3 * Copyright (c) 1991-2001, 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_ WARN_DEPRECATED, "Use of %s is deprecated", s);
324 * Deprecate a comma-less variable list.
330 deprecate("comma-less variable list");
334 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
335 * utf16-to-utf8-reversed.
338 #ifdef PERL_CR_FILTER
342 register char *s = SvPVX(sv);
343 register char *e = s + SvCUR(sv);
344 /* outer loop optimized to do nothing if there are no CR-LFs */
346 if (*s++ == '\r' && *s == '\n') {
347 /* hit a CR-LF, need to copy the rest */
348 register char *d = s - 1;
351 if (*s == '\r' && s[1] == '\n')
362 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
364 I32 count = FILTER_READ(idx+1, sv, maxlen);
365 if (count > 0 && !maxlen)
373 * Initialize variables. Uses the Perl save_stack to save its state (for
374 * recursive calls to the parser).
378 Perl_lex_start(pTHX_ SV *line)
383 SAVEI32(PL_lex_dojoin);
384 SAVEI32(PL_lex_brackets);
385 SAVEI32(PL_lex_casemods);
386 SAVEI32(PL_lex_starts);
387 SAVEI32(PL_lex_state);
388 SAVEVPTR(PL_lex_inpat);
389 SAVEI32(PL_lex_inwhat);
390 if (PL_lex_state == LEX_KNOWNEXT) {
391 I32 toke = PL_nexttoke;
392 while (--toke >= 0) {
393 SAVEI32(PL_nexttype[toke]);
394 SAVEVPTR(PL_nextval[toke]);
396 SAVEI32(PL_nexttoke);
398 SAVECOPLINE(PL_curcop);
401 SAVEPPTR(PL_oldbufptr);
402 SAVEPPTR(PL_oldoldbufptr);
403 SAVEPPTR(PL_last_lop);
404 SAVEPPTR(PL_last_uni);
405 SAVEPPTR(PL_linestart);
406 SAVESPTR(PL_linestr);
407 SAVEPPTR(PL_lex_brackstack);
408 SAVEPPTR(PL_lex_casestack);
409 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
410 SAVESPTR(PL_lex_stuff);
411 SAVEI32(PL_lex_defer);
412 SAVEI32(PL_sublex_info.sub_inwhat);
413 SAVESPTR(PL_lex_repl);
415 SAVEINT(PL_lex_expect);
417 PL_lex_state = LEX_NORMAL;
421 New(899, PL_lex_brackstack, 120, char);
422 New(899, PL_lex_casestack, 12, char);
423 SAVEFREEPV(PL_lex_brackstack);
424 SAVEFREEPV(PL_lex_casestack);
426 *PL_lex_casestack = '\0';
429 PL_lex_stuff = Nullsv;
430 PL_lex_repl = Nullsv;
434 PL_sublex_info.sub_inwhat = 0;
436 if (SvREADONLY(PL_linestr))
437 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
438 s = SvPV(PL_linestr, len);
439 if (len && s[len-1] != ';') {
440 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
441 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
442 sv_catpvn(PL_linestr, "\n;", 2);
444 SvTEMP_off(PL_linestr);
445 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
446 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
447 PL_last_lop = PL_last_uni = Nullch;
453 * Finalizer for lexing operations. Must be called when the parser is
454 * done with the lexer.
460 PL_doextract = FALSE;
465 * This subroutine has nothing to do with tilting, whether at windmills
466 * or pinball tables. Its name is short for "increment line". It
467 * increments the current line number in CopLINE(PL_curcop) and checks
468 * to see whether the line starts with a comment of the form
469 * # line 500 "foo.pm"
470 * If so, it sets the current line number and file to the values in the comment.
474 S_incline(pTHX_ char *s)
481 CopLINE_inc(PL_curcop);
484 while (SPACE_OR_TAB(*s)) s++;
485 if (strnEQ(s, "line", 4))
489 if (SPACE_OR_TAB(*s))
493 while (SPACE_OR_TAB(*s)) s++;
499 while (SPACE_OR_TAB(*s))
501 if (*s == '"' && (t = strchr(s+1, '"'))) {
506 for (t = s; !isSPACE(*t); t++) ;
509 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
511 if (*e != '\n' && *e != '\0')
512 return; /* false alarm */
518 Safefree(CopFILE(PL_curcop));
520 SvREFCNT_dec(CopFILEGV(PL_curcop));
522 CopFILE_set(PL_curcop, s);
525 CopLINE_set(PL_curcop, atoi(n)-1);
530 * Called to gobble the appropriate amount and type of whitespace.
531 * Skips comments as well.
535 S_skipspace(pTHX_ register char *s)
537 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
538 while (s < PL_bufend && SPACE_OR_TAB(*s))
544 SSize_t oldprevlen, oldoldprevlen;
545 SSize_t oldloplen = 0, oldunilen = 0;
546 while (s < PL_bufend && isSPACE(*s)) {
547 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
552 if (s < PL_bufend && *s == '#') {
553 while (s < PL_bufend && *s != '\n')
557 if (PL_in_eval && !PL_rsfp) {
564 /* only continue to recharge the buffer if we're at the end
565 * of the buffer, we're not reading from a source filter, and
566 * we're in normal lexing mode
568 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
569 PL_lex_state == LEX_FORMLINE)
572 /* try to recharge the buffer */
573 if ((s = filter_gets(PL_linestr, PL_rsfp,
574 (prevlen = SvCUR(PL_linestr)))) == Nullch)
576 /* end of file. Add on the -p or -n magic */
577 if (PL_minus_n || PL_minus_p) {
578 sv_setpv(PL_linestr,PL_minus_p ?
579 ";}continue{print or die qq(-p destination: $!\\n)" :
581 sv_catpv(PL_linestr,";}");
582 PL_minus_n = PL_minus_p = 0;
585 sv_setpv(PL_linestr,";");
587 /* reset variables for next time we lex */
588 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
590 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
591 PL_last_lop = PL_last_uni = Nullch;
593 /* Close the filehandle. Could be from -P preprocessor,
594 * STDIN, or a regular file. If we were reading code from
595 * STDIN (because the commandline held no -e or filename)
596 * then we don't close it, we reset it so the code can
597 * read from STDIN too.
600 if (PL_preprocess && !PL_in_eval)
601 (void)PerlProc_pclose(PL_rsfp);
602 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
603 PerlIO_clearerr(PL_rsfp);
605 (void)PerlIO_close(PL_rsfp);
610 /* not at end of file, so we only read another line */
611 /* make corresponding updates to old pointers, for yyerror() */
612 oldprevlen = PL_oldbufptr - PL_bufend;
613 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
615 oldunilen = PL_last_uni - PL_bufend;
617 oldloplen = PL_last_lop - PL_bufend;
618 PL_linestart = PL_bufptr = s + prevlen;
619 PL_bufend = s + SvCUR(PL_linestr);
621 PL_oldbufptr = s + oldprevlen;
622 PL_oldoldbufptr = s + oldoldprevlen;
624 PL_last_uni = s + oldunilen;
626 PL_last_lop = s + oldloplen;
629 /* debugger active and we're not compiling the debugger code,
630 * so store the line into the debugger's array of lines
632 if (PERLDB_LINE && PL_curstash != PL_debstash) {
633 SV *sv = NEWSV(85,0);
635 sv_upgrade(sv, SVt_PVMG);
636 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
639 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
646 * Check the unary operators to ensure there's no ambiguity in how they're
647 * used. An ambiguous piece of code would be:
649 * This doesn't mean rand() + 5. Because rand() is a unary operator,
650 * the +5 is its argument.
659 if (PL_oldoldbufptr != PL_last_uni)
661 while (isSPACE(*PL_last_uni))
663 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
664 if ((t = strchr(s, '(')) && t < PL_bufptr)
666 if (ckWARN_d(WARN_AMBIGUOUS)){
669 Perl_warner(aTHX_ WARN_AMBIGUOUS,
670 "Warning: Use of \"%s\" without parens is ambiguous",
676 /* workaround to replace the UNI() macro with a function. Only the
677 * hints/uts.sh file mentions this. Other comments elsewhere in the
678 * source indicate Microport Unix might need it too.
684 #define UNI(f) return uni(f,s)
687 S_uni(pTHX_ I32 f, char *s)
692 PL_last_uni = PL_oldbufptr;
703 #endif /* CRIPPLED_CC */
706 * LOP : macro to build a list operator. Its behaviour has been replaced
707 * with a subroutine, S_lop() for which LOP is just another name.
710 #define LOP(f,x) return lop(f,x,s)
714 * Build a list operator (or something that might be one). The rules:
715 * - if we have a next token, then it's a list operator [why?]
716 * - if the next thing is an opening paren, then it's a function
717 * - else it's a list operator
721 S_lop(pTHX_ I32 f, int x, char *s)
728 PL_last_lop = PL_oldbufptr;
743 * When the lexer realizes it knows the next token (for instance,
744 * it is reordering tokens for the parser) then it can call S_force_next
745 * to know what token to return the next time the lexer is called. Caller
746 * will need to set PL_nextval[], and possibly PL_expect to ensure the lexer
747 * handles the token correctly.
751 S_force_next(pTHX_ I32 type)
753 PL_nexttype[PL_nexttoke] = type;
755 if (PL_lex_state != LEX_KNOWNEXT) {
756 PL_lex_defer = PL_lex_state;
757 PL_lex_expect = PL_expect;
758 PL_lex_state = LEX_KNOWNEXT;
764 * When the lexer knows the next thing is a word (for instance, it has
765 * just seen -> and it knows that the next char is a word char, then
766 * it calls S_force_word to stick the next word into the PL_next lookahead.
769 * char *start : buffer position (must be within PL_linestr)
770 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
771 * int check_keyword : if true, Perl checks to make sure the word isn't
772 * a keyword (do this if the word is a label, e.g. goto FOO)
773 * int allow_pack : if true, : characters will also be allowed (require,
775 * int allow_initial_tick : used by the "sub" lexer only.
779 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
784 start = skipspace(start);
786 if (isIDFIRST_lazy_if(s,UTF) ||
787 (allow_pack && *s == ':') ||
788 (allow_initial_tick && *s == '\'') )
790 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
791 if (check_keyword && keyword(PL_tokenbuf, len))
793 if (token == METHOD) {
798 PL_expect = XOPERATOR;
801 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST,0, newSVpv(PL_tokenbuf,0));
802 PL_nextval[PL_nexttoke].opval->op_private |= OPpCONST_BARE;
810 * Called when the lexer wants $foo *foo &foo etc, but the program
811 * text only contains the "foo" portion. The first argument is a pointer
812 * to the "foo", and the second argument is the type symbol to prefix.
813 * Forces the next token to be a "WORD".
814 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
818 S_force_ident(pTHX_ register char *s, int kind)
821 OP* o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
822 PL_nextval[PL_nexttoke].opval = o;
825 o->op_private = OPpCONST_ENTERED;
826 /* XXX see note in pp_entereval() for why we forgo typo
827 warnings if the symbol must be introduced in an eval.
829 gv_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
830 kind == '$' ? SVt_PV :
831 kind == '@' ? SVt_PVAV :
832 kind == '%' ? SVt_PVHV :
840 Perl_str_to_version(pTHX_ SV *sv)
845 char *start = SvPVx(sv,len);
846 bool utf = SvUTF8(sv) ? TRUE : FALSE;
847 char *end = start + len;
848 while (start < end) {
852 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
857 retval += ((NV)n)/nshift;
866 * Forces the next token to be a version number.
867 * If the next token appears to be an invalid version number, (e.g. "v2b"),
868 * and if "guessing" is TRUE, then no new token is created (and the caller
869 * must use an alternative parsing method).
873 S_force_version(pTHX_ char *s, int guessing)
875 OP *version = Nullop;
884 while (isDIGIT(*d) || *d == '_' || *d == '.')
886 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
888 s = scan_num(s, &yylval);
889 version = yylval.opval;
890 ver = cSVOPx(version)->op_sv;
891 if (SvPOK(ver) && !SvNIOK(ver)) {
892 (void)SvUPGRADE(ver, SVt_PVNV);
893 SvNVX(ver) = str_to_version(ver);
894 SvNOK_on(ver); /* hint that it is a version */
901 /* NOTE: The parser sees the package name and the VERSION swapped */
902 PL_nextval[PL_nexttoke].opval = version;
910 * Tokenize a quoted string passed in as an SV. It finds the next
911 * chunk, up to end of string or a backslash. It may make a new
912 * SV containing that chunk (if HINT_NEW_STRING is on). It also
917 S_tokeq(pTHX_ SV *sv)
928 s = SvPV_force(sv, len);
929 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
932 while (s < send && *s != '\\')
937 if ( PL_hints & HINT_NEW_STRING ) {
938 pv = sv_2mortal(newSVpvn(SvPVX(pv), len));
944 if (s + 1 < send && (s[1] == '\\'))
945 s++; /* all that, just for this */
950 SvCUR_set(sv, d - SvPVX(sv));
952 if ( PL_hints & HINT_NEW_STRING )
953 return new_constant(NULL, 0, "q", sv, pv, "q");
958 * Now come three functions related to double-quote context,
959 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
960 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
961 * interact with PL_lex_state, and create fake ( ... ) argument lists
962 * to handle functions and concatenation.
963 * They assume that whoever calls them will be setting up a fake
964 * join call, because each subthing puts a ',' after it. This lets
967 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
969 * (I'm not sure whether the spurious commas at the end of lcfirst's
970 * arguments and join's arguments are created or not).
975 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
977 * Pattern matching will set PL_lex_op to the pattern-matching op to
978 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
980 * OP_CONST and OP_READLINE are easy--just make the new op and return.
982 * Everything else becomes a FUNC.
984 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
985 * had an OP_CONST or OP_READLINE). This just sets us up for a
986 * call to S_sublex_push().
992 register I32 op_type = yylval.ival;
994 if (op_type == OP_NULL) {
995 yylval.opval = PL_lex_op;
999 if (op_type == OP_CONST || op_type == OP_READLINE) {
1000 SV *sv = tokeq(PL_lex_stuff);
1002 if (SvTYPE(sv) == SVt_PVIV) {
1003 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1009 nsv = newSVpvn(p, len);
1015 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1016 PL_lex_stuff = Nullsv;
1020 PL_sublex_info.super_state = PL_lex_state;
1021 PL_sublex_info.sub_inwhat = op_type;
1022 PL_sublex_info.sub_op = PL_lex_op;
1023 PL_lex_state = LEX_INTERPPUSH;
1027 yylval.opval = PL_lex_op;
1037 * Create a new scope to save the lexing state. The scope will be
1038 * ended in S_sublex_done. Returns a '(', starting the function arguments
1039 * to the uc, lc, etc. found before.
1040 * Sets PL_lex_state to LEX_INTERPCONCAT.
1048 PL_lex_state = PL_sublex_info.super_state;
1049 SAVEI32(PL_lex_dojoin);
1050 SAVEI32(PL_lex_brackets);
1051 SAVEI32(PL_lex_casemods);
1052 SAVEI32(PL_lex_starts);
1053 SAVEI32(PL_lex_state);
1054 SAVEVPTR(PL_lex_inpat);
1055 SAVEI32(PL_lex_inwhat);
1056 SAVECOPLINE(PL_curcop);
1057 SAVEPPTR(PL_bufptr);
1058 SAVEPPTR(PL_bufend);
1059 SAVEPPTR(PL_oldbufptr);
1060 SAVEPPTR(PL_oldoldbufptr);
1061 SAVEPPTR(PL_last_lop);
1062 SAVEPPTR(PL_last_uni);
1063 SAVEPPTR(PL_linestart);
1064 SAVESPTR(PL_linestr);
1065 SAVEPPTR(PL_lex_brackstack);
1066 SAVEPPTR(PL_lex_casestack);
1068 PL_linestr = PL_lex_stuff;
1069 PL_lex_stuff = Nullsv;
1071 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1072 = SvPVX(PL_linestr);
1073 PL_bufend += SvCUR(PL_linestr);
1074 PL_last_lop = PL_last_uni = Nullch;
1075 SAVEFREESV(PL_linestr);
1077 PL_lex_dojoin = FALSE;
1078 PL_lex_brackets = 0;
1079 New(899, PL_lex_brackstack, 120, char);
1080 New(899, PL_lex_casestack, 12, char);
1081 SAVEFREEPV(PL_lex_brackstack);
1082 SAVEFREEPV(PL_lex_casestack);
1083 PL_lex_casemods = 0;
1084 *PL_lex_casestack = '\0';
1086 PL_lex_state = LEX_INTERPCONCAT;
1087 CopLINE_set(PL_curcop, PL_multi_start);
1089 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1090 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1091 PL_lex_inpat = PL_sublex_info.sub_op;
1093 PL_lex_inpat = Nullop;
1100 * Restores lexer state after a S_sublex_push.
1106 if (!PL_lex_starts++) {
1107 SV *sv = newSVpvn("",0);
1108 if (SvUTF8(PL_linestr))
1110 PL_expect = XOPERATOR;
1111 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1115 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1116 PL_lex_state = LEX_INTERPCASEMOD;
1120 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1121 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1122 PL_linestr = PL_lex_repl;
1124 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1125 PL_bufend += SvCUR(PL_linestr);
1126 PL_last_lop = PL_last_uni = Nullch;
1127 SAVEFREESV(PL_linestr);
1128 PL_lex_dojoin = FALSE;
1129 PL_lex_brackets = 0;
1130 PL_lex_casemods = 0;
1131 *PL_lex_casestack = '\0';
1133 if (SvEVALED(PL_lex_repl)) {
1134 PL_lex_state = LEX_INTERPNORMAL;
1136 /* we don't clear PL_lex_repl here, so that we can check later
1137 whether this is an evalled subst; that means we rely on the
1138 logic to ensure sublex_done() is called again only via the
1139 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1142 PL_lex_state = LEX_INTERPCONCAT;
1143 PL_lex_repl = Nullsv;
1149 PL_bufend = SvPVX(PL_linestr);
1150 PL_bufend += SvCUR(PL_linestr);
1151 PL_expect = XOPERATOR;
1152 PL_sublex_info.sub_inwhat = 0;
1160 Extracts a pattern, double-quoted string, or transliteration. This
1163 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1164 processing a pattern (PL_lex_inpat is true), a transliteration
1165 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1167 Returns a pointer to the character scanned up to. Iff this is
1168 advanced from the start pointer supplied (ie if anything was
1169 successfully parsed), will leave an OP for the substring scanned
1170 in yylval. Caller must intuit reason for not parsing further
1171 by looking at the next characters herself.
1175 double-quoted style: \r and \n
1176 regexp special ones: \D \s
1178 backrefs: \1 (deprecated in substitution replacements)
1179 case and quoting: \U \Q \E
1180 stops on @ and $, but not for $ as tail anchor
1182 In transliterations:
1183 characters are VERY literal, except for - not at the start or end
1184 of the string, which indicates a range. scan_const expands the
1185 range to the full set of intermediate characters.
1187 In double-quoted strings:
1189 double-quoted style: \r and \n
1191 backrefs: \1 (deprecated)
1192 case and quoting: \U \Q \E
1195 scan_const does *not* construct ops to handle interpolated strings.
1196 It stops processing as soon as it finds an embedded $ or @ variable
1197 and leaves it to the caller to work out what's going on.
1199 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @:foo.
1201 $ in pattern could be $foo or could be tail anchor. Assumption:
1202 it's a tail anchor if $ is the last thing in the string, or if it's
1203 followed by one of ")| \n\t"
1205 \1 (backreferences) are turned into $1
1207 The structure of the code is
1208 while (there's a character to process) {
1209 handle transliteration ranges
1210 skip regexp comments
1211 skip # initiated comments in //x patterns
1212 check for embedded @foo
1213 check for embedded scalars
1215 leave intact backslashes from leave (below)
1216 deprecate \1 in strings and sub replacements
1217 handle string-changing backslashes \l \U \Q \E, etc.
1218 switch (what was escaped) {
1219 handle - in a transliteration (becomes a literal -)
1220 handle \132 octal characters
1221 handle 0x15 hex characters
1222 handle \cV (control V)
1223 handle printf backslashes (\f, \r, \n, etc)
1225 } (end if backslash)
1226 } (end while character to read)
1231 S_scan_const(pTHX_ char *start)
1233 register char *send = PL_bufend; /* end of the constant */
1234 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1235 register char *s = start; /* start of the constant */
1236 register char *d = SvPVX(sv); /* destination for copies */
1237 bool dorange = FALSE; /* are we in a translit range? */
1238 bool didrange = FALSE; /* did we just finish a range? */
1239 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1240 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1243 const char *leaveit = /* set of acceptably-backslashed characters */
1245 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1248 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1249 /* If we are doing a trans and we know we want UTF8 set expectation */
1250 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1251 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1255 while (s < send || dorange) {
1256 /* get transliterations out of the way (they're most literal) */
1257 if (PL_lex_inwhat == OP_TRANS) {
1258 /* expand a range A-Z to the full set of characters. AIE! */
1260 I32 i; /* current expanded character */
1261 I32 min; /* first character in range */
1262 I32 max; /* last character in range */
1265 char *c = (char*)utf8_hop((U8*)d, -1);
1269 *c = (char)UTF_TO_NATIVE(0xff);
1270 /* mark the range as done, and continue */
1276 i = d - SvPVX(sv); /* remember current offset */
1277 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1278 d = SvPVX(sv) + i; /* refresh d after realloc */
1279 d -= 2; /* eat the first char and the - */
1281 min = (U8)*d; /* first char in range */
1282 max = (U8)d[1]; /* last char in range */
1286 "Invalid [] range \"%c-%c\" in transliteration operator",
1287 (char)min, (char)max);
1291 if ((isLOWER(min) && isLOWER(max)) ||
1292 (isUPPER(min) && isUPPER(max))) {
1294 for (i = min; i <= max; i++)
1296 *d++ = NATIVE_TO_NEED(has_utf8,i);
1298 for (i = min; i <= max; i++)
1300 *d++ = NATIVE_TO_NEED(has_utf8,i);
1305 for (i = min; i <= max; i++)
1308 /* mark the range as done, and continue */
1314 /* range begins (ignore - as first or last char) */
1315 else if (*s == '-' && s+1 < send && s != start) {
1317 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1320 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1332 /* if we get here, we're not doing a transliteration */
1334 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1335 except for the last char, which will be done separately. */
1336 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1338 while (s < send && *s != ')')
1339 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1341 else if (s[2] == '{' /* This should match regcomp.c */
1342 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1345 char *regparse = s + (s[2] == '{' ? 3 : 4);
1348 while (count && (c = *regparse)) {
1349 if (c == '\\' && regparse[1])
1357 if (*regparse != ')') {
1358 regparse--; /* Leave one char for continuation. */
1359 yyerror("Sequence (?{...}) not terminated or not {}-balanced");
1361 while (s < regparse)
1362 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1366 /* likewise skip #-initiated comments in //x patterns */
1367 else if (*s == '#' && PL_lex_inpat &&
1368 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1369 while (s+1 < send && *s != '\n')
1370 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1373 /* check for embedded arrays
1374 (@foo, @:foo, @'foo, @{foo}, @$foo, @+, @-)
1376 else if (*s == '@' && s[1]
1377 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1380 /* check for embedded scalars. only stop if we're sure it's a
1383 else if (*s == '$') {
1384 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1386 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1387 break; /* in regexp, $ might be tail anchor */
1390 /* End of else if chain - OP_TRANS rejoin rest */
1393 if (*s == '\\' && s+1 < send) {
1396 /* some backslashes we leave behind */
1397 if (*leaveit && *s && strchr(leaveit, *s)) {
1398 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1399 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1403 /* deprecate \1 in strings and substitution replacements */
1404 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1405 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1407 if (ckWARN(WARN_SYNTAX))
1408 Perl_warner(aTHX_ WARN_SYNTAX, "\\%c better written as $%c", *s, *s);
1413 /* string-change backslash escapes */
1414 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1419 /* if we get here, it's either a quoted -, or a digit */
1422 /* quoted - in transliterations */
1424 if (PL_lex_inwhat == OP_TRANS) {
1431 if (ckWARN(WARN_MISC) && isALNUM(*s))
1432 Perl_warner(aTHX_ WARN_MISC,
1433 "Unrecognized escape \\%c passed through",
1435 /* default action is to copy the quoted character */
1436 goto default_action;
1439 /* \132 indicates an octal constant */
1440 case '0': case '1': case '2': case '3':
1441 case '4': case '5': case '6': case '7':
1445 uv = grok_oct(s, &len, &flags, NULL);
1448 goto NUM_ESCAPE_INSERT;
1450 /* \x24 indicates a hex constant */
1454 char* e = strchr(s, '}');
1455 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
1456 PERL_SCAN_DISALLOW_PREFIX;
1461 yyerror("Missing right brace on \\x{}");
1465 uv = grok_hex(s, &len, &flags, NULL);
1471 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
1472 uv = grok_hex(s, &len, &flags, NULL);
1478 /* Insert oct or hex escaped character.
1479 * There will always enough room in sv since such
1480 * escapes will be longer than any UTF-8 sequence
1481 * they can end up as. */
1483 /* We need to map to chars to ASCII before doing the tests
1486 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
1487 if (!has_utf8 && uv > 255) {
1488 /* Might need to recode whatever we have
1489 * accumulated so far if it contains any
1492 * (Can't we keep track of that and avoid
1493 * this rescan? --jhi)
1497 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
1498 if (!NATIVE_IS_INVARIANT(*c)) {
1503 STRLEN offset = d - SvPVX(sv);
1505 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
1509 while (src >= (U8 *)SvPVX(sv)) {
1510 if (!NATIVE_IS_INVARIANT(*src)) {
1511 U8 ch = NATIVE_TO_ASCII(*src);
1512 *dst-- = UTF8_EIGHT_BIT_LO(ch);
1513 *dst-- = UTF8_EIGHT_BIT_HI(ch);
1523 if (has_utf8 || uv > 255) {
1524 d = (char*)uvchr_to_utf8((U8*)d, uv);
1526 if (PL_lex_inwhat == OP_TRANS &&
1527 PL_sublex_info.sub_op) {
1528 PL_sublex_info.sub_op->op_private |=
1529 (PL_lex_repl ? OPpTRANS_FROM_UTF
1542 /* \N{LATIN SMALL LETTER A} is a named character */
1546 char* e = strchr(s, '}');
1552 yyerror("Missing right brace on \\N{}");
1556 res = newSVpvn(s + 1, e - s - 1);
1557 res = new_constant( Nullch, 0, "charnames",
1558 res, Nullsv, "\\N{...}" );
1560 sv_utf8_upgrade(res);
1561 str = SvPV(res,len);
1562 if (!has_utf8 && SvUTF8(res)) {
1563 char *ostart = SvPVX(sv);
1564 SvCUR_set(sv, d - ostart);
1567 sv_utf8_upgrade(sv);
1568 /* this just broke our allocation above... */
1569 SvGROW(sv, send - start);
1570 d = SvPVX(sv) + SvCUR(sv);
1573 if (len > e - s + 4) { /* I _guess_ 4 is \N{} --jhi */
1574 char *odest = SvPVX(sv);
1576 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
1577 d = SvPVX(sv) + (d - odest);
1579 Copy(str, d, len, char);
1586 yyerror("Missing braces on \\N{}");
1589 /* \c is a control character */
1598 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
1602 /* printf-style backslashes, formfeeds, newlines, etc */
1604 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
1607 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
1610 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
1613 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
1616 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
1619 *d++ = ASCII_TO_NEED(has_utf8,'\033');
1622 *d++ = ASCII_TO_NEED(has_utf8,'\007');
1628 } /* end if (backslash) */
1631 /* If we started with encoded form, or already know we want it
1632 and then encode the next character */
1633 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
1635 UV uv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
1636 STRLEN need = UNISKIP(NATIVE_TO_UNI(uv));
1639 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
1640 STRLEN off = d - SvPVX(sv);
1641 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
1643 d = (char*)uvchr_to_utf8((U8*)d, uv);
1647 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1649 } /* while loop to process each character */
1651 /* terminate the string and set up the sv */
1653 SvCUR_set(sv, d - SvPVX(sv));
1654 if (SvCUR(sv) >= SvLEN(sv))
1655 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
1658 if (PL_encoding && !has_utf8) {
1659 Perl_sv_recode_to_utf8(aTHX_ sv, PL_encoding);
1664 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1665 PL_sublex_info.sub_op->op_private |=
1666 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1670 /* shrink the sv if we allocated more than we used */
1671 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1672 SvLEN_set(sv, SvCUR(sv) + 1);
1673 Renew(SvPVX(sv), SvLEN(sv), char);
1676 /* return the substring (via yylval) only if we parsed anything */
1677 if (s > PL_bufptr) {
1678 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
1679 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
1681 ( PL_lex_inwhat == OP_TRANS
1683 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
1686 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1693 * Returns TRUE if there's more to the expression (e.g., a subscript),
1696 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
1698 * ->[ and ->{ return TRUE
1699 * { and [ outside a pattern are always subscripts, so return TRUE
1700 * if we're outside a pattern and it's not { or [, then return FALSE
1701 * if we're in a pattern and the first char is a {
1702 * {4,5} (any digits around the comma) returns FALSE
1703 * if we're in a pattern and the first char is a [
1705 * [SOMETHING] has a funky algorithm to decide whether it's a
1706 * character class or not. It has to deal with things like
1707 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
1708 * anything else returns TRUE
1711 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
1714 S_intuit_more(pTHX_ register char *s)
1716 if (PL_lex_brackets)
1718 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
1720 if (*s != '{' && *s != '[')
1725 /* In a pattern, so maybe we have {n,m}. */
1742 /* On the other hand, maybe we have a character class */
1745 if (*s == ']' || *s == '^')
1748 /* this is terrifying, and it works */
1749 int weight = 2; /* let's weigh the evidence */
1751 unsigned char un_char = 255, last_un_char;
1752 char *send = strchr(s,']');
1753 char tmpbuf[sizeof PL_tokenbuf * 4];
1755 if (!send) /* has to be an expression */
1758 Zero(seen,256,char);
1761 else if (isDIGIT(*s)) {
1763 if (isDIGIT(s[1]) && s[2] == ']')
1769 for (; s < send; s++) {
1770 last_un_char = un_char;
1771 un_char = (unsigned char)*s;
1776 weight -= seen[un_char] * 10;
1777 if (isALNUM_lazy_if(s+1,UTF)) {
1778 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
1779 if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
1784 else if (*s == '$' && s[1] &&
1785 strchr("[#!%*<>()-=",s[1])) {
1786 if (/*{*/ strchr("])} =",s[2]))
1795 if (strchr("wds]",s[1]))
1797 else if (seen['\''] || seen['"'])
1799 else if (strchr("rnftbxcav",s[1]))
1801 else if (isDIGIT(s[1])) {
1803 while (s[1] && isDIGIT(s[1]))
1813 if (strchr("aA01! ",last_un_char))
1815 if (strchr("zZ79~",s[1]))
1817 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
1818 weight -= 5; /* cope with negative subscript */
1821 if (!isALNUM(last_un_char) && !strchr("$@&",last_un_char) &&
1822 isALPHA(*s) && s[1] && isALPHA(s[1])) {
1827 if (keyword(tmpbuf, d - tmpbuf))
1830 if (un_char == last_un_char + 1)
1832 weight -= seen[un_char];
1837 if (weight >= 0) /* probably a character class */
1847 * Does all the checking to disambiguate
1849 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
1850 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
1852 * First argument is the stuff after the first token, e.g. "bar".
1854 * Not a method if bar is a filehandle.
1855 * Not a method if foo is a subroutine prototyped to take a filehandle.
1856 * Not a method if it's really "Foo $bar"
1857 * Method if it's "foo $bar"
1858 * Not a method if it's really "print foo $bar"
1859 * Method if it's really "foo package::" (interpreted as package->foo)
1860 * Not a method if bar is known to be a subroutne ("sub bar; foo bar")
1861 * Not a method if bar is a filehandle or package, but is quoted with
1866 S_intuit_method(pTHX_ char *start, GV *gv)
1868 char *s = start + (*start == '$');
1869 char tmpbuf[sizeof PL_tokenbuf];
1877 if ((cv = GvCVu(gv))) {
1878 char *proto = SvPVX(cv);
1888 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
1889 /* start is the beginning of the possible filehandle/object,
1890 * and s is the end of it
1891 * tmpbuf is a copy of it
1894 if (*start == '$') {
1895 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
1900 return *s == '(' ? FUNCMETH : METHOD;
1902 if (!keyword(tmpbuf, len)) {
1903 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
1908 indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
1909 if (indirgv && GvCVu(indirgv))
1911 /* filehandle or package name makes it a method */
1912 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
1914 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
1915 return 0; /* no assumptions -- "=>" quotes bearword */
1917 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
1918 newSVpvn(tmpbuf,len));
1919 PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
1923 return *s == '(' ? FUNCMETH : METHOD;
1931 * Return a string of Perl code to load the debugger. If PERL5DB
1932 * is set, it will return the contents of that, otherwise a
1933 * compile-time require of perl5db.pl.
1940 char *pdb = PerlEnv_getenv("PERL5DB");
1944 SETERRNO(0,SS$_NORMAL);
1945 return "BEGIN { require 'perl5db.pl' }";
1951 /* Encoded script support. filter_add() effectively inserts a
1952 * 'pre-processing' function into the current source input stream.
1953 * Note that the filter function only applies to the current source file
1954 * (e.g., it will not affect files 'require'd or 'use'd by this one).
1956 * The datasv parameter (which may be NULL) can be used to pass
1957 * private data to this instance of the filter. The filter function
1958 * can recover the SV using the FILTER_DATA macro and use it to
1959 * store private buffers and state information.
1961 * The supplied datasv parameter is upgraded to a PVIO type
1962 * and the IoDIRP/IoANY field is used to store the function pointer,
1963 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
1964 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
1965 * private use must be set using malloc'd pointers.
1969 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
1974 if (!PL_rsfp_filters)
1975 PL_rsfp_filters = newAV();
1977 datasv = NEWSV(255,0);
1978 if (!SvUPGRADE(datasv, SVt_PVIO))
1979 Perl_die(aTHX_ "Can't upgrade filter_add data to SVt_PVIO");
1980 IoANY(datasv) = (void *)funcp; /* stash funcp into spare field */
1981 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
1982 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
1983 (void*)funcp, SvPV_nolen(datasv)));
1984 av_unshift(PL_rsfp_filters, 1);
1985 av_store(PL_rsfp_filters, 0, datasv) ;
1990 /* Delete most recently added instance of this filter function. */
1992 Perl_filter_del(pTHX_ filter_t funcp)
1995 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", (void*)funcp));
1996 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
1998 /* if filter is on top of stack (usual case) just pop it off */
1999 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2000 if (IoANY(datasv) == (void *)funcp) {
2001 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2002 IoANY(datasv) = (void *)NULL;
2003 sv_free(av_pop(PL_rsfp_filters));
2007 /* we need to search for the correct entry and clear it */
2008 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2012 /* Invoke the n'th filter function for the current rsfp. */
2014 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2017 /* 0 = read one text line */
2022 if (!PL_rsfp_filters)
2024 if (idx > AvFILLp(PL_rsfp_filters)){ /* Any more filters? */
2025 /* Provide a default input filter to make life easy. */
2026 /* Note that we append to the line. This is handy. */
2027 DEBUG_P(PerlIO_printf(Perl_debug_log,
2028 "filter_read %d: from rsfp\n", idx));
2032 int old_len = SvCUR(buf_sv) ;
2034 /* ensure buf_sv is large enough */
2035 SvGROW(buf_sv, old_len + maxlen) ;
2036 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2037 if (PerlIO_error(PL_rsfp))
2038 return -1; /* error */
2040 return 0 ; /* end of file */
2042 SvCUR_set(buf_sv, old_len + len) ;
2045 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2046 if (PerlIO_error(PL_rsfp))
2047 return -1; /* error */
2049 return 0 ; /* end of file */
2052 return SvCUR(buf_sv);
2054 /* Skip this filter slot if filter has been deleted */
2055 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef){
2056 DEBUG_P(PerlIO_printf(Perl_debug_log,
2057 "filter_read %d: skipped (filter deleted)\n",
2059 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2061 /* Get function pointer hidden within datasv */
2062 funcp = (filter_t)IoANY(datasv);
2063 DEBUG_P(PerlIO_printf(Perl_debug_log,
2064 "filter_read %d: via function %p (%s)\n",
2065 idx, (void*)funcp, SvPV_nolen(datasv)));
2066 /* Call function. The function is expected to */
2067 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2068 /* Return: <0:error, =0:eof, >0:not eof */
2069 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2073 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2075 #ifdef PERL_CR_FILTER
2076 if (!PL_rsfp_filters) {
2077 filter_add(S_cr_textfilter,NULL);
2080 if (PL_rsfp_filters) {
2083 SvCUR_set(sv, 0); /* start with empty line */
2084 if (FILTER_READ(0, sv, 0) > 0)
2085 return ( SvPVX(sv) ) ;
2090 return (sv_gets(sv, fp, append));
2094 S_find_in_my_stash(pTHX_ char *pkgname, I32 len)
2098 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2102 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2103 (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
2105 return GvHV(gv); /* Foo:: */
2108 /* use constant CLASS => 'MyClass' */
2109 if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
2111 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2112 pkgname = SvPV_nolen(sv);
2116 return gv_stashpv(pkgname, FALSE);
2120 static char* exp_name[] =
2121 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2122 "ATTRTERM", "TERMBLOCK"
2129 Works out what to call the token just pulled out of the input
2130 stream. The yacc parser takes care of taking the ops we return and
2131 stitching them into a tree.
2137 if read an identifier
2138 if we're in a my declaration
2139 croak if they tried to say my($foo::bar)
2140 build the ops for a my() declaration
2141 if it's an access to a my() variable
2142 are we in a sort block?
2143 croak if my($a); $a <=> $b
2144 build ops for access to a my() variable
2145 if in a dq string, and they've said @foo and we can't find @foo
2147 build ops for a bareword
2148 if we already built the token before, use it.
2151 #ifdef USE_PURE_BISON
2153 Perl_yylex_r(pTHX_ YYSTYPE *lvalp, int *lcharp)
2158 yylval_pointer[yyactlevel] = lvalp;
2159 yychar_pointer[yyactlevel] = lcharp;
2160 if (yyactlevel >= YYMAXLEVEL)
2161 Perl_croak(aTHX_ "panic: YYMAXLEVEL");
2163 r = Perl_yylex(aTHX);
2173 #pragma segment Perl_yylex
2186 /* check if there's an identifier for us to look at */
2187 if (PL_pending_ident)
2188 return S_pending_ident(aTHX);
2190 /* no identifier pending identification */
2192 switch (PL_lex_state) {
2194 case LEX_NORMAL: /* Some compilers will produce faster */
2195 case LEX_INTERPNORMAL: /* code if we comment these out. */
2199 /* when we've already built the next token, just pull it out of the queue */
2202 yylval = PL_nextval[PL_nexttoke];
2204 PL_lex_state = PL_lex_defer;
2205 PL_expect = PL_lex_expect;
2206 PL_lex_defer = LEX_NORMAL;
2208 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2209 "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
2210 (IV)PL_nexttype[PL_nexttoke]); });
2212 return(PL_nexttype[PL_nexttoke]);
2214 /* interpolated case modifiers like \L \U, including \Q and \E.
2215 when we get here, PL_bufptr is at the \
2217 case LEX_INTERPCASEMOD:
2219 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
2220 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
2222 /* handle \E or end of string */
2223 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
2227 if (PL_lex_casemods) {
2228 oldmod = PL_lex_casestack[--PL_lex_casemods];
2229 PL_lex_casestack[PL_lex_casemods] = '\0';
2231 if (PL_bufptr != PL_bufend && strchr("LUQ", oldmod)) {
2233 PL_lex_state = LEX_INTERPCONCAT;
2237 if (PL_bufptr != PL_bufend)
2239 PL_lex_state = LEX_INTERPCONCAT;
2243 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2244 "### Saw case modifier at '%s'\n", PL_bufptr); });
2246 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
2247 tmp = *s, *s = s[2], s[2] = tmp; /* misordered... */
2248 if (strchr("LU", *s) &&
2249 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U')))
2251 PL_lex_casestack[--PL_lex_casemods] = '\0';
2254 if (PL_lex_casemods > 10) {
2255 char* newlb = Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
2256 if (newlb != PL_lex_casestack) {
2258 PL_lex_casestack = newlb;
2261 PL_lex_casestack[PL_lex_casemods++] = *s;
2262 PL_lex_casestack[PL_lex_casemods] = '\0';
2263 PL_lex_state = LEX_INTERPCONCAT;
2264 PL_nextval[PL_nexttoke].ival = 0;
2267 PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
2269 PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
2271 PL_nextval[PL_nexttoke].ival = OP_LC;
2273 PL_nextval[PL_nexttoke].ival = OP_UC;
2275 PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
2277 Perl_croak(aTHX_ "panic: yylex");
2280 if (PL_lex_starts) {
2289 case LEX_INTERPPUSH:
2290 return sublex_push();
2292 case LEX_INTERPSTART:
2293 if (PL_bufptr == PL_bufend)
2294 return sublex_done();
2295 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2296 "### Interpolated variable at '%s'\n", PL_bufptr); });
2298 PL_lex_dojoin = (*PL_bufptr == '@');
2299 PL_lex_state = LEX_INTERPNORMAL;
2300 if (PL_lex_dojoin) {
2301 PL_nextval[PL_nexttoke].ival = 0;
2303 #ifdef USE_5005THREADS
2304 PL_nextval[PL_nexttoke].opval = newOP(OP_THREADSV, 0);
2305 PL_nextval[PL_nexttoke].opval->op_targ = find_threadsv("\"");
2306 force_next(PRIVATEREF);
2308 force_ident("\"", '$');
2309 #endif /* USE_5005THREADS */
2310 PL_nextval[PL_nexttoke].ival = 0;
2312 PL_nextval[PL_nexttoke].ival = 0;
2314 PL_nextval[PL_nexttoke].ival = OP_JOIN; /* emulate join($", ...) */
2317 if (PL_lex_starts++) {
2323 case LEX_INTERPENDMAYBE:
2324 if (intuit_more(PL_bufptr)) {
2325 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
2331 if (PL_lex_dojoin) {
2332 PL_lex_dojoin = FALSE;
2333 PL_lex_state = LEX_INTERPCONCAT;
2336 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
2337 && SvEVALED(PL_lex_repl))
2339 if (PL_bufptr != PL_bufend)
2340 Perl_croak(aTHX_ "Bad evalled substitution pattern");
2341 PL_lex_repl = Nullsv;
2344 case LEX_INTERPCONCAT:
2346 if (PL_lex_brackets)
2347 Perl_croak(aTHX_ "panic: INTERPCONCAT");
2349 if (PL_bufptr == PL_bufend)
2350 return sublex_done();
2352 if (SvIVX(PL_linestr) == '\'') {
2353 SV *sv = newSVsv(PL_linestr);
2356 else if ( PL_hints & HINT_NEW_RE )
2357 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
2358 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2362 s = scan_const(PL_bufptr);
2364 PL_lex_state = LEX_INTERPCASEMOD;
2366 PL_lex_state = LEX_INTERPSTART;
2369 if (s != PL_bufptr) {
2370 PL_nextval[PL_nexttoke] = yylval;
2373 if (PL_lex_starts++)
2383 PL_lex_state = LEX_NORMAL;
2384 s = scan_formline(PL_bufptr);
2385 if (!PL_lex_formbrack)
2391 PL_oldoldbufptr = PL_oldbufptr;
2394 PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at %s\n",
2395 exp_name[PL_expect], s);
2401 if (isIDFIRST_lazy_if(s,UTF))
2403 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
2406 goto fake_eof; /* emulate EOF on ^D or ^Z */
2411 if (PL_lex_brackets)
2412 yyerror("Missing right curly or square bracket");
2413 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2414 "### Tokener got EOF\n");
2418 if (s++ < PL_bufend)
2419 goto retry; /* ignore stray nulls */
2422 if (!PL_in_eval && !PL_preambled) {
2423 PL_preambled = TRUE;
2424 sv_setpv(PL_linestr,incl_perldb());
2425 if (SvCUR(PL_linestr))
2426 sv_catpv(PL_linestr,";");
2428 while(AvFILLp(PL_preambleav) >= 0) {
2429 SV *tmpsv = av_shift(PL_preambleav);
2430 sv_catsv(PL_linestr, tmpsv);
2431 sv_catpv(PL_linestr, ";");
2434 sv_free((SV*)PL_preambleav);
2435 PL_preambleav = NULL;
2437 if (PL_minus_n || PL_minus_p) {
2438 sv_catpv(PL_linestr, "LINE: while (<>) {");
2440 sv_catpv(PL_linestr,"chomp;");
2443 if (strchr("/'\"", *PL_splitstr)
2444 && strchr(PL_splitstr + 1, *PL_splitstr))
2445 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
2448 s = "'~#\200\1'"; /* surely one char is unused...*/
2449 while (s[1] && strchr(PL_splitstr, *s)) s++;
2451 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s%c",
2452 "q" + (delim == '\''), delim);
2453 for (s = PL_splitstr; *s; s++) {
2455 sv_catpvn(PL_linestr, "\\", 1);
2456 sv_catpvn(PL_linestr, s, 1);
2458 Perl_sv_catpvf(aTHX_ PL_linestr, "%c);", delim);
2462 sv_catpv(PL_linestr,"our @F=split(' ');");
2465 sv_catpv(PL_linestr, "\n");
2466 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2467 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2468 PL_last_lop = PL_last_uni = Nullch;
2469 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2470 SV *sv = NEWSV(85,0);
2472 sv_upgrade(sv, SVt_PVMG);
2473 sv_setsv(sv,PL_linestr);
2476 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2481 bof = PL_rsfp ? TRUE : FALSE;
2482 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == Nullch) {
2485 if (PL_preprocess && !PL_in_eval)
2486 (void)PerlProc_pclose(PL_rsfp);
2487 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
2488 PerlIO_clearerr(PL_rsfp);
2490 (void)PerlIO_close(PL_rsfp);
2492 PL_doextract = FALSE;
2494 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
2495 sv_setpv(PL_linestr,PL_minus_p ? ";}continue{print" : "");
2496 sv_catpv(PL_linestr,";}");
2497 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2498 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2499 PL_last_lop = PL_last_uni = Nullch;
2500 PL_minus_n = PL_minus_p = 0;
2503 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2504 PL_last_lop = PL_last_uni = Nullch;
2505 sv_setpv(PL_linestr,"");
2506 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
2508 /* if it looks like the start of a BOM, check if it in fact is */
2509 else if (bof && (!*s || *(U8*)s == 0xEF || *(U8*)s >= 0xFE)) {
2510 #ifdef PERLIO_IS_STDIO
2511 # ifdef __GNU_LIBRARY__
2512 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
2513 # define FTELL_FOR_PIPE_IS_BROKEN
2517 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
2518 # define FTELL_FOR_PIPE_IS_BROKEN
2523 #ifdef FTELL_FOR_PIPE_IS_BROKEN
2524 /* This loses the possibility to detect the bof
2525 * situation on perl -P when the libc5 is being used.
2526 * Workaround? Maybe attach some extra state to PL_rsfp?
2529 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
2531 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
2534 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2535 s = swallow_bom((U8*)s);
2539 if (*s == '#' && s[1] == '!' && instr(s,"perl"))
2540 PL_doextract = FALSE;
2542 /* Incest with pod. */
2543 if (*s == '=' && strnEQ(s, "=cut", 4)) {
2544 sv_setpv(PL_linestr, "");
2545 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2546 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2547 PL_last_lop = PL_last_uni = Nullch;
2548 PL_doextract = FALSE;
2552 } while (PL_doextract);
2553 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
2554 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2555 SV *sv = NEWSV(85,0);
2557 sv_upgrade(sv, SVt_PVMG);
2558 sv_setsv(sv,PL_linestr);
2561 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2563 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2564 PL_last_lop = PL_last_uni = Nullch;
2565 if (CopLINE(PL_curcop) == 1) {
2566 while (s < PL_bufend && isSPACE(*s))
2568 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
2572 if (*s == '#' && *(s+1) == '!')
2574 #ifdef ALTERNATE_SHEBANG
2576 static char as[] = ALTERNATE_SHEBANG;
2577 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
2578 d = s + (sizeof(as) - 1);
2580 #endif /* ALTERNATE_SHEBANG */
2589 while (*d && !isSPACE(*d))
2593 #ifdef ARG_ZERO_IS_SCRIPT
2594 if (ipathend > ipath) {
2596 * HP-UX (at least) sets argv[0] to the script name,
2597 * which makes $^X incorrect. And Digital UNIX and Linux,
2598 * at least, set argv[0] to the basename of the Perl
2599 * interpreter. So, having found "#!", we'll set it right.
2601 SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV)); /* $^X */
2602 assert(SvPOK(x) || SvGMAGICAL(x));
2603 if (sv_eq(x, CopFILESV(PL_curcop))) {
2604 sv_setpvn(x, ipath, ipathend - ipath);
2607 TAINT_NOT; /* $^X is always tainted, but that's OK */
2609 #endif /* ARG_ZERO_IS_SCRIPT */
2614 d = instr(s,"perl -");
2616 d = instr(s,"perl");
2618 /* avoid getting into infinite loops when shebang
2619 * line contains "Perl" rather than "perl" */
2621 for (d = ipathend-4; d >= ipath; --d) {
2622 if ((*d == 'p' || *d == 'P')
2623 && !ibcmp(d, "perl", 4))
2633 #ifdef ALTERNATE_SHEBANG
2635 * If the ALTERNATE_SHEBANG on this system starts with a
2636 * character that can be part of a Perl expression, then if
2637 * we see it but not "perl", we're probably looking at the
2638 * start of Perl code, not a request to hand off to some
2639 * other interpreter. Similarly, if "perl" is there, but
2640 * not in the first 'word' of the line, we assume the line
2641 * contains the start of the Perl program.
2643 if (d && *s != '#') {
2645 while (*c && !strchr("; \t\r\n\f\v#", *c))
2648 d = Nullch; /* "perl" not in first word; ignore */
2650 *s = '#'; /* Don't try to parse shebang line */
2652 #endif /* ALTERNATE_SHEBANG */
2653 #ifndef MACOS_TRADITIONAL
2658 !instr(s,"indir") &&
2659 instr(PL_origargv[0],"perl"))
2665 while (s < PL_bufend && isSPACE(*s))
2667 if (s < PL_bufend) {
2668 Newz(899,newargv,PL_origargc+3,char*);
2670 while (s < PL_bufend && !isSPACE(*s))
2673 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
2676 newargv = PL_origargv;
2678 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
2679 Perl_croak(aTHX_ "Can't exec %s", ipath);
2683 U32 oldpdb = PL_perldb;
2684 bool oldn = PL_minus_n;
2685 bool oldp = PL_minus_p;
2687 while (*d && !isSPACE(*d)) d++;
2688 while (SPACE_OR_TAB(*d)) d++;
2691 bool switches_done = PL_doswitches;
2693 if (*d == 'M' || *d == 'm') {
2695 while (*d && !isSPACE(*d)) d++;
2696 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
2699 d = moreswitches(d);
2701 if ((PERLDB_LINE && !oldpdb) ||
2702 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
2703 /* if we have already added "LINE: while (<>) {",
2704 we must not do it again */
2706 sv_setpv(PL_linestr, "");
2707 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2708 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2709 PL_last_lop = PL_last_uni = Nullch;
2710 PL_preambled = FALSE;
2712 (void)gv_fetchfile(PL_origfilename);
2715 if (PL_doswitches && !switches_done) {
2716 int argc = PL_origargc;
2717 char **argv = PL_origargv;
2720 } while (argc && argv[0][0] == '-' && argv[0][1]);
2721 init_argv_symbols(argc,argv);
2727 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2729 PL_lex_state = LEX_FORMLINE;
2734 #ifdef PERL_STRICT_CR
2735 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
2737 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
2739 case ' ': case '\t': case '\f': case 013:
2740 #ifdef MACOS_TRADITIONAL
2747 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
2748 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
2749 /* handle eval qq[#line 1 "foo"\n ...] */
2750 CopLINE_dec(PL_curcop);
2754 while (s < d && *s != '\n')
2758 else if (s > d) /* Found by Ilya: feed random input to Perl. */
2759 Perl_croak(aTHX_ "panic: input overflow");
2761 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2763 PL_lex_state = LEX_FORMLINE;
2773 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
2780 while (s < PL_bufend && SPACE_OR_TAB(*s))
2783 if (strnEQ(s,"=>",2)) {
2784 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
2785 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2786 "### Saw unary minus before =>, forcing word '%s'\n", s);
2788 OPERATOR('-'); /* unary minus */
2790 PL_last_uni = PL_oldbufptr;
2792 case 'r': ftst = OP_FTEREAD; break;
2793 case 'w': ftst = OP_FTEWRITE; break;
2794 case 'x': ftst = OP_FTEEXEC; break;
2795 case 'o': ftst = OP_FTEOWNED; break;
2796 case 'R': ftst = OP_FTRREAD; break;
2797 case 'W': ftst = OP_FTRWRITE; break;
2798 case 'X': ftst = OP_FTREXEC; break;
2799 case 'O': ftst = OP_FTROWNED; break;
2800 case 'e': ftst = OP_FTIS; break;
2801 case 'z': ftst = OP_FTZERO; break;
2802 case 's': ftst = OP_FTSIZE; break;
2803 case 'f': ftst = OP_FTFILE; break;
2804 case 'd': ftst = OP_FTDIR; break;
2805 case 'l': ftst = OP_FTLINK; break;
2806 case 'p': ftst = OP_FTPIPE; break;
2807 case 'S': ftst = OP_FTSOCK; break;
2808 case 'u': ftst = OP_FTSUID; break;
2809 case 'g': ftst = OP_FTSGID; break;
2810 case 'k': ftst = OP_FTSVTX; break;
2811 case 'b': ftst = OP_FTBLK; break;
2812 case 'c': ftst = OP_FTCHR; break;
2813 case 't': ftst = OP_FTTTY; break;
2814 case 'T': ftst = OP_FTTEXT; break;
2815 case 'B': ftst = OP_FTBINARY; break;
2816 case 'M': case 'A': case 'C':
2817 gv_fetchpv("\024",TRUE, SVt_PV);
2819 case 'M': ftst = OP_FTMTIME; break;
2820 case 'A': ftst = OP_FTATIME; break;
2821 case 'C': ftst = OP_FTCTIME; break;
2829 PL_last_lop_op = ftst;
2830 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2831 "### Saw file test %c\n", (int)ftst);
2836 /* Assume it was a minus followed by a one-letter named
2837 * subroutine call (or a -bareword), then. */
2838 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2839 "### %c looked like a file test but was not\n",
2848 if (PL_expect == XOPERATOR)
2853 else if (*s == '>') {
2856 if (isIDFIRST_lazy_if(s,UTF)) {
2857 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
2865 if (PL_expect == XOPERATOR)
2868 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2870 OPERATOR('-'); /* unary minus */
2877 if (PL_expect == XOPERATOR)
2882 if (PL_expect == XOPERATOR)
2885 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2891 if (PL_expect != XOPERATOR) {
2892 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
2893 PL_expect = XOPERATOR;
2894 force_ident(PL_tokenbuf, '*');
2907 if (PL_expect == XOPERATOR) {
2911 PL_tokenbuf[0] = '%';
2912 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
2913 if (!PL_tokenbuf[1]) {
2915 yyerror("Final % should be \\% or %name");
2918 PL_pending_ident = '%';
2937 switch (PL_expect) {
2940 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
2942 PL_bufptr = s; /* update in case we back off */
2948 PL_expect = XTERMBLOCK;
2952 while (isIDFIRST_lazy_if(s,UTF)) {
2953 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
2954 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
2955 if (tmp < 0) tmp = -tmp;
2970 d = scan_str(d,TRUE,TRUE);
2972 /* MUST advance bufptr here to avoid bogus
2973 "at end of line" context messages from yyerror().
2975 PL_bufptr = s + len;
2976 yyerror("Unterminated attribute parameter in attribute list");
2979 return 0; /* EOF indicator */
2983 SV *sv = newSVpvn(s, len);
2984 sv_catsv(sv, PL_lex_stuff);
2985 attrs = append_elem(OP_LIST, attrs,
2986 newSVOP(OP_CONST, 0, sv));
2987 SvREFCNT_dec(PL_lex_stuff);
2988 PL_lex_stuff = Nullsv;
2991 if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
2992 CvLVALUE_on(PL_compcv);
2993 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
2994 CvLOCKED_on(PL_compcv);
2995 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
2996 CvMETHOD_on(PL_compcv);
2998 else if (PL_in_my == KEY_our && len == 6 && strnEQ(s, "unique", len))
2999 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
3001 /* After we've set the flags, it could be argued that
3002 we don't need to do the attributes.pm-based setting
3003 process, and shouldn't bother appending recognized
3004 flags. To experiment with that, uncomment the
3005 following "else": */
3007 attrs = append_elem(OP_LIST, attrs,
3008 newSVOP(OP_CONST, 0,
3012 if (*s == ':' && s[1] != ':')
3015 break; /* require real whitespace or :'s */
3017 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3018 if (*s != ';' && *s != tmp && (tmp != '=' || *s != ')')) {
3019 char q = ((*s == '\'') ? '"' : '\'');
3020 /* If here for an expression, and parsed no attrs, back off. */
3021 if (tmp == '=' && !attrs) {
3025 /* MUST advance bufptr here to avoid bogus "at end of line"
3026 context messages from yyerror().
3030 yyerror("Unterminated attribute list");
3032 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3040 PL_nextval[PL_nexttoke].opval = attrs;
3048 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3049 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3065 if (PL_lex_brackets <= 0)
3066 yyerror("Unmatched right square bracket");
3069 if (PL_lex_state == LEX_INTERPNORMAL) {
3070 if (PL_lex_brackets == 0) {
3071 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3072 PL_lex_state = LEX_INTERPEND;
3079 if (PL_lex_brackets > 100) {
3080 char* newlb = Renew(PL_lex_brackstack, PL_lex_brackets + 1, char);
3081 if (newlb != PL_lex_brackstack) {
3083 PL_lex_brackstack = newlb;
3086 switch (PL_expect) {
3088 if (PL_lex_formbrack) {
3092 if (PL_oldoldbufptr == PL_last_lop)
3093 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3095 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3096 OPERATOR(HASHBRACK);
3098 while (s < PL_bufend && SPACE_OR_TAB(*s))
3101 PL_tokenbuf[0] = '\0';
3102 if (d < PL_bufend && *d == '-') {
3103 PL_tokenbuf[0] = '-';
3105 while (d < PL_bufend && SPACE_OR_TAB(*d))
3108 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3109 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3111 while (d < PL_bufend && SPACE_OR_TAB(*d))
3114 char minus = (PL_tokenbuf[0] == '-');
3115 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3123 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3128 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3133 if (PL_oldoldbufptr == PL_last_lop)
3134 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3136 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3139 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
3141 /* This hack is to get the ${} in the message. */
3143 yyerror("syntax error");
3146 OPERATOR(HASHBRACK);
3148 /* This hack serves to disambiguate a pair of curlies
3149 * as being a block or an anon hash. Normally, expectation
3150 * determines that, but in cases where we're not in a
3151 * position to expect anything in particular (like inside
3152 * eval"") we have to resolve the ambiguity. This code
3153 * covers the case where the first term in the curlies is a
3154 * quoted string. Most other cases need to be explicitly
3155 * disambiguated by prepending a `+' before the opening
3156 * curly in order to force resolution as an anon hash.
3158 * XXX should probably propagate the outer expectation
3159 * into eval"" to rely less on this hack, but that could
3160 * potentially break current behavior of eval"".
3164 if (*s == '\'' || *s == '"' || *s == '`') {
3165 /* common case: get past first string, handling escapes */
3166 for (t++; t < PL_bufend && *t != *s;)
3167 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3171 else if (*s == 'q') {
3174 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3178 char open, close, term;
3181 while (t < PL_bufend && isSPACE(*t))
3185 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3189 for (t++; t < PL_bufend; t++) {
3190 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3192 else if (*t == open)
3196 for (t++; t < PL_bufend; t++) {
3197 if (*t == '\\' && t+1 < PL_bufend)
3199 else if (*t == close && --brackets <= 0)
3201 else if (*t == open)
3207 else if (isALNUM_lazy_if(t,UTF)) {
3209 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3212 while (t < PL_bufend && isSPACE(*t))
3214 /* if comma follows first term, call it an anon hash */
3215 /* XXX it could be a comma expression with loop modifiers */
3216 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3217 || (*t == '=' && t[1] == '>')))
3218 OPERATOR(HASHBRACK);
3219 if (PL_expect == XREF)
3222 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3228 yylval.ival = CopLINE(PL_curcop);
3229 if (isSPACE(*s) || *s == '#')
3230 PL_copline = NOLINE; /* invalidate current command line number */
3235 if (PL_lex_brackets <= 0)
3236 yyerror("Unmatched right curly bracket");
3238 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3239 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3240 PL_lex_formbrack = 0;
3241 if (PL_lex_state == LEX_INTERPNORMAL) {
3242 if (PL_lex_brackets == 0) {
3243 if (PL_expect & XFAKEBRACK) {
3244 PL_expect &= XENUMMASK;
3245 PL_lex_state = LEX_INTERPEND;
3247 return yylex(); /* ignore fake brackets */
3249 if (*s == '-' && s[1] == '>')
3250 PL_lex_state = LEX_INTERPENDMAYBE;
3251 else if (*s != '[' && *s != '{')
3252 PL_lex_state = LEX_INTERPEND;
3255 if (PL_expect & XFAKEBRACK) {
3256 PL_expect &= XENUMMASK;
3258 return yylex(); /* ignore fake brackets */
3268 if (PL_expect == XOPERATOR) {
3269 if (ckWARN(WARN_SEMICOLON)
3270 && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
3272 CopLINE_dec(PL_curcop);
3273 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3274 CopLINE_inc(PL_curcop);
3279 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3281 PL_expect = XOPERATOR;
3282 force_ident(PL_tokenbuf, '&');
3286 yylval.ival = (OPpENTERSUB_AMPER<<8);
3305 if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
3306 Perl_warner(aTHX_ WARN_SYNTAX, "Reversed %c= operator",(int)tmp);
3308 if (PL_expect == XSTATE && isALPHA(tmp) &&
3309 (s == PL_linestart+1 || s[-2] == '\n') )
3311 if (PL_in_eval && !PL_rsfp) {
3316 if (strnEQ(s,"=cut",4)) {
3330 PL_doextract = TRUE;
3333 if (PL_lex_brackets < PL_lex_formbrack) {
3335 #ifdef PERL_STRICT_CR
3336 for (t = s; SPACE_OR_TAB(*t); t++) ;
3338 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3340 if (*t == '\n' || *t == '#') {
3358 if (PL_expect != XOPERATOR) {
3359 if (s[1] != '<' && !strchr(s,'>'))
3362 s = scan_heredoc(s);
3364 s = scan_inputsymbol(s);
3365 TERM(sublex_start());
3370 SHop(OP_LEFT_SHIFT);
3384 SHop(OP_RIGHT_SHIFT);
3393 if (PL_expect == XOPERATOR) {
3394 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3397 return ','; /* grandfather non-comma-format format */
3401 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3402 PL_tokenbuf[0] = '@';
3403 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3404 sizeof PL_tokenbuf - 1, FALSE);
3405 if (PL_expect == XOPERATOR)
3406 no_op("Array length", s);
3407 if (!PL_tokenbuf[1])
3409 PL_expect = XOPERATOR;
3410 PL_pending_ident = '#';
3414 PL_tokenbuf[0] = '$';
3415 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3416 sizeof PL_tokenbuf - 1, FALSE);
3417 if (PL_expect == XOPERATOR)
3419 if (!PL_tokenbuf[1]) {
3421 yyerror("Final $ should be \\$ or $name");
3425 /* This kludge not intended to be bulletproof. */
3426 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3427 yylval.opval = newSVOP(OP_CONST, 0,
3428 newSViv(PL_compiling.cop_arybase));
3429 yylval.opval->op_private = OPpCONST_ARYBASE;
3435 if (PL_lex_state == LEX_NORMAL)
3438 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3441 PL_tokenbuf[0] = '@';
3442 if (ckWARN(WARN_SYNTAX)) {
3444 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3447 PL_bufptr = skipspace(PL_bufptr);
3448 while (t < PL_bufend && *t != ']')
3450 Perl_warner(aTHX_ WARN_SYNTAX,
3451 "Multidimensional syntax %.*s not supported",
3452 (t - PL_bufptr) + 1, PL_bufptr);
3456 else if (*s == '{') {
3457 PL_tokenbuf[0] = '%';
3458 if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
3459 (t = strchr(s, '}')) && (t = strchr(t, '=')))
3461 char tmpbuf[sizeof PL_tokenbuf];
3463 for (t++; isSPACE(*t); t++) ;
3464 if (isIDFIRST_lazy_if(t,UTF)) {
3465 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3466 for (; isSPACE(*t); t++) ;
3467 if (*t == ';' && get_cv(tmpbuf, FALSE))
3468 Perl_warner(aTHX_ WARN_SYNTAX,
3469 "You need to quote \"%s\"", tmpbuf);
3475 PL_expect = XOPERATOR;
3476 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3477 bool islop = (PL_last_lop == PL_oldoldbufptr);
3478 if (!islop || PL_last_lop_op == OP_GREPSTART)
3479 PL_expect = XOPERATOR;
3480 else if (strchr("$@\"'`q", *s))
3481 PL_expect = XTERM; /* e.g. print $fh "foo" */
3482 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3483 PL_expect = XTERM; /* e.g. print $fh &sub */
3484 else if (isIDFIRST_lazy_if(s,UTF)) {
3485 char tmpbuf[sizeof PL_tokenbuf];
3486 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3487 if ((tmp = keyword(tmpbuf, len))) {
3488 /* binary operators exclude handle interpretations */
3500 PL_expect = XTERM; /* e.g. print $fh length() */
3505 GV *gv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
3506 if (gv && GvCVu(gv))
3507 PL_expect = XTERM; /* e.g. print $fh subr() */
3510 else if (isDIGIT(*s))
3511 PL_expect = XTERM; /* e.g. print $fh 3 */
3512 else if (*s == '.' && isDIGIT(s[1]))
3513 PL_expect = XTERM; /* e.g. print $fh .3 */
3514 else if (strchr("/?-+", *s) && !isSPACE(s[1]) && s[1] != '=')
3515 PL_expect = XTERM; /* e.g. print $fh -1 */
3516 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3517 PL_expect = XTERM; /* print $fh <<"EOF" */
3519 PL_pending_ident = '$';
3523 if (PL_expect == XOPERATOR)
3525 PL_tokenbuf[0] = '@';
3526 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3527 if (!PL_tokenbuf[1]) {
3529 yyerror("Final @ should be \\@ or @name");
3532 if (PL_lex_state == LEX_NORMAL)
3534 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3536 PL_tokenbuf[0] = '%';
3538 /* Warn about @ where they meant $. */
3539 if (ckWARN(WARN_SYNTAX)) {
3540 if (*s == '[' || *s == '{') {
3542 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3544 if (*t == '}' || *t == ']') {
3546 PL_bufptr = skipspace(PL_bufptr);
3547 Perl_warner(aTHX_ WARN_SYNTAX,
3548 "Scalar value %.*s better written as $%.*s",
3549 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3554 PL_pending_ident = '@';
3557 case '/': /* may either be division or pattern */
3558 case '?': /* may either be conditional or pattern */
3559 if (PL_expect != XOPERATOR) {
3560 /* Disable warning on "study /blah/" */
3561 if (PL_oldoldbufptr == PL_last_uni
3562 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3563 || memNE(PL_last_uni, "study", 5)
3564 || isALNUM_lazy_if(PL_last_uni+5,UTF)))
3566 s = scan_pat(s,OP_MATCH);
3567 TERM(sublex_start());
3575 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3576 #ifdef PERL_STRICT_CR
3579 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3581 && (s == PL_linestart || s[-1] == '\n') )
3583 PL_lex_formbrack = 0;
3587 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3593 yylval.ival = OPf_SPECIAL;
3599 if (PL_expect != XOPERATOR)
3604 case '0': case '1': case '2': case '3': case '4':
3605 case '5': case '6': case '7': case '8': case '9':
3606 s = scan_num(s, &yylval);
3607 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3608 "### Saw number in '%s'\n", s);
3610 if (PL_expect == XOPERATOR)
3615 s = scan_str(s,FALSE,FALSE);
3616 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3617 "### Saw string before '%s'\n", s);
3619 if (PL_expect == XOPERATOR) {
3620 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3623 return ','; /* grandfather non-comma-format format */
3629 missingterm((char*)0);
3630 yylval.ival = OP_CONST;
3631 TERM(sublex_start());
3634 s = scan_str(s,FALSE,FALSE);
3635 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3636 "### Saw string before '%s'\n", s);
3638 if (PL_expect == XOPERATOR) {
3639 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3642 return ','; /* grandfather non-comma-format format */
3648 missingterm((char*)0);
3649 yylval.ival = OP_CONST;
3650 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3651 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
3652 yylval.ival = OP_STRINGIFY;
3656 TERM(sublex_start());
3659 s = scan_str(s,FALSE,FALSE);
3660 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3661 "### Saw backtick string before '%s'\n", s);
3663 if (PL_expect == XOPERATOR)
3664 no_op("Backticks",s);
3666 missingterm((char*)0);
3667 yylval.ival = OP_BACKTICK;
3669 TERM(sublex_start());
3673 if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
3674 Perl_warner(aTHX_ WARN_SYNTAX,"Can't use \\%c to mean $%c in expression",
3676 if (PL_expect == XOPERATOR)
3677 no_op("Backslash",s);
3681 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3685 while (isDIGIT(*start) || *start == '_')
3687 if (*start == '.' && isDIGIT(start[1])) {
3688 s = scan_num(s, &yylval);
3691 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3692 else if (!isALPHA(*start) && (PL_expect == XTERM || PL_expect == XREF || PL_expect == XSTATE)) {
3696 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3699 s = scan_num(s, &yylval);
3706 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
3745 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3747 /* Some keywords can be followed by any delimiter, including ':' */
3748 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
3749 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
3750 (PL_tokenbuf[0] == 'q' &&
3751 strchr("qwxr", PL_tokenbuf[1])))));
3753 /* x::* is just a word, unless x is "CORE" */
3754 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
3758 while (d < PL_bufend && isSPACE(*d))
3759 d++; /* no comments skipped here, or s### is misparsed */
3761 /* Is this a label? */
3762 if (!tmp && PL_expect == XSTATE
3763 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
3765 yylval.pval = savepv(PL_tokenbuf);
3770 /* Check for keywords */
3771 tmp = keyword(PL_tokenbuf, len);
3773 /* Is this a word before a => operator? */
3774 if (*d == '=' && d[1] == '>') {
3776 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf,0));
3777 yylval.opval->op_private = OPpCONST_BARE;
3778 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
3779 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3783 if (tmp < 0) { /* second-class keyword? */
3784 GV *ogv = Nullgv; /* override (winner) */
3785 GV *hgv = Nullgv; /* hidden (loser) */
3786 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
3788 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
3791 if (GvIMPORTED_CV(gv))
3793 else if (! CvMETHOD(cv))
3797 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
3798 (gv = *gvp) != (GV*)&PL_sv_undef &&
3799 GvCVu(gv) && GvIMPORTED_CV(gv))
3805 tmp = 0; /* overridden by import or by GLOBAL */
3808 && -tmp==KEY_lock /* XXX generalizable kludge */
3810 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
3812 tmp = 0; /* any sub overrides "weak" keyword */
3814 else { /* no override */
3816 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
3817 Perl_warner(aTHX_ WARN_MISC,
3818 "dump() better written as CORE::dump()");
3822 if (ckWARN(WARN_AMBIGUOUS) && hgv
3823 && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
3824 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3825 "Ambiguous call resolved as CORE::%s(), %s",
3826 GvENAME(hgv), "qualify as such or use &");
3833 default: /* not a keyword */
3837 char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
3839 /* Get the rest if it looks like a package qualifier */
3841 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
3843 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
3846 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
3847 *s == '\'' ? "'" : "::");
3852 if (PL_expect == XOPERATOR) {
3853 if (PL_bufptr == PL_linestart) {
3854 CopLINE_dec(PL_curcop);
3855 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3856 CopLINE_inc(PL_curcop);
3859 no_op("Bareword",s);
3862 /* Look for a subroutine with this name in current package,
3863 unless name is "Foo::", in which case Foo is a bearword
3864 (and a package name). */
3867 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
3869 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
3870 Perl_warner(aTHX_ WARN_BAREWORD,
3871 "Bareword \"%s\" refers to nonexistent package",
3874 PL_tokenbuf[len] = '\0';
3881 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
3884 /* if we saw a global override before, get the right name */
3887 sv = newSVpvn("CORE::GLOBAL::",14);
3888 sv_catpv(sv,PL_tokenbuf);
3891 sv = newSVpv(PL_tokenbuf,0);
3893 /* Presume this is going to be a bareword of some sort. */
3896 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3897 yylval.opval->op_private = OPpCONST_BARE;
3899 /* And if "Foo::", then that's what it certainly is. */
3904 /* See if it's the indirect object for a list operator. */
3906 if (PL_oldoldbufptr &&
3907 PL_oldoldbufptr < PL_bufptr &&
3908 (PL_oldoldbufptr == PL_last_lop
3909 || PL_oldoldbufptr == PL_last_uni) &&
3910 /* NO SKIPSPACE BEFORE HERE! */
3911 (PL_expect == XREF ||
3912 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
3914 bool immediate_paren = *s == '(';
3916 /* (Now we can afford to cross potential line boundary.) */
3919 /* Two barewords in a row may indicate method call. */
3921 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
3924 /* If not a declared subroutine, it's an indirect object. */
3925 /* (But it's an indir obj regardless for sort.) */
3927 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
3928 ((!gv || !GvCVu(gv)) &&
3929 (PL_last_lop_op != OP_MAPSTART &&
3930 PL_last_lop_op != OP_GREPSTART))))
3932 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
3937 PL_expect = XOPERATOR;
3940 /* Is this a word before a => operator? */
3941 if (*s == '=' && s[1] == '>' && !pkgname) {
3943 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
3944 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
3945 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3949 /* If followed by a paren, it's certainly a subroutine. */
3952 if (gv && GvCVu(gv)) {
3953 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
3954 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
3959 PL_nextval[PL_nexttoke].opval = yylval.opval;
3960 PL_expect = XOPERATOR;
3966 /* If followed by var or block, call it a method (unless sub) */
3968 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
3969 PL_last_lop = PL_oldbufptr;
3970 PL_last_lop_op = OP_METHOD;
3974 /* If followed by a bareword, see if it looks like indir obj. */
3976 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp = intuit_method(s,gv)))
3979 /* Not a method, so call it a subroutine (if defined) */
3981 if (gv && GvCVu(gv)) {
3983 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
3984 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3985 "Ambiguous use of -%s resolved as -&%s()",
3986 PL_tokenbuf, PL_tokenbuf);
3987 /* Check for a constant sub */
3989 if ((sv = cv_const_sv(cv))) {
3991 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
3992 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
3993 yylval.opval->op_private = 0;
3997 /* Resolve to GV now. */
3998 op_free(yylval.opval);
3999 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4000 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4001 PL_last_lop = PL_oldbufptr;
4002 PL_last_lop_op = OP_ENTERSUB;
4003 /* Is there a prototype? */
4006 char *proto = SvPV((SV*)cv, len);
4009 if (strEQ(proto, "$"))
4011 if (*proto == '&' && *s == '{') {
4012 sv_setpv(PL_subname,"__ANON__");
4016 PL_nextval[PL_nexttoke].opval = yylval.opval;
4022 /* Call it a bare word */
4024 if (PL_hints & HINT_STRICT_SUBS)
4025 yylval.opval->op_private |= OPpCONST_STRICT;
4028 if (ckWARN(WARN_RESERVED)) {
4029 if (lastchar != '-') {
4030 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4031 if (!*d && strNE(PL_tokenbuf,"main"))
4032 Perl_warner(aTHX_ WARN_RESERVED, PL_warn_reserved,
4039 if (lastchar && strchr("*%&", lastchar) && ckWARN_d(WARN_AMBIGUOUS)) {
4040 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4041 "Operator or semicolon missing before %c%s",
4042 lastchar, PL_tokenbuf);
4043 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4044 "Ambiguous use of %c resolved as operator %c",
4045 lastchar, lastchar);
4051 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4052 newSVpv(CopFILE(PL_curcop),0));
4056 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4057 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4060 case KEY___PACKAGE__:
4061 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4063 ? newSVsv(PL_curstname)
4072 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4073 char *pname = "main";
4074 if (PL_tokenbuf[2] == 'D')
4075 pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
4076 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4079 GvIOp(gv) = newIO();
4080 IoIFP(GvIOp(gv)) = PL_rsfp;
4081 #if defined(HAS_FCNTL) && defined(F_SETFD)
4083 int fd = PerlIO_fileno(PL_rsfp);
4084 fcntl(fd,F_SETFD,fd >= 3);
4087 /* Mark this internal pseudo-handle as clean */
4088 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4090 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4091 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4092 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4094 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4095 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4096 /* if the script was opened in binmode, we need to revert
4097 * it to text mode for compatibility; but only iff it has CRs
4098 * XXX this is a questionable hack at best. */
4099 if (PL_bufend-PL_bufptr > 2
4100 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4103 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4104 loc = PerlIO_tell(PL_rsfp);
4105 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4108 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
4110 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4111 #endif /* NETWARE */
4112 #ifdef PERLIO_IS_STDIO /* really? */
4113 # if defined(__BORLANDC__)
4114 /* XXX see note in do_binmode() */
4115 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
4119 PerlIO_seek(PL_rsfp, loc, 0);
4123 #ifdef PERLIO_LAYERS
4124 if (UTF && !IN_BYTES)
4125 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4138 if (PL_expect == XSTATE) {
4145 if (*s == ':' && s[1] == ':') {
4148 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4149 if (!(tmp = keyword(PL_tokenbuf, len)))
4150 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4164 LOP(OP_ACCEPT,XTERM);
4170 LOP(OP_ATAN2,XTERM);
4176 LOP(OP_BINMODE,XTERM);
4179 LOP(OP_BLESS,XTERM);
4188 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4205 if (!PL_cryptseen) {
4206 PL_cryptseen = TRUE;
4210 LOP(OP_CRYPT,XTERM);
4213 LOP(OP_CHMOD,XTERM);
4216 LOP(OP_CHOWN,XTERM);
4219 LOP(OP_CONNECT,XTERM);
4235 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4239 PL_hints |= HINT_BLOCK_SCOPE;
4249 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4250 LOP(OP_DBMOPEN,XTERM);
4256 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4263 yylval.ival = CopLINE(PL_curcop);
4277 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4278 UNIBRACK(OP_ENTEREVAL);
4293 case KEY_endhostent:
4299 case KEY_endservent:
4302 case KEY_endprotoent:
4313 yylval.ival = CopLINE(PL_curcop);
4315 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4317 if ((PL_bufend - p) >= 3 &&
4318 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4320 else if ((PL_bufend - p) >= 4 &&
4321 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
4324 if (isIDFIRST_lazy_if(p,UTF)) {
4325 p = scan_ident(p, PL_bufend,
4326 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4330 Perl_croak(aTHX_ "Missing $ on loop variable");
4335 LOP(OP_FORMLINE,XTERM);