3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
43 #define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46 #define pl_yylval (PL_parser->yylval)
48 /* YYINITDEPTH -- initial size of the parser's stacks. */
49 #define YYINITDEPTH 200
51 /* XXX temporary backwards compatibility */
52 #define PL_lex_brackets (PL_parser->lex_brackets)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XFAKEBRACK 128
130 #define XENUMMASK 127
132 #ifdef USE_UTF8_SCRIPTS
133 # define UTF (!IN_BYTES)
135 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
138 /* The maximum number of characters preceding the unrecognized one to display */
139 #define UNRECOGNIZED_PRECEDE_COUNT 10
141 /* In variables named $^X, these are the legal values for X.
142 * 1999-02-27 mjd-perl-patch@plover.com */
143 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
145 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
147 /* LEX_* are values for PL_lex_state, the state of the lexer.
148 * They are arranged oddly so that the guard on the switch statement
149 * can get by with a single comparison (if the compiler is smart enough).
152 /* #define LEX_NOTPARSING 11 is done in perl.h. */
154 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
155 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
156 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
157 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
158 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
160 /* at end of code, eg "$x" followed by: */
161 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
162 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
164 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
165 string or after \E, $foo, etc */
166 #define LEX_INTERPCONST 2 /* NOT USED */
167 #define LEX_FORMLINE 1 /* expecting a format line */
168 #define LEX_KNOWNEXT 0 /* next token known; just return it */
172 static const char* const lex_state_names[] = {
191 #include "keywords.h"
193 /* CLINE is a macro that ensures PL_copline has a sane value */
198 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
201 # define SKIPSPACE0(s) skipspace0(s)
202 # define SKIPSPACE1(s) skipspace1(s)
203 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
204 # define PEEKSPACE(s) skipspace2(s,0)
206 # define SKIPSPACE0(s) skipspace(s)
207 # define SKIPSPACE1(s) skipspace(s)
208 # define SKIPSPACE2(s,tsv) skipspace(s)
209 # define PEEKSPACE(s) skipspace(s)
213 * Convenience functions to return different tokens and prime the
214 * lexer for the next token. They all take an argument.
216 * TOKEN : generic token (used for '(', DOLSHARP, etc)
217 * OPERATOR : generic operator
218 * AOPERATOR : assignment operator
219 * PREBLOCK : beginning the block after an if, while, foreach, ...
220 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
221 * PREREF : *EXPR where EXPR is not a simple identifier
222 * TERM : expression term
223 * LOOPX : loop exiting command (goto, last, dump, etc)
224 * FTST : file test operator
225 * FUN0 : zero-argument function
226 * FUN1 : not used, except for not, which isn't a UNIOP
227 * BOop : bitwise or or xor
229 * SHop : shift operator
230 * PWop : power operator
231 * PMop : pattern-matching operator
232 * Aop : addition-level operator
233 * Mop : multiplication-level operator
234 * Eop : equality-testing operator
235 * Rop : relational operator <= != gt
237 * Also see LOP and lop() below.
240 #ifdef DEBUGGING /* Serve -DT. */
241 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
243 # define REPORT(retval) (retval)
246 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
247 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
248 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
249 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
251 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
252 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
253 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
254 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
255 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI2(f,x) { \
273 pl_yylval.ival = f; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI2(f,XTERM)
284 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
286 #define UNIBRACK(f) { \
287 pl_yylval.ival = f; \
289 PL_last_uni = PL_oldbufptr; \
291 return REPORT( (int)FUNC1 ); \
293 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
296 /* grandfather return to old style */
297 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
301 /* how to interpret the pl_yylval associated with the token */
305 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
311 static struct debug_tokens {
313 enum token_type type;
315 } const debug_tokens[] =
317 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
318 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
319 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
320 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
321 { ARROW, TOKENTYPE_NONE, "ARROW" },
322 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
323 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
324 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
325 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
326 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
327 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
328 { DO, TOKENTYPE_NONE, "DO" },
329 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
330 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
331 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
332 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
333 { ELSE, TOKENTYPE_NONE, "ELSE" },
334 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
335 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
336 { FOR, TOKENTYPE_IVAL, "FOR" },
337 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
338 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
339 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
340 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
341 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
342 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
343 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
344 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
345 { IF, TOKENTYPE_IVAL, "IF" },
346 { LABEL, TOKENTYPE_PVAL, "LABEL" },
347 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
348 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
349 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
350 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
351 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
352 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
353 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
354 { MY, TOKENTYPE_IVAL, "MY" },
355 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
356 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
357 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
358 { OROP, TOKENTYPE_IVAL, "OROP" },
359 { OROR, TOKENTYPE_NONE, "OROR" },
360 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
361 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
362 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
363 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
364 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
365 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
366 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
367 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
368 { PREINC, TOKENTYPE_NONE, "PREINC" },
369 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
370 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
371 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
372 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
373 { SUB, TOKENTYPE_NONE, "SUB" },
374 { THING, TOKENTYPE_OPVAL, "THING" },
375 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
376 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
377 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
378 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
379 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
380 { USE, TOKENTYPE_IVAL, "USE" },
381 { WHEN, TOKENTYPE_IVAL, "WHEN" },
382 { WHILE, TOKENTYPE_IVAL, "WHILE" },
383 { WORD, TOKENTYPE_OPVAL, "WORD" },
384 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
385 { 0, TOKENTYPE_NONE, NULL }
388 /* dump the returned token in rv, plus any optional arg in pl_yylval */
391 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
395 PERL_ARGS_ASSERT_TOKEREPORT;
398 const char *name = NULL;
399 enum token_type type = TOKENTYPE_NONE;
400 const struct debug_tokens *p;
401 SV* const report = newSVpvs("<== ");
403 for (p = debug_tokens; p->token; p++) {
404 if (p->token == (int)rv) {
411 Perl_sv_catpv(aTHX_ report, name);
412 else if ((char)rv > ' ' && (char)rv < '~')
413 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
415 sv_catpvs(report, "EOF");
417 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
420 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
423 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
425 case TOKENTYPE_OPNUM:
426 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
427 PL_op_name[lvalp->ival]);
430 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
432 case TOKENTYPE_OPVAL:
434 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
435 PL_op_name[lvalp->opval->op_type]);
436 if (lvalp->opval->op_type == OP_CONST) {
437 Perl_sv_catpvf(aTHX_ report, " %s",
438 SvPEEK(cSVOPx_sv(lvalp->opval)));
443 sv_catpvs(report, "(opval=null)");
446 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
452 /* print the buffer with suitable escapes */
455 S_printbuf(pTHX_ const char *const fmt, const char *const s)
457 SV* const tmp = newSVpvs("");
459 PERL_ARGS_ASSERT_PRINTBUF;
461 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
468 S_deprecate_commaless_var_list(pTHX) {
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
477 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
478 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
482 S_ao(pTHX_ int toketype)
485 if (*PL_bufptr == '=') {
487 if (toketype == ANDAND)
488 pl_yylval.ival = OP_ANDASSIGN;
489 else if (toketype == OROR)
490 pl_yylval.ival = OP_ORASSIGN;
491 else if (toketype == DORDOR)
492 pl_yylval.ival = OP_DORASSIGN;
500 * When Perl expects an operator and finds something else, no_op
501 * prints the warning. It always prints "<something> found where
502 * operator expected. It prints "Missing semicolon on previous line?"
503 * if the surprise occurs at the start of the line. "do you need to
504 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
505 * where the compiler doesn't know if foo is a method call or a function.
506 * It prints "Missing operator before end of line" if there's nothing
507 * after the missing operator, or "... before <...>" if there is something
508 * after the missing operator.
512 S_no_op(pTHX_ const char *const what, char *s)
515 char * const oldbp = PL_bufptr;
516 const bool is_first = (PL_oldbufptr == PL_linestart);
518 PERL_ARGS_ASSERT_NO_OP;
524 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
525 if (ckWARN_d(WARN_SYNTAX)) {
527 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
528 "\t(Missing semicolon on previous line?)\n");
529 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
531 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %.*s?)\n",
536 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
541 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
549 * Complain about missing quote/regexp/heredoc terminator.
550 * If it's called with NULL then it cauterizes the line buffer.
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
557 S_missingterm(pTHX_ char *s)
563 char * const nl = strrchr(s,'\n');
567 else if (isCNTRL(PL_multi_close)) {
569 tmpbuf[1] = (char)toCTRL(PL_multi_close);
574 *tmpbuf = (char)PL_multi_close;
578 q = strchr(s,'"') ? '\'' : '"';
579 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
582 #define FEATURE_IS_ENABLED(name) \
583 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
584 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
585 /* The longest string we pass in. */
586 #define MAX_FEATURE_LEN (sizeof("unicode_strings")-1)
589 * S_feature_is_enabled
590 * Check whether the named feature is enabled.
593 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
596 HV * const hinthv = GvHV(PL_hintgv);
597 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
599 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
601 assert(namelen <= MAX_FEATURE_LEN);
602 memcpy(&he_name[8], name, namelen);
604 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
608 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
609 * utf16-to-utf8-reversed.
612 #ifdef PERL_CR_FILTER
616 register const char *s = SvPVX_const(sv);
617 register const char * const e = s + SvCUR(sv);
619 PERL_ARGS_ASSERT_STRIP_RETURN;
621 /* outer loop optimized to do nothing if there are no CR-LFs */
623 if (*s++ == '\r' && *s == '\n') {
624 /* hit a CR-LF, need to copy the rest */
625 register char *d = s - 1;
628 if (*s == '\r' && s[1] == '\n')
639 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
642 if (count > 0 && !maxlen)
653 * Create a parser object and initialise its parser and lexer fields
655 * rsfp is the opened file handle to read from (if any),
657 * line holds any initial content already read from the file (or in
658 * the case of no file, such as an eval, the whole contents);
660 * new_filter indicates that this is a new file and it shouldn't inherit
661 * the filters from the current parser (ie require).
665 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
668 const char *s = NULL;
670 yy_parser *parser, *oparser;
672 /* create and initialise a parser */
674 Newxz(parser, 1, yy_parser);
675 parser->old_parser = oparser = PL_parser;
678 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
679 parser->ps = parser->stack;
680 parser->stack_size = YYINITDEPTH;
682 parser->stack->state = 0;
683 parser->yyerrstatus = 0;
684 parser->yychar = YYEMPTY; /* Cause a token to be read. */
686 /* on scope exit, free this parser and restore any outer one */
688 parser->saved_curcop = PL_curcop;
690 /* initialise lexer state */
693 parser->curforce = -1;
695 parser->nexttoke = 0;
697 parser->error_count = oparser ? oparser->error_count : 0;
698 parser->copline = NOLINE;
699 parser->lex_state = LEX_NORMAL;
700 parser->expect = XSTATE;
702 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
703 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
705 Newx(parser->lex_brackstack, 120, char);
706 Newx(parser->lex_casestack, 12, char);
707 *parser->lex_casestack = '\0';
710 s = SvPV_const(line, len);
716 parser->linestr = newSVpvs("\n;");
717 } else if (SvREADONLY(line) || s[len-1] != ';') {
718 parser->linestr = newSVsv(line);
720 sv_catpvs(parser->linestr, "\n;");
723 SvREFCNT_inc_simple_void_NN(line);
724 parser->linestr = line;
726 parser->oldoldbufptr =
729 parser->linestart = SvPVX(parser->linestr);
730 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
731 parser->last_lop = parser->last_uni = NULL;
735 /* delete a parser object */
738 Perl_parser_free(pTHX_ const yy_parser *parser)
740 PERL_ARGS_ASSERT_PARSER_FREE;
742 PL_curcop = parser->saved_curcop;
743 SvREFCNT_dec(parser->linestr);
745 if (parser->rsfp == PerlIO_stdin())
746 PerlIO_clearerr(parser->rsfp);
747 else if (parser->rsfp && (!parser->old_parser ||
748 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
749 PerlIO_close(parser->rsfp);
750 SvREFCNT_dec(parser->rsfp_filters);
752 Safefree(parser->stack);
753 Safefree(parser->lex_brackstack);
754 Safefree(parser->lex_casestack);
755 PL_parser = parser->old_parser;
762 * Finalizer for lexing operations. Must be called when the parser is
763 * done with the lexer.
770 PL_doextract = FALSE;
774 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
776 Buffer scalar containing the chunk currently under consideration of the
777 text currently being lexed. This is always a plain string scalar (for
778 which C<SvPOK> is true). It is not intended to be used as a scalar by
779 normal scalar means; instead refer to the buffer directly by the pointer
780 variables described below.
782 The lexer maintains various C<char*> pointers to things in the
783 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
784 reallocated, all of these pointers must be updated. Don't attempt to
785 do this manually, but rather use L</lex_grow_linestr> if you need to
786 reallocate the buffer.
788 The content of the text chunk in the buffer is commonly exactly one
789 complete line of input, up to and including a newline terminator,
790 but there are situations where it is otherwise. The octets of the
791 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
792 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
793 flag on this scalar, which may disagree with it.
795 For direct examination of the buffer, the variable
796 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
797 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
798 of these pointers is usually preferable to examination of the scalar
799 through normal scalar means.
801 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
803 Direct pointer to the end of the chunk of text currently being lexed, the
804 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
805 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
806 always located at the end of the buffer, and does not count as part of
807 the buffer's contents.
809 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
811 Points to the current position of lexing inside the lexer buffer.
812 Characters around this point may be freely examined, within
813 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
814 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
815 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
817 Lexing code (whether in the Perl core or not) moves this pointer past
818 the characters that it consumes. It is also expected to perform some
819 bookkeeping whenever a newline character is consumed. This movement
820 can be more conveniently performed by the function L</lex_read_to>,
821 which handles newlines appropriately.
823 Interpretation of the buffer's octets can be abstracted out by
824 using the slightly higher-level functions L</lex_peek_unichar> and
825 L</lex_read_unichar>.
827 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
829 Points to the start of the current line inside the lexer buffer.
830 This is useful for indicating at which column an error occurred, and
831 not much else. This must be updated by any lexing code that consumes
832 a newline; the function L</lex_read_to> handles this detail.
838 =for apidoc Amx|bool|lex_bufutf8
840 Indicates whether the octets in the lexer buffer
841 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
842 of Unicode characters. If not, they should be interpreted as Latin-1
843 characters. This is analogous to the C<SvUTF8> flag for scalars.
845 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
846 contains valid UTF-8. Lexing code must be robust in the face of invalid
849 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
850 is significant, but not the whole story regarding the input character
851 encoding. Normally, when a file is being read, the scalar contains octets
852 and its C<SvUTF8> flag is off, but the octets should be interpreted as
853 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
854 however, the scalar may have the C<SvUTF8> flag on, and in this case its
855 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
856 is in effect. This logic may change in the future; use this function
857 instead of implementing the logic yourself.
863 Perl_lex_bufutf8(pTHX)
869 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
871 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
872 at least I<len> octets (including terminating NUL). Returns a
873 pointer to the reallocated buffer. This is necessary before making
874 any direct modification of the buffer that would increase its length.
875 L</lex_stuff_pvn> provides a more convenient way to insert text into
878 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
879 this function updates all of the lexer's variables that point directly
886 Perl_lex_grow_linestr(pTHX_ STRLEN len)
890 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
891 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
892 linestr = PL_parser->linestr;
893 buf = SvPVX(linestr);
894 if (len <= SvLEN(linestr))
896 bufend_pos = PL_parser->bufend - buf;
897 bufptr_pos = PL_parser->bufptr - buf;
898 oldbufptr_pos = PL_parser->oldbufptr - buf;
899 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
900 linestart_pos = PL_parser->linestart - buf;
901 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
902 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
903 buf = sv_grow(linestr, len);
904 PL_parser->bufend = buf + bufend_pos;
905 PL_parser->bufptr = buf + bufptr_pos;
906 PL_parser->oldbufptr = buf + oldbufptr_pos;
907 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
908 PL_parser->linestart = buf + linestart_pos;
909 if (PL_parser->last_uni)
910 PL_parser->last_uni = buf + last_uni_pos;
911 if (PL_parser->last_lop)
912 PL_parser->last_lop = buf + last_lop_pos;
917 =for apidoc Amx|void|lex_stuff_pvn|char *pv|STRLEN len|U32 flags
919 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
920 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
921 reallocating the buffer if necessary. This means that lexing code that
922 runs later will see the characters as if they had appeared in the input.
923 It is not recommended to do this as part of normal parsing, and most
924 uses of this facility run the risk of the inserted characters being
925 interpreted in an unintended manner.
927 The string to be inserted is represented by I<len> octets starting
928 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
929 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
930 The characters are recoded for the lexer buffer, according to how the
931 buffer is currently being interpreted (L</lex_bufutf8>). If a string
932 to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
933 function is more convenient.
939 Perl_lex_stuff_pvn(pTHX_ char *pv, STRLEN len, U32 flags)
942 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
943 if (flags & ~(LEX_STUFF_UTF8))
944 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
946 if (flags & LEX_STUFF_UTF8) {
950 char *p, *e = pv+len;
951 for (p = pv; p != e; p++)
952 highhalf += !!(((U8)*p) & 0x80);
955 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
956 bufptr = PL_parser->bufptr;
957 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
958 PL_parser->bufend += len+highhalf;
959 for (p = pv; p != e; p++) {
962 *bufptr++ = (char)(0xc0 | (c >> 6));
963 *bufptr++ = (char)(0x80 | (c & 0x3f));
970 if (flags & LEX_STUFF_UTF8) {
972 char *p, *e = pv+len;
973 for (p = pv; p != e; p++) {
976 Perl_croak(aTHX_ "Lexing code attempted to stuff "
977 "non-Latin-1 character into Latin-1 input");
978 } else if (c >= 0xc2 && p+1 != e &&
979 (((U8)p[1]) & 0xc0) == 0x80) {
982 } else if (c >= 0x80) {
983 /* malformed UTF-8 */
985 SAVESPTR(PL_warnhook);
986 PL_warnhook = PERL_WARNHOOK_FATAL;
987 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
993 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
994 bufptr = PL_parser->bufptr;
995 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
996 PL_parser->bufend += len-highhalf;
997 for (p = pv; p != e; p++) {
1000 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1003 *bufptr++ = (char)c;
1008 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1009 bufptr = PL_parser->bufptr;
1010 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1011 PL_parser->bufend += len;
1012 Copy(pv, bufptr, len, char);
1018 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1020 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1021 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1022 reallocating the buffer if necessary. This means that lexing code that
1023 runs later will see the characters as if they had appeared in the input.
1024 It is not recommended to do this as part of normal parsing, and most
1025 uses of this facility run the risk of the inserted characters being
1026 interpreted in an unintended manner.
1028 The string to be inserted is the string value of I<sv>. The characters
1029 are recoded for the lexer buffer, according to how the buffer is currently
1030 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1031 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1032 need to construct a scalar.
1038 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1042 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1044 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1046 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1050 =for apidoc Amx|void|lex_unstuff|char *ptr
1052 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1053 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1054 This hides the discarded text from any lexing code that runs later,
1055 as if the text had never appeared.
1057 This is not the normal way to consume lexed text. For that, use
1064 Perl_lex_unstuff(pTHX_ char *ptr)
1068 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1069 buf = PL_parser->bufptr;
1071 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1074 bufend = PL_parser->bufend;
1076 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1077 unstuff_len = ptr - buf;
1078 Move(ptr, buf, bufend+1-ptr, char);
1079 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1080 PL_parser->bufend = bufend - unstuff_len;
1084 =for apidoc Amx|void|lex_read_to|char *ptr
1086 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1087 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1088 performing the correct bookkeeping whenever a newline character is passed.
1089 This is the normal way to consume lexed text.
1091 Interpretation of the buffer's octets can be abstracted out by
1092 using the slightly higher-level functions L</lex_peek_unichar> and
1093 L</lex_read_unichar>.
1099 Perl_lex_read_to(pTHX_ char *ptr)
1102 PERL_ARGS_ASSERT_LEX_READ_TO;
1103 s = PL_parser->bufptr;
1104 if (ptr < s || ptr > PL_parser->bufend)
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1106 for (; s != ptr; s++)
1108 CopLINE_inc(PL_curcop);
1109 PL_parser->linestart = s+1;
1111 PL_parser->bufptr = ptr;
1115 =for apidoc Amx|void|lex_discard_to|char *ptr
1117 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1118 up to I<ptr>. The remaining content of the buffer will be moved, and
1119 all pointers into the buffer updated appropriately. I<ptr> must not
1120 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1121 it is not permitted to discard text that has yet to be lexed.
1123 Normally it is not necessarily to do this directly, because it suffices to
1124 use the implicit discarding behaviour of L</lex_next_chunk> and things
1125 based on it. However, if a token stretches across multiple lines,
1126 and the lexing code has kept multiple lines of text in the buffer fof
1127 that purpose, then after completion of the token it would be wise to
1128 explicitly discard the now-unneeded earlier lines, to avoid future
1129 multi-line tokens growing the buffer without bound.
1135 Perl_lex_discard_to(pTHX_ char *ptr)
1139 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1140 buf = SvPVX(PL_parser->linestr);
1142 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1145 if (ptr > PL_parser->bufptr)
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1147 discard_len = ptr - buf;
1148 if (PL_parser->oldbufptr < ptr)
1149 PL_parser->oldbufptr = ptr;
1150 if (PL_parser->oldoldbufptr < ptr)
1151 PL_parser->oldoldbufptr = ptr;
1152 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1153 PL_parser->last_uni = NULL;
1154 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1155 PL_parser->last_lop = NULL;
1156 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1157 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1158 PL_parser->bufend -= discard_len;
1159 PL_parser->bufptr -= discard_len;
1160 PL_parser->oldbufptr -= discard_len;
1161 PL_parser->oldoldbufptr -= discard_len;
1162 if (PL_parser->last_uni)
1163 PL_parser->last_uni -= discard_len;
1164 if (PL_parser->last_lop)
1165 PL_parser->last_lop -= discard_len;
1169 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1171 Reads in the next chunk of text to be lexed, appending it to
1172 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1173 looked to the end of the current chunk and wants to know more. It is
1174 usual, but not necessary, for lexing to have consumed the entirety of
1175 the current chunk at this time.
1177 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1178 chunk (i.e., the current chunk has been entirely consumed), normally the
1179 current chunk will be discarded at the same time that the new chunk is
1180 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1181 will not be discarded. If the current chunk has not been entirely
1182 consumed, then it will not be discarded regardless of the flag.
1184 Returns true if some new text was added to the buffer, or false if the
1185 buffer has reached the end of the input text.
1190 #define LEX_FAKE_EOF 0x80000000
1193 Perl_lex_next_chunk(pTHX_ U32 flags)
1197 STRLEN old_bufend_pos, new_bufend_pos;
1198 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1199 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1200 bool got_some_for_debugger = 0;
1202 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1203 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1204 linestr = PL_parser->linestr;
1205 buf = SvPVX(linestr);
1206 if (!(flags & LEX_KEEP_PREVIOUS) &&
1207 PL_parser->bufptr == PL_parser->bufend) {
1208 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1210 if (PL_parser->last_uni != PL_parser->bufend)
1211 PL_parser->last_uni = NULL;
1212 if (PL_parser->last_lop != PL_parser->bufend)
1213 PL_parser->last_lop = NULL;
1214 last_uni_pos = last_lop_pos = 0;
1218 old_bufend_pos = PL_parser->bufend - buf;
1219 bufptr_pos = PL_parser->bufptr - buf;
1220 oldbufptr_pos = PL_parser->oldbufptr - buf;
1221 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1222 linestart_pos = PL_parser->linestart - buf;
1223 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1224 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1226 if (flags & LEX_FAKE_EOF) {
1228 } else if (!PL_parser->rsfp) {
1230 } else if (filter_gets(linestr, old_bufend_pos)) {
1232 got_some_for_debugger = 1;
1234 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1235 sv_setpvs(linestr, "");
1237 /* End of real input. Close filehandle (unless it was STDIN),
1238 * then add implicit termination.
1240 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1241 PerlIO_clearerr(PL_parser->rsfp);
1242 else if (PL_parser->rsfp)
1243 (void)PerlIO_close(PL_parser->rsfp);
1244 PL_parser->rsfp = NULL;
1245 PL_doextract = FALSE;
1247 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1250 if (!PL_in_eval && PL_minus_p) {
1252 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1253 PL_minus_n = PL_minus_p = 0;
1254 } else if (!PL_in_eval && PL_minus_n) {
1255 sv_catpvs(linestr, /*{*/";}");
1258 sv_catpvs(linestr, ";");
1261 buf = SvPVX(linestr);
1262 new_bufend_pos = SvCUR(linestr);
1263 PL_parser->bufend = buf + new_bufend_pos;
1264 PL_parser->bufptr = buf + bufptr_pos;
1265 PL_parser->oldbufptr = buf + oldbufptr_pos;
1266 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1267 PL_parser->linestart = buf + linestart_pos;
1268 if (PL_parser->last_uni)
1269 PL_parser->last_uni = buf + last_uni_pos;
1270 if (PL_parser->last_lop)
1271 PL_parser->last_lop = buf + last_lop_pos;
1272 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1273 PL_curstash != PL_debstash) {
1274 /* debugger active and we're not compiling the debugger code,
1275 * so store the line into the debugger's array of lines
1277 update_debugger_info(NULL, buf+old_bufend_pos,
1278 new_bufend_pos-old_bufend_pos);
1284 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1286 Looks ahead one (Unicode) character in the text currently being lexed.
1287 Returns the codepoint (unsigned integer value) of the next character,
1288 or -1 if lexing has reached the end of the input text. To consume the
1289 peeked character, use L</lex_read_unichar>.
1291 If the next character is in (or extends into) the next chunk of input
1292 text, the next chunk will be read in. Normally the current chunk will be
1293 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1294 then the current chunk will not be discarded.
1296 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1297 is encountered, an exception is generated.
1303 Perl_lex_peek_unichar(pTHX_ U32 flags)
1306 if (flags & ~(LEX_KEEP_PREVIOUS))
1307 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1308 s = PL_parser->bufptr;
1309 bufend = PL_parser->bufend;
1315 if (!lex_next_chunk(flags))
1317 s = PL_parser->bufptr;
1318 bufend = PL_parser->bufend;
1324 len = PL_utf8skip[head];
1325 while ((STRLEN)(bufend-s) < len) {
1326 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1328 s = PL_parser->bufptr;
1329 bufend = PL_parser->bufend;
1332 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1333 if (retlen == (STRLEN)-1) {
1334 /* malformed UTF-8 */
1336 SAVESPTR(PL_warnhook);
1337 PL_warnhook = PERL_WARNHOOK_FATAL;
1338 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1344 if (!lex_next_chunk(flags))
1346 s = PL_parser->bufptr;
1353 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1355 Reads the next (Unicode) character in the text currently being lexed.
1356 Returns the codepoint (unsigned integer value) of the character read,
1357 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1358 if lexing has reached the end of the input text. To non-destructively
1359 examine the next character, use L</lex_peek_unichar> instead.
1361 If the next character is in (or extends into) the next chunk of input
1362 text, the next chunk will be read in. Normally the current chunk will be
1363 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1364 then the current chunk will not be discarded.
1366 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1367 is encountered, an exception is generated.
1373 Perl_lex_read_unichar(pTHX_ U32 flags)
1376 if (flags & ~(LEX_KEEP_PREVIOUS))
1377 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1378 c = lex_peek_unichar(flags);
1381 CopLINE_inc(PL_curcop);
1382 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1388 =for apidoc Amx|void|lex_read_space|U32 flags
1390 Reads optional spaces, in Perl style, in the text currently being
1391 lexed. The spaces may include ordinary whitespace characters and
1392 Perl-style comments. C<#line> directives are processed if encountered.
1393 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1394 at a non-space character (or the end of the input text).
1396 If spaces extend into the next chunk of input text, the next chunk will
1397 be read in. Normally the current chunk will be discarded at the same
1398 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1399 chunk will not be discarded.
1404 #define LEX_NO_NEXT_CHUNK 0x80000000
1407 Perl_lex_read_space(pTHX_ U32 flags)
1410 bool need_incline = 0;
1411 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1412 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1415 sv_free(PL_skipwhite);
1416 PL_skipwhite = NULL;
1419 PL_skipwhite = newSVpvs("");
1420 #endif /* PERL_MAD */
1421 s = PL_parser->bufptr;
1422 bufend = PL_parser->bufend;
1428 } while (!(c == '\n' || (c == 0 && s == bufend)));
1429 } else if (c == '\n') {
1431 PL_parser->linestart = s;
1436 } else if (isSPACE(c)) {
1438 } else if (c == 0 && s == bufend) {
1442 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1443 #endif /* PERL_MAD */
1444 if (flags & LEX_NO_NEXT_CHUNK)
1446 PL_parser->bufptr = s;
1447 CopLINE_inc(PL_curcop);
1448 got_more = lex_next_chunk(flags);
1449 CopLINE_dec(PL_curcop);
1450 s = PL_parser->bufptr;
1451 bufend = PL_parser->bufend;
1454 if (need_incline && PL_parser->rsfp) {
1464 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1465 #endif /* PERL_MAD */
1466 PL_parser->bufptr = s;
1471 * This subroutine has nothing to do with tilting, whether at windmills
1472 * or pinball tables. Its name is short for "increment line". It
1473 * increments the current line number in CopLINE(PL_curcop) and checks
1474 * to see whether the line starts with a comment of the form
1475 * # line 500 "foo.pm"
1476 * If so, it sets the current line number and file to the values in the comment.
1480 S_incline(pTHX_ const char *s)
1487 PERL_ARGS_ASSERT_INCLINE;
1489 CopLINE_inc(PL_curcop);
1492 while (SPACE_OR_TAB(*s))
1494 if (strnEQ(s, "line", 4))
1498 if (SPACE_OR_TAB(*s))
1502 while (SPACE_OR_TAB(*s))
1510 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1512 while (SPACE_OR_TAB(*s))
1514 if (*s == '"' && (t = strchr(s+1, '"'))) {
1520 while (!isSPACE(*t))
1524 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1526 if (*e != '\n' && *e != '\0')
1527 return; /* false alarm */
1530 const STRLEN len = t - s;
1531 #ifndef USE_ITHREADS
1532 SV *const temp_sv = CopFILESV(PL_curcop);
1537 cf = SvPVX(temp_sv);
1538 tmplen = SvCUR(temp_sv);
1544 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1545 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1546 * to *{"::_<newfilename"} */
1547 /* However, the long form of evals is only turned on by the
1548 debugger - usually they're "(eval %lu)" */
1552 STRLEN tmplen2 = len;
1553 if (tmplen + 2 <= sizeof smallbuf)
1556 Newx(tmpbuf, tmplen + 2, char);
1559 memcpy(tmpbuf + 2, cf, tmplen);
1561 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1566 if (tmplen2 + 2 <= sizeof smallbuf)
1569 Newx(tmpbuf2, tmplen2 + 2, char);
1571 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1572 /* Either they malloc'd it, or we malloc'd it,
1573 so no prefix is present in ours. */
1578 memcpy(tmpbuf2 + 2, s, tmplen2);
1581 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1583 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1584 /* adjust ${"::_<newfilename"} to store the new file name */
1585 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1586 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1587 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1590 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1592 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1595 CopFILE_free(PL_curcop);
1596 CopFILE_setn(PL_curcop, s, len);
1598 CopLINE_set(PL_curcop, atoi(n)-1);
1602 /* skip space before PL_thistoken */
1605 S_skipspace0(pTHX_ register char *s)
1607 PERL_ARGS_ASSERT_SKIPSPACE0;
1614 PL_thiswhite = newSVpvs("");
1615 sv_catsv(PL_thiswhite, PL_skipwhite);
1616 sv_free(PL_skipwhite);
1619 PL_realtokenstart = s - SvPVX(PL_linestr);
1623 /* skip space after PL_thistoken */
1626 S_skipspace1(pTHX_ register char *s)
1628 const char *start = s;
1629 I32 startoff = start - SvPVX(PL_linestr);
1631 PERL_ARGS_ASSERT_SKIPSPACE1;
1636 start = SvPVX(PL_linestr) + startoff;
1637 if (!PL_thistoken && PL_realtokenstart >= 0) {
1638 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1639 PL_thistoken = newSVpvn(tstart, start - tstart);
1641 PL_realtokenstart = -1;
1644 PL_nextwhite = newSVpvs("");
1645 sv_catsv(PL_nextwhite, PL_skipwhite);
1646 sv_free(PL_skipwhite);
1653 S_skipspace2(pTHX_ register char *s, SV **svp)
1656 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1657 const I32 startoff = s - SvPVX(PL_linestr);
1659 PERL_ARGS_ASSERT_SKIPSPACE2;
1662 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1663 if (!PL_madskills || !svp)
1665 start = SvPVX(PL_linestr) + startoff;
1666 if (!PL_thistoken && PL_realtokenstart >= 0) {
1667 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1668 PL_thistoken = newSVpvn(tstart, start - tstart);
1669 PL_realtokenstart = -1;
1673 *svp = newSVpvs("");
1674 sv_setsv(*svp, PL_skipwhite);
1675 sv_free(PL_skipwhite);
1684 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1686 AV *av = CopFILEAVx(PL_curcop);
1688 SV * const sv = newSV_type(SVt_PVMG);
1690 sv_setsv(sv, orig_sv);
1692 sv_setpvn(sv, buf, len);
1695 av_store(av, (I32)CopLINE(PL_curcop), sv);
1701 * Called to gobble the appropriate amount and type of whitespace.
1702 * Skips comments as well.
1706 S_skipspace(pTHX_ register char *s)
1710 #endif /* PERL_MAD */
1711 PERL_ARGS_ASSERT_SKIPSPACE;
1714 sv_free(PL_skipwhite);
1715 PL_skipwhite = NULL;
1717 #endif /* PERL_MAD */
1718 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1719 while (s < PL_bufend && SPACE_OR_TAB(*s))
1722 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1724 lex_read_space(LEX_KEEP_PREVIOUS |
1725 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1726 LEX_NO_NEXT_CHUNK : 0));
1728 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1729 if (PL_linestart > PL_bufptr)
1730 PL_bufptr = PL_linestart;
1735 PL_skipwhite = newSVpvn(start, s-start);
1736 #endif /* PERL_MAD */
1742 * Check the unary operators to ensure there's no ambiguity in how they're
1743 * used. An ambiguous piece of code would be:
1745 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1746 * the +5 is its argument.
1756 if (PL_oldoldbufptr != PL_last_uni)
1758 while (isSPACE(*PL_last_uni))
1761 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1763 if ((t = strchr(s, '(')) && t < PL_bufptr)
1766 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1767 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1768 (int)(s - PL_last_uni), PL_last_uni);
1772 * LOP : macro to build a list operator. Its behaviour has been replaced
1773 * with a subroutine, S_lop() for which LOP is just another name.
1776 #define LOP(f,x) return lop(f,x,s)
1780 * Build a list operator (or something that might be one). The rules:
1781 * - if we have a next token, then it's a list operator [why?]
1782 * - if the next thing is an opening paren, then it's a function
1783 * - else it's a list operator
1787 S_lop(pTHX_ I32 f, int x, char *s)
1791 PERL_ARGS_ASSERT_LOP;
1797 PL_last_lop = PL_oldbufptr;
1798 PL_last_lop_op = (OPCODE)f;
1801 return REPORT(LSTOP);
1804 return REPORT(LSTOP);
1807 return REPORT(FUNC);
1810 return REPORT(FUNC);
1812 return REPORT(LSTOP);
1818 * Sets up for an eventual force_next(). start_force(0) basically does
1819 * an unshift, while start_force(-1) does a push. yylex removes items
1824 S_start_force(pTHX_ int where)
1828 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1829 where = PL_lasttoke;
1830 assert(PL_curforce < 0 || PL_curforce == where);
1831 if (PL_curforce != where) {
1832 for (i = PL_lasttoke; i > where; --i) {
1833 PL_nexttoke[i] = PL_nexttoke[i-1];
1837 if (PL_curforce < 0) /* in case of duplicate start_force() */
1838 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1839 PL_curforce = where;
1842 curmad('^', newSVpvs(""));
1843 CURMAD('_', PL_nextwhite);
1848 S_curmad(pTHX_ char slot, SV *sv)
1854 if (PL_curforce < 0)
1855 where = &PL_thismad;
1857 where = &PL_nexttoke[PL_curforce].next_mad;
1863 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1865 else if (PL_encoding) {
1866 sv_recode_to_utf8(sv, PL_encoding);
1871 /* keep a slot open for the head of the list? */
1872 if (slot != '_' && *where && (*where)->mad_key == '^') {
1873 (*where)->mad_key = slot;
1874 sv_free(MUTABLE_SV(((*where)->mad_val)));
1875 (*where)->mad_val = (void*)sv;
1878 addmad(newMADsv(slot, sv), where, 0);
1881 # define start_force(where) NOOP
1882 # define curmad(slot, sv) NOOP
1887 * When the lexer realizes it knows the next token (for instance,
1888 * it is reordering tokens for the parser) then it can call S_force_next
1889 * to know what token to return the next time the lexer is called. Caller
1890 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1891 * and possibly PL_expect to ensure the lexer handles the token correctly.
1895 S_force_next(pTHX_ I32 type)
1900 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1901 tokereport(type, &NEXTVAL_NEXTTOKE);
1905 if (PL_curforce < 0)
1906 start_force(PL_lasttoke);
1907 PL_nexttoke[PL_curforce].next_type = type;
1908 if (PL_lex_state != LEX_KNOWNEXT)
1909 PL_lex_defer = PL_lex_state;
1910 PL_lex_state = LEX_KNOWNEXT;
1911 PL_lex_expect = PL_expect;
1914 PL_nexttype[PL_nexttoke] = type;
1916 if (PL_lex_state != LEX_KNOWNEXT) {
1917 PL_lex_defer = PL_lex_state;
1918 PL_lex_expect = PL_expect;
1919 PL_lex_state = LEX_KNOWNEXT;
1925 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1928 SV * const sv = newSVpvn_utf8(start, len,
1931 && !is_ascii_string((const U8*)start, len)
1932 && is_utf8_string((const U8*)start, len));
1938 * When the lexer knows the next thing is a word (for instance, it has
1939 * just seen -> and it knows that the next char is a word char, then
1940 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1944 * char *start : buffer position (must be within PL_linestr)
1945 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1946 * int check_keyword : if true, Perl checks to make sure the word isn't
1947 * a keyword (do this if the word is a label, e.g. goto FOO)
1948 * int allow_pack : if true, : characters will also be allowed (require,
1949 * use, etc. do this)
1950 * int allow_initial_tick : used by the "sub" lexer only.
1954 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1960 PERL_ARGS_ASSERT_FORCE_WORD;
1962 start = SKIPSPACE1(start);
1964 if (isIDFIRST_lazy_if(s,UTF) ||
1965 (allow_pack && *s == ':') ||
1966 (allow_initial_tick && *s == '\'') )
1968 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1969 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1971 start_force(PL_curforce);
1973 curmad('X', newSVpvn(start,s-start));
1974 if (token == METHOD) {
1979 PL_expect = XOPERATOR;
1983 curmad('g', newSVpvs( "forced" ));
1984 NEXTVAL_NEXTTOKE.opval
1985 = (OP*)newSVOP(OP_CONST,0,
1986 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1987 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1995 * Called when the lexer wants $foo *foo &foo etc, but the program
1996 * text only contains the "foo" portion. The first argument is a pointer
1997 * to the "foo", and the second argument is the type symbol to prefix.
1998 * Forces the next token to be a "WORD".
1999 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2003 S_force_ident(pTHX_ register const char *s, int kind)
2007 PERL_ARGS_ASSERT_FORCE_IDENT;
2010 const STRLEN len = strlen(s);
2011 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2012 start_force(PL_curforce);
2013 NEXTVAL_NEXTTOKE.opval = o;
2016 o->op_private = OPpCONST_ENTERED;
2017 /* XXX see note in pp_entereval() for why we forgo typo
2018 warnings if the symbol must be introduced in an eval.
2020 gv_fetchpvn_flags(s, len,
2021 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2023 kind == '$' ? SVt_PV :
2024 kind == '@' ? SVt_PVAV :
2025 kind == '%' ? SVt_PVHV :
2033 Perl_str_to_version(pTHX_ SV *sv)
2038 const char *start = SvPV_const(sv,len);
2039 const char * const end = start + len;
2040 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2042 PERL_ARGS_ASSERT_STR_TO_VERSION;
2044 while (start < end) {
2048 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2053 retval += ((NV)n)/nshift;
2062 * Forces the next token to be a version number.
2063 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2064 * and if "guessing" is TRUE, then no new token is created (and the caller
2065 * must use an alternative parsing method).
2069 S_force_version(pTHX_ char *s, int guessing)
2075 I32 startoff = s - SvPVX(PL_linestr);
2078 PERL_ARGS_ASSERT_FORCE_VERSION;
2086 while (isDIGIT(*d) || *d == '_' || *d == '.')
2090 start_force(PL_curforce);
2091 curmad('X', newSVpvn(s,d-s));
2094 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
2096 #ifdef USE_LOCALE_NUMERIC
2097 char *loc = setlocale(LC_NUMERIC, "C");
2099 s = scan_num(s, &pl_yylval);
2100 #ifdef USE_LOCALE_NUMERIC
2101 setlocale(LC_NUMERIC, loc);
2103 version = pl_yylval.opval;
2104 ver = cSVOPx(version)->op_sv;
2105 if (SvPOK(ver) && !SvNIOK(ver)) {
2106 SvUPGRADE(ver, SVt_PVNV);
2107 SvNV_set(ver, str_to_version(ver));
2108 SvNOK_on(ver); /* hint that it is a version */
2111 else if (guessing) {
2114 sv_free(PL_nextwhite); /* let next token collect whitespace */
2116 s = SvPVX(PL_linestr) + startoff;
2124 if (PL_madskills && !version) {
2125 sv_free(PL_nextwhite); /* let next token collect whitespace */
2127 s = SvPVX(PL_linestr) + startoff;
2130 /* NOTE: The parser sees the package name and the VERSION swapped */
2131 start_force(PL_curforce);
2132 NEXTVAL_NEXTTOKE.opval = version;
2139 * S_force_strict_version
2140 * Forces the next token to be a version number using strict syntax rules.
2144 S_force_strict_version(pTHX_ char *s)
2149 I32 startoff = s - SvPVX(PL_linestr);
2151 const char *errstr = NULL;
2153 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2155 while (isSPACE(*s)) /* leading whitespace */
2158 if (is_STRICT_VERSION(s,&errstr)) {
2160 s = (char *)scan_version(s, ver, 0);
2161 version = newSVOP(OP_CONST, 0, ver);
2163 else if ( (*s != ';' && *s != '}' ) && (s = SKIPSPACE1(s), (*s != ';' && *s !='}' ))) {
2166 yyerror(errstr); /* version required */
2171 if (PL_madskills && !version) {
2172 sv_free(PL_nextwhite); /* let next token collect whitespace */
2174 s = SvPVX(PL_linestr) + startoff;
2177 /* NOTE: The parser sees the package name and the VERSION swapped */
2178 start_force(PL_curforce);
2179 NEXTVAL_NEXTTOKE.opval = version;
2187 * Tokenize a quoted string passed in as an SV. It finds the next
2188 * chunk, up to end of string or a backslash. It may make a new
2189 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2194 S_tokeq(pTHX_ SV *sv)
2198 register char *send;
2203 PERL_ARGS_ASSERT_TOKEQ;
2208 s = SvPV_force(sv, len);
2209 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2212 while (s < send && *s != '\\')
2217 if ( PL_hints & HINT_NEW_STRING ) {
2218 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2222 if (s + 1 < send && (s[1] == '\\'))
2223 s++; /* all that, just for this */
2228 SvCUR_set(sv, d - SvPVX_const(sv));
2230 if ( PL_hints & HINT_NEW_STRING )
2231 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2236 * Now come three functions related to double-quote context,
2237 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2238 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2239 * interact with PL_lex_state, and create fake ( ... ) argument lists
2240 * to handle functions and concatenation.
2241 * They assume that whoever calls them will be setting up a fake
2242 * join call, because each subthing puts a ',' after it. This lets
2245 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2247 * (I'm not sure whether the spurious commas at the end of lcfirst's
2248 * arguments and join's arguments are created or not).
2253 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2255 * Pattern matching will set PL_lex_op to the pattern-matching op to
2256 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2258 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2260 * Everything else becomes a FUNC.
2262 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2263 * had an OP_CONST or OP_READLINE). This just sets us up for a
2264 * call to S_sublex_push().
2268 S_sublex_start(pTHX)
2271 register const I32 op_type = pl_yylval.ival;
2273 if (op_type == OP_NULL) {
2274 pl_yylval.opval = PL_lex_op;
2278 if (op_type == OP_CONST || op_type == OP_READLINE) {
2279 SV *sv = tokeq(PL_lex_stuff);
2281 if (SvTYPE(sv) == SVt_PVIV) {
2282 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2284 const char * const p = SvPV_const(sv, len);
2285 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2289 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2290 PL_lex_stuff = NULL;
2291 /* Allow <FH> // "foo" */
2292 if (op_type == OP_READLINE)
2293 PL_expect = XTERMORDORDOR;
2296 else if (op_type == OP_BACKTICK && PL_lex_op) {
2297 /* readpipe() vas overriden */
2298 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2299 pl_yylval.opval = PL_lex_op;
2301 PL_lex_stuff = NULL;
2305 PL_sublex_info.super_state = PL_lex_state;
2306 PL_sublex_info.sub_inwhat = (U16)op_type;
2307 PL_sublex_info.sub_op = PL_lex_op;
2308 PL_lex_state = LEX_INTERPPUSH;
2312 pl_yylval.opval = PL_lex_op;
2322 * Create a new scope to save the lexing state. The scope will be
2323 * ended in S_sublex_done. Returns a '(', starting the function arguments
2324 * to the uc, lc, etc. found before.
2325 * Sets PL_lex_state to LEX_INTERPCONCAT.
2334 PL_lex_state = PL_sublex_info.super_state;
2335 SAVEBOOL(PL_lex_dojoin);
2336 SAVEI32(PL_lex_brackets);
2337 SAVEI32(PL_lex_casemods);
2338 SAVEI32(PL_lex_starts);
2339 SAVEI8(PL_lex_state);
2340 SAVEVPTR(PL_lex_inpat);
2341 SAVEI16(PL_lex_inwhat);
2342 SAVECOPLINE(PL_curcop);
2343 SAVEPPTR(PL_bufptr);
2344 SAVEPPTR(PL_bufend);
2345 SAVEPPTR(PL_oldbufptr);
2346 SAVEPPTR(PL_oldoldbufptr);
2347 SAVEPPTR(PL_last_lop);
2348 SAVEPPTR(PL_last_uni);
2349 SAVEPPTR(PL_linestart);
2350 SAVESPTR(PL_linestr);
2351 SAVEGENERICPV(PL_lex_brackstack);
2352 SAVEGENERICPV(PL_lex_casestack);
2354 PL_linestr = PL_lex_stuff;
2355 PL_lex_stuff = NULL;
2357 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2358 = SvPVX(PL_linestr);
2359 PL_bufend += SvCUR(PL_linestr);
2360 PL_last_lop = PL_last_uni = NULL;
2361 SAVEFREESV(PL_linestr);
2363 PL_lex_dojoin = FALSE;
2364 PL_lex_brackets = 0;
2365 Newx(PL_lex_brackstack, 120, char);
2366 Newx(PL_lex_casestack, 12, char);
2367 PL_lex_casemods = 0;
2368 *PL_lex_casestack = '\0';
2370 PL_lex_state = LEX_INTERPCONCAT;
2371 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2373 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2374 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2375 PL_lex_inpat = PL_sublex_info.sub_op;
2377 PL_lex_inpat = NULL;
2384 * Restores lexer state after a S_sublex_push.
2391 if (!PL_lex_starts++) {
2392 SV * const sv = newSVpvs("");
2393 if (SvUTF8(PL_linestr))
2395 PL_expect = XOPERATOR;
2396 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2400 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2401 PL_lex_state = LEX_INTERPCASEMOD;
2405 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2406 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2407 PL_linestr = PL_lex_repl;
2409 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2410 PL_bufend += SvCUR(PL_linestr);
2411 PL_last_lop = PL_last_uni = NULL;
2412 SAVEFREESV(PL_linestr);
2413 PL_lex_dojoin = FALSE;
2414 PL_lex_brackets = 0;
2415 PL_lex_casemods = 0;
2416 *PL_lex_casestack = '\0';
2418 if (SvEVALED(PL_lex_repl)) {
2419 PL_lex_state = LEX_INTERPNORMAL;
2421 /* we don't clear PL_lex_repl here, so that we can check later
2422 whether this is an evalled subst; that means we rely on the
2423 logic to ensure sublex_done() is called again only via the
2424 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2427 PL_lex_state = LEX_INTERPCONCAT;
2437 PL_endwhite = newSVpvs("");
2438 sv_catsv(PL_endwhite, PL_thiswhite);
2442 sv_setpvs(PL_thistoken,"");
2444 PL_realtokenstart = -1;
2448 PL_bufend = SvPVX(PL_linestr);
2449 PL_bufend += SvCUR(PL_linestr);
2450 PL_expect = XOPERATOR;
2451 PL_sublex_info.sub_inwhat = 0;
2459 Extracts a pattern, double-quoted string, or transliteration. This
2462 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2463 processing a pattern (PL_lex_inpat is true), a transliteration
2464 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2466 Returns a pointer to the character scanned up to. If this is
2467 advanced from the start pointer supplied (i.e. if anything was
2468 successfully parsed), will leave an OP for the substring scanned
2469 in pl_yylval. Caller must intuit reason for not parsing further
2470 by looking at the next characters herself.
2474 constants: \N{NAME} only
2475 case and quoting: \U \Q \E
2476 stops on @ and $, but not for $ as tail anchor
2478 In transliterations:
2479 characters are VERY literal, except for - not at the start or end
2480 of the string, which indicates a range. If the range is in bytes,
2481 scan_const expands the range to the full set of intermediate
2482 characters. If the range is in utf8, the hyphen is replaced with
2483 a certain range mark which will be handled by pmtrans() in op.c.
2485 In double-quoted strings:
2487 double-quoted style: \r and \n
2488 constants: \x31, etc.
2489 deprecated backrefs: \1 (in substitution replacements)
2490 case and quoting: \U \Q \E
2493 scan_const does *not* construct ops to handle interpolated strings.
2494 It stops processing as soon as it finds an embedded $ or @ variable
2495 and leaves it to the caller to work out what's going on.
2497 embedded arrays (whether in pattern or not) could be:
2498 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2500 $ in double-quoted strings must be the symbol of an embedded scalar.
2502 $ in pattern could be $foo or could be tail anchor. Assumption:
2503 it's a tail anchor if $ is the last thing in the string, or if it's
2504 followed by one of "()| \r\n\t"
2506 \1 (backreferences) are turned into $1
2508 The structure of the code is
2509 while (there's a character to process) {
2510 handle transliteration ranges
2511 skip regexp comments /(?#comment)/ and codes /(?{code})/
2512 skip #-initiated comments in //x patterns
2513 check for embedded arrays
2514 check for embedded scalars
2516 deprecate \1 in substitution replacements
2517 handle string-changing backslashes \l \U \Q \E, etc.
2518 switch (what was escaped) {
2519 handle \- in a transliteration (becomes a literal -)
2520 if a pattern and not \N{, go treat as regular character
2521 handle \132 (octal characters)
2522 handle \x15 and \x{1234} (hex characters)
2523 handle \N{name} (named characters, also \N{3,5} in a pattern)
2524 handle \cV (control characters)
2525 handle printf-style backslashes (\f, \r, \n, etc)
2528 } (end if backslash)
2529 handle regular character
2530 } (end while character to read)
2535 S_scan_const(pTHX_ char *start)
2538 register char *send = PL_bufend; /* end of the constant */
2539 SV *sv = newSV(send - start); /* sv for the constant. See
2540 note below on sizing. */
2541 register char *s = start; /* start of the constant */
2542 register char *d = SvPVX(sv); /* destination for copies */
2543 bool dorange = FALSE; /* are we in a translit range? */
2544 bool didrange = FALSE; /* did we just finish a range? */
2545 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2546 I32 this_utf8 = UTF; /* Is the source string assumed
2547 to be UTF8? But, this can
2548 show as true when the source
2549 isn't utf8, as for example
2550 when it is entirely composed
2553 /* Note on sizing: The scanned constant is placed into sv, which is
2554 * initialized by newSV() assuming one byte of output for every byte of
2555 * input. This routine expects newSV() to allocate an extra byte for a
2556 * trailing NUL, which this routine will append if it gets to the end of
2557 * the input. There may be more bytes of input than output (eg., \N{LATIN
2558 * CAPITAL LETTER A}), or more output than input if the constant ends up
2559 * recoded to utf8, but each time a construct is found that might increase
2560 * the needed size, SvGROW() is called. Its size parameter each time is
2561 * based on the best guess estimate at the time, namely the length used so
2562 * far, plus the length the current construct will occupy, plus room for
2563 * the trailing NUL, plus one byte for every input byte still unscanned */
2567 UV literal_endpoint = 0;
2568 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2571 PERL_ARGS_ASSERT_SCAN_CONST;
2573 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2574 /* If we are doing a trans and we know we want UTF8 set expectation */
2575 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2576 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2580 while (s < send || dorange) {
2582 /* get transliterations out of the way (they're most literal) */
2583 if (PL_lex_inwhat == OP_TRANS) {
2584 /* expand a range A-Z to the full set of characters. AIE! */
2586 I32 i; /* current expanded character */
2587 I32 min; /* first character in range */
2588 I32 max; /* last character in range */
2599 char * const c = (char*)utf8_hop((U8*)d, -1);
2603 *c = (char)UTF_TO_NATIVE(0xff);
2604 /* mark the range as done, and continue */
2610 i = d - SvPVX_const(sv); /* remember current offset */
2613 SvLEN(sv) + (has_utf8 ?
2614 (512 - UTF_CONTINUATION_MARK +
2617 /* How many two-byte within 0..255: 128 in UTF-8,
2618 * 96 in UTF-8-mod. */
2620 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2622 d = SvPVX(sv) + i; /* refresh d after realloc */
2626 for (j = 0; j <= 1; j++) {
2627 char * const c = (char*)utf8_hop((U8*)d, -1);
2628 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2634 max = (U8)0xff; /* only to \xff */
2635 uvmax = uv; /* \x{100} to uvmax */
2637 d = c; /* eat endpoint chars */
2642 d -= 2; /* eat the first char and the - */
2643 min = (U8)*d; /* first char in range */
2644 max = (U8)d[1]; /* last char in range */
2651 "Invalid range \"%c-%c\" in transliteration operator",
2652 (char)min, (char)max);
2656 if (literal_endpoint == 2 &&
2657 ((isLOWER(min) && isLOWER(max)) ||
2658 (isUPPER(min) && isUPPER(max)))) {
2660 for (i = min; i <= max; i++)
2662 *d++ = NATIVE_TO_NEED(has_utf8,i);
2664 for (i = min; i <= max; i++)
2666 *d++ = NATIVE_TO_NEED(has_utf8,i);
2671 for (i = min; i <= max; i++)
2674 const U8 ch = (U8)NATIVE_TO_UTF(i);
2675 if (UNI_IS_INVARIANT(ch))
2678 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2679 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2688 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2690 *d++ = (char)UTF_TO_NATIVE(0xff);
2692 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2696 /* mark the range as done, and continue */
2700 literal_endpoint = 0;
2705 /* range begins (ignore - as first or last char) */
2706 else if (*s == '-' && s+1 < send && s != start) {
2708 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2715 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2725 literal_endpoint = 0;
2726 native_range = TRUE;
2731 /* if we get here, we're not doing a transliteration */
2733 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2734 except for the last char, which will be done separately. */
2735 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2737 while (s+1 < send && *s != ')')
2738 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2740 else if (s[2] == '{' /* This should match regcomp.c */
2741 || (s[2] == '?' && s[3] == '{'))
2744 char *regparse = s + (s[2] == '{' ? 3 : 4);
2747 while (count && (c = *regparse)) {
2748 if (c == '\\' && regparse[1])
2756 if (*regparse != ')')
2757 regparse--; /* Leave one char for continuation. */
2758 while (s < regparse)
2759 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2763 /* likewise skip #-initiated comments in //x patterns */
2764 else if (*s == '#' && PL_lex_inpat &&
2765 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2766 while (s+1 < send && *s != '\n')
2767 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2770 /* check for embedded arrays
2771 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2773 else if (*s == '@' && s[1]) {
2774 if (isALNUM_lazy_if(s+1,UTF))
2776 if (strchr(":'{$", s[1]))
2778 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2779 break; /* in regexp, neither @+ nor @- are interpolated */
2782 /* check for embedded scalars. only stop if we're sure it's a
2785 else if (*s == '$') {
2786 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2788 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2790 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2791 "Possible unintended interpolation of $\\ in regex");
2793 break; /* in regexp, $ might be tail anchor */
2797 /* End of else if chain - OP_TRANS rejoin rest */
2800 if (*s == '\\' && s+1 < send) {
2801 char* e; /* Can be used for ending '}', etc. */
2805 /* deprecate \1 in strings and substitution replacements */
2806 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2807 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2809 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2814 /* string-change backslash escapes */
2815 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2819 /* In a pattern, process \N, but skip any other backslash escapes.
2820 * This is because we don't want to translate an escape sequence
2821 * into a meta symbol and have the regex compiler use the meta
2822 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2823 * in spite of this, we do have to process \N here while the proper
2824 * charnames handler is in scope. See bugs #56444 and #62056.
2825 * There is a complication because \N in a pattern may also stand
2826 * for 'match a non-nl', and not mean a charname, in which case its
2827 * processing should be deferred to the regex compiler. To be a
2828 * charname it must be followed immediately by a '{', and not look
2829 * like \N followed by a curly quantifier, i.e., not something like
2830 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2832 else if (PL_lex_inpat
2835 || regcurly(s + 1)))
2837 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2838 goto default_action;
2843 /* quoted - in transliterations */
2845 if (PL_lex_inwhat == OP_TRANS) {
2852 if ((isALPHA(*s) || isDIGIT(*s)))
2853 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2854 "Unrecognized escape \\%c passed through",
2856 /* default action is to copy the quoted character */
2857 goto default_action;
2860 /* eg. \132 indicates the octal constant 0x132 */
2861 case '0': case '1': case '2': case '3':
2862 case '4': case '5': case '6': case '7':
2866 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2869 goto NUM_ESCAPE_INSERT;
2871 /* eg. \x24 indicates the hex constant 0x24 */
2875 char* const e = strchr(s, '}');
2876 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2877 PERL_SCAN_DISALLOW_PREFIX;
2882 yyerror("Missing right brace on \\x{}");
2886 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2892 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2893 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2899 /* Insert oct or hex escaped character. There will always be
2900 * enough room in sv since such escapes will be longer than any
2901 * UTF-8 sequence they can end up as, except if they force us
2902 * to recode the rest of the string into utf8 */
2904 /* Here uv is the ordinal of the next character being added in
2905 * unicode (converted from native). */
2906 if (!UNI_IS_INVARIANT(uv)) {
2907 if (!has_utf8 && uv > 255) {
2908 /* Might need to recode whatever we have accumulated so
2909 * far if it contains any chars variant in utf8 or
2912 SvCUR_set(sv, d - SvPVX_const(sv));
2915 /* See Note on sizing above. */
2916 sv_utf8_upgrade_flags_grow(sv,
2917 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2918 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2919 d = SvPVX(sv) + SvCUR(sv);
2924 d = (char*)uvuni_to_utf8((U8*)d, uv);
2925 if (PL_lex_inwhat == OP_TRANS &&
2926 PL_sublex_info.sub_op) {
2927 PL_sublex_info.sub_op->op_private |=
2928 (PL_lex_repl ? OPpTRANS_FROM_UTF
2932 if (uv > 255 && !dorange)
2933 native_range = FALSE;
2946 /* In a non-pattern \N must be a named character, like \N{LATIN
2947 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2948 * mean to match a non-newline. For non-patterns, named
2949 * characters are converted to their string equivalents. In
2950 * patterns, named characters are not converted to their
2951 * ultimate forms for the same reasons that other escapes
2952 * aren't. Instead, they are converted to the \N{U+...} form
2953 * to get the value from the charnames that is in effect right
2954 * now, while preserving the fact that it was a named character
2955 * so that the regex compiler knows this */
2957 /* This section of code doesn't generally use the
2958 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2959 * a close examination of this macro and determined it is a
2960 * no-op except on utfebcdic variant characters. Every
2961 * character generated by this that would normally need to be
2962 * enclosed by this macro is invariant, so the macro is not
2963 * needed, and would complicate use of copy(). There are other
2964 * parts of this file where the macro is used inconsistently,
2965 * but are saved by it being a no-op */
2967 /* The structure of this section of code (besides checking for
2968 * errors and upgrading to utf8) is:
2969 * Further disambiguate between the two meanings of \N, and if
2970 * not a charname, go process it elsewhere
2971 * If of form \N{U+...}, pass it through if a pattern;
2972 * otherwise convert to utf8
2973 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
2974 * pattern; otherwise convert to utf8 */
2976 /* Here, s points to the 'N'; the test below is guaranteed to
2977 * succeed if we are being called on a pattern as we already
2978 * know from a test above that the next character is a '{'.
2979 * On a non-pattern \N must mean 'named sequence, which
2980 * requires braces */
2983 yyerror("Missing braces on \\N{}");
2988 /* If there is no matching '}', it is an error. */
2989 if (! (e = strchr(s, '}'))) {
2990 if (! PL_lex_inpat) {
2991 yyerror("Missing right brace on \\N{}");
2993 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
2998 /* Here it looks like a named character */
3002 /* XXX This block is temporary code. \N{} implies that the
3003 * pattern is to have Unicode semantics, and therefore
3004 * currently has to be encoded in utf8. By putting it in
3005 * utf8 now, we save a whole pass in the regular expression
3006 * compiler. Once that code is changed so Unicode
3007 * semantics doesn't necessarily have to be in utf8, this
3008 * block should be removed */
3010 SvCUR_set(sv, d - SvPVX_const(sv));
3013 /* See Note on sizing above. */
3014 sv_utf8_upgrade_flags_grow(sv,
3015 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3016 /* 5 = '\N{' + cur char + NUL */
3017 (STRLEN)(send - s) + 5);
3018 d = SvPVX(sv) + SvCUR(sv);
3023 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3024 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3025 | PERL_SCAN_DISALLOW_PREFIX;
3028 /* For \N{U+...}, the '...' is a unicode value even on
3029 * EBCDIC machines */
3030 s += 2; /* Skip to next char after the 'U+' */
3032 uv = grok_hex(s, &len, &flags, NULL);
3033 if (len == 0 || len != (STRLEN)(e - s)) {
3034 yyerror("Invalid hexadecimal number in \\N{U+...}");
3041 /* Pass through to the regex compiler unchanged. The
3042 * reason we evaluated the number above is to make sure
3043 * there wasn't a syntax error. */
3044 s -= 5; /* Include the '\N{U+' */
3045 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3048 else { /* Not a pattern: convert the hex to string */
3050 /* If destination is not in utf8, unconditionally
3051 * recode it to be so. This is because \N{} implies
3052 * Unicode semantics, and scalars have to be in utf8
3053 * to guarantee those semantics */
3055 SvCUR_set(sv, d - SvPVX_const(sv));
3058 /* See Note on sizing above. */
3059 sv_utf8_upgrade_flags_grow(
3061 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3062 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3063 d = SvPVX(sv) + SvCUR(sv);
3067 /* Add the string to the output */
3068 if (UNI_IS_INVARIANT(uv)) {
3071 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3074 else { /* Here is \N{NAME} but not \N{U+...}. */
3076 SV *res; /* result from charnames */
3077 const char *str; /* the string in 'res' */
3078 STRLEN len; /* its length */
3080 /* Get the value for NAME */
3081 res = newSVpvn(s, e - s);
3082 res = new_constant( NULL, 0, "charnames",
3083 /* includes all of: \N{...} */
3084 res, NULL, s - 3, e - s + 4 );
3086 /* Most likely res will be in utf8 already since the
3087 * standard charnames uses pack U, but a custom translator
3088 * can leave it otherwise, so make sure. XXX This can be
3089 * revisited to not have charnames use utf8 for characters
3090 * that don't need it when regexes don't have to be in utf8
3091 * for Unicode semantics. If doing so, remember EBCDIC */
3092 sv_utf8_upgrade(res);
3093 str = SvPV_const(res, len);
3095 /* Don't accept malformed input */
3096 if (! is_utf8_string((U8 *) str, len)) {
3097 yyerror("Malformed UTF-8 returned by \\N");
3099 else if (PL_lex_inpat) {
3101 if (! len) { /* The name resolved to an empty string */
3102 Copy("\\N{}", d, 4, char);
3106 /* In order to not lose information for the regex
3107 * compiler, pass the result in the specially made
3108 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3109 * the code points in hex of each character
3110 * returned by charnames */
3112 const char *str_end = str + len;
3113 STRLEN char_length; /* cur char's byte length */
3114 STRLEN output_length; /* and the number of bytes
3115 after this is translated
3117 const STRLEN off = d - SvPVX_const(sv);
3119 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3120 * max('U+', '.'); and 1 for NUL */
3121 char hex_string[2 * UTF8_MAXBYTES + 5];
3123 /* Get the first character of the result. */
3124 U32 uv = utf8n_to_uvuni((U8 *) str,
3129 /* The call to is_utf8_string() above hopefully
3130 * guarantees that there won't be an error. But
3131 * it's easy here to make sure. The function just
3132 * above warns and returns 0 if invalid utf8, but
3133 * it can also return 0 if the input is validly a
3134 * NUL. Disambiguate */
3135 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3136 uv = UNICODE_REPLACEMENT;
3139 /* Convert first code point to hex, including the
3140 * boiler plate before it */
3141 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3142 output_length = strlen(hex_string);
3144 /* Make sure there is enough space to hold it */
3145 d = off + SvGROW(sv, off
3147 + (STRLEN)(send - e)
3148 + 2); /* '}' + NUL */
3150 Copy(hex_string, d, output_length, char);
3153 /* For each subsequent character, append dot and
3154 * its ordinal in hex */
3155 while ((str += char_length) < str_end) {
3156 const STRLEN off = d - SvPVX_const(sv);
3157 U32 uv = utf8n_to_uvuni((U8 *) str,
3161 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3162 uv = UNICODE_REPLACEMENT;
3165 sprintf(hex_string, ".%X", (unsigned int) uv);
3166 output_length = strlen(hex_string);
3168 d = off + SvGROW(sv, off
3170 + (STRLEN)(send - e)
3171 + 2); /* '}' + NUL */
3172 Copy(hex_string, d, output_length, char);
3176 *d++ = '}'; /* Done. Add the trailing brace */
3179 else { /* Here, not in a pattern. Convert the name to a
3182 /* If destination is not in utf8, unconditionally
3183 * recode it to be so. This is because \N{} implies
3184 * Unicode semantics, and scalars have to be in utf8
3185 * to guarantee those semantics */
3187 SvCUR_set(sv, d - SvPVX_const(sv));
3190 /* See Note on sizing above. */
3191 sv_utf8_upgrade_flags_grow(sv,
3192 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3193 len + (STRLEN)(send - s) + 1);
3194 d = SvPVX(sv) + SvCUR(sv);
3196 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3198 /* See Note on sizing above. (NOTE: SvCUR() is not
3199 * set correctly here). */
3200 const STRLEN off = d - SvPVX_const(sv);
3201 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3203 Copy(str, d, len, char);
3210 native_range = FALSE; /* \N{} is defined to be Unicode */
3212 s = e + 1; /* Point to just after the '}' */
3215 /* \c is a control character */
3224 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
3227 yyerror("Missing control char name in \\c");
3231 /* printf-style backslashes, formfeeds, newlines, etc */
3233 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3236 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3239 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3242 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3245 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3248 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3251 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3257 } /* end if (backslash) */
3264 /* If we started with encoded form, or already know we want it,
3265 then encode the next character */
3266 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3270 /* One might think that it is wasted effort in the case of the
3271 * source being utf8 (this_utf8 == TRUE) to take the next character
3272 * in the source, convert it to an unsigned value, and then convert
3273 * it back again. But the source has not been validated here. The
3274 * routine that does the conversion checks for errors like
3277 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3278 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3280 SvCUR_set(sv, d - SvPVX_const(sv));
3283 /* See Note on sizing above. */
3284 sv_utf8_upgrade_flags_grow(sv,
3285 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3286 need + (STRLEN)(send - s) + 1);
3287 d = SvPVX(sv) + SvCUR(sv);
3289 } else if (need > len) {
3290 /* encoded value larger than old, may need extra space (NOTE:
3291 * SvCUR() is not set correctly here). See Note on sizing
3293 const STRLEN off = d - SvPVX_const(sv);
3294 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3298 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3300 if (uv > 255 && !dorange)
3301 native_range = FALSE;
3305 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3307 } /* while loop to process each character */
3309 /* terminate the string and set up the sv */
3311 SvCUR_set(sv, d - SvPVX_const(sv));
3312 if (SvCUR(sv) >= SvLEN(sv))
3313 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3316 if (PL_encoding && !has_utf8) {
3317 sv_recode_to_utf8(sv, PL_encoding);
3323 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3324 PL_sublex_info.sub_op->op_private |=
3325 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3329 /* shrink the sv if we allocated more than we used */
3330 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3331 SvPV_shrink_to_cur(sv);
3334 /* return the substring (via pl_yylval) only if we parsed anything */
3335 if (s > PL_bufptr) {
3336 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3337 const char *const key = PL_lex_inpat ? "qr" : "q";
3338 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3342 if (PL_lex_inwhat == OP_TRANS) {
3345 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3353 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3356 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3363 * Returns TRUE if there's more to the expression (e.g., a subscript),
3366 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3368 * ->[ and ->{ return TRUE
3369 * { and [ outside a pattern are always subscripts, so return TRUE
3370 * if we're outside a pattern and it's not { or [, then return FALSE
3371 * if we're in a pattern and the first char is a {
3372 * {4,5} (any digits around the comma) returns FALSE
3373 * if we're in a pattern and the first char is a [
3375 * [SOMETHING] has a funky algorithm to decide whether it's a
3376 * character class or not. It has to deal with things like
3377 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3378 * anything else returns TRUE
3381 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3384 S_intuit_more(pTHX_ register char *s)
3388 PERL_ARGS_ASSERT_INTUIT_MORE;
3390 if (PL_lex_brackets)
3392 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3394 if (*s != '{' && *s != '[')
3399 /* In a pattern, so maybe we have {n,m}. */
3416 /* On the other hand, maybe we have a character class */
3419 if (*s == ']' || *s == '^')
3422 /* this is terrifying, and it works */
3423 int weight = 2; /* let's weigh the evidence */
3425 unsigned char un_char = 255, last_un_char;
3426 const char * const send = strchr(s,']');
3427 char tmpbuf[sizeof PL_tokenbuf * 4];
3429 if (!send) /* has to be an expression */
3432 Zero(seen,256,char);
3435 else if (isDIGIT(*s)) {
3437 if (isDIGIT(s[1]) && s[2] == ']')
3443 for (; s < send; s++) {
3444 last_un_char = un_char;
3445 un_char = (unsigned char)*s;
3450 weight -= seen[un_char] * 10;
3451 if (isALNUM_lazy_if(s+1,UTF)) {
3453 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3454 len = (int)strlen(tmpbuf);
3455 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3460 else if (*s == '$' && s[1] &&
3461 strchr("[#!%*<>()-=",s[1])) {
3462 if (/*{*/ strchr("])} =",s[2]))
3471 if (strchr("wds]",s[1]))
3473 else if (seen[(U8)'\''] || seen[(U8)'"'])
3475 else if (strchr("rnftbxcav",s[1]))
3477 else if (isDIGIT(s[1])) {
3479 while (s[1] && isDIGIT(s[1]))
3489 if (strchr("aA01! ",last_un_char))
3491 if (strchr("zZ79~",s[1]))
3493 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3494 weight -= 5; /* cope with negative subscript */
3497 if (!isALNUM(last_un_char)
3498 && !(last_un_char == '$' || last_un_char == '@'
3499 || last_un_char == '&')
3500 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3505 if (keyword(tmpbuf, d - tmpbuf, 0))
3508 if (un_char == last_un_char + 1)
3510 weight -= seen[un_char];
3515 if (weight >= 0) /* probably a character class */
3525 * Does all the checking to disambiguate
3527 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3528 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3530 * First argument is the stuff after the first token, e.g. "bar".
3532 * Not a method if bar is a filehandle.
3533 * Not a method if foo is a subroutine prototyped to take a filehandle.
3534 * Not a method if it's really "Foo $bar"
3535 * Method if it's "foo $bar"
3536 * Not a method if it's really "print foo $bar"
3537 * Method if it's really "foo package::" (interpreted as package->foo)
3538 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3539 * Not a method if bar is a filehandle or package, but is quoted with
3544 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3547 char *s = start + (*start == '$');
3548 char tmpbuf[sizeof PL_tokenbuf];
3555 PERL_ARGS_ASSERT_INTUIT_METHOD;
3558 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3562 const char *proto = SvPVX_const(cv);
3573 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3574 /* start is the beginning of the possible filehandle/object,
3575 * and s is the end of it
3576 * tmpbuf is a copy of it
3579 if (*start == '$') {
3580 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3581 isUPPER(*PL_tokenbuf))
3584 len = start - SvPVX(PL_linestr);
3588 start = SvPVX(PL_linestr) + len;
3592 return *s == '(' ? FUNCMETH : METHOD;
3594 if (!keyword(tmpbuf, len, 0)) {
3595 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3599 soff = s - SvPVX(PL_linestr);
3603 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3604 if (indirgv && GvCVu(indirgv))
3606 /* filehandle or package name makes it a method */
3607 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3609 soff = s - SvPVX(PL_linestr);
3612 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3613 return 0; /* no assumptions -- "=>" quotes bearword */
3615 start_force(PL_curforce);
3616 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3617 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3618 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3620 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3625 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3627 return *s == '(' ? FUNCMETH : METHOD;
3633 /* Encoded script support. filter_add() effectively inserts a
3634 * 'pre-processing' function into the current source input stream.
3635 * Note that the filter function only applies to the current source file
3636 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3638 * The datasv parameter (which may be NULL) can be used to pass
3639 * private data to this instance of the filter. The filter function
3640 * can recover the SV using the FILTER_DATA macro and use it to
3641 * store private buffers and state information.
3643 * The supplied datasv parameter is upgraded to a PVIO type
3644 * and the IoDIRP/IoANY field is used to store the function pointer,
3645 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3646 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3647 * private use must be set using malloc'd pointers.
3651 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3660 if (!PL_rsfp_filters)
3661 PL_rsfp_filters = newAV();
3664 SvUPGRADE(datasv, SVt_PVIO);
3665 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3666 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3667 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3668 FPTR2DPTR(void *, IoANY(datasv)),
3669 SvPV_nolen(datasv)));
3670 av_unshift(PL_rsfp_filters, 1);
3671 av_store(PL_rsfp_filters, 0, datasv) ;
3676 /* Delete most recently added instance of this filter function. */
3678 Perl_filter_del(pTHX_ filter_t funcp)
3683 PERL_ARGS_ASSERT_FILTER_DEL;
3686 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3687 FPTR2DPTR(void*, funcp)));
3689 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3691 /* if filter is on top of stack (usual case) just pop it off */
3692 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3693 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3694 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3695 IoANY(datasv) = (void *)NULL;
3696 sv_free(av_pop(PL_rsfp_filters));
3700 /* we need to search for the correct entry and clear it */
3701 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3705 /* Invoke the idxth filter function for the current rsfp. */
3706 /* maxlen 0 = read one text line */
3708 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3713 /* This API is bad. It should have been using unsigned int for maxlen.
3714 Not sure if we want to change the API, but if not we should sanity
3715 check the value here. */
3716 const unsigned int correct_length
3725 PERL_ARGS_ASSERT_FILTER_READ;
3727 if (!PL_parser || !PL_rsfp_filters)
3729 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3730 /* Provide a default input filter to make life easy. */
3731 /* Note that we append to the line. This is handy. */
3732 DEBUG_P(PerlIO_printf(Perl_debug_log,
3733 "filter_read %d: from rsfp\n", idx));
3734 if (correct_length) {
3737 const int old_len = SvCUR(buf_sv);
3739 /* ensure buf_sv is large enough */
3740 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3741 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3742 correct_length)) <= 0) {
3743 if (PerlIO_error(PL_rsfp))
3744 return -1; /* error */
3746 return 0 ; /* end of file */
3748 SvCUR_set(buf_sv, old_len + len) ;
3749 SvPVX(buf_sv)[old_len + len] = '\0';
3752 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3753 if (PerlIO_error(PL_rsfp))
3754 return -1; /* error */
3756 return 0 ; /* end of file */
3759 return SvCUR(buf_sv);
3761 /* Skip this filter slot if filter has been deleted */
3762 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3763 DEBUG_P(PerlIO_printf(Perl_debug_log,
3764 "filter_read %d: skipped (filter deleted)\n",
3766 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3768 /* Get function pointer hidden within datasv */
3769 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3770 DEBUG_P(PerlIO_printf(Perl_debug_log,
3771 "filter_read %d: via function %p (%s)\n",
3772 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3773 /* Call function. The function is expected to */
3774 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3775 /* Return: <0:error, =0:eof, >0:not eof */
3776 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3780 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3784 PERL_ARGS_ASSERT_FILTER_GETS;
3786 #ifdef PERL_CR_FILTER
3787 if (!PL_rsfp_filters) {
3788 filter_add(S_cr_textfilter,NULL);
3791 if (PL_rsfp_filters) {
3793 SvCUR_set(sv, 0); /* start with empty line */
3794 if (FILTER_READ(0, sv, 0) > 0)
3795 return ( SvPVX(sv) ) ;
3800 return (sv_gets(sv, PL_rsfp, append));
3804 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3809 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3811 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3815 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3816 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3818 return GvHV(gv); /* Foo:: */
3821 /* use constant CLASS => 'MyClass' */
3822 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3823 if (gv && GvCV(gv)) {
3824 SV * const sv = cv_const_sv(GvCV(gv));
3826 pkgname = SvPV_const(sv, len);
3829 return gv_stashpvn(pkgname, len, 0);
3833 * S_readpipe_override
3834 * Check whether readpipe() is overriden, and generates the appropriate
3835 * optree, provided sublex_start() is called afterwards.
3838 S_readpipe_override(pTHX)
3841 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3842 pl_yylval.ival = OP_BACKTICK;
3844 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3846 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3847 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3848 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3850 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3851 append_elem(OP_LIST,
3852 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3853 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3860 * The intent of this yylex wrapper is to minimize the changes to the
3861 * tokener when we aren't interested in collecting madprops. It remains
3862 * to be seen how successful this strategy will be...
3869 char *s = PL_bufptr;
3871 /* make sure PL_thiswhite is initialized */
3875 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3876 if (PL_pending_ident)
3877 return S_pending_ident(aTHX);
3879 /* previous token ate up our whitespace? */
3880 if (!PL_lasttoke && PL_nextwhite) {
3881 PL_thiswhite = PL_nextwhite;
3885 /* isolate the token, and figure out where it is without whitespace */
3886 PL_realtokenstart = -1;
3890 assert(PL_curforce < 0);
3892 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3893 if (!PL_thistoken) {
3894 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3895 PL_thistoken = newSVpvs("");
3897 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3898 PL_thistoken = newSVpvn(tstart, s - tstart);
3901 if (PL_thismad) /* install head */
3902 CURMAD('X', PL_thistoken);
3905 /* last whitespace of a sublex? */
3906 if (optype == ')' && PL_endwhite) {
3907 CURMAD('X', PL_endwhite);
3912 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3913 if (!PL_thiswhite && !PL_endwhite && !optype) {
3914 sv_free(PL_thistoken);
3919 /* put off final whitespace till peg */
3920 if (optype == ';' && !PL_rsfp) {
3921 PL_nextwhite = PL_thiswhite;
3924 else if (PL_thisopen) {
3925 CURMAD('q', PL_thisopen);
3927 sv_free(PL_thistoken);
3931 /* Store actual token text as madprop X */
3932 CURMAD('X', PL_thistoken);
3936 /* add preceding whitespace as madprop _ */
3937 CURMAD('_', PL_thiswhite);
3941 /* add quoted material as madprop = */
3942 CURMAD('=', PL_thisstuff);
3946 /* add terminating quote as madprop Q */
3947 CURMAD('Q', PL_thisclose);
3951 /* special processing based on optype */
3955 /* opval doesn't need a TOKEN since it can already store mp */
3965 if (pl_yylval.opval)
3966 append_madprops(PL_thismad, pl_yylval.opval, 0);
3974 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3983 /* remember any fake bracket that lexer is about to discard */
3984 if (PL_lex_brackets == 1 &&
3985 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3988 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3991 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3992 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3995 break; /* don't bother looking for trailing comment */
4004 /* attach a trailing comment to its statement instead of next token */
4008 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4010 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4012 if (*s == '\n' || *s == '#') {
4013 while (s < PL_bufend && *s != '\n')
4017 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4018 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4035 /* Create new token struct. Note: opvals return early above. */
4036 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4043 S_tokenize_use(pTHX_ int is_use, char *s) {
4046 PERL_ARGS_ASSERT_TOKENIZE_USE;
4048 if (PL_expect != XSTATE)
4049 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4050 is_use ? "use" : "no"));
4052 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4053 s = force_version(s, TRUE);
4054 if (*s == ';' || *s == '}'
4055 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4056 start_force(PL_curforce);
4057 NEXTVAL_NEXTTOKE.opval = NULL;
4060 else if (*s == 'v') {
4061 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4062 s = force_version(s, FALSE);
4066 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4067 s = force_version(s, FALSE);
4069 pl_yylval.ival = is_use;
4073 static const char* const exp_name[] =
4074 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4075 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4082 Works out what to call the token just pulled out of the input
4083 stream. The yacc parser takes care of taking the ops we return and
4084 stitching them into a tree.
4090 if read an identifier
4091 if we're in a my declaration
4092 croak if they tried to say my($foo::bar)
4093 build the ops for a my() declaration
4094 if it's an access to a my() variable
4095 are we in a sort block?
4096 croak if my($a); $a <=> $b
4097 build ops for access to a my() variable
4098 if in a dq string, and they've said @foo and we can't find @foo
4100 build ops for a bareword
4101 if we already built the token before, use it.
4106 #pragma segment Perl_yylex
4112 register char *s = PL_bufptr;
4118 /* orig_keyword, gvp, and gv are initialized here because
4119 * jump to the label just_a_word_zero can bypass their
4120 * initialization later. */
4121 I32 orig_keyword = 0;
4126 SV* tmp = newSVpvs("");
4127 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4128 (IV)CopLINE(PL_curcop),
4129 lex_state_names[PL_lex_state],
4130 exp_name[PL_expect],
4131 pv_display(tmp, s, strlen(s), 0, 60));
4134 /* check if there's an identifier for us to look at */
4135 if (PL_pending_ident)
4136 return REPORT(S_pending_ident(aTHX));
4138 /* no identifier pending identification */
4140 switch (PL_lex_state) {
4142 case LEX_NORMAL: /* Some compilers will produce faster */
4143 case LEX_INTERPNORMAL: /* code if we comment these out. */
4147 /* when we've already built the next token, just pull it out of the queue */
4151 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4153 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4154 PL_nexttoke[PL_lasttoke].next_mad = 0;
4155 if (PL_thismad && PL_thismad->mad_key == '_') {
4156 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4157 PL_thismad->mad_val = 0;
4158 mad_free(PL_thismad);
4163 PL_lex_state = PL_lex_defer;
4164 PL_expect = PL_lex_expect;
4165 PL_lex_defer = LEX_NORMAL;
4166 if (!PL_nexttoke[PL_lasttoke].next_type)
4171 pl_yylval = PL_nextval[PL_nexttoke];
4173 PL_lex_state = PL_lex_defer;
4174 PL_expect = PL_lex_expect;
4175 PL_lex_defer = LEX_NORMAL;
4179 /* FIXME - can these be merged? */
4180 return(PL_nexttoke[PL_lasttoke].next_type);
4182 return REPORT(PL_nexttype[PL_nexttoke]);
4185 /* interpolated case modifiers like \L \U, including \Q and \E.
4186 when we get here, PL_bufptr is at the \
4188 case LEX_INTERPCASEMOD:
4190 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4191 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4193 /* handle \E or end of string */
4194 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4196 if (PL_lex_casemods) {
4197 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4198 PL_lex_casestack[PL_lex_casemods] = '\0';
4200 if (PL_bufptr != PL_bufend
4201 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4203 PL_lex_state = LEX_INTERPCONCAT;
4206 PL_thistoken = newSVpvs("\\E");
4212 while (PL_bufptr != PL_bufend &&
4213 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4215 PL_thiswhite = newSVpvs("");
4216 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4220 if (PL_bufptr != PL_bufend)
4223 PL_lex_state = LEX_INTERPCONCAT;
4227 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4228 "### Saw case modifier\n"); });
4230 if (s[1] == '\\' && s[2] == 'E') {
4233 PL_thiswhite = newSVpvs("");
4234 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4237 PL_lex_state = LEX_INTERPCONCAT;
4242 if (!PL_madskills) /* when just compiling don't need correct */
4243 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4244 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4245 if ((*s == 'L' || *s == 'U') &&
4246 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4247 PL_lex_casestack[--PL_lex_casemods] = '\0';
4250 if (PL_lex_casemods > 10)
4251 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4252 PL_lex_casestack[PL_lex_casemods++] = *s;
4253 PL_lex_casestack[PL_lex_casemods] = '\0';
4254 PL_lex_state = LEX_INTERPCONCAT;
4255 start_force(PL_curforce);
4256 NEXTVAL_NEXTTOKE.ival = 0;
4258 start_force(PL_curforce);
4260 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4262 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4264 NEXTVAL_NEXTTOKE.ival = OP_LC;
4266 NEXTVAL_NEXTTOKE.ival = OP_UC;
4268 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4270 Perl_croak(aTHX_ "panic: yylex");
4272 SV* const tmpsv = newSVpvs("\\ ");
4273 /* replace the space with the character we want to escape
4275 SvPVX(tmpsv)[1] = *s;
4281 if (PL_lex_starts) {
4287 sv_free(PL_thistoken);
4288 PL_thistoken = newSVpvs("");
4291 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4292 if (PL_lex_casemods == 1 && PL_lex_inpat)
4301 case LEX_INTERPPUSH:
4302 return REPORT(sublex_push());
4304 case LEX_INTERPSTART:
4305 if (PL_bufptr == PL_bufend)
4306 return REPORT(sublex_done());
4307 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4308 "### Interpolated variable\n"); });
4310 PL_lex_dojoin = (*PL_bufptr == '@');
4311 PL_lex_state = LEX_INTERPNORMAL;
4312 if (PL_lex_dojoin) {
4313 start_force(PL_curforce);
4314 NEXTVAL_NEXTTOKE.ival = 0;
4316 start_force(PL_curforce);
4317 force_ident("\"", '$');
4318 start_force(PL_curforce);
4319 NEXTVAL_NEXTTOKE.ival = 0;
4321 start_force(PL_curforce);
4322 NEXTVAL_NEXTTOKE.ival = 0;
4324 start_force(PL_curforce);
4325 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4328 if (PL_lex_starts++) {
4333 sv_free(PL_thistoken);
4334 PL_thistoken = newSVpvs("");
4337 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4338 if (!PL_lex_casemods && PL_lex_inpat)
4345 case LEX_INTERPENDMAYBE:
4346 if (intuit_more(PL_bufptr)) {
4347 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4353 if (PL_lex_dojoin) {
4354 PL_lex_dojoin = FALSE;
4355 PL_lex_state = LEX_INTERPCONCAT;
4359 sv_free(PL_thistoken);
4360 PL_thistoken = newSVpvs("");
4365 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4366 && SvEVALED(PL_lex_repl))
4368 if (PL_bufptr != PL_bufend)
4369 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4373 case LEX_INTERPCONCAT:
4375 if (PL_lex_brackets)
4376 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4378 if (PL_bufptr == PL_bufend)
4379 return REPORT(sublex_done());
4381 if (SvIVX(PL_linestr) == '\'') {
4382 SV *sv = newSVsv(PL_linestr);
4385 else if ( PL_hints & HINT_NEW_RE )
4386 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4387 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4391 s = scan_const(PL_bufptr);
4393 PL_lex_state = LEX_INTERPCASEMOD;
4395 PL_lex_state = LEX_INTERPSTART;
4398 if (s != PL_bufptr) {
4399 start_force(PL_curforce);
4401 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4403 NEXTVAL_NEXTTOKE = pl_yylval;
4406 if (PL_lex_starts++) {
4410 sv_free(PL_thistoken);
4411 PL_thistoken = newSVpvs("");
4414 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4415 if (!PL_lex_casemods && PL_lex_inpat)
4428 PL_lex_state = LEX_NORMAL;
4429 s = scan_formline(PL_bufptr);
4430 if (!PL_lex_formbrack)
4436 PL_oldoldbufptr = PL_oldbufptr;
4442 sv_free(PL_thistoken);
4445 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4449 if (isIDFIRST_lazy_if(s,UTF))
4452 unsigned char c = *s;
4453 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4454 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4455 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4460 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4464 goto fake_eof; /* emulate EOF on ^D or ^Z */
4473 if (PL_lex_brackets) {
4474 yyerror((const char *)
4476 ? "Format not terminated"
4477 : "Missing right curly or square bracket"));
4479 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4480 "### Tokener got EOF\n");
4484 if (s++ < PL_bufend)
4485 goto retry; /* ignore stray nulls */
4488 if (!PL_in_eval && !PL_preambled) {
4489 PL_preambled = TRUE;
4495 /* Generate a string of Perl code to load the debugger.
4496 * If PERL5DB is set, it will return the contents of that,
4497 * otherwise a compile-time require of perl5db.pl. */
4499 const char * const pdb = PerlEnv_getenv("PERL5DB");
4502 sv_setpv(PL_linestr, pdb);
4503 sv_catpvs(PL_linestr,";");
4505 SETERRNO(0,SS_NORMAL);
4506 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4509 sv_setpvs(PL_linestr,"");
4510 if (PL_preambleav) {
4511 SV **svp = AvARRAY(PL_preambleav);
4512 SV **const end = svp + AvFILLp(PL_preambleav);
4514 sv_catsv(PL_linestr, *svp);
4516 sv_catpvs(PL_linestr, ";");
4518 sv_free(MUTABLE_SV(PL_preambleav));
4519 PL_preambleav = NULL;
4522 sv_catpvs(PL_linestr,
4523 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4524 if (PL_minus_n || PL_minus_p) {
4525 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4527 sv_catpvs(PL_linestr,"chomp;");
4530 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4531 || *PL_splitstr == '"')
4532 && strchr(PL_splitstr + 1, *PL_splitstr))
4533 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4535 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4536 bytes can be used as quoting characters. :-) */
4537 const char *splits = PL_splitstr;
4538 sv_catpvs(PL_linestr, "our @F=split(q\0");
4541 if (*splits == '\\')
4542 sv_catpvn(PL_linestr, splits, 1);
4543 sv_catpvn(PL_linestr, splits, 1);
4544 } while (*splits++);
4545 /* This loop will embed the trailing NUL of
4546 PL_linestr as the last thing it does before
4548 sv_catpvs(PL_linestr, ");");
4552 sv_catpvs(PL_linestr,"our @F=split(' ');");
4555 sv_catpvs(PL_linestr, "\n");
4556 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4557 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4558 PL_last_lop = PL_last_uni = NULL;
4559 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4560 update_debugger_info(PL_linestr, NULL, 0);
4565 bof = PL_rsfp ? TRUE : FALSE;
4568 fake_eof = LEX_FAKE_EOF;
4570 PL_bufptr = PL_bufend;
4571 CopLINE_inc(PL_curcop);
4572 if (!lex_next_chunk(fake_eof)) {
4573 CopLINE_dec(PL_curcop);
4575 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4577 CopLINE_dec(PL_curcop);
4580 PL_realtokenstart = -1;
4583 /* If it looks like the start of a BOM or raw UTF-16,
4584 * check if it in fact is. */
4585 if (bof && PL_rsfp &&
4590 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4592 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4593 s = swallow_bom((U8*)s);
4597 /* Incest with pod. */
4600 sv_catsv(PL_thiswhite, PL_linestr);
4602 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4603 sv_setpvs(PL_linestr, "");
4604 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4605 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4606 PL_last_lop = PL_last_uni = NULL;
4607 PL_doextract = FALSE;
4612 } while (PL_doextract);
4613 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4614 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4615 PL_last_lop = PL_last_uni = NULL;
4616 if (CopLINE(PL_curcop) == 1) {
4617 while (s < PL_bufend && isSPACE(*s))
4619 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4623 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4627 if (*s == '#' && *(s+1) == '!')
4629 #ifdef ALTERNATE_SHEBANG
4631 static char const as[] = ALTERNATE_SHEBANG;
4632 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4633 d = s + (sizeof(as) - 1);
4635 #endif /* ALTERNATE_SHEBANG */
4644 while (*d && !isSPACE(*d))
4648 #ifdef ARG_ZERO_IS_SCRIPT
4649 if (ipathend > ipath) {
4651 * HP-UX (at least) sets argv[0] to the script name,
4652 * which makes $^X incorrect. And Digital UNIX and Linux,
4653 * at least, set argv[0] to the basename of the Perl
4654 * interpreter. So, having found "#!", we'll set it right.
4656 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4658 assert(SvPOK(x) || SvGMAGICAL(x));
4659 if (sv_eq(x, CopFILESV(PL_curcop))) {
4660 sv_setpvn(x, ipath, ipathend - ipath);
4666 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4667 const char * const lstart = SvPV_const(x,llen);
4669 bstart += blen - llen;
4670 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4671 sv_setpvn(x, ipath, ipathend - ipath);
4676 TAINT_NOT; /* $^X is always tainted, but that's OK */
4678 #endif /* ARG_ZERO_IS_SCRIPT */
4683 d = instr(s,"perl -");
4685 d = instr(s,"perl");
4687 /* avoid getting into infinite loops when shebang
4688 * line contains "Perl" rather than "perl" */
4690 for (d = ipathend-4; d >= ipath; --d) {
4691 if ((*d == 'p' || *d == 'P')
4692 && !ibcmp(d, "perl", 4))
4702 #ifdef ALTERNATE_SHEBANG
4704 * If the ALTERNATE_SHEBANG on this system starts with a
4705 * character that can be part of a Perl expression, then if
4706 * we see it but not "perl", we're probably looking at the
4707 * start of Perl code, not a request to hand off to some
4708 * other interpreter. Similarly, if "perl" is there, but
4709 * not in the first 'word' of the line, we assume the line
4710 * contains the start of the Perl program.
4712 if (d && *s != '#') {
4713 const char *c = ipath;
4714 while (*c && !strchr("; \t\r\n\f\v#", *c))
4717 d = NULL; /* "perl" not in first word; ignore */
4719 *s = '#'; /* Don't try to parse shebang line */
4721 #endif /* ALTERNATE_SHEBANG */
4726 !instr(s,"indir") &&
4727 instr(PL_origargv[0],"perl"))
4734 while (s < PL_bufend && isSPACE(*s))
4736 if (s < PL_bufend) {
4737 Newx(newargv,PL_origargc+3,char*);
4739 while (s < PL_bufend && !isSPACE(*s))
4742 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4745 newargv = PL_origargv;
4748 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4750 Perl_croak(aTHX_ "Can't exec %s", ipath);
4753 while (*d && !isSPACE(*d))
4755 while (SPACE_OR_TAB(*d))
4759 const bool switches_done = PL_doswitches;
4760 const U32 oldpdb = PL_perldb;
4761 const bool oldn = PL_minus_n;
4762 const bool oldp = PL_minus_p;
4766 bool baduni = FALSE;
4768 const char *d2 = d1 + 1;
4769 if (parse_unicode_opts((const char **)&d2)
4773 if (baduni || *d1 == 'M' || *d1 == 'm') {
4774 const char * const m = d1;
4775 while (*d1 && !isSPACE(*d1))
4777 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4780 d1 = moreswitches(d1);
4782 if (PL_doswitches && !switches_done) {
4783 int argc = PL_origargc;
4784 char **argv = PL_origargv;
4787 } while (argc && argv[0][0] == '-' && argv[0][1]);
4788 init_argv_symbols(argc,argv);
4790 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4791 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4792 /* if we have already added "LINE: while (<>) {",
4793 we must not do it again */
4795 sv_setpvs(PL_linestr, "");
4796 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4797 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4798 PL_last_lop = PL_last_uni = NULL;
4799 PL_preambled = FALSE;
4800 if (PERLDB_LINE || PERLDB_SAVESRC)
4801 (void)gv_fetchfile(PL_origfilename);
4808 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4810 PL_lex_state = LEX_FORMLINE;
4815 #ifdef PERL_STRICT_CR
4816 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4818 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4820 case ' ': case '\t': case '\f': case 013:
4822 PL_realtokenstart = -1;
4824 PL_thiswhite = newSVpvs("");
4825 sv_catpvn(PL_thiswhite, s, 1);
4832 PL_realtokenstart = -1;
4836 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4837 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4838 /* handle eval qq[#line 1 "foo"\n ...] */
4839 CopLINE_dec(PL_curcop);
4842 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4844 if (!PL_in_eval || PL_rsfp)
4849 while (d < PL_bufend && *d != '\n')
4853 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4854 Perl_croak(aTHX_ "panic: input overflow");
4857 PL_thiswhite = newSVpvn(s, d - s);
4862 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4864 PL_lex_state = LEX_FORMLINE;
4870 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4871 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4874 TOKEN(PEG); /* make sure any #! line is accessible */
4879 /* if (PL_madskills && PL_lex_formbrack) { */
4881 while (d < PL_bufend && *d != '\n')
4885 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4886 Perl_croak(aTHX_ "panic: input overflow");
4887 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4889 PL_thiswhite = newSVpvs("");
4890 if (CopLINE(PL_curcop) == 1) {
4891 sv_setpvs(PL_thiswhite, "");
4894 sv_catpvn(PL_thiswhite, s, d - s);
4908 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4916 while (s < PL_bufend && SPACE_OR_TAB(*s))
4919 if (strnEQ(s,"=>",2)) {
4920 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4921 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4922 OPERATOR('-'); /* unary minus */
4924 PL_last_uni = PL_oldbufptr;
4926 case 'r': ftst = OP_FTEREAD; break;
4927 case 'w': ftst = OP_FTEWRITE; break;
4928 case 'x': ftst = OP_FTEEXEC; break;
4929 case 'o': ftst = OP_FTEOWNED; break;
4930 case 'R': ftst = OP_FTRREAD; break;
4931 case 'W': ftst = OP_FTRWRITE; break;
4932 case 'X': ftst = OP_FTREXEC; break;
4933 case 'O': ftst = OP_FTROWNED; break;
4934 case 'e': ftst = OP_FTIS; break;
4935 case 'z': ftst = OP_FTZERO; break;
4936 case 's': ftst = OP_FTSIZE; break;
4937 case 'f': ftst = OP_FTFILE; break;
4938 case 'd': ftst = OP_FTDIR; break;
4939 case 'l': ftst = OP_FTLINK; break;
4940 case 'p': ftst = OP_FTPIPE; break;
4941 case 'S': ftst = OP_FTSOCK; break;
4942 case 'u': ftst = OP_FTSUID; break;
4943 case 'g': ftst = OP_FTSGID; break;
4944 case 'k': ftst = OP_FTSVTX; break;
4945 case 'b': ftst = OP_FTBLK; break;
4946 case 'c': ftst = OP_FTCHR; break;
4947 case 't': ftst = OP_FTTTY; break;
4948 case 'T': ftst = OP_FTTEXT; break;
4949 case 'B': ftst = OP_FTBINARY; break;
4950 case 'M': case 'A': case 'C':
4951 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4953 case 'M': ftst = OP_FTMTIME; break;
4954 case 'A': ftst = OP_FTATIME; break;
4955 case 'C': ftst = OP_FTCTIME; break;
4963 PL_last_lop_op = (OPCODE)ftst;
4964 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4965 "### Saw file test %c\n", (int)tmp);
4970 /* Assume it was a minus followed by a one-letter named
4971 * subroutine call (or a -bareword), then. */
4972 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4973 "### '-%c' looked like a file test but was not\n",
4980 const char tmp = *s++;
4983 if (PL_expect == XOPERATOR)
4988 else if (*s == '>') {
4991 if (isIDFIRST_lazy_if(s,UTF)) {
4992 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5000 if (PL_expect == XOPERATOR)
5003 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5005 OPERATOR('-'); /* unary minus */
5011 const char tmp = *s++;
5014 if (PL_expect == XOPERATOR)
5019 if (PL_expect == XOPERATOR)
5022 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5029 if (PL_expect != XOPERATOR) {
5030 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5031 PL_expect = XOPERATOR;
5032 force_ident(PL_tokenbuf, '*');
5045 if (PL_expect == XOPERATOR) {
5049 PL_tokenbuf[0] = '%';
5050 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5051 sizeof PL_tokenbuf - 1, FALSE);
5052 if (!PL_tokenbuf[1]) {
5055 PL_pending_ident = '%';
5064 const char tmp = *s++;
5069 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5076 const char tmp = *s++;
5082 goto just_a_word_zero_gv;
5085 switch (PL_expect) {
5091 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5093 PL_bufptr = s; /* update in case we back off */
5095 deprecate(":= for an empty attribute list");
5102 PL_expect = XTERMBLOCK;
5105 stuffstart = s - SvPVX(PL_linestr) - 1;
5109 while (isIDFIRST_lazy_if(s,UTF)) {
5112 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5113 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5114 if (tmp < 0) tmp = -tmp;
5129 sv = newSVpvn(s, len);
5131 d = scan_str(d,TRUE,TRUE);
5133 /* MUST advance bufptr here to avoid bogus
5134 "at end of line" context messages from yyerror().
5136 PL_bufptr = s + len;
5137 yyerror("Unterminated attribute parameter in attribute list");
5141 return REPORT(0); /* EOF indicator */
5145 sv_catsv(sv, PL_lex_stuff);
5146 attrs = append_elem(OP_LIST, attrs,
5147 newSVOP(OP_CONST, 0, sv));
5148 SvREFCNT_dec(PL_lex_stuff);
5149 PL_lex_stuff = NULL;
5152 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5154 if (PL_in_my == KEY_our) {
5155 deprecate(":unique");
5158 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5161 /* NOTE: any CV attrs applied here need to be part of
5162 the CVf_BUILTIN_ATTRS define in cv.h! */
5163 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5165 CvLVALUE_on(PL_compcv);
5167 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5169 deprecate(":locked");
5171 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5173 CvMETHOD_on(PL_compcv);
5175 /* After we've set the flags, it could be argued that
5176 we don't need to do the attributes.pm-based setting
5177 process, and shouldn't bother appending recognized
5178 flags. To experiment with that, uncomment the
5179 following "else". (Note that's already been
5180 uncommented. That keeps the above-applied built-in
5181 attributes from being intercepted (and possibly
5182 rejected) by a package's attribute routines, but is
5183 justified by the performance win for the common case
5184 of applying only built-in attributes.) */
5186 attrs = append_elem(OP_LIST, attrs,
5187 newSVOP(OP_CONST, 0,
5191 if (*s == ':' && s[1] != ':')
5194 break; /* require real whitespace or :'s */
5195 /* XXX losing whitespace on sequential attributes here */
5199 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5200 if (*s != ';' && *s != '}' && *s != tmp
5201 && (tmp != '=' || *s != ')')) {
5202 const char q = ((*s == '\'') ? '"' : '\'');
5203 /* If here for an expression, and parsed no attrs, back
5205 if (tmp == '=' && !attrs) {
5209 /* MUST advance bufptr here to avoid bogus "at end of line"
5210 context messages from yyerror().
5213 yyerror( (const char *)
5215 ? Perl_form(aTHX_ "Invalid separator character "
5216 "%c%c%c in attribute list", q, *s, q)
5217 : "Unterminated attribute list" ) );
5225 start_force(PL_curforce);
5226 NEXTVAL_NEXTTOKE.opval = attrs;
5227 CURMAD('_', PL_nextwhite);
5232 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5233 (s - SvPVX(PL_linestr)) - stuffstart);
5241 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5242 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5250 const char tmp = *s++;
5255 const char tmp = *s++;
5263 if (PL_lex_brackets <= 0)
5264 yyerror("Unmatched right square bracket");
5267 if (PL_lex_state == LEX_INTERPNORMAL) {
5268 if (PL_lex_brackets == 0) {
5269 if (*s == '-' && s[1] == '>')
5270 PL_lex_state = LEX_INTERPENDMAYBE;
5271 else if (*s != '[' && *s != '{')
5272 PL_lex_state = LEX_INTERPEND;
5279 if (PL_lex_brackets > 100) {
5280 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5282 switch (PL_expect) {
5284 if (PL_lex_formbrack) {
5288 if (PL_oldoldbufptr == PL_last_lop)
5289 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5291 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5292 OPERATOR(HASHBRACK);
5294 while (s < PL_bufend && SPACE_OR_TAB(*s))
5297 PL_tokenbuf[0] = '\0';
5298 if (d < PL_bufend && *d == '-') {
5299 PL_tokenbuf[0] = '-';
5301 while (d < PL_bufend && SPACE_OR_TAB(*d))
5304 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5305 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5307 while (d < PL_bufend && SPACE_OR_TAB(*d))
5310 const char minus = (PL_tokenbuf[0] == '-');
5311 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5319 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5324 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5329 if (PL_oldoldbufptr == PL_last_lop)
5330 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5332 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5335 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5337 /* This hack is to get the ${} in the message. */
5339 yyerror("syntax error");
5342 OPERATOR(HASHBRACK);
5344 /* This hack serves to disambiguate a pair of curlies
5345 * as being a block or an anon hash. Normally, expectation
5346 * determines that, but in cases where we're not in a
5347 * position to expect anything in particular (like inside
5348 * eval"") we have to resolve the ambiguity. This code
5349 * covers the case where the first term in the curlies is a
5350 * quoted string. Most other cases need to be explicitly
5351 * disambiguated by prepending a "+" before the opening
5352 * curly in order to force resolution as an anon hash.
5354 * XXX should probably propagate the outer expectation
5355 * into eval"" to rely less on this hack, but that could
5356 * potentially break current behavior of eval"".
5360 if (*s == '\'' || *s == '"' || *s == '`') {
5361 /* common case: get past first string, handling escapes */
5362 for (t++; t < PL_bufend && *t != *s;)
5363 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5367 else if (*s == 'q') {
5370 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5373 /* skip q//-like construct */
5375 char open, close, term;
5378 while (t < PL_bufend && isSPACE(*t))
5380 /* check for q => */
5381 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5382 OPERATOR(HASHBRACK);
5386 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5390 for (t++; t < PL_bufend; t++) {
5391 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5393 else if (*t == open)
5397 for (t++; t < PL_bufend; t++) {
5398 if (*t == '\\' && t+1 < PL_bufend)
5400 else if (*t == close && --brackets <= 0)
5402 else if (*t == open)
5409 /* skip plain q word */
5410 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5413 else if (isALNUM_lazy_if(t,UTF)) {
5415 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5418 while (t < PL_bufend && isSPACE(*t))
5420 /* if comma follows first term, call it an anon hash */
5421 /* XXX it could be a comma expression with loop modifiers */
5422 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5423 || (*t == '=' && t[1] == '>')))
5424 OPERATOR(HASHBRACK);
5425 if (PL_expect == XREF)
5428 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5434 pl_yylval.ival = CopLINE(PL_curcop);
5435 if (isSPACE(*s) || *s == '#')
5436 PL_copline = NOLINE; /* invalidate current command line number */
5441 if (PL_lex_brackets <= 0)
5442 yyerror("Unmatched right curly bracket");
5444 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5445 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5446 PL_lex_formbrack = 0;
5447 if (PL_lex_state == LEX_INTERPNORMAL) {
5448 if (PL_lex_brackets == 0) {
5449 if (PL_expect & XFAKEBRACK) {
5450 PL_expect &= XENUMMASK;
5451 PL_lex_state = LEX_INTERPEND;
5456 PL_thiswhite = newSVpvs("");
5457 sv_catpvs(PL_thiswhite,"}");
5460 return yylex(); /* ignore fake brackets */
5462 if (*s == '-' && s[1] == '>')
5463 PL_lex_state = LEX_INTERPENDMAYBE;
5464 else if (*s != '[' && *s != '{')
5465 PL_lex_state = LEX_INTERPEND;
5468 if (PL_expect & XFAKEBRACK) {
5469 PL_expect &= XENUMMASK;
5471 return yylex(); /* ignore fake brackets */
5473 start_force(PL_curforce);
5475 curmad('X', newSVpvn(s-1,1));
5476 CURMAD('_', PL_thiswhite);
5481 PL_thistoken = newSVpvs("");
5489 if (PL_expect == XOPERATOR) {
5490 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5491 && isIDFIRST_lazy_if(s,UTF))
5493 CopLINE_dec(PL_curcop);
5494 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5495 CopLINE_inc(PL_curcop);
5500 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5502 PL_expect = XOPERATOR;
5503 force_ident(PL_tokenbuf, '&');
5507 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5519 const char tmp = *s++;
5526 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5527 && strchr("+-*/%.^&|<",tmp))
5528 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5529 "Reversed %c= operator",(int)tmp);
5531 if (PL_expect == XSTATE && isALPHA(tmp) &&
5532 (s == PL_linestart+1 || s[-2] == '\n') )
5534 if (PL_in_eval && !PL_rsfp) {
5539 if (strnEQ(s,"=cut",4)) {
5555 PL_thiswhite = newSVpvs("");
5556 sv_catpvn(PL_thiswhite, PL_linestart,
5557 PL_bufend - PL_linestart);
5561 PL_doextract = TRUE;
5565 if (PL_lex_brackets < PL_lex_formbrack) {
5567 #ifdef PERL_STRICT_CR
5568 while (SPACE_OR_TAB(*t))
5570 while (SPACE_OR_TAB(*t) || *t == '\r')
5573 if (*t == '\n' || *t == '#') {
5584 const char tmp = *s++;
5586 /* was this !=~ where !~ was meant?
5587 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5589 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5590 const char *t = s+1;
5592 while (t < PL_bufend && isSPACE(*t))
5595 if (*t == '/' || *t == '?' ||
5596 ((*t == 'm' || *t == 's' || *t == 'y')
5597 && !isALNUM(t[1])) ||
5598 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5599 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5600 "!=~ should be !~");
5610 if (PL_expect != XOPERATOR) {
5611 if (s[1] != '<' && !strchr(s,'>'))
5614 s = scan_heredoc(s);
5616 s = scan_inputsymbol(s);
5617 TERM(sublex_start());
5623 SHop(OP_LEFT_SHIFT);
5637 const char tmp = *s++;
5639 SHop(OP_RIGHT_SHIFT);
5640 else if (tmp == '=')
5649 if (PL_expect == XOPERATOR) {
5650 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5651 return deprecate_commaless_var_list();
5655 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5656 PL_tokenbuf[0] = '@';
5657 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5658 sizeof PL_tokenbuf - 1, FALSE);
5659 if (PL_expect == XOPERATOR)
5660 no_op("Array length", s);
5661 if (!PL_tokenbuf[1])
5663 PL_expect = XOPERATOR;
5664 PL_pending_ident = '#';
5668 PL_tokenbuf[0] = '$';
5669 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5670 sizeof PL_tokenbuf - 1, FALSE);
5671 if (PL_expect == XOPERATOR)
5673 if (!PL_tokenbuf[1]) {
5675 yyerror("Final $ should be \\$ or $name");
5679 /* This kludge not intended to be bulletproof. */
5680 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5681 pl_yylval.opval = newSVOP(OP_CONST, 0,
5682 newSViv(CopARYBASE_get(&PL_compiling)));
5683 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5689 const char tmp = *s;
5690 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5693 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5694 && intuit_more(s)) {
5696 PL_tokenbuf[0] = '@';
5697 if (ckWARN(WARN_SYNTAX)) {
5700 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5703 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5704 while (t < PL_bufend && *t != ']')
5706 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5707 "Multidimensional syntax %.*s not supported",
5708 (int)((t - PL_bufptr) + 1), PL_bufptr);
5712 else if (*s == '{') {
5714 PL_tokenbuf[0] = '%';
5715 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5716 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5718 char tmpbuf[sizeof PL_tokenbuf];
5721 } while (isSPACE(*t));
5722 if (isIDFIRST_lazy_if(t,UTF)) {
5724 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5728 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5729 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5730 "You need to quote \"%s\"",
5737 PL_expect = XOPERATOR;
5738 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5739 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5740 if (!islop || PL_last_lop_op == OP_GREPSTART)
5741 PL_expect = XOPERATOR;
5742 else if (strchr("$@\"'`q", *s))
5743 PL_expect = XTERM; /* e.g. print $fh "foo" */
5744 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5745 PL_expect = XTERM; /* e.g. print $fh &sub */
5746 else if (isIDFIRST_lazy_if(s,UTF)) {
5747 char tmpbuf[sizeof PL_tokenbuf];
5749 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5750 if ((t2 = keyword(tmpbuf, len, 0))) {
5751 /* binary operators exclude handle interpretations */
5763 PL_expect = XTERM; /* e.g. print $fh length() */
5768 PL_expect = XTERM; /* e.g. print $fh subr() */
5771 else if (isDIGIT(*s))
5772 PL_expect = XTERM; /* e.g. print $fh 3 */
5773 else if (*s == '.' && isDIGIT(s[1]))
5774 PL_expect = XTERM; /* e.g. print $fh .3 */
5775 else if ((*s == '?' || *s == '-' || *s == '+')
5776 && !isSPACE(s[1]) && s[1] != '=')
5777 PL_expect = XTERM; /* e.g. print $fh -1 */
5778 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5780 PL_expect = XTERM; /* e.g. print $fh /.../
5781 XXX except DORDOR operator
5783 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5785 PL_expect = XTERM; /* print $fh <<"EOF" */
5788 PL_pending_ident = '$';
5792 if (PL_expect == XOPERATOR)
5794 PL_tokenbuf[0] = '@';
5795 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5796 if (!PL_tokenbuf[1]) {
5799 if (PL_lex_state == LEX_NORMAL)
5801 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5803 PL_tokenbuf[0] = '%';
5805 /* Warn about @ where they meant $. */
5806 if (*s == '[' || *s == '{') {
5807 if (ckWARN(WARN_SYNTAX)) {
5808 const char *t = s + 1;
5809 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5811 if (*t == '}' || *t == ']') {
5813 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5814 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5815 "Scalar value %.*s better written as $%.*s",
5816 (int)(t-PL_bufptr), PL_bufptr,
5817 (int)(t-PL_bufptr-1), PL_bufptr+1);
5822 PL_pending_ident = '@';
5825 case '/': /* may be division, defined-or, or pattern */
5826 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5830 case '?': /* may either be conditional or pattern */
5831 if (PL_expect == XOPERATOR) {
5839 /* A // operator. */
5849 /* Disable warning on "study /blah/" */
5850 if (PL_oldoldbufptr == PL_last_uni
5851 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5852 || memNE(PL_last_uni, "study", 5)
5853 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5856 s = scan_pat(s,OP_MATCH);
5857 TERM(sublex_start());
5861 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5862 #ifdef PERL_STRICT_CR
5865 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5867 && (s == PL_linestart || s[-1] == '\n') )
5869 PL_lex_formbrack = 0;
5873 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5877 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5883 pl_yylval.ival = OPf_SPECIAL;
5892 case '0': case '1': case '2': case '3': case '4':
5893 case '5': case '6': case '7': case '8': case '9':
5894 s = scan_num(s, &pl_yylval);
5895 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5896 if (PL_expect == XOPERATOR)
5901 s = scan_str(s,!!PL_madskills,FALSE);
5902 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5903 if (PL_expect == XOPERATOR) {
5904 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5905 return deprecate_commaless_var_list();
5912 pl_yylval.ival = OP_CONST;
5913 TERM(sublex_start());
5916 s = scan_str(s,!!PL_madskills,FALSE);
5917 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5918 if (PL_expect == XOPERATOR) {
5919 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5920 return deprecate_commaless_var_list();
5927 pl_yylval.ival = OP_CONST;
5928 /* FIXME. I think that this can be const if char *d is replaced by
5929 more localised variables. */
5930 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5931 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5932 pl_yylval.ival = OP_STRINGIFY;
5936 TERM(sublex_start());
5939 s = scan_str(s,!!PL_madskills,FALSE);
5940 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5941 if (PL_expect == XOPERATOR)
5942 no_op("Backticks",s);
5945 readpipe_override();
5946 TERM(sublex_start());
5950 if (PL_lex_inwhat && isDIGIT(*s))
5951 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5953 if (PL_expect == XOPERATOR)
5954 no_op("Backslash",s);
5958 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5959 char *start = s + 2;
5960 while (isDIGIT(*start) || *start == '_')
5962 if (*start == '.' && isDIGIT(start[1])) {
5963 s = scan_num(s, &pl_yylval);
5966 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5967 else if (!isALPHA(*start) && (PL_expect == XTERM
5968 || PL_expect == XREF || PL_expect == XSTATE
5969 || PL_expect == XTERMORDORDOR)) {
5970 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5972 s = scan_num(s, &pl_yylval);
5979 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6022 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6024 /* Some keywords can be followed by any delimiter, including ':' */
6025 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6026 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6027 (PL_tokenbuf[0] == 'q' &&
6028 strchr("qwxr", PL_tokenbuf[1])))));
6030 /* x::* is just a word, unless x is "CORE" */
6031 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6035 while (d < PL_bufend && isSPACE(*d))
6036 d++; /* no comments skipped here, or s### is misparsed */
6038 /* Is this a word before a => operator? */
6039 if (*d == '=' && d[1] == '>') {
6042 = (OP*)newSVOP(OP_CONST, 0,
6043 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6044 pl_yylval.opval->op_private = OPpCONST_BARE;
6048 /* Check for plugged-in keyword */
6052 char *saved_bufptr = PL_bufptr;
6054 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
6056 if (result == KEYWORD_PLUGIN_DECLINE) {
6057 /* not a plugged-in keyword */
6058 PL_bufptr = saved_bufptr;
6059 } else if (result == KEYWORD_PLUGIN_STMT) {
6060 pl_yylval.opval = o;
6063 return REPORT(PLUGSTMT);
6064 } else if (result == KEYWORD_PLUGIN_EXPR) {
6065 pl_yylval.opval = o;
6067 PL_expect = XOPERATOR;
6068 return REPORT(PLUGEXPR);
6070 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6075 /* Check for built-in keyword */
6076 tmp = keyword(PL_tokenbuf, len, 0);
6078 /* Is this a label? */
6079 if (!anydelim && PL_expect == XSTATE
6080 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6082 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
6084 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6089 if (tmp < 0) { /* second-class keyword? */
6090 GV *ogv = NULL; /* override (winner) */
6091 GV *hgv = NULL; /* hidden (loser) */
6092 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6094 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6097 if (GvIMPORTED_CV(gv))
6099 else if (! CvMETHOD(cv))
6103 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6104 (gv = *gvp) && isGV_with_GP(gv) &&
6105 GvCVu(gv) && GvIMPORTED_CV(gv))
6112 tmp = 0; /* overridden by import or by GLOBAL */
6115 && -tmp==KEY_lock /* XXX generalizable kludge */
6118 tmp = 0; /* any sub overrides "weak" keyword */
6120 else { /* no override */
6122 if (tmp == KEY_dump) {
6123 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6124 "dump() better written as CORE::dump()");
6128 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6129 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6130 "Ambiguous call resolved as CORE::%s(), %s",
6131 GvENAME(hgv), "qualify as such or use &");
6138 default: /* not a keyword */
6139 /* Trade off - by using this evil construction we can pull the
6140 variable gv into the block labelled keylookup. If not, then
6141 we have to give it function scope so that the goto from the
6142 earlier ':' case doesn't bypass the initialisation. */
6144 just_a_word_zero_gv:
6152 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6156 SV *nextPL_nextwhite = 0;
6160 /* Get the rest if it looks like a package qualifier */
6162 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6164 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6167 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6168 *s == '\'' ? "'" : "::");
6173 if (PL_expect == XOPERATOR) {
6174 if (PL_bufptr == PL_linestart) {
6175 CopLINE_dec(PL_curcop);
6176 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6177 CopLINE_inc(PL_curcop);
6180 no_op("Bareword",s);
6183 /* Look for a subroutine with this name in current package,
6184 unless name is "Foo::", in which case Foo is a bearword
6185 (and a package name). */
6187 if (len > 2 && !PL_madskills &&
6188 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6190 if (ckWARN(WARN_BAREWORD)
6191 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6192 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6193 "Bareword \"%s\" refers to nonexistent package",
6196 PL_tokenbuf[len] = '\0';
6202 /* Mustn't actually add anything to a symbol table.
6203 But also don't want to "initialise" any placeholder
6204 constants that might already be there into full
6205 blown PVGVs with attached PVCV. */
6206 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6207 GV_NOADD_NOINIT, SVt_PVCV);
6212 /* if we saw a global override before, get the right name */
6215 sv = newSVpvs("CORE::GLOBAL::");
6216 sv_catpv(sv,PL_tokenbuf);
6219 /* If len is 0, newSVpv does strlen(), which is correct.
6220 If len is non-zero, then it will be the true length,
6221 and so the scalar will be created correctly. */
6222 sv = newSVpv(PL_tokenbuf,len);
6225 if (PL_madskills && !PL_thistoken) {
6226 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6227 PL_thistoken = newSVpvn(start,s - start);
6228 PL_realtokenstart = s - SvPVX(PL_linestr);
6232 /* Presume this is going to be a bareword of some sort. */
6235 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6236 pl_yylval.opval->op_private = OPpCONST_BARE;
6237 /* UTF-8 package name? */
6238 if (UTF && !IN_BYTES &&
6239 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6242 /* And if "Foo::", then that's what it certainly is. */
6249 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6250 const_op->op_private = OPpCONST_BARE;
6251 rv2cv_op = newCVREF(0, const_op);
6253 if (rv2cv_op->op_type == OP_RV2CV &&
6254 (rv2cv_op->op_flags & OPf_KIDS)) {
6255 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6256 switch (rv_op->op_type) {
6258 SV *sv = cSVOPx_sv(rv_op);
6259 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6263 GV *gv = cGVOPx_gv(rv_op);
6264 CV *maybe_cv = GvCVu(gv);
6265 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6271 /* See if it's the indirect object for a list operator. */
6273 if (PL_oldoldbufptr &&
6274 PL_oldoldbufptr < PL_bufptr &&
6275 (PL_oldoldbufptr == PL_last_lop
6276 || PL_oldoldbufptr == PL_last_uni) &&
6277 /* NO SKIPSPACE BEFORE HERE! */
6278 (PL_expect == XREF ||
6279 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6281 bool immediate_paren = *s == '(';
6283 /* (Now we can afford to cross potential line boundary.) */
6284 s = SKIPSPACE2(s,nextPL_nextwhite);
6286 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6289 /* Two barewords in a row may indicate method call. */
6291 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6292 (tmp = intuit_method(s, gv, cv))) {
6297 /* If not a declared subroutine, it's an indirect object. */
6298 /* (But it's an indir obj regardless for sort.) */
6299 /* Also, if "_" follows a filetest operator, it's a bareword */
6302 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6304 (PL_last_lop_op != OP_MAPSTART &&
6305 PL_last_lop_op != OP_GREPSTART))))
6306 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6307 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6310 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6315 PL_expect = XOPERATOR;
6318 s = SKIPSPACE2(s,nextPL_nextwhite);
6319 PL_nextwhite = nextPL_nextwhite;
6324 /* Is this a word before a => operator? */
6325 if (*s == '=' && s[1] == '>' && !pkgname) {
6328 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6329 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6330 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6334 /* If followed by a paren, it's certainly a subroutine. */
6339 while (SPACE_OR_TAB(*d))
6341 if (*d == ')' && (sv = cv_const_sv(cv))) {
6348 PL_nextwhite = PL_thiswhite;
6351 start_force(PL_curforce);
6353 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6354 PL_expect = XOPERATOR;
6357 PL_nextwhite = nextPL_nextwhite;
6358 curmad('X', PL_thistoken);
6359 PL_thistoken = newSVpvs("");
6368 /* If followed by var or block, call it a method (unless sub) */
6370 if ((*s == '$' || *s == '{') && !cv) {
6372 PL_last_lop = PL_oldbufptr;
6373 PL_last_lop_op = OP_METHOD;
6377 /* If followed by a bareword, see if it looks like indir obj. */
6380 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6381 && (tmp = intuit_method(s, gv, cv))) {
6386 /* Not a method, so call it a subroutine (if defined) */
6389 if (lastchar == '-')
6390 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6391 "Ambiguous use of -%s resolved as -&%s()",
6392 PL_tokenbuf, PL_tokenbuf);
6393 /* Check for a constant sub */
6394 if ((sv = cv_const_sv(cv))) {
6397 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6398 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6399 pl_yylval.opval->op_private = 0;
6403 op_free(pl_yylval.opval);
6404 pl_yylval.opval = rv2cv_op;
6405 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6406 PL_last_lop = PL_oldbufptr;
6407 PL_last_lop_op = OP_ENTERSUB;
6408 /* Is there a prototype? */
6416 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6419 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6421 while (*proto == ';')
6423 if (*proto == '&' && *s == '{') {
6425 sv_setpvs(PL_subname, "__ANON__");
6427 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6434 PL_nextwhite = PL_thiswhite;
6437 start_force(PL_curforce);
6438 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6441 PL_nextwhite = nextPL_nextwhite;
6442 curmad('X', PL_thistoken);
6443 PL_thistoken = newSVpvs("");
6450 /* Guess harder when madskills require "best effort". */
6451 if (PL_madskills && (!gv || !GvCVu(gv))) {
6452 int probable_sub = 0;
6453 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6455 else if (isALPHA(*s)) {
6459 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6460 if (!keyword(tmpbuf, tmplen, 0))
6463 while (d < PL_bufend && isSPACE(*d))
6465 if (*d == '=' && d[1] == '>')
6470 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6471 op_free(pl_yylval.opval);
6472 pl_yylval.opval = rv2cv_op;
6473 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6474 PL_last_lop = PL_oldbufptr;
6475 PL_last_lop_op = OP_ENTERSUB;
6476 PL_nextwhite = PL_thiswhite;
6478 start_force(PL_curforce);
6479 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6481 PL_nextwhite = nextPL_nextwhite;
6482 curmad('X', PL_thistoken);
6483 PL_thistoken = newSVpvs("");
6488 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6495 /* Call it a bare word */
6497 if (PL_hints & HINT_STRICT_SUBS)
6498 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6501 /* after "print" and similar functions (corresponding to
6502 * "F? L" in opcode.pl), whatever wasn't already parsed as
6503 * a filehandle should be subject to "strict subs".
6504 * Likewise for the optional indirect-object argument to system
6505 * or exec, which can't be a bareword */
6506 if ((PL_last_lop_op == OP_PRINT
6507 || PL_last_lop_op == OP_PRTF
6508 || PL_last_lop_op == OP_SAY
6509 || PL_last_lop_op == OP_SYSTEM
6510 || PL_last_lop_op == OP_EXEC)
6511 && (PL_hints & HINT_STRICT_SUBS))
6512 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6513 if (lastchar != '-') {
6514 if (ckWARN(WARN_RESERVED)) {
6518 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6519 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6527 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6528 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6529 "Operator or semicolon missing before %c%s",
6530 lastchar, PL_tokenbuf);
6531 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6532 "Ambiguous use of %c resolved as operator %c",
6533 lastchar, lastchar);
6539 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6540 newSVpv(CopFILE(PL_curcop),0));
6544 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6545 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6548 case KEY___PACKAGE__:
6549 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6551 ? newSVhek(HvNAME_HEK(PL_curstash))
6558 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6559 const char *pname = "main";
6560 if (PL_tokenbuf[2] == 'D')
6561 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6562 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6566 GvIOp(gv) = newIO();
6567 IoIFP(GvIOp(gv)) = PL_rsfp;
6568 #if defined(HAS_FCNTL) && defined(F_SETFD)
6570 const int fd = PerlIO_fileno(PL_rsfp);
6571 fcntl(fd,F_SETFD,fd >= 3);
6574 /* Mark this internal pseudo-handle as clean */
6575 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6576 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6577 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6579 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6580 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6581 /* if the script was opened in binmode, we need to revert
6582 * it to text mode for compatibility; but only iff it has CRs
6583 * XXX this is a questionable hack at best. */
6584 if (PL_bufend-PL_bufptr > 2
6585 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6588 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6589 loc = PerlIO_tell(PL_rsfp);
6590 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6593 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6595 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6596 #endif /* NETWARE */
6597 #ifdef PERLIO_IS_STDIO /* really? */
6598 # if defined(__BORLANDC__)
6599 /* XXX see note in do_binmode() */
6600 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6604 PerlIO_seek(PL_rsfp, loc, 0);
6608 #ifdef PERLIO_LAYERS
6611 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6612 else if (PL_encoding) {
6619 XPUSHs(PL_encoding);
6621 call_method("name", G_SCALAR);
6625 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6626 Perl_form(aTHX_ ":encoding(%"SVf")",
6635 if (PL_realtokenstart >= 0) {
6636 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6638 PL_endwhite = newSVpvs("");
6639 sv_catsv(PL_endwhite, PL_thiswhite);
6641 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6642 PL_realtokenstart = -1;
6644 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6660 if (PL_expect == XSTATE) {
6667 if (*s == ':' && s[1] == ':') {
6670 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6671 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6672 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6675 else if (tmp == KEY_require || tmp == KEY_do)
6676 /* that's a way to remember we saw "CORE::" */
6689 LOP(OP_ACCEPT,XTERM);
6695 LOP(OP_ATAN2,XTERM);
6701 LOP(OP_BINMODE,XTERM);
6704 LOP(OP_BLESS,XTERM);
6713 /* When 'use switch' is in effect, continue has a dual
6714 life as a control operator. */
6716 if (!FEATURE_IS_ENABLED("switch"))
6719 /* We have to disambiguate the two senses of
6720 "continue". If the next token is a '{' then
6721 treat it as the start of a continue block;
6722 otherwise treat it as a control operator.
6734 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6751 if (!PL_cryptseen) {
6752 PL_cryptseen = TRUE;
6756 LOP(OP_CRYPT,XTERM);
6759 LOP(OP_CHMOD,XTERM);
6762 LOP(OP_CHOWN,XTERM);
6765 LOP(OP_CONNECT,XTERM);
6784 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6785 if (orig_keyword == KEY_do) {
6794 PL_hints |= HINT_BLOCK_SCOPE;
6804 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6805 LOP(OP_DBMOPEN,XTERM);
6811 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6818 pl_yylval.ival = CopLINE(PL_curcop);
6834 if (*s == '{') { /* block eval */
6835 PL_expect = XTERMBLOCK;
6836 UNIBRACK(OP_ENTERTRY);
6838 else { /* string eval */
6840 UNIBRACK(OP_ENTEREVAL);
6855 case KEY_endhostent:
6861 case KEY_endservent:
6864 case KEY_endprotoent:
6875 pl_yylval.ival = CopLINE(PL_curcop);
6877 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6880 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6883 if ((PL_bufend - p) >= 3 &&
6884 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6886 else if ((PL_bufend - p) >= 4 &&
6887 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6890 if (isIDFIRST_lazy_if(p,UTF)) {
6891 p = scan_ident(p, PL_bufend,
6892 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6896 Perl_croak(aTHX_ "Missing $ on loop variable");
6898 s = SvPVX(PL_linestr) + soff;
6904 LOP(OP_FORMLINE,XTERM);
6910 LOP(OP_FCNTL,XTERM);
6916 LOP(OP_FLOCK,XTERM);
6925 LOP(OP_GREPSTART, XREF);
6928 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6943 case KEY_getpriority:
6944 LOP(OP_GETPRIORITY,XTERM);
6946 case KEY_getprotobyname:
6949 case KEY_getprotobynumber:
6950 LOP(OP_GPBYNUMBER,XTERM);
6952 case KEY_getprotoent:
6964 case KEY_getpeername:
6965 UNI(OP_GETPEERNAME);
6967 case KEY_gethostbyname:
6970 case KEY_gethostbyaddr:
6971 LOP(OP_GHBYADDR,XTERM);
6973 case KEY_gethostent:
6976 case KEY_getnetbyname:
6979 case KEY_getnetbyaddr:
6980 LOP(OP_GNBYADDR,XTERM);
6985 case KEY_getservbyname:
6986 LOP(OP_GSBYNAME,XTERM);
6988 case KEY_getservbyport:
6989 LOP(OP_GSBYPORT,XTERM);
6991 case KEY_getservent:
6994 case KEY_getsockname:
6995 UNI(OP_GETSOCKNAME);
6997 case KEY_getsockopt:
6998 LOP(OP_GSOCKOPT,XTERM);
7013 pl_yylval.ival = CopLINE(PL_curcop);
7023 pl_yylval.ival = CopLINE(PL_curcop);
7027 LOP(OP_INDEX,XTERM);
7033 LOP(OP_IOCTL,XTERM);
7045 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7077 LOP(OP_LISTEN,XTERM);
7086 s = scan_pat(s,OP_MATCH);
7087 TERM(sublex_start());
7090 LOP(OP_MAPSTART, XREF);
7093 LOP(OP_MKDIR,XTERM);
7096 LOP(OP_MSGCTL,XTERM);
7099 LOP(OP_MSGGET,XTERM);
7102 LOP(OP_MSGRCV,XTERM);
7105 LOP(OP_MSGSND,XTERM);
7110 PL_in_my = (U16)tmp;
7112 if (isIDFIRST_lazy_if(s,UTF)) {
7116 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7117 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7119 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7120 if (!PL_in_my_stash) {
7123 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7127 if (PL_madskills) { /* just add type to declarator token */
7128 sv_catsv(PL_thistoken, PL_nextwhite);
7130 sv_catpvn(PL_thistoken, start, s - start);
7138 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7145 s = tokenize_use(0, s);
7149 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7156 if (isIDFIRST_lazy_if(s,UTF)) {
7158 for (d = s; isALNUM_lazy_if(d,UTF);)
7160 for (t=d; isSPACE(*t);)
7162 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7164 && !(t[0] == '=' && t[1] == '>')
7166 int parms_len = (int)(d-s);
7167 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7168 "Precedence problem: open %.*s should be open(%.*s)",
7169 parms_len, s, parms_len, s);
7175 pl_yylval.ival = OP_OR;
7185 LOP(OP_OPEN_DIR,XTERM);
7188 checkcomma(s,PL_tokenbuf,"filehandle");
7192 checkcomma(s,PL_tokenbuf,"filehandle");
7211 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7213 s = force_strict_version(s);
7217 LOP(OP_PIPE_OP,XTERM);
7220 s = scan_str(s,!!PL_madskills,FALSE);
7223 pl_yylval.ival = OP_CONST;
7224 TERM(sublex_start());
7230 s = scan_str(s,!!PL_madskills,FALSE);
7233 PL_expect = XOPERATOR;
7235 if (SvCUR(PL_lex_stuff)) {
7238 d = SvPV_force(PL_lex_stuff, len);
7240 for (; isSPACE(*d) && len; --len, ++d)
7245 if (!warned && ckWARN(WARN_QW)) {
7246 for (; !isSPACE(*d) && len; --len, ++d) {
7248 Perl_warner(aTHX_ packWARN(WARN_QW),
7249 "Possible attempt to separate words with commas");
7252 else if (*d == '#') {
7253 Perl_warner(aTHX_ packWARN(WARN_QW),
7254 "Possible attempt to put comments in qw() list");
7260 for (; !isSPACE(*d) && len; --len, ++d)
7263 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7264 words = append_elem(OP_LIST, words,
7265 newSVOP(OP_CONST, 0, tokeq(sv)));
7269 start_force(PL_curforce);
7270 NEXTVAL_NEXTTOKE.opval = words;
7275 SvREFCNT_dec(PL_lex_stuff);
7276 PL_lex_stuff = NULL;
7282 s = scan_str(s,!!PL_madskills,FALSE);
7285 pl_yylval.ival = OP_STRINGIFY;
7286 if (SvIVX(PL_lex_stuff) == '\'')
7287 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7288 TERM(sublex_start());
7291 s = scan_pat(s,OP_QR);
7292 TERM(sublex_start());
7295 s = scan_str(s,!!PL_madskills,FALSE);
7298 readpipe_override();
7299 TERM(sublex_start());
7307 s = force_version(s, FALSE);
7309 else if (*s != 'v' || !isDIGIT(s[1])
7310 || (s = force_version(s, TRUE), *s == 'v'))
7312 *PL_tokenbuf = '\0';
7313 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7314 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7315 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7317 yyerror("<> should be quotes");
7319 if (orig_keyword == KEY_require) {
7327 PL_last_uni = PL_oldbufptr;
7328 PL_last_lop_op = OP_REQUIRE;
7330 return REPORT( (int)REQUIRE );
7336 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7340 LOP(OP_RENAME,XTERM);
7349 LOP(OP_RINDEX,XTERM);
7358 UNIDOR(OP_READLINE);
7361 UNIDOR(OP_BACKTICK);
7370 LOP(OP_REVERSE,XTERM);
7373 UNIDOR(OP_READLINK);
7380 if (pl_yylval.opval)
7381 TERM(sublex_start());
7383 TOKEN(1); /* force error */
7386 checkcomma(s,PL_tokenbuf,"filehandle");
7396 LOP(OP_SELECT,XTERM);
7402 LOP(OP_SEMCTL,XTERM);
7405 LOP(OP_SEMGET,XTERM);
7408 LOP(OP_SEMOP,XTERM);
7414 LOP(OP_SETPGRP,XTERM);
7416 case KEY_setpriority:
7417 LOP(OP_SETPRIORITY,XTERM);
7419 case KEY_sethostent:
7425 case KEY_setservent:
7428 case KEY_setprotoent:
7438 LOP(OP_SEEKDIR,XTERM);
7440 case KEY_setsockopt:
7441 LOP(OP_SSOCKOPT,XTERM);
7447 LOP(OP_SHMCTL,XTERM);
7450 LOP(OP_SHMGET,XTERM);
7453 LOP(OP_SHMREAD,XTERM);
7456 LOP(OP_SHMWRITE,XTERM);
7459 LOP(OP_SHUTDOWN,XTERM);
7468 LOP(OP_SOCKET,XTERM);
7470 case KEY_socketpair:
7471 LOP(OP_SOCKPAIR,XTERM);
7474 checkcomma(s,PL_tokenbuf,"subroutine name");
7476 if (*s == ';' || *s == ')') /* probably a close */
7477 Perl_croak(aTHX_ "sort is now a reserved word");
7479 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7483 LOP(OP_SPLIT,XTERM);
7486 LOP(OP_SPRINTF,XTERM);
7489 LOP(OP_SPLICE,XTERM);
7504 LOP(OP_SUBSTR,XTERM);
7510 char tmpbuf[sizeof PL_tokenbuf];
7511 SSize_t tboffset = 0;
7512 expectation attrful;
7513 bool have_name, have_proto;
7514 const int key = tmp;
7519 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7520 SV *subtoken = newSVpvn(tstart, s - tstart);
7524 s = SKIPSPACE2(s,tmpwhite);
7529 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7530 (*s == ':' && s[1] == ':'))
7533 SV *nametoke = NULL;
7537 attrful = XATTRBLOCK;
7538 /* remember buffer pos'n for later force_word */
7539 tboffset = s - PL_oldbufptr;
7540 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7543 nametoke = newSVpvn(s, d - s);
7545 if (memchr(tmpbuf, ':', len))
7546 sv_setpvn(PL_subname, tmpbuf, len);
7548 sv_setsv(PL_subname,PL_curstname);
7549 sv_catpvs(PL_subname,"::");
7550 sv_catpvn(PL_subname,tmpbuf,len);
7557 CURMAD('X', nametoke);
7558 CURMAD('_', tmpwhite);
7559 (void) force_word(PL_oldbufptr + tboffset, WORD,
7562 s = SKIPSPACE2(d,tmpwhite);
7569 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7570 PL_expect = XTERMBLOCK;
7571 attrful = XATTRTERM;
7572 sv_setpvs(PL_subname,"?");
7576 if (key == KEY_format) {
7578 PL_lex_formbrack = PL_lex_brackets + 1;
7580 PL_thistoken = subtoken;
7584 (void) force_word(PL_oldbufptr + tboffset, WORD,
7590 /* Look for a prototype */
7593 bool bad_proto = FALSE;
7594 bool in_brackets = FALSE;
7595 char greedy_proto = ' ';
7596 bool proto_after_greedy_proto = FALSE;
7597 bool must_be_last = FALSE;
7598 bool underscore = FALSE;
7599 bool seen_underscore = FALSE;
7600 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7602 s = scan_str(s,!!PL_madskills,FALSE);
7604 Perl_croak(aTHX_ "Prototype not terminated");
7605 /* strip spaces and check for bad characters */
7606 d = SvPVX(PL_lex_stuff);
7608 for (p = d; *p; ++p) {
7612 if (warnillegalproto) {
7614 proto_after_greedy_proto = TRUE;
7615 if (!strchr("$@%*;[]&\\_", *p)) {
7627 else if ( *p == ']' ) {
7628 in_brackets = FALSE;
7630 else if ( (*p == '@' || *p == '%') &&
7631 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7633 must_be_last = TRUE;
7636 else if ( *p == '_' ) {
7637 underscore = seen_underscore = TRUE;
7644 if (proto_after_greedy_proto)
7645 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7646 "Prototype after '%c' for %"SVf" : %s",
7647 greedy_proto, SVfARG(PL_subname), d);
7649 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7650 "Illegal character %sin prototype for %"SVf" : %s",
7651 seen_underscore ? "after '_' " : "",
7652 SVfARG(PL_subname), d);
7653 SvCUR_set(PL_lex_stuff, tmp);
7658 CURMAD('q', PL_thisopen);
7659 CURMAD('_', tmpwhite);
7660 CURMAD('=', PL_thisstuff);
7661 CURMAD('Q', PL_thisclose);
7662 NEXTVAL_NEXTTOKE.opval =
7663 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7664 PL_lex_stuff = NULL;
7667 s = SKIPSPACE2(s,tmpwhite);
7675 if (*s == ':' && s[1] != ':')
7676 PL_expect = attrful;
7677 else if (*s != '{' && key == KEY_sub) {
7679 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7680 else if (*s != ';' && *s != '}')
7681 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7688 curmad('^', newSVpvs(""));
7689 CURMAD('_', tmpwhite);
7693 PL_thistoken = subtoken;
7696 NEXTVAL_NEXTTOKE.opval =
7697 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7698 PL_lex_stuff = NULL;
7704 sv_setpvs(PL_subname, "__ANON__");
7706 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7710 (void) force_word(PL_oldbufptr + tboffset, WORD,
7719 LOP(OP_SYSTEM,XREF);
7722 LOP(OP_SYMLINK,XTERM);
7725 LOP(OP_SYSCALL,XTERM);
7728 LOP(OP_SYSOPEN,XTERM);
7731 LOP(OP_SYSSEEK,XTERM);
7734 LOP(OP_SYSREAD,XTERM);
7737 LOP(OP_SYSWRITE,XTERM);
7741 TERM(sublex_start());
7762 LOP(OP_TRUNCATE,XTERM);
7774 pl_yylval.ival = CopLINE(PL_curcop);
7778 pl_yylval.ival = CopLINE(PL_curcop);
7782 LOP(OP_UNLINK,XTERM);
7788 LOP(OP_UNPACK,XTERM);
7791 LOP(OP_UTIME,XTERM);
7797 LOP(OP_UNSHIFT,XTERM);
7800 s = tokenize_use(1, s);
7810 pl_yylval.ival = CopLINE(PL_curcop);
7814 pl_yylval.ival = CopLINE(PL_curcop);
7818 PL_hints |= HINT_BLOCK_SCOPE;
7825 LOP(OP_WAITPID,XTERM);
7834 ctl_l[0] = toCTRL('L');
7836 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7839 /* Make sure $^L is defined */
7840 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7845 if (PL_expect == XOPERATOR)
7851 pl_yylval.ival = OP_XOR;
7856 TERM(sublex_start());
7861 #pragma segment Main
7865 S_pending_ident(pTHX)
7870 /* pit holds the identifier we read and pending_ident is reset */
7871 char pit = PL_pending_ident;
7872 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7873 /* All routes through this function want to know if there is a colon. */
7874 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7875 PL_pending_ident = 0;
7877 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7878 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7879 "### Pending identifier '%s'\n", PL_tokenbuf); });
7881 /* if we're in a my(), we can't allow dynamics here.
7882 $foo'bar has already been turned into $foo::bar, so
7883 just check for colons.
7885 if it's a legal name, the OP is a PADANY.
7888 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7890 yyerror(Perl_form(aTHX_ "No package name allowed for "
7891 "variable %s in \"our\"",
7893 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7897 yyerror(Perl_form(aTHX_ PL_no_myglob,
7898 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7900 pl_yylval.opval = newOP(OP_PADANY, 0);
7901 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7907 build the ops for accesses to a my() variable.
7909 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7910 then used in a comparison. This catches most, but not
7911 all cases. For instance, it catches
7912 sort { my($a); $a <=> $b }
7914 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7915 (although why you'd do that is anyone's guess).
7920 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7921 if (tmp != NOT_IN_PAD) {
7922 /* might be an "our" variable" */
7923 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7924 /* build ops for a bareword */
7925 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7926 HEK * const stashname = HvNAME_HEK(stash);
7927 SV * const sym = newSVhek(stashname);
7928 sv_catpvs(sym, "::");
7929 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7930 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7931 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7934 ? (GV_ADDMULTI | GV_ADDINEVAL)
7937 ((PL_tokenbuf[0] == '$') ? SVt_PV
7938 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7943 /* if it's a sort block and they're naming $a or $b */
7944 if (PL_last_lop_op == OP_SORT &&
7945 PL_tokenbuf[0] == '$' &&
7946 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7949 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7950 d < PL_bufend && *d != '\n';
7953 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7954 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7960 pl_yylval.opval = newOP(OP_PADANY, 0);
7961 pl_yylval.opval->op_targ = tmp;
7967 Whine if they've said @foo in a doublequoted string,
7968 and @foo isn't a variable we can find in the symbol
7971 if (ckWARN(WARN_AMBIGUOUS) &&
7972 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7973 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7975 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7976 /* DO NOT warn for @- and @+ */
7977 && !( PL_tokenbuf[2] == '\0' &&
7978 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7981 /* Downgraded from fatal to warning 20000522 mjd */
7982 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7983 "Possible unintended interpolation of %s in string",
7988 /* build ops for a bareword */
7989 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7991 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7993 PL_tokenbuf + 1, tokenbuf_len - 1,
7994 /* If the identifier refers to a stash, don't autovivify it.
7995 * Change 24660 had the side effect of causing symbol table
7996 * hashes to always be defined, even if they were freshly
7997 * created and the only reference in the entire program was
7998 * the single statement with the defined %foo::bar:: test.
7999 * It appears that all code in the wild doing this actually
8000 * wants to know whether sub-packages have been loaded, so
8001 * by avoiding auto-vivifying symbol tables, we ensure that
8002 * defined %foo::bar:: continues to be false, and the existing
8003 * tests still give the expected answers, even though what
8004 * they're actually testing has now changed subtly.
8006 (*PL_tokenbuf == '%'
8007 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
8010 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
8011 ((PL_tokenbuf[0] == '$') ? SVt_PV
8012 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8018 * The following code was generated by perl_keyword.pl.
8022 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8026 PERL_ARGS_ASSERT_KEYWORD;
8030 case 1: /* 5 tokens of length 1 */
8062 case 2: /* 18 tokens of length 2 */
8208 case 3: /* 29 tokens of length 3 */
8212 if (name[1] == 'N' &&
8275 if (name[1] == 'i' &&
8307 if (name[1] == 'o' &&
8316 if (name[1] == 'e' &&
8325 if (name[1] == 'n' &&
8334 if (name[1] == 'o' &&
8343 if (name[1] == 'a' &&
8352 if (name[1] == 'o' &&
8414 if (name[1] == 'e' &&
8428 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8454 if (name[1] == 'i' &&
8463 if (name[1] == 's' &&
8472 if (name[1] == 'e' &&
8481 if (name[1] == 'o' &&
8493 case 4: /* 41 tokens of length 4 */
8497 if (name[1] == 'O' &&
8507 if (name[1] == 'N' &&
8517 if (name[1] == 'i' &&
8527 if (name[1] == 'h' &&
8537 if (name[1] == 'u' &&
8550 if (name[2] == 'c' &&
8559 if (name[2] == 's' &&
8568 if (name[2] == 'a' &&
8604 if (name[1] == 'o' &&
8617 if (name[2] == 't' &&
8626 if (name[2] == 'o' &&
8635 if (name[2] == 't' &&
8644 if (name[2] == 'e' &&
8657 if (name[1] == 'o' &&
8670 if (name[2] == 'y' &&
8679 if (name[2] == 'l' &&
8695 if (name[2] == 's' &&
8704 if (name[2] == 'n' &&
8713 if (name[2] == 'c' &&
8726 if (name[1] == 'e' &&
8736 if (name[1] == 'p' &&
8749 if (name[2] == 'c' &&
8758 if (name[2] == 'p' &&
8767 if (name[2] == 's' &&
8783 if (name[2] == 'n' &&
8853 if (name[2] == 'r' &&
8862 if (name[2] == 'r' &&
8871 if (name[2] == 'a' &&
8887 if (name[2] == 'l' &&
8949 if (name[2] == 'e' &&
8952 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8965 case 5: /* 39 tokens of length 5 */
8969 if (name[1] == 'E' &&
8980 if (name[1] == 'H' &&
8994 if (name[2] == 'a' &&
9004 if (name[2] == 'a' &&
9021 if (name[2] == 'e' &&
9031 if (name[2] == 'e' &&
9035 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9051 if (name[3] == 'i' &&
9060 if (name[3] == 'o' &&
9096 if (name[2] == 'o' &&
9106 if (name[2] == 'y' &&
9120 if (name[1] == 'l' &&
9134 if (name[2] == 'n' &&
9144 if (name[2] == 'o' &&
9158 if (name[1] == 'i' &&
9163 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9172 if (name[2] == 'd' &&
9182 if (name[2] == 'c' &&
9199 if (name[2] == 'c' &&
9209 if (name[2] == 't' &&
9223 if (name[1] == 'k' &&
9234 if (name[1] == 'r' &&
9248 if (name[2] == 's' &&
9258 if (name[2] == 'd' &&
9275 if (name[2] == 'm' &&
9285 if (name[2] == 'i' &&
9295 if (name[2] == 'e' &&
9305 if (name[2] == 'l' &&
9315 if (name[2] == 'a' &&
9328 if (name[3] == 't' &&
9331 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9337 if (name[3] == 'd' &&
9354 if (name[1] == 'i' &&
9368 if (name[2] == 'a' &&
9381 if (name[3] == 'e' &&
9416 if (name[2] == 'i' &&
9433 if (name[2] == 'i' &&
9443 if (name[2] == 'i' &&
9460 case 6: /* 33 tokens of length 6 */
9464 if (name[1] == 'c' &&
9479 if (name[2] == 'l' &&
9490 if (name[2] == 'r' &&
9505 if (name[1] == 'e' &&
9520 if (name[2] == 's' &&
9525 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9531 if (name[2] == 'i' &&
9549 if (name[2] == 'l' &&
9560 if (name[2] == 'r' &&
9575 if (name[1] == 'm' &&
9590 if (name[2] == 'n' &&
9601 if (name[2] == 's' &&
9616 if (name[1] == 's' &&
9622 if (name[4] == 't' &&
9631 if (name[4] == 'e' &&
9640 if (name[4] == 'c' &&
9649 if (name[4] == 'n' &&
9665 if (name[1] == 'r' &&
9683 if (name[3] == 'a' &&
9693 if (name[3] == 'u' &&
9707 if (name[2] == 'n' &&
9725 if (name[2] == 'a' &&
9739 if (name[3] == 'e' &&
9752 if (name[4] == 't' &&
9761 if (name[4] == 'e' &&
9783 if (name[4] == 't' &&
9792 if (name[4] == 'e' &&
9808 if (name[2] == 'c' &&
9819 if (name[2] == 'l' &&
9830 if (name[2] == 'b' &&
9841 if (name[2] == 's' &&
9864 if (name[4] == 's' &&
9873 if (name[4] == 'n' &&
9886 if (name[3] == 'a' &&
9903 if (name[1] == 'a' &&
9918 case 7: /* 29 tokens of length 7 */
9922 if (name[1] == 'E' &&
9935 if (name[1] == '_' &&
9948 if (name[1] == 'i' &&
9955 return -KEY_binmode;
9961 if (name[1] == 'o' &&
9968 return -KEY_connect;
9977 if (name[2] == 'm' &&
9983 return -KEY_dbmopen;
9994 if (name[4] == 'u' &&
9998 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10004 if (name[4] == 'n' &&
10008 return KEY_defined;
10025 if (name[1] == 'o' &&
10032 return KEY_foreach;
10038 if (name[1] == 'e' &&
10045 if (name[5] == 'r' &&
10048 return -KEY_getpgrp;
10054 if (name[5] == 'i' &&
10057 return -KEY_getppid;
10070 if (name[1] == 'c' &&
10077 return -KEY_lcfirst;
10083 if (name[1] == 'p' &&
10090 return -KEY_opendir;
10096 if (name[1] == 'a' &&
10103 return KEY_package;
10109 if (name[1] == 'e')
10114 if (name[3] == 'd' &&
10119 return -KEY_readdir;
10125 if (name[3] == 'u' &&
10130 return KEY_require;
10136 if (name[3] == 'e' &&
10141 return -KEY_reverse;
10160 if (name[3] == 'k' &&
10165 return -KEY_seekdir;
10171 if (name[3] == 'p' &&
10176 return -KEY_setpgrp;
10186 if (name[2] == 'm' &&
10192 return -KEY_shmread;
10198 if (name[2] == 'r' &&
10204 return -KEY_sprintf;
10213 if (name[3] == 'l' &&
10218 return -KEY_symlink;
10227 if (name[4] == 'a' &&
10231 return -KEY_syscall;
10237 if (name[4] == 'p' &&
10241 return -KEY_sysopen;
10247 if (name[4] == 'e' &&
10251 return -KEY_sysread;
10257 if (name[4] == 'e' &&
10261 return -KEY_sysseek;
10279 if (name[1] == 'e' &&
10286 return -KEY_telldir;
10295 if (name[2] == 'f' &&
10301 return -KEY_ucfirst;
10307 if (name[2] == 's' &&
10313 return -KEY_unshift;
10323 if (name[1] == 'a' &&
10330 return -KEY_waitpid;
10339 case 8: /* 26 tokens of length 8 */
10343 if (name[1] == 'U' &&
10351 return KEY_AUTOLOAD;
10357 if (name[1] == '_')
10362 if (name[3] == 'A' &&
10368 return KEY___DATA__;
10374 if (name[3] == 'I' &&
10380 return -KEY___FILE__;
10386 if (name[3] == 'I' &&
10392 return -KEY___LINE__;
10408 if (name[2] == 'o' &&
10415 return -KEY_closedir;
10421 if (name[2] == 'n' &&
10428 return -KEY_continue;
10438 if (name[1] == 'b' &&
10446 return -KEY_dbmclose;
10452 if (name[1] == 'n' &&
10458 if (name[4] == 'r' &&
10463 return -KEY_endgrent;
10469 if (name[4] == 'w' &&
10474 return -KEY_endpwent;
10487 if (name[1] == 'o' &&
10495 return -KEY_formline;
10501 if (name[1] == 'e' &&
10507 if (name[4] == 'r')
10512 if (name[6] == 'n' &&
10515 return -KEY_getgrent;
10521 if (name[6] == 'i' &&
10524 return -KEY_getgrgid;
10530 if (name[6] == 'a' &&
10533 return -KEY_getgrnam;
10546 if (name[4] == 'o' &&
10551 return -KEY_getlogin;
10557 if (name[4] == 'w')
10562 if (name[6] == 'n' &&
10565 return -KEY_getpwent;
10571 if (name[6] == 'a' &&
10574 return -KEY_getpwnam;
10580 if (name[6] == 'i' &&
10583 return -KEY_getpwuid;
10603 if (name[1] == 'e' &&
10610 if (name[5] == 'i' &&
10617 return -KEY_readline;
10622 return -KEY_readlink;
10633 if (name[5] == 'i' &&
10637 return -KEY_readpipe;
10653 if (name[2] == 't')
10658 if (name[4] == 'r' &&
10663 return -KEY_setgrent;
10669 if (name[4] == 'w' &&
10674 return -KEY_setpwent;
10690 if (name[3] == 'w' &&
10696 return -KEY_shmwrite;
10702 if (name[3] == 't' &&
10708 return -KEY_shutdown;
10718 if (name[2] == 's' &&
10725 return -KEY_syswrite;
10735 if (name[1] == 'r' &&
10743 return -KEY_truncate;
10752 case 9: /* 9 tokens of length 9 */
10756 if (name[1] == 'N' &&
10765 return KEY_UNITCHECK;
10771 if (name[1] == 'n' &&
10780 return -KEY_endnetent;
10786 if (name[1] == 'e' &&
10795 return -KEY_getnetent;
10801 if (name[1] == 'o' &&
10810 return -KEY_localtime;
10816 if (name[1] == 'r' &&
10825 return KEY_prototype;
10831 if (name[1] == 'u' &&
10840 return -KEY_quotemeta;
10846 if (name[1] == 'e' &&
10855 return -KEY_rewinddir;
10861 if (name[1] == 'e' &&
10870 return -KEY_setnetent;
10876 if (name[1] == 'a' &&
10885 return -KEY_wantarray;
10894 case 10: /* 9 tokens of length 10 */
10898 if (name[1] == 'n' &&
10904 if (name[4] == 'o' &&
10911 return -KEY_endhostent;
10917 if (name[4] == 'e' &&
10924 return -KEY_endservent;
10937 if (name[1] == 'e' &&
10943 if (name[4] == 'o' &&
10950 return -KEY_gethostent;
10959 if (name[5] == 'r' &&
10965 return -KEY_getservent;
10971 if (name[5] == 'c' &&
10977 return -KEY_getsockopt;
10997 if (name[2] == 't')
11002 if (name[4] == 'o' &&
11009 return -KEY_sethostent;
11018 if (name[5] == 'r' &&
11024 return -KEY_setservent;
11030 if (name[5] == 'c' &&
11036 return -KEY_setsockopt;
11053 if (name[2] == 'c' &&
11062 return -KEY_socketpair;
11075 case 11: /* 8 tokens of length 11 */
11079 if (name[1] == '_' &&
11089 { /* __PACKAGE__ */
11090 return -KEY___PACKAGE__;
11096 if (name[1] == 'n' &&
11106 { /* endprotoent */
11107 return -KEY_endprotoent;
11113 if (name[1] == 'e' &&
11122 if (name[5] == 'e' &&
11128 { /* getpeername */
11129 return -KEY_getpeername;
11138 if (name[6] == 'o' &&
11143 { /* getpriority */
11144 return -KEY_getpriority;
11150 if (name[6] == 't' &&
11155 { /* getprotoent */
11156 return -KEY_getprotoent;
11170 if (name[4] == 'o' &&
11177 { /* getsockname */
11178 return -KEY_getsockname;
11191 if (name[1] == 'e' &&
11199 if (name[6] == 'o' &&
11204 { /* setpriority */
11205 return -KEY_setpriority;
11211 if (name[6] == 't' &&
11216 { /* setprotoent */
11217 return -KEY_setprotoent;
11233 case 12: /* 2 tokens of length 12 */
11234 if (name[0] == 'g' &&
11246 if (name[9] == 'd' &&
11249 { /* getnetbyaddr */
11250 return -KEY_getnetbyaddr;
11256 if (name[9] == 'a' &&
11259 { /* getnetbyname */
11260 return -KEY_getnetbyname;
11272 case 13: /* 4 tokens of length 13 */
11273 if (name[0] == 'g' &&
11280 if (name[4] == 'o' &&
11289 if (name[10] == 'd' &&
11292 { /* gethostbyaddr */
11293 return -KEY_gethostbyaddr;
11299 if (name[10] == 'a' &&
11302 { /* gethostbyname */
11303 return -KEY_gethostbyname;
11316 if (name[4] == 'e' &&
11325 if (name[10] == 'a' &&
11328 { /* getservbyname */
11329 return -KEY_getservbyname;
11335 if (name[10] == 'o' &&
11338 { /* getservbyport */
11339 return -KEY_getservbyport;
11358 case 14: /* 1 tokens of length 14 */
11359 if (name[0] == 'g' &&
11373 { /* getprotobyname */
11374 return -KEY_getprotobyname;
11379 case 16: /* 1 tokens of length 16 */
11380 if (name[0] == 'g' &&
11396 { /* getprotobynumber */
11397 return -KEY_getprotobynumber;
11411 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11415 PERL_ARGS_ASSERT_CHECKCOMMA;
11417 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11418 if (ckWARN(WARN_SYNTAX)) {
11421 for (w = s+2; *w && level; w++) {
11424 else if (*w == ')')
11427 while (isSPACE(*w))
11429 /* the list of chars below is for end of statements or
11430 * block / parens, boolean operators (&&, ||, //) and branch
11431 * constructs (or, and, if, until, unless, while, err, for).
11432 * Not a very solid hack... */
11433 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11434 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11435 "%s (...) interpreted as function",name);
11438 while (s < PL_bufend && isSPACE(*s))
11442 while (s < PL_bufend && isSPACE(*s))
11444 if (isIDFIRST_lazy_if(s,UTF)) {
11445 const char * const w = s++;
11446 while (isALNUM_lazy_if(s,UTF))
11448 while (s < PL_bufend && isSPACE(*s))
11452 if (keyword(w, s - w, 0))
11455 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11456 if (gv && GvCVu(gv))
11458 Perl_croak(aTHX_ "No comma allowed after %s", what);
11463 /* Either returns sv, or mortalizes sv and returns a new SV*.
11464 Best used as sv=new_constant(..., sv, ...).
11465 If s, pv are NULL, calls subroutine with one argument,
11466 and type is used with error messages only. */
11469 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11470 SV *sv, SV *pv, const char *type, STRLEN typelen)
11473 HV * const table = GvHV(PL_hintgv); /* ^H */
11477 const char *why1 = "", *why2 = "", *why3 = "";
11479 PERL_ARGS_ASSERT_NEW_CONSTANT;
11481 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11484 why2 = (const char *)
11485 (strEQ(key,"charnames")
11486 ? "(possibly a missing \"use charnames ...\")"
11488 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11489 (type ? type: "undef"), why2);
11491 /* This is convoluted and evil ("goto considered harmful")
11492 * but I do not understand the intricacies of all the different
11493 * failure modes of %^H in here. The goal here is to make
11494 * the most probable error message user-friendly. --jhi */
11499 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11500 (type ? type: "undef"), why1, why2, why3);
11502 yyerror(SvPVX_const(msg));
11507 /* charnames doesn't work well if there have been errors found */
11508 if (PL_error_count > 0 && strEQ(key,"charnames")) return res;
11510 cvp = hv_fetch(table, key, keylen, FALSE);
11511 if (!cvp || !SvOK(*cvp)) {
11514 why3 = "} is not defined";
11517 sv_2mortal(sv); /* Parent created it permanently */
11520 pv = newSVpvn_flags(s, len, SVs_TEMP);
11522 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11524 typesv = &PL_sv_undef;
11526 PUSHSTACKi(PERLSI_OVERLOAD);
11538 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11542 /* Check the eval first */
11543 if (!PL_in_eval && SvTRUE(ERRSV)) {
11544 sv_catpvs(ERRSV, "Propagated");
11545 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11547 res = SvREFCNT_inc_simple(sv);
11551 SvREFCNT_inc_simple_void(res);
11560 why1 = "Call to &{$^H{";
11562 why3 = "}} did not return a defined value";
11570 /* Returns a NUL terminated string, with the length of the string written to
11574 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11577 register char *d = dest;
11578 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11580 PERL_ARGS_ASSERT_SCAN_WORD;
11584 Perl_croak(aTHX_ ident_too_long);
11585 if (isALNUM(*s)) /* UTF handled below */
11587 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11592 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11596 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11597 char *t = s + UTF8SKIP(s);
11599 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11603 Perl_croak(aTHX_ ident_too_long);
11604 Copy(s, d, len, char);
11617 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11620 char *bracket = NULL;
11622 register char *d = dest;
11623 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11625 PERL_ARGS_ASSERT_SCAN_IDENT;
11630 while (isDIGIT(*s)) {
11632 Perl_croak(aTHX_ ident_too_long);
11639 Perl_croak(aTHX_ ident_too_long);
11640 if (isALNUM(*s)) /* UTF handled below */
11642 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11647 else if (*s == ':' && s[1] == ':') {
11651 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11652 char *t = s + UTF8SKIP(s);
11653 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11655 if (d + (t - s) > e)
11656 Perl_croak(aTHX_ ident_too_long);
11657 Copy(s, d, t - s, char);
11668 if (PL_lex_state != LEX_NORMAL)
11669 PL_lex_state = LEX_INTERPENDMAYBE;
11672 if (*s == '$' && s[1] &&
11673 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11686 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11691 if (isSPACE(s[-1])) {
11693 const char ch = *s++;
11694 if (!SPACE_OR_TAB(ch)) {
11700 if (isIDFIRST_lazy_if(d,UTF)) {
11704 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11705 end += UTF8SKIP(end);
11706 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11707 end += UTF8SKIP(end);
11709 Copy(s, d, end - s, char);
11714 while ((isALNUM(*s) || *s == ':') && d < e)
11717 Perl_croak(aTHX_ ident_too_long);
11720 while (s < send && SPACE_OR_TAB(*s))
11722 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11723 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11724 const char * const brack =
11726 ((*s == '[') ? "[...]" : "{...}");
11727 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11728 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11729 funny, dest, brack, funny, dest, brack);
11732 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11736 /* Handle extended ${^Foo} variables
11737 * 1999-02-27 mjd-perl-patch@plover.com */
11738 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11742 while (isALNUM(*s) && d < e) {
11746 Perl_croak(aTHX_ ident_too_long);
11751 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11752 PL_lex_state = LEX_INTERPEND;
11755 if (PL_lex_state == LEX_NORMAL) {
11756 if (ckWARN(WARN_AMBIGUOUS) &&
11757 (keyword(dest, d - dest, 0)
11758 || get_cvn_flags(dest, d - dest, 0)))
11762 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11763 "Ambiguous use of %c{%s} resolved to %c%s",
11764 funny, dest, funny, dest);
11769 s = bracket; /* let the parser handle it */
11773 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11774 PL_lex_state = LEX_INTERPEND;
11779 S_pmflag(U32 pmfl, const char ch) {
11781 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11782 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11783 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11784 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11785 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11791 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11793 PERL_ARGS_ASSERT_PMFLAG;
11795 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11796 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11799 *pmfl = S_pmflag(*pmfl, (char)ch);
11804 S_scan_pat(pTHX_ char *start, I32 type)
11808 char *s = scan_str(start,!!PL_madskills,FALSE);
11809 const char * const valid_flags =
11810 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11815 PERL_ARGS_ASSERT_SCAN_PAT;
11818 const char * const delimiter = skipspace(start);
11822 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11823 : "Search pattern not terminated" ));
11826 pm = (PMOP*)newPMOP(type, 0);
11827 if (PL_multi_open == '?') {
11828 /* This is the only point in the code that sets PMf_ONCE: */
11829 pm->op_pmflags |= PMf_ONCE;
11831 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11832 allows us to restrict the list needed by reset to just the ??
11834 assert(type != OP_TRANS);
11836 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11839 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11842 elements = mg->mg_len / sizeof(PMOP**);
11843 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11844 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11845 mg->mg_len = elements * sizeof(PMOP**);
11846 PmopSTASH_set(pm,PL_curstash);
11852 while (*s && strchr(valid_flags, *s))
11853 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11855 if (PL_madskills && modstart != s) {
11856 SV* tmptoken = newSVpvn(modstart, s - modstart);
11857 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11860 /* issue a warning if /c is specified,but /g is not */
11861 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11863 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11864 "Use of /c modifier is meaningless without /g" );
11867 PL_lex_op = (OP*)pm;
11868 pl_yylval.ival = OP_MATCH;
11873 S_scan_subst(pTHX_ char *start)
11884 PERL_ARGS_ASSERT_SCAN_SUBST;
11886 pl_yylval.ival = OP_NULL;
11888 s = scan_str(start,!!PL_madskills,FALSE);
11891 Perl_croak(aTHX_ "Substitution pattern not terminated");
11893 if (s[-1] == PL_multi_open)
11896 if (PL_madskills) {
11897 CURMAD('q', PL_thisopen);
11898 CURMAD('_', PL_thiswhite);
11899 CURMAD('E', PL_thisstuff);
11900 CURMAD('Q', PL_thisclose);
11901 PL_realtokenstart = s - SvPVX(PL_linestr);
11905 first_start = PL_multi_start;
11906 s = scan_str(s,!!PL_madskills,FALSE);
11908 if (PL_lex_stuff) {
11909 SvREFCNT_dec(PL_lex_stuff);
11910 PL_lex_stuff = NULL;
11912 Perl_croak(aTHX_ "Substitution replacement not terminated");
11914 PL_multi_start = first_start; /* so whole substitution is taken together */
11916 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11919 if (PL_madskills) {
11920 CURMAD('z', PL_thisopen);
11921 CURMAD('R', PL_thisstuff);
11922 CURMAD('Z', PL_thisclose);
11928 if (*s == EXEC_PAT_MOD) {
11932 else if (strchr(S_PAT_MODS, *s))
11933 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11939 if (PL_madskills) {
11941 curmad('m', newSVpvn(modstart, s - modstart));
11942 append_madprops(PL_thismad, (OP*)pm, 0);
11946 if ((pm->op_pmflags & PMf_CONTINUE)) {
11947 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11951 SV * const repl = newSVpvs("");
11953 PL_sublex_info.super_bufptr = s;
11954 PL_sublex_info.super_bufend = PL_bufend;
11956 pm->op_pmflags |= PMf_EVAL;
11959 sv_catpvs(repl, "eval ");
11961 sv_catpvs(repl, "do ");
11963 sv_catpvs(repl, "{");
11964 sv_catsv(repl, PL_lex_repl);
11965 if (strchr(SvPVX(PL_lex_repl), '#'))
11966 sv_catpvs(repl, "\n");
11967 sv_catpvs(repl, "}");
11969 SvREFCNT_dec(PL_lex_repl);
11970 PL_lex_repl = repl;
11973 PL_lex_op = (OP*)pm;
11974 pl_yylval.ival = OP_SUBST;
11979 S_scan_trans(pTHX_ char *start)
11992 PERL_ARGS_ASSERT_SCAN_TRANS;
11994 pl_yylval.ival = OP_NULL;
11996 s = scan_str(start,!!PL_madskills,FALSE);
11998 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12000 if (s[-1] == PL_multi_open)
12003 if (PL_madskills) {
12004 CURMAD('q', PL_thisopen);
12005 CURMAD('_', PL_thiswhite);
12006 CURMAD('E', PL_thisstuff);
12007 CURMAD('Q', PL_thisclose);
12008 PL_realtokenstart = s - SvPVX(PL_linestr);
12012 s = scan_str(s,!!PL_madskills,FALSE);
12014 if (PL_lex_stuff) {
12015 SvREFCNT_dec(PL_lex_stuff);
12016 PL_lex_stuff = NULL;
12018 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12020 if (PL_madskills) {
12021 CURMAD('z', PL_thisopen);
12022 CURMAD('R', PL_thisstuff);
12023 CURMAD('Z', PL_thisclose);
12026 complement = del = squash = 0;
12033 complement = OPpTRANS_COMPLEMENT;
12036 del = OPpTRANS_DELETE;
12039 squash = OPpTRANS_SQUASH;
12048 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12049 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12050 o->op_private &= ~OPpTRANS_ALL;
12051 o->op_private |= del|squash|complement|
12052 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12053 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12056 pl_yylval.ival = OP_TRANS;
12059 if (PL_madskills) {
12061 curmad('m', newSVpvn(modstart, s - modstart));
12062 append_madprops(PL_thismad, o, 0);
12071 S_scan_heredoc(pTHX_ register char *s)
12075 I32 op_type = OP_SCALAR;
12079 const char *found_newline;
12083 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12085 I32 stuffstart = s - SvPVX(PL_linestr);
12088 PL_realtokenstart = -1;
12091 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12095 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12099 while (SPACE_OR_TAB(*peek))
12101 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12104 s = delimcpy(d, e, s, PL_bufend, term, &len);
12114 if (!isALNUM_lazy_if(s,UTF))
12115 deprecate("bare << to mean <<\"\"");
12116 for (; isALNUM_lazy_if(s,UTF); s++) {
12121 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12122 Perl_croak(aTHX_ "Delimiter for here document is too long");
12125 len = d - PL_tokenbuf;
12128 if (PL_madskills) {
12129 tstart = PL_tokenbuf + !outer;
12130 PL_thisclose = newSVpvn(tstart, len - !outer);
12131 tstart = SvPVX(PL_linestr) + stuffstart;
12132 PL_thisopen = newSVpvn(tstart, s - tstart);
12133 stuffstart = s - SvPVX(PL_linestr);
12136 #ifndef PERL_STRICT_CR
12137 d = strchr(s, '\r');
12139 char * const olds = s;
12141 while (s < PL_bufend) {
12147 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12156 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12163 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12164 herewas = newSVpvn(s,PL_bufend-s);
12168 herewas = newSVpvn(s-1,found_newline-s+1);
12171 herewas = newSVpvn(s,found_newline-s);
12175 if (PL_madskills) {
12176 tstart = SvPVX(PL_linestr) + stuffstart;
12178 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12180 PL_thisstuff = newSVpvn(tstart, s - tstart);
12183 s += SvCUR(herewas);
12186 stuffstart = s - SvPVX(PL_linestr);
12192 tmpstr = newSV_type(SVt_PVIV);
12193 SvGROW(tmpstr, 80);
12194 if (term == '\'') {
12195 op_type = OP_CONST;
12196 SvIV_set(tmpstr, -1);
12198 else if (term == '`') {
12199 op_type = OP_BACKTICK;
12200 SvIV_set(tmpstr, '\\');
12204 PL_multi_start = CopLINE(PL_curcop);
12205 PL_multi_open = PL_multi_close = '<';
12206 term = *PL_tokenbuf;
12207 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12208 char * const bufptr = PL_sublex_info.super_bufptr;
12209 char * const bufend = PL_sublex_info.super_bufend;
12210 char * const olds = s - SvCUR(herewas);
12211 s = strchr(bufptr, '\n');
12215 while (s < bufend &&
12216 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12218 CopLINE_inc(PL_curcop);
12221 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12222 missingterm(PL_tokenbuf);
12224 sv_setpvn(herewas,bufptr,d-bufptr+1);
12225 sv_setpvn(tmpstr,d+1,s-d);
12227 sv_catpvn(herewas,s,bufend-s);
12228 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12235 while (s < PL_bufend &&
12236 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12238 CopLINE_inc(PL_curcop);
12240 if (s >= PL_bufend) {
12241 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12242 missingterm(PL_tokenbuf);
12244 sv_setpvn(tmpstr,d+1,s-d);
12246 if (PL_madskills) {
12248 sv_catpvn(PL_thisstuff, d + 1, s - d);
12250 PL_thisstuff = newSVpvn(d + 1, s - d);
12251 stuffstart = s - SvPVX(PL_linestr);
12255 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12257 sv_catpvn(herewas,s,PL_bufend-s);
12258 sv_setsv(PL_linestr,herewas);
12259 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12260 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12261 PL_last_lop = PL_last_uni = NULL;
12264 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12265 while (s >= PL_bufend) { /* multiple line string? */
12267 if (PL_madskills) {
12268 tstart = SvPVX(PL_linestr) + stuffstart;
12270 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12272 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12276 CopLINE_inc(PL_curcop);
12277 if (!outer || !lex_next_chunk(0)) {
12278 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12279 missingterm(PL_tokenbuf);
12281 CopLINE_dec(PL_curcop);
12284 stuffstart = s - SvPVX(PL_linestr);
12286 CopLINE_inc(PL_curcop);
12287 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12288 PL_last_lop = PL_last_uni = NULL;
12289 #ifndef PERL_STRICT_CR
12290 if (PL_bufend - PL_linestart >= 2) {
12291 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12292 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12294 PL_bufend[-2] = '\n';
12296 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12298 else if (PL_bufend[-1] == '\r')
12299 PL_bufend[-1] = '\n';
12301 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12302 PL_bufend[-1] = '\n';
12304 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12305 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12306 *(SvPVX(PL_linestr) + off ) = ' ';
12307 sv_catsv(PL_linestr,herewas);
12308 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12309 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12313 sv_catsv(tmpstr,PL_linestr);
12318 PL_multi_end = CopLINE(PL_curcop);
12319 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12320 SvPV_shrink_to_cur(tmpstr);
12322 SvREFCNT_dec(herewas);
12324 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12326 else if (PL_encoding)
12327 sv_recode_to_utf8(tmpstr, PL_encoding);
12329 PL_lex_stuff = tmpstr;
12330 pl_yylval.ival = op_type;
12334 /* scan_inputsymbol
12335 takes: current position in input buffer
12336 returns: new position in input buffer
12337 side-effects: pl_yylval and lex_op are set.
12342 <FH> read from filehandle
12343 <pkg::FH> read from package qualified filehandle
12344 <pkg'FH> read from package qualified filehandle
12345 <$fh> read from filehandle in $fh
12346 <*.h> filename glob
12351 S_scan_inputsymbol(pTHX_ char *start)
12354 register char *s = start; /* current position in buffer */
12357 char *d = PL_tokenbuf; /* start of temp holding space */
12358 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12360 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12362 end = strchr(s, '\n');
12365 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12367 /* die if we didn't have space for the contents of the <>,
12368 or if it didn't end, or if we see a newline
12371 if (len >= (I32)sizeof PL_tokenbuf)
12372 Perl_croak(aTHX_ "Excessively long <> operator");
12374 Perl_croak(aTHX_ "Unterminated <> operator");
12379 Remember, only scalar variables are interpreted as filehandles by
12380 this code. Anything more complex (e.g., <$fh{$num}>) will be
12381 treated as a glob() call.
12382 This code makes use of the fact that except for the $ at the front,
12383 a scalar variable and a filehandle look the same.
12385 if (*d == '$' && d[1]) d++;
12387 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12388 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12391 /* If we've tried to read what we allow filehandles to look like, and
12392 there's still text left, then it must be a glob() and not a getline.
12393 Use scan_str to pull out the stuff between the <> and treat it
12394 as nothing more than a string.
12397 if (d - PL_tokenbuf != len) {
12398 pl_yylval.ival = OP_GLOB;
12399 s = scan_str(start,!!PL_madskills,FALSE);
12401 Perl_croak(aTHX_ "Glob not terminated");
12405 bool readline_overriden = FALSE;
12408 /* we're in a filehandle read situation */
12411 /* turn <> into <ARGV> */
12413 Copy("ARGV",d,5,char);
12415 /* Check whether readline() is overriden */
12416 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12418 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12420 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12421 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12422 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12423 readline_overriden = TRUE;
12425 /* if <$fh>, create the ops to turn the variable into a
12429 /* try to find it in the pad for this block, otherwise find
12430 add symbol table ops
12432 const PADOFFSET tmp = pad_findmy(d, len, 0);
12433 if (tmp != NOT_IN_PAD) {
12434 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12435 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12436 HEK * const stashname = HvNAME_HEK(stash);
12437 SV * const sym = sv_2mortal(newSVhek(stashname));
12438 sv_catpvs(sym, "::");
12439 sv_catpv(sym, d+1);
12444 OP * const o = newOP(OP_PADSV, 0);
12446 PL_lex_op = readline_overriden
12447 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12448 append_elem(OP_LIST, o,
12449 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12450 : (OP*)newUNOP(OP_READLINE, 0, o);
12459 ? (GV_ADDMULTI | GV_ADDINEVAL)
12462 PL_lex_op = readline_overriden
12463 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12464 append_elem(OP_LIST,
12465 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12466 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12467 : (OP*)newUNOP(OP_READLINE, 0,
12468 newUNOP(OP_RV2SV, 0,
12469 newGVOP(OP_GV, 0, gv)));
12471 if (!readline_overriden)
12472 PL_lex_op->op_flags |= OPf_SPECIAL;
12473 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12474 pl_yylval.ival = OP_NULL;
12477 /* If it's none of the above, it must be a literal filehandle
12478 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12480 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12481 PL_lex_op = readline_overriden
12482 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12483 append_elem(OP_LIST,
12484 newGVOP(OP_GV, 0, gv),
12485 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12486 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12487 pl_yylval.ival = OP_NULL;
12496 takes: start position in buffer
12497 keep_quoted preserve \ on the embedded delimiter(s)
12498 keep_delims preserve the delimiters around the string
12499 returns: position to continue reading from buffer
12500 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12501 updates the read buffer.
12503 This subroutine pulls a string out of the input. It is called for:
12504 q single quotes q(literal text)
12505 ' single quotes 'literal text'
12506 qq double quotes qq(interpolate $here please)
12507 " double quotes "interpolate $here please"
12508 qx backticks qx(/bin/ls -l)
12509 ` backticks `/bin/ls -l`
12510 qw quote words @EXPORT_OK = qw( func() $spam )
12511 m// regexp match m/this/
12512 s/// regexp substitute s/this/that/
12513 tr/// string transliterate tr/this/that/
12514 y/// string transliterate y/this/that/
12515 ($*@) sub prototypes sub foo ($)
12516 (stuff) sub attr parameters sub foo : attr(stuff)
12517 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12519 In most of these cases (all but <>, patterns and transliterate)
12520 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12521 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12522 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12525 It skips whitespace before the string starts, and treats the first
12526 character as the delimiter. If the delimiter is one of ([{< then
12527 the corresponding "close" character )]}> is used as the closing
12528 delimiter. It allows quoting of delimiters, and if the string has
12529 balanced delimiters ([{<>}]) it allows nesting.
12531 On success, the SV with the resulting string is put into lex_stuff or,
12532 if that is already non-NULL, into lex_repl. The second case occurs only
12533 when parsing the RHS of the special constructs s/// and tr/// (y///).
12534 For convenience, the terminating delimiter character is stuffed into
12539 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12542 SV *sv; /* scalar value: string */
12543 const char *tmps; /* temp string, used for delimiter matching */
12544 register char *s = start; /* current position in the buffer */
12545 register char term; /* terminating character */
12546 register char *to; /* current position in the sv's data */
12547 I32 brackets = 1; /* bracket nesting level */
12548 bool has_utf8 = FALSE; /* is there any utf8 content? */
12549 I32 termcode; /* terminating char. code */
12550 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12551 STRLEN termlen; /* length of terminating string */
12552 int last_off = 0; /* last position for nesting bracket */
12558 PERL_ARGS_ASSERT_SCAN_STR;
12560 /* skip space before the delimiter */
12566 if (PL_realtokenstart >= 0) {
12567 stuffstart = PL_realtokenstart;
12568 PL_realtokenstart = -1;
12571 stuffstart = start - SvPVX(PL_linestr);
12573 /* mark where we are, in case we need to report errors */
12576 /* after skipping whitespace, the next character is the terminator */
12579 termcode = termstr[0] = term;
12583 termcode = utf8_to_uvchr((U8*)s, &termlen);
12584 Copy(s, termstr, termlen, U8);
12585 if (!UTF8_IS_INVARIANT(term))
12589 /* mark where we are */
12590 PL_multi_start = CopLINE(PL_curcop);
12591 PL_multi_open = term;
12593 /* find corresponding closing delimiter */
12594 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12595 termcode = termstr[0] = term = tmps[5];
12597 PL_multi_close = term;
12599 /* create a new SV to hold the contents. 79 is the SV's initial length.
12600 What a random number. */
12601 sv = newSV_type(SVt_PVIV);
12603 SvIV_set(sv, termcode);
12604 (void)SvPOK_only(sv); /* validate pointer */
12606 /* move past delimiter and try to read a complete string */
12608 sv_catpvn(sv, s, termlen);
12611 tstart = SvPVX(PL_linestr) + stuffstart;
12612 if (!PL_thisopen && !keep_delims) {
12613 PL_thisopen = newSVpvn(tstart, s - tstart);
12614 stuffstart = s - SvPVX(PL_linestr);
12618 if (PL_encoding && !UTF) {
12622 int offset = s - SvPVX_const(PL_linestr);
12623 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12624 &offset, (char*)termstr, termlen);
12625 const char * const ns = SvPVX_const(PL_linestr) + offset;
12626 char * const svlast = SvEND(sv) - 1;
12628 for (; s < ns; s++) {
12629 if (*s == '\n' && !PL_rsfp)
12630 CopLINE_inc(PL_curcop);
12633 goto read_more_line;
12635 /* handle quoted delimiters */
12636 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12638 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12640 if ((svlast-1 - t) % 2) {
12641 if (!keep_quoted) {
12642 *(svlast-1) = term;
12644 SvCUR_set(sv, SvCUR(sv) - 1);
12649 if (PL_multi_open == PL_multi_close) {
12655 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12656 /* At here, all closes are "was quoted" one,
12657 so we don't check PL_multi_close. */
12659 if (!keep_quoted && *(t+1) == PL_multi_open)
12664 else if (*t == PL_multi_open)
12672 SvCUR_set(sv, w - SvPVX_const(sv));
12674 last_off = w - SvPVX(sv);
12675 if (--brackets <= 0)
12680 if (!keep_delims) {
12681 SvCUR_set(sv, SvCUR(sv) - 1);
12687 /* extend sv if need be */
12688 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12689 /* set 'to' to the next character in the sv's string */
12690 to = SvPVX(sv)+SvCUR(sv);
12692 /* if open delimiter is the close delimiter read unbridle */
12693 if (PL_multi_open == PL_multi_close) {
12694 for (; s < PL_bufend; s++,to++) {
12695 /* embedded newlines increment the current line number */
12696 if (*s == '\n' && !PL_rsfp)
12697 CopLINE_inc(PL_curcop);
12698 /* handle quoted delimiters */
12699 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12700 if (!keep_quoted && s[1] == term)
12702 /* any other quotes are simply copied straight through */
12706 /* terminate when run out of buffer (the for() condition), or
12707 have found the terminator */
12708 else if (*s == term) {
12711 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12714 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12720 /* if the terminator isn't the same as the start character (e.g.,
12721 matched brackets), we have to allow more in the quoting, and
12722 be prepared for nested brackets.
12725 /* read until we run out of string, or we find the terminator */
12726 for (; s < PL_bufend; s++,to++) {
12727 /* embedded newlines increment the line count */
12728 if (*s == '\n' && !PL_rsfp)
12729 CopLINE_inc(PL_curcop);
12730 /* backslashes can escape the open or closing characters */
12731 if (*s == '\\' && s+1 < PL_bufend) {
12732 if (!keep_quoted &&
12733 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12738 /* allow nested opens and closes */
12739 else if (*s == PL_multi_close && --brackets <= 0)
12741 else if (*s == PL_multi_open)
12743 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12748 /* terminate the copied string and update the sv's end-of-string */
12750 SvCUR_set(sv, to - SvPVX_const(sv));
12753 * this next chunk reads more into the buffer if we're not done yet
12757 break; /* handle case where we are done yet :-) */
12759 #ifndef PERL_STRICT_CR
12760 if (to - SvPVX_const(sv) >= 2) {
12761 if ((to[-2] == '\r' && to[-1] == '\n') ||
12762 (to[-2] == '\n' && to[-1] == '\r'))
12766 SvCUR_set(sv, to - SvPVX_const(sv));
12768 else if (to[-1] == '\r')
12771 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12776 /* if we're out of file, or a read fails, bail and reset the current
12777 line marker so we can report where the unterminated string began
12780 if (PL_madskills) {
12781 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12783 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12785 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12788 CopLINE_inc(PL_curcop);
12789 PL_bufptr = PL_bufend;
12790 if (!lex_next_chunk(0)) {
12792 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12801 /* at this point, we have successfully read the delimited string */
12803 if (!PL_encoding || UTF) {
12805 if (PL_madskills) {
12806 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12807 const int len = s - tstart;
12809 sv_catpvn(PL_thisstuff, tstart, len);
12811 PL_thisstuff = newSVpvn(tstart, len);
12812 if (!PL_thisclose && !keep_delims)
12813 PL_thisclose = newSVpvn(s,termlen);
12818 sv_catpvn(sv, s, termlen);
12823 if (PL_madskills) {
12824 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12825 const int len = s - tstart - termlen;
12827 sv_catpvn(PL_thisstuff, tstart, len);
12829 PL_thisstuff = newSVpvn(tstart, len);
12830 if (!PL_thisclose && !keep_delims)
12831 PL_thisclose = newSVpvn(s - termlen,termlen);
12835 if (has_utf8 || PL_encoding)
12838 PL_multi_end = CopLINE(PL_curcop);
12840 /* if we allocated too much space, give some back */
12841 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12842 SvLEN_set(sv, SvCUR(sv) + 1);
12843 SvPV_renew(sv, SvLEN(sv));
12846 /* decide whether this is the first or second quoted string we've read
12859 takes: pointer to position in buffer
12860 returns: pointer to new position in buffer
12861 side-effects: builds ops for the constant in pl_yylval.op
12863 Read a number in any of the formats that Perl accepts:
12865 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12866 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12869 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12871 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12874 If it reads a number without a decimal point or an exponent, it will
12875 try converting the number to an integer and see if it can do so
12876 without loss of precision.
12880 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12883 register const char *s = start; /* current position in buffer */
12884 register char *d; /* destination in temp buffer */
12885 register char *e; /* end of temp buffer */
12886 NV nv; /* number read, as a double */
12887 SV *sv = NULL; /* place to put the converted number */
12888 bool floatit; /* boolean: int or float? */
12889 const char *lastub = NULL; /* position of last underbar */
12890 static char const number_too_long[] = "Number too long";
12892 PERL_ARGS_ASSERT_SCAN_NUM;
12894 /* We use the first character to decide what type of number this is */
12898 Perl_croak(aTHX_ "panic: scan_num");
12900 /* if it starts with a 0, it could be an octal number, a decimal in
12901 0.13 disguise, or a hexadecimal number, or a binary number. */
12905 u holds the "number so far"
12906 shift the power of 2 of the base
12907 (hex == 4, octal == 3, binary == 1)
12908 overflowed was the number more than we can hold?
12910 Shift is used when we add a digit. It also serves as an "are
12911 we in octal/hex/binary?" indicator to disallow hex characters
12912 when in octal mode.
12917 bool overflowed = FALSE;
12918 bool just_zero = TRUE; /* just plain 0 or binary number? */
12919 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12920 static const char* const bases[5] =
12921 { "", "binary", "", "octal", "hexadecimal" };
12922 static const char* const Bases[5] =
12923 { "", "Binary", "", "Octal", "Hexadecimal" };
12924 static const char* const maxima[5] =
12926 "0b11111111111111111111111111111111",
12930 const char *base, *Base, *max;
12932 /* check for hex */
12937 } else if (s[1] == 'b') {
12942 /* check for a decimal in disguise */
12943 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12945 /* so it must be octal */
12952 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12953 "Misplaced _ in number");
12957 base = bases[shift];
12958 Base = Bases[shift];
12959 max = maxima[shift];
12961 /* read the rest of the number */
12963 /* x is used in the overflow test,
12964 b is the digit we're adding on. */
12969 /* if we don't mention it, we're done */
12973 /* _ are ignored -- but warned about if consecutive */
12975 if (lastub && s == lastub + 1)
12976 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12977 "Misplaced _ in number");
12981 /* 8 and 9 are not octal */
12982 case '8': case '9':
12984 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12988 case '2': case '3': case '4':
12989 case '5': case '6': case '7':
12991 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12994 case '0': case '1':
12995 b = *s++ & 15; /* ASCII digit -> value of digit */
12999 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13000 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13001 /* make sure they said 0x */
13004 b = (*s++ & 7) + 9;
13006 /* Prepare to put the digit we have onto the end
13007 of the number so far. We check for overflows.
13013 x = u << shift; /* make room for the digit */
13015 if ((x >> shift) != u
13016 && !(PL_hints & HINT_NEW_BINARY)) {
13019 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13020 "Integer overflow in %s number",
13023 u = x | b; /* add the digit to the end */
13026 n *= nvshift[shift];
13027 /* If an NV has not enough bits in its
13028 * mantissa to represent an UV this summing of
13029 * small low-order numbers is a waste of time
13030 * (because the NV cannot preserve the
13031 * low-order bits anyway): we could just
13032 * remember when did we overflow and in the
13033 * end just multiply n by the right
13041 /* if we get here, we had success: make a scalar value from
13046 /* final misplaced underbar check */
13047 if (s[-1] == '_') {
13048 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13053 if (n > 4294967295.0)
13054 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13055 "%s number > %s non-portable",
13061 if (u > 0xffffffff)
13062 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13063 "%s number > %s non-portable",
13068 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13069 sv = new_constant(start, s - start, "integer",
13070 sv, NULL, NULL, 0);
13071 else if (PL_hints & HINT_NEW_BINARY)
13072 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13077 handle decimal numbers.
13078 we're also sent here when we read a 0 as the first digit
13080 case '1': case '2': case '3': case '4': case '5':
13081 case '6': case '7': case '8': case '9': case '.':
13084 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13087 /* read next group of digits and _ and copy into d */
13088 while (isDIGIT(*s) || *s == '_') {
13089 /* skip underscores, checking for misplaced ones
13093 if (lastub && s == lastub + 1)
13094 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13095 "Misplaced _ in number");
13099 /* check for end of fixed-length buffer */
13101 Perl_croak(aTHX_ number_too_long);
13102 /* if we're ok, copy the character */
13107 /* final misplaced underbar check */
13108 if (lastub && s == lastub + 1) {
13109 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13112 /* read a decimal portion if there is one. avoid
13113 3..5 being interpreted as the number 3. followed
13116 if (*s == '.' && s[1] != '.') {
13121 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13122 "Misplaced _ in number");
13126 /* copy, ignoring underbars, until we run out of digits.
13128 for (; isDIGIT(*s) || *s == '_'; s++) {
13129 /* fixed length buffer check */
13131 Perl_croak(aTHX_ number_too_long);
13133 if (lastub && s == lastub + 1)
13134 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13135 "Misplaced _ in number");
13141 /* fractional part ending in underbar? */
13142 if (s[-1] == '_') {
13143 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13144 "Misplaced _ in number");
13146 if (*s == '.' && isDIGIT(s[1])) {
13147 /* oops, it's really a v-string, but without the "v" */
13153 /* read exponent part, if present */
13154 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13158 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13159 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13161 /* stray preinitial _ */
13163 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13164 "Misplaced _ in number");
13168 /* allow positive or negative exponent */
13169 if (*s == '+' || *s == '-')
13172 /* stray initial _ */
13174 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13175 "Misplaced _ in number");
13179 /* read digits of exponent */
13180 while (isDIGIT(*s) || *s == '_') {
13183 Perl_croak(aTHX_ number_too_long);
13187 if (((lastub && s == lastub + 1) ||
13188 (!isDIGIT(s[1]) && s[1] != '_')))
13189 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13190 "Misplaced _ in number");
13197 /* make an sv from the string */
13201 We try to do an integer conversion first if no characters
13202 indicating "float" have been found.
13207 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13209 if (flags == IS_NUMBER_IN_UV) {
13211 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
13214 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13215 if (uv <= (UV) IV_MIN)
13216 sv_setiv(sv, -(IV)uv);
13223 /* terminate the string */
13225 nv = Atof(PL_tokenbuf);
13230 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13231 const char *const key = floatit ? "float" : "integer";
13232 const STRLEN keylen = floatit ? 5 : 7;
13233 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13234 key, keylen, sv, NULL, NULL, 0);
13238 /* if it starts with a v, it could be a v-string */
13241 sv = newSV(5); /* preallocate storage space */
13242 s = scan_vstring(s, PL_bufend, sv);
13246 /* make the op for the constant and return */
13249 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13251 lvalp->opval = NULL;
13257 S_scan_formline(pTHX_ register char *s)
13260 register char *eol;
13262 SV * const stuff = newSVpvs("");
13263 bool needargs = FALSE;
13264 bool eofmt = FALSE;
13266 char *tokenstart = s;
13267 SV* savewhite = NULL;
13269 if (PL_madskills) {
13270 savewhite = PL_thiswhite;
13275 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13277 while (!needargs) {
13280 #ifdef PERL_STRICT_CR
13281 while (SPACE_OR_TAB(*t))
13284 while (SPACE_OR_TAB(*t) || *t == '\r')
13287 if (*t == '\n' || t == PL_bufend) {
13292 if (PL_in_eval && !PL_rsfp) {
13293 eol = (char *) memchr(s,'\n',PL_bufend-s);
13298 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13300 for (t = s; t < eol; t++) {
13301 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13303 goto enough; /* ~~ must be first line in formline */
13305 if (*t == '@' || *t == '^')
13309 sv_catpvn(stuff, s, eol-s);
13310 #ifndef PERL_STRICT_CR
13311 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13312 char *end = SvPVX(stuff) + SvCUR(stuff);
13315 SvCUR_set(stuff, SvCUR(stuff) - 1);
13326 if (PL_madskills) {
13328 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13330 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13333 PL_bufptr = PL_bufend;
13334 CopLINE_inc(PL_curcop);
13335 got_some = lex_next_chunk(0);
13336 CopLINE_dec(PL_curcop);
13339 tokenstart = PL_bufptr;
13347 if (SvCUR(stuff)) {
13350 PL_lex_state = LEX_NORMAL;
13351 start_force(PL_curforce);
13352 NEXTVAL_NEXTTOKE.ival = 0;
13356 PL_lex_state = LEX_FORMLINE;
13358 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13360 else if (PL_encoding)
13361 sv_recode_to_utf8(stuff, PL_encoding);
13363 start_force(PL_curforce);
13364 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13366 start_force(PL_curforce);
13367 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13371 SvREFCNT_dec(stuff);
13373 PL_lex_formbrack = 0;
13377 if (PL_madskills) {
13379 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13381 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13382 PL_thiswhite = savewhite;
13389 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13392 const I32 oldsavestack_ix = PL_savestack_ix;
13393 CV* const outsidecv = PL_compcv;
13396 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13398 SAVEI32(PL_subline);
13399 save_item(PL_subname);
13400 SAVESPTR(PL_compcv);
13402 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13403 CvFLAGS(PL_compcv) |= flags;
13405 PL_subline = CopLINE(PL_curcop);
13406 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13407 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13408 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13410 return oldsavestack_ix;
13414 #pragma segment Perl_yylex
13417 S_yywarn(pTHX_ const char *const s)
13421 PERL_ARGS_ASSERT_YYWARN;
13423 PL_in_eval |= EVAL_WARNONLY;
13425 PL_in_eval &= ~EVAL_WARNONLY;
13430 Perl_yyerror(pTHX_ const char *const s)
13433 const char *where = NULL;
13434 const char *context = NULL;
13437 int yychar = PL_parser->yychar;
13439 PERL_ARGS_ASSERT_YYERROR;
13441 if (!yychar || (yychar == ';' && !PL_rsfp))
13443 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13444 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13445 PL_oldbufptr != PL_bufptr) {
13448 The code below is removed for NetWare because it abends/crashes on NetWare
13449 when the script has error such as not having the closing quotes like:
13450 if ($var eq "value)
13451 Checking of white spaces is anyway done in NetWare code.
13454 while (isSPACE(*PL_oldoldbufptr))
13457 context = PL_oldoldbufptr;
13458 contlen = PL_bufptr - PL_oldoldbufptr;
13460 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13461 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13464 The code below is removed for NetWare because it abends/crashes on NetWare
13465 when the script has error such as not having the closing quotes like:
13466 if ($var eq "value)
13467 Checking of white spaces is anyway done in NetWare code.
13470 while (isSPACE(*PL_oldbufptr))
13473 context = PL_oldbufptr;
13474 contlen = PL_bufptr - PL_oldbufptr;
13476 else if (yychar > 255)
13477 where = "next token ???";
13478 else if (yychar == -2) { /* YYEMPTY */
13479 if (PL_lex_state == LEX_NORMAL ||
13480 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13481 where = "at end of line";
13482 else if (PL_lex_inpat)
13483 where = "within pattern";
13485 where = "within string";
13488 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13490 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13491 else if (isPRINT_LC(yychar)) {
13492 const char string = yychar;
13493 sv_catpvn(where_sv, &string, 1);
13496 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13497 where = SvPVX_const(where_sv);
13499 msg = sv_2mortal(newSVpv(s, 0));
13500 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13501 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13503 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13505 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13506 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13507 Perl_sv_catpvf(aTHX_ msg,
13508 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13509 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13512 if (PL_in_eval & EVAL_WARNONLY) {
13513 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13517 if (PL_error_count >= 10) {
13518 if (PL_in_eval && SvCUR(ERRSV))
13519 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13520 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13522 Perl_croak(aTHX_ "%s has too many errors.\n",
13523 OutCopFILE(PL_curcop));
13526 PL_in_my_stash = NULL;
13530 #pragma segment Main
13534 S_swallow_bom(pTHX_ U8 *s)
13537 const STRLEN slen = SvCUR(PL_linestr);
13539 PERL_ARGS_ASSERT_SWALLOW_BOM;
13543 if (s[1] == 0xFE) {
13544 /* UTF-16 little-endian? (or UTF-32LE?) */
13545 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13546 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13547 #ifndef PERL_NO_UTF16_FILTER
13548 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13550 if (PL_bufend > (char*)s) {
13551 s = add_utf16_textfilter(s, TRUE);
13554 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13559 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13560 #ifndef PERL_NO_UTF16_FILTER
13561 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13563 if (PL_bufend > (char *)s) {
13564 s = add_utf16_textfilter(s, FALSE);
13567 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13572 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13573 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13574 s += 3; /* UTF-8 */
13580 if (s[2] == 0xFE && s[3] == 0xFF) {
13581 /* UTF-32 big-endian */
13582 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13585 else if (s[2] == 0 && s[3] != 0) {
13588 * are a good indicator of UTF-16BE. */
13589 #ifndef PERL_NO_UTF16_FILTER
13590 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13591 s = add_utf16_textfilter(s, FALSE);
13593 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13599 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13600 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13601 s += 4; /* UTF-8 */
13607 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13610 * are a good indicator of UTF-16LE. */
13611 #ifndef PERL_NO_UTF16_FILTER
13612 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13613 s = add_utf16_textfilter(s, TRUE);
13615 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13623 #ifndef PERL_NO_UTF16_FILTER
13625 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13628 SV *const filter = FILTER_DATA(idx);
13629 /* We re-use this each time round, throwing the contents away before we
13631 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13632 SV *const utf8_buffer = filter;
13633 IV status = IoPAGE(filter);
13634 const bool reverse = (bool) IoLINES(filter);
13637 /* As we're automatically added, at the lowest level, and hence only called
13638 from this file, we can be sure that we're not called in block mode. Hence
13639 don't bother writing code to deal with block mode. */
13641 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13644 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13646 DEBUG_P(PerlIO_printf(Perl_debug_log,
13647 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13648 FPTR2DPTR(void *, S_utf16_textfilter),
13649 reverse ? 'l' : 'b', idx, maxlen, status,
13650 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13657 /* First, look in our buffer of existing UTF-8 data: */
13658 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13662 } else if (status == 0) {
13664 IoPAGE(filter) = 0;
13665 nl = SvEND(utf8_buffer);
13668 STRLEN got = nl - SvPVX(utf8_buffer);
13669 /* Did we have anything to append? */
13671 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13672 /* Everything else in this code works just fine if SVp_POK isn't
13673 set. This, however, needs it, and we need it to work, else
13674 we loop infinitely because the buffer is never consumed. */
13675 sv_chop(utf8_buffer, nl);
13679 /* OK, not a complete line there, so need to read some more UTF-16.
13680 Read an extra octect if the buffer currently has an odd number. */
13684 if (SvCUR(utf16_buffer) >= 2) {
13685 /* Location of the high octet of the last complete code point.
13686 Gosh, UTF-16 is a pain. All the benefits of variable length,
13687 *coupled* with all the benefits of partial reads and
13689 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13690 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13692 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13696 /* We have the first half of a surrogate. Read more. */
13697 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13700 status = FILTER_READ(idx + 1, utf16_buffer,
13701 160 + (SvCUR(utf16_buffer) & 1));
13702 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13703 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13706 IoPAGE(filter) = status;
13711 chars = SvCUR(utf16_buffer) >> 1;
13712 have = SvCUR(utf8_buffer);
13713 SvGROW(utf8_buffer, have + chars * 3 + 1);
13716 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13717 (U8*)SvPVX_const(utf8_buffer) + have,
13718 chars * 2, &newlen);
13720 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13721 (U8*)SvPVX_const(utf8_buffer) + have,
13722 chars * 2, &newlen);
13724 SvCUR_set(utf8_buffer, have + newlen);
13727 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13728 it's private to us, and utf16_to_utf8{,reversed} take a
13729 (pointer,length) pair, rather than a NUL-terminated string. */
13730 if(SvCUR(utf16_buffer) & 1) {
13731 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13732 SvCUR_set(utf16_buffer, 1);
13734 SvCUR_set(utf16_buffer, 0);
13737 DEBUG_P(PerlIO_printf(Perl_debug_log,
13738 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13740 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13741 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13746 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13748 SV *filter = filter_add(S_utf16_textfilter, NULL);
13750 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13751 sv_setpvs(filter, "");
13752 IoLINES(filter) = reversed;
13753 IoPAGE(filter) = 1; /* Not EOF */
13755 /* Sadly, we have to return a valid pointer, come what may, so we have to
13756 ignore any error return from this. */
13757 SvCUR_set(PL_linestr, 0);
13758 if (FILTER_READ(0, PL_linestr, 0)) {
13759 SvUTF8_on(PL_linestr);
13761 SvUTF8_on(PL_linestr);
13763 PL_bufend = SvEND(PL_linestr);
13764 return (U8*)SvPVX(PL_linestr);
13769 Returns a pointer to the next character after the parsed
13770 vstring, as well as updating the passed in sv.
13772 Function must be called like
13775 s = scan_vstring(s,e,sv);
13777 where s and e are the start and end of the string.
13778 The sv should already be large enough to store the vstring
13779 passed in, for performance reasons.
13784 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13787 const char *pos = s;
13788 const char *start = s;
13790 PERL_ARGS_ASSERT_SCAN_VSTRING;
13792 if (*pos == 'v') pos++; /* get past 'v' */
13793 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13795 if ( *pos != '.') {
13796 /* this may not be a v-string if followed by => */
13797 const char *next = pos;
13798 while (next < e && isSPACE(*next))
13800 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13801 /* return string not v-string */
13802 sv_setpvn(sv,(char *)s,pos-s);
13803 return (char *)pos;
13807 if (!isALPHA(*pos)) {
13808 U8 tmpbuf[UTF8_MAXBYTES+1];
13811 s++; /* get past 'v' */
13816 /* this is atoi() that tolerates underscores */
13819 const char *end = pos;
13821 while (--end >= s) {
13823 const UV orev = rev;
13824 rev += (*end - '0') * mult;
13827 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13828 "Integer overflow in decimal number");
13832 if (rev > 0x7FFFFFFF)
13833 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13835 /* Append native character for the rev point */
13836 tmpend = uvchr_to_utf8(tmpbuf, rev);
13837 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13838 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13840 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13846 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13850 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13857 Perl_keyword_plugin_standard(pTHX_
13858 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13860 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13861 PERL_UNUSED_CONTEXT;
13862 PERL_UNUSED_ARG(keyword_ptr);
13863 PERL_UNUSED_ARG(keyword_len);
13864 PERL_UNUSED_ARG(op_ptr);
13865 return KEYWORD_PLUGIN_DECLINE;
13870 * c-indentation-style: bsd
13871 * c-basic-offset: 4
13872 * indent-tabs-mode: t
13875 * ex: set ts=8 sts=4 sw=4 noet: