3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 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[] = "Identifier too long";
30 static const char commaless_variable_list[] = "comma-less variable list";
32 static void restore_rsfp(pTHX_ void *f);
33 #ifndef PERL_NO_UTF16_FILTER
34 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
35 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
40 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
42 # define CURMAD(slot,sv)
43 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
46 #define XFAKEBRACK 128
49 #ifdef USE_UTF8_SCRIPTS
50 # define UTF (!IN_BYTES)
52 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
55 /* In variables named $^X, these are the legal values for X.
56 * 1999-02-27 mjd-perl-patch@plover.com */
57 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
59 /* On MacOS, respect nonbreaking spaces */
60 #ifdef MACOS_TRADITIONAL
61 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
63 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
66 /* LEX_* are values for PL_lex_state, the state of the lexer.
67 * They are arranged oddly so that the guard on the switch statement
68 * can get by with a single comparison (if the compiler is smart enough).
71 /* #define LEX_NOTPARSING 11 is done in perl.h. */
73 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
74 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
75 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
76 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
77 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
79 /* at end of code, eg "$x" followed by: */
80 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
81 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
83 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
84 string or after \E, $foo, etc */
85 #define LEX_INTERPCONST 2 /* NOT USED */
86 #define LEX_FORMLINE 1 /* expecting a format line */
87 #define LEX_KNOWNEXT 0 /* next token known; just return it */
91 static const char* const lex_state_names[] = {
110 #include "keywords.h"
112 /* CLINE is a macro that ensures PL_copline has a sane value */
117 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
120 # define SKIPSPACE0(s) skipspace0(s)
121 # define SKIPSPACE1(s) skipspace1(s)
122 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
123 # define PEEKSPACE(s) skipspace2(s,0)
125 # define SKIPSPACE0(s) skipspace(s)
126 # define SKIPSPACE1(s) skipspace(s)
127 # define SKIPSPACE2(s,tsv) skipspace(s)
128 # define PEEKSPACE(s) skipspace(s)
132 * Convenience functions to return different tokens and prime the
133 * lexer for the next token. They all take an argument.
135 * TOKEN : generic token (used for '(', DOLSHARP, etc)
136 * OPERATOR : generic operator
137 * AOPERATOR : assignment operator
138 * PREBLOCK : beginning the block after an if, while, foreach, ...
139 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
140 * PREREF : *EXPR where EXPR is not a simple identifier
141 * TERM : expression term
142 * LOOPX : loop exiting command (goto, last, dump, etc)
143 * FTST : file test operator
144 * FUN0 : zero-argument function
145 * FUN1 : not used, except for not, which isn't a UNIOP
146 * BOop : bitwise or or xor
148 * SHop : shift operator
149 * PWop : power operator
150 * PMop : pattern-matching operator
151 * Aop : addition-level operator
152 * Mop : multiplication-level operator
153 * Eop : equality-testing operator
154 * Rop : relational operator <= != gt
156 * Also see LOP and lop() below.
159 #ifdef DEBUGGING /* Serve -DT. */
160 # define REPORT(retval) tokereport((I32)retval)
162 # define REPORT(retval) (retval)
165 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
166 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
167 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
168 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
170 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
171 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
172 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
173 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
174 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
175 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
176 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
177 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
178 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
179 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
180 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
181 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
182 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
183 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
184 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
186 /* This bit of chicanery makes a unary function followed by
187 * a parenthesis into a function with one argument, highest precedence.
188 * The UNIDOR macro is for unary functions that can be followed by the //
189 * operator (such as C<shift // 0>).
191 #define UNI2(f,x) { \
195 PL_last_uni = PL_oldbufptr; \
196 PL_last_lop_op = f; \
198 return REPORT( (int)FUNC1 ); \
200 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
202 #define UNI(f) UNI2(f,XTERM)
203 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
205 #define UNIBRACK(f) { \
208 PL_last_uni = PL_oldbufptr; \
210 return REPORT( (int)FUNC1 ); \
212 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
215 /* grandfather return to old style */
216 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
220 /* how to interpret the yylval associated with the token */
224 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
230 static struct debug_tokens {
232 enum token_type type;
234 } const debug_tokens[] =
236 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
237 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
238 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
239 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
240 { ARROW, TOKENTYPE_NONE, "ARROW" },
241 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
242 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
243 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
244 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
245 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
246 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
247 { DO, TOKENTYPE_NONE, "DO" },
248 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
249 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
250 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
251 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
252 { ELSE, TOKENTYPE_NONE, "ELSE" },
253 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
254 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
255 { FOR, TOKENTYPE_IVAL, "FOR" },
256 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
257 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
258 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
259 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
260 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
261 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
262 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
263 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
264 { IF, TOKENTYPE_IVAL, "IF" },
265 { LABEL, TOKENTYPE_PVAL, "LABEL" },
266 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
267 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
268 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
269 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
270 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
271 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
272 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
273 { MY, TOKENTYPE_IVAL, "MY" },
274 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
275 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
276 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
277 { OROP, TOKENTYPE_IVAL, "OROP" },
278 { OROR, TOKENTYPE_NONE, "OROR" },
279 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
280 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
281 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
282 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
283 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
284 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
285 { PREINC, TOKENTYPE_NONE, "PREINC" },
286 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
287 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
288 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
289 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
290 { SUB, TOKENTYPE_NONE, "SUB" },
291 { THING, TOKENTYPE_OPVAL, "THING" },
292 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
293 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
294 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
295 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
296 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
297 { USE, TOKENTYPE_IVAL, "USE" },
298 { WHEN, TOKENTYPE_IVAL, "WHEN" },
299 { WHILE, TOKENTYPE_IVAL, "WHILE" },
300 { WORD, TOKENTYPE_OPVAL, "WORD" },
301 { 0, TOKENTYPE_NONE, NULL }
304 /* dump the returned token in rv, plus any optional arg in yylval */
307 S_tokereport(pTHX_ I32 rv)
311 const char *name = NULL;
312 enum token_type type = TOKENTYPE_NONE;
313 const struct debug_tokens *p;
314 SV* const report = newSVpvs("<== ");
316 for (p = debug_tokens; p->token; p++) {
317 if (p->token == (int)rv) {
324 Perl_sv_catpv(aTHX_ report, name);
325 else if ((char)rv > ' ' && (char)rv < '~')
326 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
328 sv_catpvs(report, "EOF");
330 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
333 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
336 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
338 case TOKENTYPE_OPNUM:
339 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
340 PL_op_name[yylval.ival]);
343 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
345 case TOKENTYPE_OPVAL:
347 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
348 PL_op_name[yylval.opval->op_type]);
349 if (yylval.opval->op_type == OP_CONST) {
350 Perl_sv_catpvf(aTHX_ report, " %s",
351 SvPEEK(cSVOPx_sv(yylval.opval)));
356 sv_catpvs(report, "(opval=null)");
359 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
365 /* print the buffer with suitable escapes */
368 S_printbuf(pTHX_ const char* fmt, const char* s)
370 SV* const tmp = newSVpvs("");
371 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
380 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
381 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
385 S_ao(pTHX_ int toketype)
388 if (*PL_bufptr == '=') {
390 if (toketype == ANDAND)
391 yylval.ival = OP_ANDASSIGN;
392 else if (toketype == OROR)
393 yylval.ival = OP_ORASSIGN;
394 else if (toketype == DORDOR)
395 yylval.ival = OP_DORASSIGN;
403 * When Perl expects an operator and finds something else, no_op
404 * prints the warning. It always prints "<something> found where
405 * operator expected. It prints "Missing semicolon on previous line?"
406 * if the surprise occurs at the start of the line. "do you need to
407 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
408 * where the compiler doesn't know if foo is a method call or a function.
409 * It prints "Missing operator before end of line" if there's nothing
410 * after the missing operator, or "... before <...>" if there is something
411 * after the missing operator.
415 S_no_op(pTHX_ const char *what, char *s)
418 char * const oldbp = PL_bufptr;
419 const bool is_first = (PL_oldbufptr == PL_linestart);
425 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
426 if (ckWARN_d(WARN_SYNTAX)) {
428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
429 "\t(Missing semicolon on previous line?)\n");
430 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
432 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
434 if (t < PL_bufptr && isSPACE(*t))
435 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
436 "\t(Do you need to predeclare %.*s?)\n",
437 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
441 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
442 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
450 * Complain about missing quote/regexp/heredoc terminator.
451 * If it's called with NULL then it cauterizes the line buffer.
452 * If we're in a delimited string and the delimiter is a control
453 * character, it's reformatted into a two-char sequence like ^C.
458 S_missingterm(pTHX_ char *s)
464 char * const nl = strrchr(s,'\n');
470 iscntrl(PL_multi_close)
472 PL_multi_close < 32 || PL_multi_close == 127
476 tmpbuf[1] = (char)toCTRL(PL_multi_close);
481 *tmpbuf = (char)PL_multi_close;
485 q = strchr(s,'"') ? '\'' : '"';
486 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
489 #define FEATURE_IS_ENABLED(name) \
490 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
491 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
493 * S_feature_is_enabled
494 * Check whether the named feature is enabled.
497 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
500 HV * const hinthv = GvHV(PL_hintgv);
501 char he_name[32] = "feature_";
502 (void) my_strlcpy(&he_name[8], name, 24);
504 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
512 Perl_deprecate(pTHX_ const char *s)
514 if (ckWARN(WARN_DEPRECATED))
515 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
519 Perl_deprecate_old(pTHX_ const char *s)
521 /* This function should NOT be called for any new deprecated warnings */
522 /* Use Perl_deprecate instead */
524 /* It is here to maintain backward compatibility with the pre-5.8 */
525 /* warnings category hierarchy. The "deprecated" category used to */
526 /* live under the "syntax" category. It is now a top-level category */
527 /* in its own right. */
529 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
530 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
531 "Use of %s is deprecated", s);
535 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
536 * utf16-to-utf8-reversed.
539 #ifdef PERL_CR_FILTER
543 register const char *s = SvPVX_const(sv);
544 register const char * const e = s + SvCUR(sv);
545 /* outer loop optimized to do nothing if there are no CR-LFs */
547 if (*s++ == '\r' && *s == '\n') {
548 /* hit a CR-LF, need to copy the rest */
549 register char *d = s - 1;
552 if (*s == '\r' && s[1] == '\n')
563 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
565 const I32 count = FILTER_READ(idx+1, sv, maxlen);
566 if (count > 0 && !maxlen)
574 * Initialize variables. Uses the Perl save_stack to save its state (for
575 * recursive calls to the parser).
579 Perl_lex_start(pTHX_ SV *line)
585 SAVEI32(PL_lex_dojoin);
586 SAVEI32(PL_lex_brackets);
587 SAVEI32(PL_lex_casemods);
588 SAVEI32(PL_lex_starts);
589 SAVEI32(PL_lex_state);
590 SAVEVPTR(PL_lex_inpat);
591 SAVEI32(PL_lex_inwhat);
593 if (PL_lex_state == LEX_KNOWNEXT) {
594 I32 toke = PL_lasttoke;
595 while (--toke >= 0) {
596 SAVEI32(PL_nexttoke[toke].next_type);
597 SAVEVPTR(PL_nexttoke[toke].next_val);
599 SAVEVPTR(PL_nexttoke[toke].next_mad);
601 SAVEI32(PL_lasttoke);
604 SAVESPTR(PL_thistoken);
605 SAVESPTR(PL_thiswhite);
606 SAVESPTR(PL_nextwhite);
607 SAVESPTR(PL_thisopen);
608 SAVESPTR(PL_thisclose);
609 SAVESPTR(PL_thisstuff);
610 SAVEVPTR(PL_thismad);
611 SAVEI32(PL_realtokenstart);
612 SAVEI32(PL_faketokens);
614 SAVEI32(PL_curforce);
616 if (PL_lex_state == LEX_KNOWNEXT) {
617 I32 toke = PL_nexttoke;
618 while (--toke >= 0) {
619 SAVEI32(PL_nexttype[toke]);
620 SAVEVPTR(PL_nextval[toke]);
622 SAVEI32(PL_nexttoke);
625 SAVECOPLINE(PL_curcop);
628 SAVEPPTR(PL_oldbufptr);
629 SAVEPPTR(PL_oldoldbufptr);
630 SAVEPPTR(PL_last_lop);
631 SAVEPPTR(PL_last_uni);
632 SAVEPPTR(PL_linestart);
633 SAVESPTR(PL_linestr);
634 SAVEGENERICPV(PL_lex_brackstack);
635 SAVEGENERICPV(PL_lex_casestack);
636 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
637 SAVESPTR(PL_lex_stuff);
638 SAVEI32(PL_lex_defer);
639 SAVEI32(PL_sublex_info.sub_inwhat);
640 SAVESPTR(PL_lex_repl);
642 SAVEINT(PL_lex_expect);
644 PL_lex_state = LEX_NORMAL;
648 Newx(PL_lex_brackstack, 120, char);
649 Newx(PL_lex_casestack, 12, char);
651 *PL_lex_casestack = '\0';
663 PL_sublex_info.sub_inwhat = 0;
665 s = SvPV_const(PL_linestr, len);
666 if (SvREADONLY(PL_linestr) || !len || s[len-1] != ';') {
667 PL_linestr = sv_2mortal(len ? newSVsv(PL_linestr) : newSVpvn(s, 0));
668 if (!len || s[len-1] != ';')
669 sv_catpvs(PL_linestr, "\n;");
671 SvTEMP_off(PL_linestr);
672 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
673 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
674 PL_last_lop = PL_last_uni = NULL;
680 * Finalizer for lexing operations. Must be called when the parser is
681 * done with the lexer.
688 PL_doextract = FALSE;
693 * This subroutine has nothing to do with tilting, whether at windmills
694 * or pinball tables. Its name is short for "increment line". It
695 * increments the current line number in CopLINE(PL_curcop) and checks
696 * to see whether the line starts with a comment of the form
697 * # line 500 "foo.pm"
698 * If so, it sets the current line number and file to the values in the comment.
702 S_incline(pTHX_ char *s)
710 CopLINE_inc(PL_curcop);
713 while (SPACE_OR_TAB(*s))
715 if (strnEQ(s, "line", 4))
719 if (SPACE_OR_TAB(*s))
723 while (SPACE_OR_TAB(*s))
731 while (SPACE_OR_TAB(*s))
733 if (*s == '"' && (t = strchr(s+1, '"'))) {
743 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
745 if (*e != '\n' && *e != '\0')
746 return; /* false alarm */
752 const char * const cf = CopFILE(PL_curcop);
753 STRLEN tmplen = cf ? strlen(cf) : 0;
754 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
755 /* must copy *{"::_<(eval N)[oldfilename:L]"}
756 * to *{"::_<newfilename"} */
757 char smallbuf[256], smallbuf2[256];
758 char *tmpbuf, *tmpbuf2;
760 STRLEN tmplen2 = strlen(s);
761 if (tmplen + 3 < sizeof smallbuf)
764 Newx(tmpbuf, tmplen + 3, char);
765 if (tmplen2 + 3 < sizeof smallbuf2)
768 Newx(tmpbuf2, tmplen2 + 3, char);
769 tmpbuf[0] = tmpbuf2[0] = '_';
770 tmpbuf[1] = tmpbuf2[1] = '<';
771 memcpy(tmpbuf + 2, cf, ++tmplen);
772 memcpy(tmpbuf2 + 2, s, ++tmplen2);
774 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
776 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
778 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
779 /* adjust ${"::_<newfilename"} to store the new file name */
780 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
781 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
782 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
785 if (tmpbuf != smallbuf) Safefree(tmpbuf);
786 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
789 CopFILE_free(PL_curcop);
790 CopFILE_set(PL_curcop, s);
793 CopLINE_set(PL_curcop, atoi(n)-1);
797 /* skip space before PL_thistoken */
800 S_skipspace0(pTHX_ register char *s)
807 PL_thiswhite = newSVpvs("");
808 sv_catsv(PL_thiswhite, PL_skipwhite);
809 sv_free(PL_skipwhite);
812 PL_realtokenstart = s - SvPVX(PL_linestr);
816 /* skip space after PL_thistoken */
819 S_skipspace1(pTHX_ register char *s)
821 const char *start = s;
822 I32 startoff = start - SvPVX(PL_linestr);
827 start = SvPVX(PL_linestr) + startoff;
828 if (!PL_thistoken && PL_realtokenstart >= 0) {
829 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
830 PL_thistoken = newSVpvn(tstart, start - tstart);
832 PL_realtokenstart = -1;
835 PL_nextwhite = newSVpvs("");
836 sv_catsv(PL_nextwhite, PL_skipwhite);
837 sv_free(PL_skipwhite);
844 S_skipspace2(pTHX_ register char *s, SV **svp)
847 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
848 const I32 startoff = s - SvPVX(PL_linestr);
851 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
852 if (!PL_madskills || !svp)
854 start = SvPVX(PL_linestr) + startoff;
855 if (!PL_thistoken && PL_realtokenstart >= 0) {
856 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
857 PL_thistoken = newSVpvn(tstart, start - tstart);
858 PL_realtokenstart = -1;
863 sv_setsv(*svp, PL_skipwhite);
864 sv_free(PL_skipwhite);
874 * Called to gobble the appropriate amount and type of whitespace.
875 * Skips comments as well.
879 S_skipspace(pTHX_ register char *s)
884 int startoff = s - SvPVX(PL_linestr);
887 sv_free(PL_skipwhite);
892 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
893 while (s < PL_bufend && SPACE_OR_TAB(*s))
903 SSize_t oldprevlen, oldoldprevlen;
904 SSize_t oldloplen = 0, oldunilen = 0;
905 while (s < PL_bufend && isSPACE(*s)) {
906 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
911 if (s < PL_bufend && *s == '#') {
912 while (s < PL_bufend && *s != '\n')
916 if (PL_in_eval && !PL_rsfp) {
923 /* only continue to recharge the buffer if we're at the end
924 * of the buffer, we're not reading from a source filter, and
925 * we're in normal lexing mode
927 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
928 PL_lex_state == LEX_FORMLINE)
935 /* try to recharge the buffer */
937 curoff = s - SvPVX(PL_linestr);
940 if ((s = filter_gets(PL_linestr, PL_rsfp,
941 (prevlen = SvCUR(PL_linestr)))) == NULL)
944 if (PL_madskills && curoff != startoff) {
946 PL_skipwhite = newSVpvs("");
947 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
951 /* mustn't throw out old stuff yet if madpropping */
952 SvCUR(PL_linestr) = curoff;
953 s = SvPVX(PL_linestr) + curoff;
955 if (curoff && s[-1] == '\n')
959 /* end of file. Add on the -p or -n magic */
960 /* XXX these shouldn't really be added here, can't set PL_faketokens */
964 ";}continue{print or die qq(-p destination: $!\\n);}");
967 ";}continue{print or die qq(-p destination: $!\\n);}");
969 PL_minus_n = PL_minus_p = 0;
971 else if (PL_minus_n) {
973 sv_catpvn(PL_linestr, ";}", 2);
975 sv_setpvn(PL_linestr, ";}", 2);
981 sv_catpvn(PL_linestr,";", 1);
983 sv_setpvn(PL_linestr,";", 1);
986 /* reset variables for next time we lex */
987 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
993 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
994 PL_last_lop = PL_last_uni = NULL;
996 /* Close the filehandle. Could be from -P preprocessor,
997 * STDIN, or a regular file. If we were reading code from
998 * STDIN (because the commandline held no -e or filename)
999 * then we don't close it, we reset it so the code can
1000 * read from STDIN too.
1003 if (PL_preprocess && !PL_in_eval)
1004 (void)PerlProc_pclose(PL_rsfp);
1005 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1006 PerlIO_clearerr(PL_rsfp);
1008 (void)PerlIO_close(PL_rsfp);
1013 /* not at end of file, so we only read another line */
1014 /* make corresponding updates to old pointers, for yyerror() */
1015 oldprevlen = PL_oldbufptr - PL_bufend;
1016 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1018 oldunilen = PL_last_uni - PL_bufend;
1020 oldloplen = PL_last_lop - PL_bufend;
1021 PL_linestart = PL_bufptr = s + prevlen;
1022 PL_bufend = s + SvCUR(PL_linestr);
1024 PL_oldbufptr = s + oldprevlen;
1025 PL_oldoldbufptr = s + oldoldprevlen;
1027 PL_last_uni = s + oldunilen;
1029 PL_last_lop = s + oldloplen;
1032 /* debugger active and we're not compiling the debugger code,
1033 * so store the line into the debugger's array of lines
1035 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1036 SV * const sv = newSV(0);
1038 sv_upgrade(sv, SVt_PVMG);
1039 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1042 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1050 PL_skipwhite = newSVpvs("");
1051 curoff = s - SvPVX(PL_linestr);
1052 if (curoff - startoff)
1053 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1062 * Check the unary operators to ensure there's no ambiguity in how they're
1063 * used. An ambiguous piece of code would be:
1065 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1066 * the +5 is its argument.
1076 if (PL_oldoldbufptr != PL_last_uni)
1078 while (isSPACE(*PL_last_uni))
1081 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1083 if ((t = strchr(s, '(')) && t < PL_bufptr)
1086 if (ckWARN_d(WARN_AMBIGUOUS)){
1087 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1088 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1089 (int)(s - PL_last_uni), PL_last_uni);
1094 * LOP : macro to build a list operator. Its behaviour has been replaced
1095 * with a subroutine, S_lop() for which LOP is just another name.
1098 #define LOP(f,x) return lop(f,x,s)
1102 * Build a list operator (or something that might be one). The rules:
1103 * - if we have a next token, then it's a list operator [why?]
1104 * - if the next thing is an opening paren, then it's a function
1105 * - else it's a list operator
1109 S_lop(pTHX_ I32 f, int x, char *s)
1116 PL_last_lop = PL_oldbufptr;
1117 PL_last_lop_op = (OPCODE)f;
1120 return REPORT(LSTOP);
1123 return REPORT(LSTOP);
1126 return REPORT(FUNC);
1129 return REPORT(FUNC);
1131 return REPORT(LSTOP);
1137 * Sets up for an eventual force_next(). start_force(0) basically does
1138 * an unshift, while start_force(-1) does a push. yylex removes items
1143 S_start_force(pTHX_ int where)
1147 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1148 where = PL_lasttoke;
1149 assert(PL_curforce < 0 || PL_curforce == where);
1150 if (PL_curforce != where) {
1151 for (i = PL_lasttoke; i > where; --i) {
1152 PL_nexttoke[i] = PL_nexttoke[i-1];
1156 if (PL_curforce < 0) /* in case of duplicate start_force() */
1157 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1158 PL_curforce = where;
1161 curmad('^', newSVpvs(""));
1162 CURMAD('_', PL_nextwhite);
1167 S_curmad(pTHX_ char slot, SV *sv)
1173 if (PL_curforce < 0)
1174 where = &PL_thismad;
1176 where = &PL_nexttoke[PL_curforce].next_mad;
1179 sv_setpvn(sv, "", 0);
1182 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1184 else if (PL_encoding) {
1185 sv_recode_to_utf8(sv, PL_encoding);
1190 /* keep a slot open for the head of the list? */
1191 if (slot != '_' && *where && (*where)->mad_key == '^') {
1192 (*where)->mad_key = slot;
1193 sv_free((*where)->mad_val);
1194 (*where)->mad_val = (void*)sv;
1197 addmad(newMADsv(slot, sv), where, 0);
1200 # define start_force(where) NOOP
1201 # define curmad(slot, sv) NOOP
1206 * When the lexer realizes it knows the next token (for instance,
1207 * it is reordering tokens for the parser) then it can call S_force_next
1208 * to know what token to return the next time the lexer is called. Caller
1209 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1210 * and possibly PL_expect to ensure the lexer handles the token correctly.
1214 S_force_next(pTHX_ I32 type)
1218 if (PL_curforce < 0)
1219 start_force(PL_lasttoke);
1220 PL_nexttoke[PL_curforce].next_type = type;
1221 if (PL_lex_state != LEX_KNOWNEXT)
1222 PL_lex_defer = PL_lex_state;
1223 PL_lex_state = LEX_KNOWNEXT;
1224 PL_lex_expect = PL_expect;
1227 PL_nexttype[PL_nexttoke] = type;
1229 if (PL_lex_state != LEX_KNOWNEXT) {
1230 PL_lex_defer = PL_lex_state;
1231 PL_lex_expect = PL_expect;
1232 PL_lex_state = LEX_KNOWNEXT;
1238 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1241 SV * const sv = newSVpvn(start,len);
1242 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1249 * When the lexer knows the next thing is a word (for instance, it has
1250 * just seen -> and it knows that the next char is a word char, then
1251 * it calls S_force_word to stick the next word into the PL_next lookahead.
1254 * char *start : buffer position (must be within PL_linestr)
1255 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1256 * int check_keyword : if true, Perl checks to make sure the word isn't
1257 * a keyword (do this if the word is a label, e.g. goto FOO)
1258 * int allow_pack : if true, : characters will also be allowed (require,
1259 * use, etc. do this)
1260 * int allow_initial_tick : used by the "sub" lexer only.
1264 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1270 start = SKIPSPACE1(start);
1272 if (isIDFIRST_lazy_if(s,UTF) ||
1273 (allow_pack && *s == ':') ||
1274 (allow_initial_tick && *s == '\'') )
1276 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1277 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1279 start_force(PL_curforce);
1281 curmad('X', newSVpvn(start,s-start));
1282 if (token == METHOD) {
1287 PL_expect = XOPERATOR;
1290 NEXTVAL_NEXTTOKE.opval
1291 = (OP*)newSVOP(OP_CONST,0,
1292 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1293 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1301 * Called when the lexer wants $foo *foo &foo etc, but the program
1302 * text only contains the "foo" portion. The first argument is a pointer
1303 * to the "foo", and the second argument is the type symbol to prefix.
1304 * Forces the next token to be a "WORD".
1305 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1309 S_force_ident(pTHX_ register const char *s, int kind)
1313 const STRLEN len = strlen(s);
1314 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1315 start_force(PL_curforce);
1316 NEXTVAL_NEXTTOKE.opval = o;
1319 o->op_private = OPpCONST_ENTERED;
1320 /* XXX see note in pp_entereval() for why we forgo typo
1321 warnings if the symbol must be introduced in an eval.
1323 gv_fetchpvn_flags(s, len,
1324 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1326 kind == '$' ? SVt_PV :
1327 kind == '@' ? SVt_PVAV :
1328 kind == '%' ? SVt_PVHV :
1336 Perl_str_to_version(pTHX_ SV *sv)
1341 const char *start = SvPV_const(sv,len);
1342 const char * const end = start + len;
1343 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1344 while (start < end) {
1348 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1353 retval += ((NV)n)/nshift;
1362 * Forces the next token to be a version number.
1363 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1364 * and if "guessing" is TRUE, then no new token is created (and the caller
1365 * must use an alternative parsing method).
1369 S_force_version(pTHX_ char *s, int guessing)
1375 I32 startoff = s - SvPVX(PL_linestr);
1384 while (isDIGIT(*d) || *d == '_' || *d == '.')
1388 start_force(PL_curforce);
1389 curmad('X', newSVpvn(s,d-s));
1392 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1394 s = scan_num(s, &yylval);
1395 version = yylval.opval;
1396 ver = cSVOPx(version)->op_sv;
1397 if (SvPOK(ver) && !SvNIOK(ver)) {
1398 SvUPGRADE(ver, SVt_PVNV);
1399 SvNV_set(ver, str_to_version(ver));
1400 SvNOK_on(ver); /* hint that it is a version */
1403 else if (guessing) {
1406 sv_free(PL_nextwhite); /* let next token collect whitespace */
1408 s = SvPVX(PL_linestr) + startoff;
1416 if (PL_madskills && !version) {
1417 sv_free(PL_nextwhite); /* let next token collect whitespace */
1419 s = SvPVX(PL_linestr) + startoff;
1422 /* NOTE: The parser sees the package name and the VERSION swapped */
1423 start_force(PL_curforce);
1424 NEXTVAL_NEXTTOKE.opval = version;
1432 * Tokenize a quoted string passed in as an SV. It finds the next
1433 * chunk, up to end of string or a backslash. It may make a new
1434 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1439 S_tokeq(pTHX_ SV *sv)
1443 register char *send;
1451 s = SvPV_force(sv, len);
1452 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1455 while (s < send && *s != '\\')
1460 if ( PL_hints & HINT_NEW_STRING ) {
1461 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1467 if (s + 1 < send && (s[1] == '\\'))
1468 s++; /* all that, just for this */
1473 SvCUR_set(sv, d - SvPVX_const(sv));
1475 if ( PL_hints & HINT_NEW_STRING )
1476 return new_constant(NULL, 0, "q", sv, pv, "q");
1481 * Now come three functions related to double-quote context,
1482 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1483 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1484 * interact with PL_lex_state, and create fake ( ... ) argument lists
1485 * to handle functions and concatenation.
1486 * They assume that whoever calls them will be setting up a fake
1487 * join call, because each subthing puts a ',' after it. This lets
1490 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1492 * (I'm not sure whether the spurious commas at the end of lcfirst's
1493 * arguments and join's arguments are created or not).
1498 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1500 * Pattern matching will set PL_lex_op to the pattern-matching op to
1501 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1503 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1505 * Everything else becomes a FUNC.
1507 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1508 * had an OP_CONST or OP_READLINE). This just sets us up for a
1509 * call to S_sublex_push().
1513 S_sublex_start(pTHX)
1516 register const I32 op_type = yylval.ival;
1518 if (op_type == OP_NULL) {
1519 yylval.opval = PL_lex_op;
1523 if (op_type == OP_CONST || op_type == OP_READLINE) {
1524 SV *sv = tokeq(PL_lex_stuff);
1526 if (SvTYPE(sv) == SVt_PVIV) {
1527 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1529 const char * const p = SvPV_const(sv, len);
1530 SV * const nsv = newSVpvn(p, len);
1536 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1537 PL_lex_stuff = NULL;
1538 /* Allow <FH> // "foo" */
1539 if (op_type == OP_READLINE)
1540 PL_expect = XTERMORDORDOR;
1543 else if (op_type == OP_BACKTICK && PL_lex_op) {
1544 /* readpipe() vas overriden */
1545 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1546 yylval.opval = PL_lex_op;
1548 PL_lex_stuff = NULL;
1552 PL_sublex_info.super_state = PL_lex_state;
1553 PL_sublex_info.sub_inwhat = op_type;
1554 PL_sublex_info.sub_op = PL_lex_op;
1555 PL_lex_state = LEX_INTERPPUSH;
1559 yylval.opval = PL_lex_op;
1569 * Create a new scope to save the lexing state. The scope will be
1570 * ended in S_sublex_done. Returns a '(', starting the function arguments
1571 * to the uc, lc, etc. found before.
1572 * Sets PL_lex_state to LEX_INTERPCONCAT.
1581 PL_lex_state = PL_sublex_info.super_state;
1582 SAVEI32(PL_lex_dojoin);
1583 SAVEI32(PL_lex_brackets);
1584 SAVEI32(PL_lex_casemods);
1585 SAVEI32(PL_lex_starts);
1586 SAVEI32(PL_lex_state);
1587 SAVEVPTR(PL_lex_inpat);
1588 SAVEI32(PL_lex_inwhat);
1589 SAVECOPLINE(PL_curcop);
1590 SAVEPPTR(PL_bufptr);
1591 SAVEPPTR(PL_bufend);
1592 SAVEPPTR(PL_oldbufptr);
1593 SAVEPPTR(PL_oldoldbufptr);
1594 SAVEPPTR(PL_last_lop);
1595 SAVEPPTR(PL_last_uni);
1596 SAVEPPTR(PL_linestart);
1597 SAVESPTR(PL_linestr);
1598 SAVEGENERICPV(PL_lex_brackstack);
1599 SAVEGENERICPV(PL_lex_casestack);
1601 PL_linestr = PL_lex_stuff;
1602 PL_lex_stuff = NULL;
1604 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1605 = SvPVX(PL_linestr);
1606 PL_bufend += SvCUR(PL_linestr);
1607 PL_last_lop = PL_last_uni = NULL;
1608 SAVEFREESV(PL_linestr);
1610 PL_lex_dojoin = FALSE;
1611 PL_lex_brackets = 0;
1612 Newx(PL_lex_brackstack, 120, char);
1613 Newx(PL_lex_casestack, 12, char);
1614 PL_lex_casemods = 0;
1615 *PL_lex_casestack = '\0';
1617 PL_lex_state = LEX_INTERPCONCAT;
1618 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1620 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1621 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1622 PL_lex_inpat = PL_sublex_info.sub_op;
1624 PL_lex_inpat = NULL;
1631 * Restores lexer state after a S_sublex_push.
1638 if (!PL_lex_starts++) {
1639 SV * const sv = newSVpvs("");
1640 if (SvUTF8(PL_linestr))
1642 PL_expect = XOPERATOR;
1643 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1647 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1648 PL_lex_state = LEX_INTERPCASEMOD;
1652 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1653 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1654 PL_linestr = PL_lex_repl;
1656 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1657 PL_bufend += SvCUR(PL_linestr);
1658 PL_last_lop = PL_last_uni = NULL;
1659 SAVEFREESV(PL_linestr);
1660 PL_lex_dojoin = FALSE;
1661 PL_lex_brackets = 0;
1662 PL_lex_casemods = 0;
1663 *PL_lex_casestack = '\0';
1665 if (SvEVALED(PL_lex_repl)) {
1666 PL_lex_state = LEX_INTERPNORMAL;
1668 /* we don't clear PL_lex_repl here, so that we can check later
1669 whether this is an evalled subst; that means we rely on the
1670 logic to ensure sublex_done() is called again only via the
1671 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1674 PL_lex_state = LEX_INTERPCONCAT;
1684 PL_endwhite = newSVpvs("");
1685 sv_catsv(PL_endwhite, PL_thiswhite);
1689 sv_setpvn(PL_thistoken,"",0);
1691 PL_realtokenstart = -1;
1695 PL_bufend = SvPVX(PL_linestr);
1696 PL_bufend += SvCUR(PL_linestr);
1697 PL_expect = XOPERATOR;
1698 PL_sublex_info.sub_inwhat = 0;
1706 Extracts a pattern, double-quoted string, or transliteration. This
1709 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1710 processing a pattern (PL_lex_inpat is true), a transliteration
1711 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1713 Returns a pointer to the character scanned up to. If this is
1714 advanced from the start pointer supplied (i.e. if anything was
1715 successfully parsed), will leave an OP for the substring scanned
1716 in yylval. Caller must intuit reason for not parsing further
1717 by looking at the next characters herself.
1721 double-quoted style: \r and \n
1722 regexp special ones: \D \s
1725 case and quoting: \U \Q \E
1726 stops on @ and $, but not for $ as tail anchor
1728 In transliterations:
1729 characters are VERY literal, except for - not at the start or end
1730 of the string, which indicates a range. If the range is in bytes,
1731 scan_const expands the range to the full set of intermediate
1732 characters. If the range is in utf8, the hyphen is replaced with
1733 a certain range mark which will be handled by pmtrans() in op.c.
1735 In double-quoted strings:
1737 double-quoted style: \r and \n
1739 deprecated backrefs: \1 (in substitution replacements)
1740 case and quoting: \U \Q \E
1743 scan_const does *not* construct ops to handle interpolated strings.
1744 It stops processing as soon as it finds an embedded $ or @ variable
1745 and leaves it to the caller to work out what's going on.
1747 embedded arrays (whether in pattern or not) could be:
1748 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1750 $ in double-quoted strings must be the symbol of an embedded scalar.
1752 $ in pattern could be $foo or could be tail anchor. Assumption:
1753 it's a tail anchor if $ is the last thing in the string, or if it's
1754 followed by one of "()| \r\n\t"
1756 \1 (backreferences) are turned into $1
1758 The structure of the code is
1759 while (there's a character to process) {
1760 handle transliteration ranges
1761 skip regexp comments /(?#comment)/ and codes /(?{code})/
1762 skip #-initiated comments in //x patterns
1763 check for embedded arrays
1764 check for embedded scalars
1766 leave intact backslashes from leaveit (below)
1767 deprecate \1 in substitution replacements
1768 handle string-changing backslashes \l \U \Q \E, etc.
1769 switch (what was escaped) {
1770 handle \- in a transliteration (becomes a literal -)
1771 handle \132 (octal characters)
1772 handle \x15 and \x{1234} (hex characters)
1773 handle \N{name} (named characters)
1774 handle \cV (control characters)
1775 handle printf-style backslashes (\f, \r, \n, etc)
1777 } (end if backslash)
1778 } (end while character to read)
1783 S_scan_const(pTHX_ char *start)
1786 register char *send = PL_bufend; /* end of the constant */
1787 SV *sv = newSV(send - start); /* sv for the constant */
1788 register char *s = start; /* start of the constant */
1789 register char *d = SvPVX(sv); /* destination for copies */
1790 bool dorange = FALSE; /* are we in a translit range? */
1791 bool didrange = FALSE; /* did we just finish a range? */
1792 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1793 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1796 UV literal_endpoint = 0;
1797 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1800 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1801 /* If we are doing a trans and we know we want UTF8 set expectation */
1802 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1803 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1807 while (s < send || dorange) {
1808 /* get transliterations out of the way (they're most literal) */
1809 if (PL_lex_inwhat == OP_TRANS) {
1810 /* expand a range A-Z to the full set of characters. AIE! */
1812 I32 i; /* current expanded character */
1813 I32 min; /* first character in range */
1814 I32 max; /* last character in range */
1825 char * const c = (char*)utf8_hop((U8*)d, -1);
1829 *c = (char)UTF_TO_NATIVE(0xff);
1830 /* mark the range as done, and continue */
1836 i = d - SvPVX_const(sv); /* remember current offset */
1839 SvLEN(sv) + (has_utf8 ?
1840 (512 - UTF_CONTINUATION_MARK +
1843 /* How many two-byte within 0..255: 128 in UTF-8,
1844 * 96 in UTF-8-mod. */
1846 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1848 d = SvPVX(sv) + i; /* refresh d after realloc */
1852 for (j = 0; j <= 1; j++) {
1853 char * const c = (char*)utf8_hop((U8*)d, -1);
1854 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1860 max = (U8)0xff; /* only to \xff */
1861 uvmax = uv; /* \x{100} to uvmax */
1863 d = c; /* eat endpoint chars */
1868 d -= 2; /* eat the first char and the - */
1869 min = (U8)*d; /* first char in range */
1870 max = (U8)d[1]; /* last char in range */
1877 "Invalid range \"%c-%c\" in transliteration operator",
1878 (char)min, (char)max);
1882 if (literal_endpoint == 2 &&
1883 ((isLOWER(min) && isLOWER(max)) ||
1884 (isUPPER(min) && isUPPER(max)))) {
1886 for (i = min; i <= max; i++)
1888 *d++ = NATIVE_TO_NEED(has_utf8,i);
1890 for (i = min; i <= max; i++)
1892 *d++ = NATIVE_TO_NEED(has_utf8,i);
1897 for (i = min; i <= max; i++)
1900 const U8 ch = (U8)NATIVE_TO_UTF(i);
1901 if (UNI_IS_INVARIANT(ch))
1904 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1905 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1914 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1916 *d++ = (char)UTF_TO_NATIVE(0xff);
1918 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1922 /* mark the range as done, and continue */
1926 literal_endpoint = 0;
1931 /* range begins (ignore - as first or last char) */
1932 else if (*s == '-' && s+1 < send && s != start) {
1934 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1941 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1951 literal_endpoint = 0;
1952 native_range = TRUE;
1957 /* if we get here, we're not doing a transliteration */
1959 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1960 except for the last char, which will be done separately. */
1961 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1963 while (s+1 < send && *s != ')')
1964 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1966 else if (s[2] == '{' /* This should match regcomp.c */
1967 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1970 char *regparse = s + (s[2] == '{' ? 3 : 4);
1973 while (count && (c = *regparse)) {
1974 if (c == '\\' && regparse[1])
1982 if (*regparse != ')')
1983 regparse--; /* Leave one char for continuation. */
1984 while (s < regparse)
1985 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1989 /* likewise skip #-initiated comments in //x patterns */
1990 else if (*s == '#' && PL_lex_inpat &&
1991 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1992 while (s+1 < send && *s != '\n')
1993 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1996 /* check for embedded arrays
1997 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1999 else if (*s == '@' && s[1]) {
2000 if (isALNUM_lazy_if(s+1,UTF))
2002 if (strchr(":'{$", s[1]))
2004 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2005 break; /* in regexp, neither @+ nor @- are interpolated */
2008 /* check for embedded scalars. only stop if we're sure it's a
2011 else if (*s == '$') {
2012 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2014 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2015 break; /* in regexp, $ might be tail anchor */
2018 /* End of else if chain - OP_TRANS rejoin rest */
2021 if (*s == '\\' && s+1 < send) {
2024 /* deprecate \1 in strings and substitution replacements */
2025 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2026 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2028 if (ckWARN(WARN_SYNTAX))
2029 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2034 /* string-change backslash escapes */
2035 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2039 /* skip any other backslash escapes in a pattern */
2040 else if (PL_lex_inpat) {
2041 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2042 goto default_action;
2045 /* if we get here, it's either a quoted -, or a digit */
2048 /* quoted - in transliterations */
2050 if (PL_lex_inwhat == OP_TRANS) {
2057 if ((isALPHA(*s) || isDIGIT(*s)) &&
2059 Perl_warner(aTHX_ packWARN(WARN_MISC),
2060 "Unrecognized escape \\%c passed through",
2062 /* default action is to copy the quoted character */
2063 goto default_action;
2066 /* \132 indicates an octal constant */
2067 case '0': case '1': case '2': case '3':
2068 case '4': case '5': case '6': case '7':
2072 uv = grok_oct(s, &len, &flags, NULL);
2075 goto NUM_ESCAPE_INSERT;
2077 /* \x24 indicates a hex constant */
2081 char* const e = strchr(s, '}');
2082 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2083 PERL_SCAN_DISALLOW_PREFIX;
2088 yyerror("Missing right brace on \\x{}");
2092 uv = grok_hex(s, &len, &flags, NULL);
2098 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2099 uv = grok_hex(s, &len, &flags, NULL);
2105 /* Insert oct or hex escaped character.
2106 * There will always enough room in sv since such
2107 * escapes will be longer than any UTF-8 sequence
2108 * they can end up as. */
2110 /* We need to map to chars to ASCII before doing the tests
2113 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2114 if (!has_utf8 && uv > 255) {
2115 /* Might need to recode whatever we have
2116 * accumulated so far if it contains any
2119 * (Can't we keep track of that and avoid
2120 * this rescan? --jhi)
2124 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2125 if (!NATIVE_IS_INVARIANT(*c)) {
2130 const STRLEN offset = d - SvPVX_const(sv);
2132 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2136 while (src >= (const U8 *)SvPVX_const(sv)) {
2137 if (!NATIVE_IS_INVARIANT(*src)) {
2138 const U8 ch = NATIVE_TO_ASCII(*src);
2139 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2140 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2150 if (has_utf8 || uv > 255) {
2151 d = (char*)uvchr_to_utf8((U8*)d, uv);
2153 if (PL_lex_inwhat == OP_TRANS &&
2154 PL_sublex_info.sub_op) {
2155 PL_sublex_info.sub_op->op_private |=
2156 (PL_lex_repl ? OPpTRANS_FROM_UTF
2160 if (uv > 255 && !dorange)
2161 native_range = FALSE;
2173 /* \N{LATIN SMALL LETTER A} is a named character */
2177 char* e = strchr(s, '}');
2184 yyerror("Missing right brace on \\N{}");
2188 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2190 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2191 PERL_SCAN_DISALLOW_PREFIX;
2194 uv = grok_hex(s, &len, &flags, NULL);
2195 if ( e > s && len != (STRLEN)(e - s) ) {
2199 goto NUM_ESCAPE_INSERT;
2201 res = newSVpvn(s + 1, e - s - 1);
2202 type = newSVpvn(s - 2,e - s + 3);
2203 res = new_constant( NULL, 0, "charnames",
2204 res, NULL, SvPVX(type) );
2207 sv_utf8_upgrade(res);
2208 str = SvPV_const(res,len);
2209 #ifdef EBCDIC_NEVER_MIND
2210 /* charnames uses pack U and that has been
2211 * recently changed to do the below uni->native
2212 * mapping, so this would be redundant (and wrong,
2213 * the code point would be doubly converted).
2214 * But leave this in just in case the pack U change
2215 * gets revoked, but the semantics is still
2216 * desireable for charnames. --jhi */
2218 UV uv = utf8_to_uvchr((const U8*)str, 0);
2221 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2223 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2224 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2225 str = SvPV_const(res, len);
2229 if (!has_utf8 && SvUTF8(res)) {
2230 const char * const ostart = SvPVX_const(sv);
2231 SvCUR_set(sv, d - ostart);
2234 sv_utf8_upgrade(sv);
2235 /* this just broke our allocation above... */
2236 SvGROW(sv, (STRLEN)(send - start));
2237 d = SvPVX(sv) + SvCUR(sv);
2240 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2241 const char * const odest = SvPVX_const(sv);
2243 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2244 d = SvPVX(sv) + (d - odest);
2248 native_range = FALSE; /* \N{} is guessed to be Unicode */
2250 Copy(str, d, len, char);
2257 yyerror("Missing braces on \\N{}");
2260 /* \c is a control character */
2269 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2272 yyerror("Missing control char name in \\c");
2276 /* printf-style backslashes, formfeeds, newlines, etc */
2278 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2281 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2284 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2287 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2290 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2293 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2296 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2302 } /* end if (backslash) */
2309 /* If we started with encoded form, or already know we want it
2310 and then encode the next character */
2311 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2313 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2314 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2317 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2318 const STRLEN off = d - SvPVX_const(sv);
2319 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2321 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2324 if (uv > 255 && !dorange)
2325 native_range = FALSE;
2329 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2331 } /* while loop to process each character */
2333 /* terminate the string and set up the sv */
2335 SvCUR_set(sv, d - SvPVX_const(sv));
2336 if (SvCUR(sv) >= SvLEN(sv))
2337 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2340 if (PL_encoding && !has_utf8) {
2341 sv_recode_to_utf8(sv, PL_encoding);
2347 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2348 PL_sublex_info.sub_op->op_private |=
2349 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2353 /* shrink the sv if we allocated more than we used */
2354 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2355 SvPV_shrink_to_cur(sv);
2358 /* return the substring (via yylval) only if we parsed anything */
2359 if (s > PL_bufptr) {
2360 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2361 sv = new_constant(start, s - start,
2362 (const char *)(PL_lex_inpat ? "qr" : "q"),
2365 (( PL_lex_inwhat == OP_TRANS
2367 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2370 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2377 * Returns TRUE if there's more to the expression (e.g., a subscript),
2380 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2382 * ->[ and ->{ return TRUE
2383 * { and [ outside a pattern are always subscripts, so return TRUE
2384 * if we're outside a pattern and it's not { or [, then return FALSE
2385 * if we're in a pattern and the first char is a {
2386 * {4,5} (any digits around the comma) returns FALSE
2387 * if we're in a pattern and the first char is a [
2389 * [SOMETHING] has a funky algorithm to decide whether it's a
2390 * character class or not. It has to deal with things like
2391 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2392 * anything else returns TRUE
2395 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2398 S_intuit_more(pTHX_ register char *s)
2401 if (PL_lex_brackets)
2403 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2405 if (*s != '{' && *s != '[')
2410 /* In a pattern, so maybe we have {n,m}. */
2427 /* On the other hand, maybe we have a character class */
2430 if (*s == ']' || *s == '^')
2433 /* this is terrifying, and it works */
2434 int weight = 2; /* let's weigh the evidence */
2436 unsigned char un_char = 255, last_un_char;
2437 const char * const send = strchr(s,']');
2438 char tmpbuf[sizeof PL_tokenbuf * 4];
2440 if (!send) /* has to be an expression */
2443 Zero(seen,256,char);
2446 else if (isDIGIT(*s)) {
2448 if (isDIGIT(s[1]) && s[2] == ']')
2454 for (; s < send; s++) {
2455 last_un_char = un_char;
2456 un_char = (unsigned char)*s;
2461 weight -= seen[un_char] * 10;
2462 if (isALNUM_lazy_if(s+1,UTF)) {
2464 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2465 len = (int)strlen(tmpbuf);
2466 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2471 else if (*s == '$' && s[1] &&
2472 strchr("[#!%*<>()-=",s[1])) {
2473 if (/*{*/ strchr("])} =",s[2]))
2482 if (strchr("wds]",s[1]))
2484 else if (seen[(U8)'\''] || seen[(U8)'"'])
2486 else if (strchr("rnftbxcav",s[1]))
2488 else if (isDIGIT(s[1])) {
2490 while (s[1] && isDIGIT(s[1]))
2500 if (strchr("aA01! ",last_un_char))
2502 if (strchr("zZ79~",s[1]))
2504 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2505 weight -= 5; /* cope with negative subscript */
2508 if (!isALNUM(last_un_char)
2509 && !(last_un_char == '$' || last_un_char == '@'
2510 || last_un_char == '&')
2511 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2516 if (keyword(tmpbuf, d - tmpbuf, 0))
2519 if (un_char == last_un_char + 1)
2521 weight -= seen[un_char];
2526 if (weight >= 0) /* probably a character class */
2536 * Does all the checking to disambiguate
2538 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2539 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2541 * First argument is the stuff after the first token, e.g. "bar".
2543 * Not a method if bar is a filehandle.
2544 * Not a method if foo is a subroutine prototyped to take a filehandle.
2545 * Not a method if it's really "Foo $bar"
2546 * Method if it's "foo $bar"
2547 * Not a method if it's really "print foo $bar"
2548 * Method if it's really "foo package::" (interpreted as package->foo)
2549 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2550 * Not a method if bar is a filehandle or package, but is quoted with
2555 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2558 char *s = start + (*start == '$');
2559 char tmpbuf[sizeof PL_tokenbuf];
2567 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2571 const char *proto = SvPVX_const(cv);
2582 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2583 /* start is the beginning of the possible filehandle/object,
2584 * and s is the end of it
2585 * tmpbuf is a copy of it
2588 if (*start == '$') {
2589 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2592 len = start - SvPVX(PL_linestr);
2596 start = SvPVX(PL_linestr) + len;
2600 return *s == '(' ? FUNCMETH : METHOD;
2602 if (!keyword(tmpbuf, len, 0)) {
2603 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2607 soff = s - SvPVX(PL_linestr);
2611 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2612 if (indirgv && GvCVu(indirgv))
2614 /* filehandle or package name makes it a method */
2615 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2617 soff = s - SvPVX(PL_linestr);
2620 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2621 return 0; /* no assumptions -- "=>" quotes bearword */
2623 start_force(PL_curforce);
2624 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2625 newSVpvn(tmpbuf,len));
2626 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2628 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2633 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2635 return *s == '(' ? FUNCMETH : METHOD;
2643 * Return a string of Perl code to load the debugger. If PERL5DB
2644 * is set, it will return the contents of that, otherwise a
2645 * compile-time require of perl5db.pl.
2653 const char * const pdb = PerlEnv_getenv("PERL5DB");
2657 SETERRNO(0,SS_NORMAL);
2658 return "BEGIN { require 'perl5db.pl' }";
2664 /* Encoded script support. filter_add() effectively inserts a
2665 * 'pre-processing' function into the current source input stream.
2666 * Note that the filter function only applies to the current source file
2667 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2669 * The datasv parameter (which may be NULL) can be used to pass
2670 * private data to this instance of the filter. The filter function
2671 * can recover the SV using the FILTER_DATA macro and use it to
2672 * store private buffers and state information.
2674 * The supplied datasv parameter is upgraded to a PVIO type
2675 * and the IoDIRP/IoANY field is used to store the function pointer,
2676 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2677 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2678 * private use must be set using malloc'd pointers.
2682 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2688 if (!PL_rsfp_filters)
2689 PL_rsfp_filters = newAV();
2692 SvUPGRADE(datasv, SVt_PVIO);
2693 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2694 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2695 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2696 FPTR2DPTR(void *, IoANY(datasv)),
2697 SvPV_nolen(datasv)));
2698 av_unshift(PL_rsfp_filters, 1);
2699 av_store(PL_rsfp_filters, 0, datasv) ;
2704 /* Delete most recently added instance of this filter function. */
2706 Perl_filter_del(pTHX_ filter_t funcp)
2712 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2713 FPTR2DPTR(void*, funcp)));
2715 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2717 /* if filter is on top of stack (usual case) just pop it off */
2718 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2719 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2720 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2721 IoANY(datasv) = (void *)NULL;
2722 sv_free(av_pop(PL_rsfp_filters));
2726 /* we need to search for the correct entry and clear it */
2727 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2731 /* Invoke the idxth filter function for the current rsfp. */
2732 /* maxlen 0 = read one text line */
2734 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2739 /* This API is bad. It should have been using unsigned int for maxlen.
2740 Not sure if we want to change the API, but if not we should sanity
2741 check the value here. */
2742 const unsigned int correct_length
2751 if (!PL_rsfp_filters)
2753 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2754 /* Provide a default input filter to make life easy. */
2755 /* Note that we append to the line. This is handy. */
2756 DEBUG_P(PerlIO_printf(Perl_debug_log,
2757 "filter_read %d: from rsfp\n", idx));
2758 if (correct_length) {
2761 const int old_len = SvCUR(buf_sv);
2763 /* ensure buf_sv is large enough */
2764 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2765 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2766 correct_length)) <= 0) {
2767 if (PerlIO_error(PL_rsfp))
2768 return -1; /* error */
2770 return 0 ; /* end of file */
2772 SvCUR_set(buf_sv, old_len + len) ;
2775 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2776 if (PerlIO_error(PL_rsfp))
2777 return -1; /* error */
2779 return 0 ; /* end of file */
2782 return SvCUR(buf_sv);
2784 /* Skip this filter slot if filter has been deleted */
2785 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2786 DEBUG_P(PerlIO_printf(Perl_debug_log,
2787 "filter_read %d: skipped (filter deleted)\n",
2789 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2791 /* Get function pointer hidden within datasv */
2792 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2793 DEBUG_P(PerlIO_printf(Perl_debug_log,
2794 "filter_read %d: via function %p (%s)\n",
2795 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2796 /* Call function. The function is expected to */
2797 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2798 /* Return: <0:error, =0:eof, >0:not eof */
2799 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2803 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2806 #ifdef PERL_CR_FILTER
2807 if (!PL_rsfp_filters) {
2808 filter_add(S_cr_textfilter,NULL);
2811 if (PL_rsfp_filters) {
2813 SvCUR_set(sv, 0); /* start with empty line */
2814 if (FILTER_READ(0, sv, 0) > 0)
2815 return ( SvPVX(sv) ) ;
2820 return (sv_gets(sv, fp, append));
2824 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2829 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2833 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2834 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2836 return GvHV(gv); /* Foo:: */
2839 /* use constant CLASS => 'MyClass' */
2840 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2841 if (gv && GvCV(gv)) {
2842 SV * const sv = cv_const_sv(GvCV(gv));
2844 pkgname = SvPV_nolen_const(sv);
2847 return gv_stashpv(pkgname, FALSE);
2851 * S_readpipe_override
2852 * Check whether readpipe() is overriden, and generates the appropriate
2853 * optree, provided sublex_start() is called afterwards.
2856 S_readpipe_override(pTHX)
2859 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2860 yylval.ival = OP_BACKTICK;
2862 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2864 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2865 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2866 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2868 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2869 append_elem(OP_LIST,
2870 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2871 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2881 * The intent of this yylex wrapper is to minimize the changes to the
2882 * tokener when we aren't interested in collecting madprops. It remains
2883 * to be seen how successful this strategy will be...
2890 char *s = PL_bufptr;
2892 /* make sure PL_thiswhite is initialized */
2896 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2897 if (PL_pending_ident)
2898 return S_pending_ident(aTHX);
2900 /* previous token ate up our whitespace? */
2901 if (!PL_lasttoke && PL_nextwhite) {
2902 PL_thiswhite = PL_nextwhite;
2906 /* isolate the token, and figure out where it is without whitespace */
2907 PL_realtokenstart = -1;
2911 assert(PL_curforce < 0);
2913 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2914 if (!PL_thistoken) {
2915 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2916 PL_thistoken = newSVpvs("");
2918 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2919 PL_thistoken = newSVpvn(tstart, s - tstart);
2922 if (PL_thismad) /* install head */
2923 CURMAD('X', PL_thistoken);
2926 /* last whitespace of a sublex? */
2927 if (optype == ')' && PL_endwhite) {
2928 CURMAD('X', PL_endwhite);
2933 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2934 if (!PL_thiswhite && !PL_endwhite && !optype) {
2935 sv_free(PL_thistoken);
2940 /* put off final whitespace till peg */
2941 if (optype == ';' && !PL_rsfp) {
2942 PL_nextwhite = PL_thiswhite;
2945 else if (PL_thisopen) {
2946 CURMAD('q', PL_thisopen);
2948 sv_free(PL_thistoken);
2952 /* Store actual token text as madprop X */
2953 CURMAD('X', PL_thistoken);
2957 /* add preceding whitespace as madprop _ */
2958 CURMAD('_', PL_thiswhite);
2962 /* add quoted material as madprop = */
2963 CURMAD('=', PL_thisstuff);
2967 /* add terminating quote as madprop Q */
2968 CURMAD('Q', PL_thisclose);
2972 /* special processing based on optype */
2976 /* opval doesn't need a TOKEN since it can already store mp */
2987 append_madprops(PL_thismad, yylval.opval, 0);
2995 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3004 /* remember any fake bracket that lexer is about to discard */
3005 if (PL_lex_brackets == 1 &&
3006 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3009 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3012 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3013 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3016 break; /* don't bother looking for trailing comment */
3025 /* attach a trailing comment to its statement instead of next token */
3029 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3031 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3033 if (*s == '\n' || *s == '#') {
3034 while (s < PL_bufend && *s != '\n')
3038 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3039 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3056 /* Create new token struct. Note: opvals return early above. */
3057 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3064 S_tokenize_use(pTHX_ int is_use, char *s) {
3066 if (PL_expect != XSTATE)
3067 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3068 is_use ? "use" : "no"));
3070 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3071 s = force_version(s, TRUE);
3072 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3073 start_force(PL_curforce);
3074 NEXTVAL_NEXTTOKE.opval = NULL;
3077 else if (*s == 'v') {
3078 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3079 s = force_version(s, FALSE);
3083 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3084 s = force_version(s, FALSE);
3086 yylval.ival = is_use;
3090 static const char* const exp_name[] =
3091 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3092 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3099 Works out what to call the token just pulled out of the input
3100 stream. The yacc parser takes care of taking the ops we return and
3101 stitching them into a tree.
3107 if read an identifier
3108 if we're in a my declaration
3109 croak if they tried to say my($foo::bar)
3110 build the ops for a my() declaration
3111 if it's an access to a my() variable
3112 are we in a sort block?
3113 croak if my($a); $a <=> $b
3114 build ops for access to a my() variable
3115 if in a dq string, and they've said @foo and we can't find @foo
3117 build ops for a bareword
3118 if we already built the token before, use it.
3123 #pragma segment Perl_yylex
3129 register char *s = PL_bufptr;
3134 /* orig_keyword, gvp, and gv are initialized here because
3135 * jump to the label just_a_word_zero can bypass their
3136 * initialization later. */
3137 I32 orig_keyword = 0;
3142 SV* tmp = newSVpvs("");
3143 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3144 (IV)CopLINE(PL_curcop),
3145 lex_state_names[PL_lex_state],
3146 exp_name[PL_expect],
3147 pv_display(tmp, s, strlen(s), 0, 60));
3150 /* check if there's an identifier for us to look at */
3151 if (PL_pending_ident)
3152 return REPORT(S_pending_ident(aTHX));
3154 /* no identifier pending identification */
3156 switch (PL_lex_state) {
3158 case LEX_NORMAL: /* Some compilers will produce faster */
3159 case LEX_INTERPNORMAL: /* code if we comment these out. */
3163 /* when we've already built the next token, just pull it out of the queue */
3167 yylval = PL_nexttoke[PL_lasttoke].next_val;
3169 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3170 PL_nexttoke[PL_lasttoke].next_mad = 0;
3171 if (PL_thismad && PL_thismad->mad_key == '_') {
3172 PL_thiswhite = (SV*)PL_thismad->mad_val;
3173 PL_thismad->mad_val = 0;
3174 mad_free(PL_thismad);
3179 PL_lex_state = PL_lex_defer;
3180 PL_expect = PL_lex_expect;
3181 PL_lex_defer = LEX_NORMAL;
3182 if (!PL_nexttoke[PL_lasttoke].next_type)
3187 yylval = PL_nextval[PL_nexttoke];
3189 PL_lex_state = PL_lex_defer;
3190 PL_expect = PL_lex_expect;
3191 PL_lex_defer = LEX_NORMAL;
3195 /* FIXME - can these be merged? */
3196 return(PL_nexttoke[PL_lasttoke].next_type);
3198 return REPORT(PL_nexttype[PL_nexttoke]);
3201 /* interpolated case modifiers like \L \U, including \Q and \E.
3202 when we get here, PL_bufptr is at the \
3204 case LEX_INTERPCASEMOD:
3206 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3207 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3209 /* handle \E or end of string */
3210 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3212 if (PL_lex_casemods) {
3213 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3214 PL_lex_casestack[PL_lex_casemods] = '\0';
3216 if (PL_bufptr != PL_bufend
3217 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3219 PL_lex_state = LEX_INTERPCONCAT;
3222 PL_thistoken = newSVpvs("\\E");
3228 while (PL_bufptr != PL_bufend &&
3229 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3231 PL_thiswhite = newSVpvs("");
3232 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3236 if (PL_bufptr != PL_bufend)
3239 PL_lex_state = LEX_INTERPCONCAT;
3243 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3244 "### Saw case modifier\n"); });
3246 if (s[1] == '\\' && s[2] == 'E') {
3249 PL_thiswhite = newSVpvs("");
3250 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3253 PL_lex_state = LEX_INTERPCONCAT;
3258 if (!PL_madskills) /* when just compiling don't need correct */
3259 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3260 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3261 if ((*s == 'L' || *s == 'U') &&
3262 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3263 PL_lex_casestack[--PL_lex_casemods] = '\0';
3266 if (PL_lex_casemods > 10)
3267 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3268 PL_lex_casestack[PL_lex_casemods++] = *s;
3269 PL_lex_casestack[PL_lex_casemods] = '\0';
3270 PL_lex_state = LEX_INTERPCONCAT;
3271 start_force(PL_curforce);
3272 NEXTVAL_NEXTTOKE.ival = 0;
3274 start_force(PL_curforce);
3276 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3278 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3280 NEXTVAL_NEXTTOKE.ival = OP_LC;
3282 NEXTVAL_NEXTTOKE.ival = OP_UC;
3284 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3286 Perl_croak(aTHX_ "panic: yylex");
3288 SV* const tmpsv = newSVpvs("");
3289 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3295 if (PL_lex_starts) {
3301 sv_free(PL_thistoken);
3302 PL_thistoken = newSVpvs("");
3305 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3306 if (PL_lex_casemods == 1 && PL_lex_inpat)
3315 case LEX_INTERPPUSH:
3316 return REPORT(sublex_push());
3318 case LEX_INTERPSTART:
3319 if (PL_bufptr == PL_bufend)
3320 return REPORT(sublex_done());
3321 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3322 "### Interpolated variable\n"); });
3324 PL_lex_dojoin = (*PL_bufptr == '@');
3325 PL_lex_state = LEX_INTERPNORMAL;
3326 if (PL_lex_dojoin) {
3327 start_force(PL_curforce);
3328 NEXTVAL_NEXTTOKE.ival = 0;
3330 start_force(PL_curforce);
3331 force_ident("\"", '$');
3332 start_force(PL_curforce);
3333 NEXTVAL_NEXTTOKE.ival = 0;
3335 start_force(PL_curforce);
3336 NEXTVAL_NEXTTOKE.ival = 0;
3338 start_force(PL_curforce);
3339 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3342 if (PL_lex_starts++) {
3347 sv_free(PL_thistoken);
3348 PL_thistoken = newSVpvs("");
3351 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3352 if (!PL_lex_casemods && PL_lex_inpat)
3359 case LEX_INTERPENDMAYBE:
3360 if (intuit_more(PL_bufptr)) {
3361 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3367 if (PL_lex_dojoin) {
3368 PL_lex_dojoin = FALSE;
3369 PL_lex_state = LEX_INTERPCONCAT;
3373 sv_free(PL_thistoken);
3374 PL_thistoken = newSVpvs("");
3379 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3380 && SvEVALED(PL_lex_repl))
3382 if (PL_bufptr != PL_bufend)
3383 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3387 case LEX_INTERPCONCAT:
3389 if (PL_lex_brackets)
3390 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3392 if (PL_bufptr == PL_bufend)
3393 return REPORT(sublex_done());
3395 if (SvIVX(PL_linestr) == '\'') {
3396 SV *sv = newSVsv(PL_linestr);
3399 else if ( PL_hints & HINT_NEW_RE )
3400 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3401 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3405 s = scan_const(PL_bufptr);
3407 PL_lex_state = LEX_INTERPCASEMOD;
3409 PL_lex_state = LEX_INTERPSTART;
3412 if (s != PL_bufptr) {
3413 start_force(PL_curforce);
3415 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3417 NEXTVAL_NEXTTOKE = yylval;
3420 if (PL_lex_starts++) {
3424 sv_free(PL_thistoken);
3425 PL_thistoken = newSVpvs("");
3428 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3429 if (!PL_lex_casemods && PL_lex_inpat)
3442 PL_lex_state = LEX_NORMAL;
3443 s = scan_formline(PL_bufptr);
3444 if (!PL_lex_formbrack)
3450 PL_oldoldbufptr = PL_oldbufptr;
3456 sv_free(PL_thistoken);
3459 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3463 if (isIDFIRST_lazy_if(s,UTF))
3465 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3468 goto fake_eof; /* emulate EOF on ^D or ^Z */
3477 if (PL_lex_brackets) {
3478 yyerror((const char *)
3480 ? "Format not terminated"
3481 : "Missing right curly or square bracket"));
3483 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3484 "### Tokener got EOF\n");
3488 if (s++ < PL_bufend)
3489 goto retry; /* ignore stray nulls */
3492 if (!PL_in_eval && !PL_preambled) {
3493 PL_preambled = TRUE;
3498 sv_setpv(PL_linestr,incl_perldb());
3499 if (SvCUR(PL_linestr))
3500 sv_catpvs(PL_linestr,";");
3502 while(AvFILLp(PL_preambleav) >= 0) {
3503 SV *tmpsv = av_shift(PL_preambleav);
3504 sv_catsv(PL_linestr, tmpsv);
3505 sv_catpvs(PL_linestr, ";");
3508 sv_free((SV*)PL_preambleav);
3509 PL_preambleav = NULL;
3511 if (PL_minus_n || PL_minus_p) {
3512 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3514 sv_catpvs(PL_linestr,"chomp;");
3517 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3518 || *PL_splitstr == '"')
3519 && strchr(PL_splitstr + 1, *PL_splitstr))
3520 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3522 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3523 bytes can be used as quoting characters. :-) */
3524 const char *splits = PL_splitstr;
3525 sv_catpvs(PL_linestr, "our @F=split(q\0");
3528 if (*splits == '\\')
3529 sv_catpvn(PL_linestr, splits, 1);
3530 sv_catpvn(PL_linestr, splits, 1);
3531 } while (*splits++);
3532 /* This loop will embed the trailing NUL of
3533 PL_linestr as the last thing it does before
3535 sv_catpvs(PL_linestr, ");");
3539 sv_catpvs(PL_linestr,"our @F=split(' ');");
3543 sv_catpvs(PL_linestr,"use feature ':5.10';");
3544 sv_catpvs(PL_linestr, "\n");
3545 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3546 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3547 PL_last_lop = PL_last_uni = NULL;
3548 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3549 SV * const sv = newSV(0);
3551 sv_upgrade(sv, SVt_PVMG);
3552 sv_setsv(sv,PL_linestr);
3555 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3560 bof = PL_rsfp ? TRUE : FALSE;
3561 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3564 PL_realtokenstart = -1;
3567 if (PL_preprocess && !PL_in_eval)
3568 (void)PerlProc_pclose(PL_rsfp);
3569 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3570 PerlIO_clearerr(PL_rsfp);
3572 (void)PerlIO_close(PL_rsfp);
3574 PL_doextract = FALSE;
3576 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3581 sv_setpv(PL_linestr,
3584 ? ";}continue{print;}" : ";}"));
3585 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3586 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3587 PL_last_lop = PL_last_uni = NULL;
3588 PL_minus_n = PL_minus_p = 0;
3591 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3592 PL_last_lop = PL_last_uni = NULL;
3593 sv_setpvn(PL_linestr,"",0);
3594 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3596 /* If it looks like the start of a BOM or raw UTF-16,
3597 * check if it in fact is. */
3603 #ifdef PERLIO_IS_STDIO
3604 # ifdef __GNU_LIBRARY__
3605 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3606 # define FTELL_FOR_PIPE_IS_BROKEN
3610 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3611 # define FTELL_FOR_PIPE_IS_BROKEN
3616 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3617 /* This loses the possibility to detect the bof
3618 * situation on perl -P when the libc5 is being used.
3619 * Workaround? Maybe attach some extra state to PL_rsfp?
3622 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3624 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3627 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3628 s = swallow_bom((U8*)s);
3632 /* Incest with pod. */
3635 sv_catsv(PL_thiswhite, PL_linestr);
3637 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3638 sv_setpvn(PL_linestr, "", 0);
3639 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3640 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3641 PL_last_lop = PL_last_uni = NULL;
3642 PL_doextract = FALSE;
3646 } while (PL_doextract);
3647 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3648 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3649 SV * const sv = newSV(0);
3651 sv_upgrade(sv, SVt_PVMG);
3652 sv_setsv(sv,PL_linestr);
3655 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3657 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3658 PL_last_lop = PL_last_uni = NULL;
3659 if (CopLINE(PL_curcop) == 1) {
3660 while (s < PL_bufend && isSPACE(*s))
3662 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3666 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3670 if (*s == '#' && *(s+1) == '!')
3672 #ifdef ALTERNATE_SHEBANG
3674 static char const as[] = ALTERNATE_SHEBANG;
3675 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3676 d = s + (sizeof(as) - 1);
3678 #endif /* ALTERNATE_SHEBANG */
3687 while (*d && !isSPACE(*d))
3691 #ifdef ARG_ZERO_IS_SCRIPT
3692 if (ipathend > ipath) {
3694 * HP-UX (at least) sets argv[0] to the script name,
3695 * which makes $^X incorrect. And Digital UNIX and Linux,
3696 * at least, set argv[0] to the basename of the Perl
3697 * interpreter. So, having found "#!", we'll set it right.
3699 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3701 assert(SvPOK(x) || SvGMAGICAL(x));
3702 if (sv_eq(x, CopFILESV(PL_curcop))) {
3703 sv_setpvn(x, ipath, ipathend - ipath);
3709 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3710 const char * const lstart = SvPV_const(x,llen);
3712 bstart += blen - llen;
3713 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3714 sv_setpvn(x, ipath, ipathend - ipath);
3719 TAINT_NOT; /* $^X is always tainted, but that's OK */
3721 #endif /* ARG_ZERO_IS_SCRIPT */
3726 d = instr(s,"perl -");
3728 d = instr(s,"perl");
3730 /* avoid getting into infinite loops when shebang
3731 * line contains "Perl" rather than "perl" */
3733 for (d = ipathend-4; d >= ipath; --d) {
3734 if ((*d == 'p' || *d == 'P')
3735 && !ibcmp(d, "perl", 4))
3745 #ifdef ALTERNATE_SHEBANG
3747 * If the ALTERNATE_SHEBANG on this system starts with a
3748 * character that can be part of a Perl expression, then if
3749 * we see it but not "perl", we're probably looking at the
3750 * start of Perl code, not a request to hand off to some
3751 * other interpreter. Similarly, if "perl" is there, but
3752 * not in the first 'word' of the line, we assume the line
3753 * contains the start of the Perl program.
3755 if (d && *s != '#') {
3756 const char *c = ipath;
3757 while (*c && !strchr("; \t\r\n\f\v#", *c))
3760 d = NULL; /* "perl" not in first word; ignore */
3762 *s = '#'; /* Don't try to parse shebang line */
3764 #endif /* ALTERNATE_SHEBANG */
3765 #ifndef MACOS_TRADITIONAL
3770 !instr(s,"indir") &&
3771 instr(PL_origargv[0],"perl"))
3778 while (s < PL_bufend && isSPACE(*s))
3780 if (s < PL_bufend) {
3781 Newxz(newargv,PL_origargc+3,char*);
3783 while (s < PL_bufend && !isSPACE(*s))
3786 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3789 newargv = PL_origargv;
3792 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3794 Perl_croak(aTHX_ "Can't exec %s", ipath);
3798 while (*d && !isSPACE(*d))
3800 while (SPACE_OR_TAB(*d))
3804 const bool switches_done = PL_doswitches;
3805 const U32 oldpdb = PL_perldb;
3806 const bool oldn = PL_minus_n;
3807 const bool oldp = PL_minus_p;
3810 if (*d == 'M' || *d == 'm' || *d == 'C') {
3811 const char * const m = d;
3812 while (*d && !isSPACE(*d))
3814 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3817 d = moreswitches(d);
3819 if (PL_doswitches && !switches_done) {
3820 int argc = PL_origargc;
3821 char **argv = PL_origargv;
3824 } while (argc && argv[0][0] == '-' && argv[0][1]);
3825 init_argv_symbols(argc,argv);
3827 if ((PERLDB_LINE && !oldpdb) ||
3828 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3829 /* if we have already added "LINE: while (<>) {",
3830 we must not do it again */
3832 sv_setpvn(PL_linestr, "", 0);
3833 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3834 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3835 PL_last_lop = PL_last_uni = NULL;
3836 PL_preambled = FALSE;
3838 (void)gv_fetchfile(PL_origfilename);
3845 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3847 PL_lex_state = LEX_FORMLINE;
3852 #ifdef PERL_STRICT_CR
3853 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3855 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3857 case ' ': case '\t': case '\f': case 013:
3858 #ifdef MACOS_TRADITIONAL
3862 PL_realtokenstart = -1;
3871 PL_realtokenstart = -1;
3875 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3876 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3877 /* handle eval qq[#line 1 "foo"\n ...] */
3878 CopLINE_dec(PL_curcop);
3881 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3883 if (!PL_in_eval || PL_rsfp)
3888 while (d < PL_bufend && *d != '\n')
3892 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3893 Perl_croak(aTHX_ "panic: input overflow");
3896 PL_thiswhite = newSVpvn(s, d - s);
3901 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3903 PL_lex_state = LEX_FORMLINE;
3909 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3910 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3913 TOKEN(PEG); /* make sure any #! line is accessible */
3918 /* if (PL_madskills && PL_lex_formbrack) { */
3920 while (d < PL_bufend && *d != '\n')
3924 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3925 Perl_croak(aTHX_ "panic: input overflow");
3926 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3928 PL_thiswhite = newSVpvs("");
3929 if (CopLINE(PL_curcop) == 1) {
3930 sv_setpvn(PL_thiswhite, "", 0);
3933 sv_catpvn(PL_thiswhite, s, d - s);
3947 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3955 while (s < PL_bufend && SPACE_OR_TAB(*s))
3958 if (strnEQ(s,"=>",2)) {
3959 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3960 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3961 OPERATOR('-'); /* unary minus */
3963 PL_last_uni = PL_oldbufptr;
3965 case 'r': ftst = OP_FTEREAD; break;
3966 case 'w': ftst = OP_FTEWRITE; break;
3967 case 'x': ftst = OP_FTEEXEC; break;
3968 case 'o': ftst = OP_FTEOWNED; break;
3969 case 'R': ftst = OP_FTRREAD; break;
3970 case 'W': ftst = OP_FTRWRITE; break;
3971 case 'X': ftst = OP_FTREXEC; break;
3972 case 'O': ftst = OP_FTROWNED; break;
3973 case 'e': ftst = OP_FTIS; break;
3974 case 'z': ftst = OP_FTZERO; break;
3975 case 's': ftst = OP_FTSIZE; break;
3976 case 'f': ftst = OP_FTFILE; break;
3977 case 'd': ftst = OP_FTDIR; break;
3978 case 'l': ftst = OP_FTLINK; break;
3979 case 'p': ftst = OP_FTPIPE; break;
3980 case 'S': ftst = OP_FTSOCK; break;
3981 case 'u': ftst = OP_FTSUID; break;
3982 case 'g': ftst = OP_FTSGID; break;
3983 case 'k': ftst = OP_FTSVTX; break;
3984 case 'b': ftst = OP_FTBLK; break;
3985 case 'c': ftst = OP_FTCHR; break;
3986 case 't': ftst = OP_FTTTY; break;
3987 case 'T': ftst = OP_FTTEXT; break;
3988 case 'B': ftst = OP_FTBINARY; break;
3989 case 'M': case 'A': case 'C':
3990 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3992 case 'M': ftst = OP_FTMTIME; break;
3993 case 'A': ftst = OP_FTATIME; break;
3994 case 'C': ftst = OP_FTCTIME; break;
4002 PL_last_lop_op = (OPCODE)ftst;
4003 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4004 "### Saw file test %c\n", (int)tmp);
4009 /* Assume it was a minus followed by a one-letter named
4010 * subroutine call (or a -bareword), then. */
4011 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4012 "### '-%c' looked like a file test but was not\n",
4019 const char tmp = *s++;
4022 if (PL_expect == XOPERATOR)
4027 else if (*s == '>') {
4030 if (isIDFIRST_lazy_if(s,UTF)) {
4031 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4039 if (PL_expect == XOPERATOR)
4042 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4044 OPERATOR('-'); /* unary minus */
4050 const char tmp = *s++;
4053 if (PL_expect == XOPERATOR)
4058 if (PL_expect == XOPERATOR)
4061 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4068 if (PL_expect != XOPERATOR) {
4069 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4070 PL_expect = XOPERATOR;
4071 force_ident(PL_tokenbuf, '*');
4084 if (PL_expect == XOPERATOR) {
4088 PL_tokenbuf[0] = '%';
4089 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4090 if (!PL_tokenbuf[1]) {
4093 PL_pending_ident = '%';
4104 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
4105 && FEATURE_IS_ENABLED("~~"))
4112 const char tmp = *s++;
4118 goto just_a_word_zero_gv;
4121 switch (PL_expect) {
4127 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4129 PL_bufptr = s; /* update in case we back off */
4135 PL_expect = XTERMBLOCK;
4138 stuffstart = s - SvPVX(PL_linestr) - 1;
4142 while (isIDFIRST_lazy_if(s,UTF)) {
4145 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4146 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4147 if (tmp < 0) tmp = -tmp;
4162 sv = newSVpvn(s, len);
4164 d = scan_str(d,TRUE,TRUE);
4166 /* MUST advance bufptr here to avoid bogus
4167 "at end of line" context messages from yyerror().
4169 PL_bufptr = s + len;
4170 yyerror("Unterminated attribute parameter in attribute list");
4174 return REPORT(0); /* EOF indicator */
4178 sv_catsv(sv, PL_lex_stuff);
4179 attrs = append_elem(OP_LIST, attrs,
4180 newSVOP(OP_CONST, 0, sv));
4181 SvREFCNT_dec(PL_lex_stuff);
4182 PL_lex_stuff = NULL;
4185 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4187 if (PL_in_my == KEY_our) {
4189 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4191 /* skip to avoid loading attributes.pm */
4193 deprecate(":unique");
4196 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4199 /* NOTE: any CV attrs applied here need to be part of
4200 the CVf_BUILTIN_ATTRS define in cv.h! */
4201 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4203 CvLVALUE_on(PL_compcv);
4205 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4207 CvLOCKED_on(PL_compcv);
4209 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4211 CvMETHOD_on(PL_compcv);
4213 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4215 CvASSERTION_on(PL_compcv);
4217 /* After we've set the flags, it could be argued that
4218 we don't need to do the attributes.pm-based setting
4219 process, and shouldn't bother appending recognized
4220 flags. To experiment with that, uncomment the
4221 following "else". (Note that's already been
4222 uncommented. That keeps the above-applied built-in
4223 attributes from being intercepted (and possibly
4224 rejected) by a package's attribute routines, but is
4225 justified by the performance win for the common case
4226 of applying only built-in attributes.) */
4228 attrs = append_elem(OP_LIST, attrs,
4229 newSVOP(OP_CONST, 0,
4233 if (*s == ':' && s[1] != ':')
4236 break; /* require real whitespace or :'s */
4237 /* XXX losing whitespace on sequential attributes here */
4241 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4242 if (*s != ';' && *s != '}' && *s != tmp
4243 && (tmp != '=' || *s != ')')) {
4244 const char q = ((*s == '\'') ? '"' : '\'');
4245 /* If here for an expression, and parsed no attrs, back
4247 if (tmp == '=' && !attrs) {
4251 /* MUST advance bufptr here to avoid bogus "at end of line"
4252 context messages from yyerror().
4255 yyerror( (const char *)
4257 ? Perl_form(aTHX_ "Invalid separator character "
4258 "%c%c%c in attribute list", q, *s, q)
4259 : "Unterminated attribute list" ) );
4267 start_force(PL_curforce);
4268 NEXTVAL_NEXTTOKE.opval = attrs;
4269 CURMAD('_', PL_nextwhite);
4274 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4275 (s - SvPVX(PL_linestr)) - stuffstart);
4283 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4284 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4292 const char tmp = *s++;
4297 const char tmp = *s++;
4305 if (PL_lex_brackets <= 0)
4306 yyerror("Unmatched right square bracket");
4309 if (PL_lex_state == LEX_INTERPNORMAL) {
4310 if (PL_lex_brackets == 0) {
4311 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4312 PL_lex_state = LEX_INTERPEND;
4319 if (PL_lex_brackets > 100) {
4320 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4322 switch (PL_expect) {
4324 if (PL_lex_formbrack) {
4328 if (PL_oldoldbufptr == PL_last_lop)
4329 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4331 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4332 OPERATOR(HASHBRACK);
4334 while (s < PL_bufend && SPACE_OR_TAB(*s))
4337 PL_tokenbuf[0] = '\0';
4338 if (d < PL_bufend && *d == '-') {
4339 PL_tokenbuf[0] = '-';
4341 while (d < PL_bufend && SPACE_OR_TAB(*d))
4344 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4345 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4347 while (d < PL_bufend && SPACE_OR_TAB(*d))
4350 const char minus = (PL_tokenbuf[0] == '-');
4351 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4359 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4364 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4369 if (PL_oldoldbufptr == PL_last_lop)
4370 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4372 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4375 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4377 /* This hack is to get the ${} in the message. */
4379 yyerror("syntax error");
4382 OPERATOR(HASHBRACK);
4384 /* This hack serves to disambiguate a pair of curlies
4385 * as being a block or an anon hash. Normally, expectation
4386 * determines that, but in cases where we're not in a
4387 * position to expect anything in particular (like inside
4388 * eval"") we have to resolve the ambiguity. This code
4389 * covers the case where the first term in the curlies is a
4390 * quoted string. Most other cases need to be explicitly
4391 * disambiguated by prepending a "+" before the opening
4392 * curly in order to force resolution as an anon hash.
4394 * XXX should probably propagate the outer expectation
4395 * into eval"" to rely less on this hack, but that could
4396 * potentially break current behavior of eval"".
4400 if (*s == '\'' || *s == '"' || *s == '`') {
4401 /* common case: get past first string, handling escapes */
4402 for (t++; t < PL_bufend && *t != *s;)
4403 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4407 else if (*s == 'q') {
4410 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4413 /* skip q//-like construct */
4415 char open, close, term;
4418 while (t < PL_bufend && isSPACE(*t))
4420 /* check for q => */
4421 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4422 OPERATOR(HASHBRACK);
4426 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4430 for (t++; t < PL_bufend; t++) {
4431 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4433 else if (*t == open)
4437 for (t++; t < PL_bufend; t++) {
4438 if (*t == '\\' && t+1 < PL_bufend)
4440 else if (*t == close && --brackets <= 0)
4442 else if (*t == open)
4449 /* skip plain q word */
4450 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4453 else if (isALNUM_lazy_if(t,UTF)) {
4455 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4458 while (t < PL_bufend && isSPACE(*t))
4460 /* if comma follows first term, call it an anon hash */
4461 /* XXX it could be a comma expression with loop modifiers */
4462 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4463 || (*t == '=' && t[1] == '>')))
4464 OPERATOR(HASHBRACK);
4465 if (PL_expect == XREF)
4468 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4474 yylval.ival = CopLINE(PL_curcop);
4475 if (isSPACE(*s) || *s == '#')
4476 PL_copline = NOLINE; /* invalidate current command line number */
4481 if (PL_lex_brackets <= 0)
4482 yyerror("Unmatched right curly bracket");
4484 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4485 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4486 PL_lex_formbrack = 0;
4487 if (PL_lex_state == LEX_INTERPNORMAL) {
4488 if (PL_lex_brackets == 0) {
4489 if (PL_expect & XFAKEBRACK) {
4490 PL_expect &= XENUMMASK;
4491 PL_lex_state = LEX_INTERPEND;
4496 PL_thiswhite = newSVpvs("");
4497 sv_catpvn(PL_thiswhite,"}",1);
4500 return yylex(); /* ignore fake brackets */
4502 if (*s == '-' && s[1] == '>')
4503 PL_lex_state = LEX_INTERPENDMAYBE;
4504 else if (*s != '[' && *s != '{')
4505 PL_lex_state = LEX_INTERPEND;
4508 if (PL_expect & XFAKEBRACK) {
4509 PL_expect &= XENUMMASK;
4511 return yylex(); /* ignore fake brackets */
4513 start_force(PL_curforce);
4515 curmad('X', newSVpvn(s-1,1));
4516 CURMAD('_', PL_thiswhite);
4521 PL_thistoken = newSVpvs("");
4529 if (PL_expect == XOPERATOR) {
4530 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4531 && isIDFIRST_lazy_if(s,UTF))
4533 CopLINE_dec(PL_curcop);
4534 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4535 CopLINE_inc(PL_curcop);
4540 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4542 PL_expect = XOPERATOR;
4543 force_ident(PL_tokenbuf, '&');
4547 yylval.ival = (OPpENTERSUB_AMPER<<8);
4559 const char tmp = *s++;
4566 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4567 && strchr("+-*/%.^&|<",tmp))
4568 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4569 "Reversed %c= operator",(int)tmp);
4571 if (PL_expect == XSTATE && isALPHA(tmp) &&
4572 (s == PL_linestart+1 || s[-2] == '\n') )
4574 if (PL_in_eval && !PL_rsfp) {
4579 if (strnEQ(s,"=cut",4)) {
4595 PL_thiswhite = newSVpvs("");
4596 sv_catpvn(PL_thiswhite, PL_linestart,
4597 PL_bufend - PL_linestart);
4601 PL_doextract = TRUE;
4605 if (PL_lex_brackets < PL_lex_formbrack) {
4607 #ifdef PERL_STRICT_CR
4608 while (SPACE_OR_TAB(*t))
4610 while (SPACE_OR_TAB(*t) || *t == '\r')
4613 if (*t == '\n' || *t == '#') {
4624 const char tmp = *s++;
4626 /* was this !=~ where !~ was meant?
4627 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4629 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4630 const char *t = s+1;
4632 while (t < PL_bufend && isSPACE(*t))
4635 if (*t == '/' || *t == '?' ||
4636 ((*t == 'm' || *t == 's' || *t == 'y')
4637 && !isALNUM(t[1])) ||
4638 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4639 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4640 "!=~ should be !~");
4650 if (PL_expect != XOPERATOR) {
4651 if (s[1] != '<' && !strchr(s,'>'))
4654 s = scan_heredoc(s);
4656 s = scan_inputsymbol(s);
4657 TERM(sublex_start());
4663 SHop(OP_LEFT_SHIFT);
4677 const char tmp = *s++;
4679 SHop(OP_RIGHT_SHIFT);
4680 else if (tmp == '=')
4689 if (PL_expect == XOPERATOR) {
4690 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4692 deprecate_old(commaless_variable_list);
4693 return REPORT(','); /* grandfather non-comma-format format */
4697 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4698 PL_tokenbuf[0] = '@';
4699 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4700 sizeof PL_tokenbuf - 1, FALSE);
4701 if (PL_expect == XOPERATOR)
4702 no_op("Array length", s);
4703 if (!PL_tokenbuf[1])
4705 PL_expect = XOPERATOR;
4706 PL_pending_ident = '#';
4710 PL_tokenbuf[0] = '$';
4711 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4712 sizeof PL_tokenbuf - 1, FALSE);
4713 if (PL_expect == XOPERATOR)
4715 if (!PL_tokenbuf[1]) {
4717 yyerror("Final $ should be \\$ or $name");
4721 /* This kludge not intended to be bulletproof. */
4722 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4723 yylval.opval = newSVOP(OP_CONST, 0,
4724 newSViv(CopARYBASE_get(&PL_compiling)));
4725 yylval.opval->op_private = OPpCONST_ARYBASE;
4731 const char tmp = *s;
4732 if (PL_lex_state == LEX_NORMAL)
4735 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4736 && intuit_more(s)) {
4738 PL_tokenbuf[0] = '@';
4739 if (ckWARN(WARN_SYNTAX)) {
4742 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4745 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4746 while (t < PL_bufend && *t != ']')
4748 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4749 "Multidimensional syntax %.*s not supported",
4750 (int)((t - PL_bufptr) + 1), PL_bufptr);
4754 else if (*s == '{') {
4756 PL_tokenbuf[0] = '%';
4757 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4758 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4760 char tmpbuf[sizeof PL_tokenbuf];
4763 } while (isSPACE(*t));
4764 if (isIDFIRST_lazy_if(t,UTF)) {
4766 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4770 if (*t == ';' && get_cv(tmpbuf, FALSE))
4771 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4772 "You need to quote \"%s\"",
4779 PL_expect = XOPERATOR;
4780 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4781 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4782 if (!islop || PL_last_lop_op == OP_GREPSTART)
4783 PL_expect = XOPERATOR;
4784 else if (strchr("$@\"'`q", *s))
4785 PL_expect = XTERM; /* e.g. print $fh "foo" */
4786 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4787 PL_expect = XTERM; /* e.g. print $fh &sub */
4788 else if (isIDFIRST_lazy_if(s,UTF)) {
4789 char tmpbuf[sizeof PL_tokenbuf];
4791 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4792 if ((t2 = keyword(tmpbuf, len, 0))) {
4793 /* binary operators exclude handle interpretations */
4805 PL_expect = XTERM; /* e.g. print $fh length() */
4810 PL_expect = XTERM; /* e.g. print $fh subr() */
4813 else if (isDIGIT(*s))
4814 PL_expect = XTERM; /* e.g. print $fh 3 */
4815 else if (*s == '.' && isDIGIT(s[1]))
4816 PL_expect = XTERM; /* e.g. print $fh .3 */
4817 else if ((*s == '?' || *s == '-' || *s == '+')
4818 && !isSPACE(s[1]) && s[1] != '=')
4819 PL_expect = XTERM; /* e.g. print $fh -1 */
4820 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4822 PL_expect = XTERM; /* e.g. print $fh /.../
4823 XXX except DORDOR operator
4825 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4827 PL_expect = XTERM; /* print $fh <<"EOF" */
4830 PL_pending_ident = '$';
4834 if (PL_expect == XOPERATOR)
4836 PL_tokenbuf[0] = '@';
4837 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4838 if (!PL_tokenbuf[1]) {
4841 if (PL_lex_state == LEX_NORMAL)
4843 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4845 PL_tokenbuf[0] = '%';
4847 /* Warn about @ where they meant $. */
4848 if (*s == '[' || *s == '{') {
4849 if (ckWARN(WARN_SYNTAX)) {
4850 const char *t = s + 1;
4851 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4853 if (*t == '}' || *t == ']') {
4855 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4856 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4857 "Scalar value %.*s better written as $%.*s",
4858 (int)(t-PL_bufptr), PL_bufptr,
4859 (int)(t-PL_bufptr-1), PL_bufptr+1);
4864 PL_pending_ident = '@';
4867 case '/': /* may be division, defined-or, or pattern */
4868 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4872 case '?': /* may either be conditional or pattern */
4873 if(PL_expect == XOPERATOR) {
4881 /* A // operator. */
4891 /* Disable warning on "study /blah/" */
4892 if (PL_oldoldbufptr == PL_last_uni
4893 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4894 || memNE(PL_last_uni, "study", 5)
4895 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4898 s = scan_pat(s,OP_MATCH);
4899 TERM(sublex_start());
4903 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4904 #ifdef PERL_STRICT_CR
4907 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4909 && (s == PL_linestart || s[-1] == '\n') )
4911 PL_lex_formbrack = 0;
4915 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
4921 yylval.ival = OPf_SPECIAL;
4927 if (PL_expect != XOPERATOR)
4932 case '0': case '1': case '2': case '3': case '4':
4933 case '5': case '6': case '7': case '8': case '9':
4934 s = scan_num(s, &yylval);
4935 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
4936 if (PL_expect == XOPERATOR)
4941 s = scan_str(s,!!PL_madskills,FALSE);
4942 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4943 if (PL_expect == XOPERATOR) {
4944 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4946 deprecate_old(commaless_variable_list);
4947 return REPORT(','); /* grandfather non-comma-format format */
4954 yylval.ival = OP_CONST;
4955 TERM(sublex_start());
4958 s = scan_str(s,!!PL_madskills,FALSE);
4959 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
4960 if (PL_expect == XOPERATOR) {
4961 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4963 deprecate_old(commaless_variable_list);
4964 return REPORT(','); /* grandfather non-comma-format format */
4971 yylval.ival = OP_CONST;
4972 /* FIXME. I think that this can be const if char *d is replaced by
4973 more localised variables. */
4974 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
4975 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
4976 yylval.ival = OP_STRINGIFY;
4980 TERM(sublex_start());
4983 s = scan_str(s,!!PL_madskills,FALSE);
4984 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
4985 if (PL_expect == XOPERATOR)
4986 no_op("Backticks",s);
4989 readpipe_override();
4990 TERM(sublex_start());
4994 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
4995 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
4997 if (PL_expect == XOPERATOR)
4998 no_op("Backslash",s);
5002 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5003 char *start = s + 2;
5004 while (isDIGIT(*start) || *start == '_')
5006 if (*start == '.' && isDIGIT(start[1])) {
5007 s = scan_num(s, &yylval);
5010 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5011 else if (!isALPHA(*start) && (PL_expect == XTERM
5012 || PL_expect == XREF || PL_expect == XSTATE
5013 || PL_expect == XTERMORDORDOR)) {
5014 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5015 const char c = *start;
5018 gv = gv_fetchpv(s, 0, SVt_PVCV);
5021 s = scan_num(s, &yylval);
5028 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5070 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5072 /* Some keywords can be followed by any delimiter, including ':' */
5073 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5074 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5075 (PL_tokenbuf[0] == 'q' &&
5076 strchr("qwxr", PL_tokenbuf[1])))));
5078 /* x::* is just a word, unless x is "CORE" */
5079 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5083 while (d < PL_bufend && isSPACE(*d))
5084 d++; /* no comments skipped here, or s### is misparsed */
5086 /* Is this a label? */
5087 if (!tmp && PL_expect == XSTATE
5088 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5090 yylval.pval = savepv(PL_tokenbuf);
5095 /* Check for keywords */
5096 tmp = keyword(PL_tokenbuf, len, 0);
5098 /* Is this a word before a => operator? */
5099 if (*d == '=' && d[1] == '>') {
5102 = (OP*)newSVOP(OP_CONST, 0,
5103 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5104 yylval.opval->op_private = OPpCONST_BARE;
5108 if (tmp < 0) { /* second-class keyword? */
5109 GV *ogv = NULL; /* override (winner) */
5110 GV *hgv = NULL; /* hidden (loser) */
5111 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5113 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5116 if (GvIMPORTED_CV(gv))
5118 else if (! CvMETHOD(cv))
5122 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5123 (gv = *gvp) != (GV*)&PL_sv_undef &&
5124 GvCVu(gv) && GvIMPORTED_CV(gv))
5131 tmp = 0; /* overridden by import or by GLOBAL */
5134 && -tmp==KEY_lock /* XXX generalizable kludge */
5136 && !hv_fetchs(GvHVn(PL_incgv), "Thread.pm", FALSE))
5138 tmp = 0; /* any sub overrides "weak" keyword */
5140 else { /* no override */
5142 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5143 Perl_warner(aTHX_ packWARN(WARN_MISC),
5144 "dump() better written as CORE::dump()");
5148 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5149 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5150 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5151 "Ambiguous call resolved as CORE::%s(), %s",
5152 GvENAME(hgv), "qualify as such or use &");
5159 default: /* not a keyword */
5160 /* Trade off - by using this evil construction we can pull the
5161 variable gv into the block labelled keylookup. If not, then
5162 we have to give it function scope so that the goto from the
5163 earlier ':' case doesn't bypass the initialisation. */
5165 just_a_word_zero_gv:
5173 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5176 SV *nextPL_nextwhite = 0;
5180 /* Get the rest if it looks like a package qualifier */
5182 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5184 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5187 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5188 *s == '\'' ? "'" : "::");
5193 if (PL_expect == XOPERATOR) {
5194 if (PL_bufptr == PL_linestart) {
5195 CopLINE_dec(PL_curcop);
5196 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5197 CopLINE_inc(PL_curcop);
5200 no_op("Bareword",s);
5203 /* Look for a subroutine with this name in current package,
5204 unless name is "Foo::", in which case Foo is a bearword
5205 (and a package name). */
5207 if (len > 2 && !PL_madskills &&
5208 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5210 if (ckWARN(WARN_BAREWORD)
5211 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5212 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5213 "Bareword \"%s\" refers to nonexistent package",
5216 PL_tokenbuf[len] = '\0';
5222 /* Mustn't actually add anything to a symbol table.
5223 But also don't want to "initialise" any placeholder
5224 constants that might already be there into full
5225 blown PVGVs with attached PVCV. */
5226 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5227 GV_NOADD_NOINIT, SVt_PVCV);
5232 /* if we saw a global override before, get the right name */
5235 sv = newSVpvs("CORE::GLOBAL::");
5236 sv_catpv(sv,PL_tokenbuf);
5239 /* If len is 0, newSVpv does strlen(), which is correct.
5240 If len is non-zero, then it will be the true length,
5241 and so the scalar will be created correctly. */
5242 sv = newSVpv(PL_tokenbuf,len);
5245 if (PL_madskills && !PL_thistoken) {
5246 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5247 PL_thistoken = newSVpv(start,s - start);
5248 PL_realtokenstart = s - SvPVX(PL_linestr);
5252 /* Presume this is going to be a bareword of some sort. */
5255 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5256 yylval.opval->op_private = OPpCONST_BARE;
5257 /* UTF-8 package name? */
5258 if (UTF && !IN_BYTES &&
5259 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5262 /* And if "Foo::", then that's what it certainly is. */
5267 /* Do the explicit type check so that we don't need to force
5268 the initialisation of the symbol table to have a real GV.
5269 Beware - gv may not really be a PVGV, cv may not really be
5270 a PVCV, (because of the space optimisations that gv_init
5271 understands) But they're true if for this symbol there is
5272 respectively a typeglob and a subroutine.
5274 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5275 /* Real typeglob, so get the real subroutine: */
5277 /* A proxy for a subroutine in this package? */
5278 : SvOK(gv) ? (CV *) gv : NULL)
5281 /* See if it's the indirect object for a list operator. */
5283 if (PL_oldoldbufptr &&
5284 PL_oldoldbufptr < PL_bufptr &&
5285 (PL_oldoldbufptr == PL_last_lop
5286 || PL_oldoldbufptr == PL_last_uni) &&
5287 /* NO SKIPSPACE BEFORE HERE! */
5288 (PL_expect == XREF ||
5289 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5291 bool immediate_paren = *s == '(';
5293 /* (Now we can afford to cross potential line boundary.) */
5294 s = SKIPSPACE2(s,nextPL_nextwhite);
5296 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5299 /* Two barewords in a row may indicate method call. */
5301 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5302 (tmp = intuit_method(s, gv, cv)))
5305 /* If not a declared subroutine, it's an indirect object. */
5306 /* (But it's an indir obj regardless for sort.) */
5307 /* Also, if "_" follows a filetest operator, it's a bareword */
5310 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5312 (PL_last_lop_op != OP_MAPSTART &&
5313 PL_last_lop_op != OP_GREPSTART))))
5314 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5315 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5318 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5323 PL_expect = XOPERATOR;
5326 s = SKIPSPACE2(s,nextPL_nextwhite);
5327 PL_nextwhite = nextPL_nextwhite;
5332 /* Is this a word before a => operator? */
5333 if (*s == '=' && s[1] == '>' && !pkgname) {
5335 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5336 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5337 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5341 /* If followed by a paren, it's certainly a subroutine. */
5346 while (SPACE_OR_TAB(*d))
5348 if (*d == ')' && (sv = gv_const_sv(gv))) {
5352 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5353 sv_catpvn(PL_thistoken, par, s - par);
5355 sv_free(PL_nextwhite);
5365 PL_nextwhite = PL_thiswhite;
5368 start_force(PL_curforce);
5370 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5371 PL_expect = XOPERATOR;
5374 PL_nextwhite = nextPL_nextwhite;
5375 curmad('X', PL_thistoken);
5376 PL_thistoken = newSVpvs("");
5384 /* If followed by var or block, call it a method (unless sub) */
5386 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5387 PL_last_lop = PL_oldbufptr;
5388 PL_last_lop_op = OP_METHOD;
5392 /* If followed by a bareword, see if it looks like indir obj. */
5395 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5396 && (tmp = intuit_method(s, gv, cv)))
5399 /* Not a method, so call it a subroutine (if defined) */
5402 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5403 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5404 "Ambiguous use of -%s resolved as -&%s()",
5405 PL_tokenbuf, PL_tokenbuf);
5406 /* Check for a constant sub */
5407 if ((sv = gv_const_sv(gv))) {
5409 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5410 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5411 yylval.opval->op_private = 0;
5415 /* Resolve to GV now. */
5416 if (SvTYPE(gv) != SVt_PVGV) {
5417 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5418 assert (SvTYPE(gv) == SVt_PVGV);
5419 /* cv must have been some sort of placeholder, so
5420 now needs replacing with a real code reference. */
5424 op_free(yylval.opval);
5425 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5426 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5427 PL_last_lop = PL_oldbufptr;
5428 PL_last_lop_op = OP_ENTERSUB;
5429 /* Is there a prototype? */
5437 const char *proto = SvPV_const((SV*)cv, protolen);
5440 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5442 while (*proto == ';')
5444 if (*proto == '&' && *s == '{') {
5445 sv_setpv(PL_subname,
5448 "__ANON__" : "__ANON__::__ANON__"));
5455 PL_nextwhite = PL_thiswhite;
5458 start_force(PL_curforce);
5459 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5462 PL_nextwhite = nextPL_nextwhite;
5463 curmad('X', PL_thistoken);
5464 PL_thistoken = newSVpvs("");
5471 /* Guess harder when madskills require "best effort". */
5472 if (PL_madskills && (!gv || !GvCVu(gv))) {
5473 int probable_sub = 0;
5474 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5476 else if (isALPHA(*s)) {
5480 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5481 if (!keyword(tmpbuf, tmplen, 0))
5484 while (d < PL_bufend && isSPACE(*d))
5486 if (*d == '=' && d[1] == '>')
5491 gv = gv_fetchpv(PL_tokenbuf, TRUE, SVt_PVCV);
5492 op_free(yylval.opval);
5493 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5494 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5495 PL_last_lop = PL_oldbufptr;
5496 PL_last_lop_op = OP_ENTERSUB;
5497 PL_nextwhite = PL_thiswhite;
5499 start_force(PL_curforce);
5500 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5502 PL_nextwhite = nextPL_nextwhite;
5503 curmad('X', PL_thistoken);
5504 PL_thistoken = newSVpvs("");
5509 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5516 /* Call it a bare word */
5518 if (PL_hints & HINT_STRICT_SUBS)
5519 yylval.opval->op_private |= OPpCONST_STRICT;
5522 if (lastchar != '-') {
5523 if (ckWARN(WARN_RESERVED)) {
5527 if (!*d && !gv_stashpv(PL_tokenbuf,FALSE))
5528 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5535 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5536 && ckWARN_d(WARN_AMBIGUOUS)) {
5537 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5538 "Operator or semicolon missing before %c%s",
5539 lastchar, PL_tokenbuf);
5540 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5541 "Ambiguous use of %c resolved as operator %c",
5542 lastchar, lastchar);
5548 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5549 newSVpv(CopFILE(PL_curcop),0));
5553 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5554 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5557 case KEY___PACKAGE__:
5558 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5560 ? newSVhek(HvNAME_HEK(PL_curstash))
5567 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5568 const char *pname = "main";
5569 if (PL_tokenbuf[2] == 'D')
5570 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5571 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5575 GvIOp(gv) = newIO();
5576 IoIFP(GvIOp(gv)) = PL_rsfp;
5577 #if defined(HAS_FCNTL) && defined(F_SETFD)
5579 const int fd = PerlIO_fileno(PL_rsfp);
5580 fcntl(fd,F_SETFD,fd >= 3);
5583 /* Mark this internal pseudo-handle as clean */
5584 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5586 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5587 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5588 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5590 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5591 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5592 /* if the script was opened in binmode, we need to revert
5593 * it to text mode for compatibility; but only iff it has CRs
5594 * XXX this is a questionable hack at best. */
5595 if (PL_bufend-PL_bufptr > 2
5596 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5599 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5600 loc = PerlIO_tell(PL_rsfp);
5601 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5604 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5606 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5607 #endif /* NETWARE */
5608 #ifdef PERLIO_IS_STDIO /* really? */
5609 # if defined(__BORLANDC__)
5610 /* XXX see note in do_binmode() */
5611 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5615 PerlIO_seek(PL_rsfp, loc, 0);
5619 #ifdef PERLIO_LAYERS
5622 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5623 else if (PL_encoding) {
5630 XPUSHs(PL_encoding);
5632 call_method("name", G_SCALAR);
5636 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5637 Perl_form(aTHX_ ":encoding(%"SVf")",
5646 if (PL_realtokenstart >= 0) {
5647 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5649 PL_endwhite = newSVpvs("");
5650 sv_catsv(PL_endwhite, PL_thiswhite);
5652 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5653 PL_realtokenstart = -1;
5655 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5656 SvCUR(PL_endwhite))) != Nullch) ;
5671 if (PL_expect == XSTATE) {
5678 if (*s == ':' && s[1] == ':') {
5681 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5682 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5683 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5686 else if (tmp == KEY_require || tmp == KEY_do)
5687 /* that's a way to remember we saw "CORE::" */
5700 LOP(OP_ACCEPT,XTERM);
5706 LOP(OP_ATAN2,XTERM);
5712 LOP(OP_BINMODE,XTERM);
5715 LOP(OP_BLESS,XTERM);
5724 /* When 'use switch' is in effect, continue has a dual
5725 life as a control operator. */
5727 if (!FEATURE_IS_ENABLED("switch"))
5730 /* We have to disambiguate the two senses of
5731 "continue". If the next token is a '{' then
5732 treat it as the start of a continue block;
5733 otherwise treat it as a control operator.
5745 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5762 if (!PL_cryptseen) {
5763 PL_cryptseen = TRUE;
5767 LOP(OP_CRYPT,XTERM);
5770 LOP(OP_CHMOD,XTERM);
5773 LOP(OP_CHOWN,XTERM);
5776 LOP(OP_CONNECT,XTERM);
5795 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5796 if (orig_keyword == KEY_do) {
5805 PL_hints |= HINT_BLOCK_SCOPE;
5815 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5816 LOP(OP_DBMOPEN,XTERM);
5822 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5829 yylval.ival = CopLINE(PL_curcop);
5845 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5846 UNIBRACK(OP_ENTEREVAL);
5864 case KEY_endhostent:
5870 case KEY_endservent:
5873 case KEY_endprotoent:
5884 yylval.ival = CopLINE(PL_curcop);
5886 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5889 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5892 if ((PL_bufend - p) >= 3 &&
5893 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5895 else if ((PL_bufend - p) >= 4 &&
5896 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5899 if (isIDFIRST_lazy_if(p,UTF)) {
5900 p = scan_ident(p, PL_bufend,
5901 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5905 Perl_croak(aTHX_ "Missing $ on loop variable");
5907 s = SvPVX(PL_linestr) + soff;
5913 LOP(OP_FORMLINE,XTERM);
5919 LOP(OP_FCNTL,XTERM);
5925 LOP(OP_FLOCK,XTERM);
5934 LOP(OP_GREPSTART, XREF);
5937 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5952 case KEY_getpriority:
5953 LOP(OP_GETPRIORITY,XTERM);
5955 case KEY_getprotobyname:
5958 case KEY_getprotobynumber:
5959 LOP(OP_GPBYNUMBER,XTERM);
5961 case KEY_getprotoent:
5973 case KEY_getpeername:
5974 UNI(OP_GETPEERNAME);
5976 case KEY_gethostbyname:
5979 case KEY_gethostbyaddr:
5980 LOP(OP_GHBYADDR,XTERM);
5982 case KEY_gethostent:
5985 case KEY_getnetbyname:
5988 case KEY_getnetbyaddr:
5989 LOP(OP_GNBYADDR,XTERM);
5994 case KEY_getservbyname:
5995 LOP(OP_GSBYNAME,XTERM);
5997 case KEY_getservbyport:
5998 LOP(OP_GSBYPORT,XTERM);
6000 case KEY_getservent:
6003 case KEY_getsockname:
6004 UNI(OP_GETSOCKNAME);
6006 case KEY_getsockopt:
6007 LOP(OP_GSOCKOPT,XTERM);
6022 yylval.ival = CopLINE(PL_curcop);
6033 yylval.ival = CopLINE(PL_curcop);
6037 LOP(OP_INDEX,XTERM);
6043 LOP(OP_IOCTL,XTERM);
6055 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6087 LOP(OP_LISTEN,XTERM);
6096 s = scan_pat(s,OP_MATCH);
6097 TERM(sublex_start());
6100 LOP(OP_MAPSTART, XREF);
6103 LOP(OP_MKDIR,XTERM);
6106 LOP(OP_MSGCTL,XTERM);
6109 LOP(OP_MSGGET,XTERM);
6112 LOP(OP_MSGRCV,XTERM);
6115 LOP(OP_MSGSND,XTERM);
6122 if (isIDFIRST_lazy_if(s,UTF)) {
6126 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6127 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6129 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6130 if (!PL_in_my_stash) {
6133 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6137 if (PL_madskills) { /* just add type to declarator token */
6138 sv_catsv(PL_thistoken, PL_nextwhite);
6140 sv_catpvn(PL_thistoken, start, s - start);
6148 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6155 s = tokenize_use(0, s);
6159 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6166 if (isIDFIRST_lazy_if(s,UTF)) {
6168 for (d = s; isALNUM_lazy_if(d,UTF);)
6170 for (t=d; isSPACE(*t);)
6172 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6174 && !(t[0] == '=' && t[1] == '>')
6176 int parms_len = (int)(d-s);
6177 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6178 "Precedence problem: open %.*s should be open(%.*s)",
6179 parms_len, s, parms_len, s);
6185 yylval.ival = OP_OR;
6195 LOP(OP_OPEN_DIR,XTERM);
6198 checkcomma(s,PL_tokenbuf,"filehandle");
6202 checkcomma(s,PL_tokenbuf,"filehandle");
6221 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6225 LOP(OP_PIPE_OP,XTERM);
6228 s = scan_str(s,!!PL_madskills,FALSE);
6231 yylval.ival = OP_CONST;
6232 TERM(sublex_start());
6238 s = scan_str(s,!!PL_madskills,FALSE);
6241 PL_expect = XOPERATOR;
6243 if (SvCUR(PL_lex_stuff)) {
6246 d = SvPV_force(PL_lex_stuff, len);
6248 for (; isSPACE(*d) && len; --len, ++d)
6253 if (!warned && ckWARN(WARN_QW)) {
6254 for (; !isSPACE(*d) && len; --len, ++d) {
6256 Perl_warner(aTHX_ packWARN(WARN_QW),
6257 "Possible attempt to separate words with commas");
6260 else if (*d == '#') {
6261 Perl_warner(aTHX_ packWARN(WARN_QW),
6262 "Possible attempt to put comments in qw() list");
6268 for (; !isSPACE(*d) && len; --len, ++d)
6271 sv = newSVpvn(b, d-b);
6272 if (DO_UTF8(PL_lex_stuff))
6274 words = append_elem(OP_LIST, words,
6275 newSVOP(OP_CONST, 0, tokeq(sv)));
6279 start_force(PL_curforce);
6280 NEXTVAL_NEXTTOKE.opval = words;
6285 SvREFCNT_dec(PL_lex_stuff);
6286 PL_lex_stuff = NULL;
6292 s = scan_str(s,!!PL_madskills,FALSE);
6295 yylval.ival = OP_STRINGIFY;
6296 if (SvIVX(PL_lex_stuff) == '\'')
6297 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6298 TERM(sublex_start());
6301 s = scan_pat(s,OP_QR);
6302 TERM(sublex_start());
6305 s = scan_str(s,!!PL_madskills,FALSE);
6308 readpipe_override();
6309 TERM(sublex_start());
6317 s = force_version(s, FALSE);
6319 else if (*s != 'v' || !isDIGIT(s[1])
6320 || (s = force_version(s, TRUE), *s == 'v'))
6322 *PL_tokenbuf = '\0';
6323 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6324 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6325 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), TRUE);
6327 yyerror("<> should be quotes");
6329 if (orig_keyword == KEY_require) {
6337 PL_last_uni = PL_oldbufptr;
6338 PL_last_lop_op = OP_REQUIRE;
6340 return REPORT( (int)REQUIRE );
6346 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6350 LOP(OP_RENAME,XTERM);
6359 LOP(OP_RINDEX,XTERM);
6369 UNIDOR(OP_READLINE);
6382 LOP(OP_REVERSE,XTERM);
6385 UNIDOR(OP_READLINK);
6393 TERM(sublex_start());
6395 TOKEN(1); /* force error */
6398 checkcomma(s,PL_tokenbuf,"filehandle");
6408 LOP(OP_SELECT,XTERM);
6414 LOP(OP_SEMCTL,XTERM);
6417 LOP(OP_SEMGET,XTERM);
6420 LOP(OP_SEMOP,XTERM);
6426 LOP(OP_SETPGRP,XTERM);
6428 case KEY_setpriority:
6429 LOP(OP_SETPRIORITY,XTERM);
6431 case KEY_sethostent:
6437 case KEY_setservent:
6440 case KEY_setprotoent:
6450 LOP(OP_SEEKDIR,XTERM);
6452 case KEY_setsockopt:
6453 LOP(OP_SSOCKOPT,XTERM);
6459 LOP(OP_SHMCTL,XTERM);
6462 LOP(OP_SHMGET,XTERM);
6465 LOP(OP_SHMREAD,XTERM);
6468 LOP(OP_SHMWRITE,XTERM);
6471 LOP(OP_SHUTDOWN,XTERM);
6480 LOP(OP_SOCKET,XTERM);
6482 case KEY_socketpair:
6483 LOP(OP_SOCKPAIR,XTERM);
6486 checkcomma(s,PL_tokenbuf,"subroutine name");
6488 if (*s == ';' || *s == ')') /* probably a close */
6489 Perl_croak(aTHX_ "sort is now a reserved word");
6491 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6495 LOP(OP_SPLIT,XTERM);
6498 LOP(OP_SPRINTF,XTERM);
6501 LOP(OP_SPLICE,XTERM);
6516 LOP(OP_SUBSTR,XTERM);
6522 char tmpbuf[sizeof PL_tokenbuf];
6523 SSize_t tboffset = 0;
6524 expectation attrful;
6525 bool have_name, have_proto;
6526 const int key = tmp;
6531 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6532 SV *subtoken = newSVpvn(tstart, s - tstart);
6536 s = SKIPSPACE2(s,tmpwhite);
6541 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6542 (*s == ':' && s[1] == ':'))
6549 attrful = XATTRBLOCK;
6550 /* remember buffer pos'n for later force_word */
6551 tboffset = s - PL_oldbufptr;
6552 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6555 nametoke = newSVpvn(s, d - s);
6557 if (strchr(tmpbuf, ':'))
6558 sv_setpv(PL_subname, tmpbuf);
6560 sv_setsv(PL_subname,PL_curstname);
6561 sv_catpvs(PL_subname,"::");
6562 sv_catpvn(PL_subname,tmpbuf,len);
6569 CURMAD('X', nametoke);
6570 CURMAD('_', tmpwhite);
6571 (void) force_word(PL_oldbufptr + tboffset, WORD,
6574 s = SKIPSPACE2(d,tmpwhite);
6581 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6582 PL_expect = XTERMBLOCK;
6583 attrful = XATTRTERM;
6584 sv_setpvn(PL_subname,"?",1);
6588 if (key == KEY_format) {
6590 PL_lex_formbrack = PL_lex_brackets + 1;
6592 PL_thistoken = subtoken;
6596 (void) force_word(PL_oldbufptr + tboffset, WORD,
6602 /* Look for a prototype */
6605 bool bad_proto = FALSE;
6606 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6608 s = scan_str(s,!!PL_madskills,FALSE);
6610 Perl_croak(aTHX_ "Prototype not terminated");
6611 /* strip spaces and check for bad characters */
6612 d = SvPVX(PL_lex_stuff);
6614 for (p = d; *p; ++p) {
6617 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6623 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6624 "Illegal character in prototype for %"SVf" : %s",
6625 (void*)PL_subname, d);
6626 SvCUR_set(PL_lex_stuff, tmp);
6631 CURMAD('q', PL_thisopen);
6632 CURMAD('_', tmpwhite);
6633 CURMAD('=', PL_thisstuff);
6634 CURMAD('Q', PL_thisclose);
6635 NEXTVAL_NEXTTOKE.opval =
6636 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6637 PL_lex_stuff = Nullsv;
6640 s = SKIPSPACE2(s,tmpwhite);
6648 if (*s == ':' && s[1] != ':')
6649 PL_expect = attrful;
6650 else if (*s != '{' && key == KEY_sub) {
6652 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6654 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, (void*)PL_subname);
6661 curmad('^', newSVpvs(""));
6662 CURMAD('_', tmpwhite);
6666 PL_thistoken = subtoken;
6669 NEXTVAL_NEXTTOKE.opval =
6670 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6671 PL_lex_stuff = NULL;
6676 sv_setpv(PL_subname,
6678 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6682 (void) force_word(PL_oldbufptr + tboffset, WORD,
6692 LOP(OP_SYSTEM,XREF);
6695 LOP(OP_SYMLINK,XTERM);
6698 LOP(OP_SYSCALL,XTERM);
6701 LOP(OP_SYSOPEN,XTERM);
6704 LOP(OP_SYSSEEK,XTERM);
6707 LOP(OP_SYSREAD,XTERM);
6710 LOP(OP_SYSWRITE,XTERM);
6714 TERM(sublex_start());
6735 LOP(OP_TRUNCATE,XTERM);
6747 yylval.ival = CopLINE(PL_curcop);
6751 yylval.ival = CopLINE(PL_curcop);
6755 LOP(OP_UNLINK,XTERM);
6761 LOP(OP_UNPACK,XTERM);
6764 LOP(OP_UTIME,XTERM);
6770 LOP(OP_UNSHIFT,XTERM);
6773 s = tokenize_use(1, s);
6783 yylval.ival = CopLINE(PL_curcop);
6787 yylval.ival = CopLINE(PL_curcop);
6791 PL_hints |= HINT_BLOCK_SCOPE;
6798 LOP(OP_WAITPID,XTERM);
6807 ctl_l[0] = toCTRL('L');
6809 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6812 /* Make sure $^L is defined */
6813 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6818 if (PL_expect == XOPERATOR)
6824 yylval.ival = OP_XOR;
6829 TERM(sublex_start());
6834 #pragma segment Main
6838 S_pending_ident(pTHX)
6843 /* pit holds the identifier we read and pending_ident is reset */
6844 char pit = PL_pending_ident;
6845 PL_pending_ident = 0;
6847 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6848 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6849 "### Pending identifier '%s'\n", PL_tokenbuf); });
6851 /* if we're in a my(), we can't allow dynamics here.
6852 $foo'bar has already been turned into $foo::bar, so
6853 just check for colons.
6855 if it's a legal name, the OP is a PADANY.
6858 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6859 if (strchr(PL_tokenbuf,':'))
6860 yyerror(Perl_form(aTHX_ "No package name allowed for "
6861 "variable %s in \"our\"",
6863 tmp = allocmy(PL_tokenbuf);
6866 if (strchr(PL_tokenbuf,':'))
6867 yyerror(Perl_form(aTHX_ PL_no_myglob,
6868 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6870 yylval.opval = newOP(OP_PADANY, 0);
6871 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6877 build the ops for accesses to a my() variable.
6879 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6880 then used in a comparison. This catches most, but not
6881 all cases. For instance, it catches
6882 sort { my($a); $a <=> $b }
6884 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6885 (although why you'd do that is anyone's guess).
6888 if (!strchr(PL_tokenbuf,':')) {
6890 tmp = pad_findmy(PL_tokenbuf);
6891 if (tmp != NOT_IN_PAD) {
6892 /* might be an "our" variable" */
6893 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6894 /* build ops for a bareword */
6895 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6896 HEK * const stashname = HvNAME_HEK(stash);
6897 SV * const sym = newSVhek(stashname);
6898 sv_catpvs(sym, "::");
6899 sv_catpv(sym, PL_tokenbuf+1);
6900 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6901 yylval.opval->op_private = OPpCONST_ENTERED;
6904 ? (GV_ADDMULTI | GV_ADDINEVAL)
6907 ((PL_tokenbuf[0] == '$') ? SVt_PV
6908 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6913 /* if it's a sort block and they're naming $a or $b */
6914 if (PL_last_lop_op == OP_SORT &&
6915 PL_tokenbuf[0] == '$' &&
6916 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6919 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6920 d < PL_bufend && *d != '\n';
6923 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6924 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6930 yylval.opval = newOP(OP_PADANY, 0);
6931 yylval.opval->op_targ = tmp;
6937 Whine if they've said @foo in a doublequoted string,
6938 and @foo isn't a variable we can find in the symbol
6941 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
6942 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
6943 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
6944 && ckWARN(WARN_AMBIGUOUS))
6946 /* Downgraded from fatal to warning 20000522 mjd */
6947 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6948 "Possible unintended interpolation of %s in string",
6953 /* build ops for a bareword */
6954 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
6955 yylval.opval->op_private = OPpCONST_ENTERED;
6958 /* If the identifier refers to a stash, don't autovivify it.
6959 * Change 24660 had the side effect of causing symbol table
6960 * hashes to always be defined, even if they were freshly
6961 * created and the only reference in the entire program was
6962 * the single statement with the defined %foo::bar:: test.
6963 * It appears that all code in the wild doing this actually
6964 * wants to know whether sub-packages have been loaded, so
6965 * by avoiding auto-vivifying symbol tables, we ensure that
6966 * defined %foo::bar:: continues to be false, and the existing
6967 * tests still give the expected answers, even though what
6968 * they're actually testing has now changed subtly.
6970 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
6972 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
6973 ((PL_tokenbuf[0] == '$') ? SVt_PV
6974 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6980 * The following code was generated by perl_keyword.pl.
6984 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
6989 case 1: /* 5 tokens of length 1 */
7021 case 2: /* 18 tokens of length 2 */
7167 case 3: /* 29 tokens of length 3 */
7171 if (name[1] == 'N' &&
7234 if (name[1] == 'i' &&
7256 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7274 if (name[1] == 'o' &&
7283 if (name[1] == 'e' &&
7292 if (name[1] == 'n' &&
7301 if (name[1] == 'o' &&
7310 if (name[1] == 'a' &&
7319 if (name[1] == 'o' &&
7381 if (name[1] == 'e' &&
7395 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7421 if (name[1] == 'i' &&
7430 if (name[1] == 's' &&
7439 if (name[1] == 'e' &&
7448 if (name[1] == 'o' &&
7460 case 4: /* 41 tokens of length 4 */
7464 if (name[1] == 'O' &&
7474 if (name[1] == 'N' &&
7484 if (name[1] == 'i' &&
7494 if (name[1] == 'h' &&
7504 if (name[1] == 'u' &&
7517 if (name[2] == 'c' &&
7526 if (name[2] == 's' &&
7535 if (name[2] == 'a' &&
7571 if (name[1] == 'o' &&
7584 if (name[2] == 't' &&
7593 if (name[2] == 'o' &&
7602 if (name[2] == 't' &&
7611 if (name[2] == 'e' &&
7624 if (name[1] == 'o' &&
7637 if (name[2] == 'y' &&
7646 if (name[2] == 'l' &&
7662 if (name[2] == 's' &&
7671 if (name[2] == 'n' &&
7680 if (name[2] == 'c' &&
7693 if (name[1] == 'e' &&
7703 if (name[1] == 'p' &&
7716 if (name[2] == 'c' &&
7725 if (name[2] == 'p' &&
7734 if (name[2] == 's' &&
7750 if (name[2] == 'n' &&
7820 if (name[2] == 'r' &&
7829 if (name[2] == 'r' &&
7838 if (name[2] == 'a' &&
7854 if (name[2] == 'l' &&
7916 if (name[2] == 'e' &&
7919 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7932 case 5: /* 39 tokens of length 5 */
7936 if (name[1] == 'E' &&
7947 if (name[1] == 'H' &&
7961 if (name[2] == 'a' &&
7971 if (name[2] == 'a' &&
7988 if (name[2] == 'e' &&
7998 if (name[2] == 'e' &&
8002 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8018 if (name[3] == 'i' &&
8027 if (name[3] == 'o' &&
8063 if (name[2] == 'o' &&
8073 if (name[2] == 'y' &&
8087 if (name[1] == 'l' &&
8101 if (name[2] == 'n' &&
8111 if (name[2] == 'o' &&
8125 if (name[1] == 'i' &&
8130 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8139 if (name[2] == 'd' &&
8149 if (name[2] == 'c' &&
8166 if (name[2] == 'c' &&
8176 if (name[2] == 't' &&
8190 if (name[1] == 'k' &&
8201 if (name[1] == 'r' &&
8215 if (name[2] == 's' &&
8225 if (name[2] == 'd' &&
8242 if (name[2] == 'm' &&
8252 if (name[2] == 'i' &&
8262 if (name[2] == 'e' &&
8272 if (name[2] == 'l' &&
8282 if (name[2] == 'a' &&
8295 if (name[3] == 't' &&
8298 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8304 if (name[3] == 'd' &&
8321 if (name[1] == 'i' &&
8335 if (name[2] == 'a' &&
8348 if (name[3] == 'e' &&
8383 if (name[2] == 'i' &&
8400 if (name[2] == 'i' &&
8410 if (name[2] == 'i' &&
8427 case 6: /* 33 tokens of length 6 */
8431 if (name[1] == 'c' &&
8446 if (name[2] == 'l' &&
8457 if (name[2] == 'r' &&
8472 if (name[1] == 'e' &&
8487 if (name[2] == 's' &&
8492 if(ckWARN_d(WARN_SYNTAX))
8493 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8499 if (name[2] == 'i' &&
8517 if (name[2] == 'l' &&
8528 if (name[2] == 'r' &&
8543 if (name[1] == 'm' &&
8558 if (name[2] == 'n' &&
8569 if (name[2] == 's' &&
8584 if (name[1] == 's' &&
8590 if (name[4] == 't' &&
8599 if (name[4] == 'e' &&
8608 if (name[4] == 'c' &&
8617 if (name[4] == 'n' &&
8633 if (name[1] == 'r' &&
8651 if (name[3] == 'a' &&
8661 if (name[3] == 'u' &&
8675 if (name[2] == 'n' &&
8693 if (name[2] == 'a' &&
8707 if (name[3] == 'e' &&
8720 if (name[4] == 't' &&
8729 if (name[4] == 'e' &&
8751 if (name[4] == 't' &&
8760 if (name[4] == 'e' &&
8776 if (name[2] == 'c' &&
8787 if (name[2] == 'l' &&
8798 if (name[2] == 'b' &&
8809 if (name[2] == 's' &&
8832 if (name[4] == 's' &&
8841 if (name[4] == 'n' &&
8854 if (name[3] == 'a' &&
8871 if (name[1] == 'a' &&
8886 case 7: /* 29 tokens of length 7 */
8890 if (name[1] == 'E' &&
8903 if (name[1] == '_' &&
8916 if (name[1] == 'i' &&
8923 return -KEY_binmode;
8929 if (name[1] == 'o' &&
8936 return -KEY_connect;
8945 if (name[2] == 'm' &&
8951 return -KEY_dbmopen;
8962 if (name[4] == 'u' &&
8966 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
8972 if (name[4] == 'n' &&
8993 if (name[1] == 'o' &&
9006 if (name[1] == 'e' &&
9013 if (name[5] == 'r' &&
9016 return -KEY_getpgrp;
9022 if (name[5] == 'i' &&
9025 return -KEY_getppid;
9038 if (name[1] == 'c' &&
9045 return -KEY_lcfirst;
9051 if (name[1] == 'p' &&
9058 return -KEY_opendir;
9064 if (name[1] == 'a' &&
9082 if (name[3] == 'd' &&
9087 return -KEY_readdir;
9093 if (name[3] == 'u' &&
9104 if (name[3] == 'e' &&
9109 return -KEY_reverse;
9128 if (name[3] == 'k' &&
9133 return -KEY_seekdir;
9139 if (name[3] == 'p' &&
9144 return -KEY_setpgrp;
9154 if (name[2] == 'm' &&
9160 return -KEY_shmread;
9166 if (name[2] == 'r' &&
9172 return -KEY_sprintf;
9181 if (name[3] == 'l' &&
9186 return -KEY_symlink;
9195 if (name[4] == 'a' &&
9199 return -KEY_syscall;
9205 if (name[4] == 'p' &&
9209 return -KEY_sysopen;
9215 if (name[4] == 'e' &&
9219 return -KEY_sysread;
9225 if (name[4] == 'e' &&
9229 return -KEY_sysseek;
9247 if (name[1] == 'e' &&
9254 return -KEY_telldir;
9263 if (name[2] == 'f' &&
9269 return -KEY_ucfirst;
9275 if (name[2] == 's' &&
9281 return -KEY_unshift;
9291 if (name[1] == 'a' &&
9298 return -KEY_waitpid;
9307 case 8: /* 26 tokens of length 8 */
9311 if (name[1] == 'U' &&
9319 return KEY_AUTOLOAD;
9330 if (name[3] == 'A' &&
9336 return KEY___DATA__;
9342 if (name[3] == 'I' &&
9348 return -KEY___FILE__;
9354 if (name[3] == 'I' &&
9360 return -KEY___LINE__;
9376 if (name[2] == 'o' &&
9383 return -KEY_closedir;
9389 if (name[2] == 'n' &&
9396 return -KEY_continue;
9406 if (name[1] == 'b' &&
9414 return -KEY_dbmclose;
9420 if (name[1] == 'n' &&
9426 if (name[4] == 'r' &&
9431 return -KEY_endgrent;
9437 if (name[4] == 'w' &&
9442 return -KEY_endpwent;
9455 if (name[1] == 'o' &&
9463 return -KEY_formline;
9469 if (name[1] == 'e' &&
9480 if (name[6] == 'n' &&
9483 return -KEY_getgrent;
9489 if (name[6] == 'i' &&
9492 return -KEY_getgrgid;
9498 if (name[6] == 'a' &&
9501 return -KEY_getgrnam;
9514 if (name[4] == 'o' &&
9519 return -KEY_getlogin;
9530 if (name[6] == 'n' &&
9533 return -KEY_getpwent;
9539 if (name[6] == 'a' &&
9542 return -KEY_getpwnam;
9548 if (name[6] == 'i' &&
9551 return -KEY_getpwuid;
9571 if (name[1] == 'e' &&
9578 if (name[5] == 'i' &&
9585 return -KEY_readline;
9590 return -KEY_readlink;
9601 if (name[5] == 'i' &&
9605 return -KEY_readpipe;
9626 if (name[4] == 'r' &&
9631 return -KEY_setgrent;
9637 if (name[4] == 'w' &&
9642 return -KEY_setpwent;
9658 if (name[3] == 'w' &&
9664 return -KEY_shmwrite;
9670 if (name[3] == 't' &&
9676 return -KEY_shutdown;
9686 if (name[2] == 's' &&
9693 return -KEY_syswrite;
9703 if (name[1] == 'r' &&
9711 return -KEY_truncate;
9720 case 9: /* 9 tokens of length 9 */
9724 if (name[1] == 'N' &&
9733 return KEY_UNITCHECK;
9739 if (name[1] == 'n' &&
9748 return -KEY_endnetent;
9754 if (name[1] == 'e' &&
9763 return -KEY_getnetent;
9769 if (name[1] == 'o' &&
9778 return -KEY_localtime;
9784 if (name[1] == 'r' &&
9793 return KEY_prototype;
9799 if (name[1] == 'u' &&
9808 return -KEY_quotemeta;
9814 if (name[1] == 'e' &&
9823 return -KEY_rewinddir;
9829 if (name[1] == 'e' &&
9838 return -KEY_setnetent;
9844 if (name[1] == 'a' &&
9853 return -KEY_wantarray;
9862 case 10: /* 9 tokens of length 10 */
9866 if (name[1] == 'n' &&
9872 if (name[4] == 'o' &&
9879 return -KEY_endhostent;
9885 if (name[4] == 'e' &&
9892 return -KEY_endservent;
9905 if (name[1] == 'e' &&
9911 if (name[4] == 'o' &&
9918 return -KEY_gethostent;
9927 if (name[5] == 'r' &&
9933 return -KEY_getservent;
9939 if (name[5] == 'c' &&
9945 return -KEY_getsockopt;
9970 if (name[4] == 'o' &&
9977 return -KEY_sethostent;
9986 if (name[5] == 'r' &&
9992 return -KEY_setservent;
9998 if (name[5] == 'c' &&
10004 return -KEY_setsockopt;
10021 if (name[2] == 'c' &&
10030 return -KEY_socketpair;
10043 case 11: /* 8 tokens of length 11 */
10047 if (name[1] == '_' &&
10057 { /* __PACKAGE__ */
10058 return -KEY___PACKAGE__;
10064 if (name[1] == 'n' &&
10074 { /* endprotoent */
10075 return -KEY_endprotoent;
10081 if (name[1] == 'e' &&
10090 if (name[5] == 'e' &&
10096 { /* getpeername */
10097 return -KEY_getpeername;
10106 if (name[6] == 'o' &&
10111 { /* getpriority */
10112 return -KEY_getpriority;
10118 if (name[6] == 't' &&
10123 { /* getprotoent */
10124 return -KEY_getprotoent;
10138 if (name[4] == 'o' &&
10145 { /* getsockname */
10146 return -KEY_getsockname;
10159 if (name[1] == 'e' &&
10167 if (name[6] == 'o' &&
10172 { /* setpriority */
10173 return -KEY_setpriority;
10179 if (name[6] == 't' &&
10184 { /* setprotoent */
10185 return -KEY_setprotoent;
10201 case 12: /* 2 tokens of length 12 */
10202 if (name[0] == 'g' &&
10214 if (name[9] == 'd' &&
10217 { /* getnetbyaddr */
10218 return -KEY_getnetbyaddr;
10224 if (name[9] == 'a' &&
10227 { /* getnetbyname */
10228 return -KEY_getnetbyname;
10240 case 13: /* 4 tokens of length 13 */
10241 if (name[0] == 'g' &&
10248 if (name[4] == 'o' &&
10257 if (name[10] == 'd' &&
10260 { /* gethostbyaddr */
10261 return -KEY_gethostbyaddr;
10267 if (name[10] == 'a' &&
10270 { /* gethostbyname */
10271 return -KEY_gethostbyname;
10284 if (name[4] == 'e' &&
10293 if (name[10] == 'a' &&
10296 { /* getservbyname */
10297 return -KEY_getservbyname;
10303 if (name[10] == 'o' &&
10306 { /* getservbyport */
10307 return -KEY_getservbyport;
10326 case 14: /* 1 tokens of length 14 */
10327 if (name[0] == 'g' &&
10341 { /* getprotobyname */
10342 return -KEY_getprotobyname;
10347 case 16: /* 1 tokens of length 16 */
10348 if (name[0] == 'g' &&
10364 { /* getprotobynumber */
10365 return -KEY_getprotobynumber;
10379 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10383 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10384 if (ckWARN(WARN_SYNTAX)) {
10387 for (w = s+2; *w && level; w++) {
10390 else if (*w == ')')
10393 while (isSPACE(*w))
10395 if (!*w || !strchr(";|})]oaiuw!=", *w)) /* an advisory hack only... */
10396 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10397 "%s (...) interpreted as function",name);
10400 while (s < PL_bufend && isSPACE(*s))
10404 while (s < PL_bufend && isSPACE(*s))
10406 if (isIDFIRST_lazy_if(s,UTF)) {
10407 const char * const w = s++;
10408 while (isALNUM_lazy_if(s,UTF))
10410 while (s < PL_bufend && isSPACE(*s))
10414 if (keyword(w, s - w, 0))
10417 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10418 if (gv && GvCVu(gv))
10420 Perl_croak(aTHX_ "No comma allowed after %s", what);
10425 /* Either returns sv, or mortalizes sv and returns a new SV*.
10426 Best used as sv=new_constant(..., sv, ...).
10427 If s, pv are NULL, calls subroutine with one argument,
10428 and type is used with error messages only. */
10431 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10435 HV * const table = GvHV(PL_hintgv); /* ^H */
10439 const char *why1 = "", *why2 = "", *why3 = "";
10441 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10444 why2 = (const char *)
10445 (strEQ(key,"charnames")
10446 ? "(possibly a missing \"use charnames ...\")"
10448 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10449 (type ? type: "undef"), why2);
10451 /* This is convoluted and evil ("goto considered harmful")
10452 * but I do not understand the intricacies of all the different
10453 * failure modes of %^H in here. The goal here is to make
10454 * the most probable error message user-friendly. --jhi */
10459 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10460 (type ? type: "undef"), why1, why2, why3);
10462 yyerror(SvPVX_const(msg));
10466 cvp = hv_fetch(table, key, strlen(key), FALSE);
10467 if (!cvp || !SvOK(*cvp)) {
10470 why3 = "} is not defined";
10473 sv_2mortal(sv); /* Parent created it permanently */
10476 pv = sv_2mortal(newSVpvn(s, len));
10478 typesv = sv_2mortal(newSVpv(type, 0));
10480 typesv = &PL_sv_undef;
10482 PUSHSTACKi(PERLSI_OVERLOAD);
10494 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10498 /* Check the eval first */
10499 if (!PL_in_eval && SvTRUE(ERRSV)) {
10500 sv_catpvs(ERRSV, "Propagated");
10501 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10503 res = SvREFCNT_inc_simple(sv);
10507 SvREFCNT_inc_simple_void(res);
10516 why1 = "Call to &{$^H{";
10518 why3 = "}} did not return a defined value";
10526 /* Returns a NUL terminated string, with the length of the string written to
10530 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10533 register char *d = dest;
10534 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10537 Perl_croak(aTHX_ ident_too_long);
10538 if (isALNUM(*s)) /* UTF handled below */
10540 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10545 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10549 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10550 char *t = s + UTF8SKIP(s);
10552 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10556 Perl_croak(aTHX_ ident_too_long);
10557 Copy(s, d, len, char);
10570 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10573 char *bracket = NULL;
10575 register char *d = dest;
10576 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10581 while (isDIGIT(*s)) {
10583 Perl_croak(aTHX_ ident_too_long);
10590 Perl_croak(aTHX_ ident_too_long);
10591 if (isALNUM(*s)) /* UTF handled below */
10593 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10598 else if (*s == ':' && s[1] == ':') {
10602 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10603 char *t = s + UTF8SKIP(s);
10604 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10606 if (d + (t - s) > e)
10607 Perl_croak(aTHX_ ident_too_long);
10608 Copy(s, d, t - s, char);
10619 if (PL_lex_state != LEX_NORMAL)
10620 PL_lex_state = LEX_INTERPENDMAYBE;
10623 if (*s == '$' && s[1] &&
10624 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10637 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10642 if (isSPACE(s[-1])) {
10644 const char ch = *s++;
10645 if (!SPACE_OR_TAB(ch)) {
10651 if (isIDFIRST_lazy_if(d,UTF)) {
10655 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10656 end += UTF8SKIP(end);
10657 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10658 end += UTF8SKIP(end);
10660 Copy(s, d, end - s, char);
10665 while ((isALNUM(*s) || *s == ':') && d < e)
10668 Perl_croak(aTHX_ ident_too_long);
10671 while (s < send && SPACE_OR_TAB(*s))
10673 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10674 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10675 const char * const brack =
10677 ((*s == '[') ? "[...]" : "{...}");
10678 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10679 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10680 funny, dest, brack, funny, dest, brack);
10683 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10687 /* Handle extended ${^Foo} variables
10688 * 1999-02-27 mjd-perl-patch@plover.com */
10689 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10693 while (isALNUM(*s) && d < e) {
10697 Perl_croak(aTHX_ ident_too_long);
10702 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10703 PL_lex_state = LEX_INTERPEND;
10706 if (PL_lex_state == LEX_NORMAL) {
10707 if (ckWARN(WARN_AMBIGUOUS) &&
10708 (keyword(dest, d - dest, 0) || get_cv(dest, FALSE)))
10712 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10713 "Ambiguous use of %c{%s} resolved to %c%s",
10714 funny, dest, funny, dest);
10719 s = bracket; /* let the parser handle it */
10723 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10724 PL_lex_state = LEX_INTERPEND;
10729 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10731 PERL_UNUSED_CONTEXT;
10734 else if (ch == 'g')
10735 *pmfl |= PMf_GLOBAL;
10736 else if (ch == 'c')
10737 *pmfl |= PMf_CONTINUE;
10738 else if (ch == 'o')
10740 else if (ch == 'm')
10741 *pmfl |= PMf_MULTILINE;
10742 else if (ch == 's')
10743 *pmfl |= PMf_SINGLELINE;
10744 else if (ch == 'x')
10745 *pmfl |= PMf_EXTENDED;
10749 S_scan_pat(pTHX_ char *start, I32 type)
10753 char *s = scan_str(start,!!PL_madskills,FALSE);
10754 const char * const valid_flags =
10755 (const char *)((type == OP_QR) ? "iomsx" : "iogcmsx");
10762 const char * const delimiter = skipspace(start);
10766 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10767 : "Search pattern not terminated" ));
10770 pm = (PMOP*)newPMOP(type, 0);
10771 if (PL_multi_open == '?')
10772 pm->op_pmflags |= PMf_ONCE;
10776 while (*s && strchr(valid_flags, *s))
10777 pmflag(&pm->op_pmflags,*s++);
10779 if (PL_madskills && modstart != s) {
10780 SV* tmptoken = newSVpvn(modstart, s - modstart);
10781 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10784 /* issue a warning if /c is specified,but /g is not */
10785 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10786 && ckWARN(WARN_REGEXP))
10788 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless without /g" );
10791 pm->op_pmpermflags = pm->op_pmflags;
10793 PL_lex_op = (OP*)pm;
10794 yylval.ival = OP_MATCH;
10799 S_scan_subst(pTHX_ char *start)
10810 yylval.ival = OP_NULL;
10812 s = scan_str(start,!!PL_madskills,FALSE);
10815 Perl_croak(aTHX_ "Substitution pattern not terminated");
10817 if (s[-1] == PL_multi_open)
10820 if (PL_madskills) {
10821 CURMAD('q', PL_thisopen);
10822 CURMAD('_', PL_thiswhite);
10823 CURMAD('E', PL_thisstuff);
10824 CURMAD('Q', PL_thisclose);
10825 PL_realtokenstart = s - SvPVX(PL_linestr);
10829 first_start = PL_multi_start;
10830 s = scan_str(s,!!PL_madskills,FALSE);
10832 if (PL_lex_stuff) {
10833 SvREFCNT_dec(PL_lex_stuff);
10834 PL_lex_stuff = NULL;
10836 Perl_croak(aTHX_ "Substitution replacement not terminated");
10838 PL_multi_start = first_start; /* so whole substitution is taken together */
10840 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10843 if (PL_madskills) {
10844 CURMAD('z', PL_thisopen);
10845 CURMAD('R', PL_thisstuff);
10846 CURMAD('Z', PL_thisclose);
10856 else if (strchr("iogcmsx", *s))
10857 pmflag(&pm->op_pmflags,*s++);
10863 if (PL_madskills) {
10865 curmad('m', newSVpvn(modstart, s - modstart));
10866 append_madprops(PL_thismad, (OP*)pm, 0);
10870 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10871 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10875 SV * const repl = newSVpvs("");
10877 PL_sublex_info.super_bufptr = s;
10878 PL_sublex_info.super_bufend = PL_bufend;
10880 pm->op_pmflags |= PMf_EVAL;
10882 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10883 sv_catpvs(repl, "{");
10884 sv_catsv(repl, PL_lex_repl);
10885 if (strchr(SvPVX(PL_lex_repl), '#'))
10886 sv_catpvs(repl, "\n");
10887 sv_catpvs(repl, "}");
10889 SvREFCNT_dec(PL_lex_repl);
10890 PL_lex_repl = repl;
10893 pm->op_pmpermflags = pm->op_pmflags;
10894 PL_lex_op = (OP*)pm;
10895 yylval.ival = OP_SUBST;
10900 S_scan_trans(pTHX_ char *start)
10913 yylval.ival = OP_NULL;
10915 s = scan_str(start,!!PL_madskills,FALSE);
10917 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10919 if (s[-1] == PL_multi_open)
10922 if (PL_madskills) {
10923 CURMAD('q', PL_thisopen);
10924 CURMAD('_', PL_thiswhite);
10925 CURMAD('E', PL_thisstuff);
10926 CURMAD('Q', PL_thisclose);
10927 PL_realtokenstart = s - SvPVX(PL_linestr);
10931 s = scan_str(s,!!PL_madskills,FALSE);
10933 if (PL_lex_stuff) {
10934 SvREFCNT_dec(PL_lex_stuff);
10935 PL_lex_stuff = NULL;
10937 Perl_croak(aTHX_ "Transliteration replacement not terminated");
10939 if (PL_madskills) {
10940 CURMAD('z', PL_thisopen);
10941 CURMAD('R', PL_thisstuff);
10942 CURMAD('Z', PL_thisclose);
10945 complement = del = squash = 0;
10952 complement = OPpTRANS_COMPLEMENT;
10955 del = OPpTRANS_DELETE;
10958 squash = OPpTRANS_SQUASH;
10967 Newx(tbl, complement&&!del?258:256, short);
10968 o = newPVOP(OP_TRANS, 0, (char*)tbl);
10969 o->op_private &= ~OPpTRANS_ALL;
10970 o->op_private |= del|squash|complement|
10971 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
10972 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
10975 yylval.ival = OP_TRANS;
10978 if (PL_madskills) {
10980 curmad('m', newSVpvn(modstart, s - modstart));
10981 append_madprops(PL_thismad, o, 0);
10990 S_scan_heredoc(pTHX_ register char *s)
10994 I32 op_type = OP_SCALAR;
10998 const char *found_newline;
11002 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11004 I32 stuffstart = s - SvPVX(PL_linestr);
11007 PL_realtokenstart = -1;
11012 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11016 while (SPACE_OR_TAB(*peek))
11018 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11021 s = delimcpy(d, e, s, PL_bufend, term, &len);
11031 if (!isALNUM_lazy_if(s,UTF))
11032 deprecate_old("bare << to mean <<\"\"");
11033 for (; isALNUM_lazy_if(s,UTF); s++) {
11038 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11039 Perl_croak(aTHX_ "Delimiter for here document is too long");
11042 len = d - PL_tokenbuf;
11045 if (PL_madskills) {
11046 tstart = PL_tokenbuf + !outer;
11047 PL_thisclose = newSVpvn(tstart, len - !outer);
11048 tstart = SvPVX(PL_linestr) + stuffstart;
11049 PL_thisopen = newSVpvn(tstart, s - tstart);
11050 stuffstart = s - SvPVX(PL_linestr);
11053 #ifndef PERL_STRICT_CR
11054 d = strchr(s, '\r');
11056 char * const olds = s;
11058 while (s < PL_bufend) {
11064 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11073 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11080 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11081 herewas = newSVpvn(s,PL_bufend-s);
11085 herewas = newSVpvn(s-1,found_newline-s+1);
11088 herewas = newSVpvn(s,found_newline-s);
11092 if (PL_madskills) {
11093 tstart = SvPVX(PL_linestr) + stuffstart;
11095 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11097 PL_thisstuff = newSVpvn(tstart, s - tstart);
11100 s += SvCUR(herewas);
11103 stuffstart = s - SvPVX(PL_linestr);
11109 tmpstr = newSV(79);
11110 sv_upgrade(tmpstr, SVt_PVIV);
11111 if (term == '\'') {
11112 op_type = OP_CONST;
11113 SvIV_set(tmpstr, -1);
11115 else if (term == '`') {
11116 op_type = OP_BACKTICK;
11117 SvIV_set(tmpstr, '\\');
11121 PL_multi_start = CopLINE(PL_curcop);
11122 PL_multi_open = PL_multi_close = '<';
11123 term = *PL_tokenbuf;
11124 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11125 char * const bufptr = PL_sublex_info.super_bufptr;
11126 char * const bufend = PL_sublex_info.super_bufend;
11127 char * const olds = s - SvCUR(herewas);
11128 s = strchr(bufptr, '\n');
11132 while (s < bufend &&
11133 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11135 CopLINE_inc(PL_curcop);
11138 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11139 missingterm(PL_tokenbuf);
11141 sv_setpvn(herewas,bufptr,d-bufptr+1);
11142 sv_setpvn(tmpstr,d+1,s-d);
11144 sv_catpvn(herewas,s,bufend-s);
11145 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11152 while (s < PL_bufend &&
11153 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11155 CopLINE_inc(PL_curcop);
11157 if (s >= PL_bufend) {
11158 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11159 missingterm(PL_tokenbuf);
11161 sv_setpvn(tmpstr,d+1,s-d);
11163 if (PL_madskills) {
11165 sv_catpvn(PL_thisstuff, d + 1, s - d);
11167 PL_thisstuff = newSVpvn(d + 1, s - d);
11168 stuffstart = s - SvPVX(PL_linestr);
11172 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11174 sv_catpvn(herewas,s,PL_bufend-s);
11175 sv_setsv(PL_linestr,herewas);
11176 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11177 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11178 PL_last_lop = PL_last_uni = NULL;
11181 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11182 while (s >= PL_bufend) { /* multiple line string? */
11184 if (PL_madskills) {
11185 tstart = SvPVX(PL_linestr) + stuffstart;
11187 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11189 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11193 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11194 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11195 missingterm(PL_tokenbuf);
11198 stuffstart = s - SvPVX(PL_linestr);
11200 CopLINE_inc(PL_curcop);
11201 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11202 PL_last_lop = PL_last_uni = NULL;
11203 #ifndef PERL_STRICT_CR
11204 if (PL_bufend - PL_linestart >= 2) {
11205 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11206 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11208 PL_bufend[-2] = '\n';
11210 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11212 else if (PL_bufend[-1] == '\r')
11213 PL_bufend[-1] = '\n';
11215 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11216 PL_bufend[-1] = '\n';
11218 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11219 SV * const sv = newSV(0);
11221 sv_upgrade(sv, SVt_PVMG);
11222 sv_setsv(sv,PL_linestr);
11223 (void)SvIOK_on(sv);
11225 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop),sv);
11227 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11228 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11229 *(SvPVX(PL_linestr) + off ) = ' ';
11230 sv_catsv(PL_linestr,herewas);
11231 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11232 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11236 sv_catsv(tmpstr,PL_linestr);
11241 PL_multi_end = CopLINE(PL_curcop);
11242 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11243 SvPV_shrink_to_cur(tmpstr);
11245 SvREFCNT_dec(herewas);
11247 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11249 else if (PL_encoding)
11250 sv_recode_to_utf8(tmpstr, PL_encoding);
11252 PL_lex_stuff = tmpstr;
11253 yylval.ival = op_type;
11257 /* scan_inputsymbol
11258 takes: current position in input buffer
11259 returns: new position in input buffer
11260 side-effects: yylval and lex_op are set.
11265 <FH> read from filehandle
11266 <pkg::FH> read from package qualified filehandle
11267 <pkg'FH> read from package qualified filehandle
11268 <$fh> read from filehandle in $fh
11269 <*.h> filename glob
11274 S_scan_inputsymbol(pTHX_ char *start)
11277 register char *s = start; /* current position in buffer */
11281 char *d = PL_tokenbuf; /* start of temp holding space */
11282 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11284 end = strchr(s, '\n');
11287 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11289 /* die if we didn't have space for the contents of the <>,
11290 or if it didn't end, or if we see a newline
11293 if (len >= (I32)sizeof PL_tokenbuf)
11294 Perl_croak(aTHX_ "Excessively long <> operator");
11296 Perl_croak(aTHX_ "Unterminated <> operator");
11301 Remember, only scalar variables are interpreted as filehandles by
11302 this code. Anything more complex (e.g., <$fh{$num}>) will be
11303 treated as a glob() call.
11304 This code makes use of the fact that except for the $ at the front,
11305 a scalar variable and a filehandle look the same.
11307 if (*d == '$' && d[1]) d++;
11309 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11310 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11313 /* If we've tried to read what we allow filehandles to look like, and
11314 there's still text left, then it must be a glob() and not a getline.
11315 Use scan_str to pull out the stuff between the <> and treat it
11316 as nothing more than a string.
11319 if (d - PL_tokenbuf != len) {
11320 yylval.ival = OP_GLOB;
11322 s = scan_str(start,!!PL_madskills,FALSE);
11324 Perl_croak(aTHX_ "Glob not terminated");
11328 bool readline_overriden = FALSE;
11331 /* we're in a filehandle read situation */
11334 /* turn <> into <ARGV> */
11336 Copy("ARGV",d,5,char);
11338 /* Check whether readline() is overriden */
11339 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11341 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11343 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11344 && (gv_readline = *gvp) != (GV*)&PL_sv_undef
11345 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11346 readline_overriden = TRUE;
11348 /* if <$fh>, create the ops to turn the variable into a
11352 /* try to find it in the pad for this block, otherwise find
11353 add symbol table ops
11355 const PADOFFSET tmp = pad_findmy(d);
11356 if (tmp != NOT_IN_PAD) {
11357 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11358 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11359 HEK * const stashname = HvNAME_HEK(stash);
11360 SV * const sym = sv_2mortal(newSVhek(stashname));
11361 sv_catpvs(sym, "::");
11362 sv_catpv(sym, d+1);
11367 OP * const o = newOP(OP_PADSV, 0);
11369 PL_lex_op = readline_overriden
11370 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11371 append_elem(OP_LIST, o,
11372 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11373 : (OP*)newUNOP(OP_READLINE, 0, o);
11382 ? (GV_ADDMULTI | GV_ADDINEVAL)
11385 PL_lex_op = readline_overriden
11386 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11387 append_elem(OP_LIST,
11388 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11389 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11390 : (OP*)newUNOP(OP_READLINE, 0,
11391 newUNOP(OP_RV2SV, 0,
11392 newGVOP(OP_GV, 0, gv)));
11394 if (!readline_overriden)
11395 PL_lex_op->op_flags |= OPf_SPECIAL;
11396 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11397 yylval.ival = OP_NULL;
11400 /* If it's none of the above, it must be a literal filehandle
11401 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11403 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11404 PL_lex_op = readline_overriden
11405 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11406 append_elem(OP_LIST,
11407 newGVOP(OP_GV, 0, gv),
11408 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11409 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11410 yylval.ival = OP_NULL;
11419 takes: start position in buffer
11420 keep_quoted preserve \ on the embedded delimiter(s)
11421 keep_delims preserve the delimiters around the string
11422 returns: position to continue reading from buffer
11423 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11424 updates the read buffer.
11426 This subroutine pulls a string out of the input. It is called for:
11427 q single quotes q(literal text)
11428 ' single quotes 'literal text'
11429 qq double quotes qq(interpolate $here please)
11430 " double quotes "interpolate $here please"
11431 qx backticks qx(/bin/ls -l)
11432 ` backticks `/bin/ls -l`
11433 qw quote words @EXPORT_OK = qw( func() $spam )
11434 m// regexp match m/this/
11435 s/// regexp substitute s/this/that/
11436 tr/// string transliterate tr/this/that/
11437 y/// string transliterate y/this/that/
11438 ($*@) sub prototypes sub foo ($)
11439 (stuff) sub attr parameters sub foo : attr(stuff)
11440 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11442 In most of these cases (all but <>, patterns and transliterate)
11443 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11444 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11445 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11448 It skips whitespace before the string starts, and treats the first
11449 character as the delimiter. If the delimiter is one of ([{< then
11450 the corresponding "close" character )]}> is used as the closing
11451 delimiter. It allows quoting of delimiters, and if the string has
11452 balanced delimiters ([{<>}]) it allows nesting.
11454 On success, the SV with the resulting string is put into lex_stuff or,
11455 if that is already non-NULL, into lex_repl. The second case occurs only
11456 when parsing the RHS of the special constructs s/// and tr/// (y///).
11457 For convenience, the terminating delimiter character is stuffed into
11462 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11465 SV *sv; /* scalar value: string */
11466 const char *tmps; /* temp string, used for delimiter matching */
11467 register char *s = start; /* current position in the buffer */
11468 register char term; /* terminating character */
11469 register char *to; /* current position in the sv's data */
11470 I32 brackets = 1; /* bracket nesting level */
11471 bool has_utf8 = FALSE; /* is there any utf8 content? */
11472 I32 termcode; /* terminating char. code */
11473 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11474 STRLEN termlen; /* length of terminating string */
11475 int last_off = 0; /* last position for nesting bracket */
11481 /* skip space before the delimiter */
11487 if (PL_realtokenstart >= 0) {
11488 stuffstart = PL_realtokenstart;
11489 PL_realtokenstart = -1;
11492 stuffstart = start - SvPVX(PL_linestr);
11494 /* mark where we are, in case we need to report errors */
11497 /* after skipping whitespace, the next character is the terminator */
11500 termcode = termstr[0] = term;
11504 termcode = utf8_to_uvchr((U8*)s, &termlen);
11505 Copy(s, termstr, termlen, U8);
11506 if (!UTF8_IS_INVARIANT(term))
11510 /* mark where we are */
11511 PL_multi_start = CopLINE(PL_curcop);
11512 PL_multi_open = term;
11514 /* find corresponding closing delimiter */
11515 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11516 termcode = termstr[0] = term = tmps[5];
11518 PL_multi_close = term;
11520 /* create a new SV to hold the contents. 79 is the SV's initial length.
11521 What a random number. */
11523 sv_upgrade(sv, SVt_PVIV);
11524 SvIV_set(sv, termcode);
11525 (void)SvPOK_only(sv); /* validate pointer */
11527 /* move past delimiter and try to read a complete string */
11529 sv_catpvn(sv, s, termlen);
11532 tstart = SvPVX(PL_linestr) + stuffstart;
11533 if (!PL_thisopen && !keep_delims) {
11534 PL_thisopen = newSVpvn(tstart, s - tstart);
11535 stuffstart = s - SvPVX(PL_linestr);
11539 if (PL_encoding && !UTF) {
11543 int offset = s - SvPVX_const(PL_linestr);
11544 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11545 &offset, (char*)termstr, termlen);
11546 const char * const ns = SvPVX_const(PL_linestr) + offset;
11547 char * const svlast = SvEND(sv) - 1;
11549 for (; s < ns; s++) {
11550 if (*s == '\n' && !PL_rsfp)
11551 CopLINE_inc(PL_curcop);
11554 goto read_more_line;
11556 /* handle quoted delimiters */
11557 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11559 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11561 if ((svlast-1 - t) % 2) {
11562 if (!keep_quoted) {
11563 *(svlast-1) = term;
11565 SvCUR_set(sv, SvCUR(sv) - 1);
11570 if (PL_multi_open == PL_multi_close) {
11576 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11577 /* At here, all closes are "was quoted" one,
11578 so we don't check PL_multi_close. */
11580 if (!keep_quoted && *(t+1) == PL_multi_open)
11585 else if (*t == PL_multi_open)
11593 SvCUR_set(sv, w - SvPVX_const(sv));
11595 last_off = w - SvPVX(sv);
11596 if (--brackets <= 0)
11601 if (!keep_delims) {
11602 SvCUR_set(sv, SvCUR(sv) - 1);
11608 /* extend sv if need be */
11609 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11610 /* set 'to' to the next character in the sv's string */
11611 to = SvPVX(sv)+SvCUR(sv);
11613 /* if open delimiter is the close delimiter read unbridle */
11614 if (PL_multi_open == PL_multi_close) {
11615 for (; s < PL_bufend; s++,to++) {
11616 /* embedded newlines increment the current line number */
11617 if (*s == '\n' && !PL_rsfp)
11618 CopLINE_inc(PL_curcop);
11619 /* handle quoted delimiters */
11620 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11621 if (!keep_quoted && s[1] == term)
11623 /* any other quotes are simply copied straight through */
11627 /* terminate when run out of buffer (the for() condition), or
11628 have found the terminator */
11629 else if (*s == term) {
11632 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11635 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11641 /* if the terminator isn't the same as the start character (e.g.,
11642 matched brackets), we have to allow more in the quoting, and
11643 be prepared for nested brackets.
11646 /* read until we run out of string, or we find the terminator */
11647 for (; s < PL_bufend; s++,to++) {
11648 /* embedded newlines increment the line count */
11649 if (*s == '\n' && !PL_rsfp)
11650 CopLINE_inc(PL_curcop);
11651 /* backslashes can escape the open or closing characters */
11652 if (*s == '\\' && s+1 < PL_bufend) {
11653 if (!keep_quoted &&
11654 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11659 /* allow nested opens and closes */
11660 else if (*s == PL_multi_close && --brackets <= 0)
11662 else if (*s == PL_multi_open)
11664 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11669 /* terminate the copied string and update the sv's end-of-string */
11671 SvCUR_set(sv, to - SvPVX_const(sv));
11674 * this next chunk reads more into the buffer if we're not done yet
11678 break; /* handle case where we are done yet :-) */
11680 #ifndef PERL_STRICT_CR
11681 if (to - SvPVX_const(sv) >= 2) {
11682 if ((to[-2] == '\r' && to[-1] == '\n') ||
11683 (to[-2] == '\n' && to[-1] == '\r'))
11687 SvCUR_set(sv, to - SvPVX_const(sv));
11689 else if (to[-1] == '\r')
11692 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11697 /* if we're out of file, or a read fails, bail and reset the current
11698 line marker so we can report where the unterminated string began
11701 if (PL_madskills) {
11702 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11704 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11706 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11710 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11712 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11718 /* we read a line, so increment our line counter */
11719 CopLINE_inc(PL_curcop);
11721 /* update debugger info */
11722 if (PERLDB_LINE && PL_curstash != PL_debstash) {
11723 SV * const line_sv = newSV(0);
11725 sv_upgrade(line_sv, SVt_PVMG);
11726 sv_setsv(line_sv,PL_linestr);
11727 (void)SvIOK_on(line_sv);
11728 SvIV_set(line_sv, 0);
11729 av_store(CopFILEAVx(PL_curcop), (I32)CopLINE(PL_curcop), line_sv);
11732 /* having changed the buffer, we must update PL_bufend */
11733 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11734 PL_last_lop = PL_last_uni = NULL;
11737 /* at this point, we have successfully read the delimited string */
11739 if (!PL_encoding || UTF) {
11741 if (PL_madskills) {
11742 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11743 const int len = s - start;
11745 sv_catpvn(PL_thisstuff, tstart, len);
11747 PL_thisstuff = newSVpvn(tstart, len);
11748 if (!PL_thisclose && !keep_delims)
11749 PL_thisclose = newSVpvn(s,termlen);
11754 sv_catpvn(sv, s, termlen);
11759 if (PL_madskills) {
11760 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11761 const int len = s - tstart - termlen;
11763 sv_catpvn(PL_thisstuff, tstart, len);
11765 PL_thisstuff = newSVpvn(tstart, len);
11766 if (!PL_thisclose && !keep_delims)
11767 PL_thisclose = newSVpvn(s - termlen,termlen);
11771 if (has_utf8 || PL_encoding)
11774 PL_multi_end = CopLINE(PL_curcop);
11776 /* if we allocated too much space, give some back */
11777 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11778 SvLEN_set(sv, SvCUR(sv) + 1);
11779 SvPV_renew(sv, SvLEN(sv));
11782 /* decide whether this is the first or second quoted string we've read
11795 takes: pointer to position in buffer
11796 returns: pointer to new position in buffer
11797 side-effects: builds ops for the constant in yylval.op
11799 Read a number in any of the formats that Perl accepts:
11801 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11802 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11805 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11807 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11810 If it reads a number without a decimal point or an exponent, it will
11811 try converting the number to an integer and see if it can do so
11812 without loss of precision.
11816 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11819 register const char *s = start; /* current position in buffer */
11820 register char *d; /* destination in temp buffer */
11821 register char *e; /* end of temp buffer */
11822 NV nv; /* number read, as a double */
11823 SV *sv = NULL; /* place to put the converted number */
11824 bool floatit; /* boolean: int or float? */
11825 const char *lastub = NULL; /* position of last underbar */
11826 static char const number_too_long[] = "Number too long";
11828 /* We use the first character to decide what type of number this is */
11832 Perl_croak(aTHX_ "panic: scan_num");
11834 /* if it starts with a 0, it could be an octal number, a decimal in
11835 0.13 disguise, or a hexadecimal number, or a binary number. */
11839 u holds the "number so far"
11840 shift the power of 2 of the base
11841 (hex == 4, octal == 3, binary == 1)
11842 overflowed was the number more than we can hold?
11844 Shift is used when we add a digit. It also serves as an "are
11845 we in octal/hex/binary?" indicator to disallow hex characters
11846 when in octal mode.
11851 bool overflowed = FALSE;
11852 bool just_zero = TRUE; /* just plain 0 or binary number? */
11853 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11854 static const char* const bases[5] =
11855 { "", "binary", "", "octal", "hexadecimal" };
11856 static const char* const Bases[5] =
11857 { "", "Binary", "", "Octal", "Hexadecimal" };
11858 static const char* const maxima[5] =
11860 "0b11111111111111111111111111111111",
11864 const char *base, *Base, *max;
11866 /* check for hex */
11871 } else if (s[1] == 'b') {
11876 /* check for a decimal in disguise */
11877 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11879 /* so it must be octal */
11886 if (ckWARN(WARN_SYNTAX))
11887 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11888 "Misplaced _ in number");
11892 base = bases[shift];
11893 Base = Bases[shift];
11894 max = maxima[shift];
11896 /* read the rest of the number */
11898 /* x is used in the overflow test,
11899 b is the digit we're adding on. */
11904 /* if we don't mention it, we're done */
11908 /* _ are ignored -- but warned about if consecutive */
11910 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11911 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11912 "Misplaced _ in number");
11916 /* 8 and 9 are not octal */
11917 case '8': case '9':
11919 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11923 case '2': case '3': case '4':
11924 case '5': case '6': case '7':
11926 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11929 case '0': case '1':
11930 b = *s++ & 15; /* ASCII digit -> value of digit */
11934 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11935 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11936 /* make sure they said 0x */
11939 b = (*s++ & 7) + 9;
11941 /* Prepare to put the digit we have onto the end
11942 of the number so far. We check for overflows.
11948 x = u << shift; /* make room for the digit */
11950 if ((x >> shift) != u
11951 && !(PL_hints & HINT_NEW_BINARY)) {
11954 if (ckWARN_d(WARN_OVERFLOW))
11955 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
11956 "Integer overflow in %s number",
11959 u = x | b; /* add the digit to the end */
11962 n *= nvshift[shift];
11963 /* If an NV has not enough bits in its
11964 * mantissa to represent an UV this summing of
11965 * small low-order numbers is a waste of time
11966 * (because the NV cannot preserve the
11967 * low-order bits anyway): we could just
11968 * remember when did we overflow and in the
11969 * end just multiply n by the right
11977 /* if we get here, we had success: make a scalar value from
11982 /* final misplaced underbar check */
11983 if (s[-1] == '_') {
11984 if (ckWARN(WARN_SYNTAX))
11985 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
11990 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
11991 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
11992 "%s number > %s non-portable",
11998 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
11999 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12000 "%s number > %s non-portable",
12005 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12006 sv = new_constant(start, s - start, "integer",
12008 else if (PL_hints & HINT_NEW_BINARY)
12009 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12014 handle decimal numbers.
12015 we're also sent here when we read a 0 as the first digit
12017 case '1': case '2': case '3': case '4': case '5':
12018 case '6': case '7': case '8': case '9': case '.':
12021 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12024 /* read next group of digits and _ and copy into d */
12025 while (isDIGIT(*s) || *s == '_') {
12026 /* skip underscores, checking for misplaced ones
12030 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12031 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12032 "Misplaced _ in number");
12036 /* check for end of fixed-length buffer */
12038 Perl_croak(aTHX_ number_too_long);
12039 /* if we're ok, copy the character */
12044 /* final misplaced underbar check */
12045 if (lastub && s == lastub + 1) {
12046 if (ckWARN(WARN_SYNTAX))
12047 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12050 /* read a decimal portion if there is one. avoid
12051 3..5 being interpreted as the number 3. followed
12054 if (*s == '.' && s[1] != '.') {
12059 if (ckWARN(WARN_SYNTAX))
12060 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12061 "Misplaced _ in number");
12065 /* copy, ignoring underbars, until we run out of digits.
12067 for (; isDIGIT(*s) || *s == '_'; s++) {
12068 /* fixed length buffer check */
12070 Perl_croak(aTHX_ number_too_long);
12072 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12073 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12074 "Misplaced _ in number");
12080 /* fractional part ending in underbar? */
12081 if (s[-1] == '_') {
12082 if (ckWARN(WARN_SYNTAX))
12083 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12084 "Misplaced _ in number");
12086 if (*s == '.' && isDIGIT(s[1])) {
12087 /* oops, it's really a v-string, but without the "v" */
12093 /* read exponent part, if present */
12094 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12098 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12099 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12101 /* stray preinitial _ */
12103 if (ckWARN(WARN_SYNTAX))
12104 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12105 "Misplaced _ in number");
12109 /* allow positive or negative exponent */
12110 if (*s == '+' || *s == '-')
12113 /* stray initial _ */
12115 if (ckWARN(WARN_SYNTAX))
12116 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12117 "Misplaced _ in number");
12121 /* read digits of exponent */
12122 while (isDIGIT(*s) || *s == '_') {
12125 Perl_croak(aTHX_ number_too_long);
12129 if (((lastub && s == lastub + 1) ||
12130 (!isDIGIT(s[1]) && s[1] != '_'))
12131 && ckWARN(WARN_SYNTAX))
12132 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12133 "Misplaced _ in number");
12140 /* make an sv from the string */
12144 We try to do an integer conversion first if no characters
12145 indicating "float" have been found.
12150 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12152 if (flags == IS_NUMBER_IN_UV) {
12154 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12157 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12158 if (uv <= (UV) IV_MIN)
12159 sv_setiv(sv, -(IV)uv);
12166 /* terminate the string */
12168 nv = Atof(PL_tokenbuf);
12172 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12173 (PL_hints & HINT_NEW_INTEGER) )
12174 sv = new_constant(PL_tokenbuf,
12177 (floatit ? "float" : "integer"),
12181 /* if it starts with a v, it could be a v-string */
12184 sv = newSV(5); /* preallocate storage space */
12185 s = scan_vstring(s,sv);
12189 /* make the op for the constant and return */
12192 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12194 lvalp->opval = NULL;
12200 S_scan_formline(pTHX_ register char *s)
12203 register char *eol;
12205 SV * const stuff = newSVpvs("");
12206 bool needargs = FALSE;
12207 bool eofmt = FALSE;
12209 char *tokenstart = s;
12212 if (PL_madskills) {
12213 savewhite = PL_thiswhite;
12218 while (!needargs) {
12221 #ifdef PERL_STRICT_CR
12222 while (SPACE_OR_TAB(*t))
12225 while (SPACE_OR_TAB(*t) || *t == '\r')
12228 if (*t == '\n' || t == PL_bufend) {
12233 if (PL_in_eval && !PL_rsfp) {
12234 eol = (char *) memchr(s,'\n',PL_bufend-s);
12239 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12241 for (t = s; t < eol; t++) {
12242 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12244 goto enough; /* ~~ must be first line in formline */
12246 if (*t == '@' || *t == '^')
12250 sv_catpvn(stuff, s, eol-s);
12251 #ifndef PERL_STRICT_CR
12252 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12253 char *end = SvPVX(stuff) + SvCUR(stuff);
12256 SvCUR_set(stuff, SvCUR(stuff) - 1);
12266 if (PL_madskills) {
12268 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12270 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12273 s = filter_gets(PL_linestr, PL_rsfp, 0);
12275 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12277 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12279 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12280 PL_last_lop = PL_last_uni = NULL;
12289 if (SvCUR(stuff)) {
12292 PL_lex_state = LEX_NORMAL;
12293 start_force(PL_curforce);
12294 NEXTVAL_NEXTTOKE.ival = 0;
12298 PL_lex_state = LEX_FORMLINE;
12300 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12302 else if (PL_encoding)
12303 sv_recode_to_utf8(stuff, PL_encoding);
12305 start_force(PL_curforce);
12306 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12308 start_force(PL_curforce);
12309 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12313 SvREFCNT_dec(stuff);
12315 PL_lex_formbrack = 0;
12319 if (PL_madskills) {
12321 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12323 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12324 PL_thiswhite = savewhite;
12336 PL_cshlen = strlen(PL_cshname);
12338 #if defined(USE_ITHREADS)
12339 PERL_UNUSED_CONTEXT;
12345 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12348 const I32 oldsavestack_ix = PL_savestack_ix;
12349 CV* const outsidecv = PL_compcv;
12352 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12354 SAVEI32(PL_subline);
12355 save_item(PL_subname);
12356 SAVESPTR(PL_compcv);
12358 PL_compcv = (CV*)newSV(0);
12359 sv_upgrade((SV *)PL_compcv, is_format ? SVt_PVFM : SVt_PVCV);
12360 CvFLAGS(PL_compcv) |= flags;
12362 PL_subline = CopLINE(PL_curcop);
12363 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12364 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12365 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12367 return oldsavestack_ix;
12371 #pragma segment Perl_yylex
12374 Perl_yywarn(pTHX_ const char *s)
12377 PL_in_eval |= EVAL_WARNONLY;
12379 PL_in_eval &= ~EVAL_WARNONLY;
12384 Perl_yyerror(pTHX_ const char *s)
12387 const char *where = NULL;
12388 const char *context = NULL;
12392 if (!yychar || (yychar == ';' && !PL_rsfp))
12394 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12395 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12396 PL_oldbufptr != PL_bufptr) {
12399 The code below is removed for NetWare because it abends/crashes on NetWare
12400 when the script has error such as not having the closing quotes like:
12401 if ($var eq "value)
12402 Checking of white spaces is anyway done in NetWare code.
12405 while (isSPACE(*PL_oldoldbufptr))
12408 context = PL_oldoldbufptr;
12409 contlen = PL_bufptr - PL_oldoldbufptr;
12411 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12412 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12415 The code below is removed for NetWare because it abends/crashes on NetWare
12416 when the script has error such as not having the closing quotes like:
12417 if ($var eq "value)
12418 Checking of white spaces is anyway done in NetWare code.
12421 while (isSPACE(*PL_oldbufptr))
12424 context = PL_oldbufptr;
12425 contlen = PL_bufptr - PL_oldbufptr;
12427 else if (yychar > 255)
12428 where = "next token ???";
12429 else if (yychar == -2) { /* YYEMPTY */
12430 if (PL_lex_state == LEX_NORMAL ||
12431 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12432 where = "at end of line";
12433 else if (PL_lex_inpat)
12434 where = "within pattern";
12436 where = "within string";
12439 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12441 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12442 else if (isPRINT_LC(yychar))
12443 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12445 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12446 where = SvPVX_const(where_sv);
12448 msg = sv_2mortal(newSVpv(s, 0));
12449 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12450 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12452 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12454 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12455 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12456 Perl_sv_catpvf(aTHX_ msg,
12457 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12458 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12461 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12462 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, (void*)msg);
12465 if (PL_error_count >= 10) {
12466 if (PL_in_eval && SvCUR(ERRSV))
12467 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12468 (void*)ERRSV, OutCopFILE(PL_curcop));
12470 Perl_croak(aTHX_ "%s has too many errors.\n",
12471 OutCopFILE(PL_curcop));
12474 PL_in_my_stash = NULL;
12478 #pragma segment Main
12482 S_swallow_bom(pTHX_ U8 *s)
12485 const STRLEN slen = SvCUR(PL_linestr);
12488 if (s[1] == 0xFE) {
12489 /* UTF-16 little-endian? (or UTF32-LE?) */
12490 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12491 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12492 #ifndef PERL_NO_UTF16_FILTER
12493 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12496 if (PL_bufend > (char*)s) {
12500 filter_add(utf16rev_textfilter, NULL);
12501 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12502 utf16_to_utf8_reversed(s, news,
12503 PL_bufend - (char*)s - 1,
12505 sv_setpvn(PL_linestr, (const char*)news, newlen);
12507 s = (U8*)SvPVX(PL_linestr);
12508 Copy(news, s, newlen, U8);
12512 SvUTF8_on(PL_linestr);
12513 s = (U8*)SvPVX(PL_linestr);
12515 /* FIXME - is this a general bug fix? */
12518 PL_bufend = SvPVX(PL_linestr) + newlen;
12521 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12526 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12527 #ifndef PERL_NO_UTF16_FILTER
12528 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12531 if (PL_bufend > (char *)s) {
12535 filter_add(utf16_textfilter, NULL);
12536 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12537 utf16_to_utf8(s, news,
12538 PL_bufend - (char*)s,
12540 sv_setpvn(PL_linestr, (const char*)news, newlen);
12542 SvUTF8_on(PL_linestr);
12543 s = (U8*)SvPVX(PL_linestr);
12544 PL_bufend = SvPVX(PL_linestr) + newlen;
12547 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12552 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12553 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12554 s += 3; /* UTF-8 */
12560 if (s[2] == 0xFE && s[3] == 0xFF) {
12561 /* UTF-32 big-endian */
12562 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12565 else if (s[2] == 0 && s[3] != 0) {
12568 * are a good indicator of UTF-16BE. */
12569 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12575 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12576 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12577 s += 4; /* UTF-8 */
12583 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12586 * are a good indicator of UTF-16LE. */
12587 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12596 * Restore a source filter.
12600 restore_rsfp(pTHX_ void *f)
12603 PerlIO * const fp = (PerlIO*)f;
12605 if (PL_rsfp == PerlIO_stdin())
12606 PerlIO_clearerr(PL_rsfp);
12607 else if (PL_rsfp && (PL_rsfp != fp))
12608 PerlIO_close(PL_rsfp);
12612 #ifndef PERL_NO_UTF16_FILTER
12614 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12617 const STRLEN old = SvCUR(sv);
12618 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12619 DEBUG_P(PerlIO_printf(Perl_debug_log,
12620 "utf16_textfilter(%p): %d %d (%d)\n",
12621 FPTR2DPTR(void *, utf16_textfilter),
12622 idx, maxlen, (int) count));
12626 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12627 Copy(SvPVX_const(sv), tmps, old, char);
12628 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12629 SvCUR(sv) - old, &newlen);
12630 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12632 DEBUG_P({sv_dump(sv);});
12637 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12640 const STRLEN old = SvCUR(sv);
12641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12642 DEBUG_P(PerlIO_printf(Perl_debug_log,
12643 "utf16rev_textfilter(%p): %d %d (%d)\n",
12644 FPTR2DPTR(void *, utf16rev_textfilter),
12645 idx, maxlen, (int) count));
12649 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12650 Copy(SvPVX_const(sv), tmps, old, char);
12651 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12652 SvCUR(sv) - old, &newlen);
12653 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12655 DEBUG_P({ sv_dump(sv); });
12661 Returns a pointer to the next character after the parsed
12662 vstring, as well as updating the passed in sv.
12664 Function must be called like
12667 s = scan_vstring(s,sv);
12669 The sv should already be large enough to store the vstring
12670 passed in, for performance reasons.
12675 Perl_scan_vstring(pTHX_ const char *s, SV *sv)
12678 const char *pos = s;
12679 const char *start = s;
12680 if (*pos == 'v') pos++; /* get past 'v' */
12681 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12683 if ( *pos != '.') {
12684 /* this may not be a v-string if followed by => */
12685 const char *next = pos;
12686 while (next < PL_bufend && isSPACE(*next))
12688 if ((PL_bufend - next) >= 2 && *next == '=' && next[1] == '>' ) {
12689 /* return string not v-string */
12690 sv_setpvn(sv,(char *)s,pos-s);
12691 return (char *)pos;
12695 if (!isALPHA(*pos)) {
12696 U8 tmpbuf[UTF8_MAXBYTES+1];
12699 s++; /* get past 'v' */
12701 sv_setpvn(sv, "", 0);
12704 /* this is atoi() that tolerates underscores */
12707 const char *end = pos;
12709 while (--end >= s) {
12711 const UV orev = rev;
12712 rev += (*end - '0') * mult;
12714 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12715 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12716 "Integer overflow in decimal number");
12720 if (rev > 0x7FFFFFFF)
12721 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12723 /* Append native character for the rev point */
12724 tmpend = uvchr_to_utf8(tmpbuf, rev);
12725 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12726 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12728 if (pos + 1 < PL_bufend && *pos == '.' && isDIGIT(pos[1]))
12734 while (pos < PL_bufend && (isDIGIT(*pos) || *pos == '_'))
12738 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12746 * c-indentation-style: bsd
12747 * c-basic-offset: 4
12748 * indent-tabs-mode: t
12751 * ex: set ts=8 sts=4 sw=4 noet: