3 * Copyright (c) 1991-2002, Larry Wall
5 * You may distribute under the terms of either the GNU General Public
6 * License or the Artistic License, as specified in the README file.
11 * "It all comes from here, the stench and the peril." --Frodo
15 * This file is the lexer for Perl. It's closely linked to the
18 * The main routine is yylex(), which returns the next token.
22 #define PERL_IN_TOKE_C
25 #define yychar PL_yychar
26 #define yylval PL_yylval
28 static char ident_too_long[] = "Identifier too long";
30 static void restore_rsfp(pTHX_ void *f);
31 #ifndef PERL_NO_UTF16_FILTER
32 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
33 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
36 #define XFAKEBRACK 128
39 #ifdef USE_UTF8_SCRIPTS
40 # define UTF (!IN_BYTES)
42 # ifdef EBCDIC /* For now 'use utf8' does not affect tokenizer on EBCDIC */
43 # define UTF (PL_linestr && DO_UTF8(PL_linestr))
45 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
49 /* In variables named $^X, these are the legal values for X.
50 * 1999-02-27 mjd-perl-patch@plover.com */
51 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
53 /* On MacOS, respect nonbreaking spaces */
54 #ifdef MACOS_TRADITIONAL
55 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
57 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
60 /* LEX_* are values for PL_lex_state, the state of the lexer.
61 * They are arranged oddly so that the guard on the switch statement
62 * can get by with a single comparison (if the compiler is smart enough).
65 /* #define LEX_NOTPARSING 11 is done in perl.h. */
68 #define LEX_INTERPNORMAL 9
69 #define LEX_INTERPCASEMOD 8
70 #define LEX_INTERPPUSH 7
71 #define LEX_INTERPSTART 6
72 #define LEX_INTERPEND 5
73 #define LEX_INTERPENDMAYBE 4
74 #define LEX_INTERPCONCAT 3
75 #define LEX_INTERPCONST 2
76 #define LEX_FORMLINE 1
77 #define LEX_KNOWNEXT 0
85 # define YYMAXLEVEL 100
87 YYSTYPE* yylval_pointer[YYMAXLEVEL];
88 int* yychar_pointer[YYMAXLEVEL];
92 # define yylval (*yylval_pointer[yyactlevel])
93 # define yychar (*yychar_pointer[yyactlevel])
94 # define PERL_YYLEX_PARAM yylval_pointer[yyactlevel],yychar_pointer[yyactlevel]
96 # define yylex() Perl_yylex_r(aTHX_ yylval_pointer[yyactlevel],yychar_pointer[yyactlevel])
101 /* CLINE is a macro that ensures PL_copline has a sane value */
106 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
109 * Convenience functions to return different tokens and prime the
110 * lexer for the next token. They all take an argument.
112 * TOKEN : generic token (used for '(', DOLSHARP, etc)
113 * OPERATOR : generic operator
114 * AOPERATOR : assignment operator
115 * PREBLOCK : beginning the block after an if, while, foreach, ...
116 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
117 * PREREF : *EXPR where EXPR is not a simple identifier
118 * TERM : expression term
119 * LOOPX : loop exiting command (goto, last, dump, etc)
120 * FTST : file test operator
121 * FUN0 : zero-argument function
122 * FUN1 : not used, except for not, which isn't a UNIOP
123 * BOop : bitwise or or xor
125 * SHop : shift operator
126 * PWop : power operator
127 * PMop : pattern-matching operator
128 * Aop : addition-level operator
129 * Mop : multiplication-level operator
130 * Eop : equality-testing operator
131 * Rop : relational operator <= != gt
133 * Also see LOP and lop() below.
136 /* Note that REPORT() and REPORT2() will be expressions that supply
137 * their own trailing comma, not suitable for statements as such. */
138 #ifdef DEBUGGING /* Serve -DT. */
139 # define REPORT(x,retval) tokereport(x,s,(int)retval),
140 # define REPORT2(x,retval) tokereport(x,s, yylval.ival),
142 # define REPORT(x,retval)
143 # define REPORT2(x,retval)
146 #define TOKEN(retval) return (REPORT2("token",retval) PL_bufptr = s,(int)retval)
147 #define OPERATOR(retval) return (REPORT2("operator",retval) PL_expect = XTERM, PL_bufptr = s,(int)retval)
148 #define AOPERATOR(retval) return ao((REPORT2("aop",retval) PL_expect = XTERM, PL_bufptr = s,(int)retval))
149 #define PREBLOCK(retval) return (REPORT2("preblock",retval) PL_expect = XBLOCK,PL_bufptr = s,(int)retval)
150 #define PRETERMBLOCK(retval) return (REPORT2("pretermblock",retval) PL_expect = XTERMBLOCK,PL_bufptr = s,(int)retval)
151 #define PREREF(retval) return (REPORT2("preref",retval) PL_expect = XREF,PL_bufptr = s,(int)retval)
152 #define TERM(retval) return (CLINE, REPORT2("term",retval) PL_expect = XOPERATOR, PL_bufptr = s,(int)retval)
153 #define LOOPX(f) return(yylval.ival=f, REPORT("loopx",f) PL_expect = XTERM,PL_bufptr = s,(int)LOOPEX)
154 #define FTST(f) return(yylval.ival=f, REPORT("ftst",f) PL_expect = XTERM,PL_bufptr = s,(int)UNIOP)
155 #define FUN0(f) return(yylval.ival = f, REPORT("fun0",f) PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC0)
156 #define FUN1(f) return(yylval.ival = f, REPORT("fun1",f) PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC1)
157 #define BOop(f) return ao((yylval.ival=f, REPORT("bitorop",f) PL_expect = XTERM,PL_bufptr = s,(int)BITOROP))
158 #define BAop(f) return ao((yylval.ival=f, REPORT("bitandop",f) PL_expect = XTERM,PL_bufptr = s,(int)BITANDOP))
159 #define SHop(f) return ao((yylval.ival=f, REPORT("shiftop",f) PL_expect = XTERM,PL_bufptr = s,(int)SHIFTOP))
160 #define PWop(f) return ao((yylval.ival=f, REPORT("powop",f) PL_expect = XTERM,PL_bufptr = s,(int)POWOP))
161 #define PMop(f) return(yylval.ival=f, REPORT("matchop",f) PL_expect = XTERM,PL_bufptr = s,(int)MATCHOP)
162 #define Aop(f) return ao((yylval.ival=f, REPORT("add",f) PL_expect = XTERM,PL_bufptr = s,(int)ADDOP))
163 #define Mop(f) return ao((yylval.ival=f, REPORT("mul",f) PL_expect = XTERM,PL_bufptr = s,(int)MULOP))
164 #define Eop(f) return(yylval.ival=f, REPORT("eq",f) PL_expect = XTERM,PL_bufptr = s,(int)EQOP)
165 #define Rop(f) return(yylval.ival=f, REPORT("rel",f) PL_expect = XTERM,PL_bufptr = s,(int)RELOP)
167 /* This bit of chicanery makes a unary function followed by
168 * a parenthesis into a function with one argument, highest precedence.
170 #define UNI(f) return(yylval.ival = f, \
174 PL_last_uni = PL_oldbufptr, \
175 PL_last_lop_op = f, \
176 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
178 #define UNIBRACK(f) return(yylval.ival = f, \
181 PL_last_uni = PL_oldbufptr, \
182 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
184 /* grandfather return to old style */
185 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
190 S_tokereport(pTHX_ char *thing, char* s, I32 rv)
193 SV* report = newSVpv(thing, 0);
194 Perl_sv_catpvf(aTHX_ report, ":line %d:%"IVdf":", CopLINE(PL_curcop),
197 if (s - PL_bufptr > 0)
198 sv_catpvn(report, PL_bufptr, s - PL_bufptr);
200 if (PL_oldbufptr && *PL_oldbufptr)
201 sv_catpv(report, PL_tokenbuf);
203 PerlIO_printf(Perl_debug_log, "### %s\n", SvPV_nolen(report));
212 * This subroutine detects &&= and ||= and turns an ANDAND or OROR
213 * into an OP_ANDASSIGN or OP_ORASSIGN
217 S_ao(pTHX_ int toketype)
219 if (*PL_bufptr == '=') {
221 if (toketype == ANDAND)
222 yylval.ival = OP_ANDASSIGN;
223 else if (toketype == OROR)
224 yylval.ival = OP_ORASSIGN;
232 * When Perl expects an operator and finds something else, no_op
233 * prints the warning. It always prints "<something> found where
234 * operator expected. It prints "Missing semicolon on previous line?"
235 * if the surprise occurs at the start of the line. "do you need to
236 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
237 * where the compiler doesn't know if foo is a method call or a function.
238 * It prints "Missing operator before end of line" if there's nothing
239 * after the missing operator, or "... before <...>" if there is something
240 * after the missing operator.
244 S_no_op(pTHX_ char *what, char *s)
246 char *oldbp = PL_bufptr;
247 bool is_first = (PL_oldbufptr == PL_linestart);
253 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
255 Perl_warn(aTHX_ "\t(Missing semicolon on previous line?)\n");
256 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
258 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
259 if (t < PL_bufptr && isSPACE(*t))
260 Perl_warn(aTHX_ "\t(Do you need to predeclare %.*s?)\n",
261 t - PL_oldoldbufptr, PL_oldoldbufptr);
265 Perl_warn(aTHX_ "\t(Missing operator before %.*s?)\n", s - oldbp, oldbp);
272 * Complain about missing quote/regexp/heredoc terminator.
273 * If it's called with (char *)NULL then it cauterizes the line buffer.
274 * If we're in a delimited string and the delimiter is a control
275 * character, it's reformatted into a two-char sequence like ^C.
280 S_missingterm(pTHX_ char *s)
285 char *nl = strrchr(s,'\n');
291 iscntrl(PL_multi_close)
293 PL_multi_close < 32 || PL_multi_close == 127
297 tmpbuf[1] = toCTRL(PL_multi_close);
303 *tmpbuf = PL_multi_close;
307 q = strchr(s,'"') ? '\'' : '"';
308 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
316 Perl_deprecate(pTHX_ char *s)
318 if (ckWARN(WARN_DEPRECATED))
319 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
323 Perl_deprecate_old(pTHX_ char *s)
325 /* This function should NOT be called for any new deprecated warnings */
326 /* Use Perl_deprecate instead */
328 /* It is here to maintain backward compatibility with the pre-5.8 */
329 /* warnings category hierarchy. The "deprecated" category used to */
330 /* live under the "syntax" category. It is now a top-level category */
331 /* in its own right. */
333 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
334 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
335 "Use of %s is deprecated", s);
340 * Deprecate a comma-less variable list.
346 deprecate_old("comma-less variable list");
350 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
351 * utf16-to-utf8-reversed.
354 #ifdef PERL_CR_FILTER
358 register char *s = SvPVX(sv);
359 register char *e = s + SvCUR(sv);
360 /* outer loop optimized to do nothing if there are no CR-LFs */
362 if (*s++ == '\r' && *s == '\n') {
363 /* hit a CR-LF, need to copy the rest */
364 register char *d = s - 1;
367 if (*s == '\r' && s[1] == '\n')
378 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
380 I32 count = FILTER_READ(idx+1, sv, maxlen);
381 if (count > 0 && !maxlen)
389 * Initialize variables. Uses the Perl save_stack to save its state (for
390 * recursive calls to the parser).
394 Perl_lex_start(pTHX_ SV *line)
399 SAVEI32(PL_lex_dojoin);
400 SAVEI32(PL_lex_brackets);
401 SAVEI32(PL_lex_casemods);
402 SAVEI32(PL_lex_starts);
403 SAVEI32(PL_lex_state);
404 SAVEVPTR(PL_lex_inpat);
405 SAVEI32(PL_lex_inwhat);
406 if (PL_lex_state == LEX_KNOWNEXT) {
407 I32 toke = PL_nexttoke;
408 while (--toke >= 0) {
409 SAVEI32(PL_nexttype[toke]);
410 SAVEVPTR(PL_nextval[toke]);
412 SAVEI32(PL_nexttoke);
414 SAVECOPLINE(PL_curcop);
417 SAVEPPTR(PL_oldbufptr);
418 SAVEPPTR(PL_oldoldbufptr);
419 SAVEPPTR(PL_last_lop);
420 SAVEPPTR(PL_last_uni);
421 SAVEPPTR(PL_linestart);
422 SAVESPTR(PL_linestr);
423 SAVEPPTR(PL_lex_brackstack);
424 SAVEPPTR(PL_lex_casestack);
425 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
426 SAVESPTR(PL_lex_stuff);
427 SAVEI32(PL_lex_defer);
428 SAVEI32(PL_sublex_info.sub_inwhat);
429 SAVESPTR(PL_lex_repl);
431 SAVEINT(PL_lex_expect);
433 PL_lex_state = LEX_NORMAL;
437 New(899, PL_lex_brackstack, 120, char);
438 New(899, PL_lex_casestack, 12, char);
439 SAVEFREEPV(PL_lex_brackstack);
440 SAVEFREEPV(PL_lex_casestack);
442 *PL_lex_casestack = '\0';
445 PL_lex_stuff = Nullsv;
446 PL_lex_repl = Nullsv;
450 PL_sublex_info.sub_inwhat = 0;
452 if (SvREADONLY(PL_linestr))
453 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
454 s = SvPV(PL_linestr, len);
455 if (len && s[len-1] != ';') {
456 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
457 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
458 sv_catpvn(PL_linestr, "\n;", 2);
460 SvTEMP_off(PL_linestr);
461 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
462 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
463 PL_last_lop = PL_last_uni = Nullch;
469 * Finalizer for lexing operations. Must be called when the parser is
470 * done with the lexer.
476 PL_doextract = FALSE;
481 * This subroutine has nothing to do with tilting, whether at windmills
482 * or pinball tables. Its name is short for "increment line". It
483 * increments the current line number in CopLINE(PL_curcop) and checks
484 * to see whether the line starts with a comment of the form
485 * # line 500 "foo.pm"
486 * If so, it sets the current line number and file to the values in the comment.
490 S_incline(pTHX_ char *s)
497 CopLINE_inc(PL_curcop);
500 while (SPACE_OR_TAB(*s)) s++;
501 if (strnEQ(s, "line", 4))
505 if (SPACE_OR_TAB(*s))
509 while (SPACE_OR_TAB(*s)) s++;
515 while (SPACE_OR_TAB(*s))
517 if (*s == '"' && (t = strchr(s+1, '"'))) {
522 for (t = s; !isSPACE(*t); t++) ;
525 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
527 if (*e != '\n' && *e != '\0')
528 return; /* false alarm */
533 CopFILE_free(PL_curcop);
534 CopFILE_set(PL_curcop, s);
537 CopLINE_set(PL_curcop, atoi(n)-1);
542 * Called to gobble the appropriate amount and type of whitespace.
543 * Skips comments as well.
547 S_skipspace(pTHX_ register char *s)
549 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
550 while (s < PL_bufend && SPACE_OR_TAB(*s))
556 SSize_t oldprevlen, oldoldprevlen;
557 SSize_t oldloplen = 0, oldunilen = 0;
558 while (s < PL_bufend && isSPACE(*s)) {
559 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
564 if (s < PL_bufend && *s == '#') {
565 while (s < PL_bufend && *s != '\n')
569 if (PL_in_eval && !PL_rsfp) {
576 /* only continue to recharge the buffer if we're at the end
577 * of the buffer, we're not reading from a source filter, and
578 * we're in normal lexing mode
580 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
581 PL_lex_state == LEX_FORMLINE)
584 /* try to recharge the buffer */
585 if ((s = filter_gets(PL_linestr, PL_rsfp,
586 (prevlen = SvCUR(PL_linestr)))) == Nullch)
588 /* end of file. Add on the -p or -n magic */
589 if (PL_minus_n || PL_minus_p) {
590 sv_setpv(PL_linestr,PL_minus_p ?
591 ";}continue{print or die qq(-p destination: $!\\n)" :
593 sv_catpv(PL_linestr,";}");
594 PL_minus_n = PL_minus_p = 0;
597 sv_setpv(PL_linestr,";");
599 /* reset variables for next time we lex */
600 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
602 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
603 PL_last_lop = PL_last_uni = Nullch;
605 /* Close the filehandle. Could be from -P preprocessor,
606 * STDIN, or a regular file. If we were reading code from
607 * STDIN (because the commandline held no -e or filename)
608 * then we don't close it, we reset it so the code can
609 * read from STDIN too.
612 if (PL_preprocess && !PL_in_eval)
613 (void)PerlProc_pclose(PL_rsfp);
614 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
615 PerlIO_clearerr(PL_rsfp);
617 (void)PerlIO_close(PL_rsfp);
622 /* not at end of file, so we only read another line */
623 /* make corresponding updates to old pointers, for yyerror() */
624 oldprevlen = PL_oldbufptr - PL_bufend;
625 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
627 oldunilen = PL_last_uni - PL_bufend;
629 oldloplen = PL_last_lop - PL_bufend;
630 PL_linestart = PL_bufptr = s + prevlen;
631 PL_bufend = s + SvCUR(PL_linestr);
633 PL_oldbufptr = s + oldprevlen;
634 PL_oldoldbufptr = s + oldoldprevlen;
636 PL_last_uni = s + oldunilen;
638 PL_last_lop = s + oldloplen;
641 /* debugger active and we're not compiling the debugger code,
642 * so store the line into the debugger's array of lines
644 if (PERLDB_LINE && PL_curstash != PL_debstash) {
645 SV *sv = NEWSV(85,0);
647 sv_upgrade(sv, SVt_PVMG);
648 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
651 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
658 * Check the unary operators to ensure there's no ambiguity in how they're
659 * used. An ambiguous piece of code would be:
661 * This doesn't mean rand() + 5. Because rand() is a unary operator,
662 * the +5 is its argument.
671 if (PL_oldoldbufptr != PL_last_uni)
673 while (isSPACE(*PL_last_uni))
675 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
676 if ((t = strchr(s, '(')) && t < PL_bufptr)
678 if (ckWARN_d(WARN_AMBIGUOUS)){
681 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
682 "Warning: Use of \"%s\" without parens is ambiguous",
689 * LOP : macro to build a list operator. Its behaviour has been replaced
690 * with a subroutine, S_lop() for which LOP is just another name.
693 #define LOP(f,x) return lop(f,x,s)
697 * Build a list operator (or something that might be one). The rules:
698 * - if we have a next token, then it's a list operator [why?]
699 * - if the next thing is an opening paren, then it's a function
700 * - else it's a list operator
704 S_lop(pTHX_ I32 f, int x, char *s)
711 PL_last_lop = PL_oldbufptr;
726 * When the lexer realizes it knows the next token (for instance,
727 * it is reordering tokens for the parser) then it can call S_force_next
728 * to know what token to return the next time the lexer is called. Caller
729 * will need to set PL_nextval[], and possibly PL_expect to ensure the lexer
730 * handles the token correctly.
734 S_force_next(pTHX_ I32 type)
736 PL_nexttype[PL_nexttoke] = type;
738 if (PL_lex_state != LEX_KNOWNEXT) {
739 PL_lex_defer = PL_lex_state;
740 PL_lex_expect = PL_expect;
741 PL_lex_state = LEX_KNOWNEXT;
747 * When the lexer knows the next thing is a word (for instance, it has
748 * just seen -> and it knows that the next char is a word char, then
749 * it calls S_force_word to stick the next word into the PL_next lookahead.
752 * char *start : buffer position (must be within PL_linestr)
753 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
754 * int check_keyword : if true, Perl checks to make sure the word isn't
755 * a keyword (do this if the word is a label, e.g. goto FOO)
756 * int allow_pack : if true, : characters will also be allowed (require,
758 * int allow_initial_tick : used by the "sub" lexer only.
762 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
767 start = skipspace(start);
769 if (isIDFIRST_lazy_if(s,UTF) ||
770 (allow_pack && *s == ':') ||
771 (allow_initial_tick && *s == '\'') )
773 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
774 if (check_keyword && keyword(PL_tokenbuf, len))
776 if (token == METHOD) {
781 PL_expect = XOPERATOR;
784 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST,0, newSVpv(PL_tokenbuf,0));
785 PL_nextval[PL_nexttoke].opval->op_private |= OPpCONST_BARE;
793 * Called when the lexer wants $foo *foo &foo etc, but the program
794 * text only contains the "foo" portion. The first argument is a pointer
795 * to the "foo", and the second argument is the type symbol to prefix.
796 * Forces the next token to be a "WORD".
797 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
801 S_force_ident(pTHX_ register char *s, int kind)
804 OP* o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
805 PL_nextval[PL_nexttoke].opval = o;
808 o->op_private = OPpCONST_ENTERED;
809 /* XXX see note in pp_entereval() for why we forgo typo
810 warnings if the symbol must be introduced in an eval.
812 gv_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
813 kind == '$' ? SVt_PV :
814 kind == '@' ? SVt_PVAV :
815 kind == '%' ? SVt_PVHV :
823 Perl_str_to_version(pTHX_ SV *sv)
828 char *start = SvPVx(sv,len);
829 bool utf = SvUTF8(sv) ? TRUE : FALSE;
830 char *end = start + len;
831 while (start < end) {
835 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
840 retval += ((NV)n)/nshift;
849 * Forces the next token to be a version number.
850 * If the next token appears to be an invalid version number, (e.g. "v2b"),
851 * and if "guessing" is TRUE, then no new token is created (and the caller
852 * must use an alternative parsing method).
856 S_force_version(pTHX_ char *s, int guessing)
858 OP *version = Nullop;
867 while (isDIGIT(*d) || *d == '_' || *d == '.')
869 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
871 s = scan_num(s, &yylval);
872 version = yylval.opval;
873 ver = cSVOPx(version)->op_sv;
874 if (SvPOK(ver) && !SvNIOK(ver)) {
875 (void)SvUPGRADE(ver, SVt_PVNV);
876 SvNVX(ver) = str_to_version(ver);
877 SvNOK_on(ver); /* hint that it is a version */
884 /* NOTE: The parser sees the package name and the VERSION swapped */
885 PL_nextval[PL_nexttoke].opval = version;
893 * Tokenize a quoted string passed in as an SV. It finds the next
894 * chunk, up to end of string or a backslash. It may make a new
895 * SV containing that chunk (if HINT_NEW_STRING is on). It also
900 S_tokeq(pTHX_ SV *sv)
911 s = SvPV_force(sv, len);
912 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
915 while (s < send && *s != '\\')
920 if ( PL_hints & HINT_NEW_STRING ) {
921 pv = sv_2mortal(newSVpvn(SvPVX(pv), len));
927 if (s + 1 < send && (s[1] == '\\'))
928 s++; /* all that, just for this */
933 SvCUR_set(sv, d - SvPVX(sv));
935 if ( PL_hints & HINT_NEW_STRING )
936 return new_constant(NULL, 0, "q", sv, pv, "q");
941 * Now come three functions related to double-quote context,
942 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
943 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
944 * interact with PL_lex_state, and create fake ( ... ) argument lists
945 * to handle functions and concatenation.
946 * They assume that whoever calls them will be setting up a fake
947 * join call, because each subthing puts a ',' after it. This lets
950 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
952 * (I'm not sure whether the spurious commas at the end of lcfirst's
953 * arguments and join's arguments are created or not).
958 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
960 * Pattern matching will set PL_lex_op to the pattern-matching op to
961 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
963 * OP_CONST and OP_READLINE are easy--just make the new op and return.
965 * Everything else becomes a FUNC.
967 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
968 * had an OP_CONST or OP_READLINE). This just sets us up for a
969 * call to S_sublex_push().
975 register I32 op_type = yylval.ival;
977 if (op_type == OP_NULL) {
978 yylval.opval = PL_lex_op;
982 if (op_type == OP_CONST || op_type == OP_READLINE) {
983 SV *sv = tokeq(PL_lex_stuff);
985 if (SvTYPE(sv) == SVt_PVIV) {
986 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
992 nsv = newSVpvn(p, len);
998 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
999 PL_lex_stuff = Nullsv;
1003 PL_sublex_info.super_state = PL_lex_state;
1004 PL_sublex_info.sub_inwhat = op_type;
1005 PL_sublex_info.sub_op = PL_lex_op;
1006 PL_lex_state = LEX_INTERPPUSH;
1010 yylval.opval = PL_lex_op;
1020 * Create a new scope to save the lexing state. The scope will be
1021 * ended in S_sublex_done. Returns a '(', starting the function arguments
1022 * to the uc, lc, etc. found before.
1023 * Sets PL_lex_state to LEX_INTERPCONCAT.
1031 PL_lex_state = PL_sublex_info.super_state;
1032 SAVEI32(PL_lex_dojoin);
1033 SAVEI32(PL_lex_brackets);
1034 SAVEI32(PL_lex_casemods);
1035 SAVEI32(PL_lex_starts);
1036 SAVEI32(PL_lex_state);
1037 SAVEVPTR(PL_lex_inpat);
1038 SAVEI32(PL_lex_inwhat);
1039 SAVECOPLINE(PL_curcop);
1040 SAVEPPTR(PL_bufptr);
1041 SAVEPPTR(PL_bufend);
1042 SAVEPPTR(PL_oldbufptr);
1043 SAVEPPTR(PL_oldoldbufptr);
1044 SAVEPPTR(PL_last_lop);
1045 SAVEPPTR(PL_last_uni);
1046 SAVEPPTR(PL_linestart);
1047 SAVESPTR(PL_linestr);
1048 SAVEPPTR(PL_lex_brackstack);
1049 SAVEPPTR(PL_lex_casestack);
1051 PL_linestr = PL_lex_stuff;
1052 PL_lex_stuff = Nullsv;
1054 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1055 = SvPVX(PL_linestr);
1056 PL_bufend += SvCUR(PL_linestr);
1057 PL_last_lop = PL_last_uni = Nullch;
1058 SAVEFREESV(PL_linestr);
1060 PL_lex_dojoin = FALSE;
1061 PL_lex_brackets = 0;
1062 New(899, PL_lex_brackstack, 120, char);
1063 New(899, PL_lex_casestack, 12, char);
1064 SAVEFREEPV(PL_lex_brackstack);
1065 SAVEFREEPV(PL_lex_casestack);
1066 PL_lex_casemods = 0;
1067 *PL_lex_casestack = '\0';
1069 PL_lex_state = LEX_INTERPCONCAT;
1070 CopLINE_set(PL_curcop, PL_multi_start);
1072 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1073 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1074 PL_lex_inpat = PL_sublex_info.sub_op;
1076 PL_lex_inpat = Nullop;
1083 * Restores lexer state after a S_sublex_push.
1089 if (!PL_lex_starts++) {
1090 SV *sv = newSVpvn("",0);
1091 if (SvUTF8(PL_linestr))
1093 PL_expect = XOPERATOR;
1094 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1098 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1099 PL_lex_state = LEX_INTERPCASEMOD;
1103 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1104 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1105 PL_linestr = PL_lex_repl;
1107 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1108 PL_bufend += SvCUR(PL_linestr);
1109 PL_last_lop = PL_last_uni = Nullch;
1110 SAVEFREESV(PL_linestr);
1111 PL_lex_dojoin = FALSE;
1112 PL_lex_brackets = 0;
1113 PL_lex_casemods = 0;
1114 *PL_lex_casestack = '\0';
1116 if (SvEVALED(PL_lex_repl)) {
1117 PL_lex_state = LEX_INTERPNORMAL;
1119 /* we don't clear PL_lex_repl here, so that we can check later
1120 whether this is an evalled subst; that means we rely on the
1121 logic to ensure sublex_done() is called again only via the
1122 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1125 PL_lex_state = LEX_INTERPCONCAT;
1126 PL_lex_repl = Nullsv;
1132 PL_bufend = SvPVX(PL_linestr);
1133 PL_bufend += SvCUR(PL_linestr);
1134 PL_expect = XOPERATOR;
1135 PL_sublex_info.sub_inwhat = 0;
1143 Extracts a pattern, double-quoted string, or transliteration. This
1146 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1147 processing a pattern (PL_lex_inpat is true), a transliteration
1148 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1150 Returns a pointer to the character scanned up to. Iff this is
1151 advanced from the start pointer supplied (ie if anything was
1152 successfully parsed), will leave an OP for the substring scanned
1153 in yylval. Caller must intuit reason for not parsing further
1154 by looking at the next characters herself.
1158 double-quoted style: \r and \n
1159 regexp special ones: \D \s
1161 backrefs: \1 (deprecated in substitution replacements)
1162 case and quoting: \U \Q \E
1163 stops on @ and $, but not for $ as tail anchor
1165 In transliterations:
1166 characters are VERY literal, except for - not at the start or end
1167 of the string, which indicates a range. scan_const expands the
1168 range to the full set of intermediate characters.
1170 In double-quoted strings:
1172 double-quoted style: \r and \n
1174 backrefs: \1 (deprecated)
1175 case and quoting: \U \Q \E
1178 scan_const does *not* construct ops to handle interpolated strings.
1179 It stops processing as soon as it finds an embedded $ or @ variable
1180 and leaves it to the caller to work out what's going on.
1182 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @:foo.
1184 $ in pattern could be $foo or could be tail anchor. Assumption:
1185 it's a tail anchor if $ is the last thing in the string, or if it's
1186 followed by one of ")| \n\t"
1188 \1 (backreferences) are turned into $1
1190 The structure of the code is
1191 while (there's a character to process) {
1192 handle transliteration ranges
1193 skip regexp comments
1194 skip # initiated comments in //x patterns
1195 check for embedded @foo
1196 check for embedded scalars
1198 leave intact backslashes from leave (below)
1199 deprecate \1 in strings and sub replacements
1200 handle string-changing backslashes \l \U \Q \E, etc.
1201 switch (what was escaped) {
1202 handle - in a transliteration (becomes a literal -)
1203 handle \132 octal characters
1204 handle 0x15 hex characters
1205 handle \cV (control V)
1206 handle printf backslashes (\f, \r, \n, etc)
1208 } (end if backslash)
1209 } (end while character to read)
1214 S_scan_const(pTHX_ char *start)
1216 register char *send = PL_bufend; /* end of the constant */
1217 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1218 register char *s = start; /* start of the constant */
1219 register char *d = SvPVX(sv); /* destination for copies */
1220 bool dorange = FALSE; /* are we in a translit range? */
1221 bool didrange = FALSE; /* did we just finish a range? */
1222 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1223 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1226 const char *leaveit = /* set of acceptably-backslashed characters */
1228 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1231 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1232 /* If we are doing a trans and we know we want UTF8 set expectation */
1233 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1234 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1238 while (s < send || dorange) {
1239 /* get transliterations out of the way (they're most literal) */
1240 if (PL_lex_inwhat == OP_TRANS) {
1241 /* expand a range A-Z to the full set of characters. AIE! */
1243 I32 i; /* current expanded character */
1244 I32 min; /* first character in range */
1245 I32 max; /* last character in range */
1248 char *c = (char*)utf8_hop((U8*)d, -1);
1252 *c = (char)UTF_TO_NATIVE(0xff);
1253 /* mark the range as done, and continue */
1259 i = d - SvPVX(sv); /* remember current offset */
1260 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1261 d = SvPVX(sv) + i; /* refresh d after realloc */
1262 d -= 2; /* eat the first char and the - */
1264 min = (U8)*d; /* first char in range */
1265 max = (U8)d[1]; /* last char in range */
1269 "Invalid [] range \"%c-%c\" in transliteration operator",
1270 (char)min, (char)max);
1274 if ((isLOWER(min) && isLOWER(max)) ||
1275 (isUPPER(min) && isUPPER(max))) {
1277 for (i = min; i <= max; i++)
1279 *d++ = NATIVE_TO_NEED(has_utf8,i);
1281 for (i = min; i <= max; i++)
1283 *d++ = NATIVE_TO_NEED(has_utf8,i);
1288 for (i = min; i <= max; i++)
1291 /* mark the range as done, and continue */
1297 /* range begins (ignore - as first or last char) */
1298 else if (*s == '-' && s+1 < send && s != start) {
1300 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1303 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1315 /* if we get here, we're not doing a transliteration */
1317 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1318 except for the last char, which will be done separately. */
1319 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1321 while (s < send && *s != ')')
1322 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1324 else if (s[2] == '{' /* This should match regcomp.c */
1325 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1328 char *regparse = s + (s[2] == '{' ? 3 : 4);
1331 while (count && (c = *regparse)) {
1332 if (c == '\\' && regparse[1])
1340 if (*regparse != ')') {
1341 regparse--; /* Leave one char for continuation. */
1342 yyerror("Sequence (?{...}) not terminated or not {}-balanced");
1344 while (s < regparse)
1345 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1349 /* likewise skip #-initiated comments in //x patterns */
1350 else if (*s == '#' && PL_lex_inpat &&
1351 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1352 while (s+1 < send && *s != '\n')
1353 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1356 /* check for embedded arrays
1357 (@foo, @:foo, @'foo, @{foo}, @$foo, @+, @-)
1359 else if (*s == '@' && s[1]
1360 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1363 /* check for embedded scalars. only stop if we're sure it's a
1366 else if (*s == '$') {
1367 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1369 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1370 break; /* in regexp, $ might be tail anchor */
1373 /* End of else if chain - OP_TRANS rejoin rest */
1376 if (*s == '\\' && s+1 < send) {
1379 /* some backslashes we leave behind */
1380 if (*leaveit && *s && strchr(leaveit, *s)) {
1381 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1382 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1386 /* deprecate \1 in strings and substitution replacements */
1387 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1388 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1390 if (ckWARN(WARN_SYNTAX))
1391 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1396 /* string-change backslash escapes */
1397 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1402 /* if we get here, it's either a quoted -, or a digit */
1405 /* quoted - in transliterations */
1407 if (PL_lex_inwhat == OP_TRANS) {
1414 if (ckWARN(WARN_MISC) &&
1417 Perl_warner(aTHX_ packWARN(WARN_MISC),
1418 "Unrecognized escape \\%c passed through",
1420 /* default action is to copy the quoted character */
1421 goto default_action;
1424 /* \132 indicates an octal constant */
1425 case '0': case '1': case '2': case '3':
1426 case '4': case '5': case '6': case '7':
1430 uv = grok_oct(s, &len, &flags, NULL);
1433 goto NUM_ESCAPE_INSERT;
1435 /* \x24 indicates a hex constant */
1439 char* e = strchr(s, '}');
1440 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
1441 PERL_SCAN_DISALLOW_PREFIX;
1446 yyerror("Missing right brace on \\x{}");
1450 uv = grok_hex(s, &len, &flags, NULL);
1456 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
1457 uv = grok_hex(s, &len, &flags, NULL);
1463 /* Insert oct or hex escaped character.
1464 * There will always enough room in sv since such
1465 * escapes will be longer than any UTF-8 sequence
1466 * they can end up as. */
1468 /* We need to map to chars to ASCII before doing the tests
1471 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
1472 if (!has_utf8 && uv > 255) {
1473 /* Might need to recode whatever we have
1474 * accumulated so far if it contains any
1477 * (Can't we keep track of that and avoid
1478 * this rescan? --jhi)
1482 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
1483 if (!NATIVE_IS_INVARIANT(*c)) {
1488 STRLEN offset = d - SvPVX(sv);
1490 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
1494 while (src >= (U8 *)SvPVX(sv)) {
1495 if (!NATIVE_IS_INVARIANT(*src)) {
1496 U8 ch = NATIVE_TO_ASCII(*src);
1497 *dst-- = UTF8_EIGHT_BIT_LO(ch);
1498 *dst-- = UTF8_EIGHT_BIT_HI(ch);
1508 if (has_utf8 || uv > 255) {
1509 d = (char*)uvchr_to_utf8((U8*)d, uv);
1511 if (PL_lex_inwhat == OP_TRANS &&
1512 PL_sublex_info.sub_op) {
1513 PL_sublex_info.sub_op->op_private |=
1514 (PL_lex_repl ? OPpTRANS_FROM_UTF
1527 /* \N{LATIN SMALL LETTER A} is a named character */
1531 char* e = strchr(s, '}');
1537 yyerror("Missing right brace on \\N{}");
1541 res = newSVpvn(s + 1, e - s - 1);
1542 res = new_constant( Nullch, 0, "charnames",
1543 res, Nullsv, "\\N{...}" );
1545 sv_utf8_upgrade(res);
1546 str = SvPV(res,len);
1547 #ifdef EBCDIC_NEVER_MIND
1548 /* charnames uses pack U and that has been
1549 * recently changed to do the below uni->native
1550 * mapping, so this would be redundant (and wrong,
1551 * the code point would be doubly converted).
1552 * But leave this in just in case the pack U change
1553 * gets revoked, but the semantics is still
1554 * desireable for charnames. --jhi */
1556 UV uv = utf8_to_uvchr((U8*)str, 0);
1559 U8 tmpbuf[UTF8_MAXLEN+1], *d;
1561 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
1562 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
1563 str = SvPV(res, len);
1567 if (!has_utf8 && SvUTF8(res)) {
1568 char *ostart = SvPVX(sv);
1569 SvCUR_set(sv, d - ostart);
1572 sv_utf8_upgrade(sv);
1573 /* this just broke our allocation above... */
1574 SvGROW(sv, send - start);
1575 d = SvPVX(sv) + SvCUR(sv);
1578 if (len > e - s + 4) { /* I _guess_ 4 is \N{} --jhi */
1579 char *odest = SvPVX(sv);
1581 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
1582 d = SvPVX(sv) + (d - odest);
1584 Copy(str, d, len, char);
1591 yyerror("Missing braces on \\N{}");
1594 /* \c is a control character */
1603 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
1607 /* printf-style backslashes, formfeeds, newlines, etc */
1609 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
1612 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
1615 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
1618 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
1621 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
1624 *d++ = ASCII_TO_NEED(has_utf8,'\033');
1627 *d++ = ASCII_TO_NEED(has_utf8,'\007');
1633 } /* end if (backslash) */
1636 /* If we started with encoded form, or already know we want it
1637 and then encode the next character */
1638 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
1640 UV uv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
1641 STRLEN need = UNISKIP(NATIVE_TO_UNI(uv));
1644 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
1645 STRLEN off = d - SvPVX(sv);
1646 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
1648 d = (char*)uvchr_to_utf8((U8*)d, uv);
1652 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1654 } /* while loop to process each character */
1656 /* terminate the string and set up the sv */
1658 SvCUR_set(sv, d - SvPVX(sv));
1659 if (SvCUR(sv) >= SvLEN(sv))
1660 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
1663 if (PL_encoding && !has_utf8) {
1664 sv_recode_to_utf8(sv, PL_encoding);
1669 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1670 PL_sublex_info.sub_op->op_private |=
1671 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1675 /* shrink the sv if we allocated more than we used */
1676 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1677 SvLEN_set(sv, SvCUR(sv) + 1);
1678 Renew(SvPVX(sv), SvLEN(sv), char);
1681 /* return the substring (via yylval) only if we parsed anything */
1682 if (s > PL_bufptr) {
1683 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
1684 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
1686 ( PL_lex_inwhat == OP_TRANS
1688 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
1691 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1698 * Returns TRUE if there's more to the expression (e.g., a subscript),
1701 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
1703 * ->[ and ->{ return TRUE
1704 * { and [ outside a pattern are always subscripts, so return TRUE
1705 * if we're outside a pattern and it's not { or [, then return FALSE
1706 * if we're in a pattern and the first char is a {
1707 * {4,5} (any digits around the comma) returns FALSE
1708 * if we're in a pattern and the first char is a [
1710 * [SOMETHING] has a funky algorithm to decide whether it's a
1711 * character class or not. It has to deal with things like
1712 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
1713 * anything else returns TRUE
1716 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
1719 S_intuit_more(pTHX_ register char *s)
1721 if (PL_lex_brackets)
1723 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
1725 if (*s != '{' && *s != '[')
1730 /* In a pattern, so maybe we have {n,m}. */
1747 /* On the other hand, maybe we have a character class */
1750 if (*s == ']' || *s == '^')
1753 /* this is terrifying, and it works */
1754 int weight = 2; /* let's weigh the evidence */
1756 unsigned char un_char = 255, last_un_char;
1757 char *send = strchr(s,']');
1758 char tmpbuf[sizeof PL_tokenbuf * 4];
1760 if (!send) /* has to be an expression */
1763 Zero(seen,256,char);
1766 else if (isDIGIT(*s)) {
1768 if (isDIGIT(s[1]) && s[2] == ']')
1774 for (; s < send; s++) {
1775 last_un_char = un_char;
1776 un_char = (unsigned char)*s;
1781 weight -= seen[un_char] * 10;
1782 if (isALNUM_lazy_if(s+1,UTF)) {
1783 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
1784 if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
1789 else if (*s == '$' && s[1] &&
1790 strchr("[#!%*<>()-=",s[1])) {
1791 if (/*{*/ strchr("])} =",s[2]))
1800 if (strchr("wds]",s[1]))
1802 else if (seen['\''] || seen['"'])
1804 else if (strchr("rnftbxcav",s[1]))
1806 else if (isDIGIT(s[1])) {
1808 while (s[1] && isDIGIT(s[1]))
1818 if (strchr("aA01! ",last_un_char))
1820 if (strchr("zZ79~",s[1]))
1822 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
1823 weight -= 5; /* cope with negative subscript */
1826 if (!isALNUM(last_un_char) && !strchr("$@&",last_un_char) &&
1827 isALPHA(*s) && s[1] && isALPHA(s[1])) {
1832 if (keyword(tmpbuf, d - tmpbuf))
1835 if (un_char == last_un_char + 1)
1837 weight -= seen[un_char];
1842 if (weight >= 0) /* probably a character class */
1852 * Does all the checking to disambiguate
1854 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
1855 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
1857 * First argument is the stuff after the first token, e.g. "bar".
1859 * Not a method if bar is a filehandle.
1860 * Not a method if foo is a subroutine prototyped to take a filehandle.
1861 * Not a method if it's really "Foo $bar"
1862 * Method if it's "foo $bar"
1863 * Not a method if it's really "print foo $bar"
1864 * Method if it's really "foo package::" (interpreted as package->foo)
1865 * Not a method if bar is known to be a subroutne ("sub bar; foo bar")
1866 * Not a method if bar is a filehandle or package, but is quoted with
1871 S_intuit_method(pTHX_ char *start, GV *gv)
1873 char *s = start + (*start == '$');
1874 char tmpbuf[sizeof PL_tokenbuf];
1882 if ((cv = GvCVu(gv))) {
1883 char *proto = SvPVX(cv);
1893 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
1894 /* start is the beginning of the possible filehandle/object,
1895 * and s is the end of it
1896 * tmpbuf is a copy of it
1899 if (*start == '$') {
1900 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
1905 return *s == '(' ? FUNCMETH : METHOD;
1907 if (!keyword(tmpbuf, len)) {
1908 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
1913 indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
1914 if (indirgv && GvCVu(indirgv))
1916 /* filehandle or package name makes it a method */
1917 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
1919 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
1920 return 0; /* no assumptions -- "=>" quotes bearword */
1922 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
1923 newSVpvn(tmpbuf,len));
1924 PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
1928 return *s == '(' ? FUNCMETH : METHOD;
1936 * Return a string of Perl code to load the debugger. If PERL5DB
1937 * is set, it will return the contents of that, otherwise a
1938 * compile-time require of perl5db.pl.
1945 char *pdb = PerlEnv_getenv("PERL5DB");
1949 SETERRNO(0,SS$_NORMAL);
1950 return "BEGIN { require 'perl5db.pl' }";
1956 /* Encoded script support. filter_add() effectively inserts a
1957 * 'pre-processing' function into the current source input stream.
1958 * Note that the filter function only applies to the current source file
1959 * (e.g., it will not affect files 'require'd or 'use'd by this one).
1961 * The datasv parameter (which may be NULL) can be used to pass
1962 * private data to this instance of the filter. The filter function
1963 * can recover the SV using the FILTER_DATA macro and use it to
1964 * store private buffers and state information.
1966 * The supplied datasv parameter is upgraded to a PVIO type
1967 * and the IoDIRP/IoANY field is used to store the function pointer,
1968 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
1969 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
1970 * private use must be set using malloc'd pointers.
1974 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
1979 if (!PL_rsfp_filters)
1980 PL_rsfp_filters = newAV();
1982 datasv = NEWSV(255,0);
1983 if (!SvUPGRADE(datasv, SVt_PVIO))
1984 Perl_die(aTHX_ "Can't upgrade filter_add data to SVt_PVIO");
1985 IoANY(datasv) = (void *)funcp; /* stash funcp into spare field */
1986 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
1987 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
1988 (void*)funcp, SvPV_nolen(datasv)));
1989 av_unshift(PL_rsfp_filters, 1);
1990 av_store(PL_rsfp_filters, 0, datasv) ;
1995 /* Delete most recently added instance of this filter function. */
1997 Perl_filter_del(pTHX_ filter_t funcp)
2000 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", (void*)funcp));
2001 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2003 /* if filter is on top of stack (usual case) just pop it off */
2004 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2005 if (IoANY(datasv) == (void *)funcp) {
2006 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2007 IoANY(datasv) = (void *)NULL;
2008 sv_free(av_pop(PL_rsfp_filters));
2012 /* we need to search for the correct entry and clear it */
2013 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2017 /* Invoke the n'th filter function for the current rsfp. */
2019 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2022 /* 0 = read one text line */
2027 if (!PL_rsfp_filters)
2029 if (idx > AvFILLp(PL_rsfp_filters)){ /* Any more filters? */
2030 /* Provide a default input filter to make life easy. */
2031 /* Note that we append to the line. This is handy. */
2032 DEBUG_P(PerlIO_printf(Perl_debug_log,
2033 "filter_read %d: from rsfp\n", idx));
2037 int old_len = SvCUR(buf_sv) ;
2039 /* ensure buf_sv is large enough */
2040 SvGROW(buf_sv, old_len + maxlen) ;
2041 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2042 if (PerlIO_error(PL_rsfp))
2043 return -1; /* error */
2045 return 0 ; /* end of file */
2047 SvCUR_set(buf_sv, old_len + len) ;
2050 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2051 if (PerlIO_error(PL_rsfp))
2052 return -1; /* error */
2054 return 0 ; /* end of file */
2057 return SvCUR(buf_sv);
2059 /* Skip this filter slot if filter has been deleted */
2060 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef){
2061 DEBUG_P(PerlIO_printf(Perl_debug_log,
2062 "filter_read %d: skipped (filter deleted)\n",
2064 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2066 /* Get function pointer hidden within datasv */
2067 funcp = (filter_t)IoANY(datasv);
2068 DEBUG_P(PerlIO_printf(Perl_debug_log,
2069 "filter_read %d: via function %p (%s)\n",
2070 idx, (void*)funcp, SvPV_nolen(datasv)));
2071 /* Call function. The function is expected to */
2072 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2073 /* Return: <0:error, =0:eof, >0:not eof */
2074 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2078 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2080 #ifdef PERL_CR_FILTER
2081 if (!PL_rsfp_filters) {
2082 filter_add(S_cr_textfilter,NULL);
2085 if (PL_rsfp_filters) {
2088 SvCUR_set(sv, 0); /* start with empty line */
2089 if (FILTER_READ(0, sv, 0) > 0)
2090 return ( SvPVX(sv) ) ;
2095 return (sv_gets(sv, fp, append));
2099 S_find_in_my_stash(pTHX_ char *pkgname, I32 len)
2103 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2107 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2108 (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
2110 return GvHV(gv); /* Foo:: */
2113 /* use constant CLASS => 'MyClass' */
2114 if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
2116 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2117 pkgname = SvPV_nolen(sv);
2121 return gv_stashpv(pkgname, FALSE);
2125 static char* exp_name[] =
2126 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2127 "ATTRTERM", "TERMBLOCK"
2134 Works out what to call the token just pulled out of the input
2135 stream. The yacc parser takes care of taking the ops we return and
2136 stitching them into a tree.
2142 if read an identifier
2143 if we're in a my declaration
2144 croak if they tried to say my($foo::bar)
2145 build the ops for a my() declaration
2146 if it's an access to a my() variable
2147 are we in a sort block?
2148 croak if my($a); $a <=> $b
2149 build ops for access to a my() variable
2150 if in a dq string, and they've said @foo and we can't find @foo
2152 build ops for a bareword
2153 if we already built the token before, use it.
2156 #ifdef USE_PURE_BISON
2158 Perl_yylex_r(pTHX_ YYSTYPE *lvalp, int *lcharp)
2163 yylval_pointer[yyactlevel] = lvalp;
2164 yychar_pointer[yyactlevel] = lcharp;
2165 if (yyactlevel >= YYMAXLEVEL)
2166 Perl_croak(aTHX_ "panic: YYMAXLEVEL");
2168 r = Perl_yylex(aTHX);
2178 #pragma segment Perl_yylex
2191 /* check if there's an identifier for us to look at */
2192 if (PL_pending_ident)
2193 return S_pending_ident(aTHX);
2195 /* no identifier pending identification */
2197 switch (PL_lex_state) {
2199 case LEX_NORMAL: /* Some compilers will produce faster */
2200 case LEX_INTERPNORMAL: /* code if we comment these out. */
2204 /* when we've already built the next token, just pull it out of the queue */
2207 yylval = PL_nextval[PL_nexttoke];
2209 PL_lex_state = PL_lex_defer;
2210 PL_expect = PL_lex_expect;
2211 PL_lex_defer = LEX_NORMAL;
2213 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2214 "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
2215 (IV)PL_nexttype[PL_nexttoke]); });
2217 return(PL_nexttype[PL_nexttoke]);
2219 /* interpolated case modifiers like \L \U, including \Q and \E.
2220 when we get here, PL_bufptr is at the \
2222 case LEX_INTERPCASEMOD:
2224 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
2225 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
2227 /* handle \E or end of string */
2228 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
2232 if (PL_lex_casemods) {
2233 oldmod = PL_lex_casestack[--PL_lex_casemods];
2234 PL_lex_casestack[PL_lex_casemods] = '\0';
2236 if (PL_bufptr != PL_bufend && strchr("LUQ", oldmod)) {
2238 PL_lex_state = LEX_INTERPCONCAT;
2242 if (PL_bufptr != PL_bufend)
2244 PL_lex_state = LEX_INTERPCONCAT;
2248 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2249 "### Saw case modifier at '%s'\n", PL_bufptr); });
2251 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
2252 tmp = *s, *s = s[2], s[2] = tmp; /* misordered... */
2253 if (strchr("LU", *s) &&
2254 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U')))
2256 PL_lex_casestack[--PL_lex_casemods] = '\0';
2259 if (PL_lex_casemods > 10) {
2260 char* newlb = Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
2261 if (newlb != PL_lex_casestack) {
2263 PL_lex_casestack = newlb;
2266 PL_lex_casestack[PL_lex_casemods++] = *s;
2267 PL_lex_casestack[PL_lex_casemods] = '\0';
2268 PL_lex_state = LEX_INTERPCONCAT;
2269 PL_nextval[PL_nexttoke].ival = 0;
2272 PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
2274 PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
2276 PL_nextval[PL_nexttoke].ival = OP_LC;
2278 PL_nextval[PL_nexttoke].ival = OP_UC;
2280 PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
2282 Perl_croak(aTHX_ "panic: yylex");
2285 if (PL_lex_starts) {
2294 case LEX_INTERPPUSH:
2295 return sublex_push();
2297 case LEX_INTERPSTART:
2298 if (PL_bufptr == PL_bufend)
2299 return sublex_done();
2300 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2301 "### Interpolated variable at '%s'\n", PL_bufptr); });
2303 PL_lex_dojoin = (*PL_bufptr == '@');
2304 PL_lex_state = LEX_INTERPNORMAL;
2305 if (PL_lex_dojoin) {
2306 PL_nextval[PL_nexttoke].ival = 0;
2308 #ifdef USE_5005THREADS
2309 PL_nextval[PL_nexttoke].opval = newOP(OP_THREADSV, 0);
2310 PL_nextval[PL_nexttoke].opval->op_targ = find_threadsv("\"");
2311 force_next(PRIVATEREF);
2313 force_ident("\"", '$');
2314 #endif /* USE_5005THREADS */
2315 PL_nextval[PL_nexttoke].ival = 0;
2317 PL_nextval[PL_nexttoke].ival = 0;
2319 PL_nextval[PL_nexttoke].ival = OP_JOIN; /* emulate join($", ...) */
2322 if (PL_lex_starts++) {
2328 case LEX_INTERPENDMAYBE:
2329 if (intuit_more(PL_bufptr)) {
2330 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
2336 if (PL_lex_dojoin) {
2337 PL_lex_dojoin = FALSE;
2338 PL_lex_state = LEX_INTERPCONCAT;
2341 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
2342 && SvEVALED(PL_lex_repl))
2344 if (PL_bufptr != PL_bufend)
2345 Perl_croak(aTHX_ "Bad evalled substitution pattern");
2346 PL_lex_repl = Nullsv;
2349 case LEX_INTERPCONCAT:
2351 if (PL_lex_brackets)
2352 Perl_croak(aTHX_ "panic: INTERPCONCAT");
2354 if (PL_bufptr == PL_bufend)
2355 return sublex_done();
2357 if (SvIVX(PL_linestr) == '\'') {
2358 SV *sv = newSVsv(PL_linestr);
2361 else if ( PL_hints & HINT_NEW_RE )
2362 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
2363 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2367 s = scan_const(PL_bufptr);
2369 PL_lex_state = LEX_INTERPCASEMOD;
2371 PL_lex_state = LEX_INTERPSTART;
2374 if (s != PL_bufptr) {
2375 PL_nextval[PL_nexttoke] = yylval;
2378 if (PL_lex_starts++)
2388 PL_lex_state = LEX_NORMAL;
2389 s = scan_formline(PL_bufptr);
2390 if (!PL_lex_formbrack)
2396 PL_oldoldbufptr = PL_oldbufptr;
2399 PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at %s\n",
2400 exp_name[PL_expect], s);
2406 if (isIDFIRST_lazy_if(s,UTF))
2408 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
2411 goto fake_eof; /* emulate EOF on ^D or ^Z */
2416 if (PL_lex_brackets)
2417 yyerror("Missing right curly or square bracket");
2418 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2419 "### Tokener got EOF\n");
2423 if (s++ < PL_bufend)
2424 goto retry; /* ignore stray nulls */
2427 if (!PL_in_eval && !PL_preambled) {
2428 PL_preambled = TRUE;
2429 sv_setpv(PL_linestr,incl_perldb());
2430 if (SvCUR(PL_linestr))
2431 sv_catpv(PL_linestr,";");
2433 while(AvFILLp(PL_preambleav) >= 0) {
2434 SV *tmpsv = av_shift(PL_preambleav);
2435 sv_catsv(PL_linestr, tmpsv);
2436 sv_catpv(PL_linestr, ";");
2439 sv_free((SV*)PL_preambleav);
2440 PL_preambleav = NULL;
2442 if (PL_minus_n || PL_minus_p) {
2443 sv_catpv(PL_linestr, "LINE: while (<>) {");
2445 sv_catpv(PL_linestr,"chomp;");
2448 if (strchr("/'\"", *PL_splitstr)
2449 && strchr(PL_splitstr + 1, *PL_splitstr))
2450 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
2453 s = "'~#\200\1'"; /* surely one char is unused...*/
2454 while (s[1] && strchr(PL_splitstr, *s)) s++;
2456 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s%c",
2457 "q" + (delim == '\''), delim);
2458 for (s = PL_splitstr; *s; s++) {
2460 sv_catpvn(PL_linestr, "\\", 1);
2461 sv_catpvn(PL_linestr, s, 1);
2463 Perl_sv_catpvf(aTHX_ PL_linestr, "%c);", delim);
2467 sv_catpv(PL_linestr,"our @F=split(' ');");
2470 sv_catpv(PL_linestr, "\n");
2471 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2472 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2473 PL_last_lop = PL_last_uni = Nullch;
2474 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2475 SV *sv = NEWSV(85,0);
2477 sv_upgrade(sv, SVt_PVMG);
2478 sv_setsv(sv,PL_linestr);
2481 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2486 bof = PL_rsfp ? TRUE : FALSE;
2487 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == Nullch) {
2490 if (PL_preprocess && !PL_in_eval)
2491 (void)PerlProc_pclose(PL_rsfp);
2492 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
2493 PerlIO_clearerr(PL_rsfp);
2495 (void)PerlIO_close(PL_rsfp);
2497 PL_doextract = FALSE;
2499 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
2500 sv_setpv(PL_linestr,PL_minus_p ? ";}continue{print" : "");
2501 sv_catpv(PL_linestr,";}");
2502 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2503 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2504 PL_last_lop = PL_last_uni = Nullch;
2505 PL_minus_n = PL_minus_p = 0;
2508 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2509 PL_last_lop = PL_last_uni = Nullch;
2510 sv_setpv(PL_linestr,"");
2511 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
2513 /* if it looks like the start of a BOM, check if it in fact is */
2514 else if (bof && (!*s || *(U8*)s == 0xEF || *(U8*)s >= 0xFE)) {
2515 #ifdef PERLIO_IS_STDIO
2516 # ifdef __GNU_LIBRARY__
2517 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
2518 # define FTELL_FOR_PIPE_IS_BROKEN
2522 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
2523 # define FTELL_FOR_PIPE_IS_BROKEN
2528 #ifdef FTELL_FOR_PIPE_IS_BROKEN
2529 /* This loses the possibility to detect the bof
2530 * situation on perl -P when the libc5 is being used.
2531 * Workaround? Maybe attach some extra state to PL_rsfp?
2534 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
2536 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
2539 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2540 s = swallow_bom((U8*)s);
2544 /* Incest with pod. */
2545 if (*s == '=' && strnEQ(s, "=cut", 4)) {
2546 sv_setpv(PL_linestr, "");
2547 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2548 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2549 PL_last_lop = PL_last_uni = Nullch;
2550 PL_doextract = FALSE;
2554 } while (PL_doextract);
2555 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
2556 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2557 SV *sv = NEWSV(85,0);
2559 sv_upgrade(sv, SVt_PVMG);
2560 sv_setsv(sv,PL_linestr);
2563 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2565 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2566 PL_last_lop = PL_last_uni = Nullch;
2567 if (CopLINE(PL_curcop) == 1) {
2568 while (s < PL_bufend && isSPACE(*s))
2570 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
2574 if (*s == '#' && *(s+1) == '!')
2576 #ifdef ALTERNATE_SHEBANG
2578 static char as[] = ALTERNATE_SHEBANG;
2579 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
2580 d = s + (sizeof(as) - 1);
2582 #endif /* ALTERNATE_SHEBANG */
2591 while (*d && !isSPACE(*d))
2595 #ifdef ARG_ZERO_IS_SCRIPT
2596 if (ipathend > ipath) {
2598 * HP-UX (at least) sets argv[0] to the script name,
2599 * which makes $^X incorrect. And Digital UNIX and Linux,
2600 * at least, set argv[0] to the basename of the Perl
2601 * interpreter. So, having found "#!", we'll set it right.
2603 SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV)); /* $^X */
2604 assert(SvPOK(x) || SvGMAGICAL(x));
2605 if (sv_eq(x, CopFILESV(PL_curcop))) {
2606 sv_setpvn(x, ipath, ipathend - ipath);
2609 TAINT_NOT; /* $^X is always tainted, but that's OK */
2611 #endif /* ARG_ZERO_IS_SCRIPT */
2616 d = instr(s,"perl -");
2618 d = instr(s,"perl");
2620 /* avoid getting into infinite loops when shebang
2621 * line contains "Perl" rather than "perl" */
2623 for (d = ipathend-4; d >= ipath; --d) {
2624 if ((*d == 'p' || *d == 'P')
2625 && !ibcmp(d, "perl", 4))
2635 #ifdef ALTERNATE_SHEBANG
2637 * If the ALTERNATE_SHEBANG on this system starts with a
2638 * character that can be part of a Perl expression, then if
2639 * we see it but not "perl", we're probably looking at the
2640 * start of Perl code, not a request to hand off to some
2641 * other interpreter. Similarly, if "perl" is there, but
2642 * not in the first 'word' of the line, we assume the line
2643 * contains the start of the Perl program.
2645 if (d && *s != '#') {
2647 while (*c && !strchr("; \t\r\n\f\v#", *c))
2650 d = Nullch; /* "perl" not in first word; ignore */
2652 *s = '#'; /* Don't try to parse shebang line */
2654 #endif /* ALTERNATE_SHEBANG */
2655 #ifndef MACOS_TRADITIONAL
2660 !instr(s,"indir") &&
2661 instr(PL_origargv[0],"perl"))
2667 while (s < PL_bufend && isSPACE(*s))
2669 if (s < PL_bufend) {
2670 Newz(899,newargv,PL_origargc+3,char*);
2672 while (s < PL_bufend && !isSPACE(*s))
2675 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
2678 newargv = PL_origargv;
2680 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
2681 Perl_croak(aTHX_ "Can't exec %s", ipath);
2685 U32 oldpdb = PL_perldb;
2686 bool oldn = PL_minus_n;
2687 bool oldp = PL_minus_p;
2689 while (*d && !isSPACE(*d)) d++;
2690 while (SPACE_OR_TAB(*d)) d++;
2693 bool switches_done = PL_doswitches;
2695 if (*d == 'M' || *d == 'm') {
2697 while (*d && !isSPACE(*d)) d++;
2698 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
2701 d = moreswitches(d);
2703 if ((PERLDB_LINE && !oldpdb) ||
2704 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
2705 /* if we have already added "LINE: while (<>) {",
2706 we must not do it again */
2708 sv_setpv(PL_linestr, "");
2709 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2710 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2711 PL_last_lop = PL_last_uni = Nullch;
2712 PL_preambled = FALSE;
2714 (void)gv_fetchfile(PL_origfilename);
2717 if (PL_doswitches && !switches_done) {
2718 int argc = PL_origargc;
2719 char **argv = PL_origargv;
2722 } while (argc && argv[0][0] == '-' && argv[0][1]);
2723 init_argv_symbols(argc,argv);
2729 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2731 PL_lex_state = LEX_FORMLINE;
2736 #ifdef PERL_STRICT_CR
2737 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
2739 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
2741 case ' ': case '\t': case '\f': case 013:
2742 #ifdef MACOS_TRADITIONAL
2749 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
2750 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
2751 /* handle eval qq[#line 1 "foo"\n ...] */
2752 CopLINE_dec(PL_curcop);
2756 while (s < d && *s != '\n')
2760 else if (s > d) /* Found by Ilya: feed random input to Perl. */
2761 Perl_croak(aTHX_ "panic: input overflow");
2763 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2765 PL_lex_state = LEX_FORMLINE;
2775 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
2782 while (s < PL_bufend && SPACE_OR_TAB(*s))
2785 if (strnEQ(s,"=>",2)) {
2786 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
2787 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2788 "### Saw unary minus before =>, forcing word '%s'\n", s);
2790 OPERATOR('-'); /* unary minus */
2792 PL_last_uni = PL_oldbufptr;
2794 case 'r': ftst = OP_FTEREAD; break;
2795 case 'w': ftst = OP_FTEWRITE; break;
2796 case 'x': ftst = OP_FTEEXEC; break;
2797 case 'o': ftst = OP_FTEOWNED; break;
2798 case 'R': ftst = OP_FTRREAD; break;
2799 case 'W': ftst = OP_FTRWRITE; break;
2800 case 'X': ftst = OP_FTREXEC; break;
2801 case 'O': ftst = OP_FTROWNED; break;
2802 case 'e': ftst = OP_FTIS; break;
2803 case 'z': ftst = OP_FTZERO; break;
2804 case 's': ftst = OP_FTSIZE; break;
2805 case 'f': ftst = OP_FTFILE; break;
2806 case 'd': ftst = OP_FTDIR; break;
2807 case 'l': ftst = OP_FTLINK; break;
2808 case 'p': ftst = OP_FTPIPE; break;
2809 case 'S': ftst = OP_FTSOCK; break;
2810 case 'u': ftst = OP_FTSUID; break;
2811 case 'g': ftst = OP_FTSGID; break;
2812 case 'k': ftst = OP_FTSVTX; break;
2813 case 'b': ftst = OP_FTBLK; break;
2814 case 'c': ftst = OP_FTCHR; break;
2815 case 't': ftst = OP_FTTTY; break;
2816 case 'T': ftst = OP_FTTEXT; break;
2817 case 'B': ftst = OP_FTBINARY; break;
2818 case 'M': case 'A': case 'C':
2819 gv_fetchpv("\024",TRUE, SVt_PV);
2821 case 'M': ftst = OP_FTMTIME; break;
2822 case 'A': ftst = OP_FTATIME; break;
2823 case 'C': ftst = OP_FTCTIME; break;
2831 PL_last_lop_op = ftst;
2832 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2833 "### Saw file test %c\n", (int)ftst);
2838 /* Assume it was a minus followed by a one-letter named
2839 * subroutine call (or a -bareword), then. */
2840 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2841 "### %c looked like a file test but was not\n",
2850 if (PL_expect == XOPERATOR)
2855 else if (*s == '>') {
2858 if (isIDFIRST_lazy_if(s,UTF)) {
2859 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
2867 if (PL_expect == XOPERATOR)
2870 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2872 OPERATOR('-'); /* unary minus */
2879 if (PL_expect == XOPERATOR)
2884 if (PL_expect == XOPERATOR)
2887 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2893 if (PL_expect != XOPERATOR) {
2894 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
2895 PL_expect = XOPERATOR;
2896 force_ident(PL_tokenbuf, '*');
2909 if (PL_expect == XOPERATOR) {
2913 PL_tokenbuf[0] = '%';
2914 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
2915 if (!PL_tokenbuf[1]) {
2917 yyerror("Final % should be \\% or %name");
2920 PL_pending_ident = '%';
2939 switch (PL_expect) {
2942 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
2944 PL_bufptr = s; /* update in case we back off */
2950 PL_expect = XTERMBLOCK;
2954 while (isIDFIRST_lazy_if(s,UTF)) {
2955 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
2956 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
2957 if (tmp < 0) tmp = -tmp;
2972 d = scan_str(d,TRUE,TRUE);
2974 /* MUST advance bufptr here to avoid bogus
2975 "at end of line" context messages from yyerror().
2977 PL_bufptr = s + len;
2978 yyerror("Unterminated attribute parameter in attribute list");
2981 return 0; /* EOF indicator */
2985 SV *sv = newSVpvn(s, len);
2986 sv_catsv(sv, PL_lex_stuff);
2987 attrs = append_elem(OP_LIST, attrs,
2988 newSVOP(OP_CONST, 0, sv));
2989 SvREFCNT_dec(PL_lex_stuff);
2990 PL_lex_stuff = Nullsv;
2993 /* NOTE: any CV attrs applied here need to be part of
2994 the CVf_BUILTIN_ATTRS define in cv.h! */
2995 if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
2996 CvLVALUE_on(PL_compcv);
2997 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
2998 CvLOCKED_on(PL_compcv);
2999 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
3000 CvMETHOD_on(PL_compcv);
3002 else if (PL_in_my == KEY_our && len == 6 &&
3003 strnEQ(s, "unique", len))
3004 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
3006 /* After we've set the flags, it could be argued that
3007 we don't need to do the attributes.pm-based setting
3008 process, and shouldn't bother appending recognized
3009 flags. To experiment with that, uncomment the
3010 following "else". (Note that's already been
3011 uncommented. That keeps the above-applied built-in
3012 attributes from being intercepted (and possibly
3013 rejected) by a package's attribute routines, but is
3014 justified by the performance win for the common case
3015 of applying only built-in attributes.) */
3017 attrs = append_elem(OP_LIST, attrs,
3018 newSVOP(OP_CONST, 0,
3022 if (*s == ':' && s[1] != ':')
3025 break; /* require real whitespace or :'s */
3027 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3028 if (*s != ';' && *s != tmp && (tmp != '=' || *s != ')')) {
3029 char q = ((*s == '\'') ? '"' : '\'');
3030 /* If here for an expression, and parsed no attrs, back off. */
3031 if (tmp == '=' && !attrs) {
3035 /* MUST advance bufptr here to avoid bogus "at end of line"
3036 context messages from yyerror().
3040 yyerror("Unterminated attribute list");
3042 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3050 PL_nextval[PL_nexttoke].opval = attrs;
3058 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3059 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3075 if (PL_lex_brackets <= 0)
3076 yyerror("Unmatched right square bracket");
3079 if (PL_lex_state == LEX_INTERPNORMAL) {
3080 if (PL_lex_brackets == 0) {
3081 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3082 PL_lex_state = LEX_INTERPEND;
3089 if (PL_lex_brackets > 100) {
3090 char* newlb = Renew(PL_lex_brackstack, PL_lex_brackets + 1, char);
3091 if (newlb != PL_lex_brackstack) {
3093 PL_lex_brackstack = newlb;
3096 switch (PL_expect) {
3098 if (PL_lex_formbrack) {
3102 if (PL_oldoldbufptr == PL_last_lop)
3103 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3105 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3106 OPERATOR(HASHBRACK);
3108 while (s < PL_bufend && SPACE_OR_TAB(*s))
3111 PL_tokenbuf[0] = '\0';
3112 if (d < PL_bufend && *d == '-') {
3113 PL_tokenbuf[0] = '-';
3115 while (d < PL_bufend && SPACE_OR_TAB(*d))
3118 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3119 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3121 while (d < PL_bufend && SPACE_OR_TAB(*d))
3124 char minus = (PL_tokenbuf[0] == '-');
3125 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3133 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3138 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3143 if (PL_oldoldbufptr == PL_last_lop)
3144 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3146 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3149 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
3151 /* This hack is to get the ${} in the message. */
3153 yyerror("syntax error");
3156 OPERATOR(HASHBRACK);
3158 /* This hack serves to disambiguate a pair of curlies
3159 * as being a block or an anon hash. Normally, expectation
3160 * determines that, but in cases where we're not in a
3161 * position to expect anything in particular (like inside
3162 * eval"") we have to resolve the ambiguity. This code
3163 * covers the case where the first term in the curlies is a
3164 * quoted string. Most other cases need to be explicitly
3165 * disambiguated by prepending a `+' before the opening
3166 * curly in order to force resolution as an anon hash.
3168 * XXX should probably propagate the outer expectation
3169 * into eval"" to rely less on this hack, but that could
3170 * potentially break current behavior of eval"".
3174 if (*s == '\'' || *s == '"' || *s == '`') {
3175 /* common case: get past first string, handling escapes */
3176 for (t++; t < PL_bufend && *t != *s;)
3177 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3181 else if (*s == 'q') {
3184 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3188 char open, close, term;
3191 while (t < PL_bufend && isSPACE(*t))
3195 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3199 for (t++; t < PL_bufend; t++) {
3200 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3202 else if (*t == open)
3206 for (t++; t < PL_bufend; t++) {
3207 if (*t == '\\' && t+1 < PL_bufend)
3209 else if (*t == close && --brackets <= 0)
3211 else if (*t == open)
3217 else if (isALNUM_lazy_if(t,UTF)) {
3219 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3222 while (t < PL_bufend && isSPACE(*t))
3224 /* if comma follows first term, call it an anon hash */
3225 /* XXX it could be a comma expression with loop modifiers */
3226 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3227 || (*t == '=' && t[1] == '>')))
3228 OPERATOR(HASHBRACK);
3229 if (PL_expect == XREF)
3232 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3238 yylval.ival = CopLINE(PL_curcop);
3239 if (isSPACE(*s) || *s == '#')
3240 PL_copline = NOLINE; /* invalidate current command line number */
3245 if (PL_lex_brackets <= 0)
3246 yyerror("Unmatched right curly bracket");
3248 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3249 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3250 PL_lex_formbrack = 0;
3251 if (PL_lex_state == LEX_INTERPNORMAL) {
3252 if (PL_lex_brackets == 0) {
3253 if (PL_expect & XFAKEBRACK) {
3254 PL_expect &= XENUMMASK;
3255 PL_lex_state = LEX_INTERPEND;
3257 return yylex(); /* ignore fake brackets */
3259 if (*s == '-' && s[1] == '>')
3260 PL_lex_state = LEX_INTERPENDMAYBE;
3261 else if (*s != '[' && *s != '{')
3262 PL_lex_state = LEX_INTERPEND;
3265 if (PL_expect & XFAKEBRACK) {
3266 PL_expect &= XENUMMASK;
3268 return yylex(); /* ignore fake brackets */
3278 if (PL_expect == XOPERATOR) {
3279 if (ckWARN(WARN_SEMICOLON)
3280 && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
3282 CopLINE_dec(PL_curcop);
3283 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
3284 CopLINE_inc(PL_curcop);
3289 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3291 PL_expect = XOPERATOR;
3292 force_ident(PL_tokenbuf, '&');
3296 yylval.ival = (OPpENTERSUB_AMPER<<8);
3315 if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
3316 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Reversed %c= operator",(int)tmp);
3318 if (PL_expect == XSTATE && isALPHA(tmp) &&
3319 (s == PL_linestart+1 || s[-2] == '\n') )
3321 if (PL_in_eval && !PL_rsfp) {
3326 if (strnEQ(s,"=cut",4)) {
3340 PL_doextract = TRUE;
3343 if (PL_lex_brackets < PL_lex_formbrack) {
3345 #ifdef PERL_STRICT_CR
3346 for (t = s; SPACE_OR_TAB(*t); t++) ;
3348 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3350 if (*t == '\n' || *t == '#') {
3368 if (PL_expect != XOPERATOR) {
3369 if (s[1] != '<' && !strchr(s,'>'))
3372 s = scan_heredoc(s);
3374 s = scan_inputsymbol(s);
3375 TERM(sublex_start());
3380 SHop(OP_LEFT_SHIFT);
3394 SHop(OP_RIGHT_SHIFT);
3403 if (PL_expect == XOPERATOR) {
3404 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3407 return ','; /* grandfather non-comma-format format */
3411 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3412 PL_tokenbuf[0] = '@';
3413 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3414 sizeof PL_tokenbuf - 1, FALSE);
3415 if (PL_expect == XOPERATOR)
3416 no_op("Array length", s);
3417 if (!PL_tokenbuf[1])
3419 PL_expect = XOPERATOR;
3420 PL_pending_ident = '#';
3424 PL_tokenbuf[0] = '$';
3425 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3426 sizeof PL_tokenbuf - 1, FALSE);
3427 if (PL_expect == XOPERATOR)
3429 if (!PL_tokenbuf[1]) {
3431 yyerror("Final $ should be \\$ or $name");
3435 /* This kludge not intended to be bulletproof. */
3436 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3437 yylval.opval = newSVOP(OP_CONST, 0,
3438 newSViv(PL_compiling.cop_arybase));
3439 yylval.opval->op_private = OPpCONST_ARYBASE;
3445 if (PL_lex_state == LEX_NORMAL)
3448 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3451 PL_tokenbuf[0] = '@';
3452 if (ckWARN(WARN_SYNTAX)) {
3454 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3457 PL_bufptr = skipspace(PL_bufptr);
3458 while (t < PL_bufend && *t != ']')
3460 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3461 "Multidimensional syntax %.*s not supported",
3462 (t - PL_bufptr) + 1, PL_bufptr);
3466 else if (*s == '{') {
3467 PL_tokenbuf[0] = '%';
3468 if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
3469 (t = strchr(s, '}')) && (t = strchr(t, '=')))
3471 char tmpbuf[sizeof PL_tokenbuf];
3473 for (t++; isSPACE(*t); t++) ;
3474 if (isIDFIRST_lazy_if(t,UTF)) {
3475 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3476 for (; isSPACE(*t); t++) ;
3477 if (*t == ';' && get_cv(tmpbuf, FALSE))
3478 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3479 "You need to quote \"%s\"", tmpbuf);
3485 PL_expect = XOPERATOR;
3486 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3487 bool islop = (PL_last_lop == PL_oldoldbufptr);
3488 if (!islop || PL_last_lop_op == OP_GREPSTART)
3489 PL_expect = XOPERATOR;
3490 else if (strchr("$@\"'`q", *s))
3491 PL_expect = XTERM; /* e.g. print $fh "foo" */
3492 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3493 PL_expect = XTERM; /* e.g. print $fh &sub */
3494 else if (isIDFIRST_lazy_if(s,UTF)) {
3495 char tmpbuf[sizeof PL_tokenbuf];
3496 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3497 if ((tmp = keyword(tmpbuf, len))) {
3498 /* binary operators exclude handle interpretations */
3510 PL_expect = XTERM; /* e.g. print $fh length() */
3515 GV *gv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
3516 if (gv && GvCVu(gv))
3517 PL_expect = XTERM; /* e.g. print $fh subr() */
3520 else if (isDIGIT(*s))
3521 PL_expect = XTERM; /* e.g. print $fh 3 */
3522 else if (*s == '.' && isDIGIT(s[1]))
3523 PL_expect = XTERM; /* e.g. print $fh .3 */
3524 else if (strchr("/?-+", *s) && !isSPACE(s[1]) && s[1] != '=')
3525 PL_expect = XTERM; /* e.g. print $fh -1 */
3526 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3527 PL_expect = XTERM; /* print $fh <<"EOF" */
3529 PL_pending_ident = '$';
3533 if (PL_expect == XOPERATOR)
3535 PL_tokenbuf[0] = '@';
3536 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3537 if (!PL_tokenbuf[1]) {
3539 yyerror("Final @ should be \\@ or @name");
3542 if (PL_lex_state == LEX_NORMAL)
3544 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3546 PL_tokenbuf[0] = '%';
3548 /* Warn about @ where they meant $. */
3549 if (ckWARN(WARN_SYNTAX)) {
3550 if (*s == '[' || *s == '{') {
3552 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3554 if (*t == '}' || *t == ']') {
3556 PL_bufptr = skipspace(PL_bufptr);
3557 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3558 "Scalar value %.*s better written as $%.*s",
3559 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3564 PL_pending_ident = '@';
3567 case '/': /* may either be division or pattern */
3568 case '?': /* may either be conditional or pattern */
3569 if (PL_expect != XOPERATOR) {
3570 /* Disable warning on "study /blah/" */
3571 if (PL_oldoldbufptr == PL_last_uni
3572 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3573 || memNE(PL_last_uni, "study", 5)
3574 || isALNUM_lazy_if(PL_last_uni+5,UTF)))
3576 s = scan_pat(s,OP_MATCH);
3577 TERM(sublex_start());
3585 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3586 #ifdef PERL_STRICT_CR
3589 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3591 && (s == PL_linestart || s[-1] == '\n') )
3593 PL_lex_formbrack = 0;
3597 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3603 yylval.ival = OPf_SPECIAL;
3609 if (PL_expect != XOPERATOR)
3614 case '0': case '1': case '2': case '3': case '4':
3615 case '5': case '6': case '7': case '8': case '9':
3616 s = scan_num(s, &yylval);
3617 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3618 "### Saw number in '%s'\n", s);
3620 if (PL_expect == XOPERATOR)
3625 s = scan_str(s,FALSE,FALSE);
3626 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3627 "### Saw string before '%s'\n", s);
3629 if (PL_expect == XOPERATOR) {
3630 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3633 return ','; /* grandfather non-comma-format format */
3639 missingterm((char*)0);
3640 yylval.ival = OP_CONST;
3641 TERM(sublex_start());
3644 s = scan_str(s,FALSE,FALSE);
3645 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3646 "### Saw string before '%s'\n", s);
3648 if (PL_expect == XOPERATOR) {
3649 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3652 return ','; /* grandfather non-comma-format format */
3658 missingterm((char*)0);
3659 yylval.ival = OP_CONST;
3660 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3661 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
3662 yylval.ival = OP_STRINGIFY;
3666 TERM(sublex_start());
3669 s = scan_str(s,FALSE,FALSE);
3670 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3671 "### Saw backtick string before '%s'\n", s);
3673 if (PL_expect == XOPERATOR)
3674 no_op("Backticks",s);
3676 missingterm((char*)0);
3677 yylval.ival = OP_BACKTICK;
3679 TERM(sublex_start());
3683 if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
3684 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
3686 if (PL_expect == XOPERATOR)
3687 no_op("Backslash",s);
3691 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3695 while (isDIGIT(*start) || *start == '_')
3697 if (*start == '.' && isDIGIT(start[1])) {
3698 s = scan_num(s, &yylval);
3701 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3702 else if (!isALPHA(*start) && (PL_expect == XTERM || PL_expect == XREF || PL_expect == XSTATE)) {
3706 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3709 s = scan_num(s, &yylval);
3716 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
3755 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3757 /* Some keywords can be followed by any delimiter, including ':' */
3758 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
3759 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
3760 (PL_tokenbuf[0] == 'q' &&
3761 strchr("qwxr", PL_tokenbuf[1])))));
3763 /* x::* is just a word, unless x is "CORE" */
3764 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
3768 while (d < PL_bufend && isSPACE(*d))
3769 d++; /* no comments skipped here, or s### is misparsed */
3771 /* Is this a label? */
3772 if (!tmp && PL_expect == XSTATE
3773 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
3775 yylval.pval = savepv(PL_tokenbuf);
3780 /* Check for keywords */
3781 tmp = keyword(PL_tokenbuf, len);
3783 /* Is this a word before a => operator? */
3784 if (*d == '=' && d[1] == '>') {
3786 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf,0));
3787 yylval.opval->op_private = OPpCONST_BARE;
3788 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
3789 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3793 if (tmp < 0) { /* second-class keyword? */
3794 GV *ogv = Nullgv; /* override (winner) */
3795 GV *hgv = Nullgv; /* hidden (loser) */
3796 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
3798 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
3801 if (GvIMPORTED_CV(gv))
3803 else if (! CvMETHOD(cv))
3807 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
3808 (gv = *gvp) != (GV*)&PL_sv_undef &&
3809 GvCVu(gv) && GvIMPORTED_CV(gv))
3815 tmp = 0; /* overridden by import or by GLOBAL */
3818 && -tmp==KEY_lock /* XXX generalizable kludge */
3820 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
3822 tmp = 0; /* any sub overrides "weak" keyword */
3824 else { /* no override */
3826 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
3827 Perl_warner(aTHX_ packWARN(WARN_MISC),
3828 "dump() better written as CORE::dump()");
3832 if (ckWARN(WARN_AMBIGUOUS) && hgv
3833 && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
3834 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3835 "Ambiguous call resolved as CORE::%s(), %s",
3836 GvENAME(hgv), "qualify as such or use &");
3843 default: /* not a keyword */
3847 char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
3849 /* Get the rest if it looks like a package qualifier */
3851 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
3853 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
3856 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
3857 *s == '\'' ? "'" : "::");
3862 if (PL_expect == XOPERATOR) {
3863 if (PL_bufptr == PL_linestart) {
3864 CopLINE_dec(PL_curcop);
3865 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
3866 CopLINE_inc(PL_curcop);
3869 no_op("Bareword",s);
3872 /* Look for a subroutine with this name in current package,
3873 unless name is "Foo::", in which case Foo is a bearword
3874 (and a package name). */
3877 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
3879 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
3880 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
3881 "Bareword \"%s\" refers to nonexistent package",
3884 PL_tokenbuf[len] = '\0';
3891 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
3894 /* if we saw a global override before, get the right name */
3897 sv = newSVpvn("CORE::GLOBAL::",14);
3898 sv_catpv(sv,PL_tokenbuf);
3901 sv = newSVpv(PL_tokenbuf,0);
3903 /* Presume this is going to be a bareword of some sort. */
3906 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3907 yylval.opval->op_private = OPpCONST_BARE;
3909 /* And if "Foo::", then that's what it certainly is. */
3914 /* See if it's the indirect object for a list operator. */
3916 if (PL_oldoldbufptr &&
3917 PL_oldoldbufptr < PL_bufptr &&
3918 (PL_oldoldbufptr == PL_last_lop
3919 || PL_oldoldbufptr == PL_last_uni) &&
3920 /* NO SKIPSPACE BEFORE HERE! */
3921 (PL_expect == XREF ||
3922 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
3924 bool immediate_paren = *s == '(';
3926 /* (Now we can afford to cross potential line boundary.) */
3929 /* Two barewords in a row may indicate method call. */
3931 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
3934 /* If not a declared subroutine, it's an indirect object. */
3935 /* (But it's an indir obj regardless for sort.) */
3937 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
3938 ((!gv || !GvCVu(gv)) &&
3939 (PL_last_lop_op != OP_MAPSTART &&
3940 PL_last_lop_op != OP_GREPSTART))))
3942 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
3947 PL_expect = XOPERATOR;
3950 /* Is this a word before a => operator? */
3951 if (*s == '=' && s[1] == '>' && !pkgname) {
3953 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
3954 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
3955 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3959 /* If followed by a paren, it's certainly a subroutine. */
3962 if (gv && GvCVu(gv)) {
3963 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
3964 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
3969 PL_nextval[PL_nexttoke].opval = yylval.opval;
3970 PL_expect = XOPERATOR;
3976 /* If followed by var or block, call it a method (unless sub) */
3978 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
3979 PL_last_lop = PL_oldbufptr;
3980 PL_last_lop_op = OP_METHOD;
3984 /* If followed by a bareword, see if it looks like indir obj. */
3986 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp = intuit_method(s,gv)))
3989 /* Not a method, so call it a subroutine (if defined) */
3991 if (gv && GvCVu(gv)) {
3993 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
3994 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3995 "Ambiguous use of -%s resolved as -&%s()",
3996 PL_tokenbuf, PL_tokenbuf);
3997 /* Check for a constant sub */
3999 if ((sv = cv_const_sv(cv))) {
4001 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
4002 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
4003 yylval.opval->op_private = 0;
4007 /* Resolve to GV now. */
4008 op_free(yylval.opval);
4009 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4010 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4011 PL_last_lop = PL_oldbufptr;
4012 PL_last_lop_op = OP_ENTERSUB;
4013 /* Is there a prototype? */
4016 char *proto = SvPV((SV*)cv, len);
4019 if (strEQ(proto, "$"))
4021 if (*proto == '&' && *s == '{') {
4022 sv_setpv(PL_subname, PL_curstash ?
4023 "__ANON__" : "__ANON__::__ANON__");
4027 PL_nextval[PL_nexttoke].opval = yylval.opval;
4033 /* Call it a bare word */
4035 if (PL_hints & HINT_STRICT_SUBS)
4036 yylval.opval->op_private |= OPpCONST_STRICT;
4039 if (ckWARN(WARN_RESERVED)) {
4040 if (lastchar != '-') {
4041 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4042 if (!*d && strNE(PL_tokenbuf,"main"))
4043 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
4050 if (lastchar && strchr("*%&", lastchar) && ckWARN_d(WARN_AMBIGUOUS)) {
4051 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4052 "Operator or semicolon missing before %c%s",
4053 lastchar, PL_tokenbuf);
4054 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4055 "Ambiguous use of %c resolved as operator %c",
4056 lastchar, lastchar);
4062 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4063 newSVpv(CopFILE(PL_curcop),0));
4067 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4068 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4071 case KEY___PACKAGE__:
4072 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4074 ? newSVsv(PL_curstname)
4083 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4084 char *pname = "main";
4085 if (PL_tokenbuf[2] == 'D')
4086 pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
4087 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4090 GvIOp(gv) = newIO();
4091 IoIFP(GvIOp(gv)) = PL_rsfp;
4092 #if defined(HAS_FCNTL) && defined(F_SETFD)
4094 int fd = PerlIO_fileno(PL_rsfp);
4095 fcntl(fd,F_SETFD,fd >= 3);
4098 /* Mark this internal pseudo-handle as clean */
4099 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4101 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4102 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4103 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4105 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4106 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4107 /* if the script was opened in binmode, we need to revert
4108 * it to text mode for compatibility; but only iff it has CRs
4109 * XXX this is a questionable hack at best. */
4110 if (PL_bufend-PL_bufptr > 2
4111 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4114 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4115 loc = PerlIO_tell(PL_rsfp);
4116 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4119 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
4121 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4122 #endif /* NETWARE */
4123 #ifdef PERLIO_IS_STDIO /* really? */
4124 # if defined(__BORLANDC__)
4125 /* XXX see note in do_binmode() */
4126 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
4130 PerlIO_seek(PL_rsfp, loc, 0);
4134 #ifdef PERLIO_LAYERS
4135 if (UTF && !IN_BYTES)
4136 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4149 if (PL_expect == XSTATE) {
4156 if (*s == ':' && s[1] == ':') {
4159 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4160 if (!(tmp = keyword(PL_tokenbuf, len)))
4161 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4175 LOP(OP_ACCEPT,XTERM);
4181 LOP(OP_ATAN2,XTERM);
4187 LOP(OP_BINMODE,XTERM);
4190 LOP(OP_BLESS,XTERM);
4199 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4216 if (!PL_cryptseen) {
4217 PL_cryptseen = TRUE;
4221 LOP(OP_CRYPT,XTERM);
4224 LOP(OP_CHMOD,XTERM);
4227 LOP(OP_CHOWN,XTERM);
4230 LOP(OP_CONNECT,XTERM);
4246 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4250 PL_hints |= HINT_BLOCK_SCOPE;
4260 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4261 LOP(OP_DBMOPEN,XTERM);
4267 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4274 yylval.ival = CopLINE(PL_curcop);
4288 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4289 UNIBRACK(OP_ENTEREVAL);
4304 case KEY_endhostent:
4310 case KEY_endservent:
4313 case KEY_endprotoent:
4324 yylval.ival = CopLINE(PL_curcop);
4326 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4328 if ((PL_bufend - p) >= 3 &&
4329 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4331 else if ((PL_bufend - p) >= 4 &&