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 C<line> and C<rsfp>. C<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 C<line>
658 does not affect parsing. C<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 C<line> comes first and must consist of complete lines of input,
661 and C<rsfp> supplies the remainder of the source.
663 The C<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 C<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 C<len> octets starting
952 at C<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 C<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++ = EIGHT_BIT_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 C<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 C<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 C<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 C<ptr>. Text following C<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 C<ptr>. This advances L</PL_parser-E<gt>bufptr> to match C<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 C<ptr>. The remaining content of the buffer will be moved, and
1183 all pointers into the buffer updated appropriately. C<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 C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, 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)
1276 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1278 if (PL_parser->last_uni != PL_parser->bufend)
1279 PL_parser->last_uni = NULL;
1280 if (PL_parser->last_lop != PL_parser->bufend)
1281 PL_parser->last_lop = NULL;
1282 last_uni_pos = last_lop_pos = 0;
1286 old_bufend_pos = PL_parser->bufend - buf;
1287 bufptr_pos = PL_parser->bufptr - buf;
1288 oldbufptr_pos = PL_parser->oldbufptr - buf;
1289 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1290 linestart_pos = PL_parser->linestart - buf;
1291 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1292 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1294 if (flags & LEX_FAKE_EOF) {
1296 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1298 } else if (filter_gets(linestr, old_bufend_pos)) {
1300 got_some_for_debugger = 1;
1301 } else if (flags & LEX_NO_TERM) {
1304 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1305 sv_setpvs(linestr, "");
1307 /* End of real input. Close filehandle (unless it was STDIN),
1308 * then add implicit termination.
1310 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1311 PerlIO_clearerr(PL_parser->rsfp);
1312 else if (PL_parser->rsfp)
1313 (void)PerlIO_close(PL_parser->rsfp);
1314 PL_parser->rsfp = NULL;
1315 PL_parser->in_pod = PL_parser->filtered = 0;
1316 if (!PL_in_eval && PL_minus_p) {
1318 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1319 PL_minus_n = PL_minus_p = 0;
1320 } else if (!PL_in_eval && PL_minus_n) {
1321 sv_catpvs(linestr, /*{*/";}");
1324 sv_catpvs(linestr, ";");
1327 buf = SvPVX(linestr);
1328 new_bufend_pos = SvCUR(linestr);
1329 PL_parser->bufend = buf + new_bufend_pos;
1330 PL_parser->bufptr = buf + bufptr_pos;
1331 PL_parser->oldbufptr = buf + oldbufptr_pos;
1332 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1333 PL_parser->linestart = buf + linestart_pos;
1334 if (PL_parser->last_uni)
1335 PL_parser->last_uni = buf + last_uni_pos;
1336 if (PL_parser->last_lop)
1337 PL_parser->last_lop = buf + last_lop_pos;
1338 if (PL_parser->preambling != NOLINE) {
1339 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1340 PL_parser->preambling = NOLINE;
1342 if ( got_some_for_debugger
1343 && PERLDB_LINE_OR_SAVESRC
1344 && PL_curstash != PL_debstash)
1346 /* debugger active and we're not compiling the debugger code,
1347 * so store the line into the debugger's array of lines
1349 update_debugger_info(NULL, buf+old_bufend_pos,
1350 new_bufend_pos-old_bufend_pos);
1356 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1358 Looks ahead one (Unicode) character in the text currently being lexed.
1359 Returns the codepoint (unsigned integer value) of the next character,
1360 or -1 if lexing has reached the end of the input text. To consume the
1361 peeked character, use L</lex_read_unichar>.
1363 If the next character is in (or extends into) the next chunk of input
1364 text, the next chunk will be read in. Normally the current chunk will be
1365 discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1366 bit set, then the current chunk will not be discarded.
1368 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1369 is encountered, an exception is generated.
1375 Perl_lex_peek_unichar(pTHX_ U32 flags)
1379 if (flags & ~(LEX_KEEP_PREVIOUS))
1380 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1381 s = PL_parser->bufptr;
1382 bufend = PL_parser->bufend;
1388 if (!lex_next_chunk(flags))
1390 s = PL_parser->bufptr;
1391 bufend = PL_parser->bufend;
1394 if (UTF8_IS_INVARIANT(head))
1396 if (UTF8_IS_START(head)) {
1397 len = UTF8SKIP(&head);
1398 while ((STRLEN)(bufend-s) < len) {
1399 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1401 s = PL_parser->bufptr;
1402 bufend = PL_parser->bufend;
1405 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1406 if (retlen == (STRLEN)-1) {
1407 /* malformed UTF-8 */
1409 SAVESPTR(PL_warnhook);
1410 PL_warnhook = PERL_WARNHOOK_FATAL;
1411 utf8n_to_uvchr((U8*)s, bufend-s, NULL, 0);
1417 if (!lex_next_chunk(flags))
1419 s = PL_parser->bufptr;
1426 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1428 Reads the next (Unicode) character in the text currently being lexed.
1429 Returns the codepoint (unsigned integer value) of the character read,
1430 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1431 if lexing has reached the end of the input text. To non-destructively
1432 examine the next character, use L</lex_peek_unichar> instead.
1434 If the next character is in (or extends into) the next chunk of input
1435 text, the next chunk will be read in. Normally the current chunk will be
1436 discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1437 bit set, then the current chunk will not be discarded.
1439 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1440 is encountered, an exception is generated.
1446 Perl_lex_read_unichar(pTHX_ U32 flags)
1449 if (flags & ~(LEX_KEEP_PREVIOUS))
1450 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1451 c = lex_peek_unichar(flags);
1454 COPLINE_INC_WITH_HERELINES;
1456 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1458 ++(PL_parser->bufptr);
1464 =for apidoc Amx|void|lex_read_space|U32 flags
1466 Reads optional spaces, in Perl style, in the text currently being
1467 lexed. The spaces may include ordinary whitespace characters and
1468 Perl-style comments. C<#line> directives are processed if encountered.
1469 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1470 at a non-space character (or the end of the input text).
1472 If spaces extend into the next chunk of input text, the next chunk will
1473 be read in. Normally the current chunk will be discarded at the same
1474 time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
1475 chunk will not be discarded.
1480 #define LEX_NO_INCLINE 0x40000000
1481 #define LEX_NO_NEXT_CHUNK 0x80000000
1484 Perl_lex_read_space(pTHX_ U32 flags)
1487 const bool can_incline = !(flags & LEX_NO_INCLINE);
1488 bool need_incline = 0;
1489 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
1490 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1491 s = PL_parser->bufptr;
1492 bufend = PL_parser->bufend;
1498 } while (!(c == '\n' || (c == 0 && s == bufend)));
1499 } else if (c == '\n') {
1502 PL_parser->linestart = s;
1508 } else if (isSPACE(c)) {
1510 } else if (c == 0 && s == bufend) {
1513 if (flags & LEX_NO_NEXT_CHUNK)
1515 PL_parser->bufptr = s;
1516 l = CopLINE(PL_curcop);
1517 CopLINE(PL_curcop) += PL_parser->herelines + 1;
1518 got_more = lex_next_chunk(flags);
1519 CopLINE_set(PL_curcop, l);
1520 s = PL_parser->bufptr;
1521 bufend = PL_parser->bufend;
1524 if (can_incline && need_incline && PL_parser->rsfp) {
1534 PL_parser->bufptr = s;
1539 =for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1541 This function performs syntax checking on a prototype, C<proto>.
1542 If C<warn> is true, any illegal characters or mismatched brackets
1543 will trigger illegalproto warnings, declaring that they were
1544 detected in the prototype for C<name>.
1546 The return value is C<true> if this is a valid prototype, and
1547 C<false> if it is not, regardless of whether C<warn> was C<true> or
1550 Note that C<NULL> is a valid C<proto> and will always return C<true>.
1557 Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn)
1559 STRLEN len, origlen;
1560 char *p = proto ? SvPV(proto, len) : NULL;
1561 bool bad_proto = FALSE;
1562 bool in_brackets = FALSE;
1563 bool after_slash = FALSE;
1564 char greedy_proto = ' ';
1565 bool proto_after_greedy_proto = FALSE;
1566 bool must_be_last = FALSE;
1567 bool underscore = FALSE;
1568 bool bad_proto_after_underscore = FALSE;
1570 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1576 for (; len--; p++) {
1579 proto_after_greedy_proto = TRUE;
1581 if (!strchr(";@%", *p))
1582 bad_proto_after_underscore = TRUE;
1585 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1592 in_brackets = FALSE;
1593 else if ((*p == '@' || *p == '%')
1597 must_be_last = TRUE;
1606 after_slash = FALSE;
1611 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
1614 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1615 origlen, UNI_DISPLAY_ISPRINT)
1616 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1618 if (proto_after_greedy_proto)
1619 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1620 "Prototype after '%c' for %"SVf" : %s",
1621 greedy_proto, SVfARG(name), p);
1623 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1624 "Missing ']' in prototype for %"SVf" : %s",
1627 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1628 "Illegal character in prototype for %"SVf" : %s",
1630 if (bad_proto_after_underscore)
1631 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
1632 "Illegal character after '_' in prototype for %"SVf" : %s",
1636 return (! (proto_after_greedy_proto || bad_proto) );
1641 * This subroutine has nothing to do with tilting, whether at windmills
1642 * or pinball tables. Its name is short for "increment line". It
1643 * increments the current line number in CopLINE(PL_curcop) and checks
1644 * to see whether the line starts with a comment of the form
1645 * # line 500 "foo.pm"
1646 * If so, it sets the current line number and file to the values in the comment.
1650 S_incline(pTHX_ const char *s)
1658 PERL_ARGS_ASSERT_INCLINE;
1660 COPLINE_INC_WITH_HERELINES;
1661 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1662 && s+1 == PL_bufend && *s == ';') {
1663 /* fake newline in string eval */
1664 CopLINE_dec(PL_curcop);
1669 while (SPACE_OR_TAB(*s))
1671 if (strnEQ(s, "line", 4))
1675 if (SPACE_OR_TAB(*s))
1679 while (SPACE_OR_TAB(*s))
1687 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1689 while (SPACE_OR_TAB(*s))
1691 if (*s == '"' && (t = strchr(s+1, '"'))) {
1697 while (!isSPACE(*t))
1701 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1703 if (*e != '\n' && *e != '\0')
1704 return; /* false alarm */
1706 if (!grok_atoUV(n, &uv, &e))
1708 line_num = ((line_t)uv) - 1;
1711 const STRLEN len = t - s;
1713 if (!PL_rsfp && !PL_parser->filtered) {
1714 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1715 * to *{"::_<newfilename"} */
1716 /* However, the long form of evals is only turned on by the
1717 debugger - usually they're "(eval %lu)" */
1718 GV * const cfgv = CopFILEGV(PL_curcop);
1721 STRLEN tmplen2 = len;
1725 if (tmplen2 + 2 <= sizeof smallbuf)
1728 Newx(tmpbuf2, tmplen2 + 2, char);
1733 memcpy(tmpbuf2 + 2, s, tmplen2);
1736 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1738 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1739 /* adjust ${"::_<newfilename"} to store the new file name */
1740 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1741 /* The line number may differ. If that is the case,
1742 alias the saved lines that are in the array.
1743 Otherwise alias the whole array. */
1744 if (CopLINE(PL_curcop) == line_num) {
1745 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1746 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
1748 else if (GvAV(cfgv)) {
1749 AV * const av = GvAV(cfgv);
1750 const I32 start = CopLINE(PL_curcop)+1;
1751 I32 items = AvFILLp(av) - start;
1753 AV * const av2 = GvAVn(gv2);
1754 SV **svp = AvARRAY(av) + start;
1755 I32 l = (I32)line_num+1;
1757 av_store(av2, l++, SvREFCNT_inc(*svp++));
1762 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1765 CopFILE_free(PL_curcop);
1766 CopFILE_setn(PL_curcop, s, len);
1768 CopLINE_set(PL_curcop, line_num);
1771 #define skipspace(s) skipspace_flags(s, 0)
1775 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1777 AV *av = CopFILEAVx(PL_curcop);
1780 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1782 sv = *av_fetch(av, 0, 1);
1783 SvUPGRADE(sv, SVt_PVMG);
1785 if (!SvPOK(sv)) sv_setpvs(sv,"");
1787 sv_catsv(sv, orig_sv);
1789 sv_catpvn(sv, buf, len);
1794 if (PL_parser->preambling == NOLINE)
1795 av_store(av, CopLINE(PL_curcop), sv);
1801 * Called to gobble the appropriate amount and type of whitespace.
1802 * Skips comments as well.
1806 S_skipspace_flags(pTHX_ char *s, U32 flags)
1808 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
1809 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1810 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
1813 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1815 lex_read_space(flags | LEX_KEEP_PREVIOUS |
1816 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
1817 LEX_NO_NEXT_CHUNK : 0));
1819 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1820 if (PL_linestart > PL_bufptr)
1821 PL_bufptr = PL_linestart;
1829 * Check the unary operators to ensure there's no ambiguity in how they're
1830 * used. An ambiguous piece of code would be:
1832 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1833 * the +5 is its argument.
1842 if (PL_oldoldbufptr != PL_last_uni)
1844 while (isSPACE(*PL_last_uni))
1847 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1848 s += UTF ? UTF8SKIP(s) : 1;
1849 if ((t = strchr(s, '(')) && t < PL_bufptr)
1852 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1853 "Warning: Use of \"%"UTF8f"\" without parentheses is ambiguous",
1854 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
1858 * LOP : macro to build a list operator. Its behaviour has been replaced
1859 * with a subroutine, S_lop() for which LOP is just another name.
1862 #define LOP(f,x) return lop(f,x,s)
1866 * Build a list operator (or something that might be one). The rules:
1867 * - if we have a next token, then it's a list operator (no parens) for
1868 * which the next token has already been parsed; e.g.,
1871 * - if the next thing is an opening paren, then it's a function
1872 * - else it's a list operator
1876 S_lop(pTHX_ I32 f, int x, char *s)
1878 PERL_ARGS_ASSERT_LOP;
1883 PL_last_lop = PL_oldbufptr;
1884 PL_last_lop_op = (OPCODE)f;
1889 return REPORT(FUNC);
1892 return REPORT(FUNC);
1895 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1896 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1897 return REPORT(LSTOP);
1903 * When the lexer realizes it knows the next token (for instance,
1904 * it is reordering tokens for the parser) then it can call S_force_next
1905 * to know what token to return the next time the lexer is called. Caller
1906 * will need to set PL_nextval[] and possibly PL_expect to ensure
1907 * the lexer handles the token correctly.
1911 S_force_next(pTHX_ I32 type)
1915 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1916 tokereport(type, &NEXTVAL_NEXTTOKE);
1919 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
1920 PL_nexttype[PL_nexttoke] = type;
1922 if (PL_lex_state != LEX_KNOWNEXT) {
1923 PL_lex_defer = PL_lex_state;
1924 PL_lex_state = LEX_KNOWNEXT;
1931 * This subroutine handles postfix deref syntax after the arrow has already
1932 * been emitted. @* $* etc. are emitted as two separate token right here.
1933 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
1934 * only the first, leaving yylex to find the next.
1938 S_postderef(pTHX_ int const funny, char const next)
1940 assert(funny == DOLSHARP || strchr("$@%&*", funny));
1941 assert(strchr("*[{", next));
1943 PL_expect = XOPERATOR;
1944 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
1945 assert('@' == funny || '$' == funny || DOLSHARP == funny);
1946 PL_lex_state = LEX_INTERPEND;
1947 force_next(POSTJOIN);
1953 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
1954 && !PL_lex_brackets)
1956 PL_expect = XOPERATOR;
1965 int yyc = PL_parser->yychar;
1966 if (yyc != YYEMPTY) {
1968 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1969 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1970 PL_lex_allbrackets--;
1972 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1973 } else if (yyc == '('/*)*/) {
1974 PL_lex_allbrackets--;
1979 PL_parser->yychar = YYEMPTY;
1984 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1986 SV * const sv = newSVpvn_utf8(start, len,
1989 && !is_invariant_string((const U8*)start, len)
1990 && is_utf8_string((const U8*)start, len));
1996 * When the lexer knows the next thing is a word (for instance, it has
1997 * just seen -> and it knows that the next char is a word char, then
1998 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2002 * char *start : buffer position (must be within PL_linestr)
2003 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2004 * int check_keyword : if true, Perl checks to make sure the word isn't
2005 * a keyword (do this if the word is a label, e.g. goto FOO)
2006 * int allow_pack : if true, : characters will also be allowed (require,
2007 * use, etc. do this)
2011 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
2016 PERL_ARGS_ASSERT_FORCE_WORD;
2018 start = skipspace(start);
2020 if (isIDFIRST_lazy_if(s,UTF)
2021 || (allow_pack && *s == ':') )
2023 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2024 if (check_keyword) {
2025 char *s2 = PL_tokenbuf;
2027 if (allow_pack && len > 6 && strnEQ(s2, "CORE::", 6))
2029 if (keyword(s2, len2, 0))
2032 if (token == METHOD) {
2037 PL_expect = XOPERATOR;
2040 NEXTVAL_NEXTTOKE.opval
2041 = (OP*)newSVOP(OP_CONST,0,
2042 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2043 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2051 * Called when the lexer wants $foo *foo &foo etc, but the program
2052 * text only contains the "foo" portion. The first argument is a pointer
2053 * to the "foo", and the second argument is the type symbol to prefix.
2054 * Forces the next token to be a "WORD".
2055 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2059 S_force_ident(pTHX_ const char *s, int kind)
2061 PERL_ARGS_ASSERT_FORCE_IDENT;
2064 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2065 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2066 UTF ? SVf_UTF8 : 0));
2067 NEXTVAL_NEXTTOKE.opval = o;
2070 o->op_private = OPpCONST_ENTERED;
2071 /* XXX see note in pp_entereval() for why we forgo typo
2072 warnings if the symbol must be introduced in an eval.
2074 gv_fetchpvn_flags(s, len,
2075 (PL_in_eval ? GV_ADDMULTI
2076 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2077 kind == '$' ? SVt_PV :
2078 kind == '@' ? SVt_PVAV :
2079 kind == '%' ? SVt_PVHV :
2087 S_force_ident_maybe_lex(pTHX_ char pit)
2089 NEXTVAL_NEXTTOKE.ival = pit;
2094 Perl_str_to_version(pTHX_ SV *sv)
2099 const char *start = SvPV_const(sv,len);
2100 const char * const end = start + len;
2101 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2103 PERL_ARGS_ASSERT_STR_TO_VERSION;
2105 while (start < end) {
2109 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2114 retval += ((NV)n)/nshift;
2123 * Forces the next token to be a version number.
2124 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2125 * and if "guessing" is TRUE, then no new token is created (and the caller
2126 * must use an alternative parsing method).
2130 S_force_version(pTHX_ char *s, int guessing)
2135 PERL_ARGS_ASSERT_FORCE_VERSION;
2143 while (isDIGIT(*d) || *d == '_' || *d == '.')
2145 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2147 s = scan_num(s, &pl_yylval);
2148 version = pl_yylval.opval;
2149 ver = cSVOPx(version)->op_sv;
2150 if (SvPOK(ver) && !SvNIOK(ver)) {
2151 SvUPGRADE(ver, SVt_PVNV);
2152 SvNV_set(ver, str_to_version(ver));
2153 SvNOK_on(ver); /* hint that it is a version */
2156 else if (guessing) {
2161 /* NOTE: The parser sees the package name and the VERSION swapped */
2162 NEXTVAL_NEXTTOKE.opval = version;
2169 * S_force_strict_version
2170 * Forces the next token to be a version number using strict syntax rules.
2174 S_force_strict_version(pTHX_ char *s)
2177 const char *errstr = NULL;
2179 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2181 while (isSPACE(*s)) /* leading whitespace */
2184 if (is_STRICT_VERSION(s,&errstr)) {
2186 s = (char *)scan_version(s, ver, 0);
2187 version = newSVOP(OP_CONST, 0, ver);
2189 else if ((*s != ';' && *s != '{' && *s != '}' )
2190 && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
2194 yyerror(errstr); /* version required */
2198 /* NOTE: The parser sees the package name and the VERSION swapped */
2199 NEXTVAL_NEXTTOKE.opval = version;
2207 * Tokenize a quoted string passed in as an SV. It finds the next
2208 * chunk, up to end of string or a backslash. It may make a new
2209 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2214 S_tokeq(pTHX_ SV *sv)
2221 PERL_ARGS_ASSERT_TOKEQ;
2225 assert (!SvIsCOW(sv));
2226 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
2230 /* This is relying on the SV being "well formed" with a trailing '\0' */
2231 while (s < send && !(*s == '\\' && s[1] == '\\'))
2236 if ( PL_hints & HINT_NEW_STRING ) {
2237 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2238 SVs_TEMP | SvUTF8(sv));
2242 if (s + 1 < send && (s[1] == '\\'))
2243 s++; /* all that, just for this */
2248 SvCUR_set(sv, d - SvPVX_const(sv));
2250 if ( PL_hints & HINT_NEW_STRING )
2251 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2256 * Now come three functions related to double-quote context,
2257 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2258 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2259 * interact with PL_lex_state, and create fake ( ... ) argument lists
2260 * to handle functions and concatenation.
2264 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2269 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2271 * Pattern matching will set PL_lex_op to the pattern-matching op to
2272 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2274 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2276 * Everything else becomes a FUNC.
2278 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2279 * had an OP_CONST or OP_READLINE). This just sets us up for a
2280 * call to S_sublex_push().
2284 S_sublex_start(pTHX)
2286 const I32 op_type = pl_yylval.ival;
2288 if (op_type == OP_NULL) {
2289 pl_yylval.opval = PL_lex_op;
2293 if (op_type == OP_CONST) {
2294 SV *sv = PL_lex_stuff;
2295 PL_lex_stuff = NULL;
2298 if (SvTYPE(sv) == SVt_PVIV) {
2299 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2301 const char * const p = SvPV_const(sv, len);
2302 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2306 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2310 PL_sublex_info.super_state = PL_lex_state;
2311 PL_sublex_info.sub_inwhat = (U16)op_type;
2312 PL_sublex_info.sub_op = PL_lex_op;
2313 PL_lex_state = LEX_INTERPPUSH;
2317 pl_yylval.opval = PL_lex_op;
2327 * Create a new scope to save the lexing state. The scope will be
2328 * ended in S_sublex_done. Returns a '(', starting the function arguments
2329 * to the uc, lc, etc. found before.
2330 * Sets PL_lex_state to LEX_INTERPCONCAT.
2337 const bool is_heredoc = PL_multi_close == '<';
2340 PL_lex_state = PL_sublex_info.super_state;
2341 SAVEI8(PL_lex_dojoin);
2342 SAVEI32(PL_lex_brackets);
2343 SAVEI32(PL_lex_allbrackets);
2344 SAVEI32(PL_lex_formbrack);
2345 SAVEI8(PL_lex_fakeeof);
2346 SAVEI32(PL_lex_casemods);
2347 SAVEI32(PL_lex_starts);
2348 SAVEI8(PL_lex_state);
2349 SAVEI8(PL_lex_defer);
2350 SAVESPTR(PL_lex_repl);
2351 SAVEVPTR(PL_lex_inpat);
2352 SAVEI16(PL_lex_inwhat);
2355 SAVECOPLINE(PL_curcop);
2356 SAVEI32(PL_multi_end);
2357 SAVEI32(PL_parser->herelines);
2358 PL_parser->herelines = 0;
2360 SAVEI8(PL_multi_close);
2361 SAVEPPTR(PL_bufptr);
2362 SAVEPPTR(PL_bufend);
2363 SAVEPPTR(PL_oldbufptr);
2364 SAVEPPTR(PL_oldoldbufptr);
2365 SAVEPPTR(PL_last_lop);
2366 SAVEPPTR(PL_last_uni);
2367 SAVEPPTR(PL_linestart);
2368 SAVESPTR(PL_linestr);
2369 SAVEGENERICPV(PL_lex_brackstack);
2370 SAVEGENERICPV(PL_lex_casestack);
2371 SAVEGENERICPV(PL_parser->lex_shared);
2372 SAVEBOOL(PL_parser->lex_re_reparsing);
2373 SAVEI32(PL_copline);
2375 /* The here-doc parser needs to be able to peek into outer lexing
2376 scopes to find the body of the here-doc. So we put PL_linestr and
2377 PL_bufptr into lex_shared, to ‘share’ those values.
2379 PL_parser->lex_shared->ls_linestr = PL_linestr;
2380 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2382 PL_linestr = PL_lex_stuff;
2383 PL_lex_repl = PL_sublex_info.repl;
2384 PL_lex_stuff = NULL;
2385 PL_sublex_info.repl = NULL;
2387 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2388 set for an inner quote-like operator and then an error causes scope-
2389 popping. We must not have a PL_lex_stuff value left dangling, as
2390 that breaks assumptions elsewhere. See bug #123617. */
2391 SAVEGENERICSV(PL_lex_stuff);
2392 SAVEGENERICSV(PL_sublex_info.repl);
2394 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2395 = SvPVX(PL_linestr);
2396 PL_bufend += SvCUR(PL_linestr);
2397 PL_last_lop = PL_last_uni = NULL;
2398 SAVEFREESV(PL_linestr);
2399 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2401 PL_lex_dojoin = FALSE;
2402 PL_lex_brackets = PL_lex_formbrack = 0;
2403 PL_lex_allbrackets = 0;
2404 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2405 Newx(PL_lex_brackstack, 120, char);
2406 Newx(PL_lex_casestack, 12, char);
2407 PL_lex_casemods = 0;
2408 *PL_lex_casestack = '\0';
2410 PL_lex_state = LEX_INTERPCONCAT;
2412 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2413 PL_copline = NOLINE;
2415 Newxz(shared, 1, LEXSHARED);
2416 shared->ls_prev = PL_parser->lex_shared;
2417 PL_parser->lex_shared = shared;
2419 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2420 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2421 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2422 PL_lex_inpat = PL_sublex_info.sub_op;
2424 PL_lex_inpat = NULL;
2426 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2427 PL_in_eval &= ~EVAL_RE_REPARSING;
2434 * Restores lexer state after a S_sublex_push.
2440 if (!PL_lex_starts++) {
2441 SV * const sv = newSVpvs("");
2442 if (SvUTF8(PL_linestr))
2444 PL_expect = XOPERATOR;
2445 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2449 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2450 PL_lex_state = LEX_INTERPCASEMOD;
2454 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2455 assert(PL_lex_inwhat != OP_TRANSR);
2457 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
2458 PL_linestr = PL_lex_repl;
2460 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2461 PL_bufend += SvCUR(PL_linestr);
2462 PL_last_lop = PL_last_uni = NULL;
2463 PL_lex_dojoin = FALSE;
2464 PL_lex_brackets = 0;
2465 PL_lex_allbrackets = 0;
2466 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2467 PL_lex_casemods = 0;
2468 *PL_lex_casestack = '\0';
2470 if (SvEVALED(PL_lex_repl)) {
2471 PL_lex_state = LEX_INTERPNORMAL;
2473 /* we don't clear PL_lex_repl here, so that we can check later
2474 whether this is an evalled subst; that means we rely on the
2475 logic to ensure sublex_done() is called again only via the
2476 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2479 PL_lex_state = LEX_INTERPCONCAT;
2482 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2483 CopLINE(PL_curcop) +=
2484 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xpad_cop_seq.xlow
2485 + PL_parser->herelines;
2486 PL_parser->herelines = 0;
2491 const line_t l = CopLINE(PL_curcop);
2493 if (PL_multi_close == '<')
2494 PL_parser->herelines += l - PL_multi_end;
2495 PL_bufend = SvPVX(PL_linestr);
2496 PL_bufend += SvCUR(PL_linestr);
2497 PL_expect = XOPERATOR;
2502 PERL_STATIC_INLINE SV*
2503 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2505 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2506 * interior, hence to the "}". Finds what the name resolves to, returning
2507 * an SV* containing it; NULL if no valid one found */
2509 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2516 const U8* first_bad_char_loc;
2517 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2519 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2524 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2526 &first_bad_char_loc))
2528 /* If warnings are on, this will print a more detailed analysis of what
2529 * is wrong than the error message below */
2530 utf8n_to_uvchr(first_bad_char_loc,
2531 e - ((char *) first_bad_char_loc),
2534 /* We deliberately don't try to print the malformed character, which
2535 * might not print very well; it also may be just the first of many
2536 * malformations, so don't print what comes after it */
2537 yyerror_pv(Perl_form(aTHX_
2538 "Malformed UTF-8 character immediately after '%.*s'",
2539 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr),
2544 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2545 /* include the <}> */
2546 e - backslash_ptr + 1);
2548 SvREFCNT_dec_NN(res);
2552 /* See if the charnames handler is the Perl core's, and if so, we can skip
2553 * the validation needed for a user-supplied one, as Perl's does its own
2555 table = GvHV(PL_hintgv); /* ^H */
2556 cvp = hv_fetchs(table, "charnames", FALSE);
2557 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2558 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
2560 const char * const name = HvNAME(stash);
2561 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2562 && strEQ(name, "_charnames")) {
2567 /* Here, it isn't Perl's charname handler. We can't rely on a
2568 * user-supplied handler to validate the input name. For non-ut8 input,
2569 * look to see that the first character is legal. Then loop through the
2570 * rest checking that each is a continuation */
2572 /* This code makes the reasonable assumption that the only Latin1-range
2573 * characters that begin a character name alias are alphabetic, otherwise
2574 * would have to create a isCHARNAME_BEGIN macro */
2577 if (! isALPHAU(*s)) {
2582 if (! isCHARNAME_CONT(*s)) {
2585 if (*s == ' ' && *(s-1) == ' ') {
2588 if ((U8) *s == NBSP_NATIVE && ckWARN_d(WARN_DEPRECATED)) {
2589 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2590 "NO-BREAK SPACE in a charnames "
2591 "alias definition is deprecated");
2597 /* Similarly for utf8. For invariants can check directly; for other
2598 * Latin1, can calculate their code point and check; otherwise use a
2600 if (UTF8_IS_INVARIANT(*s)) {
2601 if (! isALPHAU(*s)) {
2605 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2606 if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
2612 if (! PL_utf8_charname_begin) {
2613 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2614 PL_utf8_charname_begin = _core_swash_init("utf8",
2615 "_Perl_Charname_Begin",
2617 1, 0, NULL, &flags);
2619 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2626 if (UTF8_IS_INVARIANT(*s)) {
2627 if (! isCHARNAME_CONT(*s)) {
2630 if (*s == ' ' && *(s-1) == ' ') {
2635 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2636 if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
2640 if (*s == *NBSP_UTF8
2641 && *(s+1) == *(NBSP_UTF8+1)
2642 && ckWARN_d(WARN_DEPRECATED))
2644 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
2645 "NO-BREAK SPACE in a charnames "
2646 "alias definition is deprecated");
2651 if (! PL_utf8_charname_continue) {
2652 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2653 PL_utf8_charname_continue = _core_swash_init("utf8",
2654 "_Perl_Charname_Continue",
2656 1, 0, NULL, &flags);
2658 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2665 if (*(s-1) == ' ') {
2668 "charnames alias definitions may not contain trailing "
2669 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2670 (int)(s - backslash_ptr + 1), backslash_ptr,
2671 (int)(e - s + 1), s + 1
2673 UTF ? SVf_UTF8 : 0);
2677 if (SvUTF8(res)) { /* Don't accept malformed input */
2678 const U8* first_bad_char_loc;
2680 const char* const str = SvPV_const(res, len);
2681 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2682 /* If warnings are on, this will print a more detailed analysis of
2683 * what is wrong than the error message below */
2684 utf8n_to_uvchr(first_bad_char_loc,
2685 (char *) first_bad_char_loc - str,
2688 /* We deliberately don't try to print the malformed character,
2689 * which might not print very well; it also may be just the first
2690 * of many malformations, so don't print what comes after it */
2693 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2694 (int) (e - backslash_ptr + 1), backslash_ptr,
2695 (int) ((char *) first_bad_char_loc - str), str
2706 /* The final %.*s makes sure that should the trailing NUL be missing
2707 * that this print won't run off the end of the string */
2710 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2711 (int)(s - backslash_ptr + 1), backslash_ptr,
2712 (int)(e - s + 1), s + 1
2714 UTF ? SVf_UTF8 : 0);
2721 "charnames alias definitions may not contain a sequence of "
2722 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2723 (int)(s - backslash_ptr + 1), backslash_ptr,
2724 (int)(e - s + 1), s + 1
2726 UTF ? SVf_UTF8 : 0);
2733 Extracts the next constant part of a pattern, double-quoted string,
2734 or transliteration. This is terrifying code.
2736 For example, in parsing the double-quoted string "ab\x63$d", it would
2737 stop at the '$' and return an OP_CONST containing 'abc'.
2739 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2740 processing a pattern (PL_lex_inpat is true), a transliteration
2741 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2743 Returns a pointer to the character scanned up to. If this is
2744 advanced from the start pointer supplied (i.e. if anything was
2745 successfully parsed), will leave an OP_CONST for the substring scanned
2746 in pl_yylval. Caller must intuit reason for not parsing further
2747 by looking at the next characters herself.
2751 \N{FOO} => \N{U+hex_for_character_FOO}
2752 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2755 all other \-char, including \N and \N{ apart from \N{ABC}
2758 @ and $ where it appears to be a var, but not for $ as tail anchor
2762 In transliterations:
2763 characters are VERY literal, except for - not at the start or end
2764 of the string, which indicates a range. If the range is in bytes,
2765 scan_const expands the range to the full set of intermediate
2766 characters. If the range is in utf8, the hyphen is replaced with
2767 a certain range mark which will be handled by pmtrans() in op.c.
2769 In double-quoted strings:
2771 double-quoted style: \r and \n
2772 constants: \x31, etc.
2773 deprecated backrefs: \1 (in substitution replacements)
2774 case and quoting: \U \Q \E
2777 scan_const does *not* construct ops to handle interpolated strings.
2778 It stops processing as soon as it finds an embedded $ or @ variable
2779 and leaves it to the caller to work out what's going on.
2781 embedded arrays (whether in pattern or not) could be:
2782 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2784 $ in double-quoted strings must be the symbol of an embedded scalar.
2786 $ in pattern could be $foo or could be tail anchor. Assumption:
2787 it's a tail anchor if $ is the last thing in the string, or if it's
2788 followed by one of "()| \r\n\t"
2790 \1 (backreferences) are turned into $1 in substitutions
2792 The structure of the code is
2793 while (there's a character to process) {
2794 handle transliteration ranges
2795 skip regexp comments /(?#comment)/ and codes /(?{code})/
2796 skip #-initiated comments in //x patterns
2797 check for embedded arrays
2798 check for embedded scalars
2800 deprecate \1 in substitution replacements
2801 handle string-changing backslashes \l \U \Q \E, etc.
2802 switch (what was escaped) {
2803 handle \- in a transliteration (becomes a literal -)
2804 if a pattern and not \N{, go treat as regular character
2805 handle \132 (octal characters)
2806 handle \x15 and \x{1234} (hex characters)
2807 handle \N{name} (named characters, also \N{3,5} in a pattern)
2808 handle \cV (control characters)
2809 handle printf-style backslashes (\f, \r, \n, etc)
2812 } (end if backslash)
2813 handle regular character
2814 } (end while character to read)
2819 S_scan_const(pTHX_ char *start)
2821 char *send = PL_bufend; /* end of the constant */
2822 SV *sv = newSV(send - start); /* sv for the constant. See note below
2824 char *s = start; /* start of the constant */
2825 char *d = SvPVX(sv); /* destination for copies */
2826 bool dorange = FALSE; /* are we in a translit range? */
2827 bool didrange = FALSE; /* did we just finish a range? */
2828 bool in_charclass = FALSE; /* within /[...]/ */
2829 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2830 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2831 UTF8? But, this can show as true
2832 when the source isn't utf8, as for
2833 example when it is entirely composed
2835 SV *res; /* result from charnames */
2836 STRLEN offset_to_max; /* The offset in the output to where the range
2837 high-end character is temporarily placed */
2839 /* Note on sizing: The scanned constant is placed into sv, which is
2840 * initialized by newSV() assuming one byte of output for every byte of
2841 * input. This routine expects newSV() to allocate an extra byte for a
2842 * trailing NUL, which this routine will append if it gets to the end of
2843 * the input. There may be more bytes of input than output (eg., \N{LATIN
2844 * CAPITAL LETTER A}), or more output than input if the constant ends up
2845 * recoded to utf8, but each time a construct is found that might increase
2846 * the needed size, SvGROW() is called. Its size parameter each time is
2847 * based on the best guess estimate at the time, namely the length used so
2848 * far, plus the length the current construct will occupy, plus room for
2849 * the trailing NUL, plus one byte for every input byte still unscanned */
2851 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2854 int backslash_N = 0; /* ? was the character from \N{} */
2855 int non_portable_endpoint = 0; /* ? In a range is an endpoint
2856 platform-specific like \x65 */
2859 PERL_ARGS_ASSERT_SCAN_CONST;
2861 assert(PL_lex_inwhat != OP_TRANSR);
2862 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2863 /* If we are doing a trans and we know we want UTF8 set expectation */
2864 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2865 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2868 /* Protect sv from errors and fatal warnings. */
2869 ENTER_with_name("scan_const");
2873 || dorange /* Handle tr/// range at right edge of input */
2876 /* get transliterations out of the way (they're most literal) */
2877 if (PL_lex_inwhat == OP_TRANS) {
2879 /* But there isn't any special handling necessary unless there is a
2880 * range, so for most cases we just drop down and handle the value
2881 * as any other. There are two exceptions.
2883 * 1. A minus sign indicates that we are actually going to have
2884 * a range. In this case, skip the '-', set a flag, then drop
2885 * down to handle what should be the end range value.
2886 * 2. After we've handled that value, the next time through, that
2887 * flag is set and we fix up the range.
2889 * Ranges entirely within Latin1 are expanded out entirely, in
2890 * order to avoid the significant overhead of making a swash.
2891 * Ranges that extend above Latin1 have to have a swash, so there
2892 * is no advantage to abbreviate them here, so they are stored here
2893 * as Min, ILLEGAL_UTF8_BYTE, Max. The illegal byte signifies a
2894 * hyphen without any possible ambiguity. On EBCDIC machines, if
2895 * the range is expressed as Unicode, the Latin1 portion is
2896 * expanded out even if the entire range extends above Latin1.
2897 * This is because each code point in it has to be processed here
2898 * individually to get its native translation */
2902 /* Here, we don't think we're in a range. If we've processed
2903 * at least one character, then see if this next one is a '-',
2904 * indicating the previous one was the start of a range. But
2905 * don't bother if we're too close to the end for the minus to
2907 if (*s != '-' || s >= send - 1 || s == start) {
2909 /* A regular character. Process like any other, but first
2910 * clear any flags */
2914 non_portable_endpoint = 0;
2917 /* Drops down to generic code to process current byte */
2920 if (didrange) { /* Something like y/A-C-Z// */
2921 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2926 s++; /* Skip past the minus */
2928 /* d now points to where the end-range character will be
2929 * placed. Save it so won't have to go finding it later,
2930 * and drop down to get that character. (Actually we
2931 * instead save the offset, to handle the case where a
2932 * realloc in the meantime could change the actual
2933 * pointer). We'll finish processing the range the next
2934 * time through the loop */
2935 offset_to_max = d - SvPVX_const(sv);
2937 } /* End of not a range */
2939 /* Here we have parsed a range. Now must handle it. At this
2941 * 'sv' is a SV* that contains the output string we are
2942 * constructing. The final two characters in that string
2943 * are the range start and range end, in order.
2944 * 'd' points to just beyond the range end in the 'sv' string,
2945 * where we would next place something
2946 * 'offset_to_max' is the offset in 'sv' at which the character
2947 * before 'd' begins.
2949 const char * max_ptr = SvPVX_const(sv) + offset_to_max;
2950 const char * min_ptr;
2952 IV range_max; /* last character in range */
2955 #ifndef EBCDIC /* Not meaningful except in EBCDIC, so initialize to false */
2956 const bool convert_unicode = FALSE;
2957 const IV real_range_max = 0;
2959 bool convert_unicode;
2960 IV real_range_max = 0;
2963 /* Get the range-ends code point values. */
2965 /* We know the utf8 is valid, because we just constructed
2966 * it ourselves in previous loop iterations */
2967 min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
2968 range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
2969 range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);
2972 min_ptr = max_ptr - 1;
2973 range_min = * (U8*) min_ptr;
2974 range_max = * (U8*) max_ptr;
2978 /* On EBCDIC platforms, we may have to deal with portable
2979 * ranges. These happen if at least one range endpoint is a
2980 * Unicode value (\N{...}), or if the range is a subset of
2981 * [A-Z] or [a-z], and both ends are literal characters,
2982 * like 'A', and not like \x{C1} */
2983 if ((convert_unicode
2984 = cBOOL(backslash_N) /* \N{} forces Unicode, hence
2986 || ( ! non_portable_endpoint
2987 && (( isLOWER_A(range_min) && isLOWER_A(range_max))
2988 || (isUPPER_A(range_min) && isUPPER_A(range_max))))
2991 /* Special handling is needed for these portable ranges.
2992 * They are defined to all be in Unicode terms, which
2993 * include all Unicode code points between the end points.
2994 * Convert to Unicode to get the Unicode range. Later we
2995 * will convert each code point in the range back to
2997 range_min = NATIVE_TO_UNI(range_min);
2998 range_max = NATIVE_TO_UNI(range_max);
3002 if (range_min > range_max) {
3003 if (convert_unicode) {
3004 /* Need to convert back to native for meaningful
3005 * messages for this platform */
3006 range_min = UNI_TO_NATIVE(range_min);
3007 range_max = UNI_TO_NATIVE(range_max);
3010 /* Use the characters themselves for the error message if
3011 * ASCII printables; otherwise some visible representation
3013 if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
3015 "Invalid range \"%c-%c\" in transliteration operator",
3016 (char)range_min, (char)range_max);
3018 else if (convert_unicode) {
3019 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3021 "Invalid range \"\\N{U+%04"UVXf"}-\\N{U+%04"UVXf"}\""
3022 " in transliteration operator",
3023 range_min, range_max);
3026 /* diag_listed_as: Invalid range "%s" in transliteration operator */
3028 "Invalid range \"\\x{%04"UVXf"}-\\x{%04"UVXf"}\""
3029 " in transliteration operator",
3030 range_min, range_max);
3036 /* We try to avoid creating a swash. If the upper end of
3037 * this range is below 256, this range won't force a swash;
3038 * otherwise it does force a swash, and as long as we have
3039 * to have one, we might as well not expand things out.
3040 * But if it's EBCDIC, we may have to look at each
3041 * character below 256 if we have to convert to/from
3045 && (range_min > 255 || ! convert_unicode)
3048 /* Move the high character one byte to the right; then
3049 * insert between it and the range begin, an illegal
3050 * byte which serves to indicate this is a range (using
3051 * a '-' could be ambiguous). */
3053 while (e-- > max_ptr) {
3056 *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
3060 /* Here, we're going to expand out the range. For EBCDIC
3061 * the range can extend above 255 (not so in ASCII), so
3062 * for EBCDIC, split it into the parts above and below
3065 if (range_max > 255) {
3066 real_range_max = range_max;
3072 /* Here we need to expand out the string to contain each
3073 * character in the range. Grow the output to handle this */
3075 save_offset = min_ptr - SvPVX_const(sv);
3077 /* The base growth is the number of code points in the range */
3078 grow = range_max - range_min + 1;
3081 /* But if the output is UTF-8, some of those characters may
3082 * need two bytes (since the maximum range value here is
3083 * 255, the max bytes per character is two). On ASCII
3084 * platforms, it's not much trouble to get an accurate
3085 * count of what's needed. But on EBCDIC, the ones that
3086 * need 2 bytes are scattered around, so just use a worst
3087 * case value instead of calculating for that platform. */
3091 /* Only those above 127 require 2 bytes. This may be
3092 * everything in the range, or not */
3093 if (range_min > 127) {
3096 else if (range_max > 127) {
3097 grow += range_max - 127;
3102 /* Subtract 3 for the bytes that were already accounted for
3103 * (min, max, and the hyphen) */
3104 SvGROW(sv, SvLEN(sv) + grow - 3);
3105 d = SvPVX(sv) + save_offset; /* refresh d after realloc */
3107 /* Here, we expand out the range. On ASCII platforms, the
3108 * compiler should optimize out the 'convert_unicode==TRUE'
3109 * portion of this */
3110 if (convert_unicode) {
3113 /* Recall that the min and max are now in Unicode terms, so
3114 * we have to convert each character to its native
3117 for (i = range_min; i <= range_max; i++) {
3118 append_utf8_from_native_byte(LATIN1_TO_NATIVE((U8) i),
3123 for (i = range_min; i <= range_max; i++) {
3124 *d++ = (char)LATIN1_TO_NATIVE((U8) i);
3131 /* Here, no conversions are necessary, which means that the
3132 * first character in the range is already in 'd' and
3133 * valid, so we can skip overwriting it */
3136 for (i = range_min + 1; i <= range_max; i++) {
3137 append_utf8_from_native_byte((U8) i, (U8 **) &d);
3142 for (i = range_min + 1; i <= range_max; i++) {
3148 /* (Compilers should optimize this out for non-EBCDIC). If the
3149 * original range extended above 255, add in that portion */
3150 if (real_range_max) {
3151 *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
3152 *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
3153 if (real_range_max > 0x101)
3154 *d++ = (char) ILLEGAL_UTF8_BYTE;
3155 if (real_range_max > 0x100)
3156 d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
3160 /* mark the range as done, and continue */
3164 non_portable_endpoint = 0;
3168 } /* End of is a range */
3169 } /* End of transliteration. Joins main code after these else's */
3170 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3173 while (s1 >= start && *s1-- == '\\')
3176 in_charclass = TRUE;
3179 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3182 while (s1 >= start && *s1-- == '\\')
3185 in_charclass = FALSE;
3188 /* skip for regexp comments /(?#comment)/, except for the last
3189 * char, which will be done separately.
3190 * Stop on (?{..}) and friends */
3192 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
3194 while (s+1 < send && *s != ')')
3197 else if (!PL_lex_casemods
3198 && ( s[2] == '{' /* This should match regcomp.c */
3199 || (s[2] == '?' && s[3] == '{')))
3205 /* likewise skip #-initiated comments in //x patterns */
3209 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
3211 while (s+1 < send && *s != '\n')
3215 /* no further processing of single-quoted regex */
3216 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3217 goto default_action;
3219 /* check for embedded arrays
3220 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3222 else if (*s == '@' && s[1]) {
3223 if (UTF ? isIDFIRST_utf8((U8*)s+1) : isWORDCHAR_A(s[1]))
3225 if (strchr(":'{$", s[1]))
3227 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3228 break; /* in regexp, neither @+ nor @- are interpolated */
3231 /* check for embedded scalars. only stop if we're sure it's a
3234 else if (*s == '$') {
3235 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3237 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3239 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3240 "Possible unintended interpolation of $\\ in regex");
3242 break; /* in regexp, $ might be tail anchor */
3246 /* End of else if chain - OP_TRANS rejoin rest */
3249 if (*s == '\\' && s+1 < send) {
3250 char* e; /* Can be used for ending '}', etc. */
3254 /* warn on \1 - \9 in substitution replacements, but note that \11
3255 * is an octal; and \19 is \1 followed by '9' */
3256 if (PL_lex_inwhat == OP_SUBST
3262 /* diag_listed_as: \%d better written as $%d */
3263 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3268 /* string-change backslash escapes */
3269 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3273 /* In a pattern, process \N, but skip any other backslash escapes.
3274 * This is because we don't want to translate an escape sequence
3275 * into a meta symbol and have the regex compiler use the meta
3276 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3277 * in spite of this, we do have to process \N here while the proper
3278 * charnames handler is in scope. See bugs #56444 and #62056.
3280 * There is a complication because \N in a pattern may also stand
3281 * for 'match a non-nl', and not mean a charname, in which case its
3282 * processing should be deferred to the regex compiler. To be a
3283 * charname it must be followed immediately by a '{', and not look
3284 * like \N followed by a curly quantifier, i.e., not something like
3285 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3287 else if (PL_lex_inpat
3290 || regcurly(s + 1)))
3293 goto default_action;
3299 if ((isALPHANUMERIC(*s)))
3300 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3301 "Unrecognized escape \\%c passed through",
3303 /* default action is to copy the quoted character */
3304 goto default_action;
3307 /* eg. \132 indicates the octal constant 0132 */
3308 case '0': case '1': case '2': case '3':
3309 case '4': case '5': case '6': case '7':
3311 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3313 uv = grok_oct(s, &len, &flags, NULL);
3315 if (len < 3 && s < send && isDIGIT(*s)
3316 && ckWARN(WARN_MISC))
3318 Perl_warner(aTHX_ packWARN(WARN_MISC),
3319 "%s", form_short_octal_warning(s, len));
3322 goto NUM_ESCAPE_INSERT;
3324 /* eg. \o{24} indicates the octal constant \024 */
3329 bool valid = grok_bslash_o(&s, &uv, &error,
3330 TRUE, /* Output warning */
3331 FALSE, /* Not strict */
3332 TRUE, /* Output warnings for
3339 goto NUM_ESCAPE_INSERT;
3342 /* eg. \x24 indicates the hex constant 0x24 */
3347 bool valid = grok_bslash_x(&s, &uv, &error,
3348 TRUE, /* Output warning */
3349 FALSE, /* Not strict */
3350 TRUE, /* Output warnings for
3360 /* Insert oct or hex escaped character. */
3362 /* Here uv is the ordinal of the next character being added */
3363 if (UVCHR_IS_INVARIANT(uv)) {
3367 if (!has_utf8 && uv > 255) {
3368 /* Might need to recode whatever we have accumulated so
3369 * far if it contains any chars variant in utf8 or
3372 SvCUR_set(sv, d - SvPVX_const(sv));
3375 /* See Note on sizing above. */
3376 sv_utf8_upgrade_flags_grow(
3378 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE
3379 /* Above-latin1 in string
3380 * implies no encoding */
3381 |SV_UTF8_NO_ENCODING,
3382 UVCHR_SKIP(uv) + (STRLEN)(send - s) + 1);
3383 d = SvPVX(sv) + SvCUR(sv);
3388 /* Usually, there will already be enough room in 'sv'
3389 * since such escapes are likely longer than any UTF-8
3390 * sequence they can end up as. This isn't the case on
3391 * EBCDIC where \x{40000000} contains 12 bytes, and the
3392 * UTF-8 for it contains 14. And, we have to allow for
3393 * a trailing NUL. It probably can't happen on ASCII
3394 * platforms, but be safe */
3395 const STRLEN needed = d - SvPVX(sv) + UVCHR_SKIP(uv)
3397 if (UNLIKELY(needed > SvLEN(sv))) {
3398 SvCUR_set(sv, d - SvPVX_const(sv));
3399 d = sv_grow(sv, needed) + SvCUR(sv);
3402 d = (char*)uvchr_to_utf8((U8*)d, uv);
3403 if (PL_lex_inwhat == OP_TRANS
3404 && PL_sublex_info.sub_op)
3406 PL_sublex_info.sub_op->op_private |=
3407 (PL_lex_repl ? OPpTRANS_FROM_UTF
3416 non_portable_endpoint++;
3421 /* In a non-pattern \N must be like \N{U+0041}, or it can be a
3422 * named character, like \N{LATIN SMALL LETTER A}, or a named
3423 * sequence, like \N{LATIN CAPITAL LETTER A WITH MACRON AND
3424 * GRAVE} (except y/// can't handle the latter, croaking). For
3425 * convenience all three forms are referred to as "named
3426 * characters" below.
3428 * For patterns, \N also can mean to match a non-newline. Code
3429 * before this 'switch' statement should already have handled
3430 * this situation, and hence this code only has to deal with
3431 * the named character cases.
3433 * For non-patterns, the named characters are converted to
3434 * their string equivalents. In patterns, named characters are
3435 * not converted to their ultimate forms for the same reasons
3436 * that other escapes aren't. Instead, they are converted to
3437 * the \N{U+...} form to get the value from the charnames that
3438 * is in effect right now, while preserving the fact that it
3439 * was a named character, so that the regex compiler knows
3442 * The structure of this section of code (besides checking for
3443 * errors and upgrading to utf8) is:
3444 * If the named character is of the form \N{U+...}, pass it
3445 * through if a pattern; otherwise convert the code point
3447 * Otherwise must be some \N{NAME}: convert to
3448 * \N{U+c1.c2...} if a pattern; otherwise convert to utf8
3450 * Transliteration is an exception. The conversion to utf8 is
3451 * only done if the code point requires it to be representable.
3453 * Here, 's' points to the 'N'; the test below is guaranteed to
3454 * succeed if we are being called on a pattern, as we already
3455 * know from a test above that the next character is a '{'. A
3456 * non-pattern \N must mean 'named character', which requires
3460 yyerror("Missing braces on \\N{}");
3465 /* If there is no matching '}', it is an error. */
3466 if (! (e = strchr(s, '}'))) {
3467 if (! PL_lex_inpat) {
3468 yyerror("Missing right brace on \\N{}");
3470 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N");
3475 /* Here it looks like a named character */
3477 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3478 s += 2; /* Skip to next char after the 'U+' */
3481 /* In patterns, we can have \N{U+xxxx.yyyy.zzzz...} */
3482 /* Check the syntax. */
3485 if (!isXDIGIT(*s)) {
3488 "Invalid hexadecimal number in \\N{U+...}"
3496 else if ((*s == '.' || *s == '_')
3502 /* Pass everything through unchanged.
3503 * +1 is for the '}' */
3504 Copy(orig_s, d, e - orig_s + 1, char);
3505 d += e - orig_s + 1;
3507 else { /* Not a pattern: convert the hex to string */
3508 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3509 | PERL_SCAN_SILENT_ILLDIGIT
3510 | PERL_SCAN_DISALLOW_PREFIX;
3512 uv = grok_hex(s, &len, &flags, NULL);
3513 if (len == 0 || (len != (STRLEN)(e - s)))
3516 /* For non-tr///, if the destination is not in utf8,
3517 * unconditionally recode it to be so. This is
3518 * because \N{} implies Unicode semantics, and scalars
3519 * have to be in utf8 to guarantee those semantics.
3520 * tr/// doesn't care about Unicode rules, so no need
3521 * there to upgrade to UTF-8 for small enough code
3523 if (! has_utf8 && ( uv > 0xFF
3524 || PL_lex_inwhat != OP_TRANS))
3526 SvCUR_set(sv, d - SvPVX_const(sv));
3529 /* See Note on sizing above. */
3530 sv_utf8_upgrade_flags_grow(
3532 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3533 UVCHR_SKIP(uv) + (STRLEN)(send - e) + 1);
3534 d = SvPVX(sv) + SvCUR(sv);
3538 /* Add the (Unicode) code point to the output. */
3539 if (OFFUNI_IS_INVARIANT(uv)) {
3540 *d++ = (char) LATIN1_TO_NATIVE(uv);
3543 d = (char*) uvoffuni_to_utf8_flags((U8*)d, uv, 0);
3547 else /* Here is \N{NAME} but not \N{U+...}. */
3548 if ((res = get_and_check_backslash_N_name(s, e)))
3551 const char *str = SvPV_const(res, len);
3554 if (! len) { /* The name resolved to an empty string */
3555 Copy("\\N{}", d, 4, char);
3559 /* In order to not lose information for the regex
3560 * compiler, pass the result in the specially made
3561 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3562 * the code points in hex of each character
3563 * returned by charnames */
3565 const char *str_end = str + len;
3566 const STRLEN off = d - SvPVX_const(sv);
3568 if (! SvUTF8(res)) {
3569 /* For the non-UTF-8 case, we can determine the
3570 * exact length needed without having to parse
3571 * through the string. Each character takes up
3572 * 2 hex digits plus either a trailing dot or
3574 const char initial_text[] = "\\N{U+";
3575 const STRLEN initial_len = sizeof(initial_text)
3577 d = off + SvGROW(sv, off
3580 /* +1 for trailing NUL */
3583 + (STRLEN)(send - e));
3584 Copy(initial_text, d, initial_len, char);
3586 while (str < str_end) {
3589 my_snprintf(hex_string,
3593 /* The regex compiler is
3594 * expecting Unicode, not
3596 NATIVE_TO_LATIN1(*str));
3597 PERL_MY_SNPRINTF_POST_GUARD(len,
3598 sizeof(hex_string));
3599 Copy(hex_string, d, 3, char);
3603 d--; /* Below, we will overwrite the final
3604 dot with a right brace */
3607 STRLEN char_length; /* cur char's byte length */
3609 /* and the number of bytes after this is
3610 * translated into hex digits */
3611 STRLEN output_length;
3613 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3614 * for max('U+', '.'); and 1 for NUL */
3615 char hex_string[2 * UTF8_MAXBYTES + 5];
3617 /* Get the first character of the result. */
3618 U32 uv = utf8n_to_uvchr((U8 *) str,
3622 /* Convert first code point to Unicode hex,
3623 * including the boiler plate before it. */
3625 my_snprintf(hex_string, sizeof(hex_string),
3627 (unsigned int) NATIVE_TO_UNI(uv));
3629 /* Make sure there is enough space to hold it */
3630 d = off + SvGROW(sv, off
3632 + (STRLEN)(send - e)
3633 + 2); /* '}' + NUL */
3635 Copy(hex_string, d, output_length, char);
3638 /* For each subsequent character, append dot and
3639 * its Unicode code point in hex */
3640 while ((str += char_length) < str_end) {
3641 const STRLEN off = d - SvPVX_const(sv);
3642 U32 uv = utf8n_to_uvchr((U8 *) str,
3647 my_snprintf(hex_string,
3650 (unsigned int) NATIVE_TO_UNI(uv));
3652 d = off + SvGROW(sv, off
3654 + (STRLEN)(send - e)
3655 + 2); /* '}' + NUL */
3656 Copy(hex_string, d, output_length, char);
3661 *d++ = '}'; /* Done. Add the trailing brace */
3664 else { /* Here, not in a pattern. Convert the name to a
3667 if (PL_lex_inwhat == OP_TRANS) {
3668 str = SvPV_const(res, len);
3669 if (len > ((SvUTF8(res))
3673 yyerror(Perl_form(aTHX_
3674 "%.*s must not be a named sequence"
3675 " in transliteration operator",
3676 /* +1 to include the "}" */
3677 (int) (e + 1 - start), start));
3678 goto end_backslash_N;
3681 else if (! SvUTF8(res)) {
3682 /* Make sure \N{} return is UTF-8. This is because
3683 * \N{} implies Unicode semantics, and scalars have to
3684 * be in utf8 to guarantee those semantics; but not
3685 * needed in tr/// */
3686 sv_utf8_upgrade_flags(res, SV_UTF8_NO_ENCODING);
3687 str = SvPV_const(res, len);
3690 /* Upgrade destination to be utf8 if this new
3692 if (! has_utf8 && SvUTF8(res)) {
3693 SvCUR_set(sv, d - SvPVX_const(sv));
3696 /* See Note on sizing above. */
3697 sv_utf8_upgrade_flags_grow(sv,
3698 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3699 len + (STRLEN)(send - s) + 1);
3700 d = SvPVX(sv) + SvCUR(sv);
3702 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3704 /* See Note on sizing above. (NOTE: SvCUR() is not
3705 * set correctly here). */
3706 const STRLEN off = d - SvPVX_const(sv);
3707 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3709 Copy(str, d, len, char);
3715 } /* End \N{NAME} */
3719 backslash_N++; /* \N{} is defined to be Unicode */
3721 s = e + 1; /* Point to just after the '}' */
3724 /* \c is a control character */
3728 *d++ = grok_bslash_c(*s++, 1);
3731 yyerror("Missing control char name in \\c");
3734 non_portable_endpoint++;
3738 /* printf-style backslashes, formfeeds, newlines, etc */
3764 } /* end if (backslash) */
3767 /* If we started with encoded form, or already know we want it,
3768 then encode the next character */
3769 if (! NATIVE_BYTE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3772 /* One might think that it is wasted effort in the case of the
3773 * source being utf8 (this_utf8 == TRUE) to take the next character
3774 * in the source, convert it to an unsigned value, and then convert
3775 * it back again. But the source has not been validated here. The
3776 * routine that does the conversion checks for errors like
3779 const UV nextuv = (this_utf8)
3780 ? utf8n_to_uvchr((U8*)s, send - s, &len, 0)
3782 const STRLEN need = UVCHR_SKIP(nextuv);
3784 SvCUR_set(sv, d - SvPVX_const(sv));
3787 /* See Note on sizing above. */
3788 sv_utf8_upgrade_flags_grow(sv,
3789 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3790 need + (STRLEN)(send - s) + 1);
3791 d = SvPVX(sv) + SvCUR(sv);
3793 } else if (need > len) {
3794 /* encoded value larger than old, may need extra space (NOTE:
3795 * SvCUR() is not set correctly here). See Note on sizing
3797 const STRLEN off = d - SvPVX_const(sv);
3798 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3802 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3807 } /* while loop to process each character */
3809 /* terminate the string and set up the sv */
3811 SvCUR_set(sv, d - SvPVX_const(sv));
3812 if (SvCUR(sv) >= SvLEN(sv))
3813 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3814 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3817 if (IN_ENCODING && !has_utf8) {
3818 sv_recode_to_utf8(sv, _get_encoding());
3824 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3825 PL_sublex_info.sub_op->op_private |=
3826 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3830 /* shrink the sv if we allocated more than we used */
3831 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3832 SvPV_shrink_to_cur(sv);
3835 /* return the substring (via pl_yylval) only if we parsed anything */
3838 for (; s2 < s; s2++) {
3840 COPLINE_INC_WITH_HERELINES;
3842 SvREFCNT_inc_simple_void_NN(sv);
3843 if ( (PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ))
3844 && ! PL_parser->lex_re_reparsing)
3846 const char *const key = PL_lex_inpat ? "qr" : "q";
3847 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3851 if (PL_lex_inwhat == OP_TRANS) {
3854 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3857 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3865 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3868 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3870 LEAVE_with_name("scan_const");
3875 * Returns TRUE if there's more to the expression (e.g., a subscript),
3878 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3880 * ->[ and ->{ return TRUE
3881 * ->$* ->$#* ->@* ->@[ ->@{ return TRUE if postderef_qq is enabled
3882 * { and [ outside a pattern are always subscripts, so return TRUE
3883 * if we're outside a pattern and it's not { or [, then return FALSE
3884 * if we're in a pattern and the first char is a {
3885 * {4,5} (any digits around the comma) returns FALSE
3886 * if we're in a pattern and the first char is a [
3888 * [SOMETHING] has a funky algorithm to decide whether it's a
3889 * character class or not. It has to deal with things like
3890 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3891 * anything else returns TRUE
3894 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3897 S_intuit_more(pTHX_ char *s)
3899 PERL_ARGS_ASSERT_INTUIT_MORE;
3901 if (PL_lex_brackets)
3903 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3905 if (*s == '-' && s[1] == '>'
3906 && FEATURE_POSTDEREF_QQ_IS_ENABLED
3907 && ( (s[2] == '$' && (s[3] == '*' || (s[3] == '#' && s[4] == '*')))
3908 ||(s[2] == '@' && strchr("*[{",s[3])) ))
3910 if (*s != '{' && *s != '[')
3915 /* In a pattern, so maybe we have {n,m}. */
3923 /* On the other hand, maybe we have a character class */
3926 if (*s == ']' || *s == '^')
3929 /* this is terrifying, and it works */
3932 const char * const send = strchr(s,']');
3933 unsigned char un_char, last_un_char;
3934 char tmpbuf[sizeof PL_tokenbuf * 4];
3936 if (!send) /* has to be an expression */
3938 weight = 2; /* let's weigh the evidence */
3942 else if (isDIGIT(*s)) {
3944 if (isDIGIT(s[1]) && s[2] == ']')
3950 Zero(seen,256,char);
3952 for (; s < send; s++) {
3953 last_un_char = un_char;
3954 un_char = (unsigned char)*s;
3959 weight -= seen[un_char] * 10;
3960 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3962 char *tmp = PL_bufend;
3963 PL_bufend = (char*)send;
3964 scan_ident(s, tmpbuf, sizeof tmpbuf, FALSE);
3966 len = (int)strlen(tmpbuf);
3967 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3968 UTF ? SVf_UTF8 : 0, SVt_PV))
3975 && strchr("[#!%*<>()-=",s[1]))
3977 if (/*{*/ strchr("])} =",s[2]))
3986 if (strchr("wds]",s[1]))
3988 else if (seen[(U8)'\''] || seen[(U8)'"'])
3990 else if (strchr("rnftbxcav",s[1]))
3992 else if (isDIGIT(s[1])) {
3994 while (s[1] && isDIGIT(s[1]))
4004 if (strchr("aA01! ",last_un_char))
4006 if (strchr("zZ79~",s[1]))
4008 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
4009 weight -= 5; /* cope with negative subscript */
4012 if (!isWORDCHAR(last_un_char)
4013 && !(last_un_char == '$' || last_un_char == '@'
4014 || last_un_char == '&')
4015 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
4019 if (keyword(d, s - d, 0))
4022 if (un_char == last_un_char + 1)
4024 weight -= seen[un_char];
4029 if (weight >= 0) /* probably a character class */
4039 * Does all the checking to disambiguate
4041 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
4042 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
4044 * First argument is the stuff after the first token, e.g. "bar".
4046 * Not a method if foo is a filehandle.
4047 * Not a method if foo is a subroutine prototyped to take a filehandle.
4048 * Not a method if it's really "Foo $bar"
4049 * Method if it's "foo $bar"
4050 * Not a method if it's really "print foo $bar"
4051 * Method if it's really "foo package::" (interpreted as package->foo)
4052 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
4053 * Not a method if bar is a filehandle or package, but is quoted with
4058 S_intuit_method(pTHX_ char *start, SV *ioname, CV *cv)
4060 char *s = start + (*start == '$');
4061 char tmpbuf[sizeof PL_tokenbuf];
4064 /* Mustn't actually add anything to a symbol table.
4065 But also don't want to "initialise" any placeholder
4066 constants that might already be there into full
4067 blown PVGVs with attached PVCV. */
4069 ioname ? gv_fetchsv(ioname, GV_NOADD_NOINIT, SVt_PVCV) : NULL;
4071 PERL_ARGS_ASSERT_INTUIT_METHOD;
4073 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
4075 if (cv && SvPOK(cv)) {
4076 const char *proto = CvPROTO(cv);
4078 while (*proto && (isSPACE(*proto) || *proto == ';'))
4085 if (*start == '$') {
4086 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY
4087 || isUPPER(*PL_tokenbuf))
4092 return *s == '(' ? FUNCMETH : METHOD;
4095 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4096 /* start is the beginning of the possible filehandle/object,
4097 * and s is the end of it
4098 * tmpbuf is a copy of it (but with single quotes as double colons)
4101 if (!keyword(tmpbuf, len, 0)) {
4102 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
4107 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
4108 if (indirgv && GvCVu(indirgv))
4110 /* filehandle or package name makes it a method */
4111 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
4113 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
4114 return 0; /* no assumptions -- "=>" quotes bareword */
4116 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
4117 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
4118 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
4122 return *s == '(' ? FUNCMETH : METHOD;
4128 /* Encoded script support. filter_add() effectively inserts a
4129 * 'pre-processing' function into the current source input stream.
4130 * Note that the filter function only applies to the current source file
4131 * (e.g., it will not affect files 'require'd or 'use'd by this one).
4133 * The datasv parameter (which may be NULL) can be used to pass
4134 * private data to this instance of the filter. The filter function
4135 * can recover the SV using the FILTER_DATA macro and use it to
4136 * store private buffers and state information.
4138 * The supplied datasv parameter is upgraded to a PVIO type
4139 * and the IoDIRP/IoANY field is used to store the function pointer,
4140 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
4141 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
4142 * private use must be set using malloc'd pointers.
4146 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
4154 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
4155 Perl_croak(aTHX_ "Source filters apply only to byte streams");
4157 if (!PL_rsfp_filters)
4158 PL_rsfp_filters = newAV();
4161 SvUPGRADE(datasv, SVt_PVIO);
4162 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
4163 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
4164 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
4165 FPTR2DPTR(void *, IoANY(datasv)),
4166 SvPV_nolen(datasv)));
4167 av_unshift(PL_rsfp_filters, 1);
4168 av_store(PL_rsfp_filters, 0, datasv) ;
4170 !PL_parser->filtered
4171 && PL_parser->lex_flags & LEX_EVALBYTES
4172 && PL_bufptr < PL_bufend
4174 const char *s = PL_bufptr;
4175 while (s < PL_bufend) {
4177 SV *linestr = PL_parser->linestr;
4178 char *buf = SvPVX(linestr);
4179 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
4180 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
4181 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
4182 STRLEN const linestart_pos = PL_parser->linestart - buf;
4183 STRLEN const last_uni_pos =
4184 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
4185 STRLEN const last_lop_pos =
4186 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
4187 av_push(PL_rsfp_filters, linestr);
4188 PL_parser->linestr =
4189 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
4190 buf = SvPVX(PL_parser->linestr);
4191 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4192 PL_parser->bufptr = buf + bufptr_pos;
4193 PL_parser->oldbufptr = buf + oldbufptr_pos;
4194 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4195 PL_parser->linestart = buf + linestart_pos;
4196 if (PL_parser->last_uni)
4197 PL_parser->last_uni = buf + last_uni_pos;
4198 if (PL_parser->last_lop)
4199 PL_parser->last_lop = buf + last_lop_pos;
4200 SvLEN(linestr) = SvCUR(linestr);
4201 SvCUR(linestr) = s-SvPVX(linestr);
4202 PL_parser->filtered = 1;
4212 /* Delete most recently added instance of this filter function. */
4214 Perl_filter_del(pTHX_ filter_t funcp)
4218 PERL_ARGS_ASSERT_FILTER_DEL;
4221 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4222 FPTR2DPTR(void*, funcp)));
4224 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4226 /* if filter is on top of stack (usual case) just pop it off */
4227 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4228 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4229 sv_free(av_pop(PL_rsfp_filters));
4233 /* we need to search for the correct entry and clear it */
4234 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4238 /* Invoke the idxth filter function for the current rsfp. */
4239 /* maxlen 0 = read one text line */
4241 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4245 /* This API is bad. It should have been using unsigned int for maxlen.
4246 Not sure if we want to change the API, but if not we should sanity
4247 check the value here. */
4248 unsigned int correct_length = maxlen < 0 ? PERL_INT_MAX : maxlen;
4250 PERL_ARGS_ASSERT_FILTER_READ;
4252 if (!PL_parser || !PL_rsfp_filters)
4254 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4255 /* Provide a default input filter to make life easy. */
4256 /* Note that we append to the line. This is handy. */
4257 DEBUG_P(PerlIO_printf(Perl_debug_log,
4258 "filter_read %d: from rsfp\n", idx));
4259 if (correct_length) {
4262 const int old_len = SvCUR(buf_sv);
4264 /* ensure buf_sv is large enough */
4265 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4266 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4267 correct_length)) <= 0) {
4268 if (PerlIO_error(PL_rsfp))
4269 return -1; /* error */
4271 return 0 ; /* end of file */
4273 SvCUR_set(buf_sv, old_len + len) ;
4274 SvPVX(buf_sv)[old_len + len] = '\0';
4277 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4278 if (PerlIO_error(PL_rsfp))
4279 return -1; /* error */
4281 return 0 ; /* end of file */
4284 return SvCUR(buf_sv);
4286 /* Skip this filter slot if filter has been deleted */
4287 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4288 DEBUG_P(PerlIO_printf(Perl_debug_log,
4289 "filter_read %d: skipped (filter deleted)\n",
4291 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4293 if (SvTYPE(datasv) != SVt_PVIO) {
4294 if (correct_length) {
4296 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4297 if (!remainder) return 0; /* eof */
4298 if (correct_length > remainder) correct_length = remainder;
4299 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4300 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4303 const char *s = SvEND(datasv);
4304 const char *send = SvPVX(datasv) + SvLEN(datasv);
4312 if (s == send) return 0; /* eof */
4313 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4314 SvCUR_set(datasv, s-SvPVX(datasv));
4316 return SvCUR(buf_sv);
4318 /* Get function pointer hidden within datasv */
4319 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4320 DEBUG_P(PerlIO_printf(Perl_debug_log,
4321 "filter_read %d: via function %p (%s)\n",
4322 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4323 /* Call function. The function is expected to */
4324 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4325 /* Return: <0:error, =0:eof, >0:not eof */
4326 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4330 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4332 PERL_ARGS_ASSERT_FILTER_GETS;
4334 #ifdef PERL_CR_FILTER
4335 if (!PL_rsfp_filters) {
4336 filter_add(S_cr_textfilter,NULL);
4339 if (PL_rsfp_filters) {
4341 SvCUR_set(sv, 0); /* start with empty line */
4342 if (FILTER_READ(0, sv, 0) > 0)
4343 return ( SvPVX(sv) ) ;
4348 return (sv_gets(sv, PL_rsfp, append));
4352 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4356 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4358 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4362 && (pkgname[len - 2] == ':' && pkgname[len - 1] == ':')
4363 && (gv = gv_fetchpvn_flags(pkgname,
4365 ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4367 return GvHV(gv); /* Foo:: */
4370 /* use constant CLASS => 'MyClass' */
4371 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4372 if (gv && GvCV(gv)) {
4373 SV * const sv = cv_const_sv(GvCV(gv));
4375 return gv_stashsv(sv, 0);
4378 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4383 S_tokenize_use(pTHX_ int is_use, char *s) {
4384 PERL_ARGS_ASSERT_TOKENIZE_USE;
4386 if (PL_expect != XSTATE)
4387 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4388 is_use ? "use" : "no"));
4391 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4392 s = force_version(s, TRUE);
4393 if (*s == ';' || *s == '}'
4394 || (s = skipspace(s), (*s == ';' || *s == '}'))) {
4395 NEXTVAL_NEXTTOKE.opval = NULL;
4398 else if (*s == 'v') {
4399 s = force_word(s,WORD,FALSE,TRUE);
4400 s = force_version(s, FALSE);
4404 s = force_word(s,WORD,FALSE,TRUE);
4405 s = force_version(s, FALSE);
4407 pl_yylval.ival = is_use;
4411 static const char* const exp_name[] =
4412 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4413 "ATTRTERM", "TERMBLOCK", "XBLOCKTERM", "POSTDEREF",
4418 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4420 S_word_takes_any_delimeter(char *p, STRLEN len)
4422 return (len == 1 && strchr("msyq", p[0]))
4424 && ((p[0] == 't' && p[1] == 'r')
4425 || (p[0] == 'q' && strchr("qwxr", p[1]))));
4429 S_check_scalar_slice(pTHX_ char *s)
4432 while (*s == ' ' || *s == '\t') s++;
4433 if (*s == 'q' && s[1] == 'w'
4434 && !isWORDCHAR_lazy_if(s+2,UTF))
4436 while (*s && (isWORDCHAR_lazy_if(s,UTF) || strchr(" \t$#+-'\"", *s)))
4437 s += UTF ? UTF8SKIP(s) : 1;
4438 if (*s == '}' || *s == ']')
4439 pl_yylval.ival = OPpSLICEWARNING;
4445 Works out what to call the token just pulled out of the input
4446 stream. The yacc parser takes care of taking the ops we return and
4447 stitching them into a tree.
4450 The type of the next token
4453 Switch based on the current state:
4454 - if we already built the token before, use it
4455 - if we have a case modifier in a string, deal with that
4456 - handle other cases of interpolation inside a string
4457 - scan the next line if we are inside a format
4458 In the normal state switch on the next character:
4460 if alphabetic, go to key lookup
4461 unrecoginized character - croak
4462 - 0/4/26: handle end-of-line or EOF
4463 - cases for whitespace
4464 - \n and #: handle comments and line numbers
4465 - various operators, brackets and sigils
4468 - 'v': vstrings (or go to key lookup)
4469 - 'x' repetition operator (or go to key lookup)
4470 - other ASCII alphanumerics (key lookup begins here):
4473 scan built-in keyword (but do nothing with it yet)
4474 check for statement label
4475 check for lexical subs
4476 goto just_a_word if there is one
4477 see whether built-in keyword is overridden
4478 switch on keyword number:
4479 - default: just_a_word:
4480 not a built-in keyword; handle bareword lookup
4481 disambiguate between method and sub call
4482 fall back to bareword
4483 - cases for built-in keywords
4491 char *s = PL_bufptr;
4495 const bool saw_infix_sigil = cBOOL(PL_parser->saw_infix_sigil);
4499 /* orig_keyword, gvp, and gv are initialized here because
4500 * jump to the label just_a_word_zero can bypass their
4501 * initialization later. */
4502 I32 orig_keyword = 0;
4507 SV* tmp = newSVpvs("");
4508 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4509 (IV)CopLINE(PL_curcop),
4510 lex_state_names[PL_lex_state],
4511 exp_name[PL_expect],
4512 pv_display(tmp, s, strlen(s), 0, 60));
4516 /* when we've already built the next token, just pull it out of the queue */
4519 pl_yylval = PL_nextval[PL_nexttoke];
4521 PL_lex_state = PL_lex_defer;
4522 PL_lex_defer = LEX_NORMAL;
4526 next_type = PL_nexttype[PL_nexttoke];
4527 if (next_type & (7<<24)) {
4528 if (next_type & (1<<24)) {
4529 if (PL_lex_brackets > 100)
4530 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4531 PL_lex_brackstack[PL_lex_brackets++] =
4532 (char) ((next_type >> 16) & 0xff);
4534 if (next_type & (2<<24))
4535 PL_lex_allbrackets++;
4536 if (next_type & (4<<24))
4537 PL_lex_allbrackets--;
4538 next_type &= 0xffff;
4540 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4544 switch (PL_lex_state) {
4546 case LEX_INTERPNORMAL:
4549 /* interpolated case modifiers like \L \U, including \Q and \E.
4550 when we get here, PL_bufptr is at the \
4552 case LEX_INTERPCASEMOD:
4554 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4556 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4557 PL_bufptr, PL_bufend, *PL_bufptr);
4559 /* handle \E or end of string */
4560 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4562 if (PL_lex_casemods) {
4563 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4564 PL_lex_casestack[PL_lex_casemods] = '\0';
4566 if (PL_bufptr != PL_bufend
4567 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4568 || oldmod == 'F')) {
4570 PL_lex_state = LEX_INTERPCONCAT;
4572 PL_lex_allbrackets--;
4575 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4576 /* Got an unpaired \E */
4577 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4578 "Useless use of \\E");
4580 if (PL_bufptr != PL_bufend)
4582 PL_lex_state = LEX_INTERPCONCAT;
4586 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4587 "### Saw case modifier\n"); });
4589 if (s[1] == '\\' && s[2] == 'E') {
4591 PL_lex_state = LEX_INTERPCONCAT;
4596 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4597 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4598 if ((*s == 'L' || *s == 'U' || *s == 'F')
4599 && (strchr(PL_lex_casestack, 'L')
4600 || strchr(PL_lex_casestack, 'U')
4601 || strchr(PL_lex_casestack, 'F')))
4603 PL_lex_casestack[--PL_lex_casemods] = '\0';
4604 PL_lex_allbrackets--;
4607 if (PL_lex_casemods > 10)
4608 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4609 PL_lex_casestack[PL_lex_casemods++] = *s;
4610 PL_lex_casestack[PL_lex_casemods] = '\0';
4611 PL_lex_state = LEX_INTERPCONCAT;
4612 NEXTVAL_NEXTTOKE.ival = 0;
4613 force_next((2<<24)|'(');
4615 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4617 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4619 NEXTVAL_NEXTTOKE.ival = OP_LC;
4621 NEXTVAL_NEXTTOKE.ival = OP_UC;
4623 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4625 NEXTVAL_NEXTTOKE.ival = OP_FC;
4627 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4631 if (PL_lex_starts) {
4634 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4635 if (PL_lex_casemods == 1 && PL_lex_inpat)
4638 AopNOASSIGN(OP_CONCAT);
4644 case LEX_INTERPPUSH:
4645 return REPORT(sublex_push());
4647 case LEX_INTERPSTART:
4648 if (PL_bufptr == PL_bufend)
4649 return REPORT(sublex_done());
4650 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4651 "### Interpolated variable\n"); });
4653 /* for /@a/, we leave the joining for the regex engine to do
4654 * (unless we're within \Q etc) */
4655 PL_lex_dojoin = (*PL_bufptr == '@'
4656 && (!PL_lex_inpat || PL_lex_casemods));
4657 PL_lex_state = LEX_INTERPNORMAL;
4658 if (PL_lex_dojoin) {
4659 NEXTVAL_NEXTTOKE.ival = 0;
4661 force_ident("\"", '$');
4662 NEXTVAL_NEXTTOKE.ival = 0;
4664 NEXTVAL_NEXTTOKE.ival = 0;
4665 force_next((2<<24)|'(');
4666 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4669 /* Convert (?{...}) and friends to 'do {...}' */
4670 if (PL_lex_inpat && *PL_bufptr == '(') {
4671 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4673 if (*PL_bufptr != '{')
4675 PL_expect = XTERMBLOCK;
4679 if (PL_lex_starts++) {
4681 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4682 if (!PL_lex_casemods && PL_lex_inpat)
4685 AopNOASSIGN(OP_CONCAT);
4689 case LEX_INTERPENDMAYBE:
4690 if (intuit_more(PL_bufptr)) {
4691 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4697 /* Treat state as LEX_NORMAL if we have no inner lexing scope.
4698 XXX This hack can be removed if we stop setting PL_lex_state to
4699 LEX_KNOWNEXT, as can the hack under LEX_INTREPCONCAT below. */
4700 if (UNLIKELY(!PL_lex_inwhat)) {
4701 PL_lex_state = LEX_NORMAL;
4705 if (PL_lex_dojoin) {
4706 const U8 dojoin_was = PL_lex_dojoin;
4707 PL_lex_dojoin = FALSE;
4708 PL_lex_state = LEX_INTERPCONCAT;
4709 PL_lex_allbrackets--;
4710 return REPORT(dojoin_was == 1 ? ')' : POSTJOIN);
4712 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4713 && SvEVALED(PL_lex_repl))
4715 if (PL_bufptr != PL_bufend)
4716 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4719 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4720 re_eval_str. If the here-doc body’s length equals the previous
4721 value of re_eval_start, re_eval_start will now be null. So
4722 check re_eval_str as well. */
4723 if (PL_parser->lex_shared->re_eval_start
4724 || PL_parser->lex_shared->re_eval_str) {
4726 if (*PL_bufptr != ')')
4727 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4729 /* having compiled a (?{..}) expression, return the original
4730 * text too, as a const */
4731 if (PL_parser->lex_shared->re_eval_str) {
4732 sv = PL_parser->lex_shared->re_eval_str;
4733 PL_parser->lex_shared->re_eval_str = NULL;
4735 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4736 SvPV_shrink_to_cur(sv);
4738 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4739 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4740 NEXTVAL_NEXTTOKE.opval =
4741 (OP*)newSVOP(OP_CONST, 0,
4744 PL_parser->lex_shared->re_eval_start = NULL;
4750 case LEX_INTERPCONCAT:
4752 if (PL_lex_brackets)
4753 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4754 (long) PL_lex_brackets);
4756 /* Treat state as LEX_NORMAL when not in an inner lexing scope.
4757 XXX This hack can be removed if we stop setting PL_lex_state to
4759 if (UNLIKELY(!PL_lex_inwhat)) {
4760 PL_lex_state = LEX_NORMAL;
4764 if (PL_bufptr == PL_bufend)
4765 return REPORT(sublex_done());
4767 /* m'foo' still needs to be parsed for possible (?{...}) */
4768 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4769 SV *sv = newSVsv(PL_linestr);
4771 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4775 s = scan_const(PL_bufptr);
4777 PL_lex_state = LEX_INTERPCASEMOD;
4779 PL_lex_state = LEX_INTERPSTART;
4782 if (s != PL_bufptr) {
4783 NEXTVAL_NEXTTOKE = pl_yylval;
4786 if (PL_lex_starts++) {
4787 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4788 if (!PL_lex_casemods && PL_lex_inpat)
4791 AopNOASSIGN(OP_CONCAT);
4801 s = scan_formline(PL_bufptr);
4802 if (!PL_lex_formbrack)
4811 /* We really do *not* want PL_linestr ever becoming a COW. */
4812 assert (!SvIsCOW(PL_linestr));
4814 PL_oldoldbufptr = PL_oldbufptr;
4816 PL_parser->saw_infix_sigil = 0;
4821 if (UTF ? isIDFIRST_utf8((U8*)s) : isALNUMC(*s))
4824 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4825 const char *c = UTF ? sv_uni_display(dsv, newSVpvn_flags(s,
4827 SVs_TEMP | SVf_UTF8),
4828 10, UNI_DISPLAY_ISPRINT)
4829 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4830 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4831 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4832 d = UTF ? (char *) utf8_hop((U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4836 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"UTF8f"<-- HERE near column %d", c,
4837 UTF8fARG(UTF, (s - d), d),
4842 goto fake_eof; /* emulate EOF on ^D or ^Z */
4844 if ((!PL_rsfp || PL_lex_inwhat)
4845 && (!PL_parser->filtered || s+1 < PL_bufend)) {
4849 && PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF)
4851 yyerror((const char *)
4853 ? "Format not terminated"
4854 : "Missing right curly or square bracket"));
4856 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4857 "### Tokener got EOF\n");
4861 if (s++ < PL_bufend)
4862 goto retry; /* ignore stray nulls */
4865 if (!PL_in_eval && !PL_preambled) {
4866 PL_preambled = TRUE;
4868 /* Generate a string of Perl code to load the debugger.
4869 * If PERL5DB is set, it will return the contents of that,
4870 * otherwise a compile-time require of perl5db.pl. */
4872 const char * const pdb = PerlEnv_getenv("PERL5DB");
4875 sv_setpv(PL_linestr, pdb);
4876 sv_catpvs(PL_linestr,";");
4878 SETERRNO(0,SS_NORMAL);
4879 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4881 PL_parser->preambling = CopLINE(PL_curcop);
4883 sv_setpvs(PL_linestr,"");
4884 if (PL_preambleav) {
4885 SV **svp = AvARRAY(PL_preambleav);
4886 SV **const end = svp + AvFILLp(PL_preambleav);
4888 sv_catsv(PL_linestr, *svp);
4890 sv_catpvs(PL_linestr, ";");
4892 sv_free(MUTABLE_SV(PL_preambleav));
4893 PL_preambleav = NULL;
4896 sv_catpvs(PL_linestr,
4897 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4898 if (PL_minus_n || PL_minus_p) {
4899 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4901 sv_catpvs(PL_linestr,"chomp;");
4904 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4905 || *PL_splitstr == '"')
4906 && strchr(PL_splitstr + 1, *PL_splitstr))
4907 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4909 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4910 bytes can be used as quoting characters. :-) */
4911 const char *splits = PL_splitstr;
4912 sv_catpvs(PL_linestr, "our @F=split(q\0");
4915 if (*splits == '\\')
4916 sv_catpvn(PL_linestr, splits, 1);
4917 sv_catpvn(PL_linestr, splits, 1);
4918 } while (*splits++);
4919 /* This loop will embed the trailing NUL of
4920 PL_linestr as the last thing it does before
4922 sv_catpvs(PL_linestr, ");");
4926 sv_catpvs(PL_linestr,"our @F=split(' ');");
4929 sv_catpvs(PL_linestr, "\n");
4930 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4931 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4932 PL_last_lop = PL_last_uni = NULL;
4933 if (PERLDB_LINE_OR_SAVESRC && PL_curstash != PL_debstash)
4934 update_debugger_info(PL_linestr, NULL, 0);
4939 bof = PL_rsfp ? TRUE : FALSE;
4942 fake_eof = LEX_FAKE_EOF;
4944 PL_bufptr = PL_bufend;
4945 COPLINE_INC_WITH_HERELINES;
4946 if (!lex_next_chunk(fake_eof)) {
4947 CopLINE_dec(PL_curcop);
4949 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4951 CopLINE_dec(PL_curcop);
4953 /* If it looks like the start of a BOM or raw UTF-16,
4954 * check if it in fact is. */
4957 || *(U8*)s == BOM_UTF8_FIRST_BYTE
4961 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4962 bof = (offset == (Off_t)SvCUR(PL_linestr));
4963 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4964 /* offset may include swallowed CR */
4966 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4969 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4970 s = swallow_bom((U8*)s);
4973 if (PL_parser->in_pod) {
4974 /* Incest with pod. */
4975 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4976 sv_setpvs(PL_linestr, "");
4977 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4978 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4979 PL_last_lop = PL_last_uni = NULL;
4980 PL_parser->in_pod = 0;
4983 if (PL_rsfp || PL_parser->filtered)
4985 } while (PL_parser->in_pod);
4986 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4987 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4988 PL_last_lop = PL_last_uni = NULL;
4989 if (CopLINE(PL_curcop) == 1) {
4990 while (s < PL_bufend && isSPACE(*s))
4992 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4996 if (*s == '#' && *(s+1) == '!')
4998 #ifdef ALTERNATE_SHEBANG
5000 static char const as[] = ALTERNATE_SHEBANG;
5001 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5002 d = s + (sizeof(as) - 1);
5004 #endif /* ALTERNATE_SHEBANG */
5013 while (*d && !isSPACE(*d))
5017 #ifdef ARG_ZERO_IS_SCRIPT
5018 if (ipathend > ipath) {
5020 * HP-UX (at least) sets argv[0] to the script name,
5021 * which makes $^X incorrect. And Digital UNIX and Linux,
5022 * at least, set argv[0] to the basename of the Perl
5023 * interpreter. So, having found "#!", we'll set it right.
5025 SV* copfilesv = CopFILESV(PL_curcop);
5028 GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5030 assert(SvPOK(x) || SvGMAGICAL(x));
5031 if (sv_eq(x, copfilesv)) {
5032 sv_setpvn(x, ipath, ipathend - ipath);
5038 const char *bstart = SvPV_const(copfilesv, blen);
5039 const char * const lstart = SvPV_const(x, llen);
5041 bstart += blen - llen;
5042 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5043 sv_setpvn(x, ipath, ipathend - ipath);
5050 /* Anything to do if no copfilesv? */
5052 TAINT_NOT; /* $^X is always tainted, but that's OK */
5054 #endif /* ARG_ZERO_IS_SCRIPT */
5059 d = instr(s,"perl -");
5061 d = instr(s,"perl");
5062 if (d && d[4] == '6')
5065 /* avoid getting into infinite loops when shebang
5066 * line contains "Perl" rather than "perl" */
5068 for (d = ipathend-4; d >= ipath; --d) {
5069 if (isALPHA_FOLD_EQ(*d, 'p')
5070 && !ibcmp(d, "perl", 4))
5080 #ifdef ALTERNATE_SHEBANG
5082 * If the ALTERNATE_SHEBANG on this system starts with a
5083 * character that can be part of a Perl expression, then if
5084 * we see it but not "perl", we're probably looking at the
5085 * start of Perl code, not a request to hand off to some
5086 * other interpreter. Similarly, if "perl" is there, but
5087 * not in the first 'word' of the line, we assume the line
5088 * contains the start of the Perl program.
5090 if (d && *s != '#') {
5091 const char *c = ipath;
5092 while (*c && !strchr("; \t\r\n\f\v#", *c))
5095 d = NULL; /* "perl" not in first word; ignore */
5097 *s = '#'; /* Don't try to parse shebang line */
5099 #endif /* ALTERNATE_SHEBANG */
5104 && !instr(s,"indir")
5105 && instr(PL_origargv[0],"perl"))
5112 while (s < PL_bufend && isSPACE(*s))
5114 if (s < PL_bufend) {
5115 Newx(newargv,PL_origargc+3,char*);
5117 while (s < PL_bufend && !isSPACE(*s))
5120 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5123 newargv = PL_origargv;
5126 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5128 Perl_croak(aTHX_ "Can't exec %s", ipath);
5131 while (*d && !isSPACE(*d))
5133 while (SPACE_OR_TAB(*d))
5137 const bool switches_done = PL_doswitches;
5138 const U32 oldpdb = PL_perldb;
5139 const bool oldn = PL_minus_n;
5140 const bool oldp = PL_minus_p;
5144 bool baduni = FALSE;
5146 const char *d2 = d1 + 1;
5147 if (parse_unicode_opts((const char **)&d2)
5151 if (baduni || isALPHA_FOLD_EQ(*d1, 'M')) {
5152 const char * const m = d1;
5153 while (*d1 && !isSPACE(*d1))
5155 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5158 d1 = moreswitches(d1);
5160 if (PL_doswitches && !switches_done) {
5161 int argc = PL_origargc;
5162 char **argv = PL_origargv;
5165 } while (argc && argv[0][0] == '-' && argv[0][1]);
5166 init_argv_symbols(argc,argv);
5168 if ( (PERLDB_LINE_OR_SAVESRC && !oldpdb)
5169 || ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5170 /* if we have already added "LINE: while (<>) {",
5171 we must not do it again */
5173 sv_setpvs(PL_linestr, "");
5174 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5175 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5176 PL_last_lop = PL_last_uni = NULL;
5177 PL_preambled = FALSE;
5178 if (PERLDB_LINE_OR_SAVESRC)
5179 (void)gv_fetchfile(PL_origfilename);
5186 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5187 PL_lex_state = LEX_FORMLINE;
5188 force_next(FORMRBRACK);
5193 #ifdef PERL_STRICT_CR
5194 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5196 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5198 case ' ': case '\t': case '\f': case '\v':
5203 if (PL_lex_state != LEX_NORMAL
5204 || (PL_in_eval && !PL_rsfp && !PL_parser->filtered))
5206 const bool in_comment = *s == '#';
5207 if (*s == '#' && s == PL_linestart && PL_in_eval
5208 && !PL_rsfp && !PL_parser->filtered) {
5209 /* handle eval qq[#line 1 "foo"\n ...] */
5210 CopLINE_dec(PL_curcop);
5214 while (d < PL_bufend && *d != '\n')
5218 else if (d > PL_bufend)
5219 /* Found by Ilya: feed random input to Perl. */
5220 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5223 if (in_comment && d == PL_bufend
5224 && PL_lex_state == LEX_INTERPNORMAL
5225 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5226 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5229 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5230 PL_lex_state = LEX_FORMLINE;
5231 force_next(FORMRBRACK);
5236 while (s < PL_bufend && *s != '\n')
5244 else if (s > PL_bufend)
5245 /* Found by Ilya: feed random input to Perl. */
5246 Perl_croak(aTHX_ "panic: input overflow");
5250 if (s[1] && isALPHA(s[1]) && !isWORDCHAR(s[2])) {
5258 while (s < PL_bufend && SPACE_OR_TAB(*s))
5261 if (strnEQ(s,"=>",2)) {
5262 s = force_word(PL_bufptr,WORD,FALSE,FALSE);
5263 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5264 OPERATOR('-'); /* unary minus */
5267 case 'r': ftst = OP_FTEREAD; break;
5268 case 'w': ftst = OP_FTEWRITE; break;
5269 case 'x': ftst = OP_FTEEXEC; break;
5270 case 'o': ftst = OP_FTEOWNED; break;
5271 case 'R': ftst = OP_FTRREAD; break;
5272 case 'W': ftst = OP_FTRWRITE; break;
5273 case 'X': ftst = OP_FTREXEC; break;
5274 case 'O': ftst = OP_FTROWNED; break;
5275 case 'e': ftst = OP_FTIS; break;
5276 case 'z': ftst = OP_FTZERO; break;
5277 case 's': ftst = OP_FTSIZE; break;
5278 case 'f': ftst = OP_FTFILE; break;
5279 case 'd': ftst = OP_FTDIR; break;
5280 case 'l': ftst = OP_FTLINK; break;
5281 case 'p': ftst = OP_FTPIPE; break;
5282 case 'S': ftst = OP_FTSOCK; break;
5283 case 'u': ftst = OP_FTSUID; break;
5284 case 'g': ftst = OP_FTSGID; break;
5285 case 'k': ftst = OP_FTSVTX; break;
5286 case 'b': ftst = OP_FTBLK; break;
5287 case 'c': ftst = OP_FTCHR; break;
5288 case 't': ftst = OP_FTTTY; break;
5289 case 'T': ftst = OP_FTTEXT; break;
5290 case 'B': ftst = OP_FTBINARY; break;
5291 case 'M': case 'A': case 'C':
5292 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5294 case 'M': ftst = OP_FTMTIME; break;
5295 case 'A': ftst = OP_FTATIME; break;
5296 case 'C': ftst = OP_FTCTIME; break;
5304 PL_last_uni = PL_oldbufptr;
5305 PL_last_lop_op = (OPCODE)ftst;
5306 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5307 "### Saw file test %c\n", (int)tmp);
5312 /* Assume it was a minus followed by a one-letter named
5313 * subroutine call (or a -bareword), then. */
5314 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5315 "### '-%c' looked like a file test but was not\n",
5322 const char tmp = *s++;
5325 if (PL_expect == XOPERATOR)
5330 else if (*s == '>') {
5333 if (((*s == '$' || *s == '&') && s[1] == '*')
5334 ||(*s == '$' && s[1] == '#' && s[2] == '*')
5335 ||((*s == '@' || *s == '%') && strchr("*[{", s[1]))
5336 ||(*s == '*' && (s[1] == '*' || s[1] == '{'))
5339 PL_expect = XPOSTDEREF;
5342 if (isIDFIRST_lazy_if(s,UTF)) {
5343 s = force_word(s,METHOD,FALSE,TRUE);
5351 if (PL_expect == XOPERATOR) {
5353 && !PL_lex_allbrackets
5354 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5362 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5364 OPERATOR('-'); /* unary minus */
5370 const char tmp = *s++;
5373 if (PL_expect == XOPERATOR)
5378 if (PL_expect == XOPERATOR) {
5380 && !PL_lex_allbrackets
5381 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5389 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5396 if (PL_expect == XPOSTDEREF) POSTDEREF('*');
5397 if (PL_expect != XOPERATOR) {
5398 s = scan_ident(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5399 PL_expect = XOPERATOR;
5400 force_ident(PL_tokenbuf, '*');
5408 if (*s == '=' && !PL_lex_allbrackets
5409 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5417 && !PL_lex_allbrackets
5418 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5423 PL_parser->saw_infix_sigil = 1;
5428 if (PL_expect == XOPERATOR) {
5430 && !PL_lex_allbrackets
5431 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5436 PL_parser->saw_infix_sigil = 1;
5439 else if (PL_expect == XPOSTDEREF) POSTDEREF('%');
5440 PL_tokenbuf[0] = '%';
5441 s = scan_ident(s, PL_tokenbuf + 1,
5442 sizeof PL_tokenbuf - 1, FALSE);
5444 if (!PL_tokenbuf[1]) {
5447 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5449 PL_tokenbuf[0] = '@';
5451 PL_expect = XOPERATOR;
5452 force_ident_maybe_lex('%');
5457 bof = FEATURE_BITWISE_IS_ENABLED;
5458 if (bof && s[1] == '.')
5460 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5461 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5467 BOop(bof ? d == s-2 ? OP_SBIT_XOR : OP_NBIT_XOR : OP_BIT_XOR);
5469 if (PL_lex_brackets > 100)
5470 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5471 PL_lex_brackstack[PL_lex_brackets++] = 0;
5472 PL_lex_allbrackets++;
5474 const char tmp = *s++;
5479 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5481 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5484 Perl_ck_warner_d(aTHX_
5485 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
5486 "Smartmatch is experimental");
5490 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.') {
5492 BCop(OP_SCOMPLEMENT);
5494 BCop(bof ? OP_NCOMPLEMENT : OP_COMPLEMENT);
5496 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5503 goto just_a_word_zero_gv;
5509 switch (PL_expect) {
5511 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5513 PL_bufptr = s; /* update in case we back off */
5516 "Use of := for an empty attribute list is not allowed");
5523 PL_expect = XTERMBLOCK;
5527 while (isIDFIRST_lazy_if(s,UTF)) {
5530 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5531 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5532 if (tmp < 0) tmp = -tmp;
5547 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5549 d = scan_str(d,TRUE,TRUE,FALSE,NULL);
5550 COPLINE_SET_FROM_MULTI_END;
5552 /* MUST advance bufptr here to avoid bogus
5553 "at end of line" context messages from yyerror().
5555 PL_bufptr = s + len;
5556 yyerror("Unterminated attribute parameter in attribute list");
5560 return REPORT(0); /* EOF indicator */
5564 sv_catsv(sv, PL_lex_stuff);
5565 attrs = op_append_elem(OP_LIST, attrs,
5566 newSVOP(OP_CONST, 0, sv));
5567 SvREFCNT_dec_NN(PL_lex_stuff);
5568 PL_lex_stuff = NULL;
5571 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5573 if (PL_in_my == KEY_our) {
5574 deprecate(":unique");
5577 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5580 /* NOTE: any CV attrs applied here need to be part of
5581 the CVf_BUILTIN_ATTRS define in cv.h! */
5582 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5584 CvLVALUE_on(PL_compcv);
5586 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5588 deprecate(":locked");
5590 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5592 CvMETHOD_on(PL_compcv);
5594 else if (!PL_in_my && len == 5
5595 && strnEQ(SvPVX(sv), "const", len))
5598 Perl_ck_warner_d(aTHX_
5599 packWARN(WARN_EXPERIMENTAL__CONST_ATTR),
5600 ":const is experimental"
5602 CvANONCONST_on(PL_compcv);
5603 if (!CvANON(PL_compcv))
5604 yyerror(":const is not permitted on named "
5607 /* After we've set the flags, it could be argued that
5608 we don't need to do the attributes.pm-based setting
5609 process, and shouldn't bother appending recognized
5610 flags. To experiment with that, uncomment the
5611 following "else". (Note that's already been
5612 uncommented. That keeps the above-applied built-in
5613 attributes from being intercepted (and possibly
5614 rejected) by a package's attribute routines, but is
5615 justified by the performance win for the common case
5616 of applying only built-in attributes.) */
5618 attrs = op_append_elem(OP_LIST, attrs,
5619 newSVOP(OP_CONST, 0,
5623 if (*s == ':' && s[1] != ':')
5626 break; /* require real whitespace or :'s */
5627 /* XXX losing whitespace on sequential attributes here */
5632 && !(PL_expect == XOPERATOR
5633 ? (*s == '=' || *s == ')')
5634 : (*s == '{' || *s == '(')))
5636 const char q = ((*s == '\'') ? '"' : '\'');
5637 /* If here for an expression, and parsed no attrs, back
5639 if (PL_expect == XOPERATOR && !attrs) {
5643 /* MUST advance bufptr here to avoid bogus "at end of line"
5644 context messages from yyerror().
5647 yyerror( (const char *)
5649 ? Perl_form(aTHX_ "Invalid separator character "
5650 "%c%c%c in attribute list", q, *s, q)
5651 : "Unterminated attribute list" ) );
5659 NEXTVAL_NEXTTOKE.opval = attrs;
5665 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5669 PL_lex_allbrackets--;
5673 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5674 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5678 PL_lex_allbrackets++;
5681 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5688 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5691 PL_lex_allbrackets--;
5697 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5700 if (PL_lex_brackets <= 0)
5701 /* diag_listed_as: Unmatched right %s bracket */
5702 yyerror("Unmatched right square bracket");
5705 PL_lex_allbrackets--;
5706 if (PL_lex_state == LEX_INTERPNORMAL) {
5707 if (PL_lex_brackets == 0) {
5708 if (*s == '-' && s[1] == '>')
5709 PL_lex_state = LEX_INTERPENDMAYBE;
5710 else if (*s != '[' && *s != '{')
5711 PL_lex_state = LEX_INTERPEND;
5718 if (PL_lex_brackets > 100) {
5719 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5721 switch (PL_expect) {
5724 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5725 PL_lex_allbrackets++;
5726 OPERATOR(HASHBRACK);
5728 while (s < PL_bufend && SPACE_OR_TAB(*s))
5731 PL_tokenbuf[0] = '\0';
5732 if (d < PL_bufend && *d == '-') {
5733 PL_tokenbuf[0] = '-';
5735 while (d < PL_bufend && SPACE_OR_TAB(*d))
5738 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5739 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5741 while (d < PL_bufend && SPACE_OR_TAB(*d))
5744 const char minus = (PL_tokenbuf[0] == '-');
5745 s = force_word(s + minus, WORD, FALSE, TRUE);
5753 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5754 PL_lex_allbrackets++;
5759 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5760 PL_lex_allbrackets++;
5764 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5765 PL_lex_allbrackets++;
5770 if (PL_oldoldbufptr == PL_last_lop)
5771 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5773 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5774 PL_lex_allbrackets++;
5777 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5779 /* This hack is to get the ${} in the message. */
5781 yyerror("syntax error");
5784 OPERATOR(HASHBRACK);
5786 if (PL_expect == XREF && PL_oldoldbufptr != PL_last_lop) {
5787 /* ${...} or @{...} etc., but not print {...}
5788 * Skip the disambiguation and treat this as a block.
5790 goto block_expectation;
5792 /* This hack serves to disambiguate a pair of curlies
5793 * as being a block or an anon hash. Normally, expectation
5794 * determines that, but in cases where we're not in a
5795 * position to expect anything in particular (like inside
5796 * eval"") we have to resolve the ambiguity. This code
5797 * covers the case where the first term in the curlies is a
5798 * quoted string. Most other cases need to be explicitly
5799 * disambiguated by prepending a "+" before the opening
5800 * curly in order to force resolution as an anon hash.
5802 * XXX should probably propagate the outer expectation
5803 * into eval"" to rely less on this hack, but that could
5804 * potentially break current behavior of eval"".
5808 if (*s == '\'' || *s == '"' || *s == '`') {
5809 /* common case: get past first string, handling escapes */
5810 for (t++; t < PL_bufend && *t != *s;)
5815 else if (*s == 'q') {
5818 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5819 && !isWORDCHAR(*t))))
5821 /* skip q//-like construct */
5823 char open, close, term;
5826 while (t < PL_bufend && isSPACE(*t))
5828 /* check for q => */
5829 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5830 OPERATOR(HASHBRACK);
5834 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5838 for (t++; t < PL_bufend; t++) {
5839 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5841 else if (*t == open)
5845 for (t++; t < PL_bufend; t++) {
5846 if (*t == '\\' && t+1 < PL_bufend)
5848 else if (*t == close && --brackets <= 0)
5850 else if (*t == open)
5857 /* skip plain q word */
5858 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5861 else if (isWORDCHAR_lazy_if(t,UTF)) {
5863 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
5866 while (t < PL_bufend && isSPACE(*t))
5868 /* if comma follows first term, call it an anon hash */
5869 /* XXX it could be a comma expression with loop modifiers */
5870 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5871 || (*t == '=' && t[1] == '>')))
5872 OPERATOR(HASHBRACK);
5873 if (PL_expect == XREF)
5876 /* If there is an opening brace or 'sub:', treat it
5877 as a term to make ${{...}}{k} and &{sub:attr...}
5878 dwim. Otherwise, treat it as a statement, so
5879 map {no strict; ...} works.
5886 if (strnEQ(s, "sub", 3)) {
5897 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5903 pl_yylval.ival = CopLINE(PL_curcop);
5904 PL_copline = NOLINE; /* invalidate current command line number */
5905 TOKEN(formbrack ? '=' : '{');
5907 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5911 if (PL_lex_brackets <= 0)
5912 /* diag_listed_as: Unmatched right %s bracket */
5913 yyerror("Unmatched right curly bracket");
5915 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5916 PL_lex_allbrackets--;
5917 if (PL_lex_state == LEX_INTERPNORMAL) {
5918 if (PL_lex_brackets == 0) {
5919 if (PL_expect & XFAKEBRACK) {
5920 PL_expect &= XENUMMASK;
5921 PL_lex_state = LEX_INTERPEND;
5923 return yylex(); /* ignore fake brackets */
5925 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5926 && SvEVALED(PL_lex_repl))
5927 PL_lex_state = LEX_INTERPEND;
5928 else if (*s == '-' && s[1] == '>')
5929 PL_lex_state = LEX_INTERPENDMAYBE;
5930 else if (*s != '[' && *s != '{')
5931 PL_lex_state = LEX_INTERPEND;
5934 if (PL_expect & XFAKEBRACK) {
5935 PL_expect &= XENUMMASK;
5937 return yylex(); /* ignore fake brackets */
5939 force_next(formbrack ? '.' : '}');
5940 if (formbrack) LEAVE;
5941 if (formbrack == 2) { /* means . where arguments were expected */
5947 if (PL_expect == XPOSTDEREF) POSTDEREF('&');
5950 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5951 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5958 if (PL_expect == XOPERATOR) {
5959 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5960 && isIDFIRST_lazy_if(s,UTF))
5962 CopLINE_dec(PL_curcop);
5963 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5964 CopLINE_inc(PL_curcop);
5967 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.')
5969 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5970 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5976 PL_parser->saw_infix_sigil = 1;
5977 BAop(bof ? OP_NBIT_AND : OP_BIT_AND);
5983 PL_tokenbuf[0] = '&';
5984 s = scan_ident(s - 1, PL_tokenbuf + 1,
5985 sizeof PL_tokenbuf - 1, TRUE);
5986 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5987 if (PL_tokenbuf[1]) {
5988 force_ident_maybe_lex('&');
5997 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5998 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6006 if ((bof = FEATURE_BITWISE_IS_ENABLED) && *s == '.')
6008 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6009 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6013 BOop(bof ? s == d ? OP_NBIT_OR : OP_SBIT_OR : OP_BIT_OR);
6017 const char tmp = *s++;
6019 if (!PL_lex_allbrackets
6020 && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6028 if (!PL_lex_allbrackets
6029 && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
6038 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6039 && strchr("+-*/%.^&|<",tmp))
6040 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6041 "Reversed %c= operator",(int)tmp);
6043 if (PL_expect == XSTATE
6045 && (s == PL_linestart+1 || s[-2] == '\n') )
6047 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6048 || PL_lex_state != LEX_NORMAL) {
6053 if (strnEQ(s,"=cut",4)) {
6067 PL_parser->in_pod = 1;
6071 if (PL_expect == XBLOCK) {
6073 #ifdef PERL_STRICT_CR
6074 while (SPACE_OR_TAB(*t))
6076 while (SPACE_OR_TAB(*t) || *t == '\r')
6079 if (*t == '\n' || *t == '#') {
6082 SAVEI8(PL_parser->form_lex_state);
6083 SAVEI32(PL_lex_formbrack);
6084 PL_parser->form_lex_state = PL_lex_state;
6085 PL_lex_formbrack = PL_lex_brackets + 1;
6089 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6098 const char tmp = *s++;
6100 /* was this !=~ where !~ was meant?
6101 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6103 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6104 const char *t = s+1;
6106 while (t < PL_bufend && isSPACE(*t))
6109 if (*t == '/' || *t == '?'
6110 || ((*t == 'm' || *t == 's' || *t == 'y')
6111 && !isWORDCHAR(t[1]))
6112 || (*t == 't' && t[1] == 'r' && !isWORDCHAR(t[2])))
6113 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6114 "!=~ should be !~");
6116 if (!PL_lex_allbrackets
6117 && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6130 if (PL_expect != XOPERATOR) {
6131 if (s[1] != '<' && !strchr(s,'>'))
6133 if (s[1] == '<' && s[2] != '>')
6134 s = scan_heredoc(s);
6136 s = scan_inputsymbol(s);
6137 PL_expect = XOPERATOR;
6138 TOKEN(sublex_start());
6144 if (*s == '=' && !PL_lex_allbrackets
6145 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
6150 SHop(OP_LEFT_SHIFT);
6155 if (!PL_lex_allbrackets
6156 && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6164 if (!PL_lex_allbrackets
6165 && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6174 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6182 const char tmp = *s++;
6184 if (*s == '=' && !PL_lex_allbrackets
6185 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
6190 SHop(OP_RIGHT_SHIFT);
6192 else if (tmp == '=') {
6193 if (!PL_lex_allbrackets
6194 && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
6203 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6212 if (PL_expect == XOPERATOR) {
6213 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6214 return deprecate_commaless_var_list();
6217 else if (PL_expect == XPOSTDEREF) {
6220 POSTDEREF(DOLSHARP);
6225 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6226 PL_tokenbuf[0] = '@';
6227 s = scan_ident(s + 1, PL_tokenbuf + 1,
6228 sizeof PL_tokenbuf - 1, FALSE);
6229 if (PL_expect == XOPERATOR) {
6231 if (PL_bufptr > s) {
6233 PL_bufptr = PL_oldbufptr;
6235 no_op("Array length", d);
6237 if (!PL_tokenbuf[1])
6239 PL_expect = XOPERATOR;
6240 force_ident_maybe_lex('#');
6244 PL_tokenbuf[0] = '$';
6245 s = scan_ident(s, PL_tokenbuf + 1,
6246 sizeof PL_tokenbuf - 1, FALSE);
6247 if (PL_expect == XOPERATOR) {
6249 if (PL_bufptr > s) {
6251 PL_bufptr = PL_oldbufptr;
6255 if (!PL_tokenbuf[1]) {
6257 yyerror("Final $ should be \\$ or $name");
6263 const char tmp = *s;
6264 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6267 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6268 && intuit_more(s)) {
6270 PL_tokenbuf[0] = '@';
6271 if (ckWARN(WARN_SYNTAX)) {
6274 while (isSPACE(*t) || isWORDCHAR_lazy_if(t,UTF) || *t == '$')
6275 t += UTF ? UTF8SKIP(t) : 1;
6277 PL_bufptr = skipspace(PL_bufptr); /* XXX can realloc */
6278 while (t < PL_bufend && *t != ']')
6280 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6281 "Multidimensional syntax %"UTF8f" not supported",
6282 UTF8fARG(UTF,(int)((t - PL_bufptr) + 1), PL_bufptr));
6286 else if (*s == '{') {
6288 PL_tokenbuf[0] = '%';
6289 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6290 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6292 char tmpbuf[sizeof PL_tokenbuf];
6295 } while (isSPACE(*t));
6296 if (isIDFIRST_lazy_if(t,UTF)) {
6298 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6303 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6304 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6305 "You need to quote \"%"UTF8f"\"",
6306 UTF8fARG(UTF, len, tmpbuf));
6312 PL_expect = XOPERATOR;
6313 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6314 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6315 if (!islop || PL_last_lop_op == OP_GREPSTART)
6316 PL_expect = XOPERATOR;
6317 else if (strchr("$@\"'`q", *s))
6318 PL_expect = XTERM; /* e.g. print $fh "foo" */
6319 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6320 PL_expect = XTERM; /* e.g. print $fh &sub */
6321 else if (isIDFIRST_lazy_if(s,UTF)) {
6322 char tmpbuf[sizeof PL_tokenbuf];
6324 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6325 if ((t2 = keyword(tmpbuf, len, 0))) {
6326 /* binary operators exclude handle interpretations */
6338 PL_expect = XTERM; /* e.g. print $fh length() */
6343 PL_expect = XTERM; /* e.g. print $fh subr() */
6346 else if (isDIGIT(*s))
6347 PL_expect = XTERM; /* e.g. print $fh 3 */
6348 else if (*s == '.' && isDIGIT(s[1]))
6349 PL_expect = XTERM; /* e.g. print $fh .3 */
6350 else if ((*s == '?' || *s == '-' || *s == '+')
6351 && !isSPACE(s[1]) && s[1] != '=')
6352 PL_expect = XTERM; /* e.g. print $fh -1 */
6353 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6355 PL_expect = XTERM; /* e.g. print $fh /.../
6356 XXX except DORDOR operator
6358 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6360 PL_expect = XTERM; /* print $fh <<"EOF" */
6363 force_ident_maybe_lex('$');
6367 if (PL_expect == XPOSTDEREF)
6369 PL_tokenbuf[0] = '@';
6370 s = scan_ident(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6371 if (PL_expect == XOPERATOR) {
6373 if (PL_bufptr > s) {
6375 PL_bufptr = PL_oldbufptr;
6380 if (!PL_tokenbuf[1]) {
6383 if (PL_lex_state == LEX_NORMAL)
6385 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6387 PL_tokenbuf[0] = '%';
6389 /* Warn about @ where they meant $. */
6390 if (*s == '[' || *s == '{') {
6391 if (ckWARN(WARN_SYNTAX)) {
6392 S_check_scalar_slice(aTHX_ s);
6396 PL_expect = XOPERATOR;
6397 force_ident_maybe_lex('@');
6400 case '/': /* may be division, defined-or, or pattern */
6401 if ((PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR) && s[1] == '/') {
6402 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6403 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6408 else if (PL_expect == XOPERATOR) {
6410 if (*s == '=' && !PL_lex_allbrackets
6411 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
6419 /* Disable warning on "study /blah/" */
6420 if (PL_oldoldbufptr == PL_last_uni
6421 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6422 || memNE(PL_last_uni, "study", 5)
6423 || isWORDCHAR_lazy_if(PL_last_uni+5,UTF)
6426 s = scan_pat(s,OP_MATCH);
6427 TERM(sublex_start());
6430 case '?': /* conditional */
6432 if (!PL_lex_allbrackets
6433 && PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE)
6438 PL_lex_allbrackets++;
6442 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6443 #ifdef PERL_STRICT_CR
6446 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6448 && (s == PL_linestart || s[-1] == '\n') )
6451 formbrack = 2; /* dot seen where arguments expected */
6454 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6458 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6461 if (!PL_lex_allbrackets
6462 && PL_lex_fakeeof >= LEX_FAKEEOF_RANGE)
6470 pl_yylval.ival = OPf_SPECIAL;
6476 if (*s == '=' && !PL_lex_allbrackets
6477 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
6485 case '0': case '1': case '2': case '3': case '4':
6486 case '5': case '6': case '7': case '8': case '9':
6487 s = scan_num(s, &pl_yylval);
6488 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6489 if (PL_expect == XOPERATOR)
6494 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6497 COPLINE_SET_FROM_MULTI_END;
6498 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6499 if (PL_expect == XOPERATOR) {
6500 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6501 return deprecate_commaless_var_list();
6506 pl_yylval.ival = OP_CONST;
6507 TERM(sublex_start());
6510 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6513 printbuf("### Saw string before %s\n", s);
6515 PerlIO_printf(Perl_debug_log,
6516 "### Saw unterminated string\n");
6518 if (PL_expect == XOPERATOR) {
6519 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6520 return deprecate_commaless_var_list();
6527 pl_yylval.ival = OP_CONST;
6528 /* FIXME. I think that this can be const if char *d is replaced by
6529 more localised variables. */
6530 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6531 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6532 pl_yylval.ival = OP_STRINGIFY;
6536 if (pl_yylval.ival == OP_CONST)
6537 COPLINE_SET_FROM_MULTI_END;
6538 TERM(sublex_start());
6541 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
6544 printbuf("### Saw backtick string before %s\n", s);
6546 PerlIO_printf(Perl_debug_log,
6547 "### Saw unterminated backtick string\n");
6549 if (PL_expect == XOPERATOR)
6550 no_op("Backticks",s);
6553 pl_yylval.ival = OP_BACKTICK;
6554 TERM(sublex_start());
6558 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6560 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6562 if (PL_expect == XOPERATOR)
6563 no_op("Backslash",s);
6567 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6568 char *start = s + 2;
6569 while (isDIGIT(*start) || *start == '_')
6571 if (*start == '.' && isDIGIT(start[1])) {
6572 s = scan_num(s, &pl_yylval);
6575 else if ((*start == ':' && start[1] == ':')
6576 || (PL_expect == XSTATE && *start == ':'))
6578 else if (PL_expect == XSTATE) {
6580 while (d < PL_bufend && isSPACE(*d)) d++;
6581 if (*d == ':') goto keylookup;
6583 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6584 if (!isALPHA(*start) && (PL_expect == XTERM
6585 || PL_expect == XREF || PL_expect == XSTATE
6586 || PL_expect == XTERMORDORDOR)) {
6587 GV *const gv = gv_fetchpvn_flags(s, start - s,
6588 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6590 s = scan_num(s, &pl_yylval);
6597 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6650 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6652 /* Some keywords can be followed by any delimiter, including ':' */
6653 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6655 /* x::* is just a word, unless x is "CORE" */
6656 if (!anydelim && *s == ':' && s[1] == ':') {
6657 if (strEQ(PL_tokenbuf, "CORE")) goto case_KEY_CORE;
6662 while (d < PL_bufend && isSPACE(*d))
6663 d++; /* no comments skipped here, or s### is misparsed */
6665 /* Is this a word before a => operator? */
6666 if (*d == '=' && d[1] == '>') {
6670 = (OP*)newSVOP(OP_CONST, 0,
6671 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6672 pl_yylval.opval->op_private = OPpCONST_BARE;
6676 /* Check for plugged-in keyword */
6680 char *saved_bufptr = PL_bufptr;
6682 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6684 if (result == KEYWORD_PLUGIN_DECLINE) {
6685 /* not a plugged-in keyword */
6686 PL_bufptr = saved_bufptr;
6687 } else if (result == KEYWORD_PLUGIN_STMT) {
6688 pl_yylval.opval = o;
6690 if (!PL_nexttoke) PL_expect = XSTATE;
6691 return REPORT(PLUGSTMT);
6692 } else if (result == KEYWORD_PLUGIN_EXPR) {
6693 pl_yylval.opval = o;
6695 if (!PL_nexttoke) PL_expect = XOPERATOR;
6696 return REPORT(PLUGEXPR);
6698 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6703 /* Check for built-in keyword */
6704 tmp = keyword(PL_tokenbuf, len, 0);
6706 /* Is this a label? */
6707 if (!anydelim && PL_expect == XSTATE
6708 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6710 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6711 pl_yylval.pval[len] = '\0';
6712 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6717 /* Check for lexical sub */
6718 if (PL_expect != XOPERATOR) {
6719 char tmpbuf[sizeof PL_tokenbuf + 1];
6721 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6722 off = pad_findmy_pvn(tmpbuf, len+1, 0);
6723 if (off != NOT_IN_PAD) {
6724 assert(off); /* we assume this is boolean-true below */
6725 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6726 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6727 HEK * const stashname = HvNAME_HEK(stash);
6728 sv = newSVhek(stashname);
6729 sv_catpvs(sv, "::");
6730 sv_catpvn_flags(sv, PL_tokenbuf, len,
6731 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6732 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6742 rv2cv_op = newOP(OP_PADANY, 0);
6743 rv2cv_op->op_targ = off;
6744 cv = find_lexical_cv(off);
6752 if (tmp < 0) { /* second-class keyword? */
6753 GV *ogv = NULL; /* override (winner) */
6754 GV *hgv = NULL; /* hidden (loser) */
6755 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6757 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6758 (UTF ? SVf_UTF8 : 0)|GV_NOTQUAL,
6760 && (cv = GvCVu(gv)))
6762 if (GvIMPORTED_CV(gv))
6764 else if (! CvMETHOD(cv))
6768 && (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6771 && (isGV_with_GP(gv)
6772 ? GvCVu(gv) && GvIMPORTED_CV(gv)
6773 : SvPCS_IMPORTED(gv)
6774 && (gv_init(gv, PL_globalstash, PL_tokenbuf,
6782 tmp = 0; /* overridden by import or by GLOBAL */
6785 && -tmp==KEY_lock /* XXX generalizable kludge */
6788 tmp = 0; /* any sub overrides "weak" keyword */
6790 else { /* no override */
6792 if (tmp == KEY_dump) {
6793 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6794 "dump() better written as CORE::dump()");
6798 if (hgv && tmp != KEY_x) /* never ambiguous */
6799 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6800 "Ambiguous call resolved as CORE::%s(), "
6801 "qualify as such or use &",
6806 if (tmp && tmp != KEY___DATA__ && tmp != KEY___END__
6807 && (!anydelim || *s != '#')) {
6808 /* no override, and not s### either; skipspace is safe here
6809 * check for => on following line */
6811 STRLEN bufoff = PL_bufptr - SvPVX(PL_linestr);
6812 STRLEN soff = s - SvPVX(PL_linestr);
6813 s = skipspace_flags(s, LEX_NO_INCLINE);
6814 arrow = *s == '=' && s[1] == '>';
6815 PL_bufptr = SvPVX(PL_linestr) + bufoff;
6816 s = SvPVX(PL_linestr) + soff;
6824 default: /* not a keyword */
6825 /* Trade off - by using this evil construction we can pull the
6826 variable gv into the block labelled keylookup. If not, then
6827 we have to give it function scope so that the goto from the
6828 earlier ':' case doesn't bypass the initialisation. */
6830 just_a_word_zero_gv:
6842 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6846 /* Get the rest if it looks like a package qualifier */
6848 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6850 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6853 Perl_croak(aTHX_ "Bad name after %"UTF8f"%s",
6854 UTF8fARG(UTF, len, PL_tokenbuf),
6855 *s == '\'' ? "'" : "::");
6860 if (PL_expect == XOPERATOR) {
6861 if (PL_bufptr == PL_linestart) {
6862 CopLINE_dec(PL_curcop);
6863 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6864 CopLINE_inc(PL_curcop);
6867 no_op("Bareword",s);
6870 /* See if the name is "Foo::",
6871 in which case Foo is a bareword
6872 (and a package name). */
6875 && PL_tokenbuf[len - 2] == ':'
6876 && PL_tokenbuf[len - 1] == ':')
6878 if (ckWARN(WARN_BAREWORD)
6879 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6880 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6881 "Bareword \"%"UTF8f"\" refers to nonexistent package",
6882 UTF8fARG(UTF, len, PL_tokenbuf));
6884 PL_tokenbuf[len] = '\0';
6893 /* if we saw a global override before, get the right name */
6896 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6899 SV * const tmp_sv = sv;
6900 sv = newSVpvs("CORE::GLOBAL::");
6901 sv_catsv(sv, tmp_sv);
6902 SvREFCNT_dec(tmp_sv);
6906 /* Presume this is going to be a bareword of some sort. */
6908 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6909 pl_yylval.opval->op_private = OPpCONST_BARE;
6911 /* And if "Foo::", then that's what it certainly is. */
6917 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6918 const_op->op_private = OPpCONST_BARE;
6920 newCVREF(OPpMAY_RETURN_CONSTANT<<8, const_op);
6924 : SvROK(gv) && SvTYPE(SvRV(gv)) == SVt_PVCV
6927 : rv2cv_op_cv(rv2cv_op, RV2CVOPCV_RETURN_STUB);
6930 /* Use this var to track whether intuit_method has been
6931 called. intuit_method returns 0 or > 255. */
6934 /* See if it's the indirect object for a list operator. */
6937 && PL_oldoldbufptr < PL_bufptr
6938 && (PL_oldoldbufptr == PL_last_lop
6939 || PL_oldoldbufptr == PL_last_uni)
6940 && /* NO SKIPSPACE BEFORE HERE! */
6942 || ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7)
6945 bool immediate_paren = *s == '(';
6947 /* (Now we can afford to cross potential line boundary.) */
6950 /* Two barewords in a row may indicate method call. */
6952 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$')
6953 && (tmp = intuit_method(s, lex ? NULL : sv, cv)))
6958 /* If not a declared subroutine, it's an indirect object. */
6959 /* (But it's an indir obj regardless for sort.) */
6960 /* Also, if "_" follows a filetest operator, it's a bareword */
6963 ( !immediate_paren && (PL_last_lop_op == OP_SORT
6965 && (PL_last_lop_op != OP_MAPSTART
6966 && PL_last_lop_op != OP_GREPSTART))))
6967 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6968 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK)
6972 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6977 PL_expect = XOPERATOR;
6980 /* Is this a word before a => operator? */
6981 if (*s == '=' && s[1] == '>' && !pkgname) {
6984 if (gvp || (lex && !off)) {
6985 assert (cSVOPx(pl_yylval.opval)->op_sv == sv);
6986 /* This is our own scalar, created a few lines
6987 above, so this is safe. */
6989 sv_setpv(sv, PL_tokenbuf);
6990 if (UTF && !IN_BYTES
6991 && is_utf8_string((U8*)PL_tokenbuf, len))
6998 /* If followed by a paren, it's certainly a subroutine. */
7003 while (SPACE_OR_TAB(*d))
7005 if (*d == ')' && (sv = cv_const_sv_or_av(cv))) {
7010 NEXTVAL_NEXTTOKE.opval =
7011 off ? rv2cv_op : pl_yylval.opval;
7013 op_free(pl_yylval.opval), force_next(PRIVATEREF);
7014 else op_free(rv2cv_op), force_next(WORD);
7019 /* If followed by var or block, call it a method (unless sub) */
7021 if ((*s == '$' || *s == '{') && !cv) {
7023 PL_last_lop = PL_oldbufptr;
7024 PL_last_lop_op = OP_METHOD;
7025 if (!PL_lex_allbrackets
7026 && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7028 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7030 PL_expect = XBLOCKTERM;
7032 return REPORT(METHOD);
7035 /* If followed by a bareword, see if it looks like indir obj. */
7037 if (tmp == 1 && !orig_keyword
7038 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7039 && (tmp = intuit_method(s, lex ? NULL : sv, cv))) {
7042 assert(cSVOPx(pl_yylval.opval)->op_sv == sv);
7044 sv_setpvn(sv, PL_tokenbuf, len);
7045 if (UTF && !IN_BYTES
7046 && is_utf8_string((U8*)PL_tokenbuf, len))
7048 else SvUTF8_off(sv);
7051 if (tmp == METHOD && !PL_lex_allbrackets
7052 && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7054 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7059 /* Not a method, so call it a subroutine (if defined) */
7062 /* Check for a constant sub */
7063 if ((sv = cv_const_sv_or_av(cv))) {
7066 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7067 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7068 if (SvTYPE(sv) == SVt_PVAV)
7069 pl_yylval.opval = newUNOP(OP_RV2AV, OPf_PARENS,
7072 pl_yylval.opval->op_private = 0;
7073 pl_yylval.opval->op_folded = 1;
7074 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7079 op_free(pl_yylval.opval);
7081 off ? (OP *)newCVREF(0, rv2cv_op) : rv2cv_op;
7082 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7083 PL_last_lop = PL_oldbufptr;
7084 PL_last_lop_op = OP_ENTERSUB;
7085 /* Is there a prototype? */
7089 STRLEN protolen = CvPROTOLEN(cv);
7090 const char *proto = CvPROTO(cv);
7092 proto = S_strip_spaces(aTHX_ proto, &protolen);
7095 if ((optional = *proto == ';'))
7098 while (*proto == ';');
7102 *proto == '$' || *proto == '_'
7103 || *proto == '*' || *proto == '+'
7108 *proto == '\\' && proto[1] && proto[2] == '\0'
7111 UNIPROTO(UNIOPSUB,optional);
7112 if (*proto == '\\' && proto[1] == '[') {
7113 const char *p = proto + 2;
7114 while(*p && *p != ']')
7116 if(*p == ']' && !p[1])
7117 UNIPROTO(UNIOPSUB,optional);
7119 if (*proto == '&' && *s == '{') {
7121 sv_setpvs(PL_subname, "__ANON__");
7123 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7124 if (!PL_lex_allbrackets
7125 && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7127 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7132 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7134 force_next(off ? PRIVATEREF : WORD);
7135 if (!PL_lex_allbrackets
7136 && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7138 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7143 /* Call it a bare word */
7145 if (PL_hints & HINT_STRICT_SUBS)
7146 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7149 /* after "print" and similar functions (corresponding to
7150 * "F? L" in opcode.pl), whatever wasn't already parsed as
7151 * a filehandle should be subject to "strict subs".
7152 * Likewise for the optional indirect-object argument to system
7153 * or exec, which can't be a bareword */
7154 if ((PL_last_lop_op == OP_PRINT
7155 || PL_last_lop_op == OP_PRTF
7156 || PL_last_lop_op == OP_SAY
7157 || PL_last_lop_op == OP_SYSTEM
7158 || PL_last_lop_op == OP_EXEC)
7159 && (PL_hints & HINT_STRICT_SUBS))
7160 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7161 if (lastchar != '-') {
7162 if (ckWARN(WARN_RESERVED)) {
7166 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7168 /* PL_warn_reserved is constant */
7169 GCC_DIAG_IGNORE(-Wformat-nonliteral);
7170 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7180 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
7181 && saw_infix_sigil) {
7182 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7183 "Operator or semicolon missing before %c%"UTF8f,
7185 UTF8fARG(UTF, strlen(PL_tokenbuf),
7187 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7188 "Ambiguous use of %c resolved as operator %c",
7189 lastchar, lastchar);
7196 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7201 (OP*)newSVOP(OP_CONST, 0,
7202 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7205 case KEY___PACKAGE__:
7207 (OP*)newSVOP(OP_CONST, 0,
7209 ? newSVhek(HvNAME_HEK(PL_curstash))
7216 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7217 HV * const stash = PL_tokenbuf[2] == 'D' && PL_curstash
7220 gv = (GV *)*hv_fetchs(stash, "DATA", 1);
7222 gv_init(gv,stash,"DATA",4,0);
7225 GvIOp(gv) = newIO();
7226 IoIFP(GvIOp(gv)) = PL_rsfp;
7227 #if defined(HAS_FCNTL) && defined(F_SETFD) && defined(FD_CLOEXEC)
7229 const int fd = PerlIO_fileno(PL_rsfp);
7231 fcntl(fd,F_SETFD, FD_CLOEXEC);
7235 /* Mark this internal pseudo-handle as clean */
7236 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7237 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7238 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7240 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7241 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7242 /* if the script was opened in binmode, we need to revert
7243 * it to text mode for compatibility; but only iff it has CRs
7244 * XXX this is a questionable hack at best. */
7245 if (PL_bufend-PL_bufptr > 2
7246 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7249 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7250 loc = PerlIO_tell(PL_rsfp);
7251 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7254 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7256 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7257 #endif /* NETWARE */
7259 PerlIO_seek(PL_rsfp, loc, 0);
7263 #ifdef PERLIO_LAYERS
7266 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7267 else if (IN_ENCODING) {
7273 XPUSHs(_get_encoding());
7275 call_method("name", G_SCALAR);
7279 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7280 Perl_form(aTHX_ ":encoding(%"SVf")",
7293 FUN0OP(CvCLONE(PL_compcv)
7294 ? newOP(OP_RUNCV, 0)
7295 : newPVOP(OP_RUNCV,0,NULL));
7304 if (PL_expect == XSTATE) {
7315 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7316 if ((*s == ':' && s[1] == ':')
7317 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7321 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7325 Perl_croak(aTHX_ "CORE::%"UTF8f" is not a keyword",
7326 UTF8fARG(UTF, len, PL_tokenbuf));
7329 else if (tmp == KEY_require || tmp == KEY_do
7331 /* that's a way to remember we saw "CORE::" */
7343 LOP(OP_ACCEPT,XTERM);
7346 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7351 LOP(OP_ATAN2,XTERM);
7357 LOP(OP_BINMODE,XTERM);
7360 LOP(OP_BLESS,XTERM);
7369 /* We have to disambiguate the two senses of
7370 "continue". If the next token is a '{' then
7371 treat it as the start of a continue block;
7372 otherwise treat it as a control operator.
7382 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7392 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7401 if (!PL_cryptseen) {
7402 PL_cryptseen = TRUE;
7406 LOP(OP_CRYPT,XTERM);
7409 LOP(OP_CHMOD,XTERM);
7412 LOP(OP_CHOWN,XTERM);
7415 LOP(OP_CONNECT,XTERM);
7435 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7437 if (len && (len != 4 || strNE(PL_tokenbuf+1, "CORE"))
7438 && !keyword(PL_tokenbuf + 1, len, 0)) {
7441 force_ident_maybe_lex('&');
7446 if (orig_keyword == KEY_do) {
7455 PL_hints |= HINT_BLOCK_SCOPE;
7465 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7466 STR_WITH_LEN("NDBM_File::"),
7467 STR_WITH_LEN("DB_File::"),
7468 STR_WITH_LEN("GDBM_File::"),
7469 STR_WITH_LEN("SDBM_File::"),
7470 STR_WITH_LEN("ODBM_File::"),
7472 LOP(OP_DBMOPEN,XTERM);
7484 pl_yylval.ival = CopLINE(PL_curcop);
7488 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7500 if (*s == '{') { /* block eval */
7501 PL_expect = XTERMBLOCK;
7502 UNIBRACK(OP_ENTERTRY);
7504 else { /* string eval */
7506 UNIBRACK(OP_ENTEREVAL);
7511 UNIBRACK(-OP_ENTEREVAL);
7525 case KEY_endhostent:
7531 case KEY_endservent:
7534 case KEY_endprotoent:
7545 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7547 pl_yylval.ival = CopLINE(PL_curcop);
7549 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7552 if ((PL_bufend - p) >= 3
7553 && strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7557 else if ((PL_bufend - p) >= 4
7558 && strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7561 /* skip optional package name, as in "for my abc $x (..)" */
7562 if (isIDFIRST_lazy_if(p,UTF)) {
7563 p = scan_word(p, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7567 Perl_croak(aTHX_ "Missing $ on loop variable");
7572 LOP(OP_FORMLINE,XTERM);
7581 LOP(OP_FCNTL,XTERM);
7587 LOP(OP_FLOCK,XTERM);
7590 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7595 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7600 LOP(OP_GREPSTART, XREF);
7617 case KEY_getpriority:
7618 LOP(OP_GETPRIORITY,XTERM);
7620 case KEY_getprotobyname:
7623 case KEY_getprotobynumber:
7624 LOP(OP_GPBYNUMBER,XTERM);
7626 case KEY_getprotoent:
7638 case KEY_getpeername:
7639 UNI(OP_GETPEERNAME);
7641 case KEY_gethostbyname:
7644 case KEY_gethostbyaddr:
7645 LOP(OP_GHBYADDR,XTERM);
7647 case KEY_gethostent:
7650 case KEY_getnetbyname:
7653 case KEY_getnetbyaddr:
7654 LOP(OP_GNBYADDR,XTERM);
7659 case KEY_getservbyname:
7660 LOP(OP_GSBYNAME,XTERM);
7662 case KEY_getservbyport:
7663 LOP(OP_GSBYPORT,XTERM);
7665 case KEY_getservent:
7668 case KEY_getsockname:
7669 UNI(OP_GETSOCKNAME);
7671 case KEY_getsockopt:
7672 LOP(OP_GSOCKOPT,XTERM);
7687 pl_yylval.ival = CopLINE(PL_curcop);
7688 Perl_ck_warner_d(aTHX_
7689 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
7690 "given is experimental");
7695 orig_keyword==KEY_glob ? -OP_GLOB : OP_GLOB,
7703 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7705 pl_yylval.ival = CopLINE(PL_curcop);
7709 LOP(OP_INDEX,XTERM);
7715 LOP(OP_IOCTL,XTERM);
7743 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7748 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7762 LOP(OP_LISTEN,XTERM);
7771 s = scan_pat(s,OP_MATCH);
7772 TERM(sublex_start());
7775 LOP(OP_MAPSTART, XREF);
7778 LOP(OP_MKDIR,XTERM);
7781 LOP(OP_MSGCTL,XTERM);
7784 LOP(OP_MSGGET,XTERM);
7787 LOP(OP_MSGRCV,XTERM);
7790 LOP(OP_MSGSND,XTERM);
7796 yyerror(Perl_form(aTHX_
7797 "Can't redeclare \"%s\" in \"%s\"",
7798 tmp == KEY_my ? "my" :
7799 tmp == KEY_state ? "state" : "our",
7800 PL_in_my == KEY_my ? "my" :
7801 PL_in_my == KEY_state ? "state" : "our"));
7803 PL_in_my = (U16)tmp;
7805 if (isIDFIRST_lazy_if(s,UTF)) {
7806 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7807 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7809 if (!FEATURE_LEXSUBS_IS_ENABLED)
7811 "Experimental \"%s\" subs not enabled",
7812 tmp == KEY_my ? "my" :
7813 tmp == KEY_state ? "state" : "our");
7814 Perl_ck_warner_d(aTHX_
7815 packWARN(WARN_EXPERIMENTAL__LEXICAL_SUBS),
7816 "The lexical_subs feature is experimental");
7819 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7820 if (!PL_in_my_stash) {
7824 len = my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7825 PERL_MY_SNPRINTF_POST_GUARD(len, sizeof(tmpbuf));
7826 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7836 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7841 s = tokenize_use(0, s);
7845 if (*s == '(' || (s = skipspace(s), *s == '('))
7848 if (!PL_lex_allbrackets
7849 && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7851 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7858 if (isIDFIRST_lazy_if(s,UTF)) {
7860 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE,
7862 for (t=d; isSPACE(*t);)
7864 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7866 && !(t[0] == '=' && t[1] == '>')
7867 && !(t[0] == ':' && t[1] == ':')
7868 && !keyword(s, d-s, 0)
7870 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7871 "Precedence problem: open %"UTF8f" should be open(%"UTF8f")",
7872 UTF8fARG(UTF, d-s, s), UTF8fARG(UTF, d-s, s));
7878 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7880 pl_yylval.ival = OP_OR;
7890 LOP(OP_OPEN_DIR,XTERM);
7893 checkcomma(s,PL_tokenbuf,"filehandle");
7897 checkcomma(s,PL_tokenbuf,"filehandle");
7916 s = force_word(s,WORD,FALSE,TRUE);
7918 s = force_strict_version(s);
7922 LOP(OP_PIPE_OP,XTERM);
7925 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7928 COPLINE_SET_FROM_MULTI_END;
7929 pl_yylval.ival = OP_CONST;
7930 TERM(sublex_start());
7937 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7940 COPLINE_SET_FROM_MULTI_END;
7941 PL_expect = XOPERATOR;
7942 if (SvCUR(PL_lex_stuff)) {
7943 int warned_comma = !ckWARN(WARN_QW);
7944 int warned_comment = warned_comma;
7945 d = SvPV_force(PL_lex_stuff, len);
7947 for (; isSPACE(*d) && len; --len, ++d)
7952 if (!warned_comma || !warned_comment) {
7953 for (; !isSPACE(*d) && len; --len, ++d) {
7954 if (!warned_comma && *d == ',') {
7955 Perl_warner(aTHX_ packWARN(WARN_QW),
7956 "Possible attempt to separate words with commas");
7959 else if (!warned_comment && *d == '#') {
7960 Perl_warner(aTHX_ packWARN(WARN_QW),
7961 "Possible attempt to put comments in qw() list");
7967 for (; !isSPACE(*d) && len; --len, ++d)
7970 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7971 words = op_append_elem(OP_LIST, words,
7972 newSVOP(OP_CONST, 0, tokeq(sv)));
7977 words = newNULLLIST();
7978 SvREFCNT_dec_NN(PL_lex_stuff);
7979 PL_lex_stuff = NULL;
7980 PL_expect = XOPERATOR;
7981 pl_yylval.opval = sawparens(words);
7986 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
7989 pl_yylval.ival = OP_STRINGIFY;
7990 if (SvIVX(PL_lex_stuff) == '\'')
7991 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7992 TERM(sublex_start());
7995 s = scan_pat(s,OP_QR);
7996 TERM(sublex_start());
7999 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8002 pl_yylval.ival = OP_BACKTICK;
8003 TERM(sublex_start());
8011 s = force_version(s, FALSE);
8013 else if (*s != 'v' || !isDIGIT(s[1])
8014 || (s = force_version(s, TRUE), *s == 'v'))
8016 *PL_tokenbuf = '\0';
8017 s = force_word(s,WORD,TRUE,TRUE);
8018 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8019 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8020 GV_ADD | (UTF ? SVf_UTF8 : 0));
8022 yyerror("<> at require-statement should be quotes");
8024 if (orig_keyword == KEY_require) {
8030 PL_expect = PL_nexttoke ? XOPERATOR : XTERM;
8032 PL_last_uni = PL_oldbufptr;
8033 PL_last_lop_op = OP_REQUIRE;
8035 return REPORT( (int)REQUIRE );
8044 LOP(OP_RENAME,XTERM);
8053 LOP(OP_RINDEX,XTERM);
8062 UNIDOR(OP_READLINE);
8065 UNIDOR(OP_BACKTICK);
8074 LOP(OP_REVERSE,XTERM);
8077 UNIDOR(OP_READLINK);
8084 if (pl_yylval.opval)
8085 TERM(sublex_start());
8087 TOKEN(1); /* force error */
8090 checkcomma(s,PL_tokenbuf,"filehandle");
8100 LOP(OP_SELECT,XTERM);
8106 LOP(OP_SEMCTL,XTERM);
8109 LOP(OP_SEMGET,XTERM);
8112 LOP(OP_SEMOP,XTERM);
8118 LOP(OP_SETPGRP,XTERM);
8120 case KEY_setpriority:
8121 LOP(OP_SETPRIORITY,XTERM);
8123 case KEY_sethostent:
8129 case KEY_setservent:
8132 case KEY_setprotoent:
8142 LOP(OP_SEEKDIR,XTERM);
8144 case KEY_setsockopt:
8145 LOP(OP_SSOCKOPT,XTERM);
8151 LOP(OP_SHMCTL,XTERM);
8154 LOP(OP_SHMGET,XTERM);
8157 LOP(OP_SHMREAD,XTERM);
8160 LOP(OP_SHMWRITE,XTERM);
8163 LOP(OP_SHUTDOWN,XTERM);
8172 LOP(OP_SOCKET,XTERM);
8174 case KEY_socketpair:
8175 LOP(OP_SOCKPAIR,XTERM);
8178 checkcomma(s,PL_tokenbuf,"subroutine name");
8181 s = force_word(s,WORD,TRUE,TRUE);
8185 LOP(OP_SPLIT,XTERM);
8188 LOP(OP_SPRINTF,XTERM);
8191 LOP(OP_SPLICE,XTERM);
8206 LOP(OP_SUBSTR,XTERM);
8212 char * const tmpbuf = PL_tokenbuf + 1;
8213 expectation attrful;
8214 bool have_name, have_proto;
8215 const int key = tmp;
8216 SV *format_name = NULL;
8221 if (isIDFIRST_lazy_if(s,UTF)
8223 || (*s == ':' && s[1] == ':'))
8227 attrful = XATTRBLOCK;
8228 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8230 if (key == KEY_format)
8231 format_name = S_newSV_maybe_utf8(aTHX_ s, d - s);
8233 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8235 PL_tokenbuf, len + 1, 0
8237 sv_setpvn(PL_subname, tmpbuf, len);
8239 sv_setsv(PL_subname,PL_curstname);
8240 sv_catpvs(PL_subname,"::");
8241 sv_catpvn(PL_subname,tmpbuf,len);
8243 if (SvUTF8(PL_linestr))
8244 SvUTF8_on(PL_subname);
8251 if (key == KEY_my || key == KEY_our || key==KEY_state)
8254 /* diag_listed_as: Missing name in "%s sub" */
8256 "Missing name in \"%s\"", PL_bufptr);
8258 PL_expect = XTERMBLOCK;
8259 attrful = XATTRTERM;
8260 sv_setpvs(PL_subname,"?");
8264 if (key == KEY_format) {
8266 NEXTVAL_NEXTTOKE.opval
8267 = (OP*)newSVOP(OP_CONST,0, format_name);
8268 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
8274 /* Look for a prototype */
8275 if (*s == '(' && !FEATURE_SIGNATURES_IS_ENABLED) {
8276 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
8277 COPLINE_SET_FROM_MULTI_END;
8279 Perl_croak(aTHX_ "Prototype not terminated");
8280 (void)validate_proto(PL_subname, PL_lex_stuff, ckWARN(WARN_ILLEGALPROTO));
8288 if (*s == ':' && s[1] != ':')
8289 PL_expect = attrful;
8290 else if ((*s != '{' && *s != '(') && key != KEY_format) {
8291 assert(key == KEY_sub || key == KEY_AUTOLOAD ||
8292 key == KEY_DESTROY || key == KEY_BEGIN ||
8293 key == KEY_UNITCHECK || key == KEY_CHECK ||
8294 key == KEY_INIT || key == KEY_END ||
8295 key == KEY_my || key == KEY_state ||
8298 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8299 else if (*s != ';' && *s != '}')
8300 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8304 NEXTVAL_NEXTTOKE.opval =
8305 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8306 PL_lex_stuff = NULL;
8311 sv_setpvs(PL_subname, "__ANON__");
8313 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8316 force_ident_maybe_lex('&');
8321 LOP(OP_SYSTEM,XREF);
8324 LOP(OP_SYMLINK,XTERM);
8327 LOP(OP_SYSCALL,XTERM);
8330 LOP(OP_SYSOPEN,XTERM);
8333 LOP(OP_SYSSEEK,XTERM);
8336 LOP(OP_SYSREAD,XTERM);
8339 LOP(OP_SYSWRITE,XTERM);
8344 TERM(sublex_start());
8365 LOP(OP_TRUNCATE,XTERM);
8377 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8379 pl_yylval.ival = CopLINE(PL_curcop);
8383 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8385 pl_yylval.ival = CopLINE(PL_curcop);
8389 LOP(OP_UNLINK,XTERM);
8395 LOP(OP_UNPACK,XTERM);
8398 LOP(OP_UTIME,XTERM);
8404 LOP(OP_UNSHIFT,XTERM);
8407 s = tokenize_use(1, s);
8417 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8419 pl_yylval.ival = CopLINE(PL_curcop);
8420 Perl_ck_warner_d(aTHX_
8421 packWARN(WARN_EXPERIMENTAL__SMARTMATCH),
8422 "when is experimental");
8426 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8428 pl_yylval.ival = CopLINE(PL_curcop);
8432 PL_hints |= HINT_BLOCK_SCOPE;
8439 LOP(OP_WAITPID,XTERM);
8445 /* Make sure $^L is defined. 0x0C is CTRL-L on ASCII platforms, and
8446 * we use the same number on EBCDIC */
8447 gv_fetchpvs("\x0C", GV_ADD|GV_NOTQUAL, SVt_PV);
8451 if (PL_expect == XOPERATOR) {
8452 if (*s == '=' && !PL_lex_allbrackets
8453 && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8463 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8465 pl_yylval.ival = OP_XOR;
8474 Looks up an identifier in the pad or in a package
8477 PRIVATEREF if this is a lexical name.
8478 WORD if this belongs to a package.
8481 if we're in a my declaration
8482 croak if they tried to say my($foo::bar)
8483 build the ops for a my() declaration
8484 if it's an access to a my() variable
8485 build ops for access to a my() variable
8486 if in a dq string, and they've said @foo and we can't find @foo
8488 build ops for a bareword
8492 S_pending_ident(pTHX)
8495 const char pit = (char)pl_yylval.ival;
8496 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8497 /* All routes through this function want to know if there is a colon. */
8498 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8500 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8501 "### Pending identifier '%s'\n", PL_tokenbuf); });
8503 /* if we're in a my(), we can't allow dynamics here.
8504 $foo'bar has already been turned into $foo::bar, so
8505 just check for colons.
8507 if it's a legal name, the OP is a PADANY.
8510 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8512 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8513 "variable %s in \"our\"",
8514 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8515 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8519 /* "my" variable %s can't be in a package */
8520 /* PL_no_myglob is constant */
8521 GCC_DIAG_IGNORE(-Wformat-nonliteral);
8522 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8523 PL_in_my == KEY_my ? "my" : "state",
8524 *PL_tokenbuf == '&' ? "subroutin" : "variabl",
8526 UTF ? SVf_UTF8 : 0);
8530 pl_yylval.opval = newOP(OP_PADANY, 0);
8531 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8532 UTF ? SVf_UTF8 : 0);
8538 build the ops for accesses to a my() variable.
8543 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8545 if (tmp != NOT_IN_PAD) {
8546 /* might be an "our" variable" */
8547 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8548 /* build ops for a bareword */
8549 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8550 HEK * const stashname = HvNAME_HEK(stash);
8551 SV * const sym = newSVhek(stashname);
8552 sv_catpvs(sym, "::");
8553 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8554 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8555 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8559 ((PL_tokenbuf[0] == '$') ? SVt_PV
8560 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8565 pl_yylval.opval = newOP(OP_PADANY, 0);
8566 pl_yylval.opval->op_targ = tmp;
8572 Whine if they've said @foo in a doublequoted string,
8573 and @foo isn't a variable we can find in the symbol
8576 if (ckWARN(WARN_AMBIGUOUS)
8578 && PL_lex_state != LEX_NORMAL
8579 && !PL_lex_brackets)
8581 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8582 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8583 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8584 /* DO NOT warn for @- and @+ */
8585 && !( PL_tokenbuf[2] == '\0'
8586 && ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8589 /* Downgraded from fatal to warning 20000522 mjd */
8590 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8591 "Possible unintended interpolation of %"UTF8f
8593 UTF8fARG(UTF, tokenbuf_len, PL_tokenbuf));
8597 /* build ops for a bareword */
8598 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8599 newSVpvn_flags(PL_tokenbuf + 1,
8601 UTF ? SVf_UTF8 : 0 ));
8602 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8604 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8605 (PL_in_eval ? GV_ADDMULTI : GV_ADD)
8606 | ( UTF ? SVf_UTF8 : 0 ),
8607 ((PL_tokenbuf[0] == '$') ? SVt_PV
8608 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8614 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8616 PERL_ARGS_ASSERT_CHECKCOMMA;
8618 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8619 if (ckWARN(WARN_SYNTAX)) {
8622 for (w = s+2; *w && level; w++) {
8630 /* the list of chars below is for end of statements or
8631 * block / parens, boolean operators (&&, ||, //) and branch
8632 * constructs (or, and, if, until, unless, while, err, for).
8633 * Not a very solid hack... */
8634 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8635 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8636 "%s (...) interpreted as function",name);
8639 while (s < PL_bufend && isSPACE(*s))
8643 while (s < PL_bufend && isSPACE(*s))
8645 if (isIDFIRST_lazy_if(s,UTF)) {
8646 const char * const w = s;
8647 s += UTF ? UTF8SKIP(s) : 1;
8648 while (isWORDCHAR_lazy_if(s,UTF))
8649 s += UTF ? UTF8SKIP(s) : 1;
8650 while (s < PL_bufend && isSPACE(*s))
8655 if (keyword(w, s - w, 0))
8658 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8659 if (gv && GvCVu(gv))
8663 Copy(w, tmpbuf+1, s - w, char);
8665 off = pad_findmy_pvn(tmpbuf, s-w+1, 0);
8666 if (off != NOT_IN_PAD) return;
8668 Perl_croak(aTHX_ "No comma allowed after %s", what);
8673 /* S_new_constant(): do any overload::constant lookup.
8675 Either returns sv, or mortalizes/frees sv and returns a new SV*.
8676 Best used as sv=new_constant(..., sv, ...).
8677 If s, pv are NULL, calls subroutine with one argument,
8678 and <type> is used with error messages only.
8679 <type> is assumed to be well formed UTF-8 */
8682 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8683 SV *sv, SV *pv, const char *type, STRLEN typelen)
8686 HV * table = GvHV(PL_hintgv); /* ^H */
8691 const char *why1 = "", *why2 = "", *why3 = "";
8693 PERL_ARGS_ASSERT_NEW_CONSTANT;
8694 /* We assume that this is true: */
8695 if (*key == 'c') { assert (strEQ(key, "charnames")); }
8698 /* charnames doesn't work well if there have been errors found */
8699 if (PL_error_count > 0 && *key == 'c')
8701 SvREFCNT_dec_NN(sv);
8702 return &PL_sv_undef;
8705 sv_2mortal(sv); /* Parent created it permanently */
8707 || ! (PL_hints & HINT_LOCALIZE_HH)
8708 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8713 /* Here haven't found what we're looking for. If it is charnames,
8714 * perhaps it needs to be loaded. Try doing that before giving up */
8716 Perl_load_module(aTHX_
8718 newSVpvs("_charnames"),
8719 /* version parameter; no need to specify it, as if
8720 * we get too early a version, will fail anyway,
8721 * not being able to find '_charnames' */
8726 assert(sp == PL_stack_sp);
8727 table = GvHV(PL_hintgv);
8729 && (PL_hints & HINT_LOCALIZE_HH)
8730 && (cvp = hv_fetch(table, key, keylen, FALSE))
8736 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8737 msg = Perl_form(aTHX_
8738 "Constant(%.*s) unknown",
8739 (int)(type ? typelen : len),
8745 why3 = "} is not defined";
8748 msg = Perl_form(aTHX_
8749 /* The +3 is for '\N{'; -4 for that, plus '}' */
8750 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
8754 msg = Perl_form(aTHX_ "Constant(%.*s): %s%s%s",
8755 (int)(type ? typelen : len),
8756 (type ? type: s), why1, why2, why3);
8759 yyerror_pv(msg, UTF ? SVf_UTF8 : 0);
8760 return SvREFCNT_inc_simple_NN(sv);
8765 pv = newSVpvn_flags(s, len, SVs_TEMP);
8767 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8769 typesv = &PL_sv_undef;
8771 PUSHSTACKi(PERLSI_OVERLOAD);
8783 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8787 /* Check the eval first */
8788 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
8790 const char * errstr;
8791 sv_catpvs(errsv, "Propagated");
8792 errstr = SvPV_const(errsv, errlen);
8793 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
8795 res = SvREFCNT_inc_simple_NN(sv);
8799 SvREFCNT_inc_simple_void_NN(res);
8808 why1 = "Call to &{$^H{";
8810 why3 = "}} did not return a defined value";
8812 (void)sv_2mortal(sv);
8819 PERL_STATIC_INLINE void
8820 S_parse_ident(pTHX_ char **s, char **d, char * const e, int allow_package, bool is_utf8) {
8821 PERL_ARGS_ASSERT_PARSE_IDENT;
8825 Perl_croak(aTHX_ "%s", ident_too_long);
8826 if (is_utf8 && isIDFIRST_utf8((U8*)*s)) {
8827 /* The UTF-8 case must come first, otherwise things
8828 * like c\N{COMBINING TILDE} would start failing, as the
8829 * isWORDCHAR_A case below would gobble the 'c' up.
8832 char *t = *s + UTF8SKIP(*s);
8833 while (isIDCONT_utf8((U8*)t))
8835 if (*d + (t - *s) > e)
8836 Perl_croak(aTHX_ "%s", ident_too_long);
8837 Copy(*s, *d, t - *s, char);
8841 else if ( isWORDCHAR_A(**s) ) {
8844 } while (isWORDCHAR_A(**s) && *d < e);
8846 else if (allow_package && **s == '\'' && isIDFIRST_lazy_if(*s+1,is_utf8)) {
8851 else if (allow_package && **s == ':' && (*s)[1] == ':'
8852 /* Disallow things like Foo::$bar. For the curious, this is
8853 * the code path that triggers the "Bad name after" warning
8854 * when looking for barewords.
8856 && (*s)[2] != '$') {
8866 /* Returns a NUL terminated string, with the length of the string written to
8870 S_scan_word(pTHX_ char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8873 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8874 bool is_utf8 = cBOOL(UTF);
8876 PERL_ARGS_ASSERT_SCAN_WORD;
8878 parse_ident(&s, &d, e, allow_package, is_utf8);
8884 /* Is the byte 'd' a legal single character identifier name? 'u' is true
8885 * iff Unicode semantics are to be used. The legal ones are any of:
8886 * a) all ASCII characters except:
8887 * 1) control and space-type ones, like NUL, SOH, \t, and SPACE;
8889 * The final case currently doesn't get this far in the program, so we
8890 * don't test for it. If that were to change, it would be ok to allow it.
8891 * c) When not under Unicode rules, any upper Latin1 character
8892 * d) Otherwise, when unicode rules are used, all XIDS characters.
8894 * Because all ASCII characters have the same representation whether
8895 * encoded in UTF-8 or not, we can use the foo_A macros below and '\0' and
8896 * '{' without knowing if is UTF-8 or not.
8897 * EBCDIC already uses the rules that ASCII platforms will use after the
8898 * deprecation cycle; see comment below about the deprecation. */
8900 # define VALID_LEN_ONE_IDENT(s, is_utf8) \
8901 (isGRAPH_A(*(s)) || ((is_utf8) \
8902 ? isIDFIRST_utf8((U8*) (s)) \
8904 && LIKELY((U8) *(s) != LATIN1_TO_NATIVE(0xAD)))))
8906 # define VALID_LEN_ONE_IDENT(s, is_utf8) \
8907 (isGRAPH_A(*(s)) || ((is_utf8) \
8908 ? isIDFIRST_utf8((U8*) (s)) \
8909 : ! isASCII_utf8((U8*) (s))))
8913 S_scan_ident(pTHX_ char *s, char *dest, STRLEN destlen, I32 ck_uni)
8915 I32 herelines = PL_parser->herelines;
8916 SSize_t bracket = -1;
8919 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8920 bool is_utf8 = cBOOL(UTF);
8921 I32 orig_copline = 0, tmp_copline = 0;
8923 PERL_ARGS_ASSERT_SCAN_IDENT;
8925 if (isSPACE(*s) || !*s)
8928 while (isDIGIT(*s)) {
8930 Perl_croak(aTHX_ "%s", ident_too_long);
8934 else { /* See if it is a "normal" identifier */
8935 parse_ident(&s, &d, e, 1, is_utf8);
8940 /* Either a digit variable, or parse_ident() found an identifier
8941 (anything valid as a bareword), so job done and return. */
8942 if (PL_lex_state != LEX_NORMAL)
8943 PL_lex_state = LEX_INTERPENDMAYBE;
8947 /* Here, it is not a run-of-the-mill identifier name */
8949 if (*s == '$' && s[1]
8950 && (isIDFIRST_lazy_if(s+1,is_utf8)
8951 || isDIGIT_A((U8)s[1])
8954 || strnEQ(s+1,"::",2)) )
8956 /* Dereferencing a value in a scalar variable.
8957 The alternatives are different syntaxes for a scalar variable.
8958 Using ' as a leading package separator isn't allowed. :: is. */
8961 /* Handle the opening { of @{...}, &{...}, *{...}, %{...}, ${...} */
8963 bracket = s - SvPVX(PL_linestr);
8965 orig_copline = CopLINE(PL_curcop);
8966 if (s < PL_bufend && isSPACE(*s)) {
8970 if ((s <= PL_bufend - (is_utf8)
8973 && VALID_LEN_ONE_IDENT(s, is_utf8))
8975 /* Deprecate all non-graphic characters. Include SHY as a non-graphic,
8976 * because often it has no graphic representation. (We can't get to
8977 * here with SHY when 'is_utf8' is true, so no need to include a UTF-8
8980 ? ! isGRAPH_utf8( (U8*) s)
8981 : (! isGRAPH_L1( (U8) *s)
8982 || UNLIKELY((U8) *(s) == LATIN1_TO_NATIVE(0xAD))))
8984 deprecate("literal non-graphic characters in variable names");
8988 const STRLEN skip = UTF8SKIP(s);
8991 for ( i = 0; i < skip; i++ )
8999 /* Convert $^F, ${^F} and the ^F of ${^FOO} to control characters */
9000 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9004 /* Warn about ambiguous code after unary operators if {...} notation isn't
9005 used. There's no difference in ambiguity; it's merely a heuristic
9006 about when not to warn. */
9007 else if (ck_uni && bracket == -1)
9009 if (bracket != -1) {
9010 /* If we were processing {...} notation then... */
9011 if (isIDFIRST_lazy_if(d,is_utf8)) {
9012 /* if it starts as a valid identifier, assume that it is one.
9013 (the later check for } being at the expected point will trap
9014 cases where this doesn't pan out.) */
9015 d += is_utf8 ? UTF8SKIP(d) : 1;
9016 parse_ident(&s, &d, e, 1, is_utf8);
9018 tmp_copline = CopLINE(PL_curcop);
9019 if (s < PL_bufend && isSPACE(*s)) {
9022 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9023 /* ${foo[0]} and ${foo{bar}} notation. */
9024 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9025 const char * const brack =
9027 ((*s == '[') ? "[...]" : "{...}");
9028 orig_copline = CopLINE(PL_curcop);
9029 CopLINE_set(PL_curcop, tmp_copline);
9030 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9031 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9032 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9033 funny, dest, brack, funny, dest, brack);
9034 CopLINE_set(PL_curcop, orig_copline);
9037 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9038 PL_lex_allbrackets++;
9042 /* Handle extended ${^Foo} variables
9043 * 1999-02-27 mjd-perl-patch@plover.com */
9044 else if (! isPRINT(*d) /* isCNTRL(d), plus all non-ASCII */
9048 while (isWORDCHAR(*s) && d < e) {
9052 Perl_croak(aTHX_ "%s", ident_too_long);
9057 tmp_copline = CopLINE(PL_curcop);
9058 if (s < PL_bufend && isSPACE(*s)) {
9062 /* Expect to find a closing } after consuming any trailing whitespace.
9066 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9067 PL_lex_state = LEX_INTERPEND;
9070 if (PL_lex_state == LEX_NORMAL) {
9071 if (ckWARN(WARN_AMBIGUOUS)
9072 && (keyword(dest, d - dest, 0)
9073 || get_cvn_flags(dest, d - dest, is_utf8
9077 SV *tmp = newSVpvn_flags( dest, d - dest,
9078 SVs_TEMP | (is_utf8 ? SVf_UTF8 : 0) );
9081 orig_copline = CopLINE(PL_curcop);
9082 CopLINE_set(PL_curcop, tmp_copline);
9083 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9084 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9085 funny, SVfARG(tmp), funny, SVfARG(tmp));
9086 CopLINE_set(PL_curcop, orig_copline);
9091 /* Didn't find the closing } at the point we expected, so restore
9092 state such that the next thing to process is the opening { and */
9093 s = SvPVX(PL_linestr) + bracket; /* let the parser handle it */
9094 CopLINE_set(PL_curcop, orig_copline);
9095 PL_parser->herelines = herelines;
9099 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9100 PL_lex_state = LEX_INTERPEND;
9105 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset, unsigned int * x_mod_count) {
9107 /* Adds, subtracts to/from 'pmfl' based on the next regex modifier flag
9108 * found in the parse starting at 's', based on the subset that are valid
9109 * in this context input to this routine in 'valid_flags'. Advances s.
9110 * Returns TRUE if the input should be treated as a valid flag, so the next
9111 * char may be as well; otherwise FALSE. 'charset' should point to a NUL
9112 * upon first call on the current regex. This routine will set it to any
9113 * charset modifier found. The caller shouldn't change it. This way,
9114 * another charset modifier encountered in the parse can be detected as an
9115 * error, as we have decided to allow only one */
9118 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9120 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9121 if (isWORDCHAR_lazy_if(*s, UTF)) {
9122 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9123 UTF ? SVf_UTF8 : 0);
9125 /* Pretend that it worked, so will continue processing before
9134 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, *x_mod_count);
9135 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9136 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9137 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9138 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9139 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9140 case LOCALE_PAT_MOD:
9142 goto multiple_charsets;
9144 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9147 case UNICODE_PAT_MOD:
9149 goto multiple_charsets;
9151 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9154 case ASCII_RESTRICT_PAT_MOD:
9156 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9160 /* Error if previous modifier wasn't an 'a', but if it was, see
9161 * if, and accept, a second occurrence (only) */
9163 || get_regex_charset(*pmfl)
9164 != REGEX_ASCII_RESTRICTED_CHARSET)
9166 goto multiple_charsets;
9168 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9172 case DEPENDS_PAT_MOD:
9174 goto multiple_charsets;
9176 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9185 if (*charset != c) {
9186 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9188 else if (c == 'a') {
9189 /* diag_listed_as: Regexp modifier "/%c" may appear a maximum of twice */
9190 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9193 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9196 /* Pretend that it worked, so will continue processing before dieing */
9202 S_scan_pat(pTHX_ char *start, I32 type)
9206 const char * const valid_flags =
9207 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9208 char charset = '\0'; /* character set modifier */
9209 unsigned int x_mod_count = 0;
9211 PERL_ARGS_ASSERT_SCAN_PAT;
9213 s = scan_str(start,TRUE,FALSE, (PL_in_eval & EVAL_RE_REPARSING), NULL);
9215 Perl_croak(aTHX_ "Search pattern not terminated");
9217 pm = (PMOP*)newPMOP(type, 0);
9218 if (PL_multi_open == '?') {
9219 /* This is the only point in the code that sets PMf_ONCE: */
9220 pm->op_pmflags |= PMf_ONCE;
9222 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9223 allows us to restrict the list needed by reset to just the ??
9225 assert(type != OP_TRANS);
9227 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9230 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9233 elements = mg->mg_len / sizeof(PMOP**);
9234 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9235 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9236 mg->mg_len = elements * sizeof(PMOP**);
9237 PmopSTASH_set(pm,PL_curstash);
9241 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9242 * anon CV. False positives like qr/[(?{]/ are harmless */
9244 if (type == OP_QR) {
9246 char *e, *p = SvPV(PL_lex_stuff, len);
9248 for (; p < e; p++) {
9249 if (p[0] == '(' && p[1] == '?'
9250 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9252 pm->op_pmflags |= PMf_HAS_CV;
9256 pm->op_pmflags |= PMf_IS_QR;
9259 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags),
9260 &s, &charset, &x_mod_count))
9262 /* issue a warning if /c is specified,but /g is not */
9263 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9265 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9266 "Use of /c modifier is meaningless without /g" );
9269 STD_PMMOD_FLAGS_PARSE_X_WARN(x_mod_count);
9271 PL_lex_op = (OP*)pm;
9272 pl_yylval.ival = OP_MATCH;
9277 S_scan_subst(pTHX_ char *start)
9284 char charset = '\0'; /* character set modifier */
9285 unsigned int x_mod_count = 0;
9288 PERL_ARGS_ASSERT_SCAN_SUBST;
9290 pl_yylval.ival = OP_NULL;
9292 s = scan_str(start, TRUE, FALSE, FALSE, &t);
9295 Perl_croak(aTHX_ "Substitution pattern not terminated");
9299 first_start = PL_multi_start;
9300 first_line = CopLINE(PL_curcop);
9301 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
9303 SvREFCNT_dec_NN(PL_lex_stuff);
9304 PL_lex_stuff = NULL;
9305 Perl_croak(aTHX_ "Substitution replacement not terminated");
9307 PL_multi_start = first_start; /* so whole substitution is taken together */
9309 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9313 if (*s == EXEC_PAT_MOD) {
9317 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags),
9318 &s, &charset, &x_mod_count))
9324 STD_PMMOD_FLAGS_PARSE_X_WARN(x_mod_count);
9326 if ((pm->op_pmflags & PMf_CONTINUE)) {
9327 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9331 SV * const repl = newSVpvs("");
9334 pm->op_pmflags |= PMf_EVAL;
9337 sv_catpvs(repl, "eval ");
9339 sv_catpvs(repl, "do ");
9341 sv_catpvs(repl, "{");
9342 sv_catsv(repl, PL_sublex_info.repl);
9343 sv_catpvs(repl, "}");
9345 SvREFCNT_dec(PL_sublex_info.repl);
9346 PL_sublex_info.repl = repl;
9348 if (CopLINE(PL_curcop) != first_line) {
9349 sv_upgrade(PL_sublex_info.repl, SVt_PVNV);
9350 ((XPVNV*)SvANY(PL_sublex_info.repl))->xnv_u.xpad_cop_seq.xlow =
9351 CopLINE(PL_curcop) - first_line;
9352 CopLINE_set(PL_curcop, first_line);
9355 PL_lex_op = (OP*)pm;
9356 pl_yylval.ival = OP_SUBST;
9361 S_scan_trans(pTHX_ char *start)
9368 bool nondestruct = 0;
9371 PERL_ARGS_ASSERT_SCAN_TRANS;
9373 pl_yylval.ival = OP_NULL;
9375 s = scan_str(start,FALSE,FALSE,FALSE,&t);
9377 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9381 s = scan_str(s,FALSE,FALSE,FALSE,NULL);
9383 SvREFCNT_dec_NN(PL_lex_stuff);
9384 PL_lex_stuff = NULL;
9385 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9388 complement = del = squash = 0;
9392 complement = OPpTRANS_COMPLEMENT;
9395 del = OPpTRANS_DELETE;
9398 squash = OPpTRANS_SQUASH;
9410 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9411 o->op_private &= ~OPpTRANS_ALL;
9412 o->op_private |= del|squash|complement|
9413 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9414 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9417 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9424 Takes a pointer to the first < in <<FOO.
9425 Returns a pointer to the byte following <<FOO.
9427 This function scans a heredoc, which involves different methods
9428 depending on whether we are in a string eval, quoted construct, etc.
9429 This is because PL_linestr could containing a single line of input, or
9430 a whole string being evalled, or the contents of the current quote-
9433 The two basic methods are:
9434 - Steal lines from the input stream
9435 - Scan the heredoc in PL_linestr and remove it therefrom
9437 In a file scope or filtered eval, the first method is used; in a
9438 string eval, the second.
9440 In a quote-like operator, we have to choose between the two,
9441 depending on where we can find a newline. We peek into outer lex-
9442 ing scopes until we find one with a newline in it. If we reach the
9443 outermost lexing scope and it is a file, we use the stream method.
9444 Otherwise it is treated as an eval.
9448 S_scan_heredoc(pTHX_ char *s)
9450 I32 op_type = OP_SCALAR;
9457 const bool infile = PL_rsfp || PL_parser->filtered;
9458 const line_t origline = CopLINE(PL_curcop);
9459 LEXSHARED *shared = PL_parser->lex_shared;
9461 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9464 d = PL_tokenbuf + 1;
9465 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9466 *PL_tokenbuf = '\n';
9468 while (SPACE_OR_TAB(*peek))
9470 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9473 s = delimcpy(d, e, s, PL_bufend, term, &len);
9475 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9481 /* <<\FOO is equivalent to <<'FOO' */
9485 if (!isWORDCHAR_lazy_if(s,UTF))
9486 deprecate("bare << to mean <<\"\"");
9488 while (isWORDCHAR_lazy_if(peek,UTF)) {
9489 peek += UTF ? UTF8SKIP(peek) : 1;
9491 len = (peek - s >= e - d) ? (e - d) : (peek - s);
9492 Copy(s, d, len, char);
9496 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9497 Perl_croak(aTHX_ "Delimiter for here document is too long");
9500 len = d - PL_tokenbuf;
9502 #ifndef PERL_STRICT_CR
9503 d = strchr(s, '\r');
9505 char * const olds = s;
9507 while (s < PL_bufend) {
9513 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9522 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9527 tmpstr = newSV_type(SVt_PVIV);
9531 SvIV_set(tmpstr, -1);
9533 else if (term == '`') {
9534 op_type = OP_BACKTICK;
9535 SvIV_set(tmpstr, '\\');
9538 PL_multi_start = origline + 1 + PL_parser->herelines;
9539 PL_multi_open = PL_multi_close = '<';
9540 /* inside a string eval or quote-like operator */
9541 if (!infile || PL_lex_inwhat) {
9544 char * const olds = s;
9545 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9546 /* These two fields are not set until an inner lexing scope is
9547 entered. But we need them set here. */
9548 shared->ls_bufptr = s;
9549 shared->ls_linestr = PL_linestr;
9551 /* Look for a newline. If the current buffer does not have one,
9552 peek into the line buffer of the parent lexing scope, going
9553 up as many levels as necessary to find one with a newline
9556 while (!(s = (char *)memchr(
9557 (void *)shared->ls_bufptr, '\n',
9558 SvEND(shared->ls_linestr)-shared->ls_bufptr
9560 shared = shared->ls_prev;
9561 /* shared is only null if we have gone beyond the outermost
9562 lexing scope. In a file, we will have broken out of the
9563 loop in the previous iteration. In an eval, the string buf-
9564 fer ends with "\n;", so the while condition above will have
9565 evaluated to false. So shared can never be null. Or so you
9566 might think. Odd syntax errors like s;@{<<; can gobble up
9567 the implicit semicolon at the end of a flie, causing the
9568 file handle to be closed even when we are not in a string
9569 eval. So shared may be null in that case. */
9570 if (UNLIKELY(!shared))
9572 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9573 most lexing scope. In a file, shared->ls_linestr at that
9574 level is just one line, so there is no body to steal. */
9575 if (infile && !shared->ls_prev) {
9581 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9584 linestr = shared->ls_linestr;
9585 bufend = SvEND(linestr);
9587 while (s < bufend - len + 1
9588 && memNE(s,PL_tokenbuf,len) )
9591 ++PL_parser->herelines;
9593 if (s >= bufend - len + 1) {
9596 sv_setpvn(tmpstr,d+1,s-d);
9598 /* the preceding stmt passes a newline */
9599 PL_parser->herelines++;
9601 /* s now points to the newline after the heredoc terminator.
9602 d points to the newline before the body of the heredoc.
9605 /* We are going to modify linestr in place here, so set
9606 aside copies of the string if necessary for re-evals or
9608 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9609 check shared->re_eval_str. */
9610 if (shared->re_eval_start || shared->re_eval_str) {
9611 /* Set aside the rest of the regexp */
9612 if (!shared->re_eval_str)
9613 shared->re_eval_str =
9614 newSVpvn(shared->re_eval_start,
9615 bufend - shared->re_eval_start);
9616 shared->re_eval_start -= s-d;
9619 && CxTYPE(cx) == CXt_EVAL
9620 && CxOLD_OP_TYPE(cx) == OP_ENTEREVAL
9621 && cx->blk_eval.cur_text == linestr)
9623 cx->blk_eval.cur_text = newSVsv(linestr);
9624 SvSCREAM_on(cx->blk_eval.cur_text);
9626 /* Copy everything from s onwards back to d. */
9627 Move(s,d,bufend-s + 1,char);
9628 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9629 /* Setting PL_bufend only applies when we have not dug deeper
9630 into other scopes, because sublex_done sets PL_bufend to
9631 SvEND(PL_linestr). */
9632 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9638 char *oldbufptr_save;
9640 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9641 term = PL_tokenbuf[1];
9643 linestr_save = PL_linestr; /* must restore this afterwards */
9644 d = s; /* and this */
9645 oldbufptr_save = PL_oldbufptr;
9646 PL_linestr = newSVpvs("");
9647 PL_bufend = SvPVX(PL_linestr);
9649 PL_bufptr = PL_bufend;
9650 CopLINE_set(PL_curcop,
9651 origline + 1 + PL_parser->herelines);
9652 if (!lex_next_chunk(LEX_NO_TERM)
9653 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9654 /* Simply freeing linestr_save might seem simpler here, as it
9655 does not matter what PL_linestr points to, since we are
9656 about to croak; but in a quote-like op, linestr_save
9657 will have been prospectively freed already, via
9658 SAVEFREESV(PL_linestr) in sublex_push, so it’s easier to
9659 restore PL_linestr. */
9660 SvREFCNT_dec_NN(PL_linestr);
9661 PL_linestr = linestr_save;
9662 PL_oldbufptr = oldbufptr_save;
9665 CopLINE_set(PL_curcop, origline);
9666 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9667 s = lex_grow_linestr(SvLEN(PL_linestr) + 3);
9668 /* ^That should be enough to avoid this needing to grow: */
9669 sv_catpvs(PL_linestr, "\n\0");
9670 assert(s == SvPVX(PL_linestr));
9671 PL_bufend = SvEND(PL_linestr);
9674 PL_parser->herelines++;
9675 PL_last_lop = PL_last_uni = NULL;
9676 #ifndef PERL_STRICT_CR
9677 if (PL_bufend - PL_linestart >= 2) {
9678 if ( (PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n')
9679 || (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9681 PL_bufend[-2] = '\n';
9683 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9685 else if (PL_bufend[-1] == '\r')
9686 PL_bufend[-1] = '\n';
9688 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9689 PL_bufend[-1] = '\n';
9691 if (*s == term && PL_bufend-s >= len
9692 && memEQ(s,PL_tokenbuf + 1,len)) {
9693 SvREFCNT_dec(PL_linestr);
9694 PL_linestr = linestr_save;
9695 PL_linestart = SvPVX(linestr_save);
9696 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9697 PL_oldbufptr = oldbufptr_save;
9702 sv_catsv(tmpstr,PL_linestr);
9706 PL_multi_end = origline + PL_parser->herelines;
9707 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9708 SvPV_shrink_to_cur(tmpstr);
9711 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9713 else if (IN_ENCODING)
9714 sv_recode_to_utf8(tmpstr, _get_encoding());
9716 PL_lex_stuff = tmpstr;
9717 pl_yylval.ival = op_type;
9721 SvREFCNT_dec(tmpstr);
9722 CopLINE_set(PL_curcop, origline);
9723 missingterm(PL_tokenbuf + 1);
9727 takes: current position in input buffer
9728 returns: new position in input buffer
9729 side-effects: pl_yylval and lex_op are set.
9734 <<>> read from ARGV without magic open
9735 <FH> read from filehandle
9736 <pkg::FH> read from package qualified filehandle
9737 <pkg'FH> read from package qualified filehandle
9738 <$fh> read from filehandle in $fh
9744 S_scan_inputsymbol(pTHX_ char *start)
9746 char *s = start; /* current position in buffer */
9749 bool nomagicopen = FALSE;
9750 char *d = PL_tokenbuf; /* start of temp holding space */
9751 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9753 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9755 end = strchr(s, '\n');
9758 if (s[1] == '<' && s[2] == '>' && s[3] == '>') {
9765 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9767 /* die if we didn't have space for the contents of the <>,
9768 or if it didn't end, or if we see a newline
9771 if (len >= (I32)sizeof PL_tokenbuf)
9772 Perl_croak(aTHX_ "Excessively long <> operator");
9774 Perl_croak(aTHX_ "Unterminated <> operator");
9779 Remember, only scalar variables are interpreted as filehandles by
9780 this code. Anything more complex (e.g., <$fh{$num}>) will be
9781 treated as a glob() call.
9782 This code makes use of the fact that except for the $ at the front,
9783 a scalar variable and a filehandle look the same.
9785 if (*d == '$' && d[1]) d++;
9787 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9788 while (*d && (isWORDCHAR_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9789 d += UTF ? UTF8SKIP(d) : 1;
9791 /* If we've tried to read what we allow filehandles to look like, and
9792 there's still text left, then it must be a glob() and not a getline.
9793 Use scan_str to pull out the stuff between the <> and treat it
9794 as nothing more than a string.
9797 if (d - PL_tokenbuf != len) {
9798 pl_yylval.ival = OP_GLOB;
9799 s = scan_str(start,FALSE,FALSE,FALSE,NULL);
9801 Perl_croak(aTHX_ "Glob not terminated");
9805 bool readline_overriden = FALSE;
9807 /* we're in a filehandle read situation */
9810 /* turn <> into <ARGV> */
9812 Copy("ARGV",d,5,char);
9814 /* Check whether readline() is overriden */
9815 if ((gv_readline = gv_override("readline",8)))
9816 readline_overriden = TRUE;
9818 /* if <$fh>, create the ops to turn the variable into a
9822 /* try to find it in the pad for this block, otherwise find
9823 add symbol table ops
9825 const PADOFFSET tmp = pad_findmy_pvn(d, len, 0);
9826 if (tmp != NOT_IN_PAD) {
9827 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9828 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9829 HEK * const stashname = HvNAME_HEK(stash);
9830 SV * const sym = sv_2mortal(newSVhek(stashname));
9831 sv_catpvs(sym, "::");
9837 OP * const o = newOP(OP_PADSV, 0);
9839 PL_lex_op = readline_overriden
9840 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9841 op_append_elem(OP_LIST, o,
9842 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9843 : (OP*)newUNOP(OP_READLINE, 0, o);
9851 GV_ADDMULTI | ( UTF ? SVf_UTF8 : 0 ),
9853 PL_lex_op = readline_overriden
9854 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9855 op_append_elem(OP_LIST,
9856 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9857 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9858 : (OP*)newUNOP(OP_READLINE, 0,
9859 newUNOP(OP_RV2SV, 0,
9860 newGVOP(OP_GV, 0, gv)));
9862 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9863 pl_yylval.ival = OP_NULL;
9866 /* If it's none of the above, it must be a literal filehandle
9867 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9869 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9870 PL_lex_op = readline_overriden
9871 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9872 op_append_elem(OP_LIST,
9873 newGVOP(OP_GV, 0, gv),
9874 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9875 : (OP*)newUNOP(OP_READLINE, nomagicopen ? OPf_SPECIAL : 0, newGVOP(OP_GV, 0, gv));
9876 pl_yylval.ival = OP_NULL;
9886 start position in buffer
9887 keep_bracketed_quoted preserve \ quoting of embedded delimiters, but
9888 only if they are of the open/close form
9889 keep_delims preserve the delimiters around the string
9890 re_reparse compiling a run-time /(?{})/:
9891 collapse // to /, and skip encoding src
9892 delimp if non-null, this is set to the position of
9893 the closing delimiter, or just after it if
9894 the closing and opening delimiters differ
9895 (i.e., the opening delimiter of a substitu-
9897 returns: position to continue reading from buffer
9898 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9899 updates the read buffer.
9901 This subroutine pulls a string out of the input. It is called for:
9902 q single quotes q(literal text)
9903 ' single quotes 'literal text'
9904 qq double quotes qq(interpolate $here please)
9905 " double quotes "interpolate $here please"
9906 qx backticks qx(/bin/ls -l)
9907 ` backticks `/bin/ls -l`
9908 qw quote words @EXPORT_OK = qw( func() $spam )
9909 m// regexp match m/this/
9910 s/// regexp substitute s/this/that/
9911 tr/// string transliterate tr/this/that/
9912 y/// string transliterate y/this/that/
9913 ($*@) sub prototypes sub foo ($)
9914 (stuff) sub attr parameters sub foo : attr(stuff)
9915 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9917 In most of these cases (all but <>, patterns and transliterate)
9918 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9919 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9920 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9923 It skips whitespace before the string starts, and treats the first
9924 character as the delimiter. If the delimiter is one of ([{< then
9925 the corresponding "close" character )]}> is used as the closing
9926 delimiter. It allows quoting of delimiters, and if the string has
9927 balanced delimiters ([{<>}]) it allows nesting.
9929 On success, the SV with the resulting string is put into lex_stuff or,
9930 if that is already non-NULL, into lex_repl. The second case occurs only
9931 when parsing the RHS of the special constructs s/// and tr/// (y///).
9932 For convenience, the terminating delimiter character is stuffed into
9937 S_scan_str(pTHX_ char *start, int keep_bracketed_quoted, int keep_delims, int re_reparse,
9941 SV *sv; /* scalar value: string */
9942 const char *tmps; /* temp string, used for delimiter matching */
9943 char *s = start; /* current position in the buffer */
9944 char term; /* terminating character */
9945 char *to; /* current position in the sv's data */
9946 I32 brackets = 1; /* bracket nesting level */
9947 bool has_utf8 = FALSE; /* is there any utf8 content? */
9948 I32 termcode; /* terminating char. code */
9949 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9950 STRLEN termlen; /* length of terminating string */
9951 int last_off = 0; /* last position for nesting bracket */
9954 PERL_ARGS_ASSERT_SCAN_STR;
9956 /* skip space before the delimiter */
9961 /* mark where we are, in case we need to report errors */
9964 /* after skipping whitespace, the next character is the terminator */
9967 termcode = termstr[0] = term;
9971 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
9972 Copy(s, termstr, termlen, U8);
9973 if (!UTF8_IS_INVARIANT(term))
9977 /* mark where we are */
9978 PL_multi_start = CopLINE(PL_curcop);
9979 PL_multi_open = term;
9980 herelines = PL_parser->herelines;
9982 /* find corresponding closing delimiter */
9983 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9984 termcode = termstr[0] = term = tmps[5];
9986 PL_multi_close = term;
9988 if (PL_multi_open == PL_multi_close) {
9989 keep_bracketed_quoted = FALSE;
9992 /* create a new SV to hold the contents. 79 is the SV's initial length.
9993 What a random number. */
9994 sv = newSV_type(SVt_PVIV);
9996 SvIV_set(sv, termcode);
9997 (void)SvPOK_only(sv); /* validate pointer */
9999 /* move past delimiter and try to read a complete string */
10001 sv_catpvn(sv, s, termlen);
10004 if (IN_ENCODING && !UTF && !re_reparse) {
10008 int offset = s - SvPVX_const(PL_linestr);
10009 const bool found = sv_cat_decode(sv, _get_encoding(), PL_linestr,
10010 &offset, (char*)termstr, termlen);
10014 if (SvIsCOW(PL_linestr)) {
10015 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos;
10016 STRLEN oldoldbufptr_pos, linestart_pos, last_uni_pos;
10017 STRLEN last_lop_pos, re_eval_start_pos, s_pos;
10018 char *buf = SvPVX(PL_linestr);
10019 bufend_pos = PL_parser->bufend - buf;
10020 bufptr_pos = PL_parser->bufptr - buf;
10021 oldbufptr_pos = PL_parser->oldbufptr - buf;
10022 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
10023 linestart_pos = PL_parser->linestart - buf;
10024 last_uni_pos = PL_parser->last_uni
10025 ? PL_parser->last_uni - buf
10027 last_lop_pos = PL_parser->last_lop
10028 ? PL_parser->last_lop - buf
10030 re_eval_start_pos =
10031 PL_parser->lex_shared->re_eval_start ?
10032 PL_parser->lex_shared->re_eval_start - buf : 0;
10035 sv_force_normal(PL_linestr);
10037 buf = SvPVX(PL_linestr);
10038 PL_parser->bufend = buf + bufend_pos;
10039 PL_parser->bufptr = buf + bufptr_pos;
10040 PL_parser->oldbufptr = buf + oldbufptr_pos;
10041 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
10042 PL_parser->linestart = buf + linestart_pos;
10043 if (PL_parser->last_uni)
10044 PL_parser->last_uni = buf + last_uni_pos;
10045 if (PL_parser->last_lop)
10046 PL_parser->last_lop = buf + last_lop_pos;
10047 if (PL_parser->lex_shared->re_eval_start)
10048 PL_parser->lex_shared->re_eval_start =
10049 buf + re_eval_start_pos;
10052 ns = SvPVX_const(PL_linestr) + offset;
10053 svlast = SvEND(sv) - 1;
10055 for (; s < ns; s++) {
10056 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10057 COPLINE_INC_WITH_HERELINES;
10060 goto read_more_line;
10062 /* handle quoted delimiters */
10063 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10065 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10067 if ((svlast-1 - t) % 2) {
10068 if (!keep_bracketed_quoted) {
10069 *(svlast-1) = term;
10071 SvCUR_set(sv, SvCUR(sv) - 1);
10076 if (PL_multi_open == PL_multi_close) {
10082 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10083 /* At here, all closes are "was quoted" one,
10084 so we don't check PL_multi_close. */
10086 if (!keep_bracketed_quoted && *(t+1) == PL_multi_open)
10091 else if (*t == PL_multi_open)
10099 SvCUR_set(sv, w - SvPVX_const(sv));
10101 last_off = w - SvPVX(sv);
10102 if (--brackets <= 0)
10107 if (!keep_delims) {
10108 SvCUR_set(sv, SvCUR(sv) - 1);
10114 /* extend sv if need be */
10115 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10116 /* set 'to' to the next character in the sv's string */
10117 to = SvPVX(sv)+SvCUR(sv);
10119 /* if open delimiter is the close delimiter read unbridle */
10120 if (PL_multi_open == PL_multi_close) {
10121 for (; s < PL_bufend; s++,to++) {
10122 /* embedded newlines increment the current line number */
10123 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10124 COPLINE_INC_WITH_HERELINES;
10125 /* handle quoted delimiters */
10126 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10127 if (!keep_bracketed_quoted
10129 || (re_reparse && s[1] == '\\'))
10132 else /* any other quotes are simply copied straight through */
10135 /* terminate when run out of buffer (the for() condition), or
10136 have found the terminator */
10137 else if (*s == term) {
10140 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10143 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10149 /* if the terminator isn't the same as the start character (e.g.,
10150 matched brackets), we have to allow more in the quoting, and
10151 be prepared for nested brackets.
10154 /* read until we run out of string, or we find the terminator */
10155 for (; s < PL_bufend; s++,to++) {
10156 /* embedded newlines increment the line count */
10157 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10158 COPLINE_INC_WITH_HERELINES;
10159 /* backslashes can escape the open or closing characters */
10160 if (*s == '\\' && s+1 < PL_bufend) {
10161 if (!keep_bracketed_quoted
10162 && ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10169 /* allow nested opens and closes */
10170 else if (*s == PL_multi_close && --brackets <= 0)
10172 else if (*s == PL_multi_open)
10174 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10179 /* terminate the copied string and update the sv's end-of-string */
10181 SvCUR_set(sv, to - SvPVX_const(sv));
10184 * this next chunk reads more into the buffer if we're not done yet
10188 break; /* handle case where we are done yet :-) */
10190 #ifndef PERL_STRICT_CR
10191 if (to - SvPVX_const(sv) >= 2) {
10192 if ( (to[-2] == '\r' && to[-1] == '\n')
10193 || (to[-2] == '\n' && to[-1] == '\r'))
10197 SvCUR_set(sv, to - SvPVX_const(sv));
10199 else if (to[-1] == '\r')
10202 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10207 /* if we're out of file, or a read fails, bail and reset the current
10208 line marker so we can report where the unterminated string began
10210 COPLINE_INC_WITH_HERELINES;
10211 PL_bufptr = PL_bufend;
10212 if (!lex_next_chunk(0)) {
10214 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10220 /* at this point, we have successfully read the delimited string */
10222 if (!IN_ENCODING || UTF || re_reparse) {
10225 sv_catpvn(sv, s, termlen);
10228 if (has_utf8 || (IN_ENCODING && !re_reparse))
10231 PL_multi_end = CopLINE(PL_curcop);
10232 CopLINE_set(PL_curcop, PL_multi_start);
10233 PL_parser->herelines = herelines;
10235 /* if we allocated too much space, give some back */
10236 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10237 SvLEN_set(sv, SvCUR(sv) + 1);
10238 SvPV_renew(sv, SvLEN(sv));
10241 /* decide whether this is the first or second quoted string we've read
10246 PL_sublex_info.repl = sv;
10249 if (delimp) *delimp = PL_multi_open == PL_multi_close ? s-termlen : s;
10255 takes: pointer to position in buffer
10256 returns: pointer to new position in buffer
10257 side-effects: builds ops for the constant in pl_yylval.op
10259 Read a number in any of the formats that Perl accepts:
10261 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10262 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10263 0b[01](_?[01])* binary integers
10264 0[0-7](_?[0-7])* octal integers
10265 0x[0-9A-Fa-f](_?[0-9A-Fa-f])* hexadecimal integers
10266 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*(?:\.\d*)?p[+-]?[0-9]+ hexadecimal floats
10268 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10271 If it reads a number without a decimal point or an exponent, it will
10272 try converting the number to an integer and see if it can do so
10273 without loss of precision.
10277 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10279 const char *s = start; /* current position in buffer */
10280 char *d; /* destination in temp buffer */
10281 char *e; /* end of temp buffer */
10282 NV nv; /* number read, as a double */
10283 SV *sv = NULL; /* place to put the converted number */
10284 bool floatit; /* boolean: int or float? */
10285 const char *lastub = NULL; /* position of last underbar */
10286 static const char* const number_too_long = "Number too long";
10287 /* Hexadecimal floating point.
10289 * In many places (where we have quads and NV is IEEE 754 double)
10290 * we can fit the mantissa bits of a NV into an unsigned quad.
10291 * (Note that UVs might not be quads even when we have quads.)
10292 * This will not work everywhere, though (either no quads, or
10293 * using long doubles), in which case we have to resort to NV,
10294 * which will probably mean horrible loss of precision due to
10295 * multiple fp operations. */
10296 bool hexfp = FALSE;
10297 int total_bits = 0;
10298 #if NVSIZE == 8 && defined(HAS_QUAD) && defined(Uquad_t)
10299 # define HEXFP_UQUAD
10300 Uquad_t hexfp_uquad = 0;
10301 int hexfp_frac_bits = 0;
10306 NV hexfp_mult = 1.0;
10307 UV high_non_zero = 0; /* highest digit */
10309 PERL_ARGS_ASSERT_SCAN_NUM;
10311 /* We use the first character to decide what type of number this is */
10315 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10317 /* if it starts with a 0, it could be an octal number, a decimal in
10318 0.13 disguise, or a hexadecimal number, or a binary number. */
10322 u holds the "number so far"
10323 shift the power of 2 of the base
10324 (hex == 4, octal == 3, binary == 1)
10325 overflowed was the number more than we can hold?
10327 Shift is used when we add a digit. It also serves as an "are
10328 we in octal/hex/binary?" indicator to disallow hex characters
10329 when in octal mode.
10334 bool overflowed = FALSE;
10335 bool just_zero = TRUE; /* just plain 0 or binary number? */
10336 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10337 static const char* const bases[5] =
10338 { "", "binary", "", "octal", "hexadecimal" };
10339 static const char* const Bases[5] =
10340 { "", "Binary", "", "Octal", "Hexadecimal" };
10341 static const char* const maxima[5] =
10343 "0b11111111111111111111111111111111",
10347 const char *base, *Base, *max;
10349 /* check for hex */
10350 if (isALPHA_FOLD_EQ(s[1], 'x')) {
10354 } else if (isALPHA_FOLD_EQ(s[1], 'b')) {
10359 /* check for a decimal in disguise */
10360 else if (s[1] == '.' || isALPHA_FOLD_EQ(s[1], 'e'))
10362 /* so it must be octal */
10369 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10370 "Misplaced _ in number");
10374 base = bases[shift];
10375 Base = Bases[shift];
10376 max = maxima[shift];
10378 /* read the rest of the number */
10380 /* x is used in the overflow test,
10381 b is the digit we're adding on. */
10386 /* if we don't mention it, we're done */
10390 /* _ are ignored -- but warned about if consecutive */
10392 if (lastub && s == lastub + 1)
10393 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10394 "Misplaced _ in number");
10398 /* 8 and 9 are not octal */
10399 case '8': case '9':
10401 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10405 case '2': case '3': case '4':
10406 case '5': case '6': case '7':
10408 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10411 case '0': case '1':
10412 b = *s++ & 15; /* ASCII digit -> value of digit */
10416 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10417 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10418 /* make sure they said 0x */
10421 b = (*s++ & 7) + 9;
10423 /* Prepare to put the digit we have onto the end
10424 of the number so far. We check for overflows.
10430 x = u << shift; /* make room for the digit */
10432 total_bits += shift;
10434 if ((x >> shift) != u
10435 && !(PL_hints & HINT_NEW_BINARY)) {
10438 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10439 "Integer overflow in %s number",
10442 u = x | b; /* add the digit to the end */
10445 n *= nvshift[shift];
10446 /* If an NV has not enough bits in its
10447 * mantissa to represent an UV this summing of
10448 * small low-order numbers is a waste of time
10449 * (because the NV cannot preserve the
10450 * low-order bits anyway): we could just
10451 * remember when did we overflow and in the
10452 * end just multiply n by the right
10457 if (high_non_zero == 0 && b > 0)
10460 /* this could be hexfp, but peek ahead
10461 * to avoid matching ".." */
10462 if (UNLIKELY(HEXFP_PEEK(s))) {
10470 /* if we get here, we had success: make a scalar value from
10475 /* final misplaced underbar check */
10476 if (s[-1] == '_') {
10477 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10480 if (UNLIKELY(HEXFP_PEEK(s))) {
10481 /* Do sloppy (on the underbars) but quick detection
10482 * (and value construction) for hexfp, the decimal
10483 * detection will shortly be more thorough with the
10484 * underbar checks. */
10488 #else /* HEXFP_NV */
10493 NV mult = 1 / 16.0;
10495 for (h++; (isXDIGIT(*h) || *h == '_'); h++) {
10496 if (isXDIGIT(*h)) {
10497 U8 b = XDIGIT_VALUE(*h);
10498 total_bits += shift;
10499 if (total_bits < NV_MANT_DIG) {
10501 hexfp_uquad <<= shift;
10503 hexfp_frac_bits += shift;
10504 #else /* HEXFP_NV */
10505 hexfp_nv += b * mult;
10508 } else if (total_bits - shift < NV_MANT_DIG) {
10509 /* A hexdigit straddling the edge of
10510 * mantissa. We can try grabbing as
10511 * many as possible bits. */
10515 } else if (b & 2) {
10518 } else if (b & 4) {
10521 } else if (b & 8) {
10526 hexfp_uquad <<= shift2;
10528 hexfp_frac_bits += shift2;
10529 #else /* HEXFP_NV */
10530 PERL_UNUSED_VAR(shift2);
10531 hexfp_nv += b * mult;
10539 if (total_bits >= 4) {
10540 if (high_non_zero < 0x8)
10542 if (high_non_zero < 0x4)
10544 if (high_non_zero < 0x2)
10548 if (total_bits > 0 && (isALPHA_FOLD_EQ(*h, 'p'))) {
10549 bool negexp = FALSE;
10553 else if (*h == '-') {
10559 while (isDIGIT(*h) || *h == '_') {
10562 hexfp_exp += *h - '0';
10565 && -hexfp_exp < NV_MIN_EXP - 1) {
10566 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10567 "Hexadecimal float: exponent underflow");
10573 && hexfp_exp > NV_MAX_EXP - 1) {
10574 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10575 "Hexadecimal float: exponent overflow");
10583 hexfp_exp = -hexfp_exp;
10585 hexfp_exp -= hexfp_frac_bits;
10587 hexfp_mult = pow(2.0, hexfp_exp);
10595 if (n > 4294967295.0)
10596 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10597 "%s number > %s non-portable",
10603 if (u > 0xffffffff)
10604 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10605 "%s number > %s non-portable",
10610 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10611 sv = new_constant(start, s - start, "integer",
10612 sv, NULL, NULL, 0);
10613 else if (PL_hints & HINT_NEW_BINARY)
10614 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10619 handle decimal numbers.
10620 we're also sent here when we read a 0 as the first digit
10622 case '1': case '2': case '3': case '4': case '5':
10623 case '6': case '7': case '8': case '9': case '.':
10626 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10635 /* read next group of digits and _ and copy into d */
10638 || UNLIKELY(hexfp && isXDIGIT(*s)))
10640 /* skip underscores, checking for misplaced ones
10644 if (lastub && s == lastub + 1)
10645 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10646 "Misplaced _ in number");
10650 /* check for end of fixed-length buffer */
10652 Perl_croak(aTHX_ "%s", number_too_long);
10653 /* if we're ok, copy the character */
10658 /* final misplaced underbar check */
10659 if (lastub && s == lastub + 1) {
10660 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10663 /* read a decimal portion if there is one. avoid
10664 3..5 being interpreted as the number 3. followed
10667 if (*s == '.' && s[1] != '.') {
10672 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10673 "Misplaced _ in number");
10677 /* copy, ignoring underbars, until we run out of digits.
10681 || UNLIKELY(hexfp && isXDIGIT(*s));
10684 /* fixed length buffer check */
10686 Perl_croak(aTHX_ "%s", number_too_long);
10688 if (lastub && s == lastub + 1)
10689 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10690 "Misplaced _ in number");
10696 /* fractional part ending in underbar? */
10697 if (s[-1] == '_') {
10698 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10699 "Misplaced _ in number");
10701 if (*s == '.' && isDIGIT(s[1])) {
10702 /* oops, it's really a v-string, but without the "v" */
10708 /* read exponent part, if present */
10709 if ((isALPHA_FOLD_EQ(*s, 'e')
10710 || UNLIKELY(hexfp && isALPHA_FOLD_EQ(*s, 'p')))
10711 && strchr("+-0123456789_", s[1]))
10715 /* regardless of whether user said 3E5 or 3e5, use lower 'e',
10716 ditto for p (hexfloats) */
10717 if ((isALPHA_FOLD_EQ(*s, 'e'))) {
10718 /* At least some Mach atof()s don't grok 'E' */
10721 else if (UNLIKELY(hexfp && (isALPHA_FOLD_EQ(*s, 'p')))) {
10728 /* stray preinitial _ */
10730 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10731 "Misplaced _ in number");
10735 /* allow positive or negative exponent */
10736 if (*s == '+' || *s == '-')
10739 /* stray initial _ */
10741 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10742 "Misplaced _ in number");
10746 /* read digits of exponent */
10747 while (isDIGIT(*s) || *s == '_') {
10750 Perl_croak(aTHX_ "%s", number_too_long);
10754 if (((lastub && s == lastub + 1)
10755 || (!isDIGIT(s[1]) && s[1] != '_')))
10756 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10757 "Misplaced _ in number");
10765 We try to do an integer conversion first if no characters
10766 indicating "float" have been found.
10771 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10773 if (flags == IS_NUMBER_IN_UV) {
10775 sv = newSViv(uv); /* Prefer IVs over UVs. */
10778 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10779 if (uv <= (UV) IV_MIN)
10780 sv = newSViv(-(IV)uv);
10787 STORE_LC_NUMERIC_UNDERLYING_SET_STANDARD();
10788 /* terminate the string */
10790 if (UNLIKELY(hexfp)) {
10791 # ifdef NV_MANT_DIG
10792 if (total_bits > NV_MANT_DIG)
10793 Perl_ck_warner(aTHX_ packWARN(WARN_OVERFLOW),
10794 "Hexadecimal float: mantissa overflow");
10797 nv = hexfp_uquad * hexfp_mult;
10798 #else /* HEXFP_NV */
10799 nv = hexfp_nv * hexfp_mult;
10802 nv = Atof(PL_tokenbuf);
10804 RESTORE_LC_NUMERIC_UNDERLYING();
10809 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10810 const char *const key = floatit ? "float" : "integer";
10811 const STRLEN keylen = floatit ? 5 : 7;
10812 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10813 key, keylen, sv, NULL, NULL, 0);
10817 /* if it starts with a v, it could be a v-string */
10820 sv = newSV(5); /* preallocate storage space */
10821 ENTER_with_name("scan_vstring");
10823 s = scan_vstring(s, PL_bufend, sv);
10824 SvREFCNT_inc_simple_void_NN(sv);
10825 LEAVE_with_name("scan_vstring");
10829 /* make the op for the constant and return */
10832 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10834 lvalp->opval = NULL;
10840 S_scan_formline(pTHX_ char *s)
10844 SV * const stuff = newSVpvs("");
10845 bool needargs = FALSE;
10846 bool eofmt = FALSE;
10848 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10850 while (!needargs) {
10853 #ifdef PERL_STRICT_CR
10854 while (SPACE_OR_TAB(*t))
10857 while (SPACE_OR_TAB(*t) || *t == '\r')
10860 if (*t == '\n' || t == PL_bufend) {
10865 eol = (char *) memchr(s,'\n',PL_bufend-s);
10869 for (t = s; t < eol; t++) {
10870 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10872 goto enough; /* ~~ must be first line in formline */
10874 if (*t == '@' || *t == '^')
10878 sv_catpvn(stuff, s, eol-s);
10879 #ifndef PERL_STRICT_CR
10880 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10881 char *end = SvPVX(stuff) + SvCUR(stuff);
10884 SvCUR_set(stuff, SvCUR(stuff) - 1);
10892 if ((PL_rsfp || PL_parser->filtered)
10893 && PL_parser->form_lex_state == LEX_NORMAL) {
10895 PL_bufptr = PL_bufend;
10896 COPLINE_INC_WITH_HERELINES;
10897 got_some = lex_next_chunk(0);
10898 CopLINE_dec(PL_curcop);
10906 if (!SvCUR(stuff) || needargs)
10907 PL_lex_state = PL_parser->form_lex_state;
10908 if (SvCUR(stuff)) {
10909 PL_expect = XSTATE;
10911 const char *s2 = s;
10912 while (*s2 == '\r' || *s2 == ' ' || *s2 == '\t' || *s2 == '\f'
10916 PL_expect = XTERMBLOCK;
10917 NEXTVAL_NEXTTOKE.ival = 0;
10920 NEXTVAL_NEXTTOKE.ival = 0;
10921 force_next(FORMLBRACK);
10924 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10926 else if (IN_ENCODING)
10927 sv_recode_to_utf8(stuff, _get_encoding());
10929 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10933 SvREFCNT_dec(stuff);
10935 PL_lex_formbrack = 0;
10941 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10943 const I32 oldsavestack_ix = PL_savestack_ix;
10944 CV* const outsidecv = PL_compcv;
10946 SAVEI32(PL_subline);
10947 save_item(PL_subname);
10948 SAVESPTR(PL_compcv);
10950 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10951 CvFLAGS(PL_compcv) |= flags;
10953 PL_subline = CopLINE(PL_curcop);
10954 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10955 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10956 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10957 if (outsidecv && CvPADLIST(outsidecv))
10958 CvPADLIST(PL_compcv)->xpadl_outid = CvPADLIST(outsidecv)->xpadl_id;
10960 return oldsavestack_ix;
10964 S_yywarn(pTHX_ const char *const s, U32 flags)
10966 PERL_ARGS_ASSERT_YYWARN;
10968 PL_in_eval |= EVAL_WARNONLY;
10969 yyerror_pv(s, flags);
10974 Perl_yyerror(pTHX_ const char *const s)
10976 PERL_ARGS_ASSERT_YYERROR;
10977 return yyerror_pvn(s, strlen(s), 0);
10981 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
10983 PERL_ARGS_ASSERT_YYERROR_PV;
10984 return yyerror_pvn(s, strlen(s), flags);
10988 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
10990 const char *context = NULL;
10993 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
10994 int yychar = PL_parser->yychar;
10996 PERL_ARGS_ASSERT_YYERROR_PVN;
10998 if (!yychar || (yychar == ';' && !PL_rsfp))
10999 sv_catpvs(where_sv, "at EOF");
11000 else if ( PL_oldoldbufptr
11001 && PL_bufptr > PL_oldoldbufptr
11002 && PL_bufptr - PL_oldoldbufptr < 200
11003 && PL_oldoldbufptr != PL_oldbufptr
11004 && PL_oldbufptr != PL_bufptr)
11008 The code below is removed for NetWare because it abends/crashes on NetWare
11009 when the script has error such as not having the closing quotes like:
11010 if ($var eq "value)
11011 Checking of white spaces is anyway done in NetWare code.
11014 while (isSPACE(*PL_oldoldbufptr))
11017 context = PL_oldoldbufptr;
11018 contlen = PL_bufptr - PL_oldoldbufptr;
11020 else if ( PL_oldbufptr
11021 && PL_bufptr > PL_oldbufptr
11022 && PL_bufptr - PL_oldbufptr < 200
11023 && PL_oldbufptr != PL_bufptr) {
11026 The code below is removed for NetWare because it abends/crashes on NetWare
11027 when the script has error such as not having the closing quotes like:
11028 if ($var eq "value)
11029 Checking of white spaces is anyway done in NetWare code.
11032 while (isSPACE(*PL_oldbufptr))
11035 context = PL_oldbufptr;
11036 contlen = PL_bufptr - PL_oldbufptr;
11038 else if (yychar > 255)
11039 sv_catpvs(where_sv, "next token ???");
11040 else if (yychar == YYEMPTY) {
11041 if ( PL_lex_state == LEX_NORMAL
11042 || (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11043 sv_catpvs(where_sv, "at end of line");
11044 else if (PL_lex_inpat)
11045 sv_catpvs(where_sv, "within pattern");
11047 sv_catpvs(where_sv, "within string");
11050 sv_catpvs(where_sv, "next char ");
11052 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11053 else if (isPRINT_LC(yychar)) {
11054 const char string = yychar;
11055 sv_catpvn(where_sv, &string, 1);
11058 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11060 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
11061 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11062 OutCopFILE(PL_curcop),
11063 (IV)(PL_parser->preambling == NOLINE
11064 ? CopLINE(PL_curcop)
11065 : PL_parser->preambling));
11067 Perl_sv_catpvf(aTHX_ msg, "near \"%"UTF8f"\"\n",
11068 UTF8fARG(UTF, contlen, context));
11070 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11071 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11072 Perl_sv_catpvf(aTHX_ msg,
11073 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11074 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11077 if (PL_in_eval & EVAL_WARNONLY) {
11078 PL_in_eval &= ~EVAL_WARNONLY;
11079 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11083 if (PL_error_count >= 10) {
11085 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
11086 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11087 SVfARG(errsv), OutCopFILE(PL_curcop));
11089 Perl_croak(aTHX_ "%s has too many errors.\n",
11090 OutCopFILE(PL_curcop));
11093 PL_in_my_stash = NULL;
11098 S_swallow_bom(pTHX_ U8 *s)
11100 const STRLEN slen = SvCUR(PL_linestr);
11102 PERL_ARGS_ASSERT_SWALLOW_BOM;
11106 if (s[1] == 0xFE) {
11107 /* UTF-16 little-endian? (or UTF-32LE?) */
11108 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11109 /* diag_listed_as: Unsupported script encoding %s */
11110 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11111 #ifndef PERL_NO_UTF16_FILTER
11112 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11114 if (PL_bufend > (char*)s) {
11115 s = add_utf16_textfilter(s, TRUE);
11118 /* diag_listed_as: Unsupported script encoding %s */
11119 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11124 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11125 #ifndef PERL_NO_UTF16_FILTER
11126 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11128 if (PL_bufend > (char *)s) {
11129 s = add_utf16_textfilter(s, FALSE);
11132 /* diag_listed_as: Unsupported script encoding %s */
11133 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11137 case BOM_UTF8_FIRST_BYTE: {
11138 const STRLEN len = sizeof(BOM_UTF8_TAIL) - 1; /* Exclude trailing NUL */
11139 if (slen > len && memEQ(s+1, BOM_UTF8_TAIL, len)) {
11140 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11141 s += len + 1; /* UTF-8 */
11148 if (s[2] == 0xFE && s[3] == 0xFF) {
11149 /* UTF-32 big-endian */
11150 /* diag_listed_as: Unsupported script encoding %s */
11151 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11154 else if (s[2] == 0 && s[3] != 0) {
11157 * are a good indicator of UTF-16BE. */
11158 #ifndef PERL_NO_UTF16_FILTER
11159 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11160 s = add_utf16_textfilter(s, FALSE);
11162 /* diag_listed_as: Unsupported script encoding %s */
11163 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11170 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11173 * are a good indicator of UTF-16LE. */
11174 #ifndef PERL_NO_UTF16_FILTER
11175 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11176 s = add_utf16_textfilter(s, TRUE);
11178 /* diag_listed_as: Unsupported script encoding %s */
11179 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11187 #ifndef PERL_NO_UTF16_FILTER
11189 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11191 SV *const filter = FILTER_DATA(idx);
11192 /* We re-use this each time round, throwing the contents away before we
11194 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11195 SV *const utf8_buffer = filter;
11196 IV status = IoPAGE(filter);
11197 const bool reverse = cBOOL(IoLINES(filter));
11200 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11202 /* As we're automatically added, at the lowest level, and hence only called
11203 from this file, we can be sure that we're not called in block mode. Hence
11204 don't bother writing code to deal with block mode. */
11206 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11209 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11211 DEBUG_P(PerlIO_printf(Perl_debug_log,
11212 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11213 FPTR2DPTR(void *, S_utf16_textfilter),
11214 reverse ? 'l' : 'b', idx, maxlen, status,
11215 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11222 /* First, look in our buffer of existing UTF-8 data: */
11223 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11227 } else if (status == 0) {
11229 IoPAGE(filter) = 0;
11230 nl = SvEND(utf8_buffer);
11233 STRLEN got = nl - SvPVX(utf8_buffer);
11234 /* Did we have anything to append? */
11236 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11237 /* Everything else in this code works just fine if SVp_POK isn't
11238 set. This, however, needs it, and we need it to work, else
11239 we loop infinitely because the buffer is never consumed. */
11240 sv_chop(utf8_buffer, nl);
11244 /* OK, not a complete line there, so need to read some more UTF-16.
11245 Read an extra octect if the buffer currently has an odd number. */
11249 if (SvCUR(utf16_buffer) >= 2) {
11250 /* Location of the high octet of the last complete code point.
11251 Gosh, UTF-16 is a pain. All the benefits of variable length,
11252 *coupled* with all the benefits of partial reads and
11254 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11255 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11257 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11261 /* We have the first half of a surrogate. Read more. */
11262 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11265 status = FILTER_READ(idx + 1, utf16_buffer,
11266 160 + (SvCUR(utf16_buffer) & 1));
11267 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11268 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11271 IoPAGE(filter) = status;
11276 chars = SvCUR(utf16_buffer) >> 1;
11277 have = SvCUR(utf8_buffer);
11278 SvGROW(utf8_buffer, have + chars * 3 + 1);
11281 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11282 (U8*)SvPVX_const(utf8_buffer) + have,
11283 chars * 2, &newlen);
11285 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11286 (U8*)SvPVX_const(utf8_buffer) + have,
11287 chars * 2, &newlen);
11289 SvCUR_set(utf8_buffer, have + newlen);
11292 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11293 it's private to us, and utf16_to_utf8{,reversed} take a
11294 (pointer,length) pair, rather than a NUL-terminated string. */
11295 if(SvCUR(utf16_buffer) & 1) {
11296 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11297 SvCUR_set(utf16_buffer, 1);
11299 SvCUR_set(utf16_buffer, 0);
11302 DEBUG_P(PerlIO_printf(Perl_debug_log,
11303 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11305 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11306 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11311 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11313 SV *filter = filter_add(S_utf16_textfilter, NULL);
11315 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11317 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11318 sv_setpvs(filter, "");
11319 IoLINES(filter) = reversed;
11320 IoPAGE(filter) = 1; /* Not EOF */
11322 /* Sadly, we have to return a valid pointer, come what may, so we have to
11323 ignore any error return from this. */
11324 SvCUR_set(PL_linestr, 0);
11325 if (FILTER_READ(0, PL_linestr, 0)) {
11326 SvUTF8_on(PL_linestr);
11328 SvUTF8_on(PL_linestr);
11330 PL_bufend = SvEND(PL_linestr);
11331 return (U8*)SvPVX(PL_linestr);
11336 Returns a pointer to the next character after the parsed
11337 vstring, as well as updating the passed in sv.
11339 Function must be called like
11341 sv = sv_2mortal(newSV(5));
11342 s = scan_vstring(s,e,sv);
11344 where s and e are the start and end of the string.
11345 The sv should already be large enough to store the vstring
11346 passed in, for performance reasons.
11348 This function may croak if fatal warnings are enabled in the
11349 calling scope, hence the sv_2mortal in the example (to prevent
11350 a leak). Make sure to do SvREFCNT_inc afterwards if you use
11356 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11358 const char *pos = s;
11359 const char *start = s;
11361 PERL_ARGS_ASSERT_SCAN_VSTRING;
11363 if (*pos == 'v') pos++; /* get past 'v' */
11364 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11366 if ( *pos != '.') {
11367 /* this may not be a v-string if followed by => */
11368 const char *next = pos;
11369 while (next < e && isSPACE(*next))
11371 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11372 /* return string not v-string */
11373 sv_setpvn(sv,(char *)s,pos-s);
11374 return (char *)pos;
11378 if (!isALPHA(*pos)) {
11379 U8 tmpbuf[UTF8_MAXBYTES+1];
11382 s++; /* get past 'v' */
11387 /* this is atoi() that tolerates underscores */
11390 const char *end = pos;
11392 while (--end >= s) {
11394 const UV orev = rev;
11395 rev += (*end - '0') * mult;
11398 /* diag_listed_as: Integer overflow in %s number */
11399 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11400 "Integer overflow in decimal number");
11404 /* Append native character for the rev point */
11405 tmpend = uvchr_to_utf8(tmpbuf, rev);
11406 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11407 if (!UVCHR_IS_INVARIANT(rev))
11409 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11415 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11419 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11426 Perl_keyword_plugin_standard(pTHX_
11427 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11429 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11430 PERL_UNUSED_CONTEXT;
11431 PERL_UNUSED_ARG(keyword_ptr);
11432 PERL_UNUSED_ARG(keyword_len);
11433 PERL_UNUSED_ARG(op_ptr);
11434 return KEYWORD_PLUGIN_DECLINE;
11437 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11439 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11441 SAVEI32(PL_lex_brackets);
11442 if (PL_lex_brackets > 100)
11443 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11444 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11445 SAVEI32(PL_lex_allbrackets);
11446 PL_lex_allbrackets = 0;
11447 SAVEI8(PL_lex_fakeeof);
11448 PL_lex_fakeeof = (U8)fakeeof;
11449 if(yyparse(gramtype) && !PL_parser->error_count)
11450 qerror(Perl_mess(aTHX_ "Parse error"));
11453 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11455 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11459 SAVEVPTR(PL_eval_root);
11460 PL_eval_root = NULL;
11461 parse_recdescent(gramtype, fakeeof);
11467 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11469 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11472 if (flags & ~PARSE_OPTIONAL)
11473 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11474 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11475 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11476 if (!PL_parser->error_count)
11477 qerror(Perl_mess(aTHX_ "Parse error"));
11478 exprop = newOP(OP_NULL, 0);
11484 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11486 Parse a Perl arithmetic expression. This may contain operators of precedence
11487 down to the bit shift operators. The expression must be followed (and thus
11488 terminated) either by a comparison or lower-precedence operator or by
11489 something that would normally terminate an expression such as semicolon.
11490 If C<flags> has the C<PARSE_OPTIONAL> bit set, then the expression is optional,
11491 otherwise it is mandatory. It is up to the caller to ensure that the
11492 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11493 the source of the code to be parsed and the lexical context for the
11496 The op tree representing the expression is returned. If an optional
11497 expression is absent, a null pointer is returned, otherwise the pointer
11500 If an error occurs in parsing or compilation, in most cases a valid op
11501 tree is returned anyway. The error is reflected in the parser state,
11502 normally resulting in a single exception at the top level of parsing
11503 which covers all the compilation errors that occurred. Some compilation
11504 errors, however, will throw an exception immediately.
11510 Perl_parse_arithexpr(pTHX_ U32 flags)
11512 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11516 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11518 Parse a Perl term expression. This may contain operators of precedence
11519 down to the assignment operators. The expression must be followed (and thus
11520 terminated) either by a comma or lower-precedence operator or by
11521 something that would normally terminate an expression such as semicolon.
11522 If C<flags> has the C<PARSE_OPTIONAL> bit set, then the expression is optional,
11523 otherwise it is mandatory. It is up to the caller to ensure that the
11524 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11525 the source of the code to be parsed and the lexical context for the
11528 The op tree representing the expression is returned. If an optional
11529 expression is absent, a null pointer is returned, otherwise the pointer
11532 If an error occurs in parsing or compilation, in most cases a valid op
11533 tree is returned anyway. The error is reflected in the parser state,
11534 normally resulting in a single exception at the top level of parsing
11535 which covers all the compilation errors that occurred. Some compilation
11536 errors, however, will throw an exception immediately.
11542 Perl_parse_termexpr(pTHX_ U32 flags)
11544 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11548 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11550 Parse a Perl list expression. This may contain operators of precedence
11551 down to the comma operator. The expression must be followed (and thus
11552 terminated) either by a low-precedence logic operator such as C<or> or by
11553 something that would normally terminate an expression such as semicolon.
11554 If C<flags> has the C<PARSE_OPTIONAL> bit set, then the expression is optional,
11555 otherwise it is mandatory. It is up to the caller to ensure that the
11556 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11557 the source of the code to be parsed and the lexical context for the
11560 The op tree representing the expression is returned. If an optional
11561 expression is absent, a null pointer is returned, otherwise the pointer
11564 If an error occurs in parsing or compilation, in most cases a valid op
11565 tree is returned anyway. The error is reflected in the parser state,
11566 normally resulting in a single exception at the top level of parsing
11567 which covers all the compilation errors that occurred. Some compilation
11568 errors, however, will throw an exception immediately.
11574 Perl_parse_listexpr(pTHX_ U32 flags)
11576 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11580 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11582 Parse a single complete Perl expression. This allows the full
11583 expression grammar, including the lowest-precedence operators such
11584 as C<or>. The expression must be followed (and thus terminated) by a
11585 token that an expression would normally be terminated by: end-of-file,
11586 closing bracketing punctuation, semicolon, or one of the keywords that
11587 signals a postfix expression-statement modifier. If C<flags> has the
11588 C<PARSE_OPTIONAL> bit set, then the expression is optional, otherwise it is
11589 mandatory. It is up to the caller to ensure that the dynamic parser
11590 state (L</PL_parser> et al) is correctly set to reflect the source of
11591 the code to be parsed and the lexical context for the expression.
11593 The op tree representing the expression is returned. If an optional
11594 expression is absent, a null pointer is returned, otherwise the pointer
11597 If an error occurs in parsing or compilation, in most cases a valid op
11598 tree is returned anyway. The error is reflected in the parser state,
11599 normally resulting in a single exception at the top level of parsing
11600 which covers all the compilation errors that occurred. Some compilation
11601 errors, however, will throw an exception immediately.
11607 Perl_parse_fullexpr(pTHX_ U32 flags)
11609 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11613 =for apidoc Amx|OP *|parse_block|U32 flags
11615 Parse a single complete Perl code block. This consists of an opening
11616 brace, a sequence of statements, and a closing brace. The block
11617 constitutes a lexical scope, so C<my> variables and various compile-time
11618 effects can be contained within it. It is up to the caller to ensure
11619 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11620 reflect the source of the code to be parsed and the lexical context for
11623 The op tree representing the code block is returned. This is always a
11624 real op, never a null pointer. It will normally be a C<lineseq> list,
11625 including C<nextstate> or equivalent ops. No ops to construct any kind
11626 of runtime scope are included by virtue of it being a block.
11628 If an error occurs in parsing or compilation, in most cases a valid op
11629 tree (most likely null) is returned anyway. The error is reflected in
11630 the parser state, normally resulting in a single exception at the top
11631 level of parsing which covers all the compilation errors that occurred.
11632 Some compilation errors, however, will throw an exception immediately.
11634 The C<flags> parameter is reserved for future use, and must always
11641 Perl_parse_block(pTHX_ U32 flags)
11644 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11645 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11649 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11651 Parse a single unadorned Perl statement. This may be a normal imperative
11652 statement or a declaration that has compile-time effect. It does not
11653 include any label or other affixture. It is up to the caller to ensure
11654 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11655 reflect the source of the code to be parsed and the lexical context for
11658 The op tree representing the statement is returned. This may be a
11659 null pointer if the statement is null, for example if it was actually
11660 a subroutine definition (which has compile-time side effects). If not
11661 null, it will be ops directly implementing the statement, suitable to
11662 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11663 equivalent op (except for those embedded in a scope contained entirely
11664 within the statement).
11666 If an error occurs in parsing or compilation, in most cases a valid op
11667 tree (most likely null) is returned anyway. The error is reflected in
11668 the parser state, normally resulting in a single exception at the top
11669 level of parsing which covers all the compilation errors that occurred.
11670 Some compilation errors, however, will throw an exception immediately.
11672 The C<flags> parameter is reserved for future use, and must always
11679 Perl_parse_barestmt(pTHX_ U32 flags)
11682 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11683 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11687 =for apidoc Amx|SV *|parse_label|U32 flags
11689 Parse a single label, possibly optional, of the type that may prefix a
11690 Perl statement. It is up to the caller to ensure that the dynamic parser
11691 state (L</PL_parser> et al) is correctly set to reflect the source of
11692 the code to be parsed. If C<flags> has the C<PARSE_OPTIONAL> bit set, then the
11693 label is optional, otherwise it is mandatory.
11695 The name of the label is returned in the form of a fresh scalar. If an
11696 optional label is absent, a null pointer is returned.
11698 If an error occurs in parsing, which can only occur if the label is
11699 mandatory, a valid label is returned anyway. The error is reflected in
11700 the parser state, normally resulting in a single exception at the top
11701 level of parsing which covers all the compilation errors that occurred.
11707 Perl_parse_label(pTHX_ U32 flags)
11709 if (flags & ~PARSE_OPTIONAL)
11710 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11711 if (PL_lex_state == LEX_KNOWNEXT) {
11712 PL_parser->yychar = yylex();
11713 if (PL_parser->yychar == LABEL) {
11714 char * const lpv = pl_yylval.pval;
11715 STRLEN llen = strlen(lpv);
11716 PL_parser->yychar = YYEMPTY;
11717 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11724 STRLEN wlen, bufptr_pos;
11727 if (!isIDFIRST_lazy_if(s, UTF))
11729 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11730 if (word_takes_any_delimeter(s, wlen))
11732 bufptr_pos = s - SvPVX(PL_linestr);
11734 lex_read_space(LEX_KEEP_PREVIOUS);
11736 s = SvPVX(PL_linestr) + bufptr_pos;
11737 if (t[0] == ':' && t[1] != ':') {
11738 PL_oldoldbufptr = PL_oldbufptr;
11741 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11745 if (flags & PARSE_OPTIONAL) {
11748 qerror(Perl_mess(aTHX_ "Parse error"));
11749 return newSVpvs("x");
11756 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11758 Parse a single complete Perl statement. This may be a normal imperative
11759 statement or a declaration that has compile-time effect, and may include
11760 optional labels. It is up to the caller to ensure that the dynamic
11761 parser state (L</PL_parser> et al) is correctly set to reflect the source
11762 of the code to be parsed and the lexical context for the statement.
11764 The op tree representing the statement is returned. This may be a
11765 null pointer if the statement is null, for example if it was actually
11766 a subroutine definition (which has compile-time side effects). If not
11767 null, it will be the result of a L</newSTATEOP> call, normally including
11768 a C<nextstate> or equivalent op.
11770 If an error occurs in parsing or compilation, in most cases a valid op
11771 tree (most likely null) is returned anyway. The error is reflected in
11772 the parser state, normally resulting in a single exception at the top
11773 level of parsing which covers all the compilation errors that occurred.
11774 Some compilation errors, however, will throw an exception immediately.
11776 The C<flags> parameter is reserved for future use, and must always
11783 Perl_parse_fullstmt(pTHX_ U32 flags)
11786 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11787 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11791 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11793 Parse a sequence of zero or more Perl statements. These may be normal
11794 imperative statements, including optional labels, or declarations
11795 that have compile-time effect, or any mixture thereof. The statement
11796 sequence ends when a closing brace or end-of-file is encountered in a
11797 place where a new statement could have validly started. It is up to
11798 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11799 is correctly set to reflect the source of the code to be parsed and the
11800 lexical context for the statements.
11802 The op tree representing the statement sequence is returned. This may
11803 be a null pointer if the statements were all null, for example if there
11804 were no statements or if there were only subroutine definitions (which
11805 have compile-time side effects). If not null, it will be a C<lineseq>
11806 list, normally including C<nextstate> or equivalent ops.
11808 If an error occurs in parsing or compilation, in most cases a valid op
11809 tree is returned anyway. The error is reflected in the parser state,
11810 normally resulting in a single exception at the top level of parsing
11811 which covers all the compilation errors that occurred. Some compilation
11812 errors, however, will throw an exception immediately.
11814 The C<flags> parameter is reserved for future use, and must always
11821 Perl_parse_stmtseq(pTHX_ U32 flags)
11826 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11827 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11828 c = lex_peek_unichar(0);
11829 if (c != -1 && c != /*{*/'}')
11830 qerror(Perl_mess(aTHX_ "Parse error"));
11834 #define lex_token_boundary() S_lex_token_boundary(aTHX)
11836 S_lex_token_boundary(pTHX)
11838 PL_oldoldbufptr = PL_oldbufptr;
11839 PL_oldbufptr = PL_bufptr;
11842 #define parse_opt_lexvar() S_parse_opt_lexvar(aTHX)
11844 S_parse_opt_lexvar(pTHX)
11849 lex_token_boundary();
11850 sigil = lex_read_unichar(0);
11851 if (lex_peek_unichar(0) == '#') {
11852 qerror(Perl_mess(aTHX_ "Parse error"));
11856 c = lex_peek_unichar(0);
11857 if (c == -1 || !(UTF ? isIDFIRST_uni(c) : isIDFIRST_A(c)))
11860 d = PL_tokenbuf + 1;
11861 PL_tokenbuf[0] = (char)sigil;
11862 parse_ident(&s, &d, PL_tokenbuf + sizeof(PL_tokenbuf) - 1, 0, cBOOL(UTF));
11864 if (d == PL_tokenbuf+1)
11866 var = newOP(sigil == '$' ? OP_PADSV : sigil == '@' ? OP_PADAV : OP_PADHV,
11867 OPf_MOD | (OPpLVAL_INTRO<<8));
11868 var->op_targ = allocmy(PL_tokenbuf, d - PL_tokenbuf, UTF ? SVf_UTF8 : 0);
11873 Perl_parse_subsignature(pTHX)
11876 int prev_type = 0, pos = 0, min_arity = 0, max_arity = 0;
11877 OP *initops = NULL;
11879 c = lex_peek_unichar(0);
11880 while (c != /*(*/')') {
11884 if (prev_type == 2)
11885 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11886 var = parse_opt_lexvar();
11888 newBINOP(OP_AELEM, 0,
11889 ref(newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv)),
11891 newSVOP(OP_CONST, 0, newSViv(pos))) :
11894 c = lex_peek_unichar(0);
11896 lex_token_boundary();
11897 lex_read_unichar(0);
11899 c = lex_peek_unichar(0);
11900 if (c == ',' || c == /*(*/')') {
11902 qerror(Perl_mess(aTHX_ "Optional parameter "
11903 "lacks default expression"));
11905 OP *defexpr = parse_termexpr(0);
11906 if (defexpr->op_type == OP_UNDEF
11907 && !(defexpr->op_flags & OPf_KIDS))
11913 scalar(newUNOP(OP_RV2AV, 0,
11914 newGVOP(OP_GV, 0, PL_defgv))),
11915 newSVOP(OP_CONST, 0, newSViv(pos+1)));
11917 newCONDOP(0, ifop, expr, defexpr) :
11918 newLOGOP(OP_OR, 0, ifop, defexpr);
11923 if (prev_type == 1)
11924 qerror(Perl_mess(aTHX_ "Mandatory parameter "
11925 "follows optional parameter"));
11927 min_arity = pos + 1;
11929 if (var) expr = newASSIGNOP(OPf_STACKED, var, 0, expr);
11931 initops = op_append_list(OP_LINESEQ, initops,
11932 newSTATEOP(0, NULL, expr));
11938 if (prev_type == 2)
11939 qerror(Perl_mess(aTHX_ "Slurpy parameter not last"));
11940 var = parse_opt_lexvar();
11942 OP *chkop = newLOGOP((pos & 1) ? OP_OR : OP_AND, 0,
11943 newBINOP(OP_BIT_AND, 0,
11944 scalar(newUNOP(OP_RV2AV, 0,
11945 newGVOP(OP_GV, 0, PL_defgv))),
11946 newSVOP(OP_CONST, 0, newSViv(1))),
11947 op_convert_list(OP_DIE, 0,
11948 op_convert_list(OP_SPRINTF, 0,
11949 op_append_list(OP_LIST,
11950 newSVOP(OP_CONST, 0,
11951 newSVpvs("Odd name/value argument for subroutine at %s line %d.\n")),
11953 op_append_list(OP_LIST,
11954 newSVOP(OP_CONST, 0, newSViv(1)),
11955 newSVOP(OP_CONST, 0, newSViv(2))),
11956 newOP(OP_CALLER, 0))))));
11957 if (pos != min_arity)
11958 chkop = newLOGOP(OP_AND, 0,
11960 scalar(newUNOP(OP_RV2AV, 0,
11961 newGVOP(OP_GV, 0, PL_defgv))),
11962 newSVOP(OP_CONST, 0, newSViv(pos))),
11964 initops = op_append_list(OP_LINESEQ,
11965 newSTATEOP(0, NULL, chkop),
11970 op_prepend_elem(OP_ASLICE,
11971 newOP(OP_PUSHMARK, 0),
11972 newLISTOP(OP_ASLICE, 0,
11974 newSVOP(OP_CONST, 0, newSViv(pos)),
11975 newUNOP(OP_AV2ARYLEN, 0,
11976 ref(newUNOP(OP_RV2AV, 0,
11977 newGVOP(OP_GV, 0, PL_defgv)),
11979 ref(newUNOP(OP_RV2AV, 0,
11980 newGVOP(OP_GV, 0, PL_defgv)),
11982 newUNOP(OP_RV2AV, 0, newGVOP(OP_GV, 0, PL_defgv));
11983 initops = op_append_list(OP_LINESEQ, initops,
11984 newSTATEOP(0, NULL,
11985 newASSIGNOP(OPf_STACKED, var, 0, slice)));
11992 qerror(Perl_mess(aTHX_ "Parse error"));
11996 c = lex_peek_unichar(0);
11998 case /*(*/')': break;
12001 lex_token_boundary();
12002 lex_read_unichar(0);
12004 c = lex_peek_unichar(0);
12005 } while (c == ',');
12011 if (min_arity != 0) {
12012 initops = op_append_list(OP_LINESEQ,
12013 newSTATEOP(0, NULL,
12016 scalar(newUNOP(OP_RV2AV, 0,
12017 newGVOP(OP_GV, 0, PL_defgv))),
12018 newSVOP(OP_CONST, 0, newSViv(min_arity))),
12019 op_convert_list(OP_DIE, 0,
12020 op_convert_list(OP_SPRINTF, 0,
12021 op_append_list(OP_LIST,
12022 newSVOP(OP_CONST, 0,
12023 newSVpvs("Too few arguments for subroutine at %s line %d.\n")),
12025 op_append_list(OP_LIST,
12026 newSVOP(OP_CONST, 0, newSViv(1)),
12027 newSVOP(OP_CONST, 0, newSViv(2))),
12028 newOP(OP_CALLER, 0))))))),
12031 if (max_arity != -1) {
12032 initops = op_append_list(OP_LINESEQ,
12033 newSTATEOP(0, NULL,
12036 scalar(newUNOP(OP_RV2AV, 0,
12037 newGVOP(OP_GV, 0, PL_defgv))),
12038 newSVOP(OP_CONST, 0, newSViv(max_arity))),
12039 op_convert_list(OP_DIE, 0,
12040 op_convert_list(OP_SPRINTF, 0,
12041 op_append_list(OP_LIST,
12042 newSVOP(OP_CONST, 0,
12043 newSVpvs("Too many arguments for subroutine at %s line %d.\n")),
12045 op_append_list(OP_LIST,
12046 newSVOP(OP_CONST, 0, newSViv(1)),
12047 newSVOP(OP_CONST, 0, newSViv(2))),
12048 newOP(OP_CALLER, 0))))))),
12055 * ex: set ts=8 sts=4 sw=4 et: