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("switch")-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 = SKIPSPACE1(s), *s == ';')) {
3811 start_force(PL_curforce);
3812 NEXTVAL_NEXTTOKE.opval = NULL;
3815 else if (*s == 'v') {
3816 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3817 s = force_version(s, FALSE);
3821 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3822 s = force_version(s, FALSE);
3824 pl_yylval.ival = is_use;
3828 static const char* const exp_name[] =
3829 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3830 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3837 Works out what to call the token just pulled out of the input
3838 stream. The yacc parser takes care of taking the ops we return and
3839 stitching them into a tree.
3845 if read an identifier
3846 if we're in a my declaration
3847 croak if they tried to say my($foo::bar)
3848 build the ops for a my() declaration
3849 if it's an access to a my() variable
3850 are we in a sort block?
3851 croak if my($a); $a <=> $b
3852 build ops for access to a my() variable
3853 if in a dq string, and they've said @foo and we can't find @foo
3855 build ops for a bareword
3856 if we already built the token before, use it.
3861 #pragma segment Perl_yylex
3867 register char *s = PL_bufptr;
3873 /* orig_keyword, gvp, and gv are initialized here because
3874 * jump to the label just_a_word_zero can bypass their
3875 * initialization later. */
3876 I32 orig_keyword = 0;
3881 SV* tmp = newSVpvs("");
3882 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3883 (IV)CopLINE(PL_curcop),
3884 lex_state_names[PL_lex_state],
3885 exp_name[PL_expect],
3886 pv_display(tmp, s, strlen(s), 0, 60));
3889 /* check if there's an identifier for us to look at */
3890 if (PL_pending_ident)
3891 return REPORT(S_pending_ident(aTHX));
3893 /* no identifier pending identification */
3895 switch (PL_lex_state) {
3897 case LEX_NORMAL: /* Some compilers will produce faster */
3898 case LEX_INTERPNORMAL: /* code if we comment these out. */
3902 /* when we've already built the next token, just pull it out of the queue */
3906 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3908 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3909 PL_nexttoke[PL_lasttoke].next_mad = 0;
3910 if (PL_thismad && PL_thismad->mad_key == '_') {
3911 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3912 PL_thismad->mad_val = 0;
3913 mad_free(PL_thismad);
3918 PL_lex_state = PL_lex_defer;
3919 PL_expect = PL_lex_expect;
3920 PL_lex_defer = LEX_NORMAL;
3921 if (!PL_nexttoke[PL_lasttoke].next_type)
3926 pl_yylval = PL_nextval[PL_nexttoke];
3928 PL_lex_state = PL_lex_defer;
3929 PL_expect = PL_lex_expect;
3930 PL_lex_defer = LEX_NORMAL;
3934 /* FIXME - can these be merged? */
3935 return(PL_nexttoke[PL_lasttoke].next_type);
3937 return REPORT(PL_nexttype[PL_nexttoke]);
3940 /* interpolated case modifiers like \L \U, including \Q and \E.
3941 when we get here, PL_bufptr is at the \
3943 case LEX_INTERPCASEMOD:
3945 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3946 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3948 /* handle \E or end of string */
3949 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3951 if (PL_lex_casemods) {
3952 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3953 PL_lex_casestack[PL_lex_casemods] = '\0';
3955 if (PL_bufptr != PL_bufend
3956 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3958 PL_lex_state = LEX_INTERPCONCAT;
3961 PL_thistoken = newSVpvs("\\E");
3967 while (PL_bufptr != PL_bufend &&
3968 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3970 PL_thiswhite = newSVpvs("");
3971 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3975 if (PL_bufptr != PL_bufend)
3978 PL_lex_state = LEX_INTERPCONCAT;
3982 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3983 "### Saw case modifier\n"); });
3985 if (s[1] == '\\' && s[2] == 'E') {
3988 PL_thiswhite = newSVpvs("");
3989 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3992 PL_lex_state = LEX_INTERPCONCAT;
3997 if (!PL_madskills) /* when just compiling don't need correct */
3998 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3999 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4000 if ((*s == 'L' || *s == 'U') &&
4001 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4002 PL_lex_casestack[--PL_lex_casemods] = '\0';
4005 if (PL_lex_casemods > 10)
4006 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4007 PL_lex_casestack[PL_lex_casemods++] = *s;
4008 PL_lex_casestack[PL_lex_casemods] = '\0';
4009 PL_lex_state = LEX_INTERPCONCAT;
4010 start_force(PL_curforce);
4011 NEXTVAL_NEXTTOKE.ival = 0;
4013 start_force(PL_curforce);
4015 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4017 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4019 NEXTVAL_NEXTTOKE.ival = OP_LC;
4021 NEXTVAL_NEXTTOKE.ival = OP_UC;
4023 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4025 Perl_croak(aTHX_ "panic: yylex");
4027 SV* const tmpsv = newSVpvs("\\ ");
4028 /* replace the space with the character we want to escape
4030 SvPVX(tmpsv)[1] = *s;
4036 if (PL_lex_starts) {
4042 sv_free(PL_thistoken);
4043 PL_thistoken = newSVpvs("");
4046 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4047 if (PL_lex_casemods == 1 && PL_lex_inpat)
4056 case LEX_INTERPPUSH:
4057 return REPORT(sublex_push());
4059 case LEX_INTERPSTART:
4060 if (PL_bufptr == PL_bufend)
4061 return REPORT(sublex_done());
4062 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4063 "### Interpolated variable\n"); });
4065 PL_lex_dojoin = (*PL_bufptr == '@');
4066 PL_lex_state = LEX_INTERPNORMAL;
4067 if (PL_lex_dojoin) {
4068 start_force(PL_curforce);
4069 NEXTVAL_NEXTTOKE.ival = 0;
4071 start_force(PL_curforce);
4072 force_ident("\"", '$');
4073 start_force(PL_curforce);
4074 NEXTVAL_NEXTTOKE.ival = 0;
4076 start_force(PL_curforce);
4077 NEXTVAL_NEXTTOKE.ival = 0;
4079 start_force(PL_curforce);
4080 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4083 if (PL_lex_starts++) {
4088 sv_free(PL_thistoken);
4089 PL_thistoken = newSVpvs("");
4092 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4093 if (!PL_lex_casemods && PL_lex_inpat)
4100 case LEX_INTERPENDMAYBE:
4101 if (intuit_more(PL_bufptr)) {
4102 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4108 if (PL_lex_dojoin) {
4109 PL_lex_dojoin = FALSE;
4110 PL_lex_state = LEX_INTERPCONCAT;
4114 sv_free(PL_thistoken);
4115 PL_thistoken = newSVpvs("");
4120 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4121 && SvEVALED(PL_lex_repl))
4123 if (PL_bufptr != PL_bufend)
4124 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4128 case LEX_INTERPCONCAT:
4130 if (PL_lex_brackets)
4131 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4133 if (PL_bufptr == PL_bufend)
4134 return REPORT(sublex_done());
4136 if (SvIVX(PL_linestr) == '\'') {
4137 SV *sv = newSVsv(PL_linestr);
4140 else if ( PL_hints & HINT_NEW_RE )
4141 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4142 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4146 s = scan_const(PL_bufptr);
4148 PL_lex_state = LEX_INTERPCASEMOD;
4150 PL_lex_state = LEX_INTERPSTART;
4153 if (s != PL_bufptr) {
4154 start_force(PL_curforce);
4156 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4158 NEXTVAL_NEXTTOKE = pl_yylval;
4161 if (PL_lex_starts++) {
4165 sv_free(PL_thistoken);
4166 PL_thistoken = newSVpvs("");
4169 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4170 if (!PL_lex_casemods && PL_lex_inpat)
4183 PL_lex_state = LEX_NORMAL;
4184 s = scan_formline(PL_bufptr);
4185 if (!PL_lex_formbrack)
4191 PL_oldoldbufptr = PL_oldbufptr;
4197 sv_free(PL_thistoken);
4200 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4204 if (isIDFIRST_lazy_if(s,UTF))
4207 unsigned char c = *s;
4208 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4209 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4210 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4215 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4219 goto fake_eof; /* emulate EOF on ^D or ^Z */
4228 if (PL_lex_brackets) {
4229 yyerror((const char *)
4231 ? "Format not terminated"
4232 : "Missing right curly or square bracket"));
4234 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4235 "### Tokener got EOF\n");
4239 if (s++ < PL_bufend)
4240 goto retry; /* ignore stray nulls */
4243 if (!PL_in_eval && !PL_preambled) {
4244 PL_preambled = TRUE;
4250 /* Generate a string of Perl code to load the debugger.
4251 * If PERL5DB is set, it will return the contents of that,
4252 * otherwise a compile-time require of perl5db.pl. */
4254 const char * const pdb = PerlEnv_getenv("PERL5DB");
4257 sv_setpv(PL_linestr, pdb);
4258 sv_catpvs(PL_linestr,";");
4260 SETERRNO(0,SS_NORMAL);
4261 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4264 sv_setpvs(PL_linestr,"");
4265 if (PL_preambleav) {
4266 SV **svp = AvARRAY(PL_preambleav);
4267 SV **const end = svp + AvFILLp(PL_preambleav);
4269 sv_catsv(PL_linestr, *svp);
4271 sv_catpvs(PL_linestr, ";");
4273 sv_free(MUTABLE_SV(PL_preambleav));
4274 PL_preambleav = NULL;
4277 sv_catpvs(PL_linestr,
4278 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4279 if (PL_minus_n || PL_minus_p) {
4280 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4282 sv_catpvs(PL_linestr,"chomp;");
4285 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4286 || *PL_splitstr == '"')
4287 && strchr(PL_splitstr + 1, *PL_splitstr))
4288 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4290 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4291 bytes can be used as quoting characters. :-) */
4292 const char *splits = PL_splitstr;
4293 sv_catpvs(PL_linestr, "our @F=split(q\0");
4296 if (*splits == '\\')
4297 sv_catpvn(PL_linestr, splits, 1);
4298 sv_catpvn(PL_linestr, splits, 1);
4299 } while (*splits++);
4300 /* This loop will embed the trailing NUL of
4301 PL_linestr as the last thing it does before
4303 sv_catpvs(PL_linestr, ");");
4307 sv_catpvs(PL_linestr,"our @F=split(' ');");
4310 sv_catpvs(PL_linestr, "\n");
4311 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4312 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4313 PL_last_lop = PL_last_uni = NULL;
4314 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4315 update_debugger_info(PL_linestr, NULL, 0);
4320 bof = PL_rsfp ? TRUE : FALSE;
4323 fake_eof = LEX_FAKE_EOF;
4325 PL_bufptr = PL_bufend;
4326 CopLINE_inc(PL_curcop);
4327 if (!lex_next_chunk(fake_eof)) {
4328 CopLINE_dec(PL_curcop);
4330 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4332 CopLINE_dec(PL_curcop);
4335 PL_realtokenstart = -1;
4338 /* If it looks like the start of a BOM or raw UTF-16,
4339 * check if it in fact is. */
4340 if (bof && PL_rsfp &&
4345 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4347 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4348 s = swallow_bom((U8*)s);
4352 /* Incest with pod. */
4355 sv_catsv(PL_thiswhite, PL_linestr);
4357 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4358 sv_setpvs(PL_linestr, "");
4359 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4360 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4361 PL_last_lop = PL_last_uni = NULL;
4362 PL_doextract = FALSE;
4366 } while (PL_doextract);
4367 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4368 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4369 PL_last_lop = PL_last_uni = NULL;
4370 if (CopLINE(PL_curcop) == 1) {
4371 while (s < PL_bufend && isSPACE(*s))
4373 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4377 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4381 if (*s == '#' && *(s+1) == '!')
4383 #ifdef ALTERNATE_SHEBANG
4385 static char const as[] = ALTERNATE_SHEBANG;
4386 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4387 d = s + (sizeof(as) - 1);
4389 #endif /* ALTERNATE_SHEBANG */
4398 while (*d && !isSPACE(*d))
4402 #ifdef ARG_ZERO_IS_SCRIPT
4403 if (ipathend > ipath) {
4405 * HP-UX (at least) sets argv[0] to the script name,
4406 * which makes $^X incorrect. And Digital UNIX and Linux,
4407 * at least, set argv[0] to the basename of the Perl
4408 * interpreter. So, having found "#!", we'll set it right.
4410 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4412 assert(SvPOK(x) || SvGMAGICAL(x));
4413 if (sv_eq(x, CopFILESV(PL_curcop))) {
4414 sv_setpvn(x, ipath, ipathend - ipath);
4420 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4421 const char * const lstart = SvPV_const(x,llen);
4423 bstart += blen - llen;
4424 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4425 sv_setpvn(x, ipath, ipathend - ipath);
4430 TAINT_NOT; /* $^X is always tainted, but that's OK */
4432 #endif /* ARG_ZERO_IS_SCRIPT */
4437 d = instr(s,"perl -");
4439 d = instr(s,"perl");
4441 /* avoid getting into infinite loops when shebang
4442 * line contains "Perl" rather than "perl" */
4444 for (d = ipathend-4; d >= ipath; --d) {
4445 if ((*d == 'p' || *d == 'P')
4446 && !ibcmp(d, "perl", 4))
4456 #ifdef ALTERNATE_SHEBANG
4458 * If the ALTERNATE_SHEBANG on this system starts with a
4459 * character that can be part of a Perl expression, then if
4460 * we see it but not "perl", we're probably looking at the
4461 * start of Perl code, not a request to hand off to some
4462 * other interpreter. Similarly, if "perl" is there, but
4463 * not in the first 'word' of the line, we assume the line
4464 * contains the start of the Perl program.
4466 if (d && *s != '#') {
4467 const char *c = ipath;
4468 while (*c && !strchr("; \t\r\n\f\v#", *c))
4471 d = NULL; /* "perl" not in first word; ignore */
4473 *s = '#'; /* Don't try to parse shebang line */
4475 #endif /* ALTERNATE_SHEBANG */
4480 !instr(s,"indir") &&
4481 instr(PL_origargv[0],"perl"))
4488 while (s < PL_bufend && isSPACE(*s))
4490 if (s < PL_bufend) {
4491 Newx(newargv,PL_origargc+3,char*);
4493 while (s < PL_bufend && !isSPACE(*s))
4496 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4499 newargv = PL_origargv;
4502 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4504 Perl_croak(aTHX_ "Can't exec %s", ipath);
4507 while (*d && !isSPACE(*d))
4509 while (SPACE_OR_TAB(*d))
4513 const bool switches_done = PL_doswitches;
4514 const U32 oldpdb = PL_perldb;
4515 const bool oldn = PL_minus_n;
4516 const bool oldp = PL_minus_p;
4520 bool baduni = FALSE;
4522 const char *d2 = d1 + 1;
4523 if (parse_unicode_opts((const char **)&d2)
4527 if (baduni || *d1 == 'M' || *d1 == 'm') {
4528 const char * const m = d1;
4529 while (*d1 && !isSPACE(*d1))
4531 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4534 d1 = moreswitches(d1);
4536 if (PL_doswitches && !switches_done) {
4537 int argc = PL_origargc;
4538 char **argv = PL_origargv;
4541 } while (argc && argv[0][0] == '-' && argv[0][1]);
4542 init_argv_symbols(argc,argv);
4544 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4545 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4546 /* if we have already added "LINE: while (<>) {",
4547 we must not do it again */
4549 sv_setpvs(PL_linestr, "");
4550 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4551 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4552 PL_last_lop = PL_last_uni = NULL;
4553 PL_preambled = FALSE;
4554 if (PERLDB_LINE || PERLDB_SAVESRC)
4555 (void)gv_fetchfile(PL_origfilename);
4562 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4564 PL_lex_state = LEX_FORMLINE;
4569 #ifdef PERL_STRICT_CR
4570 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4572 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4574 case ' ': case '\t': case '\f': case 013:
4576 PL_realtokenstart = -1;
4578 PL_thiswhite = newSVpvs("");
4579 sv_catpvn(PL_thiswhite, s, 1);
4586 PL_realtokenstart = -1;
4590 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4591 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4592 /* handle eval qq[#line 1 "foo"\n ...] */
4593 CopLINE_dec(PL_curcop);
4596 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4598 if (!PL_in_eval || PL_rsfp)
4603 while (d < PL_bufend && *d != '\n')
4607 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4608 Perl_croak(aTHX_ "panic: input overflow");
4611 PL_thiswhite = newSVpvn(s, d - s);
4616 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4618 PL_lex_state = LEX_FORMLINE;
4624 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4625 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4628 TOKEN(PEG); /* make sure any #! line is accessible */
4633 /* if (PL_madskills && PL_lex_formbrack) { */
4635 while (d < PL_bufend && *d != '\n')
4639 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4640 Perl_croak(aTHX_ "panic: input overflow");
4641 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4643 PL_thiswhite = newSVpvs("");
4644 if (CopLINE(PL_curcop) == 1) {
4645 sv_setpvs(PL_thiswhite, "");
4648 sv_catpvn(PL_thiswhite, s, d - s);
4662 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4670 while (s < PL_bufend && SPACE_OR_TAB(*s))
4673 if (strnEQ(s,"=>",2)) {
4674 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4675 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4676 OPERATOR('-'); /* unary minus */
4678 PL_last_uni = PL_oldbufptr;
4680 case 'r': ftst = OP_FTEREAD; break;
4681 case 'w': ftst = OP_FTEWRITE; break;
4682 case 'x': ftst = OP_FTEEXEC; break;
4683 case 'o': ftst = OP_FTEOWNED; break;
4684 case 'R': ftst = OP_FTRREAD; break;
4685 case 'W': ftst = OP_FTRWRITE; break;
4686 case 'X': ftst = OP_FTREXEC; break;
4687 case 'O': ftst = OP_FTROWNED; break;
4688 case 'e': ftst = OP_FTIS; break;
4689 case 'z': ftst = OP_FTZERO; break;
4690 case 's': ftst = OP_FTSIZE; break;
4691 case 'f': ftst = OP_FTFILE; break;
4692 case 'd': ftst = OP_FTDIR; break;
4693 case 'l': ftst = OP_FTLINK; break;
4694 case 'p': ftst = OP_FTPIPE; break;
4695 case 'S': ftst = OP_FTSOCK; break;
4696 case 'u': ftst = OP_FTSUID; break;
4697 case 'g': ftst = OP_FTSGID; break;
4698 case 'k': ftst = OP_FTSVTX; break;
4699 case 'b': ftst = OP_FTBLK; break;
4700 case 'c': ftst = OP_FTCHR; break;
4701 case 't': ftst = OP_FTTTY; break;
4702 case 'T': ftst = OP_FTTEXT; break;
4703 case 'B': ftst = OP_FTBINARY; break;
4704 case 'M': case 'A': case 'C':
4705 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4707 case 'M': ftst = OP_FTMTIME; break;
4708 case 'A': ftst = OP_FTATIME; break;
4709 case 'C': ftst = OP_FTCTIME; break;
4717 PL_last_lop_op = (OPCODE)ftst;
4718 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4719 "### Saw file test %c\n", (int)tmp);
4724 /* Assume it was a minus followed by a one-letter named
4725 * subroutine call (or a -bareword), then. */
4726 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4727 "### '-%c' looked like a file test but was not\n",
4734 const char tmp = *s++;
4737 if (PL_expect == XOPERATOR)
4742 else if (*s == '>') {
4745 if (isIDFIRST_lazy_if(s,UTF)) {
4746 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4754 if (PL_expect == XOPERATOR)
4757 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4759 OPERATOR('-'); /* unary minus */
4765 const char tmp = *s++;
4768 if (PL_expect == XOPERATOR)
4773 if (PL_expect == XOPERATOR)
4776 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4783 if (PL_expect != XOPERATOR) {
4784 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4785 PL_expect = XOPERATOR;
4786 force_ident(PL_tokenbuf, '*');
4799 if (PL_expect == XOPERATOR) {
4803 PL_tokenbuf[0] = '%';
4804 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4805 sizeof PL_tokenbuf - 1, FALSE);
4806 if (!PL_tokenbuf[1]) {
4809 PL_pending_ident = '%';
4818 const char tmp = *s++;
4823 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4830 const char tmp = *s++;
4836 goto just_a_word_zero_gv;
4839 switch (PL_expect) {
4845 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4847 PL_bufptr = s; /* update in case we back off */
4849 deprecate(":= for an empty attribute list");
4856 PL_expect = XTERMBLOCK;
4859 stuffstart = s - SvPVX(PL_linestr) - 1;
4863 while (isIDFIRST_lazy_if(s,UTF)) {
4866 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4867 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4868 if (tmp < 0) tmp = -tmp;
4883 sv = newSVpvn(s, len);
4885 d = scan_str(d,TRUE,TRUE);
4887 /* MUST advance bufptr here to avoid bogus
4888 "at end of line" context messages from yyerror().
4890 PL_bufptr = s + len;
4891 yyerror("Unterminated attribute parameter in attribute list");
4895 return REPORT(0); /* EOF indicator */
4899 sv_catsv(sv, PL_lex_stuff);
4900 attrs = append_elem(OP_LIST, attrs,
4901 newSVOP(OP_CONST, 0, sv));
4902 SvREFCNT_dec(PL_lex_stuff);
4903 PL_lex_stuff = NULL;
4906 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4908 if (PL_in_my == KEY_our) {
4909 deprecate(":unique");
4912 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4915 /* NOTE: any CV attrs applied here need to be part of
4916 the CVf_BUILTIN_ATTRS define in cv.h! */
4917 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4919 CvLVALUE_on(PL_compcv);
4921 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4923 deprecate(":locked");
4925 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4927 CvMETHOD_on(PL_compcv);
4929 /* After we've set the flags, it could be argued that
4930 we don't need to do the attributes.pm-based setting
4931 process, and shouldn't bother appending recognized
4932 flags. To experiment with that, uncomment the
4933 following "else". (Note that's already been
4934 uncommented. That keeps the above-applied built-in
4935 attributes from being intercepted (and possibly
4936 rejected) by a package's attribute routines, but is
4937 justified by the performance win for the common case
4938 of applying only built-in attributes.) */
4940 attrs = append_elem(OP_LIST, attrs,
4941 newSVOP(OP_CONST, 0,
4945 if (*s == ':' && s[1] != ':')
4948 break; /* require real whitespace or :'s */
4949 /* XXX losing whitespace on sequential attributes here */
4953 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4954 if (*s != ';' && *s != '}' && *s != tmp
4955 && (tmp != '=' || *s != ')')) {
4956 const char q = ((*s == '\'') ? '"' : '\'');
4957 /* If here for an expression, and parsed no attrs, back
4959 if (tmp == '=' && !attrs) {
4963 /* MUST advance bufptr here to avoid bogus "at end of line"
4964 context messages from yyerror().
4967 yyerror( (const char *)
4969 ? Perl_form(aTHX_ "Invalid separator character "
4970 "%c%c%c in attribute list", q, *s, q)
4971 : "Unterminated attribute list" ) );
4979 start_force(PL_curforce);
4980 NEXTVAL_NEXTTOKE.opval = attrs;
4981 CURMAD('_', PL_nextwhite);
4986 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4987 (s - SvPVX(PL_linestr)) - stuffstart);
4995 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4996 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5004 const char tmp = *s++;
5009 const char tmp = *s++;
5017 if (PL_lex_brackets <= 0)
5018 yyerror("Unmatched right square bracket");
5021 if (PL_lex_state == LEX_INTERPNORMAL) {
5022 if (PL_lex_brackets == 0) {
5023 if (*s == '-' && s[1] == '>')
5024 PL_lex_state = LEX_INTERPENDMAYBE;
5025 else if (*s != '[' && *s != '{')
5026 PL_lex_state = LEX_INTERPEND;
5033 if (PL_lex_brackets > 100) {
5034 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5036 switch (PL_expect) {
5038 if (PL_lex_formbrack) {
5042 if (PL_oldoldbufptr == PL_last_lop)
5043 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5045 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5046 OPERATOR(HASHBRACK);
5048 while (s < PL_bufend && SPACE_OR_TAB(*s))
5051 PL_tokenbuf[0] = '\0';
5052 if (d < PL_bufend && *d == '-') {
5053 PL_tokenbuf[0] = '-';
5055 while (d < PL_bufend && SPACE_OR_TAB(*d))
5058 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5059 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5061 while (d < PL_bufend && SPACE_OR_TAB(*d))
5064 const char minus = (PL_tokenbuf[0] == '-');
5065 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5073 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5078 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5083 if (PL_oldoldbufptr == PL_last_lop)
5084 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5086 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5089 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5091 /* This hack is to get the ${} in the message. */
5093 yyerror("syntax error");
5096 OPERATOR(HASHBRACK);
5098 /* This hack serves to disambiguate a pair of curlies
5099 * as being a block or an anon hash. Normally, expectation
5100 * determines that, but in cases where we're not in a
5101 * position to expect anything in particular (like inside
5102 * eval"") we have to resolve the ambiguity. This code
5103 * covers the case where the first term in the curlies is a
5104 * quoted string. Most other cases need to be explicitly
5105 * disambiguated by prepending a "+" before the opening
5106 * curly in order to force resolution as an anon hash.
5108 * XXX should probably propagate the outer expectation
5109 * into eval"" to rely less on this hack, but that could
5110 * potentially break current behavior of eval"".
5114 if (*s == '\'' || *s == '"' || *s == '`') {
5115 /* common case: get past first string, handling escapes */
5116 for (t++; t < PL_bufend && *t != *s;)
5117 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5121 else if (*s == 'q') {
5124 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5127 /* skip q//-like construct */
5129 char open, close, term;
5132 while (t < PL_bufend && isSPACE(*t))
5134 /* check for q => */
5135 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5136 OPERATOR(HASHBRACK);
5140 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5144 for (t++; t < PL_bufend; t++) {
5145 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5147 else if (*t == open)
5151 for (t++; t < PL_bufend; t++) {
5152 if (*t == '\\' && t+1 < PL_bufend)
5154 else if (*t == close && --brackets <= 0)
5156 else if (*t == open)
5163 /* skip plain q word */
5164 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5167 else if (isALNUM_lazy_if(t,UTF)) {
5169 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5172 while (t < PL_bufend && isSPACE(*t))
5174 /* if comma follows first term, call it an anon hash */
5175 /* XXX it could be a comma expression with loop modifiers */
5176 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5177 || (*t == '=' && t[1] == '>')))
5178 OPERATOR(HASHBRACK);
5179 if (PL_expect == XREF)
5182 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5188 pl_yylval.ival = CopLINE(PL_curcop);
5189 if (isSPACE(*s) || *s == '#')
5190 PL_copline = NOLINE; /* invalidate current command line number */
5195 if (PL_lex_brackets <= 0)
5196 yyerror("Unmatched right curly bracket");
5198 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5199 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5200 PL_lex_formbrack = 0;
5201 if (PL_lex_state == LEX_INTERPNORMAL) {
5202 if (PL_lex_brackets == 0) {
5203 if (PL_expect & XFAKEBRACK) {
5204 PL_expect &= XENUMMASK;
5205 PL_lex_state = LEX_INTERPEND;
5210 PL_thiswhite = newSVpvs("");
5211 sv_catpvs(PL_thiswhite,"}");
5214 return yylex(); /* ignore fake brackets */
5216 if (*s == '-' && s[1] == '>')
5217 PL_lex_state = LEX_INTERPENDMAYBE;
5218 else if (*s != '[' && *s != '{')
5219 PL_lex_state = LEX_INTERPEND;
5222 if (PL_expect & XFAKEBRACK) {
5223 PL_expect &= XENUMMASK;
5225 return yylex(); /* ignore fake brackets */
5227 start_force(PL_curforce);
5229 curmad('X', newSVpvn(s-1,1));
5230 CURMAD('_', PL_thiswhite);
5235 PL_thistoken = newSVpvs("");
5243 if (PL_expect == XOPERATOR) {
5244 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5245 && isIDFIRST_lazy_if(s,UTF))
5247 CopLINE_dec(PL_curcop);
5248 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5249 CopLINE_inc(PL_curcop);
5254 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5256 PL_expect = XOPERATOR;
5257 force_ident(PL_tokenbuf, '&');
5261 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5273 const char tmp = *s++;
5280 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5281 && strchr("+-*/%.^&|<",tmp))
5282 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5283 "Reversed %c= operator",(int)tmp);
5285 if (PL_expect == XSTATE && isALPHA(tmp) &&
5286 (s == PL_linestart+1 || s[-2] == '\n') )
5288 if (PL_in_eval && !PL_rsfp) {
5293 if (strnEQ(s,"=cut",4)) {
5309 PL_thiswhite = newSVpvs("");
5310 sv_catpvn(PL_thiswhite, PL_linestart,
5311 PL_bufend - PL_linestart);
5315 PL_doextract = TRUE;
5319 if (PL_lex_brackets < PL_lex_formbrack) {
5321 #ifdef PERL_STRICT_CR
5322 while (SPACE_OR_TAB(*t))
5324 while (SPACE_OR_TAB(*t) || *t == '\r')
5327 if (*t == '\n' || *t == '#') {
5338 const char tmp = *s++;
5340 /* was this !=~ where !~ was meant?
5341 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5343 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5344 const char *t = s+1;
5346 while (t < PL_bufend && isSPACE(*t))
5349 if (*t == '/' || *t == '?' ||
5350 ((*t == 'm' || *t == 's' || *t == 'y')
5351 && !isALNUM(t[1])) ||
5352 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5353 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5354 "!=~ should be !~");
5364 if (PL_expect != XOPERATOR) {
5365 if (s[1] != '<' && !strchr(s,'>'))
5368 s = scan_heredoc(s);
5370 s = scan_inputsymbol(s);
5371 TERM(sublex_start());
5377 SHop(OP_LEFT_SHIFT);
5391 const char tmp = *s++;
5393 SHop(OP_RIGHT_SHIFT);
5394 else if (tmp == '=')
5403 if (PL_expect == XOPERATOR) {
5404 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5405 return deprecate_commaless_var_list();
5409 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5410 PL_tokenbuf[0] = '@';
5411 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5412 sizeof PL_tokenbuf - 1, FALSE);
5413 if (PL_expect == XOPERATOR)
5414 no_op("Array length", s);
5415 if (!PL_tokenbuf[1])
5417 PL_expect = XOPERATOR;
5418 PL_pending_ident = '#';
5422 PL_tokenbuf[0] = '$';
5423 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5424 sizeof PL_tokenbuf - 1, FALSE);
5425 if (PL_expect == XOPERATOR)
5427 if (!PL_tokenbuf[1]) {
5429 yyerror("Final $ should be \\$ or $name");
5433 /* This kludge not intended to be bulletproof. */
5434 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5435 pl_yylval.opval = newSVOP(OP_CONST, 0,
5436 newSViv(CopARYBASE_get(&PL_compiling)));
5437 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5443 const char tmp = *s;
5444 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5447 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5448 && intuit_more(s)) {
5450 PL_tokenbuf[0] = '@';
5451 if (ckWARN(WARN_SYNTAX)) {
5454 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5457 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5458 while (t < PL_bufend && *t != ']')
5460 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5461 "Multidimensional syntax %.*s not supported",
5462 (int)((t - PL_bufptr) + 1), PL_bufptr);
5466 else if (*s == '{') {
5468 PL_tokenbuf[0] = '%';
5469 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5470 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5472 char tmpbuf[sizeof PL_tokenbuf];
5475 } while (isSPACE(*t));
5476 if (isIDFIRST_lazy_if(t,UTF)) {
5478 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5482 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5483 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5484 "You need to quote \"%s\"",
5491 PL_expect = XOPERATOR;
5492 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5493 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5494 if (!islop || PL_last_lop_op == OP_GREPSTART)
5495 PL_expect = XOPERATOR;
5496 else if (strchr("$@\"'`q", *s))
5497 PL_expect = XTERM; /* e.g. print $fh "foo" */
5498 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5499 PL_expect = XTERM; /* e.g. print $fh &sub */
5500 else if (isIDFIRST_lazy_if(s,UTF)) {
5501 char tmpbuf[sizeof PL_tokenbuf];
5503 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5504 if ((t2 = keyword(tmpbuf, len, 0))) {
5505 /* binary operators exclude handle interpretations */
5517 PL_expect = XTERM; /* e.g. print $fh length() */
5522 PL_expect = XTERM; /* e.g. print $fh subr() */
5525 else if (isDIGIT(*s))
5526 PL_expect = XTERM; /* e.g. print $fh 3 */
5527 else if (*s == '.' && isDIGIT(s[1]))
5528 PL_expect = XTERM; /* e.g. print $fh .3 */
5529 else if ((*s == '?' || *s == '-' || *s == '+')
5530 && !isSPACE(s[1]) && s[1] != '=')
5531 PL_expect = XTERM; /* e.g. print $fh -1 */
5532 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5534 PL_expect = XTERM; /* e.g. print $fh /.../
5535 XXX except DORDOR operator
5537 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5539 PL_expect = XTERM; /* print $fh <<"EOF" */
5542 PL_pending_ident = '$';
5546 if (PL_expect == XOPERATOR)
5548 PL_tokenbuf[0] = '@';
5549 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5550 if (!PL_tokenbuf[1]) {
5553 if (PL_lex_state == LEX_NORMAL)
5555 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5557 PL_tokenbuf[0] = '%';
5559 /* Warn about @ where they meant $. */
5560 if (*s == '[' || *s == '{') {
5561 if (ckWARN(WARN_SYNTAX)) {
5562 const char *t = s + 1;
5563 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5565 if (*t == '}' || *t == ']') {
5567 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5568 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5569 "Scalar value %.*s better written as $%.*s",
5570 (int)(t-PL_bufptr), PL_bufptr,
5571 (int)(t-PL_bufptr-1), PL_bufptr+1);
5576 PL_pending_ident = '@';
5579 case '/': /* may be division, defined-or, or pattern */
5580 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5584 case '?': /* may either be conditional or pattern */
5585 if (PL_expect == XOPERATOR) {
5593 /* A // operator. */
5603 /* Disable warning on "study /blah/" */
5604 if (PL_oldoldbufptr == PL_last_uni
5605 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5606 || memNE(PL_last_uni, "study", 5)
5607 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5610 s = scan_pat(s,OP_MATCH);
5611 TERM(sublex_start());
5615 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5616 #ifdef PERL_STRICT_CR
5619 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5621 && (s == PL_linestart || s[-1] == '\n') )
5623 PL_lex_formbrack = 0;
5627 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5631 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5637 pl_yylval.ival = OPf_SPECIAL;
5643 if (PL_expect != XOPERATOR)
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 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6591 UNIBRACK(OP_ENTEREVAL);
6605 case KEY_endhostent:
6611 case KEY_endservent:
6614 case KEY_endprotoent:
6625 pl_yylval.ival = CopLINE(PL_curcop);
6627 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6630 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6633 if ((PL_bufend - p) >= 3 &&
6634 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6636 else if ((PL_bufend - p) >= 4 &&
6637 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6640 if (isIDFIRST_lazy_if(p,UTF)) {
6641 p = scan_ident(p, PL_bufend,
6642 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6646 Perl_croak(aTHX_ "Missing $ on loop variable");
6648 s = SvPVX(PL_linestr) + soff;
6654 LOP(OP_FORMLINE,XTERM);
6660 LOP(OP_FCNTL,XTERM);
6666 LOP(OP_FLOCK,XTERM);
6675 LOP(OP_GREPSTART, XREF);
6678 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6693 case KEY_getpriority:
6694 LOP(OP_GETPRIORITY,XTERM);
6696 case KEY_getprotobyname:
6699 case KEY_getprotobynumber:
6700 LOP(OP_GPBYNUMBER,XTERM);
6702 case KEY_getprotoent:
6714 case KEY_getpeername:
6715 UNI(OP_GETPEERNAME);
6717 case KEY_gethostbyname:
6720 case KEY_gethostbyaddr:
6721 LOP(OP_GHBYADDR,XTERM);
6723 case KEY_gethostent:
6726 case KEY_getnetbyname:
6729 case KEY_getnetbyaddr:
6730 LOP(OP_GNBYADDR,XTERM);
6735 case KEY_getservbyname:
6736 LOP(OP_GSBYNAME,XTERM);
6738 case KEY_getservbyport:
6739 LOP(OP_GSBYPORT,XTERM);
6741 case KEY_getservent:
6744 case KEY_getsockname:
6745 UNI(OP_GETSOCKNAME);
6747 case KEY_getsockopt:
6748 LOP(OP_GSOCKOPT,XTERM);
6763 pl_yylval.ival = CopLINE(PL_curcop);
6773 pl_yylval.ival = CopLINE(PL_curcop);
6777 LOP(OP_INDEX,XTERM);
6783 LOP(OP_IOCTL,XTERM);
6795 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6827 LOP(OP_LISTEN,XTERM);
6836 s = scan_pat(s,OP_MATCH);
6837 TERM(sublex_start());
6840 LOP(OP_MAPSTART, XREF);
6843 LOP(OP_MKDIR,XTERM);
6846 LOP(OP_MSGCTL,XTERM);
6849 LOP(OP_MSGGET,XTERM);
6852 LOP(OP_MSGRCV,XTERM);
6855 LOP(OP_MSGSND,XTERM);
6860 PL_in_my = (U16)tmp;
6862 if (isIDFIRST_lazy_if(s,UTF)) {
6866 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6867 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6869 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6870 if (!PL_in_my_stash) {
6873 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6877 if (PL_madskills) { /* just add type to declarator token */
6878 sv_catsv(PL_thistoken, PL_nextwhite);
6880 sv_catpvn(PL_thistoken, start, s - start);
6888 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6895 s = tokenize_use(0, s);
6899 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6906 if (isIDFIRST_lazy_if(s,UTF)) {
6908 for (d = s; isALNUM_lazy_if(d,UTF);)
6910 for (t=d; isSPACE(*t);)
6912 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6914 && !(t[0] == '=' && t[1] == '>')
6916 int parms_len = (int)(d-s);
6917 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6918 "Precedence problem: open %.*s should be open(%.*s)",
6919 parms_len, s, parms_len, s);
6925 pl_yylval.ival = OP_OR;
6935 LOP(OP_OPEN_DIR,XTERM);
6938 checkcomma(s,PL_tokenbuf,"filehandle");
6942 checkcomma(s,PL_tokenbuf,"filehandle");
6961 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6962 s = force_version(s, FALSE);
6966 LOP(OP_PIPE_OP,XTERM);
6969 s = scan_str(s,!!PL_madskills,FALSE);
6972 pl_yylval.ival = OP_CONST;
6973 TERM(sublex_start());
6979 s = scan_str(s,!!PL_madskills,FALSE);
6982 PL_expect = XOPERATOR;
6984 if (SvCUR(PL_lex_stuff)) {
6987 d = SvPV_force(PL_lex_stuff, len);
6989 for (; isSPACE(*d) && len; --len, ++d)
6994 if (!warned && ckWARN(WARN_QW)) {
6995 for (; !isSPACE(*d) && len; --len, ++d) {
6997 Perl_warner(aTHX_ packWARN(WARN_QW),
6998 "Possible attempt to separate words with commas");
7001 else if (*d == '#') {
7002 Perl_warner(aTHX_ packWARN(WARN_QW),
7003 "Possible attempt to put comments in qw() list");
7009 for (; !isSPACE(*d) && len; --len, ++d)
7012 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7013 words = append_elem(OP_LIST, words,
7014 newSVOP(OP_CONST, 0, tokeq(sv)));
7018 start_force(PL_curforce);
7019 NEXTVAL_NEXTTOKE.opval = words;
7024 SvREFCNT_dec(PL_lex_stuff);
7025 PL_lex_stuff = NULL;
7031 s = scan_str(s,!!PL_madskills,FALSE);
7034 pl_yylval.ival = OP_STRINGIFY;
7035 if (SvIVX(PL_lex_stuff) == '\'')
7036 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7037 TERM(sublex_start());
7040 s = scan_pat(s,OP_QR);
7041 TERM(sublex_start());
7044 s = scan_str(s,!!PL_madskills,FALSE);
7047 readpipe_override();
7048 TERM(sublex_start());
7056 s = force_version(s, FALSE);
7058 else if (*s != 'v' || !isDIGIT(s[1])
7059 || (s = force_version(s, TRUE), *s == 'v'))
7061 *PL_tokenbuf = '\0';
7062 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7063 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7064 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7066 yyerror("<> should be quotes");
7068 if (orig_keyword == KEY_require) {
7076 PL_last_uni = PL_oldbufptr;
7077 PL_last_lop_op = OP_REQUIRE;
7079 return REPORT( (int)REQUIRE );
7085 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7089 LOP(OP_RENAME,XTERM);
7098 LOP(OP_RINDEX,XTERM);
7107 UNIDOR(OP_READLINE);
7110 UNIDOR(OP_BACKTICK);
7119 LOP(OP_REVERSE,XTERM);
7122 UNIDOR(OP_READLINK);
7129 if (pl_yylval.opval)
7130 TERM(sublex_start());
7132 TOKEN(1); /* force error */
7135 checkcomma(s,PL_tokenbuf,"filehandle");
7145 LOP(OP_SELECT,XTERM);
7151 LOP(OP_SEMCTL,XTERM);
7154 LOP(OP_SEMGET,XTERM);
7157 LOP(OP_SEMOP,XTERM);
7163 LOP(OP_SETPGRP,XTERM);
7165 case KEY_setpriority:
7166 LOP(OP_SETPRIORITY,XTERM);
7168 case KEY_sethostent:
7174 case KEY_setservent:
7177 case KEY_setprotoent:
7187 LOP(OP_SEEKDIR,XTERM);
7189 case KEY_setsockopt:
7190 LOP(OP_SSOCKOPT,XTERM);
7196 LOP(OP_SHMCTL,XTERM);
7199 LOP(OP_SHMGET,XTERM);
7202 LOP(OP_SHMREAD,XTERM);
7205 LOP(OP_SHMWRITE,XTERM);
7208 LOP(OP_SHUTDOWN,XTERM);
7217 LOP(OP_SOCKET,XTERM);
7219 case KEY_socketpair:
7220 LOP(OP_SOCKPAIR,XTERM);
7223 checkcomma(s,PL_tokenbuf,"subroutine name");
7225 if (*s == ';' || *s == ')') /* probably a close */
7226 Perl_croak(aTHX_ "sort is now a reserved word");
7228 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7232 LOP(OP_SPLIT,XTERM);
7235 LOP(OP_SPRINTF,XTERM);
7238 LOP(OP_SPLICE,XTERM);
7253 LOP(OP_SUBSTR,XTERM);
7259 char tmpbuf[sizeof PL_tokenbuf];
7260 SSize_t tboffset = 0;
7261 expectation attrful;
7262 bool have_name, have_proto;
7263 const int key = tmp;
7268 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7269 SV *subtoken = newSVpvn(tstart, s - tstart);
7273 s = SKIPSPACE2(s,tmpwhite);
7278 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7279 (*s == ':' && s[1] == ':'))
7282 SV *nametoke = NULL;
7286 attrful = XATTRBLOCK;
7287 /* remember buffer pos'n for later force_word */
7288 tboffset = s - PL_oldbufptr;
7289 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7292 nametoke = newSVpvn(s, d - s);
7294 if (memchr(tmpbuf, ':', len))
7295 sv_setpvn(PL_subname, tmpbuf, len);
7297 sv_setsv(PL_subname,PL_curstname);
7298 sv_catpvs(PL_subname,"::");
7299 sv_catpvn(PL_subname,tmpbuf,len);
7306 CURMAD('X', nametoke);
7307 CURMAD('_', tmpwhite);
7308 (void) force_word(PL_oldbufptr + tboffset, WORD,
7311 s = SKIPSPACE2(d,tmpwhite);
7318 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7319 PL_expect = XTERMBLOCK;
7320 attrful = XATTRTERM;
7321 sv_setpvs(PL_subname,"?");
7325 if (key == KEY_format) {
7327 PL_lex_formbrack = PL_lex_brackets + 1;
7329 PL_thistoken = subtoken;
7333 (void) force_word(PL_oldbufptr + tboffset, WORD,
7339 /* Look for a prototype */
7342 bool bad_proto = FALSE;
7343 bool in_brackets = FALSE;
7344 char greedy_proto = ' ';
7345 bool proto_after_greedy_proto = FALSE;
7346 bool must_be_last = FALSE;
7347 bool underscore = FALSE;
7348 bool seen_underscore = FALSE;
7349 const bool warnsyntax = ckWARN(WARN_SYNTAX);
7351 s = scan_str(s,!!PL_madskills,FALSE);
7353 Perl_croak(aTHX_ "Prototype not terminated");
7354 /* strip spaces and check for bad characters */
7355 d = SvPVX(PL_lex_stuff);
7357 for (p = d; *p; ++p) {
7363 proto_after_greedy_proto = TRUE;
7364 if (!strchr("$@%*;[]&\\_", *p)) {
7376 else if ( *p == ']' ) {
7377 in_brackets = FALSE;
7379 else if ( (*p == '@' || *p == '%') &&
7380 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7382 must_be_last = TRUE;
7385 else if ( *p == '_' ) {
7386 underscore = seen_underscore = TRUE;
7393 if (proto_after_greedy_proto)
7394 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7395 "Prototype after '%c' for %"SVf" : %s",
7396 greedy_proto, SVfARG(PL_subname), d);
7398 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7399 "Illegal character %sin prototype for %"SVf" : %s",
7400 seen_underscore ? "after '_' " : "",
7401 SVfARG(PL_subname), d);
7402 SvCUR_set(PL_lex_stuff, tmp);
7407 CURMAD('q', PL_thisopen);
7408 CURMAD('_', tmpwhite);
7409 CURMAD('=', PL_thisstuff);
7410 CURMAD('Q', PL_thisclose);
7411 NEXTVAL_NEXTTOKE.opval =
7412 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7413 PL_lex_stuff = NULL;
7416 s = SKIPSPACE2(s,tmpwhite);
7424 if (*s == ':' && s[1] != ':')
7425 PL_expect = attrful;
7426 else if (*s != '{' && key == KEY_sub) {
7428 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7430 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7437 curmad('^', newSVpvs(""));
7438 CURMAD('_', tmpwhite);
7442 PL_thistoken = subtoken;
7445 NEXTVAL_NEXTTOKE.opval =
7446 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7447 PL_lex_stuff = NULL;
7453 sv_setpvs(PL_subname, "__ANON__");
7455 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7459 (void) force_word(PL_oldbufptr + tboffset, WORD,
7468 LOP(OP_SYSTEM,XREF);
7471 LOP(OP_SYMLINK,XTERM);
7474 LOP(OP_SYSCALL,XTERM);
7477 LOP(OP_SYSOPEN,XTERM);
7480 LOP(OP_SYSSEEK,XTERM);
7483 LOP(OP_SYSREAD,XTERM);
7486 LOP(OP_SYSWRITE,XTERM);
7490 TERM(sublex_start());
7511 LOP(OP_TRUNCATE,XTERM);
7523 pl_yylval.ival = CopLINE(PL_curcop);
7527 pl_yylval.ival = CopLINE(PL_curcop);
7531 LOP(OP_UNLINK,XTERM);
7537 LOP(OP_UNPACK,XTERM);
7540 LOP(OP_UTIME,XTERM);
7546 LOP(OP_UNSHIFT,XTERM);
7549 s = tokenize_use(1, s);
7559 pl_yylval.ival = CopLINE(PL_curcop);
7563 pl_yylval.ival = CopLINE(PL_curcop);
7567 PL_hints |= HINT_BLOCK_SCOPE;
7574 LOP(OP_WAITPID,XTERM);
7583 ctl_l[0] = toCTRL('L');
7585 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7588 /* Make sure $^L is defined */
7589 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7594 if (PL_expect == XOPERATOR)
7600 pl_yylval.ival = OP_XOR;
7605 TERM(sublex_start());
7610 #pragma segment Main
7614 S_pending_ident(pTHX)
7619 /* pit holds the identifier we read and pending_ident is reset */
7620 char pit = PL_pending_ident;
7621 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7622 /* All routes through this function want to know if there is a colon. */
7623 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7624 PL_pending_ident = 0;
7626 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7627 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7628 "### Pending identifier '%s'\n", PL_tokenbuf); });
7630 /* if we're in a my(), we can't allow dynamics here.
7631 $foo'bar has already been turned into $foo::bar, so
7632 just check for colons.
7634 if it's a legal name, the OP is a PADANY.
7637 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7639 yyerror(Perl_form(aTHX_ "No package name allowed for "
7640 "variable %s in \"our\"",
7642 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7646 yyerror(Perl_form(aTHX_ PL_no_myglob,
7647 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7649 pl_yylval.opval = newOP(OP_PADANY, 0);
7650 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7656 build the ops for accesses to a my() variable.
7658 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7659 then used in a comparison. This catches most, but not
7660 all cases. For instance, it catches
7661 sort { my($a); $a <=> $b }
7663 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7664 (although why you'd do that is anyone's guess).
7669 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7670 if (tmp != NOT_IN_PAD) {
7671 /* might be an "our" variable" */
7672 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7673 /* build ops for a bareword */
7674 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7675 HEK * const stashname = HvNAME_HEK(stash);
7676 SV * const sym = newSVhek(stashname);
7677 sv_catpvs(sym, "::");
7678 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7679 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7680 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7683 ? (GV_ADDMULTI | GV_ADDINEVAL)
7686 ((PL_tokenbuf[0] == '$') ? SVt_PV
7687 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7692 /* if it's a sort block and they're naming $a or $b */
7693 if (PL_last_lop_op == OP_SORT &&
7694 PL_tokenbuf[0] == '$' &&
7695 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7698 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7699 d < PL_bufend && *d != '\n';
7702 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7703 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7709 pl_yylval.opval = newOP(OP_PADANY, 0);
7710 pl_yylval.opval->op_targ = tmp;
7716 Whine if they've said @foo in a doublequoted string,
7717 and @foo isn't a variable we can find in the symbol
7720 if (ckWARN(WARN_AMBIGUOUS) &&
7721 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7722 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7724 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7725 /* DO NOT warn for @- and @+ */
7726 && !( PL_tokenbuf[2] == '\0' &&
7727 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7730 /* Downgraded from fatal to warning 20000522 mjd */
7731 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7732 "Possible unintended interpolation of %s in string",
7737 /* build ops for a bareword */
7738 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7740 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7742 PL_tokenbuf + 1, tokenbuf_len - 1,
7743 /* If the identifier refers to a stash, don't autovivify it.
7744 * Change 24660 had the side effect of causing symbol table
7745 * hashes to always be defined, even if they were freshly
7746 * created and the only reference in the entire program was
7747 * the single statement with the defined %foo::bar:: test.
7748 * It appears that all code in the wild doing this actually
7749 * wants to know whether sub-packages have been loaded, so
7750 * by avoiding auto-vivifying symbol tables, we ensure that
7751 * defined %foo::bar:: continues to be false, and the existing
7752 * tests still give the expected answers, even though what
7753 * they're actually testing has now changed subtly.
7755 (*PL_tokenbuf == '%'
7756 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7759 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7760 ((PL_tokenbuf[0] == '$') ? SVt_PV
7761 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7767 * The following code was generated by perl_keyword.pl.
7771 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7775 PERL_ARGS_ASSERT_KEYWORD;
7779 case 1: /* 5 tokens of length 1 */
7811 case 2: /* 18 tokens of length 2 */
7957 case 3: /* 29 tokens of length 3 */
7961 if (name[1] == 'N' &&
8024 if (name[1] == 'i' &&
8056 if (name[1] == 'o' &&
8065 if (name[1] == 'e' &&
8074 if (name[1] == 'n' &&
8083 if (name[1] == 'o' &&
8092 if (name[1] == 'a' &&
8101 if (name[1] == 'o' &&
8163 if (name[1] == 'e' &&
8177 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8203 if (name[1] == 'i' &&
8212 if (name[1] == 's' &&
8221 if (name[1] == 'e' &&
8230 if (name[1] == 'o' &&
8242 case 4: /* 41 tokens of length 4 */
8246 if (name[1] == 'O' &&
8256 if (name[1] == 'N' &&
8266 if (name[1] == 'i' &&
8276 if (name[1] == 'h' &&
8286 if (name[1] == 'u' &&
8299 if (name[2] == 'c' &&
8308 if (name[2] == 's' &&
8317 if (name[2] == 'a' &&
8353 if (name[1] == 'o' &&
8366 if (name[2] == 't' &&
8375 if (name[2] == 'o' &&
8384 if (name[2] == 't' &&
8393 if (name[2] == 'e' &&
8406 if (name[1] == 'o' &&
8419 if (name[2] == 'y' &&
8428 if (name[2] == 'l' &&
8444 if (name[2] == 's' &&
8453 if (name[2] == 'n' &&
8462 if (name[2] == 'c' &&
8475 if (name[1] == 'e' &&
8485 if (name[1] == 'p' &&
8498 if (name[2] == 'c' &&
8507 if (name[2] == 'p' &&
8516 if (name[2] == 's' &&
8532 if (name[2] == 'n' &&
8602 if (name[2] == 'r' &&
8611 if (name[2] == 'r' &&
8620 if (name[2] == 'a' &&
8636 if (name[2] == 'l' &&
8698 if (name[2] == 'e' &&
8701 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8714 case 5: /* 39 tokens of length 5 */
8718 if (name[1] == 'E' &&
8729 if (name[1] == 'H' &&
8743 if (name[2] == 'a' &&
8753 if (name[2] == 'a' &&
8770 if (name[2] == 'e' &&
8780 if (name[2] == 'e' &&
8784 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8800 if (name[3] == 'i' &&
8809 if (name[3] == 'o' &&
8845 if (name[2] == 'o' &&
8855 if (name[2] == 'y' &&
8869 if (name[1] == 'l' &&
8883 if (name[2] == 'n' &&
8893 if (name[2] == 'o' &&
8907 if (name[1] == 'i' &&
8912 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8921 if (name[2] == 'd' &&
8931 if (name[2] == 'c' &&
8948 if (name[2] == 'c' &&
8958 if (name[2] == 't' &&
8972 if (name[1] == 'k' &&
8983 if (name[1] == 'r' &&
8997 if (name[2] == 's' &&
9007 if (name[2] == 'd' &&
9024 if (name[2] == 'm' &&
9034 if (name[2] == 'i' &&
9044 if (name[2] == 'e' &&
9054 if (name[2] == 'l' &&
9064 if (name[2] == 'a' &&
9077 if (name[3] == 't' &&
9080 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9086 if (name[3] == 'd' &&
9103 if (name[1] == 'i' &&
9117 if (name[2] == 'a' &&
9130 if (name[3] == 'e' &&
9165 if (name[2] == 'i' &&
9182 if (name[2] == 'i' &&
9192 if (name[2] == 'i' &&
9209 case 6: /* 33 tokens of length 6 */
9213 if (name[1] == 'c' &&
9228 if (name[2] == 'l' &&
9239 if (name[2] == 'r' &&
9254 if (name[1] == 'e' &&
9269 if (name[2] == 's' &&
9274 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9280 if (name[2] == 'i' &&
9298 if (name[2] == 'l' &&
9309 if (name[2] == 'r' &&
9324 if (name[1] == 'm' &&
9339 if (name[2] == 'n' &&
9350 if (name[2] == 's' &&
9365 if (name[1] == 's' &&
9371 if (name[4] == 't' &&
9380 if (name[4] == 'e' &&
9389 if (name[4] == 'c' &&
9398 if (name[4] == 'n' &&
9414 if (name[1] == 'r' &&
9432 if (name[3] == 'a' &&
9442 if (name[3] == 'u' &&
9456 if (name[2] == 'n' &&
9474 if (name[2] == 'a' &&
9488 if (name[3] == 'e' &&
9501 if (name[4] == 't' &&
9510 if (name[4] == 'e' &&
9532 if (name[4] == 't' &&
9541 if (name[4] == 'e' &&
9557 if (name[2] == 'c' &&
9568 if (name[2] == 'l' &&
9579 if (name[2] == 'b' &&
9590 if (name[2] == 's' &&
9613 if (name[4] == 's' &&
9622 if (name[4] == 'n' &&
9635 if (name[3] == 'a' &&
9652 if (name[1] == 'a' &&
9667 case 7: /* 29 tokens of length 7 */
9671 if (name[1] == 'E' &&
9684 if (name[1] == '_' &&
9697 if (name[1] == 'i' &&
9704 return -KEY_binmode;
9710 if (name[1] == 'o' &&
9717 return -KEY_connect;
9726 if (name[2] == 'm' &&
9732 return -KEY_dbmopen;
9743 if (name[4] == 'u' &&
9747 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9753 if (name[4] == 'n' &&
9774 if (name[1] == 'o' &&
9787 if (name[1] == 'e' &&
9794 if (name[5] == 'r' &&
9797 return -KEY_getpgrp;
9803 if (name[5] == 'i' &&
9806 return -KEY_getppid;
9819 if (name[1] == 'c' &&
9826 return -KEY_lcfirst;
9832 if (name[1] == 'p' &&
9839 return -KEY_opendir;
9845 if (name[1] == 'a' &&
9863 if (name[3] == 'd' &&
9868 return -KEY_readdir;
9874 if (name[3] == 'u' &&
9885 if (name[3] == 'e' &&
9890 return -KEY_reverse;
9909 if (name[3] == 'k' &&
9914 return -KEY_seekdir;
9920 if (name[3] == 'p' &&
9925 return -KEY_setpgrp;
9935 if (name[2] == 'm' &&
9941 return -KEY_shmread;
9947 if (name[2] == 'r' &&
9953 return -KEY_sprintf;
9962 if (name[3] == 'l' &&
9967 return -KEY_symlink;
9976 if (name[4] == 'a' &&
9980 return -KEY_syscall;
9986 if (name[4] == 'p' &&
9990 return -KEY_sysopen;
9996 if (name[4] == 'e' &&
10000 return -KEY_sysread;
10006 if (name[4] == 'e' &&
10010 return -KEY_sysseek;
10028 if (name[1] == 'e' &&
10035 return -KEY_telldir;
10044 if (name[2] == 'f' &&
10050 return -KEY_ucfirst;
10056 if (name[2] == 's' &&
10062 return -KEY_unshift;
10072 if (name[1] == 'a' &&
10079 return -KEY_waitpid;
10088 case 8: /* 26 tokens of length 8 */
10092 if (name[1] == 'U' &&
10100 return KEY_AUTOLOAD;
10106 if (name[1] == '_')
10111 if (name[3] == 'A' &&
10117 return KEY___DATA__;
10123 if (name[3] == 'I' &&
10129 return -KEY___FILE__;
10135 if (name[3] == 'I' &&
10141 return -KEY___LINE__;
10157 if (name[2] == 'o' &&
10164 return -KEY_closedir;
10170 if (name[2] == 'n' &&
10177 return -KEY_continue;
10187 if (name[1] == 'b' &&
10195 return -KEY_dbmclose;
10201 if (name[1] == 'n' &&
10207 if (name[4] == 'r' &&
10212 return -KEY_endgrent;
10218 if (name[4] == 'w' &&
10223 return -KEY_endpwent;
10236 if (name[1] == 'o' &&
10244 return -KEY_formline;
10250 if (name[1] == 'e' &&
10256 if (name[4] == 'r')
10261 if (name[6] == 'n' &&
10264 return -KEY_getgrent;
10270 if (name[6] == 'i' &&
10273 return -KEY_getgrgid;
10279 if (name[6] == 'a' &&
10282 return -KEY_getgrnam;
10295 if (name[4] == 'o' &&
10300 return -KEY_getlogin;
10306 if (name[4] == 'w')
10311 if (name[6] == 'n' &&
10314 return -KEY_getpwent;
10320 if (name[6] == 'a' &&
10323 return -KEY_getpwnam;
10329 if (name[6] == 'i' &&
10332 return -KEY_getpwuid;
10352 if (name[1] == 'e' &&
10359 if (name[5] == 'i' &&
10366 return -KEY_readline;
10371 return -KEY_readlink;
10382 if (name[5] == 'i' &&
10386 return -KEY_readpipe;
10402 if (name[2] == 't')
10407 if (name[4] == 'r' &&
10412 return -KEY_setgrent;
10418 if (name[4] == 'w' &&
10423 return -KEY_setpwent;
10439 if (name[3] == 'w' &&
10445 return -KEY_shmwrite;
10451 if (name[3] == 't' &&
10457 return -KEY_shutdown;
10467 if (name[2] == 's' &&
10474 return -KEY_syswrite;
10484 if (name[1] == 'r' &&
10492 return -KEY_truncate;
10501 case 9: /* 9 tokens of length 9 */
10505 if (name[1] == 'N' &&
10514 return KEY_UNITCHECK;
10520 if (name[1] == 'n' &&
10529 return -KEY_endnetent;
10535 if (name[1] == 'e' &&
10544 return -KEY_getnetent;
10550 if (name[1] == 'o' &&
10559 return -KEY_localtime;
10565 if (name[1] == 'r' &&
10574 return KEY_prototype;
10580 if (name[1] == 'u' &&
10589 return -KEY_quotemeta;
10595 if (name[1] == 'e' &&
10604 return -KEY_rewinddir;
10610 if (name[1] == 'e' &&
10619 return -KEY_setnetent;
10625 if (name[1] == 'a' &&
10634 return -KEY_wantarray;
10643 case 10: /* 9 tokens of length 10 */
10647 if (name[1] == 'n' &&
10653 if (name[4] == 'o' &&
10660 return -KEY_endhostent;
10666 if (name[4] == 'e' &&
10673 return -KEY_endservent;
10686 if (name[1] == 'e' &&
10692 if (name[4] == 'o' &&
10699 return -KEY_gethostent;
10708 if (name[5] == 'r' &&
10714 return -KEY_getservent;
10720 if (name[5] == 'c' &&
10726 return -KEY_getsockopt;
10746 if (name[2] == 't')
10751 if (name[4] == 'o' &&
10758 return -KEY_sethostent;
10767 if (name[5] == 'r' &&
10773 return -KEY_setservent;
10779 if (name[5] == 'c' &&
10785 return -KEY_setsockopt;
10802 if (name[2] == 'c' &&
10811 return -KEY_socketpair;
10824 case 11: /* 8 tokens of length 11 */
10828 if (name[1] == '_' &&
10838 { /* __PACKAGE__ */
10839 return -KEY___PACKAGE__;
10845 if (name[1] == 'n' &&
10855 { /* endprotoent */
10856 return -KEY_endprotoent;
10862 if (name[1] == 'e' &&
10871 if (name[5] == 'e' &&
10877 { /* getpeername */
10878 return -KEY_getpeername;
10887 if (name[6] == 'o' &&
10892 { /* getpriority */
10893 return -KEY_getpriority;
10899 if (name[6] == 't' &&
10904 { /* getprotoent */
10905 return -KEY_getprotoent;
10919 if (name[4] == 'o' &&
10926 { /* getsockname */
10927 return -KEY_getsockname;
10940 if (name[1] == 'e' &&
10948 if (name[6] == 'o' &&
10953 { /* setpriority */
10954 return -KEY_setpriority;
10960 if (name[6] == 't' &&
10965 { /* setprotoent */
10966 return -KEY_setprotoent;
10982 case 12: /* 2 tokens of length 12 */
10983 if (name[0] == 'g' &&
10995 if (name[9] == 'd' &&
10998 { /* getnetbyaddr */
10999 return -KEY_getnetbyaddr;
11005 if (name[9] == 'a' &&
11008 { /* getnetbyname */
11009 return -KEY_getnetbyname;
11021 case 13: /* 4 tokens of length 13 */
11022 if (name[0] == 'g' &&
11029 if (name[4] == 'o' &&
11038 if (name[10] == 'd' &&
11041 { /* gethostbyaddr */
11042 return -KEY_gethostbyaddr;
11048 if (name[10] == 'a' &&
11051 { /* gethostbyname */
11052 return -KEY_gethostbyname;
11065 if (name[4] == 'e' &&
11074 if (name[10] == 'a' &&
11077 { /* getservbyname */
11078 return -KEY_getservbyname;
11084 if (name[10] == 'o' &&
11087 { /* getservbyport */
11088 return -KEY_getservbyport;
11107 case 14: /* 1 tokens of length 14 */
11108 if (name[0] == 'g' &&
11122 { /* getprotobyname */
11123 return -KEY_getprotobyname;
11128 case 16: /* 1 tokens of length 16 */
11129 if (name[0] == 'g' &&
11145 { /* getprotobynumber */
11146 return -KEY_getprotobynumber;
11160 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11164 PERL_ARGS_ASSERT_CHECKCOMMA;
11166 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11167 if (ckWARN(WARN_SYNTAX)) {
11170 for (w = s+2; *w && level; w++) {
11173 else if (*w == ')')
11176 while (isSPACE(*w))
11178 /* the list of chars below is for end of statements or
11179 * block / parens, boolean operators (&&, ||, //) and branch
11180 * constructs (or, and, if, until, unless, while, err, for).
11181 * Not a very solid hack... */
11182 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11183 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11184 "%s (...) interpreted as function",name);
11187 while (s < PL_bufend && isSPACE(*s))
11191 while (s < PL_bufend && isSPACE(*s))
11193 if (isIDFIRST_lazy_if(s,UTF)) {
11194 const char * const w = s++;
11195 while (isALNUM_lazy_if(s,UTF))
11197 while (s < PL_bufend && isSPACE(*s))
11201 if (keyword(w, s - w, 0))
11204 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11205 if (gv && GvCVu(gv))
11207 Perl_croak(aTHX_ "No comma allowed after %s", what);
11212 /* Either returns sv, or mortalizes sv and returns a new SV*.
11213 Best used as sv=new_constant(..., sv, ...).
11214 If s, pv are NULL, calls subroutine with one argument,
11215 and type is used with error messages only. */
11218 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11219 SV *sv, SV *pv, const char *type, STRLEN typelen)
11222 HV * const table = GvHV(PL_hintgv); /* ^H */
11226 const char *why1 = "", *why2 = "", *why3 = "";
11228 PERL_ARGS_ASSERT_NEW_CONSTANT;
11230 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11233 why2 = (const char *)
11234 (strEQ(key,"charnames")
11235 ? "(possibly a missing \"use charnames ...\")"
11237 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11238 (type ? type: "undef"), why2);
11240 /* This is convoluted and evil ("goto considered harmful")
11241 * but I do not understand the intricacies of all the different
11242 * failure modes of %^H in here. The goal here is to make
11243 * the most probable error message user-friendly. --jhi */
11248 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11249 (type ? type: "undef"), why1, why2, why3);
11251 yyerror(SvPVX_const(msg));
11255 cvp = hv_fetch(table, key, keylen, FALSE);
11256 if (!cvp || !SvOK(*cvp)) {
11259 why3 = "} is not defined";
11262 sv_2mortal(sv); /* Parent created it permanently */
11265 pv = newSVpvn_flags(s, len, SVs_TEMP);
11267 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11269 typesv = &PL_sv_undef;
11271 PUSHSTACKi(PERLSI_OVERLOAD);
11283 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11287 /* Check the eval first */
11288 if (!PL_in_eval && SvTRUE(ERRSV)) {
11289 sv_catpvs(ERRSV, "Propagated");
11290 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11292 res = SvREFCNT_inc_simple(sv);
11296 SvREFCNT_inc_simple_void(res);
11305 why1 = "Call to &{$^H{";
11307 why3 = "}} did not return a defined value";
11315 /* Returns a NUL terminated string, with the length of the string written to
11319 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11322 register char *d = dest;
11323 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11325 PERL_ARGS_ASSERT_SCAN_WORD;
11329 Perl_croak(aTHX_ ident_too_long);
11330 if (isALNUM(*s)) /* UTF handled below */
11332 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11337 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11341 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11342 char *t = s + UTF8SKIP(s);
11344 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11348 Perl_croak(aTHX_ ident_too_long);
11349 Copy(s, d, len, char);
11362 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11365 char *bracket = NULL;
11367 register char *d = dest;
11368 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
11370 PERL_ARGS_ASSERT_SCAN_IDENT;
11375 while (isDIGIT(*s)) {
11377 Perl_croak(aTHX_ ident_too_long);
11384 Perl_croak(aTHX_ ident_too_long);
11385 if (isALNUM(*s)) /* UTF handled below */
11387 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11392 else if (*s == ':' && s[1] == ':') {
11396 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11397 char *t = s + UTF8SKIP(s);
11398 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11400 if (d + (t - s) > e)
11401 Perl_croak(aTHX_ ident_too_long);
11402 Copy(s, d, t - s, char);
11413 if (PL_lex_state != LEX_NORMAL)
11414 PL_lex_state = LEX_INTERPENDMAYBE;
11417 if (*s == '$' && s[1] &&
11418 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11431 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11436 if (isSPACE(s[-1])) {
11438 const char ch = *s++;
11439 if (!SPACE_OR_TAB(ch)) {
11445 if (isIDFIRST_lazy_if(d,UTF)) {
11449 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11450 end += UTF8SKIP(end);
11451 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11452 end += UTF8SKIP(end);
11454 Copy(s, d, end - s, char);
11459 while ((isALNUM(*s) || *s == ':') && d < e)
11462 Perl_croak(aTHX_ ident_too_long);
11465 while (s < send && SPACE_OR_TAB(*s))
11467 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11468 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11469 const char * const brack =
11471 ((*s == '[') ? "[...]" : "{...}");
11472 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11473 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11474 funny, dest, brack, funny, dest, brack);
11477 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11481 /* Handle extended ${^Foo} variables
11482 * 1999-02-27 mjd-perl-patch@plover.com */
11483 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11487 while (isALNUM(*s) && d < e) {
11491 Perl_croak(aTHX_ ident_too_long);
11496 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11497 PL_lex_state = LEX_INTERPEND;
11500 if (PL_lex_state == LEX_NORMAL) {
11501 if (ckWARN(WARN_AMBIGUOUS) &&
11502 (keyword(dest, d - dest, 0)
11503 || get_cvn_flags(dest, d - dest, 0)))
11507 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11508 "Ambiguous use of %c{%s} resolved to %c%s",
11509 funny, dest, funny, dest);
11514 s = bracket; /* let the parser handle it */
11518 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11519 PL_lex_state = LEX_INTERPEND;
11524 S_pmflag(U32 pmfl, const char ch) {
11526 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11527 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11528 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11529 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11530 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11536 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11538 PERL_ARGS_ASSERT_PMFLAG;
11540 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11541 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11544 *pmfl = S_pmflag(*pmfl, (char)ch);
11549 S_scan_pat(pTHX_ char *start, I32 type)
11553 char *s = scan_str(start,!!PL_madskills,FALSE);
11554 const char * const valid_flags =
11555 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11560 PERL_ARGS_ASSERT_SCAN_PAT;
11563 const char * const delimiter = skipspace(start);
11567 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11568 : "Search pattern not terminated" ));
11571 pm = (PMOP*)newPMOP(type, 0);
11572 if (PL_multi_open == '?') {
11573 /* This is the only point in the code that sets PMf_ONCE: */
11574 pm->op_pmflags |= PMf_ONCE;
11576 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11577 allows us to restrict the list needed by reset to just the ??
11579 assert(type != OP_TRANS);
11581 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11584 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11587 elements = mg->mg_len / sizeof(PMOP**);
11588 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11589 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11590 mg->mg_len = elements * sizeof(PMOP**);
11591 PmopSTASH_set(pm,PL_curstash);
11597 while (*s && strchr(valid_flags, *s))
11598 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11600 if (PL_madskills && modstart != s) {
11601 SV* tmptoken = newSVpvn(modstart, s - modstart);
11602 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11605 /* issue a warning if /c is specified,but /g is not */
11606 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11608 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11609 "Use of /c modifier is meaningless without /g" );
11612 PL_lex_op = (OP*)pm;
11613 pl_yylval.ival = OP_MATCH;
11618 S_scan_subst(pTHX_ char *start)
11629 PERL_ARGS_ASSERT_SCAN_SUBST;
11631 pl_yylval.ival = OP_NULL;
11633 s = scan_str(start,!!PL_madskills,FALSE);
11636 Perl_croak(aTHX_ "Substitution pattern not terminated");
11638 if (s[-1] == PL_multi_open)
11641 if (PL_madskills) {
11642 CURMAD('q', PL_thisopen);
11643 CURMAD('_', PL_thiswhite);
11644 CURMAD('E', PL_thisstuff);
11645 CURMAD('Q', PL_thisclose);
11646 PL_realtokenstart = s - SvPVX(PL_linestr);
11650 first_start = PL_multi_start;
11651 s = scan_str(s,!!PL_madskills,FALSE);
11653 if (PL_lex_stuff) {
11654 SvREFCNT_dec(PL_lex_stuff);
11655 PL_lex_stuff = NULL;
11657 Perl_croak(aTHX_ "Substitution replacement not terminated");
11659 PL_multi_start = first_start; /* so whole substitution is taken together */
11661 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11664 if (PL_madskills) {
11665 CURMAD('z', PL_thisopen);
11666 CURMAD('R', PL_thisstuff);
11667 CURMAD('Z', PL_thisclose);
11673 if (*s == EXEC_PAT_MOD) {
11677 else if (strchr(S_PAT_MODS, *s))
11678 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11684 if (PL_madskills) {
11686 curmad('m', newSVpvn(modstart, s - modstart));
11687 append_madprops(PL_thismad, (OP*)pm, 0);
11691 if ((pm->op_pmflags & PMf_CONTINUE)) {
11692 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11696 SV * const repl = newSVpvs("");
11698 PL_sublex_info.super_bufptr = s;
11699 PL_sublex_info.super_bufend = PL_bufend;
11701 pm->op_pmflags |= PMf_EVAL;
11704 sv_catpvs(repl, "eval ");
11706 sv_catpvs(repl, "do ");
11708 sv_catpvs(repl, "{");
11709 sv_catsv(repl, PL_lex_repl);
11710 if (strchr(SvPVX(PL_lex_repl), '#'))
11711 sv_catpvs(repl, "\n");
11712 sv_catpvs(repl, "}");
11714 SvREFCNT_dec(PL_lex_repl);
11715 PL_lex_repl = repl;
11718 PL_lex_op = (OP*)pm;
11719 pl_yylval.ival = OP_SUBST;
11724 S_scan_trans(pTHX_ char *start)
11737 PERL_ARGS_ASSERT_SCAN_TRANS;
11739 pl_yylval.ival = OP_NULL;
11741 s = scan_str(start,!!PL_madskills,FALSE);
11743 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11745 if (s[-1] == PL_multi_open)
11748 if (PL_madskills) {
11749 CURMAD('q', PL_thisopen);
11750 CURMAD('_', PL_thiswhite);
11751 CURMAD('E', PL_thisstuff);
11752 CURMAD('Q', PL_thisclose);
11753 PL_realtokenstart = s - SvPVX(PL_linestr);
11757 s = scan_str(s,!!PL_madskills,FALSE);
11759 if (PL_lex_stuff) {
11760 SvREFCNT_dec(PL_lex_stuff);
11761 PL_lex_stuff = NULL;
11763 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11765 if (PL_madskills) {
11766 CURMAD('z', PL_thisopen);
11767 CURMAD('R', PL_thisstuff);
11768 CURMAD('Z', PL_thisclose);
11771 complement = del = squash = 0;
11778 complement = OPpTRANS_COMPLEMENT;
11781 del = OPpTRANS_DELETE;
11784 squash = OPpTRANS_SQUASH;
11793 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11794 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11795 o->op_private &= ~OPpTRANS_ALL;
11796 o->op_private |= del|squash|complement|
11797 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11798 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11801 pl_yylval.ival = OP_TRANS;
11804 if (PL_madskills) {
11806 curmad('m', newSVpvn(modstart, s - modstart));
11807 append_madprops(PL_thismad, o, 0);
11816 S_scan_heredoc(pTHX_ register char *s)
11820 I32 op_type = OP_SCALAR;
11824 const char *found_newline;
11828 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11830 I32 stuffstart = s - SvPVX(PL_linestr);
11833 PL_realtokenstart = -1;
11836 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11840 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11844 while (SPACE_OR_TAB(*peek))
11846 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11849 s = delimcpy(d, e, s, PL_bufend, term, &len);
11859 if (!isALNUM_lazy_if(s,UTF))
11860 deprecate("bare << to mean <<\"\"");
11861 for (; isALNUM_lazy_if(s,UTF); s++) {
11866 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11867 Perl_croak(aTHX_ "Delimiter for here document is too long");
11870 len = d - PL_tokenbuf;
11873 if (PL_madskills) {
11874 tstart = PL_tokenbuf + !outer;
11875 PL_thisclose = newSVpvn(tstart, len - !outer);
11876 tstart = SvPVX(PL_linestr) + stuffstart;
11877 PL_thisopen = newSVpvn(tstart, s - tstart);
11878 stuffstart = s - SvPVX(PL_linestr);
11881 #ifndef PERL_STRICT_CR
11882 d = strchr(s, '\r');
11884 char * const olds = s;
11886 while (s < PL_bufend) {
11892 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11901 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11908 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11909 herewas = newSVpvn(s,PL_bufend-s);
11913 herewas = newSVpvn(s-1,found_newline-s+1);
11916 herewas = newSVpvn(s,found_newline-s);
11920 if (PL_madskills) {
11921 tstart = SvPVX(PL_linestr) + stuffstart;
11923 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11925 PL_thisstuff = newSVpvn(tstart, s - tstart);
11928 s += SvCUR(herewas);
11931 stuffstart = s - SvPVX(PL_linestr);
11937 tmpstr = newSV_type(SVt_PVIV);
11938 SvGROW(tmpstr, 80);
11939 if (term == '\'') {
11940 op_type = OP_CONST;
11941 SvIV_set(tmpstr, -1);
11943 else if (term == '`') {
11944 op_type = OP_BACKTICK;
11945 SvIV_set(tmpstr, '\\');
11949 PL_multi_start = CopLINE(PL_curcop);
11950 PL_multi_open = PL_multi_close = '<';
11951 term = *PL_tokenbuf;
11952 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11953 char * const bufptr = PL_sublex_info.super_bufptr;
11954 char * const bufend = PL_sublex_info.super_bufend;
11955 char * const olds = s - SvCUR(herewas);
11956 s = strchr(bufptr, '\n');
11960 while (s < bufend &&
11961 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11963 CopLINE_inc(PL_curcop);
11966 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11967 missingterm(PL_tokenbuf);
11969 sv_setpvn(herewas,bufptr,d-bufptr+1);
11970 sv_setpvn(tmpstr,d+1,s-d);
11972 sv_catpvn(herewas,s,bufend-s);
11973 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11980 while (s < PL_bufend &&
11981 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11983 CopLINE_inc(PL_curcop);
11985 if (s >= PL_bufend) {
11986 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11987 missingterm(PL_tokenbuf);
11989 sv_setpvn(tmpstr,d+1,s-d);
11991 if (PL_madskills) {
11993 sv_catpvn(PL_thisstuff, d + 1, s - d);
11995 PL_thisstuff = newSVpvn(d + 1, s - d);
11996 stuffstart = s - SvPVX(PL_linestr);
12000 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12002 sv_catpvn(herewas,s,PL_bufend-s);
12003 sv_setsv(PL_linestr,herewas);
12004 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12005 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12006 PL_last_lop = PL_last_uni = NULL;
12009 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12010 while (s >= PL_bufend) { /* multiple line string? */
12012 if (PL_madskills) {
12013 tstart = SvPVX(PL_linestr) + stuffstart;
12015 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12017 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12021 CopLINE_inc(PL_curcop);
12022 if (!outer || !lex_next_chunk(0)) {
12023 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12024 missingterm(PL_tokenbuf);
12026 CopLINE_dec(PL_curcop);
12029 stuffstart = s - SvPVX(PL_linestr);
12031 CopLINE_inc(PL_curcop);
12032 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12033 PL_last_lop = PL_last_uni = NULL;
12034 #ifndef PERL_STRICT_CR
12035 if (PL_bufend - PL_linestart >= 2) {
12036 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12037 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12039 PL_bufend[-2] = '\n';
12041 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12043 else if (PL_bufend[-1] == '\r')
12044 PL_bufend[-1] = '\n';
12046 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12047 PL_bufend[-1] = '\n';
12049 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12050 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12051 *(SvPVX(PL_linestr) + off ) = ' ';
12052 sv_catsv(PL_linestr,herewas);
12053 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12054 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12058 sv_catsv(tmpstr,PL_linestr);
12063 PL_multi_end = CopLINE(PL_curcop);
12064 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12065 SvPV_shrink_to_cur(tmpstr);
12067 SvREFCNT_dec(herewas);
12069 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12071 else if (PL_encoding)
12072 sv_recode_to_utf8(tmpstr, PL_encoding);
12074 PL_lex_stuff = tmpstr;
12075 pl_yylval.ival = op_type;
12079 /* scan_inputsymbol
12080 takes: current position in input buffer
12081 returns: new position in input buffer
12082 side-effects: pl_yylval and lex_op are set.
12087 <FH> read from filehandle
12088 <pkg::FH> read from package qualified filehandle
12089 <pkg'FH> read from package qualified filehandle
12090 <$fh> read from filehandle in $fh
12091 <*.h> filename glob
12096 S_scan_inputsymbol(pTHX_ char *start)
12099 register char *s = start; /* current position in buffer */
12102 char *d = PL_tokenbuf; /* start of temp holding space */
12103 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12105 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12107 end = strchr(s, '\n');
12110 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12112 /* die if we didn't have space for the contents of the <>,
12113 or if it didn't end, or if we see a newline
12116 if (len >= (I32)sizeof PL_tokenbuf)
12117 Perl_croak(aTHX_ "Excessively long <> operator");
12119 Perl_croak(aTHX_ "Unterminated <> operator");
12124 Remember, only scalar variables are interpreted as filehandles by
12125 this code. Anything more complex (e.g., <$fh{$num}>) will be
12126 treated as a glob() call.
12127 This code makes use of the fact that except for the $ at the front,
12128 a scalar variable and a filehandle look the same.
12130 if (*d == '$' && d[1]) d++;
12132 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12133 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12136 /* If we've tried to read what we allow filehandles to look like, and
12137 there's still text left, then it must be a glob() and not a getline.
12138 Use scan_str to pull out the stuff between the <> and treat it
12139 as nothing more than a string.
12142 if (d - PL_tokenbuf != len) {
12143 pl_yylval.ival = OP_GLOB;
12144 s = scan_str(start,!!PL_madskills,FALSE);
12146 Perl_croak(aTHX_ "Glob not terminated");
12150 bool readline_overriden = FALSE;
12153 /* we're in a filehandle read situation */
12156 /* turn <> into <ARGV> */
12158 Copy("ARGV",d,5,char);
12160 /* Check whether readline() is overriden */
12161 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12163 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12165 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12166 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12167 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12168 readline_overriden = TRUE;
12170 /* if <$fh>, create the ops to turn the variable into a
12174 /* try to find it in the pad for this block, otherwise find
12175 add symbol table ops
12177 const PADOFFSET tmp = pad_findmy(d, len, 0);
12178 if (tmp != NOT_IN_PAD) {
12179 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12180 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12181 HEK * const stashname = HvNAME_HEK(stash);
12182 SV * const sym = sv_2mortal(newSVhek(stashname));
12183 sv_catpvs(sym, "::");
12184 sv_catpv(sym, d+1);
12189 OP * const o = newOP(OP_PADSV, 0);
12191 PL_lex_op = readline_overriden
12192 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12193 append_elem(OP_LIST, o,
12194 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12195 : (OP*)newUNOP(OP_READLINE, 0, o);
12204 ? (GV_ADDMULTI | GV_ADDINEVAL)
12207 PL_lex_op = readline_overriden
12208 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12209 append_elem(OP_LIST,
12210 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12211 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12212 : (OP*)newUNOP(OP_READLINE, 0,
12213 newUNOP(OP_RV2SV, 0,
12214 newGVOP(OP_GV, 0, gv)));
12216 if (!readline_overriden)
12217 PL_lex_op->op_flags |= OPf_SPECIAL;
12218 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12219 pl_yylval.ival = OP_NULL;
12222 /* If it's none of the above, it must be a literal filehandle
12223 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12225 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12226 PL_lex_op = readline_overriden
12227 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12228 append_elem(OP_LIST,
12229 newGVOP(OP_GV, 0, gv),
12230 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12231 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12232 pl_yylval.ival = OP_NULL;
12241 takes: start position in buffer
12242 keep_quoted preserve \ on the embedded delimiter(s)
12243 keep_delims preserve the delimiters around the string
12244 returns: position to continue reading from buffer
12245 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12246 updates the read buffer.
12248 This subroutine pulls a string out of the input. It is called for:
12249 q single quotes q(literal text)
12250 ' single quotes 'literal text'
12251 qq double quotes qq(interpolate $here please)
12252 " double quotes "interpolate $here please"
12253 qx backticks qx(/bin/ls -l)
12254 ` backticks `/bin/ls -l`
12255 qw quote words @EXPORT_OK = qw( func() $spam )
12256 m// regexp match m/this/
12257 s/// regexp substitute s/this/that/
12258 tr/// string transliterate tr/this/that/
12259 y/// string transliterate y/this/that/
12260 ($*@) sub prototypes sub foo ($)
12261 (stuff) sub attr parameters sub foo : attr(stuff)
12262 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12264 In most of these cases (all but <>, patterns and transliterate)
12265 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12266 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12267 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12270 It skips whitespace before the string starts, and treats the first
12271 character as the delimiter. If the delimiter is one of ([{< then
12272 the corresponding "close" character )]}> is used as the closing
12273 delimiter. It allows quoting of delimiters, and if the string has
12274 balanced delimiters ([{<>}]) it allows nesting.
12276 On success, the SV with the resulting string is put into lex_stuff or,
12277 if that is already non-NULL, into lex_repl. The second case occurs only
12278 when parsing the RHS of the special constructs s/// and tr/// (y///).
12279 For convenience, the terminating delimiter character is stuffed into
12284 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12287 SV *sv; /* scalar value: string */
12288 const char *tmps; /* temp string, used for delimiter matching */
12289 register char *s = start; /* current position in the buffer */
12290 register char term; /* terminating character */
12291 register char *to; /* current position in the sv's data */
12292 I32 brackets = 1; /* bracket nesting level */
12293 bool has_utf8 = FALSE; /* is there any utf8 content? */
12294 I32 termcode; /* terminating char. code */
12295 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12296 STRLEN termlen; /* length of terminating string */
12297 int last_off = 0; /* last position for nesting bracket */
12303 PERL_ARGS_ASSERT_SCAN_STR;
12305 /* skip space before the delimiter */
12311 if (PL_realtokenstart >= 0) {
12312 stuffstart = PL_realtokenstart;
12313 PL_realtokenstart = -1;
12316 stuffstart = start - SvPVX(PL_linestr);
12318 /* mark where we are, in case we need to report errors */
12321 /* after skipping whitespace, the next character is the terminator */
12324 termcode = termstr[0] = term;
12328 termcode = utf8_to_uvchr((U8*)s, &termlen);
12329 Copy(s, termstr, termlen, U8);
12330 if (!UTF8_IS_INVARIANT(term))
12334 /* mark where we are */
12335 PL_multi_start = CopLINE(PL_curcop);
12336 PL_multi_open = term;
12338 /* find corresponding closing delimiter */
12339 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12340 termcode = termstr[0] = term = tmps[5];
12342 PL_multi_close = term;
12344 /* create a new SV to hold the contents. 79 is the SV's initial length.
12345 What a random number. */
12346 sv = newSV_type(SVt_PVIV);
12348 SvIV_set(sv, termcode);
12349 (void)SvPOK_only(sv); /* validate pointer */
12351 /* move past delimiter and try to read a complete string */
12353 sv_catpvn(sv, s, termlen);
12356 tstart = SvPVX(PL_linestr) + stuffstart;
12357 if (!PL_thisopen && !keep_delims) {
12358 PL_thisopen = newSVpvn(tstart, s - tstart);
12359 stuffstart = s - SvPVX(PL_linestr);
12363 if (PL_encoding && !UTF) {
12367 int offset = s - SvPVX_const(PL_linestr);
12368 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12369 &offset, (char*)termstr, termlen);
12370 const char * const ns = SvPVX_const(PL_linestr) + offset;
12371 char * const svlast = SvEND(sv) - 1;
12373 for (; s < ns; s++) {
12374 if (*s == '\n' && !PL_rsfp)
12375 CopLINE_inc(PL_curcop);
12378 goto read_more_line;
12380 /* handle quoted delimiters */
12381 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12383 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12385 if ((svlast-1 - t) % 2) {
12386 if (!keep_quoted) {
12387 *(svlast-1) = term;
12389 SvCUR_set(sv, SvCUR(sv) - 1);
12394 if (PL_multi_open == PL_multi_close) {
12400 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12401 /* At here, all closes are "was quoted" one,
12402 so we don't check PL_multi_close. */
12404 if (!keep_quoted && *(t+1) == PL_multi_open)
12409 else if (*t == PL_multi_open)
12417 SvCUR_set(sv, w - SvPVX_const(sv));
12419 last_off = w - SvPVX(sv);
12420 if (--brackets <= 0)
12425 if (!keep_delims) {
12426 SvCUR_set(sv, SvCUR(sv) - 1);
12432 /* extend sv if need be */
12433 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12434 /* set 'to' to the next character in the sv's string */
12435 to = SvPVX(sv)+SvCUR(sv);
12437 /* if open delimiter is the close delimiter read unbridle */
12438 if (PL_multi_open == PL_multi_close) {
12439 for (; s < PL_bufend; s++,to++) {
12440 /* embedded newlines increment the current line number */
12441 if (*s == '\n' && !PL_rsfp)
12442 CopLINE_inc(PL_curcop);
12443 /* handle quoted delimiters */
12444 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12445 if (!keep_quoted && s[1] == term)
12447 /* any other quotes are simply copied straight through */
12451 /* terminate when run out of buffer (the for() condition), or
12452 have found the terminator */
12453 else if (*s == term) {
12456 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12459 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12465 /* if the terminator isn't the same as the start character (e.g.,
12466 matched brackets), we have to allow more in the quoting, and
12467 be prepared for nested brackets.
12470 /* read until we run out of string, or we find the terminator */
12471 for (; s < PL_bufend; s++,to++) {
12472 /* embedded newlines increment the line count */
12473 if (*s == '\n' && !PL_rsfp)
12474 CopLINE_inc(PL_curcop);
12475 /* backslashes can escape the open or closing characters */
12476 if (*s == '\\' && s+1 < PL_bufend) {
12477 if (!keep_quoted &&
12478 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12483 /* allow nested opens and closes */
12484 else if (*s == PL_multi_close && --brackets <= 0)
12486 else if (*s == PL_multi_open)
12488 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12493 /* terminate the copied string and update the sv's end-of-string */
12495 SvCUR_set(sv, to - SvPVX_const(sv));
12498 * this next chunk reads more into the buffer if we're not done yet
12502 break; /* handle case where we are done yet :-) */
12504 #ifndef PERL_STRICT_CR
12505 if (to - SvPVX_const(sv) >= 2) {
12506 if ((to[-2] == '\r' && to[-1] == '\n') ||
12507 (to[-2] == '\n' && to[-1] == '\r'))
12511 SvCUR_set(sv, to - SvPVX_const(sv));
12513 else if (to[-1] == '\r')
12516 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12521 /* if we're out of file, or a read fails, bail and reset the current
12522 line marker so we can report where the unterminated string began
12525 if (PL_madskills) {
12526 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12528 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12530 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12533 CopLINE_inc(PL_curcop);
12534 PL_bufptr = PL_bufend;
12535 if (!lex_next_chunk(0)) {
12537 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12546 /* at this point, we have successfully read the delimited string */
12548 if (!PL_encoding || UTF) {
12550 if (PL_madskills) {
12551 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12552 const int len = s - tstart;
12554 sv_catpvn(PL_thisstuff, tstart, len);
12556 PL_thisstuff = newSVpvn(tstart, len);
12557 if (!PL_thisclose && !keep_delims)
12558 PL_thisclose = newSVpvn(s,termlen);
12563 sv_catpvn(sv, s, termlen);
12568 if (PL_madskills) {
12569 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12570 const int len = s - tstart - termlen;
12572 sv_catpvn(PL_thisstuff, tstart, len);
12574 PL_thisstuff = newSVpvn(tstart, len);
12575 if (!PL_thisclose && !keep_delims)
12576 PL_thisclose = newSVpvn(s - termlen,termlen);
12580 if (has_utf8 || PL_encoding)
12583 PL_multi_end = CopLINE(PL_curcop);
12585 /* if we allocated too much space, give some back */
12586 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12587 SvLEN_set(sv, SvCUR(sv) + 1);
12588 SvPV_renew(sv, SvLEN(sv));
12591 /* decide whether this is the first or second quoted string we've read
12604 takes: pointer to position in buffer
12605 returns: pointer to new position in buffer
12606 side-effects: builds ops for the constant in pl_yylval.op
12608 Read a number in any of the formats that Perl accepts:
12610 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12611 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12614 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12616 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12619 If it reads a number without a decimal point or an exponent, it will
12620 try converting the number to an integer and see if it can do so
12621 without loss of precision.
12625 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12628 register const char *s = start; /* current position in buffer */
12629 register char *d; /* destination in temp buffer */
12630 register char *e; /* end of temp buffer */
12631 NV nv; /* number read, as a double */
12632 SV *sv = NULL; /* place to put the converted number */
12633 bool floatit; /* boolean: int or float? */
12634 const char *lastub = NULL; /* position of last underbar */
12635 static char const number_too_long[] = "Number too long";
12637 PERL_ARGS_ASSERT_SCAN_NUM;
12639 /* We use the first character to decide what type of number this is */
12643 Perl_croak(aTHX_ "panic: scan_num");
12645 /* if it starts with a 0, it could be an octal number, a decimal in
12646 0.13 disguise, or a hexadecimal number, or a binary number. */
12650 u holds the "number so far"
12651 shift the power of 2 of the base
12652 (hex == 4, octal == 3, binary == 1)
12653 overflowed was the number more than we can hold?
12655 Shift is used when we add a digit. It also serves as an "are
12656 we in octal/hex/binary?" indicator to disallow hex characters
12657 when in octal mode.
12662 bool overflowed = FALSE;
12663 bool just_zero = TRUE; /* just plain 0 or binary number? */
12664 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12665 static const char* const bases[5] =
12666 { "", "binary", "", "octal", "hexadecimal" };
12667 static const char* const Bases[5] =
12668 { "", "Binary", "", "Octal", "Hexadecimal" };
12669 static const char* const maxima[5] =
12671 "0b11111111111111111111111111111111",
12675 const char *base, *Base, *max;
12677 /* check for hex */
12682 } else if (s[1] == 'b') {
12687 /* check for a decimal in disguise */
12688 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12690 /* so it must be octal */
12697 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12698 "Misplaced _ in number");
12702 base = bases[shift];
12703 Base = Bases[shift];
12704 max = maxima[shift];
12706 /* read the rest of the number */
12708 /* x is used in the overflow test,
12709 b is the digit we're adding on. */
12714 /* if we don't mention it, we're done */
12718 /* _ are ignored -- but warned about if consecutive */
12720 if (lastub && s == lastub + 1)
12721 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12722 "Misplaced _ in number");
12726 /* 8 and 9 are not octal */
12727 case '8': case '9':
12729 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12733 case '2': case '3': case '4':
12734 case '5': case '6': case '7':
12736 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12739 case '0': case '1':
12740 b = *s++ & 15; /* ASCII digit -> value of digit */
12744 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12745 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12746 /* make sure they said 0x */
12749 b = (*s++ & 7) + 9;
12751 /* Prepare to put the digit we have onto the end
12752 of the number so far. We check for overflows.
12758 x = u << shift; /* make room for the digit */
12760 if ((x >> shift) != u
12761 && !(PL_hints & HINT_NEW_BINARY)) {
12764 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12765 "Integer overflow in %s number",
12768 u = x | b; /* add the digit to the end */
12771 n *= nvshift[shift];
12772 /* If an NV has not enough bits in its
12773 * mantissa to represent an UV this summing of
12774 * small low-order numbers is a waste of time
12775 * (because the NV cannot preserve the
12776 * low-order bits anyway): we could just
12777 * remember when did we overflow and in the
12778 * end just multiply n by the right
12786 /* if we get here, we had success: make a scalar value from
12791 /* final misplaced underbar check */
12792 if (s[-1] == '_') {
12793 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12798 if (n > 4294967295.0)
12799 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12800 "%s number > %s non-portable",
12806 if (u > 0xffffffff)
12807 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12808 "%s number > %s non-portable",
12813 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12814 sv = new_constant(start, s - start, "integer",
12815 sv, NULL, NULL, 0);
12816 else if (PL_hints & HINT_NEW_BINARY)
12817 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12822 handle decimal numbers.
12823 we're also sent here when we read a 0 as the first digit
12825 case '1': case '2': case '3': case '4': case '5':
12826 case '6': case '7': case '8': case '9': case '.':
12829 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12832 /* read next group of digits and _ and copy into d */
12833 while (isDIGIT(*s) || *s == '_') {
12834 /* skip underscores, checking for misplaced ones
12838 if (lastub && s == lastub + 1)
12839 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12840 "Misplaced _ in number");
12844 /* check for end of fixed-length buffer */
12846 Perl_croak(aTHX_ number_too_long);
12847 /* if we're ok, copy the character */
12852 /* final misplaced underbar check */
12853 if (lastub && s == lastub + 1) {
12854 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12857 /* read a decimal portion if there is one. avoid
12858 3..5 being interpreted as the number 3. followed
12861 if (*s == '.' && s[1] != '.') {
12866 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12867 "Misplaced _ in number");
12871 /* copy, ignoring underbars, until we run out of digits.
12873 for (; isDIGIT(*s) || *s == '_'; s++) {
12874 /* fixed length buffer check */
12876 Perl_croak(aTHX_ number_too_long);
12878 if (lastub && s == lastub + 1)
12879 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12880 "Misplaced _ in number");
12886 /* fractional part ending in underbar? */
12887 if (s[-1] == '_') {
12888 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12889 "Misplaced _ in number");
12891 if (*s == '.' && isDIGIT(s[1])) {
12892 /* oops, it's really a v-string, but without the "v" */
12898 /* read exponent part, if present */
12899 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12903 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12904 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12906 /* stray preinitial _ */
12908 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12909 "Misplaced _ in number");
12913 /* allow positive or negative exponent */
12914 if (*s == '+' || *s == '-')
12917 /* stray initial _ */
12919 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12920 "Misplaced _ in number");
12924 /* read digits of exponent */
12925 while (isDIGIT(*s) || *s == '_') {
12928 Perl_croak(aTHX_ number_too_long);
12932 if (((lastub && s == lastub + 1) ||
12933 (!isDIGIT(s[1]) && s[1] != '_')))
12934 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12935 "Misplaced _ in number");
12942 /* make an sv from the string */
12946 We try to do an integer conversion first if no characters
12947 indicating "float" have been found.
12952 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12954 if (flags == IS_NUMBER_IN_UV) {
12956 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12959 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12960 if (uv <= (UV) IV_MIN)
12961 sv_setiv(sv, -(IV)uv);
12968 /* terminate the string */
12970 nv = Atof(PL_tokenbuf);
12975 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12976 const char *const key = floatit ? "float" : "integer";
12977 const STRLEN keylen = floatit ? 5 : 7;
12978 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12979 key, keylen, sv, NULL, NULL, 0);
12983 /* if it starts with a v, it could be a v-string */
12986 sv = newSV(5); /* preallocate storage space */
12987 s = scan_vstring(s, PL_bufend, sv);
12991 /* make the op for the constant and return */
12994 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12996 lvalp->opval = NULL;
13002 S_scan_formline(pTHX_ register char *s)
13005 register char *eol;
13007 SV * const stuff = newSVpvs("");
13008 bool needargs = FALSE;
13009 bool eofmt = FALSE;
13011 char *tokenstart = s;
13012 SV* savewhite = NULL;
13014 if (PL_madskills) {
13015 savewhite = PL_thiswhite;
13020 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13022 while (!needargs) {
13025 #ifdef PERL_STRICT_CR
13026 while (SPACE_OR_TAB(*t))
13029 while (SPACE_OR_TAB(*t) || *t == '\r')
13032 if (*t == '\n' || t == PL_bufend) {
13037 if (PL_in_eval && !PL_rsfp) {
13038 eol = (char *) memchr(s,'\n',PL_bufend-s);
13043 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13045 for (t = s; t < eol; t++) {
13046 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13048 goto enough; /* ~~ must be first line in formline */
13050 if (*t == '@' || *t == '^')
13054 sv_catpvn(stuff, s, eol-s);
13055 #ifndef PERL_STRICT_CR
13056 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13057 char *end = SvPVX(stuff) + SvCUR(stuff);
13060 SvCUR_set(stuff, SvCUR(stuff) - 1);
13071 if (PL_madskills) {
13073 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13075 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13078 PL_bufptr = PL_bufend;
13079 CopLINE_inc(PL_curcop);
13080 got_some = lex_next_chunk(0);
13081 CopLINE_dec(PL_curcop);
13084 tokenstart = PL_bufptr;
13092 if (SvCUR(stuff)) {
13095 PL_lex_state = LEX_NORMAL;
13096 start_force(PL_curforce);
13097 NEXTVAL_NEXTTOKE.ival = 0;
13101 PL_lex_state = LEX_FORMLINE;
13103 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13105 else if (PL_encoding)
13106 sv_recode_to_utf8(stuff, PL_encoding);
13108 start_force(PL_curforce);
13109 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13111 start_force(PL_curforce);
13112 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13116 SvREFCNT_dec(stuff);
13118 PL_lex_formbrack = 0;
13122 if (PL_madskills) {
13124 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13126 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13127 PL_thiswhite = savewhite;
13134 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13137 const I32 oldsavestack_ix = PL_savestack_ix;
13138 CV* const outsidecv = PL_compcv;
13141 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13143 SAVEI32(PL_subline);
13144 save_item(PL_subname);
13145 SAVESPTR(PL_compcv);
13147 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13148 CvFLAGS(PL_compcv) |= flags;
13150 PL_subline = CopLINE(PL_curcop);
13151 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13152 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13153 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13155 return oldsavestack_ix;
13159 #pragma segment Perl_yylex
13162 S_yywarn(pTHX_ const char *const s)
13166 PERL_ARGS_ASSERT_YYWARN;
13168 PL_in_eval |= EVAL_WARNONLY;
13170 PL_in_eval &= ~EVAL_WARNONLY;
13175 Perl_yyerror(pTHX_ const char *const s)
13178 const char *where = NULL;
13179 const char *context = NULL;
13182 int yychar = PL_parser->yychar;
13184 PERL_ARGS_ASSERT_YYERROR;
13186 if (!yychar || (yychar == ';' && !PL_rsfp))
13188 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13189 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13190 PL_oldbufptr != PL_bufptr) {
13193 The code below is removed for NetWare because it abends/crashes on NetWare
13194 when the script has error such as not having the closing quotes like:
13195 if ($var eq "value)
13196 Checking of white spaces is anyway done in NetWare code.
13199 while (isSPACE(*PL_oldoldbufptr))
13202 context = PL_oldoldbufptr;
13203 contlen = PL_bufptr - PL_oldoldbufptr;
13205 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13206 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13209 The code below is removed for NetWare because it abends/crashes on NetWare
13210 when the script has error such as not having the closing quotes like:
13211 if ($var eq "value)
13212 Checking of white spaces is anyway done in NetWare code.
13215 while (isSPACE(*PL_oldbufptr))
13218 context = PL_oldbufptr;
13219 contlen = PL_bufptr - PL_oldbufptr;
13221 else if (yychar > 255)
13222 where = "next token ???";
13223 else if (yychar == -2) { /* YYEMPTY */
13224 if (PL_lex_state == LEX_NORMAL ||
13225 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13226 where = "at end of line";
13227 else if (PL_lex_inpat)
13228 where = "within pattern";
13230 where = "within string";
13233 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13235 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13236 else if (isPRINT_LC(yychar)) {
13237 const char string = yychar;
13238 sv_catpvn(where_sv, &string, 1);
13241 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13242 where = SvPVX_const(where_sv);
13244 msg = sv_2mortal(newSVpv(s, 0));
13245 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13246 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13248 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13250 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13251 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13252 Perl_sv_catpvf(aTHX_ msg,
13253 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13254 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13257 if (PL_in_eval & EVAL_WARNONLY) {
13258 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13262 if (PL_error_count >= 10) {
13263 if (PL_in_eval && SvCUR(ERRSV))
13264 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13265 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13267 Perl_croak(aTHX_ "%s has too many errors.\n",
13268 OutCopFILE(PL_curcop));
13271 PL_in_my_stash = NULL;
13275 #pragma segment Main
13279 S_swallow_bom(pTHX_ U8 *s)
13282 const STRLEN slen = SvCUR(PL_linestr);
13284 PERL_ARGS_ASSERT_SWALLOW_BOM;
13288 if (s[1] == 0xFE) {
13289 /* UTF-16 little-endian? (or UTF32-LE?) */
13290 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13291 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
13292 #ifndef PERL_NO_UTF16_FILTER
13293 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
13295 if (PL_bufend > (char*)s) {
13296 s = add_utf16_textfilter(s, TRUE);
13299 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
13304 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13305 #ifndef PERL_NO_UTF16_FILTER
13306 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13308 if (PL_bufend > (char *)s) {
13309 s = add_utf16_textfilter(s, FALSE);
13312 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
13317 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13318 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13319 s += 3; /* UTF-8 */
13325 if (s[2] == 0xFE && s[3] == 0xFF) {
13326 /* UTF-32 big-endian */
13327 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
13330 else if (s[2] == 0 && s[3] != 0) {
13333 * are a good indicator of UTF-16BE. */
13334 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13335 s = add_utf16_textfilter(s, FALSE);
13340 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13341 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13342 s += 4; /* UTF-8 */
13348 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13351 * are a good indicator of UTF-16LE. */
13352 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13353 s = add_utf16_textfilter(s, TRUE);
13360 #ifndef PERL_NO_UTF16_FILTER
13362 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13365 SV *const filter = FILTER_DATA(idx);
13366 /* We re-use this each time round, throwing the contents away before we
13368 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13369 SV *const utf8_buffer = filter;
13370 IV status = IoPAGE(filter);
13371 const bool reverse = (bool) IoLINES(filter);
13374 /* As we're automatically added, at the lowest level, and hence only called
13375 from this file, we can be sure that we're not called in block mode. Hence
13376 don't bother writing code to deal with block mode. */
13378 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13381 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13383 DEBUG_P(PerlIO_printf(Perl_debug_log,
13384 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13385 FPTR2DPTR(void *, S_utf16_textfilter),
13386 reverse ? 'l' : 'b', idx, maxlen, status,
13387 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13394 /* First, look in our buffer of existing UTF-8 data: */
13395 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13399 } else if (status == 0) {
13401 IoPAGE(filter) = 0;
13402 nl = SvEND(utf8_buffer);
13405 STRLEN got = nl - SvPVX(utf8_buffer);
13406 /* Did we have anything to append? */
13408 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13409 /* Everything else in this code works just fine if SVp_POK isn't
13410 set. This, however, needs it, and we need it to work, else
13411 we loop infinitely because the buffer is never consumed. */
13412 sv_chop(utf8_buffer, nl);
13416 /* OK, not a complete line there, so need to read some more UTF-16.
13417 Read an extra octect if the buffer currently has an odd number. */
13421 if (SvCUR(utf16_buffer) >= 2) {
13422 /* Location of the high octet of the last complete code point.
13423 Gosh, UTF-16 is a pain. All the benefits of variable length,
13424 *coupled* with all the benefits of partial reads and
13426 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13427 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13429 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13433 /* We have the first half of a surrogate. Read more. */
13434 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13437 status = FILTER_READ(idx + 1, utf16_buffer,
13438 160 + (SvCUR(utf16_buffer) & 1));
13439 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13440 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13443 IoPAGE(filter) = status;
13448 chars = SvCUR(utf16_buffer) >> 1;
13449 have = SvCUR(utf8_buffer);
13450 SvGROW(utf8_buffer, have + chars * 3 + 1);
13453 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13454 (U8*)SvPVX_const(utf8_buffer) + have,
13455 chars * 2, &newlen);
13457 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13458 (U8*)SvPVX_const(utf8_buffer) + have,
13459 chars * 2, &newlen);
13461 SvCUR_set(utf8_buffer, have + newlen);
13464 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13465 it's private to us, and utf16_to_utf8{,reversed} take a
13466 (pointer,length) pair, rather than a NUL-terminated string. */
13467 if(SvCUR(utf16_buffer) & 1) {
13468 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13469 SvCUR_set(utf16_buffer, 1);
13471 SvCUR_set(utf16_buffer, 0);
13474 DEBUG_P(PerlIO_printf(Perl_debug_log,
13475 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13477 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13478 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13483 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13485 SV *filter = filter_add(S_utf16_textfilter, NULL);
13487 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13488 sv_setpvs(filter, "");
13489 IoLINES(filter) = reversed;
13490 IoPAGE(filter) = 1; /* Not EOF */
13492 /* Sadly, we have to return a valid pointer, come what may, so we have to
13493 ignore any error return from this. */
13494 SvCUR_set(PL_linestr, 0);
13495 if (FILTER_READ(0, PL_linestr, 0)) {
13496 SvUTF8_on(PL_linestr);
13498 SvUTF8_on(PL_linestr);
13500 PL_bufend = SvEND(PL_linestr);
13501 return (U8*)SvPVX(PL_linestr);
13506 Returns a pointer to the next character after the parsed
13507 vstring, as well as updating the passed in sv.
13509 Function must be called like
13512 s = scan_vstring(s,e,sv);
13514 where s and e are the start and end of the string.
13515 The sv should already be large enough to store the vstring
13516 passed in, for performance reasons.
13521 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13524 const char *pos = s;
13525 const char *start = s;
13527 PERL_ARGS_ASSERT_SCAN_VSTRING;
13529 if (*pos == 'v') pos++; /* get past 'v' */
13530 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13532 if ( *pos != '.') {
13533 /* this may not be a v-string if followed by => */
13534 const char *next = pos;
13535 while (next < e && isSPACE(*next))
13537 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13538 /* return string not v-string */
13539 sv_setpvn(sv,(char *)s,pos-s);
13540 return (char *)pos;
13544 if (!isALPHA(*pos)) {
13545 U8 tmpbuf[UTF8_MAXBYTES+1];
13548 s++; /* get past 'v' */
13553 /* this is atoi() that tolerates underscores */
13556 const char *end = pos;
13558 while (--end >= s) {
13560 const UV orev = rev;
13561 rev += (*end - '0') * mult;
13564 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13565 "Integer overflow in decimal number");
13569 if (rev > 0x7FFFFFFF)
13570 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13572 /* Append native character for the rev point */
13573 tmpend = uvchr_to_utf8(tmpbuf, rev);
13574 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13575 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13577 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13583 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13587 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13594 Perl_keyword_plugin_standard(pTHX_
13595 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13597 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13598 PERL_UNUSED_CONTEXT;
13599 PERL_UNUSED_ARG(keyword_ptr);
13600 PERL_UNUSED_ARG(keyword_len);
13601 PERL_UNUSED_ARG(op_ptr);
13602 return KEYWORD_PLUGIN_DECLINE;
13607 * c-indentation-style: bsd
13608 * c-basic-offset: 4
13609 * indent-tabs-mode: t
13612 * ex: set ts=8 sts=4 sw=4 noet: