, 0, ver);
2185 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2186 (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
2190 yyerror(errstr); /* version required */
2194 /* NOTE: The parser sees the package name and the VERSION swapped */
2195 NEXTVAL_NEXTTOKE.opval = version;
2203 * Tokenize a quoted string passed in as an SV. It finds the next
2204 * chunk, up to end of string or a backslash. It may make a new
2205 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2210 S_tokeq(pTHX_ SV *sv)
2217 PERL_ARGS_ASSERT_TOKEQ;
2221 assert (!SvIsCOW(sv));
2222 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
2226 /* This is relying on the SV being "well formed" with a trailing '\0' */
2227 while (s < send && !(*s == '\\' && s[1] == '\\'))
2232 if ( PL_hints & HINT_NEW_STRING ) {
2233 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2234 SVs_TEMP | SvUTF8(sv));
2238 if (s + 1 < send && (s[1] == '\\'))
2239 s++; /* all that, just for this */
2244 SvCUR_set(sv, d - SvPVX_const(sv));
2246 if ( PL_hints & HINT_NEW_STRING )
2247 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2252 * Now come three functions related to double-quote context,
2253 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2254 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2255 * interact with PL_lex_state, and create fake ( ... ) argument lists
2256 * to handle functions and concatenation.
2260 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2265 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2267 * Pattern matching will set PL_lex_op to the pattern-matching op to
2268 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2270 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2272 * Everything else becomes a FUNC.
2274 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2275 * had an OP_CONST or OP_READLINE). This just sets us up for a
2276 * call to S_sublex_push().
2280 S_sublex_start(pTHX)
2282 const I32 op_type = pl_yylval.ival;
2284 if (op_type == OP_NULL) {
2285 pl_yylval.opval = PL_lex_op;
2289 if (op_type == OP_CONST) {
2290 SV *sv = PL_lex_stuff;
2291 PL_lex_stuff = NULL;
2294 if (SvTYPE(sv) == SVt_PVIV) {
2295 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2297 const char * const p = SvPV_const(sv, len);
2298 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2302 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2306 PL_sublex_info.super_state = PL_lex_state;
2307 PL_sublex_info.sub_inwhat = (U16)op_type;
2308 PL_sublex_info.sub_op = PL_lex_op;
2309 PL_lex_state = LEX_INTERPPUSH;
2313 pl_yylval.opval = PL_lex_op;
2323 * Create a new scope to save the lexing state. The scope will be
2324 * ended in S_sublex_done. Returns a '(', starting the function arguments
2325 * to the uc, lc, etc. found before.
2326 * Sets PL_lex_state to LEX_INTERPCONCAT.
2333 const bool is_heredoc = PL_multi_close == '<';
2336 PL_lex_state = PL_sublex_info.super_state;
2337 SAVEI8(PL_lex_dojoin);
2338 SAVEI32(PL_lex_brackets);
2339 SAVEI32(PL_lex_allbrackets);
2340 SAVEI32(PL_lex_formbrack);
2341 SAVEI8(PL_lex_fakeeof);
2342 SAVEI32(PL_lex_casemods);
2343 SAVEI32(PL_lex_starts);
2344 SAVEI8(PL_lex_state);
2345 SAVEI8(PL_lex_defer);
2346 SAVESPTR(PL_lex_repl);
2347 SAVEVPTR(PL_lex_inpat);
2348 SAVEI16(PL_lex_inwhat);
2351 SAVECOPLINE(PL_curcop);
2352 SAVEI32(PL_multi_end);
2353 SAVEI32(PL_parser->herelines);
2354 PL_parser->herelines = 0;
2356 SAVEI8(PL_multi_close);
2357 SAVEPPTR(PL_bufptr);
2358 SAVEPPTR(PL_bufend);
2359 SAVEPPTR(PL_oldbufptr);
2360 SAVEPPTR(PL_oldoldbufptr);
2361 SAVEPPTR(PL_last_lop);
2362 SAVEPPTR(PL_last_uni);
2363 SAVEPPTR(PL_linestart);
2364 SAVESPTR(PL_linestr);
2365 SAVEGENERICPV(PL_lex_brackstack);
2366 SAVEGENERICPV(PL_lex_casestack);
2367 SAVEGENERICPV(PL_parser->lex_shared);
2368 SAVEBOOL(PL_parser->lex_re_reparsing);
2369 SAVEI32(PL_copline);
2371 /* The here-doc parser needs to be able to peek into outer lexing
2372 scopes to find the body of the here-doc. So we put PL_linestr and
2373 PL_bufptr into lex_shared, to ‘share’ those values.
2375 PL_parser->lex_shared->ls_linestr = PL_linestr;
2376 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2378 PL_linestr = PL_lex_stuff;
2379 PL_lex_repl = PL_sublex_info.repl;
2380 PL_lex_stuff = NULL;
2381 PL_sublex_info.repl = NULL;
2383 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2384 set for an inner quote-like operator and then an error causes scope-
2385 popping. We must not have a PL_lex_stuff value left dangling, as
2386 that breaks assumptions elsewhere. See bug #123617. */
2387 SAVEGENERICSV(PL_lex_stuff);
2388 SAVEGENERICSV(PL_sublex_info.repl);
2390 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2391 = SvPVX(PL_linestr);
2392 PL_bufend += SvCUR(PL_linestr);
2393 PL_last_lop = PL_last_uni = NULL;
2394 SAVEFREESV(PL_linestr);
2395 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2397 PL_lex_dojoin = FALSE;
2398 PL_lex_brackets = PL_lex_formbrack = 0;
2399 PL_lex_allbrackets = 0;
2400 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2401 Newx(PL_lex_brackstack, 120, char);
2402 Newx(PL_lex_casestack, 12, char);
2403 PL_lex_casemods = 0;
2404 *PL_lex_casestack = '\0';
2406 PL_lex_state = LEX_INTERPCONCAT;
2408 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2409 PL_copline = NOLINE;
2411 Newxz(shared, 1, LEXSHARED);
2412 shared->ls_prev = PL_parser->lex_shared;
2413 PL_parser->lex_shared = shared;
2415 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2416 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2417 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2418 PL_lex_inpat = PL_sublex_info.sub_op;
2420 PL_lex_inpat = NULL;
2422 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2423 PL_in_eval &= ~EVAL_RE_REPARSING;
2430 * Restores lexer state after a S_sublex_push.
2436 if (!PL_lex_starts++) {
2437 SV * const sv = newSVpvs("");
2438 if (SvUTF8(PL_linestr))
2440 PL_expect = XOPERATOR;
2441 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2445 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2446 PL_lex_state = LEX_INTERPCASEMOD;
2450 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2451 assert(PL_lex_inwhat != OP_TRANSR);
2453 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
2454 PL_linestr = PL_lex_repl;
2456 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2457 PL_bufend += SvCUR(PL_linestr);
2458 PL_last_lop = PL_last_uni = NULL;
2459 PL_lex_dojoin = FALSE;
2460 PL_lex_brackets = 0;
2461 PL_lex_allbrackets = 0;
2462 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2463 PL_lex_casemods = 0;
2464 *PL_lex_casestack = '\0';
2466 if (SvEVALED(PL_lex_repl)) {
2467 PL_lex_state = LEX_INTERPNORMAL;
2469 /* we don't clear PL_lex_repl here, so that we can check later
2470 whether this is an evalled subst; that means we rely on the
2471 logic to ensure sublex_done() is called again only via the
2472 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2475 PL_lex_state = LEX_INTERPCONCAT;
2478 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2479 CopLINE(PL_curcop) +=
2480 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xpad_cop_seq.xlow
2481 + PL_parser->herelines;
2482 PL_parser->herelines = 0;
2487 const line_t l = CopLINE(PL_curcop);
2489 if (PL_multi_close == '<')
2490 PL_parser->herelines += l - PL_multi_end;
2491 PL_bufend = SvPVX(PL_linestr);
2492 PL_bufend += SvCUR(PL_linestr);
2493 PL_expect = XOPERATOR;
2498 PERL_STATIC_INLINE SV*
2499 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2501 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2502 * interior, hence to the "}". Finds what the name resolves to, returning
2503 * an SV* containing it; NULL if no valid one found */
2505 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2512 const U8* first_bad_char_loc;
2513 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2515 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2520 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2522 &first_bad_char_loc))
2524 /* If warnings are on, this will print a more detailed analysis of what
2525 * is wrong than the error message below */
2526 utf8n_to_uvchr(first_bad_char_loc,
2527 e - ((char *) first_bad_char_loc),
2530 /* We deliberately don't try to print the malformed character, which
2531 * might not print very well; it also may be just the first of many
2532 * malformations, so don't print what comes after it */
2533 yyerror_pv(Perl_form(aTHX_
2534 "Malformed UTF-8 character immediately after '%.*s'",
2535 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr),
2540 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2541 /* include the <}> */
2542 e - backslash_ptr + 1);
2544 SvREFCNT_dec_NN(res);
2548 /* See if the charnames handler is the Perl core's, and if so, we can skip
2549 * the validation needed for a user-supplied one, as Perl's does its own
2551 table = GvHV(PL_hintgv); /* ^H */
2552 cvp = hv_fetchs(table, "charnames", FALSE);
2553 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2554 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
2556 const char * const name = HvNAME(stash);
2557 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2558 && strEQ(name, "_charnames")) {
2563 /* Here, it isn't Perl's charname handler. We can't rely on a
2564 * user-supplied handler to validate the input name. For non-ut8 input,
2565 * look to see that the first character is legal. Then loop through the
2566 * rest checking that each is a continuation */
2568 /* This code makes the reasonable assumption that the only Latin1-range
2569 * characters that begin a character name alias are alphabetic, otherwise
2570 * would have to create a isCHARNAME_BEGIN macro */
2573 if (! isALPHAU(*s)) {
2578 if (! isCHARNAME_CONT(*s)) {
2581 if (*s == ' ' && *(s-1) == ' ') {
2584 if ((U8) *s == NBSP_NATIVE && ckWARN_d(WARN_DEPRECATED)) {
2585 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2586 "NO-BREAK SPACE in a charnames "
2587 "alias definition is deprecated");
2593 /* Similarly for utf8. For invariants can check directly; for other
2594 * Latin1, can calculate their code point and check; otherwise use a
2596 if (UTF8_IS_INVARIANT(*s)) {
2597 if (! isALPHAU(*s)) {
2601 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2602 if (! isALPHAU(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1)))) {
2608 if (! PL_utf8_charname_begin) {
2609 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2610 PL_utf8_charname_begin = _core_swash_init("utf8",
2611 "_Perl_Charname_Begin",
2613 1, 0, NULL, &flags);
2615 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2622 if (UTF8_IS_INVARIANT(*s)) {
2623 if (! isCHARNAME_CONT(*s)) {
2626 if (*s == ' ' && *(s-1) == ' ') {
2631 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2632 if (! isCHARNAME_CONT(TWO_BYTE_UTF8_TO_NATIVE(*s, *(s+1))))
2636 if (*s == *NBSP_UTF8
2637 && *(s+1) == *(NBSP_UTF8+1)
2638 && ckWARN_d(WARN_DEPRECATED))
2640 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2641 "NO-BREAK SPACE in a charnames "
2642 "alias definition is deprecated");
2647 if (! PL_utf8_charname_continue) {
2648 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2649 PL_utf8_charname_continue = _core_swash_init("utf8",
2650 "_Perl_Charname_Continue",
2652 1, 0, NULL, &flags);
2654 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2661 if (*(s-1) == ' ') {
2664 "charnames alias definitions may not contain trailing "
2665 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2666 (int)(s - backslash_ptr + 1), backslash_ptr,
2667 (int)(e - s + 1), s + 1
2669 UTF ? SVf_UTF8 : 0);
2673 if (SvUTF8(res)) { /* Don't accept malformed input */
2674 const U8* first_bad_char_loc;
2676 const char* const str = SvPV_const(res, len);
2677 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2678 /* If warnings are on, this will print a more detailed analysis of
2679 * what is wrong than the error message below */
2680 utf8n_to_uvchr(first_bad_char_loc,
2681 (char *) first_bad_char_loc - str,
2684 /* We deliberately don't try to print the malformed character,
2685 * which might not print very well; it also may be just the first
2686 * of many malformations, so don't print what comes after it */
2689 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2690 (int) (e - backslash_ptr + 1), backslash_ptr,
2691 (int) ((char *) first_bad_char_loc - str), str
2702 /* The final %.*s makes sure that should the trailing NUL be missing
2703 * that this print won't run off the end of the string */
2706 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2707 (int)(s - backslash_ptr + 1), backslash_ptr,
2708 (int)(e - s + 1), s + 1
2710 UTF ? SVf_UTF8 : 0);
2717 "charnames alias definitions may not contain a sequence of "
2718 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2719 (int)(s - backslash_ptr + 1), backslash_ptr,
2720 (int)(e - s + 1), s + 1
2722 UTF ? SVf_UTF8 : 0);
2729 Extracts the next constant part of a pattern, double-quoted string,
2730 or transliteration. This is terrifying code.
2732 For example, in parsing the double-quoted string "ab\x63$d", it would
2733 stop at the '$' and return an OP_CONST containing 'abc'.
2735 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2736 processing a pattern (PL_lex_inpat is true), a transliteration
2737 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2739 Returns a pointer to the character scanned up to. If this is
2740 advanced from the start pointer supplied (i.e. if anything was
2741 successfully parsed), will leave an OP_CONST for the substring scanned
2742 in pl_yylval. Caller must intuit reason for not parsing further
2743 by looking at the next characters herself.
2747 \N{FOO} => \N{U+hex_for_character_FOO}
2748 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2751 all other \-char, including \N and \N{ apart from \N{ABC}
2754 @ and $ where it appears to be a var, but not for $ as tail anchor
2759 In transliterations:
2760 characters are VERY literal, except for - not at the start or end
2761 of the string, which indicates a range. If the range is in bytes,
2762 scan_const expands the range to the full set of intermediate
2763 characters. If the range is in utf8, the hyphen is replaced with
2764 a certain range mark which will be handled by pmtrans() in op.c.
2766 In double-quoted strings:
2768 double-quoted style: \r and \n
2769 constants: \x31, etc.
2770 deprecated backrefs: \1 (in substitution replacements)
2771 case and quoting: \U \Q \E
2774 scan_const does *not* construct ops to handle interpolated strings.
2775 It stops processing as soon as it finds an embedded $ or @ variable
2776 and leaves it to the caller to work out what's going on.
2778 embedded arrays (whether in pattern or not) could be:
2779 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2781 $ in double-quoted strings must be the symbol of an embedded scalar.
2783 $ in pattern could be $foo or could be tail anchor. Assumption:
2784 it's a tail anchor if $ is the last thing in the string, or if it's
2785 followed by one of "()| \r\n\t"
2787 \1 (backreferences) are turned into $1 in substitutions
2789 The structure of the code is
2790 while (there's a character to process) {
2791 handle transliteration ranges
2792 skip regexp comments /(?#comment)/ and codes /(?{code})/
2793 skip #-initiated comments in //x patterns
2794 check for embedded arrays
2795 check for embedded scalars
2797 deprecate \1 in substitution replacements
2798 handle string-changing backslashes \l \U \Q \E, etc.
2799 switch (what was escaped) {
2800 handle \- in a transliteration (becomes a literal -)
2801 if a pattern and not \N{, go treat as regular character
2802 handle \132 (octal characters)
2803 handle \x15 and \x{1234} (hex characters)
2804 handle \N{name} (named characters, also \N{3,5} in a pattern)
2805 handle \cV (control characters)
2806 handle printf-style backslashes (\f, \r, \n, etc)
2809 } (end if backslash)
2810 handle regular character
2811 } (end while character to read)
2816 S_scan_const(pTHX_ char *start)
2818 char *send = PL_bufend; /* end of the constant */
2819 SV *sv = newSV(send - start); /* sv for the constant. See note below
2821 char *s = start; /* start of the constant */
2822 char *d = SvPVX(sv); /* destination for copies */
2823 bool dorange = FALSE; /* are we in a translit range? */
2824 bool didrange = FALSE; /* did we just finish a range? */
2825 bool in_charclass = FALSE; /* within /[...]/ */
2826 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2827 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2828 UTF8? But, this can show as true
2829 when the source isn't utf8, as for
2830 example when it is entirely composed
2832 SV *res; /* result from charnames */
2834 /* Note on sizing: The scanned constant is placed into sv, which is
2835 * initialized by newSV() assuming one byte of output for every byte of
2836 * input. This routine expects newSV() to allocate an extra byte for a
2837 * trailing NUL, which this routine will append if it gets to the end of
2838 * the input. There may be more bytes of input than output (eg., \N{LATIN
2839 * CAPITAL LETTER A}), or more output than input if the constant ends up
2840 * recoded to utf8, but each time a construct is found that might increase
2841 * the needed size, SvGROW() is called. Its size parameter each time is
2842 * based on the best guess estimate at the time, namely the length used so
2843 * far, plus the length the current construct will occupy, plus room for
2844 * the trailing NUL, plus one byte for every input byte still unscanned */
2846 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2849 UV literal_endpoint = 0;
2850 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2853 PERL_ARGS_ASSERT_SCAN_CONST;
2855 assert(PL_lex_inwhat != OP_TRANSR);
2856 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2857 /* If we are doing a trans and we know we want UTF8 set expectation */
2858 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2859 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2862 /* Protect sv from errors and fatal warnings. */
2863 ENTER_with_name("scan_const");
2866 while (s < send || dorange) {
2868 /* get transliterations out of the way (they're most literal) */
2869 if (PL_lex_inwhat == OP_TRANS) {
2870 /* expand a range A-Z to the full set of characters. AIE! */
2872 I32 i; /* current expanded character */
2873 I32 min; /* first character in range */
2874 I32 max; /* last character in range */
2885 char * const c = (char*)utf8_hop((U8*)d, -1);
2889 *c = (char) ILLEGAL_UTF8_BYTE;
2890 /* mark the range as done, and continue */
2896 i = d - SvPVX_const(sv); /* remember current offset */
2899 SvLEN(sv) + ((has_utf8)
2900 ? (512 - UTF_CONTINUATION_MARK
2903 /* How many two-byte within 0..255: 128 in UTF-8,
2904 * 96 in UTF-8-mod. */
2906 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2908 d = SvPVX(sv) + i; /* refresh d after realloc */
2912 for (j = 0; j <= 1; j++) {
2913 char * const c = (char*)utf8_hop((U8*)d, -1);
2914 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2920 max = (U8)0xff; /* only to \xff */
2921 uvmax = uv; /* \x{100} to uvmax */
2923 d = c; /* eat endpoint chars */
2928 d -= 2; /* eat the first char and the - */
2929 min = (U8)*d; /* first char in range */
2930 max = (U8)d[1]; /* last char in range */
2937 "Invalid range \"%c-%c\" in transliteration operator",
2938 (char)min, (char)max);
2942 /* Because of the discontinuities in EBCDIC A-Z and a-z, expand
2943 * any subsets of these ranges into individual characters */
2944 if (literal_endpoint == 2 &&
2945 ((isLOWER_A(min) && isLOWER_A(max)) ||
2946 (isUPPER_A(min) && isUPPER_A(max))))
2948 for (i = min; i <= max; i++) {
2955 for (i = min; i <= max; i++)
2958 append_utf8_from_native_byte(i, &d);
2966 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2968 *d++ = (char) ILLEGAL_UTF8_BYTE;
2970 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2974 /* mark the range as done, and continue */
2978 literal_endpoint = 0;
2983 /* range begins (ignore - as first or last char) */
2984 else if (*s == '-' && s+1 < send && s != start) {
2986 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2993 *d++ = (char) ILLEGAL_UTF8_BYTE; /* use illegal utf8 byte--see pmtrans */
3003 literal_endpoint = 0;
3004 native_range = TRUE;
3009 /* if we get here, we're not doing a transliteration */
3011 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3014 while (s1 >= start && *s1-- == '\\')
3017 in_charclass = TRUE;
3020 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3023 while (s1 >= start && *s1-- == '\\')
3026 in_charclass = FALSE;
3029 /* skip for regexp comments /(?#comment)/, except for the last
3030 * char, which will be done separately.
3031 * Stop on (?{..}) and friends */
3033 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
3035 while (s+1 < send && *s != ')')
3038 else if (!PL_lex_casemods &&
3039 ( s[2] == '{' /* This should match regcomp.c */
3040 || (s[2] == '?' && s[3] == '{')))
3046 /* likewise skip #-initiated comments in //x patterns */
3047 else if (*s == '#' && PL_lex_inpat && !in_charclass &&
3048 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3049 while (s+1 < send && *s != '\n')
3053 /* no further processing of single-quoted regex */
3054 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3055 goto default_action;
3057 /* check for embedded arrays
3058 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3060 else if (*s == '@' && s[1]) {
3061 if (UTF ? isIDFIRST_utf8((U8*)s+1) : isWORDCHAR_A(s[1]))
3063 if (strchr(":'{$", s[1]))
3065 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3066 break; /* in regexp, neither @+ nor @- are interpolated */
3069 /* check for embedded scalars. only stop if we're sure it's a
3072 else if (*s == '$') {
3073 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3075 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3077 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3078 "Possible unintended interpolation of $\\ in regex");
3080 break; /* in regexp, $ might be tail anchor */
3084 /* End of else if chain - OP_TRANS rejoin rest */
3087 if (*s == '\\' && s+1 < send) {
3088 char* e; /* Can be used for ending '}', etc. */
3092 /* warn on \1 - \9 in substitution replacements, but note that \11
3093 * is an octal; and \19 is \1 followed by '9' */
3094 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3095 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3097 /* diag_listed_as: \%d better written as $%d */
3098 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3103 /* string-change backslash escapes */
3104 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3108 /* In a pattern, process \N, but skip any other backslash escapes.
3109 * This is because we don't want to translate an escape sequence
3110 * into a meta symbol and have the regex compiler use the meta
3111 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3112 * in spite of this, we do have to process \N here while the proper
3113 * charnames handler is in scope. See bugs #56444 and #62056.
3115 * There is a complication because \N in a pattern may also stand
3116 * for 'match a non-nl', and not mean a charname, in which case its
3117 * processing should be deferred to the regex compiler. To be a
3118 * charname it must be followed immediately by a '{', and not look
3119 * like \N followed by a curly quantifier, i.e., not something like
3120 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3122 else if (PL_lex_inpat
3125 || regcurly(s + 1)))
3128 goto default_action;
3133 /* quoted - in transliterations */
3135 if (PL_lex_inwhat == OP_TRANS) {
3142 if ((isALPHANUMERIC(*s)))
3143 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3144 "Unrecognized escape \\%c passed through",
3146 /* default action is to copy the quoted character */
3147 goto default_action;
3150 /* eg. \132 indicates the octal constant 0132 */
3151 case '0': case '1': case '2': case '3':
3152 case '4': case '5': case '6': case '7':
3154 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3156 uv = grok_oct(s, &len, &flags, NULL);
3158 if (len < 3 && s < send && isDIGIT(*s)
3159 && ckWARN(WARN_MISC))
3161 Perl_warner(aTHX_ packWARN(WARN_MISC),
3162 "%s", form_short_octal_warning(s, len));
3165 goto NUM_ESCAPE_INSERT;
3167 /* eg. \o{24} indicates the octal constant \024 */
3172 bool valid = grok_bslash_o(&s, &uv, &error,
3173 TRUE, /* Output warning */
3174 FALSE, /* Not strict */
3175 TRUE, /* Output warnings for
3182 goto NUM_ESCAPE_INSERT;
3185 /* eg. \x24 indicates the hex constant 0x24 */
3190 bool valid = grok_bslash_x(&s, &uv, &error,
3191 TRUE, /* Output warning */
3192 FALSE, /* Not strict */
3193 TRUE, /* Output warnings for
3203 /* Insert oct or hex escaped character. There will always be
3204 * enough room in sv since such escapes will be longer than any
3205 * UTF-8 sequence they can end up as, except if they force us
3206 * to recode the rest of the string into utf8 */
3208 /* Here uv is the ordinal of the next character being added */
3209 if (!UVCHR_IS_INVARIANT(uv)) {
3210 if (!has_utf8 && uv > 255) {
3211 /* Might need to recode whatever we have accumulated so
3212 * far if it contains any chars variant in utf8 or
3215 SvCUR_set(sv, d - SvPVX_const(sv));
3218 /* See Note on sizing above. */
3219 sv_utf8_upgrade_flags_grow(
3221 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE
3222 /* Above-latin1 in string
3223 * implies no encoding */
3224 |SV_UTF8_NO_ENCODING,
3225 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3226 d = SvPVX(sv) + SvCUR(sv);
3231 d = (char*)uvchr_to_utf8((U8*)d, uv);
3232 if (PL_lex_inwhat == OP_TRANS &&
3233 PL_sublex_info.sub_op) {
3234 PL_sublex_info.sub_op->op_private |=
3235 (PL_lex_repl ? OPpTRANS_FROM_UTF
3239 if (uv > 255 && !dorange)
3240 native_range = FALSE;
3253 /* In a non-pattern \N must be like \N{U+0041}, or it can be a
3254 * named character, like \N{LATIN SMALL LETTER A}, or a named
3255 * sequence, like \N{LATIN CAPITAL LETTER A WITH MACRON AND
3256 * GRAVE}. For convenience all three forms are referred to as
3257 * "named characters" below.
3259 * For patterns, \N also can mean to match a non-newline. Code
3260 * before this 'switch' statement should already have handled
3261 * this situation, and hence this code only has to deal with
3262 * the named character cases.
3264 * For non-patterns, the named characters are converted to
3265 * their string equivalents. In patterns, named characters are
3266 * not converted to their ultimate forms for the same reasons
3267 * that other escapes aren't. Instead, they are converted to
3268 * the \N{U+...} form to get the value from the charnames that
3269 * is in effect right now, while preserving the fact that it
3270 * was a named character, so that the regex compiler knows
3273 * The structure of this section of code (besides checking for
3274 * errors and upgrading to utf8) is:
3275 * If the named character is of the form \N{U+...}, pass it
3276 * through if a pattern; otherwise convert the code point
3278 * Otherwise must be some \N{NAME}: convert to \N{U+c1.c2...}
3279 * if a pattern; otherwise convert to utf8
3281 * Here, 's' points to the 'N'; the test below is guaranteed to
3282 * succeed if we are being called on a pattern, as we already
3283 * know from a test above that the next character is a '{'. A
3284 * non-pattern \N must mean 'named character', which requires
3288 yyerror("Missing braces on \\N{}");
3293 /* If there is no matching '}', it is an error. */
3294 if (! (e = strchr(s, '}'))) {
3295 if (! PL_lex_inpat) {
3296 yyerror("Missing right brace on \\N{}");
3298 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
3303 /* Here it looks like a named character */
3305 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3306 s += 2; /* Skip to next char after the 'U+' */
3309 /* In patterns, we can have \N{U+xxxx.yyyy.zzzz...} */
3310 /* Check the syntax. */
3313 if (!isXDIGIT(*s)) {
3316 "Invalid hexadecimal number in \\N{U+...}"
3324 else if ((*s == '.' || *s == '_')
3330 /* Pass everything through unchanged.
3331 * +1 is for the '}' */
3332 Copy(orig_s, d, e - orig_s + 1, char);
3333 d += e - orig_s + 1;
3335 else { /* Not a pattern: convert the hex to string */
3336 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3337 | PERL_SCAN_SILENT_ILLDIGIT
3338 | PERL_SCAN_DISALLOW_PREFIX;
3340 uv = grok_hex(s, &len, &flags, NULL);
3341 if (len == 0 || (len != (STRLEN)(e - s)))
3344 /* If the destination is not in utf8, unconditionally
3345 * recode it to be so. This is because \N{} implies
3346 * Unicode semantics, and scalars have to be in utf8
3347 * to guarantee those semantics */
3349 SvCUR_set(sv, d - SvPVX_const(sv));
3352 /* See Note on sizing above. */
3353 sv_utf8_upgrade_flags_grow(
3355 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3356 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3357 d = SvPVX(sv) + SvCUR(sv);
3361 /* Add the (Unicode) code point to the output. */
3362 if (UNI_IS_INVARIANT(uv)) {
3363 *d++ = (char) LATIN1_TO_NATIVE(uv);
3366 d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
3370 else /* Here is \N{NAME} but not \N{U+...}. */
3371 if ((res = get_and_check_backslash_N_name(s, e)))
3374 const char *str = SvPV_const(res, len);
3377 if (! len) { /* The name resolved to an empty string */
3378 Copy("\\N{}", d, 4, char);
3382 /* In order to not lose information for the regex
3383 * compiler, pass the result in the specially made
3384 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3385 * the code points in hex of each character
3386 * returned by charnames */
3388 const char *str_end = str + len;
3389 const STRLEN off = d - SvPVX_const(sv);
3391 if (! SvUTF8(res)) {
3392 /* For the non-UTF-8 case, we can determine the
3393 * exact length needed without having to parse
3394 * through the string. Each character takes up
3395 * 2 hex digits plus either a trailing dot or
3397 const char initial_text[] = "\\N{U+";
3398 const STRLEN initial_len = sizeof(initial_text)
3400 d = off + SvGROW(sv, off
3403 /* +1 for trailing NUL */
3406 + (STRLEN)(send - e));
3407 Copy(initial_text, d, initial_len, char);
3409 while (str < str_end) {
3412 my_snprintf(hex_string,
3416 /* The regex compiler is
3417 * expecting Unicode, not
3419 (U8) NATIVE_TO_LATIN1(*str));
3420 PERL_MY_SNPRINTF_POST_GUARD(len,
3421 sizeof(hex_string));
3422 Copy(hex_string, d, 3, char);
3426 d--; /* Below, we will overwrite the final
3427 dot with a right brace */
3430 STRLEN char_length; /* cur char's byte length */
3432 /* and the number of bytes after this is
3433 * translated into hex digits */
3434 STRLEN output_length;
3436 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3437 * for max('U+', '.'); and 1 for NUL */
3438 char hex_string[2 * UTF8_MAXBYTES + 5];
3440 /* Get the first character of the result. */
3441 U32 uv = utf8n_to_uvchr((U8 *) str,
3445 /* Convert first code point to Unicode hex,
3446 * including the boiler plate before it. */
3448 my_snprintf(hex_string, sizeof(hex_string),
3450 (unsigned int) NATIVE_TO_UNI(uv));
3452 /* Make sure there is enough space to hold it */
3453 d = off + SvGROW(sv, off
3455 + (STRLEN)(send - e)
3456 + 2); /* '}' + NUL */
3458 Copy(hex_string, d, output_length, char);
3461 /* For each subsequent character, append dot and
3462 * its Unicode code point in hex */
3463 while ((str += char_length) < str_end) {
3464 const STRLEN off = d - SvPVX_const(sv);
3465 U32 uv = utf8n_to_uvchr((U8 *) str,
3470 my_snprintf(hex_string,
3473 (unsigned int) NATIVE_TO_UNI(uv));
3475 d = off + SvGROW(sv, off
3477 + (STRLEN)(send - e)
3478 + 2); /* '}' + NUL */
3479 Copy(hex_string, d, output_length, char);
3484 *d++ = '}'; /* Done. Add the trailing brace */
3487 else { /* Here, not in a pattern. Convert the name to a
3490 /* If destination is not in utf8, unconditionally
3491 * recode it to be so. This is because \N{} implies
3492 * Unicode semantics, and scalars have to be in utf8
3493 * to guarantee those semantics */
3495 SvCUR_set(sv, d - SvPVX_const(sv));
3498 /* See Note on sizing above. */
3499 sv_utf8_upgrade_flags_grow(sv,
3500 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3501 len + (STRLEN)(send - s) + 1);
3502 d = SvPVX(sv) + SvCUR(sv);
3504 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3506 /* See Note on sizing above. (NOTE: SvCUR() is not
3507 * set correctly here). */
3508 const STRLEN off = d - SvPVX_const(sv);
3509 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3511 if (! SvUTF8(res)) { /* Make sure \N{} return is UTF-8 */
3512 sv_utf8_upgrade_flags(res, SV_UTF8_NO_ENCODING);
3513 str = SvPV_const(res, len);
3515 Copy(str, d, len, char);
3521 } /* End \N{NAME} */
3524 native_range = FALSE; /* \N{} is defined to be Unicode */
3526 s = e + 1; /* Point to just after the '}' */
3529 /* \c is a control character */
3533 *d++ = grok_bslash_c(*s++, 1);
3536 yyerror("Missing control char name in \\c");
3540 /* printf-style backslashes, formfeeds, newlines, etc */
3566 } /* end if (backslash) */
3573 /* If we started with encoded form, or already know we want it,
3574 then encode the next character */
3575 if (! NATIVE_BYTE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3579 /* One might think that it is wasted effort in the case of the
3580 * source being utf8 (this_utf8 == TRUE) to take the next character
3581 * in the source, convert it to an unsigned value, and then convert
3582 * it back again. But the source has not been validated here. The
3583 * routine that does the conversion checks for errors like
3586 const UV nextuv = (this_utf8)
3587 ? utf8n_to_uvchr((U8*)s, send - s, &len, 0)
3589 const STRLEN need = UNISKIP(nextuv);
3591 SvCUR_set(sv, d - SvPVX_const(sv));
3594 /* See Note on sizing above. */
3595 sv_utf8_upgrade_flags_grow(sv,
3596 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3597 need + (STRLEN)(send - s) + 1);
3598 d = SvPVX(sv) + SvCUR(sv);
3600 } else if (need > len) {
3601 /* encoded value larger than old, may need extra space (NOTE:
3602 * SvCUR() is not set correctly here). See Note on sizing
3604 const STRLEN off = d - SvPVX_const(sv);
3605 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3609 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3611 if (uv > 255 && !dorange)
3612 native_range = FALSE;
3618 } /* while loop to process each character */
3620 /* terminate the string and set up the sv */
3622 SvCUR_set(sv, d - SvPVX_const(sv));
3623 if (SvCUR(sv) >= SvLEN(sv))
3624 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3625 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3628 if (IN_ENCODING && !has_utf8) {
3629 sv_recode_to_utf8(sv, _get_encoding());
3635 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3636 PL_sublex_info.sub_op->op_private |=
3637 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3641 /* shrink the sv if we allocated more than we used */
3642 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3643 SvPV_shrink_to_cur(sv);
3646 /* return the substring (via pl_yylval) only if we parsed anything */
3649 for (; s2 < s; s2++) {
3651 COPLINE_INC_WITH_HERELINES;
3653 SvREFCNT_inc_simple_void_NN(sv);
3654 if ( (PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ))
3655 && ! PL_parser->lex_re_reparsing)
3657 const char *const key = PL_lex_inpat ? "qr" : "q";
3658 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3662 if (PL_lex_inwhat == OP_TRANS) {
3665 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3668 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3676 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3679 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3681 LEAVE_with_name("scan_const");
3686 * Returns TRUE if there's more to the expression (e.g., a subscript),
3689 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3691 * ->[ and ->{ return TRUE
3692 * ->$* ->$#* ->@* ->@[ ->@{ return TRUE if postderef_qq is enabled
3693 * { and [ outside a pattern are always subscripts, so return TRUE
3694 * if we're outside a pattern and it's not { or [, then return FALSE
3695 * if we're in a pattern and the first char is a {
3696 * {4,5} (any digits around the comma) returns FALSE
3697 * if we're in a pattern and the first char is a [
3699 * [SOMETHING] has a funky algorithm to decide whether it's a
3700 * character class or not. It has to deal with things like
3701 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3702 * anything else returns TRUE
3705 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3708 S_intuit_more(pTHX_ char *s)
3710 PERL_ARGS_ASSERT_INTUIT_MORE;
3712 if (PL_lex_brackets)
3714 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3716 if (*s == '-' && s[1] == '>'
3717 && FEATURE_POSTDEREF_QQ_IS_ENABLED
3718 && ( (s[2] == '$' && (s[3] == '*' || (s[3] == '#' && s[4] == '*')))
3719 ||(s[2] == '@' && strchr("*[{",s[3])) ))
3721 if (*s != '{' && *s != '[')
3726 /* In a pattern, so maybe we have {n,m}. */
3734 /* On the other hand, maybe we have a character class */
3737 if (*s == ']' || *s == '^')
3740 /* this is terrifying, and it works */
3743 const char * const send = strchr(s,']');
3744 unsigned char un_char, last_un_char;
3745 char tmpbuf[sizeof PL_tokenbuf * 4];
3747 if (!send) /* has to be an expression */
3749 weight = 2; /* let's weigh the evidence */
3753 else if (isDIGIT(*s)) {
3755 if (isDIGIT(s[1]) && s[2] == ']')
3761 Zero(seen,256,char);
3763 for (; s < send; s++) {
3764 last_un_char = un_char;
3765 un_char = (unsigned char)*s;
3770 weight -= seen[un_char] * 10;
3771 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3773 char *tmp = PL_bufend;
3774 PL_bufend = (char*)send;
3775 scan_ident(s, tmpbuf, sizeof tmpbuf, FALSE);
3777 len = (int)strlen(tmpbuf);
3778 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3779 UTF ? SVf_UTF8 : 0, SVt_PV))
3784 else if (*s == '$' && s[1] &&
3785 strchr("[#!%*<>()-=",s[1])) {
3786 if (/*{*/ strchr("])} =",s[2]))
3795 if (strchr("wds]",s[1]))
3797 else if (seen[(U8)'\''] || seen[(U8)'"'])
3799 else if (strchr("rnftbxcav",s[1]))
3801 else if (isDIGIT(s[1])) {
3803 while (s[1] && isDIGIT(s[1]))
3813 if (strchr("aA01! ",last_un_char))
3815 if (strchr("zZ79~",s[1]))
3817 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3818 weight -= 5; /* cope with negative subscript */
3821 if (!isWORDCHAR(last_un_char)
3822 && !(last_un_char == '$' || last_un_char == '@'
3823 || last_un_char == '&')
3824 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3828 if (keyword(d, s - d, 0))
3831 if (un_char == last_un_char + 1)
3833 weight -= seen[un_char];
3838 if (weight >= 0) /* probably a character class */
3848 * Does all the checking to disambiguate
3850 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3851 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3853 * First argument is the stuff after the first token, e.g. "bar".
3855 * Not a method if foo is a filehandle.
3856 * Not a method if foo is a subroutine prototyped to take a filehandle.
3857 * Not a method if it's really "Foo $bar"
3858 * Method if it's "foo $bar"
3859 * Not a method if it's really "print foo $bar"
3860 * Method if it's really "foo package::" (interpreted as package->foo)
3861 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3862 * Not a method if bar is a filehandle or package, but is quoted with
3867 S_intuit_method(pTHX_ char *start, SV *ioname, CV *cv)
3869 char *s = start + (*start == '$');
3870 char tmpbuf[sizeof PL_tokenbuf];
3873 /* Mustn't actually add anything to a symbol table.
3874 But also don't want to "initialise" any placeholder
3875 constants that might already be there into full
3876 blown PVGVs with attached PVCV. */
3878 ioname ? gv_fetchsv(ioname, GV_NOADD_NOINIT, SVt_PVCV) : NULL;
3880 PERL_ARGS_ASSERT_INTUIT_METHOD;
3882 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3884 if (cv && SvPOK(cv)) {
3885 const char *proto = CvPROTO(cv);
3887 while (*proto && (isSPACE(*proto) || *proto == ';'))
3894 if (*start == '$') {
3895 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3896 isUPPER(*PL_tokenbuf))
3901 return *s == '(' ? FUNCMETH : METHOD;
3904 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3905 /* start is the beginning of the possible filehandle/object,
3906 * and s is the end of it
3907 * tmpbuf is a copy of it (but with single quotes as double colons)
3910 if (!keyword(tmpbuf, len, 0)) {
3911 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3916 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3917 if (indirgv && GvCVu(indirgv))
3919 /* filehandle or package name makes it a method */
3920 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3922 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3923 return 0; /* no assumptions -- "=>" quotes bareword */
3925 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3926 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3927 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3931 return *s == '(' ? FUNCMETH : METHOD;
3937 /* Encoded script support. filter_add() effectively inserts a
3938 * 'pre-processing' function into the current source input stream.
3939 * Note that the filter function only applies to the current source file
3940 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3942 * The datasv parameter (which may be NULL) can be used to pass
3943 * private data to this instance of the filter. The filter function
3944 * can recover the SV using the FILTER_DATA macro and use it to
3945 * store private buffers and state information.
3947 * The supplied datasv parameter is upgraded to a PVIO type
3948 * and the IoDIRP/IoANY field is used to store the function pointer,
3949 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3950 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3951 * private use must be set using malloc'd pointers.
3955 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3963 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3964 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3966 if (!PL_rsfp_filters)
3967 PL_rsfp_filters = newAV();
3970 SvUPGRADE(datasv, SVt_PVIO);
3971 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3972 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3973 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3974 FPTR2DPTR(void *, IoANY(datasv)),
3975 SvPV_nolen(datasv)));
3976 av_unshift(PL_rsfp_filters, 1);
3977 av_store(PL_rsfp_filters, 0, datasv) ;
3979 !PL_parser->filtered
3980 && PL_parser->lex_flags & LEX_EVALBYTES
3981 && PL_bufptr < PL_bufend
3983 const char *s = PL_bufptr;
3984 while (s < PL_bufend) {
3986 SV *linestr = PL_parser->linestr;
3987 char *buf = SvPVX(linestr);
3988 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3989 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3990 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3991 STRLEN const linestart_pos = PL_parser->linestart - buf;
3992 STRLEN const last_uni_pos =
3993 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3994 STRLEN const last_lop_pos =
3995 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3996 av_push(PL_rsfp_filters, linestr);
3997 PL_parser->linestr =
3998 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3999 buf = SvPVX(PL_parser->linestr);
4000 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4001 PL_parser->bufptr = buf + bufptr_pos;
4002 PL_parser->oldbufptr = buf + oldbufptr_pos;
4003 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4004 PL_parser->linestart = buf + linestart_pos;
4005 if (PL_parser->last_uni)
4006 PL_parser->last_uni = buf + last_uni_pos;
4007 if (PL_parser->last_lop)
4008 PL_parser->last_lop = buf + last_lop_pos;
4009 SvLEN(linestr) = SvCUR(linestr);
4010 SvCUR(linestr) = s-SvPVX(linestr);
4011 PL_parser->filtered = 1;
4021 /* Delete most recently added instance of this filter function. */
4023 Perl_filter_del(pTHX_ filter_t funcp)
4027 PERL_ARGS_ASSERT_FILTER_DEL;
4030 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4031 FPTR2DPTR(void*, funcp)));
4033 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4035 /* if filter is on top of stack (usual case) just pop it off */
4036 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4037 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4038 sv_free(av_pop(PL_rsfp_filters));
4042 /* we need to search for the correct entry and clear it */
4043 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4047 /* Invoke the idxth filter function for the current rsfp. */
4048 /* maxlen 0 = read one text line */
4050 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4054 /* This API is bad. It should have been using unsigned int for maxlen.
4055 Not sure if we want to change the API, but if not we should sanity
4056 check the value here. */
4057 unsigned int correct_length = maxlen < 0 ? PERL_INT_MAX : maxlen;
4059 PERL_ARGS_ASSERT_FILTER_READ;
4061 if (!PL_parser || !PL_rsfp_filters)
4063 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4064 /* Provide a default input filter to make life easy. */
4065 /* Note that we append to the line. This is handy. */
4066 DEBUG_P(PerlIO_printf(Perl_debug_log,
4067 "filter_read %d: from rsfp\n", idx));
4068 if (correct_length) {
4071 const int old_len = SvCUR(buf_sv);
4073 /* ensure buf_sv is large enough */
4074 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4075 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4076 correct_length)) <= 0) {
4077 if (PerlIO_error(PL_rsfp))
4078 return -1; /* error */
4080 return 0 ; /* end of file */
4082 SvCUR_set(buf_sv, old_len + len) ;
4083 SvPVX(buf_sv)[old_len + len] = '\0';
4086 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4087 if (PerlIO_error(PL_rsfp))
4088 return -1; /* error */
4090 return 0 ; /* end of file */
4093 return SvCUR(buf_sv);
4095 /* Skip this filter slot if filter has been deleted */
4096 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4097 DEBUG_P(PerlIO_printf(Perl_debug_log,
4098 "filter_read %d: skipped (filter deleted)\n",
4100 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4102 if (SvTYPE(datasv) != SVt_PVIO) {
4103 if (correct_length) {
4105 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4106 if (!remainder) return 0; /* eof */
4107 if (correct_length > remainder) correct_length = remainder;
4108 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4109 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4112 const char *s = SvEND(datasv);
4113 const char *send = SvPVX(datasv) + SvLEN(datasv);
4121 if (s == send) return 0; /* eof */
4122 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4123 SvCUR_set(datasv, s-SvPVX(datasv));
4125 return SvCUR(buf_sv);
4127 /* Get function pointer hidden within datasv */
4128 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4129 DEBUG_P(PerlIO_printf(Perl_debug_log,
4130 "filter_read %d: via function %p (%s)\n",
4131 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4132 /* Call function. The function is expected to */
4133 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4134 /* Return: <0:error, =0:eof, >0:not eof */
4135 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4139 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4141 PERL_ARGS_ASSERT_FILTER_GETS;
4143 #ifdef PERL_CR_FILTER
4144 if (!PL_rsfp_filters) {
4145 filter_add(S_cr_textfilter,NULL);
4148 if (PL_rsfp_filters) {
4150 SvCUR_set(sv, 0); /* start with empty line */
4151 if (FILTER_READ(0, sv, 0) > 0)
4152 return ( SvPVX(sv) ) ;
4157 return (sv_gets(sv, PL_rsfp, append));
4161 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4165 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4167 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4171 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4172 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4174 return GvHV(gv); /* Foo:: */
4177 /* use constant CLASS => 'MyClass' */
4178 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4179 if (gv && GvCV(gv)) {
4180 SV * const sv = cv_const_sv(GvCV(gv));
4182 return gv_stashsv(sv, 0);
4185 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4190 S_tokenize_use(pTHX_ int is_use, char *s) {
4191 PERL_ARGS_ASSERT_TOKENIZE_USE;
4193 if (PL_expect != XSTATE)
4194 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4195 is_use ? "use" : "no"));
4198 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4199 s = force_version(s, TRUE);
4200 if (*s == ';' || *s == '}'
4201 || (s = skipspace(s), (*s == ';' || *s == '}'))) {
4202 NEXTVAL_NEXTTOKE.opval = NULL;
4205 else if (*s == 'v') {
4206 s = force_word(s,WORD,FALSE,TRUE);
4207 s = force_version(s, FALSE);
4211 s = force_word(s,WORD,FALSE,TRUE);
4212 s = force_version(s, FALSE);
4214 pl_yylval.ival = is_use;
4218 static const char* const exp_name[] =
4219 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4220 "ATTRTERM", "TERMBLOCK", "XBLOCKTERM", "POSTDEREF",
4225 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4227 S_word_takes_any_delimeter(char *p, STRLEN len)
4229 return (len == 1 && strchr("msyq", p[0])) ||
4231 (p[0] == 't' && p[1] == 'r') ||
4232 (p[0] == 'q' && strchr("qwxr", p[1]))));
4236 S_check_scalar_slice(pTHX_ char *s)
4239 while (*s == ' ' || *s == '\t') s++;
4240 if (*s == 'q' && s[1] == 'w'
4241 && !isWORDCHAR_lazy_if(s+2,UTF))
4243 while (*s && (isWORDCHAR_lazy_if(s,UTF) || strchr(" \t$#+-'\"", *s)))
4244 s += UTF ? UTF8SKIP(s) : 1;
4245 if (*s == '}' || *s == ']')
4246 pl_yylval.ival = OPpSLICEWARNING;
4252 Works out what to call the token just pulled out of the input
4253 stream. The yacc parser takes care of taking the ops we return and
4254 stitching them into a tree.
4257 The type of the next token
4260 Switch based on the current state:
4261 - if we already built the token before, use it
4262 - if we have a case modifier in a string, deal with that
4263 - handle other cases of interpolation inside a string
4264 - scan the next line if we are inside a format
4265 In the normal state switch on the next character:
4267 if alphabetic, go to key lookup
4268 unrecoginized character - croak
4269 - 0/4/26: handle end-of-line or EOF
4270 - cases for whitespace
4271 - \n and #: handle comments and line numbers
4272 - various operators, brackets and sigils
4275 - 'v': vstrings (or go to key lookup)
4276 - 'x' repetition operator (or go to key lookup)
4277 - other ASCII alphanumerics (key lookup begins here):
4280 scan built-in keyword (but do nothing with it yet)
4281 check for statement label
4282 check for lexical subs
4283 goto just_a_word if there is one
4284 see whether built-in keyword is overridden
4285 switch on keyword number:
4286 - default: just_a_word:
4287 not a built-in keyword; handle bareword lookup
4288 disambiguate between method and sub call
4289 fall back to bareword
4290 - cases for built-in keywords
4298 char *s = PL_bufptr;
4302 const bool saw_infix_sigil = cBOOL(PL_parser->saw_infix_sigil);
4306 /* orig_keyword, gvp, and gv are initialized here because
4307 * jump to the label just_a_word_zero can bypass their
4308 * initialization later. */
4309 I32 orig_keyword = 0;
4314 SV* tmp = newSVpvs("");
4315 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4316 (IV)CopLINE(PL_curcop),
4317 lex_state_names[PL_lex_state],
4318 exp_name[PL_expect],
4319 pv_display(tmp, s, strlen(s), 0, 60));
4323 /* when we've already built the next token, just pull it out of the queue */
4326 pl_yylval = PL_nextval[PL_nexttoke];
4328 PL_lex_state = PL_lex_defer;
4329 PL_lex_defer = LEX_NORMAL;
4333 next_type = PL_nexttype[PL_nexttoke];
4334 if (next_type & (7<<24)) {
4335 if (next_type & (1<<24)) {
4336 if (PL_lex_brackets > 100)
4337 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4338 PL_lex_brackstack[PL_lex_brackets++] =
4339 (char) ((next_type >> 16) & 0xff);
4341 if (next_type & (2<<24))
4342 PL_lex_allbrackets++;
4343 if (next_type & (4<<24))
4344 PL_lex_allbrackets--;
4345 next_type &= 0xffff;
4347 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4351 switch (PL_lex_state) {
4353 case LEX_INTERPNORMAL:
4356 /* interpolated case modifiers like \L \U, including \Q and \E.
4357 when we get here, PL_bufptr is at the \
4359 case LEX_INTERPCASEMOD:
4361 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4363 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4364 PL_bufptr, PL_bufend, *PL_bufptr);
4366 /* handle \E or end of string */
4367 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4369 if (PL_lex_casemods) {
4370 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4371 PL_lex_casestack[PL_lex_casemods] = '\0';
4373 if (PL_bufptr != PL_bufend
4374 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4375 || oldmod == 'F')) {
4377 PL_lex_state = LEX_INTERPCONCAT;
4379 PL_lex_allbrackets--;
4382 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4383 /* Got an unpaired \E */
4384 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4385 "Useless use of \\E");
4387 if (PL_bufptr != PL_bufend)
4389 PL_lex_state = LEX_INTERPCONCAT;
4393 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4394 "### Saw case modifier\n"); });
4396 if (s[1] == '\\' && s[2] == 'E') {
4398 PL_lex_state = LEX_INTERPCONCAT;
4403 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4404 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4405 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4406 (strchr(PL_lex_casestack, 'L')
4407 || strchr(PL_lex_casestack, 'U')
4408 || strchr(PL_lex_casestack, 'F'))) {
4409 PL_lex_casestack[--PL_lex_casemods] = '\0';
4410 PL_lex_allbrackets--;
4413 if (PL_lex_casemods > 10)
4414 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4415 PL_lex_casestack[PL_lex_casemods++] = *s;
4416 PL_lex_casestack[PL_lex_casemods] = '\0';
4417 PL_lex_state = LEX_INTERPCONCAT;
4418 NEXTVAL_NEXTTOKE.ival = 0;
4419 force_next((2<<24)|'(');
4421 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4423 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4425 NEXTVAL_NEXTTOKE.ival = OP_LC;
4427 NEXTVAL_NEXTTOKE.ival = OP_UC;
4429 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4431 NEXTVAL_NEXTTOKE.ival = OP_FC;
4433 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4437 if (PL_lex_starts) {
4440 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4441 if (PL_lex_casemods == 1 && PL_lex_inpat)
4444 AopNOASSIGN(OP_CONCAT);
4450 case LEX_INTERPPUSH:
4451 return REPORT(sublex_push());
4453 case LEX_INTERPSTART:
4454 if (PL_bufptr == PL_bufend)
4455 return REPORT(sublex_done());
4456 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4457 "### Interpolated variable\n"); });
4459 /* for /@a/, we leave the joining for the regex engine to do
4460 * (unless we're within \Q etc) */
4461 PL_lex_dojoin = (*PL_bufptr == '@'
4462 && (!PL_lex_inpat || PL_lex_casemods));
4463 PL_lex_state = LEX_INTERPNORMAL;
4464 if (PL_lex_dojoin) {
4465 NEXTVAL_NEXTTOKE.ival = 0;
4467 force_ident("\"", '$');
4468 NEXTVAL_NEXTTOKE.ival = 0;
4470 NEXTVAL_NEXTTOKE.ival = 0;
4471 force_next((2<<24)|'(');
4472 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4475 /* Convert (?{...}) and friends to 'do {...}' */
4476 if (PL_lex_inpat && *PL_bufptr == '(') {
4477 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4479 if (*PL_bufptr != '{')
4481 PL_expect = XTERMBLOCK;
4485 if (PL_lex_starts++) {
4487 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4488 if (!PL_lex_casemods && PL_lex_inpat)
4491 AopNOASSIGN(OP_CONCAT);
4495 case LEX_INTERPENDMAYBE:
4496 if (intuit_more(PL_bufptr)) {
4497 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4503 /* Treat state as LEX_NORMAL if we have no inner lexing scope.
4504 XXX This hack can be removed if we stop setting PL_lex_state to
4505 LEX_KNOWNEXT, as can the hack under LEX_INTREPCONCAT below. */
4506 if (UNLIKELY(!PL_lex_inwhat)) {
4507 PL_lex_state = LEX_NORMAL;
4511 if (PL_lex_dojoin) {
4512 const U8 dojoin_was = PL_lex_dojoin;
4513 PL_lex_dojoin = FALSE;
4514 PL_lex_state = LEX_INTERPCONCAT;
4515 PL_lex_allbrackets--;
4516 return REPORT(dojoin_was == 1 ? ')' : POSTJOIN);
4518 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4519 && SvEVALED(PL_lex_repl))
4521 if (PL_bufptr != PL_bufend)
4522 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4525 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4526 re_eval_str. If the here-doc body’s length equals the previous
4527 value of re_eval_start, re_eval_start will now be null. So
4528 check re_eval_str as well. */
4529 if (PL_parser->lex_shared->re_eval_start
4530 || PL_parser->lex_shared->re_eval_str) {
4532 if (*PL_bufptr != ')')
4533 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4535 /* having compiled a (?{..}) expression, return the original
4536 * text too, as a const */
4537 if (PL_parser->lex_shared->re_eval_str) {
4538 sv = PL_parser->lex_shared->re_eval_str;
4539 PL_parser->lex_shared->re_eval_str = NULL;
4541 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4542 SvPV_shrink_to_cur(sv);
4544 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4545 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4546 NEXTVAL_NEXTTOKE.opval =
4547 (OP*)newSVOP(OP_CONST, 0,
4550 PL_parser->lex_shared->re_eval_start = NULL;
4556 case LEX_INTERPCONCAT:
4558 if (PL_lex_brackets)
4559 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4560 (long) PL_lex_brackets);
4562 /* Treat state as LEX_NORMAL when not in an inner lexing scope.
4563 XXX This hack can be removed if we stop setting PL_lex_state to
4565 if (UNLIKELY(!PL_lex_inwhat)) {
4566 PL_lex_state = LEX_NORMAL;
4570 if (PL_bufptr == PL_bufend)
4571 return REPORT(sublex_done());
4573 /* m'foo' still needs to be parsed for possible (?{...}) */
4574 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4575 SV *sv = newSVsv(PL_linestr);
4577 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4581 s = scan_const(PL_bufptr);
4583 PL_lex_state = LEX_INTERPCASEMOD;
4585 PL_lex_state = LEX_INTERPSTART;
4588 if (s != PL_bufptr) {
4589 NEXTVAL_NEXTTOKE = pl_yylval;
4592 if (PL_lex_starts++) {
4593 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4594 if (!PL_lex_casemods && PL_lex_inpat)
4597 AopNOASSIGN(OP_CONCAT);
4607 s = scan_formline(PL_bufptr);
4608 if (!PL_lex_formbrack)
4617 /* We really do *not* want PL_linestr ever becoming a COW. */
4618 assert (!SvIsCOW(PL_linestr));
4620 PL_oldoldbufptr = PL_oldbufptr;
4622 PL_parser->saw_infix_sigil = 0;
4627 if (UTF ? isIDFIRST_utf8((U8*)s) : isALNUMC(*s))
4630 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4631 const char *c = UTF ? sv_uni_display(dsv, newSVpvn_flags(s,
4633 SVs_TEMP | SVf_UTF8),
4634 10, UNI_DISPLAY_ISPRINT)
4635 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4636 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4637 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4638 d = UTF ? (char *) utf8_hop((U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4642 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"UTF8f"<-- HERE near column %d", c,
4643 UTF8fARG(UTF, (s - d), d),
4648 goto fake_eof; /* emulate EOF on ^D or ^Z */
4650 if ((!PL_rsfp || PL_lex_inwhat)
4651 && (!PL_parser->filtered || s+1 < PL_bufend)) {
4654 if (PL_lex_brackets &&
4655 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4656 yyerror((const char *)
4658 ? "Format not terminated"
4659 : "Missing right curly or square bracket"));
4661 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4662 "### Tokener got EOF\n");
4666 if (s++ < PL_bufend)
4667 goto retry; /* ignore stray nulls */
4670 if (!PL_in_eval && !PL_preambled) {
4671 PL_preambled = TRUE;
4673 /* Generate a string of Perl code to load the debugger.
4674 * If PERL5DB is set, it will return the contents of that,
4675 * otherwise a compile-time require of perl5db.pl. */
4677 const char * const pdb = PerlEnv_getenv("PERL5DB");
4680 sv_setpv(PL_linestr, pdb);
4681 sv_catpvs(PL_linestr,";");
4683 SETERRNO(0,SS_NORMAL);
4684 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4686 PL_parser->preambling = CopLINE(PL_curcop);
4688 sv_setpvs(PL_linestr,"");
4689 if (PL_preambleav) {
4690 SV **svp = AvARRAY(PL_preambleav);
4691 SV **const end = svp + AvFILLp(PL_preambleav);
4693 sv_catsv(PL_linestr, *svp);
4695 sv_catpvs(PL_linestr, ";");
4697 sv_free(MUTABLE_SV(PL_preambleav));
4698 PL_preambleav = NULL;
4701 sv_catpvs(PL_linestr,
4702 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4703 if (PL_minus_n || PL_minus_p) {
4704 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4706 sv_catpvs(PL_linestr,"chomp;");
4709 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4710 || *PL_splitstr == '"')
4711 && strchr(PL_splitstr + 1, *PL_splitstr))
4712 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4714 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4715 bytes can be used as quoting characters. :-) */
4716 const char *splits = PL_splitstr;
4717 sv_catpvs(PL_linestr, "our @F=split(q\0");
4720 if (*splits == '\\')
4721 sv_catpvn(PL_linestr, splits, 1);
4722 sv_catpvn(PL_linestr, splits, 1);
4723 } while (*splits++);
4724 /* This loop will embed the trailing NUL of
4725 PL_linestr as the last thing it does before
4727 sv_catpvs(PL_linestr, ");");
4731 sv_catpvs(PL_linestr,"our @F=split(' ');");
4734 sv_catpvs(PL_linestr, "\n");
4735 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4736 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4737 PL_last_lop = PL_last_uni = NULL;
4738 if (PERLDB_LINE_OR_SAVESRC && PL_curstash != PL_debstash)
4739 update_debugger_info(PL_linestr, NULL, 0);
4744 bof = PL_rsfp ? TRUE : FALSE;
4747 fake_eof = LEX_FAKE_EOF;
4749 PL_bufptr = PL_bufend;
4750 COPLINE_INC_WITH_HERELINES;
4751 if (!lex_next_chunk(fake_eof)) {
4752 CopLINE_dec(PL_curcop);
4754 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4756 CopLINE_dec(PL_curcop);
4758 /* If it looks like the start of a BOM or raw UTF-16,
4759 * check if it in fact is. */
4760 if (bof && PL_rsfp &&
4762 *(U8*)s == BOM_UTF8_FIRST_BYTE ||
4765 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4766 bof = (offset == (Off_t)SvCUR(PL_linestr));
4767 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4768 /* offset may include swallowed CR */
4770 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4773 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4774 s = swallow_bom((U8*)s);
4777 if (PL_parser->in_pod) {
4778 /* Incest with pod. */
4779 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4780 sv_setpvs(PL_linestr, "");
4781 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4782 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4783 PL_last_lop = PL_last_uni = NULL;
4784 PL_parser->in_pod = 0;
4787 if (PL_rsfp || PL_parser->filtered)
4789 } while (PL_parser->in_pod);
4790 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4791 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4792 PL_last_lop = PL_last_uni = NULL;
4793 if (CopLINE(PL_curcop) == 1) {
4794 while (s < PL_bufend && isSPACE(*s))
4796 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4800 if (*s == '#' && *(s+1) == '!')
4802 #ifdef ALTERNATE_SHEBANG
4804 static char const as[] = ALTERNATE_SHEBANG;
4805 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4806 d = s + (sizeof(as) - 1);
4808 #endif /* ALTERNATE_SHEBANG */
4817 while (*d && !isSPACE(*d))
4821 #ifdef ARG_ZERO_IS_SCRIPT
4822 if (ipathend > ipath) {
4824 * HP-UX (at least) sets argv[0] to the script name,
4825 * which makes $^X incorrect. And Digital UNIX and Linux,
4826 * at least, set argv[0] to the basename of the Perl
4827 * interpreter. So, having found "#!", we'll set it right.
4829 SV* copfilesv = CopFILESV(PL_curcop);
4832 GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4834 assert(SvPOK(x) || SvGMAGICAL(x));
4835 if (sv_eq(x, copfilesv)) {
4836 sv_setpvn(x, ipath, ipathend - ipath);
4842 const char *bstart = SvPV_const(copfilesv, blen);
4843 const char * const lstart = SvPV_const(x, llen);
4845 bstart += blen - llen;
4846 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4847 sv_setpvn(x, ipath, ipathend - ipath);
4854 /* Anything to do if no copfilesv? */
4856 TAINT_NOT; /* $^X is always tainted, but that's OK */
4858 #endif /* ARG_ZERO_IS_SCRIPT */
4863 d = instr(s,"perl -");
4865 d = instr(s,"perl");
4866 if (d && d[4] == '6')
4869 /* avoid getting into infinite loops when shebang
4870 * line contains "Perl" rather than "perl" */
4872 for (d = ipathend-4; d >= ipath; --d) {
4873 if (isALPHA_FOLD_EQ(*d, 'p')
4874 && !ibcmp(d, "perl", 4))
4884 #ifdef ALTERNATE_SHEBANG
4886 * If the ALTERNATE_SHEBANG on this system starts with a
4887 * character that can be part of a Perl expression, then if
4888 * we see it but not "perl", we're probably looking at the
4889 * start of Perl code, not a request to hand off to some
4890 * other interpreter. Similarly, if "perl" is there, but
4891 * not in the first 'word' of the line, we assume the line
4892 * contains the start of the Perl program.
4894 if (d && *s != '#') {
4895 const char *c = ipath;
4896 while (*c && !strchr("; \t\r\n\f\v#", *c))
4899 d = NULL; /* "perl" not in first word; ignore */
4901 *s = '#'; /* Don't try to parse shebang line */
4903 #endif /* ALTERNATE_SHEBANG */
4908 !instr(s,"indir") &&
4909 instr(PL_origargv[0],"perl"))
4916 while (s < PL_bufend && isSPACE(*s))
4918 if (s < PL_bufend) {
4919 Newx(newargv,PL_origargc+3,char*);
4921 while (s < PL_bufend && !isSPACE(*s))
4924 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4927 newargv = PL_origargv;
4930 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4932 Perl_croak(aTHX_ "Can't exec %s", ipath);
4935 while (*d && !isSPACE(*d))
4937 while (SPACE_OR_TAB(*d))
4941 const bool switches_done = PL_doswitches;
4942 const U32 oldpdb = PL_perldb;
4943 const bool oldn = PL_minus_n;
4944 const bool oldp = PL_minus_p;
4948 bool baduni = FALSE;
4950 const char *d2 = d1 + 1;
4951 if (parse_unicode_opts((const char **)&d2)
4955 if (baduni || isALPHA_FOLD_EQ(*d1, 'M')) {
4956 const char * const m = d1;
4957 while (*d1 && !isSPACE(*d1))
4959 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4962 d1 = moreswitches(d1);
4964 if (PL_doswitches && !switches_done) {
4965 int argc = PL_origargc;
4966 char **argv = PL_origargv;
4969 } while (argc && argv[0][0] == '-' && argv[0][1]);
4970 init_argv_symbols(argc,argv);
4972 if ((PERLDB_LINE_OR_SAVESRC && !oldpdb) ||
4973 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4974 /* if we have already added "LINE: while (<>) {",
4975 we must not do it again */
4977 sv_setpvs(PL_linestr, "");
4978 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4979 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4980 PL_last_lop = PL_last_uni = NULL;
4981 PL_preambled = FALSE;
4982 if (PERLDB_LINE_OR_SAVESRC)
4983 (void)gv_fetchfile(PL_origfilename);
4990 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4991 PL_lex_state = LEX_FORMLINE;
4992 force_next(FORMRBRACK);
4997 #ifdef PERL_STRICT_CR
4998 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5000 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5002 case ' ': case '\t': case '\f': case '\v':
5007 if (PL_lex_state != LEX_NORMAL ||
5008 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5009 const bool in_comment = *s == '#';
5010 if (*s == '#' && s == PL_linestart && PL_in_eval
5011 && !PL_rsfp && !PL_parser->filtered) {
5012 /* handle eval qq[#line 1 "foo"\n ...] */
5013 CopLINE_dec(PL_curcop);
5017 while (d < PL_bufend && *d != '\n')
5021 else if (d > PL_bufend)
5022 /* Found by Ilya: feed random input to Perl. */
5023 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5026 if (in_comment && d == PL_bufend
5027 && PL_lex_state == LEX_INTERPNORMAL
5028 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5029 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5032 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5033 PL_lex_state = LEX_FORMLINE;
5034 force_next(FORMRBRACK);
5039 while (s < PL_bufend && *s != '\n')
5047 else if (s > PL_bufend)
5048 /* Found by Ilya: feed random input to Perl. */
5049 Perl_croak(aTHX_ "panic: input overflow");
5053 if (s[1] && isALPHA(s[1]) && !isWORDCHAR(s[2])) {
5061 while (s < PL_bufend && SPACE_OR_TAB(*s))
5064 if (strnEQ(s,"=>",2)) {
5065 s = force_word(PL_bufptr,WORD,FALSE,FALSE);
5066 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5067 OPERATOR('-'); /* unary minus */
5070 case 'r': ftst = OP_FTEREAD; break;
5071 case 'w': ftst = OP_FTEWRITE; break;
5072 case 'x': ftst = OP_FTEEXEC; break;
5073 case 'o': ftst = OP_FTEOWNED; break;
5074 case 'R': ftst = OP_FTRREAD; break;
5075 case 'W': ftst = OP_FTRWRITE; break;
5076 case 'X': ftst = OP_FTREXEC; break;
5077 case 'O': ftst = OP_FTROWNED; break;
5078 case 'e': ftst = OP_FTIS; break;
5079 case 'z': ftst = OP_FTZERO; break;
5080 case 's': ftst = OP_FTSIZE; break;
5081 case 'f': ftst = OP_FTFILE; break;
5082 case 'd': ftst = OP_FTDIR; break;
5083 case 'l': ftst = OP_FTLINK; break;
5084 case 'p': ftst = OP_FTPIPE; break;
5085 case 'S': ftst = OP_FTSOCK; break;
5086 case 'u': ftst = OP_FTSUID; break;
5087 case 'g': ftst = OP_FTSGID; break;
5088 case 'k': ftst = OP_FTSVTX; break;
5089 case 'b': ftst = OP_FTBLK; break;
5090 case 'c': ftst = OP_FTCHR; break;
5091 case 't': ftst = OP_FTTTY; break;
5092 case 'T': ftst = OP_FTTEXT; break;
5093 case 'B': ftst = OP_FTBINARY; break;
5094 case 'M': case 'A': case 'C':
5095 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5097 case 'M': ftst = OP_FTMTIME; break;
5098 case 'A': ftst = OP_FTATIME; break;
5099 case 'C': ftst = OP_FTCTIME; break;
5107 PL_last_uni = PL_oldbufptr;
5108 PL_last_lop_op = (OPCODE)ftst;
5109 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5110 "### Saw file test %c\n", (int)tmp);
5115 /* Assume it was a minus followed by a one-letter named
5116 * subroutine call (or a -bareword), then. */
5117 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5118 "### '-%c' looked like a file test but was not\n",
5125 const char tmp = *s++;
5128 if (PL_expect == XOPERATOR)
5133 else if (*s == '>') {
5136 if (((*s == '$' || *s == '&') && s[1] == '*')
5137 ||(*s == '$' && s[1] == '#' && s[2] == '*')
5138 ||((*s == '@' || *s == '%') && strchr("*[{", s[1]))
5139 ||(*s == '*' && (s[1] == '*' || s[1] == '{'))
5142 PL_expect = XPOSTDEREF;
5145 if (isIDFIRST_lazy_if(s,UTF)) {
5146 s = force_word(s,METHOD,FALSE,TRUE);
5154 if (PL_expect == XOPERATOR) {
5155 if (*s == '=' && !PL_lex_allbrackets &&
5156 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5163 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5165 OPERATOR('-'); /* unary minus */
5171 const char tmp = *s++;
5174 if (PL_expect == XOPERATOR)
5179 if (PL_expect == XOPERATOR) {
5180 if (*s == '=' && !PL_lex_allbrackets &&
5181 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5188 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5195 if (PL_expect == XPOSTDEREF) POSTDEREF('*');
5196 if (PL_expect != XOPERATOR) {
5197 s = scan_ident(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5198 PL_expect = XOPERATOR;
5199 force_ident(PL_tokenbuf, '*');
5207 if (*s == '=' && !PL_lex_allbrackets &&
5208 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5214 if (*s == '=' && !PL_lex_allbrackets &&
5215 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5219 PL_parser->saw_infix_sigil = 1;
5224 if (PL_expect == XOPERATOR) {
5225 if (s[1] == '=' && !PL_lex_allbrackets &&
5226 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5229 PL_parser->saw_infix_sigil = 1;
5232 else if (PL_expect == XPOSTDEREF) POSTDEREF('%');
5233 PL_tokenbuf[0] = '%';
5234 s = scan_ident(s, PL_tokenbuf + 1,
5235 sizeof PL_tokenbuf - 1, FALSE);
5237 if (!PL_tokenbuf[1]) {
5240 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5242 PL_tokenbuf[0] = '@';
5244 PL_expect = XOPERATOR;
5245 force_ident_maybe_lex('%');
5250 bof = FEATURE_BITWISE_IS_ENABLED;
5251 if (bof && s[1] == '.')
5253 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5254 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5260 BOop(bof ? d == s-2 ? OP_SBIT_XOR : OP_NBIT_XOR : OP_BIT_XOR);
5262 if (PL_lex_brackets > 100)
5263 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5264 PL_lex_brackstack[PL_lex_brackets++] = 0;
5265 PL_lex_allbrackets++;
5267 const char tmp = *s++;
5272 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5274 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5277 Perl_ck_warner_d(aTHX_
5278 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
5279 "Smartmatch is experimental");
5283 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.') {
5285 BCop(OP_SCOMPLEMENT);
5287 BCop(bof ? OP_NCOMPLEMENT : OP_COMPLEMENT);
5289 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5296 goto just_a_word_zero_gv;
5302 switch (PL_expect) {
5304 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5306 PL_bufptr = s; /* update in case we back off */
5309 "Use of := for an empty attribute list is not allowed");
5316 PL_expect = XTERMBLOCK;
5320 while (isIDFIRST_lazy_if(s,UTF)) {
5323 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5324 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5325 if (tmp < 0) tmp = -tmp;
5340 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5342 d = scan_str(d,TRUE,TRUE,FALSE,NULL);
5343 COPLINE_SET_FROM_MULTI_END;
5345 /* MUST advance bufptr here to avoid bogus
5346 "at end of line" context messages from yyerror().
5348 PL_bufptr = s + len;
5349 yyerror("Unterminated attribute parameter in attribute list");
5353 return REPORT(0); /* EOF indicator */
5357 sv_catsv(sv, PL_lex_stuff);
5358 attrs = op_append_elem(OP_LIST, attrs,
5359 newSVOP(OP_CONST, 0, sv));
5360 SvREFCNT_dec_NN(PL_lex_stuff);
5361 PL_lex_stuff = NULL;
5364 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5366 if (PL_in_my == KEY_our) {
5367 deprecate(":unique");
5370 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5373 /* NOTE: any CV attrs applied here need to be part of
5374 the CVf_BUILTIN_ATTRS define in cv.h! */
5375 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5377 CvLVALUE_on(PL_compcv);
5379 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5381 deprecate(":locked");
5383 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5385 CvMETHOD_on(PL_compcv);
5387 else if (!PL_in_my && len == 5
5388 && strnEQ(SvPVX(sv), "const", len))
5391 Perl_ck_warner_d(aTHX_
5392 packWARN(WARN_EXPERIMENTAL__CONST_ATTR),
5393 ":const is experimental"
5395 CvANONCONST_on(PL_compcv);
5396 if (!CvANON(PL_compcv))
5397 yyerror(":const is not permitted on named "
5400 /* After we've set the flags, it could be argued that
5401 we don't need to do the attributes.pm-based setting
5402 process, and shouldn't bother appending recognized
5403 flags. To experiment with that, uncomment the
5404 following "else". (Note that's already been
5405 uncommented. That keeps the above-applied built-in
5406 attributes from being intercepted (and possibly
5407 rejected) by a package's attribute routines, but is
5408 justified by the performance win for the common case
5409 of applying only built-in attributes.) */
5411 attrs = op_append_elem(OP_LIST, attrs,
5412 newSVOP(OP_CONST, 0,
5416 if (*s == ':' && s[1] != ':')
5419 break; /* require real whitespace or :'s */
5420 /* XXX losing whitespace on sequential attributes here */
5423 if (*s != ';' && *s != '}' &&
5424 !(PL_expect == XOPERATOR
5425 ? (*s == '=' || *s == ')')
5426 : (*s == '{' || *s == '('))) {
5427 const char q = ((*s == '\'') ? '"' : '\'');
5428 /* If here for an expression, and parsed no attrs, back
5430 if (PL_expect == XOPERATOR && !attrs) {
5434 /* MUST advance bufptr here to avoid bogus "at end of line"
5435 context messages from yyerror().
5438 yyerror( (const char *)
5440 ? Perl_form(aTHX_ "Invalid separator character "
5441 "%c%c%c in attribute list", q, *s, q)
5442 : "Unterminated attribute list" ) );
5450 NEXTVAL_NEXTTOKE.opval = attrs;
5456 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5460 PL_lex_allbrackets--;
5464 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5465 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5469 PL_lex_allbrackets++;
5472 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5479 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5482 PL_lex_allbrackets--;
5488 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5491 if (PL_lex_brackets <= 0)
5492 /* diag_listed_as: Unmatched right %s bracket */
5493 yyerror("Unmatched right square bracket");
5496 PL_lex_allbrackets--;
5497 if (PL_lex_state == LEX_INTERPNORMAL) {
5498 if (PL_lex_brackets == 0) {
5499 if (*s == '-' && s[1] == '>')
5500 PL_lex_state = LEX_INTERPENDMAYBE;
5501 else if (*s != '[' && *s != '{')
5502 PL_lex_state = LEX_INTERPEND;
5509 if (PL_lex_brackets > 100) {
5510 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5512 switch (PL_expect) {
5515 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5516 PL_lex_allbrackets++;
5517 OPERATOR(HASHBRACK);
5519 while (s < PL_bufend && SPACE_OR_TAB(*s))
5522 PL_tokenbuf[0] = '\0';
5523 if (d < PL_bufend && *d == '-') {
5524 PL_tokenbuf[0] = '-';
5526 while (d < PL_bufend && SPACE_OR_TAB(*d))
5529 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5530 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5532 while (d < PL_bufend && SPACE_OR_TAB(*d))
5535 const char minus = (PL_tokenbuf[0] == '-');
5536 s = force_word(s + minus, WORD, FALSE, TRUE);
5544 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5545 PL_lex_allbrackets++;
5550 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5551 PL_lex_allbrackets++;
5555 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5556 PL_lex_allbrackets++;
5561 if (PL_oldoldbufptr == PL_last_lop)
5562 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5564 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5565 PL_lex_allbrackets++;
5568 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5570 /* This hack is to get the ${} in the message. */
5572 yyerror("syntax error");
5575 OPERATOR(HASHBRACK);
5577 if (PL_expect == XREF && PL_oldoldbufptr != PL_last_lop) {
5578 /* ${...} or @{...} etc., but not print {...}
5579 * Skip the disambiguation and treat this as a block.
5581 goto block_expectation;
5583 /* This hack serves to disambiguate a pair of curlies
5584 * as being a block or an anon hash. Normally, expectation
5585 * determines that, but in cases where we're not in a
5586 * position to expect anything in particular (like inside
5587 * eval"") we have to resolve the ambiguity. This code
5588 * covers the case where the first term in the curlies is a
5589 * quoted string. Most other cases need to be explicitly
5590 * disambiguated by prepending a "+" before the opening
5591 * curly in order to force resolution as an anon hash.
5593 * XXX should probably propagate the outer expectation
5594 * into eval"" to rely less on this hack, but that could
5595 * potentially break current behavior of eval"".
5599 if (*s == '\'' || *s == '"' || *s == '`') {
5600 /* common case: get past first string, handling escapes */
5601 for (t++; t < PL_bufend && *t != *s;)
5606 else if (*s == 'q') {
5609 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5610 && !isWORDCHAR(*t))))
5612 /* skip q//-like construct */
5614 char open, close, term;
5617 while (t < PL_bufend && isSPACE(*t))
5619 /* check for q => */
5620 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5621 OPERATOR(HASHBRACK);
5625 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5629 for (t++; t < PL_bufend; t++) {
5630 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5632 else if (*t == open)
5636 for (t++; t < PL_bufend; t++) {
5637 if (*t == '\\' && t+1 < PL_bufend)
5639 else if (*t == close && --brackets <= 0)
5641 else if (*t == open)
5648 /* skip plain q word */
5649 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5652 else if (isWORDCHAR_lazy_if(t,UTF)) {
5654 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5657 while (t < PL_bufend && isSPACE(*t))
5659 /* if comma follows first term, call it an anon hash */
5660 /* XXX it could be a comma expression with loop modifiers */
5661 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5662 || (*t == '=' && t[1] == '>')))
5663 OPERATOR(HASHBRACK);
5664 if (PL_expect == XREF)
5667 /* If there is an opening brace or 'sub:', treat it
5668 as a term to make ${{...}}{k} and &{sub:attr...}
5669 dwim. Otherwise, treat it as a statement, so
5670 map {no strict; ...} works.
5677 if (strnEQ(s, "sub", 3)) {
5688 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5694 pl_yylval.ival = CopLINE(PL_curcop);
5695 PL_copline = NOLINE; /* invalidate current command line number */
5696 TOKEN(formbrack ? '=' : '{');
5698 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5702 if (PL_lex_brackets <= 0)
5703 /* diag_listed_as: Unmatched right %s bracket */
5704 yyerror("Unmatched right curly bracket");
5706 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5707 PL_lex_allbrackets--;
5708 if (PL_lex_state == LEX_INTERPNORMAL) {
5709 if (PL_lex_brackets == 0) {
5710 if (PL_expect & XFAKEBRACK) {
5711 PL_expect &= XENUMMASK;
5712 PL_lex_state = LEX_INTERPEND;
5714 return yylex(); /* ignore fake brackets */
5716 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5717 && SvEVALED(PL_lex_repl))
5718 PL_lex_state = LEX_INTERPEND;
5719 else if (*s == '-' && s[1] == '>')
5720 PL_lex_state = LEX_INTERPENDMAYBE;
5721 else if (*s != '[' && *s != '{')
5722 PL_lex_state = LEX_INTERPEND;
5725 if (PL_expect & XFAKEBRACK) {
5726 PL_expect &= XENUMMASK;
5728 return yylex(); /* ignore fake brackets */
5730 force_next(formbrack ? '.' : '}');
5731 if (formbrack) LEAVE;
5732 if (formbrack == 2) { /* means . where arguments were expected */
5738 if (PL_expect == XPOSTDEREF) POSTDEREF('&');
5741 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5742 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5749 if (PL_expect == XOPERATOR) {
5750 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5751 && isIDFIRST_lazy_if(s,UTF))
5753 CopLINE_dec(PL_curcop);
5754 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5755 CopLINE_inc(PL_curcop);
5758 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.')
5760 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5761 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5767 PL_parser->saw_infix_sigil = 1;
5768 BAop(bof ? OP_NBIT_AND : OP_BIT_AND);
5774 PL_tokenbuf[0] = '&';
5775 s = scan_ident(s - 1, PL_tokenbuf + 1,
5776 sizeof PL_tokenbuf - 1, TRUE);
5777 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5778 if (PL_tokenbuf[1]) {
5779 force_ident_maybe_lex('&');
5788 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5789 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5797 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.')
5799 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5800 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5804 BOop(bof ? s == d ? OP_NBIT_OR : OP_SBIT_OR : OP_BIT_OR);
5808 const char tmp = *s++;
5810 if (!PL_lex_allbrackets &&
5811 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5818 if (!PL_lex_allbrackets &&
5819 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5827 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5828 && strchr("+-*/%.^&|<",tmp))
5829 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5830 "Reversed %c= operator",(int)tmp);
5832 if (PL_expect == XSTATE && isALPHA(tmp) &&
5833 (s == PL_linestart+1 || s[-2] == '\n') )
5835 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
5836 || PL_lex_state != LEX_NORMAL) {
5841 if (strnEQ(s,"=cut",4)) {
5855 PL_parser->in_pod = 1;
5859 if (PL_expect == XBLOCK) {
5861 #ifdef PERL_STRICT_CR
5862 while (SPACE_OR_TAB(*t))
5864 while (SPACE_OR_TAB(*t) || *t == '\r')
5867 if (*t == '\n' || *t == '#') {
5870 SAVEI8(PL_parser->form_lex_state);
5871 SAVEI32(PL_lex_formbrack);
5872 PL_parser->form_lex_state = PL_lex_state;
5873 PL_lex_formbrack = PL_lex_brackets + 1;
5877 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5886 const char tmp = *s++;
5888 /* was this !=~ where !~ was meant?
5889 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5891 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5892 const char *t = s+1;
5894 while (t < PL_bufend && isSPACE(*t))
5897 if (*t == '/' || *t == '?' ||
5898 ((*t == 'm' || *t == 's' || *t == 'y')
5899 && !isWORDCHAR(t[1])) ||
5900 (*t == 't' && t[1] == 'r' && !isWORDCHAR(t[2])))
5901 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5902 "!=~ should be !~");
5904 if (!PL_lex_allbrackets &&
5905 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5917 if (PL_expect != XOPERATOR) {
5918 if (s[1] != '<' && !strchr(s,'>'))
5920 if (s[1] == '<' && s[2] != '>')
5921 s = scan_heredoc(s);
5923 s = scan_inputsymbol(s);
5924 PL_expect = XOPERATOR;
5925 TOKEN(sublex_start());
5931 if (*s == '=' && !PL_lex_allbrackets &&
5932 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5936 SHop(OP_LEFT_SHIFT);
5941 if (!PL_lex_allbrackets &&
5942 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5949 if (!PL_lex_allbrackets &&
5950 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5958 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5966 const char tmp = *s++;
5968 if (*s == '=' && !PL_lex_allbrackets &&
5969 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5973 SHop(OP_RIGHT_SHIFT);
5975 else if (tmp == '=') {
5976 if (!PL_lex_allbrackets &&
5977 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5985 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5994 if (PL_expect == XOPERATOR) {
5995 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5996 return deprecate_commaless_var_list();
5999 else if (PL_expect == XPOSTDEREF) {
6002 POSTDEREF(DOLSHARP);
6007 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6008 PL_tokenbuf[0] = '@';
6009 s = scan_ident(s + 1, PL_tokenbuf + 1,
6010 sizeof PL_tokenbuf - 1, FALSE);
6011 if (PL_expect == XOPERATOR) {
6013 if (PL_bufptr > s) {
6015 PL_bufptr = PL_oldbufptr;
6017 no_op("Array length", d);
6019 if (!PL_tokenbuf[1])
6021 PL_expect = XOPERATOR;
6022 force_ident_maybe_lex('#');
6026 PL_tokenbuf[0] = '$';
6027 s = scan_ident(s, PL_tokenbuf + 1,
6028 sizeof PL_tokenbuf - 1, FALSE);
6029 if (PL_expect == XOPERATOR) {
6031 if (PL_bufptr > s) {
6033 PL_bufptr = PL_oldbufptr;
6037 if (!PL_tokenbuf[1]) {
6039 yyerror("Final $ should be \\$ or $name");
6045 const char tmp = *s;
6046 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6049 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6050 && intuit_more(s)) {
6052 PL_tokenbuf[0] = '@';
6053 if (ckWARN(WARN_SYNTAX)) {
6056 while (isSPACE(*t) || isWORDCHAR_lazy_if(t,UTF) || *t == '$')
6057 t += UTF ? UTF8SKIP(t) : 1;
6059 PL_bufptr = skipspace(PL_bufptr); /* XXX can realloc */
6060 while (t < PL_bufend && *t != ']')
6062 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6063 "Multidimensional syntax %"UTF8f" not supported",
6064 UTF8fARG(UTF,(int)((t - PL_bufptr) + 1), PL_bufptr));
6068 else if (*s == '{') {
6070 PL_tokenbuf[0] = '%';
6071 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6072 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6074 char tmpbuf[sizeof PL_tokenbuf];
6077 } while (isSPACE(*t));
6078 if (isIDFIRST_lazy_if(t,UTF)) {
6080 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6085 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6086 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6087 "You need to quote \"%"UTF8f"\"",
6088 UTF8fARG(UTF, len, tmpbuf));
6094 PL_expect = XOPERATOR;
6095 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6096 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6097 if (!islop || PL_last_lop_op == OP_GREPSTART)
6098 PL_expect = XOPERATOR;
6099 else if (strchr("$@\"'`q", *s))
6100 PL_expect = XTERM; /* e.g. print $fh "foo" */
6101 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6102 PL_expect = XTERM; /* e.g. print $fh &sub */
6103 else if (isIDFIRST_lazy_if(s,UTF)) {
6104 char tmpbuf[sizeof PL_tokenbuf];
6106 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6107 if ((t2 = keyword(tmpbuf, len, 0))) {
6108 /* binary operators exclude handle interpretations */
6120 PL_expect = XTERM; /* e.g. print $fh length() */
6125 PL_expect = XTERM; /* e.g. print $fh subr() */
6128 else if (isDIGIT(*s))
6129 PL_expect = XTERM; /* e.g. print $fh 3 */
6130 else if (*s == '.' && isDIGIT(s[1]))
6131 PL_expect = XTERM; /* e.g. print $fh .3 */
6132 else if ((*s == '?' || *s == '-' || *s == '+')
6133 && !isSPACE(s[1]) && s[1] != '=')
6134 PL_expect = XTERM; /* e.g. print $fh -1 */
6135 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6137 PL_expect = XTERM; /* e.g. print $fh /.../
6138 XXX except DORDOR operator
6140 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6142 PL_expect = XTERM; /* print $fh <<"EOF" */
6145 force_ident_maybe_lex('$');
6149 if (PL_expect == XOPERATOR)
6151 else if (PL_expect == XPOSTDEREF) POSTDEREF('@');
6152 PL_tokenbuf[0] = '@';
6153 s = scan_ident(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6155 if (!PL_tokenbuf[1]) {
6158 if (PL_lex_state == LEX_NORMAL)
6160 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6162 PL_tokenbuf[0] = '%';
6164 /* Warn about @ where they meant $. */
6165 if (*s == '[' || *s == '{') {
6166 if (ckWARN(WARN_SYNTAX)) {
6167 S_check_scalar_slice(aTHX_ s);
6171 PL_expect = XOPERATOR;
6172 force_ident_maybe_lex('@');
6175 case '/': /* may be division, defined-or, or pattern */
6176 if ((PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR) && s[1] == '/') {
6177 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6178 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6183 else if (PL_expect == XOPERATOR) {
6185 if (*s == '=' && !PL_lex_allbrackets &&
6186 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6193 /* Disable warning on "study /blah/" */
6194 if (PL_oldoldbufptr == PL_last_uni
6195 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6196 || memNE(PL_last_uni, "study", 5)
6197 || isWORDCHAR_lazy_if(PL_last_uni+5,UTF)
6200 s = scan_pat(s,OP_MATCH);
6201 TERM(sublex_start());
6204 case '?': /* conditional */
6206 if (!PL_lex_allbrackets &&
6207 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6211 PL_lex_allbrackets++;
6215 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6216 #ifdef PERL_STRICT_CR
6219 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6221 && (s == PL_linestart || s[-1] == '\n') )
6224 formbrack = 2; /* dot seen where arguments expected */
6227 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6231 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6234 if (!PL_lex_allbrackets &&
6235 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6242 pl_yylval.ival = OPf_SPECIAL;
6248 if (*s == '=' && !PL_lex_allbrackets &&
6249 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6256 case '0': case '1': case '2': case '3': case '4':
6257 case '5': case '6': case '7': case '8': case '9':
6258 s = scan_num(s, &pl_yylval);
6259 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6260 if (PL_expect == XOPERATOR)
6265 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6268 COPLINE_SET_FROM_MULTI_END;
6269 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6270 if (PL_expect == XOPERATOR) {
6271 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6272 return deprecate_commaless_var_list();
6277 pl_yylval.ival = OP_CONST;
6278 TERM(sublex_start());
6281 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6284 printbuf("### Saw string before %s\n", s);
6286 PerlIO_printf(Perl_debug_log,
6287 "### Saw unterminated string\n");
6289 if (PL_expect == XOPERATOR) {
6290 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6291 return deprecate_commaless_var_list();
6298 pl_yylval.ival = OP_CONST;
6299 /* FIXME. I think that this can be const if char *d is replaced by
6300 more localised variables. */
6301 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6302 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6303 pl_yylval.ival = OP_STRINGIFY;
6307 if (pl_yylval.ival == OP_CONST)
6308 COPLINE_SET_FROM_MULTI_END;
6309 TERM(sublex_start());
6312 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6315 printbuf("### Saw backtick string before %s\n", s);
6317 PerlIO_printf(Perl_debug_log,
6318 "### Saw unterminated backtick string\n");
6320 if (PL_expect == XOPERATOR)
6321 no_op("Backticks",s);
6324 pl_yylval.ival = OP_BACKTICK;
6325 TERM(sublex_start());
6329 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6331 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6333 if (PL_expect == XOPERATOR)
6334 no_op("Backslash",s);
6338 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6339 char *start = s + 2;
6340 while (isDIGIT(*start) || *start == '_')
6342 if (*start == '.' && isDIGIT(start[1])) {
6343 s = scan_num(s, &pl_yylval);
6346 else if ((*start == ':' && start[1] == ':')
6347 || (PL_expect == XSTATE && *start == ':'))
6349 else if (PL_expect == XSTATE) {
6351 while (d < PL_bufend && isSPACE(*d)) d++;
6352 if (*d == ':') goto keylookup;
6354 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6355 if (!isALPHA(*start) && (PL_expect == XTERM
6356 || PL_expect == XREF || PL_expect == XSTATE
6357 || PL_expect == XTERMORDORDOR)) {
6358 GV *const gv = gv_fetchpvn_flags(s, start - s,
6359 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6361 s = scan_num(s, &pl_yylval);
6368 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6421 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6423 /* Some keywords can be followed by any delimiter, including ':' */
6424 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6426 /* x::* is just a word, unless x is "CORE" */
6427 if (!anydelim && *s == ':' && s[1] == ':') {
6428 if (strEQ(PL_tokenbuf, "CORE")) goto case_KEY_CORE;
6433 while (d < PL_bufend && isSPACE(*d))
6434 d++; /* no comments skipped here, or s### is misparsed */
6436 /* Is this a word before a => operator? */
6437 if (*d == '=' && d[1] == '>') {
6441 = (OP*)newSVOP(OP_CONST, 0,
6442 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6443 pl_yylval.opval->op_private = OPpCONST_BARE;
6447 /* Check for plugged-in keyword */
6451 char *saved_bufptr = PL_bufptr;
6453 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6455 if (result == KEYWORD_PLUGIN_DECLINE) {
6456 /* not a plugged-in keyword */
6457 PL_bufptr = saved_bufptr;
6458 } else if (result == KEYWORD_PLUGIN_STMT) {
6459 pl_yylval.opval = o;
6461 if (!PL_nexttoke) PL_expect = XSTATE;
6462 return REPORT(PLUGSTMT);
6463 } else if (result == KEYWORD_PLUGIN_EXPR) {
6464 pl_yylval.opval = o;
6466 if (!PL_nexttoke) PL_expect = XOPERATOR;
6467 return REPORT(PLUGEXPR);
6469 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6474 /* Check for built-in keyword */
6475 tmp = keyword(PL_tokenbuf, len, 0);
6477 /* Is this a label? */
6478 if (!anydelim && PL_expect == XSTATE
6479 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6481 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6482 pl_yylval.pval[len] = '\0';
6483 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6488 /* Check for lexical sub */
6489 if (PL_expect != XOPERATOR) {
6490 char tmpbuf[sizeof PL_tokenbuf + 1];
6492 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6493 off = pad_findmy_pvn(tmpbuf, len+1, 0);
6494 if (off != NOT_IN_PAD) {
6495 assert(off); /* we assume this is boolean-true below */
6496 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6497 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6498 HEK * const stashname = HvNAME_HEK(stash);
6499 sv = newSVhek(stashname);
6500 sv_catpvs(sv, "::");
6501 sv_catpvn_flags(sv, PL_tokenbuf, len,
6502 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6503 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6513 rv2cv_op = newOP(OP_PADANY, 0);
6514 rv2cv_op->op_targ = off;
6515 cv = find_lexical_cv(off);
6523 if (tmp < 0) { /* second-class keyword? */
6524 GV *ogv = NULL; /* override (winner) */
6525 GV *hgv = NULL; /* hidden (loser) */
6526 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6528 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6529 (UTF ? SVf_UTF8 : 0)|GV_NOTQUAL,
6533 if (GvIMPORTED_CV(gv))
6535 else if (! CvMETHOD(cv))
6539 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6543 ? GvCVu(gv) && GvIMPORTED_CV(gv)
6544 : SvPCS_IMPORTED(gv)
6545 && (gv_init(gv, PL_globalstash, PL_tokenbuf,
6554 tmp = 0; /* overridden by import or by GLOBAL */
6557 && -tmp==KEY_lock /* XXX generalizable kludge */
6560 tmp = 0; /* any sub overrides "weak" keyword */
6562 else { /* no override */
6564 if (tmp == KEY_dump) {
6565 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6566 "dump() better written as CORE::dump()");
6570 if (hgv && tmp != KEY_x) /* never ambiguous */
6571 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6572 "Ambiguous call resolved as CORE::%s(), "
6573 "qualify as such or use &",
6578 if (tmp && tmp != KEY___DATA__ && tmp != KEY___END__
6579 && (!anydelim || *s != '#')) {
6580 /* no override, and not s### either; skipspace is safe here
6581 * check for => on following line */
6583 STRLEN bufoff = PL_bufptr - SvPVX(PL_linestr);
6584 STRLEN soff = s - SvPVX(PL_linestr);
6585 s = skipspace_flags(s, LEX_NO_INCLINE);
6586 arrow = *s == '=' && s[1] == '>';
6587 PL_bufptr = SvPVX(PL_linestr) + bufoff;
6588 s = SvPVX(PL_linestr) + soff;
6596 default: /* not a keyword */
6597 /* Trade off - by using this evil construction we can pull the
6598 variable gv into the block labelled keylookup. If not, then
6599 we have to give it function scope so that the goto from the
6600 earlier ':' case doesn't bypass the initialisation. */
6602 just_a_word_zero_gv:
6614 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6618 /* Get the rest if it looks like a package qualifier */
6620 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6622 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6625 Perl_croak(aTHX_ "Bad name after %"UTF8f"%s",
6626 UTF8fARG(UTF, len, PL_tokenbuf),
6627 *s == '\'' ? "'" : "::");
6632 if (PL_expect == XOPERATOR) {
6633 if (PL_bufptr == PL_linestart) {
6634 CopLINE_dec(PL_curcop);
6635 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6636 CopLINE_inc(PL_curcop);
6639 no_op("Bareword",s);
6642 /* See if the name is "Foo::",
6643 in which case Foo is a bareword
6644 (and a package name). */
6647 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6649 if (ckWARN(WARN_BAREWORD)
6650 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6651 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6652 "Bareword \"%"UTF8f"\" refers to nonexistent package",
6653 UTF8fARG(UTF, len, PL_tokenbuf));
6655 PL_tokenbuf[len] = '\0';
6664 /* if we saw a global override before, get the right name */
6667 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6670 SV * const tmp_sv = sv;
6671 sv = newSVpvs("CORE::GLOBAL::");
6672 sv_catsv(sv, tmp_sv);
6673 SvREFCNT_dec(tmp_sv);
6677 /* Presume this is going to be a bareword of some sort. */
6679 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6680 pl_yylval.opval->op_private = OPpCONST_BARE;
6682 /* And if "Foo::", then that's what it certainly is. */
6688 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6689 const_op->op_private = OPpCONST_BARE;
6691 newCVREF(OPpMAY_RETURN_CONSTANT<<8, const_op);
6695 : SvROK(gv) && SvTYPE(SvRV(gv)) == SVt_PVCV
6698 : rv2cv_op_cv(rv2cv_op, RV2CVOPCV_RETURN_STUB);
6701 /* Use this var to track whether intuit_method has been
6702 called. intuit_method returns 0 or > 255. */
6705 /* See if it's the indirect object for a list operator. */
6707 if (PL_oldoldbufptr &&
6708 PL_oldoldbufptr < PL_bufptr &&
6709 (PL_oldoldbufptr == PL_last_lop
6710 || PL_oldoldbufptr == PL_last_uni) &&
6711 /* NO SKIPSPACE BEFORE HERE! */
6712 (PL_expect == XREF ||
6713 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6715 bool immediate_paren = *s == '(';
6717 /* (Now we can afford to cross potential line boundary.) */
6720 /* Two barewords in a row may indicate method call. */
6722 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6723 (tmp = intuit_method(s, lex ? NULL : sv, cv))) {
6727 /* If not a declared subroutine, it's an indirect object. */
6728 /* (But it's an indir obj regardless for sort.) */
6729 /* Also, if "_" follows a filetest operator, it's a bareword */
6732 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6734 (PL_last_lop_op != OP_MAPSTART &&
6735 PL_last_lop_op != OP_GREPSTART))))
6736 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6737 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6740 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6745 PL_expect = XOPERATOR;
6748 /* Is this a word before a => operator? */
6749 if (*s == '=' && s[1] == '>' && !pkgname) {
6752 if (gvp || (lex && !off)) {
6753 assert (cSVOPx(pl_yylval.opval)->op_sv == sv);
6754 /* This is our own scalar, created a few lines
6755 above, so this is safe. */
6757 sv_setpv(sv, PL_tokenbuf);
6758 if (UTF && !IN_BYTES
6759 && is_utf8_string((U8*)PL_tokenbuf, len))
6766 /* If followed by a paren, it's certainly a subroutine. */
6771 while (SPACE_OR_TAB(*d))
6773 if (*d == ')' && (sv = cv_const_sv_or_av(cv))) {
6778 NEXTVAL_NEXTTOKE.opval =
6779 off ? rv2cv_op : pl_yylval.opval;
6781 op_free(pl_yylval.opval), force_next(PRIVATEREF);
6782 else op_free(rv2cv_op), force_next(WORD);
6787 /* If followed by var or block, call it a method (unless sub) */
6789 if ((*s == '$' || *s == '{') && !cv) {
6791 PL_last_lop = PL_oldbufptr;
6792 PL_last_lop_op = OP_METHOD;
6793 if (!PL_lex_allbrackets &&
6794 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6795 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6796 PL_expect = XBLOCKTERM;
6798 return REPORT(METHOD);
6801 /* If followed by a bareword, see if it looks like indir obj. */
6803 if (tmp == 1 && !orig_keyword
6804 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6805 && (tmp = intuit_method(s, lex ? NULL : sv, cv))) {
6808 assert(cSVOPx(pl_yylval.opval)->op_sv == sv);
6810 sv_setpvn(sv, PL_tokenbuf, len);
6811 if (UTF && !IN_BYTES
6812 && is_utf8_string((U8*)PL_tokenbuf, len))
6814 else SvUTF8_off(sv);
6817 if (tmp == METHOD && !PL_lex_allbrackets &&
6818 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6819 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6823 /* Not a method, so call it a subroutine (if defined) */
6826 /* Check for a constant sub */
6827 if ((sv = cv_const_sv_or_av(cv))) {
6830 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6831 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6832 if (SvTYPE(sv) == SVt_PVAV)
6833 pl_yylval.opval = newUNOP(OP_RV2AV, OPf_PARENS,
6836 pl_yylval.opval->op_private = 0;
6837 pl_yylval.opval->op_folded = 1;
6838 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6843 op_free(pl_yylval.opval);
6845 off ? (OP *)newCVREF(0, rv2cv_op) : rv2cv_op;
6846 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6847 PL_last_lop = PL_oldbufptr;
6848 PL_last_lop_op = OP_ENTERSUB;
6849 /* Is there a prototype? */
6853 STRLEN protolen = CvPROTOLEN(cv);
6854 const char *proto = CvPROTO(cv);
6856 proto = S_strip_spaces(aTHX_ proto, &protolen);
6859 if ((optional = *proto == ';'))
6862 while (*proto == ';');
6866 *proto == '$' || *proto == '_'
6867 || *proto == '*' || *proto == '+'
6872 *proto == '\\' && proto[1] && proto[2] == '\0'
6875 UNIPROTO(UNIOPSUB,optional);
6876 if (*proto == '\\' && proto[1] == '[') {
6877 const char *p = proto + 2;
6878 while(*p && *p != ']')
6880 if(*p == ']' && !p[1])
6881 UNIPROTO(UNIOPSUB,optional);
6883 if (*proto == '&' && *s == '{') {
6885 sv_setpvs(PL_subname, "__ANON__");
6887 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6888 if (!PL_lex_allbrackets &&
6889 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6890 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6894 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6896 force_next(off ? PRIVATEREF : WORD);
6897 if (!PL_lex_allbrackets &&
6898 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6899 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6903 /* Call it a bare word */
6905 if (PL_hints & HINT_STRICT_SUBS)
6906 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6909 /* after "print" and similar functions (corresponding to
6910 * "F? L" in opcode.pl), whatever wasn't already parsed as
6911 * a filehandle should be subject to "strict subs".
6912 * Likewise for the optional indirect-object argument to system
6913 * or exec, which can't be a bareword */
6914 if ((PL_last_lop_op == OP_PRINT
6915 || PL_last_lop_op == OP_PRTF
6916 || PL_last_lop_op == OP_SAY
6917 || PL_last_lop_op == OP_SYSTEM
6918 || PL_last_lop_op == OP_EXEC)
6919 && (PL_hints & HINT_STRICT_SUBS))
6920 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6921 if (lastchar != '-') {
6922 if (ckWARN(WARN_RESERVED)) {
6926 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
6928 /* PL_warn_reserved is constant */
6929 GCC_DIAG_IGNORE(-Wformat-nonliteral);
6930 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6940 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
6941 && saw_infix_sigil) {
6942 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6943 "Operator or semicolon missing before %c%"UTF8f,
6945 UTF8fARG(UTF, strlen(PL_tokenbuf),
6947 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6948 "Ambiguous use of %c resolved as operator %c",
6949 lastchar, lastchar);
6956 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
6961 (OP*)newSVOP(OP_CONST, 0,
6962 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
6965 case KEY___PACKAGE__:
6967 (OP*)newSVOP(OP_CONST, 0,
6969 ? newSVhek(HvNAME_HEK(PL_curstash))
6976 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6977 HV * const stash = PL_tokenbuf[2] == 'D' && PL_curstash
6980 gv = (GV *)*hv_fetchs(stash, "DATA", 1);
6982 gv_init(gv,stash,"DATA",4,0);
6985 GvIOp(gv) = newIO();
6986 IoIFP(GvIOp(gv)) = PL_rsfp;
6987 #if defined(HAS_FCNTL) && defined(F_SETFD)
6989 const int fd = PerlIO_fileno(PL_rsfp);
6990 fcntl(fd,F_SETFD,fd >= 3);
6993 /* Mark this internal pseudo-handle as clean */
6994 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6995 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6996 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6998 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6999 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7000 /* if the script was opened in binmode, we need to revert
7001 * it to text mode for compatibility; but only iff it has CRs
7002 * XXX this is a questionable hack at best. */
7003 if (PL_bufend-PL_bufptr > 2
7004 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7007 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7008 loc = PerlIO_tell(PL_rsfp);
7009 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7012 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7014 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7015 #endif /* NETWARE */
7017 PerlIO_seek(PL_rsfp, loc, 0);
7021 #ifdef PERLIO_LAYERS
7024 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7025 else if (IN_ENCODING) {
7031 XPUSHs(_get_encoding());
7033 call_method("name", G_SCALAR);
7037 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7038 Perl_form(aTHX_ ":encoding(%"SVf")",
7051 FUN0OP(CvCLONE(PL_compcv)
7052 ? newOP(OP_RUNCV, 0)
7053 : newPVOP(OP_RUNCV,0,NULL));
7062 if (PL_expect == XSTATE) {
7073 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7074 if ((*s == ':' && s[1] == ':')
7075 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7079 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7083 Perl_croak(aTHX_ "CORE::%"UTF8f" is not a keyword",
7084 UTF8fARG(UTF, len, PL_tokenbuf));
7087 else if (tmp == KEY_require || tmp == KEY_do
7089 /* that's a way to remember we saw "CORE::" */
7101 LOP(OP_ACCEPT,XTERM);
7104 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7109 LOP(OP_ATAN2,XTERM);
7115 LOP(OP_BINMODE,XTERM);
7118 LOP(OP_BLESS,XTERM);
7127 /* We have to disambiguate the two senses of
7128 "continue". If the next token is a '{' then
7129 treat it as the start of a continue block;
7130 otherwise treat it as a control operator.
7140 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7150 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7159 if (!PL_cryptseen) {
7160 PL_cryptseen = TRUE;
7164 LOP(OP_CRYPT,XTERM);
7167 LOP(OP_CHMOD,XTERM);
7170 LOP(OP_CHOWN,XTERM);
7173 LOP(OP_CONNECT,XTERM);
7193 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7195 if (len && (len != 4 || strNE(PL_tokenbuf+1, "CORE"))
7196 && !keyword(PL_tokenbuf + 1, len, 0)) {
7199 force_ident_maybe_lex('&');
7204 if (orig_keyword == KEY_do) {
7213 PL_hints |= HINT_BLOCK_SCOPE;
7223 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7224 STR_WITH_LEN("NDBM_File::"),
7225 STR_WITH_LEN("DB_File::"),
7226 STR_WITH_LEN("GDBM_File::"),
7227 STR_WITH_LEN("SDBM_File::"),
7228 STR_WITH_LEN("ODBM_File::"),
7230 LOP(OP_DBMOPEN,XTERM);
7242 pl_yylval.ival = CopLINE(PL_curcop);
7246 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7258 if (*s == '{') { /* block eval */
7259 PL_expect = XTERMBLOCK;
7260 UNIBRACK(OP_ENTERTRY);
7262 else { /* string eval */
7264 UNIBRACK(OP_ENTEREVAL);
7269 UNIBRACK(-OP_ENTEREVAL);
7283 case KEY_endhostent:
7289 case KEY_endservent:
7292 case KEY_endprotoent:
7303 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7305 pl_yylval.ival = CopLINE(PL_curcop);
7307 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7310 if ((PL_bufend - p) >= 3 &&
7311 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7313 else if ((PL_bufend - p) >= 4 &&
7314 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7317 /* skip optional package name, as in "for my abc $x (..)" */
7318 if (isIDFIRST_lazy_if(p,UTF)) {
7319 p = scan_word(p, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7323 Perl_croak(aTHX_ "Missing $ on loop variable");
7328 LOP(OP_FORMLINE,XTERM);
7337 LOP(OP_FCNTL,XTERM);
7343 LOP(OP_FLOCK,XTERM);
7346 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7351 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7356 LOP(OP_GREPSTART, XREF);
7373 case KEY_getpriority:
7374 LOP(OP_GETPRIORITY,XTERM);
7376 case KEY_getprotobyname:
7379 case KEY_getprotobynumber:
7380 LOP(OP_GPBYNUMBER,XTERM);
7382 case KEY_getprotoent:
7394 case KEY_getpeername:
7395 UNI(OP_GETPEERNAME);
7397 case KEY_gethostbyname:
7400 case KEY_gethostbyaddr:
7401 LOP(OP_GHBYADDR,XTERM);
7403 case KEY_gethostent:
7406 case KEY_getnetbyname:
7409 case KEY_getnetbyaddr:
7410 LOP(OP_GNBYADDR,XTERM);
7415 case KEY_getservbyname:
7416 LOP(OP_GSBYNAME,XTERM);
7418 case KEY_getservbyport:
7419 LOP(OP_GSBYPORT,XTERM);
7421 case KEY_getservent:
7424 case KEY_getsockname:
7425 UNI(OP_GETSOCKNAME);
7427 case KEY_getsockopt:
7428 LOP(OP_GSOCKOPT,XTERM);
7443 pl_yylval.ival = CopLINE(PL_curcop);
7444 Perl_ck_warner_d(aTHX_
7445 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
7446 "given is experimental");
7451 orig_keyword==KEY_glob ? -OP_GLOB : OP_GLOB,
7459 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7461 pl_yylval.ival = CopLINE(PL_curcop);
7465 LOP(OP_INDEX,XTERM);
7471 LOP(OP_IOCTL,XTERM);
7499 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7504 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7518 LOP(OP_LISTEN,XTERM);
7527 s = scan_pat(s,OP_MATCH);
7528 TERM(sublex_start());
7531 LOP(OP_MAPSTART, XREF);
7534 LOP(OP_MKDIR,XTERM);
7537 LOP(OP_MSGCTL,XTERM);
7540 LOP(OP_MSGGET,XTERM);
7543 LOP(OP_MSGRCV,XTERM);
7546 LOP(OP_MSGSND,XTERM);
7551 PL_in_my = (U16)tmp;
7553 if (isIDFIRST_lazy_if(s,UTF)) {
7554 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7555 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7557 if (!FEATURE_LEXSUBS_IS_ENABLED)
7559 "Experimental \"%s\" subs not enabled",
7560 tmp == KEY_my ? "my" :
7561 tmp == KEY_state ? "state" : "our");
7562 Perl_ck_warner_d(aTHX_
7563 packWARN(WARN_EXPERIMENTAL__LEXICAL_SUBS),
7564 "The lexical_subs feature is experimental");
7567 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7568 if (!PL_in_my_stash) {
7572 len = my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7573 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(tmpbuf));
7574 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7584 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7589 s = tokenize_use(0, s);
7593 if (*s == '(' || (s = skipspace(s), *s == '('))
7596 if (!PL_lex_allbrackets &&
7597 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7598 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7604 if (isIDFIRST_lazy_if(s,UTF)) {
7606 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE,
7608 for (t=d; isSPACE(*t);)
7610 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7612 && !(t[0] == '=' && t[1] == '>')
7613 && !(t[0] == ':' && t[1] == ':')
7614 && !keyword(s, d-s, 0)
7616 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7617 "Precedence problem: open %"UTF8f" should be open(%"UTF8f")",
7618 UTF8fARG(UTF, d-s, s), UTF8fARG(UTF, d-s, s));
7624 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7626 pl_yylval.ival = OP_OR;
7636 LOP(OP_OPEN_DIR,XTERM);
7639 checkcomma(s,PL_tokenbuf,"filehandle");
7643 checkcomma(s,PL_tokenbuf,"filehandle");
7662 s = force_word(s,WORD,FALSE,TRUE);
7664 s = force_strict_version(s);
7668 LOP(OP_PIPE_OP,XTERM);
7671 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7674 COPLINE_SET_FROM_MULTI_END;
7675 pl_yylval.ival = OP_CONST;
7676 TERM(sublex_start());
7683 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7686 COPLINE_SET_FROM_MULTI_END;
7687 PL_expect = XOPERATOR;
7688 if (SvCUR(PL_lex_stuff)) {
7689 int warned_comma = !ckWARN(WARN_QW);
7690 int warned_comment = warned_comma;
7691 d = SvPV_force(PL_lex_stuff, len);
7693 for (; isSPACE(*d) && len; --len, ++d)
7698 if (!warned_comma || !warned_comment) {
7699 for (; !isSPACE(*d) && len; --len, ++d) {
7700 if (!warned_comma && *d == ',') {
7701 Perl_warner(aTHX_ packWARN(WARN_QW),
7702 "Possible attempt to separate words with commas");
7705 else if (!warned_comment && *d == '#') {
7706 Perl_warner(aTHX_ packWARN(WARN_QW),
7707 "Possible attempt to put comments in qw() list");
7713 for (; !isSPACE(*d) && len; --len, ++d)
7716 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7717 words = op_append_elem(OP_LIST, words,
7718 newSVOP(OP_CONST, 0, tokeq(sv)));
7723 words = newNULLLIST();
7724 SvREFCNT_dec_NN(PL_lex_stuff);
7725 PL_lex_stuff = NULL;
7726 PL_expect = XOPERATOR;
7727 pl_yylval.opval = sawparens(words);
7732 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7735 pl_yylval.ival = OP_STRINGIFY;
7736 if (SvIVX(PL_lex_stuff) == '\'')
7737 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7738 TERM(sublex_start());
7741 s = scan_pat(s,OP_QR);
7742 TERM(sublex_start());
7745 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7748 pl_yylval.ival = OP_BACKTICK;
7749 TERM(sublex_start());
7757 s = force_version(s, FALSE);
7759 else if (*s != 'v' || !isDIGIT(s[1])
7760 || (s = force_version(s, TRUE), *s == 'v'))
7762 *PL_tokenbuf = '\0';
7763 s = force_word(s,WORD,TRUE,TRUE);
7764 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7765 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
7766 GV_ADD | (UTF ? SVf_UTF8 : 0));
7768 yyerror("<> at require-statement should be quotes");
7770 if (orig_keyword == KEY_require) {
7776 PL_expect = PL_nexttoke ? XOPERATOR : XTERM;
7778 PL_last_uni = PL_oldbufptr;
7779 PL_last_lop_op = OP_REQUIRE;
7781 return REPORT( (int)REQUIRE );
7790 LOP(OP_RENAME,XTERM);
7799 LOP(OP_RINDEX,XTERM);
7808 UNIDOR(OP_READLINE);
7811 UNIDOR(OP_BACKTICK);
7820 LOP(OP_REVERSE,XTERM);
7823 UNIDOR(OP_READLINK);
7830 if (pl_yylval.opval)
7831 TERM(sublex_start());
7833 TOKEN(1); /* force error */
7836 checkcomma(s,PL_tokenbuf,"filehandle");
7846 LOP(OP_SELECT,XTERM);
7852 LOP(OP_SEMCTL,XTERM);
7855 LOP(OP_SEMGET,XTERM);
7858 LOP(OP_SEMOP,XTERM);
7864 LOP(OP_SETPGRP,XTERM);
7866 case KEY_setpriority:
7867 LOP(OP_SETPRIORITY,XTERM);
7869 case KEY_sethostent:
7875 case KEY_setservent:
7878 case KEY_setprotoent:
7888 LOP(OP_SEEKDIR,XTERM);
7890 case KEY_setsockopt:
7891 LOP(OP_SSOCKOPT,XTERM);
7897 LOP(OP_SHMCTL,XTERM);
7900 LOP(OP_SHMGET,XTERM);
7903 LOP(OP_SHMREAD,XTERM);
7906 LOP(OP_SHMWRITE,XTERM);
7909 LOP(OP_SHUTDOWN,XTERM);
7918 LOP(OP_SOCKET,XTERM);
7920 case KEY_socketpair:
7921 LOP(OP_SOCKPAIR,XTERM);
7924 checkcomma(s,PL_tokenbuf,"subroutine name");
7927 s = force_word(s,WORD,TRUE,TRUE);
7931 LOP(OP_SPLIT,XTERM);
7934 LOP(OP_SPRINTF,XTERM);
7937 LOP(OP_SPLICE,XTERM);
7952 LOP(OP_SUBSTR,XTERM);
7958 char * const tmpbuf = PL_tokenbuf + 1;
7959 expectation attrful;
7960 bool have_name, have_proto;
7961 const int key = tmp;
7962 SV *format_name = NULL;
7967 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7968 (*s == ':' && s[1] == ':'))
7972 attrful = XATTRBLOCK;
7973 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
7975 if (key == KEY_format)
7976 format_name = S_newSV_maybe_utf8(aTHX_ s, d - s);
7978 if (memchr(tmpbuf, ':', len) || key != KEY_sub
7980 PL_tokenbuf, len + 1, 0
7982 sv_setpvn(PL_subname, tmpbuf, len);
7984 sv_setsv(PL_subname,PL_curstname);
7985 sv_catpvs(PL_subname,"::");
7986 sv_catpvn(PL_subname,tmpbuf,len);
7988 if (SvUTF8(PL_linestr))
7989 SvUTF8_on(PL_subname);
7996 if (key == KEY_my || key == KEY_our || key==KEY_state)
7999 /* diag_listed_as: Missing name in "%s sub" */
8001 "Missing name in \"%s\"", PL_bufptr);
8003 PL_expect = XTERMBLOCK;
8004 attrful = XATTRTERM;
8005 sv_setpvs(PL_subname,"?");
8009 if (key == KEY_format) {
8011 NEXTVAL_NEXTTOKE.opval
8012 = (OP*)newSVOP(OP_CONST,0, format_name);
8013 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
8019 /* Look for a prototype */
8020 if (*s == '(' && !FEATURE_SIGNATURES_IS_ENABLED) {
8021 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8022 COPLINE_SET_FROM_MULTI_END;
8024 Perl_croak(aTHX_ "Prototype not terminated");
8025 (void)validate_proto(PL_subname, PL_lex_stuff, ckWARN(WARN_ILLEGALPROTO));
8033 if (*s == ':' && s[1] != ':')
8034 PL_expect = attrful;
8035 else if ((*s != '{' && *s != '(') && key == KEY_sub) {
8037 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8038 else if (*s != ';' && *s != '}')
8039 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8043 NEXTVAL_NEXTTOKE.opval =
8044 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8045 PL_lex_stuff = NULL;
8050 sv_setpvs(PL_subname, "__ANON__");
8052 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8055 force_ident_maybe_lex('&');
8060 LOP(OP_SYSTEM,XREF);
8063 LOP(OP_SYMLINK,XTERM);
8066 LOP(OP_SYSCALL,XTERM);
8069 LOP(OP_SYSOPEN,XTERM);
8072 LOP(OP_SYSSEEK,XTERM);
8075 LOP(OP_SYSREAD,XTERM);
8078 LOP(OP_SYSWRITE,XTERM);
8083 TERM(sublex_start());
8104 LOP(OP_TRUNCATE,XTERM);
8116 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8118 pl_yylval.ival = CopLINE(PL_curcop);
8122 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8124 pl_yylval.ival = CopLINE(PL_curcop);
8128 LOP(OP_UNLINK,XTERM);
8134 LOP(OP_UNPACK,XTERM);
8137 LOP(OP_UTIME,XTERM);
8143 LOP(OP_UNSHIFT,XTERM);
8146 s = tokenize_use(1, s);
8156 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8158 pl_yylval.ival = CopLINE(PL_curcop);
8159 Perl_ck_warner_d(aTHX_
8160 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
8161 "when is experimental");
8165 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8167 pl_yylval.ival = CopLINE(PL_curcop);
8171 PL_hints |= HINT_BLOCK_SCOPE;
8178 LOP(OP_WAITPID,XTERM);
8184 /* Make sure $^L is defined. 0x0C is CTRL-L on ASCII platforms, and
8185 * we use the same number on EBCDIC */
8186 gv_fetchpvs("\x0C", GV_ADD|GV_NOTQUAL, SVt_PV);
8190 if (PL_expect == XOPERATOR) {
8191 if (*s == '=' && !PL_lex_allbrackets &&
8192 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8200 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8202 pl_yylval.ival = OP_XOR;
8211 Looks up an identifier in the pad or in a package
8214 PRIVATEREF if this is a lexical name.
8215 WORD if this belongs to a package.
8218 if we're in a my declaration
8219 croak if they tried to say my($foo::bar)
8220 build the ops for a my() declaration
8221 if it's an access to a my() variable
8222 build ops for access to a my() variable
8223 if in a dq string, and they've said @foo and we can't find @foo
8225 build ops for a bareword
8229 S_pending_ident(pTHX)
8232 const char pit = (char)pl_yylval.ival;
8233 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8234 /* All routes through this function want to know if there is a colon. */
8235 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8237 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8238 "### Pending identifier '%s'\n", PL_tokenbuf); });
8240 /* if we're in a my(), we can't allow dynamics here.
8241 $foo'bar has already been turned into $foo::bar, so
8242 just check for colons.
8244 if it's a legal name, the OP is a PADANY.
8247 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8249 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8250 "variable %s in \"our\"",
8251 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8252 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8256 /* "my" variable %s can't be in a package */
8257 /* PL_no_myglob is constant */
8258 GCC_DIAG_IGNORE(-Wformat-nonliteral);
8259 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8260 PL_in_my == KEY_my ? "my" : "state",
8261 *PL_tokenbuf == '&' ? "subroutin" : "variabl",
8263 UTF ? SVf_UTF8 : 0);
8267 pl_yylval.opval = newOP(OP_PADANY, 0);
8268 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8269 UTF ? SVf_UTF8 : 0);
8275 build the ops for accesses to a my() variable.
8280 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8282 if (tmp != NOT_IN_PAD) {
8283 /* might be an "our" variable" */
8284 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8285 /* build ops for a bareword */
8286 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8287 HEK * const stashname = HvNAME_HEK(stash);
8288 SV * const sym = newSVhek(stashname);
8289 sv_catpvs(sym, "::");
8290 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8291 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8292 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8296 ((PL_tokenbuf[0] == '$') ? SVt_PV
8297 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8302 pl_yylval.opval = newOP(OP_PADANY, 0);
8303 pl_yylval.opval->op_targ = tmp;
8309 Whine if they've said @foo in a doublequoted string,
8310 and @foo isn't a variable we can find in the symbol
8313 if (ckWARN(WARN_AMBIGUOUS) &&
8314 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8315 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8316 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8317 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8318 /* DO NOT warn for @- and @+ */
8319 && !( PL_tokenbuf[2] == '\0' &&
8320 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8323 /* Downgraded from fatal to warning 20000522 mjd */
8324 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8325 "Possible unintended interpolation of %"UTF8f
8327 UTF8fARG(UTF, tokenbuf_len, PL_tokenbuf));
8331 /* build ops for a bareword */
8332 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8333 newSVpvn_flags(PL_tokenbuf + 1,
8335 UTF ? SVf_UTF8 : 0 ));
8336 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8338 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8339 (PL_in_eval ? GV_ADDMULTI : GV_ADD)
8340 | ( UTF ? SVf_UTF8 : 0 ),
8341 ((PL_tokenbuf[0] == '$') ? SVt_PV
8342 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8348 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8350 PERL_ARGS_ASSERT_CHECKCOMMA;
8352 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8353 if (ckWARN(WARN_SYNTAX)) {
8356 for (w = s+2; *w && level; w++) {
8364 /* the list of chars below is for end of statements or
8365 * block / parens, boolean operators (&&, ||, //) and branch
8366 * constructs (or, and, if, until, unless, while, err, for).
8367 * Not a very solid hack... */
8368 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8369 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8370 "%s (...) interpreted as function",name);
8373 while (s < PL_bufend && isSPACE(*s))
8377 while (s < PL_bufend && isSPACE(*s))
8379 if (isIDFIRST_lazy_if(s,UTF)) {
8380 const char * const w = s;
8381 s += UTF ? UTF8SKIP(s) : 1;
8382 while (isWORDCHAR_lazy_if(s,UTF))
8383 s += UTF ? UTF8SKIP(s) : 1;
8384 while (s < PL_bufend && isSPACE(*s))
8389 if (keyword(w, s - w, 0))
8392 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8393 if (gv && GvCVu(gv))
8397 Copy(w, tmpbuf+1, s - w, char);
8399 off = pad_findmy_pvn(tmpbuf, s-w+1, 0);
8400 if (off != NOT_IN_PAD) return;
8402 Perl_croak(aTHX_ "No comma allowed after %s", what);
8407 /* S_new_constant(): do any overload::constant lookup.
8409 Either returns sv, or mortalizes/frees sv and returns a new SV*.
8410 Best used as sv=new_constant(..., sv, ...).
8411 If s, pv are NULL, calls subroutine with one argument,
8412 and <type> is used with error messages only.
8413 <type> is assumed to be well formed UTF-8 */
8416 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8417 SV *sv, SV *pv, const char *type, STRLEN typelen)
8420 HV * table = GvHV(PL_hintgv); /* ^H */
8425 const char *why1 = "", *why2 = "", *why3 = "";
8427 PERL_ARGS_ASSERT_NEW_CONSTANT;
8428 /* We assume that this is true: */
8429 if (*key == 'c') { assert (strEQ(key, "charnames")); }
8432 /* charnames doesn't work well if there have been errors found */
8433 if (PL_error_count > 0 && *key == 'c')
8435 SvREFCNT_dec_NN(sv);
8436 return &PL_sv_undef;
8439 sv_2mortal(sv); /* Parent created it permanently */
8441 || ! (PL_hints & HINT_LOCALIZE_HH)
8442 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8447 /* Here haven't found what we're looking for. If it is charnames,
8448 * perhaps it needs to be loaded. Try doing that before giving up */
8450 Perl_load_module(aTHX_
8452 newSVpvs("_charnames"),
8453 /* version parameter; no need to specify it, as if
8454 * we get too early a version, will fail anyway,
8455 * not being able to find '_charnames' */
8460 assert(sp == PL_stack_sp);
8461 table = GvHV(PL_hintgv);
8463 && (PL_hints & HINT_LOCALIZE_HH)
8464 && (cvp = hv_fetch(table, key, keylen, FALSE))
8470 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8471 msg = Perl_form(aTHX_
8472 "Constant(%.*s) unknown",
8473 (int)(type ? typelen : len),
8479 why3 = "} is not defined";
8482 msg = Perl_form(aTHX_
8483 /* The +3 is for '\N{'; -4 for that, plus '}' */
8484 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
8488 msg = Perl_form(aTHX_ "Constant(%.*s): %s%s%s",
8489 (int)(type ? typelen : len),
8490 (type ? type: s), why1, why2, why3);
8493 yyerror_pv(msg, UTF ? SVf_UTF8 : 0);
8494 return SvREFCNT_inc_simple_NN(sv);
8499 pv = newSVpvn_flags(s, len, SVs_TEMP);
8501 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8503 typesv = &PL_sv_undef;
8505 PUSHSTACKi(PERLSI_OVERLOAD);
8517 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8521 /* Check the eval first */
8522 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
8524 const char * errstr;
8525 sv_catpvs(errsv, "Propagated");
8526 errstr = SvPV_const(errsv, errlen);
8527 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
8529 res = SvREFCNT_inc_simple_NN(sv);
8533 SvREFCNT_inc_simple_void_NN(res);
8542 why1 = "Call to &{$^H{";
8544 why3 = "}} did not return a defined value";
8546 (void)sv_2mortal(sv);
8553 PERL_STATIC_INLINE void
8554 S_parse_ident(pTHX_ char **s, char **d, char * const e, int allow_package, bool is_utf8) {
8555 PERL_ARGS_ASSERT_PARSE_IDENT;
8559 Perl_croak(aTHX_ "%s", ident_too_long);
8560 if (is_utf8 && isIDFIRST_utf8((U8*)*s)) {
8561 /* The UTF-8 case must come first, otherwise things
8562 * like c\N{COMBINING TILDE} would start failing, as the
8563 * isWORDCHAR_A case below would gobble the 'c' up.
8566 char *t = *s + UTF8SKIP(*s);
8567 while (isIDCONT_utf8((U8*)t))
8569 if (*d + (t - *s) > e)
8570 Perl_croak(aTHX_ "%s", ident_too_long);
8571 Copy(*s, *d, t - *s, char);
8575 else if ( isWORDCHAR_A(**s) ) {
8578 } while (isWORDCHAR_A(**s) && *d < e);
8580 else if (allow_package && **s == '\'' && isIDFIRST_lazy_if(*s+1,is_utf8)) {
8585 else if (allow_package && **s == ':' && (*s)[1] == ':'
8586 /* Disallow things like Foo::$bar. For the curious, this is
8587 * the code path that triggers the "Bad name after" warning
8588 * when looking for barewords.
8590 && (*s)[2] != '$') {
8600 /* Returns a NUL terminated string, with the length of the string written to
8604 S_scan_word(pTHX_ char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8607 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8608 bool is_utf8 = cBOOL(UTF);
8610 PERL_ARGS_ASSERT_SCAN_WORD;
8612 parse_ident(&s, &d, e, allow_package, is_utf8);
8618 /* Is the byte 'd' a legal single character identifier name? 'u' is true
8619 * iff Unicode semantics are to be used. The legal ones are any of:
8620 * a) all ASCII characters except:
8621 * 1) control and space-type ones, like NUL, SOH, \t, and SPACE;
8623 * The final case currently doesn't get this far in the program, so we
8624 * don't test for it. If that were to change, it would be ok to allow it.
8625 * c) When not under Unicode rules, any upper Latin1 character
8626 * d) Otherwise, when unicode rules are used, all XIDS characters.
8628 * Because all ASCII characters have the same representation whether
8629 * encoded in UTF-8 or not, we can use the foo_A macros below and '\0' and
8630 * '{' without knowing if is UTF-8 or not.
8631 * EBCDIC already uses the rules that ASCII platforms will use after the
8632 * deprecation cycle; see comment below about the deprecation. */
8634 # define VALID_LEN_ONE_IDENT(s, is_utf8) \
8635 (isGRAPH_A(*(s)) || ((is_utf8) \
8636 ? isIDFIRST_utf8((U8*) (s)) \
8638 && LIKELY((U8) *(s) != LATIN1_TO_NATIVE(0xAD)))))
8640 # define VALID_LEN_ONE_IDENT(s, is_utf8) \
8641 (isGRAPH_A(*(s)) || ((is_utf8) \
8642 ? isIDFIRST_utf8((U8*) (s)) \
8643 : ! isASCII_utf8((U8*) (s))))
8647 S_scan_ident(pTHX_ char *s, char *dest, STRLEN destlen, I32 ck_uni)
8649 I32 herelines = PL_parser->herelines;
8650 SSize_t bracket = -1;
8653 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8654 bool is_utf8 = cBOOL(UTF);
8655 I32 orig_copline = 0, tmp_copline = 0;
8657 PERL_ARGS_ASSERT_SCAN_IDENT;
8659 if (isSPACE(*s) || !*s)
8662 while (isDIGIT(*s)) {
8664 Perl_croak(aTHX_ "%s", ident_too_long);
8668 else { /* See if it is a "normal" identifier */
8669 parse_ident(&s, &d, e, 1, is_utf8);
8674 /* Either a digit variable, or parse_ident() found an identifier
8675 (anything valid as a bareword), so job done and return. */
8676 if (PL_lex_state != LEX_NORMAL)
8677 PL_lex_state = LEX_INTERPENDMAYBE;
8681 /* Here, it is not a run-of-the-mill identifier name */
8683 if (*s == '$' && s[1] &&
8684 (isIDFIRST_lazy_if(s+1,is_utf8)
8685 || isDIGIT_A((U8)s[1])
8688 || strnEQ(s+1,"::",2)) )
8690 /* Dereferencing a value in a scalar variable.
8691 The alternatives are different syntaxes for a scalar variable.
8692 Using ' as a leading package separator isn't allowed. :: is. */
8695 /* Handle the opening { of @{...}, &{...}, *{...}, %{...}, ${...} */
8697 bracket = s - SvPVX(PL_linestr);
8699 orig_copline = CopLINE(PL_curcop);
8700 if (s < PL_bufend && isSPACE(*s)) {
8704 if ((s <= PL_bufend - (is_utf8)
8707 && VALID_LEN_ONE_IDENT(s, is_utf8))
8709 /* Deprecate all non-graphic characters. Include SHY as a non-graphic,
8710 * because often it has no graphic representation. (We can't get to
8711 * here with SHY when 'is_utf8' is true, so no need to include a UTF-8
8714 ? ! isGRAPH_utf8( (U8*) s)
8715 : (! isGRAPH_L1( (U8) *s)
8716 || UNLIKELY((U8) *(s) == LATIN1_TO_NATIVE(0xAD))))
8718 deprecate("literal non-graphic characters in variable names");
8722 const STRLEN skip = UTF8SKIP(s);
8725 for ( i = 0; i < skip; i++ )
8733 /* Convert $^F, ${^F} and the ^F of ${^FOO} to control characters */
8734 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8738 /* Warn about ambiguous code after unary operators if {...} notation isn't
8739 used. There's no difference in ambiguity; it's merely a heuristic
8740 about when not to warn. */
8741 else if (ck_uni && bracket == -1)
8743 if (bracket != -1) {
8744 /* If we were processing {...} notation then... */
8745 if (isIDFIRST_lazy_if(d,is_utf8)) {
8746 /* if it starts as a valid identifier, assume that it is one.
8747 (the later check for } being at the expected point will trap
8748 cases where this doesn't pan out.) */
8749 d += is_utf8 ? UTF8SKIP(d) : 1;
8750 parse_ident(&s, &d, e, 1, is_utf8);
8752 tmp_copline = CopLINE(PL_curcop);
8753 if (s < PL_bufend && isSPACE(*s)) {
8756 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8757 /* ${foo[0]} and ${foo{bar}} notation. */
8758 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8759 const char * const brack =
8761 ((*s == '[') ? "[...]" : "{...}");
8762 orig_copline = CopLINE(PL_curcop);
8763 CopLINE_set(PL_curcop, tmp_copline);
8764 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8765 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8766 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8767 funny, dest, brack, funny, dest, brack);
8768 CopLINE_set(PL_curcop, orig_copline);
8771 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8772 PL_lex_allbrackets++;
8776 /* Handle extended ${^Foo} variables
8777 * 1999-02-27 mjd-perl-patch@plover.com */
8778 else if (! isPRINT(*d) /* isCNTRL(d), plus all non-ASCII */
8782 while (isWORDCHAR(*s) && d < e) {
8786 Perl_croak(aTHX_ "%s", ident_too_long);
8791 tmp_copline = CopLINE(PL_curcop);
8792 if (s < PL_bufend && isSPACE(*s)) {
8796 /* Expect to find a closing } after consuming any trailing whitespace.
8800 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8801 PL_lex_state = LEX_INTERPEND;
8804 if (PL_lex_state == LEX_NORMAL) {
8805 if (ckWARN(WARN_AMBIGUOUS) &&
8806 (keyword(dest, d - dest, 0)
8807 || get_cvn_flags(dest, d - dest, is_utf8 ? SVf_UTF8 : 0)))
8809 SV *tmp = newSVpvn_flags( dest, d - dest,
8810 SVs_TEMP | (is_utf8 ? SVf_UTF8 : 0) );
8813 orig_copline = CopLINE(PL_curcop);
8814 CopLINE_set(PL_curcop, tmp_copline);
8815 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8816 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
8817 funny, SVfARG(tmp), funny, SVfARG(tmp));
8818 CopLINE_set(PL_curcop, orig_copline);
8823 /* Didn't find the closing } at the point we expected, so restore
8824 state such that the next thing to process is the opening { and */
8825 s = SvPVX(PL_linestr) + bracket; /* let the parser handle it */
8826 CopLINE_set(PL_curcop, orig_copline);
8827 PL_parser->herelines = herelines;
8831 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8832 PL_lex_state = LEX_INTERPEND;
8837 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset, unsigned int * x_mod_count) {
8839 /* Adds, subtracts to/from 'pmfl' based on the next regex modifier flag
8840 * found in the parse starting at 's', based on the subset that are valid
8841 * in this context input to this routine in 'valid_flags'. Advances s.
8842 * Returns TRUE if the input should be treated as a valid flag, so the next
8843 * char may be as well; otherwise FALSE. 'charset' should point to a NUL
8844 * upon first call on the current regex. This routine will set it to any
8845 * charset modifier found. The caller shouldn't change it. This way,
8846 * another charset modifier encountered in the parse can be detected as an
8847 * error, as we have decided to allow only one */
8850 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
8852 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
8853 if (isWORDCHAR_lazy_if(*s, UTF)) {
8854 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
8855 UTF ? SVf_UTF8 : 0);
8857 /* Pretend that it worked, so will continue processing before
8866 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, *x_mod_count);
8867 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8868 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8869 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8870 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8871 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8872 case LOCALE_PAT_MOD:
8874 goto multiple_charsets;
8876 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8879 case UNICODE_PAT_MOD:
8881 goto multiple_charsets;
8883 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8886 case ASCII_RESTRICT_PAT_MOD:
8888 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8892 /* Error if previous modifier wasn't an 'a', but if it was, see
8893 * if, and accept, a second occurrence (only) */
8895 || get_regex_charset(*pmfl)
8896 != REGEX_ASCII_RESTRICTED_CHARSET)
8898 goto multiple_charsets;
8900 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8904 case DEPENDS_PAT_MOD:
8906 goto multiple_charsets;
8908 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8917 if (*charset != c) {
8918 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8920 else if (c == 'a') {
8921 /* diag_listed_as: Regexp modifier "/%c" may appear a maximum of twice */
8922 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
8925 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8928 /* Pretend that it worked, so will continue processing before dieing */
8934 S_scan_pat(pTHX_ char *start, I32 type)
8938 const char * const valid_flags =
8939 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8940 char charset = '\0'; /* character set modifier */
8941 unsigned int x_mod_count = 0;
8943 PERL_ARGS_ASSERT_SCAN_PAT;
8945 s = scan_str(start,TRUE,FALSE, (PL_in_eval & EVAL_RE_REPARSING), NULL);
8947 Perl_croak(aTHX_ "Search pattern not terminated");
8949 pm = (PMOP*)newPMOP(type, 0);
8950 if (PL_multi_open == '?') {
8951 /* This is the only point in the code that sets PMf_ONCE: */
8952 pm->op_pmflags |= PMf_ONCE;
8954 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8955 allows us to restrict the list needed by reset to just the ??
8957 assert(type != OP_TRANS);
8959 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8962 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8965 elements = mg->mg_len / sizeof(PMOP**);
8966 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8967 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8968 mg->mg_len = elements * sizeof(PMOP**);
8969 PmopSTASH_set(pm,PL_curstash);
8973 /* if qr/...(?{..}).../, then need to parse the pattern within a new
8974 * anon CV. False positives like qr/[(?{]/ are harmless */
8976 if (type == OP_QR) {
8978 char *e, *p = SvPV(PL_lex_stuff, len);
8980 for (; p < e; p++) {
8981 if (p[0] == '(' && p[1] == '?'
8982 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
8984 pm->op_pmflags |= PMf_HAS_CV;
8988 pm->op_pmflags |= PMf_IS_QR;
8991 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags),
8992 &s, &charset, &x_mod_count))
8994 /* issue a warning if /c is specified,but /g is not */
8995 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8997 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8998 "Use of /c modifier is meaningless without /g" );
9001 STD_PMMOD_FLAGS_PARSE_X_WARN(x_mod_count);
9003 PL_lex_op = (OP*)pm;
9004 pl_yylval.ival = OP_MATCH;
9009 S_scan_subst(pTHX_ char *start)
9016 char charset = '\0'; /* character set modifier */
9017 unsigned int x_mod_count = 0;
9020 PERL_ARGS_ASSERT_SCAN_SUBST;
9022 pl_yylval.ival = OP_NULL;
9024 s = scan_str(start, TRUE, FALSE, FALSE, &t);
9027 Perl_croak(aTHX_ "Substitution pattern not terminated");
9031 first_start = PL_multi_start;
9032 first_line = CopLINE(PL_curcop);
9033 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
9035 SvREFCNT_dec_NN(PL_lex_stuff);
9036 PL_lex_stuff = NULL;
9037 Perl_croak(aTHX_ "Substitution replacement not terminated");
9039 PL_multi_start = first_start; /* so whole substitution is taken together */
9041 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9045 if (*s == EXEC_PAT_MOD) {
9049 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags),
9050 &s, &charset, &x_mod_count))
9056 STD_PMMOD_FLAGS_PARSE_X_WARN(x_mod_count);
9058 if ((pm->op_pmflags & PMf_CONTINUE)) {
9059 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9063 SV * const repl = newSVpvs("");
9066 pm->op_pmflags |= PMf_EVAL;
9069 sv_catpvs(repl, "eval ");
9071 sv_catpvs(repl, "do ");
9073 sv_catpvs(repl, "{");
9074 sv_catsv(repl, PL_sublex_info.repl);
9075 sv_catpvs(repl, "}");
9077 SvREFCNT_dec(PL_sublex_info.repl);
9078 PL_sublex_info.repl = repl;
9080 if (CopLINE(PL_curcop) != first_line) {
9081 sv_upgrade(PL_sublex_info.repl, SVt_PVNV);
9082 ((XPVNV*)SvANY(PL_sublex_info.repl))->xnv_u.xpad_cop_seq.xlow =
9083 CopLINE(PL_curcop) - first_line;
9084 CopLINE_set(PL_curcop, first_line);
9087 PL_lex_op = (OP*)pm;
9088 pl_yylval.ival = OP_SUBST;
9093 S_scan_trans(pTHX_ char *start)
9100 bool nondestruct = 0;
9103 PERL_ARGS_ASSERT_SCAN_TRANS;
9105 pl_yylval.ival = OP_NULL;
9107 s = scan_str(start,FALSE,FALSE,FALSE,&t);
9109 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9113 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
9115 SvREFCNT_dec_NN(PL_lex_stuff);
9116 PL_lex_stuff = NULL;
9117 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9120 complement = del = squash = 0;
9124 complement = OPpTRANS_COMPLEMENT;
9127 del = OPpTRANS_DELETE;
9130 squash = OPpTRANS_SQUASH;
9142 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9143 o->op_private &= ~OPpTRANS_ALL;
9144 o->op_private |= del|squash|complement|
9145 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9146 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9149 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9156 Takes a pointer to the first < in <<FOO.
9157 Returns a pointer to the byte following <<FOO.
9159 This function scans a heredoc, which involves different methods
9160 depending on whether we are in a string eval, quoted construct, etc.
9161 This is because PL_linestr could containing a single line of input, or
9162 a whole string being evalled, or the contents of the current quote-
9165 The two basic methods are:
9166 - Steal lines from the input stream
9167 - Scan the heredoc in PL_linestr and remove it therefrom
9169 In a file scope or filtered eval, the first method is used; in a
9170 string eval, the second.
9172 In a quote-like operator, we have to choose between the two,
9173 depending on where we can find a newline. We peek into outer lex-
9174 ing scopes until we find one with a newline in it. If we reach the
9175 outermost lexing scope and it is a file, we use the stream method.
9176 Otherwise it is treated as an eval.
9180 S_scan_heredoc(pTHX_ char *s)
9182 I32 op_type = OP_SCALAR;
9189 const bool infile = PL_rsfp || PL_parser->filtered;
9190 const line_t origline = CopLINE(PL_curcop);
9191 LEXSHARED *shared = PL_parser->lex_shared;
9193 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9196 d = PL_tokenbuf + 1;
9197 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9198 *PL_tokenbuf = '\n';
9200 while (SPACE_OR_TAB(*peek))
9202 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9205 s = delimcpy(d, e, s, PL_bufend, term, &len);
9207 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9213 /* <<\FOO is equivalent to <<'FOO' */
9217 if (!isWORDCHAR_lazy_if(s,UTF))
9218 deprecate("bare << to mean <<\"\"");
9220 while (isWORDCHAR_lazy_if(peek,UTF)) {
9221 peek += UTF ? UTF8SKIP(peek) : 1;
9223 len = (peek - s >= e - d) ? (e - d) : (peek - s);
9224 Copy(s, d, len, char);
9228 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9229 Perl_croak(aTHX_ "Delimiter for here document is too long");
9232 len = d - PL_tokenbuf;
9234 #ifndef PERL_STRICT_CR
9235 d = strchr(s, '\r');
9237 char * const olds = s;
9239 while (s < PL_bufend) {
9245 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9254 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9259 tmpstr = newSV_type(SVt_PVIV);
9263 SvIV_set(tmpstr, -1);
9265 else if (term == '`') {
9266 op_type = OP_BACKTICK;
9267 SvIV_set(tmpstr, '\\');
9270 PL_multi_start = origline + 1 + PL_parser->herelines;
9271 PL_multi_open = PL_multi_close = '<';
9272 /* inside a string eval or quote-like operator */
9273 if (!infile || PL_lex_inwhat) {
9276 char * const olds = s;
9277 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9278 /* These two fields are not set until an inner lexing scope is
9279 entered. But we need them set here. */
9280 shared->ls_bufptr = s;
9281 shared->ls_linestr = PL_linestr;
9283 /* Look for a newline. If the current buffer does not have one,
9284 peek into the line buffer of the parent lexing scope, going
9285 up as many levels as necessary to find one with a newline
9288 while (!(s = (char *)memchr(
9289 (void *)shared->ls_bufptr, '\n',
9290 SvEND(shared->ls_linestr)-shared->ls_bufptr
9292 shared = shared->ls_prev;
9293 /* shared is only null if we have gone beyond the outermost
9294 lexing scope. In a file, we will have broken out of the
9295 loop in the previous iteration. In an eval, the string buf-
9296 fer ends with "\n;", so the while condition above will have
9297 evaluated to false. So shared can never be null. Or so you
9298 might think. Odd syntax errors like s;@{<<; can gobble up
9299 the implicit semicolon at the end of a flie, causing the
9300 file handle to be closed even when we are not in a string
9301 eval. So shared may be null in that case. */
9302 if (UNLIKELY(!shared))
9304 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9305 most lexing scope. In a file, shared->ls_linestr at that
9306 level is just one line, so there is no body to steal. */
9307 if (infile && !shared->ls_prev) {
9313 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9316 linestr = shared->ls_linestr;
9317 bufend = SvEND(linestr);
9319 while (s < bufend - len + 1 &&
9320 memNE(s,PL_tokenbuf,len) ) {
9322 ++PL_parser->herelines;
9324 if (s >= bufend - len + 1) {
9327 sv_setpvn(tmpstr,d+1,s-d);
9329 /* the preceding stmt passes a newline */
9330 PL_parser->herelines++;
9332 /* s now points to the newline after the heredoc terminator.
9333 d points to the newline before the body of the heredoc.
9336 /* We are going to modify linestr in place here, so set
9337 aside copies of the string if necessary for re-evals or
9339 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9340 check shared->re_eval_str. */
9341 if (shared->re_eval_start || shared->re_eval_str) {
9342 /* Set aside the rest of the regexp */
9343 if (!shared->re_eval_str)
9344 shared->re_eval_str =
9345 newSVpvn(shared->re_eval_start,
9346 bufend - shared->re_eval_start);
9347 shared->re_eval_start -= s-d;
9349 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9350 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9351 cx->blk_eval.cur_text == linestr)
9353 cx->blk_eval.cur_text = newSVsv(linestr);
9354 SvSCREAM_on(cx->blk_eval.cur_text);
9356 /* Copy everything from s onwards back to d. */
9357 Move(s,d,bufend-s + 1,char);
9358 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9359 /* Setting PL_bufend only applies when we have not dug deeper
9360 into other scopes, because sublex_done sets PL_bufend to
9361 SvEND(PL_linestr). */
9362 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9369 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9370 term = PL_tokenbuf[1];
9372 linestr_save = PL_linestr; /* must restore this afterwards */
9373 d = s; /* and this */
9374 PL_linestr = newSVpvs("");
9375 PL_bufend = SvPVX(PL_linestr);
9377 PL_bufptr = PL_bufend;
9378 CopLINE_set(PL_curcop,
9379 origline + 1 + PL_parser->herelines);
9380 if (!lex_next_chunk(LEX_NO_TERM)
9381 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9382 /* Simply freeing linestr_save might seem simpler here, as it
9383 does not matter what PL_linestr points to, since we are
9384 about to croak; but in a quote-like op, linestr_save
9385 will have been prospectively freed already, via
9386 SAVEFREESV(PL_linestr) in sublex_push, so it’s easier to
9387 restore PL_linestr. */
9388 SvREFCNT_dec_NN(PL_linestr);
9389 PL_linestr = linestr_save;
9392 CopLINE_set(PL_curcop, origline);
9393 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9394 s = lex_grow_linestr(SvLEN(PL_linestr) + 3);
9395 /* ^That should be enough to avoid this needing to grow: */
9396 sv_catpvs(PL_linestr, "\n\0");
9397 assert(s == SvPVX(PL_linestr));
9398 PL_bufend = SvEND(PL_linestr);
9401 PL_parser->herelines++;
9402 PL_last_lop = PL_last_uni = NULL;
9403 #ifndef PERL_STRICT_CR
9404 if (PL_bufend - PL_linestart >= 2) {
9405 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9406 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9408 PL_bufend[-2] = '\n';
9410 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9412 else if (PL_bufend[-1] == '\r')
9413 PL_bufend[-1] = '\n';
9415 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9416 PL_bufend[-1] = '\n';
9418 if (*s == term && PL_bufend-s >= len
9419 && memEQ(s,PL_tokenbuf + 1,len)) {
9420 SvREFCNT_dec(PL_linestr);
9421 PL_linestr = linestr_save;
9422 PL_linestart = SvPVX(linestr_save);
9423 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9428 sv_catsv(tmpstr,PL_linestr);
9432 PL_multi_end = origline + PL_parser->herelines;
9433 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9434 SvPV_shrink_to_cur(tmpstr);
9437 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9439 else if (IN_ENCODING)
9440 sv_recode_to_utf8(tmpstr, _get_encoding());
9442 PL_lex_stuff = tmpstr;
9443 pl_yylval.ival = op_type;
9447 SvREFCNT_dec(tmpstr);
9448 CopLINE_set(PL_curcop, origline);
9449 missingterm(PL_tokenbuf + 1);
9453 takes: current position in input buffer
9454 returns: new position in input buffer
9455 side-effects: pl_yylval and lex_op are set.
9460 <<>> read from ARGV without magic open
9461 <FH> read from filehandle
9462 <pkg::FH> read from package qualified filehandle
9463 <pkg'FH> read from package qualified filehandle
9464 <$fh> read from filehandle in $fh
9470 S_scan_inputsymbol(pTHX_ char *start)
9472 char *s = start; /* current position in buffer */
9475 bool nomagicopen = FALSE;
9476 char *d = PL_tokenbuf; /* start of temp holding space */
9477 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9479 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9481 end = strchr(s, '\n');
9484 if (s[1] == '<' && s[2] == '>' && s[3] == '>') {
9491 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9493 /* die if we didn't have space for the contents of the <>,
9494 or if it didn't end, or if we see a newline
9497 if (len >= (I32)sizeof PL_tokenbuf)
9498 Perl_croak(aTHX_ "Excessively long <> operator");
9500 Perl_croak(aTHX_ "Unterminated <> operator");
9505 Remember, only scalar variables are interpreted as filehandles by
9506 this code. Anything more complex (e.g., <$fh{$num}>) will be
9507 treated as a glob() call.
9508 This code makes use of the fact that except for the $ at the front,
9509 a scalar variable and a filehandle look the same.
9511 if (*d == '$' && d[1]) d++;
9513 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9514 while (*d && (isWORDCHAR_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9515 d += UTF ? UTF8SKIP(d) : 1;
9517 /* If we've tried to read what we allow filehandles to look like, and
9518 there's still text left, then it must be a glob() and not a getline.
9519 Use scan_str to pull out the stuff between the <> and treat it
9520 as nothing more than a string.
9523 if (d - PL_tokenbuf != len) {
9524 pl_yylval.ival = OP_GLOB;
9525 s = scan_str(start,FALSE,FALSE,FALSE,NULL);
9527 Perl_croak(aTHX_ "Glob not terminated");
9531 bool readline_overriden = FALSE;
9533 /* we're in a filehandle read situation */
9536 /* turn <> into <ARGV> */
9538 Copy("ARGV",d,5,char);
9540 /* Check whether readline() is overriden */
9541 if ((gv_readline = gv_override("readline",8)))
9542 readline_overriden = TRUE;
9544 /* if <$fh>, create the ops to turn the variable into a
9548 /* try to find it in the pad for this block, otherwise find
9549 add symbol table ops
9551 const PADOFFSET tmp = pad_findmy_pvn(d, len, 0);
9552 if (tmp != NOT_IN_PAD) {
9553 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9554 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9555 HEK * const stashname = HvNAME_HEK(stash);
9556 SV * const sym = sv_2mortal(newSVhek(stashname));
9557 sv_catpvs(sym, "::");
9563 OP * const o = newOP(OP_PADSV, 0);
9565 PL_lex_op = readline_overriden
9566 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9567 op_append_elem(OP_LIST, o,
9568 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9569 : (OP*)newUNOP(OP_READLINE, 0, o);
9577 GV_ADDMULTI | ( UTF ? SVf_UTF8 : 0 ),
9579 PL_lex_op = readline_overriden
9580 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9581 op_append_elem(OP_LIST,
9582 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9583 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9584 : (OP*)newUNOP(OP_READLINE, 0,
9585 newUNOP(OP_RV2SV, 0,
9586 newGVOP(OP_GV, 0, gv)));
9588 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9589 pl_yylval.ival = OP_NULL;
9592 /* If it's none of the above, it must be a literal filehandle
9593 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9595 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9596 PL_lex_op = readline_overriden
9597 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9598 op_append_elem(OP_LIST,
9599 newGVOP(OP_GV, 0, gv),
9600 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9601 : (OP*)newUNOP(OP_READLINE, nomagicopen ? OPf_SPECIAL : 0, newGVOP(OP_GV, 0, gv));
9602 pl_yylval.ival = OP_NULL;
9612 start position in buffer
9613 keep_bracketed_quoted preserve \ quoting of embedded delimiters, but
9614 only if they are of the open/close form
9615 keep_delims preserve the delimiters around the string
9616 re_reparse compiling a run-time /(?{})/:
9617 collapse // to /, and skip encoding src
9618 delimp if non-null, this is set to the position of
9619 the closing delimiter, or just after it if
9620 the closing and opening delimiters differ
9621 (i.e., the opening delimiter of a substitu-
9623 returns: position to continue reading from buffer
9624 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9625 updates the read buffer.
9627 This subroutine pulls a string out of the input. It is called for:
9628 q single quotes q(literal text)
9629 ' single quotes 'literal text'
9630 qq double quotes qq(interpolate $here please)
9631 " double quotes "interpolate $here please"
9632 qx backticks qx(/bin/ls -l)
9633 ` backticks `/bin/ls -l`
9634 qw quote words @EXPORT_OK = qw( func() $spam )
9635 m// regexp match m/this/
9636 s/// regexp substitute s/this/that/
9637 tr/// string transliterate tr/this/that/
9638 y/// string transliterate y/this/that/
9639 ($*@) sub prototypes sub foo ($)
9640 (stuff) sub attr parameters sub foo : attr(stuff)
9641 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9643 In most of these cases (all but <>, patterns and transliterate)
9644 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9645 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9646 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9649 It skips whitespace before the string starts, and treats the first
9650 character as the delimiter. If the delimiter is one of ([{< then
9651 the corresponding "close" character )]}> is used as the closing
9652 delimiter. It allows quoting of delimiters, and if the string has
9653 balanced delimiters ([{<>}]) it allows nesting.
9655 On success, the SV with the resulting string is put into lex_stuff or,
9656 if that is already non-NULL, into lex_repl. The second case occurs only
9657 when parsing the RHS of the special constructs s/// and tr/// (y///).
9658 For convenience, the terminating delimiter character is stuffed into
9663 S_scan_str(pTHX_ char *start, int keep_bracketed_quoted, int keep_delims, int re_reparse,
9667 SV *sv; /* scalar value: string */
9668 const char *tmps; /* temp string, used for delimiter matching */
9669 char *s = start; /* current position in the buffer */
9670 char term; /* terminating character */
9671 char *to; /* current position in the sv's data */
9672 I32 brackets = 1; /* bracket nesting level */
9673 bool has_utf8 = FALSE; /* is there any utf8 content? */
9674 I32 termcode; /* terminating char. code */
9675 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9676 STRLEN termlen; /* length of terminating string */
9677 int last_off = 0; /* last position for nesting bracket */
9680 PERL_ARGS_ASSERT_SCAN_STR;
9682 /* skip space before the delimiter */
9687 /* mark where we are, in case we need to report errors */
9690 /* after skipping whitespace, the next character is the terminator */
9693 termcode = termstr[0] = term;
9697 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
9698 Copy(s, termstr, termlen, U8);
9699 if (!UTF8_IS_INVARIANT(term))
9703 /* mark where we are */
9704 PL_multi_start = CopLINE(PL_curcop);
9705 PL_multi_open = term;
9706 herelines = PL_parser->herelines;
9708 /* find corresponding closing delimiter */
9709 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9710 termcode = termstr[0] = term = tmps[5];
9712 PL_multi_close = term;
9714 if (PL_multi_open == PL_multi_close) {
9715 keep_bracketed_quoted = FALSE;
9718 /* create a new SV to hold the contents. 79 is the SV's initial length.
9719 What a random number. */
9720 sv = newSV_type(SVt_PVIV);
9722 SvIV_set(sv, termcode);
9723 (void)SvPOK_only(sv); /* validate pointer */
9725 /* move past delimiter and try to read a complete string */
9727 sv_catpvn(sv, s, termlen);
9730 if (IN_ENCODING && !UTF && !re_reparse) {
9734 int offset = s - SvPVX_const(PL_linestr);
9735 const bool found = sv_cat_decode(sv, _get_encoding(), PL_linestr,
9736 &offset, (char*)termstr, termlen);
9740 if (SvIsCOW(PL_linestr)) {
9741 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos;
9742 STRLEN oldoldbufptr_pos, linestart_pos, last_uni_pos;
9743 STRLEN last_lop_pos, re_eval_start_pos, s_pos;
9744 char *buf = SvPVX(PL_linestr);
9745 bufend_pos = PL_parser->bufend - buf;
9746 bufptr_pos = PL_parser->bufptr - buf;
9747 oldbufptr_pos = PL_parser->oldbufptr - buf;
9748 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
9749 linestart_pos = PL_parser->linestart - buf;
9750 last_uni_pos = PL_parser->last_uni
9751 ? PL_parser->last_uni - buf
9753 last_lop_pos = PL_parser->last_lop
9754 ? PL_parser->last_lop - buf
9757 PL_parser->lex_shared->re_eval_start ?
9758 PL_parser->lex_shared->re_eval_start - buf : 0;
9761 sv_force_normal(PL_linestr);
9763 buf = SvPVX(PL_linestr);
9764 PL_parser->bufend = buf + bufend_pos;
9765 PL_parser->bufptr = buf + bufptr_pos;
9766 PL_parser->oldbufptr = buf + oldbufptr_pos;
9767 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
9768 PL_parser->linestart = buf + linestart_pos;
9769 if (PL_parser->last_uni)
9770 PL_parser->last_uni = buf + last_uni_pos;
9771 if (PL_parser->last_lop)
9772 PL_parser->last_lop = buf + last_lop_pos;
9773 if (PL_parser->lex_shared->re_eval_start)
9774 PL_parser->lex_shared->re_eval_start =
9775 buf + re_eval_start_pos;
9778 ns = SvPVX_const(PL_linestr) + offset;
9779 svlast = SvEND(sv) - 1;
9781 for (; s < ns; s++) {
9782 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9783 COPLINE_INC_WITH_HERELINES;
9786 goto read_more_line;
9788 /* handle quoted delimiters */
9789 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9791 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9793 if ((svlast-1 - t) % 2) {
9794 if (!keep_bracketed_quoted) {
9797 SvCUR_set(sv, SvCUR(sv) - 1);
9802 if (PL_multi_open == PL_multi_close) {
9808 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9809 /* At here, all closes are "was quoted" one,
9810 so we don't check PL_multi_close. */
9812 if (!keep_bracketed_quoted && *(t+1) == PL_multi_open)
9817 else if (*t == PL_multi_open)
9825 SvCUR_set(sv, w - SvPVX_const(sv));
9827 last_off = w - SvPVX(sv);
9828 if (--brackets <= 0)
9834 SvCUR_set(sv, SvCUR(sv) - 1);
9840 /* extend sv if need be */
9841 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9842 /* set 'to' to the next character in the sv's string */
9843 to = SvPVX(sv)+SvCUR(sv);
9845 /* if open delimiter is the close delimiter read unbridle */
9846 if (PL_multi_open == PL_multi_close) {
9847 for (; s < PL_bufend; s++,to++) {
9848 /* embedded newlines increment the current line number */
9849 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9850 COPLINE_INC_WITH_HERELINES;
9851 /* handle quoted delimiters */
9852 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9853 if (!keep_bracketed_quoted
9855 || (re_reparse && s[1] == '\\'))
9858 else /* any other quotes are simply copied straight through */
9861 /* terminate when run out of buffer (the for() condition), or
9862 have found the terminator */
9863 else if (*s == term) {
9866 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9869 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9875 /* if the terminator isn't the same as the start character (e.g.,
9876 matched brackets), we have to allow more in the quoting, and
9877 be prepared for nested brackets.
9880 /* read until we run out of string, or we find the terminator */
9881 for (; s < PL_bufend; s++,to++) {
9882 /* embedded newlines increment the line count */
9883 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9884 COPLINE_INC_WITH_HERELINES;
9885 /* backslashes can escape the open or closing characters */
9886 if (*s == '\\' && s+1 < PL_bufend) {
9887 if (!keep_bracketed_quoted &&
9888 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9895 /* allow nested opens and closes */
9896 else if (*s == PL_multi_close && --brackets <= 0)
9898 else if (*s == PL_multi_open)
9900 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9905 /* terminate the copied string and update the sv's end-of-string */
9907 SvCUR_set(sv, to - SvPVX_const(sv));
9910 * this next chunk reads more into the buffer if we're not done yet
9914 break; /* handle case where we are done yet :-) */
9916 #ifndef PERL_STRICT_CR
9917 if (to - SvPVX_const(sv) >= 2) {
9918 if ((to[-2] == '\r' && to[-1] == '\n') ||
9919 (to[-2] == '\n' && to[-1] == '\r'))
9923 SvCUR_set(sv, to - SvPVX_const(sv));
9925 else if (to[-1] == '\r')
9928 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9933 /* if we're out of file, or a read fails, bail and reset the current
9934 line marker so we can report where the unterminated string began
9936 COPLINE_INC_WITH_HERELINES;
9937 PL_bufptr = PL_bufend;
9938 if (!lex_next_chunk(0)) {
9940 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9946 /* at this point, we have successfully read the delimited string */
9948 if (!IN_ENCODING || UTF || re_reparse) {
9951 sv_catpvn(sv, s, termlen);
9954 if (has_utf8 || (IN_ENCODING && !re_reparse))
9957 PL_multi_end = CopLINE(PL_curcop);
9958 CopLINE_set(PL_curcop, PL_multi_start);
9959 PL_parser->herelines = herelines;
9961 /* if we allocated too much space, give some back */
9962 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9963 SvLEN_set(sv, SvCUR(sv) + 1);
9964 SvPV_renew(sv, SvLEN(sv));
9967 /* decide whether this is the first or second quoted string we've read
9972 PL_sublex_info.repl = sv;
9975 if (delimp) *delimp = PL_multi_open == PL_multi_close ? s-termlen : s;
9981 takes: pointer to position in buffer
9982 returns: pointer to new position in buffer
9983 side-effects: builds ops for the constant in pl_yylval.op
9985 Read a number in any of the formats that Perl accepts:
9987 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
9988 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
9989 0b[01](_?[01])* binary integers
9990 0[0-7](_?[0-7])* octal integers
9991 0x[0-9A-Fa-f](_?[0-9A-Fa-f])* hexadecimal integers
9992 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*(?:\.\d*)?p[+-]?[0-9]+ hexadecimal floats
9994 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
9997 If it reads a number without a decimal point or an exponent, it will
9998 try converting the number to an integer and see if it can do so
9999 without loss of precision.
10003 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10005 const char *s = start; /* current position in buffer */
10006 char *d; /* destination in temp buffer */
10007 char *e; /* end of temp buffer */
10008 NV nv; /* number read, as a double */
10009 SV *sv = NULL; /* place to put the converted number */
10010 bool floatit; /* boolean: int or float? */
10011 const char *lastub = NULL; /* position of last underbar */
10012 static const char* const number_too_long = "Number too long";
10013 /* Hexadecimal floating point.
10015 * In many places (where we have quads and NV is IEEE 754 double)
10016 * we can fit the mantissa bits of a NV into an unsigned quad.
10017 * (Note that UVs might not be quads even when we have quads.)
10018 * This will not work everywhere, though (either no quads, or
10019 * using long doubles), in which case we have to resort to NV,
10020 * which will probably mean horrible loss of precision due to
10021 * multiple fp operations. */
10022 bool hexfp = FALSE;
10023 int total_bits = 0;
10024 #if NVSIZE == 8 && defined(HAS_QUAD) && defined(Uquad_t)
10025 # define HEXFP_UQUAD
10026 Uquad_t hexfp_uquad = 0;
10027 int hexfp_frac_bits = 0;
10032 NV hexfp_mult = 1.0;
10033 UV high_non_zero = 0; /* highest digit */
10035 PERL_ARGS_ASSERT_SCAN_NUM;
10037 /* We use the first character to decide what type of number this is */
10041 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10043 /* if it starts with a 0, it could be an octal number, a decimal in
10044 0.13 disguise, or a hexadecimal number, or a binary number. */
10048 u holds the "number so far"
10049 shift the power of 2 of the base
10050 (hex == 4, octal == 3, binary == 1)
10051 overflowed was the number more than we can hold?
10053 Shift is used when we add a digit. It also serves as an "are
10054 we in octal/hex/binary?" indicator to disallow hex characters
10055 when in octal mode.
10060 bool overflowed = FALSE;
10061 bool just_zero = TRUE; /* just plain 0 or binary number? */
10062 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10063 static const char* const bases[5] =
10064 { "", "binary", "", "octal", "hexadecimal" };
10065 static const char* const Bases[5] =
10066 { "", "Binary", "", "Octal", "Hexadecimal" };
10067 static const char* const maxima[5] =
10069 "0b11111111111111111111111111111111",
10073 const char *base, *Base, *max;
10075 /* check for hex */
10076 if (isALPHA_FOLD_EQ(s[1], 'x')) {
10080 } else if (isALPHA_FOLD_EQ(s[1], 'b')) {
10085 /* check for a decimal in disguise */
10086 else if (s[1] == '.' || isALPHA_FOLD_EQ(s[1], 'e'))
10088 /* so it must be octal */
10095 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10096 "Misplaced _ in number");
10100 base = bases[shift];
10101 Base = Bases[shift];
10102 max = maxima[shift];
10104 /* read the rest of the number */
10106 /* x is used in the overflow test,
10107 b is the digit we're adding on. */
10112 /* if we don't mention it, we're done */
10116 /* _ are ignored -- but warned about if consecutive */
10118 if (lastub && s == lastub + 1)
10119 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10120 "Misplaced _ in number");
10124 /* 8 and 9 are not octal */
10125 case '8': case '9':
10127 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10131 case '2': case '3': case '4':
10132 case '5': case '6': case '7':
10134 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10137 case '0': case '1':
10138 b = *s++ & 15; /* ASCII digit -> value of digit */
10142 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10143 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10144 /* make sure they said 0x */
10147 b = (*s++ & 7) + 9;
10149 /* Prepare to put the digit we have onto the end
10150 of the number so far. We check for overflows.
10156 x = u << shift; /* make room for the digit */
10158 total_bits += shift;
10160 if ((x >> shift) != u
10161 && !(PL_hints & HINT_NEW_BINARY)) {
10164 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10165 "Integer overflow in %s number",
10168 u = x | b; /* add the digit to the end */
10171 n *= nvshift[shift];
10172 /* If an NV has not enough bits in its
10173 * mantissa to represent an UV this summing of
10174 * small low-order numbers is a waste of time
10175 * (because the NV cannot preserve the
10176 * low-order bits anyway): we could just
10177 * remember when did we overflow and in the
10178 * end just multiply n by the right
10183 if (high_non_zero == 0 && b > 0)
10186 /* this could be hexfp, but peek ahead
10187 * to avoid matching ".." */
10188 if (UNLIKELY(HEXFP_PEEK(s))) {
10196 /* if we get here, we had success: make a scalar value from
10201 /* final misplaced underbar check */
10202 if (s[-1] == '_') {
10203 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10206 if (UNLIKELY(HEXFP_PEEK(s))) {
10207 /* Do sloppy (on the underbars) but quick detection
10208 * (and value construction) for hexfp, the decimal
10209 * detection will shortly be more thorough with the
10210 * underbar checks. */
10214 #else /* HEXFP_NV */
10219 NV mult = 1 / 16.0;
10222 while (isXDIGIT(*h) || *h == '_') {
10223 if (isXDIGIT(*h)) {
10224 U8 b = XDIGIT_VALUE(*h);
10225 total_bits += shift;
10227 hexfp_uquad <<= shift;
10229 hexfp_frac_bits += shift;
10230 #else /* HEXFP_NV */
10231 hexfp_nv += b * mult;
10239 if (total_bits >= 4) {
10240 if (high_non_zero < 0x8)
10242 if (high_non_zero < 0x4)
10244 if (high_non_zero < 0x2)
10248 if (total_bits > 0 && (isALPHA_FOLD_EQ(*h, 'p'))) {
10249 bool negexp = FALSE;
10253 else if (*h == '-') {
10259 while (isDIGIT(*h) || *h == '_') {
10262 hexfp_exp += *h - '0';
10265 -hexfp_exp < NV_MIN_EXP - 1) {
10266 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10267 "Hexadecimal float: exponent underflow");
10274 hexfp_exp > NV_MAX_EXP - 1) {
10275 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10276 "Hexadecimal float: exponent overflow");
10285 hexfp_exp = -hexfp_exp;
10287 hexfp_exp -= hexfp_frac_bits;
10289 hexfp_mult = pow(2.0, hexfp_exp);
10297 if (n > 4294967295.0)
10298 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10299 "%s number > %s non-portable",
10305 if (u > 0xffffffff)
10306 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10307 "%s number > %s non-portable",
10312 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10313 sv = new_constant(start, s - start, "integer",
10314 sv, NULL, NULL, 0);
10315 else if (PL_hints & HINT_NEW_BINARY)
10316 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10321 handle decimal numbers.
10322 we're also sent here when we read a 0 as the first digit
10324 case '1': case '2': case '3': case '4': case '5':
10325 case '6': case '7': case '8': case '9': case '.':
10328 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10337 /* read next group of digits and _ and copy into d */
10338 while (isDIGIT(*s) || *s == '_' ||
10339 UNLIKELY(hexfp && isXDIGIT(*s))) {
10340 /* skip underscores, checking for misplaced ones
10344 if (lastub && s == lastub + 1)
10345 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10346 "Misplaced _ in number");
10350 /* check for end of fixed-length buffer */
10352 Perl_croak(aTHX_ "%s", number_too_long);
10353 /* if we're ok, copy the character */
10358 /* final misplaced underbar check */
10359 if (lastub && s == lastub + 1) {
10360 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10363 /* read a decimal portion if there is one. avoid
10364 3..5 being interpreted as the number 3. followed
10367 if (*s == '.' && s[1] != '.') {
10372 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10373 "Misplaced _ in number");
10377 /* copy, ignoring underbars, until we run out of digits.
10379 for (; isDIGIT(*s) || *s == '_' ||
10380 UNLIKELY(hexfp && isXDIGIT(*s));
10382 /* fixed length buffer check */
10384 Perl_croak(aTHX_ "%s", number_too_long);
10386 if (lastub && s == lastub + 1)
10387 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10388 "Misplaced _ in number");
10394 /* fractional part ending in underbar? */
10395 if (s[-1] == '_') {
10396 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10397 "Misplaced _ in number");
10399 if (*s == '.' && isDIGIT(s[1])) {
10400 /* oops, it's really a v-string, but without the "v" */
10406 /* read exponent part, if present */
10407 if ((isALPHA_FOLD_EQ(*s, 'e')
10408 || UNLIKELY(hexfp && isALPHA_FOLD_EQ(*s, 'p')))
10409 && strchr("+-0123456789_", s[1]))
10413 /* regardless of whether user said 3E5 or 3e5, use lower 'e',
10414 ditto for p (hexfloats) */
10415 if ((isALPHA_FOLD_EQ(*s, 'e'))) {
10416 /* At least some Mach atof()s don't grok 'E' */
10419 else if (UNLIKELY(hexfp && (isALPHA_FOLD_EQ(*s, 'p')))) {
10426 /* stray preinitial _ */
10428 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10429 "Misplaced _ in number");
10433 /* allow positive or negative exponent */
10434 if (*s == '+' || *s == '-')
10437 /* stray initial _ */
10439 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10440 "Misplaced _ in number");
10444 /* read digits of exponent */
10445 while (isDIGIT(*s) || *s == '_') {
10448 Perl_croak(aTHX_ "%s", number_too_long);
10452 if (((lastub && s == lastub + 1) ||
10453 (!isDIGIT(s[1]) && s[1] != '_')))
10454 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10455 "Misplaced _ in number");
10463 We try to do an integer conversion first if no characters
10464 indicating "float" have been found.
10469 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10471 if (flags == IS_NUMBER_IN_UV) {
10473 sv = newSViv(uv); /* Prefer IVs over UVs. */
10476 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10477 if (uv <= (UV) IV_MIN)
10478 sv = newSViv(-(IV)uv);
10485 STORE_LC_NUMERIC_UNDERLYING_SET_STANDARD();
10486 /* terminate the string */
10488 if (UNLIKELY(hexfp)) {
10489 # ifdef NV_MANT_DIG
10490 if (total_bits > NV_MANT_DIG)
10491 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10492 "Hexadecimal float: mantissa overflow");
10495 nv = hexfp_uquad * hexfp_mult;
10496 #else /* HEXFP_NV */
10497 nv = hexfp_nv * hexfp_mult;
10500 nv = Atof(PL_tokenbuf);
10502 RESTORE_LC_NUMERIC_UNDERLYING();
10507 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10508 const char *const key = floatit ? "float" : "integer";
10509 const STRLEN keylen = floatit ? 5 : 7;
10510 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10511 key, keylen, sv, NULL, NULL, 0);
10515 /* if it starts with a v, it could be a v-string */
10518 sv = newSV(5); /* preallocate storage space */
10519 ENTER_with_name("scan_vstring");
10521 s = scan_vstring(s, PL_bufend, sv);
10522 SvREFCNT_inc_simple_void_NN(sv);
10523 LEAVE_with_name("scan_vstring");
10527 /* make the op for the constant and return */
10530 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10532 lvalp->opval = NULL;
10538 S_scan_formline(pTHX_ char *s)
10542 SV * const stuff = newSVpvs("");
10543 bool needargs = FALSE;
10544 bool eofmt = FALSE;
10546 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10548 while (!needargs) {
10551 #ifdef PERL_STRICT_CR
10552 while (SPACE_OR_TAB(*t))
10555 while (SPACE_OR_TAB(*t) || *t == '\r')
10558 if (*t == '\n' || t == PL_bufend) {
10563 eol = (char *) memchr(s,'\n',PL_bufend-s);
10567 for (t = s; t < eol; t++) {
10568 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10570 goto enough; /* ~~ must be first line in formline */
10572 if (*t == '@' || *t == '^')
10576 sv_catpvn(stuff, s, eol-s);
10577 #ifndef PERL_STRICT_CR
10578 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10579 char *end = SvPVX(stuff) + SvCUR(stuff);
10582 SvCUR_set(stuff, SvCUR(stuff) - 1);
10590 if ((PL_rsfp || PL_parser->filtered)
10591 && PL_parser->form_lex_state == LEX_NORMAL) {
10593 PL_bufptr = PL_bufend;
10594 COPLINE_INC_WITH_HERELINES;
10595 got_some = lex_next_chunk(0);
10596 CopLINE_dec(PL_curcop);
10604 if (!SvCUR(stuff) || needargs)
10605 PL_lex_state = PL_parser->form_lex_state;
10606 if (SvCUR(stuff)) {
10607 PL_expect = XSTATE;
10609 const char *s2 = s;
10610 while (*s2 == '\r' || *s2 == ' ' || *s2 == '\t' || *s2 == '\f'
10614 PL_expect = XTERMBLOCK;
10615 NEXTVAL_NEXTTOKE.ival = 0;
10618 NEXTVAL_NEXTTOKE.ival = 0;
10619 force_next(FORMLBRACK);
10622 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10624 else if (IN_ENCODING)
10625 sv_recode_to_utf8(stuff, _get_encoding());
10627 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10631 SvREFCNT_dec(stuff);
10633 PL_lex_formbrack = 0;
10639 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10641 const I32 oldsavestack_ix = PL_savestack_ix;
10642 CV* const outsidecv = PL_compcv;
10644 SAVEI32(PL_subline);
10645 save_item(PL_subname);
10646 SAVESPTR(PL_compcv);
10648 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10649 CvFLAGS(PL_compcv) |= flags;
10651 PL_subline = CopLINE(PL_curcop);
10652 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10653 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10654 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10655 if (outsidecv && CvPADLIST(outsidecv))
10656 CvPADLIST(PL_compcv)->xpadl_outid = CvPADLIST(outsidecv)->xpadl_id;
10658 return oldsavestack_ix;
10662 S_yywarn(pTHX_ const char *const s, U32 flags)
10664 PERL_ARGS_ASSERT_YYWARN;
10666 PL_in_eval |= EVAL_WARNONLY;
10667 yyerror_pv(s, flags);
10672 Perl_yyerror(pTHX_ const char *const s)
10674 PERL_ARGS_ASSERT_YYERROR;
10675 return yyerror_pvn(s, strlen(s), 0);
10679 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
10681 PERL_ARGS_ASSERT_YYERROR_PV;
10682 return yyerror_pvn(s, strlen(s), flags);
10686 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
10688 const char *context = NULL;
10691 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
10692 int yychar = PL_parser->yychar;
10694 PERL_ARGS_ASSERT_YYERROR_PVN;
10696 if (!yychar || (yychar == ';' && !PL_rsfp))
10697 sv_catpvs(where_sv, "at EOF");
10698 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10699 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10700 PL_oldbufptr != PL_bufptr) {
10703 The code below is removed for NetWare because it abends/crashes on NetWare
10704 when the script has error such as not having the closing quotes like:
10705 if ($var eq "value)
10706 Checking of white spaces is anyway done in NetWare code.
10709 while (isSPACE(*PL_oldoldbufptr))
10712 context = PL_oldoldbufptr;
10713 contlen = PL_bufptr - PL_oldoldbufptr;
10715 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10716 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10719 The code below is removed for NetWare because it abends/crashes on NetWare
10720 when the script has error such as not having the closing quotes like:
10721 if ($var eq "value)
10722 Checking of white spaces is anyway done in NetWare code.
10725 while (isSPACE(*PL_oldbufptr))
10728 context = PL_oldbufptr;
10729 contlen = PL_bufptr - PL_oldbufptr;
10731 else if (yychar > 255)
10732 sv_catpvs(where_sv, "next token ???");
10733 else if (yychar == YYEMPTY) {
10734 if (PL_lex_state == LEX_NORMAL ||
10735 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10736 sv_catpvs(where_sv, "at end of line");
10737 else if (PL_lex_inpat)
10738 sv_catpvs(where_sv, "within pattern");
10740 sv_catpvs(where_sv, "within string");
10743 sv_catpvs(where_sv, "next char ");
10745 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10746 else if (isPRINT_LC(yychar)) {
10747 const char string = yychar;
10748 sv_catpvn(where_sv, &string, 1);
10751 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10753 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
10754 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10755 OutCopFILE(PL_curcop),
10756 (IV)(PL_parser->preambling == NOLINE
10757 ? CopLINE(PL_curcop)
10758 : PL_parser->preambling));
10760 Perl_sv_catpvf(aTHX_ msg, "near \"%"UTF8f"\"\n",
10761 UTF8fARG(UTF, contlen, context));
10763 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
10764 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10765 Perl_sv_catpvf(aTHX_ msg,
10766 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10767 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10770 if (PL_in_eval & EVAL_WARNONLY) {
10771 PL_in_eval &= ~EVAL_WARNONLY;
10772 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10776 if (PL_error_count >= 10) {
10778 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
10779 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10780 SVfARG(errsv), OutCopFILE(PL_curcop));
10782 Perl_croak(aTHX_ "%s has too many errors.\n",
10783 OutCopFILE(PL_curcop));
10786 PL_in_my_stash = NULL;
10791 S_swallow_bom(pTHX_ U8 *s)
10793 const STRLEN slen = SvCUR(PL_linestr);
10795 PERL_ARGS_ASSERT_SWALLOW_BOM;
10799 if (s[1] == 0xFE) {
10800 /* UTF-16 little-endian? (or UTF-32LE?) */
10801 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10802 /* diag_listed_as: Unsupported script encoding %s */
10803 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10804 #ifndef PERL_NO_UTF16_FILTER
10805 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10807 if (PL_bufend > (char*)s) {
10808 s = add_utf16_textfilter(s, TRUE);
10811 /* diag_listed_as: Unsupported script encoding %s */
10812 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10817 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10818 #ifndef PERL_NO_UTF16_FILTER
10819 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10821 if (PL_bufend > (char *)s) {
10822 s = add_utf16_textfilter(s, FALSE);
10825 /* diag_listed_as: Unsupported script encoding %s */
10826 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10830 case BOM_UTF8_FIRST_BYTE: {
10831 const STRLEN len = sizeof(BOM_UTF8_TAIL) - 1; /* Exclude trailing NUL */
10832 if (slen > len && memEQ(s+1, BOM_UTF8_TAIL, len)) {
10833 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10834 s += len + 1; /* UTF-8 */
10841 if (s[2] == 0xFE && s[3] == 0xFF) {
10842 /* UTF-32 big-endian */
10843 /* diag_listed_as: Unsupported script encoding %s */
10844 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10847 else if (s[2] == 0 && s[3] != 0) {
10850 * are a good indicator of UTF-16BE. */
10851 #ifndef PERL_NO_UTF16_FILTER
10852 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10853 s = add_utf16_textfilter(s, FALSE);
10855 /* diag_listed_as: Unsupported script encoding %s */
10856 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10863 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10866 * are a good indicator of UTF-16LE. */
10867 #ifndef PERL_NO_UTF16_FILTER
10868 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10869 s = add_utf16_textfilter(s, TRUE);
10871 /* diag_listed_as: Unsupported script encoding %s */
10872 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10880 #ifndef PERL_NO_UTF16_FILTER
10882 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10884 SV *const filter = FILTER_DATA(idx);
10885 /* We re-use this each time round, throwing the contents away before we
10887 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10888 SV *const utf8_buffer = filter;
10889 IV status = IoPAGE(filter);
10890 const bool reverse = cBOOL(IoLINES(filter));
10893 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10895 /* As we're automatically added, at the lowest level, and hence only called
10896 from this file, we can be sure that we're not called in block mode. Hence
10897 don't bother writing code to deal with block mode. */
10899 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10902 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10904 DEBUG_P(PerlIO_printf(Perl_debug_log,
10905 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10906 FPTR2DPTR(void *, S_utf16_textfilter),
10907 reverse ? 'l' : 'b', idx, maxlen, status,
10908 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10915 /* First, look in our buffer of existing UTF-8 data: */
10916 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10920 } else if (status == 0) {
10922 IoPAGE(filter) = 0;
10923 nl = SvEND(utf8_buffer);
10926 STRLEN got = nl - SvPVX(utf8_buffer);
10927 /* Did we have anything to append? */
10929 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10930 /* Everything else in this code works just fine if SVp_POK isn't
10931 set. This, however, needs it, and we need it to work, else
10932 we loop infinitely because the buffer is never consumed. */
10933 sv_chop(utf8_buffer, nl);
10937 /* OK, not a complete line there, so need to read some more UTF-16.
10938 Read an extra octect if the buffer currently has an odd number. */
10942 if (SvCUR(utf16_buffer) >= 2) {
10943 /* Location of the high octet of the last complete code point.
10944 Gosh, UTF-16 is a pain. All the benefits of variable length,
10945 *coupled* with all the benefits of partial reads and
10947 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10948 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10950 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10954 /* We have the first half of a surrogate. Read more. */
10955 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10958 status = FILTER_READ(idx + 1, utf16_buffer,
10959 160 + (SvCUR(utf16_buffer) & 1));
10960 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10961 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10964 IoPAGE(filter) = status;
10969 chars = SvCUR(utf16_buffer) >> 1;
10970 have = SvCUR(utf8_buffer);
10971 SvGROW(utf8_buffer, have + chars * 3 + 1);
10974 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10975 (U8*)SvPVX_const(utf8_buffer) + have,
10976 chars * 2, &newlen);
10978 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10979 (U8*)SvPVX_const(utf8_buffer) + have,
10980 chars * 2, &newlen);
10982 SvCUR_set(utf8_buffer, have + newlen);
10985 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10986 it's private to us, and utf16_to_utf8{,reversed} take a
10987 (pointer,length) pair, rather than a NUL-terminated string. */
10988 if(SvCUR(utf16_buffer) & 1) {
10989 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10990 SvCUR_set(utf16_buffer, 1);
10992 SvCUR_set(utf16_buffer, 0);
10995 DEBUG_P(PerlIO_printf(Perl_debug_log,
10996 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10998 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10999 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11004 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11006 SV *filter = filter_add(S_utf16_textfilter, NULL);
11008 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11010 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11011 sv_setpvs(filter, "");
11012 IoLINES(filter) = reversed;
11013 IoPAGE(filter) = 1; /* Not EOF */
11015 /* Sadly, we have to return a valid pointer, come what may, so we have to
11016 ignore any error return from this. */
11017 SvCUR_set(PL_linestr, 0);
11018 if (FILTER_READ(0, PL_linestr, 0)) {
11019 SvUTF8_on(PL_linestr);
11021 SvUTF8_on(PL_linestr);
11023 PL_bufend = SvEND(PL_linestr);
11024 return (U8*)SvPVX(PL_linestr);
11029 Returns a pointer to the next character after the parsed
11030 vstring, as well as updating the passed in sv.
11032 Function must be called like
11034 sv = sv_2mortal(newSV(5));
11035 s = scan_vstring(s,e,sv);
11037 where s and e are the start and end of the string.
11038 The sv should already be large enough to store the vstring
11039 passed in, for performance reasons.
11041 This function may croak if fatal warnings are enabled in the
11042 calling scope, hence the sv_2mortal in the example (to prevent
11043 a leak). Make sure to do SvREFCNT_inc afterwards if you use
11049 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11051 const char *pos = s;
11052 const char *start = s;
11054 PERL_ARGS_ASSERT_SCAN_VSTRING;
11056 if (*pos == 'v') pos++; /* get past 'v' */
11057 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11059 if ( *pos != '.') {
11060 /* this may not be a v-string if followed by => */
11061 const char *next = pos;
11062 while (next < e && isSPACE(*next))
11064 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11065 /* return string not v-string */
11066 sv_setpvn(sv,(char *)s,pos-s);
11067 return (char *)pos;
11071 if (!isALPHA(*pos)) {
11072 U8 tmpbuf[UTF8_MAXBYTES+1];
11075 s++; /* get past 'v' */
11080 /* this is atoi() that tolerates underscores */
11083 const char *end = pos;
11085 while (--end >= s) {
11087 const UV orev = rev;
11088 rev += (*end - '0') * mult;
11091 /* diag_listed_as: Integer overflow in %s number */
11092 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11093 "Integer overflow in decimal number");
11097 if (rev > 0x7FFFFFFF)
11098 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11100 /* Append native character for the rev point */
11101 tmpend = uvchr_to_utf8(tmpbuf, rev);
11102 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11103 if (!UVCHR_IS_INVARIANT(rev))
11105 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11111 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11115 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11122 Perl_keyword_plugin_standard(pTHX_
11123 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11125 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11126 PERL_UNUSED_CONTEXT;
11127 PERL_UNUSED_ARG(keyword_ptr);
11128 PERL_UNUSED_ARG(keyword_len);
11129 PERL_UNUSED_ARG(op_ptr);
11130 return KEYWORD_PLUGIN_DECLINE;
11133 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11135 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11137 SAVEI32(PL_lex_brackets);
11138 if (PL_lex_brackets > 100)
11139 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11140 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11141 SAVEI32(PL_lex_allbrackets);
11142 PL_lex_allbrackets = 0;
11143 SAVEI8(PL_lex_fakeeof);
11144 PL_lex_fakeeof = (U8)fakeeof;
11145 if(yyparse(gramtype) && !PL_parser->error_count)
11146 qerror(Perl_mess(aTHX_ "Parse error"));
11149 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11151 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11155 SAVEVPTR(PL_eval_root);
11156 PL_eval_root = NULL;
11157 parse_recdescent(gramtype, fakeeof);
11163 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11165 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11168 if (flags & ~PARSE_OPTIONAL)
11169 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11170 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11171 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11172 if (!PL_parser->error_count)
11173 qerror(Perl_mess(aTHX_ "Parse error"));
11174 exprop = newOP(OP_NULL, 0);
11180 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11182 Parse a Perl arithmetic expression. This may contain operators of precedence
11183 down to the bit shift operators. The expression must be followed (and thus
11184 terminated) either by a comparison or lower-precedence operator or by
11185 something that would normally terminate an expression such as semicolon.
11186 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11187 otherwise it is mandatory. It is up to the caller to ensure that the
11188 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11189 the source of the code to be parsed and the lexical context for the
11192 The op tree representing the expression is returned. If an optional
11193 expression is absent, a null pointer is returned, otherwise the pointer
11196 If an error occurs in parsing or compilation, in most cases a valid op
11197 tree is returned anyway. The error is reflected in the parser state,
11198 normally resulting in a single exception at the top level of parsing
11199 which covers all the compilation errors that occurred. Some compilation
11200 errors, however, will throw an exception immediately.
11206 Perl_parse_arithexpr(pTHX_ U32 flags)
11208 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11212 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11214 Parse a Perl term expression. This may contain operators of precedence
11215 down to the assignment operators. The expression must be followed (and thus
11216 terminated) either by a comma or lower-precedence operator or by
11217 something that would normally terminate an expression such as semicolon.
11218 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11219 otherwise it is mandatory. It is up to the caller to ensure that the
11220 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11221 the source of the code to be parsed and the lexical context for the
11224 The op tree representing the expression is returned. If an optional
11225 expression is absent, a null pointer is returned, otherwise the pointer
11228 If an error occurs in parsing or compilation, in most cases a valid op
11229 tree is returned anyway. The error is reflected in the parser state,
11230 normally resulting in a single exception at the top level of parsing
11231 which covers all the compilation errors that occurred. Some compilation
11232 errors, however, will throw an exception immediately.
11238 Perl_parse_termexpr(pTHX_ U32 flags)
11240 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11244 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11246 Parse a Perl list expression. This may contain operators of precedence
11247 down to the comma operator. The expression must be followed (and thus
11248 terminated) either by a low-precedence logic operator such as C<or> or by
11249 something that would normally terminate an expression such as semicolon.
11250 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11251 otherwise it is mandatory. It is up to the caller to ensure that the
11252 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11253 the source of the code to be parsed and the lexical context for the
11256 The op tree representing the expression is returned. If an optional
11257 expression is absent, a null pointer is returned, otherwise the pointer
11260 If an error occurs in parsing or compilation, in most cases a valid op
11261 tree is returned anyway. The error is reflected in the parser state,
11262 normally resulting in a single exception at the top level of parsing
11263 which covers all the compilation errors that occurred. Some compilation
11264 errors, however, will throw an exception immediately.
11270 Perl_parse_listexpr(pTHX_ U32 flags)
11272 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11276 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11278 Parse a single complete Perl expression. This allows the full
11279 expression grammar, including the lowest-precedence operators such
11280 as C<or>. The expression must be followed (and thus terminated) by a
11281 token that an expression would normally be terminated by: end-of-file,
11282 closing bracketing punctuation, semicolon, or one of the keywords that
11283 signals a postfix expression-statement modifier. If I<flags> includes
11284 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11285 mandatory. It is up to the caller to ensure that the dynamic parser
11286 state (L</PL_parser> et al) is correctly set to reflect the source of
11287 the code to be parsed and the lexical context for the expression.
11289 The op tree representing the expression is returned. If an optional
11290 expression is absent, a null pointer is returned, otherwise the pointer
11293 If an error occurs in parsing or compilation, in most cases a valid op
11294 tree is returned anyway. The error is reflected in the parser state,
11295 normally resulting in a single exception at the top level of parsing
11296 which covers all the compilation errors that occurred. Some compilation
11297 errors, however, will throw an exception immediately.
11303 Perl_parse_fullexpr(pTHX_ U32 flags)
11305 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11309 =for apidoc Amx|OP *|parse_block|U32 flags
11311 Parse a single complete Perl code block. This consists of an opening
11312 brace, a sequence of statements, and a closing brace. The block
11313 constitutes a lexical scope, so C<my> variables and various compile-time
11314 effects can be contained within it. It is up to the caller to ensure
11315 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11316 reflect the source of the code to be parsed and the lexical context for
11319 The op tree representing the code block is returned. This is always a
11320 real op, never a null pointer. It will normally be a C<lineseq> list,
11321 including C<nextstate> or equivalent ops. No ops to construct any kind
11322 of runtime scope are included by virtue of it being a block.
11324 If an error occurs in parsing or compilation, in most cases a valid op
11325 tree (most likely null) is returned anyway. The error is reflected in
11326 the parser state, normally resulting in a single exception at the top
11327 level of parsing which covers all the compilation errors that occurred.
11328 Some compilation errors, however, will throw an exception immediately.
11330 The I<flags> parameter is reserved for future use, and must always
11337 Perl_parse_block(pTHX_ U32 flags)
11340 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11341 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11345 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11347 Parse a single unadorned Perl statement. This may be a normal imperative
11348 statement or a declaration that has compile-time effect. It does not
11349 include any label or other affixture. It is up to the caller to ensure
11350 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11351 reflect the source of the code to be parsed and the lexical context for
11354 The op tree representing the statement is returned. This may be a
11355 null pointer if the statement is null, for example if it was actually
11356 a subroutine definition (which has compile-time side effects). If not
11357 null, it will be ops directly implementing the statement, suitable to
11358 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11359 equivalent op (except for those embedded in a scope contained entirely
11360 within the statement).
11362 If an error occurs in parsing or compilation, in most cases a valid op
11363 tree (most likely null) is returned anyway. The error is reflected in
11364 the parser state, normally resulting in a single exception at the top
11365 level of parsing which covers all the compilation errors that occurred.
11366 Some compilation errors, however, will throw an exception immediately.
11368 The I<flags> parameter is reserved for future use, and must always
11375 Perl_parse_barestmt(pTHX_ U32 flags)
11378 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11379 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11383 =for apidoc Amx|SV *|parse_label|U32 flags
11385 Parse a single label, possibly optional, of the type that may prefix a
11386 Perl statement. It is up to the caller to ensure that the dynamic parser
11387 state (L</PL_parser> et al) is correctly set to reflect the source of
11388 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11389 label is optional, otherwise it is mandatory.
11391 The name of the label is returned in the form of a fresh scalar. If an
11392 optional label is absent, a null pointer is returned.
11394 If an error occurs in parsing, which can only occur if the label is
11395 mandatory, a valid label is returned anyway. The error is reflected in
11396 the parser state, normally resulting in a single exception at the top
11397 level of parsing which covers all the compilation errors that occurred.
11403 Perl_parse_label(pTHX_ U32 flags)
11405 if (flags & ~PARSE_OPTIONAL)
11406 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11407 if (PL_lex_state == LEX_KNOWNEXT) {
11408 PL_parser->yychar = yylex();
11409 if (PL_parser->yychar == LABEL) {
11410 char * const lpv = pl_yylval.pval;
11411 STRLEN llen = strlen(lpv);
11412 PL_parser->yychar = YYEMPTY;
11413 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11420 STRLEN wlen, bufptr_pos;
11423 if (!isIDFIRST_lazy_if(s, UTF))
11425 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11426 if (word_takes_any_delimeter(s, wlen))
11428 bufptr_pos = s - SvPVX(PL_linestr);
11430 lex_read_space(LEX_KEEP_PREVIOUS);
11432 s = SvPVX(PL_linestr) + bufptr_pos;
11433 if (t[0] == ':' && t[1] != ':') {
11434 PL_oldoldbufptr = PL_oldbufptr;
11437 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11441 if (flags & PARSE_OPTIONAL) {
11444 qerror(Perl_mess(aTHX_ "Parse error"));
11445 return newSVpvs("x");
11452 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11454 Parse a single complete Perl statement. This may be a normal imperative
11455 statement or a declaration that has compile-time effect, and may include
11456 optional labels. It is up to the caller to ensure that the dynamic
11457 parser state (L</PL_parser> et al) is correctly set to reflect the source
11458 of the code to be parsed and the lexical context for the statement.
11460 The op tree representing the statement is returned. This may be a
11461 null pointer if the statement is null, for example if it was actually
11462 a subroutine definition (which has compile-time side effects). If not
11463 null, it will be the result of a L</newSTATEOP> call, normally including
11464 a C<nextstate> or equivalent op.
11466 If an error occurs in parsing or compilation, in most cases a valid op
11467 tree (most likely null) is returned anyway. The error is reflected in
11468 the parser state, normally resulting in a single exception at the top
11469 level of parsing which covers all the compilation errors that occurred.
11470 Some compilation errors, however, will throw an exception immediately.
11472 The I<flags> parameter is reserved for future use, and must always
11479 Perl_parse_fullstmt(pTHX_ U32 flags)
11482 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11483 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11487 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11489 Parse a sequence of zero or more Perl statements. These may be normal
11490 imperative statements, including optional labels, or declarations
11491 that have compile-time effect, or any mixture thereof. The statement
11492 sequence ends when a closing brace or end-of-file is encountered in a
11493 place where a new statement could have validly started. It is up to
11494 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11495 is correctly set to reflect the source of the code to be parsed and the
11496 lexical context for the statements.
11498 The op tree representing the statement sequence is returned. This may
11499 be a null pointer if the statements were all null, for example if there
11500 were no statements or if there were only subroutine definitions (which
11501 have compile-time side effects). If not null, it will be a C<lineseq>
11502 list, normally including C<nextstate> or equivalent ops.
11504 If an error occurs in parsing or compilation, in most cases a valid op
11505 tree is returned anyway. The error is reflected in the parser state,
11506 normally resulting in a single exception at the top level of parsing
11507 which covers all the compilation errors that occurred. Some compilation
11508 errors, however, will throw an exception immediately.
11510 The I<flags> parameter is reserved for future use, and must always
11517 Perl_parse_stmtseq(pTHX_ U32 flags)
11522 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11523 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11524 c = lex_peek_unichar(0);
11525 if (c != -1 && c != /*{*/'}')
11526 qerror(Perl_mess(aTHX_ "Parse error"));
11530 #define lex_token_boundary() S_lex_token_boundary(aTHX)
11532 S_lex_token_boundary(pTHX)
11534 PL_oldoldbufptr = PL_oldbufptr;
11535 PL_oldbufptr = PL_bufptr;
11538 #define parse_opt_lexvar() S_parse_opt_lexvar(aTHX)
11540 S_parse_opt_lexvar(pTHX)
11545 lex_token_boundary();
11546 sigil = lex_read_unichar(0);
11547 if (lex_peek_unichar(0) == '#') {
11548 qerror(Perl_mess(aTHX_ "Parse error"));
11552 c = lex_peek_unichar(0);
11553 if (c == -1 || !(UTF ? isIDFIRST_uni(c) : isIDFIRST_A(c)))
11556 d = PL_tokenbuf + 1;
11557 PL_tokenbuf[0] = (char)sigil;
11558 parse_ident(&s, &d, PL_tokenbuf + sizeof(PL_tokenbuf) - 1, 0, cBOOL(UTF));
11560 if (d == PL_tokenbuf+1)
11562 var = newOP(sigil == '$' ? OP_PADSV : sigil == '@' ? OP_PADAV : OP_PADHV,
11563 OPf_MOD | (OPpLVAL_INTRO<<8));
11564 var->op_targ = allocmy(PL_tokenbuf, d - PL_tokenbuf, UTF ? SVf_UTF8 : 0);
11569 Perl_parse_subsignature(pTHX)
11572 int prev_type = 0, pos = 0, min_arity = 0, max_arity = 0;
11573 OP *initops = NULL;
11575 c = lex_peek_unichar(0);
11576 while (c != /*(*/')') {
11580 if (prev_type == 2)
11581 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11582 var = parse_opt_lexvar();
11584 newBINOP(OP_AELEM, 0,
11585 ref(newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv)),
11587 newSVOP(OP_CONST, 0, newSViv(pos))) :
11590 c = lex_peek_unichar(0);
11592 lex_token_boundary();
11593 lex_read_unichar(0);
11595 c = lex_peek_unichar(0);
11596 if (c == ',' || c == /*(*/')') {
11598 qerror(Perl_mess(aTHX_ "Optional parameter "
11599 "lacks default expression"));
11601 OP *defexpr = parse_termexpr(0);
11602 if (defexpr->op_type == OP_UNDEF &&
11603 !(defexpr->op_flags & OPf_KIDS)) {
11608 scalar(newUNOP(OP_RV2AV, 0,
11609 newGVOP(OP_GV, 0, PL_defgv))),
11610 newSVOP(OP_CONST, 0, newSViv(pos+1)));
11612 newCONDOP(0, ifop, expr, defexpr) :
11613 newLOGOP(OP_OR, 0, ifop, defexpr);
11618 if (prev_type == 1)
11619 qerror(Perl_mess(aTHX_ "Mandatory parameter "
11620 "follows optional parameter"));
11622 min_arity = pos + 1;
11624 if (var) expr = newASSIGNOP(OPf_STACKED, var, 0, expr);
11626 initops = op_append_list(OP_LINESEQ, initops,
11627 newSTATEOP(0, NULL, expr));
11633 if (prev_type == 2)
11634 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11635 var = parse_opt_lexvar();
11637 OP *chkop = newLOGOP((pos & 1) ? OP_OR : OP_AND, 0,
11638 newBINOP(OP_BIT_AND, 0,
11639 scalar(newUNOP(OP_RV2AV, 0,
11640 newGVOP(OP_GV, 0, PL_defgv))),
11641 newSVOP(OP_CONST, 0, newSViv(1))),
11642 op_convert_list(OP_DIE, 0,
11643 op_convert_list(OP_SPRINTF, 0,
11644 op_append_list(OP_LIST,
11645 newSVOP(OP_CONST, 0,
11646 newSVpvs("Odd name/value argument for subroutine at %s line %d.\n")),
11648 op_append_list(OP_LIST,
11649 newSVOP(OP_CONST, 0, newSViv(1)),
11650 newSVOP(OP_CONST, 0, newSViv(2))),
11651 newOP(OP_CALLER, 0))))));
11652 if (pos != min_arity)
11653 chkop = newLOGOP(OP_AND, 0,
11655 scalar(newUNOP(OP_RV2AV, 0,
11656 newGVOP(OP_GV, 0, PL_defgv))),
11657 newSVOP(OP_CONST, 0, newSViv(pos))),
11659 initops = op_append_list(OP_LINESEQ,
11660 newSTATEOP(0, NULL, chkop),
11665 op_prepend_elem(OP_ASLICE,
11666 newOP(OP_PUSHMARK, 0),
11667 newLISTOP(OP_ASLICE, 0,
11669 newSVOP(OP_CONST, 0, newSViv(pos)),
11670 newUNOP(OP_AV2ARYLEN, 0,
11671 ref(newUNOP(OP_RV2AV, 0,
11672 newGVOP(OP_GV, 0, PL_defgv)),
11674 ref(newUNOP(OP_RV2AV, 0,
11675 newGVOP(OP_GV, 0, PL_defgv)),
11677 newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv));
11678 initops = op_append_list(OP_LINESEQ, initops,
11679 newSTATEOP(0, NULL,
11680 newASSIGNOP(OPf_STACKED, var, 0, slice)));
11687 qerror(Perl_mess(aTHX_ "Parse error"));
11691 c = lex_peek_unichar(0);
11693 case /*(*/')': break;
11696 lex_token_boundary();
11697 lex_read_unichar(0);
11699 c = lex_peek_unichar(0);
11700 } while (c == ',');
11706 if (min_arity != 0) {
11707 initops = op_append_list(OP_LINESEQ,
11708 newSTATEOP(0, NULL,
11711 scalar(newUNOP(OP_RV2AV, 0,
11712 newGVOP(OP_GV, 0, PL_defgv))),
11713 newSVOP(OP_CONST, 0, newSViv(min_arity))),
11714 op_convert_list(OP_DIE, 0,
11715 op_convert_list(OP_SPRINTF, 0,
11716 op_append_list(OP_LIST,
11717 newSVOP(OP_CONST, 0,
11718 newSVpvs("Too few arguments for subroutine at %s line %d.\n")),
11720 op_append_list(OP_LIST,
11721 newSVOP(OP_CONST, 0, newSViv(1)),
11722 newSVOP(OP_CONST, 0, newSViv(2))),
11723 newOP(OP_CALLER, 0))))))),
11726 if (max_arity != -1) {
11727 initops = op_append_list(OP_LINESEQ,
11728 newSTATEOP(0, NULL,
11731 scalar(newUNOP(OP_RV2AV, 0,
11732 newGVOP(OP_GV, 0, PL_defgv))),
11733 newSVOP(OP_CONST, 0, newSViv(max_arity))),
11734 op_convert_list(OP_DIE, 0,
11735 op_convert_list(OP_SPRINTF, 0,
11736 op_append_list(OP_LIST,
11737 newSVOP(OP_CONST, 0,
11738 newSVpvs("Too many arguments for subroutine at %s line %d.\n")),
11740 op_append_list(OP_LIST,
11741 newSVOP(OP_CONST, 0, newSViv(1)),
11742 newSVOP(OP_CONST, 0, newSViv(2))),
11743 newOP(OP_CALLER, 0))))))),
11750 * ex: set ts=8 sts=4 sw=4 et: