3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yychar (*PL_yycharp)
27 #define yylval (*PL_yylvalp)
29 static const char ident_too_long[] =
30 "Identifier too long";
31 static const char c_without_g[] =
32 "Use of /c modifier is meaningless without /g";
33 static const char c_in_subst[] =
34 "Use of /c modifier is meaningless in s///";
36 static void restore_rsfp(pTHX_ void *f);
37 #ifndef PERL_NO_UTF16_FILTER
38 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
42 #define XFAKEBRACK 128
45 #ifdef USE_UTF8_SCRIPTS
46 # define UTF (!IN_BYTES)
48 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
51 /* In variables named $^X, these are the legal values for X.
52 * 1999-02-27 mjd-perl-patch@plover.com */
53 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
55 /* On MacOS, respect nonbreaking spaces */
56 #ifdef MACOS_TRADITIONAL
57 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
59 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
62 /* LEX_* are values for PL_lex_state, the state of the lexer.
63 * They are arranged oddly so that the guard on the switch statement
64 * can get by with a single comparison (if the compiler is smart enough).
67 /* #define LEX_NOTPARSING 11 is done in perl.h. */
70 #define LEX_INTERPNORMAL 9
71 #define LEX_INTERPCASEMOD 8
72 #define LEX_INTERPPUSH 7
73 #define LEX_INTERPSTART 6
74 #define LEX_INTERPEND 5
75 #define LEX_INTERPENDMAYBE 4
76 #define LEX_INTERPCONCAT 3
77 #define LEX_INTERPCONST 2
78 #define LEX_FORMLINE 1
79 #define LEX_KNOWNEXT 0
82 static const char* const lex_state_names[] = {
101 #include "keywords.h"
103 /* CLINE is a macro that ensures PL_copline has a sane value */
108 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
111 * Convenience functions to return different tokens and prime the
112 * lexer for the next token. They all take an argument.
114 * TOKEN : generic token (used for '(', DOLSHARP, etc)
115 * OPERATOR : generic operator
116 * AOPERATOR : assignment operator
117 * PREBLOCK : beginning the block after an if, while, foreach, ...
118 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
119 * PREREF : *EXPR where EXPR is not a simple identifier
120 * TERM : expression term
121 * LOOPX : loop exiting command (goto, last, dump, etc)
122 * FTST : file test operator
123 * FUN0 : zero-argument function
124 * FUN1 : not used, except for not, which isn't a UNIOP
125 * BOop : bitwise or or xor
127 * SHop : shift operator
128 * PWop : power operator
129 * PMop : pattern-matching operator
130 * Aop : addition-level operator
131 * Mop : multiplication-level operator
132 * Eop : equality-testing operator
133 * Rop : relational operator <= != gt
135 * Also see LOP and lop() below.
138 #ifdef DEBUGGING /* Serve -DT. */
139 # define REPORT(retval) tokereport(s,(int)retval)
141 # define REPORT(retval) (retval)
144 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
145 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
146 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
147 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
148 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
149 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
150 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
151 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
152 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
153 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
154 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
155 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
156 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
157 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
158 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
159 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
160 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
161 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
162 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
163 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
165 /* This bit of chicanery makes a unary function followed by
166 * a parenthesis into a function with one argument, highest precedence.
167 * The UNIDOR macro is for unary functions that can be followed by the //
168 * operator (such as C<shift // 0>).
170 #define UNI2(f,x) { \
174 PL_last_uni = PL_oldbufptr; \
175 PL_last_lop_op = f; \
177 return REPORT( (int)FUNC1 ); \
179 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
181 #define UNI(f) UNI2(f,XTERM)
182 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
184 #define UNIBRACK(f) { \
187 PL_last_uni = PL_oldbufptr; \
189 return REPORT( (int)FUNC1 ); \
191 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
194 /* grandfather return to old style */
195 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
199 /* how to interpret the yylval associated with the token */
203 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
209 static struct debug_tokens { const int token, type; const char *name; }
210 const debug_tokens[] =
212 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
213 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
214 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
215 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
216 { ARROW, TOKENTYPE_NONE, "ARROW" },
217 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
218 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
219 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
220 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
221 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
222 { DO, TOKENTYPE_NONE, "DO" },
223 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
224 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
225 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
226 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
227 { ELSE, TOKENTYPE_NONE, "ELSE" },
228 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
229 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
230 { FOR, TOKENTYPE_IVAL, "FOR" },
231 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
232 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
233 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
234 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
235 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
236 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
237 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
238 { IF, TOKENTYPE_IVAL, "IF" },
239 { LABEL, TOKENTYPE_PVAL, "LABEL" },
240 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
241 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
242 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
243 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
244 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
245 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
246 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
247 { MY, TOKENTYPE_IVAL, "MY" },
248 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
249 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
250 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
251 { OROP, TOKENTYPE_IVAL, "OROP" },
252 { OROR, TOKENTYPE_NONE, "OROR" },
253 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
254 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
255 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
256 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
257 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
258 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
259 { PREINC, TOKENTYPE_NONE, "PREINC" },
260 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
261 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
262 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
263 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
264 { SUB, TOKENTYPE_NONE, "SUB" },
265 { THING, TOKENTYPE_OPVAL, "THING" },
266 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
267 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
268 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
269 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
270 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
271 { USE, TOKENTYPE_IVAL, "USE" },
272 { WHILE, TOKENTYPE_IVAL, "WHILE" },
273 { WORD, TOKENTYPE_OPVAL, "WORD" },
274 { 0, TOKENTYPE_NONE, 0 }
277 /* dump the returned token in rv, plus any optional arg in yylval */
280 S_tokereport(pTHX_ const char* s, I32 rv)
283 const char *name = Nullch;
284 enum token_type type = TOKENTYPE_NONE;
285 const struct debug_tokens *p;
286 SV* const report = newSVpvn("<== ", 4);
288 for (p = debug_tokens; p->token; p++) {
289 if (p->token == (int)rv) {
296 Perl_sv_catpv(aTHX_ report, name);
297 else if ((char)rv > ' ' && (char)rv < '~')
298 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
300 Perl_sv_catpv(aTHX_ report, "EOF");
302 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
305 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
308 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
310 case TOKENTYPE_OPNUM:
311 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
312 PL_op_name[yylval.ival]);
315 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
317 case TOKENTYPE_OPVAL:
319 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
320 PL_op_name[yylval.opval->op_type]);
322 Perl_sv_catpv(aTHX_ report, "(opval=null)");
325 Perl_sv_catpvf(aTHX_ report, " at line %"IVdf" [", (IV)CopLINE(PL_curcop));
326 if (s - PL_bufptr > 0)
327 sv_catpvn(report, PL_bufptr, s - PL_bufptr);
329 if (PL_oldbufptr && *PL_oldbufptr)
330 sv_catpv(report, PL_tokenbuf);
332 PerlIO_printf(Perl_debug_log, "### %s]\n", SvPV_nolen_const(report));
342 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
343 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
347 S_ao(pTHX_ int toketype)
349 if (*PL_bufptr == '=') {
351 if (toketype == ANDAND)
352 yylval.ival = OP_ANDASSIGN;
353 else if (toketype == OROR)
354 yylval.ival = OP_ORASSIGN;
355 else if (toketype == DORDOR)
356 yylval.ival = OP_DORASSIGN;
364 * When Perl expects an operator and finds something else, no_op
365 * prints the warning. It always prints "<something> found where
366 * operator expected. It prints "Missing semicolon on previous line?"
367 * if the surprise occurs at the start of the line. "do you need to
368 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
369 * where the compiler doesn't know if foo is a method call or a function.
370 * It prints "Missing operator before end of line" if there's nothing
371 * after the missing operator, or "... before <...>" if there is something
372 * after the missing operator.
376 S_no_op(pTHX_ const char *what, char *s)
378 char * const oldbp = PL_bufptr;
379 const bool is_first = (PL_oldbufptr == PL_linestart);
385 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
386 if (ckWARN_d(WARN_SYNTAX)) {
388 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
389 "\t(Missing semicolon on previous line?)\n");
390 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
392 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
393 if (t < PL_bufptr && isSPACE(*t))
394 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
395 "\t(Do you need to predeclare %.*s?)\n",
396 t - PL_oldoldbufptr, PL_oldoldbufptr);
400 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
401 "\t(Missing operator before %.*s?)\n", s - oldbp, oldbp);
409 * Complain about missing quote/regexp/heredoc terminator.
410 * If it's called with (char *)NULL then it cauterizes the line buffer.
411 * If we're in a delimited string and the delimiter is a control
412 * character, it's reformatted into a two-char sequence like ^C.
417 S_missingterm(pTHX_ char *s)
422 char * const nl = strrchr(s,'\n');
428 iscntrl(PL_multi_close)
430 PL_multi_close < 32 || PL_multi_close == 127
434 tmpbuf[1] = toCTRL(PL_multi_close);
439 *tmpbuf = (char)PL_multi_close;
443 q = strchr(s,'"') ? '\'' : '"';
444 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
452 Perl_deprecate(pTHX_ const char *s)
454 if (ckWARN(WARN_DEPRECATED))
455 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
459 Perl_deprecate_old(pTHX_ const char *s)
461 /* This function should NOT be called for any new deprecated warnings */
462 /* Use Perl_deprecate instead */
464 /* It is here to maintain backward compatibility with the pre-5.8 */
465 /* warnings category hierarchy. The "deprecated" category used to */
466 /* live under the "syntax" category. It is now a top-level category */
467 /* in its own right. */
469 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
470 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
471 "Use of %s is deprecated", s);
476 * Deprecate a comma-less variable list.
482 deprecate_old("comma-less variable list");
486 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
487 * utf16-to-utf8-reversed.
490 #ifdef PERL_CR_FILTER
494 register const char *s = SvPVX_const(sv);
495 register const char * const e = s + SvCUR(sv);
496 /* outer loop optimized to do nothing if there are no CR-LFs */
498 if (*s++ == '\r' && *s == '\n') {
499 /* hit a CR-LF, need to copy the rest */
500 register char *d = s - 1;
503 if (*s == '\r' && s[1] == '\n')
514 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
516 const I32 count = FILTER_READ(idx+1, sv, maxlen);
517 if (count > 0 && !maxlen)
525 * Initialize variables. Uses the Perl save_stack to save its state (for
526 * recursive calls to the parser).
530 Perl_lex_start(pTHX_ SV *line)
535 SAVEI32(PL_lex_dojoin);
536 SAVEI32(PL_lex_brackets);
537 SAVEI32(PL_lex_casemods);
538 SAVEI32(PL_lex_starts);
539 SAVEI32(PL_lex_state);
540 SAVEVPTR(PL_lex_inpat);
541 SAVEI32(PL_lex_inwhat);
542 if (PL_lex_state == LEX_KNOWNEXT) {
543 I32 toke = PL_nexttoke;
544 while (--toke >= 0) {
545 SAVEI32(PL_nexttype[toke]);
546 SAVEVPTR(PL_nextval[toke]);
548 SAVEI32(PL_nexttoke);
550 SAVECOPLINE(PL_curcop);
553 SAVEPPTR(PL_oldbufptr);
554 SAVEPPTR(PL_oldoldbufptr);
555 SAVEPPTR(PL_last_lop);
556 SAVEPPTR(PL_last_uni);
557 SAVEPPTR(PL_linestart);
558 SAVESPTR(PL_linestr);
559 SAVEGENERICPV(PL_lex_brackstack);
560 SAVEGENERICPV(PL_lex_casestack);
561 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
562 SAVESPTR(PL_lex_stuff);
563 SAVEI32(PL_lex_defer);
564 SAVEI32(PL_sublex_info.sub_inwhat);
565 SAVESPTR(PL_lex_repl);
567 SAVEINT(PL_lex_expect);
569 PL_lex_state = LEX_NORMAL;
573 Newx(PL_lex_brackstack, 120, char);
574 Newx(PL_lex_casestack, 12, char);
576 *PL_lex_casestack = '\0';
579 PL_lex_stuff = Nullsv;
580 PL_lex_repl = Nullsv;
584 PL_sublex_info.sub_inwhat = 0;
586 if (SvREADONLY(PL_linestr))
587 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
588 s = SvPV_const(PL_linestr, len);
589 if (!len || s[len-1] != ';') {
590 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
591 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
592 sv_catpvn(PL_linestr, "\n;", 2);
594 SvTEMP_off(PL_linestr);
595 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
596 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
597 PL_last_lop = PL_last_uni = Nullch;
603 * Finalizer for lexing operations. Must be called when the parser is
604 * done with the lexer.
610 PL_doextract = FALSE;
615 * This subroutine has nothing to do with tilting, whether at windmills
616 * or pinball tables. Its name is short for "increment line". It
617 * increments the current line number in CopLINE(PL_curcop) and checks
618 * to see whether the line starts with a comment of the form
619 * # line 500 "foo.pm"
620 * If so, it sets the current line number and file to the values in the comment.
624 S_incline(pTHX_ char *s)
631 CopLINE_inc(PL_curcop);
634 while (SPACE_OR_TAB(*s)) s++;
635 if (strnEQ(s, "line", 4))
639 if (SPACE_OR_TAB(*s))
643 while (SPACE_OR_TAB(*s)) s++;
649 while (SPACE_OR_TAB(*s))
651 if (*s == '"' && (t = strchr(s+1, '"'))) {
656 for (t = s; !isSPACE(*t); t++) ;
659 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
661 if (*e != '\n' && *e != '\0')
662 return; /* false alarm */
667 CopFILE_free(PL_curcop);
668 CopFILE_set(PL_curcop, s);
671 CopLINE_set(PL_curcop, atoi(n)-1);
676 * Called to gobble the appropriate amount and type of whitespace.
677 * Skips comments as well.
681 S_skipspace(pTHX_ register char *s)
683 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
684 while (s < PL_bufend && SPACE_OR_TAB(*s))
690 SSize_t oldprevlen, oldoldprevlen;
691 SSize_t oldloplen = 0, oldunilen = 0;
692 while (s < PL_bufend && isSPACE(*s)) {
693 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
698 if (s < PL_bufend && *s == '#') {
699 while (s < PL_bufend && *s != '\n')
703 if (PL_in_eval && !PL_rsfp) {
710 /* only continue to recharge the buffer if we're at the end
711 * of the buffer, we're not reading from a source filter, and
712 * we're in normal lexing mode
714 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
715 PL_lex_state == LEX_FORMLINE)
718 /* try to recharge the buffer */
719 if ((s = filter_gets(PL_linestr, PL_rsfp,
720 (prevlen = SvCUR(PL_linestr)))) == Nullch)
722 /* end of file. Add on the -p or -n magic */
725 ";}continue{print or die qq(-p destination: $!\\n);}");
726 PL_minus_n = PL_minus_p = 0;
728 else if (PL_minus_n) {
729 sv_setpvn(PL_linestr, ";}", 2);
733 sv_setpvn(PL_linestr,";", 1);
735 /* reset variables for next time we lex */
736 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
738 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
739 PL_last_lop = PL_last_uni = Nullch;
741 /* Close the filehandle. Could be from -P preprocessor,
742 * STDIN, or a regular file. If we were reading code from
743 * STDIN (because the commandline held no -e or filename)
744 * then we don't close it, we reset it so the code can
745 * read from STDIN too.
748 if (PL_preprocess && !PL_in_eval)
749 (void)PerlProc_pclose(PL_rsfp);
750 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
751 PerlIO_clearerr(PL_rsfp);
753 (void)PerlIO_close(PL_rsfp);
758 /* not at end of file, so we only read another line */
759 /* make corresponding updates to old pointers, for yyerror() */
760 oldprevlen = PL_oldbufptr - PL_bufend;
761 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
763 oldunilen = PL_last_uni - PL_bufend;
765 oldloplen = PL_last_lop - PL_bufend;
766 PL_linestart = PL_bufptr = s + prevlen;
767 PL_bufend = s + SvCUR(PL_linestr);
769 PL_oldbufptr = s + oldprevlen;
770 PL_oldoldbufptr = s + oldoldprevlen;
772 PL_last_uni = s + oldunilen;
774 PL_last_lop = s + oldloplen;
777 /* debugger active and we're not compiling the debugger code,
778 * so store the line into the debugger's array of lines
780 if (PERLDB_LINE && PL_curstash != PL_debstash) {
781 SV * const sv = NEWSV(85,0);
783 sv_upgrade(sv, SVt_PVMG);
784 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
787 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
794 * Check the unary operators to ensure there's no ambiguity in how they're
795 * used. An ambiguous piece of code would be:
797 * This doesn't mean rand() + 5. Because rand() is a unary operator,
798 * the +5 is its argument.
807 if (PL_oldoldbufptr != PL_last_uni)
809 while (isSPACE(*PL_last_uni))
811 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
812 if ((t = strchr(s, '(')) && t < PL_bufptr)
814 if (ckWARN_d(WARN_AMBIGUOUS)){
817 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
818 "Warning: Use of \"%s\" without parentheses is ambiguous",
825 * LOP : macro to build a list operator. Its behaviour has been replaced
826 * with a subroutine, S_lop() for which LOP is just another name.
829 #define LOP(f,x) return lop(f,x,s)
833 * Build a list operator (or something that might be one). The rules:
834 * - if we have a next token, then it's a list operator [why?]
835 * - if the next thing is an opening paren, then it's a function
836 * - else it's a list operator
840 S_lop(pTHX_ I32 f, int x, char *s)
846 PL_last_lop = PL_oldbufptr;
847 PL_last_lop_op = (OPCODE)f;
849 return REPORT(LSTOP);
856 return REPORT(LSTOP);
861 * When the lexer realizes it knows the next token (for instance,
862 * it is reordering tokens for the parser) then it can call S_force_next
863 * to know what token to return the next time the lexer is called. Caller
864 * will need to set PL_nextval[], and possibly PL_expect to ensure the lexer
865 * handles the token correctly.
869 S_force_next(pTHX_ I32 type)
871 PL_nexttype[PL_nexttoke] = type;
873 if (PL_lex_state != LEX_KNOWNEXT) {
874 PL_lex_defer = PL_lex_state;
875 PL_lex_expect = PL_expect;
876 PL_lex_state = LEX_KNOWNEXT;
881 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
883 SV * const sv = newSVpvn(start,len);
884 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
891 * When the lexer knows the next thing is a word (for instance, it has
892 * just seen -> and it knows that the next char is a word char, then
893 * it calls S_force_word to stick the next word into the PL_next lookahead.
896 * char *start : buffer position (must be within PL_linestr)
897 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
898 * int check_keyword : if true, Perl checks to make sure the word isn't
899 * a keyword (do this if the word is a label, e.g. goto FOO)
900 * int allow_pack : if true, : characters will also be allowed (require,
902 * int allow_initial_tick : used by the "sub" lexer only.
906 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
911 start = skipspace(start);
913 if (isIDFIRST_lazy_if(s,UTF) ||
914 (allow_pack && *s == ':') ||
915 (allow_initial_tick && *s == '\'') )
917 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
918 if (check_keyword && keyword(PL_tokenbuf, len))
920 if (token == METHOD) {
925 PL_expect = XOPERATOR;
928 PL_nextval[PL_nexttoke].opval
929 = (OP*)newSVOP(OP_CONST,0,
930 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
931 PL_nextval[PL_nexttoke].opval->op_private |= OPpCONST_BARE;
939 * Called when the lexer wants $foo *foo &foo etc, but the program
940 * text only contains the "foo" portion. The first argument is a pointer
941 * to the "foo", and the second argument is the type symbol to prefix.
942 * Forces the next token to be a "WORD".
943 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
947 S_force_ident(pTHX_ register const char *s, int kind)
950 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpv(s,0));
951 PL_nextval[PL_nexttoke].opval = o;
954 o->op_private = OPpCONST_ENTERED;
955 /* XXX see note in pp_entereval() for why we forgo typo
956 warnings if the symbol must be introduced in an eval.
958 gv_fetchpv(s, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
959 kind == '$' ? SVt_PV :
960 kind == '@' ? SVt_PVAV :
961 kind == '%' ? SVt_PVHV :
969 Perl_str_to_version(pTHX_ SV *sv)
974 const char *start = SvPV_const(sv,len);
975 const char * const end = start + len;
976 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
977 while (start < end) {
981 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
986 retval += ((NV)n)/nshift;
995 * Forces the next token to be a version number.
996 * If the next token appears to be an invalid version number, (e.g. "v2b"),
997 * and if "guessing" is TRUE, then no new token is created (and the caller
998 * must use an alternative parsing method).
1002 S_force_version(pTHX_ char *s, int guessing)
1004 OP *version = Nullop;
1013 while (isDIGIT(*d) || *d == '_' || *d == '.')
1015 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1017 s = scan_num(s, &yylval);
1018 version = yylval.opval;
1019 ver = cSVOPx(version)->op_sv;
1020 if (SvPOK(ver) && !SvNIOK(ver)) {
1021 SvUPGRADE(ver, SVt_PVNV);
1022 SvNV_set(ver, str_to_version(ver));
1023 SvNOK_on(ver); /* hint that it is a version */
1030 /* NOTE: The parser sees the package name and the VERSION swapped */
1031 PL_nextval[PL_nexttoke].opval = version;
1039 * Tokenize a quoted string passed in as an SV. It finds the next
1040 * chunk, up to end of string or a backslash. It may make a new
1041 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1046 S_tokeq(pTHX_ SV *sv)
1049 register char *send;
1057 s = SvPV_force(sv, len);
1058 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1061 while (s < send && *s != '\\')
1066 if ( PL_hints & HINT_NEW_STRING ) {
1067 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1073 if (s + 1 < send && (s[1] == '\\'))
1074 s++; /* all that, just for this */
1079 SvCUR_set(sv, d - SvPVX_const(sv));
1081 if ( PL_hints & HINT_NEW_STRING )
1082 return new_constant(NULL, 0, "q", sv, pv, "q");
1087 * Now come three functions related to double-quote context,
1088 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1089 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1090 * interact with PL_lex_state, and create fake ( ... ) argument lists
1091 * to handle functions and concatenation.
1092 * They assume that whoever calls them will be setting up a fake
1093 * join call, because each subthing puts a ',' after it. This lets
1096 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1098 * (I'm not sure whether the spurious commas at the end of lcfirst's
1099 * arguments and join's arguments are created or not).
1104 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1106 * Pattern matching will set PL_lex_op to the pattern-matching op to
1107 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1109 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1111 * Everything else becomes a FUNC.
1113 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1114 * had an OP_CONST or OP_READLINE). This just sets us up for a
1115 * call to S_sublex_push().
1119 S_sublex_start(pTHX)
1121 const register I32 op_type = yylval.ival;
1123 if (op_type == OP_NULL) {
1124 yylval.opval = PL_lex_op;
1128 if (op_type == OP_CONST || op_type == OP_READLINE) {
1129 SV *sv = tokeq(PL_lex_stuff);
1131 if (SvTYPE(sv) == SVt_PVIV) {
1132 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1134 const char *p = SvPV_const(sv, len);
1135 SV * const nsv = newSVpvn(p, len);
1141 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1142 PL_lex_stuff = Nullsv;
1143 /* Allow <FH> // "foo" */
1144 if (op_type == OP_READLINE)
1145 PL_expect = XTERMORDORDOR;
1149 PL_sublex_info.super_state = PL_lex_state;
1150 PL_sublex_info.sub_inwhat = op_type;
1151 PL_sublex_info.sub_op = PL_lex_op;
1152 PL_lex_state = LEX_INTERPPUSH;
1156 yylval.opval = PL_lex_op;
1166 * Create a new scope to save the lexing state. The scope will be
1167 * ended in S_sublex_done. Returns a '(', starting the function arguments
1168 * to the uc, lc, etc. found before.
1169 * Sets PL_lex_state to LEX_INTERPCONCAT.
1178 PL_lex_state = PL_sublex_info.super_state;
1179 SAVEI32(PL_lex_dojoin);
1180 SAVEI32(PL_lex_brackets);
1181 SAVEI32(PL_lex_casemods);
1182 SAVEI32(PL_lex_starts);
1183 SAVEI32(PL_lex_state);
1184 SAVEVPTR(PL_lex_inpat);
1185 SAVEI32(PL_lex_inwhat);
1186 SAVECOPLINE(PL_curcop);
1187 SAVEPPTR(PL_bufptr);
1188 SAVEPPTR(PL_bufend);
1189 SAVEPPTR(PL_oldbufptr);
1190 SAVEPPTR(PL_oldoldbufptr);
1191 SAVEPPTR(PL_last_lop);
1192 SAVEPPTR(PL_last_uni);
1193 SAVEPPTR(PL_linestart);
1194 SAVESPTR(PL_linestr);
1195 SAVEGENERICPV(PL_lex_brackstack);
1196 SAVEGENERICPV(PL_lex_casestack);
1198 PL_linestr = PL_lex_stuff;
1199 PL_lex_stuff = Nullsv;
1201 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1202 = SvPVX(PL_linestr);
1203 PL_bufend += SvCUR(PL_linestr);
1204 PL_last_lop = PL_last_uni = Nullch;
1205 SAVEFREESV(PL_linestr);
1207 PL_lex_dojoin = FALSE;
1208 PL_lex_brackets = 0;
1209 Newx(PL_lex_brackstack, 120, char);
1210 Newx(PL_lex_casestack, 12, char);
1211 PL_lex_casemods = 0;
1212 *PL_lex_casestack = '\0';
1214 PL_lex_state = LEX_INTERPCONCAT;
1215 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1217 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1218 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1219 PL_lex_inpat = PL_sublex_info.sub_op;
1221 PL_lex_inpat = Nullop;
1228 * Restores lexer state after a S_sublex_push.
1235 if (!PL_lex_starts++) {
1236 SV * const sv = newSVpvn("",0);
1237 if (SvUTF8(PL_linestr))
1239 PL_expect = XOPERATOR;
1240 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1244 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1245 PL_lex_state = LEX_INTERPCASEMOD;
1249 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1250 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1251 PL_linestr = PL_lex_repl;
1253 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1254 PL_bufend += SvCUR(PL_linestr);
1255 PL_last_lop = PL_last_uni = Nullch;
1256 SAVEFREESV(PL_linestr);
1257 PL_lex_dojoin = FALSE;
1258 PL_lex_brackets = 0;
1259 PL_lex_casemods = 0;
1260 *PL_lex_casestack = '\0';
1262 if (SvEVALED(PL_lex_repl)) {
1263 PL_lex_state = LEX_INTERPNORMAL;
1265 /* we don't clear PL_lex_repl here, so that we can check later
1266 whether this is an evalled subst; that means we rely on the
1267 logic to ensure sublex_done() is called again only via the
1268 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1271 PL_lex_state = LEX_INTERPCONCAT;
1272 PL_lex_repl = Nullsv;
1278 PL_bufend = SvPVX(PL_linestr);
1279 PL_bufend += SvCUR(PL_linestr);
1280 PL_expect = XOPERATOR;
1281 PL_sublex_info.sub_inwhat = 0;
1289 Extracts a pattern, double-quoted string, or transliteration. This
1292 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1293 processing a pattern (PL_lex_inpat is true), a transliteration
1294 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1296 Returns a pointer to the character scanned up to. Iff this is
1297 advanced from the start pointer supplied (ie if anything was
1298 successfully parsed), will leave an OP for the substring scanned
1299 in yylval. Caller must intuit reason for not parsing further
1300 by looking at the next characters herself.
1304 double-quoted style: \r and \n
1305 regexp special ones: \D \s
1307 backrefs: \1 (deprecated in substitution replacements)
1308 case and quoting: \U \Q \E
1309 stops on @ and $, but not for $ as tail anchor
1311 In transliterations:
1312 characters are VERY literal, except for - not at the start or end
1313 of the string, which indicates a range. scan_const expands the
1314 range to the full set of intermediate characters.
1316 In double-quoted strings:
1318 double-quoted style: \r and \n
1320 backrefs: \1 (deprecated)
1321 case and quoting: \U \Q \E
1324 scan_const does *not* construct ops to handle interpolated strings.
1325 It stops processing as soon as it finds an embedded $ or @ variable
1326 and leaves it to the caller to work out what's going on.
1328 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1330 $ in pattern could be $foo or could be tail anchor. Assumption:
1331 it's a tail anchor if $ is the last thing in the string, or if it's
1332 followed by one of ")| \n\t"
1334 \1 (backreferences) are turned into $1
1336 The structure of the code is
1337 while (there's a character to process) {
1338 handle transliteration ranges
1339 skip regexp comments
1340 skip # initiated comments in //x patterns
1341 check for embedded @foo
1342 check for embedded scalars
1344 leave intact backslashes from leave (below)
1345 deprecate \1 in strings and sub replacements
1346 handle string-changing backslashes \l \U \Q \E, etc.
1347 switch (what was escaped) {
1348 handle - in a transliteration (becomes a literal -)
1349 handle \132 octal characters
1350 handle 0x15 hex characters
1351 handle \cV (control V)
1352 handle printf backslashes (\f, \r, \n, etc)
1354 } (end if backslash)
1355 } (end while character to read)
1360 S_scan_const(pTHX_ char *start)
1362 register char *send = PL_bufend; /* end of the constant */
1363 SV *sv = NEWSV(93, send - start); /* sv for the constant */
1364 register char *s = start; /* start of the constant */
1365 register char *d = SvPVX(sv); /* destination for copies */
1366 bool dorange = FALSE; /* are we in a translit range? */
1367 bool didrange = FALSE; /* did we just finish a range? */
1368 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1369 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1372 UV literal_endpoint = 0;
1375 const char *leaveit = /* set of acceptably-backslashed characters */
1377 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1380 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1381 /* If we are doing a trans and we know we want UTF8 set expectation */
1382 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1383 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1387 while (s < send || dorange) {
1388 /* get transliterations out of the way (they're most literal) */
1389 if (PL_lex_inwhat == OP_TRANS) {
1390 /* expand a range A-Z to the full set of characters. AIE! */
1392 I32 i; /* current expanded character */
1393 I32 min; /* first character in range */
1394 I32 max; /* last character in range */
1397 char * const c = (char*)utf8_hop((U8*)d, -1);
1401 *c = (char)UTF_TO_NATIVE(0xff);
1402 /* mark the range as done, and continue */
1408 i = d - SvPVX_const(sv); /* remember current offset */
1409 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1410 d = SvPVX(sv) + i; /* refresh d after realloc */
1411 d -= 2; /* eat the first char and the - */
1413 min = (U8)*d; /* first char in range */
1414 max = (U8)d[1]; /* last char in range */
1418 "Invalid range \"%c-%c\" in transliteration operator",
1419 (char)min, (char)max);
1423 if (literal_endpoint == 2 &&
1424 ((isLOWER(min) && isLOWER(max)) ||
1425 (isUPPER(min) && isUPPER(max)))) {
1427 for (i = min; i <= max; i++)
1429 *d++ = NATIVE_TO_NEED(has_utf8,i);
1431 for (i = min; i <= max; i++)
1433 *d++ = NATIVE_TO_NEED(has_utf8,i);
1438 for (i = min; i <= max; i++)
1441 /* mark the range as done, and continue */
1445 literal_endpoint = 0;
1450 /* range begins (ignore - as first or last char) */
1451 else if (*s == '-' && s+1 < send && s != start) {
1453 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1456 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1466 literal_endpoint = 0;
1471 /* if we get here, we're not doing a transliteration */
1473 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1474 except for the last char, which will be done separately. */
1475 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1477 while (s+1 < send && *s != ')')
1478 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1480 else if (s[2] == '{' /* This should match regcomp.c */
1481 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1484 char *regparse = s + (s[2] == '{' ? 3 : 4);
1487 while (count && (c = *regparse)) {
1488 if (c == '\\' && regparse[1])
1496 if (*regparse != ')')
1497 regparse--; /* Leave one char for continuation. */
1498 while (s < regparse)
1499 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1503 /* likewise skip #-initiated comments in //x patterns */
1504 else if (*s == '#' && PL_lex_inpat &&
1505 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1506 while (s+1 < send && *s != '\n')
1507 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1510 /* check for embedded arrays
1511 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1513 else if (*s == '@' && s[1]
1514 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1517 /* check for embedded scalars. only stop if we're sure it's a
1520 else if (*s == '$') {
1521 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1523 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1524 break; /* in regexp, $ might be tail anchor */
1527 /* End of else if chain - OP_TRANS rejoin rest */
1530 if (*s == '\\' && s+1 < send) {
1533 /* some backslashes we leave behind */
1534 if (*leaveit && *s && strchr(leaveit, *s)) {
1535 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1536 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1540 /* deprecate \1 in strings and substitution replacements */
1541 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1542 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1544 if (ckWARN(WARN_SYNTAX))
1545 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1550 /* string-change backslash escapes */
1551 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1556 /* if we get here, it's either a quoted -, or a digit */
1559 /* quoted - in transliterations */
1561 if (PL_lex_inwhat == OP_TRANS) {
1571 Perl_warner(aTHX_ packWARN(WARN_MISC),
1572 "Unrecognized escape \\%c passed through",
1574 /* default action is to copy the quoted character */
1575 goto default_action;
1578 /* \132 indicates an octal constant */
1579 case '0': case '1': case '2': case '3':
1580 case '4': case '5': case '6': case '7':
1584 uv = grok_oct(s, &len, &flags, NULL);
1587 goto NUM_ESCAPE_INSERT;
1589 /* \x24 indicates a hex constant */
1593 char* const e = strchr(s, '}');
1594 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
1595 PERL_SCAN_DISALLOW_PREFIX;
1600 yyerror("Missing right brace on \\x{}");
1604 uv = grok_hex(s, &len, &flags, NULL);
1610 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
1611 uv = grok_hex(s, &len, &flags, NULL);
1617 /* Insert oct or hex escaped character.
1618 * There will always enough room in sv since such
1619 * escapes will be longer than any UTF-8 sequence
1620 * they can end up as. */
1622 /* We need to map to chars to ASCII before doing the tests
1625 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
1626 if (!has_utf8 && uv > 255) {
1627 /* Might need to recode whatever we have
1628 * accumulated so far if it contains any
1631 * (Can't we keep track of that and avoid
1632 * this rescan? --jhi)
1636 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
1637 if (!NATIVE_IS_INVARIANT(*c)) {
1642 const STRLEN offset = d - SvPVX_const(sv);
1644 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
1648 while (src >= (const U8 *)SvPVX_const(sv)) {
1649 if (!NATIVE_IS_INVARIANT(*src)) {
1650 const U8 ch = NATIVE_TO_ASCII(*src);
1651 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
1652 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
1662 if (has_utf8 || uv > 255) {
1663 d = (char*)uvchr_to_utf8((U8*)d, uv);
1665 if (PL_lex_inwhat == OP_TRANS &&
1666 PL_sublex_info.sub_op) {
1667 PL_sublex_info.sub_op->op_private |=
1668 (PL_lex_repl ? OPpTRANS_FROM_UTF
1681 /* \N{LATIN SMALL LETTER A} is a named character */
1685 char* e = strchr(s, '}');
1691 yyerror("Missing right brace on \\N{}");
1695 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
1697 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
1698 PERL_SCAN_DISALLOW_PREFIX;
1701 uv = grok_hex(s, &len, &flags, NULL);
1703 goto NUM_ESCAPE_INSERT;
1705 res = newSVpvn(s + 1, e - s - 1);
1706 res = new_constant( Nullch, 0, "charnames",
1707 res, Nullsv, "\\N{...}" );
1709 sv_utf8_upgrade(res);
1710 str = SvPV_const(res,len);
1711 #ifdef EBCDIC_NEVER_MIND
1712 /* charnames uses pack U and that has been
1713 * recently changed to do the below uni->native
1714 * mapping, so this would be redundant (and wrong,
1715 * the code point would be doubly converted).
1716 * But leave this in just in case the pack U change
1717 * gets revoked, but the semantics is still
1718 * desireable for charnames. --jhi */
1720 UV uv = utf8_to_uvchr((const U8*)str, 0);
1723 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
1725 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
1726 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
1727 str = SvPV_const(res, len);
1731 if (!has_utf8 && SvUTF8(res)) {
1732 const char * const ostart = SvPVX_const(sv);
1733 SvCUR_set(sv, d - ostart);
1736 sv_utf8_upgrade(sv);
1737 /* this just broke our allocation above... */
1738 SvGROW(sv, (STRLEN)(send - start));
1739 d = SvPVX(sv) + SvCUR(sv);
1742 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
1743 const char * const odest = SvPVX_const(sv);
1745 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
1746 d = SvPVX(sv) + (d - odest);
1748 Copy(str, d, len, char);
1755 yyerror("Missing braces on \\N{}");
1758 /* \c is a control character */
1767 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
1770 yyerror("Missing control char name in \\c");
1774 /* printf-style backslashes, formfeeds, newlines, etc */
1776 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
1779 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
1782 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
1785 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
1788 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
1791 *d++ = ASCII_TO_NEED(has_utf8,'\033');
1794 *d++ = ASCII_TO_NEED(has_utf8,'\007');
1800 } /* end if (backslash) */
1807 /* If we started with encoded form, or already know we want it
1808 and then encode the next character */
1809 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
1811 const UV uv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
1812 const STRLEN need = UNISKIP(NATIVE_TO_UNI(uv));
1815 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
1816 const STRLEN off = d - SvPVX_const(sv);
1817 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
1819 d = (char*)uvchr_to_utf8((U8*)d, uv);
1823 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1825 } /* while loop to process each character */
1827 /* terminate the string and set up the sv */
1829 SvCUR_set(sv, d - SvPVX_const(sv));
1830 if (SvCUR(sv) >= SvLEN(sv))
1831 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
1834 if (PL_encoding && !has_utf8) {
1835 sv_recode_to_utf8(sv, PL_encoding);
1841 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1842 PL_sublex_info.sub_op->op_private |=
1843 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1847 /* shrink the sv if we allocated more than we used */
1848 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1849 SvPV_shrink_to_cur(sv);
1852 /* return the substring (via yylval) only if we parsed anything */
1853 if (s > PL_bufptr) {
1854 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
1855 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
1857 ( PL_lex_inwhat == OP_TRANS
1859 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
1862 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1869 * Returns TRUE if there's more to the expression (e.g., a subscript),
1872 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
1874 * ->[ and ->{ return TRUE
1875 * { and [ outside a pattern are always subscripts, so return TRUE
1876 * if we're outside a pattern and it's not { or [, then return FALSE
1877 * if we're in a pattern and the first char is a {
1878 * {4,5} (any digits around the comma) returns FALSE
1879 * if we're in a pattern and the first char is a [
1881 * [SOMETHING] has a funky algorithm to decide whether it's a
1882 * character class or not. It has to deal with things like
1883 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
1884 * anything else returns TRUE
1887 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
1890 S_intuit_more(pTHX_ register char *s)
1892 if (PL_lex_brackets)
1894 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
1896 if (*s != '{' && *s != '[')
1901 /* In a pattern, so maybe we have {n,m}. */
1918 /* On the other hand, maybe we have a character class */
1921 if (*s == ']' || *s == '^')
1924 /* this is terrifying, and it works */
1925 int weight = 2; /* let's weigh the evidence */
1927 unsigned char un_char = 255, last_un_char;
1928 const char * const send = strchr(s,']');
1929 char tmpbuf[sizeof PL_tokenbuf * 4];
1931 if (!send) /* has to be an expression */
1934 Zero(seen,256,char);
1937 else if (isDIGIT(*s)) {
1939 if (isDIGIT(s[1]) && s[2] == ']')
1945 for (; s < send; s++) {
1946 last_un_char = un_char;
1947 un_char = (unsigned char)*s;
1952 weight -= seen[un_char] * 10;
1953 if (isALNUM_lazy_if(s+1,UTF)) {
1954 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
1955 if ((int)strlen(tmpbuf) > 1 && gv_fetchpv(tmpbuf,FALSE, SVt_PV))
1960 else if (*s == '$' && s[1] &&
1961 strchr("[#!%*<>()-=",s[1])) {
1962 if (/*{*/ strchr("])} =",s[2]))
1971 if (strchr("wds]",s[1]))
1973 else if (seen['\''] || seen['"'])
1975 else if (strchr("rnftbxcav",s[1]))
1977 else if (isDIGIT(s[1])) {
1979 while (s[1] && isDIGIT(s[1]))
1989 if (strchr("aA01! ",last_un_char))
1991 if (strchr("zZ79~",s[1]))
1993 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
1994 weight -= 5; /* cope with negative subscript */
1997 if (!isALNUM(last_un_char)
1998 && !(last_un_char == '$' || last_un_char == '@'
1999 || last_un_char == '&')
2000 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2005 if (keyword(tmpbuf, d - tmpbuf))
2008 if (un_char == last_un_char + 1)
2010 weight -= seen[un_char];
2015 if (weight >= 0) /* probably a character class */
2025 * Does all the checking to disambiguate
2027 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2028 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2030 * First argument is the stuff after the first token, e.g. "bar".
2032 * Not a method if bar is a filehandle.
2033 * Not a method if foo is a subroutine prototyped to take a filehandle.
2034 * Not a method if it's really "Foo $bar"
2035 * Method if it's "foo $bar"
2036 * Not a method if it's really "print foo $bar"
2037 * Method if it's really "foo package::" (interpreted as package->foo)
2038 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2039 * Not a method if bar is a filehandle or package, but is quoted with
2044 S_intuit_method(pTHX_ char *start, GV *gv)
2046 char *s = start + (*start == '$');
2047 char tmpbuf[sizeof PL_tokenbuf];
2055 if ((cv = GvCVu(gv))) {
2056 const char *proto = SvPVX_const(cv);
2066 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2067 /* start is the beginning of the possible filehandle/object,
2068 * and s is the end of it
2069 * tmpbuf is a copy of it
2072 if (*start == '$') {
2073 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2078 return *s == '(' ? FUNCMETH : METHOD;
2080 if (!keyword(tmpbuf, len)) {
2081 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2086 indirgv = gv_fetchpv(tmpbuf, FALSE, SVt_PVCV);
2087 if (indirgv && GvCVu(indirgv))
2089 /* filehandle or package name makes it a method */
2090 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2092 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2093 return 0; /* no assumptions -- "=>" quotes bearword */
2095 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0,
2096 newSVpvn(tmpbuf,len));
2097 PL_nextval[PL_nexttoke].opval->op_private = OPpCONST_BARE;
2101 return *s == '(' ? FUNCMETH : METHOD;
2109 * Return a string of Perl code to load the debugger. If PERL5DB
2110 * is set, it will return the contents of that, otherwise a
2111 * compile-time require of perl5db.pl.
2118 const char * const pdb = PerlEnv_getenv("PERL5DB");
2122 SETERRNO(0,SS_NORMAL);
2123 return "BEGIN { require 'perl5db.pl' }";
2129 /* Encoded script support. filter_add() effectively inserts a
2130 * 'pre-processing' function into the current source input stream.
2131 * Note that the filter function only applies to the current source file
2132 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2134 * The datasv parameter (which may be NULL) can be used to pass
2135 * private data to this instance of the filter. The filter function
2136 * can recover the SV using the FILTER_DATA macro and use it to
2137 * store private buffers and state information.
2139 * The supplied datasv parameter is upgraded to a PVIO type
2140 * and the IoDIRP/IoANY field is used to store the function pointer,
2141 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2142 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2143 * private use must be set using malloc'd pointers.
2147 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2152 if (!PL_rsfp_filters)
2153 PL_rsfp_filters = newAV();
2155 datasv = NEWSV(255,0);
2156 SvUPGRADE(datasv, SVt_PVIO);
2157 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2158 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2159 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2160 IoANY(datasv), SvPV_nolen(datasv)));
2161 av_unshift(PL_rsfp_filters, 1);
2162 av_store(PL_rsfp_filters, 0, datasv) ;
2167 /* Delete most recently added instance of this filter function. */
2169 Perl_filter_del(pTHX_ filter_t funcp)
2174 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2176 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2178 /* if filter is on top of stack (usual case) just pop it off */
2179 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2180 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2181 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2182 IoANY(datasv) = (void *)NULL;
2183 sv_free(av_pop(PL_rsfp_filters));
2187 /* we need to search for the correct entry and clear it */
2188 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2192 /* Invoke the idxth filter function for the current rsfp. */
2193 /* maxlen 0 = read one text line */
2195 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2200 if (!PL_rsfp_filters)
2202 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2203 /* Provide a default input filter to make life easy. */
2204 /* Note that we append to the line. This is handy. */
2205 DEBUG_P(PerlIO_printf(Perl_debug_log,
2206 "filter_read %d: from rsfp\n", idx));
2210 const int old_len = SvCUR(buf_sv);
2212 /* ensure buf_sv is large enough */
2213 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2214 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2215 if (PerlIO_error(PL_rsfp))
2216 return -1; /* error */
2218 return 0 ; /* end of file */
2220 SvCUR_set(buf_sv, old_len + len) ;
2223 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2224 if (PerlIO_error(PL_rsfp))
2225 return -1; /* error */
2227 return 0 ; /* end of file */
2230 return SvCUR(buf_sv);
2232 /* Skip this filter slot if filter has been deleted */
2233 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2234 DEBUG_P(PerlIO_printf(Perl_debug_log,
2235 "filter_read %d: skipped (filter deleted)\n",
2237 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2239 /* Get function pointer hidden within datasv */
2240 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2241 DEBUG_P(PerlIO_printf(Perl_debug_log,
2242 "filter_read %d: via function %p (%s)\n",
2243 idx, datasv, SvPV_nolen_const(datasv)));
2244 /* Call function. The function is expected to */
2245 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2246 /* Return: <0:error, =0:eof, >0:not eof */
2247 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2251 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2253 #ifdef PERL_CR_FILTER
2254 if (!PL_rsfp_filters) {
2255 filter_add(S_cr_textfilter,NULL);
2258 if (PL_rsfp_filters) {
2260 SvCUR_set(sv, 0); /* start with empty line */
2261 if (FILTER_READ(0, sv, 0) > 0)
2262 return ( SvPVX(sv) ) ;
2267 return (sv_gets(sv, fp, append));
2271 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2275 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2279 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2280 (gv = gv_fetchpv(pkgname, FALSE, SVt_PVHV)))
2282 return GvHV(gv); /* Foo:: */
2285 /* use constant CLASS => 'MyClass' */
2286 if ((gv = gv_fetchpv(pkgname, FALSE, SVt_PVCV))) {
2288 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2289 pkgname = SvPV_nolen_const(sv);
2293 return gv_stashpv(pkgname, FALSE);
2297 static const char* const exp_name[] =
2298 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2299 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2306 Works out what to call the token just pulled out of the input
2307 stream. The yacc parser takes care of taking the ops we return and
2308 stitching them into a tree.
2314 if read an identifier
2315 if we're in a my declaration
2316 croak if they tried to say my($foo::bar)
2317 build the ops for a my() declaration
2318 if it's an access to a my() variable
2319 are we in a sort block?
2320 croak if my($a); $a <=> $b
2321 build ops for access to a my() variable
2322 if in a dq string, and they've said @foo and we can't find @foo
2324 build ops for a bareword
2325 if we already built the token before, use it.
2330 #pragma segment Perl_yylex
2335 register char *s = PL_bufptr;
2342 I32 orig_keyword = 0;
2345 PerlIO_printf(Perl_debug_log, "### LEX_%s\n",
2346 lex_state_names[PL_lex_state]);
2348 /* check if there's an identifier for us to look at */
2349 if (PL_pending_ident)
2350 return REPORT(S_pending_ident(aTHX));
2352 /* no identifier pending identification */
2354 switch (PL_lex_state) {
2356 case LEX_NORMAL: /* Some compilers will produce faster */
2357 case LEX_INTERPNORMAL: /* code if we comment these out. */
2361 /* when we've already built the next token, just pull it out of the queue */
2364 yylval = PL_nextval[PL_nexttoke];
2366 PL_lex_state = PL_lex_defer;
2367 PL_expect = PL_lex_expect;
2368 PL_lex_defer = LEX_NORMAL;
2370 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2371 "### Next token after '%s' was known, type %"IVdf"\n", PL_bufptr,
2372 (IV)PL_nexttype[PL_nexttoke]); });
2374 return REPORT(PL_nexttype[PL_nexttoke]);
2376 /* interpolated case modifiers like \L \U, including \Q and \E.
2377 when we get here, PL_bufptr is at the \
2379 case LEX_INTERPCASEMOD:
2381 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
2382 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
2384 /* handle \E or end of string */
2385 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
2387 if (PL_lex_casemods) {
2388 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
2389 PL_lex_casestack[PL_lex_casemods] = '\0';
2391 if (PL_bufptr != PL_bufend
2392 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
2394 PL_lex_state = LEX_INTERPCONCAT;
2398 if (PL_bufptr != PL_bufend)
2400 PL_lex_state = LEX_INTERPCONCAT;
2404 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2405 "### Saw case modifier at '%s'\n", PL_bufptr); });
2407 if (s[1] == '\\' && s[2] == 'E') {
2409 PL_lex_state = LEX_INTERPCONCAT;
2413 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
2414 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
2415 if ((*s == 'L' || *s == 'U') &&
2416 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
2417 PL_lex_casestack[--PL_lex_casemods] = '\0';
2420 if (PL_lex_casemods > 10)
2421 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
2422 PL_lex_casestack[PL_lex_casemods++] = *s;
2423 PL_lex_casestack[PL_lex_casemods] = '\0';
2424 PL_lex_state = LEX_INTERPCONCAT;
2425 PL_nextval[PL_nexttoke].ival = 0;
2428 PL_nextval[PL_nexttoke].ival = OP_LCFIRST;
2430 PL_nextval[PL_nexttoke].ival = OP_UCFIRST;
2432 PL_nextval[PL_nexttoke].ival = OP_LC;
2434 PL_nextval[PL_nexttoke].ival = OP_UC;
2436 PL_nextval[PL_nexttoke].ival = OP_QUOTEMETA;
2438 Perl_croak(aTHX_ "panic: yylex");
2442 if (PL_lex_starts) {
2445 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
2446 if (PL_lex_casemods == 1 && PL_lex_inpat)
2455 case LEX_INTERPPUSH:
2456 return REPORT(sublex_push());
2458 case LEX_INTERPSTART:
2459 if (PL_bufptr == PL_bufend)
2460 return REPORT(sublex_done());
2461 DEBUG_T({ PerlIO_printf(Perl_debug_log,
2462 "### Interpolated variable at '%s'\n", PL_bufptr); });
2464 PL_lex_dojoin = (*PL_bufptr == '@');
2465 PL_lex_state = LEX_INTERPNORMAL;
2466 if (PL_lex_dojoin) {
2467 PL_nextval[PL_nexttoke].ival = 0;
2469 force_ident("\"", '$');
2470 PL_nextval[PL_nexttoke].ival = 0;
2472 PL_nextval[PL_nexttoke].ival = 0;
2474 PL_nextval[PL_nexttoke].ival = OP_JOIN; /* emulate join($", ...) */
2477 if (PL_lex_starts++) {
2479 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
2480 if (!PL_lex_casemods && PL_lex_inpat)
2487 case LEX_INTERPENDMAYBE:
2488 if (intuit_more(PL_bufptr)) {
2489 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
2495 if (PL_lex_dojoin) {
2496 PL_lex_dojoin = FALSE;
2497 PL_lex_state = LEX_INTERPCONCAT;
2500 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
2501 && SvEVALED(PL_lex_repl))
2503 if (PL_bufptr != PL_bufend)
2504 Perl_croak(aTHX_ "Bad evalled substitution pattern");
2505 PL_lex_repl = Nullsv;
2508 case LEX_INTERPCONCAT:
2510 if (PL_lex_brackets)
2511 Perl_croak(aTHX_ "panic: INTERPCONCAT");
2513 if (PL_bufptr == PL_bufend)
2514 return REPORT(sublex_done());
2516 if (SvIVX(PL_linestr) == '\'') {
2517 SV *sv = newSVsv(PL_linestr);
2520 else if ( PL_hints & HINT_NEW_RE )
2521 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
2522 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2526 s = scan_const(PL_bufptr);
2528 PL_lex_state = LEX_INTERPCASEMOD;
2530 PL_lex_state = LEX_INTERPSTART;
2533 if (s != PL_bufptr) {
2534 PL_nextval[PL_nexttoke] = yylval;
2537 if (PL_lex_starts++) {
2538 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
2539 if (!PL_lex_casemods && PL_lex_inpat)
2552 PL_lex_state = LEX_NORMAL;
2553 s = scan_formline(PL_bufptr);
2554 if (!PL_lex_formbrack)
2560 PL_oldoldbufptr = PL_oldbufptr;
2563 PerlIO_printf(Perl_debug_log, "### Tokener expecting %s at [%s]\n",
2564 exp_name[PL_expect], s);
2570 if (isIDFIRST_lazy_if(s,UTF))
2572 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
2575 goto fake_eof; /* emulate EOF on ^D or ^Z */
2580 if (PL_lex_brackets) {
2581 if (PL_lex_formbrack)
2582 yyerror("Format not terminated");
2584 yyerror("Missing right curly or square bracket");
2586 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2587 "### Tokener got EOF\n");
2591 if (s++ < PL_bufend)
2592 goto retry; /* ignore stray nulls */
2595 if (!PL_in_eval && !PL_preambled) {
2596 PL_preambled = TRUE;
2597 sv_setpv(PL_linestr,incl_perldb());
2598 if (SvCUR(PL_linestr))
2599 sv_catpvn(PL_linestr,";", 1);
2601 while(AvFILLp(PL_preambleav) >= 0) {
2602 SV *tmpsv = av_shift(PL_preambleav);
2603 sv_catsv(PL_linestr, tmpsv);
2604 sv_catpvn(PL_linestr, ";", 1);
2607 sv_free((SV*)PL_preambleav);
2608 PL_preambleav = NULL;
2610 if (PL_minus_n || PL_minus_p) {
2611 sv_catpv(PL_linestr, "LINE: while (<>) {");
2613 sv_catpv(PL_linestr,"chomp;");
2616 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
2617 || *PL_splitstr == '"')
2618 && strchr(PL_splitstr + 1, *PL_splitstr))
2619 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
2621 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
2622 bytes can be used as quoting characters. :-) */
2623 /* The count here deliberately includes the NUL
2624 that terminates the C string constant. This
2625 embeds the opening NUL into the string. */
2626 const char *splits = PL_splitstr;
2627 sv_catpvn(PL_linestr, "our @F=split(q", 15);
2630 if (*splits == '\\')
2631 sv_catpvn(PL_linestr, splits, 1);
2632 sv_catpvn(PL_linestr, splits, 1);
2633 } while (*splits++);
2634 /* This loop will embed the trailing NUL of
2635 PL_linestr as the last thing it does before
2637 sv_catpvn(PL_linestr, ");", 2);
2641 sv_catpv(PL_linestr,"our @F=split(' ');");
2644 sv_catpvn(PL_linestr, "\n", 1);
2645 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2646 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2647 PL_last_lop = PL_last_uni = Nullch;
2648 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2649 SV * const sv = NEWSV(85,0);
2651 sv_upgrade(sv, SVt_PVMG);
2652 sv_setsv(sv,PL_linestr);
2655 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2660 bof = PL_rsfp ? TRUE : FALSE;
2661 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == Nullch) {
2664 if (PL_preprocess && !PL_in_eval)
2665 (void)PerlProc_pclose(PL_rsfp);
2666 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
2667 PerlIO_clearerr(PL_rsfp);
2669 (void)PerlIO_close(PL_rsfp);
2671 PL_doextract = FALSE;
2673 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
2674 sv_setpv(PL_linestr,PL_minus_p
2675 ? ";}continue{print;}" : ";}");
2676 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2677 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2678 PL_last_lop = PL_last_uni = Nullch;
2679 PL_minus_n = PL_minus_p = 0;
2682 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2683 PL_last_lop = PL_last_uni = Nullch;
2684 sv_setpvn(PL_linestr,"",0);
2685 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
2687 /* If it looks like the start of a BOM or raw UTF-16,
2688 * check if it in fact is. */
2694 #ifdef PERLIO_IS_STDIO
2695 # ifdef __GNU_LIBRARY__
2696 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
2697 # define FTELL_FOR_PIPE_IS_BROKEN
2701 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
2702 # define FTELL_FOR_PIPE_IS_BROKEN
2707 #ifdef FTELL_FOR_PIPE_IS_BROKEN
2708 /* This loses the possibility to detect the bof
2709 * situation on perl -P when the libc5 is being used.
2710 * Workaround? Maybe attach some extra state to PL_rsfp?
2713 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
2715 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
2718 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2719 s = swallow_bom((U8*)s);
2723 /* Incest with pod. */
2724 if (*s == '=' && strnEQ(s, "=cut", 4)) {
2725 sv_setpvn(PL_linestr, "", 0);
2726 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2727 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2728 PL_last_lop = PL_last_uni = Nullch;
2729 PL_doextract = FALSE;
2733 } while (PL_doextract);
2734 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
2735 if (PERLDB_LINE && PL_curstash != PL_debstash) {
2736 SV * const sv = NEWSV(85,0);
2738 sv_upgrade(sv, SVt_PVMG);
2739 sv_setsv(sv,PL_linestr);
2742 av_store(CopFILEAV(PL_curcop),(I32)CopLINE(PL_curcop),sv);
2744 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2745 PL_last_lop = PL_last_uni = Nullch;
2746 if (CopLINE(PL_curcop) == 1) {
2747 while (s < PL_bufend && isSPACE(*s))
2749 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
2753 if (*s == '#' && *(s+1) == '!')
2755 #ifdef ALTERNATE_SHEBANG
2757 static char const as[] = ALTERNATE_SHEBANG;
2758 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
2759 d = s + (sizeof(as) - 1);
2761 #endif /* ALTERNATE_SHEBANG */
2770 while (*d && !isSPACE(*d))
2774 #ifdef ARG_ZERO_IS_SCRIPT
2775 if (ipathend > ipath) {
2777 * HP-UX (at least) sets argv[0] to the script name,
2778 * which makes $^X incorrect. And Digital UNIX and Linux,
2779 * at least, set argv[0] to the basename of the Perl
2780 * interpreter. So, having found "#!", we'll set it right.
2782 SV *x = GvSV(gv_fetchpv("\030", TRUE, SVt_PV)); /* $^X */
2783 assert(SvPOK(x) || SvGMAGICAL(x));
2784 if (sv_eq(x, CopFILESV(PL_curcop))) {
2785 sv_setpvn(x, ipath, ipathend - ipath);
2791 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
2792 const char * const lstart = SvPV_const(x,llen);
2794 bstart += blen - llen;
2795 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
2796 sv_setpvn(x, ipath, ipathend - ipath);
2801 TAINT_NOT; /* $^X is always tainted, but that's OK */
2803 #endif /* ARG_ZERO_IS_SCRIPT */
2808 d = instr(s,"perl -");
2810 d = instr(s,"perl");
2812 /* avoid getting into infinite loops when shebang
2813 * line contains "Perl" rather than "perl" */
2815 for (d = ipathend-4; d >= ipath; --d) {
2816 if ((*d == 'p' || *d == 'P')
2817 && !ibcmp(d, "perl", 4))
2827 #ifdef ALTERNATE_SHEBANG
2829 * If the ALTERNATE_SHEBANG on this system starts with a
2830 * character that can be part of a Perl expression, then if
2831 * we see it but not "perl", we're probably looking at the
2832 * start of Perl code, not a request to hand off to some
2833 * other interpreter. Similarly, if "perl" is there, but
2834 * not in the first 'word' of the line, we assume the line
2835 * contains the start of the Perl program.
2837 if (d && *s != '#') {
2838 const char *c = ipath;
2839 while (*c && !strchr("; \t\r\n\f\v#", *c))
2842 d = Nullch; /* "perl" not in first word; ignore */
2844 *s = '#'; /* Don't try to parse shebang line */
2846 #endif /* ALTERNATE_SHEBANG */
2847 #ifndef MACOS_TRADITIONAL
2852 !instr(s,"indir") &&
2853 instr(PL_origargv[0],"perl"))
2860 while (s < PL_bufend && isSPACE(*s))
2862 if (s < PL_bufend) {
2863 Newxz(newargv,PL_origargc+3,char*);
2865 while (s < PL_bufend && !isSPACE(*s))
2868 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
2871 newargv = PL_origargv;
2874 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
2876 Perl_croak(aTHX_ "Can't exec %s", ipath);
2880 const U32 oldpdb = PL_perldb;
2881 const bool oldn = PL_minus_n;
2882 const bool oldp = PL_minus_p;
2884 while (*d && !isSPACE(*d)) d++;
2885 while (SPACE_OR_TAB(*d)) d++;
2888 const bool switches_done = PL_doswitches;
2890 if (*d == 'M' || *d == 'm' || *d == 'C') {
2891 const char * const m = d;
2892 while (*d && !isSPACE(*d)) d++;
2893 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
2896 d = moreswitches(d);
2898 if (PL_doswitches && !switches_done) {
2899 int argc = PL_origargc;
2900 char **argv = PL_origargv;
2903 } while (argc && argv[0][0] == '-' && argv[0][1]);
2904 init_argv_symbols(argc,argv);
2906 if ((PERLDB_LINE && !oldpdb) ||
2907 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
2908 /* if we have already added "LINE: while (<>) {",
2909 we must not do it again */
2911 sv_setpvn(PL_linestr, "", 0);
2912 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
2913 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
2914 PL_last_lop = PL_last_uni = Nullch;
2915 PL_preambled = FALSE;
2917 (void)gv_fetchfile(PL_origfilename);
2920 if (PL_doswitches && !switches_done) {
2921 int argc = PL_origargc;
2922 char **argv = PL_origargv;
2925 } while (argc && argv[0][0] == '-' && argv[0][1]);
2926 init_argv_symbols(argc,argv);
2932 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2934 PL_lex_state = LEX_FORMLINE;
2939 #ifdef PERL_STRICT_CR
2940 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
2942 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
2944 case ' ': case '\t': case '\f': case 013:
2945 #ifdef MACOS_TRADITIONAL
2952 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
2953 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
2954 /* handle eval qq[#line 1 "foo"\n ...] */
2955 CopLINE_dec(PL_curcop);
2959 while (s < d && *s != '\n')
2963 else if (s > d) /* Found by Ilya: feed random input to Perl. */
2964 Perl_croak(aTHX_ "panic: input overflow");
2966 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
2968 PL_lex_state = LEX_FORMLINE;
2978 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
2985 while (s < PL_bufend && SPACE_OR_TAB(*s))
2988 if (strnEQ(s,"=>",2)) {
2989 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
2990 DEBUG_T( { PerlIO_printf(Perl_debug_log,
2991 "### Saw unary minus before =>, forcing word '%s'\n", s);
2993 OPERATOR('-'); /* unary minus */
2995 PL_last_uni = PL_oldbufptr;
2997 case 'r': ftst = OP_FTEREAD; break;
2998 case 'w': ftst = OP_FTEWRITE; break;
2999 case 'x': ftst = OP_FTEEXEC; break;
3000 case 'o': ftst = OP_FTEOWNED; break;
3001 case 'R': ftst = OP_FTRREAD; break;
3002 case 'W': ftst = OP_FTRWRITE; break;
3003 case 'X': ftst = OP_FTREXEC; break;
3004 case 'O': ftst = OP_FTROWNED; break;
3005 case 'e': ftst = OP_FTIS; break;
3006 case 'z': ftst = OP_FTZERO; break;
3007 case 's': ftst = OP_FTSIZE; break;
3008 case 'f': ftst = OP_FTFILE; break;
3009 case 'd': ftst = OP_FTDIR; break;
3010 case 'l': ftst = OP_FTLINK; break;
3011 case 'p': ftst = OP_FTPIPE; break;
3012 case 'S': ftst = OP_FTSOCK; break;
3013 case 'u': ftst = OP_FTSUID; break;
3014 case 'g': ftst = OP_FTSGID; break;
3015 case 'k': ftst = OP_FTSVTX; break;
3016 case 'b': ftst = OP_FTBLK; break;
3017 case 'c': ftst = OP_FTCHR; break;
3018 case 't': ftst = OP_FTTTY; break;
3019 case 'T': ftst = OP_FTTEXT; break;
3020 case 'B': ftst = OP_FTBINARY; break;
3021 case 'M': case 'A': case 'C':
3022 gv_fetchpv("\024",TRUE, SVt_PV);
3024 case 'M': ftst = OP_FTMTIME; break;
3025 case 'A': ftst = OP_FTATIME; break;
3026 case 'C': ftst = OP_FTCTIME; break;
3034 PL_last_lop_op = (OPCODE)ftst;
3035 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3036 "### Saw file test %c\n", (int)ftst);
3041 /* Assume it was a minus followed by a one-letter named
3042 * subroutine call (or a -bareword), then. */
3043 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3044 "### '-%c' looked like a file test but was not\n",
3053 if (PL_expect == XOPERATOR)
3058 else if (*s == '>') {
3061 if (isIDFIRST_lazy_if(s,UTF)) {
3062 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3070 if (PL_expect == XOPERATOR)
3073 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3075 OPERATOR('-'); /* unary minus */
3082 if (PL_expect == XOPERATOR)
3087 if (PL_expect == XOPERATOR)
3090 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3096 if (PL_expect != XOPERATOR) {
3097 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3098 PL_expect = XOPERATOR;
3099 force_ident(PL_tokenbuf, '*');
3112 if (PL_expect == XOPERATOR) {
3116 PL_tokenbuf[0] = '%';
3117 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3118 if (!PL_tokenbuf[1]) {
3121 PL_pending_ident = '%';
3140 switch (PL_expect) {
3143 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3145 PL_bufptr = s; /* update in case we back off */
3151 PL_expect = XTERMBLOCK;
3155 while (isIDFIRST_lazy_if(s,UTF)) {
3156 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3157 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
3158 if (tmp < 0) tmp = -tmp;
3174 d = scan_str(d,TRUE,TRUE);
3176 /* MUST advance bufptr here to avoid bogus
3177 "at end of line" context messages from yyerror().
3179 PL_bufptr = s + len;
3180 yyerror("Unterminated attribute parameter in attribute list");
3183 return REPORT(0); /* EOF indicator */
3187 SV *sv = newSVpvn(s, len);
3188 sv_catsv(sv, PL_lex_stuff);
3189 attrs = append_elem(OP_LIST, attrs,
3190 newSVOP(OP_CONST, 0, sv));
3191 SvREFCNT_dec(PL_lex_stuff);
3192 PL_lex_stuff = Nullsv;
3195 if (len == 6 && strnEQ(s, "unique", len)) {
3196 if (PL_in_my == KEY_our)
3198 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
3200 ; /* skip to avoid loading attributes.pm */
3203 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
3206 /* NOTE: any CV attrs applied here need to be part of
3207 the CVf_BUILTIN_ATTRS define in cv.h! */
3208 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
3209 CvLVALUE_on(PL_compcv);
3210 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
3211 CvLOCKED_on(PL_compcv);
3212 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
3213 CvMETHOD_on(PL_compcv);
3214 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
3215 CvASSERTION_on(PL_compcv);
3216 /* After we've set the flags, it could be argued that
3217 we don't need to do the attributes.pm-based setting
3218 process, and shouldn't bother appending recognized
3219 flags. To experiment with that, uncomment the
3220 following "else". (Note that's already been
3221 uncommented. That keeps the above-applied built-in
3222 attributes from being intercepted (and possibly
3223 rejected) by a package's attribute routines, but is
3224 justified by the performance win for the common case
3225 of applying only built-in attributes.) */
3227 attrs = append_elem(OP_LIST, attrs,
3228 newSVOP(OP_CONST, 0,
3232 if (*s == ':' && s[1] != ':')
3235 break; /* require real whitespace or :'s */
3237 tmp = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
3238 if (*s != ';' && *s != '}' && *s != tmp && (tmp != '=' || *s != ')')) {
3239 const char q = ((*s == '\'') ? '"' : '\'');
3240 /* If here for an expression, and parsed no attrs, back off. */
3241 if (tmp == '=' && !attrs) {
3245 /* MUST advance bufptr here to avoid bogus "at end of line"
3246 context messages from yyerror().
3250 yyerror("Unterminated attribute list");
3252 yyerror(Perl_form(aTHX_ "Invalid separator character %c%c%c in attribute list",
3260 PL_nextval[PL_nexttoke].opval = attrs;
3268 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
3269 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
3286 if (PL_lex_brackets <= 0)
3287 yyerror("Unmatched right square bracket");
3290 if (PL_lex_state == LEX_INTERPNORMAL) {
3291 if (PL_lex_brackets == 0) {
3292 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
3293 PL_lex_state = LEX_INTERPEND;
3300 if (PL_lex_brackets > 100) {
3301 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
3303 switch (PL_expect) {
3305 if (PL_lex_formbrack) {
3309 if (PL_oldoldbufptr == PL_last_lop)
3310 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3312 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3313 OPERATOR(HASHBRACK);
3315 while (s < PL_bufend && SPACE_OR_TAB(*s))
3318 PL_tokenbuf[0] = '\0';
3319 if (d < PL_bufend && *d == '-') {
3320 PL_tokenbuf[0] = '-';
3322 while (d < PL_bufend && SPACE_OR_TAB(*d))
3325 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
3326 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
3328 while (d < PL_bufend && SPACE_OR_TAB(*d))
3331 const char minus = (PL_tokenbuf[0] == '-');
3332 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
3340 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
3345 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3350 if (PL_oldoldbufptr == PL_last_lop)
3351 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
3353 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
3356 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
3358 /* This hack is to get the ${} in the message. */
3360 yyerror("syntax error");
3363 OPERATOR(HASHBRACK);
3365 /* This hack serves to disambiguate a pair of curlies
3366 * as being a block or an anon hash. Normally, expectation
3367 * determines that, but in cases where we're not in a
3368 * position to expect anything in particular (like inside
3369 * eval"") we have to resolve the ambiguity. This code
3370 * covers the case where the first term in the curlies is a
3371 * quoted string. Most other cases need to be explicitly
3372 * disambiguated by prepending a "+" before the opening
3373 * curly in order to force resolution as an anon hash.
3375 * XXX should probably propagate the outer expectation
3376 * into eval"" to rely less on this hack, but that could
3377 * potentially break current behavior of eval"".
3381 if (*s == '\'' || *s == '"' || *s == '`') {
3382 /* common case: get past first string, handling escapes */
3383 for (t++; t < PL_bufend && *t != *s;)
3384 if (*t++ == '\\' && (*t == '\\' || *t == *s))
3388 else if (*s == 'q') {
3391 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
3394 /* skip q//-like construct */
3396 char open, close, term;
3399 while (t < PL_bufend && isSPACE(*t))
3401 /* check for q => */
3402 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
3403 OPERATOR(HASHBRACK);
3407 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
3411 for (t++; t < PL_bufend; t++) {
3412 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
3414 else if (*t == open)
3418 for (t++; t < PL_bufend; t++) {
3419 if (*t == '\\' && t+1 < PL_bufend)
3421 else if (*t == close && --brackets <= 0)
3423 else if (*t == open)
3430 /* skip plain q word */
3431 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3434 else if (isALNUM_lazy_if(t,UTF)) {
3436 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
3439 while (t < PL_bufend && isSPACE(*t))
3441 /* if comma follows first term, call it an anon hash */
3442 /* XXX it could be a comma expression with loop modifiers */
3443 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
3444 || (*t == '=' && t[1] == '>')))
3445 OPERATOR(HASHBRACK);
3446 if (PL_expect == XREF)
3449 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
3455 yylval.ival = CopLINE(PL_curcop);
3456 if (isSPACE(*s) || *s == '#')
3457 PL_copline = NOLINE; /* invalidate current command line number */
3462 if (PL_lex_brackets <= 0)
3463 yyerror("Unmatched right curly bracket");
3465 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
3466 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
3467 PL_lex_formbrack = 0;
3468 if (PL_lex_state == LEX_INTERPNORMAL) {
3469 if (PL_lex_brackets == 0) {
3470 if (PL_expect & XFAKEBRACK) {
3471 PL_expect &= XENUMMASK;
3472 PL_lex_state = LEX_INTERPEND;
3474 return yylex(); /* ignore fake brackets */
3476 if (*s == '-' && s[1] == '>')
3477 PL_lex_state = LEX_INTERPENDMAYBE;
3478 else if (*s != '[' && *s != '{')
3479 PL_lex_state = LEX_INTERPEND;
3482 if (PL_expect & XFAKEBRACK) {
3483 PL_expect &= XENUMMASK;
3485 return yylex(); /* ignore fake brackets */
3495 if (PL_expect == XOPERATOR) {
3496 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
3497 && isIDFIRST_lazy_if(s,UTF))
3499 CopLINE_dec(PL_curcop);
3500 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
3501 CopLINE_inc(PL_curcop);
3506 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3508 PL_expect = XOPERATOR;
3509 force_ident(PL_tokenbuf, '&');
3513 yylval.ival = (OPpENTERSUB_AMPER<<8);
3532 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX) && strchr("+-*/%.^&|<",tmp))
3533 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Reversed %c= operator",(int)tmp);
3535 if (PL_expect == XSTATE && isALPHA(tmp) &&
3536 (s == PL_linestart+1 || s[-2] == '\n') )
3538 if (PL_in_eval && !PL_rsfp) {
3543 if (strnEQ(s,"=cut",4)) {
3557 PL_doextract = TRUE;
3560 if (PL_lex_brackets < PL_lex_formbrack) {
3562 #ifdef PERL_STRICT_CR
3563 for (t = s; SPACE_OR_TAB(*t); t++) ;
3565 for (t = s; SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
3567 if (*t == '\n' || *t == '#') {
3579 /* was this !=~ where !~ was meant?
3580 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
3582 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
3583 const char *t = s+1;
3585 while (t < PL_bufend && isSPACE(*t))
3588 if (*t == '/' || *t == '?' ||
3589 ((*t == 'm' || *t == 's' || *t == 'y') && !isALNUM(t[1])) ||
3590 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
3591 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3592 "!=~ should be !~");
3601 if (PL_expect != XOPERATOR) {
3602 if (s[1] != '<' && !strchr(s,'>'))
3605 s = scan_heredoc(s);
3607 s = scan_inputsymbol(s);
3608 TERM(sublex_start());
3613 SHop(OP_LEFT_SHIFT);
3627 SHop(OP_RIGHT_SHIFT);
3636 if (PL_expect == XOPERATOR) {
3637 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3640 return REPORT(','); /* grandfather non-comma-format format */
3644 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
3645 PL_tokenbuf[0] = '@';
3646 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
3647 sizeof PL_tokenbuf - 1, FALSE);
3648 if (PL_expect == XOPERATOR)
3649 no_op("Array length", s);
3650 if (!PL_tokenbuf[1])
3652 PL_expect = XOPERATOR;
3653 PL_pending_ident = '#';
3657 PL_tokenbuf[0] = '$';
3658 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
3659 sizeof PL_tokenbuf - 1, FALSE);
3660 if (PL_expect == XOPERATOR)
3662 if (!PL_tokenbuf[1]) {
3664 yyerror("Final $ should be \\$ or $name");
3668 /* This kludge not intended to be bulletproof. */
3669 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
3670 yylval.opval = newSVOP(OP_CONST, 0,
3671 newSViv(PL_compiling.cop_arybase));
3672 yylval.opval->op_private = OPpCONST_ARYBASE;
3678 if (PL_lex_state == LEX_NORMAL)
3681 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3683 PL_tokenbuf[0] = '@';
3684 if (ckWARN(WARN_SYNTAX)) {
3687 isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$';
3690 PL_bufptr = skipspace(PL_bufptr);
3691 while (t < PL_bufend && *t != ']')
3693 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3694 "Multidimensional syntax %.*s not supported",
3695 (t - PL_bufptr) + 1, PL_bufptr);
3699 else if (*s == '{') {
3701 PL_tokenbuf[0] = '%';
3702 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
3703 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
3705 char tmpbuf[sizeof PL_tokenbuf];
3706 for (t++; isSPACE(*t); t++) ;
3707 if (isIDFIRST_lazy_if(t,UTF)) {
3709 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE, &len);
3710 for (; isSPACE(*t); t++) ;
3711 if (*t == ';' && get_cv(tmpbuf, FALSE))
3712 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3713 "You need to quote \"%s\"", tmpbuf);
3719 PL_expect = XOPERATOR;
3720 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
3721 const bool islop = (PL_last_lop == PL_oldoldbufptr);
3722 if (!islop || PL_last_lop_op == OP_GREPSTART)
3723 PL_expect = XOPERATOR;
3724 else if (strchr("$@\"'`q", *s))
3725 PL_expect = XTERM; /* e.g. print $fh "foo" */
3726 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
3727 PL_expect = XTERM; /* e.g. print $fh &sub */
3728 else if (isIDFIRST_lazy_if(s,UTF)) {
3729 char tmpbuf[sizeof PL_tokenbuf];
3730 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3731 if ((tmp = keyword(tmpbuf, len))) {
3732 /* binary operators exclude handle interpretations */
3744 PL_expect = XTERM; /* e.g. print $fh length() */
3749 PL_expect = XTERM; /* e.g. print $fh subr() */
3752 else if (isDIGIT(*s))
3753 PL_expect = XTERM; /* e.g. print $fh 3 */
3754 else if (*s == '.' && isDIGIT(s[1]))
3755 PL_expect = XTERM; /* e.g. print $fh .3 */
3756 else if ((*s == '?' || *s == '-' || *s == '+')
3757 && !isSPACE(s[1]) && s[1] != '=')
3758 PL_expect = XTERM; /* e.g. print $fh -1 */
3759 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '=' && s[1] != '/')
3760 PL_expect = XTERM; /* e.g. print $fh /.../
3761 XXX except DORDOR operator */
3762 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2]) && s[2] != '=')
3763 PL_expect = XTERM; /* print $fh <<"EOF" */
3765 PL_pending_ident = '$';
3769 if (PL_expect == XOPERATOR)
3771 PL_tokenbuf[0] = '@';
3772 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
3773 if (!PL_tokenbuf[1]) {
3776 if (PL_lex_state == LEX_NORMAL)
3778 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
3780 PL_tokenbuf[0] = '%';
3782 /* Warn about @ where they meant $. */
3783 if (*s == '[' || *s == '{') {
3784 if (ckWARN(WARN_SYNTAX)) {
3785 const char *t = s + 1;
3786 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
3788 if (*t == '}' || *t == ']') {
3790 PL_bufptr = skipspace(PL_bufptr);
3791 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3792 "Scalar value %.*s better written as $%.*s",
3793 t-PL_bufptr, PL_bufptr, t-PL_bufptr-1, PL_bufptr+1);
3798 PL_pending_ident = '@';
3801 case '/': /* may be division, defined-or, or pattern */
3802 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
3806 case '?': /* may either be conditional or pattern */
3807 if(PL_expect == XOPERATOR) {
3815 /* A // operator. */
3825 /* Disable warning on "study /blah/" */
3826 if (PL_oldoldbufptr == PL_last_uni
3827 && (*PL_last_uni != 's' || s - PL_last_uni < 5
3828 || memNE(PL_last_uni, "study", 5)
3829 || isALNUM_lazy_if(PL_last_uni+5,UTF)
3832 s = scan_pat(s,OP_MATCH);
3833 TERM(sublex_start());
3837 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
3838 #ifdef PERL_STRICT_CR
3841 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
3843 && (s == PL_linestart || s[-1] == '\n') )
3845 PL_lex_formbrack = 0;
3849 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
3855 yylval.ival = OPf_SPECIAL;
3861 if (PL_expect != XOPERATOR)
3866 case '0': case '1': case '2': case '3': case '4':
3867 case '5': case '6': case '7': case '8': case '9':
3868 s = scan_num(s, &yylval);
3869 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3870 "### Saw number in '%s'\n", s);
3872 if (PL_expect == XOPERATOR)
3877 s = scan_str(s,FALSE,FALSE);
3878 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3879 "### Saw string before '%s'\n", s);
3881 if (PL_expect == XOPERATOR) {
3882 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3885 return REPORT(','); /* grandfather non-comma-format format */
3891 missingterm((char*)0);
3892 yylval.ival = OP_CONST;
3893 TERM(sublex_start());
3896 s = scan_str(s,FALSE,FALSE);
3897 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3898 "### Saw string before '%s'\n", s);
3900 if (PL_expect == XOPERATOR) {
3901 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
3904 return REPORT(','); /* grandfather non-comma-format format */
3910 missingterm((char*)0);
3911 yylval.ival = OP_CONST;
3912 /* FIXME. I think that this can be const if char *d is replaced by
3913 more localised variables. */
3914 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
3915 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
3916 yylval.ival = OP_STRINGIFY;
3920 TERM(sublex_start());
3923 s = scan_str(s,FALSE,FALSE);
3924 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3925 "### Saw backtick string before '%s'\n", s);
3927 if (PL_expect == XOPERATOR)
3928 no_op("Backticks",s);
3930 missingterm((char*)0);
3931 yylval.ival = OP_BACKTICK;
3933 TERM(sublex_start());
3937 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
3938 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
3940 if (PL_expect == XOPERATOR)
3941 no_op("Backslash",s);
3945 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
3946 char *start = s + 2;
3947 while (isDIGIT(*start) || *start == '_')
3949 if (*start == '.' && isDIGIT(start[1])) {
3950 s = scan_num(s, &yylval);
3953 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
3954 else if (!isALPHA(*start) && (PL_expect == XTERM
3955 || PL_expect == XREF || PL_expect == XSTATE
3956 || PL_expect == XTERMORDORDOR)) {
3957 const char c = *start;
3960 gv = gv_fetchpv(s, FALSE, SVt_PVCV);
3963 s = scan_num(s, &yylval);
3970 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
4010 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4012 /* Some keywords can be followed by any delimiter, including ':' */
4013 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
4014 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
4015 (PL_tokenbuf[0] == 'q' &&
4016 strchr("qwxr", PL_tokenbuf[1])))));
4018 /* x::* is just a word, unless x is "CORE" */
4019 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
4023 while (d < PL_bufend && isSPACE(*d))
4024 d++; /* no comments skipped here, or s### is misparsed */
4026 /* Is this a label? */
4027 if (!tmp && PL_expect == XSTATE
4028 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
4030 yylval.pval = savepv(PL_tokenbuf);
4035 /* Check for keywords */
4036 tmp = keyword(PL_tokenbuf, len);
4038 /* Is this a word before a => operator? */
4039 if (*d == '=' && d[1] == '>') {
4042 = (OP*)newSVOP(OP_CONST, 0,
4043 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
4044 yylval.opval->op_private = OPpCONST_BARE;
4048 if (tmp < 0) { /* second-class keyword? */
4049 GV *ogv = Nullgv; /* override (winner) */
4050 GV *hgv = Nullgv; /* hidden (loser) */
4051 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
4053 if ((gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV)) &&
4056 if (GvIMPORTED_CV(gv))
4058 else if (! CvMETHOD(cv))
4062 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
4063 (gv = *gvp) != (GV*)&PL_sv_undef &&
4064 GvCVu(gv) && GvIMPORTED_CV(gv))
4071 tmp = 0; /* overridden by import or by GLOBAL */
4074 && -tmp==KEY_lock /* XXX generalizable kludge */
4076 && !hv_fetch(GvHVn(PL_incgv), "Thread.pm", 9, FALSE))
4078 tmp = 0; /* any sub overrides "weak" keyword */
4083 && PL_expect != XOPERATOR
4084 && PL_expect != XTERMORDORDOR)
4086 /* any sub overrides the "err" keyword, except when really an
4087 * operator is expected */
4090 else { /* no override */
4092 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
4093 Perl_warner(aTHX_ packWARN(WARN_MISC),
4094 "dump() better written as CORE::dump()");
4098 if (hgv && tmp != KEY_x && tmp != KEY_CORE
4099 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
4100 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4101 "Ambiguous call resolved as CORE::%s(), %s",
4102 GvENAME(hgv), "qualify as such or use &");
4109 default: /* not a keyword */
4113 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
4115 /* Get the rest if it looks like a package qualifier */
4117 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
4119 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
4122 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
4123 *s == '\'' ? "'" : "::");
4128 if (PL_expect == XOPERATOR) {
4129 if (PL_bufptr == PL_linestart) {
4130 CopLINE_dec(PL_curcop);
4131 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4132 CopLINE_inc(PL_curcop);
4135 no_op("Bareword",s);
4138 /* Look for a subroutine with this name in current package,
4139 unless name is "Foo::", in which case Foo is a bearword
4140 (and a package name). */
4143 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
4145 if (ckWARN(WARN_BAREWORD) && ! gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVHV))
4146 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
4147 "Bareword \"%s\" refers to nonexistent package",
4150 PL_tokenbuf[len] = '\0';
4157 gv = gv_fetchpv(PL_tokenbuf, FALSE, SVt_PVCV);
4160 /* if we saw a global override before, get the right name */
4163 sv = newSVpvn("CORE::GLOBAL::",14);
4164 sv_catpv(sv,PL_tokenbuf);
4167 /* If len is 0, newSVpv does strlen(), which is correct.
4168 If len is non-zero, then it will be the true length,
4169 and so the scalar will be created correctly. */
4170 sv = newSVpv(PL_tokenbuf,len);
4173 /* Presume this is going to be a bareword of some sort. */
4176 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4177 yylval.opval->op_private = OPpCONST_BARE;
4178 /* UTF-8 package name? */
4179 if (UTF && !IN_BYTES &&
4180 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
4183 /* And if "Foo::", then that's what it certainly is. */
4188 /* See if it's the indirect object for a list operator. */
4190 if (PL_oldoldbufptr &&
4191 PL_oldoldbufptr < PL_bufptr &&
4192 (PL_oldoldbufptr == PL_last_lop
4193 || PL_oldoldbufptr == PL_last_uni) &&
4194 /* NO SKIPSPACE BEFORE HERE! */
4195 (PL_expect == XREF ||
4196 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
4198 bool immediate_paren = *s == '(';
4200 /* (Now we can afford to cross potential line boundary.) */
4203 /* Two barewords in a row may indicate method call. */
4205 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') && (tmp=intuit_method(s,gv)))
4208 /* If not a declared subroutine, it's an indirect object. */
4209 /* (But it's an indir obj regardless for sort.) */
4211 if ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
4212 ((!gv || !GvCVu(gv)) &&
4213 (PL_last_lop_op != OP_MAPSTART &&
4214 PL_last_lop_op != OP_GREPSTART))))
4216 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
4221 PL_expect = XOPERATOR;
4224 /* Is this a word before a => operator? */
4225 if (*s == '=' && s[1] == '>' && !pkgname) {
4227 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
4228 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
4229 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
4233 /* If followed by a paren, it's certainly a subroutine. */
4236 if (gv && GvCVu(gv)) {
4237 for (d = s + 1; SPACE_OR_TAB(*d); d++) ;
4238 if (*d == ')' && (sv = cv_const_sv(GvCV(gv)))) {
4243 PL_nextval[PL_nexttoke].opval = yylval.opval;
4244 PL_expect = XOPERATOR;
4250 /* If followed by var or block, call it a method (unless sub) */
4252 if ((*s == '$' || *s == '{') && (!gv || !GvCVu(gv))) {
4253 PL_last_lop = PL_oldbufptr;
4254 PL_last_lop_op = OP_METHOD;
4258 /* If followed by a bareword, see if it looks like indir obj. */
4261 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
4262 && (tmp = intuit_method(s,gv)))
4265 /* Not a method, so call it a subroutine (if defined) */
4267 if (gv && GvCVu(gv)) {
4269 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
4270 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4271 "Ambiguous use of -%s resolved as -&%s()",
4272 PL_tokenbuf, PL_tokenbuf);
4273 /* Check for a constant sub */
4275 if ((sv = cv_const_sv(cv))) {
4277 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
4278 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc(sv);
4279 yylval.opval->op_private = 0;
4283 /* Resolve to GV now. */
4284 op_free(yylval.opval);
4285 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
4286 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
4287 PL_last_lop = PL_oldbufptr;
4288 PL_last_lop_op = OP_ENTERSUB;
4289 /* Is there a prototype? */
4292 const char *proto = SvPV_const((SV*)cv, len);
4295 if (*proto == '$' && proto[1] == '\0')
4297 while (*proto == ';')
4299 if (*proto == '&' && *s == '{') {
4300 sv_setpv(PL_subname, PL_curstash ?
4301 "__ANON__" : "__ANON__::__ANON__");
4305 PL_nextval[PL_nexttoke].opval = yylval.opval;
4311 /* Call it a bare word */
4313 if (PL_hints & HINT_STRICT_SUBS)
4314 yylval.opval->op_private |= OPpCONST_STRICT;
4317 if (lastchar != '-') {
4318 if (ckWARN(WARN_RESERVED)) {
4319 for (d = PL_tokenbuf; *d && isLOWER(*d); d++) ;
4320 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
4321 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
4328 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
4329 && ckWARN_d(WARN_AMBIGUOUS)) {
4330 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4331 "Operator or semicolon missing before %c%s",
4332 lastchar, PL_tokenbuf);
4333 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
4334 "Ambiguous use of %c resolved as operator %c",
4335 lastchar, lastchar);
4341 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4342 newSVpv(CopFILE(PL_curcop),0));
4346 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4347 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
4350 case KEY___PACKAGE__:
4351 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
4353 ? newSVhek(HvNAME_HEK(PL_curstash))
4360 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
4361 const char *pname = "main";
4362 if (PL_tokenbuf[2] == 'D')
4363 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
4364 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), TRUE, SVt_PVIO);
4367 GvIOp(gv) = newIO();
4368 IoIFP(GvIOp(gv)) = PL_rsfp;
4369 #if defined(HAS_FCNTL) && defined(F_SETFD)
4371 const int fd = PerlIO_fileno(PL_rsfp);
4372 fcntl(fd,F_SETFD,fd >= 3);
4375 /* Mark this internal pseudo-handle as clean */
4376 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
4378 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
4379 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
4380 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
4382 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
4383 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
4384 /* if the script was opened in binmode, we need to revert
4385 * it to text mode for compatibility; but only iff it has CRs
4386 * XXX this is a questionable hack at best. */
4387 if (PL_bufend-PL_bufptr > 2
4388 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
4391 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
4392 loc = PerlIO_tell(PL_rsfp);
4393 (void)PerlIO_seek(PL_rsfp, 0L, 0);
4396 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
4398 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
4399 #endif /* NETWARE */
4400 #ifdef PERLIO_IS_STDIO /* really? */
4401 # if defined(__BORLANDC__)
4402 /* XXX see note in do_binmode() */
4403 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
4407 PerlIO_seek(PL_rsfp, loc, 0);
4411 #ifdef PERLIO_LAYERS
4414 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
4415 else if (PL_encoding) {
4422 XPUSHs(PL_encoding);
4424 call_method("name", G_SCALAR);
4428 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
4429 Perl_form(aTHX_ ":encoding(%"SVf")",
4447 if (PL_expect == XSTATE) {
4454 if (*s == ':' && s[1] == ':') {
4457 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4458 if (!(tmp = keyword(PL_tokenbuf, len)))
4459 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
4473 LOP(OP_ACCEPT,XTERM);
4479 LOP(OP_ATAN2,XTERM);
4485 LOP(OP_BINMODE,XTERM);
4488 LOP(OP_BLESS,XTERM);
4497 (void)gv_fetchpv("ENV",TRUE, SVt_PVHV); /* may use HOME */
4514 if (!PL_cryptseen) {
4515 PL_cryptseen = TRUE;
4519 LOP(OP_CRYPT,XTERM);
4522 LOP(OP_CHMOD,XTERM);
4525 LOP(OP_CHOWN,XTERM);
4528 LOP(OP_CONNECT,XTERM);
4544 s = force_word(s,WORD,TRUE,TRUE,FALSE);
4548 PL_hints |= HINT_BLOCK_SCOPE;
4558 gv_fetchpv("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
4559 LOP(OP_DBMOPEN,XTERM);
4565 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4572 yylval.ival = CopLINE(PL_curcop);
4586 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
4587 UNIBRACK(OP_ENTEREVAL);
4605 case KEY_endhostent:
4611 case KEY_endservent:
4614 case KEY_endprotoent:
4625 yylval.ival = CopLINE(PL_curcop);
4627 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
4629 if ((PL_bufend - p) >= 3 &&
4630 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
4632 else if ((PL_bufend - p) >= 4 &&
4633 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
4636 if (isIDFIRST_lazy_if(p,UTF)) {
4637 p = scan_ident(p, PL_bufend,
4638 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4642 Perl_croak(aTHX_ "Missing $ on loop variable");
4647 LOP(OP_FORMLINE,XTERM);
4653 LOP(OP_FCNTL,XTERM);
4659 LOP(OP_FLOCK,XTERM);
4668 LOP(OP_GREPSTART, XREF);
4671 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4686 case KEY_getpriority:
4687 LOP(OP_GETPRIORITY,XTERM);
4689 case KEY_getprotobyname:
4692 case KEY_getprotobynumber:
4693 LOP(OP_GPBYNUMBER,XTERM);
4695 case KEY_getprotoent:
4707 case KEY_getpeername:
4708 UNI(OP_GETPEERNAME);
4710 case KEY_gethostbyname:
4713 case KEY_gethostbyaddr:
4714 LOP(OP_GHBYADDR,XTERM);
4716 case KEY_gethostent:
4719 case KEY_getnetbyname:
4722 case KEY_getnetbyaddr:
4723 LOP(OP_GNBYADDR,XTERM);
4728 case KEY_getservbyname:
4729 LOP(OP_GSBYNAME,XTERM);
4731 case KEY_getservbyport:
4732 LOP(OP_GSBYPORT,XTERM);
4734 case KEY_getservent:
4737 case KEY_getsockname:
4738 UNI(OP_GETSOCKNAME);
4740 case KEY_getsockopt:
4741 LOP(OP_GSOCKOPT,XTERM);
4763 yylval.ival = CopLINE(PL_curcop);
4767 LOP(OP_INDEX,XTERM);
4773 LOP(OP_IOCTL,XTERM);
4785 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4817 LOP(OP_LISTEN,XTERM);
4826 s = scan_pat(s,OP_MATCH);
4827 TERM(sublex_start());
4830 LOP(OP_MAPSTART, XREF);
4833 LOP(OP_MKDIR,XTERM);
4836 LOP(OP_MSGCTL,XTERM);
4839 LOP(OP_MSGGET,XTERM);
4842 LOP(OP_MSGRCV,XTERM);
4845 LOP(OP_MSGSND,XTERM);
4851 if (isIDFIRST_lazy_if(s,UTF)) {
4852 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
4853 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
4855 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
4856 if (!PL_in_my_stash) {
4859 sprintf(tmpbuf, "No such class %.1000s", PL_tokenbuf);
4867 s = force_word(s,WORD,TRUE,FALSE,FALSE);
4874 if (PL_expect != XSTATE)
4875 yyerror("\"no\" not allowed in expression");
4876 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4877 s = force_version(s, FALSE);
4882 if (*s == '(' || (s = skipspace(s), *s == '('))
4889 if (isIDFIRST_lazy_if(s,UTF)) {
4891 for (d = s; isALNUM_lazy_if(d,UTF); d++) ;
4892 for (t=d; *t && isSPACE(*t); t++) ;
4893 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
4895 && !(t[0] == '=' && t[1] == '>')
4897 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
4898 "Precedence problem: open %.*s should be open(%.*s)",
4899 d - s, s, d - s, s);
4905 yylval.ival = OP_OR;
4915 LOP(OP_OPEN_DIR,XTERM);
4918 checkcomma(s,PL_tokenbuf,"filehandle");
4922 checkcomma(s,PL_tokenbuf,"filehandle");
4941 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4945 LOP(OP_PIPE_OP,XTERM);
4948 s = scan_str(s,FALSE,FALSE);
4950 missingterm((char*)0);
4951 yylval.ival = OP_CONST;
4952 TERM(sublex_start());
4958 s = scan_str(s,FALSE,FALSE);
4960 missingterm((char*)0);
4961 PL_expect = XOPERATOR;
4963 if (SvCUR(PL_lex_stuff)) {
4966 d = SvPV_force(PL_lex_stuff, len);
4969 for (; isSPACE(*d) && len; --len, ++d) ;
4972 if (!warned && ckWARN(WARN_QW)) {
4973 for (; !isSPACE(*d) && len; --len, ++d) {
4975 Perl_warner(aTHX_ packWARN(WARN_QW),
4976 "Possible attempt to separate words with commas");
4979 else if (*d == '#') {
4980 Perl_warner(aTHX_ packWARN(WARN_QW),
4981 "Possible attempt to put comments in qw() list");
4987 for (; !isSPACE(*d) && len; --len, ++d) ;
4989 sv = newSVpvn(b, d-b);
4990 if (DO_UTF8(PL_lex_stuff))
4992 words = append_elem(OP_LIST, words,
4993 newSVOP(OP_CONST, 0, tokeq(sv)));
4997 PL_nextval[PL_nexttoke].opval = words;
5002 SvREFCNT_dec(PL_lex_stuff);
5003 PL_lex_stuff = Nullsv;
5009 s = scan_str(s,FALSE,FALSE);
5011 missingterm((char*)0);
5012 yylval.ival = OP_STRINGIFY;
5013 if (SvIVX(PL_lex_stuff) == '\'')
5014 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
5015 TERM(sublex_start());
5018 s = scan_pat(s,OP_QR);
5019 TERM(sublex_start());
5022 s = scan_str(s,FALSE,FALSE);
5024 missingterm((char*)0);
5025 yylval.ival = OP_BACKTICK;
5027 TERM(sublex_start());
5035 s = force_version(s, FALSE);
5037 else if (*s != 'v' || !isDIGIT(s[1])
5038 || (s = force_version(s, TRUE), *s == 'v'))
5040 *PL_tokenbuf = '\0';
5041 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5042 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
5043 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
5045 yyerror("<> should be quotes");
5053 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5057 LOP(OP_RENAME,XTERM);
5066 LOP(OP_RINDEX,XTERM);
5076 UNIDOR(OP_READLINE);
5089 LOP(OP_REVERSE,XTERM);
5092 UNIDOR(OP_READLINK);
5100 TERM(sublex_start());
5102 TOKEN(1); /* force error */
5111 LOP(OP_SELECT,XTERM);
5117 LOP(OP_SEMCTL,XTERM);
5120 LOP(OP_SEMGET,XTERM);
5123 LOP(OP_SEMOP,XTERM);
5129 LOP(OP_SETPGRP,XTERM);
5131 case KEY_setpriority:
5132 LOP(OP_SETPRIORITY,XTERM);
5134 case KEY_sethostent:
5140 case KEY_setservent:
5143 case KEY_setprotoent:
5153 LOP(OP_SEEKDIR,XTERM);
5155 case KEY_setsockopt:
5156 LOP(OP_SSOCKOPT,XTERM);
5162 LOP(OP_SHMCTL,XTERM);
5165 LOP(OP_SHMGET,XTERM);
5168 LOP(OP_SHMREAD,XTERM);
5171 LOP(OP_SHMWRITE,XTERM);
5174 LOP(OP_SHUTDOWN,XTERM);
5183 LOP(OP_SOCKET,XTERM);
5185 case KEY_socketpair:
5186 LOP(OP_SOCKPAIR,XTERM);
5189 checkcomma(s,PL_tokenbuf,"subroutine name");
5191 if (*s == ';' || *s == ')') /* probably a close */
5192 Perl_croak(aTHX_ "sort is now a reserved word");
5194 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5198 LOP(OP_SPLIT,XTERM);
5201 LOP(OP_SPRINTF,XTERM);
5204 LOP(OP_SPLICE,XTERM);
5219 LOP(OP_SUBSTR,XTERM);
5225 char tmpbuf[sizeof PL_tokenbuf];
5226 SSize_t tboffset = 0;
5227 expectation attrful;
5228 bool have_name, have_proto, bad_proto;
5229 const int key = tmp;
5233 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
5234 (*s == ':' && s[1] == ':'))
5237 attrful = XATTRBLOCK;
5238 /* remember buffer pos'n for later force_word */
5239 tboffset = s - PL_oldbufptr;
5240 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5241 if (strchr(tmpbuf, ':'))
5242 sv_setpv(PL_subname, tmpbuf);
5244 sv_setsv(PL_subname,PL_curstname);
5245 sv_catpvn(PL_subname,"::",2);
5246 sv_catpvn(PL_subname,tmpbuf,len);
5253 Perl_croak(aTHX_ "Missing name in \"my sub\"");
5254 PL_expect = XTERMBLOCK;
5255 attrful = XATTRTERM;
5256 sv_setpvn(PL_subname,"?",1);
5260 if (key == KEY_format) {
5262 PL_lex_formbrack = PL_lex_brackets + 1;
5264 (void) force_word(PL_oldbufptr + tboffset, WORD,
5269 /* Look for a prototype */
5273 s = scan_str(s,FALSE,FALSE);
5275 Perl_croak(aTHX_ "Prototype not terminated");
5276 /* strip spaces and check for bad characters */
5277 d = SvPVX(PL_lex_stuff);
5280 for (p = d; *p; ++p) {
5283 if (!strchr("$@%*;[]&\\", *p))
5288 if (bad_proto && ckWARN(WARN_SYNTAX))
5289 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5290 "Illegal character in prototype for %"SVf" : %s",
5292 SvCUR_set(PL_lex_stuff, tmp);
5300 if (*s == ':' && s[1] != ':')
5301 PL_expect = attrful;
5302 else if (*s != '{' && key == KEY_sub) {
5304 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
5306 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, PL_subname);
5310 PL_nextval[PL_nexttoke].opval =
5311 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
5312 PL_lex_stuff = Nullsv;
5316 sv_setpv(PL_subname,
5317 PL_curstash ? "__ANON__" : "__ANON__::__ANON__");
5320 (void) force_word(PL_oldbufptr + tboffset, WORD,
5329 LOP(OP_SYSTEM,XREF);
5332 LOP(OP_SYMLINK,XTERM);
5335 LOP(OP_SYSCALL,XTERM);
5338 LOP(OP_SYSOPEN,XTERM);
5341 LOP(OP_SYSSEEK,XTERM);
5344 LOP(OP_SYSREAD,XTERM);
5347 LOP(OP_SYSWRITE,XTERM);
5351 TERM(sublex_start());
5372 LOP(OP_TRUNCATE,XTERM);
5384 yylval.ival = CopLINE(PL_curcop);
5388 yylval.ival = CopLINE(PL_curcop);
5392 LOP(OP_UNLINK,XTERM);
5398 LOP(OP_UNPACK,XTERM);
5401 LOP(OP_UTIME,XTERM);
5407 LOP(OP_UNSHIFT,XTERM);
5410 if (PL_expect != XSTATE)
5411 yyerror("\"use\" not allowed in expression");
5413 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
5414 s = force_version(s, TRUE);
5415 if (*s == ';' || (s = skipspace(s), *s == ';')) {
5416 PL_nextval[PL_nexttoke].opval = Nullop;
5419 else if (*s == 'v') {
5420 s = force_word(s,WORD,FALSE,TRUE,FALSE);
5421 s = force_version(s, FALSE);
5425 s = force_word(s,WORD,FALSE,TRUE,FALSE);
5426 s = force_version(s, FALSE);
5438 yylval.ival = CopLINE(PL_curcop);
5442 PL_hints |= HINT_BLOCK_SCOPE;
5449 LOP(OP_WAITPID,XTERM);
5458 ctl_l[0] = toCTRL('L');
5460 gv_fetchpv(ctl_l,TRUE, SVt_PV);
5463 gv_fetchpv("\f",TRUE, SVt_PV); /* Make sure $^L is defined */
5468 if (PL_expect == XOPERATOR)
5474 yylval.ival = OP_XOR;
5479 TERM(sublex_start());
5484 #pragma segment Main
5488 S_pending_ident(pTHX)
5491 register I32 tmp = 0;
5492 /* pit holds the identifier we read and pending_ident is reset */
5493 char pit = PL_pending_ident;
5494 PL_pending_ident = 0;
5496 DEBUG_T({ PerlIO_printf(Perl_debug_log,
5497 "### Tokener saw identifier '%s'\n", PL_tokenbuf); });
5499 /* if we're in a my(), we can't allow dynamics here.
5500 $foo'bar has already been turned into $foo::bar, so
5501 just check for colons.
5503 if it's a legal name, the OP is a PADANY.
5506 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
5507 if (strchr(PL_tokenbuf,':'))
5508 yyerror(Perl_form(aTHX_ "No package name allowed for "
5509 "variable %s in \"our\"",
5511 tmp = allocmy(PL_tokenbuf);
5514 if (strchr(PL_tokenbuf,':'))
5515 yyerror(Perl_form(aTHX_ PL_no_myglob,PL_tokenbuf));
5517 yylval.opval = newOP(OP_PADANY, 0);
5518 yylval.opval->op_targ = allocmy(PL_tokenbuf);
5524 build the ops for accesses to a my() variable.
5526 Deny my($a) or my($b) in a sort block, *if* $a or $b is
5527 then used in a comparison. This catches most, but not
5528 all cases. For instance, it catches
5529 sort { my($a); $a <=> $b }
5531 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
5532 (although why you'd do that is anyone's guess).
5535 if (!strchr(PL_tokenbuf,':')) {
5537 tmp = pad_findmy(PL_tokenbuf);
5538 if (tmp != NOT_IN_PAD) {
5539 /* might be an "our" variable" */
5540 if (PAD_COMPNAME_FLAGS(tmp) & SVpad_OUR) {
5541 /* build ops for a bareword */
5542 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
5543 HEK * const stashname = HvNAME_HEK(stash);
5544 SV * const sym = newSVhek(stashname);
5545 sv_catpvn(sym, "::", 2);
5546 sv_catpv(sym, PL_tokenbuf+1);
5547 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
5548 yylval.opval->op_private = OPpCONST_ENTERED;
5551 ? (GV_ADDMULTI | GV_ADDINEVAL)
5554 ((PL_tokenbuf[0] == '$') ? SVt_PV
5555 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
5560 /* if it's a sort block and they're naming $a or $b */
5561 if (PL_last_lop_op == OP_SORT &&
5562 PL_tokenbuf[0] == '$' &&
5563 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
5566 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
5567 d < PL_bufend && *d != '\n';
5570 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
5571 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
5577 yylval.opval = newOP(OP_PADANY, 0);
5578 yylval.opval->op_targ = tmp;
5584 Whine if they've said @foo in a doublequoted string,
5585 and @foo isn't a variable we can find in the symbol
5588 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
5589 GV *gv = gv_fetchpv(PL_tokenbuf+1, FALSE, SVt_PVAV);
5590 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
5591 && ckWARN(WARN_AMBIGUOUS))
5593 /* Downgraded from fatal to warning 20000522 mjd */
5594 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5595 "Possible unintended interpolation of %s in string",
5600 /* build ops for a bareword */
5601 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
5602 yylval.opval->op_private = OPpCONST_ENTERED;
5603 gv_fetchpv(PL_tokenbuf+1, PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : TRUE,
5604 ((PL_tokenbuf[0] == '$') ? SVt_PV
5605 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
5611 * The following code was generated by perl_keyword.pl.
5615 Perl_keyword (pTHX_ const char *name, I32 len)
5619 case 1: /* 5 tokens of length 1 */
5651 case 2: /* 18 tokens of length 2 */
5797 case 3: /* 28 tokens of length 3 */
5801 if (name[1] == 'N' &&
5864 if (name[1] == 'i' &&
5904 if (name[1] == 'o' &&
5913 if (name[1] == 'e' &&
5922 if (name[1] == 'n' &&
5931 if (name[1] == 'o' &&
5940 if (name[1] == 'a' &&
5949 if (name[1] == 'o' &&
6011 if (name[1] == 'e' &&
6043 if (name[1] == 'i' &&
6052 if (name[1] == 's' &&
6061 if (name[1] == 'e' &&
6070 if (name[1] == 'o' &&
6082 case 4: /* 40 tokens of length 4 */
6086 if (name[1] == 'O' &&
6096 if (name[1] == 'N' &&
6106 if (name[1] == 'i' &&
6116 if (name[1] == 'h' &&
6126 if (name[1] == 'u' &&
6139 if (name[2] == 'c' &&
6148 if (name[2] == 's' &&
6157 if (name[2] == 'a' &&
6193 if (name[1] == 'o' &&
6206 if (name[2] == 't' &&
6215 if (name[2] == 'o' &&
6224 if (name[2] == 't' &&
6233 if (name[2] == 'e' &&
6246 if (name[1] == 'o' &&
6259 if (name[2] == 'y' &&
6268 if (name[2] == 'l' &&
6284 if (name[2] == 's' &&
6293 if (name[2] == 'n' &&
6302 if (name[2] == 'c' &&
6315 if (name[1] == 'e' &&
6325 if (name[1] == 'p' &&
6338 if (name[2] == 'c' &&
6347 if (name[2] == 'p' &&
6356 if (name[2] == 's' &&
6372 if (name[2] == 'n' &&
6442 if (name[2] == 'r' &&
6451 if (name[2] == 'r' &&
6460 if (name[2] == 'a' &&
6476 if (name[2] == 'l' &&
6543 case 5: /* 36 tokens of length 5 */
6547 if (name[1] == 'E' &&
6558 if (name[1] == 'H' &&
6572 if (name[2] == 'a' &&
6582 if (name[2] == 'a' &&
6596 if (name[1] == 'l' &&
6613 if (name[3] == 'i' &&
6622 if (name[3] == 'o' &&
6658 if (name[2] == 'o' &&
6668 if (name[2] == 'y' &&
6682 if (name[1] == 'l' &&
6696 if (name[2] == 'n' &&
6706 if (name[2] == 'o' &&
6723 if (name[2] == 'd' &&
6733 if (name[2] == 'c' &&
6750 if (name[2] == 'c' &&
6760 if (name[2] == 't' &&
6774 if (name[1] == 'k' &&
6785 if (name[1] == 'r' &&
6799 if (name[2] == 's' &&
6809 if (name[2] == 'd' &&
6826 if (name[2] == 'm' &&
6836 if (name[2] == 'i' &&
6846 if (name[2] == 'e' &&
6856 if (name[2] == 'l' &&
6866 if (name[2] == 'a' &&
6876 if (name[2] == 'u' &&
6890 if (name[1] == 'i' &&
6904 if (name[2] == 'a' &&
6917 if (name[3] == 'e' &&
6952 if (name[2] == 'i' &&
6969 if (name[2] == 'i' &&
6979 if (name[2] == 'i' &&
6996 case 6: /* 33 tokens of length 6 */
7000 if (name[1] == 'c' &&
7015 if (name[2] == 'l' &&
7026 if (name[2] == 'r' &&
7041 if (name[1] == 'e' &&
7056 if (name[2] == 's' &&
7061 if(ckWARN_d(WARN_SYNTAX))
7062 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
7068 if (name[2] == 'i' &&
7086 if (name[2] == 'l' &&
7097 if (name[2] == 'r' &&
7112 if (name[1] == 'm' &&
7127 if (name[2] == 'n' &&
7138 if (name[2] == 's' &&
7153 if (name[1] == 's' &&
7159 if (name[4] == 't' &&
7168 if (name[4] == 'e' &&
7177 if (name[4] == 'c' &&
7186 if (name[4] == 'n' &&
7202 if (name[1] == 'r' &&
7220 if (name[3] == 'a' &&
7230 if (name[3] == 'u' &&
7244 if (name[2] == 'n' &&
7262 if (name[2] == 'a' &&
7276 if (name[3] == 'e' &&
7289 if (name[4] == 't' &&
7298 if (name[4] == 'e' &&
7320 if (name[4] == 't' &&
7329 if (name[4] == 'e' &&
7345 if (name[2] == 'c' &&
7356 if (name[2] == 'l' &&
7367 if (name[2] == 'b' &&
7378 if (name[2] == 's' &&
7401 if (name[4] == 's' &&
7410 if (name[4] == 'n' &&
7423 if (name[3] == 'a' &&
7440 if (name[1] == 'a' &&
7455 case 7: /* 28 tokens of length 7 */
7459 if (name[1] == 'E' &&
7472 if (name[1] == '_' &&
7485 if (name[1] == 'i' &&
7492 return -KEY_binmode;
7498 if (name[1] == 'o' &&
7505 return -KEY_connect;
7514 if (name[2] == 'm' &&
7520 return -KEY_dbmopen;
7526 if (name[2] == 'f' &&
7542 if (name[1] == 'o' &&
7555 if (name[1] == 'e' &&
7562 if (name[5] == 'r' &&
7565 return -KEY_getpgrp;
7571 if (name[5] == 'i' &&
7574 return -KEY_getppid;
7587 if (name[1] == 'c' &&
7594 return -KEY_lcfirst;
7600 if (name[1] == 'p' &&
7607 return -KEY_opendir;
7613 if (name[1] == 'a' &&
7631 if (name[3] == 'd' &&
7636 return -KEY_readdir;
7642 if (name[3] == 'u' &&
7653 if (name[3] == 'e' &&
7658 return -KEY_reverse;
7677 if (name[3] == 'k' &&
7682 return -KEY_seekdir;
7688 if (name[3] == 'p' &&
7693 return -KEY_setpgrp;
7703 if (name[2] == 'm' &&
7709 return -KEY_shmread;
7715 if (name[2] == 'r' &&
7721 return -KEY_sprintf;
7730 if (name[3] == 'l' &&
7735 return -KEY_symlink;
7744 if (name[4] == 'a' &&
7748 return -KEY_syscall;
7754 if (name[4] == 'p' &&
7758 return -KEY_sysopen;
7764 if (name[4] == 'e' &&
7768 return -KEY_sysread;
7774 if (name[4] == 'e' &&
7778 return -KEY_sysseek;
7796 if (name[1] == 'e' &&
7803 return -KEY_telldir;
7812 if (name[2] == 'f' &&
7818 return -KEY_ucfirst;
7824 if (name[2] == 's' &&
7830 return -KEY_unshift;
7840 if (name[1] == 'a' &&
7847 return -KEY_waitpid;
7856 case 8: /* 26 tokens of length 8 */
7860 if (name[1] == 'U' &&
7868 return KEY_AUTOLOAD;
7879 if (name[3] == 'A' &&
7885 return KEY___DATA__;
7891 if (name[3] == 'I' &&
7897 return -KEY___FILE__;
7903 if (name[3] == 'I' &&
7909 return -KEY___LINE__;
7925 if (name[2] == 'o' &&
7932 return -KEY_closedir;
7938 if (name[2] == 'n' &&
7945 return -KEY_continue;
7955 if (name[1] == 'b' &&
7963 return -KEY_dbmclose;
7969 if (name[1] == 'n' &&
7975 if (name[4] == 'r' &&
7980 return -KEY_endgrent;
7986 if (name[4] == 'w' &&
7991 return -KEY_endpwent;
8004 if (name[1] == 'o' &&
8012 return -KEY_formline;
8018 if (name[1] == 'e' &&
8029 if (name[6] == 'n' &&
8032 return -KEY_getgrent;
8038 if (name[6] == 'i' &&
8041 return -KEY_getgrgid;
8047 if (name[6] == 'a' &&
8050 return -KEY_getgrnam;
8063 if (name[4] == 'o' &&
8068 return -KEY_getlogin;
8079 if (name[6] == 'n' &&
8082 return -KEY_getpwent;
8088 if (name[6] == 'a' &&
8091 return -KEY_getpwnam;
8097 if (name[6] == 'i' &&
8100 return -KEY_getpwuid;
8120 if (name[1] == 'e' &&
8127 if (name[5] == 'i' &&
8134 return -KEY_readline;
8139 return -KEY_readlink;
8150 if (name[5] == 'i' &&
8154 return -KEY_readpipe;
8175 if (name[4] == 'r' &&
8180 return -KEY_setgrent;
8186 if (name[4] == 'w' &&
8191 return -KEY_setpwent;
8207 if (name[3] == 'w' &&
8213 return -KEY_shmwrite;
8219 if (name[3] == 't' &&
8225 return -KEY_shutdown;
8235 if (name[2] == 's' &&
8242 return -KEY_syswrite;
8252 if (name[1] == 'r' &&
8260 return -KEY_truncate;
8269 case 9: /* 8 tokens of length 9 */
8273 if (name[1] == 'n' &&
8282 return -KEY_endnetent;
8288 if (name[1] == 'e' &&
8297 return -KEY_getnetent;
8303 if (name[1] == 'o' &&
8312 return -KEY_localtime;
8318 if (name[1] == 'r' &&
8327 return KEY_prototype;
8333 if (name[1] == 'u' &&
8342 return -KEY_quotemeta;
8348 if (name[1] == 'e' &&
8357 return -KEY_rewinddir;
8363 if (name[1] == 'e' &&
8372 return -KEY_setnetent;
8378 if (name[1] == 'a' &&
8387 return -KEY_wantarray;
8396 case 10: /* 9 tokens of length 10 */
8400 if (name[1] == 'n' &&
8406 if (name[4] == 'o' &&
8413 return -KEY_endhostent;
8419 if (name[4] == 'e' &&
8426 return -KEY_endservent;
8439 if (name[1] == 'e' &&
8445 if (name[4] == 'o' &&
8452 return -KEY_gethostent;
8461 if (name[5] == 'r' &&
8467 return -KEY_getservent;
8473 if (name[5] == 'c' &&
8479 return -KEY_getsockopt;
8504 if (name[4] == 'o' &&
8511 return -KEY_sethostent;
8520 if (name[5] == 'r' &&
8526 return -KEY_setservent;
8532 if (name[5] == 'c' &&
8538 return -KEY_setsockopt;
8555 if (name[2] == 'c' &&
8564 return -KEY_socketpair;
8577 case 11: /* 8 tokens of length 11 */
8581 if (name[1] == '_' &&
8592 return -KEY___PACKAGE__;
8598 if (name[1] == 'n' &&
8609 return -KEY_endprotoent;
8615 if (name[1] == 'e' &&
8624 if (name[5] == 'e' &&
8631 return -KEY_getpeername;
8640 if (name[6] == 'o' &&
8646 return -KEY_getpriority;
8652 if (name[6] == 't' &&
8658 return -KEY_getprotoent;
8672 if (name[4] == 'o' &&
8680 return -KEY_getsockname;
8693 if (name[1] == 'e' &&
8701 if (name[6] == 'o' &&
8707 return -KEY_setpriority;
8713 if (name[6] == 't' &&
8719 return -KEY_setprotoent;
8735 case 12: /* 2 tokens of length 12 */
8736 if (name[0] == 'g' &&
8748 if (name[9] == 'd' &&
8751 { /* getnetbyaddr */
8752 return -KEY_getnetbyaddr;
8758 if (name[9] == 'a' &&
8761 { /* getnetbyname */
8762 return -KEY_getnetbyname;
8774 case 13: /* 4 tokens of length 13 */
8775 if (name[0] == 'g' &&
8782 if (name[4] == 'o' &&
8791 if (name[10] == 'd' &&
8794 { /* gethostbyaddr */
8795 return -KEY_gethostbyaddr;
8801 if (name[10] == 'a' &&
8804 { /* gethostbyname */
8805 return -KEY_gethostbyname;
8818 if (name[4] == 'e' &&
8827 if (name[10] == 'a' &&
8830 { /* getservbyname */
8831 return -KEY_getservbyname;
8837 if (name[10] == 'o' &&
8840 { /* getservbyport */
8841 return -KEY_getservbyport;
8860 case 14: /* 1 tokens of length 14 */
8861 if (name[0] == 'g' &&
8875 { /* getprotobyname */
8876 return -KEY_getprotobyname;
8881 case 16: /* 1 tokens of length 16 */
8882 if (name[0] == 'g' &&
8898 { /* getprotobynumber */
8899 return -KEY_getprotobynumber;
8913 S_checkcomma(pTHX_ register char *s, const char *name, const char *what)
8917 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8918 if (ckWARN(WARN_SYNTAX)) {
8920 for (w = s+2; *w && level; w++) {
8927 for (; *w && isSPACE(*w); w++) ;
8928 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
8929 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8930 "%s (...) interpreted as function",name);
8933 while (s < PL_bufend && isSPACE(*s))
8937 while (s < PL_bufend && isSPACE(*s))
8939 if (isIDFIRST_lazy_if(s,UTF)) {
8941 while (isALNUM_lazy_if(s,UTF))
8943 while (s < PL_bufend && isSPACE(*s))
8947 *s = '\0'; /* XXX If we didn't do this, we could const a lot of toke.c */
8948 kw = keyword(w, s - w) || get_cv(w, FALSE) != 0;
8952 Perl_croak(aTHX_ "No comma allowed after %s", what);
8957 /* Either returns sv, or mortalizes sv and returns a new SV*.
8958 Best used as sv=new_constant(..., sv, ...).
8959 If s, pv are NULL, calls subroutine with one argument,
8960 and type is used with error messages only. */
8963 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
8967 HV * const table = GvHV(PL_hintgv); /* ^H */
8971 const char *why1 = "", *why2 = "", *why3 = "";
8973 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8976 why2 = strEQ(key,"charnames")
8977 ? "(possibly a missing \"use charnames ...\")"
8979 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
8980 (type ? type: "undef"), why2);
8982 /* This is convoluted and evil ("goto considered harmful")
8983 * but I do not understand the intricacies of all the different
8984 * failure modes of %^H in here. The goal here is to make
8985 * the most probable error message user-friendly. --jhi */
8990 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8991 (type ? type: "undef"), why1, why2, why3);
8993 yyerror(SvPVX_const(msg));
8997 cvp = hv_fetch(table, key, strlen(key), FALSE);
8998 if (!cvp || !SvOK(*cvp)) {
9001 why3 = "} is not defined";
9004 sv_2mortal(sv); /* Parent created it permanently */
9007 pv = sv_2mortal(newSVpvn(s, len));
9009 typesv = sv_2mortal(newSVpv(type, 0));
9011 typesv = &PL_sv_undef;
9013 PUSHSTACKi(PERLSI_OVERLOAD);
9025 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
9029 /* Check the eval first */
9030 if (!PL_in_eval && SvTRUE(ERRSV)) {
9031 sv_catpv(ERRSV, "Propagated");
9032 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
9034 res = SvREFCNT_inc(sv);
9038 (void)SvREFCNT_inc(res);
9047 why1 = "Call to &{$^H{";
9049 why3 = "}} did not return a defined value";
9057 /* Returns a NUL terminated string, with the length of the string written to
9061 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
9063 register char *d = dest;
9064 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
9067 Perl_croak(aTHX_ ident_too_long);
9068 if (isALNUM(*s)) /* UTF handled below */
9070 else if (*s == '\'' && allow_package && isIDFIRST_lazy_if(s+1,UTF)) {
9075 else if (*s == ':' && s[1] == ':' && allow_package && s[2] != '$') {
9079 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9080 char *t = s + UTF8SKIP(s);
9081 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9083 if (d + (t - s) > e)
9084 Perl_croak(aTHX_ ident_too_long);
9085 Copy(s, d, t - s, char);
9098 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
9102 char *bracket = Nullch;
9108 e = d + destlen - 3; /* two-character token, ending NUL */
9110 while (isDIGIT(*s)) {
9112 Perl_croak(aTHX_ ident_too_long);
9119 Perl_croak(aTHX_ ident_too_long);
9120 if (isALNUM(*s)) /* UTF handled below */
9122 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
9127 else if (*s == ':' && s[1] == ':') {
9131 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9132 char *t = s + UTF8SKIP(s);
9133 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9135 if (d + (t - s) > e)
9136 Perl_croak(aTHX_ ident_too_long);
9137 Copy(s, d, t - s, char);
9148 if (PL_lex_state != LEX_NORMAL)
9149 PL_lex_state = LEX_INTERPENDMAYBE;
9152 if (*s == '$' && s[1] &&
9153 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
9166 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9171 if (isSPACE(s[-1])) {
9173 const char ch = *s++;
9174 if (!SPACE_OR_TAB(ch)) {
9180 if (isIDFIRST_lazy_if(d,UTF)) {
9184 while ((e < send && isALNUM_lazy_if(e,UTF)) || *e == ':') {
9186 while (e < send && UTF8_IS_CONTINUED(*e) && is_utf8_mark((U8*)e))
9189 Copy(s, d, e - s, char);
9194 while ((isALNUM(*s) || *s == ':') && d < e)
9197 Perl_croak(aTHX_ ident_too_long);
9200 while (s < send && SPACE_OR_TAB(*s)) s++;
9201 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9202 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest)) {
9203 const char *brack = *s == '[' ? "[...]" : "{...}";
9204 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9205 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9206 funny, dest, brack, funny, dest, brack);
9209 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9213 /* Handle extended ${^Foo} variables
9214 * 1999-02-27 mjd-perl-patch@plover.com */
9215 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
9219 while (isALNUM(*s) && d < e) {
9223 Perl_croak(aTHX_ ident_too_long);
9228 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9229 PL_lex_state = LEX_INTERPEND;
9234 if (PL_lex_state == LEX_NORMAL) {
9235 if (ckWARN(WARN_AMBIGUOUS) &&
9236 (keyword(dest, d - dest) || get_cv(dest, FALSE)))
9238 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9239 "Ambiguous use of %c{%s} resolved to %c%s",
9240 funny, dest, funny, dest);
9245 s = bracket; /* let the parser handle it */
9249 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9250 PL_lex_state = LEX_INTERPEND;
9255 Perl_pmflag(pTHX_ U32* pmfl, int ch)
9260 *pmfl |= PMf_GLOBAL;
9262 *pmfl |= PMf_CONTINUE;
9266 *pmfl |= PMf_MULTILINE;
9268 *pmfl |= PMf_SINGLELINE;
9270 *pmfl |= PMf_EXTENDED;
9274 S_scan_pat(pTHX_ char *start, I32 type)
9277 char *s = scan_str(start,FALSE,FALSE);
9280 char * const delimiter = skipspace(start);
9281 Perl_croak(aTHX_ *delimiter == '?'
9282 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9283 : "Search pattern not terminated" );
9286 pm = (PMOP*)newPMOP(type, 0);
9287 if (PL_multi_open == '?')
9288 pm->op_pmflags |= PMf_ONCE;
9290 while (*s && strchr("iomsx", *s))
9291 pmflag(&pm->op_pmflags,*s++);
9294 while (*s && strchr("iogcmsx", *s))
9295 pmflag(&pm->op_pmflags,*s++);
9297 /* issue a warning if /c is specified,but /g is not */
9298 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
9299 && ckWARN(WARN_REGEXP))
9301 Perl_warner(aTHX_ packWARN(WARN_REGEXP), c_without_g);
9304 pm->op_pmpermflags = pm->op_pmflags;
9306 PL_lex_op = (OP*)pm;
9307 yylval.ival = OP_MATCH;
9312 S_scan_subst(pTHX_ char *start)
9320 yylval.ival = OP_NULL;
9322 s = scan_str(start,FALSE,FALSE);
9325 Perl_croak(aTHX_ "Substitution pattern not terminated");
9327 if (s[-1] == PL_multi_open)
9330 first_start = PL_multi_start;
9331 s = scan_str(s,FALSE,FALSE);
9334 SvREFCNT_dec(PL_lex_stuff);
9335 PL_lex_stuff = Nullsv;
9337 Perl_croak(aTHX_ "Substitution replacement not terminated");
9339 PL_multi_start = first_start; /* so whole substitution is taken together */
9341 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9347 else if (strchr("iogcmsx", *s))
9348 pmflag(&pm->op_pmflags,*s++);
9353 /* /c is not meaningful with s/// */
9354 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP))
9356 Perl_warner(aTHX_ packWARN(WARN_REGEXP), c_in_subst);
9361 PL_sublex_info.super_bufptr = s;
9362 PL_sublex_info.super_bufend = PL_bufend;
9364 pm->op_pmflags |= PMf_EVAL;
9365 repl = newSVpvn("",0);
9367 sv_catpv(repl, es ? "eval " : "do ");
9368 sv_catpvn(repl, "{ ", 2);
9369 sv_catsv(repl, PL_lex_repl);
9370 sv_catpvn(repl, " };", 2);
9372 SvREFCNT_dec(PL_lex_repl);
9376 pm->op_pmpermflags = pm->op_pmflags;
9377 PL_lex_op = (OP*)pm;
9378 yylval.ival = OP_SUBST;
9383 S_scan_trans(pTHX_ char *start)
9392 yylval.ival = OP_NULL;
9394 s = scan_str(start,FALSE,FALSE);
9396 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9397 if (s[-1] == PL_multi_open)
9400 s = scan_str(s,FALSE,FALSE);
9403 SvREFCNT_dec(PL_lex_stuff);
9404 PL_lex_stuff = Nullsv;
9406 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9409 complement = del = squash = 0;
9413 complement = OPpTRANS_COMPLEMENT;
9416 del = OPpTRANS_DELETE;
9419 squash = OPpTRANS_SQUASH;
9428 Newx(tbl, complement&&!del?258:256, short);
9429 o = newPVOP(OP_TRANS, 0, (char*)tbl);
9430 o->op_private &= ~OPpTRANS_ALL;
9431 o->op_private |= del|squash|complement|
9432 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9433 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9436 yylval.ival = OP_TRANS;
9441 S_scan_heredoc(pTHX_ register char *s)
9444 I32 op_type = OP_SCALAR;
9448 const char newline[] = "\n";
9449 const char *found_newline;
9453 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
9457 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9460 for (peek = s; SPACE_OR_TAB(*peek); peek++) ;
9461 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9464 s = delimcpy(d, e, s, PL_bufend, term, &len);
9474 if (!isALNUM_lazy_if(s,UTF))
9475 deprecate_old("bare << to mean <<\"\"");
9476 for (; isALNUM_lazy_if(s,UTF); s++) {
9481 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9482 Perl_croak(aTHX_ "Delimiter for here document is too long");
9485 len = d - PL_tokenbuf;
9486 #ifndef PERL_STRICT_CR
9487 d = strchr(s, '\r');
9489 char * const olds = s;
9491 while (s < PL_bufend) {
9497 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9506 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9510 if ( outer || !(found_newline = ninstr(s,PL_bufend,newline,newline+1)) ) {
9511 herewas = newSVpvn(s,PL_bufend-s);
9515 herewas = newSVpvn(s,found_newline-s);
9517 s += SvCUR(herewas);
9519 tmpstr = NEWSV(87,79);
9520 sv_upgrade(tmpstr, SVt_PVIV);
9523 SvIV_set(tmpstr, -1);
9525 else if (term == '`') {
9526 op_type = OP_BACKTICK;
9527 SvIV_set(tmpstr, '\\');
9531 PL_multi_start = CopLINE(PL_curcop);
9532 PL_multi_open = PL_multi_close = '<';
9533 term = *PL_tokenbuf;
9534 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
9535 char *bufptr = PL_sublex_info.super_bufptr;
9536 char *bufend = PL_sublex_info.super_bufend;
9537 char * const olds = s - SvCUR(herewas);
9538 s = strchr(bufptr, '\n');
9542 while (s < bufend &&
9543 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9545 CopLINE_inc(PL_curcop);
9548 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9549 missingterm(PL_tokenbuf);
9551 sv_setpvn(herewas,bufptr,d-bufptr+1);
9552 sv_setpvn(tmpstr,d+1,s-d);
9554 sv_catpvn(herewas,s,bufend-s);
9555 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9562 while (s < PL_bufend &&
9563 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9565 CopLINE_inc(PL_curcop);
9567 if (s >= PL_bufend) {
9568 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9569 missingterm(PL_tokenbuf);
9571 sv_setpvn(tmpstr,d+1,s-d);
9573 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9575 sv_catpvn(herewas,s,PL_bufend-s);
9576 sv_setsv(PL_linestr,herewas);
9577 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9578 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9579 PL_last_lop = PL_last_uni = Nullch;
9582 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
9583 while (s >= PL_bufend) { /* multiple line string? */
9585 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
9586 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9587 missingterm(PL_tokenbuf);
9589 CopLINE_inc(PL_curcop);
9590 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9591 PL_last_lop = PL_last_uni = Nullch;
9592 #ifndef PERL_STRICT_CR
9593 if (PL_bufend - PL_linestart >= 2) {
9594 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9595 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9597 PL_bufend[-2] = '\n';
9599 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9601 else if (PL_bufend[-1] == '\r')
9602 PL_bufend[-1] = '\n';
9604 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9605 PL_bufend[-1] = '\n';
9607 if (PERLDB_LINE && PL_curstash != PL_debstash) {
9608 SV *sv = NEWSV(88,0);
9610 sv_upgrade(sv, SVt_PVMG);
9611 sv_setsv(sv,PL_linestr);
9614 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop),sv);
9616 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9617 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9618 *(SvPVX(PL_linestr) + off ) = ' ';
9619 sv_catsv(PL_linestr,herewas);
9620 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9621 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9625 sv_catsv(tmpstr,PL_linestr);
9630 PL_multi_end = CopLINE(PL_curcop);
9631 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9632 SvPV_shrink_to_cur(tmpstr);
9634 SvREFCNT_dec(herewas);
9636 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9638 else if (PL_encoding)
9639 sv_recode_to_utf8(tmpstr, PL_encoding);
9641 PL_lex_stuff = tmpstr;
9642 yylval.ival = op_type;
9647 takes: current position in input buffer
9648 returns: new position in input buffer
9649 side-effects: yylval and lex_op are set.
9654 <FH> read from filehandle
9655 <pkg::FH> read from package qualified filehandle
9656 <pkg'FH> read from package qualified filehandle
9657 <$fh> read from filehandle in $fh
9663 S_scan_inputsymbol(pTHX_ char *start)
9665 register char *s = start; /* current position in buffer */
9671 d = PL_tokenbuf; /* start of temp holding space */
9672 e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9673 end = strchr(s, '\n');
9676 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9678 /* die if we didn't have space for the contents of the <>,
9679 or if it didn't end, or if we see a newline
9682 if (len >= sizeof PL_tokenbuf)
9683 Perl_croak(aTHX_ "Excessively long <> operator");
9685 Perl_croak(aTHX_ "Unterminated <> operator");
9690 Remember, only scalar variables are interpreted as filehandles by
9691 this code. Anything more complex (e.g., <$fh{$num}>) will be
9692 treated as a glob() call.
9693 This code makes use of the fact that except for the $ at the front,
9694 a scalar variable and a filehandle look the same.
9696 if (*d == '$' && d[1]) d++;
9698 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9699 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9702 /* If we've tried to read what we allow filehandles to look like, and
9703 there's still text left, then it must be a glob() and not a getline.
9704 Use scan_str to pull out the stuff between the <> and treat it
9705 as nothing more than a string.
9708 if (d - PL_tokenbuf != len) {
9709 yylval.ival = OP_GLOB;
9711 s = scan_str(start,FALSE,FALSE);
9713 Perl_croak(aTHX_ "Glob not terminated");
9717 bool readline_overriden = FALSE;
9718 GV *gv_readline = Nullgv;
9720 /* we're in a filehandle read situation */
9723 /* turn <> into <ARGV> */
9725 Copy("ARGV",d,5,char);
9727 /* Check whether readline() is overriden */
9728 if (((gv_readline = gv_fetchpv("readline", FALSE, SVt_PVCV))
9729 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9731 ((gvp = (GV**)hv_fetch(PL_globalstash, "readline", 8, FALSE))
9732 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
9733 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9734 readline_overriden = TRUE;
9736 /* if <$fh>, create the ops to turn the variable into a
9742 /* try to find it in the pad for this block, otherwise find
9743 add symbol table ops
9745 if ((tmp = pad_findmy(d)) != NOT_IN_PAD) {
9746 if (PAD_COMPNAME_FLAGS(tmp) & SVpad_OUR) {
9747 HV *stash = PAD_COMPNAME_OURSTASH(tmp);
9748 HEK *stashname = HvNAME_HEK(stash);
9749 SV *sym = sv_2mortal(newSVhek(stashname));
9750 sv_catpvn(sym, "::", 2);
9756 OP *o = newOP(OP_PADSV, 0);
9758 PL_lex_op = readline_overriden
9759 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9760 append_elem(OP_LIST, o,
9761 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9762 : (OP*)newUNOP(OP_READLINE, 0, o);
9771 ? (GV_ADDMULTI | GV_ADDINEVAL)
9774 PL_lex_op = readline_overriden
9775 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9776 append_elem(OP_LIST,
9777 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9778 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9779 : (OP*)newUNOP(OP_READLINE, 0,
9780 newUNOP(OP_RV2SV, 0,
9781 newGVOP(OP_GV, 0, gv)));
9783 if (!readline_overriden)
9784 PL_lex_op->op_flags |= OPf_SPECIAL;
9785 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
9786 yylval.ival = OP_NULL;
9789 /* If it's none of the above, it must be a literal filehandle
9790 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9792 GV *gv = gv_fetchpv(d,TRUE, SVt_PVIO);
9793 PL_lex_op = readline_overriden
9794 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9795 append_elem(OP_LIST,
9796 newGVOP(OP_GV, 0, gv),
9797 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9798 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9799 yylval.ival = OP_NULL;
9808 takes: start position in buffer
9809 keep_quoted preserve \ on the embedded delimiter(s)
9810 keep_delims preserve the delimiters around the string
9811 returns: position to continue reading from buffer
9812 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9813 updates the read buffer.
9815 This subroutine pulls a string out of the input. It is called for:
9816 q single quotes q(literal text)
9817 ' single quotes 'literal text'
9818 qq double quotes qq(interpolate $here please)
9819 " double quotes "interpolate $here please"
9820 qx backticks qx(/bin/ls -l)
9821 ` backticks `/bin/ls -l`
9822 qw quote words @EXPORT_OK = qw( func() $spam )
9823 m// regexp match m/this/
9824 s/// regexp substitute s/this/that/
9825 tr/// string transliterate tr/this/that/
9826 y/// string transliterate y/this/that/
9827 ($*@) sub prototypes sub foo ($)
9828 (stuff) sub attr parameters sub foo : attr(stuff)
9829 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9831 In most of these cases (all but <>, patterns and transliterate)
9832 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9833 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9834 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9837 It skips whitespace before the string starts, and treats the first
9838 character as the delimiter. If the delimiter is one of ([{< then
9839 the corresponding "close" character )]}> is used as the closing
9840 delimiter. It allows quoting of delimiters, and if the string has
9841 balanced delimiters ([{<>}]) it allows nesting.
9843 On success, the SV with the resulting string is put into lex_stuff or,
9844 if that is already non-NULL, into lex_repl. The second case occurs only
9845 when parsing the RHS of the special constructs s/// and tr/// (y///).
9846 For convenience, the terminating delimiter character is stuffed into
9851 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9853 SV *sv; /* scalar value: string */
9854 char *tmps; /* temp string, used for delimiter matching */
9855 register char *s = start; /* current position in the buffer */
9856 register char term; /* terminating character */
9857 register char *to; /* current position in the sv's data */
9858 I32 brackets = 1; /* bracket nesting level */
9859 bool has_utf8 = FALSE; /* is there any utf8 content? */
9860 I32 termcode; /* terminating char. code */
9861 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9862 STRLEN termlen; /* length of terminating string */
9863 char *last = NULL; /* last position for nesting bracket */
9865 /* skip space before the delimiter */
9869 /* mark where we are, in case we need to report errors */
9872 /* after skipping whitespace, the next character is the terminator */
9875 termcode = termstr[0] = term;
9879 termcode = utf8_to_uvchr((U8*)s, &termlen);
9880 Copy(s, termstr, termlen, U8);
9881 if (!UTF8_IS_INVARIANT(term))
9885 /* mark where we are */
9886 PL_multi_start = CopLINE(PL_curcop);
9887 PL_multi_open = term;
9889 /* find corresponding closing delimiter */
9890 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9891 termcode = termstr[0] = term = tmps[5];
9893 PL_multi_close = term;
9895 /* create a new SV to hold the contents. 87 is leak category, I'm
9896 assuming. 79 is the SV's initial length. What a random number. */
9898 sv_upgrade(sv, SVt_PVIV);
9899 SvIV_set(sv, termcode);
9900 (void)SvPOK_only(sv); /* validate pointer */
9902 /* move past delimiter and try to read a complete string */
9904 sv_catpvn(sv, s, termlen);
9907 if (PL_encoding && !UTF) {
9911 int offset = s - SvPVX_const(PL_linestr);
9912 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9913 &offset, (char*)termstr, termlen);
9914 const char *ns = SvPVX_const(PL_linestr) + offset;
9915 char *svlast = SvEND(sv) - 1;
9917 for (; s < ns; s++) {
9918 if (*s == '\n' && !PL_rsfp)
9919 CopLINE_inc(PL_curcop);
9922 goto read_more_line;
9924 /* handle quoted delimiters */
9925 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9927 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9929 if ((svlast-1 - t) % 2) {
9933 SvCUR_set(sv, SvCUR(sv) - 1);
9938 if (PL_multi_open == PL_multi_close) {
9946 for (t = w = last; t < svlast; w++, t++) {
9947 /* At here, all closes are "was quoted" one,
9948 so we don't check PL_multi_close. */
9950 if (!keep_quoted && *(t+1) == PL_multi_open)
9955 else if (*t == PL_multi_open)
9963 SvCUR_set(sv, w - SvPVX_const(sv));
9966 if (--brackets <= 0)
9972 SvCUR_set(sv, SvCUR(sv) - 1);
9978 /* extend sv if need be */
9979 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9980 /* set 'to' to the next character in the sv's string */
9981 to = SvPVX(sv)+SvCUR(sv);
9983 /* if open delimiter is the close delimiter read unbridle */
9984 if (PL_multi_open == PL_multi_close) {
9985 for (; s < PL_bufend; s++,to++) {
9986 /* embedded newlines increment the current line number */
9987 if (*s == '\n' && !PL_rsfp)
9988 CopLINE_inc(PL_curcop);
9989 /* handle quoted delimiters */
9990 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9991 if (!keep_quoted && s[1] == term)
9993 /* any other quotes are simply copied straight through */
9997 /* terminate when run out of buffer (the for() condition), or
9998 have found the terminator */
9999 else if (*s == term) {
10002 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10005 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10011 /* if the terminator isn't the same as the start character (e.g.,
10012 matched brackets), we have to allow more in the quoting, and
10013 be prepared for nested brackets.
10016 /* read until we run out of string, or we find the terminator */
10017 for (; s < PL_bufend; s++,to++) {
10018 /* embedded newlines increment the line count */
10019 if (*s == '\n' && !PL_rsfp)
10020 CopLINE_inc(PL_curcop);
10021 /* backslashes can escape the open or closing characters */
10022 if (*s == '\\' && s+1 < PL_bufend) {
10023 if (!keep_quoted &&
10024 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10029 /* allow nested opens and closes */
10030 else if (*s == PL_multi_close && --brackets <= 0)
10032 else if (*s == PL_multi_open)
10034 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10039 /* terminate the copied string and update the sv's end-of-string */
10041 SvCUR_set(sv, to - SvPVX_const(sv));
10044 * this next chunk reads more into the buffer if we're not done yet
10048 break; /* handle case where we are done yet :-) */
10050 #ifndef PERL_STRICT_CR
10051 if (to - SvPVX_const(sv) >= 2) {
10052 if ((to[-2] == '\r' && to[-1] == '\n') ||
10053 (to[-2] == '\n' && to[-1] == '\r'))
10057 SvCUR_set(sv, to - SvPVX_const(sv));
10059 else if (to[-1] == '\r')
10062 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10067 /* if we're out of file, or a read fails, bail and reset the current
10068 line marker so we can report where the unterminated string began
10071 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
10073 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10076 /* we read a line, so increment our line counter */
10077 CopLINE_inc(PL_curcop);
10079 /* update debugger info */
10080 if (PERLDB_LINE && PL_curstash != PL_debstash) {
10081 SV *sv = NEWSV(88,0);
10083 sv_upgrade(sv, SVt_PVMG);
10084 sv_setsv(sv,PL_linestr);
10085 (void)SvIOK_on(sv);
10087 av_store(CopFILEAV(PL_curcop), (I32)CopLINE(PL_curcop), sv);
10090 /* having changed the buffer, we must update PL_bufend */
10091 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10092 PL_last_lop = PL_last_uni = Nullch;
10095 /* at this point, we have successfully read the delimited string */
10097 if (!PL_encoding || UTF) {
10099 sv_catpvn(sv, s, termlen);
10102 if (has_utf8 || PL_encoding)
10105 PL_multi_end = CopLINE(PL_curcop);
10107 /* if we allocated too much space, give some back */
10108 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10109 SvLEN_set(sv, SvCUR(sv) + 1);
10110 SvPV_renew(sv, SvLEN(sv));
10113 /* decide whether this is the first or second quoted string we've read
10126 takes: pointer to position in buffer
10127 returns: pointer to new position in buffer
10128 side-effects: builds ops for the constant in yylval.op
10130 Read a number in any of the formats that Perl accepts:
10132 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10133 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10136 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10138 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10141 If it reads a number without a decimal point or an exponent, it will
10142 try converting the number to an integer and see if it can do so
10143 without loss of precision.
10147 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10149 register const char *s = start; /* current position in buffer */
10150 register char *d; /* destination in temp buffer */
10151 register char *e; /* end of temp buffer */
10152 NV nv; /* number read, as a double */
10153 SV *sv = Nullsv; /* place to put the converted number */
10154 bool floatit; /* boolean: int or float? */
10155 const char *lastub = 0; /* position of last underbar */
10156 static char const number_too_long[] = "Number too long";
10158 /* We use the first character to decide what type of number this is */
10162 Perl_croak(aTHX_ "panic: scan_num");
10164 /* if it starts with a 0, it could be an octal number, a decimal in
10165 0.13 disguise, or a hexadecimal number, or a binary number. */
10169 u holds the "number so far"
10170 shift the power of 2 of the base
10171 (hex == 4, octal == 3, binary == 1)
10172 overflowed was the number more than we can hold?
10174 Shift is used when we add a digit. It also serves as an "are
10175 we in octal/hex/binary?" indicator to disallow hex characters
10176 when in octal mode.
10181 bool overflowed = FALSE;
10182 bool just_zero = TRUE; /* just plain 0 or binary number? */
10183 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10184 static const char* const bases[5] =
10185 { "", "binary", "", "octal", "hexadecimal" };
10186 static const char* const Bases[5] =
10187 { "", "Binary", "", "Octal", "Hexadecimal" };
10188 static const char* const maxima[5] =
10190 "0b11111111111111111111111111111111",
10194 const char *base, *Base, *max;
10196 /* check for hex */
10201 } else if (s[1] == 'b') {
10206 /* check for a decimal in disguise */
10207 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10209 /* so it must be octal */
10216 if (ckWARN(WARN_SYNTAX))
10217 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10218 "Misplaced _ in number");
10222 base = bases[shift];
10223 Base = Bases[shift];
10224 max = maxima[shift];
10226 /* read the rest of the number */
10228 /* x is used in the overflow test,
10229 b is the digit we're adding on. */
10234 /* if we don't mention it, we're done */
10238 /* _ are ignored -- but warned about if consecutive */
10240 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
10241 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10242 "Misplaced _ in number");
10246 /* 8 and 9 are not octal */
10247 case '8': case '9':
10249 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10253 case '2': case '3': case '4':
10254 case '5': case '6': case '7':
10256 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10259 case '0': case '1':
10260 b = *s++ & 15; /* ASCII digit -> value of digit */
10264 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10265 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10266 /* make sure they said 0x */
10269 b = (*s++ & 7) + 9;
10271 /* Prepare to put the digit we have onto the end
10272 of the number so far. We check for overflows.
10278 x = u << shift; /* make room for the digit */
10280 if ((x >> shift) != u
10281 && !(PL_hints & HINT_NEW_BINARY)) {
10284 if (ckWARN_d(WARN_OVERFLOW))
10285 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
10286 "Integer overflow in %s number",
10289 u = x | b; /* add the digit to the end */
10292 n *= nvshift[shift];
10293 /* If an NV has not enough bits in its
10294 * mantissa to represent an UV this summing of
10295 * small low-order numbers is a waste of time
10296 * (because the NV cannot preserve the
10297 * low-order bits anyway): we could just
10298 * remember when did we overflow and in the
10299 * end just multiply n by the right
10307 /* if we get here, we had success: make a scalar value from
10312 /* final misplaced underbar check */
10313 if (s[-1] == '_') {
10314 if (ckWARN(WARN_SYNTAX))
10315 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10320 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
10321 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
10322 "%s number > %s non-portable",
10328 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
10329 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
10330 "%s number > %s non-portable",
10335 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10336 sv = new_constant(start, s - start, "integer",
10338 else if (PL_hints & HINT_NEW_BINARY)
10339 sv = new_constant(start, s - start, "binary", sv, Nullsv, NULL);
10344 handle decimal numbers.
10345 we're also sent here when we read a 0 as the first digit
10347 case '1': case '2': case '3': case '4': case '5':
10348 case '6': case '7': case '8': case '9': case '.':
10351 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10354 /* read next group of digits and _ and copy into d */
10355 while (isDIGIT(*s) || *s == '_') {
10356 /* skip underscores, checking for misplaced ones
10360 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
10361 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10362 "Misplaced _ in number");
10366 /* check for end of fixed-length buffer */
10368 Perl_croak(aTHX_ number_too_long);
10369 /* if we're ok, copy the character */
10374 /* final misplaced underbar check */
10375 if (lastub && s == lastub + 1) {
10376 if (ckWARN(WARN_SYNTAX))
10377 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10380 /* read a decimal portion if there is one. avoid
10381 3..5 being interpreted as the number 3. followed
10384 if (*s == '.' && s[1] != '.') {
10389 if (ckWARN(WARN_SYNTAX))
10390 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10391 "Misplaced _ in number");
10395 /* copy, ignoring underbars, until we run out of digits.
10397 for (; isDIGIT(*s) || *s == '_'; s++) {
10398 /* fixed length buffer check */
10400 Perl_croak(aTHX_ number_too_long);
10402 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
10403 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10404 "Misplaced _ in number");
10410 /* fractional part ending in underbar? */
10411 if (s[-1] == '_') {
10412 if (ckWARN(WARN_SYNTAX))
10413 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10414 "Misplaced _ in number");
10416 if (*s == '.' && isDIGIT(s[1])) {
10417 /* oops, it's really a v-string, but without the "v" */
10423 /* read exponent part, if present */
10424 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10428 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10429 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10431 /* stray preinitial _ */
10433 if (ckWARN(WARN_SYNTAX))
10434 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10435 "Misplaced _ in number");
10439 /* allow positive or negative exponent */
10440 if (*s == '+' || *s == '-')
10443 /* stray initial _ */
10445 if (ckWARN(WARN_SYNTAX))
10446 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10447 "Misplaced _ in number");
10451 /* read digits of exponent */
10452 while (isDIGIT(*s) || *s == '_') {
10455 Perl_croak(aTHX_ number_too_long);
10459 if (((lastub && s == lastub + 1) ||
10460 (!isDIGIT(s[1]) && s[1] != '_'))
10461 && ckWARN(WARN_SYNTAX))
10462 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10463 "Misplaced _ in number");
10470 /* make an sv from the string */
10474 We try to do an integer conversion first if no characters
10475 indicating "float" have been found.
10480 int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10482 if (flags == IS_NUMBER_IN_UV) {
10484 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
10487 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10488 if (uv <= (UV) IV_MIN)
10489 sv_setiv(sv, -(IV)uv);
10496 /* terminate the string */
10498 nv = Atof(PL_tokenbuf);
10502 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
10503 (PL_hints & HINT_NEW_INTEGER) )
10504 sv = new_constant(PL_tokenbuf, d - PL_tokenbuf,
10505 (floatit ? "float" : "integer"),
10509 /* if it starts with a v, it could be a v-string */
10512 sv = NEWSV(92,5); /* preallocate storage space */
10513 s = scan_vstring(s,sv);
10517 /* make the op for the constant and return */
10520 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10522 lvalp->opval = Nullop;
10528 S_scan_formline(pTHX_ register char *s)
10530 register char *eol;
10532 SV *stuff = newSVpvn("",0);
10533 bool needargs = FALSE;
10534 bool eofmt = FALSE;
10536 while (!needargs) {
10538 #ifdef PERL_STRICT_CR
10539 for (t = s+1;SPACE_OR_TAB(*t); t++) ;
10541 for (t = s+1;SPACE_OR_TAB(*t) || *t == '\r'; t++) ;
10543 if (*t == '\n' || t == PL_bufend) {
10548 if (PL_in_eval && !PL_rsfp) {
10549 eol = (char *) memchr(s,'\n',PL_bufend-s);
10554 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10556 for (t = s; t < eol; t++) {
10557 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10559 goto enough; /* ~~ must be first line in formline */
10561 if (*t == '@' || *t == '^')
10565 sv_catpvn(stuff, s, eol-s);
10566 #ifndef PERL_STRICT_CR
10567 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10568 char *end = SvPVX(stuff) + SvCUR(stuff);
10571 SvCUR_set(stuff, SvCUR(stuff) - 1);
10580 s = filter_gets(PL_linestr, PL_rsfp, 0);
10581 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
10582 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
10583 PL_last_lop = PL_last_uni = Nullch;
10592 if (SvCUR(stuff)) {
10595 PL_lex_state = LEX_NORMAL;
10596 PL_nextval[PL_nexttoke].ival = 0;
10600 PL_lex_state = LEX_FORMLINE;
10602 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10604 else if (PL_encoding)
10605 sv_recode_to_utf8(stuff, PL_encoding);
10607 PL_nextval[PL_nexttoke].opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10609 PL_nextval[PL_nexttoke].ival = OP_FORMLINE;
10613 SvREFCNT_dec(stuff);
10615 PL_lex_formbrack = 0;
10626 PL_cshlen = strlen(PL_cshname);
10631 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10633 const I32 oldsavestack_ix = PL_savestack_ix;
10634 CV* outsidecv = PL_compcv;
10637 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10639 SAVEI32(PL_subline);
10640 save_item(PL_subname);
10641 SAVESPTR(PL_compcv);
10643 PL_compcv = (CV*)NEWSV(1104,0);
10644 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
10645 CvFLAGS(PL_compcv) |= flags;
10647 PL_subline = CopLINE(PL_curcop);
10648 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10649 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc(outsidecv);
10650 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10652 return oldsavestack_ix;
10656 #pragma segment Perl_yylex
10659 Perl_yywarn(pTHX_ const char *s)
10661 PL_in_eval |= EVAL_WARNONLY;
10663 PL_in_eval &= ~EVAL_WARNONLY;
10668 Perl_yyerror(pTHX_ const char *s)
10670 const char *where = NULL;
10671 const char *context = NULL;
10675 if (!yychar || (yychar == ';' && !PL_rsfp))
10677 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10678 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10679 PL_oldbufptr != PL_bufptr) {
10682 The code below is removed for NetWare because it abends/crashes on NetWare
10683 when the script has error such as not having the closing quotes like:
10684 if ($var eq "value)
10685 Checking of white spaces is anyway done in NetWare code.
10688 while (isSPACE(*PL_oldoldbufptr))
10691 context = PL_oldoldbufptr;
10692 contlen = PL_bufptr - PL_oldoldbufptr;
10694 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10695 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10698 The code below is removed for NetWare because it abends/crashes on NetWare
10699 when the script has error such as not having the closing quotes like:
10700 if ($var eq "value)
10701 Checking of white spaces is anyway done in NetWare code.
10704 while (isSPACE(*PL_oldbufptr))
10707 context = PL_oldbufptr;
10708 contlen = PL_bufptr - PL_oldbufptr;
10710 else if (yychar > 255)
10711 where = "next token ???";
10712 else if (yychar == -2) { /* YYEMPTY */
10713 if (PL_lex_state == LEX_NORMAL ||
10714 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10715 where = "at end of line";
10716 else if (PL_lex_inpat)
10717 where = "within pattern";
10719 where = "within string";
10722 SV *where_sv = sv_2mortal(newSVpvn("next char ", 10));
10724 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10725 else if (isPRINT_LC(yychar))
10726 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
10728 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10729 where = SvPVX_const(where_sv);
10731 msg = sv_2mortal(newSVpv(s, 0));
10732 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10733 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10735 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10737 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10738 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10739 Perl_sv_catpvf(aTHX_ msg,
10740 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10741 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10744 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
10745 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, msg);
10748 if (PL_error_count >= 10) {
10749 if (PL_in_eval && SvCUR(ERRSV))
10750 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10751 ERRSV, OutCopFILE(PL_curcop));
10753 Perl_croak(aTHX_ "%s has too many errors.\n",
10754 OutCopFILE(PL_curcop));
10757 PL_in_my_stash = Nullhv;
10761 #pragma segment Main
10765 S_swallow_bom(pTHX_ U8 *s)
10767 const STRLEN slen = SvCUR(PL_linestr);
10770 if (s[1] == 0xFE) {
10771 /* UTF-16 little-endian? (or UTF32-LE?) */
10772 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10773 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
10774 #ifndef PERL_NO_UTF16_FILTER
10775 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
10778 if (PL_bufend > (char*)s) {
10782 filter_add(utf16rev_textfilter, NULL);
10783 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
10784 utf16_to_utf8_reversed(s, news,
10785 PL_bufend - (char*)s - 1,
10787 sv_setpvn(PL_linestr, (const char*)news, newlen);
10789 SvUTF8_on(PL_linestr);
10790 s = (U8*)SvPVX(PL_linestr);
10791 PL_bufend = SvPVX(PL_linestr) + newlen;
10794 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
10799 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10800 #ifndef PERL_NO_UTF16_FILTER
10801 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10804 if (PL_bufend > (char *)s) {
10808 filter_add(utf16_textfilter, NULL);
10809 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
10810 utf16_to_utf8(s, news,
10811 PL_bufend - (char*)s,
10813 sv_setpvn(PL_linestr, (const char*)news, newlen);
10815 SvUTF8_on(PL_linestr);
10816 s = (U8*)SvPVX(PL_linestr);
10817 PL_bufend = SvPVX(PL_linestr) + newlen;
10820 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
10825 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10826 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10827 s += 3; /* UTF-8 */
10833 if (s[2] == 0xFE && s[3] == 0xFF) {
10834 /* UTF-32 big-endian */
10835 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
10838 else if (s[2] == 0 && s[3] != 0) {
10841 * are a good indicator of UTF-16BE. */
10842 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10847 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10850 * are a good indicator of UTF-16LE. */
10851 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10860 * Restore a source filter.
10864 restore_rsfp(pTHX_ void *f)
10866 PerlIO *fp = (PerlIO*)f;
10868 if (PL_rsfp == PerlIO_stdin())
10869 PerlIO_clearerr(PL_rsfp);
10870 else if (PL_rsfp && (PL_rsfp != fp))
10871 PerlIO_close(PL_rsfp);
10875 #ifndef PERL_NO_UTF16_FILTER
10877 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10879 const STRLEN old = SvCUR(sv);
10880 const I32 count = FILTER_READ(idx+1, sv, maxlen);
10881 DEBUG_P(PerlIO_printf(Perl_debug_log,
10882 "utf16_textfilter(%p): %d %d (%d)\n",
10883 utf16_textfilter, idx, maxlen, (int) count));
10887 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
10888 Copy(SvPVX_const(sv), tmps, old, char);
10889 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
10890 SvCUR(sv) - old, &newlen);
10891 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
10893 DEBUG_P({sv_dump(sv);});
10898 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10900 const STRLEN old = SvCUR(sv);
10901 const I32 count = FILTER_READ(idx+1, sv, maxlen);
10902 DEBUG_P(PerlIO_printf(Perl_debug_log,
10903 "utf16rev_textfilter(%p): %d %d (%d)\n",
10904 utf16rev_textfilter, idx, maxlen, (int) count));
10908 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
10909 Copy(SvPVX_const(sv), tmps, old, char);
10910 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
10911 SvCUR(sv) - old, &newlen);
10912 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
10914 DEBUG_P({ sv_dump(sv); });
10920 Returns a pointer to the next character after the parsed
10921 vstring, as well as updating the passed in sv.
10923 Function must be called like
10926 s = scan_vstring(s,sv);
10928 The sv should already be large enough to store the vstring
10929 passed in, for performance reasons.
10934 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
10936 const char *pos = s;
10937 const char *start = s;
10938 if (*pos == 'v') pos++; /* get past 'v' */
10939 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
10941 if ( *pos != '.') {
10942 /* this may not be a v-string if followed by => */
10943 const char *next = pos;
10944 while (next < PL_bufend && isSPACE(*next))
10946 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
10947 /* return string not v-string */
10948 sv_setpvn(sv,(char *)s,pos-s);
10949 return (char *)pos;
10953 if (!isALPHA(*pos)) {
10955 U8 tmpbuf[UTF8_MAXBYTES+1];
10958 if (*s == 'v') s++; /* get past 'v' */
10960 sv_setpvn(sv, "", 0);
10965 /* this is atoi() that tolerates underscores */
10966 const char *end = pos;
10968 while (--end >= s) {
10973 rev += (*end - '0') * mult;
10975 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
10976 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
10977 "Integer overflow in decimal number");
10981 if (rev > 0x7FFFFFFF)
10982 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10984 /* Append native character for the rev point */
10985 tmpend = uvchr_to_utf8(tmpbuf, rev);
10986 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10987 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
10989 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
10995 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
10999 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11007 * c-indentation-style: bsd
11008 * c-basic-offset: 4
11009 * indent-tabs-mode: t
11012 * ex: set ts=8 sts=4 sw=4 noet: