3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
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 /* YYINITDEPTH -- initial size of the parser's stacks. */
49 #define YYINITDEPTH 200
51 /* XXX temporary backwards compatibility */
52 #define PL_lex_brackets (PL_parser->lex_brackets)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XFAKEBRACK 128
130 #define XENUMMASK 127
132 #ifdef USE_UTF8_SCRIPTS
133 # define UTF (!IN_BYTES)
135 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
138 /* The maximum number of characters preceding the unrecognized one to display */
139 #define UNRECOGNIZED_PRECEDE_COUNT 10
141 /* In variables named $^X, these are the legal values for X.
142 * 1999-02-27 mjd-perl-patch@plover.com */
143 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
145 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
147 /* LEX_* are values for PL_lex_state, the state of the lexer.
148 * They are arranged oddly so that the guard on the switch statement
149 * can get by with a single comparison (if the compiler is smart enough).
152 /* #define LEX_NOTPARSING 11 is done in perl.h. */
154 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
155 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
156 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
157 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
158 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
160 /* at end of code, eg "$x" followed by: */
161 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
162 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
164 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
165 string or after \E, $foo, etc */
166 #define LEX_INTERPCONST 2 /* NOT USED */
167 #define LEX_FORMLINE 1 /* expecting a format line */
168 #define LEX_KNOWNEXT 0 /* next token known; just return it */
172 static const char* const lex_state_names[] = {
191 #include "keywords.h"
193 /* CLINE is a macro that ensures PL_copline has a sane value */
198 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
201 # define SKIPSPACE0(s) skipspace0(s)
202 # define SKIPSPACE1(s) skipspace1(s)
203 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
204 # define PEEKSPACE(s) skipspace2(s,0)
206 # define SKIPSPACE0(s) skipspace(s)
207 # define SKIPSPACE1(s) skipspace(s)
208 # define SKIPSPACE2(s,tsv) skipspace(s)
209 # define PEEKSPACE(s) skipspace(s)
213 * Convenience functions to return different tokens and prime the
214 * lexer for the next token. They all take an argument.
216 * TOKEN : generic token (used for '(', DOLSHARP, etc)
217 * OPERATOR : generic operator
218 * AOPERATOR : assignment operator
219 * PREBLOCK : beginning the block after an if, while, foreach, ...
220 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
221 * PREREF : *EXPR where EXPR is not a simple identifier
222 * TERM : expression term
223 * LOOPX : loop exiting command (goto, last, dump, etc)
224 * FTST : file test operator
225 * FUN0 : zero-argument function
226 * FUN1 : not used, except for not, which isn't a UNIOP
227 * BOop : bitwise or or xor
229 * SHop : shift operator
230 * PWop : power operator
231 * PMop : pattern-matching operator
232 * Aop : addition-level operator
233 * Mop : multiplication-level operator
234 * Eop : equality-testing operator
235 * Rop : relational operator <= != gt
237 * Also see LOP and lop() below.
240 #ifdef DEBUGGING /* Serve -DT. */
241 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
243 # define REPORT(retval) (retval)
246 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
247 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
248 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
249 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
251 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
252 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
253 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
254 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
255 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI2(f,x) { \
273 pl_yylval.ival = f; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI2(f,XTERM)
284 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
286 #define UNIBRACK(f) { \
287 pl_yylval.ival = f; \
289 PL_last_uni = PL_oldbufptr; \
291 return REPORT( (int)FUNC1 ); \
293 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
296 /* grandfather return to old style */
297 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
301 /* how to interpret the pl_yylval associated with the token */
305 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
311 static struct debug_tokens {
313 enum token_type type;
315 } const debug_tokens[] =
317 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
318 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
319 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
320 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
321 { ARROW, TOKENTYPE_NONE, "ARROW" },
322 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
323 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
324 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
325 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
326 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
327 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
328 { DO, TOKENTYPE_NONE, "DO" },
329 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
330 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
331 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
332 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
333 { ELSE, TOKENTYPE_NONE, "ELSE" },
334 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
335 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
336 { FOR, TOKENTYPE_IVAL, "FOR" },
337 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
338 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
339 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
340 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
341 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
342 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
343 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
344 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
345 { IF, TOKENTYPE_IVAL, "IF" },
346 { LABEL, TOKENTYPE_PVAL, "LABEL" },
347 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
348 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
349 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
350 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
351 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
352 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
353 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
354 { MY, TOKENTYPE_IVAL, "MY" },
355 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
356 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
357 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
358 { OROP, TOKENTYPE_IVAL, "OROP" },
359 { OROR, TOKENTYPE_NONE, "OROR" },
360 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
361 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
362 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
363 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
364 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
365 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
366 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
367 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
368 { PREINC, TOKENTYPE_NONE, "PREINC" },
369 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
370 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
371 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
372 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
373 { SUB, TOKENTYPE_NONE, "SUB" },
374 { THING, TOKENTYPE_OPVAL, "THING" },
375 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
376 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
377 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
378 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
379 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
380 { USE, TOKENTYPE_IVAL, "USE" },
381 { WHEN, TOKENTYPE_IVAL, "WHEN" },
382 { WHILE, TOKENTYPE_IVAL, "WHILE" },
383 { WORD, TOKENTYPE_OPVAL, "WORD" },
384 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
385 { 0, TOKENTYPE_NONE, NULL }
388 /* dump the returned token in rv, plus any optional arg in pl_yylval */
391 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
395 PERL_ARGS_ASSERT_TOKEREPORT;
398 const char *name = NULL;
399 enum token_type type = TOKENTYPE_NONE;
400 const struct debug_tokens *p;
401 SV* const report = newSVpvs("<== ");
403 for (p = debug_tokens; p->token; p++) {
404 if (p->token == (int)rv) {
411 Perl_sv_catpv(aTHX_ report, name);
412 else if ((char)rv > ' ' && (char)rv < '~')
413 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
415 sv_catpvs(report, "EOF");
417 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
420 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
423 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
425 case TOKENTYPE_OPNUM:
426 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
427 PL_op_name[lvalp->ival]);
430 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
432 case TOKENTYPE_OPVAL:
434 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
435 PL_op_name[lvalp->opval->op_type]);
436 if (lvalp->opval->op_type == OP_CONST) {
437 Perl_sv_catpvf(aTHX_ report, " %s",
438 SvPEEK(cSVOPx_sv(lvalp->opval)));
443 sv_catpvs(report, "(opval=null)");
446 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
452 /* print the buffer with suitable escapes */
455 S_printbuf(pTHX_ const char *const fmt, const char *const s)
457 SV* const tmp = newSVpvs("");
459 PERL_ARGS_ASSERT_PRINTBUF;
461 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
468 S_deprecate_commaless_var_list(pTHX) {
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
477 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
478 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
482 S_ao(pTHX_ int toketype)
485 if (*PL_bufptr == '=') {
487 if (toketype == ANDAND)
488 pl_yylval.ival = OP_ANDASSIGN;
489 else if (toketype == OROR)
490 pl_yylval.ival = OP_ORASSIGN;
491 else if (toketype == DORDOR)
492 pl_yylval.ival = OP_DORASSIGN;
500 * When Perl expects an operator and finds something else, no_op
501 * prints the warning. It always prints "<something> found where
502 * operator expected. It prints "Missing semicolon on previous line?"
503 * if the surprise occurs at the start of the line. "do you need to
504 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
505 * where the compiler doesn't know if foo is a method call or a function.
506 * It prints "Missing operator before end of line" if there's nothing
507 * after the missing operator, or "... before <...>" if there is something
508 * after the missing operator.
512 S_no_op(pTHX_ const char *const what, char *s)
515 char * const oldbp = PL_bufptr;
516 const bool is_first = (PL_oldbufptr == PL_linestart);
518 PERL_ARGS_ASSERT_NO_OP;
524 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
525 if (ckWARN_d(WARN_SYNTAX)) {
527 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
528 "\t(Missing semicolon on previous line?)\n");
529 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
531 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %.*s?)\n",
536 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
541 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
549 * Complain about missing quote/regexp/heredoc terminator.
550 * If it's called with NULL then it cauterizes the line buffer.
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
557 S_missingterm(pTHX_ char *s)
563 char * const nl = strrchr(s,'\n');
567 else if (isCNTRL(PL_multi_close)) {
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);
582 #define FEATURE_IS_ENABLED(name) \
583 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
584 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
585 /* The longest string we pass in. */
586 #define MAX_FEATURE_LEN (sizeof("unicode_strings")-1)
589 * S_feature_is_enabled
590 * Check whether the named feature is enabled.
593 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
596 HV * const hinthv = GvHV(PL_hintgv);
597 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
599 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
601 assert(namelen <= MAX_FEATURE_LEN);
602 memcpy(&he_name[8], name, namelen);
604 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
608 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
609 * utf16-to-utf8-reversed.
612 #ifdef PERL_CR_FILTER
616 register const char *s = SvPVX_const(sv);
617 register const char * const e = s + SvCUR(sv);
619 PERL_ARGS_ASSERT_STRIP_RETURN;
621 /* outer loop optimized to do nothing if there are no CR-LFs */
623 if (*s++ == '\r' && *s == '\n') {
624 /* hit a CR-LF, need to copy the rest */
625 register char *d = s - 1;
628 if (*s == '\r' && s[1] == '\n')
639 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
642 if (count > 0 && !maxlen)
653 * Create a parser object and initialise its parser and lexer fields
655 * rsfp is the opened file handle to read from (if any),
657 * line holds any initial content already read from the file (or in
658 * the case of no file, such as an eval, the whole contents);
660 * new_filter indicates that this is a new file and it shouldn't inherit
661 * the filters from the current parser (ie require).
665 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
668 const char *s = NULL;
670 yy_parser *parser, *oparser;
672 /* create and initialise a parser */
674 Newxz(parser, 1, yy_parser);
675 parser->old_parser = oparser = PL_parser;
678 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
679 parser->ps = parser->stack;
680 parser->stack_size = YYINITDEPTH;
682 parser->stack->state = 0;
683 parser->yyerrstatus = 0;
684 parser->yychar = YYEMPTY; /* Cause a token to be read. */
686 /* on scope exit, free this parser and restore any outer one */
688 parser->saved_curcop = PL_curcop;
690 /* initialise lexer state */
693 parser->curforce = -1;
695 parser->nexttoke = 0;
697 parser->error_count = oparser ? oparser->error_count : 0;
698 parser->copline = NOLINE;
699 parser->lex_state = LEX_NORMAL;
700 parser->expect = XSTATE;
702 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
703 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
705 Newx(parser->lex_brackstack, 120, char);
706 Newx(parser->lex_casestack, 12, char);
707 *parser->lex_casestack = '\0';
710 s = SvPV_const(line, len);
716 parser->linestr = newSVpvs("\n;");
717 } else if (SvREADONLY(line) || s[len-1] != ';') {
718 parser->linestr = newSVsv(line);
720 sv_catpvs(parser->linestr, "\n;");
723 SvREFCNT_inc_simple_void_NN(line);
724 parser->linestr = line;
726 parser->oldoldbufptr =
729 parser->linestart = SvPVX(parser->linestr);
730 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
731 parser->last_lop = parser->last_uni = NULL;
735 /* delete a parser object */
738 Perl_parser_free(pTHX_ const yy_parser *parser)
740 PERL_ARGS_ASSERT_PARSER_FREE;
742 PL_curcop = parser->saved_curcop;
743 SvREFCNT_dec(parser->linestr);
745 if (parser->rsfp == PerlIO_stdin())
746 PerlIO_clearerr(parser->rsfp);
747 else if (parser->rsfp && (!parser->old_parser ||
748 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
749 PerlIO_close(parser->rsfp);
750 SvREFCNT_dec(parser->rsfp_filters);
752 Safefree(parser->stack);
753 Safefree(parser->lex_brackstack);
754 Safefree(parser->lex_casestack);
755 PL_parser = parser->old_parser;
762 * Finalizer for lexing operations. Must be called when the parser is
763 * done with the lexer.
770 PL_doextract = FALSE;
774 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
776 Buffer scalar containing the chunk currently under consideration of the
777 text currently being lexed. This is always a plain string scalar (for
778 which C<SvPOK> is true). It is not intended to be used as a scalar by
779 normal scalar means; instead refer to the buffer directly by the pointer
780 variables described below.
782 The lexer maintains various C<char*> pointers to things in the
783 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
784 reallocated, all of these pointers must be updated. Don't attempt to
785 do this manually, but rather use L</lex_grow_linestr> if you need to
786 reallocate the buffer.
788 The content of the text chunk in the buffer is commonly exactly one
789 complete line of input, up to and including a newline terminator,
790 but there are situations where it is otherwise. The octets of the
791 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
792 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
793 flag on this scalar, which may disagree with it.
795 For direct examination of the buffer, the variable
796 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
797 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
798 of these pointers is usually preferable to examination of the scalar
799 through normal scalar means.
801 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
803 Direct pointer to the end of the chunk of text currently being lexed, the
804 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
805 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
806 always located at the end of the buffer, and does not count as part of
807 the buffer's contents.
809 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
811 Points to the current position of lexing inside the lexer buffer.
812 Characters around this point may be freely examined, within
813 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
814 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
815 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
817 Lexing code (whether in the Perl core or not) moves this pointer past
818 the characters that it consumes. It is also expected to perform some
819 bookkeeping whenever a newline character is consumed. This movement
820 can be more conveniently performed by the function L</lex_read_to>,
821 which handles newlines appropriately.
823 Interpretation of the buffer's octets can be abstracted out by
824 using the slightly higher-level functions L</lex_peek_unichar> and
825 L</lex_read_unichar>.
827 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
829 Points to the start of the current line inside the lexer buffer.
830 This is useful for indicating at which column an error occurred, and
831 not much else. This must be updated by any lexing code that consumes
832 a newline; the function L</lex_read_to> handles this detail.
838 =for apidoc Amx|bool|lex_bufutf8
840 Indicates whether the octets in the lexer buffer
841 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
842 of Unicode characters. If not, they should be interpreted as Latin-1
843 characters. This is analogous to the C<SvUTF8> flag for scalars.
845 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
846 contains valid UTF-8. Lexing code must be robust in the face of invalid
849 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
850 is significant, but not the whole story regarding the input character
851 encoding. Normally, when a file is being read, the scalar contains octets
852 and its C<SvUTF8> flag is off, but the octets should be interpreted as
853 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
854 however, the scalar may have the C<SvUTF8> flag on, and in this case its
855 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
856 is in effect. This logic may change in the future; use this function
857 instead of implementing the logic yourself.
863 Perl_lex_bufutf8(pTHX)
869 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
871 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
872 at least I<len> octets (including terminating NUL). Returns a
873 pointer to the reallocated buffer. This is necessary before making
874 any direct modification of the buffer that would increase its length.
875 L</lex_stuff_pvn> provides a more convenient way to insert text into
878 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
879 this function updates all of the lexer's variables that point directly
886 Perl_lex_grow_linestr(pTHX_ STRLEN len)
890 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
891 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
892 linestr = PL_parser->linestr;
893 buf = SvPVX(linestr);
894 if (len <= SvLEN(linestr))
896 bufend_pos = PL_parser->bufend - buf;
897 bufptr_pos = PL_parser->bufptr - buf;
898 oldbufptr_pos = PL_parser->oldbufptr - buf;
899 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
900 linestart_pos = PL_parser->linestart - buf;
901 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
902 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
903 buf = sv_grow(linestr, len);
904 PL_parser->bufend = buf + bufend_pos;
905 PL_parser->bufptr = buf + bufptr_pos;
906 PL_parser->oldbufptr = buf + oldbufptr_pos;
907 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
908 PL_parser->linestart = buf + linestart_pos;
909 if (PL_parser->last_uni)
910 PL_parser->last_uni = buf + last_uni_pos;
911 if (PL_parser->last_lop)
912 PL_parser->last_lop = buf + last_lop_pos;
917 =for apidoc Amx|void|lex_stuff_pvn|char *pv|STRLEN len|U32 flags
919 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
920 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
921 reallocating the buffer if necessary. This means that lexing code that
922 runs later will see the characters as if they had appeared in the input.
923 It is not recommended to do this as part of normal parsing, and most
924 uses of this facility run the risk of the inserted characters being
925 interpreted in an unintended manner.
927 The string to be inserted is represented by I<len> octets starting
928 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
929 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
930 The characters are recoded for the lexer buffer, according to how the
931 buffer is currently being interpreted (L</lex_bufutf8>). If a string
932 to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
933 function is more convenient.
939 Perl_lex_stuff_pvn(pTHX_ char *pv, STRLEN len, U32 flags)
942 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
943 if (flags & ~(LEX_STUFF_UTF8))
944 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
946 if (flags & LEX_STUFF_UTF8) {
950 char *p, *e = pv+len;
951 for (p = pv; p != e; p++)
952 highhalf += !!(((U8)*p) & 0x80);
955 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
956 bufptr = PL_parser->bufptr;
957 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
958 PL_parser->bufend += len+highhalf;
959 for (p = pv; p != e; p++) {
962 *bufptr++ = (char)(0xc0 | (c >> 6));
963 *bufptr++ = (char)(0x80 | (c & 0x3f));
970 if (flags & LEX_STUFF_UTF8) {
972 char *p, *e = pv+len;
973 for (p = pv; p != e; p++) {
976 Perl_croak(aTHX_ "Lexing code attempted to stuff "
977 "non-Latin-1 character into Latin-1 input");
978 } else if (c >= 0xc2 && p+1 != e &&
979 (((U8)p[1]) & 0xc0) == 0x80) {
982 } else if (c >= 0x80) {
983 /* malformed UTF-8 */
985 SAVESPTR(PL_warnhook);
986 PL_warnhook = PERL_WARNHOOK_FATAL;
987 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
993 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
994 bufptr = PL_parser->bufptr;
995 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
996 PL_parser->bufend += len-highhalf;
997 for (p = pv; p != e; p++) {
1000 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1003 *bufptr++ = (char)c;
1008 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1009 bufptr = PL_parser->bufptr;
1010 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1011 PL_parser->bufend += len;
1012 Copy(pv, bufptr, len, char);
1018 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1020 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1021 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1022 reallocating the buffer if necessary. This means that lexing code that
1023 runs later will see the characters as if they had appeared in the input.
1024 It is not recommended to do this as part of normal parsing, and most
1025 uses of this facility run the risk of the inserted characters being
1026 interpreted in an unintended manner.
1028 The string to be inserted is the string value of I<sv>. The characters
1029 are recoded for the lexer buffer, according to how the buffer is currently
1030 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1031 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1032 need to construct a scalar.
1038 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1042 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1044 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1046 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1050 =for apidoc Amx|void|lex_unstuff|char *ptr
1052 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1053 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1054 This hides the discarded text from any lexing code that runs later,
1055 as if the text had never appeared.
1057 This is not the normal way to consume lexed text. For that, use
1064 Perl_lex_unstuff(pTHX_ char *ptr)
1068 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1069 buf = PL_parser->bufptr;
1071 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1074 bufend = PL_parser->bufend;
1076 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1077 unstuff_len = ptr - buf;
1078 Move(ptr, buf, bufend+1-ptr, char);
1079 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1080 PL_parser->bufend = bufend - unstuff_len;
1084 =for apidoc Amx|void|lex_read_to|char *ptr
1086 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1087 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1088 performing the correct bookkeeping whenever a newline character is passed.
1089 This is the normal way to consume lexed text.
1091 Interpretation of the buffer's octets can be abstracted out by
1092 using the slightly higher-level functions L</lex_peek_unichar> and
1093 L</lex_read_unichar>.
1099 Perl_lex_read_to(pTHX_ char *ptr)
1102 PERL_ARGS_ASSERT_LEX_READ_TO;
1103 s = PL_parser->bufptr;
1104 if (ptr < s || ptr > PL_parser->bufend)
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1106 for (; s != ptr; s++)
1108 CopLINE_inc(PL_curcop);
1109 PL_parser->linestart = s+1;
1111 PL_parser->bufptr = ptr;
1115 =for apidoc Amx|void|lex_discard_to|char *ptr
1117 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1118 up to I<ptr>. The remaining content of the buffer will be moved, and
1119 all pointers into the buffer updated appropriately. I<ptr> must not
1120 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1121 it is not permitted to discard text that has yet to be lexed.
1123 Normally it is not necessarily to do this directly, because it suffices to
1124 use the implicit discarding behaviour of L</lex_next_chunk> and things
1125 based on it. However, if a token stretches across multiple lines,
1126 and the lexing code has kept multiple lines of text in the buffer fof
1127 that purpose, then after completion of the token it would be wise to
1128 explicitly discard the now-unneeded earlier lines, to avoid future
1129 multi-line tokens growing the buffer without bound.
1135 Perl_lex_discard_to(pTHX_ char *ptr)
1139 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1140 buf = SvPVX(PL_parser->linestr);
1142 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1145 if (ptr > PL_parser->bufptr)
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1147 discard_len = ptr - buf;
1148 if (PL_parser->oldbufptr < ptr)
1149 PL_parser->oldbufptr = ptr;
1150 if (PL_parser->oldoldbufptr < ptr)
1151 PL_parser->oldoldbufptr = ptr;
1152 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1153 PL_parser->last_uni = NULL;
1154 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1155 PL_parser->last_lop = NULL;
1156 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1157 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1158 PL_parser->bufend -= discard_len;
1159 PL_parser->bufptr -= discard_len;
1160 PL_parser->oldbufptr -= discard_len;
1161 PL_parser->oldoldbufptr -= discard_len;
1162 if (PL_parser->last_uni)
1163 PL_parser->last_uni -= discard_len;
1164 if (PL_parser->last_lop)
1165 PL_parser->last_lop -= discard_len;
1169 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1171 Reads in the next chunk of text to be lexed, appending it to
1172 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1173 looked to the end of the current chunk and wants to know more. It is
1174 usual, but not necessary, for lexing to have consumed the entirety of
1175 the current chunk at this time.
1177 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1178 chunk (i.e., the current chunk has been entirely consumed), normally the
1179 current chunk will be discarded at the same time that the new chunk is
1180 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1181 will not be discarded. If the current chunk has not been entirely
1182 consumed, then it will not be discarded regardless of the flag.
1184 Returns true if some new text was added to the buffer, or false if the
1185 buffer has reached the end of the input text.
1190 #define LEX_FAKE_EOF 0x80000000
1193 Perl_lex_next_chunk(pTHX_ U32 flags)
1197 STRLEN old_bufend_pos, new_bufend_pos;
1198 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1199 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1200 bool got_some_for_debugger = 0;
1202 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1203 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1204 linestr = PL_parser->linestr;
1205 buf = SvPVX(linestr);
1206 if (!(flags & LEX_KEEP_PREVIOUS) &&
1207 PL_parser->bufptr == PL_parser->bufend) {
1208 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1210 if (PL_parser->last_uni != PL_parser->bufend)
1211 PL_parser->last_uni = NULL;
1212 if (PL_parser->last_lop != PL_parser->bufend)
1213 PL_parser->last_lop = NULL;
1214 last_uni_pos = last_lop_pos = 0;
1218 old_bufend_pos = PL_parser->bufend - buf;
1219 bufptr_pos = PL_parser->bufptr - buf;
1220 oldbufptr_pos = PL_parser->oldbufptr - buf;
1221 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1222 linestart_pos = PL_parser->linestart - buf;
1223 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1224 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1226 if (flags & LEX_FAKE_EOF) {
1228 } else if (!PL_parser->rsfp) {
1230 } else if (filter_gets(linestr, old_bufend_pos)) {
1232 got_some_for_debugger = 1;
1234 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1235 sv_setpvs(linestr, "");
1237 /* End of real input. Close filehandle (unless it was STDIN),
1238 * then add implicit termination.
1240 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1241 PerlIO_clearerr(PL_parser->rsfp);
1242 else if (PL_parser->rsfp)
1243 (void)PerlIO_close(PL_parser->rsfp);
1244 PL_parser->rsfp = NULL;
1245 PL_doextract = FALSE;
1247 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1250 if (!PL_in_eval && PL_minus_p) {
1252 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1253 PL_minus_n = PL_minus_p = 0;
1254 } else if (!PL_in_eval && PL_minus_n) {
1255 sv_catpvs(linestr, /*{*/";}");
1258 sv_catpvs(linestr, ";");
1261 buf = SvPVX(linestr);
1262 new_bufend_pos = SvCUR(linestr);
1263 PL_parser->bufend = buf + new_bufend_pos;
1264 PL_parser->bufptr = buf + bufptr_pos;
1265 PL_parser->oldbufptr = buf + oldbufptr_pos;
1266 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1267 PL_parser->linestart = buf + linestart_pos;
1268 if (PL_parser->last_uni)
1269 PL_parser->last_uni = buf + last_uni_pos;
1270 if (PL_parser->last_lop)
1271 PL_parser->last_lop = buf + last_lop_pos;
1272 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1273 PL_curstash != PL_debstash) {
1274 /* debugger active and we're not compiling the debugger code,
1275 * so store the line into the debugger's array of lines
1277 update_debugger_info(NULL, buf+old_bufend_pos,
1278 new_bufend_pos-old_bufend_pos);
1284 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1286 Looks ahead one (Unicode) character in the text currently being lexed.
1287 Returns the codepoint (unsigned integer value) of the next character,
1288 or -1 if lexing has reached the end of the input text. To consume the
1289 peeked character, use L</lex_read_unichar>.
1291 If the next character is in (or extends into) the next chunk of input
1292 text, the next chunk will be read in. Normally the current chunk will be
1293 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1294 then the current chunk will not be discarded.
1296 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1297 is encountered, an exception is generated.
1303 Perl_lex_peek_unichar(pTHX_ U32 flags)
1306 if (flags & ~(LEX_KEEP_PREVIOUS))
1307 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1308 s = PL_parser->bufptr;
1309 bufend = PL_parser->bufend;
1315 if (!lex_next_chunk(flags))
1317 s = PL_parser->bufptr;
1318 bufend = PL_parser->bufend;
1324 len = PL_utf8skip[head];
1325 while ((STRLEN)(bufend-s) < len) {
1326 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1328 s = PL_parser->bufptr;
1329 bufend = PL_parser->bufend;
1332 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1333 if (retlen == (STRLEN)-1) {
1334 /* malformed UTF-8 */
1336 SAVESPTR(PL_warnhook);
1337 PL_warnhook = PERL_WARNHOOK_FATAL;
1338 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1344 if (!lex_next_chunk(flags))
1346 s = PL_parser->bufptr;
1353 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1355 Reads the next (Unicode) character in the text currently being lexed.
1356 Returns the codepoint (unsigned integer value) of the character read,
1357 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1358 if lexing has reached the end of the input text. To non-destructively
1359 examine the next character, use L</lex_peek_unichar> instead.
1361 If the next character is in (or extends into) the next chunk of input
1362 text, the next chunk will be read in. Normally the current chunk will be
1363 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1364 then the current chunk will not be discarded.
1366 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1367 is encountered, an exception is generated.
1373 Perl_lex_read_unichar(pTHX_ U32 flags)
1376 if (flags & ~(LEX_KEEP_PREVIOUS))
1377 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1378 c = lex_peek_unichar(flags);
1381 CopLINE_inc(PL_curcop);
1382 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1388 =for apidoc Amx|void|lex_read_space|U32 flags
1390 Reads optional spaces, in Perl style, in the text currently being
1391 lexed. The spaces may include ordinary whitespace characters and
1392 Perl-style comments. C<#line> directives are processed if encountered.
1393 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1394 at a non-space character (or the end of the input text).
1396 If spaces extend into the next chunk of input text, the next chunk will
1397 be read in. Normally the current chunk will be discarded at the same
1398 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1399 chunk will not be discarded.
1405 Perl_lex_read_space(pTHX_ U32 flags)
1408 bool need_incline = 0;
1409 if (flags & ~(LEX_KEEP_PREVIOUS))
1410 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1413 sv_free(PL_skipwhite);
1414 PL_skipwhite = NULL;
1417 PL_skipwhite = newSVpvs("");
1418 #endif /* PERL_MAD */
1419 s = PL_parser->bufptr;
1420 bufend = PL_parser->bufend;
1426 } while (!(c == '\n' || (c == 0 && s == bufend)));
1427 } else if (c == '\n') {
1429 PL_parser->linestart = s;
1434 } else if (isSPACE(c)) {
1436 } else if (c == 0 && s == bufend) {
1440 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1441 #endif /* PERL_MAD */
1442 PL_parser->bufptr = s;
1443 CopLINE_inc(PL_curcop);
1444 got_more = lex_next_chunk(flags);
1445 CopLINE_dec(PL_curcop);
1446 s = PL_parser->bufptr;
1447 bufend = PL_parser->bufend;
1450 if (need_incline && PL_parser->rsfp) {
1460 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1461 #endif /* PERL_MAD */
1462 PL_parser->bufptr = s;
1467 * This subroutine has nothing to do with tilting, whether at windmills
1468 * or pinball tables. Its name is short for "increment line". It
1469 * increments the current line number in CopLINE(PL_curcop) and checks
1470 * to see whether the line starts with a comment of the form
1471 * # line 500 "foo.pm"
1472 * If so, it sets the current line number and file to the values in the comment.
1476 S_incline(pTHX_ const char *s)
1483 PERL_ARGS_ASSERT_INCLINE;
1485 CopLINE_inc(PL_curcop);
1488 while (SPACE_OR_TAB(*s))
1490 if (strnEQ(s, "line", 4))
1494 if (SPACE_OR_TAB(*s))
1498 while (SPACE_OR_TAB(*s))
1506 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1508 while (SPACE_OR_TAB(*s))
1510 if (*s == '"' && (t = strchr(s+1, '"'))) {
1516 while (!isSPACE(*t))
1520 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1522 if (*e != '\n' && *e != '\0')
1523 return; /* false alarm */
1526 const STRLEN len = t - s;
1527 #ifndef USE_ITHREADS
1528 SV *const temp_sv = CopFILESV(PL_curcop);
1533 cf = SvPVX(temp_sv);
1534 tmplen = SvCUR(temp_sv);
1540 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1541 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1542 * to *{"::_<newfilename"} */
1543 /* However, the long form of evals is only turned on by the
1544 debugger - usually they're "(eval %lu)" */
1548 STRLEN tmplen2 = len;
1549 if (tmplen + 2 <= sizeof smallbuf)
1552 Newx(tmpbuf, tmplen + 2, char);
1555 memcpy(tmpbuf + 2, cf, tmplen);
1557 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1562 if (tmplen2 + 2 <= sizeof smallbuf)
1565 Newx(tmpbuf2, tmplen2 + 2, char);
1567 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1568 /* Either they malloc'd it, or we malloc'd it,
1569 so no prefix is present in ours. */
1574 memcpy(tmpbuf2 + 2, s, tmplen2);
1577 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1579 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1580 /* adjust ${"::_<newfilename"} to store the new file name */
1581 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1582 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1583 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1586 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1588 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1591 CopFILE_free(PL_curcop);
1592 CopFILE_setn(PL_curcop, s, len);
1594 CopLINE_set(PL_curcop, atoi(n)-1);
1598 /* skip space before PL_thistoken */
1601 S_skipspace0(pTHX_ register char *s)
1603 PERL_ARGS_ASSERT_SKIPSPACE0;
1610 PL_thiswhite = newSVpvs("");
1611 sv_catsv(PL_thiswhite, PL_skipwhite);
1612 sv_free(PL_skipwhite);
1615 PL_realtokenstart = s - SvPVX(PL_linestr);
1619 /* skip space after PL_thistoken */
1622 S_skipspace1(pTHX_ register char *s)
1624 const char *start = s;
1625 I32 startoff = start - SvPVX(PL_linestr);
1627 PERL_ARGS_ASSERT_SKIPSPACE1;
1632 start = SvPVX(PL_linestr) + startoff;
1633 if (!PL_thistoken && PL_realtokenstart >= 0) {
1634 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1635 PL_thistoken = newSVpvn(tstart, start - tstart);
1637 PL_realtokenstart = -1;
1640 PL_nextwhite = newSVpvs("");
1641 sv_catsv(PL_nextwhite, PL_skipwhite);
1642 sv_free(PL_skipwhite);
1649 S_skipspace2(pTHX_ register char *s, SV **svp)
1652 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1653 const I32 startoff = s - SvPVX(PL_linestr);
1655 PERL_ARGS_ASSERT_SKIPSPACE2;
1658 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1659 if (!PL_madskills || !svp)
1661 start = SvPVX(PL_linestr) + startoff;
1662 if (!PL_thistoken && PL_realtokenstart >= 0) {
1663 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1664 PL_thistoken = newSVpvn(tstart, start - tstart);
1665 PL_realtokenstart = -1;
1669 *svp = newSVpvs("");
1670 sv_setsv(*svp, PL_skipwhite);
1671 sv_free(PL_skipwhite);
1680 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1682 AV *av = CopFILEAVx(PL_curcop);
1684 SV * const sv = newSV_type(SVt_PVMG);
1686 sv_setsv(sv, orig_sv);
1688 sv_setpvn(sv, buf, len);
1691 av_store(av, (I32)CopLINE(PL_curcop), sv);
1697 * Called to gobble the appropriate amount and type of whitespace.
1698 * Skips comments as well.
1702 S_skipspace(pTHX_ register char *s)
1706 #endif /* PERL_MAD */
1707 PERL_ARGS_ASSERT_SKIPSPACE;
1710 sv_free(PL_skipwhite);
1711 PL_skipwhite = NULL;
1713 #endif /* PERL_MAD */
1714 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1715 while (s < PL_bufend && SPACE_OR_TAB(*s))
1717 } else if (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE) {
1718 while (isSPACE(*s) && *s != '\n')
1723 } while (s != PL_bufend && *s != '\n');
1728 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1730 lex_read_space(LEX_KEEP_PREVIOUS);
1732 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1733 if (PL_linestart > PL_bufptr)
1734 PL_bufptr = PL_linestart;
1739 PL_skipwhite = newSVpvn(start, s-start);
1740 #endif /* PERL_MAD */
1746 * Check the unary operators to ensure there's no ambiguity in how they're
1747 * used. An ambiguous piece of code would be:
1749 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1750 * the +5 is its argument.
1760 if (PL_oldoldbufptr != PL_last_uni)
1762 while (isSPACE(*PL_last_uni))
1765 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1767 if ((t = strchr(s, '(')) && t < PL_bufptr)
1770 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1771 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1772 (int)(s - PL_last_uni), PL_last_uni);
1776 * LOP : macro to build a list operator. Its behaviour has been replaced
1777 * with a subroutine, S_lop() for which LOP is just another name.
1780 #define LOP(f,x) return lop(f,x,s)
1784 * Build a list operator (or something that might be one). The rules:
1785 * - if we have a next token, then it's a list operator [why?]
1786 * - if the next thing is an opening paren, then it's a function
1787 * - else it's a list operator
1791 S_lop(pTHX_ I32 f, int x, char *s)
1795 PERL_ARGS_ASSERT_LOP;
1801 PL_last_lop = PL_oldbufptr;
1802 PL_last_lop_op = (OPCODE)f;
1805 return REPORT(LSTOP);
1808 return REPORT(LSTOP);
1811 return REPORT(FUNC);
1814 return REPORT(FUNC);
1816 return REPORT(LSTOP);
1822 * Sets up for an eventual force_next(). start_force(0) basically does
1823 * an unshift, while start_force(-1) does a push. yylex removes items
1828 S_start_force(pTHX_ int where)
1832 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1833 where = PL_lasttoke;
1834 assert(PL_curforce < 0 || PL_curforce == where);
1835 if (PL_curforce != where) {
1836 for (i = PL_lasttoke; i > where; --i) {
1837 PL_nexttoke[i] = PL_nexttoke[i-1];
1841 if (PL_curforce < 0) /* in case of duplicate start_force() */
1842 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1843 PL_curforce = where;
1846 curmad('^', newSVpvs(""));
1847 CURMAD('_', PL_nextwhite);
1852 S_curmad(pTHX_ char slot, SV *sv)
1858 if (PL_curforce < 0)
1859 where = &PL_thismad;
1861 where = &PL_nexttoke[PL_curforce].next_mad;
1867 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1869 else if (PL_encoding) {
1870 sv_recode_to_utf8(sv, PL_encoding);
1875 /* keep a slot open for the head of the list? */
1876 if (slot != '_' && *where && (*where)->mad_key == '^') {
1877 (*where)->mad_key = slot;
1878 sv_free(MUTABLE_SV(((*where)->mad_val)));
1879 (*where)->mad_val = (void*)sv;
1882 addmad(newMADsv(slot, sv), where, 0);
1885 # define start_force(where) NOOP
1886 # define curmad(slot, sv) NOOP
1891 * When the lexer realizes it knows the next token (for instance,
1892 * it is reordering tokens for the parser) then it can call S_force_next
1893 * to know what token to return the next time the lexer is called. Caller
1894 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1895 * and possibly PL_expect to ensure the lexer handles the token correctly.
1899 S_force_next(pTHX_ I32 type)
1904 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1905 tokereport(type, &NEXTVAL_NEXTTOKE);
1909 if (PL_curforce < 0)
1910 start_force(PL_lasttoke);
1911 PL_nexttoke[PL_curforce].next_type = type;
1912 if (PL_lex_state != LEX_KNOWNEXT)
1913 PL_lex_defer = PL_lex_state;
1914 PL_lex_state = LEX_KNOWNEXT;
1915 PL_lex_expect = PL_expect;
1918 PL_nexttype[PL_nexttoke] = type;
1920 if (PL_lex_state != LEX_KNOWNEXT) {
1921 PL_lex_defer = PL_lex_state;
1922 PL_lex_expect = PL_expect;
1923 PL_lex_state = LEX_KNOWNEXT;
1929 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1932 SV * const sv = newSVpvn_utf8(start, len,
1935 && !is_ascii_string((const U8*)start, len)
1936 && is_utf8_string((const U8*)start, len));
1942 * When the lexer knows the next thing is a word (for instance, it has
1943 * just seen -> and it knows that the next char is a word char, then
1944 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1948 * char *start : buffer position (must be within PL_linestr)
1949 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1950 * int check_keyword : if true, Perl checks to make sure the word isn't
1951 * a keyword (do this if the word is a label, e.g. goto FOO)
1952 * int allow_pack : if true, : characters will also be allowed (require,
1953 * use, etc. do this)
1954 * int allow_initial_tick : used by the "sub" lexer only.
1958 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1964 PERL_ARGS_ASSERT_FORCE_WORD;
1966 start = SKIPSPACE1(start);
1968 if (isIDFIRST_lazy_if(s,UTF) ||
1969 (allow_pack && *s == ':') ||
1970 (allow_initial_tick && *s == '\'') )
1972 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1973 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1975 start_force(PL_curforce);
1977 curmad('X', newSVpvn(start,s-start));
1978 if (token == METHOD) {
1983 PL_expect = XOPERATOR;
1987 curmad('g', newSVpvs( "forced" ));
1988 NEXTVAL_NEXTTOKE.opval
1989 = (OP*)newSVOP(OP_CONST,0,
1990 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1991 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1999 * Called when the lexer wants $foo *foo &foo etc, but the program
2000 * text only contains the "foo" portion. The first argument is a pointer
2001 * to the "foo", and the second argument is the type symbol to prefix.
2002 * Forces the next token to be a "WORD".
2003 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2007 S_force_ident(pTHX_ register const char *s, int kind)
2011 PERL_ARGS_ASSERT_FORCE_IDENT;
2014 const STRLEN len = strlen(s);
2015 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2016 start_force(PL_curforce);
2017 NEXTVAL_NEXTTOKE.opval = o;
2020 o->op_private = OPpCONST_ENTERED;
2021 /* XXX see note in pp_entereval() for why we forgo typo
2022 warnings if the symbol must be introduced in an eval.
2024 gv_fetchpvn_flags(s, len,
2025 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2027 kind == '$' ? SVt_PV :
2028 kind == '@' ? SVt_PVAV :
2029 kind == '%' ? SVt_PVHV :
2037 Perl_str_to_version(pTHX_ SV *sv)
2042 const char *start = SvPV_const(sv,len);
2043 const char * const end = start + len;
2044 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2046 PERL_ARGS_ASSERT_STR_TO_VERSION;
2048 while (start < end) {
2052 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2057 retval += ((NV)n)/nshift;
2066 * Forces the next token to be a version number.
2067 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2068 * and if "guessing" is TRUE, then no new token is created (and the caller
2069 * must use an alternative parsing method).
2073 S_force_version(pTHX_ char *s, int guessing)
2079 I32 startoff = s - SvPVX(PL_linestr);
2082 PERL_ARGS_ASSERT_FORCE_VERSION;
2090 while (isDIGIT(*d) || *d == '_' || *d == '.')
2094 start_force(PL_curforce);
2095 curmad('X', newSVpvn(s,d-s));
2098 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
2100 s = scan_num(s, &pl_yylval);
2101 version = pl_yylval.opval;
2102 ver = cSVOPx(version)->op_sv;
2103 if (SvPOK(ver) && !SvNIOK(ver)) {
2104 SvUPGRADE(ver, SVt_PVNV);
2105 SvNV_set(ver, str_to_version(ver));
2106 SvNOK_on(ver); /* hint that it is a version */
2109 else if (guessing) {
2112 sv_free(PL_nextwhite); /* let next token collect whitespace */
2114 s = SvPVX(PL_linestr) + startoff;
2122 if (PL_madskills && !version) {
2123 sv_free(PL_nextwhite); /* let next token collect whitespace */
2125 s = SvPVX(PL_linestr) + startoff;
2128 /* NOTE: The parser sees the package name and the VERSION swapped */
2129 start_force(PL_curforce);
2130 NEXTVAL_NEXTTOKE.opval = version;
2138 * Tokenize a quoted string passed in as an SV. It finds the next
2139 * chunk, up to end of string or a backslash. It may make a new
2140 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2145 S_tokeq(pTHX_ SV *sv)
2149 register char *send;
2154 PERL_ARGS_ASSERT_TOKEQ;
2159 s = SvPV_force(sv, len);
2160 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2163 while (s < send && *s != '\\')
2168 if ( PL_hints & HINT_NEW_STRING ) {
2169 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2173 if (s + 1 < send && (s[1] == '\\'))
2174 s++; /* all that, just for this */
2179 SvCUR_set(sv, d - SvPVX_const(sv));
2181 if ( PL_hints & HINT_NEW_STRING )
2182 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2187 * Now come three functions related to double-quote context,
2188 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2189 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2190 * interact with PL_lex_state, and create fake ( ... ) argument lists
2191 * to handle functions and concatenation.
2192 * They assume that whoever calls them will be setting up a fake
2193 * join call, because each subthing puts a ',' after it. This lets
2196 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2198 * (I'm not sure whether the spurious commas at the end of lcfirst's
2199 * arguments and join's arguments are created or not).
2204 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2206 * Pattern matching will set PL_lex_op to the pattern-matching op to
2207 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2209 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2211 * Everything else becomes a FUNC.
2213 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2214 * had an OP_CONST or OP_READLINE). This just sets us up for a
2215 * call to S_sublex_push().
2219 S_sublex_start(pTHX)
2222 register const I32 op_type = pl_yylval.ival;
2224 if (op_type == OP_NULL) {
2225 pl_yylval.opval = PL_lex_op;
2229 if (op_type == OP_CONST || op_type == OP_READLINE) {
2230 SV *sv = tokeq(PL_lex_stuff);
2232 if (SvTYPE(sv) == SVt_PVIV) {
2233 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2235 const char * const p = SvPV_const(sv, len);
2236 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2240 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2241 PL_lex_stuff = NULL;
2242 /* Allow <FH> // "foo" */
2243 if (op_type == OP_READLINE)
2244 PL_expect = XTERMORDORDOR;
2247 else if (op_type == OP_BACKTICK && PL_lex_op) {
2248 /* readpipe() vas overriden */
2249 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2250 pl_yylval.opval = PL_lex_op;
2252 PL_lex_stuff = NULL;
2256 PL_sublex_info.super_state = PL_lex_state;
2257 PL_sublex_info.sub_inwhat = (U16)op_type;
2258 PL_sublex_info.sub_op = PL_lex_op;
2259 PL_lex_state = LEX_INTERPPUSH;
2263 pl_yylval.opval = PL_lex_op;
2273 * Create a new scope to save the lexing state. The scope will be
2274 * ended in S_sublex_done. Returns a '(', starting the function arguments
2275 * to the uc, lc, etc. found before.
2276 * Sets PL_lex_state to LEX_INTERPCONCAT.
2285 PL_lex_state = PL_sublex_info.super_state;
2286 SAVEBOOL(PL_lex_dojoin);
2287 SAVEI32(PL_lex_brackets);
2288 SAVEI32(PL_lex_casemods);
2289 SAVEI32(PL_lex_starts);
2290 SAVEI8(PL_lex_state);
2291 SAVEVPTR(PL_lex_inpat);
2292 SAVEI16(PL_lex_inwhat);
2293 SAVECOPLINE(PL_curcop);
2294 SAVEPPTR(PL_bufptr);
2295 SAVEPPTR(PL_bufend);
2296 SAVEPPTR(PL_oldbufptr);
2297 SAVEPPTR(PL_oldoldbufptr);
2298 SAVEPPTR(PL_last_lop);
2299 SAVEPPTR(PL_last_uni);
2300 SAVEPPTR(PL_linestart);
2301 SAVESPTR(PL_linestr);
2302 SAVEGENERICPV(PL_lex_brackstack);
2303 SAVEGENERICPV(PL_lex_casestack);
2305 PL_linestr = PL_lex_stuff;
2306 PL_lex_stuff = NULL;
2308 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2309 = SvPVX(PL_linestr);
2310 PL_bufend += SvCUR(PL_linestr);
2311 PL_last_lop = PL_last_uni = NULL;
2312 SAVEFREESV(PL_linestr);
2314 PL_lex_dojoin = FALSE;
2315 PL_lex_brackets = 0;
2316 Newx(PL_lex_brackstack, 120, char);
2317 Newx(PL_lex_casestack, 12, char);
2318 PL_lex_casemods = 0;
2319 *PL_lex_casestack = '\0';
2321 PL_lex_state = LEX_INTERPCONCAT;
2322 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2324 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2325 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2326 PL_lex_inpat = PL_sublex_info.sub_op;
2328 PL_lex_inpat = NULL;
2335 * Restores lexer state after a S_sublex_push.
2342 if (!PL_lex_starts++) {
2343 SV * const sv = newSVpvs("");
2344 if (SvUTF8(PL_linestr))
2346 PL_expect = XOPERATOR;
2347 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2351 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2352 PL_lex_state = LEX_INTERPCASEMOD;
2356 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2357 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2358 PL_linestr = PL_lex_repl;
2360 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2361 PL_bufend += SvCUR(PL_linestr);
2362 PL_last_lop = PL_last_uni = NULL;
2363 SAVEFREESV(PL_linestr);
2364 PL_lex_dojoin = FALSE;
2365 PL_lex_brackets = 0;
2366 PL_lex_casemods = 0;
2367 *PL_lex_casestack = '\0';
2369 if (SvEVALED(PL_lex_repl)) {
2370 PL_lex_state = LEX_INTERPNORMAL;
2372 /* we don't clear PL_lex_repl here, so that we can check later
2373 whether this is an evalled subst; that means we rely on the
2374 logic to ensure sublex_done() is called again only via the
2375 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2378 PL_lex_state = LEX_INTERPCONCAT;
2388 PL_endwhite = newSVpvs("");
2389 sv_catsv(PL_endwhite, PL_thiswhite);
2393 sv_setpvs(PL_thistoken,"");
2395 PL_realtokenstart = -1;
2399 PL_bufend = SvPVX(PL_linestr);
2400 PL_bufend += SvCUR(PL_linestr);
2401 PL_expect = XOPERATOR;
2402 PL_sublex_info.sub_inwhat = 0;
2410 Extracts a pattern, double-quoted string, or transliteration. This
2413 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2414 processing a pattern (PL_lex_inpat is true), a transliteration
2415 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2417 Returns a pointer to the character scanned up to. If this is
2418 advanced from the start pointer supplied (i.e. if anything was
2419 successfully parsed), will leave an OP for the substring scanned
2420 in pl_yylval. Caller must intuit reason for not parsing further
2421 by looking at the next characters herself.
2425 double-quoted style: \r and \n
2426 regexp special ones: \D \s
2429 case and quoting: \U \Q \E
2430 stops on @ and $, but not for $ as tail anchor
2432 In transliterations:
2433 characters are VERY literal, except for - not at the start or end
2434 of the string, which indicates a range. If the range is in bytes,
2435 scan_const expands the range to the full set of intermediate
2436 characters. If the range is in utf8, the hyphen is replaced with
2437 a certain range mark which will be handled by pmtrans() in op.c.
2439 In double-quoted strings:
2441 double-quoted style: \r and \n
2443 deprecated backrefs: \1 (in substitution replacements)
2444 case and quoting: \U \Q \E
2447 scan_const does *not* construct ops to handle interpolated strings.
2448 It stops processing as soon as it finds an embedded $ or @ variable
2449 and leaves it to the caller to work out what's going on.
2451 embedded arrays (whether in pattern or not) could be:
2452 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2454 $ in double-quoted strings must be the symbol of an embedded scalar.
2456 $ in pattern could be $foo or could be tail anchor. Assumption:
2457 it's a tail anchor if $ is the last thing in the string, or if it's
2458 followed by one of "()| \r\n\t"
2460 \1 (backreferences) are turned into $1
2462 The structure of the code is
2463 while (there's a character to process) {
2464 handle transliteration ranges
2465 skip regexp comments /(?#comment)/ and codes /(?{code})/
2466 skip #-initiated comments in //x patterns
2467 check for embedded arrays
2468 check for embedded scalars
2470 leave intact backslashes from leaveit (below)
2471 deprecate \1 in substitution replacements
2472 handle string-changing backslashes \l \U \Q \E, etc.
2473 switch (what was escaped) {
2474 handle \- in a transliteration (becomes a literal -)
2475 handle \132 (octal characters)
2476 handle \x15 and \x{1234} (hex characters)
2477 handle \N{name} (named characters)
2478 handle \cV (control characters)
2479 handle printf-style backslashes (\f, \r, \n, etc)
2482 } (end if backslash)
2483 handle regular character
2484 } (end while character to read)
2489 S_scan_const(pTHX_ char *start)
2492 register char *send = PL_bufend; /* end of the constant */
2493 SV *sv = newSV(send - start); /* sv for the constant. See
2494 note below on sizing. */
2495 register char *s = start; /* start of the constant */
2496 register char *d = SvPVX(sv); /* destination for copies */
2497 bool dorange = FALSE; /* are we in a translit range? */
2498 bool didrange = FALSE; /* did we just finish a range? */
2499 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2500 I32 this_utf8 = UTF; /* Is the source string assumed
2501 to be UTF8? But, this can
2502 show as true when the source
2503 isn't utf8, as for example
2504 when it is entirely composed
2507 /* Note on sizing: The scanned constant is placed into sv, which is
2508 * initialized by newSV() assuming one byte of output for every byte of
2509 * input. This routine expects newSV() to allocate an extra byte for a
2510 * trailing NUL, which this routine will append if it gets to the end of
2511 * the input. There may be more bytes of input than output (eg., \N{LATIN
2512 * CAPITAL LETTER A}), or more output than input if the constant ends up
2513 * recoded to utf8, but each time a construct is found that might increase
2514 * the needed size, SvGROW() is called. Its size parameter each time is
2515 * based on the best guess estimate at the time, namely the length used so
2516 * far, plus the length the current construct will occupy, plus room for
2517 * the trailing NUL, plus one byte for every input byte still unscanned */
2521 UV literal_endpoint = 0;
2522 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2525 PERL_ARGS_ASSERT_SCAN_CONST;
2527 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2528 /* If we are doing a trans and we know we want UTF8 set expectation */
2529 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2530 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2534 while (s < send || dorange) {
2535 /* get transliterations out of the way (they're most literal) */
2536 if (PL_lex_inwhat == OP_TRANS) {
2537 /* expand a range A-Z to the full set of characters. AIE! */
2539 I32 i; /* current expanded character */
2540 I32 min; /* first character in range */
2541 I32 max; /* last character in range */
2552 char * const c = (char*)utf8_hop((U8*)d, -1);
2556 *c = (char)UTF_TO_NATIVE(0xff);
2557 /* mark the range as done, and continue */
2563 i = d - SvPVX_const(sv); /* remember current offset */
2566 SvLEN(sv) + (has_utf8 ?
2567 (512 - UTF_CONTINUATION_MARK +
2570 /* How many two-byte within 0..255: 128 in UTF-8,
2571 * 96 in UTF-8-mod. */
2573 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2575 d = SvPVX(sv) + i; /* refresh d after realloc */
2579 for (j = 0; j <= 1; j++) {
2580 char * const c = (char*)utf8_hop((U8*)d, -1);
2581 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2587 max = (U8)0xff; /* only to \xff */
2588 uvmax = uv; /* \x{100} to uvmax */
2590 d = c; /* eat endpoint chars */
2595 d -= 2; /* eat the first char and the - */
2596 min = (U8)*d; /* first char in range */
2597 max = (U8)d[1]; /* last char in range */
2604 "Invalid range \"%c-%c\" in transliteration operator",
2605 (char)min, (char)max);
2609 if (literal_endpoint == 2 &&
2610 ((isLOWER(min) && isLOWER(max)) ||
2611 (isUPPER(min) && isUPPER(max)))) {
2613 for (i = min; i <= max; i++)
2615 *d++ = NATIVE_TO_NEED(has_utf8,i);
2617 for (i = min; i <= max; i++)
2619 *d++ = NATIVE_TO_NEED(has_utf8,i);
2624 for (i = min; i <= max; i++)
2627 const U8 ch = (U8)NATIVE_TO_UTF(i);
2628 if (UNI_IS_INVARIANT(ch))
2631 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2632 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2641 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2643 *d++ = (char)UTF_TO_NATIVE(0xff);
2645 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2649 /* mark the range as done, and continue */
2653 literal_endpoint = 0;
2658 /* range begins (ignore - as first or last char) */
2659 else if (*s == '-' && s+1 < send && s != start) {
2661 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2668 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2678 literal_endpoint = 0;
2679 native_range = TRUE;
2684 /* if we get here, we're not doing a transliteration */
2686 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2687 except for the last char, which will be done separately. */
2688 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2690 while (s+1 < send && *s != ')')
2691 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2693 else if (s[2] == '{' /* This should match regcomp.c */
2694 || (s[2] == '?' && s[3] == '{'))
2697 char *regparse = s + (s[2] == '{' ? 3 : 4);
2700 while (count && (c = *regparse)) {
2701 if (c == '\\' && regparse[1])
2709 if (*regparse != ')')
2710 regparse--; /* Leave one char for continuation. */
2711 while (s < regparse)
2712 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2716 /* likewise skip #-initiated comments in //x patterns */
2717 else if (*s == '#' && PL_lex_inpat &&
2718 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2719 while (s+1 < send && *s != '\n')
2720 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2723 /* check for embedded arrays
2724 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2726 else if (*s == '@' && s[1]) {
2727 if (isALNUM_lazy_if(s+1,UTF))
2729 if (strchr(":'{$", s[1]))
2731 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2732 break; /* in regexp, neither @+ nor @- are interpolated */
2735 /* check for embedded scalars. only stop if we're sure it's a
2738 else if (*s == '$') {
2739 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2741 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2743 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2744 "Possible unintended interpolation of $\\ in regex");
2746 break; /* in regexp, $ might be tail anchor */
2750 /* End of else if chain - OP_TRANS rejoin rest */
2753 if (*s == '\\' && s+1 < send) {
2756 /* deprecate \1 in strings and substitution replacements */
2757 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2758 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2760 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2765 /* string-change backslash escapes */
2766 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2770 /* skip any other backslash escapes in a pattern */
2771 else if (PL_lex_inpat) {
2772 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2773 goto default_action;
2776 /* if we get here, it's either a quoted -, or a digit */
2779 /* quoted - in transliterations */
2781 if (PL_lex_inwhat == OP_TRANS) {
2788 if ((isALPHA(*s) || isDIGIT(*s)))
2789 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2790 "Unrecognized escape \\%c passed through",
2792 /* default action is to copy the quoted character */
2793 goto default_action;
2796 /* eg. \132 indicates the octal constant 0x132 */
2797 case '0': case '1': case '2': case '3':
2798 case '4': case '5': case '6': case '7':
2802 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2805 goto NUM_ESCAPE_INSERT;
2807 /* eg. \x24 indicates the hex constant 0x24 */
2811 char* const e = strchr(s, '}');
2812 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2813 PERL_SCAN_DISALLOW_PREFIX;
2818 yyerror("Missing right brace on \\x{}");
2822 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2828 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2829 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2835 /* Insert oct, hex, or \N{U+...} escaped character. There will
2836 * always be enough room in sv since such escapes will be
2837 * longer than any UTF-8 sequence they can end up as, except if
2838 * they force us to recode the rest of the string into utf8 */
2840 /* Here uv is the ordinal of the next character being added in
2841 * unicode (converted from native). (It has to be done before
2842 * here because \N is interpreted as unicode, and oct and hex
2844 if (!UNI_IS_INVARIANT(uv)) {
2845 if (!has_utf8 && uv > 255) {
2846 /* Might need to recode whatever we have accumulated so
2847 * far if it contains any chars variant in utf8 or
2850 SvCUR_set(sv, d - SvPVX_const(sv));
2853 /* See Note on sizing above. */
2854 sv_utf8_upgrade_flags_grow(sv,
2855 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2856 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2857 d = SvPVX(sv) + SvCUR(sv);
2862 d = (char*)uvuni_to_utf8((U8*)d, uv);
2863 if (PL_lex_inwhat == OP_TRANS &&
2864 PL_sublex_info.sub_op) {
2865 PL_sublex_info.sub_op->op_private |=
2866 (PL_lex_repl ? OPpTRANS_FROM_UTF
2870 if (uv > 255 && !dorange)
2871 native_range = FALSE;
2883 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2888 char* e = strchr(s, '}');
2894 yyerror("Missing right brace on \\N{}");
2898 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2899 /* \N{U+...} The ... is a unicode value even on EBCDIC
2901 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2902 PERL_SCAN_DISALLOW_PREFIX;
2905 uv = grok_hex(s, &len, &flags, NULL);
2906 if ( e > s && len != (STRLEN)(e - s) ) {
2910 goto NUM_ESCAPE_INSERT;
2912 res = newSVpvn(s + 1, e - s - 1);
2913 res = new_constant( NULL, 0, "charnames",
2914 res, NULL, s - 2, e - s + 3 );
2916 sv_utf8_upgrade(res);
2917 str = SvPV_const(res,len);
2918 #ifdef EBCDIC_NEVER_MIND
2919 /* charnames uses pack U and that has been
2920 * recently changed to do the below uni->native
2921 * mapping, so this would be redundant (and wrong,
2922 * the code point would be doubly converted).
2923 * But leave this in just in case the pack U change
2924 * gets revoked, but the semantics is still
2925 * desireable for charnames. --jhi */
2927 UV uv = utf8_to_uvchr((const U8*)str, 0);
2930 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2932 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2933 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2934 str = SvPV_const(res, len);
2938 /* If destination is not in utf8 but this new character is,
2939 * recode the dest to utf8 */
2940 if (!has_utf8 && SvUTF8(res)) {
2941 SvCUR_set(sv, d - SvPVX_const(sv));
2944 /* See Note on sizing above. */
2945 sv_utf8_upgrade_flags_grow(sv,
2946 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2947 len + (STRLEN)(send - s) + 1);
2948 d = SvPVX(sv) + SvCUR(sv);
2950 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2952 /* See Note on sizing above. (NOTE: SvCUR() is not set
2953 * correctly here). */
2954 const STRLEN off = d - SvPVX_const(sv);
2955 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2959 native_range = FALSE; /* \N{} is guessed to be Unicode */
2961 Copy(str, d, len, char);
2968 yyerror("Missing braces on \\N{}");
2971 /* \c is a control character */
2980 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2983 yyerror("Missing control char name in \\c");
2987 /* printf-style backslashes, formfeeds, newlines, etc */
2989 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2992 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2995 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2998 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3001 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3004 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3007 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3013 } /* end if (backslash) */
3020 /* If we started with encoded form, or already know we want it,
3021 then encode the next character */
3022 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3026 /* One might think that it is wasted effort in the case of the
3027 * source being utf8 (this_utf8 == TRUE) to take the next character
3028 * in the source, convert it to an unsigned value, and then convert
3029 * it back again. But the source has not been validated here. The
3030 * routine that does the conversion checks for errors like
3033 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3034 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3036 SvCUR_set(sv, d - SvPVX_const(sv));
3039 /* See Note on sizing above. */
3040 sv_utf8_upgrade_flags_grow(sv,
3041 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3042 need + (STRLEN)(send - s) + 1);
3043 d = SvPVX(sv) + SvCUR(sv);
3045 } else if (need > len) {
3046 /* encoded value larger than old, may need extra space (NOTE:
3047 * SvCUR() is not set correctly here). See Note on sizing
3049 const STRLEN off = d - SvPVX_const(sv);
3050 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3054 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3056 if (uv > 255 && !dorange)
3057 native_range = FALSE;
3061 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3063 } /* while loop to process each character */
3065 /* terminate the string and set up the sv */
3067 SvCUR_set(sv, d - SvPVX_const(sv));
3068 if (SvCUR(sv) >= SvLEN(sv))
3069 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3072 if (PL_encoding && !has_utf8) {
3073 sv_recode_to_utf8(sv, PL_encoding);
3079 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3080 PL_sublex_info.sub_op->op_private |=
3081 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3085 /* shrink the sv if we allocated more than we used */
3086 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3087 SvPV_shrink_to_cur(sv);
3090 /* return the substring (via pl_yylval) only if we parsed anything */
3091 if (s > PL_bufptr) {
3092 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3093 const char *const key = PL_lex_inpat ? "qr" : "q";
3094 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3098 if (PL_lex_inwhat == OP_TRANS) {
3101 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3109 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3112 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3119 * Returns TRUE if there's more to the expression (e.g., a subscript),
3122 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3124 * ->[ and ->{ return TRUE
3125 * { and [ outside a pattern are always subscripts, so return TRUE
3126 * if we're outside a pattern and it's not { or [, then return FALSE
3127 * if we're in a pattern and the first char is a {
3128 * {4,5} (any digits around the comma) returns FALSE
3129 * if we're in a pattern and the first char is a [
3131 * [SOMETHING] has a funky algorithm to decide whether it's a
3132 * character class or not. It has to deal with things like
3133 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3134 * anything else returns TRUE
3137 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3140 S_intuit_more(pTHX_ register char *s)
3144 PERL_ARGS_ASSERT_INTUIT_MORE;
3146 if (PL_lex_brackets)
3148 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3150 if (*s != '{' && *s != '[')
3155 /* In a pattern, so maybe we have {n,m}. */
3172 /* On the other hand, maybe we have a character class */
3175 if (*s == ']' || *s == '^')
3178 /* this is terrifying, and it works */
3179 int weight = 2; /* let's weigh the evidence */
3181 unsigned char un_char = 255, last_un_char;
3182 const char * const send = strchr(s,']');
3183 char tmpbuf[sizeof PL_tokenbuf * 4];
3185 if (!send) /* has to be an expression */
3188 Zero(seen,256,char);
3191 else if (isDIGIT(*s)) {
3193 if (isDIGIT(s[1]) && s[2] == ']')
3199 for (; s < send; s++) {
3200 last_un_char = un_char;
3201 un_char = (unsigned char)*s;
3206 weight -= seen[un_char] * 10;
3207 if (isALNUM_lazy_if(s+1,UTF)) {
3209 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3210 len = (int)strlen(tmpbuf);
3211 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3216 else if (*s == '$' && s[1] &&
3217 strchr("[#!%*<>()-=",s[1])) {
3218 if (/*{*/ strchr("])} =",s[2]))
3227 if (strchr("wds]",s[1]))
3229 else if (seen[(U8)'\''] || seen[(U8)'"'])
3231 else if (strchr("rnftbxcav",s[1]))
3233 else if (isDIGIT(s[1])) {
3235 while (s[1] && isDIGIT(s[1]))
3245 if (strchr("aA01! ",last_un_char))
3247 if (strchr("zZ79~",s[1]))
3249 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3250 weight -= 5; /* cope with negative subscript */
3253 if (!isALNUM(last_un_char)
3254 && !(last_un_char == '$' || last_un_char == '@'
3255 || last_un_char == '&')
3256 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3261 if (keyword(tmpbuf, d - tmpbuf, 0))
3264 if (un_char == last_un_char + 1)
3266 weight -= seen[un_char];
3271 if (weight >= 0) /* probably a character class */
3281 * Does all the checking to disambiguate
3283 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3284 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3286 * First argument is the stuff after the first token, e.g. "bar".
3288 * Not a method if bar is a filehandle.
3289 * Not a method if foo is a subroutine prototyped to take a filehandle.
3290 * Not a method if it's really "Foo $bar"
3291 * Method if it's "foo $bar"
3292 * Not a method if it's really "print foo $bar"
3293 * Method if it's really "foo package::" (interpreted as package->foo)
3294 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3295 * Not a method if bar is a filehandle or package, but is quoted with
3300 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3303 char *s = start + (*start == '$');
3304 char tmpbuf[sizeof PL_tokenbuf];
3311 PERL_ARGS_ASSERT_INTUIT_METHOD;
3314 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3318 const char *proto = SvPVX_const(cv);
3329 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3330 /* start is the beginning of the possible filehandle/object,
3331 * and s is the end of it
3332 * tmpbuf is a copy of it
3335 if (*start == '$') {
3336 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3337 isUPPER(*PL_tokenbuf))
3340 len = start - SvPVX(PL_linestr);
3344 start = SvPVX(PL_linestr) + len;
3348 return *s == '(' ? FUNCMETH : METHOD;
3350 if (!keyword(tmpbuf, len, 0)) {
3351 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3355 soff = s - SvPVX(PL_linestr);
3359 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3360 if (indirgv && GvCVu(indirgv))
3362 /* filehandle or package name makes it a method */
3363 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3365 soff = s - SvPVX(PL_linestr);
3368 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3369 return 0; /* no assumptions -- "=>" quotes bearword */
3371 start_force(PL_curforce);
3372 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3373 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3374 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3376 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3381 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3383 return *s == '(' ? FUNCMETH : METHOD;
3389 /* Encoded script support. filter_add() effectively inserts a
3390 * 'pre-processing' function into the current source input stream.
3391 * Note that the filter function only applies to the current source file
3392 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3394 * The datasv parameter (which may be NULL) can be used to pass
3395 * private data to this instance of the filter. The filter function
3396 * can recover the SV using the FILTER_DATA macro and use it to
3397 * store private buffers and state information.
3399 * The supplied datasv parameter is upgraded to a PVIO type
3400 * and the IoDIRP/IoANY field is used to store the function pointer,
3401 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3402 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3403 * private use must be set using malloc'd pointers.
3407 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3416 if (!PL_rsfp_filters)
3417 PL_rsfp_filters = newAV();
3420 SvUPGRADE(datasv, SVt_PVIO);
3421 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3422 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3423 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3424 FPTR2DPTR(void *, IoANY(datasv)),
3425 SvPV_nolen(datasv)));
3426 av_unshift(PL_rsfp_filters, 1);
3427 av_store(PL_rsfp_filters, 0, datasv) ;
3432 /* Delete most recently added instance of this filter function. */
3434 Perl_filter_del(pTHX_ filter_t funcp)
3439 PERL_ARGS_ASSERT_FILTER_DEL;
3442 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3443 FPTR2DPTR(void*, funcp)));
3445 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3447 /* if filter is on top of stack (usual case) just pop it off */
3448 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3449 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3450 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3451 IoANY(datasv) = (void *)NULL;
3452 sv_free(av_pop(PL_rsfp_filters));
3456 /* we need to search for the correct entry and clear it */
3457 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3461 /* Invoke the idxth filter function for the current rsfp. */
3462 /* maxlen 0 = read one text line */
3464 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3469 /* This API is bad. It should have been using unsigned int for maxlen.
3470 Not sure if we want to change the API, but if not we should sanity
3471 check the value here. */
3472 const unsigned int correct_length
3481 PERL_ARGS_ASSERT_FILTER_READ;
3483 if (!PL_parser || !PL_rsfp_filters)
3485 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3486 /* Provide a default input filter to make life easy. */
3487 /* Note that we append to the line. This is handy. */
3488 DEBUG_P(PerlIO_printf(Perl_debug_log,
3489 "filter_read %d: from rsfp\n", idx));
3490 if (correct_length) {
3493 const int old_len = SvCUR(buf_sv);
3495 /* ensure buf_sv is large enough */
3496 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3497 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3498 correct_length)) <= 0) {
3499 if (PerlIO_error(PL_rsfp))
3500 return -1; /* error */
3502 return 0 ; /* end of file */
3504 SvCUR_set(buf_sv, old_len + len) ;
3505 SvPVX(buf_sv)[old_len + len] = '\0';
3508 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3509 if (PerlIO_error(PL_rsfp))
3510 return -1; /* error */
3512 return 0 ; /* end of file */
3515 return SvCUR(buf_sv);
3517 /* Skip this filter slot if filter has been deleted */
3518 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3519 DEBUG_P(PerlIO_printf(Perl_debug_log,
3520 "filter_read %d: skipped (filter deleted)\n",
3522 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3524 /* Get function pointer hidden within datasv */
3525 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3526 DEBUG_P(PerlIO_printf(Perl_debug_log,
3527 "filter_read %d: via function %p (%s)\n",
3528 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3529 /* Call function. The function is expected to */
3530 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3531 /* Return: <0:error, =0:eof, >0:not eof */
3532 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3536 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3540 PERL_ARGS_ASSERT_FILTER_GETS;
3542 #ifdef PERL_CR_FILTER
3543 if (!PL_rsfp_filters) {
3544 filter_add(S_cr_textfilter,NULL);
3547 if (PL_rsfp_filters) {
3549 SvCUR_set(sv, 0); /* start with empty line */
3550 if (FILTER_READ(0, sv, 0) > 0)
3551 return ( SvPVX(sv) ) ;
3556 return (sv_gets(sv, PL_rsfp, append));
3560 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3565 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3567 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3571 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3572 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3574 return GvHV(gv); /* Foo:: */
3577 /* use constant CLASS => 'MyClass' */
3578 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3579 if (gv && GvCV(gv)) {
3580 SV * const sv = cv_const_sv(GvCV(gv));
3582 pkgname = SvPV_const(sv, len);
3585 return gv_stashpvn(pkgname, len, 0);
3589 * S_readpipe_override
3590 * Check whether readpipe() is overriden, and generates the appropriate
3591 * optree, provided sublex_start() is called afterwards.
3594 S_readpipe_override(pTHX)
3597 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3598 pl_yylval.ival = OP_BACKTICK;
3600 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3602 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3603 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3604 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3606 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3607 append_elem(OP_LIST,
3608 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3609 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3616 * The intent of this yylex wrapper is to minimize the changes to the
3617 * tokener when we aren't interested in collecting madprops. It remains
3618 * to be seen how successful this strategy will be...
3625 char *s = PL_bufptr;
3627 /* make sure PL_thiswhite is initialized */
3631 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3632 if (PL_pending_ident)
3633 return S_pending_ident(aTHX);
3635 /* previous token ate up our whitespace? */
3636 if (!PL_lasttoke && PL_nextwhite) {
3637 PL_thiswhite = PL_nextwhite;
3641 /* isolate the token, and figure out where it is without whitespace */
3642 PL_realtokenstart = -1;
3646 assert(PL_curforce < 0);
3648 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3649 if (!PL_thistoken) {
3650 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3651 PL_thistoken = newSVpvs("");
3653 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3654 PL_thistoken = newSVpvn(tstart, s - tstart);
3657 if (PL_thismad) /* install head */
3658 CURMAD('X', PL_thistoken);
3661 /* last whitespace of a sublex? */
3662 if (optype == ')' && PL_endwhite) {
3663 CURMAD('X', PL_endwhite);
3668 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3669 if (!PL_thiswhite && !PL_endwhite && !optype) {
3670 sv_free(PL_thistoken);
3675 /* put off final whitespace till peg */
3676 if (optype == ';' && !PL_rsfp) {
3677 PL_nextwhite = PL_thiswhite;
3680 else if (PL_thisopen) {
3681 CURMAD('q', PL_thisopen);
3683 sv_free(PL_thistoken);
3687 /* Store actual token text as madprop X */
3688 CURMAD('X', PL_thistoken);
3692 /* add preceding whitespace as madprop _ */
3693 CURMAD('_', PL_thiswhite);
3697 /* add quoted material as madprop = */
3698 CURMAD('=', PL_thisstuff);
3702 /* add terminating quote as madprop Q */
3703 CURMAD('Q', PL_thisclose);
3707 /* special processing based on optype */
3711 /* opval doesn't need a TOKEN since it can already store mp */
3721 if (pl_yylval.opval)
3722 append_madprops(PL_thismad, pl_yylval.opval, 0);
3730 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3739 /* remember any fake bracket that lexer is about to discard */
3740 if (PL_lex_brackets == 1 &&
3741 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3744 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3747 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3748 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3751 break; /* don't bother looking for trailing comment */
3760 /* attach a trailing comment to its statement instead of next token */
3764 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3766 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3768 if (*s == '\n' || *s == '#') {
3769 while (s < PL_bufend && *s != '\n')
3773 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3774 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3791 /* Create new token struct. Note: opvals return early above. */
3792 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3799 S_tokenize_use(pTHX_ int is_use, char *s) {
3802 PERL_ARGS_ASSERT_TOKENIZE_USE;
3804 if (PL_expect != XSTATE)
3805 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3806 is_use ? "use" : "no"));
3808 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3809 s = force_version(s, TRUE);
3810 if (*s == ';' || *s == '}'
3811 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
3812 start_force(PL_curforce);
3813 NEXTVAL_NEXTTOKE.opval = NULL;
3816 else if (*s == 'v') {
3817 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3818 s = force_version(s, FALSE);
3822 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3823 s = force_version(s, FALSE);
3825 pl_yylval.ival = is_use;
3829 static const char* const exp_name[] =
3830 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3831 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3838 Works out what to call the token just pulled out of the input
3839 stream. The yacc parser takes care of taking the ops we return and
3840 stitching them into a tree.
3846 if read an identifier
3847 if we're in a my declaration
3848 croak if they tried to say my($foo::bar)
3849 build the ops for a my() declaration
3850 if it's an access to a my() variable
3851 are we in a sort block?
3852 croak if my($a); $a <=> $b
3853 build ops for access to a my() variable
3854 if in a dq string, and they've said @foo and we can't find @foo
3856 build ops for a bareword
3857 if we already built the token before, use it.
3862 #pragma segment Perl_yylex
3868 register char *s = PL_bufptr;
3874 /* orig_keyword, gvp, and gv are initialized here because
3875 * jump to the label just_a_word_zero can bypass their
3876 * initialization later. */
3877 I32 orig_keyword = 0;
3882 SV* tmp = newSVpvs("");
3883 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3884 (IV)CopLINE(PL_curcop),
3885 lex_state_names[PL_lex_state],
3886 exp_name[PL_expect],
3887 pv_display(tmp, s, strlen(s), 0, 60));
3890 /* check if there's an identifier for us to look at */
3891 if (PL_pending_ident)
3892 return REPORT(S_pending_ident(aTHX));
3894 /* no identifier pending identification */
3896 switch (PL_lex_state) {
3898 case LEX_NORMAL: /* Some compilers will produce faster */
3899 case LEX_INTERPNORMAL: /* code if we comment these out. */
3903 /* when we've already built the next token, just pull it out of the queue */
3907 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3909 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3910 PL_nexttoke[PL_lasttoke].next_mad = 0;
3911 if (PL_thismad && PL_thismad->mad_key == '_') {
3912 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3913 PL_thismad->mad_val = 0;
3914 mad_free(PL_thismad);
3919 PL_lex_state = PL_lex_defer;
3920 PL_expect = PL_lex_expect;
3921 PL_lex_defer = LEX_NORMAL;
3922 if (!PL_nexttoke[PL_lasttoke].next_type)
3927 pl_yylval = PL_nextval[PL_nexttoke];
3929 PL_lex_state = PL_lex_defer;
3930 PL_expect = PL_lex_expect;
3931 PL_lex_defer = LEX_NORMAL;
3935 /* FIXME - can these be merged? */
3936 return(PL_nexttoke[PL_lasttoke].next_type);
3938 return REPORT(PL_nexttype[PL_nexttoke]);
3941 /* interpolated case modifiers like \L \U, including \Q and \E.
3942 when we get here, PL_bufptr is at the \
3944 case LEX_INTERPCASEMOD:
3946 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3947 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3949 /* handle \E or end of string */
3950 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3952 if (PL_lex_casemods) {
3953 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3954 PL_lex_casestack[PL_lex_casemods] = '\0';
3956 if (PL_bufptr != PL_bufend
3957 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3959 PL_lex_state = LEX_INTERPCONCAT;
3962 PL_thistoken = newSVpvs("\\E");
3968 while (PL_bufptr != PL_bufend &&
3969 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3971 PL_thiswhite = newSVpvs("");
3972 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3976 if (PL_bufptr != PL_bufend)
3979 PL_lex_state = LEX_INTERPCONCAT;
3983 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3984 "### Saw case modifier\n"); });
3986 if (s[1] == '\\' && s[2] == 'E') {
3989 PL_thiswhite = newSVpvs("");
3990 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3993 PL_lex_state = LEX_INTERPCONCAT;
3998 if (!PL_madskills) /* when just compiling don't need correct */
3999 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4000 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4001 if ((*s == 'L' || *s == 'U') &&
4002 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4003 PL_lex_casestack[--PL_lex_casemods] = '\0';
4006 if (PL_lex_casemods > 10)
4007 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4008 PL_lex_casestack[PL_lex_casemods++] = *s;
4009 PL_lex_casestack[PL_lex_casemods] = '\0';
4010 PL_lex_state = LEX_INTERPCONCAT;
4011 start_force(PL_curforce);
4012 NEXTVAL_NEXTTOKE.ival = 0;
4014 start_force(PL_curforce);
4016 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4018 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4020 NEXTVAL_NEXTTOKE.ival = OP_LC;
4022 NEXTVAL_NEXTTOKE.ival = OP_UC;
4024 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4026 Perl_croak(aTHX_ "panic: yylex");
4028 SV* const tmpsv = newSVpvs("\\ ");
4029 /* replace the space with the character we want to escape
4031 SvPVX(tmpsv)[1] = *s;
4037 if (PL_lex_starts) {
4043 sv_free(PL_thistoken);
4044 PL_thistoken = newSVpvs("");
4047 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4048 if (PL_lex_casemods == 1 && PL_lex_inpat)
4057 case LEX_INTERPPUSH:
4058 return REPORT(sublex_push());
4060 case LEX_INTERPSTART:
4061 if (PL_bufptr == PL_bufend)
4062 return REPORT(sublex_done());
4063 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4064 "### Interpolated variable\n"); });
4066 PL_lex_dojoin = (*PL_bufptr == '@');
4067 PL_lex_state = LEX_INTERPNORMAL;
4068 if (PL_lex_dojoin) {
4069 start_force(PL_curforce);
4070 NEXTVAL_NEXTTOKE.ival = 0;
4072 start_force(PL_curforce);
4073 force_ident("\"", '$');
4074 start_force(PL_curforce);
4075 NEXTVAL_NEXTTOKE.ival = 0;
4077 start_force(PL_curforce);
4078 NEXTVAL_NEXTTOKE.ival = 0;
4080 start_force(PL_curforce);
4081 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4084 if (PL_lex_starts++) {
4089 sv_free(PL_thistoken);
4090 PL_thistoken = newSVpvs("");
4093 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4094 if (!PL_lex_casemods && PL_lex_inpat)
4101 case LEX_INTERPENDMAYBE:
4102 if (intuit_more(PL_bufptr)) {
4103 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4109 if (PL_lex_dojoin) {
4110 PL_lex_dojoin = FALSE;
4111 PL_lex_state = LEX_INTERPCONCAT;
4115 sv_free(PL_thistoken);
4116 PL_thistoken = newSVpvs("");
4121 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4122 && SvEVALED(PL_lex_repl))
4124 if (PL_bufptr != PL_bufend)
4125 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4129 case LEX_INTERPCONCAT:
4131 if (PL_lex_brackets)
4132 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4134 if (PL_bufptr == PL_bufend)
4135 return REPORT(sublex_done());
4137 if (SvIVX(PL_linestr) == '\'') {
4138 SV *sv = newSVsv(PL_linestr);
4141 else if ( PL_hints & HINT_NEW_RE )
4142 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4143 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4147 s = scan_const(PL_bufptr);
4149 PL_lex_state = LEX_INTERPCASEMOD;
4151 PL_lex_state = LEX_INTERPSTART;
4154 if (s != PL_bufptr) {
4155 start_force(PL_curforce);
4157 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4159 NEXTVAL_NEXTTOKE = pl_yylval;
4162 if (PL_lex_starts++) {
4166 sv_free(PL_thistoken);
4167 PL_thistoken = newSVpvs("");
4170 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4171 if (!PL_lex_casemods && PL_lex_inpat)
4184 PL_lex_state = LEX_NORMAL;
4185 s = scan_formline(PL_bufptr);
4186 if (!PL_lex_formbrack)
4192 PL_oldoldbufptr = PL_oldbufptr;
4198 sv_free(PL_thistoken);
4201 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4205 if (isIDFIRST_lazy_if(s,UTF))
4208 unsigned char c = *s;
4209 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4210 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4211 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4216 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4220 goto fake_eof; /* emulate EOF on ^D or ^Z */
4229 if (PL_lex_brackets) {
4230 yyerror((const char *)
4232 ? "Format not terminated"
4233 : "Missing right curly or square bracket"));
4235 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4236 "### Tokener got EOF\n");
4240 if (s++ < PL_bufend)
4241 goto retry; /* ignore stray nulls */
4244 if (!PL_in_eval && !PL_preambled) {
4245 PL_preambled = TRUE;
4251 /* Generate a string of Perl code to load the debugger.
4252 * If PERL5DB is set, it will return the contents of that,
4253 * otherwise a compile-time require of perl5db.pl. */
4255 const char * const pdb = PerlEnv_getenv("PERL5DB");
4258 sv_setpv(PL_linestr, pdb);
4259 sv_catpvs(PL_linestr,";");
4261 SETERRNO(0,SS_NORMAL);
4262 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4265 sv_setpvs(PL_linestr,"");
4266 if (PL_preambleav) {
4267 SV **svp = AvARRAY(PL_preambleav);
4268 SV **const end = svp + AvFILLp(PL_preambleav);
4270 sv_catsv(PL_linestr, *svp);
4272 sv_catpvs(PL_linestr, ";");
4274 sv_free(MUTABLE_SV(PL_preambleav));
4275 PL_preambleav = NULL;
4278 sv_catpvs(PL_linestr,
4279 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4280 if (PL_minus_n || PL_minus_p) {
4281 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4283 sv_catpvs(PL_linestr,"chomp;");
4286 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4287 || *PL_splitstr == '"')
4288 && strchr(PL_splitstr + 1, *PL_splitstr))
4289 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4291 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4292 bytes can be used as quoting characters. :-) */
4293 const char *splits = PL_splitstr;
4294 sv_catpvs(PL_linestr, "our @F=split(q\0");
4297 if (*splits == '\\')
4298 sv_catpvn(PL_linestr, splits, 1);
4299 sv_catpvn(PL_linestr, splits, 1);
4300 } while (*splits++);
4301 /* This loop will embed the trailing NUL of
4302 PL_linestr as the last thing it does before
4304 sv_catpvs(PL_linestr, ");");
4308 sv_catpvs(PL_linestr,"our @F=split(' ');");
4311 sv_catpvs(PL_linestr, "\n");
4312 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4313 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4314 PL_last_lop = PL_last_uni = NULL;
4315 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4316 update_debugger_info(PL_linestr, NULL, 0);
4321 bof = PL_rsfp ? TRUE : FALSE;
4324 fake_eof = LEX_FAKE_EOF;
4326 PL_bufptr = PL_bufend;
4327 CopLINE_inc(PL_curcop);
4328 if (!lex_next_chunk(fake_eof)) {
4329 CopLINE_dec(PL_curcop);
4331 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4333 CopLINE_dec(PL_curcop);
4336 PL_realtokenstart = -1;
4339 /* If it looks like the start of a BOM or raw UTF-16,
4340 * check if it in fact is. */
4341 if (bof && PL_rsfp &&
4346 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4348 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4349 s = swallow_bom((U8*)s);
4353 /* Incest with pod. */
4356 sv_catsv(PL_thiswhite, PL_linestr);
4358 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4359 sv_setpvs(PL_linestr, "");
4360 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4361 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4362 PL_last_lop = PL_last_uni = NULL;
4363 PL_doextract = FALSE;
4368 } while (PL_doextract);
4369 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4370 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4371 PL_last_lop = PL_last_uni = NULL;
4372 if (CopLINE(PL_curcop) == 1) {
4373 while (s < PL_bufend && isSPACE(*s))
4375 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4379 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4383 if (*s == '#' && *(s+1) == '!')
4385 #ifdef ALTERNATE_SHEBANG
4387 static char const as[] = ALTERNATE_SHEBANG;
4388 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4389 d = s + (sizeof(as) - 1);
4391 #endif /* ALTERNATE_SHEBANG */
4400 while (*d && !isSPACE(*d))
4404 #ifdef ARG_ZERO_IS_SCRIPT
4405 if (ipathend > ipath) {
4407 * HP-UX (at least) sets argv[0] to the script name,
4408 * which makes $^X incorrect. And Digital UNIX and Linux,
4409 * at least, set argv[0] to the basename of the Perl
4410 * interpreter. So, having found "#!", we'll set it right.
4412 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4414 assert(SvPOK(x) || SvGMAGICAL(x));
4415 if (sv_eq(x, CopFILESV(PL_curcop))) {
4416 sv_setpvn(x, ipath, ipathend - ipath);
4422 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4423 const char * const lstart = SvPV_const(x,llen);
4425 bstart += blen - llen;
4426 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4427 sv_setpvn(x, ipath, ipathend - ipath);
4432 TAINT_NOT; /* $^X is always tainted, but that's OK */
4434 #endif /* ARG_ZERO_IS_SCRIPT */
4439 d = instr(s,"perl -");
4441 d = instr(s,"perl");
4443 /* avoid getting into infinite loops when shebang
4444 * line contains "Perl" rather than "perl" */
4446 for (d = ipathend-4; d >= ipath; --d) {
4447 if ((*d == 'p' || *d == 'P')
4448 && !ibcmp(d, "perl", 4))
4458 #ifdef ALTERNATE_SHEBANG
4460 * If the ALTERNATE_SHEBANG on this system starts with a
4461 * character that can be part of a Perl expression, then if
4462 * we see it but not "perl", we're probably looking at the
4463 * start of Perl code, not a request to hand off to some
4464 * other interpreter. Similarly, if "perl" is there, but
4465 * not in the first 'word' of the line, we assume the line
4466 * contains the start of the Perl program.
4468 if (d && *s != '#') {
4469 const char *c = ipath;
4470 while (*c && !strchr("; \t\r\n\f\v#", *c))
4473 d = NULL; /* "perl" not in first word; ignore */
4475 *s = '#'; /* Don't try to parse shebang line */
4477 #endif /* ALTERNATE_SHEBANG */
4482 !instr(s,"indir") &&
4483 instr(PL_origargv[0],"perl"))
4490 while (s < PL_bufend && isSPACE(*s))
4492 if (s < PL_bufend) {
4493 Newx(newargv,PL_origargc+3,char*);
4495 while (s < PL_bufend && !isSPACE(*s))
4498 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4501 newargv = PL_origargv;
4504 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4506 Perl_croak(aTHX_ "Can't exec %s", ipath);
4509 while (*d && !isSPACE(*d))
4511 while (SPACE_OR_TAB(*d))
4515 const bool switches_done = PL_doswitches;
4516 const U32 oldpdb = PL_perldb;
4517 const bool oldn = PL_minus_n;
4518 const bool oldp = PL_minus_p;
4522 bool baduni = FALSE;
4524 const char *d2 = d1 + 1;
4525 if (parse_unicode_opts((const char **)&d2)
4529 if (baduni || *d1 == 'M' || *d1 == 'm') {
4530 const char * const m = d1;
4531 while (*d1 && !isSPACE(*d1))
4533 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4536 d1 = moreswitches(d1);
4538 if (PL_doswitches && !switches_done) {
4539 int argc = PL_origargc;
4540 char **argv = PL_origargv;
4543 } while (argc && argv[0][0] == '-' && argv[0][1]);
4544 init_argv_symbols(argc,argv);
4546 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4547 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4548 /* if we have already added "LINE: while (<>) {",
4549 we must not do it again */
4551 sv_setpvs(PL_linestr, "");
4552 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4553 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4554 PL_last_lop = PL_last_uni = NULL;
4555 PL_preambled = FALSE;
4556 if (PERLDB_LINE || PERLDB_SAVESRC)
4557 (void)gv_fetchfile(PL_origfilename);
4564 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4566 PL_lex_state = LEX_FORMLINE;
4571 #ifdef PERL_STRICT_CR
4572 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4574 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4576 case ' ': case '\t': case '\f': case 013:
4578 PL_realtokenstart = -1;
4580 PL_thiswhite = newSVpvs("");
4581 sv_catpvn(PL_thiswhite, s, 1);
4588 PL_realtokenstart = -1;
4592 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4593 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4594 /* handle eval qq[#line 1 "foo"\n ...] */
4595 CopLINE_dec(PL_curcop);
4598 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4600 if (!PL_in_eval || PL_rsfp)
4605 while (d < PL_bufend && *d != '\n')
4609 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4610 Perl_croak(aTHX_ "panic: input overflow");
4613 PL_thiswhite = newSVpvn(s, d - s);
4618 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4620 PL_lex_state = LEX_FORMLINE;
4626 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4627 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4630 TOKEN(PEG); /* make sure any #! line is accessible */
4635 /* if (PL_madskills && PL_lex_formbrack) { */
4637 while (d < PL_bufend && *d != '\n')
4641 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4642 Perl_croak(aTHX_ "panic: input overflow");
4643 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4645 PL_thiswhite = newSVpvs("");
4646 if (CopLINE(PL_curcop) == 1) {
4647 sv_setpvs(PL_thiswhite, "");
4650 sv_catpvn(PL_thiswhite, s, d - s);
4664 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4672 while (s < PL_bufend && SPACE_OR_TAB(*s))
4675 if (strnEQ(s,"=>",2)) {
4676 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4677 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4678 OPERATOR('-'); /* unary minus */
4680 PL_last_uni = PL_oldbufptr;
4682 case 'r': ftst = OP_FTEREAD; break;
4683 case 'w': ftst = OP_FTEWRITE; break;
4684 case 'x': ftst = OP_FTEEXEC; break;
4685 case 'o': ftst = OP_FTEOWNED; break;
4686 case 'R': ftst = OP_FTRREAD; break;
4687 case 'W': ftst = OP_FTRWRITE; break;
4688 case 'X': ftst = OP_FTREXEC; break;
4689 case 'O': ftst = OP_FTROWNED; break;
4690 case 'e': ftst = OP_FTIS; break;
4691 case 'z': ftst = OP_FTZERO; break;
4692 case 's': ftst = OP_FTSIZE; break;
4693 case 'f': ftst = OP_FTFILE; break;
4694 case 'd': ftst = OP_FTDIR; break;
4695 case 'l': ftst = OP_FTLINK; break;
4696 case 'p': ftst = OP_FTPIPE; break;
4697 case 'S': ftst = OP_FTSOCK; break;
4698 case 'u': ftst = OP_FTSUID; break;
4699 case 'g': ftst = OP_FTSGID; break;
4700 case 'k': ftst = OP_FTSVTX; break;
4701 case 'b': ftst = OP_FTBLK; break;
4702 case 'c': ftst = OP_FTCHR; break;
4703 case 't': ftst = OP_FTTTY; break;
4704 case 'T': ftst = OP_FTTEXT; break;
4705 case 'B': ftst = OP_FTBINARY; break;
4706 case 'M': case 'A': case 'C':
4707 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4709 case 'M': ftst = OP_FTMTIME; break;
4710 case 'A': ftst = OP_FTATIME; break;
4711 case 'C': ftst = OP_FTCTIME; break;
4719 PL_last_lop_op = (OPCODE)ftst;
4720 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4721 "### Saw file test %c\n", (int)tmp);
4726 /* Assume it was a minus followed by a one-letter named
4727 * subroutine call (or a -bareword), then. */
4728 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4729 "### '-%c' looked like a file test but was not\n",
4736 const char tmp = *s++;
4739 if (PL_expect == XOPERATOR)
4744 else if (*s == '>') {
4747 if (isIDFIRST_lazy_if(s,UTF)) {
4748 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4756 if (PL_expect == XOPERATOR)
4759 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4761 OPERATOR('-'); /* unary minus */
4767 const char tmp = *s++;
4770 if (PL_expect == XOPERATOR)
4775 if (PL_expect == XOPERATOR)
4778 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4785 if (PL_expect != XOPERATOR) {
4786 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4787 PL_expect = XOPERATOR;
4788 force_ident(PL_tokenbuf, '*');
4801 if (PL_expect == XOPERATOR) {
4805 PL_tokenbuf[0] = '%';
4806 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4807 sizeof PL_tokenbuf - 1, FALSE);
4808 if (!PL_tokenbuf[1]) {
4811 PL_pending_ident = '%';
4820 const char tmp = *s++;
4825 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4832 const char tmp = *s++;
4838 goto just_a_word_zero_gv;
4841 switch (PL_expect) {
4847 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4849 PL_bufptr = s; /* update in case we back off */
4851 deprecate(":= for an empty attribute list");
4858 PL_expect = XTERMBLOCK;
4861 stuffstart = s - SvPVX(PL_linestr) - 1;
4865 while (isIDFIRST_lazy_if(s,UTF)) {
4868 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4869 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4870 if (tmp < 0) tmp = -tmp;
4885 sv = newSVpvn(s, len);
4887 d = scan_str(d,TRUE,TRUE);
4889 /* MUST advance bufptr here to avoid bogus
4890 "at end of line" context messages from yyerror().
4892 PL_bufptr = s + len;
4893 yyerror("Unterminated attribute parameter in attribute list");
4897 return REPORT(0); /* EOF indicator */
4901 sv_catsv(sv, PL_lex_stuff);
4902 attrs = append_elem(OP_LIST, attrs,
4903 newSVOP(OP_CONST, 0, sv));
4904 SvREFCNT_dec(PL_lex_stuff);
4905 PL_lex_stuff = NULL;
4908 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4910 if (PL_in_my == KEY_our) {
4911 deprecate(":unique");
4914 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4917 /* NOTE: any CV attrs applied here need to be part of
4918 the CVf_BUILTIN_ATTRS define in cv.h! */
4919 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4921 CvLVALUE_on(PL_compcv);
4923 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4925 deprecate(":locked");
4927 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4929 CvMETHOD_on(PL_compcv);
4931 /* After we've set the flags, it could be argued that
4932 we don't need to do the attributes.pm-based setting
4933 process, and shouldn't bother appending recognized
4934 flags. To experiment with that, uncomment the
4935 following "else". (Note that's already been
4936 uncommented. That keeps the above-applied built-in
4937 attributes from being intercepted (and possibly
4938 rejected) by a package's attribute routines, but is
4939 justified by the performance win for the common case
4940 of applying only built-in attributes.) */
4942 attrs = append_elem(OP_LIST, attrs,
4943 newSVOP(OP_CONST, 0,
4947 if (*s == ':' && s[1] != ':')
4950 break; /* require real whitespace or :'s */
4951 /* XXX losing whitespace on sequential attributes here */
4955 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4956 if (*s != ';' && *s != '}' && *s != tmp
4957 && (tmp != '=' || *s != ')')) {
4958 const char q = ((*s == '\'') ? '"' : '\'');
4959 /* If here for an expression, and parsed no attrs, back
4961 if (tmp == '=' && !attrs) {
4965 /* MUST advance bufptr here to avoid bogus "at end of line"
4966 context messages from yyerror().
4969 yyerror( (const char *)
4971 ? Perl_form(aTHX_ "Invalid separator character "
4972 "%c%c%c in attribute list", q, *s, q)
4973 : "Unterminated attribute list" ) );
4981 start_force(PL_curforce);
4982 NEXTVAL_NEXTTOKE.opval = attrs;
4983 CURMAD('_', PL_nextwhite);
4988 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4989 (s - SvPVX(PL_linestr)) - stuffstart);
4997 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4998 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5006 const char tmp = *s++;
5011 const char tmp = *s++;
5019 if (PL_lex_brackets <= 0)
5020 yyerror("Unmatched right square bracket");
5023 if (PL_lex_state == LEX_INTERPNORMAL) {
5024 if (PL_lex_brackets == 0) {
5025 if (*s == '-' && s[1] == '>')
5026 PL_lex_state = LEX_INTERPENDMAYBE;
5027 else if (*s != '[' && *s != '{')
5028 PL_lex_state = LEX_INTERPEND;
5035 if (PL_lex_brackets > 100) {
5036 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5038 switch (PL_expect) {
5040 if (PL_lex_formbrack) {
5044 if (PL_oldoldbufptr == PL_last_lop)
5045 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5047 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5048 OPERATOR(HASHBRACK);
5050 while (s < PL_bufend && SPACE_OR_TAB(*s))
5053 PL_tokenbuf[0] = '\0';
5054 if (d < PL_bufend && *d == '-') {
5055 PL_tokenbuf[0] = '-';
5057 while (d < PL_bufend && SPACE_OR_TAB(*d))
5060 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5061 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5063 while (d < PL_bufend && SPACE_OR_TAB(*d))
5066 const char minus = (PL_tokenbuf[0] == '-');
5067 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5075 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5080 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5085 if (PL_oldoldbufptr == PL_last_lop)
5086 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5088 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5091 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5093 /* This hack is to get the ${} in the message. */
5095 yyerror("syntax error");
5098 OPERATOR(HASHBRACK);
5100 /* This hack serves to disambiguate a pair of curlies
5101 * as being a block or an anon hash. Normally, expectation
5102 * determines that, but in cases where we're not in a
5103 * position to expect anything in particular (like inside
5104 * eval"") we have to resolve the ambiguity. This code
5105 * covers the case where the first term in the curlies is a
5106 * quoted string. Most other cases need to be explicitly
5107 * disambiguated by prepending a "+" before the opening
5108 * curly in order to force resolution as an anon hash.
5110 * XXX should probably propagate the outer expectation
5111 * into eval"" to rely less on this hack, but that could
5112 * potentially break current behavior of eval"".
5116 if (*s == '\'' || *s == '"' || *s == '`') {
5117 /* common case: get past first string, handling escapes */
5118 for (t++; t < PL_bufend && *t != *s;)
5119 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5123 else if (*s == 'q') {
5126 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5129 /* skip q//-like construct */
5131 char open, close, term;
5134 while (t < PL_bufend && isSPACE(*t))
5136 /* check for q => */
5137 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5138 OPERATOR(HASHBRACK);
5142 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5146 for (t++; t < PL_bufend; t++) {
5147 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5149 else if (*t == open)
5153 for (t++; t < PL_bufend; t++) {
5154 if (*t == '\\' && t+1 < PL_bufend)
5156 else if (*t == close && --brackets <= 0)
5158 else if (*t == open)
5165 /* skip plain q word */
5166 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5169 else if (isALNUM_lazy_if(t,UTF)) {
5171 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5174 while (t < PL_bufend && isSPACE(*t))
5176 /* if comma follows first term, call it an anon hash */
5177 /* XXX it could be a comma expression with loop modifiers */
5178 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5179 || (*t == '=' && t[1] == '>')))
5180 OPERATOR(HASHBRACK);
5181 if (PL_expect == XREF)
5184 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5190 pl_yylval.ival = CopLINE(PL_curcop);
5191 if (isSPACE(*s) || *s == '#')
5192 PL_copline = NOLINE; /* invalidate current command line number */
5197 if (PL_lex_brackets <= 0)
5198 yyerror("Unmatched right curly bracket");
5200 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5201 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5202 PL_lex_formbrack = 0;
5203 if (PL_lex_state == LEX_INTERPNORMAL) {
5204 if (PL_lex_brackets == 0) {
5205 if (PL_expect & XFAKEBRACK) {
5206 PL_expect &= XENUMMASK;
5207 PL_lex_state = LEX_INTERPEND;
5212 PL_thiswhite = newSVpvs("");
5213 sv_catpvs(PL_thiswhite,"}");
5216 return yylex(); /* ignore fake brackets */
5218 if (*s == '-' && s[1] == '>')
5219 PL_lex_state = LEX_INTERPENDMAYBE;
5220 else if (*s != '[' && *s != '{')
5221 PL_lex_state = LEX_INTERPEND;
5224 if (PL_expect & XFAKEBRACK) {
5225 PL_expect &= XENUMMASK;
5227 return yylex(); /* ignore fake brackets */
5229 start_force(PL_curforce);
5231 curmad('X', newSVpvn(s-1,1));
5232 CURMAD('_', PL_thiswhite);
5237 PL_thistoken = newSVpvs("");
5245 if (PL_expect == XOPERATOR) {
5246 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5247 && isIDFIRST_lazy_if(s,UTF))
5249 CopLINE_dec(PL_curcop);
5250 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5251 CopLINE_inc(PL_curcop);
5256 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5258 PL_expect = XOPERATOR;
5259 force_ident(PL_tokenbuf, '&');
5263 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5275 const char tmp = *s++;
5282 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5283 && strchr("+-*/%.^&|<",tmp))
5284 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5285 "Reversed %c= operator",(int)tmp);
5287 if (PL_expect == XSTATE && isALPHA(tmp) &&
5288 (s == PL_linestart+1 || s[-2] == '\n') )
5290 if (PL_in_eval && !PL_rsfp) {
5295 if (strnEQ(s,"=cut",4)) {
5311 PL_thiswhite = newSVpvs("");
5312 sv_catpvn(PL_thiswhite, PL_linestart,
5313 PL_bufend - PL_linestart);
5317 PL_doextract = TRUE;
5321 if (PL_lex_brackets < PL_lex_formbrack) {
5323 #ifdef PERL_STRICT_CR
5324 while (SPACE_OR_TAB(*t))
5326 while (SPACE_OR_TAB(*t) || *t == '\r')
5329 if (*t == '\n' || *t == '#') {
5340 const char tmp = *s++;
5342 /* was this !=~ where !~ was meant?
5343 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5345 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5346 const char *t = s+1;
5348 while (t < PL_bufend && isSPACE(*t))
5351 if (*t == '/' || *t == '?' ||
5352 ((*t == 'm' || *t == 's' || *t == 'y')
5353 && !isALNUM(t[1])) ||
5354 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5355 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5356 "!=~ should be !~");
5366 if (PL_expect != XOPERATOR) {
5367 if (s[1] != '<' && !strchr(s,'>'))
5370 s = scan_heredoc(s);
5372 s = scan_inputsymbol(s);
5373 TERM(sublex_start());
5379 SHop(OP_LEFT_SHIFT);
5393 const char tmp = *s++;
5395 SHop(OP_RIGHT_SHIFT);
5396 else if (tmp == '=')
5405 if (PL_expect == XOPERATOR) {
5406 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5407 return deprecate_commaless_var_list();
5411 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5412 PL_tokenbuf[0] = '@';
5413 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5414 sizeof PL_tokenbuf - 1, FALSE);
5415 if (PL_expect == XOPERATOR)
5416 no_op("Array length", s);
5417 if (!PL_tokenbuf[1])
5419 PL_expect = XOPERATOR;
5420 PL_pending_ident = '#';
5424 PL_tokenbuf[0] = '$';
5425 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5426 sizeof PL_tokenbuf - 1, FALSE);
5427 if (PL_expect == XOPERATOR)
5429 if (!PL_tokenbuf[1]) {
5431 yyerror("Final $ should be \\$ or $name");
5435 /* This kludge not intended to be bulletproof. */
5436 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5437 pl_yylval.opval = newSVOP(OP_CONST, 0,
5438 newSViv(CopARYBASE_get(&PL_compiling)));
5439 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5445 const char tmp = *s;
5446 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5449 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5450 && intuit_more(s)) {
5452 PL_tokenbuf[0] = '@';
5453 if (ckWARN(WARN_SYNTAX)) {
5456 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5459 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5460 while (t < PL_bufend && *t != ']')
5462 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5463 "Multidimensional syntax %.*s not supported",
5464 (int)((t - PL_bufptr) + 1), PL_bufptr);
5468 else if (*s == '{') {
5470 PL_tokenbuf[0] = '%';
5471 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5472 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5474 char tmpbuf[sizeof PL_tokenbuf];
5477 } while (isSPACE(*t));
5478 if (isIDFIRST_lazy_if(t,UTF)) {
5480 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5484 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5485 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5486 "You need to quote \"%s\"",
5493 PL_expect = XOPERATOR;
5494 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5495 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5496 if (!islop || PL_last_lop_op == OP_GREPSTART)
5497 PL_expect = XOPERATOR;
5498 else if (strchr("$@\"'`q", *s))
5499 PL_expect = XTERM; /* e.g. print $fh "foo" */
5500 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5501 PL_expect = XTERM; /* e.g. print $fh &sub */
5502 else if (isIDFIRST_lazy_if(s,UTF)) {
5503 char tmpbuf[sizeof PL_tokenbuf];
5505 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5506 if ((t2 = keyword(tmpbuf, len, 0))) {
5507 /* binary operators exclude handle interpretations */
5519 PL_expect = XTERM; /* e.g. print $fh length() */
5524 PL_expect = XTERM; /* e.g. print $fh subr() */
5527 else if (isDIGIT(*s))
5528 PL_expect = XTERM; /* e.g. print $fh 3 */
5529 else if (*s == '.' && isDIGIT(s[1]))
5530 PL_expect = XTERM; /* e.g. print $fh .3 */
5531 else if ((*s == '?' || *s == '-' || *s == '+')
5532 && !isSPACE(s[1]) && s[1] != '=')
5533 PL_expect = XTERM; /* e.g. print $fh -1 */
5534 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5536 PL_expect = XTERM; /* e.g. print $fh /.../
5537 XXX except DORDOR operator
5539 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5541 PL_expect = XTERM; /* print $fh <<"EOF" */
5544 PL_pending_ident = '$';
5548 if (PL_expect == XOPERATOR)
5550 PL_tokenbuf[0] = '@';
5551 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5552 if (!PL_tokenbuf[1]) {
5555 if (PL_lex_state == LEX_NORMAL)
5557 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5559 PL_tokenbuf[0] = '%';
5561 /* Warn about @ where they meant $. */
5562 if (*s == '[' || *s == '{') {
5563 if (ckWARN(WARN_SYNTAX)) {
5564 const char *t = s + 1;
5565 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5567 if (*t == '}' || *t == ']') {
5569 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5570 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5571 "Scalar value %.*s better written as $%.*s",
5572 (int)(t-PL_bufptr), PL_bufptr,
5573 (int)(t-PL_bufptr-1), PL_bufptr+1);
5578 PL_pending_ident = '@';
5581 case '/': /* may be division, defined-or, or pattern */
5582 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5586 case '?': /* may either be conditional or pattern */
5587 if (PL_expect == XOPERATOR) {
5595 /* A // operator. */
5605 /* Disable warning on "study /blah/" */
5606 if (PL_oldoldbufptr == PL_last_uni
5607 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5608 || memNE(PL_last_uni, "study", 5)
5609 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5612 s = scan_pat(s,OP_MATCH);
5613 TERM(sublex_start());
5617 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5618 #ifdef PERL_STRICT_CR
5621 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5623 && (s == PL_linestart || s[-1] == '\n') )
5625 PL_lex_formbrack = 0;
5629 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5633 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5639 pl_yylval.ival = OPf_SPECIAL;
5648 case '0': case '1': case '2': case '3': case '4':
5649 case '5': case '6': case '7': case '8': case '9':
5650 s = scan_num(s, &pl_yylval);
5651 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5652 if (PL_expect == XOPERATOR)
5657 s = scan_str(s,!!PL_madskills,FALSE);
5658 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5659 if (PL_expect == XOPERATOR) {
5660 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5661 return deprecate_commaless_var_list();
5668 pl_yylval.ival = OP_CONST;
5669 TERM(sublex_start());
5672 s = scan_str(s,!!PL_madskills,FALSE);
5673 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5674 if (PL_expect == XOPERATOR) {
5675 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5676 return deprecate_commaless_var_list();
5683 pl_yylval.ival = OP_CONST;
5684 /* FIXME. I think that this can be const if char *d is replaced by
5685 more localised variables. */
5686 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5687 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5688 pl_yylval.ival = OP_STRINGIFY;
5692 TERM(sublex_start());
5695 s = scan_str(s,!!PL_madskills,FALSE);
5696 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5697 if (PL_expect == XOPERATOR)
5698 no_op("Backticks",s);
5701 readpipe_override();
5702 TERM(sublex_start());
5706 if (PL_lex_inwhat && isDIGIT(*s))
5707 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5709 if (PL_expect == XOPERATOR)
5710 no_op("Backslash",s);
5714 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5715 char *start = s + 2;
5716 while (isDIGIT(*start) || *start == '_')
5718 if (*start == '.' && isDIGIT(start[1])) {
5719 s = scan_num(s, &pl_yylval);
5722 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5723 else if (!isALPHA(*start) && (PL_expect == XTERM
5724 || PL_expect == XREF || PL_expect == XSTATE
5725 || PL_expect == XTERMORDORDOR)) {
5726 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5728 s = scan_num(s, &pl_yylval);
5735 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5778 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5780 /* Some keywords can be followed by any delimiter, including ':' */
5781 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5782 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5783 (PL_tokenbuf[0] == 'q' &&
5784 strchr("qwxr", PL_tokenbuf[1])))));
5786 /* x::* is just a word, unless x is "CORE" */
5787 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5791 while (d < PL_bufend && isSPACE(*d))
5792 d++; /* no comments skipped here, or s### is misparsed */
5794 /* Is this a word before a => operator? */
5795 if (*d == '=' && d[1] == '>') {
5798 = (OP*)newSVOP(OP_CONST, 0,
5799 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5800 pl_yylval.opval->op_private = OPpCONST_BARE;
5804 /* Check for plugged-in keyword */
5808 char *saved_bufptr = PL_bufptr;
5810 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
5812 if (result == KEYWORD_PLUGIN_DECLINE) {
5813 /* not a plugged-in keyword */
5814 PL_bufptr = saved_bufptr;
5815 } else if (result == KEYWORD_PLUGIN_STMT) {
5816 pl_yylval.opval = o;
5819 return REPORT(PLUGSTMT);
5820 } else if (result == KEYWORD_PLUGIN_EXPR) {
5821 pl_yylval.opval = o;
5823 PL_expect = XOPERATOR;
5824 return REPORT(PLUGEXPR);
5826 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
5831 /* Check for built-in keyword */
5832 tmp = keyword(PL_tokenbuf, len, 0);
5834 /* Is this a label? */
5835 if (!anydelim && PL_expect == XSTATE
5836 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5838 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
5840 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5845 if (tmp < 0) { /* second-class keyword? */
5846 GV *ogv = NULL; /* override (winner) */
5847 GV *hgv = NULL; /* hidden (loser) */
5848 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5850 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5853 if (GvIMPORTED_CV(gv))
5855 else if (! CvMETHOD(cv))
5859 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5860 (gv = *gvp) && isGV_with_GP(gv) &&
5861 GvCVu(gv) && GvIMPORTED_CV(gv))
5868 tmp = 0; /* overridden by import or by GLOBAL */
5871 && -tmp==KEY_lock /* XXX generalizable kludge */
5874 tmp = 0; /* any sub overrides "weak" keyword */
5876 else { /* no override */
5878 if (tmp == KEY_dump) {
5879 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
5880 "dump() better written as CORE::dump()");
5884 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
5885 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5886 "Ambiguous call resolved as CORE::%s(), %s",
5887 GvENAME(hgv), "qualify as such or use &");
5894 default: /* not a keyword */
5895 /* Trade off - by using this evil construction we can pull the
5896 variable gv into the block labelled keylookup. If not, then
5897 we have to give it function scope so that the goto from the
5898 earlier ':' case doesn't bypass the initialisation. */
5900 just_a_word_zero_gv:
5908 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5912 SV *nextPL_nextwhite = 0;
5916 /* Get the rest if it looks like a package qualifier */
5918 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5920 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5923 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5924 *s == '\'' ? "'" : "::");
5929 if (PL_expect == XOPERATOR) {
5930 if (PL_bufptr == PL_linestart) {
5931 CopLINE_dec(PL_curcop);
5932 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5933 CopLINE_inc(PL_curcop);
5936 no_op("Bareword",s);
5939 /* Look for a subroutine with this name in current package,
5940 unless name is "Foo::", in which case Foo is a bearword
5941 (and a package name). */
5943 if (len > 2 && !PL_madskills &&
5944 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5946 if (ckWARN(WARN_BAREWORD)
5947 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5948 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5949 "Bareword \"%s\" refers to nonexistent package",
5952 PL_tokenbuf[len] = '\0';
5958 /* Mustn't actually add anything to a symbol table.
5959 But also don't want to "initialise" any placeholder
5960 constants that might already be there into full
5961 blown PVGVs with attached PVCV. */
5962 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5963 GV_NOADD_NOINIT, SVt_PVCV);
5968 /* if we saw a global override before, get the right name */
5971 sv = newSVpvs("CORE::GLOBAL::");
5972 sv_catpv(sv,PL_tokenbuf);
5975 /* If len is 0, newSVpv does strlen(), which is correct.
5976 If len is non-zero, then it will be the true length,
5977 and so the scalar will be created correctly. */
5978 sv = newSVpv(PL_tokenbuf,len);
5981 if (PL_madskills && !PL_thistoken) {
5982 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5983 PL_thistoken = newSVpvn(start,s - start);
5984 PL_realtokenstart = s - SvPVX(PL_linestr);
5988 /* Presume this is going to be a bareword of some sort. */
5991 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5992 pl_yylval.opval->op_private = OPpCONST_BARE;
5993 /* UTF-8 package name? */
5994 if (UTF && !IN_BYTES &&
5995 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5998 /* And if "Foo::", then that's what it certainly is. */
6005 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6006 const_op->op_private = OPpCONST_BARE;
6007 rv2cv_op = newCVREF(0, const_op);
6009 if (rv2cv_op->op_type == OP_RV2CV &&
6010 (rv2cv_op->op_flags & OPf_KIDS)) {
6011 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6012 switch (rv_op->op_type) {
6014 SV *sv = cSVOPx_sv(rv_op);
6015 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6019 GV *gv = cGVOPx_gv(rv_op);
6020 CV *maybe_cv = GvCVu(gv);
6021 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6027 /* See if it's the indirect object for a list operator. */
6029 if (PL_oldoldbufptr &&
6030 PL_oldoldbufptr < PL_bufptr &&
6031 (PL_oldoldbufptr == PL_last_lop
6032 || PL_oldoldbufptr == PL_last_uni) &&
6033 /* NO SKIPSPACE BEFORE HERE! */
6034 (PL_expect == XREF ||
6035 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6037 bool immediate_paren = *s == '(';
6039 /* (Now we can afford to cross potential line boundary.) */
6040 s = SKIPSPACE2(s,nextPL_nextwhite);
6042 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6045 /* Two barewords in a row may indicate method call. */
6047 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6048 (tmp = intuit_method(s, gv, cv))) {
6053 /* If not a declared subroutine, it's an indirect object. */
6054 /* (But it's an indir obj regardless for sort.) */
6055 /* Also, if "_" follows a filetest operator, it's a bareword */
6058 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6060 (PL_last_lop_op != OP_MAPSTART &&
6061 PL_last_lop_op != OP_GREPSTART))))
6062 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6063 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6066 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6071 PL_expect = XOPERATOR;
6074 s = SKIPSPACE2(s,nextPL_nextwhite);
6075 PL_nextwhite = nextPL_nextwhite;
6080 /* Is this a word before a => operator? */
6081 if (*s == '=' && s[1] == '>' && !pkgname) {
6084 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6085 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6086 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6090 /* If followed by a paren, it's certainly a subroutine. */
6095 while (SPACE_OR_TAB(*d))
6097 if (*d == ')' && (sv = cv_const_sv(cv))) {
6104 PL_nextwhite = PL_thiswhite;
6107 start_force(PL_curforce);
6109 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6110 PL_expect = XOPERATOR;
6113 PL_nextwhite = nextPL_nextwhite;
6114 curmad('X', PL_thistoken);
6115 PL_thistoken = newSVpvs("");
6124 /* If followed by var or block, call it a method (unless sub) */
6126 if ((*s == '$' || *s == '{') && !cv) {
6128 PL_last_lop = PL_oldbufptr;
6129 PL_last_lop_op = OP_METHOD;
6133 /* If followed by a bareword, see if it looks like indir obj. */
6136 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6137 && (tmp = intuit_method(s, gv, cv))) {
6142 /* Not a method, so call it a subroutine (if defined) */
6145 if (lastchar == '-')
6146 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6147 "Ambiguous use of -%s resolved as -&%s()",
6148 PL_tokenbuf, PL_tokenbuf);
6149 /* Check for a constant sub */
6150 if ((sv = cv_const_sv(cv))) {
6153 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6154 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6155 pl_yylval.opval->op_private = 0;
6159 op_free(pl_yylval.opval);
6160 pl_yylval.opval = rv2cv_op;
6161 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6162 PL_last_lop = PL_oldbufptr;
6163 PL_last_lop_op = OP_ENTERSUB;
6164 /* Is there a prototype? */
6172 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6175 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6177 while (*proto == ';')
6179 if (*proto == '&' && *s == '{') {
6181 sv_setpvs(PL_subname, "__ANON__");
6183 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6190 PL_nextwhite = PL_thiswhite;
6193 start_force(PL_curforce);
6194 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6197 PL_nextwhite = nextPL_nextwhite;
6198 curmad('X', PL_thistoken);
6199 PL_thistoken = newSVpvs("");
6206 /* Guess harder when madskills require "best effort". */
6207 if (PL_madskills && (!gv || !GvCVu(gv))) {
6208 int probable_sub = 0;
6209 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6211 else if (isALPHA(*s)) {
6215 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6216 if (!keyword(tmpbuf, tmplen, 0))
6219 while (d < PL_bufend && isSPACE(*d))
6221 if (*d == '=' && d[1] == '>')
6226 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6227 op_free(pl_yylval.opval);
6228 pl_yylval.opval = rv2cv_op;
6229 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6230 PL_last_lop = PL_oldbufptr;
6231 PL_last_lop_op = OP_ENTERSUB;
6232 PL_nextwhite = PL_thiswhite;
6234 start_force(PL_curforce);
6235 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6237 PL_nextwhite = nextPL_nextwhite;
6238 curmad('X', PL_thistoken);
6239 PL_thistoken = newSVpvs("");
6244 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6251 /* Call it a bare word */
6253 if (PL_hints & HINT_STRICT_SUBS)
6254 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6257 /* after "print" and similar functions (corresponding to
6258 * "F? L" in opcode.pl), whatever wasn't already parsed as
6259 * a filehandle should be subject to "strict subs".
6260 * Likewise for the optional indirect-object argument to system
6261 * or exec, which can't be a bareword */
6262 if ((PL_last_lop_op == OP_PRINT
6263 || PL_last_lop_op == OP_PRTF
6264 || PL_last_lop_op == OP_SAY
6265 || PL_last_lop_op == OP_SYSTEM
6266 || PL_last_lop_op == OP_EXEC)
6267 && (PL_hints & HINT_STRICT_SUBS))
6268 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6269 if (lastchar != '-') {
6270 if (ckWARN(WARN_RESERVED)) {
6274 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6275 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6283 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6284 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6285 "Operator or semicolon missing before %c%s",
6286 lastchar, PL_tokenbuf);
6287 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6288 "Ambiguous use of %c resolved as operator %c",
6289 lastchar, lastchar);
6295 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6296 newSVpv(CopFILE(PL_curcop),0));
6300 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6301 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6304 case KEY___PACKAGE__:
6305 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6307 ? newSVhek(HvNAME_HEK(PL_curstash))
6314 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6315 const char *pname = "main";
6316 if (PL_tokenbuf[2] == 'D')
6317 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6318 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6322 GvIOp(gv) = newIO();
6323 IoIFP(GvIOp(gv)) = PL_rsfp;
6324 #if defined(HAS_FCNTL) && defined(F_SETFD)
6326 const int fd = PerlIO_fileno(PL_rsfp);
6327 fcntl(fd,F_SETFD,fd >= 3);
6330 /* Mark this internal pseudo-handle as clean */
6331 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6332 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6333 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6335 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6336 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6337 /* if the script was opened in binmode, we need to revert
6338 * it to text mode for compatibility; but only iff it has CRs
6339 * XXX this is a questionable hack at best. */
6340 if (PL_bufend-PL_bufptr > 2
6341 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6344 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6345 loc = PerlIO_tell(PL_rsfp);
6346 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6349 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6351 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6352 #endif /* NETWARE */
6353 #ifdef PERLIO_IS_STDIO /* really? */
6354 # if defined(__BORLANDC__)
6355 /* XXX see note in do_binmode() */
6356 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6360 PerlIO_seek(PL_rsfp, loc, 0);
6364 #ifdef PERLIO_LAYERS
6367 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6368 else if (PL_encoding) {
6375 XPUSHs(PL_encoding);
6377 call_method("name", G_SCALAR);
6381 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6382 Perl_form(aTHX_ ":encoding(%"SVf")",
6391 if (PL_realtokenstart >= 0) {
6392 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6394 PL_endwhite = newSVpvs("");
6395 sv_catsv(PL_endwhite, PL_thiswhite);
6397 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6398 PL_realtokenstart = -1;
6400 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6416 if (PL_expect == XSTATE) {
6423 if (*s == ':' && s[1] == ':') {
6426 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6427 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6428 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6431 else if (tmp == KEY_require || tmp == KEY_do)
6432 /* that's a way to remember we saw "CORE::" */
6445 LOP(OP_ACCEPT,XTERM);
6451 LOP(OP_ATAN2,XTERM);
6457 LOP(OP_BINMODE,XTERM);
6460 LOP(OP_BLESS,XTERM);
6469 /* When 'use switch' is in effect, continue has a dual
6470 life as a control operator. */
6472 if (!FEATURE_IS_ENABLED("switch"))
6475 /* We have to disambiguate the two senses of
6476 "continue". If the next token is a '{' then
6477 treat it as the start of a continue block;
6478 otherwise treat it as a control operator.
6490 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6507 if (!PL_cryptseen) {
6508 PL_cryptseen = TRUE;
6512 LOP(OP_CRYPT,XTERM);
6515 LOP(OP_CHMOD,XTERM);
6518 LOP(OP_CHOWN,XTERM);
6521 LOP(OP_CONNECT,XTERM);
6540 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6541 if (orig_keyword == KEY_do) {
6550 PL_hints |= HINT_BLOCK_SCOPE;
6560 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6561 LOP(OP_DBMOPEN,XTERM);
6567 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6574 pl_yylval.ival = CopLINE(PL_curcop);
6590 if (*s == '{') { /* block eval */
6591 PL_expect = XTERMBLOCK;
6592 UNIBRACK(OP_ENTERTRY);
6594 else { /* string eval */
6596 UNIBRACK(OP_ENTEREVAL);
6611 case KEY_endhostent:
6617 case KEY_endservent:
6620 case KEY_endprotoent:
6631 pl_yylval.ival = CopLINE(PL_curcop);
6633 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6636 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6639 if ((PL_bufend - p) >= 3 &&
6640 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6642 else if ((PL_bufend - p) >= 4 &&
6643 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6646 if (isIDFIRST_lazy_if(p,UTF)) {
6647 p = scan_ident(p, PL_bufend,
6648 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6652 Perl_croak(aTHX_ "Missing $ on loop variable");
6654 s = SvPVX(PL_linestr) + soff;
6660 LOP(OP_FORMLINE,XTERM);
6666 LOP(OP_FCNTL,XTERM);
6672 LOP(OP_FLOCK,XTERM);
6681 LOP(OP_GREPSTART, XREF);
6684 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6699 case KEY_getpriority:
6700 LOP(OP_GETPRIORITY,XTERM);
6702 case KEY_getprotobyname:
6705 case KEY_getprotobynumber:
6706 LOP(OP_GPBYNUMBER,XTERM);
6708 case KEY_getprotoent:
6720 case KEY_getpeername:
6721 UNI(OP_GETPEERNAME);
6723 case KEY_gethostbyname:
6726 case KEY_gethostbyaddr:
6727 LOP(OP_GHBYADDR,XTERM);
6729 case KEY_gethostent:
6732 case KEY_getnetbyname:
6735 case KEY_getnetbyaddr:
6736 LOP(OP_GNBYADDR,XTERM);
6741 case KEY_getservbyname:
6742 LOP(OP_GSBYNAME,XTERM);
6744 case KEY_getservbyport:
6745 LOP(OP_GSBYPORT,XTERM);
6747 case KEY_getservent:
6750 case KEY_getsockname:
6751 UNI(OP_GETSOCKNAME);
6753 case KEY_getsockopt:
6754 LOP(OP_GSOCKOPT,XTERM);
6769 pl_yylval.ival = CopLINE(PL_curcop);
6779 pl_yylval.ival = CopLINE(PL_curcop);
6783 LOP(OP_INDEX,XTERM);
6789 LOP(OP_IOCTL,XTERM);
6801 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6833 LOP(OP_LISTEN,XTERM);
6842 s = scan_pat(s,OP_MATCH);
6843 TERM(sublex_start());
6846 LOP(OP_MAPSTART, XREF);
6849 LOP(OP_MKDIR,XTERM);
6852 LOP(OP_MSGCTL,XTERM);
6855 LOP(OP_MSGGET,XTERM);
6858 LOP(OP_MSGRCV,XTERM);
6861 LOP(OP_MSGSND,XTERM);
6866 PL_in_my = (U16)tmp;
6868 if (isIDFIRST_lazy_if(s,UTF)) {
6872 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6873 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6875 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6876 if (!PL_in_my_stash) {
6879 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6883 if (PL_madskills) { /* just add type to declarator token */
6884 sv_catsv(PL_thistoken, PL_nextwhite);
6886 sv_catpvn(PL_thistoken, start, s - start);
6894 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6901 s = tokenize_use(0, s);
6905 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6912 if (isIDFIRST_lazy_if(s,UTF)) {
6914 for (d = s; isALNUM_lazy_if(d,UTF);)
6916 for (t=d; isSPACE(*t);)
6918 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6920 && !(t[0] == '=' && t[1] == '>')
6922 int parms_len = (int)(d-s);
6923 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6924 "Precedence problem: open %.*s should be open(%.*s)",
6925 parms_len, s, parms_len, s);
6931 pl_yylval.ival = OP_OR;
6941 LOP(OP_OPEN_DIR,XTERM);
6944 checkcomma(s,PL_tokenbuf,"filehandle");
6948 checkcomma(s,PL_tokenbuf,"filehandle");
6967 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6968 s = force_version(s, FALSE);
6972 LOP(OP_PIPE_OP,XTERM);
6975 s = scan_str(s,!!PL_madskills,FALSE);
6978 pl_yylval.ival = OP_CONST;
6979 TERM(sublex_start());
6985 s = scan_str(s,!!PL_madskills,FALSE);
6988 PL_expect = XOPERATOR;
6990 if (SvCUR(PL_lex_stuff)) {
6993 d = SvPV_force(PL_lex_stuff, len);
6995 for (; isSPACE(*d) && len; --len, ++d)
7000 if (!warned && ckWARN(WARN_QW)) {
7001 for (; !isSPACE(*d) && len; --len, ++d) {
7003 Perl_warner(aTHX_ packWARN(WARN_QW),
7004 "Possible attempt to separate words with commas");
7007 else if (*d == '#') {
7008 Perl_warner(aTHX_ packWARN(WARN_QW),
7009 "Possible attempt to put comments in qw() list");
7015 for (; !isSPACE(*d) && len; --len, ++d)
7018 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7019 words = append_elem(OP_LIST, words,
7020 newSVOP(OP_CONST, 0, tokeq(sv)));
7024 start_force(PL_curforce);
7025 NEXTVAL_NEXTTOKE.opval = words;
7030 SvREFCNT_dec(PL_lex_stuff);
7031 PL_lex_stuff = NULL;
7037 s = scan_str(s,!!PL_madskills,FALSE);
7040 pl_yylval.ival = OP_STRINGIFY;
7041 if (SvIVX(PL_lex_stuff) == '\'')
7042 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7043 TERM(sublex_start());
7046 s = scan_pat(s,OP_QR);
7047 TERM(sublex_start());
7050 s = scan_str(s,!!PL_madskills,FALSE);
7053 readpipe_override();
7054 TERM(sublex_start());
7062 s = force_version(s, FALSE);
7064 else if (*s != 'v' || !isDIGIT(s[1])
7065 || (s = force_version(s, TRUE), *s == 'v'))
7067 *PL_tokenbuf = '\0';
7068 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7069 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7070 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7072 yyerror("<> should be quotes");
7074 if (orig_keyword == KEY_require) {
7082 PL_last_uni = PL_oldbufptr;
7083 PL_last_lop_op = OP_REQUIRE;
7085 return REPORT( (int)REQUIRE );
7091 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7095 LOP(OP_RENAME,XTERM);
7104 LOP(OP_RINDEX,XTERM);
7113 UNIDOR(OP_READLINE);
7116 UNIDOR(OP_BACKTICK);
7125 LOP(OP_REVERSE,XTERM);
7128 UNIDOR(OP_READLINK);
7135 if (pl_yylval.opval)
7136 TERM(sublex_start());
7138 TOKEN(1); /* force error */
7141 checkcomma(s,PL_tokenbuf,"filehandle");
7151 LOP(OP_SELECT,XTERM);
7157 LOP(OP_SEMCTL,XTERM);
7160 LOP(OP_SEMGET,XTERM);
7163 LOP(OP_SEMOP,XTERM);
7169 LOP(OP_SETPGRP,XTERM);
7171 case KEY_setpriority:
7172 LOP(OP_SETPRIORITY,XTERM);
7174 case KEY_sethostent:
7180 case KEY_setservent:
7183 case KEY_setprotoent:
7193 LOP(OP_SEEKDIR,XTERM);
7195 case KEY_setsockopt:
7196 LOP(OP_SSOCKOPT,XTERM);
7202 LOP(OP_SHMCTL,XTERM);
7205 LOP(OP_SHMGET,XTERM);
7208 LOP(OP_SHMREAD,XTERM);
7211 LOP(OP_SHMWRITE,XTERM);
7214 LOP(OP_SHUTDOWN,XTERM);
7223 LOP(OP_SOCKET,XTERM);
7225 case KEY_socketpair:
7226 LOP(OP_SOCKPAIR,XTERM);
7229 checkcomma(s,PL_tokenbuf,"subroutine name");
7231 if (*s == ';' || *s == ')') /* probably a close */
7232 Perl_croak(aTHX_ "sort is now a reserved word");
7234 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7238 LOP(OP_SPLIT,XTERM);
7241 LOP(OP_SPRINTF,XTERM);
7244 LOP(OP_SPLICE,XTERM);
7259 LOP(OP_SUBSTR,XTERM);
7265 char tmpbuf[sizeof PL_tokenbuf];
7266 SSize_t tboffset = 0;
7267 expectation attrful;
7268 bool have_name, have_proto;
7269 const int key = tmp;
7274 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7275 SV *subtoken = newSVpvn(tstart, s - tstart);
7279 s = SKIPSPACE2(s,tmpwhite);
7284 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7285 (*s == ':' && s[1] == ':'))
7288 SV *nametoke = NULL;
7292 attrful = XATTRBLOCK;
7293 /* remember buffer pos'n for later force_word */
7294 tboffset = s - PL_oldbufptr;
7295 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7298 nametoke = newSVpvn(s, d - s);
7300 if (memchr(tmpbuf, ':', len))
7301 sv_setpvn(PL_subname, tmpbuf, len);
7303 sv_setsv(PL_subname,PL_curstname);
7304 sv_catpvs(PL_subname,"::");
7305 sv_catpvn(PL_subname,tmpbuf,len);
7312 CURMAD('X', nametoke);
7313 CURMAD('_', tmpwhite);
7314 (void) force_word(PL_oldbufptr + tboffset, WORD,
7317 s = SKIPSPACE2(d,tmpwhite);
7324 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7325 PL_expect = XTERMBLOCK;
7326 attrful = XATTRTERM;
7327 sv_setpvs(PL_subname,"?");
7331 if (key == KEY_format) {
7333 PL_lex_formbrack = PL_lex_brackets + 1;
7335 PL_thistoken = subtoken;
7339 (void) force_word(PL_oldbufptr + tboffset, WORD,
7345 /* Look for a prototype */
7348 bool bad_proto = FALSE;
7349 bool in_brackets = FALSE;
7350 char greedy_proto = ' ';
7351 bool proto_after_greedy_proto = FALSE;
7352 bool must_be_last = FALSE;
7353 bool underscore = FALSE;
7354 bool seen_underscore = FALSE;
7355 const bool warnsyntax = ckWARN(WARN_SYNTAX);
7357 s = scan_str(s,!!PL_madskills,FALSE);
7359 Perl_croak(aTHX_ "Prototype not terminated");
7360 /* strip spaces and check for bad characters */
7361 d = SvPVX(PL_lex_stuff);
7363 for (p = d; *p; ++p) {
7369 proto_after_greedy_proto = TRUE;
7370 if (!strchr("$@%*;[]&\\_", *p)) {
7382 else if ( *p == ']' ) {
7383 in_brackets = FALSE;
7385 else if ( (*p == '@' || *p == '%') &&
7386 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7388 must_be_last = TRUE;
7391 else if ( *p == '_' ) {
7392 underscore = seen_underscore = TRUE;
7399 if (proto_after_greedy_proto)
7400 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7401 "Prototype after '%c' for %"SVf" : %s",
7402 greedy_proto, SVfARG(PL_subname), d);
7404 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7405 "Illegal character %sin prototype for %"SVf" : %s",
7406 seen_underscore ? "after '_' " : "",
7407 SVfARG(PL_subname), d);
7408 SvCUR_set(PL_lex_stuff, tmp);
7413 CURMAD('q', PL_thisopen);
7414 CURMAD('_', tmpwhite);
7415 CURMAD('=', PL_thisstuff);
7416 CURMAD('Q', PL_thisclose);
7417 NEXTVAL_NEXTTOKE.opval =
7418 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7419 PL_lex_stuff = NULL;
7422 s = SKIPSPACE2(s,tmpwhite);
7430 if (*s == ':' && s[1] != ':')
7431 PL_expect = attrful;
7432 else if (*s != '{' && key == KEY_sub) {
7434 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7436 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7443 curmad('^', newSVpvs(""));
7444 CURMAD('_', tmpwhite);
7448 PL_thistoken = subtoken;
7451 NEXTVAL_NEXTTOKE.opval =
7452 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7453 PL_lex_stuff = NULL;
7459 sv_setpvs(PL_subname, "__ANON__");
7461 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7465 (void) force_word(PL_oldbufptr + tboffset, WORD,
7474 LOP(OP_SYSTEM,XREF);
7477 LOP(OP_SYMLINK,XTERM);
7480 LOP(OP_SYSCALL,XTERM);
7483 LOP(OP_SYSOPEN,XTERM);
7486 LOP(OP_SYSSEEK,XTERM);
7489 LOP(OP_SYSREAD,XTERM);
7492 LOP(OP_SYSWRITE,XTERM);
7496 TERM(sublex_start());
7517 LOP(OP_TRUNCATE,XTERM);
7529 pl_yylval.ival = CopLINE(PL_curcop);
7533 pl_yylval.ival = CopLINE(PL_curcop);
7537 LOP(OP_UNLINK,XTERM);
7543 LOP(OP_UNPACK,XTERM);
7546 LOP(OP_UTIME,XTERM);
7552 LOP(OP_UNSHIFT,XTERM);
7555 s = tokenize_use(1, s);
7565 pl_yylval.ival = CopLINE(PL_curcop);
7569 pl_yylval.ival = CopLINE(PL_curcop);
7573 PL_hints |= HINT_BLOCK_SCOPE;
7580 LOP(OP_WAITPID,XTERM);
7589 ctl_l[0] = toCTRL('L');
7591 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7594 /* Make sure $^L is defined */
7595 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7600 if (PL_expect == XOPERATOR)
7606 pl_yylval.ival = OP_XOR;
7611 TERM(sublex_start());
7616 #pragma segment Main
7620 S_pending_ident(pTHX)
7625 /* pit holds the identifier we read and pending_ident is reset */
7626 char pit = PL_pending_ident;
7627 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7628 /* All routes through this function want to know if there is a colon. */
7629 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7630 PL_pending_ident = 0;
7632 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7633 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7634 "### Pending identifier '%s'\n", PL_tokenbuf); });
7636 /* if we're in a my(), we can't allow dynamics here.
7637 $foo'bar has already been turned into $foo::bar, so
7638 just check for colons.
7640 if it's a legal name, the OP is a PADANY.
7643 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7645 yyerror(Perl_form(aTHX_ "No package name allowed for "
7646 "variable %s in \"our\"",
7648 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7652 yyerror(Perl_form(aTHX_ PL_no_myglob,
7653 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7655 pl_yylval.opval = newOP(OP_PADANY, 0);
7656 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7662 build the ops for accesses to a my() variable.
7664 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7665 then used in a comparison. This catches most, but not
7666 all cases. For instance, it catches
7667 sort { my($a); $a <=> $b }
7669 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7670 (although why you'd do that is anyone's guess).
7675 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7676 if (tmp != NOT_IN_PAD) {
7677 /* might be an "our" variable" */
7678 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7679 /* build ops for a bareword */
7680 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7681 HEK * const stashname = HvNAME_HEK(stash);
7682 SV * const sym = newSVhek(stashname);
7683 sv_catpvs(sym, "::");
7684 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7685 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7686 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7689 ? (GV_ADDMULTI | GV_ADDINEVAL)
7692 ((PL_tokenbuf[0] == '$') ? SVt_PV
7693 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7698 /* if it's a sort block and they're naming $a or $b */
7699 if (PL_last_lop_op == OP_SORT &&
7700 PL_tokenbuf[0] == '$' &&
7701 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7704 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7705 d < PL_bufend && *d != '\n';
7708 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7709 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7715 pl_yylval.opval = newOP(OP_PADANY, 0);
7716 pl_yylval.opval->op_targ = tmp;
7722 Whine if they've said @foo in a doublequoted string,
7723 and @foo isn't a variable we can find in the symbol
7726 if (ckWARN(WARN_AMBIGUOUS) &&
7727 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7728 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7730 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7731 /* DO NOT warn for @- and @+ */
7732 && !( PL_tokenbuf[2] == '\0' &&
7733 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7736 /* Downgraded from fatal to warning 20000522 mjd */
7737 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7738 "Possible unintended interpolation of %s in string",
7743 /* build ops for a bareword */
7744 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7746 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7748 PL_tokenbuf + 1, tokenbuf_len - 1,
7749 /* If the identifier refers to a stash, don't autovivify it.
7750 * Change 24660 had the side effect of causing symbol table
7751 * hashes to always be defined, even if they were freshly
7752 * created and the only reference in the entire program was
7753 * the single statement with the defined %foo::bar:: test.
7754 * It appears that all code in the wild doing this actually
7755 * wants to know whether sub-packages have been loaded, so
7756 * by avoiding auto-vivifying symbol tables, we ensure that
7757 * defined %foo::bar:: continues to be false, and the existing
7758 * tests still give the expected answers, even though what
7759 * they're actually testing has now changed subtly.
7761 (*PL_tokenbuf == '%'
7762 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7765 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7766 ((PL_tokenbuf[0] == '$') ? SVt_PV
7767 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7773 * The following code was generated by perl_keyword.pl.
7777 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7781 PERL_ARGS_ASSERT_KEYWORD;
7785 case 1: /* 5 tokens of length 1 */
7817 case 2: /* 18 tokens of length 2 */
7963 case 3: /* 29 tokens of length 3 */
7967 if (name[1] == 'N' &&
8030 if (name[1] == 'i' &&
8062 if (name[1] == 'o' &&
8071 if (name[1] == 'e' &&
8080 if (name[1] == 'n' &&
8089 if (name[1] == 'o' &&
8098 if (name[1] == 'a' &&
8107 if (name[1] == 'o' &&
8169 if (name[1] == 'e' &&
8183 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8209 if (name[1] == 'i' &&
8218 if (name[1] == 's' &&
8227 if (name[1] == 'e' &&
8236 if (name[1] == 'o' &&
8248 case 4: /* 41 tokens of length 4 */
8252 if (name[1] == 'O' &&
8262 if (name[1] == 'N' &&
8272 if (name[1] == 'i' &&
8282 if (name[1] == 'h' &&
8292 if (name[1] == 'u' &&
8305 if (name[2] == 'c' &&
8314 if (name[2] == 's' &&
8323 if (name[2] == 'a' &&
8359 if (name[1] == 'o' &&
8372 if (name[2] == 't' &&
8381 if (name[2] == 'o' &&
8390 if (name[2] == 't' &&
8399 if (name[2] == 'e' &&
8412 if (name[1] == 'o' &&
8425 if (name[2] == 'y' &&
8434 if (name[2] == 'l' &&
8450 if (name[2] == 's' &&
8459 if (name[2] == 'n' &&
8468 if (name[2] == 'c' &&
8481 if (name[1] == 'e' &&
8491 if (name[1] == 'p' &&
8504 if (name[2] == 'c' &&
8513 if (name[2] == 'p' &&
8522 if (name[2] == 's' &&
8538 if (name[2] == 'n' &&
8608 if (name[2] == 'r' &&
8617 if (name[2] == 'r' &&
8626 if (name[2] == 'a' &&
8642 if (name[2] == 'l' &&
8704 if (name[2] == 'e' &&
8707 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8720 case 5: /* 39 tokens of length 5 */
8724 if (name[1] == 'E' &&
8735 if (name[1] == 'H' &&
8749 if (name[2] == 'a' &&
8759 if (name[2] == 'a' &&
8776 if (name[2] == 'e' &&
8786 if (name[2] == 'e' &&
8790 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8806 if (name[3] == 'i' &&
8815 if (name[3] == 'o' &&
8851 if (name[2] == 'o' &&
8861 if (name[2] == 'y' &&
8875 if (name[1] == 'l' &&
8889 if (name[2] == 'n' &&
8899 if (name[2] == 'o' &&
8913 if (name[1] == 'i' &&
8918 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8927 if (name[2] == 'd' &&
8937 if (name[2] == 'c' &&
8954 if (name[2] == 'c' &&
8964 if (name[2] == 't' &&
8978 if (name[1] == 'k' &&
8989 if (name[1] == 'r' &&
9003 if (name[2] == 's' &&
9013 if (name[2] == 'd' &&
9030 if (name[2] == 'm' &&
9040 if (name[2] == 'i' &&
9050 if (name[2] == 'e' &&
9060 if (name[2] == 'l' &&
9070 if (name[2] == 'a' &&
9083 if (name[3] == 't' &&
9086 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9092 if (name[3] == 'd' &&
9109 if (name[1] == 'i' &&
9123 if (name[2] == 'a' &&
9136 if (name[3] == 'e' &&
9171 if (name[2] == 'i' &&
9188 if (name[2] == 'i' &&
9198 if (name[2] == 'i' &&
9215 case 6: /* 33 tokens of length 6 */
9219 if (name[1] == 'c' &&
9234 if (name[2] == 'l' &&
9245 if (name[2] == 'r' &&
9260 if (name[1] == 'e' &&
9275 if (name[2] == 's' &&
9280 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9286 if (name[2] == 'i' &&
9304 if (name[2] == 'l' &&
9315 if (name[2] == 'r' &&
9330 if (name[1] == 'm' &&
9345 if (name[2] == 'n' &&
9356 if (name[2] == 's' &&
9371 if (name[1] == 's' &&
9377 if (name[4] == 't' &&
9386 if (name[4] == 'e' &&
9395 if (name[4] == 'c' &&
9404 if (name[4] == 'n' &&
9420 if (name[1] == 'r' &&
9438 if (name[3] == 'a' &&
9448 if (name[3] == 'u' &&
9462 if (name[2] == 'n' &&
9480 if (name[2] == 'a' &&
9494 if (name[3] == 'e' &&
9507 if (name[4] == 't' &&
9516 if (name[4] == 'e' &&
9538 if (name[4] == 't' &&
9547 if (name[4] == 'e' &&
9563 if (name[2] == 'c' &&
9574 if (name[2] == 'l' &&
9585 if (name[2] == 'b' &&
9596 if (name[2] == 's' &&
9619 if (name[4] == 's' &&
9628 if (name[4] == 'n' &&
9641 if (name[3] == 'a' &&
9658 if (name[1] == 'a' &&
9673 case 7: /* 29 tokens of length 7 */
9677 if (name[1] == 'E' &&
9690 if (name[1] == '_' &&
9703 if (name[1] == 'i' &&
9710 return -KEY_binmode;
9716 if (name[1] == 'o' &&
9723 return -KEY_connect;
9732 if (name[2] == 'm' &&
9738 return -KEY_dbmopen;
9749 if (name[4] == 'u' &&
9753 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9759 if (name[4] == 'n' &&
9780 if (name[1] == 'o' &&
9793 if (name[1] == 'e' &&
9800 if (name[5] == 'r' &&
9803 return -KEY_getpgrp;
9809 if (name[5] == 'i' &&
9812 return -KEY_getppid;
9825 if (name[1] == 'c' &&
9832 return -KEY_lcfirst;
9838 if (name[1] == 'p' &&
9845 return -KEY_opendir;
9851 if (name[1] == 'a' &&
9869 if (name[3] == 'd' &&
9874 return -KEY_readdir;
9880 if (name[3] == 'u' &&
9891 if (name[3] == 'e' &&
9896 return -KEY_reverse;
9915 if (name[3] == 'k' &&
9920 return -KEY_seekdir;
9926 if (name[3] == 'p' &&
9931 return -KEY_setpgrp;
9941 if (name[2] == 'm' &&
9947 return -KEY_shmread;
9953 if (name[2] == 'r' &&
9959 return -KEY_sprintf;
9968 if (name[3] == 'l' &&
9973 return -KEY_symlink;
9982 if (name[4] == 'a' &&
9986 return -KEY_syscall;
9992 if (name[4] == 'p' &&
9996 return -KEY_sysopen;
10002 if (name[4] == 'e' &&
10006 return -KEY_sysread;
10012 if (name[4] == 'e' &&
10016 return -KEY_sysseek;
10034 if (name[1] == 'e' &&
10041 return -KEY_telldir;
10050 if (name[2] == 'f' &&
10056 return -KEY_ucfirst;
10062 if (name[2] == 's' &&
10068 return -KEY_unshift;
10078 if (name[1] == 'a' &&
10085 return -KEY_waitpid;
10094 case 8: /* 26 tokens of length 8 */
10098 if (name[1] == 'U' &&
10106 return KEY_AUTOLOAD;
10112 if (name[1] == '_')
10117 if (name[3] == 'A' &&
10123 return KEY___DATA__;
10129 if (name[3] == 'I' &&
10135 return -KEY___FILE__;
10141 if (name[3] == 'I' &&
10147 return -KEY___LINE__;
10163 if (name[2] == 'o' &&
10170 return -KEY_closedir;
10176 if (name[2] == 'n' &&
10183 return -KEY_continue;
10193 if (name[1] == 'b' &&
10201 return -KEY_dbmclose;
10207 if (name[1] == 'n' &&
10213 if (name[4] == 'r' &&
10218 return -KEY_endgrent;
10224 if (name[4] == 'w' &&
10229 return -KEY_endpwent;
10242 if (name[1] == 'o' &&
10250 return -KEY_formline;
10256 if (name[1] == 'e' &&
10262 if (name[4] == 'r')
10267 if (name[6] == 'n' &&
10270 return -KEY_getgrent;
10276 if (name[6] == 'i' &&
10279 return -KEY_getgrgid;
10285 if (name[6] == 'a' &&
10288 return -KEY_getgrnam;
10301 if (name[4] == 'o' &&
10306 return -KEY_getlogin;
10312 if (name[4] == 'w')
10317 if (name[6] == 'n' &&
10320 return -KEY_getpwent;
10326 if (name[6] == 'a' &&
10329 return -KEY_getpwnam;
10335 if (name[6] == 'i' &&
10338 return -KEY_getpwuid;
10358 if (name[1] == 'e' &&
10365 if (name[5] == 'i' &&
10372 return -KEY_readline;
10377 return -KEY_readlink;
10388 if (name[5] == 'i' &&
10392 return -KEY_readpipe;
10408 if (name[2] == 't')
10413 if (name[4] == 'r' &&
10418 return -KEY_setgrent;
10424 if (name[4] == 'w' &&
10429 return -KEY_setpwent;
10445 if (name[3] == 'w' &&
10451 return -KEY_shmwrite;
10457 if (name[3] == 't' &&
10463 return -KEY_shutdown;
10473 if (name[2] == 's' &&
10480 return -KEY_syswrite;
10490 if (name[1] == 'r' &&
10498 return -KEY_truncate;
10507 case 9: /* 9 tokens of length 9 */
10511 if (name[1] == 'N' &&
10520 return KEY_UNITCHECK;
10526 if (name[1] == 'n' &&
10535 return -KEY_endnetent;
10541 if (name[1] == 'e' &&
10550 return -KEY_getnetent;
10556 if (name[1] == 'o' &&
10565 return -KEY_localtime;
10571 if (name[1] == 'r' &&
10580 return KEY_prototype;
10586 if (name[1] == 'u' &&
10595 return -KEY_quotemeta;
10601 if (name[1] == 'e' &&
10610 return -KEY_rewinddir;
10616 if (name[1] == 'e' &&
10625 return -KEY_setnetent;
10631 if (name[1] == 'a' &&
10640 return -KEY_wantarray;
10649 case 10: /* 9 tokens of length 10 */
10653 if (name[1] == 'n' &&
10659 if (name[4] == 'o' &&
10666 return -KEY_endhostent;
10672 if (name[4] == 'e' &&
10679 return -KEY_endservent;
10692 if (name[1] == 'e' &&
10698 if (name[4] == 'o' &&
10705 return -KEY_gethostent;
10714 if (name[5] == 'r' &&
10720 return -KEY_getservent;
10726 if (name[5] == 'c' &&
10732 return -KEY_getsockopt;
10752 if (name[2] == 't')
10757 if (name[4] == 'o' &&
10764 return -KEY_sethostent;
10773 if (name[5] == 'r' &&
10779 return -KEY_setservent;
10785 if (name[5] == 'c' &&
10791 return -KEY_setsockopt;
10808 if (name[2] == 'c' &&
10817 return -KEY_socketpair;
10830 case 11: /* 8 tokens of length 11 */
10834 if (name[1] == '_' &&
10844 { /* __PACKAGE__ */
10845 return -KEY___PACKAGE__;
10851 if (name[1] == 'n' &&
10861 { /* endprotoent */
10862 return -KEY_endprotoent;
10868 if (name[1] == 'e' &&
10877 if (name[5] == 'e' &&
10883 { /* getpeername */
10884 return -KEY_getpeername;
10893 if (name[6] == 'o' &&
10898 { /* getpriority */
10899 return -KEY_getpriority;
10905 if (name[6] == 't' &&
10910 { /* getprotoent */
10911 return -KEY_getprotoent;
10925 if (name[4] == 'o' &&
10932 { /* getsockname */
10933 return -KEY_getsockname;
10946 if (name[1] == 'e' &&
10954 if (name[6] == 'o' &&
10959 { /* setpriority */
10960 return -KEY_setpriority;
10966 if (name[6] == 't' &&
10971 { /* setprotoent */
10972 return -KEY_setprotoent;
10988 case 12: /* 2 tokens of length 12 */
10989 if (name[0] == 'g' &&
11001 if (name[9] == 'd' &&
11004 { /* getnetbyaddr */
11005 return -KEY_getnetbyaddr;
11011 if (name[9] == 'a' &&
11014 { /* getnetbyname */
11015 return -KEY_getnetbyname;
11027 case 13: /* 4 tokens of length 13 */
11028 if (name[0] == 'g' &&
11035 if (name[4] == 'o' &&
11044 if (name[10] == 'd' &&
11047 { /* gethostbyaddr */
11048 return -KEY_gethostbyaddr;
11054 if (name[10] == 'a' &&
11057 { /* gethostbyname */
11058 return -KEY_gethostbyname;
11071 if (name[4] == 'e' &&
11080 if (name[10] == 'a' &&
11083 { /* getservbyname */
11084 return -KEY_getservbyname;
11090 if (name[10] == 'o' &&
11093 { /* getservbyport */
11094 return -KEY_getservbyport;
11113 case 14: /* 1 tokens of length 14 */
11114 if (name[0] == 'g' &&
11128 { /* getprotobyname */
11129 return -KEY_getprotobyname;
11134 case 16: /* 1 tokens of length 16 */
11135 if (name[0] == 'g' &&
11151 { /* getprotobynumber */
11152 return -KEY_getprotobynumber;
11166 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11170 PERL_ARGS_ASSERT_CHECKCOMMA;
11172 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11173 if (ckWARN(WARN_SYNTAX)) {
11176 for (w = s+2; *w && level; w++) {
11179 else if (*w == ')')
11182 while (isSPACE(*w))
11184 /* the list of chars below is for end of statements or
11185 * block / parens, boolean operators (&&, ||, //) and branch
11186 * constructs (or, and, if, until, unless, while, err, for).
11187 * Not a very solid hack... */
11188 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11189 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11190 "%s (...) interpreted as function",name);
11193 while (s < PL_bufend && isSPACE(*s))
11197 while (s < PL_bufend && isSPACE(*s))
11199 if (isIDFIRST_lazy_if(s,UTF)) {
11200 const char * const w = s++;
11201 while (isALNUM_lazy_if(s,UTF))
11203 while (s < PL_bufend && isSPACE(*s))
11207 if (keyword(w, s - w, 0))
11210 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11211 if (gv && GvCVu(gv))
11213 Perl_croak(aTHX_ "No comma allowed after %s", what);
11218 /* Either returns sv, or mortalizes sv and returns a new SV*.
11219 Best used as sv=new_constant(..., sv, ...).
11220 If s, pv are NULL, calls subroutine with one argument,
11221 and type is used with error messages only. */
11224 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11225 SV *sv, SV *pv, const char *type, STRLEN typelen)
11228 HV * const table = GvHV(PL_hintgv); /* ^H */
11232 const char *why1 = "", *why2 = "", *why3 = "";
11234 PERL_ARGS_ASSERT_NEW_CONSTANT;
11236 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11239 why2 = (const char *)
11240 (strEQ(key,"charnames")
11241 ? "(possibly a missing \"use charnames ...\")"
11243 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11244 (type ? type: "undef"), why2);
11246 /* This is convoluted and evil ("goto considered harmful")
11247 * but I do not understand the intricacies of all the different
11248 * failure modes of %^H in here. The goal here is to make
11249 * the most probable error message user-friendly. --jhi */
11254 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11255 (type ? type: "undef"), why1, why2, why3);
11257 yyerror(SvPVX_const(msg));
11261 cvp = hv_fetch(table, key, keylen, FALSE);
11262 if (!cvp || !SvOK(*cvp)) {
11265 why3 = "} is not defined";
11268 sv_2mortal(sv); /* Parent created it permanently */
11271 pv = newSVpvn_flags(s, len, SVs_TEMP);
11273 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11275 typesv = &PL_sv_undef;
11277 PUSHSTACKi(PERLSI_OVERLOAD);
11289 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11293 /* Check the eval first */
11294 if (!PL_in_eval && SvTRUE(ERRSV)) {
11295 sv_catpvs(ERRSV, "Propagated");
11296 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11298 res = SvREFCNT_inc_simple(sv);
11302 SvREFCNT_inc_simple_void(res);
11311 why1 = "Call to &{$^H{";
11313 why3 = "}} did not return a defined value";
11321 /* Returns a NUL terminated string, with the length of the string written to
11325 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11328 register char *d = dest;
11329 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11331 PERL_ARGS_ASSERT_SCAN_WORD;
11335 Perl_croak(aTHX_ ident_too_long);
11336 if (isALNUM(*s)) /* UTF handled below */
11338 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11343 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11347 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11348 char *t = s + UTF8SKIP(s);
11350 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11354 Perl_croak(aTHX_ ident_too_long);
11355 Copy(s, d, len, char);
11368 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11371 char *bracket = NULL;
11373 register char *d = dest;
11374 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11376 PERL_ARGS_ASSERT_SCAN_IDENT;
11381 while (isDIGIT(*s)) {
11383 Perl_croak(aTHX_ ident_too_long);
11390 Perl_croak(aTHX_ ident_too_long);
11391 if (isALNUM(*s)) /* UTF handled below */
11393 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11398 else if (*s == ':' && s[1] == ':') {
11402 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11403 char *t = s + UTF8SKIP(s);
11404 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11406 if (d + (t - s) > e)
11407 Perl_croak(aTHX_ ident_too_long);
11408 Copy(s, d, t - s, char);
11419 if (PL_lex_state != LEX_NORMAL)
11420 PL_lex_state = LEX_INTERPENDMAYBE;
11423 if (*s == '$' && s[1] &&
11424 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11437 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11442 if (isSPACE(s[-1])) {
11444 const char ch = *s++;
11445 if (!SPACE_OR_TAB(ch)) {
11451 if (isIDFIRST_lazy_if(d,UTF)) {
11455 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11456 end += UTF8SKIP(end);
11457 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11458 end += UTF8SKIP(end);
11460 Copy(s, d, end - s, char);
11465 while ((isALNUM(*s) || *s == ':') && d < e)
11468 Perl_croak(aTHX_ ident_too_long);
11471 while (s < send && SPACE_OR_TAB(*s))
11473 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11474 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11475 const char * const brack =
11477 ((*s == '[') ? "[...]" : "{...}");
11478 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11479 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11480 funny, dest, brack, funny, dest, brack);
11483 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11487 /* Handle extended ${^Foo} variables
11488 * 1999-02-27 mjd-perl-patch@plover.com */
11489 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11493 while (isALNUM(*s) && d < e) {
11497 Perl_croak(aTHX_ ident_too_long);
11502 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11503 PL_lex_state = LEX_INTERPEND;
11506 if (PL_lex_state == LEX_NORMAL) {
11507 if (ckWARN(WARN_AMBIGUOUS) &&
11508 (keyword(dest, d - dest, 0)
11509 || get_cvn_flags(dest, d - dest, 0)))
11513 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11514 "Ambiguous use of %c{%s} resolved to %c%s",
11515 funny, dest, funny, dest);
11520 s = bracket; /* let the parser handle it */
11524 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11525 PL_lex_state = LEX_INTERPEND;
11530 S_pmflag(U32 pmfl, const char ch) {
11532 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11533 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11534 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11535 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11536 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11542 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11544 PERL_ARGS_ASSERT_PMFLAG;
11546 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11547 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11550 *pmfl = S_pmflag(*pmfl, (char)ch);
11555 S_scan_pat(pTHX_ char *start, I32 type)
11559 char *s = scan_str(start,!!PL_madskills,FALSE);
11560 const char * const valid_flags =
11561 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11566 PERL_ARGS_ASSERT_SCAN_PAT;
11569 const char * const delimiter = skipspace(start);
11573 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11574 : "Search pattern not terminated" ));
11577 pm = (PMOP*)newPMOP(type, 0);
11578 if (PL_multi_open == '?') {
11579 /* This is the only point in the code that sets PMf_ONCE: */
11580 pm->op_pmflags |= PMf_ONCE;
11582 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11583 allows us to restrict the list needed by reset to just the ??
11585 assert(type != OP_TRANS);
11587 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11590 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11593 elements = mg->mg_len / sizeof(PMOP**);
11594 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11595 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11596 mg->mg_len = elements * sizeof(PMOP**);
11597 PmopSTASH_set(pm,PL_curstash);
11603 while (*s && strchr(valid_flags, *s))
11604 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11606 if (PL_madskills && modstart != s) {
11607 SV* tmptoken = newSVpvn(modstart, s - modstart);
11608 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11611 /* issue a warning if /c is specified,but /g is not */
11612 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11614 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11615 "Use of /c modifier is meaningless without /g" );
11618 PL_lex_op = (OP*)pm;
11619 pl_yylval.ival = OP_MATCH;
11624 S_scan_subst(pTHX_ char *start)
11635 PERL_ARGS_ASSERT_SCAN_SUBST;
11637 pl_yylval.ival = OP_NULL;
11639 s = scan_str(start,!!PL_madskills,FALSE);
11642 Perl_croak(aTHX_ "Substitution pattern not terminated");
11644 if (s[-1] == PL_multi_open)
11647 if (PL_madskills) {
11648 CURMAD('q', PL_thisopen);
11649 CURMAD('_', PL_thiswhite);
11650 CURMAD('E', PL_thisstuff);
11651 CURMAD('Q', PL_thisclose);
11652 PL_realtokenstart = s - SvPVX(PL_linestr);
11656 first_start = PL_multi_start;
11657 s = scan_str(s,!!PL_madskills,FALSE);
11659 if (PL_lex_stuff) {
11660 SvREFCNT_dec(PL_lex_stuff);
11661 PL_lex_stuff = NULL;
11663 Perl_croak(aTHX_ "Substitution replacement not terminated");
11665 PL_multi_start = first_start; /* so whole substitution is taken together */
11667 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11670 if (PL_madskills) {
11671 CURMAD('z', PL_thisopen);
11672 CURMAD('R', PL_thisstuff);
11673 CURMAD('Z', PL_thisclose);
11679 if (*s == EXEC_PAT_MOD) {
11683 else if (strchr(S_PAT_MODS, *s))
11684 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11690 if (PL_madskills) {
11692 curmad('m', newSVpvn(modstart, s - modstart));
11693 append_madprops(PL_thismad, (OP*)pm, 0);
11697 if ((pm->op_pmflags & PMf_CONTINUE)) {
11698 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11702 SV * const repl = newSVpvs("");
11704 PL_sublex_info.super_bufptr = s;
11705 PL_sublex_info.super_bufend = PL_bufend;
11707 pm->op_pmflags |= PMf_EVAL;
11710 sv_catpvs(repl, "eval ");
11712 sv_catpvs(repl, "do ");
11714 sv_catpvs(repl, "{");
11715 sv_catsv(repl, PL_lex_repl);
11716 if (strchr(SvPVX(PL_lex_repl), '#'))
11717 sv_catpvs(repl, "\n");
11718 sv_catpvs(repl, "}");
11720 SvREFCNT_dec(PL_lex_repl);
11721 PL_lex_repl = repl;
11724 PL_lex_op = (OP*)pm;
11725 pl_yylval.ival = OP_SUBST;
11730 S_scan_trans(pTHX_ char *start)
11743 PERL_ARGS_ASSERT_SCAN_TRANS;
11745 pl_yylval.ival = OP_NULL;
11747 s = scan_str(start,!!PL_madskills,FALSE);
11749 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11751 if (s[-1] == PL_multi_open)
11754 if (PL_madskills) {
11755 CURMAD('q', PL_thisopen);
11756 CURMAD('_', PL_thiswhite);
11757 CURMAD('E', PL_thisstuff);
11758 CURMAD('Q', PL_thisclose);
11759 PL_realtokenstart = s - SvPVX(PL_linestr);
11763 s = scan_str(s,!!PL_madskills,FALSE);
11765 if (PL_lex_stuff) {
11766 SvREFCNT_dec(PL_lex_stuff);
11767 PL_lex_stuff = NULL;
11769 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11771 if (PL_madskills) {
11772 CURMAD('z', PL_thisopen);
11773 CURMAD('R', PL_thisstuff);
11774 CURMAD('Z', PL_thisclose);
11777 complement = del = squash = 0;
11784 complement = OPpTRANS_COMPLEMENT;
11787 del = OPpTRANS_DELETE;
11790 squash = OPpTRANS_SQUASH;
11799 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11800 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11801 o->op_private &= ~OPpTRANS_ALL;
11802 o->op_private |= del|squash|complement|
11803 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11804 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11807 pl_yylval.ival = OP_TRANS;
11810 if (PL_madskills) {
11812 curmad('m', newSVpvn(modstart, s - modstart));
11813 append_madprops(PL_thismad, o, 0);
11822 S_scan_heredoc(pTHX_ register char *s)
11826 I32 op_type = OP_SCALAR;
11830 const char *found_newline;
11834 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11836 I32 stuffstart = s - SvPVX(PL_linestr);
11839 PL_realtokenstart = -1;
11842 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11846 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11850 while (SPACE_OR_TAB(*peek))
11852 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11855 s = delimcpy(d, e, s, PL_bufend, term, &len);
11865 if (!isALNUM_lazy_if(s,UTF))
11866 deprecate("bare << to mean <<\"\"");
11867 for (; isALNUM_lazy_if(s,UTF); s++) {
11872 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11873 Perl_croak(aTHX_ "Delimiter for here document is too long");
11876 len = d - PL_tokenbuf;
11879 if (PL_madskills) {
11880 tstart = PL_tokenbuf + !outer;
11881 PL_thisclose = newSVpvn(tstart, len - !outer);
11882 tstart = SvPVX(PL_linestr) + stuffstart;
11883 PL_thisopen = newSVpvn(tstart, s - tstart);
11884 stuffstart = s - SvPVX(PL_linestr);
11887 #ifndef PERL_STRICT_CR
11888 d = strchr(s, '\r');
11890 char * const olds = s;
11892 while (s < PL_bufend) {
11898 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11907 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11914 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11915 herewas = newSVpvn(s,PL_bufend-s);
11919 herewas = newSVpvn(s-1,found_newline-s+1);
11922 herewas = newSVpvn(s,found_newline-s);
11926 if (PL_madskills) {
11927 tstart = SvPVX(PL_linestr) + stuffstart;
11929 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11931 PL_thisstuff = newSVpvn(tstart, s - tstart);
11934 s += SvCUR(herewas);
11937 stuffstart = s - SvPVX(PL_linestr);
11943 tmpstr = newSV_type(SVt_PVIV);
11944 SvGROW(tmpstr, 80);
11945 if (term == '\'') {
11946 op_type = OP_CONST;
11947 SvIV_set(tmpstr, -1);
11949 else if (term == '`') {
11950 op_type = OP_BACKTICK;
11951 SvIV_set(tmpstr, '\\');
11955 PL_multi_start = CopLINE(PL_curcop);
11956 PL_multi_open = PL_multi_close = '<';
11957 term = *PL_tokenbuf;
11958 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11959 char * const bufptr = PL_sublex_info.super_bufptr;
11960 char * const bufend = PL_sublex_info.super_bufend;
11961 char * const olds = s - SvCUR(herewas);
11962 s = strchr(bufptr, '\n');
11966 while (s < bufend &&
11967 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11969 CopLINE_inc(PL_curcop);
11972 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11973 missingterm(PL_tokenbuf);
11975 sv_setpvn(herewas,bufptr,d-bufptr+1);
11976 sv_setpvn(tmpstr,d+1,s-d);
11978 sv_catpvn(herewas,s,bufend-s);
11979 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11986 while (s < PL_bufend &&
11987 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11989 CopLINE_inc(PL_curcop);
11991 if (s >= PL_bufend) {
11992 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11993 missingterm(PL_tokenbuf);
11995 sv_setpvn(tmpstr,d+1,s-d);
11997 if (PL_madskills) {
11999 sv_catpvn(PL_thisstuff, d + 1, s - d);
12001 PL_thisstuff = newSVpvn(d + 1, s - d);
12002 stuffstart = s - SvPVX(PL_linestr);
12006 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12008 sv_catpvn(herewas,s,PL_bufend-s);
12009 sv_setsv(PL_linestr,herewas);
12010 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12011 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12012 PL_last_lop = PL_last_uni = NULL;
12015 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12016 while (s >= PL_bufend) { /* multiple line string? */
12018 if (PL_madskills) {
12019 tstart = SvPVX(PL_linestr) + stuffstart;
12021 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12023 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12027 CopLINE_inc(PL_curcop);
12028 if (!outer || !lex_next_chunk(0)) {
12029 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12030 missingterm(PL_tokenbuf);
12032 CopLINE_dec(PL_curcop);
12035 stuffstart = s - SvPVX(PL_linestr);
12037 CopLINE_inc(PL_curcop);
12038 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12039 PL_last_lop = PL_last_uni = NULL;
12040 #ifndef PERL_STRICT_CR
12041 if (PL_bufend - PL_linestart >= 2) {
12042 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12043 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12045 PL_bufend[-2] = '\n';
12047 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12049 else if (PL_bufend[-1] == '\r')
12050 PL_bufend[-1] = '\n';
12052 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12053 PL_bufend[-1] = '\n';
12055 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12056 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12057 *(SvPVX(PL_linestr) + off ) = ' ';
12058 sv_catsv(PL_linestr,herewas);
12059 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12060 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12064 sv_catsv(tmpstr,PL_linestr);
12069 PL_multi_end = CopLINE(PL_curcop);
12070 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12071 SvPV_shrink_to_cur(tmpstr);
12073 SvREFCNT_dec(herewas);
12075 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12077 else if (PL_encoding)
12078 sv_recode_to_utf8(tmpstr, PL_encoding);
12080 PL_lex_stuff = tmpstr;
12081 pl_yylval.ival = op_type;
12085 /* scan_inputsymbol
12086 takes: current position in input buffer
12087 returns: new position in input buffer
12088 side-effects: pl_yylval and lex_op are set.
12093 <FH> read from filehandle
12094 <pkg::FH> read from package qualified filehandle
12095 <pkg'FH> read from package qualified filehandle
12096 <$fh> read from filehandle in $fh
12097 <*.h> filename glob
12102 S_scan_inputsymbol(pTHX_ char *start)
12105 register char *s = start; /* current position in buffer */
12108 char *d = PL_tokenbuf; /* start of temp holding space */
12109 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12111 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12113 end = strchr(s, '\n');
12116 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12118 /* die if we didn't have space for the contents of the <>,
12119 or if it didn't end, or if we see a newline
12122 if (len >= (I32)sizeof PL_tokenbuf)
12123 Perl_croak(aTHX_ "Excessively long <> operator");
12125 Perl_croak(aTHX_ "Unterminated <> operator");
12130 Remember, only scalar variables are interpreted as filehandles by
12131 this code. Anything more complex (e.g., <$fh{$num}>) will be
12132 treated as a glob() call.
12133 This code makes use of the fact that except for the $ at the front,
12134 a scalar variable and a filehandle look the same.
12136 if (*d == '$' && d[1]) d++;
12138 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12139 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12142 /* If we've tried to read what we allow filehandles to look like, and
12143 there's still text left, then it must be a glob() and not a getline.
12144 Use scan_str to pull out the stuff between the <> and treat it
12145 as nothing more than a string.
12148 if (d - PL_tokenbuf != len) {
12149 pl_yylval.ival = OP_GLOB;
12150 s = scan_str(start,!!PL_madskills,FALSE);
12152 Perl_croak(aTHX_ "Glob not terminated");
12156 bool readline_overriden = FALSE;
12159 /* we're in a filehandle read situation */
12162 /* turn <> into <ARGV> */
12164 Copy("ARGV",d,5,char);
12166 /* Check whether readline() is overriden */
12167 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12169 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12171 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12172 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12173 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12174 readline_overriden = TRUE;
12176 /* if <$fh>, create the ops to turn the variable into a
12180 /* try to find it in the pad for this block, otherwise find
12181 add symbol table ops
12183 const PADOFFSET tmp = pad_findmy(d, len, 0);
12184 if (tmp != NOT_IN_PAD) {
12185 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12186 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12187 HEK * const stashname = HvNAME_HEK(stash);
12188 SV * const sym = sv_2mortal(newSVhek(stashname));
12189 sv_catpvs(sym, "::");
12190 sv_catpv(sym, d+1);
12195 OP * const o = newOP(OP_PADSV, 0);
12197 PL_lex_op = readline_overriden
12198 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12199 append_elem(OP_LIST, o,
12200 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12201 : (OP*)newUNOP(OP_READLINE, 0, o);
12210 ? (GV_ADDMULTI | GV_ADDINEVAL)
12213 PL_lex_op = readline_overriden
12214 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12215 append_elem(OP_LIST,
12216 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12217 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12218 : (OP*)newUNOP(OP_READLINE, 0,
12219 newUNOP(OP_RV2SV, 0,
12220 newGVOP(OP_GV, 0, gv)));
12222 if (!readline_overriden)
12223 PL_lex_op->op_flags |= OPf_SPECIAL;
12224 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12225 pl_yylval.ival = OP_NULL;
12228 /* If it's none of the above, it must be a literal filehandle
12229 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12231 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12232 PL_lex_op = readline_overriden
12233 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12234 append_elem(OP_LIST,
12235 newGVOP(OP_GV, 0, gv),
12236 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12237 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12238 pl_yylval.ival = OP_NULL;
12247 takes: start position in buffer
12248 keep_quoted preserve \ on the embedded delimiter(s)
12249 keep_delims preserve the delimiters around the string
12250 returns: position to continue reading from buffer
12251 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12252 updates the read buffer.
12254 This subroutine pulls a string out of the input. It is called for:
12255 q single quotes q(literal text)
12256 ' single quotes 'literal text'
12257 qq double quotes qq(interpolate $here please)
12258 " double quotes "interpolate $here please"
12259 qx backticks qx(/bin/ls -l)
12260 ` backticks `/bin/ls -l`
12261 qw quote words @EXPORT_OK = qw( func() $spam )
12262 m// regexp match m/this/
12263 s/// regexp substitute s/this/that/
12264 tr/// string transliterate tr/this/that/
12265 y/// string transliterate y/this/that/
12266 ($*@) sub prototypes sub foo ($)
12267 (stuff) sub attr parameters sub foo : attr(stuff)
12268 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12270 In most of these cases (all but <>, patterns and transliterate)
12271 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12272 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12273 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12276 It skips whitespace before the string starts, and treats the first
12277 character as the delimiter. If the delimiter is one of ([{< then
12278 the corresponding "close" character )]}> is used as the closing
12279 delimiter. It allows quoting of delimiters, and if the string has
12280 balanced delimiters ([{<>}]) it allows nesting.
12282 On success, the SV with the resulting string is put into lex_stuff or,
12283 if that is already non-NULL, into lex_repl. The second case occurs only
12284 when parsing the RHS of the special constructs s/// and tr/// (y///).
12285 For convenience, the terminating delimiter character is stuffed into
12290 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12293 SV *sv; /* scalar value: string */
12294 const char *tmps; /* temp string, used for delimiter matching */
12295 register char *s = start; /* current position in the buffer */
12296 register char term; /* terminating character */
12297 register char *to; /* current position in the sv's data */
12298 I32 brackets = 1; /* bracket nesting level */
12299 bool has_utf8 = FALSE; /* is there any utf8 content? */
12300 I32 termcode; /* terminating char. code */
12301 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12302 STRLEN termlen; /* length of terminating string */
12303 int last_off = 0; /* last position for nesting bracket */
12309 PERL_ARGS_ASSERT_SCAN_STR;
12311 /* skip space before the delimiter */
12317 if (PL_realtokenstart >= 0) {
12318 stuffstart = PL_realtokenstart;
12319 PL_realtokenstart = -1;
12322 stuffstart = start - SvPVX(PL_linestr);
12324 /* mark where we are, in case we need to report errors */
12327 /* after skipping whitespace, the next character is the terminator */
12330 termcode = termstr[0] = term;
12334 termcode = utf8_to_uvchr((U8*)s, &termlen);
12335 Copy(s, termstr, termlen, U8);
12336 if (!UTF8_IS_INVARIANT(term))
12340 /* mark where we are */
12341 PL_multi_start = CopLINE(PL_curcop);
12342 PL_multi_open = term;
12344 /* find corresponding closing delimiter */
12345 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12346 termcode = termstr[0] = term = tmps[5];
12348 PL_multi_close = term;
12350 /* create a new SV to hold the contents. 79 is the SV's initial length.
12351 What a random number. */
12352 sv = newSV_type(SVt_PVIV);
12354 SvIV_set(sv, termcode);
12355 (void)SvPOK_only(sv); /* validate pointer */
12357 /* move past delimiter and try to read a complete string */
12359 sv_catpvn(sv, s, termlen);
12362 tstart = SvPVX(PL_linestr) + stuffstart;
12363 if (!PL_thisopen && !keep_delims) {
12364 PL_thisopen = newSVpvn(tstart, s - tstart);
12365 stuffstart = s - SvPVX(PL_linestr);
12369 if (PL_encoding && !UTF) {
12373 int offset = s - SvPVX_const(PL_linestr);
12374 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12375 &offset, (char*)termstr, termlen);
12376 const char * const ns = SvPVX_const(PL_linestr) + offset;
12377 char * const svlast = SvEND(sv) - 1;
12379 for (; s < ns; s++) {
12380 if (*s == '\n' && !PL_rsfp)
12381 CopLINE_inc(PL_curcop);
12384 goto read_more_line;
12386 /* handle quoted delimiters */
12387 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12389 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12391 if ((svlast-1 - t) % 2) {
12392 if (!keep_quoted) {
12393 *(svlast-1) = term;
12395 SvCUR_set(sv, SvCUR(sv) - 1);
12400 if (PL_multi_open == PL_multi_close) {
12406 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12407 /* At here, all closes are "was quoted" one,
12408 so we don't check PL_multi_close. */
12410 if (!keep_quoted && *(t+1) == PL_multi_open)
12415 else if (*t == PL_multi_open)
12423 SvCUR_set(sv, w - SvPVX_const(sv));
12425 last_off = w - SvPVX(sv);
12426 if (--brackets <= 0)
12431 if (!keep_delims) {
12432 SvCUR_set(sv, SvCUR(sv) - 1);
12438 /* extend sv if need be */
12439 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12440 /* set 'to' to the next character in the sv's string */
12441 to = SvPVX(sv)+SvCUR(sv);
12443 /* if open delimiter is the close delimiter read unbridle */
12444 if (PL_multi_open == PL_multi_close) {
12445 for (; s < PL_bufend; s++,to++) {
12446 /* embedded newlines increment the current line number */
12447 if (*s == '\n' && !PL_rsfp)
12448 CopLINE_inc(PL_curcop);
12449 /* handle quoted delimiters */
12450 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12451 if (!keep_quoted && s[1] == term)
12453 /* any other quotes are simply copied straight through */
12457 /* terminate when run out of buffer (the for() condition), or
12458 have found the terminator */
12459 else if (*s == term) {
12462 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12465 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12471 /* if the terminator isn't the same as the start character (e.g.,
12472 matched brackets), we have to allow more in the quoting, and
12473 be prepared for nested brackets.
12476 /* read until we run out of string, or we find the terminator */
12477 for (; s < PL_bufend; s++,to++) {
12478 /* embedded newlines increment the line count */
12479 if (*s == '\n' && !PL_rsfp)
12480 CopLINE_inc(PL_curcop);
12481 /* backslashes can escape the open or closing characters */
12482 if (*s == '\\' && s+1 < PL_bufend) {
12483 if (!keep_quoted &&
12484 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12489 /* allow nested opens and closes */
12490 else if (*s == PL_multi_close && --brackets <= 0)
12492 else if (*s == PL_multi_open)
12494 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12499 /* terminate the copied string and update the sv's end-of-string */
12501 SvCUR_set(sv, to - SvPVX_const(sv));
12504 * this next chunk reads more into the buffer if we're not done yet
12508 break; /* handle case where we are done yet :-) */
12510 #ifndef PERL_STRICT_CR
12511 if (to - SvPVX_const(sv) >= 2) {
12512 if ((to[-2] == '\r' && to[-1] == '\n') ||
12513 (to[-2] == '\n' && to[-1] == '\r'))
12517 SvCUR_set(sv, to - SvPVX_const(sv));
12519 else if (to[-1] == '\r')
12522 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12527 /* if we're out of file, or a read fails, bail and reset the current
12528 line marker so we can report where the unterminated string began
12531 if (PL_madskills) {
12532 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12534 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12536 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12539 CopLINE_inc(PL_curcop);
12540 PL_bufptr = PL_bufend;
12541 if (!lex_next_chunk(0)) {
12543 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12552 /* at this point, we have successfully read the delimited string */
12554 if (!PL_encoding || UTF) {
12556 if (PL_madskills) {
12557 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12558 const int len = s - tstart;
12560 sv_catpvn(PL_thisstuff, tstart, len);
12562 PL_thisstuff = newSVpvn(tstart, len);
12563 if (!PL_thisclose && !keep_delims)
12564 PL_thisclose = newSVpvn(s,termlen);
12569 sv_catpvn(sv, s, termlen);
12574 if (PL_madskills) {
12575 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12576 const int len = s - tstart - termlen;
12578 sv_catpvn(PL_thisstuff, tstart, len);
12580 PL_thisstuff = newSVpvn(tstart, len);
12581 if (!PL_thisclose && !keep_delims)
12582 PL_thisclose = newSVpvn(s - termlen,termlen);
12586 if (has_utf8 || PL_encoding)
12589 PL_multi_end = CopLINE(PL_curcop);
12591 /* if we allocated too much space, give some back */
12592 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12593 SvLEN_set(sv, SvCUR(sv) + 1);
12594 SvPV_renew(sv, SvLEN(sv));
12597 /* decide whether this is the first or second quoted string we've read
12610 takes: pointer to position in buffer
12611 returns: pointer to new position in buffer
12612 side-effects: builds ops for the constant in pl_yylval.op
12614 Read a number in any of the formats that Perl accepts:
12616 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12617 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12620 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12622 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12625 If it reads a number without a decimal point or an exponent, it will
12626 try converting the number to an integer and see if it can do so
12627 without loss of precision.
12631 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12634 register const char *s = start; /* current position in buffer */
12635 register char *d; /* destination in temp buffer */
12636 register char *e; /* end of temp buffer */
12637 NV nv; /* number read, as a double */
12638 SV *sv = NULL; /* place to put the converted number */
12639 bool floatit; /* boolean: int or float? */
12640 const char *lastub = NULL; /* position of last underbar */
12641 static char const number_too_long[] = "Number too long";
12643 PERL_ARGS_ASSERT_SCAN_NUM;
12645 /* We use the first character to decide what type of number this is */
12649 Perl_croak(aTHX_ "panic: scan_num");
12651 /* if it starts with a 0, it could be an octal number, a decimal in
12652 0.13 disguise, or a hexadecimal number, or a binary number. */
12656 u holds the "number so far"
12657 shift the power of 2 of the base
12658 (hex == 4, octal == 3, binary == 1)
12659 overflowed was the number more than we can hold?
12661 Shift is used when we add a digit. It also serves as an "are
12662 we in octal/hex/binary?" indicator to disallow hex characters
12663 when in octal mode.
12668 bool overflowed = FALSE;
12669 bool just_zero = TRUE; /* just plain 0 or binary number? */
12670 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12671 static const char* const bases[5] =
12672 { "", "binary", "", "octal", "hexadecimal" };
12673 static const char* const Bases[5] =
12674 { "", "Binary", "", "Octal", "Hexadecimal" };
12675 static const char* const maxima[5] =
12677 "0b11111111111111111111111111111111",
12681 const char *base, *Base, *max;
12683 /* check for hex */
12688 } else if (s[1] == 'b') {
12693 /* check for a decimal in disguise */
12694 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12696 /* so it must be octal */
12703 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12704 "Misplaced _ in number");
12708 base = bases[shift];
12709 Base = Bases[shift];
12710 max = maxima[shift];
12712 /* read the rest of the number */
12714 /* x is used in the overflow test,
12715 b is the digit we're adding on. */
12720 /* if we don't mention it, we're done */
12724 /* _ are ignored -- but warned about if consecutive */
12726 if (lastub && s == lastub + 1)
12727 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12728 "Misplaced _ in number");
12732 /* 8 and 9 are not octal */
12733 case '8': case '9':
12735 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12739 case '2': case '3': case '4':
12740 case '5': case '6': case '7':
12742 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12745 case '0': case '1':
12746 b = *s++ & 15; /* ASCII digit -> value of digit */
12750 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12751 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12752 /* make sure they said 0x */
12755 b = (*s++ & 7) + 9;
12757 /* Prepare to put the digit we have onto the end
12758 of the number so far. We check for overflows.
12764 x = u << shift; /* make room for the digit */
12766 if ((x >> shift) != u
12767 && !(PL_hints & HINT_NEW_BINARY)) {
12770 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12771 "Integer overflow in %s number",
12774 u = x | b; /* add the digit to the end */
12777 n *= nvshift[shift];
12778 /* If an NV has not enough bits in its
12779 * mantissa to represent an UV this summing of
12780 * small low-order numbers is a waste of time
12781 * (because the NV cannot preserve the
12782 * low-order bits anyway): we could just
12783 * remember when did we overflow and in the
12784 * end just multiply n by the right
12792 /* if we get here, we had success: make a scalar value from
12797 /* final misplaced underbar check */
12798 if (s[-1] == '_') {
12799 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12804 if (n > 4294967295.0)
12805 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12806 "%s number > %s non-portable",
12812 if (u > 0xffffffff)
12813 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12814 "%s number > %s non-portable",
12819 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12820 sv = new_constant(start, s - start, "integer",
12821 sv, NULL, NULL, 0);
12822 else if (PL_hints & HINT_NEW_BINARY)
12823 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12828 handle decimal numbers.
12829 we're also sent here when we read a 0 as the first digit
12831 case '1': case '2': case '3': case '4': case '5':
12832 case '6': case '7': case '8': case '9': case '.':
12835 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12838 /* read next group of digits and _ and copy into d */
12839 while (isDIGIT(*s) || *s == '_') {
12840 /* skip underscores, checking for misplaced ones
12844 if (lastub && s == lastub + 1)
12845 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12846 "Misplaced _ in number");
12850 /* check for end of fixed-length buffer */
12852 Perl_croak(aTHX_ number_too_long);
12853 /* if we're ok, copy the character */
12858 /* final misplaced underbar check */
12859 if (lastub && s == lastub + 1) {
12860 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12863 /* read a decimal portion if there is one. avoid
12864 3..5 being interpreted as the number 3. followed
12867 if (*s == '.' && s[1] != '.') {
12872 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12873 "Misplaced _ in number");
12877 /* copy, ignoring underbars, until we run out of digits.
12879 for (; isDIGIT(*s) || *s == '_'; s++) {
12880 /* fixed length buffer check */
12882 Perl_croak(aTHX_ number_too_long);
12884 if (lastub && s == lastub + 1)
12885 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12886 "Misplaced _ in number");
12892 /* fractional part ending in underbar? */
12893 if (s[-1] == '_') {
12894 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12895 "Misplaced _ in number");
12897 if (*s == '.' && isDIGIT(s[1])) {
12898 /* oops, it's really a v-string, but without the "v" */
12904 /* read exponent part, if present */
12905 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12909 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12910 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12912 /* stray preinitial _ */
12914 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12915 "Misplaced _ in number");
12919 /* allow positive or negative exponent */
12920 if (*s == '+' || *s == '-')
12923 /* stray initial _ */
12925 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12926 "Misplaced _ in number");
12930 /* read digits of exponent */
12931 while (isDIGIT(*s) || *s == '_') {
12934 Perl_croak(aTHX_ number_too_long);
12938 if (((lastub && s == lastub + 1) ||
12939 (!isDIGIT(s[1]) && s[1] != '_')))
12940 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12941 "Misplaced _ in number");
12948 /* make an sv from the string */
12952 We try to do an integer conversion first if no characters
12953 indicating "float" have been found.
12958 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12960 if (flags == IS_NUMBER_IN_UV) {
12962 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12965 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12966 if (uv <= (UV) IV_MIN)
12967 sv_setiv(sv, -(IV)uv);
12974 /* terminate the string */
12976 nv = Atof(PL_tokenbuf);
12981 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12982 const char *const key = floatit ? "float" : "integer";
12983 const STRLEN keylen = floatit ? 5 : 7;
12984 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12985 key, keylen, sv, NULL, NULL, 0);
12989 /* if it starts with a v, it could be a v-string */
12992 sv = newSV(5); /* preallocate storage space */
12993 s = scan_vstring(s, PL_bufend, sv);
12997 /* make the op for the constant and return */
13000 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13002 lvalp->opval = NULL;
13008 S_scan_formline(pTHX_ register char *s)
13011 register char *eol;
13013 SV * const stuff = newSVpvs("");
13014 bool needargs = FALSE;
13015 bool eofmt = FALSE;
13017 char *tokenstart = s;
13018 SV* savewhite = NULL;
13020 if (PL_madskills) {
13021 savewhite = PL_thiswhite;
13026 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13028 while (!needargs) {
13031 #ifdef PERL_STRICT_CR
13032 while (SPACE_OR_TAB(*t))
13035 while (SPACE_OR_TAB(*t) || *t == '\r')
13038 if (*t == '\n' || t == PL_bufend) {
13043 if (PL_in_eval && !PL_rsfp) {
13044 eol = (char *) memchr(s,'\n',PL_bufend-s);
13049 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13051 for (t = s; t < eol; t++) {
13052 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13054 goto enough; /* ~~ must be first line in formline */
13056 if (*t == '@' || *t == '^')
13060 sv_catpvn(stuff, s, eol-s);
13061 #ifndef PERL_STRICT_CR
13062 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13063 char *end = SvPVX(stuff) + SvCUR(stuff);
13066 SvCUR_set(stuff, SvCUR(stuff) - 1);
13077 if (PL_madskills) {
13079 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13081 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13084 PL_bufptr = PL_bufend;
13085 CopLINE_inc(PL_curcop);
13086 got_some = lex_next_chunk(0);
13087 CopLINE_dec(PL_curcop);
13090 tokenstart = PL_bufptr;
13098 if (SvCUR(stuff)) {
13101 PL_lex_state = LEX_NORMAL;
13102 start_force(PL_curforce);
13103 NEXTVAL_NEXTTOKE.ival = 0;
13107 PL_lex_state = LEX_FORMLINE;
13109 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13111 else if (PL_encoding)
13112 sv_recode_to_utf8(stuff, PL_encoding);
13114 start_force(PL_curforce);
13115 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13117 start_force(PL_curforce);
13118 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13122 SvREFCNT_dec(stuff);
13124 PL_lex_formbrack = 0;
13128 if (PL_madskills) {
13130 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13132 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13133 PL_thiswhite = savewhite;
13140 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13143 const I32 oldsavestack_ix = PL_savestack_ix;
13144 CV* const outsidecv = PL_compcv;
13147 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13149 SAVEI32(PL_subline);
13150 save_item(PL_subname);
13151 SAVESPTR(PL_compcv);
13153 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13154 CvFLAGS(PL_compcv) |= flags;
13156 PL_subline = CopLINE(PL_curcop);
13157 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13158 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13159 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13161 return oldsavestack_ix;
13165 #pragma segment Perl_yylex
13168 S_yywarn(pTHX_ const char *const s)
13172 PERL_ARGS_ASSERT_YYWARN;
13174 PL_in_eval |= EVAL_WARNONLY;
13176 PL_in_eval &= ~EVAL_WARNONLY;
13181 Perl_yyerror(pTHX_ const char *const s)
13184 const char *where = NULL;
13185 const char *context = NULL;
13188 int yychar = PL_parser->yychar;
13190 PERL_ARGS_ASSERT_YYERROR;
13192 if (!yychar || (yychar == ';' && !PL_rsfp))
13194 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13195 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13196 PL_oldbufptr != PL_bufptr) {
13199 The code below is removed for NetWare because it abends/crashes on NetWare
13200 when the script has error such as not having the closing quotes like:
13201 if ($var eq "value)
13202 Checking of white spaces is anyway done in NetWare code.
13205 while (isSPACE(*PL_oldoldbufptr))
13208 context = PL_oldoldbufptr;
13209 contlen = PL_bufptr - PL_oldoldbufptr;
13211 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13212 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13215 The code below is removed for NetWare because it abends/crashes on NetWare
13216 when the script has error such as not having the closing quotes like:
13217 if ($var eq "value)
13218 Checking of white spaces is anyway done in NetWare code.
13221 while (isSPACE(*PL_oldbufptr))
13224 context = PL_oldbufptr;
13225 contlen = PL_bufptr - PL_oldbufptr;
13227 else if (yychar > 255)
13228 where = "next token ???";
13229 else if (yychar == -2) { /* YYEMPTY */
13230 if (PL_lex_state == LEX_NORMAL ||
13231 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13232 where = "at end of line";
13233 else if (PL_lex_inpat)
13234 where = "within pattern";
13236 where = "within string";
13239 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13241 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13242 else if (isPRINT_LC(yychar)) {
13243 const char string = yychar;
13244 sv_catpvn(where_sv, &string, 1);
13247 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13248 where = SvPVX_const(where_sv);
13250 msg = sv_2mortal(newSVpv(s, 0));
13251 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13252 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13254 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13256 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13257 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13258 Perl_sv_catpvf(aTHX_ msg,
13259 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13260 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13263 if (PL_in_eval & EVAL_WARNONLY) {
13264 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13268 if (PL_error_count >= 10) {
13269 if (PL_in_eval && SvCUR(ERRSV))
13270 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13271 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13273 Perl_croak(aTHX_ "%s has too many errors.\n",
13274 OutCopFILE(PL_curcop));
13277 PL_in_my_stash = NULL;
13281 #pragma segment Main
13285 S_swallow_bom(pTHX_ U8 *s)
13288 const STRLEN slen = SvCUR(PL_linestr);
13290 PERL_ARGS_ASSERT_SWALLOW_BOM;
13294 if (s[1] == 0xFE) {
13295 /* UTF-16 little-endian? (or UTF-32LE?) */
13296 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13297 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13298 #ifndef PERL_NO_UTF16_FILTER
13299 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13301 if (PL_bufend > (char*)s) {
13302 s = add_utf16_textfilter(s, TRUE);
13305 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13310 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13311 #ifndef PERL_NO_UTF16_FILTER
13312 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13314 if (PL_bufend > (char *)s) {
13315 s = add_utf16_textfilter(s, FALSE);
13318 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13323 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13324 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13325 s += 3; /* UTF-8 */
13331 if (s[2] == 0xFE && s[3] == 0xFF) {
13332 /* UTF-32 big-endian */
13333 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13336 else if (s[2] == 0 && s[3] != 0) {
13339 * are a good indicator of UTF-16BE. */
13340 #ifndef PERL_NO_UTF16_FILTER
13341 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13342 s = add_utf16_textfilter(s, FALSE);
13344 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13350 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13351 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13352 s += 4; /* UTF-8 */
13358 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13361 * are a good indicator of UTF-16LE. */
13362 #ifndef PERL_NO_UTF16_FILTER
13363 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13364 s = add_utf16_textfilter(s, TRUE);
13366 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13374 #ifndef PERL_NO_UTF16_FILTER
13376 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13379 SV *const filter = FILTER_DATA(idx);
13380 /* We re-use this each time round, throwing the contents away before we
13382 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13383 SV *const utf8_buffer = filter;
13384 IV status = IoPAGE(filter);
13385 const bool reverse = (bool) IoLINES(filter);
13388 /* As we're automatically added, at the lowest level, and hence only called
13389 from this file, we can be sure that we're not called in block mode. Hence
13390 don't bother writing code to deal with block mode. */
13392 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13395 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13397 DEBUG_P(PerlIO_printf(Perl_debug_log,
13398 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13399 FPTR2DPTR(void *, S_utf16_textfilter),
13400 reverse ? 'l' : 'b', idx, maxlen, status,
13401 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13408 /* First, look in our buffer of existing UTF-8 data: */
13409 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13413 } else if (status == 0) {
13415 IoPAGE(filter) = 0;
13416 nl = SvEND(utf8_buffer);
13419 STRLEN got = nl - SvPVX(utf8_buffer);
13420 /* Did we have anything to append? */
13422 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13423 /* Everything else in this code works just fine if SVp_POK isn't
13424 set. This, however, needs it, and we need it to work, else
13425 we loop infinitely because the buffer is never consumed. */
13426 sv_chop(utf8_buffer, nl);
13430 /* OK, not a complete line there, so need to read some more UTF-16.
13431 Read an extra octect if the buffer currently has an odd number. */
13435 if (SvCUR(utf16_buffer) >= 2) {
13436 /* Location of the high octet of the last complete code point.
13437 Gosh, UTF-16 is a pain. All the benefits of variable length,
13438 *coupled* with all the benefits of partial reads and
13440 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13441 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13443 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13447 /* We have the first half of a surrogate. Read more. */
13448 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13451 status = FILTER_READ(idx + 1, utf16_buffer,
13452 160 + (SvCUR(utf16_buffer) & 1));
13453 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13454 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13457 IoPAGE(filter) = status;
13462 chars = SvCUR(utf16_buffer) >> 1;
13463 have = SvCUR(utf8_buffer);
13464 SvGROW(utf8_buffer, have + chars * 3 + 1);
13467 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13468 (U8*)SvPVX_const(utf8_buffer) + have,
13469 chars * 2, &newlen);
13471 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13472 (U8*)SvPVX_const(utf8_buffer) + have,
13473 chars * 2, &newlen);
13475 SvCUR_set(utf8_buffer, have + newlen);
13478 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13479 it's private to us, and utf16_to_utf8{,reversed} take a
13480 (pointer,length) pair, rather than a NUL-terminated string. */
13481 if(SvCUR(utf16_buffer) & 1) {
13482 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13483 SvCUR_set(utf16_buffer, 1);
13485 SvCUR_set(utf16_buffer, 0);
13488 DEBUG_P(PerlIO_printf(Perl_debug_log,
13489 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13491 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13492 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13497 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13499 SV *filter = filter_add(S_utf16_textfilter, NULL);
13501 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13502 sv_setpvs(filter, "");
13503 IoLINES(filter) = reversed;
13504 IoPAGE(filter) = 1; /* Not EOF */
13506 /* Sadly, we have to return a valid pointer, come what may, so we have to
13507 ignore any error return from this. */
13508 SvCUR_set(PL_linestr, 0);
13509 if (FILTER_READ(0, PL_linestr, 0)) {
13510 SvUTF8_on(PL_linestr);
13512 SvUTF8_on(PL_linestr);
13514 PL_bufend = SvEND(PL_linestr);
13515 return (U8*)SvPVX(PL_linestr);
13520 Returns a pointer to the next character after the parsed
13521 vstring, as well as updating the passed in sv.
13523 Function must be called like
13526 s = scan_vstring(s,e,sv);
13528 where s and e are the start and end of the string.
13529 The sv should already be large enough to store the vstring
13530 passed in, for performance reasons.
13535 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13538 const char *pos = s;
13539 const char *start = s;
13541 PERL_ARGS_ASSERT_SCAN_VSTRING;
13543 if (*pos == 'v') pos++; /* get past 'v' */
13544 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13546 if ( *pos != '.') {
13547 /* this may not be a v-string if followed by => */
13548 const char *next = pos;
13549 while (next < e && isSPACE(*next))
13551 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13552 /* return string not v-string */
13553 sv_setpvn(sv,(char *)s,pos-s);
13554 return (char *)pos;
13558 if (!isALPHA(*pos)) {
13559 U8 tmpbuf[UTF8_MAXBYTES+1];
13562 s++; /* get past 'v' */
13567 /* this is atoi() that tolerates underscores */
13570 const char *end = pos;
13572 while (--end >= s) {
13574 const UV orev = rev;
13575 rev += (*end - '0') * mult;
13578 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13579 "Integer overflow in decimal number");
13583 if (rev > 0x7FFFFFFF)
13584 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13586 /* Append native character for the rev point */
13587 tmpend = uvchr_to_utf8(tmpbuf, rev);
13588 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13589 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13591 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13597 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13601 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13608 Perl_keyword_plugin_standard(pTHX_
13609 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13611 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13612 PERL_UNUSED_CONTEXT;
13613 PERL_UNUSED_ARG(keyword_ptr);
13614 PERL_UNUSED_ARG(keyword_len);
13615 PERL_UNUSED_ARG(op_ptr);
13616 return KEYWORD_PLUGIN_DECLINE;
13621 * c-indentation-style: bsd
13622 * c-basic-offset: 4
13623 * indent-tabs-mode: t
13626 * ex: set ts=8 sts=4 sw=4 noet: