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(pTHXo_ void *f);
31 #ifndef PERL_NO_UTF16_FILTER
32 static I32 utf16_textfilter(pTHXo_ int idx, SV *sv, int maxlen);
33 static I32 utf16rev_textfilter(pTHXo_ int idx, SV *sv, int maxlen);
36 #define XFAKEBRACK 128
39 /*#define UTF (SvUTF8(PL_linestr) && !(PL_hints & HINT_BYTE))*/
40 #define UTF (PL_hints & HINT_UTF8)
42 /* In variables name $^X, these are the legal values for X.
43 * 1999-02-27 mjd-perl-patch@plover.com */
44 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
46 /* On MacOS, respect nonbreaking spaces */
47 #ifdef MACOS_TRADITIONAL
48 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
50 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
53 /* LEX_* are values for PL_lex_state, the state of the lexer.
54 * They are arranged oddly so that the guard on the switch statement
55 * can get by with a single comparison (if the compiler is smart enough).
58 /* #define LEX_NOTPARSING 11 is done in perl.h. */
61 #define LEX_INTERPNORMAL 9
62 #define LEX_INTERPCASEMOD 8
63 #define LEX_INTERPPUSH 7
64 #define LEX_INTERPSTART 6
65 #define LEX_INTERPEND 5
66 #define LEX_INTERPENDMAYBE 4
67 #define LEX_INTERPCONCAT 3
68 #define LEX_INTERPCONST 2
69 #define LEX_FORMLINE 1
70 #define LEX_KNOWNEXT 0
78 # define YYMAXLEVEL 100
80 YYSTYPE* yylval_pointer[YYMAXLEVEL];
81 int* yychar_pointer[YYMAXLEVEL];
85 # define yylval (*yylval_pointer[yyactlevel])
86 # define yychar (*yychar_pointer[yyactlevel])
87 # define PERL_YYLEX_PARAM yylval_pointer[yyactlevel],yychar_pointer[yyactlevel]
89 # define yylex() Perl_yylex_r(aTHX_ yylval_pointer[yyactlevel],yychar_pointer[yyactlevel])
94 /* CLINE is a macro that ensures PL_copline has a sane value */
99 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
102 * Convenience functions to return different tokens and prime the
103 * lexer for the next token. They all take an argument.
105 * TOKEN : generic token (used for '(', DOLSHARP, etc)
106 * OPERATOR : generic operator
107 * AOPERATOR : assignment operator
108 * PREBLOCK : beginning the block after an if, while, foreach, ...
109 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
110 * PREREF : *EXPR where EXPR is not a simple identifier
111 * TERM : expression term
112 * LOOPX : loop exiting command (goto, last, dump, etc)
113 * FTST : file test operator
114 * FUN0 : zero-argument function
115 * FUN1 : not used, except for not, which isn't a UNIOP
116 * BOop : bitwise or or xor
118 * SHop : shift operator
119 * PWop : power operator
120 * PMop : pattern-matching operator
121 * Aop : addition-level operator
122 * Mop : multiplication-level operator
123 * Eop : equality-testing operator
124 * Rop : relational operator <= != gt
126 * Also see LOP and lop() below.
129 #define TOKEN(retval) return (PL_bufptr = s,(int)retval)
130 #define OPERATOR(retval) return (PL_expect = XTERM,PL_bufptr = s,(int)retval)
131 #define AOPERATOR(retval) return ao((PL_expect = XTERM,PL_bufptr = s,(int)retval))
132 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s,(int)retval)
133 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s,(int)retval)
134 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s,(int)retval)
135 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR,PL_bufptr = s,(int)retval)
136 #define LOOPX(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LOOPEX)
137 #define FTST(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)UNIOP)
138 #define FUN0(f) return(yylval.ival = f,PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC0)
139 #define FUN1(f) return(yylval.ival = f,PL_expect = XOPERATOR,PL_bufptr = s,(int)FUNC1)
140 #define BOop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)BITOROP))
141 #define BAop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)BITANDOP))
142 #define SHop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)SHIFTOP))
143 #define PWop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)POWOP))
144 #define PMop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)MATCHOP)
145 #define Aop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)ADDOP))
146 #define Mop(f) return ao((yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)MULOP))
147 #define Eop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)EQOP)
148 #define Rop(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)RELOP)
150 /* This bit of chicanery makes a unary function followed by
151 * a parenthesis into a function with one argument, highest precedence.
153 #define UNI(f) return(yylval.ival = f, \
156 PL_last_uni = PL_oldbufptr, \
157 PL_last_lop_op = f, \
158 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
160 #define UNIBRACK(f) return(yylval.ival = f, \
162 PL_last_uni = PL_oldbufptr, \
163 (*s == '(' || (s = skipspace(s), *s == '(') ? (int)FUNC1 : (int)UNIOP) )
165 /* grandfather return to old style */
166 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
171 * This subroutine detects &&= and ||= and turns an ANDAND or OROR
172 * into an OP_ANDASSIGN or OP_ORASSIGN
176 S_ao(pTHX_ int toketype)
178 if (*PL_bufptr == '=') {
180 if (toketype == ANDAND)
181 yylval.ival = OP_ANDASSIGN;
182 else if (toketype == OROR)
183 yylval.ival = OP_ORASSIGN;
191 * When Perl expects an operator and finds something else, no_op
192 * prints the warning. It always prints "<something> found where
193 * operator expected. It prints "Missing semicolon on previous line?"
194 * if the surprise occurs at the start of the line. "do you need to
195 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
196 * where the compiler doesn't know if foo is a method call or a function.
197 * It prints "Missing operator before end of line" if there's nothing
198 * after the missing operator, or "... before <...>" if there is something
199 * after the missing operator.
203 S_no_op(pTHX_ char *what, char *s)
205 char *oldbp = PL_bufptr;
206 bool is_first = (PL_oldbufptr == PL_linestart);
212 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
214 Perl_warn(aTHX_ "\t(Missing semicolon on previous line?)\n");
215 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
217 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
218 if (t < PL_bufptr && isSPACE(*t))
219 Perl_warn(aTHX_ "\t(Do you need to predeclare %.*s?)\n",
220 t - PL_oldoldbufptr, PL_oldoldbufptr);
224 Perl_warn(aTHX_ "\t(Missing operator before %.*s?)\n", s - oldbp, oldbp);
231 * Complain about missing quote/regexp/heredoc terminator.
232 * If it's called with (char *)NULL then it cauterizes the line buffer.
233 * If we're in a delimited string and the delimiter is a control
234 * character, it's reformatted into a two-char sequence like ^C.
239 S_missingterm(pTHX_ char *s)
244 char *nl = strrchr(s,'\n');
250 iscntrl(PL_multi_close)
252 PL_multi_close < 32 || PL_multi_close == 127
256 tmpbuf[1] = toCTRL(PL_multi_close);
262 *tmpbuf = PL_multi_close;
266 q = strchr(s,'"') ? '\'' : '"';
267 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
275 Perl_deprecate(pTHX_ char *s)
277 if (ckWARN(WARN_DEPRECATED))
278 Perl_warner(aTHX_ WARN_DEPRECATED, "Use of %s is deprecated", s);
283 * Deprecate a comma-less variable list.
289 deprecate("comma-less variable list");
293 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
294 * utf16-to-utf8-reversed.
297 #ifdef PERL_CR_FILTER
301 register char *s = SvPVX(sv);
302 register char *e = s + SvCUR(sv);
303 /* outer loop optimized to do nothing if there are no CR-LFs */
305 if (*s++ == '\r' && *s == '\n') {
306 /* hit a CR-LF, need to copy the rest */
307 register char *d = s - 1;
310 if (*s == '\r' && s[1] == '\n')
321 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
323 I32 count = FILTER_READ(idx+1, sv, maxlen);
324 if (count > 0 && !maxlen)
332 * Initialize variables. Uses the Perl save_stack to save its state (for
333 * recursive calls to the parser).
337 Perl_lex_start(pTHX_ SV *line)
342 SAVEI32(PL_lex_dojoin);
343 SAVEI32(PL_lex_brackets);
344 SAVEI32(PL_lex_casemods);
345 SAVEI32(PL_lex_starts);
346 SAVEI32(PL_lex_state);
347 SAVEVPTR(PL_lex_inpat);
348 SAVEI32(PL_lex_inwhat);
349 if (PL_lex_state == LEX_KNOWNEXT) {
350 I32 toke = PL_nexttoke;
351 while (--toke >= 0) {
352 SAVEI32(PL_nexttype[toke]);
353 SAVEVPTR(PL_nextval[toke]);
355 SAVEI32(PL_nexttoke);
357 SAVECOPLINE(PL_curcop);
360 SAVEPPTR(PL_oldbufptr);
361 SAVEPPTR(PL_oldoldbufptr);
362 SAVEPPTR(PL_linestart);
363 SAVESPTR(PL_linestr);
364 SAVEPPTR(PL_lex_brackstack);
365 SAVEPPTR(PL_lex_casestack);
366 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
367 SAVESPTR(PL_lex_stuff);
368 SAVEI32(PL_lex_defer);
369 SAVEI32(PL_sublex_info.sub_inwhat);
370 SAVESPTR(PL_lex_repl);
372 SAVEINT(PL_lex_expect);
374 PL_lex_state = LEX_NORMAL;
378 New(899, PL_lex_brackstack, 120, char);
379 New(899, PL_lex_casestack, 12, char);
380 SAVEFREEPV(PL_lex_brackstack);
381 SAVEFREEPV(PL_lex_casestack);
383 *PL_lex_casestack = '\0';
386 PL_lex_stuff = Nullsv;
387 PL_lex_repl = Nullsv;
391 PL_sublex_info.sub_inwhat = 0;
393 if (SvREADONLY(PL_linestr))
394 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
395 s = SvPV(PL_linestr, len);
396 if (len && s[len-1] != ';') {
397 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
398 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
399 sv_catpvn(PL_linestr, "\n;", 2);
401 SvTEMP_off(PL_linestr);
402 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
403 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
405 PL_rs = newSVpvn("\n", 1);
411 * Finalizer for lexing operations. Must be called when the parser is
412 * done with the lexer.
418 PL_doextract = FALSE;
423 * This subroutine has nothing to do with tilting, whether at windmills
424 * or pinball tables. Its name is short for "increment line". It
425 * increments the current line number in CopLINE(PL_curcop) and checks
426 * to see whether the line starts with a comment of the form
427 * # line 500 "foo.pm"
428 * If so, it sets the current line number and file to the values in the comment.
432 S_incline(pTHX_ char *s)
439 CopLINE_inc(PL_curcop);
442 while (SPACE_OR_TAB(*s)) s++;
443 if (strnEQ(s, "line", 4))
447 if (*s == ' ' || *s == '\t')
451 while (SPACE_OR_TAB(*s)) s++;
457 while (SPACE_OR_TAB(*s))
459 if (*s == '"' && (t = strchr(s+1, '"'))) {
464 for (t = s; !isSPACE(*t); t++) ;
467 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
469 if (*e != '\n' && *e != '\0')
470 return; /* false alarm */
476 Safefree(CopFILE(PL_curcop));
478 SvREFCNT_dec(CopFILEGV(PL_curcop));
480 CopFILE_set(PL_curcop, s);
483 CopLINE_set(PL_curcop, atoi(n)-1);
488 * Called to gobble the appropriate amount and type of whitespace.
489 * Skips comments as well.
493 S_skipspace(pTHX_ register char *s)
495 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
496 while (s < PL_bufend && SPACE_OR_TAB(*s))
502 SSize_t oldprevlen, oldoldprevlen;
503 SSize_t oldloplen, oldunilen;
504 while (s < PL_bufend && isSPACE(*s)) {
505 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
510 if (s < PL_bufend && *s == '#') {
511 while (s < PL_bufend && *s != '\n')
515 if (PL_in_eval && !PL_rsfp) {
522 /* only continue to recharge the buffer if we're at the end
523 * of the buffer, we're not reading from a source filter, and
524 * we're in normal lexing mode
526 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
527 PL_lex_state == LEX_FORMLINE)
530 /* try to recharge the buffer */
531 if ((s = filter_gets(PL_linestr, PL_rsfp,
532 (prevlen = SvCUR(PL_linestr)))) == Nullch)
534 /* end of file. Add on the -p or -n magic */
535 if (PL_minus_n || PL_minus_p) {
536 sv_setpv(PL_linestr,PL_minus_p ?
537 ";}continue{print or die qq(-p destination: $!\\n)" :
539 sv_catpv(PL_linestr,";}");
540 PL_minus_n = PL_minus_p = 0;
543 sv_setpv(PL_linestr,";");
545 /* reset variables for next time we lex */
546 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
548 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
550 /* Close the filehandle. Could be from -P preprocessor,
551 * STDIN, or a regular file. If we were reading code from
552 * STDIN (because the commandline held no -e or filename)
553 * then we don't close it, we reset it so the code can
554 * read from STDIN too.
557 if (PL_preprocess && !PL_in_eval)
558 (void)PerlProc_pclose(PL_rsfp);
559 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
560 PerlIO_clearerr(PL_rsfp);
562 (void)PerlIO_close(PL_rsfp);
567 /* not at end of file, so we only read another line */
568 /* make corresponding updates to old pointers, for yyerror() */
569 oldprevlen = PL_oldbufptr - PL_bufend;
570 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
572 oldunilen = PL_last_uni - PL_bufend;
574 oldloplen = PL_last_lop - PL_bufend;
575 PL_linestart = PL_bufptr = s + prevlen;
576 PL_bufend = s + SvCUR(PL_linestr);
578 PL_oldbufptr = s + oldprevlen;
579 PL_oldoldbufptr = s + oldoldprevlen;
581 PL_last_uni = s + oldunilen;
583 PL_last_lop = s + oldloplen;
586 /* debugger active and we're not compiling the debugger code,
587 * so store the line into the debugger's array of lines
589 if (PERLDB_LINE && PL_curstash != PL_debstash) {
590 SV *sv = NEWSV(85,0);
592 sv_upgrade(sv, SVt_PVMG);
593 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
594 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
601 * Check the unary operators to ensure there's no ambiguity in how they're
602 * used. An ambiguous piece of code would be:
604 * This doesn't mean rand() + 5. Because rand() is a unary operator,
605 * the +5 is its argument.
614 if (PL_oldoldbufptr != PL_last_uni)
616 while (isSPACE(*PL_last_uni))
618 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
619 if ((t = strchr(s, '(')) && t < PL_bufptr)
621 if (ckWARN_d(WARN_AMBIGUOUS)){
624 Perl_warner(aTHX_ WARN_AMBIGUOUS,
625 "Warning: Use of \"%s\" without parens is ambiguous",
631 /* workaround to replace the UNI() macro with a function. Only the
632 * hints/uts.sh file mentions this. Other comments elsewhere in the
633 * source indicate Microport Unix might need it too.
639 #define UNI(f) return uni(f,s)
642 S_uni(pTHX_ I32 f, char *s)
647 PL_last_uni = PL_oldbufptr;
658 #endif /* CRIPPLED_CC */
661 * LOP : macro to build a list operator. Its behaviour has been replaced
662 * with a subroutine, S_lop() for which LOP is just another name.
665 #define LOP(f,x) return lop(f,x,s)
669 * Build a list operator (or something that might be one). The rules:
670 * - if we have a next token, then it's a list operator [why?]
671 * - if the next thing is an opening paren, then it's a function
672 * - else it's a list operator
676 S_lop(pTHX_ I32 f, int x, char *s)
682 PL_last_lop = PL_oldbufptr;
697 * When the lexer realizes it knows the next token (for instance,
698 * it is reordering tokens for the parser) then it can call S_force_next
699 * to know what token to return the next time the lexer is called. Caller
700 * will need to set PL_nextval[], and possibly PL_expect to ensure the lexer
701 * handles the token correctly.
705 S_force_next(pTHX_ I32 type)
707 PL_nexttype[PL_nexttoke] = type;
709 if (PL_lex_state != LEX_KNOWNEXT) {
710 PL_lex_defer = PL_lex_state;
711 PL_lex_expect = PL_expect;
712 PL_lex_state = LEX_KNOWNEXT;
718 * When the lexer knows the next thing is a word (for instance, it has
719 * just seen -> and it knows that the next char is a word char, then
720 * it calls S_force_word to stick the next word into the PL_next lookahead.
723 * char *start : buffer position (must be within PL_linestr)
724 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
725 * int check_keyword : if true, Perl checks to make sure the word isn't
726 * a keyword (do this if the word is a label, e.g. goto FOO)
727 * int allow_pack : if true, : characters will also be allowed (require,
729 * int allow_initial_tick : used by the "sub" lexer only.
733 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
738 start = skipspace(start);
740 if (isIDFIRST_lazy_if(s,UTF) ||
741 (allow_pack && *s == ':') ||
742 (allow_initial_tick && *s == '\'') )
744 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
745 if (check_keyword && keyword(PL_tokenbuf, len))
747 if (token == METHOD) {
752 PL_expect = XOPERATOR;
755 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST,0, newSVpv(PL_tokenbuf,0));
756 PL_nextval[PL_nexttoke].opval->op_private |= OPpCONST_BARE;
764 * Called when the lexer wants $foo *foo &foo etc, but the program
765 * text only contains the "foo" portion. The first argument is a pointer
766 * to the "foo", and the second argument is the type symbol to prefix.
767 * Forces the next token to be a "WORD".
768 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
772 S_force_ident(pTHX_ register char *s, int kind)
775 OP* o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
776 PL_nextval[PL_nexttoke].opval = o;
779 o->op_private = OPpCONST_ENTERED;
780 /* XXX see note in pp_entereval() for why we forgo typo
781 warnings if the symbol must be introduced in an eval.
783 gv_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
784 kind == '$' ? SVt_PV :
785 kind == '@' ? SVt_PVAV :
786 kind == '%' ? SVt_PVHV :
794 Perl_str_to_version(pTHX_ SV *sv)
799 char *start = SvPVx(sv,len);
800 bool utf = SvUTF8(sv) ? TRUE : FALSE;
801 char *end = start + len;
802 while (start < end) {
806 n = utf8_to_uv((U8*)start, len, &skip, 0);
811 retval += ((NV)n)/nshift;
820 * Forces the next token to be a version number.
824 S_force_version(pTHX_ char *s)
826 OP *version = Nullop;
835 for (; isDIGIT(*d) || *d == '_' || *d == '.'; d++);
836 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
838 s = scan_num(s, &yylval);
839 version = yylval.opval;
840 ver = cSVOPx(version)->op_sv;
841 if (SvPOK(ver) && !SvNIOK(ver)) {
842 (void)SvUPGRADE(ver, SVt_PVNV);
843 SvNVX(ver) = str_to_version(ver);
844 SvNOK_on(ver); /* hint that it is a version */
849 /* NOTE: The parser sees the package name and the VERSION swapped */
850 PL_nextval[PL_nexttoke].opval = version;
858 * Tokenize a quoted string passed in as an SV. It finds the next
859 * chunk, up to end of string or a backslash. It may make a new
860 * SV containing that chunk (if HINT_NEW_STRING is on). It also
865 S_tokeq(pTHX_ SV *sv)
876 s = SvPV_force(sv, len);
877 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
880 while (s < send && *s != '\\')
885 if ( PL_hints & HINT_NEW_STRING )
886 pv = sv_2mortal(newSVpvn(SvPVX(pv), len));
889 if (s + 1 < send && (s[1] == '\\'))
890 s++; /* all that, just for this */
895 SvCUR_set(sv, d - SvPVX(sv));
897 if ( PL_hints & HINT_NEW_STRING )
898 return new_constant(NULL, 0, "q", sv, pv, "q");
903 * Now come three functions related to double-quote context,
904 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
905 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
906 * interact with PL_lex_state, and create fake ( ... ) argument lists
907 * to handle functions and concatenation.
908 * They assume that whoever calls them will be setting up a fake
909 * join call, because each subthing puts a ',' after it. This lets
912 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
914 * (I'm not sure whether the spurious commas at the end of lcfirst's
915 * arguments and join's arguments are created or not).
920 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
922 * Pattern matching will set PL_lex_op to the pattern-matching op to
923 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
925 * OP_CONST and OP_READLINE are easy--just make the new op and return.
927 * Everything else becomes a FUNC.
929 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
930 * had an OP_CONST or OP_READLINE). This just sets us up for a
931 * call to S_sublex_push().
937 register I32 op_type = yylval.ival;
939 if (op_type == OP_NULL) {
940 yylval.opval = PL_lex_op;
944 if (op_type == OP_CONST || op_type == OP_READLINE) {
945 SV *sv = tokeq(PL_lex_stuff);
947 if (SvTYPE(sv) == SVt_PVIV) {
948 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
954 nsv = newSVpvn(p, len);
960 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
961 PL_lex_stuff = Nullsv;
965 PL_sublex_info.super_state = PL_lex_state;
966 PL_sublex_info.sub_inwhat = op_type;
967 PL_sublex_info.sub_op = PL_lex_op;
968 PL_lex_state = LEX_INTERPPUSH;
972 yylval.opval = PL_lex_op;
982 * Create a new scope to save the lexing state. The scope will be
983 * ended in S_sublex_done. Returns a '(', starting the function arguments
984 * to the uc, lc, etc. found before.
985 * Sets PL_lex_state to LEX_INTERPCONCAT.
993 PL_lex_state = PL_sublex_info.super_state;
994 SAVEI32(PL_lex_dojoin);
995 SAVEI32(PL_lex_brackets);
996 SAVEI32(PL_lex_casemods);
997 SAVEI32(PL_lex_starts);
998 SAVEI32(PL_lex_state);
999 SAVEVPTR(PL_lex_inpat);
1000 SAVEI32(PL_lex_inwhat);
1001 SAVECOPLINE(PL_curcop);
1002 SAVEPPTR(PL_bufptr);
1003 SAVEPPTR(PL_oldbufptr);
1004 SAVEPPTR(PL_oldoldbufptr);
1005 SAVEPPTR(PL_linestart);
1006 SAVESPTR(PL_linestr);
1007 SAVEPPTR(PL_lex_brackstack);
1008 SAVEPPTR(PL_lex_casestack);
1010 PL_linestr = PL_lex_stuff;
1011 PL_lex_stuff = Nullsv;
1013 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1014 = SvPVX(PL_linestr);
1015 PL_bufend += SvCUR(PL_linestr);
1016 SAVEFREESV(PL_linestr);
1018 PL_lex_dojoin = FALSE;
1019 PL_lex_brackets = 0;
1020 New(899, PL_lex_brackstack, 120, char);
1021 New(899, PL_lex_casestack, 12, char);
1022 SAVEFREEPV(PL_lex_brackstack);
1023 SAVEFREEPV(PL_lex_casestack);
1024 PL_lex_casemods = 0;
1025 *PL_lex_casestack = '\0';
1027 PL_lex_state = LEX_INTERPCONCAT;
1028 CopLINE_set(PL_curcop, PL_multi_start);
1030 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1031 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1032 PL_lex_inpat = PL_sublex_info.sub_op;
1034 PL_lex_inpat = Nullop;
1041 * Restores lexer state after a S_sublex_push.
1047 if (!PL_lex_starts++) {
1048 PL_expect = XOPERATOR;
1049 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn("",0));
1053 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1054 PL_lex_state = LEX_INTERPCASEMOD;
1058 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1059 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1060 PL_linestr = PL_lex_repl;
1062 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1063 PL_bufend += SvCUR(PL_linestr);
1064 SAVEFREESV(PL_linestr);
1065 PL_lex_dojoin = FALSE;
1066 PL_lex_brackets = 0;
1067 PL_lex_casemods = 0;
1068 *PL_lex_casestack = '\0';
1070 if (SvEVALED(PL_lex_repl)) {
1071 PL_lex_state = LEX_INTERPNORMAL;
1073 /* we don't clear PL_lex_repl here, so that we can check later
1074 whether this is an evalled subst; that means we rely on the
1075 logic to ensure sublex_done() is called again only via the
1076 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1079 PL_lex_state = LEX_INTERPCONCAT;
1080 PL_lex_repl = Nullsv;
1086 PL_bufend = SvPVX(PL_linestr);
1087 PL_bufend += SvCUR(PL_linestr);
1088 PL_expect = XOPERATOR;
1089 PL_sublex_info.sub_inwhat = 0;
1097 Extracts a pattern, double-quoted string, or transliteration. This
1100 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1101 processing a pattern (PL_lex_inpat is true), a transliteration
1102 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1104 Returns a pointer to the character scanned up to. Iff this is
1105 advanced from the start pointer supplied (ie if anything was
1106 successfully parsed), will leave an OP for the substring scanned
1107 in yylval. Caller must intuit reason for not parsing further
1108 by looking at the next characters herself.
1112 double-quoted style: \r and \n
1113 regexp special ones: \D \s
1115 backrefs: \1 (deprecated in substitution replacements)
1116 case and quoting: \U \Q \E
1117 stops on @ and $, but not for $ as tail anchor
1119 In transliterations:
1120 characters are VERY literal, except for - not at the start or end
1121 of the string, which indicates a range. scan_const expands the
1122 range to the full set of intermediate characters.
1124 In double-quoted strings:
1126 double-quoted style: \r and \n
1128 backrefs: \1 (deprecated)
1129 case and quoting: \U \Q \E
1132 scan_const does *not* construct ops to handle interpolated strings.
1133 It stops processing as soon as it finds an embedded $ or @ variable
1134 and leaves it to the caller to work out what's going on.
1136 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @:foo.
1138 $ in pattern could be $foo or could be tail anchor. Assumption:
1139 it's a tail anchor if $ is the last thing in the string, or if it's
1140 followed by one of ")| \n\t"
1142 \1 (backreferences) are turned into $1
1144 The structure of the code is
1145 while (there's a character to process) {
1146 handle transliteration ranges
1147 skip regexp comments
1148 skip # initiated comments in //x patterns
1149 check for embedded @foo
1150 check for embedded scalars
1152 leave intact backslashes from leave (below)
1153 deprecate \1 in strings and sub replacements
1154 handle string-changing backslashes \l \U \Q \E, etc.
1155 switch (what was escaped) {
1156 handle - in a transliteration (becomes a literal -)
1157 handle \132 octal characters
1158 handle 0x15 hex characters
1159 handle \cV (control V)
1160 handle printf backslashes (\f, \r, \n, etc)
1162 } (end if backslash)
1163 } (end while character to read)
1168 S_scan_const(pTHX_ char *start)
1170 register char *send = PL_bufend; /* end of the constant */
1171 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1172 register char *s = start; /* start of the constant */
1173 register char *d = SvPVX(sv); /* destination for copies */
1174 bool dorange = FALSE; /* are we in a translit range? */
1175 bool didrange = FALSE; /* did we just finish a range? */
1176 bool has_utf8 = FALSE; /* embedded \x{} */
1179 I32 utf = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1180 ? (PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF))
1182 I32 this_utf8 = (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op)
1183 ? (PL_sublex_info.sub_op->op_private & (PL_lex_repl ?
1184 OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF))
1186 const char *leaveit = /* set of acceptably-backslashed characters */
1188 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxcz0123456789[{]} \t\n\r\f\v#"
1191 while (s < send || dorange) {
1192 /* get transliterations out of the way (they're most literal) */
1193 if (PL_lex_inwhat == OP_TRANS) {
1194 /* expand a range A-Z to the full set of characters. AIE! */
1196 I32 i; /* current expanded character */
1197 I32 min; /* first character in range */
1198 I32 max; /* last character in range */
1200 i = d - SvPVX(sv); /* remember current offset */
1201 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1202 d = SvPVX(sv) + i; /* refresh d after realloc */
1203 d -= 2; /* eat the first char and the - */
1205 min = (U8)*d; /* first char in range */
1206 max = (U8)d[1]; /* last char in range */
1210 "Invalid [] range \"%c-%c\" in transliteration operator",
1211 (char)min, (char)max);
1215 if ((isLOWER(min) && isLOWER(max)) ||
1216 (isUPPER(min) && isUPPER(max))) {
1218 for (i = min; i <= max; i++)
1222 for (i = min; i <= max; i++)
1229 for (i = min; i <= max; i++)
1232 /* mark the range as done, and continue */
1238 /* range begins (ignore - as first or last char) */
1239 else if (*s == '-' && s+1 < send && s != start) {
1241 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1244 *d++ = (char)0xff; /* use illegal utf8 byte--see pmtrans */
1256 /* if we get here, we're not doing a transliteration */
1258 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1259 except for the last char, which will be done separately. */
1260 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1262 while (s < send && *s != ')')
1265 else if (s[2] == '{' /* This should match regcomp.c */
1266 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1269 char *regparse = s + (s[2] == '{' ? 3 : 4);
1272 while (count && (c = *regparse)) {
1273 if (c == '\\' && regparse[1])
1281 if (*regparse != ')') {
1282 regparse--; /* Leave one char for continuation. */
1283 yyerror("Sequence (?{...}) not terminated or not {}-balanced");
1285 while (s < regparse)
1290 /* likewise skip #-initiated comments in //x patterns */
1291 else if (*s == '#' && PL_lex_inpat &&
1292 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1293 while (s+1 < send && *s != '\n')
1297 /* check for embedded arrays
1298 (@foo, @:foo, @'foo, @{foo}, @$foo, @+, @-)
1300 else if (*s == '@' && s[1]
1301 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1304 /* check for embedded scalars. only stop if we're sure it's a
1307 else if (*s == '$') {
1308 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1310 if (s + 1 < send && !strchr("()| \n\t", s[1]))
1311 break; /* in regexp, $ might be tail anchor */
1315 if (*s == '\\' && s+1 < send) {
1316 bool to_be_utf8 = FALSE;
1320 /* some backslashes we leave behind */
1321 if (*leaveit && *s && strchr(leaveit, *s)) {
1327 /* deprecate \1 in strings and substitution replacements */
1328 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1329 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1331 if (ckWARN(WARN_SYNTAX))
1332 Perl_warner(aTHX_ WARN_SYNTAX, "\\%c better written as $%c", *s, *s);
1337 /* string-change backslash escapes */
1338 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1343 /* if we get here, it's either a quoted -, or a digit */
1346 /* quoted - in transliterations */
1348 if (PL_lex_inwhat == OP_TRANS) {
1355 if (ckWARN(WARN_MISC) && isALNUM(*s))
1356 Perl_warner(aTHX_ WARN_MISC,
1357 "Unrecognized escape \\%c passed through",
1359 /* default action is to copy the quoted character */
1364 /* \132 indicates an octal constant */
1365 case '0': case '1': case '2': case '3':
1366 case '4': case '5': case '6': case '7':
1368 STRLEN len = 0; /* disallow underscores */
1369 uv = (UV)scan_oct(s, 3, &len);
1372 goto NUM_ESCAPE_INSERT;
1374 /* \x24 indicates a hex constant */
1378 char* e = strchr(s, '}');
1380 yyerror("Missing right brace on \\x{}");
1384 STRLEN len = 1; /* allow underscores */
1385 uv = (UV)scan_hex(s + 1, e - s - 1, &len);
1392 STRLEN len = 0; /* disallow underscores */
1393 uv = (UV)scan_hex(s, 2, &len);
1399 /* Insert oct or hex escaped character.
1400 * There will always enough room in sv since such
1401 * escapes will be longer than any UT-F8 sequence
1402 * they can end up as. */
1404 /* This spot is wrong for EBCDIC. Characters like
1405 * the lowercase letters and digits are >127 in EBCDIC,
1406 * so here they would need to be mapped to the Unicode
1407 * repertoire. --jhi */
1410 if (!has_utf8 && (to_be_utf8 || uv > 255)) {
1411 /* Might need to recode whatever we have
1412 * accumulated so far if it contains any
1415 * (Can't we keep track of that and avoid
1416 * this rescan? --jhi)
1421 for (c = SvPVX(sv); c < d; c++) {
1422 if (UTF8_IS_CONTINUED(*c))
1426 char *old_pvx = SvPVX(sv);
1428 U8 tmpbuf[UTF8_MAXLEN+1];
1432 SvCUR(sv) + hicount + 1) +
1440 if (UTF8_IS_CONTINUED(*src)) {
1441 tmpend = uv_to_utf8(tmpbuf, (U8)*src--);
1442 dst -= tmpend - tmpbuf;
1443 Copy((char *)tmpbuf, dst+1,
1444 tmpend - tmpbuf, char);
1453 if (to_be_utf8 || (has_utf8 && uv > 127) || uv > 255) {
1454 d = (char*)uv_to_utf8((U8*)d, uv);
1466 /* \N{latin small letter a} is a named character */
1470 char* e = strchr(s, '}');
1476 yyerror("Missing right brace on \\N{}");
1480 res = newSVpvn(s + 1, e - s - 1);
1481 res = new_constant( Nullch, 0, "charnames",
1482 res, Nullsv, "\\N{...}" );
1483 str = SvPV(res,len);
1484 if (!has_utf8 && SvUTF8(res)) {
1485 char *ostart = SvPVX(sv);
1486 SvCUR_set(sv, d - ostart);
1489 sv_utf8_upgrade(sv);
1490 /* this just broke our allocation above... */
1491 SvGROW(sv, send - start);
1492 d = SvPVX(sv) + SvCUR(sv);
1495 if (len > e - s + 4) {
1496 char *odest = SvPVX(sv);
1498 SvGROW(sv, (SvCUR(sv) + len - (e - s + 4)));
1499 d = SvPVX(sv) + (d - odest);
1501 Copy(str, d, len, char);
1508 yyerror("Missing braces on \\N{}");
1511 /* \c is a control character */
1528 /* printf-style backslashes, formfeeds, newlines, etc */
1546 *d++ = '\047'; /* CP 1047 */
1549 *d++ = '\057'; /* CP 1047 */
1563 } /* end if (backslash) */
1565 /* (now in tr/// code again) */
1567 if (UTF8_IS_CONTINUED(*s) && (this_utf8 || has_utf8)) {
1568 STRLEN len = (STRLEN) -1;
1571 uv = utf8_to_uv((U8*)s, send - s, &len, 0);
1573 if (len == (STRLEN)-1) {
1574 /* Illegal UTF8 (a high-bit byte), make it valid. */
1575 char *old_pvx = SvPVX(sv);
1576 /* need space for one extra char (NOTE: SvCUR() not set here) */
1577 d = SvGROW(sv, SvLEN(sv) + 1) + (d - old_pvx);
1578 d = (char*)uv_to_utf8((U8*)d, (U8)*s++);
1589 } /* while loop to process each character */
1591 /* terminate the string and set up the sv */
1593 SvCUR_set(sv, d - SvPVX(sv));
1598 /* shrink the sv if we allocated more than we used */
1599 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1600 SvLEN_set(sv, SvCUR(sv) + 1);
1601 Renew(SvPVX(sv), SvLEN(sv), char);
1604 /* return the substring (via yylval) only if we parsed anything */
1605 if (s > PL_bufptr) {
1606 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
1607 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
1609 ( PL_lex_inwhat == OP_TRANS
1611 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
1614 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1621 * Returns TRUE if there's more to the expression (e.g., a subscript),
1624 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
1626 * ->[ and ->{ return TRUE
1627 * { and [ outside a pattern are always subscripts, so return TRUE
1628 * if we're outside a pattern and it's not { or [, then return FALSE
1629 * if we're in a pattern and the first char is a {
1630 * {4,5} (any digits around the comma) returns FALSE
1631 * if we're in a pattern and the first char is a [
1633 * [SOMETHING] has a funky algorithm to decide whether it's a
1634 * character class or not. It has to deal with things like
1635 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
1636 * anything else returns TRUE
1639 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
1642 S_intuit_more(pTHX_ register char *s)
1644 if (PL_lex_brackets)
1646 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
1648 if (*s != '{' && *s != '[')
1653 /* In a pattern, so maybe we have {n,m}. */
1670 /* On the other hand, maybe we have a character class */
1673 if (*s == ']' || *s == '^')
1676 /* this is terrifying, and it works */
1677 int weight = 2; /* let's weigh the evidence */
1679 unsigned char un_char = 255, last_un_char;
1680 char *send = strchr(s,']');
1681 char tmpbuf[sizeof PL_tokenbuf * 4];
1683 if (!send) /* has to be an expression */
1686 Zero(seen,256,char);
1689 else if (isDIGIT(*s)) {
1691 if (isDIGIT(s[1]) && s[2] == ']')
1697 for (; s < send; s++) {
1698 last_un_char = un_char;
1699 un_char = (unsigned char)*s;
1704 weight -= seen[un_char] * 10;
1705 if (isALNUM_lazy_if(s+1,UTF)) {
1706 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
1707 if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
1712 else if (*s == '$' && s[1] &&
1713 strchr("[#!%*<>()-=",s[1])) {
1714 if (/*{*/ strchr("])} =",s[2]))
1723 if (strchr("wds]",s[1]))
1725 else if (seen['\''] || seen['"'])
1727 else if (strchr("rnftbxcav",s[1]))
1729 else if (isDIGIT(s[1])) {
1731 while (s[1] && isDIGIT(s[1]))
1741 if (strchr("aA01! ",last_un_char))
1743 if (strchr("zZ79~",s[1]))
1745 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
1746 weight -= 5; /* cope with negative subscript */
1749 if (!isALNUM(last_un_char) && !strchr("$@&",last_un_char) &&
1750 isALPHA(*s) && s[1] && isALPHA(s[1])) {
1755 if (keyword(tmpbuf, d - tmpbuf))
1758 if (un_char == last_un_char + 1)
1760 weight -= seen[un_char];
1765 if (weight >= 0) /* probably a character class */
1775 * Does all the checking to disambiguate
1777 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
1778 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
1780 * First argument is the stuff after the first token, e.g. "bar".
1782 * Not a method if bar is a filehandle.
1783 * Not a method if foo is a subroutine prototyped to take a filehandle.
1784 * Not a method if it's really "Foo $bar"
1785 * Method if it's "foo $bar"
1786 * Not a method if it's really "print foo $bar"
1787 * Method if it's really "foo package::" (interpreted as package->foo)
1788 * Not a method if bar is known to be a subroutne ("sub bar; foo bar")
1789 * Not a method if bar is a filehandle or package, but is quoted with
1794 S_intuit_method(pTHX_ char *start, GV *gv)
1796 char *s = start + (*start == '$');
1797 char tmpbuf[sizeof PL_tokenbuf];
1805 if ((cv = GvCVu(gv))) {
1806 char *proto = SvPVX(cv);
1816 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
1817 /* start is the beginning of the possible filehandle/object,
1818 * and s is the end of it
1819 * tmpbuf is a copy of it
1822 if (*start == '$') {
1823 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
1828 return *s == '(' ? FUNCMETH : METHOD;
1830 if (!keyword(tmpbuf, len)) {
1831 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
1836 indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
1837 if (indirgv && GvCVu(indirgv))
1839 /* filehandle or package name makes it a method */
1840 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
1842 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
1843 return 0; /* no assumptions -- "=>" quotes bearword */
1845 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
1846 newSVpvn(tmpbuf,len));
1847 PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
1851 return *s == '(' ? FUNCMETH : METHOD;
1859 * Return a string of Perl code to load the debugger. If PERL5DB
1860 * is set, it will return the contents of that, otherwise a
1861 * compile-time require of perl5db.pl.
1868 char *pdb = PerlEnv_getenv("PERL5DB");
1872 SETERRNO(0,SS$_NORMAL);
1873 return "BEGIN { require 'perl5db.pl' }";
1879 /* Encoded script support. filter_add() effectively inserts a
1880 * 'pre-processing' function into the current source input stream.
1881 * Note that the filter function only applies to the current source file
1882 * (e.g., it will not affect files 'require'd or 'use'd by this one).
1884 * The datasv parameter (which may be NULL) can be used to pass
1885 * private data to this instance of the filter. The filter function
1886 * can recover the SV using the FILTER_DATA macro and use it to
1887 * store private buffers and state information.
1889 * The supplied datasv parameter is upgraded to a PVIO type
1890 * and the IoDIRP/IoANY field is used to store the function pointer,
1891 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
1892 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
1893 * private use must be set using malloc'd pointers.
1897 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
1902 if (!PL_rsfp_filters)
1903 PL_rsfp_filters = newAV();
1905 datasv = NEWSV(255,0);
1906 if (!SvUPGRADE(datasv, SVt_PVIO))
1907 Perl_die(aTHX_ "Can't upgrade filter_add data to SVt_PVIO");
1908 IoANY(datasv) = (void *)funcp; /* stash funcp into spare field */
1909 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
1910 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
1911 funcp, SvPV_nolen(datasv)));
1912 av_unshift(PL_rsfp_filters, 1);
1913 av_store(PL_rsfp_filters, 0, datasv) ;
1918 /* Delete most recently added instance of this filter function. */
1920 Perl_filter_del(pTHX_ filter_t funcp)
1923 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", funcp));
1924 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
1926 /* if filter is on top of stack (usual case) just pop it off */
1927 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
1928 if (IoANY(datasv) == (void *)funcp) {
1929 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
1930 IoANY(datasv) = (void *)NULL;
1931 sv_free(av_pop(PL_rsfp_filters));
1935 /* we need to search for the correct entry and clear it */
1936 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
1940 /* Invoke the n'th filter function for the current rsfp. */
1942 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
1945 /* 0 = read one text line */
1950 if (!PL_rsfp_filters)
1952 if (idx > AvFILLp(PL_rsfp_filters)){ /* Any more filters? */
1953 /* Provide a default input filter to make life easy. */
1954 /* Note that we append to the line. This is handy. */
1955 DEBUG_P(PerlIO_printf(Perl_debug_log,
1956 "filter_read %d: from rsfp\n", idx));
1960 int old_len = SvCUR(buf_sv) ;
1962 /* ensure buf_sv is large enough */
1963 SvGROW(buf_sv, old_len + maxlen) ;
1964 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
1965 if (PerlIO_error(PL_rsfp))
1966 return -1; /* error */
1968 return 0 ; /* end of file */
1970 SvCUR_set(buf_sv, old_len + len) ;
1973 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
1974 if (PerlIO_error(PL_rsfp))
1975 return -1; /* error */
1977 return 0 ; /* end of file */
1980 return SvCUR(buf_sv);
1982 /* Skip this filter slot if filter has been deleted */
1983 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef){
1984 DEBUG_P(PerlIO_printf(Perl_debug_log,
1985 "filter_read %d: skipped (filter deleted)\n",
1987 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
1989 /* Get function pointer hidden within datasv */
1990 funcp = (filter_t)IoANY(datasv);
1991 DEBUG_P(PerlIO_printf(Perl_debug_log,
1992 "filter_read %d: via function %p (%s)\n",
1993 idx, funcp, SvPV_nolen(datasv)));
1994 /* Call function. The function is expected to */
1995 /* call "FILTER_READ(idx+1, buf_sv)" first. */
1996 /* Return: <0:error, =0:eof, >0:not eof */
1997 return (*funcp)(aTHXo_ idx, buf_sv, maxlen);
2001 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2003 #ifdef PERL_CR_FILTER
2004 if (!PL_rsfp_filters) {
2005 filter_add(S_cr_textfilter,NULL);
2008 if (PL_rsfp_filters) {
2011 SvCUR_set(sv, 0); /* start with empty line */
2012 if (FILTER_READ(0, sv, 0) > 0)
2013 return ( SvPVX(sv) ) ;
2018 return (sv_gets(sv, fp, append));
2022 S_find_in_my_stash(pTHX_ char *pkgname, I32 len)
2026 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2030 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2031 (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
2033 return GvHV(gv); /* Foo:: */
2036 /* use constant CLASS => 'MyClass' */
2037 if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
2039 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2040 pkgname = SvPV_nolen(sv);
2044 return gv_stashpv(pkgname, FALSE);
2048 static char* exp_name[] =
2049 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2050 "ATTRTERM", "TERMBLOCK"
2057 Works out what to call the token just pulled out of the input
2058 stream. The yacc parser takes care of taking the ops we return and
2059 stitching them into a tree.
2065 if read an identifier
2066 if we're in a my declaration
2067 croak if they tried to say my($foo::bar)
2068 build the ops for a my() declaration
2069 if it's an access to a my() variable
2070 are we in a sort block?
2071 croak if my($a); $a <=> $b
2072 build ops for access to a my() variable
2073 if in a dq string, and they've said @foo and we can't find @foo
2075 build ops for a bareword
2076 if we already built the token before, use it.
2079 #ifdef USE_PURE_BISON
2081 #pragma segment Perl_yylex_r
2084 Perl_yylex_r(pTHX_ YYSTYPE *lvalp, int *lcharp)
2089 yylval_pointer[yyactlevel] = lvalp;
2090 yychar_pointer[yyactlevel] = lcharp;
2091 if (yyactlevel >= YYMAXLEVEL)
2092 Perl_croak(aTHX_ "panic: YYMAXLEVEL");
2094 r = Perl_yylex(aTHX);
2103 #pragma segment Perl_yylex
2116 /* check if there's an identifier for us to look at */
2117 if (PL_pending_ident) {
2118 /* pit holds the identifier we read and pending_ident is reset */
2119 char pit = PL_pending_ident;
2120 PL_pending_ident = 0;
2122 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2123 "### Tokener saw identifier '%s'\n", PL_tokenbuf); })
2125 /* if we're in a my(), we can't allow dynamics here.
2126 $foo'bar has already been turned into $foo::bar, so
2127 just check for colons.
2129 if it's a legal name, the OP is a PADANY.
2132 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
2133 if (strchr(PL_tokenbuf,':'))
2134 yyerror(Perl_form(aTHX_ "No package name allowed for "
2135 "variable %s in \"our\"",
2137 tmp = pad_allocmy(PL_tokenbuf);
2140 if (strchr(PL_tokenbuf,':'))
2141 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
2143 yylval.opval = newOP(OP_PADANY, 0);
2144 yylval.opval->op_targ = pad_allocmy(PL_tokenbuf);
2150 build the ops for accesses to a my() variable.
2152 Deny my($a) or my($b) in a sort block, *if* $a or $b is
2153 then used in a comparison. This catches most, but not
2154 all cases. For instance, it catches
2155 sort { my($a); $a <=> $b }
2157 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
2158 (although why you'd do that is anyone's guess).
2161 if (!strchr(PL_tokenbuf,':')) {
2163 /* Check for single character per-thread SVs */
2164 if (PL_tokenbuf[0] == '$' && PL_tokenbuf[2] == '\0'
2165 && !isALPHA(PL_tokenbuf[1]) /* Rule out obvious non-threadsvs */
2166 && (tmp = find_threadsv(&PL_tokenbuf[1])) != NOT_IN_PAD)
2168 yylval.opval = newOP(OP_THREADSV, 0);
2169 yylval.opval->op_targ = tmp;
2172 #endif /* USE_THREADS */
2173 if ((tmp = pad_findmy(PL_tokenbuf)) != NOT_IN_PAD) {
2174 SV *namesv = AvARRAY(PL_comppad_name)[tmp];
2175 /* might be an "our" variable" */
2176 if (SvFLAGS(namesv) & SVpad_OUR) {
2177 /* build ops for a bareword */
2178 SV *sym = newSVpv(HvNAME(GvSTASH(namesv)),0);
2179 sv_catpvn(sym, "::", 2);
2180 sv_catpv(sym, PL_tokenbuf+1);
2181 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
2182 yylval.opval->op_private = OPpCONST_ENTERED;
2183 gv_fetchpv(SvPVX(sym),
2185 ? (GV_ADDMULTI | GV_ADDINEVAL)
2188 ((PL_tokenbuf[0] == '$') ? SVt_PV
2189 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
2194 /* if it's a sort block and they're naming $a or $b */
2195 if (PL_last_lop_op == OP_SORT &&
2196 PL_tokenbuf[0] == '$' &&
2197 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
2200 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
2201 d < PL_bufend && *d != '\n';
2204 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
2205 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
2211 yylval.opval = newOP(OP_PADANY, 0);
2212 yylval.opval->op_targ = tmp;
2218 Whine if they've said @foo in a doublequoted string,
2219 and @foo isn't a variable we can find in the symbol
2222 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
2223 GV *gv = gv_fetchpv(PL_tokenbuf+1, FALSE, SVt_PVAV);
2224 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
2225 && ckWARN(WARN_AMBIGUOUS))
2227 /* Downgraded from fatal to warning 20000522 mjd */
2228 Perl_warner(aTHX_ WARN_AMBIGUOUS,
2229 "Possible unintended interpolation of %s in string",
2234 /* build ops for a bareword */
2235 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
2236 yylval.opval->op_private = OPpCONST_ENTERED;
2237 gv_fetchpv(PL_tokenbuf+1, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
2238 ((PL_tokenbuf[0] == '$') ? SVt_PV
2239 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
2244 /* no identifier pending identification */
2246 switch (PL_lex_state) {
2248 case LEX_NORMAL: /* Some compilers will produce faster */
2249 case LEX_INTERPNORMAL: /* code if we comment these out. */
2253 /* when we've already built the next token, just pull it out of the queue */
2256 yylval = PL_nextval[PL_nexttoke];
2258 PL_lex_state = PL_lex_defer;
2259 PL_expect = PL_lex_expect;
2260 PL_lex_defer = LEX_NORMAL;
2262 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2263 "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
2264 (IV)PL_nexttype[PL_nexttoke]); })
2266 return(PL_nexttype[PL_nexttoke]);
2268 /* interpolated case modifiers like \L \U, including \Q and \E.
2269 when we get here, PL_bufptr is at the \
2271 case LEX_INTERPCASEMOD:
2273 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
2274 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
2276 /* handle \E or end of string */
2277 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
2281 if (PL_lex_casemods) {
2282 oldmod = PL_lex_casestack[--PL_lex_casemods];
2283 PL_lex_casestack[PL_lex_casemods] = '\0';
2285 if (PL_bufptr != PL_bufend && strchr("LUQ", oldmod)) {
2287 PL_lex_state = LEX_INTERPCONCAT;
2291 if (PL_bufptr != PL_bufend)
2293 PL_lex_state = LEX_INTERPCONCAT;
2297 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2298 "### Saw case modifier at '%s'\n", PL_bufptr); })
2300 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
2301 tmp = *s, *s = s[2], s[2] = tmp; /* misordered... */
2302 if (strchr("LU", *s) &&
2303 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U')))
2305 PL_lex_casestack[--PL_lex_casemods] = '\0';
2308 if (PL_lex_casemods > 10) {
2309 char* newlb = Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
2310 if (newlb != PL_lex_casestack) {
2312 PL_lex_casestack = newlb;
2315 PL_lex_casestack[PL_lex_casemods++] = *s;
2316 PL_lex_casestack[PL_lex_casemods] = '\0';
2317 PL_lex_state = LEX_INTERPCONCAT;
2318 PL_nextval[PL_nexttoke].ival = 0;
2321 PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
2323 PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
2325 PL_nextval[PL_nexttoke].ival = OP_LC;
2327 PL_nextval[PL_nexttoke].ival = OP_UC;
2329 PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
2331 Perl_croak(aTHX_ "panic: yylex");
2334 if (PL_lex_starts) {
2343 case LEX_INTERPPUSH:
2344 return sublex_push();
2346 case LEX_INTERPSTART:
2347 if (PL_bufptr == PL_bufend)
2348 return sublex_done();
2349 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2350 "### Interpolated variable at '%s'\n", PL_bufptr); })
2352 PL_lex_dojoin = (*PL_bufptr == '@');
2353 PL_lex_state = LEX_INTERPNORMAL;
2354 if (PL_lex_dojoin) {
2355 PL_nextval[PL_nexttoke].ival = 0;
2358 PL_nextval[PL_nexttoke].opval = newOP(OP_THREADSV, 0);
2359 PL_nextval[PL_nexttoke].opval->op_targ = find_threadsv("\"");
2360 force_next(PRIVATEREF);
2362 force_ident("\"", '$');
2363 #endif /* USE_THREADS */
2364 PL_nextval[PL_nexttoke].ival = 0;
2366 PL_nextval[PL_nexttoke].ival = 0;
2368 PL_nextval[PL_nexttoke].ival = OP_JOIN; /* emulate join($", ...) */
2371 if (PL_lex_starts++) {
2377 case LEX_INTERPENDMAYBE:
2378 if (intuit_more(PL_bufptr)) {
2379 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
2385 if (PL_lex_dojoin) {
2386 PL_lex_dojoin = FALSE;
2387 PL_lex_state = LEX_INTERPCONCAT;
2390 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
2391 && SvEVALED(PL_lex_repl))
2393 if (PL_bufptr != PL_bufend)
2394 Perl_croak(aTHX_ "Bad evalled substitution pattern");
2395 PL_lex_repl = Nullsv;
2398 case LEX_INTERPCONCAT:
2400 if (PL_lex_brackets)
2401 Perl_croak(aTHX_ "panic: INTERPCONCAT");
2403 if (PL_bufptr == PL_bufend)
2404 return sublex_done();
2406 if (SvIVX(PL_linestr) == '\'') {
2407 SV *sv = newSVsv(PL_linestr);
2410 else if ( PL_hints & HINT_NEW_RE )
2411 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
2412 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2416 s = scan_const(PL_bufptr);
2418 PL_lex_state = LEX_INTERPCASEMOD;
2420 PL_lex_state = LEX_INTERPSTART;
2423 if (s != PL_bufptr) {
2424 PL_nextval[PL_nexttoke] = yylval;
2427 if (PL_lex_starts++)
2437 PL_lex_state = LEX_NORMAL;
2438 s = scan_formline(PL_bufptr);
2439 if (!PL_lex_formbrack)
2445 PL_oldoldbufptr = PL_oldbufptr;
2448 PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at %s\n",
2449 exp_name[PL_expect], s);
2455 if (isIDFIRST_lazy_if(s,UTF))
2457 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
2460 goto fake_eof; /* emulate EOF on ^D or ^Z */
2465 if (PL_lex_brackets)
2466 yyerror("Missing right curly or square bracket");
2467 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2468 "### Tokener got EOF\n");
2472 if (s++ < PL_bufend)
2473 goto retry; /* ignore stray nulls */
2476 if (!PL_in_eval && !PL_preambled) {
2477 PL_preambled = TRUE;
2478 sv_setpv(PL_linestr,incl_perldb());
2479 if (SvCUR(PL_linestr))
2480 sv_catpv(PL_linestr,";");
2482 while(AvFILLp(PL_preambleav) >= 0) {
2483 SV *tmpsv = av_shift(PL_preambleav);
2484 sv_catsv(PL_linestr, tmpsv);
2485 sv_catpv(PL_linestr, ";");
2488 sv_free((SV*)PL_preambleav);
2489 PL_preambleav = NULL;
2491 if (PL_minus_n || PL_minus_p) {
2492 sv_catpv(PL_linestr, "LINE: while (<>) {");
2494 sv_catpv(PL_linestr,"chomp;");
2496 GV* gv = gv_fetchpv("::F", TRUE, SVt_PVAV);
2498 GvIMPORTED_AV_on(gv);
2500 if (strchr("/'\"", *PL_splitstr)
2501 && strchr(PL_splitstr + 1, *PL_splitstr))
2502 Perl_sv_catpvf(aTHX_ PL_linestr, "@F=split(%s);", PL_splitstr);
2505 s = "'~#\200\1'"; /* surely one char is unused...*/
2506 while (s[1] && strchr(PL_splitstr, *s)) s++;
2508 Perl_sv_catpvf(aTHX_ PL_linestr, "@F=split(%s%c",
2509 "q" + (delim == '\''), delim);
2510 for (s = PL_splitstr; *s; s++) {
2512 sv_catpvn(PL_linestr, "\\", 1);
2513 sv_catpvn(PL_linestr, s, 1);
2515 Perl_sv_catpvf(aTHX_ PL_linestr, "%c);", delim);
2519 sv_catpv(PL_linestr,"@F=split(' ');");
2522 sv_catpv(PL_linestr, "\n");
2523 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2524 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2525 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2526 SV *sv = NEWSV(85,0);
2528 sv_upgrade(sv, SVt_PVMG);
2529 sv_setsv(sv,PL_linestr);
2530 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2535 bof = PL_rsfp ? TRUE : FALSE;
2537 #ifdef PERLIO_IS_STDIO
2538 # ifdef __GNU_LIBRARY__
2539 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
2540 # define FTELL_FOR_PIPE_IS_BROKEN
2544 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
2545 # define FTELL_FOR_PIPE_IS_BROKEN
2550 #ifdef FTELL_FOR_PIPE_IS_BROKEN
2551 /* This loses the possibility to detect the bof
2552 * situation on perl -P when the libc5 is being used.
2553 * Workaround? Maybe attach some extra state to PL_rsfp?
2556 bof = PerlIO_tell(PL_rsfp) == 0;
2558 bof = PerlIO_tell(PL_rsfp) == 0;
2561 s = filter_gets(PL_linestr, PL_rsfp, 0);
2565 if (PL_preprocess && !PL_in_eval)
2566 (void)PerlProc_pclose(PL_rsfp);
2567 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
2568 PerlIO_clearerr(PL_rsfp);
2570 (void)PerlIO_close(PL_rsfp);
2572 PL_doextract = FALSE;
2574 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
2575 sv_setpv(PL_linestr,PL_minus_p ? ";}continue{print" : "");
2576 sv_catpv(PL_linestr,";}");
2577 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2578 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2579 PL_minus_n = PL_minus_p = 0;
2582 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2583 sv_setpv(PL_linestr,"");
2584 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
2586 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2587 s = swallow_bom((U8*)s);
2590 if (*s == '#' && s[1] == '!' && instr(s,"perl"))
2591 PL_doextract = FALSE;
2593 /* Incest with pod. */
2594 if (*s == '=' && strnEQ(s, "=cut", 4)) {
2595 sv_setpv(PL_linestr, "");
2596 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2597 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2598 PL_doextract = FALSE;
2602 } while (PL_doextract);
2603 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
2604 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2605 SV *sv = NEWSV(85,0);
2607 sv_upgrade(sv, SVt_PVMG);
2608 sv_setsv(sv,PL_linestr);
2609 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2611 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2612 if (CopLINE(PL_curcop) == 1) {
2613 while (s < PL_bufend && isSPACE(*s))
2615 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
2619 if (*s == '#' && *(s+1) == '!')
2621 #ifdef ALTERNATE_SHEBANG
2623 static char as[] = ALTERNATE_SHEBANG;
2624 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
2625 d = s + (sizeof(as) - 1);
2627 #endif /* ALTERNATE_SHEBANG */
2636 while (*d && !isSPACE(*d))
2640 #ifdef ARG_ZERO_IS_SCRIPT
2641 if (ipathend > ipath) {
2643 * HP-UX (at least) sets argv[0] to the script name,
2644 * which makes $^X incorrect. And Digital UNIX and Linux,
2645 * at least, set argv[0] to the basename of the Perl
2646 * interpreter. So, having found "#!", we'll set it right.
2648 SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV));
2649 assert(SvPOK(x) || SvGMAGICAL(x));
2650 if (sv_eq(x, CopFILESV(PL_curcop))) {
2651 sv_setpvn(x, ipath, ipathend - ipath);
2654 TAINT_NOT; /* $^X is always tainted, but that's OK */
2656 #endif /* ARG_ZERO_IS_SCRIPT */
2661 d = instr(s,"perl -");
2663 d = instr(s,"perl");
2665 /* avoid getting into infinite loops when shebang
2666 * line contains "Perl" rather than "perl" */
2668 for (d = ipathend-4; d >= ipath; --d) {
2669 if ((*d == 'p' || *d == 'P')
2670 && !ibcmp(d, "perl", 4))
2680 #ifdef ALTERNATE_SHEBANG
2682 * If the ALTERNATE_SHEBANG on this system starts with a
2683 * character that can be part of a Perl expression, then if
2684 * we see it but not "perl", we're probably looking at the
2685 * start of Perl code, not a request to hand off to some
2686 * other interpreter. Similarly, if "perl" is there, but
2687 * not in the first 'word' of the line, we assume the line
2688 * contains the start of the Perl program.
2690 if (d && *s != '#') {
2692 while (*c && !strchr("; \t\r\n\f\v#", *c))
2695 d = Nullch; /* "perl" not in first word; ignore */
2697 *s = '#'; /* Don't try to parse shebang line */
2699 #endif /* ALTERNATE_SHEBANG */
2700 #ifndef MACOS_TRADITIONAL
2705 !instr(s,"indir") &&
2706 instr(PL_origargv[0],"perl"))
2712 while (s < PL_bufend && isSPACE(*s))
2714 if (s < PL_bufend) {
2715 Newz(899,newargv,PL_origargc+3,char*);
2717 while (s < PL_bufend && !isSPACE(*s))
2720 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
2723 newargv = PL_origargv;
2725 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
2726 Perl_croak(aTHX_ "Can't exec %s", ipath);
2730 U32 oldpdb = PL_perldb;
2731 bool oldn = PL_minus_n;
2732 bool oldp = PL_minus_p;
2734 while (*d && !isSPACE(*d)) d++;
2735 while (SPACE_OR_TAB(*d)) d++;
2739 if (*d == 'M' || *d == 'm') {
2741 while (*d && !isSPACE(*d)) d++;
2742 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
2745 d = moreswitches(d);
2747 if ((PERLDB_LINE && !oldpdb) ||
2748 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
2749 /* if we have already added "LINE: while (<>) {",
2750 we must not do it again */
2752 sv_setpv(PL_linestr, "");
2753 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2754 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2755 PL_preambled = FALSE;
2757 (void)gv_fetchfile(PL_origfilename);
2764 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2766 PL_lex_state = LEX_FORMLINE;
2771 #ifdef PERL_STRICT_CR
2772 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
2774 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
2776 case ' ': case '\t': case '\f': case 013:
2777 #ifdef MACOS_TRADITIONAL
2784 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
2785 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
2786 /* handle eval qq[#line 1 "foo"\n ...] */
2787 CopLINE_dec(PL_curcop);
2791 while (s < d && *s != '\n')
2796 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2798 PL_lex_state = LEX_FORMLINE;
2808 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
2815 while (s < PL_bufend && SPACE_OR_TAB(*s))
2818 if (strnEQ(s,"=>",2)) {
2819 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
2820 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2821 "### Saw unary minus before =>, forcing word '%s'\n", s);
2823 OPERATOR('-'); /* unary minus */
2825 PL_last_uni = PL_oldbufptr;
2827 case 'r': ftst = OP_FTEREAD; break;
2828 case 'w': ftst = OP_FTEWRITE; break;
2829 case 'x': ftst = OP_FTEEXEC; break;
2830 case 'o': ftst = OP_FTEOWNED; break;
2831 case 'R': ftst = OP_FTRREAD; break;
2832 case 'W': ftst = OP_FTRWRITE; break;
2833 case 'X': ftst = OP_FTREXEC; break;
2834 case 'O': ftst = OP_FTROWNED; break;
2835 case 'e': ftst = OP_FTIS; break;
2836 case 'z': ftst = OP_FTZERO; break;
2837 case 's': ftst = OP_FTSIZE; break;
2838 case 'f': ftst = OP_FTFILE; break;
2839 case 'd': ftst = OP_FTDIR; break;
2840 case 'l': ftst = OP_FTLINK; break;
2841 case 'p': ftst = OP_FTPIPE; break;
2842 case 'S': ftst = OP_FTSOCK; break;
2843 case 'u': ftst = OP_FTSUID; break;
2844 case 'g': ftst = OP_FTSGID; break;
2845 case 'k': ftst = OP_FTSVTX; break;
2846 case 'b': ftst = OP_FTBLK; break;
2847 case 'c': ftst = OP_FTCHR; break;
2848 case 't': ftst = OP_FTTTY; break;
2849 case 'T': ftst = OP_FTTEXT; break;
2850 case 'B': ftst = OP_FTBINARY; break;
2851 case 'M': case 'A': case 'C':
2852 gv_fetchpv("\024",TRUE, SVt_PV);
2854 case 'M': ftst = OP_FTMTIME; break;
2855 case 'A': ftst = OP_FTATIME; break;
2856 case 'C': ftst = OP_FTCTIME; break;
2864 PL_last_lop_op = ftst;
2865 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2866 "### Saw file test %c\n", ftst);
2871 /* Assume it was a minus followed by a one-letter named
2872 * subroutine call (or a -bareword), then. */
2873 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2874 "### %c looked like a file test but was not\n", ftst);
2882 if (PL_expect == XOPERATOR)
2887 else if (*s == '>') {
2890 if (isIDFIRST_lazy_if(s,UTF)) {
2891 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
2899 if (PL_expect == XOPERATOR)
2902 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2904 OPERATOR('-'); /* unary minus */
2911 if (PL_expect == XOPERATOR)
2916 if (PL_expect == XOPERATOR)
2919 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
2925 if (PL_expect != XOPERATOR) {
2926 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
2927 PL_expect = XOPERATOR;
2928 force_ident(PL_tokenbuf, '*');
2941 if (PL_expect == XOPERATOR) {
2945 PL_tokenbuf[0] = '%';
2946 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
2947 if (!PL_tokenbuf[1]) {
2949 yyerror("Final % should be \\% or %name");
2952 PL_pending_ident = '%';
2971 switch (PL_expect) {
2974 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
2976 PL_bufptr = s; /* update in case we back off */
2982 PL_expect = XTERMBLOCK;
2986 while (isIDFIRST_lazy_if(s,UTF)) {
2987 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
2988 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
2989 if (tmp < 0) tmp = -tmp;
3004 d = scan_str(d,TRUE,TRUE);
3007 SvREFCNT_dec(PL_lex_stuff);
3008 PL_lex_stuff = Nullsv;
3010 /* MUST advance bufptr here to avoid bogus
3011 "at end of line" context messages from yyerror().
3013 PL_bufptr = s + len;
3014 yyerror("Unterminated attribute parameter in attribute list");
3017 return 0; /* EOF indicator */
3021 SV *sv = newSVpvn(s, len);
3022 sv_catsv(sv, PL_lex_stuff);
3023 attrs = append_elem(OP_LIST, attrs,
3024 newSVOP(OP_CONST, 0, sv));
3025 SvREFCNT_dec(PL_lex_stuff);
3026 PL_lex_stuff = Nullsv;
3029 attrs = append_elem(OP_LIST, attrs,
3030 newSVOP(OP_CONST, 0,
3034 if (*s == ':' && s[1] != ':')
3037 break; /* require real whitespace or :'s */
3039 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3040 if (*s != ';' && *s != tmp && (tmp != '=' || *s != ')')) {
3041 char q = ((*s == '\'') ? '"' : '\'');
3042 /* If here for an expression, and parsed no attrs, back off. */
3043 if (tmp == '=' && !attrs) {
3047 /* MUST advance bufptr here to avoid bogus "at end of line"
3048 context messages from yyerror().
3052 yyerror("Unterminated attribute list");
3054 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3062 PL_nextval[PL_nexttoke].opval = attrs;
3070 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3071 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3087 if (PL_lex_brackets <= 0)
3088 yyerror("Unmatched right square bracket");
3091 if (PL_lex_state == LEX_INTERPNORMAL) {
3092 if (PL_lex_brackets == 0) {
3093 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3094 PL_lex_state = LEX_INTERPEND;
3101 if (PL_lex_brackets > 100) {
3102 char* newlb = Renew(PL_lex_brackstack, PL_lex_brackets + 1, char);
3103 if (newlb != PL_lex_brackstack) {
3105 PL_lex_brackstack = newlb;
3108 switch (PL_expect) {
3110 if (PL_lex_formbrack) {
3114 if (PL_oldoldbufptr == PL_last_lop)
3115 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3117 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3118 OPERATOR(HASHBRACK);
3120 while (s < PL_bufend && SPACE_OR_TAB(*s))
3123 PL_tokenbuf[0] = '\0';
3124 if (d < PL_bufend && *d == '-') {
3125 PL_tokenbuf[0] = '-';
3127 while (d < PL_bufend && SPACE_OR_TAB(*d))
3130 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3131 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3133 while (d < PL_bufend && SPACE_OR_TAB(*d))
3136 char minus = (PL_tokenbuf[0] == '-');
3137 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3138 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, 0) &&
3139 PL_nextval[PL_nexttoke-1].opval)
3140 SvUTF8_on(((SVOP*)PL_nextval[PL_nexttoke-1].opval)->op_sv);
3148 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3153 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3158 if (PL_oldoldbufptr == PL_last_lop)
3159 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3161 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3164 OPERATOR(HASHBRACK);
3165 /* This hack serves to disambiguate a pair of curlies
3166 * as being a block or an anon hash. Normally, expectation
3167 * determines that, but in cases where we're not in a
3168 * position to expect anything in particular (like inside
3169 * eval"") we have to resolve the ambiguity. This code
3170 * covers the case where the first term in the curlies is a
3171 * quoted string. Most other cases need to be explicitly
3172 * disambiguated by prepending a `+' before the opening
3173 * curly in order to force resolution as an anon hash.
3175 * XXX should probably propagate the outer expectation
3176 * into eval"" to rely less on this hack, but that could
3177 * potentially break current behavior of eval"".
3181 if (*s == '\'' || *s == '"' || *s == '`') {
3182 /* common case: get past first string, handling escapes */
3183 for (t++; t < PL_bufend && *t != *s;)
3184 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3188 else if (*s == 'q') {
3191 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3195 char open, close, term;
3198 while (t < PL_bufend && isSPACE(*t))
3202 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3206 for (t++; t < PL_bufend; t++) {
3207 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3209 else if (*t == open)
3213 for (t++; t < PL_bufend; t++) {
3214 if (*t == '\\' && t+1 < PL_bufend)
3216 else if (*t == close && --brackets <= 0)
3218 else if (*t == open)
3224 else if (isALNUM_lazy_if(t,UTF)) {
3226 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3229 while (t < PL_bufend && isSPACE(*t))
3231 /* if comma follows first term, call it an anon hash */
3232 /* XXX it could be a comma expression with loop modifiers */
3233 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3234 || (*t == '=' && t[1] == '>')))
3235 OPERATOR(HASHBRACK);
3236 if (PL_expect == XREF)
3239 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3245 yylval.ival = CopLINE(PL_curcop);
3246 if (isSPACE(*s) || *s == '#')
3247 PL_copline = NOLINE; /* invalidate current command line number */
3252 if (PL_lex_brackets <= 0)
3253 yyerror("Unmatched right curly bracket");
3255 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3256 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3257 PL_lex_formbrack = 0;
3258 if (PL_lex_state == LEX_INTERPNORMAL) {
3259 if (PL_lex_brackets == 0) {
3260 if (PL_expect & XFAKEBRACK) {
3261 PL_expect &= XENUMMASK;
3262 PL_lex_state = LEX_INTERPEND;
3264 return yylex(); /* ignore fake brackets */
3266 if (*s == '-' && s[1] == '>')
3267 PL_lex_state = LEX_INTERPENDMAYBE;
3268 else if (*s != '[' && *s != '{')
3269 PL_lex_state = LEX_INTERPEND;
3272 if (PL_expect & XFAKEBRACK) {
3273 PL_expect &= XENUMMASK;
3275 return yylex(); /* ignore fake brackets */
3285 if (PL_expect == XOPERATOR) {
3286 if (ckWARN(WARN_SEMICOLON)
3287 && isIDFIRST_lazy_if(s,UTF) && PL_bufptr == PL_linestart)
3289 CopLINE_dec(PL_curcop);
3290 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3291 CopLINE_inc(PL_curcop);
3296 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3298 PL_expect = XOPERATOR;
3299 force_ident(PL_tokenbuf, '&');
3303 yylval.ival = (OPpENTERSUB_AMPER<<8);
3322 if (ckWARN(WARN_SYNTAX) && tmp && isSPACE(*s) && strchr("+-*/%.^&|<",tmp))
3323 Perl_warner(aTHX_ WARN_SYNTAX, "Reversed %c= operator",(int)tmp);
3325 if (PL_expect == XSTATE && isALPHA(tmp) &&
3326 (s == PL_linestart+1 || s[-2] == '\n') )
3328 if (PL_in_eval && !PL_rsfp) {
3333 if (strnEQ(s,"=cut",4)) {
3347 PL_doextract = TRUE;
3350 if (PL_lex_brackets < PL_lex_formbrack) {
3352 #ifdef PERL_STRICT_CR
3353 for (t = s; SPACE_OR_TAB(*t); t++) ;
3355 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3357 if (*t == '\n' || *t == '#') {
3375 if (PL_expect != XOPERATOR) {
3376 if (s[1] != '<' && !strchr(s,'>'))
3379 s = scan_heredoc(s);
3381 s = scan_inputsymbol(s);
3382 TERM(sublex_start());
3387 SHop(OP_LEFT_SHIFT);
3401 SHop(OP_RIGHT_SHIFT);
3410 if (PL_expect == XOPERATOR) {
3411 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3414 return ','; /* grandfather non-comma-format format */
3418 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3419 PL_tokenbuf[0] = '@';
3420 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3421 sizeof PL_tokenbuf - 1, FALSE);
3422 if (PL_expect == XOPERATOR)
3423 no_op("Array length", s);
3424 if (!PL_tokenbuf[1])
3426 PL_expect = XOPERATOR;
3427 PL_pending_ident = '#';
3431 PL_tokenbuf[0] = '$';
3432 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3433 sizeof PL_tokenbuf - 1, FALSE);
3434 if (PL_expect == XOPERATOR)
3436 if (!PL_tokenbuf[1]) {
3438 yyerror("Final $ should be \\$ or $name");
3442 /* This kludge not intended to be bulletproof. */
3443 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3444 yylval.opval = newSVOP(OP_CONST, 0,
3445 newSViv(PL_compiling.cop_arybase));
3446 yylval.opval->op_private = OPpCONST_ARYBASE;
3452 if (PL_lex_state == LEX_NORMAL)
3455 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3458 PL_tokenbuf[0] = '@';
3459 if (ckWARN(WARN_SYNTAX)) {
3461 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3464 PL_bufptr = skipspace(PL_bufptr);
3465 while (t < PL_bufend && *t != ']')
3467 Perl_warner(aTHX_ WARN_SYNTAX,
3468 "Multidimensional syntax %.*s not supported",
3469 (t - PL_bufptr) + 1, PL_bufptr);
3473 else if (*s == '{') {
3474 PL_tokenbuf[0] = '%';
3475 if (ckWARN(WARN_SYNTAX) && strEQ(PL_tokenbuf+1, "SIG") &&
3476 (t = strchr(s, '}')) && (t = strchr(t, '=')))
3478 char tmpbuf[sizeof PL_tokenbuf];
3480 for (t++; isSPACE(*t); t++) ;
3481 if (isIDFIRST_lazy_if(t,UTF)) {
3482 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3483 for (; isSPACE(*t); t++) ;
3484 if (*t == ';' && get_cv(tmpbuf, FALSE))
3485 Perl_warner(aTHX_ WARN_SYNTAX,
3486 "You need to quote \"%s\"", tmpbuf);
3492 PL_expect = XOPERATOR;
3493 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3494 bool islop = (PL_last_lop == PL_oldoldbufptr);
3495 if (!islop || PL_last_lop_op == OP_GREPSTART)
3496 PL_expect = XOPERATOR;
3497 else if (strchr("$@\"'`q", *s))
3498 PL_expect = XTERM; /* e.g. print $fh "foo" */
3499 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3500 PL_expect = XTERM; /* e.g. print $fh &sub */
3501 else if (isIDFIRST_lazy_if(s,UTF)) {
3502 char tmpbuf[sizeof PL_tokenbuf];
3503 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3504 if ((tmp = keyword(tmpbuf, len))) {
3505 /* binary operators exclude handle interpretations */
3517 PL_expect = XTERM; /* e.g. print $fh length() */
3522 GV *gv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
3523 if (gv && GvCVu(gv))
3524 PL_expect = XTERM; /* e.g. print $fh subr() */
3527 else if (isDIGIT(*s))
3528 PL_expect = XTERM; /* e.g. print $fh 3 */
3529 else if (*s == '.' && isDIGIT(s[1]))
3530 PL_expect = XTERM; /* e.g. print $fh .3 */
3531 else if (strchr("/?-+", *s) && !isSPACE(s[1]) && s[1] != '=')
3532 PL_expect = XTERM; /* e.g. print $fh -1 */
3533 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3534 PL_expect = XTERM; /* print $fh <<"EOF" */
3536 PL_pending_ident = '$';
3540 if (PL_expect == XOPERATOR)
3542 PL_tokenbuf[0] = '@';
3543 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3544 if (!PL_tokenbuf[1]) {
3546 yyerror("Final @ should be \\@ or @name");
3549 if (PL_lex_state == LEX_NORMAL)
3551 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3553 PL_tokenbuf[0] = '%';
3555 /* Warn about @ where they meant $. */
3556 if (ckWARN(WARN_SYNTAX)) {
3557 if (*s == '[' || *s == '{') {
3559 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3561 if (*t == '}' || *t == ']') {
3563 PL_bufptr = skipspace(PL_bufptr);
3564 Perl_warner(aTHX_ WARN_SYNTAX,
3565 "Scalar value %.*s better written as $%.*s",
3566 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3571 PL_pending_ident = '@';
3574 case '/': /* may either be division or pattern */
3575 case '?': /* may either be conditional or pattern */
3576 if (PL_expect != XOPERATOR) {
3577 /* Disable warning on "study /blah/" */
3578 if (PL_oldoldbufptr == PL_last_uni
3579 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3580 || memNE(PL_last_uni, "study", 5)
3581 || isALNUM_lazy_if(PL_last_uni+5,UTF)))
3583 s = scan_pat(s,OP_MATCH);
3584 TERM(sublex_start());
3592 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3593 #ifdef PERL_STRICT_CR
3596 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3598 && (s == PL_linestart || s[-1] == '\n') )
3600 PL_lex_formbrack = 0;
3604 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3610 yylval.ival = OPf_SPECIAL;
3616 if (PL_expect != XOPERATOR)
3621 case '0': case '1': case '2': case '3': case '4':
3622 case '5': case '6': case '7': case '8': case '9':
3623 s = scan_num(s, &yylval);
3624 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3625 "### Saw number in '%s'\n", s);
3627 if (PL_expect == XOPERATOR)
3632 s = scan_str(s,FALSE,FALSE);
3633 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3634 "### Saw string in '%s'\n", s);
3636 if (PL_expect == XOPERATOR) {
3637 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3640 return ','; /* grandfather non-comma-format format */
3646 missingterm((char*)0);
3647 yylval.ival = OP_CONST;
3648 TERM(sublex_start());
3651 s = scan_str(s,FALSE,FALSE);
3652 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3653 "### Saw string in '%s'\n", s);
3655 if (PL_expect == XOPERATOR) {
3656 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3659 return ','; /* grandfather non-comma-format format */
3665 missingterm((char*)0);
3666 yylval.ival = OP_CONST;
3667 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3668 if (*d == '$' || *d == '@' || *d == '\\' || UTF8_IS_CONTINUED(*d)) {
3669 yylval.ival = OP_STRINGIFY;
3673 TERM(sublex_start());
3676 s = scan_str(s,FALSE,FALSE);
3677 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3678 "### Saw backtick string in '%s'\n", s);
3680 if (PL_expect == XOPERATOR)
3681 no_op("Backticks",s);
3683 missingterm((char*)0);
3684 yylval.ival = OP_BACKTICK;
3686 TERM(sublex_start());
3690 if (ckWARN(WARN_SYNTAX) && PL_lex_inwhat && isDIGIT(*s))
3691 Perl_warner(aTHX_ WARN_SYNTAX,"Can't use \\%c to mean $%c in expression",
3693 if (PL_expect == XOPERATOR)
3694 no_op("Backslash",s);
3698 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3702 while (isDIGIT(*start) || *start == '_')
3704 if (*start == '.' && isDIGIT(start[1])) {
3705 s = scan_num(s, &yylval);
3708 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3709 else if (!isALPHA(*start) && (PL_expect == XTERM || PL_expect == XREF)) {
3713 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3716 s = scan_num(s, &yylval);
3723 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
3762 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3764 /* Some keywords can be followed by any delimiter, including ':' */
3765 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
3766 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
3767 (PL_tokenbuf[0] == 'q' &&
3768 strchr("qwxr", PL_tokenbuf[1])))));
3770 /* x::* is just a word, unless x is "CORE" */
3771 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
3775 while (d < PL_bufend && isSPACE(*d))
3776 d++; /* no comments skipped here, or s### is misparsed */
3778 /* Is this a label? */
3779 if (!tmp && PL_expect == XSTATE
3780 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
3782 yylval.pval = savepv(PL_tokenbuf);
3787 /* Check for keywords */
3788 tmp = keyword(PL_tokenbuf, len);
3790 /* Is this a word before a => operator? */
3791 if (*d == '=' && d[1] == '>') {
3793 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf,0));
3794 yylval.opval->op_private = OPpCONST_BARE;
3795 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3796 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3800 if (tmp < 0) { /* second-class keyword? */
3801 GV *ogv = Nullgv; /* override (winner) */
3802 GV *hgv = Nullgv; /* hidden (loser) */
3803 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
3805 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
3808 if (GvIMPORTED_CV(gv))
3810 else if (! CvMETHOD(cv))
3814 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
3815 (gv = *gvp) != (GV*)&PL_sv_undef &&
3816 GvCVu(gv) && GvIMPORTED_CV(gv))
3822 tmp = 0; /* overridden by import or by GLOBAL */
3825 && -tmp==KEY_lock /* XXX generalizable kludge */
3827 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
3829 tmp = 0; /* any sub overrides "weak" keyword */
3831 else { /* no override */
3835 if (ckWARN(WARN_AMBIGUOUS) && hgv
3836 && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
3837 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3838 "Ambiguous call resolved as CORE::%s(), %s",
3839 GvENAME(hgv), "qualify as such or use &");
3846 default: /* not a keyword */
3849 char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
3851 /* Get the rest if it looks like a package qualifier */
3853 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
3855 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
3858 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
3859 *s == '\'' ? "'" : "::");
3863 if (PL_expect == XOPERATOR) {
3864 if (PL_bufptr == PL_linestart) {
3865 CopLINE_dec(PL_curcop);
3866 Perl_warner(aTHX_ WARN_SEMICOLON, PL_warn_nosemi);
3867 CopLINE_inc(PL_curcop);
3870 no_op("Bareword",s);
3873 /* Look for a subroutine with this name in current package,
3874 unless name is "Foo::", in which case Foo is a bearword
3875 (and a package name). */
3878 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
3880 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
3881 Perl_warner(aTHX_ WARN_BAREWORD,
3882 "Bareword \"%s\" refers to nonexistent package",
3885 PL_tokenbuf[len] = '\0';
3892 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
3895 /* if we saw a global override before, get the right name */
3898 sv = newSVpvn("CORE::GLOBAL::",14);
3899 sv_catpv(sv,PL_tokenbuf);
3902 sv = newSVpv(PL_tokenbuf,0);
3904 /* Presume this is going to be a bareword of some sort. */
3907 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3908 yylval.opval->op_private = OPpCONST_BARE;
3910 /* And if "Foo::", then that's what it certainly is. */
3915 /* See if it's the indirect object for a list operator. */
3917 if (PL_oldoldbufptr &&
3918 PL_oldoldbufptr < PL_bufptr &&
3919 (PL_oldoldbufptr == PL_last_lop
3920 || PL_oldoldbufptr == PL_last_uni) &&
3921 /* NO SKIPSPACE BEFORE HERE! */
3922 (PL_expect == XREF ||
3923 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
3925 bool immediate_paren = *s == '(';
3927 /* (Now we can afford to cross potential line boundary.) */
3930 /* Two barewords in a row may indicate method call. */
3932 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
3935 /* If not a declared subroutine, it's an indirect object. */
3936 /* (But it's an indir obj regardless for sort.) */
3938 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
3939 ((!gv || !GvCVu(gv)) &&
3940 (PL_last_lop_op != OP_MAPSTART &&
3941 PL_last_lop_op != OP_GREPSTART))))
3943 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
3949 PL_expect = XOPERATOR;
3952 /* Is this a word before a => operator? */
3953 if (*s == '=' && s[1] == '>') {
3955 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
3956 if (UTF && !IN_BYTE && is_utf8_string((U8*)PL_tokenbuf, len))
3957 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
3961 /* If followed by a paren, it's certainly a subroutine. */
3964 if (gv && GvCVu(gv)) {
3965 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
3966 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
3971 PL_nextval[PL_nexttoke].opval = yylval.opval;
3972 PL_expect = XOPERATOR;
3978 /* If followed by var or block, call it a method (unless sub) */
3980 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
3981 PL_last_lop = PL_oldbufptr;
3982 PL_last_lop_op = OP_METHOD;
3986 /* If followed by a bareword, see if it looks like indir obj. */
3988 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp = intuit_method(s,gv)))
3991 /* Not a method, so call it a subroutine (if defined) */
3993 if (gv && GvCVu(gv)) {
3995 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
3996 Perl_warner(aTHX_ WARN_AMBIGUOUS,
3997 "Ambiguous use of -%s resolved as -&%s()",
3998 PL_tokenbuf, PL_tokenbuf);
3999 /* Check for a constant sub */
4001 if ((sv = cv_const_sv(cv))) {
4003 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
4004 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
4005 yylval.opval->op_private = 0;
4009 /* Resolve to GV now. */
4010 op_free(yylval.opval);
4011 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4012 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4013 PL_last_lop = PL_oldbufptr;
4014 PL_last_lop_op = OP_ENTERSUB;
4015 /* Is there a prototype? */
4018 char *proto = SvPV((SV*)cv, len);
4021 if (strEQ(proto, "$"))
4023 if (*proto == '&' && *s == '{') {
4024 sv_setpv(PL_subname,"__ANON__");
4028 PL_nextval[PL_nexttoke].opval = yylval.opval;
4034 /* Call it a bare word */
4036 if (PL_hints & HINT_STRICT_SUBS)
4037 yylval.opval->op_private |= OPpCONST_STRICT;
4040 if (ckWARN(WARN_RESERVED)) {
4041 if (lastchar != '-') {
4042 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4044 Perl_warner(aTHX_ WARN_RESERVED, PL_warn_reserved,
4051 if (lastchar && strchr("*%&", lastchar) && ckWARN_d(WARN_AMBIGUOUS)) {
4052 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4053 "Operator or semicolon missing before %c%s",
4054 lastchar, PL_tokenbuf);
4055 Perl_warner(aTHX_ WARN_AMBIGUOUS,
4056 "Ambiguous use of %c resolved as operator %c",
4057 lastchar, lastchar);
4063 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4064 newSVpv(CopFILE(PL_curcop),0));
4068 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4069 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4072 case KEY___PACKAGE__:
4073 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4075 ? newSVsv(PL_curstname)
4084 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4085 char *pname = "main";
4086 if (PL_tokenbuf[2] == 'D')
4087 pname = HvNAME(PL_curstash ? PL_curstash : PL_defstash);
4088 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4091 GvIOp(gv) = newIO();
4092 IoIFP(GvIOp(gv)) = PL_rsfp;
4093 #if defined(HAS_FCNTL) && defined(F_SETFD)
4095 int fd = PerlIO_fileno(PL_rsfp);
4096 fcntl(fd,F_SETFD,fd >= 3);
4099 /* Mark this internal pseudo-handle as clean */
4100 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4102 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4103 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4104 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4106 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4107 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4108 /* if the script was opened in binmode, we need to revert
4109 * it to text mode for compatibility; but only iff it has CRs
4110 * XXX this is a questionable hack at best. */
4111 if (PL_bufend-PL_bufptr > 2
4112 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4115 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4116 loc = PerlIO_tell(PL_rsfp);
4117 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4119 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4120 #if defined(__BORLANDC__)
4121 /* XXX see note in do_binmode() */
4122 ((FILE*)PL_rsfp)->flags |= _F_BIN;
4125 PerlIO_seek(PL_rsfp, loc, 0);
4129 #ifdef PERLIO_LAYERS
4130 if (UTF && !IN_BYTE)
4131 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4144 if (PL_expect == XSTATE) {
4151 if (*s == ':' && s[1] == ':') {
4154 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4155 if (!(tmp = keyword(PL_tokenbuf, len)))
4156 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4170 LOP(OP_ACCEPT,XTERM);
4176 LOP(OP_ATAN2,XTERM);
4182 LOP(OP_BINMODE,XTERM);
4185 LOP(OP_BLESS,XTERM);
4194 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4211 if (!PL_cryptseen) {
4212 PL_cryptseen = TRUE;
4216 LOP(OP_CRYPT,XTERM);
4219 if (ckWARN(WARN_CHMOD)) {
4220 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
4221 if (*d != '0' && isDIGIT(*d))
4222 Perl_warner(aTHX_ WARN_CHMOD,
4223 "chmod() mode argument is missing initial 0");
4225 LOP(OP_CHMOD,XTERM);
4228 LOP(OP_CHOWN,XTERM);
4231 LOP(OP_CONNECT,XTERM);
4247 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4251 PL_hints |= HINT_BLOCK_SCOPE;
4261 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4262 LOP(OP_DBMOPEN,XTERM);
4268 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4275 yylval.ival = CopLINE(PL_curcop);
4289 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4290 UNIBRACK(OP_ENTEREVAL);
4305 case KEY_endhostent:
4311 case KEY_endservent:
4314 case KEY_endprotoent:
4325 yylval.ival = CopLINE(PL_curcop);
4327 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4329 if ((PL_bufend - p) >= 3 &&
4330 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4332 else if ((PL_bufend - p) >= 4 &&
4333 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
4336 if (isIDFIRST_lazy_if(p,UTF)) {
4337 p = scan_ident(p, PL_bufend,
4338 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4342 Perl_croak(aTHX_ "Missing $ on loop variable");
4347 LOP(OP_FORMLINE,XTERM);
4353 LOP(OP_FCNTL,XTERM);
4359 LOP(OP_FLOCK,XTERM);
4368 LOP(OP_GREPSTART, XREF);
4371 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4386 case KEY_getpriority:
4387 LOP(OP_GETPRIORITY,XTERM);
4389 case KEY_getprotobyname:
4392 case KEY_getprotobynumber:
4393 LOP(OP_GPBYNUMBER,XTERM);
4395 case KEY_getprotoent:
4407 case KEY_getpeername:
4408 UNI(OP_GETPEERNAME);
4410 case KEY_gethostbyname:
4413 case KEY_gethostbyaddr:
4414 LOP(OP_GHBYADDR,XTERM);
4416 case KEY_gethostent:
4419 case KEY_getnetbyname:
4422 case KEY_getnetbyaddr:
4423 LOP(OP_GNBYADDR,XTERM);
4428 case KEY_getservbyname:
4429 LOP(OP_GSBYNAME,XTERM);
4431 case KEY_getservbyport:
4432 LOP(OP_GSBYPORT,XTERM);
4434 case KEY_getservent:
4437 case KEY_getsockname:
4438 UNI(OP_GETSOCKNAME);
4440 case KEY_getsockopt:
4441 LOP(OP_GSOCKOPT,XTERM);
4463 yylval.ival = CopLINE(PL_curcop);
4467 LOP(OP_INDEX,XTERM);
4473 LOP(OP_IOCTL,XTERM);
4485 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4517 LOP(OP_LISTEN,XTERM);
4526 s = scan_pat(s,OP_MATCH);
4527 TERM(sublex_start());
4530 LOP(OP_MAPSTART, XREF);
4533 LOP(OP_MKDIR,XTERM);
4536 LOP(OP_MSGCTL,XTERM);
4539 LOP(OP_MSGGET,XTERM);
4542 LOP(OP_MSGRCV,XTERM);
4545 LOP(OP_MSGSND,XTERM);
4551 if (isIDFIRST_lazy_if(s,UTF)) {
4552 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
4553 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
4555 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
4556 if (!PL_in_my_stash) {
4559 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
4567 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4574 if (PL_expect != XSTATE)
4575 yyerror("\"no\" not allowed in expression");
4576 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4577 s = force_version(s);
4582 if (*s == '(' || (s = skipspace(s), *s == '('))
4589 if (isIDFIRST_lazy_if(s,UTF)) {
4591 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
4593 if (strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE))
4594 Perl_warner(aTHX_ WARN_PRECEDENCE,
4595 "Precedence problem: open %.*s should be open(%.*s)",
4601 yylval.ival = OP_OR;
4611 LOP(OP_OPEN_DIR,XTERM);
4614 checkcomma(s,PL_tokenbuf,"filehandle");
4618 checkcomma(s,PL_tokenbuf,"filehandle");
4637 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4641 LOP(OP_PIPE_OP,XTERM);
4644 s = scan_str(s,FALSE,FALSE);
4646 missingterm((char*)0);
4647 yylval.ival = OP_CONST;
4648 TERM(sublex_start());
4654 s = scan_str(s,FALSE,FALSE);
4656 missingterm((char*)0);
4658 if (SvCUR(PL_lex_stuff)) {
4661 d = SvPV_force(PL_lex_stuff, len);
4664 for (; isSPACE(*d) && len; --len, ++d) ;
4667 if (!warned && ckWARN(WARN_QW)) {
4668 for (; !isSPACE(*d) && len; --len, ++d) {
4670 Perl_warner(aTHX_ WARN_QW,
4671 "Possible attempt to separate words with commas");
4674 else if (*d == '#') {
4675 Perl_warner(aTHX_ WARN_QW,
4676 "Possible attempt to put comments in qw() list");
4682 for (; !isSPACE(*d) && len; --len, ++d) ;
4684 sv = newSVpvn(b, d-b);
4685 if (DO_UTF8(PL_lex_stuff))
4687 words = append_elem(OP_LIST, words,
4688 newSVOP(OP_CONST, 0, tokeq(sv)));
4692 PL_nextval[PL_nexttoke].opval = words;
4697 SvREFCNT_dec(PL_lex_stuff);
4698 PL_lex_stuff = Nullsv;
4703 s = scan_str(s,FALSE,FALSE);
4705 missingterm((char*)0);
4706 yylval.ival = OP_STRINGIFY;
4707 if (SvIVX(PL_lex_stuff) == '\'')
4708 SvIVX(PL_lex_stuff) = 0; /* qq'$foo' should intepolate */
4709 TERM(sublex_start());
4712 s = scan_pat(s,OP_QR);
4713 TERM(sublex_start());
4716 s = scan_str(s,FALSE,FALSE);
4718 missingterm((char*)0);
4719 yylval.ival = OP_BACKTICK;
4721 TERM(sublex_start());
4728 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4729 s = force_version(s);
4732 *PL_tokenbuf = '\0';
4733 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4734 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
4735 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
4737 yyerror("<> should be quotes");
4745 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4749 LOP(OP_RENAME,XTERM);
4758 LOP(OP_RINDEX,XTERM);
4781 LOP(OP_REVERSE,XTERM);
4792 TERM(sublex_start());
4794 TOKEN(1); /* force error */
4803 LOP(OP_SELECT,XTERM);
4809 LOP(OP_SEMCTL,XTERM);
4812 LOP(OP_SEMGET,XTERM);
4815 LOP(OP_SEMOP,XTERM);
4821 LOP(OP_SETPGRP,XTERM);
4823 case KEY_setpriority:
4824 LOP(OP_SETPRIORITY,XTERM);
4826 case KEY_sethostent:
4832 case KEY_setservent:
4835 case KEY_setprotoent:
4845 LOP(OP_SEEKDIR,XTERM);
4847 case KEY_setsockopt:
4848 LOP(OP_SSOCKOPT,XTERM);
4854 LOP(OP_SHMCTL,XTERM);
4857 LOP(OP_SHMGET,XTERM);
4860 LOP(OP_SHMREAD,XTERM);
4863 LOP(OP_SHMWRITE,XTERM);
4866 LOP(OP_SHUTDOWN,XTERM);
4875 LOP(OP_SOCKET,XTERM);
4877 case KEY_socketpair:
4878 LOP(OP_SOCKPAIR,XTERM);
4881 checkcomma(s,PL_tokenbuf,"subroutine name");
4883 if (*s == ';' || *s == ')') /* probably a close */
4884 Perl_croak(aTHX_ "sort is now a reserved word");
4886 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4890 LOP(OP_SPLIT,XTERM);
4893 LOP(OP_SPRINTF,XTERM);
4896 LOP(OP_SPLICE,XTERM);
4911 LOP(OP_SUBSTR,XTERM);
4917 char tmpbuf[sizeof PL_tokenbuf];
4919 expectation attrful;
4920 bool have_name, have_proto;
4925 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
4926 (*s == ':' && s[1] == ':'))
4929 attrful = XATTRBLOCK;
4930 /* remember buffer pos'n for later force_word */
4931 tboffset = s - PL_oldbufptr;
4932 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4933 if (strchr(tmpbuf, ':'))
4934 sv_setpv(PL_subname, tmpbuf);
4936 sv_setsv(PL_subname,PL_curstname);
4937 sv_catpvn(PL_subname,"::",2);
4938 sv_catpvn(PL_subname,tmpbuf,len);
4945 Perl_croak(aTHX_ "Missing name in \"my sub\"");
4946 PL_expect = XTERMBLOCK;
4947 attrful = XATTRTERM;
4948 sv_setpv(PL_subname,"?");
4952 if (key == KEY_format) {
4954 PL_lex_formbrack = PL_lex_brackets + 1;
4956 (void) force_word(PL_oldbufptr + tboffset, WORD,
4961 /* Look for a prototype */
4965 s = scan_str(s,FALSE,FALSE);
4968 SvREFCNT_dec(PL_lex_stuff);
4969 PL_lex_stuff = Nullsv;
4970 Perl_croak(aTHX_ "Prototype not terminated");
4973 d = SvPVX(PL_lex_stuff);
4975 for (p = d; *p; ++p) {
4980 SvCUR(PL_lex_stuff) = tmp;
4988 if (*s == ':' && s[1] != ':')
4989 PL_expect = attrful;
4992 PL_nextval[PL_nexttoke].opval =
4993 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
4994 PL_lex_stuff = Nullsv;
4998 sv_setpv(PL_subname,"__ANON__");
5001 (void) force_word(PL_oldbufptr + tboffset, WORD,
5010 LOP(OP_SYSTEM,XREF);
5013 LOP(OP_SYMLINK,XTERM);
5016 LOP(OP_SYSCALL,XTERM);
5019 LOP(OP_SYSOPEN,XTERM);
5022 LOP(OP_SYSSEEK,XTERM);
5025 LOP(OP_SYSREAD,XTERM);
5028 LOP(OP_SYSWRITE,XTERM);
5032 TERM(sublex_start());
5053 LOP(OP_TRUNCATE,XTERM);
5065 yylval.ival = CopLINE(PL_curcop);
5069 yylval.ival = CopLINE(PL_curcop);
5073 LOP(OP_UNLINK,XTERM);
5079 LOP(OP_UNPACK,XTERM);
5082 LOP(OP_UTIME,XTERM);
5085 if (ckWARN(WARN_UMASK)) {
5086 for (d = s; d < PL_bufend && (isSPACE(*d) || *d == '('); d++) ;
5087 if (*d != '0' && isDIGIT(*d))
5088 Perl_warner(aTHX_ WARN_UMASK,
5089 "umask: argument is missing initial 0");
5094 LOP(OP_UNSHIFT,XTERM);
5097 if (PL_expect != XSTATE)
5098 yyerror("\"use\" not allowed in expression");
5100 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
5101 s = force_version(s);
5102 if (*s == ';' || (s = skipspace(s), *s == ';')) {
5103 PL_nextval[PL_nexttoke].opval = Nullop;
5108 s = force_word(s,WORD,FALSE,TRUE,FALSE);
5109 s = force_version(s);
5121 yylval.ival = CopLINE(PL_curcop);
5125 PL_hints |= HINT_BLOCK_SCOPE;
5132 LOP(OP_WAITPID,XTERM);
5140 static char ctl_l[2];
5142 if (ctl_l[0] == '\0')
5143 ctl_l[0] = toCTRL('L');
5144 gv_fetchpv(ctl_l,TRUE, SVt_PV);
5147 gv_fetchpv("\f",TRUE, SVt_PV); /* Make sure $^L is defined */
5152 if (PL_expect == XOPERATOR)
5158 yylval.ival = OP_XOR;
5163 TERM(sublex_start());
5168 #pragma segment Main
5172 Perl_keyword(pTHX_ register char *d, I32 len)
5177 if (strEQ(d,"__FILE__")) return -KEY___FILE__;
5178 if (strEQ(d,"__LINE__")) return -KEY___LINE__;
5179 if (strEQ(d,"__PACKAGE__")) return -KEY___PACKAGE__;
5180 if (strEQ(d,"__DATA__")) return KEY___DATA__;
5181 if (strEQ(d,"__END__")) return KEY___END__;
5185 if (strEQ(d,"AUTOLOAD")) return KEY_AUTOLOAD;
5190 if (strEQ(d,"and")) return -KEY_and;
5191 if (strEQ(d,"abs")) return -KEY_abs;
5194 if (strEQ(d,"alarm")) return -KEY_alarm;
5195 if (strEQ(d,"atan2")) return -KEY_atan2;
5198 if (strEQ(d,"accept")) return -KEY_accept;
5203 if (strEQ(d,"BEGIN")) return KEY_BEGIN;
5206 if (strEQ(d,"bless")) return -KEY_bless;
5207 if (strEQ(d,"bind")) return -KEY_bind;
5208 if (strEQ(d,"binmode")) return -KEY_binmode;
5211 if (strEQ(d,"CORE")) return -KEY_CORE;
5212 if (strEQ(d,"CHECK")) return KEY_CHECK;
5217 if (strEQ(d,"cmp")) return -KEY_cmp;
5218 if (strEQ(d,"chr")) return -KEY_chr;
5219 if (strEQ(d,"cos")) return -KEY_cos;
5222 if (strEQ(d,"chop")) return -KEY_chop;
5225 if (strEQ(d,"close")) return -KEY_close;
5226 if (strEQ(d,"chdir")) return -KEY_chdir;
5227 if (strEQ(d,"chomp")) return -KEY_chomp;
5228 if (strEQ(d,"chmod")) return -KEY_chmod;
5229 if (strEQ(d,"chown")) return -KEY_chown;
5230 if (strEQ(d,"crypt")) return -KEY_crypt;
5233 if (strEQ(d,"chroot")) return -KEY_chroot;
5234 if (strEQ(d,"caller")) return -KEY_caller;
5237 if (strEQ(d,"connect")) return -KEY_connect;
5240 if (strEQ(d,"closedir")) return -KEY_closedir;
5241 if (strEQ(d,"continue")) return -KEY_continue;
5246 if (strEQ(d,"DESTROY")) return KEY_DESTROY;
5251 if (strEQ(d,"do")) return KEY_do;
5254 if (strEQ(d,"die")) return -KEY_die;
5257 if (strEQ(d,"dump")) return -KEY_dump;
5260 if (strEQ(d,"delete")) return KEY_delete;
5263 if (strEQ(d,"defined")) return KEY_defined;
5264 if (strEQ(d,"dbmopen")) return -KEY_dbmopen;
5267 if (strEQ(d,"dbmclose")) return -KEY_dbmclose;
5272 if (strEQ(d,"END")) return KEY_END;
5277 if (strEQ(d,"eq")) return -KEY_eq;
5280 if (strEQ(d,"eof")) return -KEY_eof;
5281 if (strEQ(d,"exp")) return -KEY_exp;
5284 if (strEQ(d,"else")) return KEY_else;
5285 if (strEQ(d,"exit")) return -KEY_exit;
5286 if (strEQ(d,"eval")) return KEY_eval;
5287 if (strEQ(d,"exec")) return -KEY_exec;
5288 if (strEQ(d,"each")) return -KEY_each;
5291 if (strEQ(d,"elsif")) return KEY_elsif;
5294 if (strEQ(d,"exists")) return KEY_exists;
5295 if (strEQ(d,"elseif")) Perl_warn(aTHX_ "elseif should be elsif");
5298 if (strEQ(d,"endgrent")) return -KEY_endgrent;
5299 if (strEQ(d,"endpwent")) return -KEY_endpwent;
5302 if (strEQ(d,"endnetent")) return -KEY_endnetent;
5305 if (strEQ(d,"endhostent")) return -KEY_endhostent;
5306 if (strEQ(d,"endservent")) return -KEY_endservent;
5309 if (strEQ(d,"endprotoent")) return -KEY_endprotoent;
5316 if (strEQ(d,"for")) return KEY_for;
5319 if (strEQ(d,"fork")) return -KEY_fork;
5322 if (strEQ(d,"fcntl")) return -KEY_fcntl;
5323 if (strEQ(d,"flock")) return -KEY_flock;
5326 if (strEQ(d,"format")) return KEY_format;
5327 if (strEQ(d,"fileno")) return -KEY_fileno;
5330 if (strEQ(d,"foreach")) return KEY_foreach;
5333 if (strEQ(d,"formline")) return -KEY_formline;
5338 if (strnEQ(d,"get",3)) {
5343 if (strEQ(d,"ppid")) return -KEY_getppid;
5344 if (strEQ(d,"pgrp")) return -KEY_getpgrp;
5347 if (strEQ(d,"pwent")) return -KEY_getpwent;
5348 if (strEQ(d,"pwnam")) return -KEY_getpwnam;
5349 if (strEQ(d,"pwuid")) return -KEY_getpwuid;
5352 if (strEQ(d,"peername")) return -KEY_getpeername;
5353 if (strEQ(d,"protoent")) return -KEY_getprotoent;
5354 if (strEQ(d,"priority")) return -KEY_getpriority;
5357 if (strEQ(d,"protobyname")) return -KEY_getprotobyname;
5360 if (strEQ(d,"protobynumber"))return -KEY_getprotobynumber;
5364 else if (*d == 'h') {
5365 if (strEQ(d,"hostbyname")) return -KEY_gethostbyname;
5366 if (strEQ(d,"hostbyaddr")) return -KEY_gethostbyaddr;
5367 if (strEQ(d,"hostent")) return -KEY_gethostent;
5369 else if (*d == 'n') {
5370 if (strEQ(d,"netbyname")) return -KEY_getnetbyname;
5371 if (strEQ(d,"netbyaddr")) return -KEY_getnetbyaddr;
5372 if (strEQ(d,"netent")) return -KEY_getnetent;
5374 else if (*d == 's') {
5375 if (strEQ(d,"servbyname")) return -KEY_getservbyname;
5376 if (strEQ(d,"servbyport")) return -KEY_getservbyport;
5377 if (strEQ(d,"servent")) return -KEY_getservent;
5378 if (strEQ(d,"sockname")) return -KEY_getsockname;
5379 if (strEQ(d,"sockopt")) return -KEY_getsockopt;
5381 else if (*d == 'g') {
5382 if (strEQ(d,"grent")) return -KEY_getgrent;
5383 if (strEQ(d,"grnam")) return -KEY_getgrnam;
5384 if (strEQ(d,"grgid")) return -KEY_getgrgid;
5386 else if (*d == 'l') {
5387 if (strEQ(d,"login")) return -KEY_getlogin;
5389 else if (strEQ(d,"c")) return -KEY_getc;
5394 if (strEQ(d,"gt")) return -KEY_gt;
5395 if (strEQ(d,"ge")) return -KEY_ge;
5398 if (strEQ(d,"grep")) return KEY_grep;
5399 if (strEQ(d,"goto")) return KEY_goto;
5400 if (strEQ(d,"glob")) return KEY_glob;
5403 if (strEQ(d,"gmtime")) return -KEY_gmtime;
5408 if (strEQ(d,"hex")) return -KEY_hex;
5411 if (strEQ(d,"INIT")) return KEY_INIT;
5416 if (strEQ(d,"if")) return KEY_if;
5419 if (strEQ(d,"int")) return -KEY_int;
5422 if (strEQ(d,"index")) return -KEY_index;
5423 if (strEQ(d,"ioctl")) return -KEY_ioctl;
5428 if (strEQ(d,"join")) return -KEY_join;
5432 if (strEQ(d,"keys")) return -KEY_keys;
5433 if (strEQ(d,"kill")) return -KEY_kill;
5439 if (strEQ(d,"lt")) return -KEY_lt;
5440 if (strEQ(d,"le")) return -KEY_le;
5441 if (strEQ(d,"lc")) return -KEY_lc;
5444 if (strEQ(d,"log")) return -KEY_log;
5447 if (strEQ(d,"last")) return KEY_last;
5448 if (strEQ(d,"link")) return -KEY_link;
5449 if (strEQ(d,"lock")) return -KEY_lock;
5452 if (strEQ(d,"local")) return KEY_local;
5453 if (strEQ(d,"lstat")) return -KEY_lstat;
5456 if (strEQ(d,"length")) return -KEY_length;
5457 if (strEQ(d,"listen")) return -KEY_listen;
5460 if (strEQ(d,"lcfirst")) return -KEY_lcfirst;
5463 if (strEQ(d,"localtime")) return -KEY_localtime;
5469 case 1: return KEY_m;
5471 if (strEQ(d,"my")) return KEY_my;
5474 if (strEQ(d,"map")) return KEY_map;
5477 if (strEQ(d,"mkdir")) return -KEY_mkdir;
5480 if (strEQ(d,"msgctl")) return -KEY_msgctl;
5481 if (strEQ(d,"msgget")) return -KEY_msgget;
5482 if (strEQ(d,"msgrcv")) return -KEY_msgrcv;
5483 if (strEQ(d,"msgsnd")) return -KEY_msgsnd;
5488 if (strEQ(d,"next")) return KEY_next;
5489 if (strEQ(d,"ne")) return -KEY_ne;
5490 if (strEQ(d,"not")) return -KEY_not;
5491 if (strEQ(d,"no")) return KEY_no;
5496 if (strEQ(d,"or")) return -KEY_or;
5499 if (strEQ(d,"ord")) return -KEY_ord;
5500 if (strEQ(d,"oct")) return -KEY_oct;
5501 if (strEQ(d,"our")) return KEY_our;
5504 if (strEQ(d,"open")) return -KEY_open;
5507 if (strEQ(d,"opendir")) return -KEY_opendir;
5514 if (strEQ(d,"pop")) return -KEY_pop;
5515 if (strEQ(d,"pos")) return KEY_pos;
5518 if (strEQ(d,"push")) return -KEY_push;
5519 if (strEQ(d,"pack")) return -KEY_pack;
5520 if (strEQ(d,"pipe")) return -KEY_pipe;
5523 if (strEQ(d,"print")) return KEY_print;
5526 if (strEQ(d,"printf")) return KEY_printf;
5529 if (strEQ(d,"package")) return KEY_package;
5532 if (strEQ(d,"prototype")) return KEY_prototype;
5537 if (strEQ(d,"q")) return KEY_q;
5538 if (strEQ(d,"qr")) return KEY_qr;
5539 if (strEQ(d,"qq")) return KEY_qq;
5540 if (strEQ(d,"qw")) return KEY_qw;
5541 if (strEQ(d,"qx")) return KEY_qx;
5543 else if (strEQ(d,"quotemeta")) return -KEY_quotemeta;
5548 if (strEQ(d,"ref")) return -KEY_ref;
5551 if (strEQ(d,"read")) return -KEY_read;
5552 if (strEQ(d,"rand")) return -KEY_rand;
5553 if (strEQ(d,"recv")) return -KEY_recv;
5554 if (strEQ(d,"redo")) return KEY_redo;
5557 if (strEQ(d,"rmdir")) return -KEY_rmdir;
5558 if (strEQ(d,"reset")) return -KEY_reset;
5561 if (strEQ(d,"return")) return KEY_return;
5562 if (strEQ(d,"rename")) return -KEY_rename;
5563 if (strEQ(d,"rindex")) return -KEY_rindex;
5566 if (strEQ(d,"require")) return -KEY_require;
5567 if (strEQ(d,"reverse")) return -KEY_reverse;
5568 if (strEQ(d,"readdir")) return -KEY_readdir;
5571 if (strEQ(d,"readlink")) return -KEY_readlink;
5572 if (strEQ(d,"readline")) return -KEY_readline;
5573 if (strEQ(d,"readpipe")) return -KEY_readpipe;
5576 if (strEQ(d,"rewinddir")) return -KEY_rewinddir;
5582 case 0: return KEY_s;
5584 if (strEQ(d,"scalar")) return KEY_scalar;
5589 if (strEQ(d,"seek")) return -KEY_seek;
5590 if (strEQ(d,"send")) return -KEY_send;
5593 if (strEQ(d,"semop")) return -KEY_semop;
5596 if (strEQ(d,"select")) return -KEY_select;
5597 if (strEQ(d,"semctl")) return -KEY_semctl;
5598 if (strEQ(d,"semget")) return -KEY_semget;
5601 if (strEQ(d,"setpgrp")) return -KEY_setpgrp;
5602 if (strEQ(d,"seekdir")) return -KEY_seekdir;
5605 if (strEQ(d,"setpwent")) return -KEY_setpwent;
5606 if (strEQ(d,"setgrent")) return -KEY_setgrent;
5609 if (strEQ(d,"setnetent")) return -KEY_setnetent;
5612 if (strEQ(d,"setsockopt")) return -KEY_setsockopt;
5613 if (strEQ(d,"sethostent")) return -KEY_sethostent;
5614 if (strEQ(d,"setservent")) return -KEY_setservent;
5617 if (strEQ(d,"setpriority")) return -KEY_setpriority;
5618 if (strEQ(d,"setprotoent")) return -KEY_setprotoent;
5625 if (strEQ(d,"shift")) return -KEY_shift;
5628 if (strEQ(d,"shmctl")) return -KEY_shmctl;
5629 if (strEQ(d,"shmget")) return -KEY_shmget;
5632 if (strEQ(d,"shmread")) return -KEY_shmread;
5635 if (strEQ(d,"shmwrite")) return -KEY_shmwrite;
5636 if (strEQ(d,"shutdown")) return -KEY_shutdown;
5641 if (strEQ(d,"sin")) return -KEY_sin;
5644 if (strEQ(d,"sleep")) return -KEY_sleep;
5647 if (strEQ(d,"sort")) return KEY_sort;
5648 if (strEQ(d,"socket")) return -KEY_socket;
5649 if (strEQ(d,"socketpair")) return -KEY_socketpair;
5652 if (strEQ(d,"split")) return KEY_split;
5653 if (strEQ(d,"sprintf")) return -KEY_sprintf;
5654 if (strEQ(d,"splice")) return -KEY_splice;
5657 if (strEQ(d,"sqrt")) return -KEY_sqrt;
5660 if (strEQ(d,"srand")) return -KEY_srand;
5663 if (strEQ(d,"stat")) return -KEY_stat;
5664 if (strEQ(d,"study")) return KEY_study;
5667 if (strEQ(d,"substr")) return -KEY_substr;
5668 if (strEQ(d,"sub")) return KEY_sub;
5673 if (strEQ(d,"system")) return -KEY_system;
5676 if (strEQ(d,"symlink")) return -KEY_symlink;
5677 if (strEQ(d,"syscall")) return -KEY_syscall;
5678 if (strEQ(d,"sysopen")) return -KEY_sysopen;
5679 if (strEQ(d,"sysread")) return -KEY_sysread;
5680 if (strEQ(d,"sysseek")) return -KEY_sysseek;
5683 if (strEQ(d,"syswrite")) return -KEY_syswrite;
5692 if (strEQ(d,"tr")) return KEY_tr;
5695 if (strEQ(d,"tie")) return KEY_tie;
5698 if (strEQ(d,"tell")) return -KEY_tell;
5699 if (strEQ(d,"tied")) return KEY_tied;
5700 if (strEQ(d,"time")) return -KEY_time;
5703 if (strEQ(d,"times")) return -KEY_times;
5706 if (strEQ(d,"telldir")) return -KEY_telldir;
5709 if (strEQ(d,"truncate")) return -KEY_truncate;
5716 if (strEQ(d,"uc")) return -KEY_uc;
5719 if (strEQ(d,"use")) return KEY_use;
5722 if (strEQ(d,"undef")) return KEY_undef;
5723 if (strEQ(d,"until")) return KEY_until;
5724 if (strEQ(d,"untie")) return KEY_untie;
5725 if (strEQ(d,"utime")) return -KEY_utime;
5726 if (strEQ(d,"umask")) return -KEY_umask;
5729 if (strEQ(d,"unless")) return KEY_unless;
5730 if (strEQ(d,"unpack")) return -KEY_unpack;
5731 if (strEQ(d,"unlink")) return -KEY_unlink;
5734 if (strEQ(d,"unshift")) return -KEY_unshift;
5735 if (strEQ(d,"ucfirst")) return -KEY_ucfirst;
5740 if (strEQ(d,"values")) return -KEY_values;
5741 if (strEQ(d,"vec")) return -KEY_vec;
5746 if (strEQ(d,"warn")) return -KEY_warn;
5747 if (strEQ(d,"wait")) return -KEY_wait;
5750 if (strEQ(d,"while")) return KEY_while;
5751 if (strEQ(d,"write")) return -KEY_write;
5754 if (strEQ(d,"waitpid")) return -KEY_waitpid;
5757 if (strEQ(d,"wantarray")) return -KEY_wantarray;
5762 if (len == 1) return -KEY_x;
5763 if (strEQ(d,"xor")) return -KEY_xor;
5766 if (len == 1) return KEY_y;
5775 S_checkcomma(pTHX_ register char *s, char *name, char *what)
5779 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
5780 if (ckWARN(WARN_SYNTAX)) {
5782 for (w = s+2; *w && level; w++) {
5789 for (; *w && isSPACE(*w); w++) ;
5790 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
5791 Perl_warner(aTHX_ WARN_SYNTAX,
5792 "%s (...) interpreted as function",name);
5795 while (s < PL_bufend && isSPACE(*s))
5799 while (s < PL_bufend && isSPACE(*s))
5801 if (isIDFIRST_lazy_if(s,UTF)) {
5803 while (isALNUM_lazy_if(s,UTF))
5805 while (s < PL_bufend && isSPACE(*s))
5810 kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
5814 Perl_croak(aTHX_ "No comma allowed after %s", what);
5819 /* Either returns sv, or mortalizes sv and returns a new SV*.
5820 Best used as sv=new_constant(..., sv, ...).
5821 If s, pv are NULL, calls subroutine with one argument,
5822 and type is used with error messages only. */
5825 S_new_constant(pTHX_ char *s, STRLEN len, const char *key, SV *sv, SV *pv,
5829 HV *table = GvHV(PL_hintgv); /* ^H */
5833 const char *why1, *why2, *why3;
5835 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
5838 why2 = strEQ(key,"charnames")
5839 ? "(possibly a missing \"use charnames ...\")"
5841 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
5842 (type ? type: "undef"), why2);
5844 /* This is convoluted and evil ("goto considered harmful")
5845 * but I do not understand the intricacies of all the different
5846 * failure modes of %^H in here. The goal here is to make
5847 * the most probable error message user-friendly. --jhi */
5852 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
5853 (type ? type: "undef"), why1, why2, why3);
5855 yyerror(SvPVX(msg));
5859 cvp = hv_fetch(table, key, strlen(key), FALSE);
5860 if (!cvp || !SvOK(*cvp)) {
5863 why3 = "} is not defined";
5866 sv_2mortal(sv); /* Parent created it permanently */
5869 pv = sv_2mortal(newSVpvn(s, len));
5871 typesv = sv_2mortal(newSVpv(type, 0));
5873 typesv = &PL_sv_undef;
5875 PUSHSTACKi(PERLSI_OVERLOAD);
5887 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
5891 /* Check the eval first */
5892 if (!PL_in_eval && SvTRUE(ERRSV)) {
5894 sv_catpv(ERRSV, "Propagated");
5895 yyerror(SvPV(ERRSV, n_a)); /* Duplicates the message inside eval */
5897 res = SvREFCNT_inc(sv);
5901 (void)SvREFCNT_inc(res);
5910 why1 = "Call to &{$^H{";
5912 why3 = "}} did not return a defined value";
5921 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
5923 register char *d = dest;
5924 register char *e = d + destlen - 3; /* two-character token, ending NUL */
5927 Perl_croak(aTHX_ ident_too_long);
5928 if (isALNUM(*s)) /* UTF handled below */
5930 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
5935 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
5939 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
5940 char *t = s + UTF8SKIP(s);
5941 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
5943 if (d + (t - s) > e)
5944 Perl_croak(aTHX_ ident_too_long);
5945 Copy(s, d, t - s, char);
5958 S_scan_ident(pTHX_ register char *s, register char *send, char *dest, STRLEN destlen, I32 ck_uni)
5968 e = d + destlen - 3; /* two-character token, ending NUL */
5970 while (isDIGIT(*s)) {
5972 Perl_croak(aTHX_ ident_too_long);
5979 Perl_croak(aTHX_ ident_too_long);
5980 if (isALNUM(*s)) /* UTF handled below */
5982 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
5987 else if (*s == ':' && s[1] == ':') {
5991 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
5992 char *t = s + UTF8SKIP(s);
5993 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
5995 if (d + (t - s) > e)
5996 Perl_croak(aTHX_ ident_too_long);
5997 Copy(s, d, t - s, char);
6008 if (PL_lex_state != LEX_NORMAL)
6009 PL_lex_state = LEX_INTERPENDMAYBE;
6012 if (*s == '$' && s[1] &&
6013 (isALNUM_lazy_if(s+1,UTF) || strchr("${", s[1]) || strnEQ(s+1,"::",2)) )
6026 if (*d == '^' && *s && isCONTROLVAR(*s)) {
6031 if (isSPACE(s[-1])) {
6034 if (!SPACE_OR_TAB(ch)) {
6040 if (isIDFIRST_lazy_if(d,UTF)) {
6044 while ((e < send && isALNUM_lazy_if(e,UTF)) || *e == ':') {
6046 while (e < send && UTF8_IS_CONTINUED(*e) && is_utf8_mark((U8*)e))
6049 Copy(s, d, e - s, char);
6054 while ((isALNUM(*s) || *s == ':') && d < e)
6057 Perl_croak(aTHX_ ident_too_long);
6060 while (s < send && SPACE_OR_TAB(*s)) s++;
6061 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
6062 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
6063 const char *brack = *s == '[' ? "[...]" : "{...}";
6064 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6065 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
6066 funny, dest, brack, funny, dest, brack);
6069 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
6073 /* Handle extended ${^Foo} variables
6074 * 1999-02-27 mjd-perl-patch@plover.com */
6075 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
6079 while (isALNUM(*s) && d < e) {
6083 Perl_croak(aTHX_ ident_too_long);
6088 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets)
6089 PL_lex_state = LEX_INTERPEND;
6092 if (PL_lex_state == LEX_NORMAL) {
6093 if (ckWARN(WARN_AMBIGUOUS) &&
6094 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
6096 Perl_warner(aTHX_ WARN_AMBIGUOUS,
6097 "Ambiguous use of %c{%s} resolved to %c%s",
6098 funny, dest, funny, dest);
6103 s = bracket; /* let the parser handle it */
6107 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
6108 PL_lex_state = LEX_INTERPEND;
6113 Perl_pmflag(pTHX_ U16 *pmfl, int ch)
6118 *pmfl |= PMf_GLOBAL;
6120 *pmfl |= PMf_CONTINUE;
6124 *pmfl |= PMf_MULTILINE;
6126 *pmfl |= PMf_SINGLELINE;
6128 *pmfl |= PMf_EXTENDED;
6132 S_scan_pat(pTHX_ char *start, I32 type)
6137 s = scan_str(start,FALSE,FALSE);
6140 SvREFCNT_dec(PL_lex_stuff);
6141 PL_lex_stuff = Nullsv;
6142 Perl_croak(aTHX_ "Search pattern not terminated");
6145 pm = (PMOP*)newPMOP(type, 0);
6146 if (PL_multi_open == '?')
6147 pm->op_pmflags |= PMf_ONCE;
6149 while (*s && strchr("iomsx", *s))
6150 pmflag(&pm->op_pmflags,*s++);
6153 while (*s && strchr("iogcmsx", *s))
6154 pmflag(&pm->op_pmflags,*s++);
6156 pm->op_pmpermflags = pm->op_pmflags;
6158 PL_lex_op = (OP*)pm;
6159 yylval.ival = OP_MATCH;
6164 S_scan_subst(pTHX_ char *start)
6171 yylval.ival = OP_NULL;
6173 s = scan_str(start,FALSE,FALSE);
6177 SvREFCNT_dec(PL_lex_stuff);
6178 PL_lex_stuff = Nullsv;
6179 Perl_croak(aTHX_ "Substitution pattern not terminated");
6182 if (s[-1] == PL_multi_open)
6185 first_start = PL_multi_start;
6186 s = scan_str(s,FALSE,FALSE);
6189 SvREFCNT_dec(PL_lex_stuff);
6190 PL_lex_stuff = Nullsv;
6192 SvREFCNT_dec(PL_lex_repl);
6193 PL_lex_repl = Nullsv;
6194 Perl_croak(aTHX_ "Substitution replacement not terminated");
6196 PL_multi_start = first_start; /* so whole substitution is taken together */
6198 pm = (PMOP*)newPMOP(OP_SUBST, 0);
6204 else if (strchr("iogcmsx", *s))
6205 pmflag(&pm->op_pmflags,*s++);
6212 PL_sublex_info.super_bufptr = s;
6213 PL_sublex_info.super_bufend = PL_bufend;
6215 pm->op_pmflags |= PMf_EVAL;
6216 repl = newSVpvn("",0);
6218 sv_catpv(repl, es ? "eval " : "do ");
6219 sv_catpvn(repl, "{ ", 2);
6220 sv_catsv(repl, PL_lex_repl);
6221 sv_catpvn(repl, " };", 2);
6223 SvREFCNT_dec(PL_lex_repl);
6227 pm->op_pmpermflags = pm->op_pmflags;
6228 PL_lex_op = (OP*)pm;
6229 yylval.ival = OP_SUBST;
6234 S_scan_trans(pTHX_ char *start)
6245 yylval.ival = OP_NULL;
6247 s = scan_str(start,FALSE,FALSE);
6250 SvREFCNT_dec(PL_lex_stuff);
6251 PL_lex_stuff = Nullsv;
6252 Perl_croak(aTHX_ "Transliteration pattern not terminated");
6254 if (s[-1] == PL_multi_open)
6257 s = scan_str(s,FALSE,FALSE);
6260 SvREFCNT_dec(PL_lex_stuff);
6261 PL_lex_stuff = Nullsv;
6263 SvREFCNT_dec(PL_lex_repl);
6264 PL_lex_repl = Nullsv;
6265 Perl_croak(aTHX_ "Transliteration replacement not terminated");
6268 New(803,tbl,256,short);
6269 o = newPVOP(OP_TRANS, 0, (char*)tbl);
6271 complement = del = squash = 0;
6272 while (strchr("cds", *s)) {
6274 complement = OPpTRANS_COMPLEMENT;
6276 del = OPpTRANS_DELETE;
6278 squash = OPpTRANS_SQUASH;
6281 o->op_private = del|squash|complement|
6282 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
6283 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
6286 yylval.ival = OP_TRANS;
6291 S_scan_heredoc(pTHX_ register char *s)
6294 I32 op_type = OP_SCALAR;
6301 int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
6305 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
6308 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
6309 if (*peek && strchr("`'\"",*peek)) {
6312 s = delimcpy(d, e, s, PL_bufend, term, &len);
6322 if (!isALNUM_lazy_if(s,UTF))
6323 deprecate("bare << to mean <<\"\"");
6324 for (; isALNUM_lazy_if(s,UTF); s++) {
6329 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
6330 Perl_croak(aTHX_ "Delimiter for here document is too long");
6333 len = d - PL_tokenbuf;
6334 #ifndef PERL_STRICT_CR
6335 d = strchr(s, '\r');
6339 while (s < PL_bufend) {
6345 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
6354 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6359 if (outer || !(d=ninstr(s,PL_bufend,d,d+1)))
6360 herewas = newSVpvn(s,PL_bufend-s);
6362 s--, herewas = newSVpvn(s,d-s);
6363 s += SvCUR(herewas);
6365 tmpstr = NEWSV(87,79);
6366 sv_upgrade(tmpstr, SVt_PVIV);
6371 else if (term == '`') {
6372 op_type = OP_BACKTICK;
6373 SvIVX(tmpstr) = '\\';
6377 PL_multi_start = CopLINE(PL_curcop);
6378 PL_multi_open = PL_multi_close = '<';
6379 term = *PL_tokenbuf;
6380 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
6381 char *bufptr = PL_sublex_info.super_bufptr;
6382 char *bufend = PL_sublex_info.super_bufend;
6383 char *olds = s - SvCUR(herewas);
6384 s = strchr(bufptr, '\n');
6388 while (s < bufend &&
6389 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6391 CopLINE_inc(PL_curcop);
6394 CopLINE_set(PL_curcop, PL_multi_start);
6395 missingterm(PL_tokenbuf);
6397 sv_setpvn(herewas,bufptr,d-bufptr+1);
6398 sv_setpvn(tmpstr,d+1,s-d);
6400 sv_catpvn(herewas,s,bufend-s);
6401 (void)strcpy(bufptr,SvPVX(herewas));
6408 while (s < PL_bufend &&
6409 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
6411 CopLINE_inc(PL_curcop);
6413 if (s >= PL_bufend) {
6414 CopLINE_set(PL_curcop, PL_multi_start);
6415 missingterm(PL_tokenbuf);
6417 sv_setpvn(tmpstr,d+1,s-d);
6419 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
6421 sv_catpvn(herewas,s,PL_bufend-s);
6422 sv_setsv(PL_linestr,herewas);
6423 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
6424 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6427 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
6428 while (s >= PL_bufend) { /* multiple line string? */
6430 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6431 CopLINE_set(PL_curcop, PL_multi_start);
6432 missingterm(PL_tokenbuf);
6434 CopLINE_inc(PL_curcop);
6435 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6436 #ifndef PERL_STRICT_CR
6437 if (PL_bufend - PL_linestart >= 2) {
6438 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
6439 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
6441 PL_bufend[-2] = '\n';
6443 SvCUR_set(PL_linestr, PL_bufend - SvPVX(PL_linestr));
6445 else if (PL_bufend[-1] == '\r')
6446 PL_bufend[-1] = '\n';
6448 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
6449 PL_bufend[-1] = '\n';
6451 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6452 SV *sv = NEWSV(88,0);
6454 sv_upgrade(sv, SVt_PVMG);
6455 sv_setsv(sv,PL_linestr);
6456 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop),sv);
6458 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
6461 sv_catsv(PL_linestr,herewas);
6462 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6466 sv_catsv(tmpstr,PL_linestr);
6471 PL_multi_end = CopLINE(PL_curcop);
6472 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
6473 SvLEN_set(tmpstr, SvCUR(tmpstr) + 1);
6474 Renew(SvPVX(tmpstr), SvLEN(tmpstr), char);
6476 SvREFCNT_dec(herewas);
6477 if (UTF && !IN_BYTE && is_utf8_string((U8*)SvPVX(tmpstr), SvCUR(tmpstr)))
6479 PL_lex_stuff = tmpstr;
6480 yylval.ival = op_type;
6485 takes: current position in input buffer
6486 returns: new position in input buffer
6487 side-effects: yylval and lex_op are set.
6492 <FH> read from filehandle
6493 <pkg::FH> read from package qualified filehandle
6494 <pkg'FH> read from package qualified filehandle
6495 <$fh> read from filehandle in $fh
6501 S_scan_inputsymbol(pTHX_ char *start)
6503 register char *s = start; /* current position in buffer */
6509 d = PL_tokenbuf; /* start of temp holding space */
6510 e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
6511 end = strchr(s, '\n');
6514 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
6516 /* die if we didn't have space for the contents of the <>,
6517 or if it didn't end, or if we see a newline
6520 if (len >= sizeof PL_tokenbuf)
6521 Perl_croak(aTHX_ "Excessively long <> operator");
6523 Perl_croak(aTHX_ "Unterminated <> operator");
6528 Remember, only scalar variables are interpreted as filehandles by
6529 this code. Anything more complex (e.g., <$fh{$num}>) will be
6530 treated as a glob() call.
6531 This code makes use of the fact that except for the $ at the front,
6532 a scalar variable and a filehandle look the same.
6534 if (*d == '$' && d[1]) d++;
6536 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
6537 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
6540 /* If we've tried to read what we allow filehandles to look like, and
6541 there's still text left, then it must be a glob() and not a getline.
6542 Use scan_str to pull out the stuff between the <> and treat it
6543 as nothing more than a string.
6546 if (d - PL_tokenbuf != len) {
6547 yylval.ival = OP_GLOB;
6549 s = scan_str(start,FALSE,FALSE);
6551 Perl_croak(aTHX_ "Glob not terminated");
6555 /* we're in a filehandle read situation */
6558 /* turn <> into <ARGV> */
6560 (void)strcpy(d,"ARGV");
6562 /* if <$fh>, create the ops to turn the variable into a
6568 /* try to find it in the pad for this block, otherwise find
6569 add symbol table ops
6571 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
6572 OP *o = newOP(OP_PADSV, 0);
6574 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, o);
6577 GV *gv = gv_fetchpv(d+1,TRUE, SVt_PV);
6578 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0,
6579 newUNOP(OP_RV2SV, 0,
6580 newGVOP(OP_GV, 0, gv)));
6582 PL_lex_op->op_flags |= OPf_SPECIAL;
6583 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
6584 yylval.ival = OP_NULL;
6587 /* If it's none of the above, it must be a literal filehandle
6588 (<Foo::BAR> or <FOO>) so build a simple readline OP */
6590 GV *gv = gv_fetchpv(d,TRUE, SVt_PVIO);
6591 PL_lex_op = (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
6592 yylval.ival = OP_NULL;
6601 takes: start position in buffer
6602 keep_quoted preserve \ on the embedded delimiter(s)
6603 keep_delims preserve the delimiters around the string
6604 returns: position to continue reading from buffer
6605 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
6606 updates the read buffer.
6608 This subroutine pulls a string out of the input. It is called for:
6609 q single quotes q(literal text)
6610 ' single quotes 'literal text'
6611 qq double quotes qq(interpolate $here please)
6612 " double quotes "interpolate $here please"
6613 qx backticks qx(/bin/ls -l)
6614 ` backticks `/bin/ls -l`
6615 qw quote words @EXPORT_OK = qw( func() $spam )
6616 m// regexp match m/this/
6617 s/// regexp substitute s/this/that/
6618 tr/// string transliterate tr/this/that/
6619 y/// string transliterate y/this/that/
6620 ($*@) sub prototypes sub foo ($)
6621 (stuff) sub attr parameters sub foo : attr(stuff)
6622 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
6624 In most of these cases (all but <>, patterns and transliterate)
6625 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
6626 calls scan_str(). s/// makes yylex() call scan_subst() which calls
6627 scan_str(). tr/// and y/// make yylex() call scan_trans() which
6630 It skips whitespace before the string starts, and treats the first
6631 character as the delimiter. If the delimiter is one of ([{< then
6632 the corresponding "close" character )]}> is used as the closing
6633 delimiter. It allows quoting of delimiters, and if the string has
6634 balanced delimiters ([{<>}]) it allows nesting.
6636 The lexer always reads these strings into lex_stuff, except in the
6637 case of the operators which take *two* arguments (s/// and tr///)
6638 when it checks to see if lex_stuff is full (presumably with the 1st
6639 arg to s or tr) and if so puts the string into lex_repl.
6644 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
6646 SV *sv; /* scalar value: string */
6647 char *tmps; /* temp string, used for delimiter matching */
6648 register char *s = start; /* current position in the buffer */
6649 register char term; /* terminating character */
6650 register char *to; /* current position in the sv's data */
6651 I32 brackets = 1; /* bracket nesting level */
6652 bool has_utf8 = FALSE; /* is there any utf8 content? */
6654 /* skip space before the delimiter */
6658 /* mark where we are, in case we need to report errors */
6661 /* after skipping whitespace, the next character is the terminator */
6663 if (UTF8_IS_CONTINUED(term) && UTF)
6666 /* mark where we are */
6667 PL_multi_start = CopLINE(PL_curcop);
6668 PL_multi_open = term;
6670 /* find corresponding closing delimiter */
6671 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6673 PL_multi_close = term;
6675 /* create a new SV to hold the contents. 87 is leak category, I'm
6676 assuming. 79 is the SV's initial length. What a random number. */
6678 sv_upgrade(sv, SVt_PVIV);
6680 (void)SvPOK_only(sv); /* validate pointer */
6682 /* move past delimiter and try to read a complete string */
6684 sv_catpvn(sv, s, 1);
6687 /* extend sv if need be */
6688 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
6689 /* set 'to' to the next character in the sv's string */
6690 to = SvPVX(sv)+SvCUR(sv);
6692 /* if open delimiter is the close delimiter read unbridle */
6693 if (PL_multi_open == PL_multi_close) {
6694 for (; s < PL_bufend; s++,to++) {
6695 /* embedded newlines increment the current line number */
6696 if (*s == '\n' && !PL_rsfp)
6697 CopLINE_inc(PL_curcop);
6698 /* handle quoted delimiters */
6699 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
6700 if (!keep_quoted && s[1] == term)
6702 /* any other quotes are simply copied straight through */
6706 /* terminate when run out of buffer (the for() condition), or
6707 have found the terminator */
6708 else if (*s == term)
6710 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6716 /* if the terminator isn't the same as the start character (e.g.,
6717 matched brackets), we have to allow more in the quoting, and
6718 be prepared for nested brackets.
6721 /* read until we run out of string, or we find the terminator */
6722 for (; s < PL_bufend; s++,to++) {
6723 /* embedded newlines increment the line count */
6724 if (*s == '\n' && !PL_rsfp)
6725 CopLINE_inc(PL_curcop);
6726 /* backslashes can escape the open or closing characters */
6727 if (*s == '\\' && s+1 < PL_bufend) {
6729 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
6734 /* allow nested opens and closes */
6735 else if (*s == PL_multi_close && --brackets <= 0)
6737 else if (*s == PL_multi_open)
6739 else if (!has_utf8 && UTF8_IS_CONTINUED(*s) && UTF)
6744 /* terminate the copied string and update the sv's end-of-string */
6746 SvCUR_set(sv, to - SvPVX(sv));
6749 * this next chunk reads more into the buffer if we're not done yet
6753 break; /* handle case where we are done yet :-) */
6755 #ifndef PERL_STRICT_CR
6756 if (to - SvPVX(sv) >= 2) {
6757 if ((to[-2] == '\r' && to[-1] == '\n') ||
6758 (to[-2] == '\n' && to[-1] == '\r'))
6762 SvCUR_set(sv, to - SvPVX(sv));
6764 else if (to[-1] == '\r')
6767 else if (to - SvPVX(sv) == 1 && to[-1] == '\r')
6771 /* if we're out of file, or a read fails, bail and reset the current
6772 line marker so we can report where the unterminated string began
6775 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
6777 CopLINE_set(PL_curcop, PL_multi_start);
6780 /* we read a line, so increment our line counter */
6781 CopLINE_inc(PL_curcop);
6783 /* update debugger info */
6784 if (PERLDB_LINE && PL_curstash != PL_debstash) {
6785 SV *sv = NEWSV(88,0);
6787 sv_upgrade(sv, SVt_PVMG);
6788 sv_setsv(sv,PL_linestr);
6789 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop), sv);
6792 /* having changed the buffer, we must update PL_bufend */
6793 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
6796 /* at this point, we have successfully read the delimited string */
6799 sv_catpvn(sv, s, 1);
6802 PL_multi_end = CopLINE(PL_curcop);
6805 /* if we allocated too much space, give some back */
6806 if (SvCUR(sv) + 5 < SvLEN(sv)) {
6807 SvLEN_set(sv, SvCUR(sv) + 1);
6808 Renew(SvPVX(sv), SvLEN(sv), char);
6811 /* decide whether this is the first or second quoted string we've read
6824 takes: pointer to position in buffer
6825 returns: pointer to new position in buffer
6826 side-effects: builds ops for the constant in yylval.op
6828 Read a number in any of the formats that Perl accepts:
6830 0(x[0-7A-F]+)|([0-7]+)|(b[01])
6831 [\d_]+(\.[\d_]*)?[Ee](\d+)
6833 Underbars (_) are allowed in decimal numbers. If -w is on,
6834 underbars before a decimal point must be at three digit intervals.
6836 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
6839 If it reads a number without a decimal point or an exponent, it will
6840 try converting the number to an integer and see if it can do so
6841 without loss of precision.
6845 Perl_scan_num(pTHX_ char *start, YYSTYPE* lvalp)
6847 register char *s = start; /* current position in buffer */
6848 register char *d; /* destination in temp buffer */
6849 register char *e; /* end of temp buffer */
6850 NV nv; /* number read, as a double */
6851 SV *sv = Nullsv; /* place to put the converted number */
6852 bool floatit; /* boolean: int or float? */
6853 char *lastub = 0; /* position of last underbar */
6854 static char number_too_long[] = "Number too long";
6856 /* We use the first character to decide what type of number this is */
6860 Perl_croak(aTHX_ "panic: scan_num");
6862 /* if it starts with a 0, it could be an octal number, a decimal in
6863 0.13 disguise, or a hexadecimal number, or a binary number. */
6867 u holds the "number so far"
6868 shift the power of 2 of the base
6869 (hex == 4, octal == 3, binary == 1)
6870 overflowed was the number more than we can hold?
6872 Shift is used when we add a digit. It also serves as an "are
6873 we in octal/hex/binary?" indicator to disallow hex characters
6879 bool overflowed = FALSE;
6880 static NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
6881 static char* bases[5] = { "", "binary", "", "octal",
6883 static char* Bases[5] = { "", "Binary", "", "Octal",
6885 static char *maxima[5] = { "",
6886 "0b11111111111111111111111111111111",
6890 char *base, *Base, *max;
6896 } else if (s[1] == 'b') {
6900 /* check for a decimal in disguise */
6901 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
6903 /* so it must be octal */
6907 base = bases[shift];
6908 Base = Bases[shift];
6909 max = maxima[shift];
6911 /* read the rest of the number */
6913 /* x is used in the overflow test,
6914 b is the digit we're adding on. */
6919 /* if we don't mention it, we're done */
6928 /* 8 and 9 are not octal */
6931 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
6935 case '2': case '3': case '4':
6936 case '5': case '6': case '7':
6938 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
6942 b = *s++ & 15; /* ASCII digit -> value of digit */
6946 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
6947 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
6948 /* make sure they said 0x */
6953 /* Prepare to put the digit we have onto the end
6954 of the number so far. We check for overflows.
6959 x = u << shift; /* make room for the digit */
6961 if ((x >> shift) != u
6962 && !(PL_hints & HINT_NEW_BINARY)) {
6965 if (ckWARN_d(WARN_OVERFLOW))
6966 Perl_warner(aTHX_ WARN_OVERFLOW,
6967 "Integer overflow in %s number",
6970 u = x | b; /* add the digit to the end */
6973 n *= nvshift[shift];
6974 /* If an NV has not enough bits in its
6975 * mantissa to represent an UV this summing of
6976 * small low-order numbers is a waste of time
6977 * (because the NV cannot preserve the
6978 * low-order bits anyway): we could just
6979 * remember when did we overflow and in the
6980 * end just multiply n by the right
6988 /* if we get here, we had success: make a scalar value from
6994 if (ckWARN(WARN_PORTABLE) && n > 4294967295.0)
6995 Perl_warner(aTHX_ WARN_PORTABLE,
6996 "%s number > %s non-portable",
7002 if (ckWARN(WARN_PORTABLE) && u > 0xffffffff)
7003 Perl_warner(aTHX_ WARN_PORTABLE,
7004 "%s number > %s non-portable",
7009 if (PL_hints & HINT_NEW_BINARY)
7010 sv = new_constant(start, s - start, "binary", sv, Nullsv, NULL);
7015 handle decimal numbers.
7016 we're also sent here when we read a 0 as the first digit
7018 case '1': case '2': case '3': case '4': case '5':
7019 case '6': case '7': case '8': case '9': case '.':
7022 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
7025 /* read next group of digits and _ and copy into d */
7026 while (isDIGIT(*s) || *s == '_') {
7027 /* skip underscores, checking for misplaced ones
7031 if (ckWARN(WARN_SYNTAX) && lastub && s - lastub != 3)
7032 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7036 /* check for end of fixed-length buffer */
7038 Perl_croak(aTHX_ number_too_long);
7039 /* if we're ok, copy the character */
7044 /* final misplaced underbar check */
7045 if (lastub && s - lastub != 3) {
7046 if (ckWARN(WARN_SYNTAX))
7047 Perl_warner(aTHX_ WARN_SYNTAX, "Misplaced _ in number");
7050 /* read a decimal portion if there is one. avoid
7051 3..5 being interpreted as the number 3. followed
7054 if (*s == '.' && s[1] != '.') {
7058 /* copy, ignoring underbars, until we run out of
7059 digits. Note: no misplaced underbar checks!
7061 for (; isDIGIT(*s) || *s == '_'; s++) {
7062 /* fixed length buffer check */
7064 Perl_croak(aTHX_ number_too_long);
7068 if (*s == '.' && isDIGIT(s[1])) {
7069 /* oops, it's really a v-string, but without the "v" */
7075 /* read exponent part, if present */
7076 if (*s && strchr("eE",*s) && strchr("+-0123456789",s[1])) {
7080 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
7081 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
7083 /* allow positive or negative exponent */
7084 if (*s == '+' || *s == '-')
7087 /* read digits of exponent (no underbars :-) */
7088 while (isDIGIT(*s)) {
7090 Perl_croak(aTHX_ number_too_long);
7095 /* terminate the string */
7098 /* make an sv from the string */
7101 #if defined(Strtol) && defined(Strtoul)
7104 strtol/strtoll sets errno to ERANGE if the number is too big
7105 for an integer. We try to do an integer conversion first
7106 if no characters indicating "float" have been found.
7113 if (*PL_tokenbuf == '-')
7114 iv = Strtol(PL_tokenbuf, (char**)NULL, 10);
7116 uv = Strtoul(PL_tokenbuf, (char**)NULL, 10);
7118 floatit = TRUE; /* Probably just too large. */
7119 else if (*PL_tokenbuf == '-')
7121 else if (uv <= IV_MAX)
7122 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7127 nv = Atof(PL_tokenbuf);
7132 No working strtou?ll?.
7134 Unfortunately atol() doesn't do range checks (returning
7135 LONG_MIN/LONG_MAX, and setting errno to ERANGE on overflows)
7136 everywhere [1], so we cannot use use atol() (or atoll()).
7137 If we could, they would be used, as Atol(), very much like
7138 Strtol() and Strtoul() are used above.
7140 [1] XXX Configure test needed to check for atol()
7141 (and atoll()) overflow behaviour XXX
7145 We need to do this the hard way. */
7147 nv = Atof(PL_tokenbuf);
7149 /* See if we can make do with an integer value without loss of
7150 precision. We use U_V to cast to a UV, because some
7151 compilers have issues. Then we try casting it back and see
7152 if it was the same [1]. We only do this if we know we
7153 specifically read an integer. If floatit is true, then we
7154 don't need to do the conversion at all.
7156 [1] Note that this is lossy if our NVs cannot preserve our
7157 UVs. There are metaconfig defines NV_PRESERVES_UV (a boolean)
7158 and NV_PRESERVES_UV_BITS (a number), but in general we really
7159 do hope all such potentially lossy platforms have strtou?ll?
7160 to do a lossless IV/UV conversion.
7162 Maybe could do some tricks with DBL_DIG, LDBL_DIG and
7163 DBL_MANT_DIG and LDBL_MANT_DIG (these are already available
7164 as NV_DIG and NV_MANT_DIG)?
7170 if (!floatit && (NV)uv == nv) {
7172 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
7180 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
7181 (PL_hints & HINT_NEW_INTEGER) )
7182 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
7183 (floatit ? "float" : "integer"),
7187 /* if it starts with a v, it could be a v-string */
7193 while (isDIGIT(*pos) || *pos == '_')
7195 if (!isALPHA(*pos)) {
7197 U8 tmpbuf[UTF8_MAXLEN+1];
7200 s++; /* get past 'v' */
7203 sv_setpvn(sv, "", 0);
7206 if (*s == '0' && isDIGIT(s[1]))
7207 yyerror("Octal number in vector unsupported");
7210 /* this is atoi() that tolerates underscores */
7213 while (--end >= s) {
7218 rev += (*end - '0') * mult;
7220 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
7221 Perl_warner(aTHX_ WARN_OVERFLOW,
7222 "Integer overflow in decimal number");
7225 tmpend = uv_to_utf8(tmpbuf, rev);
7226 utf8 = utf8 || rev > 127;
7227 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
7228 if (*pos == '.' && isDIGIT(pos[1]))
7234 while (isDIGIT(*pos) || *pos == '_')
7243 sv_utf8_downgrade(sv, TRUE);
7250 /* make the op for the constant and return */
7253 lvalp->opval = newSVOP(OP_CONST, 0, sv);
7255 lvalp->opval = Nullop;
7261 S_scan_formline(pTHX_ register char *s)
7265 SV *stuff = newSVpvn("",0);
7266 bool needargs = FALSE;
7269 if (*s == '.' || *s == /*{*/'}') {
7271 #ifdef PERL_STRICT_CR
7272 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
7274 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
7276 if (*t == '\n' || t == PL_bufend)
7279 if (PL_in_eval && !PL_rsfp) {
7280 eol = strchr(s,'\n');
7285 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
7287 for (t = s; t < eol; t++) {
7288 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
7290 goto enough; /* ~~ must be first line in formline */
7292 if (*t == '@' || *t == '^')
7295 sv_catpvn(stuff, s, eol-s);
7296 #ifndef PERL_STRICT_CR
7297 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
7298 char *end = SvPVX(stuff) + SvCUR(stuff);
7307 s = filter_gets(PL_linestr, PL_rsfp, 0);
7308 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
7309 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
7312 yyerror("Format not terminated");
7322 PL_lex_state = LEX_NORMAL;
7323 PL_nextval[PL_nexttoke].ival = 0;
7327 PL_lex_state = LEX_FORMLINE;
7328 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0, stuff);
7330 PL_nextval[PL_nexttoke].ival = OP_FORMLINE;
7334 SvREFCNT_dec(stuff);
7335 PL_lex_formbrack = 0;
7346 PL_cshlen = strlen(PL_cshname);
7351 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
7353 I32 oldsavestack_ix = PL_savestack_ix;
7354 CV* outsidecv = PL_compcv;
7358 assert(SvTYPE(PL_compcv) == SVt_PVCV);
7360 SAVEI32(PL_subline);
7361 save_item(PL_subname);
7364 SAVESPTR(PL_comppad_name);
7365 SAVESPTR(PL_compcv);
7366 SAVEI32(PL_comppad_name_fill);
7367 SAVEI32(PL_min_intro_pending);
7368 SAVEI32(PL_max_intro_pending);
7369 SAVEI32(PL_pad_reset_pending);
7371 PL_compcv = (CV*)NEWSV(1104,0);
7372 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
7373 CvFLAGS(PL_compcv) |= flags;
7375 PL_comppad = newAV();
7376 av_push(PL_comppad, Nullsv);
7377 PL_curpad = AvARRAY(PL_comppad);
7378 PL_comppad_name = newAV();
7379 PL_comppad_name_fill = 0;
7380 PL_min_intro_pending = 0;
7382 PL_subline = CopLINE(PL_curcop);
7384 av_store(PL_comppad_name, 0, newSVpvn("@_", 2));
7385 PL_curpad[0] = (SV*)newAV();
7386 SvPADMY_on(PL_curpad[0]); /* XXX Needed? */
7387 #endif /* USE_THREADS */
7389 comppadlist = newAV();
7390 AvREAL_off(comppadlist);
7391 av_store(comppadlist, 0, (SV*)PL_comppad_name);
7392 av_store(comppadlist, 1, (SV*)PL_comppad);
7394 CvPADLIST(PL_compcv) = comppadlist;
7395 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc(outsidecv);
7397 CvOWNER(PL_compcv) = 0;
7398 New(666, CvMUTEXP(PL_compcv), 1, perl_mutex);
7399 MUTEX_INIT(CvMUTEXP(PL_compcv));
7400 #endif /* USE_THREADS */
7402 return oldsavestack_ix;
7406 Perl_yywarn(pTHX_ char *s)
7408 PL_in_eval |= EVAL_WARNONLY;
7410 PL_in_eval &= ~EVAL_WARNONLY;
7415 Perl_yyerror(pTHX_ char *s)
7418 char *context = NULL;
7422 if (!yychar || (yychar == ';' && !PL_rsfp))
7424 else if (PL_bufptr > PL_oldoldbufptr && PL_bufptr - PL_oldoldbufptr < 200 &&
7425 PL_oldoldbufptr != PL_oldbufptr && PL_oldbufptr != PL_bufptr) {
7426 while (isSPACE(*PL_oldoldbufptr))
7428 context = PL_oldoldbufptr;
7429 contlen = PL_bufptr - PL_oldoldbufptr;
7431 else if (PL_bufptr > PL_oldbufptr && PL_bufptr - PL_oldbufptr < 200 &&
7432 PL_oldbufptr != PL_bufptr) {
7433 while (isSPACE(*PL_oldbufptr))
7435 context = PL_oldbufptr;
7436 contlen = PL_bufptr - PL_oldbufptr;
7438 else if (yychar > 255)
7439 where = "next token ???";
7440 #ifdef USE_PURE_BISON
7441 /* GNU Bison sets the value -2 */
7442 else if (yychar == -2) {
7444 else if ((yychar & 127) == 127) {
7446 if (PL_lex_state == LEX_NORMAL ||
7447 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
7448 where = "at end of line";
7449 else if (PL_lex_inpat)
7450 where = "within pattern";
7452 where = "within string";
7455 SV *where_sv = sv_2mortal(newSVpvn("next char ", 10));
7457 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
7458 else if (isPRINT_LC(yychar))
7459 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
7461 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
7462 where = SvPVX(where_sv);
7464 msg = sv_2mortal(newSVpv(s, 0));
7465 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
7466 CopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
7468 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
7470 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
7471 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
7472 Perl_sv_catpvf(aTHX_ msg,
7473 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
7474 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
7477 if (PL_in_eval & EVAL_WARNONLY)
7478 Perl_warn(aTHX_ "%"SVf, msg);
7481 if (PL_error_count >= 10) {
7482 if (PL_in_eval && SvCUR(ERRSV))
7483 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
7484 ERRSV, CopFILE(PL_curcop));
7486 Perl_croak(aTHX_ "%s has too many errors.\n",
7487 CopFILE(PL_curcop));
7490 PL_in_my_stash = Nullhv;
7495 S_swallow_bom(pTHX_ U8 *s)
7498 slen = SvCUR(PL_linestr);
7502 /* UTF-16 little-endian */
7503 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
7504 Perl_croak(aTHX_ "Unsupported script encoding");
7505 #ifndef PERL_NO_UTF16_FILTER
7506 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-LE script encoding\n"));
7508 if (PL_bufend > (char*)s) {
7512 filter_add(utf16rev_textfilter, NULL);
7513 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7514 PL_bufend = (char*)utf16_to_utf8_reversed(s, news,
7515 PL_bufend - (char*)s - 1,
7517 Copy(news, s, newlen, U8);
7518 SvCUR_set(PL_linestr, newlen);
7519 PL_bufend = SvPVX(PL_linestr) + newlen;
7520 news[newlen++] = '\0';
7524 Perl_croak(aTHX_ "Unsupported script encoding");
7529 if (s[1] == 0xFF) { /* UTF-16 big-endian */
7530 #ifndef PERL_NO_UTF16_FILTER
7531 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding\n"));
7533 if (PL_bufend > (char *)s) {
7537 filter_add(utf16_textfilter, NULL);
7538 New(898, news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
7539 PL_bufend = (char*)utf16_to_utf8(s, news,
7540 PL_bufend - (char*)s,
7542 Copy(news, s, newlen, U8);
7543 SvCUR_set(PL_linestr, newlen);
7544 PL_bufend = SvPVX(PL_linestr) + newlen;
7545 news[newlen++] = '\0';
7549 Perl_croak(aTHX_ "Unsupported script encoding");
7554 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
7555 DEBUG_p(PerlIO_printf(Perl_debug_log, "UTF-8 script encoding\n"));
7560 if (slen > 3 && s[1] == 0 && /* UTF-32 big-endian */
7561 s[2] == 0xFE && s[3] == 0xFF)
7563 Perl_croak(aTHX_ "Unsupported script encoding");
7575 * Restore a source filter.
7579 restore_rsfp(pTHXo_ void *f)
7581 PerlIO *fp = (PerlIO*)f;
7583 if (PL_rsfp == PerlIO_stdin())
7584 PerlIO_clearerr(PL_rsfp);
7585 else if (PL_rsfp && (PL_rsfp != fp))
7586 PerlIO_close(PL_rsfp);
7590 #ifndef PERL_NO_UTF16_FILTER
7592 utf16_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7594 I32 count = FILTER_READ(idx+1, sv, maxlen);
7599 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7600 if (!*SvPV_nolen(sv))
7601 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7604 tend = utf16_to_utf8((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7605 sv_usepvn(sv, (char*)tmps, tend - tmps);
7611 utf16rev_textfilter(pTHXo_ int idx, SV *sv, int maxlen)
7613 I32 count = FILTER_READ(idx+1, sv, maxlen);
7618 if (!*SvPV_nolen(sv))
7619 /* Game over, but don't feed an odd-length string to utf16_to_utf8 */
7622 New(898, tmps, SvCUR(sv) * 3 / 2 + 1, U8);
7623 tend = utf16_to_utf8_reversed((U8*)SvPVX(sv), tmps, SvCUR(sv), &newlen);
7624 sv_usepvn(sv, (char*)tmps, tend - tmps);