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 #define PERL_IN_TOKE_C
28 #define new_constant(a,b,c,d,e,f,g) \
29 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
31 #define pl_yylval (PL_parser->yylval)
33 /* YYINITDEPTH -- initial size of the parser's stacks. */
34 #define YYINITDEPTH 200
36 /* XXX temporary backwards compatibility */
37 #define PL_lex_brackets (PL_parser->lex_brackets)
38 #define PL_lex_brackstack (PL_parser->lex_brackstack)
39 #define PL_lex_casemods (PL_parser->lex_casemods)
40 #define PL_lex_casestack (PL_parser->lex_casestack)
41 #define PL_lex_defer (PL_parser->lex_defer)
42 #define PL_lex_dojoin (PL_parser->lex_dojoin)
43 #define PL_lex_expect (PL_parser->lex_expect)
44 #define PL_lex_formbrack (PL_parser->lex_formbrack)
45 #define PL_lex_inpat (PL_parser->lex_inpat)
46 #define PL_lex_inwhat (PL_parser->lex_inwhat)
47 #define PL_lex_op (PL_parser->lex_op)
48 #define PL_lex_repl (PL_parser->lex_repl)
49 #define PL_lex_starts (PL_parser->lex_starts)
50 #define PL_lex_stuff (PL_parser->lex_stuff)
51 #define PL_multi_start (PL_parser->multi_start)
52 #define PL_multi_open (PL_parser->multi_open)
53 #define PL_multi_close (PL_parser->multi_close)
54 #define PL_pending_ident (PL_parser->pending_ident)
55 #define PL_preambled (PL_parser->preambled)
56 #define PL_sublex_info (PL_parser->sublex_info)
57 #define PL_linestr (PL_parser->linestr)
58 #define PL_expect (PL_parser->expect)
59 #define PL_copline (PL_parser->copline)
60 #define PL_bufptr (PL_parser->bufptr)
61 #define PL_oldbufptr (PL_parser->oldbufptr)
62 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
63 #define PL_linestart (PL_parser->linestart)
64 #define PL_bufend (PL_parser->bufend)
65 #define PL_last_uni (PL_parser->last_uni)
66 #define PL_last_lop (PL_parser->last_lop)
67 #define PL_last_lop_op (PL_parser->last_lop_op)
68 #define PL_lex_state (PL_parser->lex_state)
69 #define PL_rsfp (PL_parser->rsfp)
70 #define PL_rsfp_filters (PL_parser->rsfp_filters)
71 #define PL_in_my (PL_parser->in_my)
72 #define PL_in_my_stash (PL_parser->in_my_stash)
73 #define PL_tokenbuf (PL_parser->tokenbuf)
74 #define PL_multi_end (PL_parser->multi_end)
75 #define PL_error_count (PL_parser->error_count)
78 # define PL_endwhite (PL_parser->endwhite)
79 # define PL_faketokens (PL_parser->faketokens)
80 # define PL_lasttoke (PL_parser->lasttoke)
81 # define PL_nextwhite (PL_parser->nextwhite)
82 # define PL_realtokenstart (PL_parser->realtokenstart)
83 # define PL_skipwhite (PL_parser->skipwhite)
84 # define PL_thisclose (PL_parser->thisclose)
85 # define PL_thismad (PL_parser->thismad)
86 # define PL_thisopen (PL_parser->thisopen)
87 # define PL_thisstuff (PL_parser->thisstuff)
88 # define PL_thistoken (PL_parser->thistoken)
89 # define PL_thiswhite (PL_parser->thiswhite)
90 # define PL_thiswhite (PL_parser->thiswhite)
91 # define PL_nexttoke (PL_parser->nexttoke)
92 # define PL_curforce (PL_parser->curforce)
94 # define PL_nexttoke (PL_parser->nexttoke)
95 # define PL_nexttype (PL_parser->nexttype)
96 # define PL_nextval (PL_parser->nextval)
99 /* This can't be done with embed.fnc, because struct yy_parser contains a
100 member named pending_ident, which clashes with the generated #define */
102 S_pending_ident(pTHX);
104 static const char ident_too_long[] = "Identifier too long";
107 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
108 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
110 # define CURMAD(slot,sv)
111 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
114 #define XFAKEBRACK 128
115 #define XENUMMASK 127
117 #ifdef USE_UTF8_SCRIPTS
118 # define UTF (!IN_BYTES)
120 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
123 /* The maximum number of characters preceding the unrecognized one to display */
124 #define UNRECOGNIZED_PRECEDE_COUNT 10
126 /* In variables named $^X, these are the legal values for X.
127 * 1999-02-27 mjd-perl-patch@plover.com */
128 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
130 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
132 /* LEX_* are values for PL_lex_state, the state of the lexer.
133 * They are arranged oddly so that the guard on the switch statement
134 * can get by with a single comparison (if the compiler is smart enough).
137 /* #define LEX_NOTPARSING 11 is done in perl.h. */
139 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
140 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
141 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
142 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
143 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
145 /* at end of code, eg "$x" followed by: */
146 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
147 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
149 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
150 string or after \E, $foo, etc */
151 #define LEX_INTERPCONST 2 /* NOT USED */
152 #define LEX_FORMLINE 1 /* expecting a format line */
153 #define LEX_KNOWNEXT 0 /* next token known; just return it */
157 static const char* const lex_state_names[] = {
176 #include "keywords.h"
178 /* CLINE is a macro that ensures PL_copline has a sane value */
183 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
186 # define SKIPSPACE0(s) skipspace0(s)
187 # define SKIPSPACE1(s) skipspace1(s)
188 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
189 # define PEEKSPACE(s) skipspace2(s,0)
191 # define SKIPSPACE0(s) skipspace(s)
192 # define SKIPSPACE1(s) skipspace(s)
193 # define SKIPSPACE2(s,tsv) skipspace(s)
194 # define PEEKSPACE(s) skipspace(s)
198 * Convenience functions to return different tokens and prime the
199 * lexer for the next token. They all take an argument.
201 * TOKEN : generic token (used for '(', DOLSHARP, etc)
202 * OPERATOR : generic operator
203 * AOPERATOR : assignment operator
204 * PREBLOCK : beginning the block after an if, while, foreach, ...
205 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
206 * PREREF : *EXPR where EXPR is not a simple identifier
207 * TERM : expression term
208 * LOOPX : loop exiting command (goto, last, dump, etc)
209 * FTST : file test operator
210 * FUN0 : zero-argument function
211 * FUN1 : not used, except for not, which isn't a UNIOP
212 * BOop : bitwise or or xor
214 * SHop : shift operator
215 * PWop : power operator
216 * PMop : pattern-matching operator
217 * Aop : addition-level operator
218 * Mop : multiplication-level operator
219 * Eop : equality-testing operator
220 * Rop : relational operator <= != gt
222 * Also see LOP and lop() below.
225 #ifdef DEBUGGING /* Serve -DT. */
226 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
228 # define REPORT(retval) (retval)
231 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
232 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
233 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
234 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
235 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
236 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
237 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
238 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
239 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
240 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
241 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
242 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
243 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
244 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
245 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
246 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
247 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
248 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
249 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
250 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
252 /* This bit of chicanery makes a unary function followed by
253 * a parenthesis into a function with one argument, highest precedence.
254 * The UNIDOR macro is for unary functions that can be followed by the //
255 * operator (such as C<shift // 0>).
257 #define UNI2(f,x) { \
258 pl_yylval.ival = f; \
261 PL_last_uni = PL_oldbufptr; \
262 PL_last_lop_op = f; \
264 return REPORT( (int)FUNC1 ); \
266 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
268 #define UNI(f) UNI2(f,XTERM)
269 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
271 #define UNIBRACK(f) { \
272 pl_yylval.ival = f; \
274 PL_last_uni = PL_oldbufptr; \
276 return REPORT( (int)FUNC1 ); \
278 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
281 /* grandfather return to old style */
282 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
286 /* how to interpret the pl_yylval associated with the token */
290 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
296 static struct debug_tokens {
298 enum token_type type;
300 } const debug_tokens[] =
302 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
303 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
304 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
305 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
306 { ARROW, TOKENTYPE_NONE, "ARROW" },
307 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
308 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
309 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
310 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
311 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
312 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
313 { DO, TOKENTYPE_NONE, "DO" },
314 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
315 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
316 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
317 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
318 { ELSE, TOKENTYPE_NONE, "ELSE" },
319 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
320 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
321 { FOR, TOKENTYPE_IVAL, "FOR" },
322 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
323 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
324 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
325 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
326 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
327 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
328 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
329 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
330 { IF, TOKENTYPE_IVAL, "IF" },
331 { LABEL, TOKENTYPE_PVAL, "LABEL" },
332 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
333 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
334 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
335 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
336 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
337 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
338 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
339 { MY, TOKENTYPE_IVAL, "MY" },
340 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
341 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
342 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
343 { OROP, TOKENTYPE_IVAL, "OROP" },
344 { OROR, TOKENTYPE_NONE, "OROR" },
345 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
346 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
347 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
348 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
349 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
350 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
351 { PREINC, TOKENTYPE_NONE, "PREINC" },
352 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
353 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
354 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
355 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
356 { SUB, TOKENTYPE_NONE, "SUB" },
357 { THING, TOKENTYPE_OPVAL, "THING" },
358 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
359 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
360 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
361 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
362 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
363 { USE, TOKENTYPE_IVAL, "USE" },
364 { WHEN, TOKENTYPE_IVAL, "WHEN" },
365 { WHILE, TOKENTYPE_IVAL, "WHILE" },
366 { WORD, TOKENTYPE_OPVAL, "WORD" },
367 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
368 { 0, TOKENTYPE_NONE, NULL }
371 /* dump the returned token in rv, plus any optional arg in pl_yylval */
374 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
378 PERL_ARGS_ASSERT_TOKEREPORT;
381 const char *name = NULL;
382 enum token_type type = TOKENTYPE_NONE;
383 const struct debug_tokens *p;
384 SV* const report = newSVpvs("<== ");
386 for (p = debug_tokens; p->token; p++) {
387 if (p->token == (int)rv) {
394 Perl_sv_catpv(aTHX_ report, name);
395 else if ((char)rv > ' ' && (char)rv < '~')
396 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
398 sv_catpvs(report, "EOF");
400 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
403 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
406 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
408 case TOKENTYPE_OPNUM:
409 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
410 PL_op_name[lvalp->ival]);
413 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
415 case TOKENTYPE_OPVAL:
417 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
418 PL_op_name[lvalp->opval->op_type]);
419 if (lvalp->opval->op_type == OP_CONST) {
420 Perl_sv_catpvf(aTHX_ report, " %s",
421 SvPEEK(cSVOPx_sv(lvalp->opval)));
426 sv_catpvs(report, "(opval=null)");
429 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
435 /* print the buffer with suitable escapes */
438 S_printbuf(pTHX_ const char *const fmt, const char *const s)
440 SV* const tmp = newSVpvs("");
442 PERL_ARGS_ASSERT_PRINTBUF;
444 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
451 S_deprecate_commaless_var_list(pTHX) {
453 deprecate("comma-less variable list");
454 return REPORT(','); /* grandfather non-comma-format format */
460 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
461 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
465 S_ao(pTHX_ int toketype)
468 if (*PL_bufptr == '=') {
470 if (toketype == ANDAND)
471 pl_yylval.ival = OP_ANDASSIGN;
472 else if (toketype == OROR)
473 pl_yylval.ival = OP_ORASSIGN;
474 else if (toketype == DORDOR)
475 pl_yylval.ival = OP_DORASSIGN;
483 * When Perl expects an operator and finds something else, no_op
484 * prints the warning. It always prints "<something> found where
485 * operator expected. It prints "Missing semicolon on previous line?"
486 * if the surprise occurs at the start of the line. "do you need to
487 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
488 * where the compiler doesn't know if foo is a method call or a function.
489 * It prints "Missing operator before end of line" if there's nothing
490 * after the missing operator, or "... before <...>" if there is something
491 * after the missing operator.
495 S_no_op(pTHX_ const char *const what, char *s)
498 char * const oldbp = PL_bufptr;
499 const bool is_first = (PL_oldbufptr == PL_linestart);
501 PERL_ARGS_ASSERT_NO_OP;
507 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
508 if (ckWARN_d(WARN_SYNTAX)) {
510 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
511 "\t(Missing semicolon on previous line?)\n");
512 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
514 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
516 if (t < PL_bufptr && isSPACE(*t))
517 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
518 "\t(Do you need to predeclare %.*s?)\n",
519 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
523 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
524 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
532 * Complain about missing quote/regexp/heredoc terminator.
533 * If it's called with NULL then it cauterizes the line buffer.
534 * If we're in a delimited string and the delimiter is a control
535 * character, it's reformatted into a two-char sequence like ^C.
540 S_missingterm(pTHX_ char *s)
546 char * const nl = strrchr(s,'\n');
550 else if (isCNTRL(PL_multi_close)) {
552 tmpbuf[1] = (char)toCTRL(PL_multi_close);
557 *tmpbuf = (char)PL_multi_close;
561 q = strchr(s,'"') ? '\'' : '"';
562 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
565 #define FEATURE_IS_ENABLED(name) \
566 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
567 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
568 /* The longest string we pass in. */
569 #define MAX_FEATURE_LEN (sizeof("switch")-1)
572 * S_feature_is_enabled
573 * Check whether the named feature is enabled.
576 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
579 HV * const hinthv = GvHV(PL_hintgv);
580 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
582 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
584 assert(namelen <= MAX_FEATURE_LEN);
585 memcpy(&he_name[8], name, namelen);
587 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
591 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
592 * utf16-to-utf8-reversed.
595 #ifdef PERL_CR_FILTER
599 register const char *s = SvPVX_const(sv);
600 register const char * const e = s + SvCUR(sv);
602 PERL_ARGS_ASSERT_STRIP_RETURN;
604 /* outer loop optimized to do nothing if there are no CR-LFs */
606 if (*s++ == '\r' && *s == '\n') {
607 /* hit a CR-LF, need to copy the rest */
608 register char *d = s - 1;
611 if (*s == '\r' && s[1] == '\n')
622 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
624 const I32 count = FILTER_READ(idx+1, sv, maxlen);
625 if (count > 0 && !maxlen)
636 * Create a parser object and initialise its parser and lexer fields
638 * rsfp is the opened file handle to read from (if any),
640 * line holds any initial content already read from the file (or in
641 * the case of no file, such as an eval, the whole contents);
643 * new_filter indicates that this is a new file and it shouldn't inherit
644 * the filters from the current parser (ie require).
648 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
651 const char *s = NULL;
653 yy_parser *parser, *oparser;
655 /* create and initialise a parser */
657 Newxz(parser, 1, yy_parser);
658 parser->old_parser = oparser = PL_parser;
661 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
662 parser->ps = parser->stack;
663 parser->stack_size = YYINITDEPTH;
665 parser->stack->state = 0;
666 parser->yyerrstatus = 0;
667 parser->yychar = YYEMPTY; /* Cause a token to be read. */
669 /* on scope exit, free this parser and restore any outer one */
671 parser->saved_curcop = PL_curcop;
673 /* initialise lexer state */
676 parser->curforce = -1;
678 parser->nexttoke = 0;
680 parser->error_count = oparser ? oparser->error_count : 0;
681 parser->copline = NOLINE;
682 parser->lex_state = LEX_NORMAL;
683 parser->expect = XSTATE;
685 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
686 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
688 Newx(parser->lex_brackstack, 120, char);
689 Newx(parser->lex_casestack, 12, char);
690 *parser->lex_casestack = '\0';
693 s = SvPV_const(line, len);
699 parser->linestr = newSVpvs("\n;");
700 } else if (SvREADONLY(line) || s[len-1] != ';') {
701 parser->linestr = newSVsv(line);
703 sv_catpvs(parser->linestr, "\n;");
706 SvREFCNT_inc_simple_void_NN(line);
707 parser->linestr = line;
709 parser->oldoldbufptr =
712 parser->linestart = SvPVX(parser->linestr);
713 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
714 parser->last_lop = parser->last_uni = NULL;
718 /* delete a parser object */
721 Perl_parser_free(pTHX_ const yy_parser *parser)
723 PERL_ARGS_ASSERT_PARSER_FREE;
725 PL_curcop = parser->saved_curcop;
726 SvREFCNT_dec(parser->linestr);
728 if (parser->rsfp == PerlIO_stdin())
729 PerlIO_clearerr(parser->rsfp);
730 else if (parser->rsfp && (!parser->old_parser ||
731 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
732 PerlIO_close(parser->rsfp);
733 SvREFCNT_dec(parser->rsfp_filters);
735 Safefree(parser->stack);
736 Safefree(parser->lex_brackstack);
737 Safefree(parser->lex_casestack);
738 PL_parser = parser->old_parser;
745 * Finalizer for lexing operations. Must be called when the parser is
746 * done with the lexer.
753 PL_doextract = FALSE;
758 * This subroutine has nothing to do with tilting, whether at windmills
759 * or pinball tables. Its name is short for "increment line". It
760 * increments the current line number in CopLINE(PL_curcop) and checks
761 * to see whether the line starts with a comment of the form
762 * # line 500 "foo.pm"
763 * If so, it sets the current line number and file to the values in the comment.
767 S_incline(pTHX_ const char *s)
774 PERL_ARGS_ASSERT_INCLINE;
776 CopLINE_inc(PL_curcop);
779 while (SPACE_OR_TAB(*s))
781 if (strnEQ(s, "line", 4))
785 if (SPACE_OR_TAB(*s))
789 while (SPACE_OR_TAB(*s))
797 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
799 while (SPACE_OR_TAB(*s))
801 if (*s == '"' && (t = strchr(s+1, '"'))) {
811 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
813 if (*e != '\n' && *e != '\0')
814 return; /* false alarm */
817 const STRLEN len = t - s;
819 SV *const temp_sv = CopFILESV(PL_curcop);
825 tmplen = SvCUR(temp_sv);
831 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
832 /* must copy *{"::_<(eval N)[oldfilename:L]"}
833 * to *{"::_<newfilename"} */
834 /* However, the long form of evals is only turned on by the
835 debugger - usually they're "(eval %lu)" */
839 STRLEN tmplen2 = len;
840 if (tmplen + 2 <= sizeof smallbuf)
843 Newx(tmpbuf, tmplen + 2, char);
846 memcpy(tmpbuf + 2, cf, tmplen);
848 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
853 if (tmplen2 + 2 <= sizeof smallbuf)
856 Newx(tmpbuf2, tmplen2 + 2, char);
858 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
859 /* Either they malloc'd it, or we malloc'd it,
860 so no prefix is present in ours. */
865 memcpy(tmpbuf2 + 2, s, tmplen2);
868 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
870 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
871 /* adjust ${"::_<newfilename"} to store the new file name */
872 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
873 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
874 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
877 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
879 if (tmpbuf != smallbuf) Safefree(tmpbuf);
882 CopFILE_free(PL_curcop);
883 CopFILE_setn(PL_curcop, s, len);
885 CopLINE_set(PL_curcop, atoi(n)-1);
889 /* skip space before PL_thistoken */
892 S_skipspace0(pTHX_ register char *s)
894 PERL_ARGS_ASSERT_SKIPSPACE0;
901 PL_thiswhite = newSVpvs("");
902 sv_catsv(PL_thiswhite, PL_skipwhite);
903 sv_free(PL_skipwhite);
906 PL_realtokenstart = s - SvPVX(PL_linestr);
910 /* skip space after PL_thistoken */
913 S_skipspace1(pTHX_ register char *s)
915 const char *start = s;
916 I32 startoff = start - SvPVX(PL_linestr);
918 PERL_ARGS_ASSERT_SKIPSPACE1;
923 start = SvPVX(PL_linestr) + startoff;
924 if (!PL_thistoken && PL_realtokenstart >= 0) {
925 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
926 PL_thistoken = newSVpvn(tstart, start - tstart);
928 PL_realtokenstart = -1;
931 PL_nextwhite = newSVpvs("");
932 sv_catsv(PL_nextwhite, PL_skipwhite);
933 sv_free(PL_skipwhite);
940 S_skipspace2(pTHX_ register char *s, SV **svp)
943 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
944 const I32 startoff = s - SvPVX(PL_linestr);
946 PERL_ARGS_ASSERT_SKIPSPACE2;
949 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
950 if (!PL_madskills || !svp)
952 start = SvPVX(PL_linestr) + startoff;
953 if (!PL_thistoken && PL_realtokenstart >= 0) {
954 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
955 PL_thistoken = newSVpvn(tstart, start - tstart);
956 PL_realtokenstart = -1;
961 sv_setsv(*svp, PL_skipwhite);
962 sv_free(PL_skipwhite);
971 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
973 AV *av = CopFILEAVx(PL_curcop);
975 SV * const sv = newSV_type(SVt_PVMG);
977 sv_setsv(sv, orig_sv);
979 sv_setpvn(sv, buf, len);
982 av_store(av, (I32)CopLINE(PL_curcop), sv);
988 * Called to gobble the appropriate amount and type of whitespace.
989 * Skips comments as well.
993 S_skipspace(pTHX_ register char *s)
998 int startoff = s - SvPVX(PL_linestr);
1000 PERL_ARGS_ASSERT_SKIPSPACE;
1003 sv_free(PL_skipwhite);
1007 PERL_ARGS_ASSERT_SKIPSPACE;
1009 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1010 while (s < PL_bufend && SPACE_OR_TAB(*s))
1020 SSize_t oldprevlen, oldoldprevlen;
1021 SSize_t oldloplen = 0, oldunilen = 0;
1022 while (s < PL_bufend && isSPACE(*s)) {
1023 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1028 if (s < PL_bufend && *s == '#') {
1029 while (s < PL_bufend && *s != '\n')
1031 if (s < PL_bufend) {
1033 if (PL_in_eval && !PL_rsfp) {
1040 /* only continue to recharge the buffer if we're at the end
1041 * of the buffer, we're not reading from a source filter, and
1042 * we're in normal lexing mode
1044 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1045 PL_lex_state == LEX_FORMLINE)
1052 /* try to recharge the buffer */
1054 curoff = s - SvPVX(PL_linestr);
1057 if ((s = filter_gets(PL_linestr, (prevlen = SvCUR(PL_linestr))))
1061 if (PL_madskills && curoff != startoff) {
1063 PL_skipwhite = newSVpvs("");
1064 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1068 /* mustn't throw out old stuff yet if madpropping */
1069 SvCUR(PL_linestr) = curoff;
1070 s = SvPVX(PL_linestr) + curoff;
1072 if (curoff && s[-1] == '\n')
1076 /* end of file. Add on the -p or -n magic */
1077 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1080 sv_catpvs(PL_linestr,
1081 ";}continue{print or die qq(-p destination: $!\\n);}");
1083 sv_setpvs(PL_linestr,
1084 ";}continue{print or die qq(-p destination: $!\\n);}");
1086 PL_minus_n = PL_minus_p = 0;
1088 else if (PL_minus_n) {
1090 sv_catpvs(PL_linestr, ";}");
1092 sv_setpvs(PL_linestr, ";}");
1098 sv_catpvs(PL_linestr,";");
1100 sv_setpvs(PL_linestr,";");
1103 /* reset variables for next time we lex */
1104 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1110 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1111 PL_last_lop = PL_last_uni = NULL;
1113 /* Close the filehandle. Could be from
1114 * STDIN, or a regular file. If we were reading code from
1115 * STDIN (because the commandline held no -e or filename)
1116 * then we don't close it, we reset it so the code can
1117 * read from STDIN too.
1120 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1121 PerlIO_clearerr(PL_rsfp);
1123 (void)PerlIO_close(PL_rsfp);
1128 /* not at end of file, so we only read another line */
1129 /* make corresponding updates to old pointers, for yyerror() */
1130 oldprevlen = PL_oldbufptr - PL_bufend;
1131 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1133 oldunilen = PL_last_uni - PL_bufend;
1135 oldloplen = PL_last_lop - PL_bufend;
1136 PL_linestart = PL_bufptr = s + prevlen;
1137 PL_bufend = s + SvCUR(PL_linestr);
1139 PL_oldbufptr = s + oldprevlen;
1140 PL_oldoldbufptr = s + oldoldprevlen;
1142 PL_last_uni = s + oldunilen;
1144 PL_last_lop = s + oldloplen;
1147 /* debugger active and we're not compiling the debugger code,
1148 * so store the line into the debugger's array of lines
1150 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
1151 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1158 PL_skipwhite = newSVpvs("");
1159 curoff = s - SvPVX(PL_linestr);
1160 if (curoff - startoff)
1161 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1170 * Check the unary operators to ensure there's no ambiguity in how they're
1171 * used. An ambiguous piece of code would be:
1173 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1174 * the +5 is its argument.
1184 if (PL_oldoldbufptr != PL_last_uni)
1186 while (isSPACE(*PL_last_uni))
1189 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1191 if ((t = strchr(s, '(')) && t < PL_bufptr)
1194 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1195 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1196 (int)(s - PL_last_uni), PL_last_uni);
1200 * LOP : macro to build a list operator. Its behaviour has been replaced
1201 * with a subroutine, S_lop() for which LOP is just another name.
1204 #define LOP(f,x) return lop(f,x,s)
1208 * Build a list operator (or something that might be one). The rules:
1209 * - if we have a next token, then it's a list operator [why?]
1210 * - if the next thing is an opening paren, then it's a function
1211 * - else it's a list operator
1215 S_lop(pTHX_ I32 f, int x, char *s)
1219 PERL_ARGS_ASSERT_LOP;
1225 PL_last_lop = PL_oldbufptr;
1226 PL_last_lop_op = (OPCODE)f;
1229 return REPORT(LSTOP);
1232 return REPORT(LSTOP);
1235 return REPORT(FUNC);
1238 return REPORT(FUNC);
1240 return REPORT(LSTOP);
1246 * Sets up for an eventual force_next(). start_force(0) basically does
1247 * an unshift, while start_force(-1) does a push. yylex removes items
1252 S_start_force(pTHX_ int where)
1256 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1257 where = PL_lasttoke;
1258 assert(PL_curforce < 0 || PL_curforce == where);
1259 if (PL_curforce != where) {
1260 for (i = PL_lasttoke; i > where; --i) {
1261 PL_nexttoke[i] = PL_nexttoke[i-1];
1265 if (PL_curforce < 0) /* in case of duplicate start_force() */
1266 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1267 PL_curforce = where;
1270 curmad('^', newSVpvs(""));
1271 CURMAD('_', PL_nextwhite);
1276 S_curmad(pTHX_ char slot, SV *sv)
1282 if (PL_curforce < 0)
1283 where = &PL_thismad;
1285 where = &PL_nexttoke[PL_curforce].next_mad;
1291 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1293 else if (PL_encoding) {
1294 sv_recode_to_utf8(sv, PL_encoding);
1299 /* keep a slot open for the head of the list? */
1300 if (slot != '_' && *where && (*where)->mad_key == '^') {
1301 (*where)->mad_key = slot;
1302 sv_free(MUTABLE_SV(((*where)->mad_val)));
1303 (*where)->mad_val = (void*)sv;
1306 addmad(newMADsv(slot, sv), where, 0);
1309 # define start_force(where) NOOP
1310 # define curmad(slot, sv) NOOP
1315 * When the lexer realizes it knows the next token (for instance,
1316 * it is reordering tokens for the parser) then it can call S_force_next
1317 * to know what token to return the next time the lexer is called. Caller
1318 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1319 * and possibly PL_expect to ensure the lexer handles the token correctly.
1323 S_force_next(pTHX_ I32 type)
1328 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1329 tokereport(type, &NEXTVAL_NEXTTOKE);
1333 if (PL_curforce < 0)
1334 start_force(PL_lasttoke);
1335 PL_nexttoke[PL_curforce].next_type = type;
1336 if (PL_lex_state != LEX_KNOWNEXT)
1337 PL_lex_defer = PL_lex_state;
1338 PL_lex_state = LEX_KNOWNEXT;
1339 PL_lex_expect = PL_expect;
1342 PL_nexttype[PL_nexttoke] = type;
1344 if (PL_lex_state != LEX_KNOWNEXT) {
1345 PL_lex_defer = PL_lex_state;
1346 PL_lex_expect = PL_expect;
1347 PL_lex_state = LEX_KNOWNEXT;
1353 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1356 SV * const sv = newSVpvn_utf8(start, len,
1359 && !is_ascii_string((const U8*)start, len)
1360 && is_utf8_string((const U8*)start, len));
1366 * When the lexer knows the next thing is a word (for instance, it has
1367 * just seen -> and it knows that the next char is a word char, then
1368 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1372 * char *start : buffer position (must be within PL_linestr)
1373 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1374 * int check_keyword : if true, Perl checks to make sure the word isn't
1375 * a keyword (do this if the word is a label, e.g. goto FOO)
1376 * int allow_pack : if true, : characters will also be allowed (require,
1377 * use, etc. do this)
1378 * int allow_initial_tick : used by the "sub" lexer only.
1382 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1388 PERL_ARGS_ASSERT_FORCE_WORD;
1390 start = SKIPSPACE1(start);
1392 if (isIDFIRST_lazy_if(s,UTF) ||
1393 (allow_pack && *s == ':') ||
1394 (allow_initial_tick && *s == '\'') )
1396 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1397 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1399 start_force(PL_curforce);
1401 curmad('X', newSVpvn(start,s-start));
1402 if (token == METHOD) {
1407 PL_expect = XOPERATOR;
1411 curmad('g', newSVpvs( "forced" ));
1412 NEXTVAL_NEXTTOKE.opval
1413 = (OP*)newSVOP(OP_CONST,0,
1414 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1415 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1423 * Called when the lexer wants $foo *foo &foo etc, but the program
1424 * text only contains the "foo" portion. The first argument is a pointer
1425 * to the "foo", and the second argument is the type symbol to prefix.
1426 * Forces the next token to be a "WORD".
1427 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1431 S_force_ident(pTHX_ register const char *s, int kind)
1435 PERL_ARGS_ASSERT_FORCE_IDENT;
1438 const STRLEN len = strlen(s);
1439 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1440 start_force(PL_curforce);
1441 NEXTVAL_NEXTTOKE.opval = o;
1444 o->op_private = OPpCONST_ENTERED;
1445 /* XXX see note in pp_entereval() for why we forgo typo
1446 warnings if the symbol must be introduced in an eval.
1448 gv_fetchpvn_flags(s, len,
1449 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1451 kind == '$' ? SVt_PV :
1452 kind == '@' ? SVt_PVAV :
1453 kind == '%' ? SVt_PVHV :
1461 Perl_str_to_version(pTHX_ SV *sv)
1466 const char *start = SvPV_const(sv,len);
1467 const char * const end = start + len;
1468 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1470 PERL_ARGS_ASSERT_STR_TO_VERSION;
1472 while (start < end) {
1476 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1481 retval += ((NV)n)/nshift;
1490 * Forces the next token to be a version number.
1491 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1492 * and if "guessing" is TRUE, then no new token is created (and the caller
1493 * must use an alternative parsing method).
1497 S_force_version(pTHX_ char *s, int guessing)
1503 I32 startoff = s - SvPVX(PL_linestr);
1506 PERL_ARGS_ASSERT_FORCE_VERSION;
1514 while (isDIGIT(*d) || *d == '_' || *d == '.')
1518 start_force(PL_curforce);
1519 curmad('X', newSVpvn(s,d-s));
1522 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1524 s = scan_num(s, &pl_yylval);
1525 version = pl_yylval.opval;
1526 ver = cSVOPx(version)->op_sv;
1527 if (SvPOK(ver) && !SvNIOK(ver)) {
1528 SvUPGRADE(ver, SVt_PVNV);
1529 SvNV_set(ver, str_to_version(ver));
1530 SvNOK_on(ver); /* hint that it is a version */
1533 else if (guessing) {
1536 sv_free(PL_nextwhite); /* let next token collect whitespace */
1538 s = SvPVX(PL_linestr) + startoff;
1546 if (PL_madskills && !version) {
1547 sv_free(PL_nextwhite); /* let next token collect whitespace */
1549 s = SvPVX(PL_linestr) + startoff;
1552 /* NOTE: The parser sees the package name and the VERSION swapped */
1553 start_force(PL_curforce);
1554 NEXTVAL_NEXTTOKE.opval = version;
1562 * Tokenize a quoted string passed in as an SV. It finds the next
1563 * chunk, up to end of string or a backslash. It may make a new
1564 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1569 S_tokeq(pTHX_ SV *sv)
1573 register char *send;
1578 PERL_ARGS_ASSERT_TOKEQ;
1583 s = SvPV_force(sv, len);
1584 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1587 while (s < send && *s != '\\')
1592 if ( PL_hints & HINT_NEW_STRING ) {
1593 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1597 if (s + 1 < send && (s[1] == '\\'))
1598 s++; /* all that, just for this */
1603 SvCUR_set(sv, d - SvPVX_const(sv));
1605 if ( PL_hints & HINT_NEW_STRING )
1606 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1611 * Now come three functions related to double-quote context,
1612 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1613 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1614 * interact with PL_lex_state, and create fake ( ... ) argument lists
1615 * to handle functions and concatenation.
1616 * They assume that whoever calls them will be setting up a fake
1617 * join call, because each subthing puts a ',' after it. This lets
1620 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1622 * (I'm not sure whether the spurious commas at the end of lcfirst's
1623 * arguments and join's arguments are created or not).
1628 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1630 * Pattern matching will set PL_lex_op to the pattern-matching op to
1631 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1633 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1635 * Everything else becomes a FUNC.
1637 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1638 * had an OP_CONST or OP_READLINE). This just sets us up for a
1639 * call to S_sublex_push().
1643 S_sublex_start(pTHX)
1646 register const I32 op_type = pl_yylval.ival;
1648 if (op_type == OP_NULL) {
1649 pl_yylval.opval = PL_lex_op;
1653 if (op_type == OP_CONST || op_type == OP_READLINE) {
1654 SV *sv = tokeq(PL_lex_stuff);
1656 if (SvTYPE(sv) == SVt_PVIV) {
1657 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1659 const char * const p = SvPV_const(sv, len);
1660 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1664 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1665 PL_lex_stuff = NULL;
1666 /* Allow <FH> // "foo" */
1667 if (op_type == OP_READLINE)
1668 PL_expect = XTERMORDORDOR;
1671 else if (op_type == OP_BACKTICK && PL_lex_op) {
1672 /* readpipe() vas overriden */
1673 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1674 pl_yylval.opval = PL_lex_op;
1676 PL_lex_stuff = NULL;
1680 PL_sublex_info.super_state = PL_lex_state;
1681 PL_sublex_info.sub_inwhat = (U16)op_type;
1682 PL_sublex_info.sub_op = PL_lex_op;
1683 PL_lex_state = LEX_INTERPPUSH;
1687 pl_yylval.opval = PL_lex_op;
1697 * Create a new scope to save the lexing state. The scope will be
1698 * ended in S_sublex_done. Returns a '(', starting the function arguments
1699 * to the uc, lc, etc. found before.
1700 * Sets PL_lex_state to LEX_INTERPCONCAT.
1709 PL_lex_state = PL_sublex_info.super_state;
1710 SAVEBOOL(PL_lex_dojoin);
1711 SAVEI32(PL_lex_brackets);
1712 SAVEI32(PL_lex_casemods);
1713 SAVEI32(PL_lex_starts);
1714 SAVEI8(PL_lex_state);
1715 SAVEVPTR(PL_lex_inpat);
1716 SAVEI16(PL_lex_inwhat);
1717 SAVECOPLINE(PL_curcop);
1718 SAVEPPTR(PL_bufptr);
1719 SAVEPPTR(PL_bufend);
1720 SAVEPPTR(PL_oldbufptr);
1721 SAVEPPTR(PL_oldoldbufptr);
1722 SAVEPPTR(PL_last_lop);
1723 SAVEPPTR(PL_last_uni);
1724 SAVEPPTR(PL_linestart);
1725 SAVESPTR(PL_linestr);
1726 SAVEGENERICPV(PL_lex_brackstack);
1727 SAVEGENERICPV(PL_lex_casestack);
1729 PL_linestr = PL_lex_stuff;
1730 PL_lex_stuff = NULL;
1732 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1733 = SvPVX(PL_linestr);
1734 PL_bufend += SvCUR(PL_linestr);
1735 PL_last_lop = PL_last_uni = NULL;
1736 SAVEFREESV(PL_linestr);
1738 PL_lex_dojoin = FALSE;
1739 PL_lex_brackets = 0;
1740 Newx(PL_lex_brackstack, 120, char);
1741 Newx(PL_lex_casestack, 12, char);
1742 PL_lex_casemods = 0;
1743 *PL_lex_casestack = '\0';
1745 PL_lex_state = LEX_INTERPCONCAT;
1746 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1748 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1749 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1750 PL_lex_inpat = PL_sublex_info.sub_op;
1752 PL_lex_inpat = NULL;
1759 * Restores lexer state after a S_sublex_push.
1766 if (!PL_lex_starts++) {
1767 SV * const sv = newSVpvs("");
1768 if (SvUTF8(PL_linestr))
1770 PL_expect = XOPERATOR;
1771 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1775 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1776 PL_lex_state = LEX_INTERPCASEMOD;
1780 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1781 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1782 PL_linestr = PL_lex_repl;
1784 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1785 PL_bufend += SvCUR(PL_linestr);
1786 PL_last_lop = PL_last_uni = NULL;
1787 SAVEFREESV(PL_linestr);
1788 PL_lex_dojoin = FALSE;
1789 PL_lex_brackets = 0;
1790 PL_lex_casemods = 0;
1791 *PL_lex_casestack = '\0';
1793 if (SvEVALED(PL_lex_repl)) {
1794 PL_lex_state = LEX_INTERPNORMAL;
1796 /* we don't clear PL_lex_repl here, so that we can check later
1797 whether this is an evalled subst; that means we rely on the
1798 logic to ensure sublex_done() is called again only via the
1799 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1802 PL_lex_state = LEX_INTERPCONCAT;
1812 PL_endwhite = newSVpvs("");
1813 sv_catsv(PL_endwhite, PL_thiswhite);
1817 sv_setpvs(PL_thistoken,"");
1819 PL_realtokenstart = -1;
1823 PL_bufend = SvPVX(PL_linestr);
1824 PL_bufend += SvCUR(PL_linestr);
1825 PL_expect = XOPERATOR;
1826 PL_sublex_info.sub_inwhat = 0;
1834 Extracts a pattern, double-quoted string, or transliteration. This
1837 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1838 processing a pattern (PL_lex_inpat is true), a transliteration
1839 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1841 Returns a pointer to the character scanned up to. If this is
1842 advanced from the start pointer supplied (i.e. if anything was
1843 successfully parsed), will leave an OP for the substring scanned
1844 in pl_yylval. Caller must intuit reason for not parsing further
1845 by looking at the next characters herself.
1849 double-quoted style: \r and \n
1850 regexp special ones: \D \s
1853 case and quoting: \U \Q \E
1854 stops on @ and $, but not for $ as tail anchor
1856 In transliterations:
1857 characters are VERY literal, except for - not at the start or end
1858 of the string, which indicates a range. If the range is in bytes,
1859 scan_const expands the range to the full set of intermediate
1860 characters. If the range is in utf8, the hyphen is replaced with
1861 a certain range mark which will be handled by pmtrans() in op.c.
1863 In double-quoted strings:
1865 double-quoted style: \r and \n
1867 deprecated backrefs: \1 (in substitution replacements)
1868 case and quoting: \U \Q \E
1871 scan_const does *not* construct ops to handle interpolated strings.
1872 It stops processing as soon as it finds an embedded $ or @ variable
1873 and leaves it to the caller to work out what's going on.
1875 embedded arrays (whether in pattern or not) could be:
1876 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1878 $ in double-quoted strings must be the symbol of an embedded scalar.
1880 $ in pattern could be $foo or could be tail anchor. Assumption:
1881 it's a tail anchor if $ is the last thing in the string, or if it's
1882 followed by one of "()| \r\n\t"
1884 \1 (backreferences) are turned into $1
1886 The structure of the code is
1887 while (there's a character to process) {
1888 handle transliteration ranges
1889 skip regexp comments /(?#comment)/ and codes /(?{code})/
1890 skip #-initiated comments in //x patterns
1891 check for embedded arrays
1892 check for embedded scalars
1894 leave intact backslashes from leaveit (below)
1895 deprecate \1 in substitution replacements
1896 handle string-changing backslashes \l \U \Q \E, etc.
1897 switch (what was escaped) {
1898 handle \- in a transliteration (becomes a literal -)
1899 handle \132 (octal characters)
1900 handle \x15 and \x{1234} (hex characters)
1901 handle \N{name} (named characters)
1902 handle \cV (control characters)
1903 handle printf-style backslashes (\f, \r, \n, etc)
1906 } (end if backslash)
1907 handle regular character
1908 } (end while character to read)
1913 S_scan_const(pTHX_ char *start)
1916 register char *send = PL_bufend; /* end of the constant */
1917 SV *sv = newSV(send - start); /* sv for the constant. See
1918 note below on sizing. */
1919 register char *s = start; /* start of the constant */
1920 register char *d = SvPVX(sv); /* destination for copies */
1921 bool dorange = FALSE; /* are we in a translit range? */
1922 bool didrange = FALSE; /* did we just finish a range? */
1923 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1924 I32 this_utf8 = UTF; /* Is the source string assumed
1925 to be UTF8? But, this can
1926 show as true when the source
1927 isn't utf8, as for example
1928 when it is entirely composed
1931 /* Note on sizing: The scanned constant is placed into sv, which is
1932 * initialized by newSV() assuming one byte of output for every byte of
1933 * input. This routine expects newSV() to allocate an extra byte for a
1934 * trailing NUL, which this routine will append if it gets to the end of
1935 * the input. There may be more bytes of input than output (eg., \N{LATIN
1936 * CAPITAL LETTER A}), or more output than input if the constant ends up
1937 * recoded to utf8, but each time a construct is found that might increase
1938 * the needed size, SvGROW() is called. Its size parameter each time is
1939 * based on the best guess estimate at the time, namely the length used so
1940 * far, plus the length the current construct will occupy, plus room for
1941 * the trailing NUL, plus one byte for every input byte still unscanned */
1945 UV literal_endpoint = 0;
1946 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1949 PERL_ARGS_ASSERT_SCAN_CONST;
1951 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1952 /* If we are doing a trans and we know we want UTF8 set expectation */
1953 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1954 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1958 while (s < send || dorange) {
1959 /* get transliterations out of the way (they're most literal) */
1960 if (PL_lex_inwhat == OP_TRANS) {
1961 /* expand a range A-Z to the full set of characters. AIE! */
1963 I32 i; /* current expanded character */
1964 I32 min; /* first character in range */
1965 I32 max; /* last character in range */
1976 char * const c = (char*)utf8_hop((U8*)d, -1);
1980 *c = (char)UTF_TO_NATIVE(0xff);
1981 /* mark the range as done, and continue */
1987 i = d - SvPVX_const(sv); /* remember current offset */
1990 SvLEN(sv) + (has_utf8 ?
1991 (512 - UTF_CONTINUATION_MARK +
1994 /* How many two-byte within 0..255: 128 in UTF-8,
1995 * 96 in UTF-8-mod. */
1997 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1999 d = SvPVX(sv) + i; /* refresh d after realloc */
2003 for (j = 0; j <= 1; j++) {
2004 char * const c = (char*)utf8_hop((U8*)d, -1);
2005 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2011 max = (U8)0xff; /* only to \xff */
2012 uvmax = uv; /* \x{100} to uvmax */
2014 d = c; /* eat endpoint chars */
2019 d -= 2; /* eat the first char and the - */
2020 min = (U8)*d; /* first char in range */
2021 max = (U8)d[1]; /* last char in range */
2028 "Invalid range \"%c-%c\" in transliteration operator",
2029 (char)min, (char)max);
2033 if (literal_endpoint == 2 &&
2034 ((isLOWER(min) && isLOWER(max)) ||
2035 (isUPPER(min) && isUPPER(max)))) {
2037 for (i = min; i <= max; i++)
2039 *d++ = NATIVE_TO_NEED(has_utf8,i);
2041 for (i = min; i <= max; i++)
2043 *d++ = NATIVE_TO_NEED(has_utf8,i);
2048 for (i = min; i <= max; i++)
2051 const U8 ch = (U8)NATIVE_TO_UTF(i);
2052 if (UNI_IS_INVARIANT(ch))
2055 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2056 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2065 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2067 *d++ = (char)UTF_TO_NATIVE(0xff);
2069 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2073 /* mark the range as done, and continue */
2077 literal_endpoint = 0;
2082 /* range begins (ignore - as first or last char) */
2083 else if (*s == '-' && s+1 < send && s != start) {
2085 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2092 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2102 literal_endpoint = 0;
2103 native_range = TRUE;
2108 /* if we get here, we're not doing a transliteration */
2110 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2111 except for the last char, which will be done separately. */
2112 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2114 while (s+1 < send && *s != ')')
2115 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2117 else if (s[2] == '{' /* This should match regcomp.c */
2118 || (s[2] == '?' && s[3] == '{'))
2121 char *regparse = s + (s[2] == '{' ? 3 : 4);
2124 while (count && (c = *regparse)) {
2125 if (c == '\\' && regparse[1])
2133 if (*regparse != ')')
2134 regparse--; /* Leave one char for continuation. */
2135 while (s < regparse)
2136 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2140 /* likewise skip #-initiated comments in //x patterns */
2141 else if (*s == '#' && PL_lex_inpat &&
2142 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2143 while (s+1 < send && *s != '\n')
2144 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2147 /* check for embedded arrays
2148 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2150 else if (*s == '@' && s[1]) {
2151 if (isALNUM_lazy_if(s+1,UTF))
2153 if (strchr(":'{$", s[1]))
2155 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2156 break; /* in regexp, neither @+ nor @- are interpolated */
2159 /* check for embedded scalars. only stop if we're sure it's a
2162 else if (*s == '$') {
2163 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2165 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2167 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2168 "Possible unintended interpolation of $\\ in regex");
2170 break; /* in regexp, $ might be tail anchor */
2174 /* End of else if chain - OP_TRANS rejoin rest */
2177 if (*s == '\\' && s+1 < send) {
2180 /* deprecate \1 in strings and substitution replacements */
2181 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2182 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2184 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2189 /* string-change backslash escapes */
2190 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2194 /* skip any other backslash escapes in a pattern */
2195 else if (PL_lex_inpat) {
2196 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2197 goto default_action;
2200 /* if we get here, it's either a quoted -, or a digit */
2203 /* quoted - in transliterations */
2205 if (PL_lex_inwhat == OP_TRANS) {
2212 if ((isALPHA(*s) || isDIGIT(*s)))
2213 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2214 "Unrecognized escape \\%c passed through",
2216 /* default action is to copy the quoted character */
2217 goto default_action;
2220 /* eg. \132 indicates the octal constant 0x132 */
2221 case '0': case '1': case '2': case '3':
2222 case '4': case '5': case '6': case '7':
2226 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2229 goto NUM_ESCAPE_INSERT;
2231 /* eg. \x24 indicates the hex constant 0x24 */
2235 char* const e = strchr(s, '}');
2236 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2237 PERL_SCAN_DISALLOW_PREFIX;
2242 yyerror("Missing right brace on \\x{}");
2246 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2252 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2253 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2259 /* Insert oct, hex, or \N{U+...} escaped character. There will
2260 * always be enough room in sv since such escapes will be
2261 * longer than any UTF-8 sequence they can end up as, except if
2262 * they force us to recode the rest of the string into utf8 */
2264 /* Here uv is the ordinal of the next character being added in
2265 * unicode (converted from native). (It has to be done before
2266 * here because \N is interpreted as unicode, and oct and hex
2268 if (!UNI_IS_INVARIANT(uv)) {
2269 if (!has_utf8 && uv > 255) {
2270 /* Might need to recode whatever we have accumulated so
2271 * far if it contains any chars variant in utf8 or
2274 SvCUR_set(sv, d - SvPVX_const(sv));
2277 /* See Note on sizing above. */
2278 sv_utf8_upgrade_flags_grow(sv,
2279 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2280 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2281 d = SvPVX(sv) + SvCUR(sv);
2286 d = (char*)uvuni_to_utf8((U8*)d, uv);
2287 if (PL_lex_inwhat == OP_TRANS &&
2288 PL_sublex_info.sub_op) {
2289 PL_sublex_info.sub_op->op_private |=
2290 (PL_lex_repl ? OPpTRANS_FROM_UTF
2294 if (uv > 255 && !dorange)
2295 native_range = FALSE;
2307 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2312 char* e = strchr(s, '}');
2318 yyerror("Missing right brace on \\N{}");
2322 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2323 /* \N{U+...} The ... is a unicode value even on EBCDIC
2325 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2326 PERL_SCAN_DISALLOW_PREFIX;
2329 uv = grok_hex(s, &len, &flags, NULL);
2330 if ( e > s && len != (STRLEN)(e - s) ) {
2334 goto NUM_ESCAPE_INSERT;
2336 res = newSVpvn(s + 1, e - s - 1);
2337 res = new_constant( NULL, 0, "charnames",
2338 res, NULL, s - 2, e - s + 3 );
2340 sv_utf8_upgrade(res);
2341 str = SvPV_const(res,len);
2342 #ifdef EBCDIC_NEVER_MIND
2343 /* charnames uses pack U and that has been
2344 * recently changed to do the below uni->native
2345 * mapping, so this would be redundant (and wrong,
2346 * the code point would be doubly converted).
2347 * But leave this in just in case the pack U change
2348 * gets revoked, but the semantics is still
2349 * desireable for charnames. --jhi */
2351 UV uv = utf8_to_uvchr((const U8*)str, 0);
2354 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2356 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2357 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2358 str = SvPV_const(res, len);
2362 /* If destination is not in utf8 but this new character is,
2363 * recode the dest to utf8 */
2364 if (!has_utf8 && SvUTF8(res)) {
2365 SvCUR_set(sv, d - SvPVX_const(sv));
2368 /* See Note on sizing above. */
2369 sv_utf8_upgrade_flags_grow(sv,
2370 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2371 len + (STRLEN)(send - s) + 1);
2372 d = SvPVX(sv) + SvCUR(sv);
2374 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2376 /* See Note on sizing above. (NOTE: SvCUR() is not set
2377 * correctly here). */
2378 const STRLEN off = d - SvPVX_const(sv);
2379 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2383 native_range = FALSE; /* \N{} is guessed to be Unicode */
2385 Copy(str, d, len, char);
2392 yyerror("Missing braces on \\N{}");
2395 /* \c is a control character */
2404 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2407 yyerror("Missing control char name in \\c");
2411 /* printf-style backslashes, formfeeds, newlines, etc */
2413 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2416 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2419 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2422 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2425 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2428 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2431 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2437 } /* end if (backslash) */
2444 /* If we started with encoded form, or already know we want it,
2445 then encode the next character */
2446 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2450 /* One might think that it is wasted effort in the case of the
2451 * source being utf8 (this_utf8 == TRUE) to take the next character
2452 * in the source, convert it to an unsigned value, and then convert
2453 * it back again. But the source has not been validated here. The
2454 * routine that does the conversion checks for errors like
2457 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2458 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2460 SvCUR_set(sv, d - SvPVX_const(sv));
2463 /* See Note on sizing above. */
2464 sv_utf8_upgrade_flags_grow(sv,
2465 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2466 need + (STRLEN)(send - s) + 1);
2467 d = SvPVX(sv) + SvCUR(sv);
2469 } else if (need > len) {
2470 /* encoded value larger than old, may need extra space (NOTE:
2471 * SvCUR() is not set correctly here). See Note on sizing
2473 const STRLEN off = d - SvPVX_const(sv);
2474 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2478 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2480 if (uv > 255 && !dorange)
2481 native_range = FALSE;
2485 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2487 } /* while loop to process each character */
2489 /* terminate the string and set up the sv */
2491 SvCUR_set(sv, d - SvPVX_const(sv));
2492 if (SvCUR(sv) >= SvLEN(sv))
2493 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2496 if (PL_encoding && !has_utf8) {
2497 sv_recode_to_utf8(sv, PL_encoding);
2503 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2504 PL_sublex_info.sub_op->op_private |=
2505 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2509 /* shrink the sv if we allocated more than we used */
2510 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2511 SvPV_shrink_to_cur(sv);
2514 /* return the substring (via pl_yylval) only if we parsed anything */
2515 if (s > PL_bufptr) {
2516 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2517 const char *const key = PL_lex_inpat ? "qr" : "q";
2518 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2522 if (PL_lex_inwhat == OP_TRANS) {
2525 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2533 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2536 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2543 * Returns TRUE if there's more to the expression (e.g., a subscript),
2546 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2548 * ->[ and ->{ return TRUE
2549 * { and [ outside a pattern are always subscripts, so return TRUE
2550 * if we're outside a pattern and it's not { or [, then return FALSE
2551 * if we're in a pattern and the first char is a {
2552 * {4,5} (any digits around the comma) returns FALSE
2553 * if we're in a pattern and the first char is a [
2555 * [SOMETHING] has a funky algorithm to decide whether it's a
2556 * character class or not. It has to deal with things like
2557 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2558 * anything else returns TRUE
2561 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2564 S_intuit_more(pTHX_ register char *s)
2568 PERL_ARGS_ASSERT_INTUIT_MORE;
2570 if (PL_lex_brackets)
2572 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2574 if (*s != '{' && *s != '[')
2579 /* In a pattern, so maybe we have {n,m}. */
2596 /* On the other hand, maybe we have a character class */
2599 if (*s == ']' || *s == '^')
2602 /* this is terrifying, and it works */
2603 int weight = 2; /* let's weigh the evidence */
2605 unsigned char un_char = 255, last_un_char;
2606 const char * const send = strchr(s,']');
2607 char tmpbuf[sizeof PL_tokenbuf * 4];
2609 if (!send) /* has to be an expression */
2612 Zero(seen,256,char);
2615 else if (isDIGIT(*s)) {
2617 if (isDIGIT(s[1]) && s[2] == ']')
2623 for (; s < send; s++) {
2624 last_un_char = un_char;
2625 un_char = (unsigned char)*s;
2630 weight -= seen[un_char] * 10;
2631 if (isALNUM_lazy_if(s+1,UTF)) {
2633 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2634 len = (int)strlen(tmpbuf);
2635 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2640 else if (*s == '$' && s[1] &&
2641 strchr("[#!%*<>()-=",s[1])) {
2642 if (/*{*/ strchr("])} =",s[2]))
2651 if (strchr("wds]",s[1]))
2653 else if (seen[(U8)'\''] || seen[(U8)'"'])
2655 else if (strchr("rnftbxcav",s[1]))
2657 else if (isDIGIT(s[1])) {
2659 while (s[1] && isDIGIT(s[1]))
2669 if (strchr("aA01! ",last_un_char))
2671 if (strchr("zZ79~",s[1]))
2673 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2674 weight -= 5; /* cope with negative subscript */
2677 if (!isALNUM(last_un_char)
2678 && !(last_un_char == '$' || last_un_char == '@'
2679 || last_un_char == '&')
2680 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2685 if (keyword(tmpbuf, d - tmpbuf, 0))
2688 if (un_char == last_un_char + 1)
2690 weight -= seen[un_char];
2695 if (weight >= 0) /* probably a character class */
2705 * Does all the checking to disambiguate
2707 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2708 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2710 * First argument is the stuff after the first token, e.g. "bar".
2712 * Not a method if bar is a filehandle.
2713 * Not a method if foo is a subroutine prototyped to take a filehandle.
2714 * Not a method if it's really "Foo $bar"
2715 * Method if it's "foo $bar"
2716 * Not a method if it's really "print foo $bar"
2717 * Method if it's really "foo package::" (interpreted as package->foo)
2718 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2719 * Not a method if bar is a filehandle or package, but is quoted with
2724 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2727 char *s = start + (*start == '$');
2728 char tmpbuf[sizeof PL_tokenbuf];
2735 PERL_ARGS_ASSERT_INTUIT_METHOD;
2738 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2742 const char *proto = SvPVX_const(cv);
2753 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2754 /* start is the beginning of the possible filehandle/object,
2755 * and s is the end of it
2756 * tmpbuf is a copy of it
2759 if (*start == '$') {
2760 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2761 isUPPER(*PL_tokenbuf))
2764 len = start - SvPVX(PL_linestr);
2768 start = SvPVX(PL_linestr) + len;
2772 return *s == '(' ? FUNCMETH : METHOD;
2774 if (!keyword(tmpbuf, len, 0)) {
2775 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2779 soff = s - SvPVX(PL_linestr);
2783 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2784 if (indirgv && GvCVu(indirgv))
2786 /* filehandle or package name makes it a method */
2787 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2789 soff = s - SvPVX(PL_linestr);
2792 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2793 return 0; /* no assumptions -- "=>" quotes bearword */
2795 start_force(PL_curforce);
2796 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2797 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
2798 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2800 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2805 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2807 return *s == '(' ? FUNCMETH : METHOD;
2813 /* Encoded script support. filter_add() effectively inserts a
2814 * 'pre-processing' function into the current source input stream.
2815 * Note that the filter function only applies to the current source file
2816 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2818 * The datasv parameter (which may be NULL) can be used to pass
2819 * private data to this instance of the filter. The filter function
2820 * can recover the SV using the FILTER_DATA macro and use it to
2821 * store private buffers and state information.
2823 * The supplied datasv parameter is upgraded to a PVIO type
2824 * and the IoDIRP/IoANY field is used to store the function pointer,
2825 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2826 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2827 * private use must be set using malloc'd pointers.
2831 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2840 if (!PL_rsfp_filters)
2841 PL_rsfp_filters = newAV();
2844 SvUPGRADE(datasv, SVt_PVIO);
2845 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2846 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2847 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2848 FPTR2DPTR(void *, IoANY(datasv)),
2849 SvPV_nolen(datasv)));
2850 av_unshift(PL_rsfp_filters, 1);
2851 av_store(PL_rsfp_filters, 0, datasv) ;
2856 /* Delete most recently added instance of this filter function. */
2858 Perl_filter_del(pTHX_ filter_t funcp)
2863 PERL_ARGS_ASSERT_FILTER_DEL;
2866 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2867 FPTR2DPTR(void*, funcp)));
2869 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2871 /* if filter is on top of stack (usual case) just pop it off */
2872 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2873 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2874 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2875 IoANY(datasv) = (void *)NULL;
2876 sv_free(av_pop(PL_rsfp_filters));
2880 /* we need to search for the correct entry and clear it */
2881 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2885 /* Invoke the idxth filter function for the current rsfp. */
2886 /* maxlen 0 = read one text line */
2888 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2893 /* This API is bad. It should have been using unsigned int for maxlen.
2894 Not sure if we want to change the API, but if not we should sanity
2895 check the value here. */
2896 const unsigned int correct_length
2905 PERL_ARGS_ASSERT_FILTER_READ;
2907 if (!PL_parser || !PL_rsfp_filters)
2909 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2910 /* Provide a default input filter to make life easy. */
2911 /* Note that we append to the line. This is handy. */
2912 DEBUG_P(PerlIO_printf(Perl_debug_log,
2913 "filter_read %d: from rsfp\n", idx));
2914 if (correct_length) {
2917 const int old_len = SvCUR(buf_sv);
2919 /* ensure buf_sv is large enough */
2920 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
2921 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2922 correct_length)) <= 0) {
2923 if (PerlIO_error(PL_rsfp))
2924 return -1; /* error */
2926 return 0 ; /* end of file */
2928 SvCUR_set(buf_sv, old_len + len) ;
2929 SvPVX(buf_sv)[old_len + len] = '\0';
2932 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2933 if (PerlIO_error(PL_rsfp))
2934 return -1; /* error */
2936 return 0 ; /* end of file */
2939 return SvCUR(buf_sv);
2941 /* Skip this filter slot if filter has been deleted */
2942 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2943 DEBUG_P(PerlIO_printf(Perl_debug_log,
2944 "filter_read %d: skipped (filter deleted)\n",
2946 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2948 /* Get function pointer hidden within datasv */
2949 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2950 DEBUG_P(PerlIO_printf(Perl_debug_log,
2951 "filter_read %d: via function %p (%s)\n",
2952 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2953 /* Call function. The function is expected to */
2954 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2955 /* Return: <0:error, =0:eof, >0:not eof */
2956 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2960 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
2964 PERL_ARGS_ASSERT_FILTER_GETS;
2966 #ifdef PERL_CR_FILTER
2967 if (!PL_rsfp_filters) {
2968 filter_add(S_cr_textfilter,NULL);
2971 if (PL_rsfp_filters) {
2973 SvCUR_set(sv, 0); /* start with empty line */
2974 if (FILTER_READ(0, sv, 0) > 0)
2975 return ( SvPVX(sv) ) ;
2980 return (sv_gets(sv, PL_rsfp, append));
2984 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2989 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
2991 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2995 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2996 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2998 return GvHV(gv); /* Foo:: */
3001 /* use constant CLASS => 'MyClass' */
3002 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3003 if (gv && GvCV(gv)) {
3004 SV * const sv = cv_const_sv(GvCV(gv));
3006 pkgname = SvPV_const(sv, len);
3009 return gv_stashpvn(pkgname, len, 0);
3013 * S_readpipe_override
3014 * Check whether readpipe() is overriden, and generates the appropriate
3015 * optree, provided sublex_start() is called afterwards.
3018 S_readpipe_override(pTHX)
3021 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3022 pl_yylval.ival = OP_BACKTICK;
3024 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3026 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3027 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3028 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3030 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3031 append_elem(OP_LIST,
3032 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3033 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3040 * The intent of this yylex wrapper is to minimize the changes to the
3041 * tokener when we aren't interested in collecting madprops. It remains
3042 * to be seen how successful this strategy will be...
3049 char *s = PL_bufptr;
3051 /* make sure PL_thiswhite is initialized */
3055 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3056 if (PL_pending_ident)
3057 return S_pending_ident(aTHX);
3059 /* previous token ate up our whitespace? */
3060 if (!PL_lasttoke && PL_nextwhite) {
3061 PL_thiswhite = PL_nextwhite;
3065 /* isolate the token, and figure out where it is without whitespace */
3066 PL_realtokenstart = -1;
3070 assert(PL_curforce < 0);
3072 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3073 if (!PL_thistoken) {
3074 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3075 PL_thistoken = newSVpvs("");
3077 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3078 PL_thistoken = newSVpvn(tstart, s - tstart);
3081 if (PL_thismad) /* install head */
3082 CURMAD('X', PL_thistoken);
3085 /* last whitespace of a sublex? */
3086 if (optype == ')' && PL_endwhite) {
3087 CURMAD('X', PL_endwhite);
3092 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3093 if (!PL_thiswhite && !PL_endwhite && !optype) {
3094 sv_free(PL_thistoken);
3099 /* put off final whitespace till peg */
3100 if (optype == ';' && !PL_rsfp) {
3101 PL_nextwhite = PL_thiswhite;
3104 else if (PL_thisopen) {
3105 CURMAD('q', PL_thisopen);
3107 sv_free(PL_thistoken);
3111 /* Store actual token text as madprop X */
3112 CURMAD('X', PL_thistoken);
3116 /* add preceding whitespace as madprop _ */
3117 CURMAD('_', PL_thiswhite);
3121 /* add quoted material as madprop = */
3122 CURMAD('=', PL_thisstuff);
3126 /* add terminating quote as madprop Q */
3127 CURMAD('Q', PL_thisclose);
3131 /* special processing based on optype */
3135 /* opval doesn't need a TOKEN since it can already store mp */
3145 if (pl_yylval.opval)
3146 append_madprops(PL_thismad, pl_yylval.opval, 0);
3154 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3163 /* remember any fake bracket that lexer is about to discard */
3164 if (PL_lex_brackets == 1 &&
3165 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3168 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3171 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3172 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3175 break; /* don't bother looking for trailing comment */
3184 /* attach a trailing comment to its statement instead of next token */
3188 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3190 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3192 if (*s == '\n' || *s == '#') {
3193 while (s < PL_bufend && *s != '\n')
3197 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3198 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3215 /* Create new token struct. Note: opvals return early above. */
3216 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3223 S_tokenize_use(pTHX_ int is_use, char *s) {
3226 PERL_ARGS_ASSERT_TOKENIZE_USE;
3228 if (PL_expect != XSTATE)
3229 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3230 is_use ? "use" : "no"));
3232 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3233 s = force_version(s, TRUE);
3234 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3235 start_force(PL_curforce);
3236 NEXTVAL_NEXTTOKE.opval = NULL;
3239 else if (*s == 'v') {
3240 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3241 s = force_version(s, FALSE);
3245 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3246 s = force_version(s, FALSE);
3248 pl_yylval.ival = is_use;
3252 static const char* const exp_name[] =
3253 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3254 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3261 Works out what to call the token just pulled out of the input
3262 stream. The yacc parser takes care of taking the ops we return and
3263 stitching them into a tree.
3269 if read an identifier
3270 if we're in a my declaration
3271 croak if they tried to say my($foo::bar)
3272 build the ops for a my() declaration
3273 if it's an access to a my() variable
3274 are we in a sort block?
3275 croak if my($a); $a <=> $b
3276 build ops for access to a my() variable
3277 if in a dq string, and they've said @foo and we can't find @foo
3279 build ops for a bareword
3280 if we already built the token before, use it.
3285 #pragma segment Perl_yylex
3291 register char *s = PL_bufptr;
3296 /* orig_keyword, gvp, and gv are initialized here because
3297 * jump to the label just_a_word_zero can bypass their
3298 * initialization later. */
3299 I32 orig_keyword = 0;
3304 SV* tmp = newSVpvs("");
3305 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3306 (IV)CopLINE(PL_curcop),
3307 lex_state_names[PL_lex_state],
3308 exp_name[PL_expect],
3309 pv_display(tmp, s, strlen(s), 0, 60));
3312 /* check if there's an identifier for us to look at */
3313 if (PL_pending_ident)
3314 return REPORT(S_pending_ident(aTHX));
3316 /* no identifier pending identification */
3318 switch (PL_lex_state) {
3320 case LEX_NORMAL: /* Some compilers will produce faster */
3321 case LEX_INTERPNORMAL: /* code if we comment these out. */
3325 /* when we've already built the next token, just pull it out of the queue */
3329 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3331 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3332 PL_nexttoke[PL_lasttoke].next_mad = 0;
3333 if (PL_thismad && PL_thismad->mad_key == '_') {
3334 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3335 PL_thismad->mad_val = 0;
3336 mad_free(PL_thismad);
3341 PL_lex_state = PL_lex_defer;
3342 PL_expect = PL_lex_expect;
3343 PL_lex_defer = LEX_NORMAL;
3344 if (!PL_nexttoke[PL_lasttoke].next_type)
3349 pl_yylval = PL_nextval[PL_nexttoke];
3351 PL_lex_state = PL_lex_defer;
3352 PL_expect = PL_lex_expect;
3353 PL_lex_defer = LEX_NORMAL;
3357 /* FIXME - can these be merged? */
3358 return(PL_nexttoke[PL_lasttoke].next_type);
3360 return REPORT(PL_nexttype[PL_nexttoke]);
3363 /* interpolated case modifiers like \L \U, including \Q and \E.
3364 when we get here, PL_bufptr is at the \
3366 case LEX_INTERPCASEMOD:
3368 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3369 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3371 /* handle \E or end of string */
3372 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3374 if (PL_lex_casemods) {
3375 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3376 PL_lex_casestack[PL_lex_casemods] = '\0';
3378 if (PL_bufptr != PL_bufend
3379 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3381 PL_lex_state = LEX_INTERPCONCAT;
3384 PL_thistoken = newSVpvs("\\E");
3390 while (PL_bufptr != PL_bufend &&
3391 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3393 PL_thiswhite = newSVpvs("");
3394 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3398 if (PL_bufptr != PL_bufend)
3401 PL_lex_state = LEX_INTERPCONCAT;
3405 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3406 "### Saw case modifier\n"); });
3408 if (s[1] == '\\' && s[2] == 'E') {
3411 PL_thiswhite = newSVpvs("");
3412 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3415 PL_lex_state = LEX_INTERPCONCAT;
3420 if (!PL_madskills) /* when just compiling don't need correct */
3421 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3422 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3423 if ((*s == 'L' || *s == 'U') &&
3424 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3425 PL_lex_casestack[--PL_lex_casemods] = '\0';
3428 if (PL_lex_casemods > 10)
3429 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3430 PL_lex_casestack[PL_lex_casemods++] = *s;
3431 PL_lex_casestack[PL_lex_casemods] = '\0';
3432 PL_lex_state = LEX_INTERPCONCAT;
3433 start_force(PL_curforce);
3434 NEXTVAL_NEXTTOKE.ival = 0;
3436 start_force(PL_curforce);
3438 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3440 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3442 NEXTVAL_NEXTTOKE.ival = OP_LC;
3444 NEXTVAL_NEXTTOKE.ival = OP_UC;
3446 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3448 Perl_croak(aTHX_ "panic: yylex");
3450 SV* const tmpsv = newSVpvs("\\ ");
3451 /* replace the space with the character we want to escape
3453 SvPVX(tmpsv)[1] = *s;
3459 if (PL_lex_starts) {
3465 sv_free(PL_thistoken);
3466 PL_thistoken = newSVpvs("");
3469 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3470 if (PL_lex_casemods == 1 && PL_lex_inpat)
3479 case LEX_INTERPPUSH:
3480 return REPORT(sublex_push());
3482 case LEX_INTERPSTART:
3483 if (PL_bufptr == PL_bufend)
3484 return REPORT(sublex_done());
3485 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3486 "### Interpolated variable\n"); });
3488 PL_lex_dojoin = (*PL_bufptr == '@');
3489 PL_lex_state = LEX_INTERPNORMAL;
3490 if (PL_lex_dojoin) {
3491 start_force(PL_curforce);
3492 NEXTVAL_NEXTTOKE.ival = 0;
3494 start_force(PL_curforce);
3495 force_ident("\"", '$');
3496 start_force(PL_curforce);
3497 NEXTVAL_NEXTTOKE.ival = 0;
3499 start_force(PL_curforce);
3500 NEXTVAL_NEXTTOKE.ival = 0;
3502 start_force(PL_curforce);
3503 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3506 if (PL_lex_starts++) {
3511 sv_free(PL_thistoken);
3512 PL_thistoken = newSVpvs("");
3515 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3516 if (!PL_lex_casemods && PL_lex_inpat)
3523 case LEX_INTERPENDMAYBE:
3524 if (intuit_more(PL_bufptr)) {
3525 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3531 if (PL_lex_dojoin) {
3532 PL_lex_dojoin = FALSE;
3533 PL_lex_state = LEX_INTERPCONCAT;
3537 sv_free(PL_thistoken);
3538 PL_thistoken = newSVpvs("");
3543 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3544 && SvEVALED(PL_lex_repl))
3546 if (PL_bufptr != PL_bufend)
3547 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3551 case LEX_INTERPCONCAT:
3553 if (PL_lex_brackets)
3554 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3556 if (PL_bufptr == PL_bufend)
3557 return REPORT(sublex_done());
3559 if (SvIVX(PL_linestr) == '\'') {
3560 SV *sv = newSVsv(PL_linestr);
3563 else if ( PL_hints & HINT_NEW_RE )
3564 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3565 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3569 s = scan_const(PL_bufptr);
3571 PL_lex_state = LEX_INTERPCASEMOD;
3573 PL_lex_state = LEX_INTERPSTART;
3576 if (s != PL_bufptr) {
3577 start_force(PL_curforce);
3579 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3581 NEXTVAL_NEXTTOKE = pl_yylval;
3584 if (PL_lex_starts++) {
3588 sv_free(PL_thistoken);
3589 PL_thistoken = newSVpvs("");
3592 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3593 if (!PL_lex_casemods && PL_lex_inpat)
3606 PL_lex_state = LEX_NORMAL;
3607 s = scan_formline(PL_bufptr);
3608 if (!PL_lex_formbrack)
3614 PL_oldoldbufptr = PL_oldbufptr;
3620 sv_free(PL_thistoken);
3623 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3627 if (isIDFIRST_lazy_if(s,UTF))
3630 unsigned char c = *s;
3631 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3632 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
3633 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
3638 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
3642 goto fake_eof; /* emulate EOF on ^D or ^Z */
3651 if (PL_lex_brackets) {
3652 yyerror((const char *)
3654 ? "Format not terminated"
3655 : "Missing right curly or square bracket"));
3657 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3658 "### Tokener got EOF\n");
3662 if (s++ < PL_bufend)
3663 goto retry; /* ignore stray nulls */
3666 if (!PL_in_eval && !PL_preambled) {
3667 PL_preambled = TRUE;
3673 /* Generate a string of Perl code to load the debugger.
3674 * If PERL5DB is set, it will return the contents of that,
3675 * otherwise a compile-time require of perl5db.pl. */
3677 const char * const pdb = PerlEnv_getenv("PERL5DB");
3680 sv_setpv(PL_linestr, pdb);
3681 sv_catpvs(PL_linestr,";");
3683 SETERRNO(0,SS_NORMAL);
3684 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3687 sv_setpvs(PL_linestr,"");
3688 if (PL_preambleav) {
3689 SV **svp = AvARRAY(PL_preambleav);
3690 SV **const end = svp + AvFILLp(PL_preambleav);
3692 sv_catsv(PL_linestr, *svp);
3694 sv_catpvs(PL_linestr, ";");
3696 sv_free(MUTABLE_SV(PL_preambleav));
3697 PL_preambleav = NULL;
3700 sv_catpvs(PL_linestr,
3701 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3702 if (PL_minus_n || PL_minus_p) {
3703 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3705 sv_catpvs(PL_linestr,"chomp;");
3708 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3709 || *PL_splitstr == '"')
3710 && strchr(PL_splitstr + 1, *PL_splitstr))
3711 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3713 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3714 bytes can be used as quoting characters. :-) */
3715 const char *splits = PL_splitstr;
3716 sv_catpvs(PL_linestr, "our @F=split(q\0");
3719 if (*splits == '\\')
3720 sv_catpvn(PL_linestr, splits, 1);
3721 sv_catpvn(PL_linestr, splits, 1);
3722 } while (*splits++);
3723 /* This loop will embed the trailing NUL of
3724 PL_linestr as the last thing it does before
3726 sv_catpvs(PL_linestr, ");");
3730 sv_catpvs(PL_linestr,"our @F=split(' ');");
3733 sv_catpvs(PL_linestr, "\n");
3734 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3735 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3736 PL_last_lop = PL_last_uni = NULL;
3737 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3738 update_debugger_info(PL_linestr, NULL, 0);
3742 bof = PL_rsfp ? TRUE : FALSE;
3743 if ((s = filter_gets(PL_linestr, 0)) == NULL) {
3746 PL_realtokenstart = -1;
3749 if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3750 PerlIO_clearerr(PL_rsfp);
3752 (void)PerlIO_close(PL_rsfp);
3754 PL_doextract = FALSE;
3756 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3762 sv_setpvs(PL_linestr, ";}continue{print;}");
3764 sv_setpvs(PL_linestr, ";}");
3765 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3766 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3767 PL_last_lop = PL_last_uni = NULL;
3768 PL_minus_n = PL_minus_p = 0;
3771 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3772 PL_last_lop = PL_last_uni = NULL;
3773 sv_setpvs(PL_linestr,"");
3774 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3776 /* If it looks like the start of a BOM or raw UTF-16,
3777 * check if it in fact is. */
3783 #ifdef PERLIO_IS_STDIO
3784 # ifdef __GNU_LIBRARY__
3785 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3786 # define FTELL_FOR_PIPE_IS_BROKEN
3790 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3791 # define FTELL_FOR_PIPE_IS_BROKEN
3796 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3798 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3799 s = swallow_bom((U8*)s);
3803 /* Incest with pod. */
3806 sv_catsv(PL_thiswhite, PL_linestr);
3808 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3809 sv_setpvs(PL_linestr, "");
3810 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3811 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3812 PL_last_lop = PL_last_uni = NULL;
3813 PL_doextract = FALSE;
3817 } while (PL_doextract);
3818 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3819 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3820 update_debugger_info(PL_linestr, NULL, 0);
3821 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3822 PL_last_lop = PL_last_uni = NULL;
3823 if (CopLINE(PL_curcop) == 1) {
3824 while (s < PL_bufend && isSPACE(*s))
3826 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3830 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3834 if (*s == '#' && *(s+1) == '!')
3836 #ifdef ALTERNATE_SHEBANG
3838 static char const as[] = ALTERNATE_SHEBANG;
3839 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3840 d = s + (sizeof(as) - 1);
3842 #endif /* ALTERNATE_SHEBANG */
3851 while (*d && !isSPACE(*d))
3855 #ifdef ARG_ZERO_IS_SCRIPT
3856 if (ipathend > ipath) {
3858 * HP-UX (at least) sets argv[0] to the script name,
3859 * which makes $^X incorrect. And Digital UNIX and Linux,
3860 * at least, set argv[0] to the basename of the Perl
3861 * interpreter. So, having found "#!", we'll set it right.
3863 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3865 assert(SvPOK(x) || SvGMAGICAL(x));
3866 if (sv_eq(x, CopFILESV(PL_curcop))) {
3867 sv_setpvn(x, ipath, ipathend - ipath);
3873 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3874 const char * const lstart = SvPV_const(x,llen);
3876 bstart += blen - llen;
3877 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3878 sv_setpvn(x, ipath, ipathend - ipath);
3883 TAINT_NOT; /* $^X is always tainted, but that's OK */
3885 #endif /* ARG_ZERO_IS_SCRIPT */
3890 d = instr(s,"perl -");
3892 d = instr(s,"perl");
3894 /* avoid getting into infinite loops when shebang
3895 * line contains "Perl" rather than "perl" */
3897 for (d = ipathend-4; d >= ipath; --d) {
3898 if ((*d == 'p' || *d == 'P')
3899 && !ibcmp(d, "perl", 4))
3909 #ifdef ALTERNATE_SHEBANG
3911 * If the ALTERNATE_SHEBANG on this system starts with a
3912 * character that can be part of a Perl expression, then if
3913 * we see it but not "perl", we're probably looking at the
3914 * start of Perl code, not a request to hand off to some
3915 * other interpreter. Similarly, if "perl" is there, but
3916 * not in the first 'word' of the line, we assume the line
3917 * contains the start of the Perl program.
3919 if (d && *s != '#') {
3920 const char *c = ipath;
3921 while (*c && !strchr("; \t\r\n\f\v#", *c))
3924 d = NULL; /* "perl" not in first word; ignore */
3926 *s = '#'; /* Don't try to parse shebang line */
3928 #endif /* ALTERNATE_SHEBANG */
3933 !instr(s,"indir") &&
3934 instr(PL_origargv[0],"perl"))
3941 while (s < PL_bufend && isSPACE(*s))
3943 if (s < PL_bufend) {
3944 Newx(newargv,PL_origargc+3,char*);
3946 while (s < PL_bufend && !isSPACE(*s))
3949 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3952 newargv = PL_origargv;
3955 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3957 Perl_croak(aTHX_ "Can't exec %s", ipath);
3960 while (*d && !isSPACE(*d))
3962 while (SPACE_OR_TAB(*d))
3966 const bool switches_done = PL_doswitches;
3967 const U32 oldpdb = PL_perldb;
3968 const bool oldn = PL_minus_n;
3969 const bool oldp = PL_minus_p;
3973 bool baduni = FALSE;
3975 const char *d2 = d1 + 1;
3976 if (parse_unicode_opts((const char **)&d2)
3980 if (baduni || *d1 == 'M' || *d1 == 'm') {
3981 const char * const m = d1;
3982 while (*d1 && !isSPACE(*d1))
3984 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3987 d1 = moreswitches(d1);
3989 if (PL_doswitches && !switches_done) {
3990 int argc = PL_origargc;
3991 char **argv = PL_origargv;
3994 } while (argc && argv[0][0] == '-' && argv[0][1]);
3995 init_argv_symbols(argc,argv);
3997 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
3998 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3999 /* if we have already added "LINE: while (<>) {",
4000 we must not do it again */
4002 sv_setpvs(PL_linestr, "");
4003 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4004 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4005 PL_last_lop = PL_last_uni = NULL;
4006 PL_preambled = FALSE;
4007 if (PERLDB_LINE || PERLDB_SAVESRC)
4008 (void)gv_fetchfile(PL_origfilename);
4015 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4017 PL_lex_state = LEX_FORMLINE;
4022 #ifdef PERL_STRICT_CR
4023 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4025 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4027 case ' ': case '\t': case '\f': case 013:
4029 PL_realtokenstart = -1;
4031 PL_thiswhite = newSVpvs("");
4032 sv_catpvn(PL_thiswhite, s, 1);
4039 PL_realtokenstart = -1;
4043 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4044 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4045 /* handle eval qq[#line 1 "foo"\n ...] */
4046 CopLINE_dec(PL_curcop);
4049 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4051 if (!PL_in_eval || PL_rsfp)
4056 while (d < PL_bufend && *d != '\n')
4060 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4061 Perl_croak(aTHX_ "panic: input overflow");
4064 PL_thiswhite = newSVpvn(s, d - s);
4069 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4071 PL_lex_state = LEX_FORMLINE;
4077 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4078 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4081 TOKEN(PEG); /* make sure any #! line is accessible */
4086 /* if (PL_madskills && PL_lex_formbrack) { */
4088 while (d < PL_bufend && *d != '\n')
4092 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4093 Perl_croak(aTHX_ "panic: input overflow");
4094 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4096 PL_thiswhite = newSVpvs("");
4097 if (CopLINE(PL_curcop) == 1) {
4098 sv_setpvs(PL_thiswhite, "");
4101 sv_catpvn(PL_thiswhite, s, d - s);
4115 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4123 while (s < PL_bufend && SPACE_OR_TAB(*s))
4126 if (strnEQ(s,"=>",2)) {
4127 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4128 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4129 OPERATOR('-'); /* unary minus */
4131 PL_last_uni = PL_oldbufptr;
4133 case 'r': ftst = OP_FTEREAD; break;
4134 case 'w': ftst = OP_FTEWRITE; break;
4135 case 'x': ftst = OP_FTEEXEC; break;
4136 case 'o': ftst = OP_FTEOWNED; break;
4137 case 'R': ftst = OP_FTRREAD; break;
4138 case 'W': ftst = OP_FTRWRITE; break;
4139 case 'X': ftst = OP_FTREXEC; break;
4140 case 'O': ftst = OP_FTROWNED; break;
4141 case 'e': ftst = OP_FTIS; break;
4142 case 'z': ftst = OP_FTZERO; break;
4143 case 's': ftst = OP_FTSIZE; break;
4144 case 'f': ftst = OP_FTFILE; break;
4145 case 'd': ftst = OP_FTDIR; break;
4146 case 'l': ftst = OP_FTLINK; break;
4147 case 'p': ftst = OP_FTPIPE; break;
4148 case 'S': ftst = OP_FTSOCK; break;
4149 case 'u': ftst = OP_FTSUID; break;
4150 case 'g': ftst = OP_FTSGID; break;
4151 case 'k': ftst = OP_FTSVTX; break;
4152 case 'b': ftst = OP_FTBLK; break;
4153 case 'c': ftst = OP_FTCHR; break;
4154 case 't': ftst = OP_FTTTY; break;
4155 case 'T': ftst = OP_FTTEXT; break;
4156 case 'B': ftst = OP_FTBINARY; break;
4157 case 'M': case 'A': case 'C':
4158 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4160 case 'M': ftst = OP_FTMTIME; break;
4161 case 'A': ftst = OP_FTATIME; break;
4162 case 'C': ftst = OP_FTCTIME; break;
4170 PL_last_lop_op = (OPCODE)ftst;
4171 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4172 "### Saw file test %c\n", (int)tmp);
4177 /* Assume it was a minus followed by a one-letter named
4178 * subroutine call (or a -bareword), then. */
4179 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4180 "### '-%c' looked like a file test but was not\n",
4187 const char tmp = *s++;
4190 if (PL_expect == XOPERATOR)
4195 else if (*s == '>') {
4198 if (isIDFIRST_lazy_if(s,UTF)) {
4199 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4207 if (PL_expect == XOPERATOR)
4210 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4212 OPERATOR('-'); /* unary minus */
4218 const char tmp = *s++;
4221 if (PL_expect == XOPERATOR)
4226 if (PL_expect == XOPERATOR)
4229 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4236 if (PL_expect != XOPERATOR) {
4237 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4238 PL_expect = XOPERATOR;
4239 force_ident(PL_tokenbuf, '*');
4252 if (PL_expect == XOPERATOR) {
4256 PL_tokenbuf[0] = '%';
4257 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4258 sizeof PL_tokenbuf - 1, FALSE);
4259 if (!PL_tokenbuf[1]) {
4262 PL_pending_ident = '%';
4271 const char tmp = *s++;
4276 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4283 const char tmp = *s++;
4289 goto just_a_word_zero_gv;
4292 switch (PL_expect) {
4298 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4300 PL_bufptr = s; /* update in case we back off */
4306 PL_expect = XTERMBLOCK;
4309 stuffstart = s - SvPVX(PL_linestr) - 1;
4313 while (isIDFIRST_lazy_if(s,UTF)) {
4316 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4317 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4318 if (tmp < 0) tmp = -tmp;
4333 sv = newSVpvn(s, len);
4335 d = scan_str(d,TRUE,TRUE);
4337 /* MUST advance bufptr here to avoid bogus
4338 "at end of line" context messages from yyerror().
4340 PL_bufptr = s + len;
4341 yyerror("Unterminated attribute parameter in attribute list");
4345 return REPORT(0); /* EOF indicator */
4349 sv_catsv(sv, PL_lex_stuff);
4350 attrs = append_elem(OP_LIST, attrs,
4351 newSVOP(OP_CONST, 0, sv));
4352 SvREFCNT_dec(PL_lex_stuff);
4353 PL_lex_stuff = NULL;
4356 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4358 if (PL_in_my == KEY_our) {
4359 deprecate(":unique");
4362 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4365 /* NOTE: any CV attrs applied here need to be part of
4366 the CVf_BUILTIN_ATTRS define in cv.h! */
4367 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4369 CvLVALUE_on(PL_compcv);
4371 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4373 deprecate(":locked");
4375 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4377 CvMETHOD_on(PL_compcv);
4379 /* After we've set the flags, it could be argued that
4380 we don't need to do the attributes.pm-based setting
4381 process, and shouldn't bother appending recognized
4382 flags. To experiment with that, uncomment the
4383 following "else". (Note that's already been
4384 uncommented. That keeps the above-applied built-in
4385 attributes from being intercepted (and possibly
4386 rejected) by a package's attribute routines, but is
4387 justified by the performance win for the common case
4388 of applying only built-in attributes.) */
4390 attrs = append_elem(OP_LIST, attrs,
4391 newSVOP(OP_CONST, 0,
4395 if (*s == ':' && s[1] != ':')
4398 break; /* require real whitespace or :'s */
4399 /* XXX losing whitespace on sequential attributes here */
4403 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4404 if (*s != ';' && *s != '}' && *s != tmp
4405 && (tmp != '=' || *s != ')')) {
4406 const char q = ((*s == '\'') ? '"' : '\'');
4407 /* If here for an expression, and parsed no attrs, back
4409 if (tmp == '=' && !attrs) {
4413 /* MUST advance bufptr here to avoid bogus "at end of line"
4414 context messages from yyerror().
4417 yyerror( (const char *)
4419 ? Perl_form(aTHX_ "Invalid separator character "
4420 "%c%c%c in attribute list", q, *s, q)
4421 : "Unterminated attribute list" ) );
4429 start_force(PL_curforce);
4430 NEXTVAL_NEXTTOKE.opval = attrs;
4431 CURMAD('_', PL_nextwhite);
4436 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4437 (s - SvPVX(PL_linestr)) - stuffstart);
4445 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4446 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4454 const char tmp = *s++;
4459 const char tmp = *s++;
4467 if (PL_lex_brackets <= 0)
4468 yyerror("Unmatched right square bracket");
4471 if (PL_lex_state == LEX_INTERPNORMAL) {
4472 if (PL_lex_brackets == 0) {
4473 if (*s == '-' && s[1] == '>')
4474 PL_lex_state = LEX_INTERPENDMAYBE;
4475 else if (*s != '[' && *s != '{')
4476 PL_lex_state = LEX_INTERPEND;
4483 if (PL_lex_brackets > 100) {
4484 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4486 switch (PL_expect) {
4488 if (PL_lex_formbrack) {
4492 if (PL_oldoldbufptr == PL_last_lop)
4493 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4495 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4496 OPERATOR(HASHBRACK);
4498 while (s < PL_bufend && SPACE_OR_TAB(*s))
4501 PL_tokenbuf[0] = '\0';
4502 if (d < PL_bufend && *d == '-') {
4503 PL_tokenbuf[0] = '-';
4505 while (d < PL_bufend && SPACE_OR_TAB(*d))
4508 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4509 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4511 while (d < PL_bufend && SPACE_OR_TAB(*d))
4514 const char minus = (PL_tokenbuf[0] == '-');
4515 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4523 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4528 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4533 if (PL_oldoldbufptr == PL_last_lop)
4534 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4536 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4539 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4541 /* This hack is to get the ${} in the message. */
4543 yyerror("syntax error");
4546 OPERATOR(HASHBRACK);
4548 /* This hack serves to disambiguate a pair of curlies
4549 * as being a block or an anon hash. Normally, expectation
4550 * determines that, but in cases where we're not in a
4551 * position to expect anything in particular (like inside
4552 * eval"") we have to resolve the ambiguity. This code
4553 * covers the case where the first term in the curlies is a
4554 * quoted string. Most other cases need to be explicitly
4555 * disambiguated by prepending a "+" before the opening
4556 * curly in order to force resolution as an anon hash.
4558 * XXX should probably propagate the outer expectation
4559 * into eval"" to rely less on this hack, but that could
4560 * potentially break current behavior of eval"".
4564 if (*s == '\'' || *s == '"' || *s == '`') {
4565 /* common case: get past first string, handling escapes */
4566 for (t++; t < PL_bufend && *t != *s;)
4567 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4571 else if (*s == 'q') {
4574 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4577 /* skip q//-like construct */
4579 char open, close, term;
4582 while (t < PL_bufend && isSPACE(*t))
4584 /* check for q => */
4585 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4586 OPERATOR(HASHBRACK);
4590 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4594 for (t++; t < PL_bufend; t++) {
4595 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4597 else if (*t == open)
4601 for (t++; t < PL_bufend; t++) {
4602 if (*t == '\\' && t+1 < PL_bufend)
4604 else if (*t == close && --brackets <= 0)
4606 else if (*t == open)
4613 /* skip plain q word */
4614 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4617 else if (isALNUM_lazy_if(t,UTF)) {
4619 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4622 while (t < PL_bufend && isSPACE(*t))
4624 /* if comma follows first term, call it an anon hash */
4625 /* XXX it could be a comma expression with loop modifiers */
4626 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4627 || (*t == '=' && t[1] == '>')))
4628 OPERATOR(HASHBRACK);
4629 if (PL_expect == XREF)
4632 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4638 pl_yylval.ival = CopLINE(PL_curcop);
4639 if (isSPACE(*s) || *s == '#')
4640 PL_copline = NOLINE; /* invalidate current command line number */
4645 if (PL_lex_brackets <= 0)
4646 yyerror("Unmatched right curly bracket");
4648 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4649 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4650 PL_lex_formbrack = 0;
4651 if (PL_lex_state == LEX_INTERPNORMAL) {
4652 if (PL_lex_brackets == 0) {
4653 if (PL_expect & XFAKEBRACK) {
4654 PL_expect &= XENUMMASK;
4655 PL_lex_state = LEX_INTERPEND;
4660 PL_thiswhite = newSVpvs("");
4661 sv_catpvs(PL_thiswhite,"}");
4664 return yylex(); /* ignore fake brackets */
4666 if (*s == '-' && s[1] == '>')
4667 PL_lex_state = LEX_INTERPENDMAYBE;
4668 else if (*s != '[' && *s != '{')
4669 PL_lex_state = LEX_INTERPEND;
4672 if (PL_expect & XFAKEBRACK) {
4673 PL_expect &= XENUMMASK;
4675 return yylex(); /* ignore fake brackets */
4677 start_force(PL_curforce);
4679 curmad('X', newSVpvn(s-1,1));
4680 CURMAD('_', PL_thiswhite);
4685 PL_thistoken = newSVpvs("");
4693 if (PL_expect == XOPERATOR) {
4694 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4695 && isIDFIRST_lazy_if(s,UTF))
4697 CopLINE_dec(PL_curcop);
4698 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
4699 CopLINE_inc(PL_curcop);
4704 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4706 PL_expect = XOPERATOR;
4707 force_ident(PL_tokenbuf, '&');
4711 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4723 const char tmp = *s++;
4730 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4731 && strchr("+-*/%.^&|<",tmp))
4732 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4733 "Reversed %c= operator",(int)tmp);
4735 if (PL_expect == XSTATE && isALPHA(tmp) &&
4736 (s == PL_linestart+1 || s[-2] == '\n') )
4738 if (PL_in_eval && !PL_rsfp) {
4743 if (strnEQ(s,"=cut",4)) {
4759 PL_thiswhite = newSVpvs("");
4760 sv_catpvn(PL_thiswhite, PL_linestart,
4761 PL_bufend - PL_linestart);
4765 PL_doextract = TRUE;
4769 if (PL_lex_brackets < PL_lex_formbrack) {
4771 #ifdef PERL_STRICT_CR
4772 while (SPACE_OR_TAB(*t))
4774 while (SPACE_OR_TAB(*t) || *t == '\r')
4777 if (*t == '\n' || *t == '#') {
4788 const char tmp = *s++;
4790 /* was this !=~ where !~ was meant?
4791 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4793 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4794 const char *t = s+1;
4796 while (t < PL_bufend && isSPACE(*t))
4799 if (*t == '/' || *t == '?' ||
4800 ((*t == 'm' || *t == 's' || *t == 'y')
4801 && !isALNUM(t[1])) ||
4802 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4803 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4804 "!=~ should be !~");
4814 if (PL_expect != XOPERATOR) {
4815 if (s[1] != '<' && !strchr(s,'>'))
4818 s = scan_heredoc(s);
4820 s = scan_inputsymbol(s);
4821 TERM(sublex_start());
4827 SHop(OP_LEFT_SHIFT);
4841 const char tmp = *s++;
4843 SHop(OP_RIGHT_SHIFT);
4844 else if (tmp == '=')
4853 if (PL_expect == XOPERATOR) {
4854 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4855 return deprecate_commaless_var_list();
4859 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4860 PL_tokenbuf[0] = '@';
4861 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4862 sizeof PL_tokenbuf - 1, FALSE);
4863 if (PL_expect == XOPERATOR)
4864 no_op("Array length", s);
4865 if (!PL_tokenbuf[1])
4867 PL_expect = XOPERATOR;
4868 PL_pending_ident = '#';
4872 PL_tokenbuf[0] = '$';
4873 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4874 sizeof PL_tokenbuf - 1, FALSE);
4875 if (PL_expect == XOPERATOR)
4877 if (!PL_tokenbuf[1]) {
4879 yyerror("Final $ should be \\$ or $name");
4883 /* This kludge not intended to be bulletproof. */
4884 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4885 pl_yylval.opval = newSVOP(OP_CONST, 0,
4886 newSViv(CopARYBASE_get(&PL_compiling)));
4887 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4893 const char tmp = *s;
4894 if (PL_lex_state == LEX_NORMAL)
4897 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4898 && intuit_more(s)) {
4900 PL_tokenbuf[0] = '@';
4901 if (ckWARN(WARN_SYNTAX)) {
4904 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4907 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4908 while (t < PL_bufend && *t != ']')
4910 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4911 "Multidimensional syntax %.*s not supported",
4912 (int)((t - PL_bufptr) + 1), PL_bufptr);
4916 else if (*s == '{') {
4918 PL_tokenbuf[0] = '%';
4919 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4920 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4922 char tmpbuf[sizeof PL_tokenbuf];
4925 } while (isSPACE(*t));
4926 if (isIDFIRST_lazy_if(t,UTF)) {
4928 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4932 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4933 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4934 "You need to quote \"%s\"",
4941 PL_expect = XOPERATOR;
4942 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4943 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4944 if (!islop || PL_last_lop_op == OP_GREPSTART)
4945 PL_expect = XOPERATOR;
4946 else if (strchr("$@\"'`q", *s))
4947 PL_expect = XTERM; /* e.g. print $fh "foo" */
4948 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4949 PL_expect = XTERM; /* e.g. print $fh &sub */
4950 else if (isIDFIRST_lazy_if(s,UTF)) {
4951 char tmpbuf[sizeof PL_tokenbuf];
4953 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4954 if ((t2 = keyword(tmpbuf, len, 0))) {
4955 /* binary operators exclude handle interpretations */
4967 PL_expect = XTERM; /* e.g. print $fh length() */
4972 PL_expect = XTERM; /* e.g. print $fh subr() */
4975 else if (isDIGIT(*s))
4976 PL_expect = XTERM; /* e.g. print $fh 3 */
4977 else if (*s == '.' && isDIGIT(s[1]))
4978 PL_expect = XTERM; /* e.g. print $fh .3 */
4979 else if ((*s == '?' || *s == '-' || *s == '+')
4980 && !isSPACE(s[1]) && s[1] != '=')
4981 PL_expect = XTERM; /* e.g. print $fh -1 */
4982 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4984 PL_expect = XTERM; /* e.g. print $fh /.../
4985 XXX except DORDOR operator
4987 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4989 PL_expect = XTERM; /* print $fh <<"EOF" */
4992 PL_pending_ident = '$';
4996 if (PL_expect == XOPERATOR)
4998 PL_tokenbuf[0] = '@';
4999 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5000 if (!PL_tokenbuf[1]) {
5003 if (PL_lex_state == LEX_NORMAL)
5005 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5007 PL_tokenbuf[0] = '%';
5009 /* Warn about @ where they meant $. */
5010 if (*s == '[' || *s == '{') {
5011 if (ckWARN(WARN_SYNTAX)) {
5012 const char *t = s + 1;
5013 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5015 if (*t == '}' || *t == ']') {
5017 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5018 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5019 "Scalar value %.*s better written as $%.*s",
5020 (int)(t-PL_bufptr), PL_bufptr,
5021 (int)(t-PL_bufptr-1), PL_bufptr+1);
5026 PL_pending_ident = '@';
5029 case '/': /* may be division, defined-or, or pattern */
5030 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5034 case '?': /* may either be conditional or pattern */
5035 if (PL_expect == XOPERATOR) {
5043 /* A // operator. */
5053 /* Disable warning on "study /blah/" */
5054 if (PL_oldoldbufptr == PL_last_uni
5055 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5056 || memNE(PL_last_uni, "study", 5)
5057 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5060 s = scan_pat(s,OP_MATCH);
5061 TERM(sublex_start());
5065 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5066 #ifdef PERL_STRICT_CR
5069 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5071 && (s == PL_linestart || s[-1] == '\n') )
5073 PL_lex_formbrack = 0;
5077 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5081 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5087 pl_yylval.ival = OPf_SPECIAL;
5093 if (PL_expect != XOPERATOR)
5098 case '0': case '1': case '2': case '3': case '4':
5099 case '5': case '6': case '7': case '8': case '9':
5100 s = scan_num(s, &pl_yylval);
5101 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5102 if (PL_expect == XOPERATOR)
5107 s = scan_str(s,!!PL_madskills,FALSE);
5108 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5109 if (PL_expect == XOPERATOR) {
5110 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5111 return deprecate_commaless_var_list();
5118 pl_yylval.ival = OP_CONST;
5119 TERM(sublex_start());
5122 s = scan_str(s,!!PL_madskills,FALSE);
5123 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5124 if (PL_expect == XOPERATOR) {
5125 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5126 return deprecate_commaless_var_list();
5133 pl_yylval.ival = OP_CONST;
5134 /* FIXME. I think that this can be const if char *d is replaced by
5135 more localised variables. */
5136 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5137 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5138 pl_yylval.ival = OP_STRINGIFY;
5142 TERM(sublex_start());
5145 s = scan_str(s,!!PL_madskills,FALSE);
5146 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5147 if (PL_expect == XOPERATOR)
5148 no_op("Backticks",s);
5151 readpipe_override();
5152 TERM(sublex_start());
5156 if (PL_lex_inwhat && isDIGIT(*s))
5157 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5159 if (PL_expect == XOPERATOR)
5160 no_op("Backslash",s);
5164 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5165 char *start = s + 2;
5166 while (isDIGIT(*start) || *start == '_')
5168 if (*start == '.' && isDIGIT(start[1])) {
5169 s = scan_num(s, &pl_yylval);
5172 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5173 else if (!isALPHA(*start) && (PL_expect == XTERM
5174 || PL_expect == XREF || PL_expect == XSTATE
5175 || PL_expect == XTERMORDORDOR)) {
5176 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5178 s = scan_num(s, &pl_yylval);
5185 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5227 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5229 /* Some keywords can be followed by any delimiter, including ':' */
5230 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5231 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5232 (PL_tokenbuf[0] == 'q' &&
5233 strchr("qwxr", PL_tokenbuf[1])))));
5235 /* x::* is just a word, unless x is "CORE" */
5236 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5240 while (d < PL_bufend && isSPACE(*d))
5241 d++; /* no comments skipped here, or s### is misparsed */
5243 /* Is this a label? */
5244 if (!tmp && PL_expect == XSTATE
5245 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5246 tmp = keyword(PL_tokenbuf, len, 0);
5248 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
5250 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5255 /* Check for keywords */
5256 tmp = keyword(PL_tokenbuf, len, 0);
5258 /* Is this a word before a => operator? */
5259 if (*d == '=' && d[1] == '>') {
5262 = (OP*)newSVOP(OP_CONST, 0,
5263 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5264 pl_yylval.opval->op_private = OPpCONST_BARE;
5268 if (tmp < 0) { /* second-class keyword? */
5269 GV *ogv = NULL; /* override (winner) */
5270 GV *hgv = NULL; /* hidden (loser) */
5271 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5273 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5276 if (GvIMPORTED_CV(gv))
5278 else if (! CvMETHOD(cv))
5282 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5283 (gv = *gvp) && isGV_with_GP(gv) &&
5284 GvCVu(gv) && GvIMPORTED_CV(gv))
5291 tmp = 0; /* overridden by import or by GLOBAL */
5294 && -tmp==KEY_lock /* XXX generalizable kludge */
5297 tmp = 0; /* any sub overrides "weak" keyword */
5299 else { /* no override */
5301 if (tmp == KEY_dump) {
5302 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
5303 "dump() better written as CORE::dump()");
5307 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
5308 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5309 "Ambiguous call resolved as CORE::%s(), %s",
5310 GvENAME(hgv), "qualify as such or use &");
5317 default: /* not a keyword */
5318 /* Trade off - by using this evil construction we can pull the
5319 variable gv into the block labelled keylookup. If not, then
5320 we have to give it function scope so that the goto from the
5321 earlier ':' case doesn't bypass the initialisation. */
5323 just_a_word_zero_gv:
5331 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5334 SV *nextPL_nextwhite = 0;
5338 /* Get the rest if it looks like a package qualifier */
5340 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5342 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5345 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5346 *s == '\'' ? "'" : "::");
5351 if (PL_expect == XOPERATOR) {
5352 if (PL_bufptr == PL_linestart) {
5353 CopLINE_dec(PL_curcop);
5354 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5355 CopLINE_inc(PL_curcop);
5358 no_op("Bareword",s);
5361 /* Look for a subroutine with this name in current package,
5362 unless name is "Foo::", in which case Foo is a bearword
5363 (and a package name). */
5365 if (len > 2 && !PL_madskills &&
5366 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5368 if (ckWARN(WARN_BAREWORD)
5369 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5370 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5371 "Bareword \"%s\" refers to nonexistent package",
5374 PL_tokenbuf[len] = '\0';
5380 /* Mustn't actually add anything to a symbol table.
5381 But also don't want to "initialise" any placeholder
5382 constants that might already be there into full
5383 blown PVGVs with attached PVCV. */
5384 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5385 GV_NOADD_NOINIT, SVt_PVCV);
5390 /* if we saw a global override before, get the right name */
5393 sv = newSVpvs("CORE::GLOBAL::");
5394 sv_catpv(sv,PL_tokenbuf);
5397 /* If len is 0, newSVpv does strlen(), which is correct.
5398 If len is non-zero, then it will be the true length,
5399 and so the scalar will be created correctly. */
5400 sv = newSVpv(PL_tokenbuf,len);
5403 if (PL_madskills && !PL_thistoken) {
5404 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5405 PL_thistoken = newSVpvn(start,s - start);
5406 PL_realtokenstart = s - SvPVX(PL_linestr);
5410 /* Presume this is going to be a bareword of some sort. */
5413 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5414 pl_yylval.opval->op_private = OPpCONST_BARE;
5415 /* UTF-8 package name? */
5416 if (UTF && !IN_BYTES &&
5417 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5420 /* And if "Foo::", then that's what it certainly is. */
5425 /* Do the explicit type check so that we don't need to force
5426 the initialisation of the symbol table to have a real GV.
5427 Beware - gv may not really be a PVGV, cv may not really be
5428 a PVCV, (because of the space optimisations that gv_init
5429 understands) But they're true if for this symbol there is
5430 respectively a typeglob and a subroutine.
5432 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5433 /* Real typeglob, so get the real subroutine: */
5435 /* A proxy for a subroutine in this package? */
5436 : SvOK(gv) ? MUTABLE_CV(gv) : NULL)
5439 /* See if it's the indirect object for a list operator. */
5441 if (PL_oldoldbufptr &&
5442 PL_oldoldbufptr < PL_bufptr &&
5443 (PL_oldoldbufptr == PL_last_lop
5444 || PL_oldoldbufptr == PL_last_uni) &&
5445 /* NO SKIPSPACE BEFORE HERE! */
5446 (PL_expect == XREF ||
5447 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5449 bool immediate_paren = *s == '(';
5451 /* (Now we can afford to cross potential line boundary.) */
5452 s = SKIPSPACE2(s,nextPL_nextwhite);
5454 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5457 /* Two barewords in a row may indicate method call. */
5459 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5460 (tmp = intuit_method(s, gv, cv)))
5463 /* If not a declared subroutine, it's an indirect object. */
5464 /* (But it's an indir obj regardless for sort.) */
5465 /* Also, if "_" follows a filetest operator, it's a bareword */
5468 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5470 (PL_last_lop_op != OP_MAPSTART &&
5471 PL_last_lop_op != OP_GREPSTART))))
5472 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5473 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5476 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5481 PL_expect = XOPERATOR;
5484 s = SKIPSPACE2(s,nextPL_nextwhite);
5485 PL_nextwhite = nextPL_nextwhite;
5490 /* Is this a word before a => operator? */
5491 if (*s == '=' && s[1] == '>' && !pkgname) {
5493 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5494 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5495 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5499 /* If followed by a paren, it's certainly a subroutine. */
5504 while (SPACE_OR_TAB(*d))
5506 if (*d == ')' && (sv = gv_const_sv(gv))) {
5513 PL_nextwhite = PL_thiswhite;
5516 start_force(PL_curforce);
5518 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5519 PL_expect = XOPERATOR;
5522 PL_nextwhite = nextPL_nextwhite;
5523 curmad('X', PL_thistoken);
5524 PL_thistoken = newSVpvs("");
5532 /* If followed by var or block, call it a method (unless sub) */
5534 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5535 PL_last_lop = PL_oldbufptr;
5536 PL_last_lop_op = OP_METHOD;
5540 /* If followed by a bareword, see if it looks like indir obj. */
5543 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5544 && (tmp = intuit_method(s, gv, cv)))
5547 /* Not a method, so call it a subroutine (if defined) */
5550 if (lastchar == '-')
5551 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5552 "Ambiguous use of -%s resolved as -&%s()",
5553 PL_tokenbuf, PL_tokenbuf);
5554 /* Check for a constant sub */
5555 if ((sv = gv_const_sv(gv))) {
5557 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5558 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5559 pl_yylval.opval->op_private = 0;
5563 /* Resolve to GV now. */
5564 if (SvTYPE(gv) != SVt_PVGV) {
5565 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5566 assert (SvTYPE(gv) == SVt_PVGV);
5567 /* cv must have been some sort of placeholder, so
5568 now needs replacing with a real code reference. */
5572 op_free(pl_yylval.opval);
5573 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5574 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5575 PL_last_lop = PL_oldbufptr;
5576 PL_last_lop_op = OP_ENTERSUB;
5577 /* Is there a prototype? */
5585 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
5588 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5590 while (*proto == ';')
5592 if (*proto == '&' && *s == '{') {
5594 sv_setpvs(PL_subname, "__ANON__");
5596 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5603 PL_nextwhite = PL_thiswhite;
5606 start_force(PL_curforce);
5607 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5610 PL_nextwhite = nextPL_nextwhite;
5611 curmad('X', PL_thistoken);
5612 PL_thistoken = newSVpvs("");
5619 /* Guess harder when madskills require "best effort". */
5620 if (PL_madskills && (!gv || !GvCVu(gv))) {
5621 int probable_sub = 0;
5622 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5624 else if (isALPHA(*s)) {
5628 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5629 if (!keyword(tmpbuf, tmplen, 0))
5632 while (d < PL_bufend && isSPACE(*d))
5634 if (*d == '=' && d[1] == '>')
5639 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5640 op_free(pl_yylval.opval);
5641 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5642 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5643 PL_last_lop = PL_oldbufptr;
5644 PL_last_lop_op = OP_ENTERSUB;
5645 PL_nextwhite = PL_thiswhite;
5647 start_force(PL_curforce);
5648 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5650 PL_nextwhite = nextPL_nextwhite;
5651 curmad('X', PL_thistoken);
5652 PL_thistoken = newSVpvs("");
5657 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5664 /* Call it a bare word */
5666 if (PL_hints & HINT_STRICT_SUBS)
5667 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5670 /* after "print" and similar functions (corresponding to
5671 * "F? L" in opcode.pl), whatever wasn't already parsed as
5672 * a filehandle should be subject to "strict subs".
5673 * Likewise for the optional indirect-object argument to system
5674 * or exec, which can't be a bareword */
5675 if ((PL_last_lop_op == OP_PRINT
5676 || PL_last_lop_op == OP_PRTF
5677 || PL_last_lop_op == OP_SAY
5678 || PL_last_lop_op == OP_SYSTEM
5679 || PL_last_lop_op == OP_EXEC)
5680 && (PL_hints & HINT_STRICT_SUBS))
5681 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5682 if (lastchar != '-') {
5683 if (ckWARN(WARN_RESERVED)) {
5687 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5688 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5695 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
5696 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5697 "Operator or semicolon missing before %c%s",
5698 lastchar, PL_tokenbuf);
5699 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5700 "Ambiguous use of %c resolved as operator %c",
5701 lastchar, lastchar);
5707 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5708 newSVpv(CopFILE(PL_curcop),0));
5712 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5713 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5716 case KEY___PACKAGE__:
5717 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5719 ? newSVhek(HvNAME_HEK(PL_curstash))
5726 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5727 const char *pname = "main";
5728 if (PL_tokenbuf[2] == 'D')
5729 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5730 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5734 GvIOp(gv) = newIO();
5735 IoIFP(GvIOp(gv)) = PL_rsfp;
5736 #if defined(HAS_FCNTL) && defined(F_SETFD)
5738 const int fd = PerlIO_fileno(PL_rsfp);
5739 fcntl(fd,F_SETFD,fd >= 3);
5742 /* Mark this internal pseudo-handle as clean */
5743 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5744 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5745 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5747 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5748 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5749 /* if the script was opened in binmode, we need to revert
5750 * it to text mode for compatibility; but only iff it has CRs
5751 * XXX this is a questionable hack at best. */
5752 if (PL_bufend-PL_bufptr > 2
5753 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5756 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5757 loc = PerlIO_tell(PL_rsfp);
5758 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5761 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5763 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5764 #endif /* NETWARE */
5765 #ifdef PERLIO_IS_STDIO /* really? */
5766 # if defined(__BORLANDC__)
5767 /* XXX see note in do_binmode() */
5768 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5772 PerlIO_seek(PL_rsfp, loc, 0);
5776 #ifdef PERLIO_LAYERS
5779 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5780 else if (PL_encoding) {
5787 XPUSHs(PL_encoding);
5789 call_method("name", G_SCALAR);
5793 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5794 Perl_form(aTHX_ ":encoding(%"SVf")",
5803 if (PL_realtokenstart >= 0) {
5804 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5806 PL_endwhite = newSVpvs("");
5807 sv_catsv(PL_endwhite, PL_thiswhite);
5809 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5810 PL_realtokenstart = -1;
5812 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
5828 if (PL_expect == XSTATE) {
5835 if (*s == ':' && s[1] == ':') {
5838 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5839 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5840 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5843 else if (tmp == KEY_require || tmp == KEY_do)
5844 /* that's a way to remember we saw "CORE::" */
5857 LOP(OP_ACCEPT,XTERM);
5863 LOP(OP_ATAN2,XTERM);
5869 LOP(OP_BINMODE,XTERM);
5872 LOP(OP_BLESS,XTERM);
5881 /* When 'use switch' is in effect, continue has a dual
5882 life as a control operator. */
5884 if (!FEATURE_IS_ENABLED("switch"))
5887 /* We have to disambiguate the two senses of
5888 "continue". If the next token is a '{' then
5889 treat it as the start of a continue block;
5890 otherwise treat it as a control operator.
5902 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5919 if (!PL_cryptseen) {
5920 PL_cryptseen = TRUE;
5924 LOP(OP_CRYPT,XTERM);
5927 LOP(OP_CHMOD,XTERM);
5930 LOP(OP_CHOWN,XTERM);
5933 LOP(OP_CONNECT,XTERM);
5952 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5953 if (orig_keyword == KEY_do) {
5962 PL_hints |= HINT_BLOCK_SCOPE;
5972 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5973 LOP(OP_DBMOPEN,XTERM);
5979 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5986 pl_yylval.ival = CopLINE(PL_curcop);
6002 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6003 UNIBRACK(OP_ENTEREVAL);
6017 case KEY_endhostent:
6023 case KEY_endservent:
6026 case KEY_endprotoent:
6037 pl_yylval.ival = CopLINE(PL_curcop);
6039 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6042 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6045 if ((PL_bufend - p) >= 3 &&
6046 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6048 else if ((PL_bufend - p) >= 4 &&
6049 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6052 if (isIDFIRST_lazy_if(p,UTF)) {
6053 p = scan_ident(p, PL_bufend,
6054 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6058 Perl_croak(aTHX_ "Missing $ on loop variable");
6060 s = SvPVX(PL_linestr) + soff;
6066 LOP(OP_FORMLINE,XTERM);
6072 LOP(OP_FCNTL,XTERM);
6078 LOP(OP_FLOCK,XTERM);
6087 LOP(OP_GREPSTART, XREF);
6090 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6105 case KEY_getpriority:
6106 LOP(OP_GETPRIORITY,XTERM);
6108 case KEY_getprotobyname:
6111 case KEY_getprotobynumber:
6112 LOP(OP_GPBYNUMBER,XTERM);
6114 case KEY_getprotoent:
6126 case KEY_getpeername:
6127 UNI(OP_GETPEERNAME);
6129 case KEY_gethostbyname:
6132 case KEY_gethostbyaddr:
6133 LOP(OP_GHBYADDR,XTERM);
6135 case KEY_gethostent:
6138 case KEY_getnetbyname:
6141 case KEY_getnetbyaddr:
6142 LOP(OP_GNBYADDR,XTERM);
6147 case KEY_getservbyname:
6148 LOP(OP_GSBYNAME,XTERM);
6150 case KEY_getservbyport:
6151 LOP(OP_GSBYPORT,XTERM);
6153 case KEY_getservent:
6156 case KEY_getsockname:
6157 UNI(OP_GETSOCKNAME);
6159 case KEY_getsockopt:
6160 LOP(OP_GSOCKOPT,XTERM);
6175 pl_yylval.ival = CopLINE(PL_curcop);
6185 pl_yylval.ival = CopLINE(PL_curcop);
6189 LOP(OP_INDEX,XTERM);
6195 LOP(OP_IOCTL,XTERM);
6207 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6239 LOP(OP_LISTEN,XTERM);
6248 s = scan_pat(s,OP_MATCH);
6249 TERM(sublex_start());
6252 LOP(OP_MAPSTART, XREF);
6255 LOP(OP_MKDIR,XTERM);
6258 LOP(OP_MSGCTL,XTERM);
6261 LOP(OP_MSGGET,XTERM);
6264 LOP(OP_MSGRCV,XTERM);
6267 LOP(OP_MSGSND,XTERM);
6272 PL_in_my = (U16)tmp;
6274 if (isIDFIRST_lazy_if(s,UTF)) {
6278 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6279 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6281 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6282 if (!PL_in_my_stash) {
6285 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6289 if (PL_madskills) { /* just add type to declarator token */
6290 sv_catsv(PL_thistoken, PL_nextwhite);
6292 sv_catpvn(PL_thistoken, start, s - start);
6300 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6307 s = tokenize_use(0, s);
6311 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6318 if (isIDFIRST_lazy_if(s,UTF)) {
6320 for (d = s; isALNUM_lazy_if(d,UTF);)
6322 for (t=d; isSPACE(*t);)
6324 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6326 && !(t[0] == '=' && t[1] == '>')
6328 int parms_len = (int)(d-s);
6329 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6330 "Precedence problem: open %.*s should be open(%.*s)",
6331 parms_len, s, parms_len, s);
6337 pl_yylval.ival = OP_OR;
6347 LOP(OP_OPEN_DIR,XTERM);
6350 checkcomma(s,PL_tokenbuf,"filehandle");
6354 checkcomma(s,PL_tokenbuf,"filehandle");
6373 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6374 s = force_version(s, FALSE);
6378 LOP(OP_PIPE_OP,XTERM);
6381 s = scan_str(s,!!PL_madskills,FALSE);
6384 pl_yylval.ival = OP_CONST;
6385 TERM(sublex_start());
6391 s = scan_str(s,!!PL_madskills,FALSE);
6394 PL_expect = XOPERATOR;
6396 if (SvCUR(PL_lex_stuff)) {
6399 d = SvPV_force(PL_lex_stuff, len);
6401 for (; isSPACE(*d) && len; --len, ++d)
6406 if (!warned && ckWARN(WARN_QW)) {
6407 for (; !isSPACE(*d) && len; --len, ++d) {
6409 Perl_warner(aTHX_ packWARN(WARN_QW),
6410 "Possible attempt to separate words with commas");
6413 else if (*d == '#') {
6414 Perl_warner(aTHX_ packWARN(WARN_QW),
6415 "Possible attempt to put comments in qw() list");
6421 for (; !isSPACE(*d) && len; --len, ++d)
6424 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6425 words = append_elem(OP_LIST, words,
6426 newSVOP(OP_CONST, 0, tokeq(sv)));
6430 start_force(PL_curforce);
6431 NEXTVAL_NEXTTOKE.opval = words;
6436 SvREFCNT_dec(PL_lex_stuff);
6437 PL_lex_stuff = NULL;
6443 s = scan_str(s,!!PL_madskills,FALSE);
6446 pl_yylval.ival = OP_STRINGIFY;
6447 if (SvIVX(PL_lex_stuff) == '\'')
6448 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6449 TERM(sublex_start());
6452 s = scan_pat(s,OP_QR);
6453 TERM(sublex_start());
6456 s = scan_str(s,!!PL_madskills,FALSE);
6459 readpipe_override();
6460 TERM(sublex_start());
6468 s = force_version(s, FALSE);
6470 else if (*s != 'v' || !isDIGIT(s[1])
6471 || (s = force_version(s, TRUE), *s == 'v'))
6473 *PL_tokenbuf = '\0';
6474 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6475 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6476 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6478 yyerror("<> should be quotes");
6480 if (orig_keyword == KEY_require) {
6488 PL_last_uni = PL_oldbufptr;
6489 PL_last_lop_op = OP_REQUIRE;
6491 return REPORT( (int)REQUIRE );
6497 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6501 LOP(OP_RENAME,XTERM);
6510 LOP(OP_RINDEX,XTERM);
6519 UNIDOR(OP_READLINE);
6522 UNIDOR(OP_BACKTICK);
6531 LOP(OP_REVERSE,XTERM);
6534 UNIDOR(OP_READLINK);
6541 if (pl_yylval.opval)
6542 TERM(sublex_start());
6544 TOKEN(1); /* force error */
6547 checkcomma(s,PL_tokenbuf,"filehandle");
6557 LOP(OP_SELECT,XTERM);
6563 LOP(OP_SEMCTL,XTERM);
6566 LOP(OP_SEMGET,XTERM);
6569 LOP(OP_SEMOP,XTERM);
6575 LOP(OP_SETPGRP,XTERM);
6577 case KEY_setpriority:
6578 LOP(OP_SETPRIORITY,XTERM);
6580 case KEY_sethostent:
6586 case KEY_setservent:
6589 case KEY_setprotoent:
6599 LOP(OP_SEEKDIR,XTERM);
6601 case KEY_setsockopt:
6602 LOP(OP_SSOCKOPT,XTERM);
6608 LOP(OP_SHMCTL,XTERM);
6611 LOP(OP_SHMGET,XTERM);
6614 LOP(OP_SHMREAD,XTERM);
6617 LOP(OP_SHMWRITE,XTERM);
6620 LOP(OP_SHUTDOWN,XTERM);
6629 LOP(OP_SOCKET,XTERM);
6631 case KEY_socketpair:
6632 LOP(OP_SOCKPAIR,XTERM);
6635 checkcomma(s,PL_tokenbuf,"subroutine name");
6637 if (*s == ';' || *s == ')') /* probably a close */
6638 Perl_croak(aTHX_ "sort is now a reserved word");
6640 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6644 LOP(OP_SPLIT,XTERM);
6647 LOP(OP_SPRINTF,XTERM);
6650 LOP(OP_SPLICE,XTERM);
6665 LOP(OP_SUBSTR,XTERM);
6671 char tmpbuf[sizeof PL_tokenbuf];
6672 SSize_t tboffset = 0;
6673 expectation attrful;
6674 bool have_name, have_proto;
6675 const int key = tmp;
6680 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6681 SV *subtoken = newSVpvn(tstart, s - tstart);
6685 s = SKIPSPACE2(s,tmpwhite);
6690 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6691 (*s == ':' && s[1] == ':'))
6694 SV *nametoke = NULL;
6698 attrful = XATTRBLOCK;
6699 /* remember buffer pos'n for later force_word */
6700 tboffset = s - PL_oldbufptr;
6701 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6704 nametoke = newSVpvn(s, d - s);
6706 if (memchr(tmpbuf, ':', len))
6707 sv_setpvn(PL_subname, tmpbuf, len);
6709 sv_setsv(PL_subname,PL_curstname);
6710 sv_catpvs(PL_subname,"::");
6711 sv_catpvn(PL_subname,tmpbuf,len);
6718 CURMAD('X', nametoke);
6719 CURMAD('_', tmpwhite);
6720 (void) force_word(PL_oldbufptr + tboffset, WORD,
6723 s = SKIPSPACE2(d,tmpwhite);
6730 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6731 PL_expect = XTERMBLOCK;
6732 attrful = XATTRTERM;
6733 sv_setpvs(PL_subname,"?");
6737 if (key == KEY_format) {
6739 PL_lex_formbrack = PL_lex_brackets + 1;
6741 PL_thistoken = subtoken;
6745 (void) force_word(PL_oldbufptr + tboffset, WORD,
6751 /* Look for a prototype */
6754 bool bad_proto = FALSE;
6755 bool in_brackets = FALSE;
6756 char greedy_proto = ' ';
6757 bool proto_after_greedy_proto = FALSE;
6758 bool must_be_last = FALSE;
6759 bool underscore = FALSE;
6760 bool seen_underscore = FALSE;
6761 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6763 s = scan_str(s,!!PL_madskills,FALSE);
6765 Perl_croak(aTHX_ "Prototype not terminated");
6766 /* strip spaces and check for bad characters */
6767 d = SvPVX(PL_lex_stuff);
6769 for (p = d; *p; ++p) {
6775 proto_after_greedy_proto = TRUE;
6776 if (!strchr("$@%*;[]&\\_", *p)) {
6788 else if ( *p == ']' ) {
6789 in_brackets = FALSE;
6791 else if ( (*p == '@' || *p == '%') &&
6792 ( tmp < 2 || d[tmp-2] != '\\' ) &&
6794 must_be_last = TRUE;
6797 else if ( *p == '_' ) {
6798 underscore = seen_underscore = TRUE;
6805 if (proto_after_greedy_proto)
6806 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6807 "Prototype after '%c' for %"SVf" : %s",
6808 greedy_proto, SVfARG(PL_subname), d);
6810 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6811 "Illegal character %sin prototype for %"SVf" : %s",
6812 seen_underscore ? "after '_' " : "",
6813 SVfARG(PL_subname), d);
6814 SvCUR_set(PL_lex_stuff, tmp);
6819 CURMAD('q', PL_thisopen);
6820 CURMAD('_', tmpwhite);
6821 CURMAD('=', PL_thisstuff);
6822 CURMAD('Q', PL_thisclose);
6823 NEXTVAL_NEXTTOKE.opval =
6824 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6825 PL_lex_stuff = NULL;
6828 s = SKIPSPACE2(s,tmpwhite);
6836 if (*s == ':' && s[1] != ':')
6837 PL_expect = attrful;
6838 else if (*s != '{' && key == KEY_sub) {
6840 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6842 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6849 curmad('^', newSVpvs(""));
6850 CURMAD('_', tmpwhite);
6854 PL_thistoken = subtoken;
6857 NEXTVAL_NEXTTOKE.opval =
6858 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6859 PL_lex_stuff = NULL;
6865 sv_setpvs(PL_subname, "__ANON__");
6867 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6871 (void) force_word(PL_oldbufptr + tboffset, WORD,
6880 LOP(OP_SYSTEM,XREF);
6883 LOP(OP_SYMLINK,XTERM);
6886 LOP(OP_SYSCALL,XTERM);
6889 LOP(OP_SYSOPEN,XTERM);
6892 LOP(OP_SYSSEEK,XTERM);
6895 LOP(OP_SYSREAD,XTERM);
6898 LOP(OP_SYSWRITE,XTERM);
6902 TERM(sublex_start());
6923 LOP(OP_TRUNCATE,XTERM);
6935 pl_yylval.ival = CopLINE(PL_curcop);
6939 pl_yylval.ival = CopLINE(PL_curcop);
6943 LOP(OP_UNLINK,XTERM);
6949 LOP(OP_UNPACK,XTERM);
6952 LOP(OP_UTIME,XTERM);
6958 LOP(OP_UNSHIFT,XTERM);
6961 s = tokenize_use(1, s);
6971 pl_yylval.ival = CopLINE(PL_curcop);
6975 pl_yylval.ival = CopLINE(PL_curcop);
6979 PL_hints |= HINT_BLOCK_SCOPE;
6986 LOP(OP_WAITPID,XTERM);
6995 ctl_l[0] = toCTRL('L');
6997 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7000 /* Make sure $^L is defined */
7001 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7006 if (PL_expect == XOPERATOR)
7012 pl_yylval.ival = OP_XOR;
7017 TERM(sublex_start());
7022 #pragma segment Main
7026 S_pending_ident(pTHX)
7031 /* pit holds the identifier we read and pending_ident is reset */
7032 char pit = PL_pending_ident;
7033 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7034 /* All routes through this function want to know if there is a colon. */
7035 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7036 PL_pending_ident = 0;
7038 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7039 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7040 "### Pending identifier '%s'\n", PL_tokenbuf); });
7042 /* if we're in a my(), we can't allow dynamics here.
7043 $foo'bar has already been turned into $foo::bar, so
7044 just check for colons.
7046 if it's a legal name, the OP is a PADANY.
7049 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7051 yyerror(Perl_form(aTHX_ "No package name allowed for "
7052 "variable %s in \"our\"",
7054 tmp = allocmy(PL_tokenbuf);
7058 yyerror(Perl_form(aTHX_ PL_no_myglob,
7059 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7061 pl_yylval.opval = newOP(OP_PADANY, 0);
7062 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
7068 build the ops for accesses to a my() variable.
7070 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7071 then used in a comparison. This catches most, but not
7072 all cases. For instance, it catches
7073 sort { my($a); $a <=> $b }
7075 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7076 (although why you'd do that is anyone's guess).
7081 tmp = pad_findmy(PL_tokenbuf);
7082 if (tmp != NOT_IN_PAD) {
7083 /* might be an "our" variable" */
7084 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7085 /* build ops for a bareword */
7086 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7087 HEK * const stashname = HvNAME_HEK(stash);
7088 SV * const sym = newSVhek(stashname);
7089 sv_catpvs(sym, "::");
7090 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7091 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7092 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7095 ? (GV_ADDMULTI | GV_ADDINEVAL)
7098 ((PL_tokenbuf[0] == '$') ? SVt_PV
7099 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7104 /* if it's a sort block and they're naming $a or $b */
7105 if (PL_last_lop_op == OP_SORT &&
7106 PL_tokenbuf[0] == '$' &&
7107 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7110 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7111 d < PL_bufend && *d != '\n';
7114 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7115 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7121 pl_yylval.opval = newOP(OP_PADANY, 0);
7122 pl_yylval.opval->op_targ = tmp;
7128 Whine if they've said @foo in a doublequoted string,
7129 and @foo isn't a variable we can find in the symbol
7132 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7133 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7135 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7136 /* DO NOT warn for @- and @+ */
7137 && !( PL_tokenbuf[2] == '\0' &&
7138 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7141 /* Downgraded from fatal to warning 20000522 mjd */
7142 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7143 "Possible unintended interpolation of %s in string",
7148 /* build ops for a bareword */
7149 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7151 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7153 PL_tokenbuf + 1, tokenbuf_len - 1,
7154 /* If the identifier refers to a stash, don't autovivify it.
7155 * Change 24660 had the side effect of causing symbol table
7156 * hashes to always be defined, even if they were freshly
7157 * created and the only reference in the entire program was
7158 * the single statement with the defined %foo::bar:: test.
7159 * It appears that all code in the wild doing this actually
7160 * wants to know whether sub-packages have been loaded, so
7161 * by avoiding auto-vivifying symbol tables, we ensure that
7162 * defined %foo::bar:: continues to be false, and the existing
7163 * tests still give the expected answers, even though what
7164 * they're actually testing has now changed subtly.
7166 (*PL_tokenbuf == '%'
7167 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7170 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7171 ((PL_tokenbuf[0] == '$') ? SVt_PV
7172 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7178 * The following code was generated by perl_keyword.pl.
7182 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7186 PERL_ARGS_ASSERT_KEYWORD;
7190 case 1: /* 5 tokens of length 1 */
7222 case 2: /* 18 tokens of length 2 */
7368 case 3: /* 29 tokens of length 3 */
7372 if (name[1] == 'N' &&
7435 if (name[1] == 'i' &&
7467 if (name[1] == 'o' &&
7476 if (name[1] == 'e' &&
7485 if (name[1] == 'n' &&
7494 if (name[1] == 'o' &&
7503 if (name[1] == 'a' &&
7512 if (name[1] == 'o' &&
7574 if (name[1] == 'e' &&
7588 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7614 if (name[1] == 'i' &&
7623 if (name[1] == 's' &&
7632 if (name[1] == 'e' &&
7641 if (name[1] == 'o' &&
7653 case 4: /* 41 tokens of length 4 */
7657 if (name[1] == 'O' &&
7667 if (name[1] == 'N' &&
7677 if (name[1] == 'i' &&
7687 if (name[1] == 'h' &&
7697 if (name[1] == 'u' &&
7710 if (name[2] == 'c' &&
7719 if (name[2] == 's' &&
7728 if (name[2] == 'a' &&
7764 if (name[1] == 'o' &&
7777 if (name[2] == 't' &&
7786 if (name[2] == 'o' &&
7795 if (name[2] == 't' &&
7804 if (name[2] == 'e' &&
7817 if (name[1] == 'o' &&
7830 if (name[2] == 'y' &&
7839 if (name[2] == 'l' &&
7855 if (name[2] == 's' &&
7864 if (name[2] == 'n' &&
7873 if (name[2] == 'c' &&
7886 if (name[1] == 'e' &&
7896 if (name[1] == 'p' &&
7909 if (name[2] == 'c' &&
7918 if (name[2] == 'p' &&
7927 if (name[2] == 's' &&
7943 if (name[2] == 'n' &&
8013 if (name[2] == 'r' &&
8022 if (name[2] == 'r' &&
8031 if (name[2] == 'a' &&
8047 if (name[2] == 'l' &&
8109 if (name[2] == 'e' &&
8112 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8125 case 5: /* 39 tokens of length 5 */
8129 if (name[1] == 'E' &&
8140 if (name[1] == 'H' &&
8154 if (name[2] == 'a' &&
8164 if (name[2] == 'a' &&
8181 if (name[2] == 'e' &&
8191 if (name[2] == 'e' &&
8195 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8211 if (name[3] == 'i' &&
8220 if (name[3] == 'o' &&
8256 if (name[2] == 'o' &&
8266 if (name[2] == 'y' &&
8280 if (name[1] == 'l' &&
8294 if (name[2] == 'n' &&
8304 if (name[2] == 'o' &&
8318 if (name[1] == 'i' &&
8323 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8332 if (name[2] == 'd' &&
8342 if (name[2] == 'c' &&
8359 if (name[2] == 'c' &&
8369 if (name[2] == 't' &&
8383 if (name[1] == 'k' &&
8394 if (name[1] == 'r' &&
8408 if (name[2] == 's' &&
8418 if (name[2] == 'd' &&
8435 if (name[2] == 'm' &&
8445 if (name[2] == 'i' &&
8455 if (name[2] == 'e' &&
8465 if (name[2] == 'l' &&
8475 if (name[2] == 'a' &&
8488 if (name[3] == 't' &&
8491 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8497 if (name[3] == 'd' &&
8514 if (name[1] == 'i' &&
8528 if (name[2] == 'a' &&
8541 if (name[3] == 'e' &&
8576 if (name[2] == 'i' &&
8593 if (name[2] == 'i' &&
8603 if (name[2] == 'i' &&
8620 case 6: /* 33 tokens of length 6 */
8624 if (name[1] == 'c' &&
8639 if (name[2] == 'l' &&
8650 if (name[2] == 'r' &&
8665 if (name[1] == 'e' &&
8680 if (name[2] == 's' &&
8685 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8691 if (name[2] == 'i' &&
8709 if (name[2] == 'l' &&
8720 if (name[2] == 'r' &&
8735 if (name[1] == 'm' &&
8750 if (name[2] == 'n' &&
8761 if (name[2] == 's' &&
8776 if (name[1] == 's' &&
8782 if (name[4] == 't' &&
8791 if (name[4] == 'e' &&
8800 if (name[4] == 'c' &&
8809 if (name[4] == 'n' &&
8825 if (name[1] == 'r' &&
8843 if (name[3] == 'a' &&
8853 if (name[3] == 'u' &&
8867 if (name[2] == 'n' &&
8885 if (name[2] == 'a' &&
8899 if (name[3] == 'e' &&
8912 if (name[4] == 't' &&
8921 if (name[4] == 'e' &&
8943 if (name[4] == 't' &&
8952 if (name[4] == 'e' &&
8968 if (name[2] == 'c' &&
8979 if (name[2] == 'l' &&
8990 if (name[2] == 'b' &&
9001 if (name[2] == 's' &&
9024 if (name[4] == 's' &&
9033 if (name[4] == 'n' &&
9046 if (name[3] == 'a' &&
9063 if (name[1] == 'a' &&
9078 case 7: /* 29 tokens of length 7 */
9082 if (name[1] == 'E' &&
9095 if (name[1] == '_' &&
9108 if (name[1] == 'i' &&
9115 return -KEY_binmode;
9121 if (name[1] == 'o' &&
9128 return -KEY_connect;
9137 if (name[2] == 'm' &&
9143 return -KEY_dbmopen;
9154 if (name[4] == 'u' &&
9158 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9164 if (name[4] == 'n' &&
9185 if (name[1] == 'o' &&
9198 if (name[1] == 'e' &&
9205 if (name[5] == 'r' &&
9208 return -KEY_getpgrp;
9214 if (name[5] == 'i' &&
9217 return -KEY_getppid;
9230 if (name[1] == 'c' &&
9237 return -KEY_lcfirst;
9243 if (name[1] == 'p' &&
9250 return -KEY_opendir;
9256 if (name[1] == 'a' &&
9274 if (name[3] == 'd' &&
9279 return -KEY_readdir;
9285 if (name[3] == 'u' &&
9296 if (name[3] == 'e' &&
9301 return -KEY_reverse;
9320 if (name[3] == 'k' &&
9325 return -KEY_seekdir;
9331 if (name[3] == 'p' &&
9336 return -KEY_setpgrp;
9346 if (name[2] == 'm' &&
9352 return -KEY_shmread;
9358 if (name[2] == 'r' &&
9364 return -KEY_sprintf;
9373 if (name[3] == 'l' &&
9378 return -KEY_symlink;
9387 if (name[4] == 'a' &&
9391 return -KEY_syscall;
9397 if (name[4] == 'p' &&
9401 return -KEY_sysopen;
9407 if (name[4] == 'e' &&
9411 return -KEY_sysread;
9417 if (name[4] == 'e' &&
9421 return -KEY_sysseek;
9439 if (name[1] == 'e' &&
9446 return -KEY_telldir;
9455 if (name[2] == 'f' &&
9461 return -KEY_ucfirst;
9467 if (name[2] == 's' &&
9473 return -KEY_unshift;
9483 if (name[1] == 'a' &&
9490 return -KEY_waitpid;
9499 case 8: /* 26 tokens of length 8 */
9503 if (name[1] == 'U' &&
9511 return KEY_AUTOLOAD;
9522 if (name[3] == 'A' &&
9528 return KEY___DATA__;
9534 if (name[3] == 'I' &&
9540 return -KEY___FILE__;
9546 if (name[3] == 'I' &&
9552 return -KEY___LINE__;
9568 if (name[2] == 'o' &&
9575 return -KEY_closedir;
9581 if (name[2] == 'n' &&
9588 return -KEY_continue;
9598 if (name[1] == 'b' &&
9606 return -KEY_dbmclose;
9612 if (name[1] == 'n' &&
9618 if (name[4] == 'r' &&
9623 return -KEY_endgrent;
9629 if (name[4] == 'w' &&
9634 return -KEY_endpwent;
9647 if (name[1] == 'o' &&
9655 return -KEY_formline;
9661 if (name[1] == 'e' &&
9672 if (name[6] == 'n' &&
9675 return -KEY_getgrent;
9681 if (name[6] == 'i' &&
9684 return -KEY_getgrgid;
9690 if (name[6] == 'a' &&
9693 return -KEY_getgrnam;
9706 if (name[4] == 'o' &&
9711 return -KEY_getlogin;
9722 if (name[6] == 'n' &&
9725 return -KEY_getpwent;
9731 if (name[6] == 'a' &&
9734 return -KEY_getpwnam;
9740 if (name[6] == 'i' &&
9743 return -KEY_getpwuid;
9763 if (name[1] == 'e' &&
9770 if (name[5] == 'i' &&
9777 return -KEY_readline;
9782 return -KEY_readlink;
9793 if (name[5] == 'i' &&
9797 return -KEY_readpipe;
9818 if (name[4] == 'r' &&
9823 return -KEY_setgrent;
9829 if (name[4] == 'w' &&
9834 return -KEY_setpwent;
9850 if (name[3] == 'w' &&
9856 return -KEY_shmwrite;
9862 if (name[3] == 't' &&
9868 return -KEY_shutdown;
9878 if (name[2] == 's' &&
9885 return -KEY_syswrite;
9895 if (name[1] == 'r' &&
9903 return -KEY_truncate;
9912 case 9: /* 9 tokens of length 9 */
9916 if (name[1] == 'N' &&
9925 return KEY_UNITCHECK;
9931 if (name[1] == 'n' &&
9940 return -KEY_endnetent;
9946 if (name[1] == 'e' &&
9955 return -KEY_getnetent;
9961 if (name[1] == 'o' &&
9970 return -KEY_localtime;
9976 if (name[1] == 'r' &&
9985 return KEY_prototype;
9991 if (name[1] == 'u' &&
10000 return -KEY_quotemeta;
10006 if (name[1] == 'e' &&
10015 return -KEY_rewinddir;
10021 if (name[1] == 'e' &&
10030 return -KEY_setnetent;
10036 if (name[1] == 'a' &&
10045 return -KEY_wantarray;
10054 case 10: /* 9 tokens of length 10 */
10058 if (name[1] == 'n' &&
10064 if (name[4] == 'o' &&
10071 return -KEY_endhostent;
10077 if (name[4] == 'e' &&
10084 return -KEY_endservent;
10097 if (name[1] == 'e' &&
10103 if (name[4] == 'o' &&
10110 return -KEY_gethostent;
10119 if (name[5] == 'r' &&
10125 return -KEY_getservent;
10131 if (name[5] == 'c' &&
10137 return -KEY_getsockopt;
10157 if (name[2] == 't')
10162 if (name[4] == 'o' &&
10169 return -KEY_sethostent;
10178 if (name[5] == 'r' &&
10184 return -KEY_setservent;
10190 if (name[5] == 'c' &&
10196 return -KEY_setsockopt;
10213 if (name[2] == 'c' &&
10222 return -KEY_socketpair;
10235 case 11: /* 8 tokens of length 11 */
10239 if (name[1] == '_' &&
10249 { /* __PACKAGE__ */
10250 return -KEY___PACKAGE__;
10256 if (name[1] == 'n' &&
10266 { /* endprotoent */
10267 return -KEY_endprotoent;
10273 if (name[1] == 'e' &&
10282 if (name[5] == 'e' &&
10288 { /* getpeername */
10289 return -KEY_getpeername;
10298 if (name[6] == 'o' &&
10303 { /* getpriority */
10304 return -KEY_getpriority;
10310 if (name[6] == 't' &&
10315 { /* getprotoent */
10316 return -KEY_getprotoent;
10330 if (name[4] == 'o' &&
10337 { /* getsockname */
10338 return -KEY_getsockname;
10351 if (name[1] == 'e' &&
10359 if (name[6] == 'o' &&
10364 { /* setpriority */
10365 return -KEY_setpriority;
10371 if (name[6] == 't' &&
10376 { /* setprotoent */
10377 return -KEY_setprotoent;
10393 case 12: /* 2 tokens of length 12 */
10394 if (name[0] == 'g' &&
10406 if (name[9] == 'd' &&
10409 { /* getnetbyaddr */
10410 return -KEY_getnetbyaddr;
10416 if (name[9] == 'a' &&
10419 { /* getnetbyname */
10420 return -KEY_getnetbyname;
10432 case 13: /* 4 tokens of length 13 */
10433 if (name[0] == 'g' &&
10440 if (name[4] == 'o' &&
10449 if (name[10] == 'd' &&
10452 { /* gethostbyaddr */
10453 return -KEY_gethostbyaddr;
10459 if (name[10] == 'a' &&
10462 { /* gethostbyname */
10463 return -KEY_gethostbyname;
10476 if (name[4] == 'e' &&
10485 if (name[10] == 'a' &&
10488 { /* getservbyname */
10489 return -KEY_getservbyname;
10495 if (name[10] == 'o' &&
10498 { /* getservbyport */
10499 return -KEY_getservbyport;
10518 case 14: /* 1 tokens of length 14 */
10519 if (name[0] == 'g' &&
10533 { /* getprotobyname */
10534 return -KEY_getprotobyname;
10539 case 16: /* 1 tokens of length 16 */
10540 if (name[0] == 'g' &&
10556 { /* getprotobynumber */
10557 return -KEY_getprotobynumber;
10571 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10575 PERL_ARGS_ASSERT_CHECKCOMMA;
10577 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10578 if (ckWARN(WARN_SYNTAX)) {
10581 for (w = s+2; *w && level; w++) {
10584 else if (*w == ')')
10587 while (isSPACE(*w))
10589 /* the list of chars below is for end of statements or
10590 * block / parens, boolean operators (&&, ||, //) and branch
10591 * constructs (or, and, if, until, unless, while, err, for).
10592 * Not a very solid hack... */
10593 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10594 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10595 "%s (...) interpreted as function",name);
10598 while (s < PL_bufend && isSPACE(*s))
10602 while (s < PL_bufend && isSPACE(*s))
10604 if (isIDFIRST_lazy_if(s,UTF)) {
10605 const char * const w = s++;
10606 while (isALNUM_lazy_if(s,UTF))
10608 while (s < PL_bufend && isSPACE(*s))
10612 if (keyword(w, s - w, 0))
10615 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10616 if (gv && GvCVu(gv))
10618 Perl_croak(aTHX_ "No comma allowed after %s", what);
10623 /* Either returns sv, or mortalizes sv and returns a new SV*.
10624 Best used as sv=new_constant(..., sv, ...).
10625 If s, pv are NULL, calls subroutine with one argument,
10626 and type is used with error messages only. */
10629 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10630 SV *sv, SV *pv, const char *type, STRLEN typelen)
10633 HV * const table = GvHV(PL_hintgv); /* ^H */
10637 const char *why1 = "", *why2 = "", *why3 = "";
10639 PERL_ARGS_ASSERT_NEW_CONSTANT;
10641 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10644 why2 = (const char *)
10645 (strEQ(key,"charnames")
10646 ? "(possibly a missing \"use charnames ...\")"
10648 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10649 (type ? type: "undef"), why2);
10651 /* This is convoluted and evil ("goto considered harmful")
10652 * but I do not understand the intricacies of all the different
10653 * failure modes of %^H in here. The goal here is to make
10654 * the most probable error message user-friendly. --jhi */
10659 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10660 (type ? type: "undef"), why1, why2, why3);
10662 yyerror(SvPVX_const(msg));
10666 cvp = hv_fetch(table, key, keylen, FALSE);
10667 if (!cvp || !SvOK(*cvp)) {
10670 why3 = "} is not defined";
10673 sv_2mortal(sv); /* Parent created it permanently */
10676 pv = newSVpvn_flags(s, len, SVs_TEMP);
10678 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10680 typesv = &PL_sv_undef;
10682 PUSHSTACKi(PERLSI_OVERLOAD);
10694 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10698 /* Check the eval first */
10699 if (!PL_in_eval && SvTRUE(ERRSV)) {
10700 sv_catpvs(ERRSV, "Propagated");
10701 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10703 res = SvREFCNT_inc_simple(sv);
10707 SvREFCNT_inc_simple_void(res);
10716 why1 = "Call to &{$^H{";
10718 why3 = "}} did not return a defined value";
10726 /* Returns a NUL terminated string, with the length of the string written to
10730 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10733 register char *d = dest;
10734 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10736 PERL_ARGS_ASSERT_SCAN_WORD;
10740 Perl_croak(aTHX_ ident_too_long);
10741 if (isALNUM(*s)) /* UTF handled below */
10743 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10748 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10752 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10753 char *t = s + UTF8SKIP(s);
10755 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10759 Perl_croak(aTHX_ ident_too_long);
10760 Copy(s, d, len, char);
10773 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10776 char *bracket = NULL;
10778 register char *d = dest;
10779 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10781 PERL_ARGS_ASSERT_SCAN_IDENT;
10786 while (isDIGIT(*s)) {
10788 Perl_croak(aTHX_ ident_too_long);
10795 Perl_croak(aTHX_ ident_too_long);
10796 if (isALNUM(*s)) /* UTF handled below */
10798 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10803 else if (*s == ':' && s[1] == ':') {
10807 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10808 char *t = s + UTF8SKIP(s);
10809 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10811 if (d + (t - s) > e)
10812 Perl_croak(aTHX_ ident_too_long);
10813 Copy(s, d, t - s, char);
10824 if (PL_lex_state != LEX_NORMAL)
10825 PL_lex_state = LEX_INTERPENDMAYBE;
10828 if (*s == '$' && s[1] &&
10829 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10842 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10847 if (isSPACE(s[-1])) {
10849 const char ch = *s++;
10850 if (!SPACE_OR_TAB(ch)) {
10856 if (isIDFIRST_lazy_if(d,UTF)) {
10860 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10861 end += UTF8SKIP(end);
10862 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10863 end += UTF8SKIP(end);
10865 Copy(s, d, end - s, char);
10870 while ((isALNUM(*s) || *s == ':') && d < e)
10873 Perl_croak(aTHX_ ident_too_long);
10876 while (s < send && SPACE_OR_TAB(*s))
10878 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10879 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10880 const char * const brack =
10882 ((*s == '[') ? "[...]" : "{...}");
10883 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10884 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10885 funny, dest, brack, funny, dest, brack);
10888 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10892 /* Handle extended ${^Foo} variables
10893 * 1999-02-27 mjd-perl-patch@plover.com */
10894 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10898 while (isALNUM(*s) && d < e) {
10902 Perl_croak(aTHX_ ident_too_long);
10907 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10908 PL_lex_state = LEX_INTERPEND;
10911 if (PL_lex_state == LEX_NORMAL) {
10912 if (ckWARN(WARN_AMBIGUOUS) &&
10913 (keyword(dest, d - dest, 0)
10914 || get_cvn_flags(dest, d - dest, 0)))
10918 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10919 "Ambiguous use of %c{%s} resolved to %c%s",
10920 funny, dest, funny, dest);
10925 s = bracket; /* let the parser handle it */
10929 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10930 PL_lex_state = LEX_INTERPEND;
10935 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10937 PERL_ARGS_ASSERT_PMFLAG;
10939 PERL_UNUSED_CONTEXT;
10941 const char c = (char)ch;
10943 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10944 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10945 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10946 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10947 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10953 S_scan_pat(pTHX_ char *start, I32 type)
10957 char *s = scan_str(start,!!PL_madskills,FALSE);
10958 const char * const valid_flags =
10959 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10964 PERL_ARGS_ASSERT_SCAN_PAT;
10967 const char * const delimiter = skipspace(start);
10971 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10972 : "Search pattern not terminated" ));
10975 pm = (PMOP*)newPMOP(type, 0);
10976 if (PL_multi_open == '?') {
10977 /* This is the only point in the code that sets PMf_ONCE: */
10978 pm->op_pmflags |= PMf_ONCE;
10980 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10981 allows us to restrict the list needed by reset to just the ??
10983 assert(type != OP_TRANS);
10985 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
10988 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
10991 elements = mg->mg_len / sizeof(PMOP**);
10992 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10993 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10994 mg->mg_len = elements * sizeof(PMOP**);
10995 PmopSTASH_set(pm,PL_curstash);
11001 while (*s && strchr(valid_flags, *s))
11002 pmflag(&pm->op_pmflags,*s++);
11004 if (PL_madskills && modstart != s) {
11005 SV* tmptoken = newSVpvn(modstart, s - modstart);
11006 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11009 /* issue a warning if /c is specified,but /g is not */
11010 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11012 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11013 "Use of /c modifier is meaningless without /g" );
11016 PL_lex_op = (OP*)pm;
11017 pl_yylval.ival = OP_MATCH;
11022 S_scan_subst(pTHX_ char *start)
11033 PERL_ARGS_ASSERT_SCAN_SUBST;
11035 pl_yylval.ival = OP_NULL;
11037 s = scan_str(start,!!PL_madskills,FALSE);
11040 Perl_croak(aTHX_ "Substitution pattern not terminated");
11042 if (s[-1] == PL_multi_open)
11045 if (PL_madskills) {
11046 CURMAD('q', PL_thisopen);
11047 CURMAD('_', PL_thiswhite);
11048 CURMAD('E', PL_thisstuff);
11049 CURMAD('Q', PL_thisclose);
11050 PL_realtokenstart = s - SvPVX(PL_linestr);
11054 first_start = PL_multi_start;
11055 s = scan_str(s,!!PL_madskills,FALSE);
11057 if (PL_lex_stuff) {
11058 SvREFCNT_dec(PL_lex_stuff);
11059 PL_lex_stuff = NULL;
11061 Perl_croak(aTHX_ "Substitution replacement not terminated");
11063 PL_multi_start = first_start; /* so whole substitution is taken together */
11065 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11068 if (PL_madskills) {
11069 CURMAD('z', PL_thisopen);
11070 CURMAD('R', PL_thisstuff);
11071 CURMAD('Z', PL_thisclose);
11077 if (*s == EXEC_PAT_MOD) {
11081 else if (strchr(S_PAT_MODS, *s))
11082 pmflag(&pm->op_pmflags,*s++);
11088 if (PL_madskills) {
11090 curmad('m', newSVpvn(modstart, s - modstart));
11091 append_madprops(PL_thismad, (OP*)pm, 0);
11095 if ((pm->op_pmflags & PMf_CONTINUE)) {
11096 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11100 SV * const repl = newSVpvs("");
11102 PL_sublex_info.super_bufptr = s;
11103 PL_sublex_info.super_bufend = PL_bufend;
11105 pm->op_pmflags |= PMf_EVAL;
11108 sv_catpvs(repl, "eval ");
11110 sv_catpvs(repl, "do ");
11112 sv_catpvs(repl, "{");
11113 sv_catsv(repl, PL_lex_repl);
11114 if (strchr(SvPVX(PL_lex_repl), '#'))
11115 sv_catpvs(repl, "\n");
11116 sv_catpvs(repl, "}");
11118 SvREFCNT_dec(PL_lex_repl);
11119 PL_lex_repl = repl;
11122 PL_lex_op = (OP*)pm;
11123 pl_yylval.ival = OP_SUBST;
11128 S_scan_trans(pTHX_ char *start)
11141 PERL_ARGS_ASSERT_SCAN_TRANS;
11143 pl_yylval.ival = OP_NULL;
11145 s = scan_str(start,!!PL_madskills,FALSE);
11147 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11149 if (s[-1] == PL_multi_open)
11152 if (PL_madskills) {
11153 CURMAD('q', PL_thisopen);
11154 CURMAD('_', PL_thiswhite);
11155 CURMAD('E', PL_thisstuff);
11156 CURMAD('Q', PL_thisclose);
11157 PL_realtokenstart = s - SvPVX(PL_linestr);
11161 s = scan_str(s,!!PL_madskills,FALSE);
11163 if (PL_lex_stuff) {
11164 SvREFCNT_dec(PL_lex_stuff);
11165 PL_lex_stuff = NULL;
11167 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11169 if (PL_madskills) {
11170 CURMAD('z', PL_thisopen);
11171 CURMAD('R', PL_thisstuff);
11172 CURMAD('Z', PL_thisclose);
11175 complement = del = squash = 0;
11182 complement = OPpTRANS_COMPLEMENT;
11185 del = OPpTRANS_DELETE;
11188 squash = OPpTRANS_SQUASH;
11197 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11198 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11199 o->op_private &= ~OPpTRANS_ALL;
11200 o->op_private |= del|squash|complement|
11201 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11202 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11205 pl_yylval.ival = OP_TRANS;
11208 if (PL_madskills) {
11210 curmad('m', newSVpvn(modstart, s - modstart));
11211 append_madprops(PL_thismad, o, 0);
11220 S_scan_heredoc(pTHX_ register char *s)
11224 I32 op_type = OP_SCALAR;
11228 const char *found_newline;
11232 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11234 I32 stuffstart = s - SvPVX(PL_linestr);
11237 PL_realtokenstart = -1;
11240 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11244 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11248 while (SPACE_OR_TAB(*peek))
11250 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11253 s = delimcpy(d, e, s, PL_bufend, term, &len);
11263 if (!isALNUM_lazy_if(s,UTF))
11264 deprecate("bare << to mean <<\"\"");
11265 for (; isALNUM_lazy_if(s,UTF); s++) {
11270 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11271 Perl_croak(aTHX_ "Delimiter for here document is too long");
11274 len = d - PL_tokenbuf;
11277 if (PL_madskills) {
11278 tstart = PL_tokenbuf + !outer;
11279 PL_thisclose = newSVpvn(tstart, len - !outer);
11280 tstart = SvPVX(PL_linestr) + stuffstart;
11281 PL_thisopen = newSVpvn(tstart, s - tstart);
11282 stuffstart = s - SvPVX(PL_linestr);
11285 #ifndef PERL_STRICT_CR
11286 d = strchr(s, '\r');
11288 char * const olds = s;
11290 while (s < PL_bufend) {
11296 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11305 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11312 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11313 herewas = newSVpvn(s,PL_bufend-s);
11317 herewas = newSVpvn(s-1,found_newline-s+1);
11320 herewas = newSVpvn(s,found_newline-s);
11324 if (PL_madskills) {
11325 tstart = SvPVX(PL_linestr) + stuffstart;
11327 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11329 PL_thisstuff = newSVpvn(tstart, s - tstart);
11332 s += SvCUR(herewas);
11335 stuffstart = s - SvPVX(PL_linestr);
11341 tmpstr = newSV_type(SVt_PVIV);
11342 SvGROW(tmpstr, 80);
11343 if (term == '\'') {
11344 op_type = OP_CONST;
11345 SvIV_set(tmpstr, -1);
11347 else if (term == '`') {
11348 op_type = OP_BACKTICK;
11349 SvIV_set(tmpstr, '\\');
11353 PL_multi_start = CopLINE(PL_curcop);
11354 PL_multi_open = PL_multi_close = '<';
11355 term = *PL_tokenbuf;
11356 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11357 char * const bufptr = PL_sublex_info.super_bufptr;
11358 char * const bufend = PL_sublex_info.super_bufend;
11359 char * const olds = s - SvCUR(herewas);
11360 s = strchr(bufptr, '\n');
11364 while (s < bufend &&
11365 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11367 CopLINE_inc(PL_curcop);
11370 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11371 missingterm(PL_tokenbuf);
11373 sv_setpvn(herewas,bufptr,d-bufptr+1);
11374 sv_setpvn(tmpstr,d+1,s-d);
11376 sv_catpvn(herewas,s,bufend-s);
11377 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11384 while (s < PL_bufend &&
11385 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11387 CopLINE_inc(PL_curcop);
11389 if (s >= PL_bufend) {
11390 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11391 missingterm(PL_tokenbuf);
11393 sv_setpvn(tmpstr,d+1,s-d);
11395 if (PL_madskills) {
11397 sv_catpvn(PL_thisstuff, d + 1, s - d);
11399 PL_thisstuff = newSVpvn(d + 1, s - d);
11400 stuffstart = s - SvPVX(PL_linestr);
11404 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11406 sv_catpvn(herewas,s,PL_bufend-s);
11407 sv_setsv(PL_linestr,herewas);
11408 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11409 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11410 PL_last_lop = PL_last_uni = NULL;
11413 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
11414 while (s >= PL_bufend) { /* multiple line string? */
11416 if (PL_madskills) {
11417 tstart = SvPVX(PL_linestr) + stuffstart;
11419 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11421 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11425 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart
11426 = filter_gets(PL_linestr, 0))) {
11427 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11428 missingterm(PL_tokenbuf);
11431 stuffstart = s - SvPVX(PL_linestr);
11433 CopLINE_inc(PL_curcop);
11434 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11435 PL_last_lop = PL_last_uni = NULL;
11436 #ifndef PERL_STRICT_CR
11437 if (PL_bufend - PL_linestart >= 2) {
11438 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11439 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11441 PL_bufend[-2] = '\n';
11443 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11445 else if (PL_bufend[-1] == '\r')
11446 PL_bufend[-1] = '\n';
11448 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11449 PL_bufend[-1] = '\n';
11451 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11452 update_debugger_info(PL_linestr, NULL, 0);
11453 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11454 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11455 *(SvPVX(PL_linestr) + off ) = ' ';
11456 sv_catsv(PL_linestr,herewas);
11457 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11458 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11462 sv_catsv(tmpstr,PL_linestr);
11467 PL_multi_end = CopLINE(PL_curcop);
11468 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11469 SvPV_shrink_to_cur(tmpstr);
11471 SvREFCNT_dec(herewas);
11473 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11475 else if (PL_encoding)
11476 sv_recode_to_utf8(tmpstr, PL_encoding);
11478 PL_lex_stuff = tmpstr;
11479 pl_yylval.ival = op_type;
11483 /* scan_inputsymbol
11484 takes: current position in input buffer
11485 returns: new position in input buffer
11486 side-effects: pl_yylval and lex_op are set.
11491 <FH> read from filehandle
11492 <pkg::FH> read from package qualified filehandle
11493 <pkg'FH> read from package qualified filehandle
11494 <$fh> read from filehandle in $fh
11495 <*.h> filename glob
11500 S_scan_inputsymbol(pTHX_ char *start)
11503 register char *s = start; /* current position in buffer */
11506 char *d = PL_tokenbuf; /* start of temp holding space */
11507 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11509 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
11511 end = strchr(s, '\n');
11514 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11516 /* die if we didn't have space for the contents of the <>,
11517 or if it didn't end, or if we see a newline
11520 if (len >= (I32)sizeof PL_tokenbuf)
11521 Perl_croak(aTHX_ "Excessively long <> operator");
11523 Perl_croak(aTHX_ "Unterminated <> operator");
11528 Remember, only scalar variables are interpreted as filehandles by
11529 this code. Anything more complex (e.g., <$fh{$num}>) will be
11530 treated as a glob() call.
11531 This code makes use of the fact that except for the $ at the front,
11532 a scalar variable and a filehandle look the same.
11534 if (*d == '$' && d[1]) d++;
11536 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11537 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11540 /* If we've tried to read what we allow filehandles to look like, and
11541 there's still text left, then it must be a glob() and not a getline.
11542 Use scan_str to pull out the stuff between the <> and treat it
11543 as nothing more than a string.
11546 if (d - PL_tokenbuf != len) {
11547 pl_yylval.ival = OP_GLOB;
11548 s = scan_str(start,!!PL_madskills,FALSE);
11550 Perl_croak(aTHX_ "Glob not terminated");
11554 bool readline_overriden = FALSE;
11557 /* we're in a filehandle read situation */
11560 /* turn <> into <ARGV> */
11562 Copy("ARGV",d,5,char);
11564 /* Check whether readline() is overriden */
11565 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11567 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11569 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11570 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11571 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11572 readline_overriden = TRUE;
11574 /* if <$fh>, create the ops to turn the variable into a
11578 /* try to find it in the pad for this block, otherwise find
11579 add symbol table ops
11581 const PADOFFSET tmp = pad_findmy(d);
11582 if (tmp != NOT_IN_PAD) {
11583 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11584 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11585 HEK * const stashname = HvNAME_HEK(stash);
11586 SV * const sym = sv_2mortal(newSVhek(stashname));
11587 sv_catpvs(sym, "::");
11588 sv_catpv(sym, d+1);
11593 OP * const o = newOP(OP_PADSV, 0);
11595 PL_lex_op = readline_overriden
11596 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11597 append_elem(OP_LIST, o,
11598 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11599 : (OP*)newUNOP(OP_READLINE, 0, o);
11608 ? (GV_ADDMULTI | GV_ADDINEVAL)
11611 PL_lex_op = readline_overriden
11612 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11613 append_elem(OP_LIST,
11614 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11615 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11616 : (OP*)newUNOP(OP_READLINE, 0,
11617 newUNOP(OP_RV2SV, 0,
11618 newGVOP(OP_GV, 0, gv)));
11620 if (!readline_overriden)
11621 PL_lex_op->op_flags |= OPf_SPECIAL;
11622 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11623 pl_yylval.ival = OP_NULL;
11626 /* If it's none of the above, it must be a literal filehandle
11627 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11629 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11630 PL_lex_op = readline_overriden
11631 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11632 append_elem(OP_LIST,
11633 newGVOP(OP_GV, 0, gv),
11634 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11635 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11636 pl_yylval.ival = OP_NULL;
11645 takes: start position in buffer
11646 keep_quoted preserve \ on the embedded delimiter(s)
11647 keep_delims preserve the delimiters around the string
11648 returns: position to continue reading from buffer
11649 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11650 updates the read buffer.
11652 This subroutine pulls a string out of the input. It is called for:
11653 q single quotes q(literal text)
11654 ' single quotes 'literal text'
11655 qq double quotes qq(interpolate $here please)
11656 " double quotes "interpolate $here please"
11657 qx backticks qx(/bin/ls -l)
11658 ` backticks `/bin/ls -l`
11659 qw quote words @EXPORT_OK = qw( func() $spam )
11660 m// regexp match m/this/
11661 s/// regexp substitute s/this/that/
11662 tr/// string transliterate tr/this/that/
11663 y/// string transliterate y/this/that/
11664 ($*@) sub prototypes sub foo ($)
11665 (stuff) sub attr parameters sub foo : attr(stuff)
11666 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11668 In most of these cases (all but <>, patterns and transliterate)
11669 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11670 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11671 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11674 It skips whitespace before the string starts, and treats the first
11675 character as the delimiter. If the delimiter is one of ([{< then
11676 the corresponding "close" character )]}> is used as the closing
11677 delimiter. It allows quoting of delimiters, and if the string has
11678 balanced delimiters ([{<>}]) it allows nesting.
11680 On success, the SV with the resulting string is put into lex_stuff or,
11681 if that is already non-NULL, into lex_repl. The second case occurs only
11682 when parsing the RHS of the special constructs s/// and tr/// (y///).
11683 For convenience, the terminating delimiter character is stuffed into
11688 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11691 SV *sv; /* scalar value: string */
11692 const char *tmps; /* temp string, used for delimiter matching */
11693 register char *s = start; /* current position in the buffer */
11694 register char term; /* terminating character */
11695 register char *to; /* current position in the sv's data */
11696 I32 brackets = 1; /* bracket nesting level */
11697 bool has_utf8 = FALSE; /* is there any utf8 content? */
11698 I32 termcode; /* terminating char. code */
11699 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11700 STRLEN termlen; /* length of terminating string */
11701 int last_off = 0; /* last position for nesting bracket */
11707 PERL_ARGS_ASSERT_SCAN_STR;
11709 /* skip space before the delimiter */
11715 if (PL_realtokenstart >= 0) {
11716 stuffstart = PL_realtokenstart;
11717 PL_realtokenstart = -1;
11720 stuffstart = start - SvPVX(PL_linestr);
11722 /* mark where we are, in case we need to report errors */
11725 /* after skipping whitespace, the next character is the terminator */
11728 termcode = termstr[0] = term;
11732 termcode = utf8_to_uvchr((U8*)s, &termlen);
11733 Copy(s, termstr, termlen, U8);
11734 if (!UTF8_IS_INVARIANT(term))
11738 /* mark where we are */
11739 PL_multi_start = CopLINE(PL_curcop);
11740 PL_multi_open = term;
11742 /* find corresponding closing delimiter */
11743 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11744 termcode = termstr[0] = term = tmps[5];
11746 PL_multi_close = term;
11748 /* create a new SV to hold the contents. 79 is the SV's initial length.
11749 What a random number. */
11750 sv = newSV_type(SVt_PVIV);
11752 SvIV_set(sv, termcode);
11753 (void)SvPOK_only(sv); /* validate pointer */
11755 /* move past delimiter and try to read a complete string */
11757 sv_catpvn(sv, s, termlen);
11760 tstart = SvPVX(PL_linestr) + stuffstart;
11761 if (!PL_thisopen && !keep_delims) {
11762 PL_thisopen = newSVpvn(tstart, s - tstart);
11763 stuffstart = s - SvPVX(PL_linestr);
11767 if (PL_encoding && !UTF) {
11771 int offset = s - SvPVX_const(PL_linestr);
11772 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11773 &offset, (char*)termstr, termlen);
11774 const char * const ns = SvPVX_const(PL_linestr) + offset;
11775 char * const svlast = SvEND(sv) - 1;
11777 for (; s < ns; s++) {
11778 if (*s == '\n' && !PL_rsfp)
11779 CopLINE_inc(PL_curcop);
11782 goto read_more_line;
11784 /* handle quoted delimiters */
11785 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11787 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11789 if ((svlast-1 - t) % 2) {
11790 if (!keep_quoted) {
11791 *(svlast-1) = term;
11793 SvCUR_set(sv, SvCUR(sv) - 1);
11798 if (PL_multi_open == PL_multi_close) {
11804 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11805 /* At here, all closes are "was quoted" one,
11806 so we don't check PL_multi_close. */
11808 if (!keep_quoted && *(t+1) == PL_multi_open)
11813 else if (*t == PL_multi_open)
11821 SvCUR_set(sv, w - SvPVX_const(sv));
11823 last_off = w - SvPVX(sv);
11824 if (--brackets <= 0)
11829 if (!keep_delims) {
11830 SvCUR_set(sv, SvCUR(sv) - 1);
11836 /* extend sv if need be */
11837 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11838 /* set 'to' to the next character in the sv's string */
11839 to = SvPVX(sv)+SvCUR(sv);
11841 /* if open delimiter is the close delimiter read unbridle */
11842 if (PL_multi_open == PL_multi_close) {
11843 for (; s < PL_bufend; s++,to++) {
11844 /* embedded newlines increment the current line number */
11845 if (*s == '\n' && !PL_rsfp)
11846 CopLINE_inc(PL_curcop);
11847 /* handle quoted delimiters */
11848 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11849 if (!keep_quoted && s[1] == term)
11851 /* any other quotes are simply copied straight through */
11855 /* terminate when run out of buffer (the for() condition), or
11856 have found the terminator */
11857 else if (*s == term) {
11860 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11863 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11869 /* if the terminator isn't the same as the start character (e.g.,
11870 matched brackets), we have to allow more in the quoting, and
11871 be prepared for nested brackets.
11874 /* read until we run out of string, or we find the terminator */
11875 for (; s < PL_bufend; s++,to++) {
11876 /* embedded newlines increment the line count */
11877 if (*s == '\n' && !PL_rsfp)
11878 CopLINE_inc(PL_curcop);
11879 /* backslashes can escape the open or closing characters */
11880 if (*s == '\\' && s+1 < PL_bufend) {
11881 if (!keep_quoted &&
11882 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11887 /* allow nested opens and closes */
11888 else if (*s == PL_multi_close && --brackets <= 0)
11890 else if (*s == PL_multi_open)
11892 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11897 /* terminate the copied string and update the sv's end-of-string */
11899 SvCUR_set(sv, to - SvPVX_const(sv));
11902 * this next chunk reads more into the buffer if we're not done yet
11906 break; /* handle case where we are done yet :-) */
11908 #ifndef PERL_STRICT_CR
11909 if (to - SvPVX_const(sv) >= 2) {
11910 if ((to[-2] == '\r' && to[-1] == '\n') ||
11911 (to[-2] == '\n' && to[-1] == '\r'))
11915 SvCUR_set(sv, to - SvPVX_const(sv));
11917 else if (to[-1] == '\r')
11920 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11925 /* if we're out of file, or a read fails, bail and reset the current
11926 line marker so we can report where the unterminated string began
11929 if (PL_madskills) {
11930 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11932 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11934 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11938 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart
11939 = filter_gets(PL_linestr, 0))) {
11941 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11947 /* we read a line, so increment our line counter */
11948 CopLINE_inc(PL_curcop);
11950 /* update debugger info */
11951 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11952 update_debugger_info(PL_linestr, NULL, 0);
11954 /* having changed the buffer, we must update PL_bufend */
11955 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11956 PL_last_lop = PL_last_uni = NULL;
11959 /* at this point, we have successfully read the delimited string */
11961 if (!PL_encoding || UTF) {
11963 if (PL_madskills) {
11964 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11965 const int len = s - tstart;
11967 sv_catpvn(PL_thisstuff, tstart, len);
11969 PL_thisstuff = newSVpvn(tstart, len);
11970 if (!PL_thisclose && !keep_delims)
11971 PL_thisclose = newSVpvn(s,termlen);
11976 sv_catpvn(sv, s, termlen);
11981 if (PL_madskills) {
11982 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11983 const int len = s - tstart - termlen;
11985 sv_catpvn(PL_thisstuff, tstart, len);
11987 PL_thisstuff = newSVpvn(tstart, len);
11988 if (!PL_thisclose && !keep_delims)
11989 PL_thisclose = newSVpvn(s - termlen,termlen);
11993 if (has_utf8 || PL_encoding)
11996 PL_multi_end = CopLINE(PL_curcop);
11998 /* if we allocated too much space, give some back */
11999 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12000 SvLEN_set(sv, SvCUR(sv) + 1);
12001 SvPV_renew(sv, SvLEN(sv));
12004 /* decide whether this is the first or second quoted string we've read
12017 takes: pointer to position in buffer
12018 returns: pointer to new position in buffer
12019 side-effects: builds ops for the constant in pl_yylval.op
12021 Read a number in any of the formats that Perl accepts:
12023 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12024 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12027 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12029 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12032 If it reads a number without a decimal point or an exponent, it will
12033 try converting the number to an integer and see if it can do so
12034 without loss of precision.
12038 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12041 register const char *s = start; /* current position in buffer */
12042 register char *d; /* destination in temp buffer */
12043 register char *e; /* end of temp buffer */
12044 NV nv; /* number read, as a double */
12045 SV *sv = NULL; /* place to put the converted number */
12046 bool floatit; /* boolean: int or float? */
12047 const char *lastub = NULL; /* position of last underbar */
12048 static char const number_too_long[] = "Number too long";
12050 PERL_ARGS_ASSERT_SCAN_NUM;
12052 /* We use the first character to decide what type of number this is */
12056 Perl_croak(aTHX_ "panic: scan_num");
12058 /* if it starts with a 0, it could be an octal number, a decimal in
12059 0.13 disguise, or a hexadecimal number, or a binary number. */
12063 u holds the "number so far"
12064 shift the power of 2 of the base
12065 (hex == 4, octal == 3, binary == 1)
12066 overflowed was the number more than we can hold?
12068 Shift is used when we add a digit. It also serves as an "are
12069 we in octal/hex/binary?" indicator to disallow hex characters
12070 when in octal mode.
12075 bool overflowed = FALSE;
12076 bool just_zero = TRUE; /* just plain 0 or binary number? */
12077 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12078 static const char* const bases[5] =
12079 { "", "binary", "", "octal", "hexadecimal" };
12080 static const char* const Bases[5] =
12081 { "", "Binary", "", "Octal", "Hexadecimal" };
12082 static const char* const maxima[5] =
12084 "0b11111111111111111111111111111111",
12088 const char *base, *Base, *max;
12090 /* check for hex */
12095 } else if (s[1] == 'b') {
12100 /* check for a decimal in disguise */
12101 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12103 /* so it must be octal */
12110 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12111 "Misplaced _ in number");
12115 base = bases[shift];
12116 Base = Bases[shift];
12117 max = maxima[shift];
12119 /* read the rest of the number */
12121 /* x is used in the overflow test,
12122 b is the digit we're adding on. */
12127 /* if we don't mention it, we're done */
12131 /* _ are ignored -- but warned about if consecutive */
12133 if (lastub && s == lastub + 1)
12134 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12135 "Misplaced _ in number");
12139 /* 8 and 9 are not octal */
12140 case '8': case '9':
12142 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12146 case '2': case '3': case '4':
12147 case '5': case '6': case '7':
12149 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12152 case '0': case '1':
12153 b = *s++ & 15; /* ASCII digit -> value of digit */
12157 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12158 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12159 /* make sure they said 0x */
12162 b = (*s++ & 7) + 9;
12164 /* Prepare to put the digit we have onto the end
12165 of the number so far. We check for overflows.
12171 x = u << shift; /* make room for the digit */
12173 if ((x >> shift) != u
12174 && !(PL_hints & HINT_NEW_BINARY)) {
12177 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12178 "Integer overflow in %s number",
12181 u = x | b; /* add the digit to the end */
12184 n *= nvshift[shift];
12185 /* If an NV has not enough bits in its
12186 * mantissa to represent an UV this summing of
12187 * small low-order numbers is a waste of time
12188 * (because the NV cannot preserve the
12189 * low-order bits anyway): we could just
12190 * remember when did we overflow and in the
12191 * end just multiply n by the right
12199 /* if we get here, we had success: make a scalar value from
12204 /* final misplaced underbar check */
12205 if (s[-1] == '_') {
12206 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12211 if (n > 4294967295.0)
12212 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12213 "%s number > %s non-portable",
12219 if (u > 0xffffffff)
12220 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12221 "%s number > %s non-portable",
12226 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12227 sv = new_constant(start, s - start, "integer",
12228 sv, NULL, NULL, 0);
12229 else if (PL_hints & HINT_NEW_BINARY)
12230 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12235 handle decimal numbers.
12236 we're also sent here when we read a 0 as the first digit
12238 case '1': case '2': case '3': case '4': case '5':
12239 case '6': case '7': case '8': case '9': case '.':
12242 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12245 /* read next group of digits and _ and copy into d */
12246 while (isDIGIT(*s) || *s == '_') {
12247 /* skip underscores, checking for misplaced ones
12251 if (lastub && s == lastub + 1)
12252 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12253 "Misplaced _ in number");
12257 /* check for end of fixed-length buffer */
12259 Perl_croak(aTHX_ number_too_long);
12260 /* if we're ok, copy the character */
12265 /* final misplaced underbar check */
12266 if (lastub && s == lastub + 1) {
12267 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12270 /* read a decimal portion if there is one. avoid
12271 3..5 being interpreted as the number 3. followed
12274 if (*s == '.' && s[1] != '.') {
12279 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12280 "Misplaced _ in number");
12284 /* copy, ignoring underbars, until we run out of digits.
12286 for (; isDIGIT(*s) || *s == '_'; s++) {
12287 /* fixed length buffer check */
12289 Perl_croak(aTHX_ number_too_long);
12291 if (lastub && s == lastub + 1)
12292 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12293 "Misplaced _ in number");
12299 /* fractional part ending in underbar? */
12300 if (s[-1] == '_') {
12301 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12302 "Misplaced _ in number");
12304 if (*s == '.' && isDIGIT(s[1])) {
12305 /* oops, it's really a v-string, but without the "v" */
12311 /* read exponent part, if present */
12312 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12316 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12317 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12319 /* stray preinitial _ */
12321 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12322 "Misplaced _ in number");
12326 /* allow positive or negative exponent */
12327 if (*s == '+' || *s == '-')
12330 /* stray initial _ */
12332 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12333 "Misplaced _ in number");
12337 /* read digits of exponent */
12338 while (isDIGIT(*s) || *s == '_') {
12341 Perl_croak(aTHX_ number_too_long);
12345 if (((lastub && s == lastub + 1) ||
12346 (!isDIGIT(s[1]) && s[1] != '_')))
12347 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12348 "Misplaced _ in number");
12355 /* make an sv from the string */
12359 We try to do an integer conversion first if no characters
12360 indicating "float" have been found.
12365 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12367 if (flags == IS_NUMBER_IN_UV) {
12369 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12372 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12373 if (uv <= (UV) IV_MIN)
12374 sv_setiv(sv, -(IV)uv);
12381 /* terminate the string */
12383 nv = Atof(PL_tokenbuf);
12388 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12389 const char *const key = floatit ? "float" : "integer";
12390 const STRLEN keylen = floatit ? 5 : 7;
12391 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12392 key, keylen, sv, NULL, NULL, 0);
12396 /* if it starts with a v, it could be a v-string */
12399 sv = newSV(5); /* preallocate storage space */
12400 s = scan_vstring(s, PL_bufend, sv);
12404 /* make the op for the constant and return */
12407 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12409 lvalp->opval = NULL;
12415 S_scan_formline(pTHX_ register char *s)
12418 register char *eol;
12420 SV * const stuff = newSVpvs("");
12421 bool needargs = FALSE;
12422 bool eofmt = FALSE;
12424 char *tokenstart = s;
12425 SV* savewhite = NULL;
12427 if (PL_madskills) {
12428 savewhite = PL_thiswhite;
12433 PERL_ARGS_ASSERT_SCAN_FORMLINE;
12435 while (!needargs) {
12438 #ifdef PERL_STRICT_CR
12439 while (SPACE_OR_TAB(*t))
12442 while (SPACE_OR_TAB(*t) || *t == '\r')
12445 if (*t == '\n' || t == PL_bufend) {
12450 if (PL_in_eval && !PL_rsfp) {
12451 eol = (char *) memchr(s,'\n',PL_bufend-s);
12456 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12458 for (t = s; t < eol; t++) {
12459 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12461 goto enough; /* ~~ must be first line in formline */
12463 if (*t == '@' || *t == '^')
12467 sv_catpvn(stuff, s, eol-s);
12468 #ifndef PERL_STRICT_CR
12469 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12470 char *end = SvPVX(stuff) + SvCUR(stuff);
12473 SvCUR_set(stuff, SvCUR(stuff) - 1);
12483 if (PL_madskills) {
12485 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12487 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12490 s = filter_gets(PL_linestr, 0);
12492 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12494 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12496 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12497 PL_last_lop = PL_last_uni = NULL;
12506 if (SvCUR(stuff)) {
12509 PL_lex_state = LEX_NORMAL;
12510 start_force(PL_curforce);
12511 NEXTVAL_NEXTTOKE.ival = 0;
12515 PL_lex_state = LEX_FORMLINE;
12517 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12519 else if (PL_encoding)
12520 sv_recode_to_utf8(stuff, PL_encoding);
12522 start_force(PL_curforce);
12523 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12525 start_force(PL_curforce);
12526 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12530 SvREFCNT_dec(stuff);
12532 PL_lex_formbrack = 0;
12536 if (PL_madskills) {
12538 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12540 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12541 PL_thiswhite = savewhite;
12548 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12551 const I32 oldsavestack_ix = PL_savestack_ix;
12552 CV* const outsidecv = PL_compcv;
12555 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12557 SAVEI32(PL_subline);
12558 save_item(PL_subname);
12559 SAVESPTR(PL_compcv);
12561 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
12562 CvFLAGS(PL_compcv) |= flags;
12564 PL_subline = CopLINE(PL_curcop);
12565 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12566 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
12567 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12569 return oldsavestack_ix;
12573 #pragma segment Perl_yylex
12576 S_yywarn(pTHX_ const char *const s)
12580 PERL_ARGS_ASSERT_YYWARN;
12582 PL_in_eval |= EVAL_WARNONLY;
12584 PL_in_eval &= ~EVAL_WARNONLY;
12589 Perl_yyerror(pTHX_ const char *const s)
12592 const char *where = NULL;
12593 const char *context = NULL;
12596 int yychar = PL_parser->yychar;
12598 PERL_ARGS_ASSERT_YYERROR;
12600 if (!yychar || (yychar == ';' && !PL_rsfp))
12602 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12603 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12604 PL_oldbufptr != PL_bufptr) {
12607 The code below is removed for NetWare because it abends/crashes on NetWare
12608 when the script has error such as not having the closing quotes like:
12609 if ($var eq "value)
12610 Checking of white spaces is anyway done in NetWare code.
12613 while (isSPACE(*PL_oldoldbufptr))
12616 context = PL_oldoldbufptr;
12617 contlen = PL_bufptr - PL_oldoldbufptr;
12619 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12620 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12623 The code below is removed for NetWare because it abends/crashes on NetWare
12624 when the script has error such as not having the closing quotes like:
12625 if ($var eq "value)
12626 Checking of white spaces is anyway done in NetWare code.
12629 while (isSPACE(*PL_oldbufptr))
12632 context = PL_oldbufptr;
12633 contlen = PL_bufptr - PL_oldbufptr;
12635 else if (yychar > 255)
12636 where = "next token ???";
12637 else if (yychar == -2) { /* YYEMPTY */
12638 if (PL_lex_state == LEX_NORMAL ||
12639 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12640 where = "at end of line";
12641 else if (PL_lex_inpat)
12642 where = "within pattern";
12644 where = "within string";
12647 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12649 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12650 else if (isPRINT_LC(yychar)) {
12651 const char string = yychar;
12652 sv_catpvn(where_sv, &string, 1);
12655 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12656 where = SvPVX_const(where_sv);
12658 msg = sv_2mortal(newSVpv(s, 0));
12659 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12660 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12662 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12664 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12665 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12666 Perl_sv_catpvf(aTHX_ msg,
12667 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12668 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12671 if (PL_in_eval & EVAL_WARNONLY) {
12672 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12676 if (PL_error_count >= 10) {
12677 if (PL_in_eval && SvCUR(ERRSV))
12678 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12679 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12681 Perl_croak(aTHX_ "%s has too many errors.\n",
12682 OutCopFILE(PL_curcop));
12685 PL_in_my_stash = NULL;
12689 #pragma segment Main
12693 S_swallow_bom(pTHX_ U8 *s)
12696 const STRLEN slen = SvCUR(PL_linestr);
12698 PERL_ARGS_ASSERT_SWALLOW_BOM;
12702 if (s[1] == 0xFE) {
12703 /* UTF-16 little-endian? (or UTF32-LE?) */
12704 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12705 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12706 #ifndef PERL_NO_UTF16_FILTER
12707 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12709 if (PL_bufend > (char*)s) {
12710 s = add_utf16_textfilter(s, TRUE);
12713 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12718 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12719 #ifndef PERL_NO_UTF16_FILTER
12720 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12722 if (PL_bufend > (char *)s) {
12723 s = add_utf16_textfilter(s, FALSE);
12726 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12731 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12732 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12733 s += 3; /* UTF-8 */
12739 if (s[2] == 0xFE && s[3] == 0xFF) {
12740 /* UTF-32 big-endian */
12741 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12744 else if (s[2] == 0 && s[3] != 0) {
12747 * are a good indicator of UTF-16BE. */
12748 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12749 s = add_utf16_textfilter(s, FALSE);
12754 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12755 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12756 s += 4; /* UTF-8 */
12762 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12765 * are a good indicator of UTF-16LE. */
12766 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12767 s = add_utf16_textfilter(s, TRUE);
12774 #ifndef PERL_NO_UTF16_FILTER
12776 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12779 const STRLEN old = SvCUR(sv);
12780 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12781 const bool reverse = IoLINES(FILTER_DATA(idx));
12783 /* As we're automatically added, at the lowest level, and hence only called
12784 from this file, we can be sure that we're not called in block mode. Hence
12785 don't bother writing code to deal with block mode. */
12787 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
12789 DEBUG_P(PerlIO_printf(Perl_debug_log,
12790 "utf16%s_textfilter(%p): %d %d (%d)\n",
12791 reverse ? "rev" : "",
12792 FPTR2DPTR(void *, S_utf16_textfilter),
12793 idx, maxlen, (int) count));
12797 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12798 Copy(SvPVX_const(sv), tmps, old, char);
12800 /* You would expect this to be utf16_to_utf8_reversed()
12801 It was, prior to 1de9afcdf18cf98bbdecaa782da93e907be6fe4e
12802 Effectively, right now, UTF-16LE is being read in off-by-one
12804 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12805 SvCUR(sv) - old, &newlen);
12807 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12808 SvCUR(sv) - old, &newlen);
12810 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12812 DEBUG_P({sv_dump(sv);});
12817 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
12822 IoLINES(filter_add(S_utf16_textfilter, NULL)) = reversed;
12823 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12825 utf16_to_utf8_reversed(s, news, PL_bufend - (char*)s - 1, &newlen);
12827 utf16_to_utf8(s, news, PL_bufend - (char*)s, &newlen);
12829 sv_setpvn(PL_linestr, (const char*)news, newlen);
12831 SvUTF8_on(PL_linestr);
12832 PL_bufend = SvPVX(PL_linestr) + newlen;
12833 return (U8*)SvPVX(PL_linestr);
12838 Returns a pointer to the next character after the parsed
12839 vstring, as well as updating the passed in sv.
12841 Function must be called like
12844 s = scan_vstring(s,e,sv);
12846 where s and e are the start and end of the string.
12847 The sv should already be large enough to store the vstring
12848 passed in, for performance reasons.
12853 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12856 const char *pos = s;
12857 const char *start = s;
12859 PERL_ARGS_ASSERT_SCAN_VSTRING;
12861 if (*pos == 'v') pos++; /* get past 'v' */
12862 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12864 if ( *pos != '.') {
12865 /* this may not be a v-string if followed by => */
12866 const char *next = pos;
12867 while (next < e && isSPACE(*next))
12869 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12870 /* return string not v-string */
12871 sv_setpvn(sv,(char *)s,pos-s);
12872 return (char *)pos;
12876 if (!isALPHA(*pos)) {
12877 U8 tmpbuf[UTF8_MAXBYTES+1];
12880 s++; /* get past 'v' */
12885 /* this is atoi() that tolerates underscores */
12888 const char *end = pos;
12890 while (--end >= s) {
12892 const UV orev = rev;
12893 rev += (*end - '0') * mult;
12896 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12897 "Integer overflow in decimal number");
12901 if (rev > 0x7FFFFFFF)
12902 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12904 /* Append native character for the rev point */
12905 tmpend = uvchr_to_utf8(tmpbuf, rev);
12906 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12907 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12909 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12915 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12919 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12927 * c-indentation-style: bsd
12928 * c-basic-offset: 4
12929 * indent-tabs-mode: t
12932 * ex: set ts=8 sts=4 sw=4 noet: