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 /* XXX these probably need to be made into PL vars */
40 static I32 realtokenstart;
41 static I32 faketokens = 0;
42 static MADPROP *thismad;
51 static I32 curforce = -1;
53 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
54 # define NEXTVAL_NEXTTOKE PL_nexttoke[curforce].next_val
56 # define CURMAD(slot,sv)
57 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
60 #define XFAKEBRACK 128
63 #ifdef USE_UTF8_SCRIPTS
64 # define UTF (!IN_BYTES)
66 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
69 /* In variables named $^X, these are the legal values for X.
70 * 1999-02-27 mjd-perl-patch@plover.com */
71 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
73 /* On MacOS, respect nonbreaking spaces */
74 #ifdef MACOS_TRADITIONAL
75 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
77 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
80 /* LEX_* are values for PL_lex_state, the state of the lexer.
81 * They are arranged oddly so that the guard on the switch statement
82 * can get by with a single comparison (if the compiler is smart enough).
85 /* #define LEX_NOTPARSING 11 is done in perl.h. */
87 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
88 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
89 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
90 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
91 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
93 /* at end of code, eg "$x" followed by: */
94 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
95 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
97 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
98 string or after \E, $foo, etc */
99 #define LEX_INTERPCONST 2 /* NOT USED */
100 #define LEX_FORMLINE 1 /* expecting a format line */
101 #define LEX_KNOWNEXT 0 /* next token known; just return it */
105 static const char* const lex_state_names[] = {
124 #include "keywords.h"
126 /* CLINE is a macro that ensures PL_copline has a sane value */
131 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
134 # define SKIPSPACE0(s) skipspace0(s)
135 # define SKIPSPACE1(s) skipspace1(s)
136 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
137 # define PEEKSPACE(s) skipspace2(s,0)
139 # define SKIPSPACE0(s) skipspace(s)
140 # define SKIPSPACE1(s) skipspace(s)
141 # define SKIPSPACE2(s,tsv) skipspace(s)
142 # define PEEKSPACE(s) skipspace(s)
146 * Convenience functions to return different tokens and prime the
147 * lexer for the next token. They all take an argument.
149 * TOKEN : generic token (used for '(', DOLSHARP, etc)
150 * OPERATOR : generic operator
151 * AOPERATOR : assignment operator
152 * PREBLOCK : beginning the block after an if, while, foreach, ...
153 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
154 * PREREF : *EXPR where EXPR is not a simple identifier
155 * TERM : expression term
156 * LOOPX : loop exiting command (goto, last, dump, etc)
157 * FTST : file test operator
158 * FUN0 : zero-argument function
159 * FUN1 : not used, except for not, which isn't a UNIOP
160 * BOop : bitwise or or xor
162 * SHop : shift operator
163 * PWop : power operator
164 * PMop : pattern-matching operator
165 * Aop : addition-level operator
166 * Mop : multiplication-level operator
167 * Eop : equality-testing operator
168 * Rop : relational operator <= != gt
170 * Also see LOP and lop() below.
173 #ifdef DEBUGGING /* Serve -DT. */
174 # define REPORT(retval) tokereport((I32)retval)
176 # define REPORT(retval) (retval)
179 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
180 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
181 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
182 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
183 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
184 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
185 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
186 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
187 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
188 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
189 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
190 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
191 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
192 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
193 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
194 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
195 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
196 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
197 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
198 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
200 /* This bit of chicanery makes a unary function followed by
201 * a parenthesis into a function with one argument, highest precedence.
202 * The UNIDOR macro is for unary functions that can be followed by the //
203 * operator (such as C<shift // 0>).
205 #define UNI2(f,x) { \
209 PL_last_uni = PL_oldbufptr; \
210 PL_last_lop_op = f; \
212 return REPORT( (int)FUNC1 ); \
214 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
216 #define UNI(f) UNI2(f,XTERM)
217 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
219 #define UNIBRACK(f) { \
222 PL_last_uni = PL_oldbufptr; \
224 return REPORT( (int)FUNC1 ); \
226 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
229 /* grandfather return to old style */
230 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
234 /* how to interpret the yylval associated with the token */
238 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
244 static struct debug_tokens {
246 enum token_type type;
248 } const debug_tokens[] =
250 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
251 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
252 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
253 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
254 { ARROW, TOKENTYPE_NONE, "ARROW" },
255 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
256 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
257 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
258 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
259 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
260 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
261 { DO, TOKENTYPE_NONE, "DO" },
262 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
263 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
264 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
265 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
266 { ELSE, TOKENTYPE_NONE, "ELSE" },
267 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
268 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
269 { FOR, TOKENTYPE_IVAL, "FOR" },
270 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
271 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
272 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
273 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
274 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
275 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
276 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
277 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
278 { IF, TOKENTYPE_IVAL, "IF" },
279 { LABEL, TOKENTYPE_PVAL, "LABEL" },
280 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
281 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
282 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
283 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
284 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
285 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
286 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
287 { MY, TOKENTYPE_IVAL, "MY" },
288 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
289 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
290 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
291 { OROP, TOKENTYPE_IVAL, "OROP" },
292 { OROR, TOKENTYPE_NONE, "OROR" },
293 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
294 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
295 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
296 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
297 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
298 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
299 { PREINC, TOKENTYPE_NONE, "PREINC" },
300 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
301 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
302 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
303 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
304 { SUB, TOKENTYPE_NONE, "SUB" },
305 { THING, TOKENTYPE_OPVAL, "THING" },
306 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
307 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
308 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
309 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
310 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
311 { USE, TOKENTYPE_IVAL, "USE" },
312 { WHEN, TOKENTYPE_IVAL, "WHEN" },
313 { WHILE, TOKENTYPE_IVAL, "WHILE" },
314 { WORD, TOKENTYPE_OPVAL, "WORD" },
315 { 0, TOKENTYPE_NONE, 0 }
318 /* dump the returned token in rv, plus any optional arg in yylval */
321 S_tokereport(pTHX_ I32 rv)
325 const char *name = NULL;
326 enum token_type type = TOKENTYPE_NONE;
327 const struct debug_tokens *p;
328 SV* const report = newSVpvs("<== ");
330 for (p = debug_tokens; p->token; p++) {
331 if (p->token == (int)rv) {
338 Perl_sv_catpv(aTHX_ report, name);
339 else if ((char)rv > ' ' && (char)rv < '~')
340 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
342 sv_catpvs(report, "EOF");
344 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
347 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
350 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
352 case TOKENTYPE_OPNUM:
353 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
354 PL_op_name[yylval.ival]);
357 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
359 case TOKENTYPE_OPVAL:
361 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
362 PL_op_name[yylval.opval->op_type]);
363 if (yylval.opval->op_type == OP_CONST) {
364 Perl_sv_catpvf(aTHX_ report, " %s",
365 SvPEEK(cSVOPx_sv(yylval.opval)));
370 sv_catpvs(report, "(opval=null)");
373 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
379 /* print the buffer with suitable escapes */
382 S_printbuf(pTHX_ const char* fmt, const char* s)
384 SV* const tmp = newSVpvs("");
385 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
394 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
395 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
399 S_ao(pTHX_ int toketype)
402 if (*PL_bufptr == '=') {
404 if (toketype == ANDAND)
405 yylval.ival = OP_ANDASSIGN;
406 else if (toketype == OROR)
407 yylval.ival = OP_ORASSIGN;
408 else if (toketype == DORDOR)
409 yylval.ival = OP_DORASSIGN;
417 * When Perl expects an operator and finds something else, no_op
418 * prints the warning. It always prints "<something> found where
419 * operator expected. It prints "Missing semicolon on previous line?"
420 * if the surprise occurs at the start of the line. "do you need to
421 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
422 * where the compiler doesn't know if foo is a method call or a function.
423 * It prints "Missing operator before end of line" if there's nothing
424 * after the missing operator, or "... before <...>" if there is something
425 * after the missing operator.
429 S_no_op(pTHX_ const char *what, char *s)
432 char * const oldbp = PL_bufptr;
433 const bool is_first = (PL_oldbufptr == PL_linestart);
439 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
440 if (ckWARN_d(WARN_SYNTAX)) {
442 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
443 "\t(Missing semicolon on previous line?)\n");
444 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
446 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
447 if (t < PL_bufptr && isSPACE(*t))
448 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
449 "\t(Do you need to predeclare %.*s?)\n",
450 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
454 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
455 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
463 * Complain about missing quote/regexp/heredoc terminator.
464 * If it's called with (char *)NULL then it cauterizes the line buffer.
465 * If we're in a delimited string and the delimiter is a control
466 * character, it's reformatted into a two-char sequence like ^C.
471 S_missingterm(pTHX_ char *s)
477 char * const nl = strrchr(s,'\n');
483 iscntrl(PL_multi_close)
485 PL_multi_close < 32 || PL_multi_close == 127
489 tmpbuf[1] = (char)toCTRL(PL_multi_close);
494 *tmpbuf = (char)PL_multi_close;
498 q = strchr(s,'"') ? '\'' : '"';
499 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
502 #define FEATURE_IS_ENABLED(name) \
503 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
504 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
506 * S_feature_is_enabled
507 * Check whether the named feature is enabled.
510 S_feature_is_enabled(pTHX_ char *name, STRLEN namelen)
513 HV * const hinthv = GvHV(PL_hintgv);
514 char he_name[32] = "feature_";
515 (void) strncpy(&he_name[8], name, 24);
517 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
525 Perl_deprecate(pTHX_ const char *s)
527 if (ckWARN(WARN_DEPRECATED))
528 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
532 Perl_deprecate_old(pTHX_ const char *s)
534 /* This function should NOT be called for any new deprecated warnings */
535 /* Use Perl_deprecate instead */
537 /* It is here to maintain backward compatibility with the pre-5.8 */
538 /* warnings category hierarchy. The "deprecated" category used to */
539 /* live under the "syntax" category. It is now a top-level category */
540 /* in its own right. */
542 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
543 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
544 "Use of %s is deprecated", s);
548 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
549 * utf16-to-utf8-reversed.
552 #ifdef PERL_CR_FILTER
556 register const char *s = SvPVX_const(sv);
557 register const char * const e = s + SvCUR(sv);
558 /* outer loop optimized to do nothing if there are no CR-LFs */
560 if (*s++ == '\r' && *s == '\n') {
561 /* hit a CR-LF, need to copy the rest */
562 register char *d = s - 1;
565 if (*s == '\r' && s[1] == '\n')
576 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
578 const I32 count = FILTER_READ(idx+1, sv, maxlen);
579 if (count > 0 && !maxlen)
587 * Initialize variables. Uses the Perl save_stack to save its state (for
588 * recursive calls to the parser).
592 Perl_lex_start(pTHX_ SV *line)
598 SAVEI32(PL_lex_dojoin);
599 SAVEI32(PL_lex_brackets);
600 SAVEI32(PL_lex_casemods);
601 SAVEI32(PL_lex_starts);
602 SAVEI32(PL_lex_state);
603 SAVEVPTR(PL_lex_inpat);
604 SAVEI32(PL_lex_inwhat);
606 if (PL_lex_state == LEX_KNOWNEXT) {
607 I32 toke = PL_lasttoke;
608 while (--toke >= 0) {
609 SAVEI32(PL_nexttoke[toke].next_type);
610 SAVEVPTR(PL_nexttoke[toke].next_val);
612 SAVEVPTR(PL_nexttoke[toke].next_mad);
614 SAVEI32(PL_lasttoke);
624 SAVEI32(realtokenstart);
629 if (PL_lex_state == LEX_KNOWNEXT) {
630 I32 toke = PL_nexttoke;
631 while (--toke >= 0) {
632 SAVEI32(PL_nexttype[toke]);
633 SAVEVPTR(PL_nextval[toke]);
635 SAVEI32(PL_nexttoke);
638 SAVECOPLINE(PL_curcop);
641 SAVEPPTR(PL_oldbufptr);
642 SAVEPPTR(PL_oldoldbufptr);
643 SAVEPPTR(PL_last_lop);
644 SAVEPPTR(PL_last_uni);
645 SAVEPPTR(PL_linestart);
646 SAVESPTR(PL_linestr);
647 SAVEGENERICPV(PL_lex_brackstack);
648 SAVEGENERICPV(PL_lex_casestack);
649 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
650 SAVESPTR(PL_lex_stuff);
651 SAVEI32(PL_lex_defer);
652 SAVEI32(PL_sublex_info.sub_inwhat);
653 SAVESPTR(PL_lex_repl);
655 SAVEINT(PL_lex_expect);
657 PL_lex_state = LEX_NORMAL;
661 Newx(PL_lex_brackstack, 120, char);
662 Newx(PL_lex_casestack, 12, char);
664 *PL_lex_casestack = '\0';
676 PL_sublex_info.sub_inwhat = 0;
678 if (SvREADONLY(PL_linestr))
679 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
680 s = SvPV_const(PL_linestr, len);
681 if (!len || s[len-1] != ';') {
682 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
683 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
684 sv_catpvs(PL_linestr, "\n;");
686 SvTEMP_off(PL_linestr);
687 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
688 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
689 PL_last_lop = PL_last_uni = NULL;
695 * Finalizer for lexing operations. Must be called when the parser is
696 * done with the lexer.
703 PL_doextract = FALSE;
708 * This subroutine has nothing to do with tilting, whether at windmills
709 * or pinball tables. Its name is short for "increment line". It
710 * increments the current line number in CopLINE(PL_curcop) and checks
711 * to see whether the line starts with a comment of the form
712 * # line 500 "foo.pm"
713 * If so, it sets the current line number and file to the values in the comment.
717 S_incline(pTHX_ char *s)
725 CopLINE_inc(PL_curcop);
728 while (SPACE_OR_TAB(*s)) s++;
729 if (strnEQ(s, "line", 4))
733 if (SPACE_OR_TAB(*s))
737 while (SPACE_OR_TAB(*s)) s++;
743 while (SPACE_OR_TAB(*s))
745 if (*s == '"' && (t = strchr(s+1, '"'))) {
750 for (t = s; !isSPACE(*t); t++) ;
753 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
755 if (*e != '\n' && *e != '\0')
756 return; /* false alarm */
762 const char * const cf = CopFILE(PL_curcop);
763 STRLEN tmplen = cf ? strlen(cf) : 0;
764 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
765 /* must copy *{"::_<(eval N)[oldfilename:L]"}
766 * to *{"::_<newfilename"} */
767 char smallbuf[256], smallbuf2[256];
768 char *tmpbuf, *tmpbuf2;
770 STRLEN tmplen2 = strlen(s);
771 if (tmplen + 3 < sizeof smallbuf)
774 Newx(tmpbuf, tmplen + 3, char);
775 if (tmplen2 + 3 < sizeof smallbuf2)
778 Newx(tmpbuf2, tmplen2 + 3, char);
779 tmpbuf[0] = tmpbuf2[0] = '_';
780 tmpbuf[1] = tmpbuf2[1] = '<';
781 memcpy(tmpbuf + 2, cf, ++tmplen);
782 memcpy(tmpbuf2 + 2, s, ++tmplen2);
784 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
786 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
788 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
789 /* adjust ${"::_<newfilename"} to store the new file name */
790 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
791 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
792 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
794 if (tmpbuf != smallbuf) Safefree(tmpbuf);
795 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
798 CopFILE_free(PL_curcop);
799 CopFILE_set(PL_curcop, s);
802 CopLINE_set(PL_curcop, atoi(n)-1);
806 /* skip space before thistoken */
809 S_skipspace0(pTHX_ register char *s)
816 thiswhite = newSVpvn("",0);
817 sv_catsv(thiswhite, skipwhite);
821 realtokenstart = s - SvPVX(PL_linestr);
825 /* skip space after thistoken */
828 S_skipspace1(pTHX_ register char *s)
831 I32 startoff = start - SvPVX(PL_linestr);
836 start = SvPVX(PL_linestr) + startoff;
837 if (!thistoken && realtokenstart >= 0) {
838 char *tstart = SvPVX(PL_linestr) + realtokenstart;
839 thistoken = newSVpvn(tstart, start - tstart);
844 nextwhite = newSVpvn("",0);
845 sv_catsv(nextwhite, skipwhite);
853 S_skipspace2(pTHX_ register char *s, SV **svp)
856 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
857 I32 startoff = start - SvPVX(PL_linestr);
859 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
860 if (!PL_madskills || !svp)
862 start = SvPVX(PL_linestr) + startoff;
863 if (!thistoken && realtokenstart >= 0) {
864 char *tstart = SvPVX(PL_linestr) + realtokenstart;
865 thistoken = newSVpvn(tstart, start - tstart);
870 *svp = newSVpvn("",0);
871 sv_setsv(*svp, skipwhite);
882 * Called to gobble the appropriate amount and type of whitespace.
883 * Skips comments as well.
887 S_skipspace(pTHX_ register char *s)
892 int startoff = s - SvPVX(PL_linestr);
900 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
901 while (s < PL_bufend && SPACE_OR_TAB(*s))
911 SSize_t oldprevlen, oldoldprevlen;
912 SSize_t oldloplen = 0, oldunilen = 0;
913 while (s < PL_bufend && isSPACE(*s)) {
914 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
919 if (s < PL_bufend && *s == '#') {
920 while (s < PL_bufend && *s != '\n')
924 if (PL_in_eval && !PL_rsfp) {
931 /* only continue to recharge the buffer if we're at the end
932 * of the buffer, we're not reading from a source filter, and
933 * we're in normal lexing mode
935 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
936 PL_lex_state == LEX_FORMLINE)
943 /* try to recharge the buffer */
945 curoff = s - SvPVX(PL_linestr);
948 if ((s = filter_gets(PL_linestr, PL_rsfp,
949 (prevlen = SvCUR(PL_linestr)))) == NULL)
952 if (PL_madskills && curoff != startoff) {
954 skipwhite = newSVpvn("",0);
955 sv_catpvn(skipwhite, SvPVX(PL_linestr) + startoff,
959 /* mustn't throw out old stuff yet if madpropping */
960 SvCUR(PL_linestr) = curoff;
961 s = SvPVX(PL_linestr) + curoff;
963 if (curoff && s[-1] == '\n')
967 /* end of file. Add on the -p or -n magic */
968 /* XXX these shouldn't really be added here, can't set faketokens */
972 ";}continue{print or die qq(-p destination: $!\\n);}");
975 ";}continue{print or die qq(-p destination: $!\\n);}");
977 PL_minus_n = PL_minus_p = 0;
979 else if (PL_minus_n) {
981 sv_catpvn(PL_linestr, ";}", 2);
983 sv_setpvn(PL_linestr, ";}", 2);
989 sv_catpvn(PL_linestr,";", 1);
991 sv_setpvn(PL_linestr,";", 1);
994 /* reset variables for next time we lex */
995 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1001 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1002 PL_last_lop = PL_last_uni = NULL;
1004 /* Close the filehandle. Could be from -P preprocessor,
1005 * STDIN, or a regular file. If we were reading code from
1006 * STDIN (because the commandline held no -e or filename)
1007 * then we don't close it, we reset it so the code can
1008 * read from STDIN too.
1011 if (PL_preprocess && !PL_in_eval)
1012 (void)PerlProc_pclose(PL_rsfp);
1013 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1014 PerlIO_clearerr(PL_rsfp);
1016 (void)PerlIO_close(PL_rsfp);
1021 /* not at end of file, so we only read another line */
1022 /* make corresponding updates to old pointers, for yyerror() */
1023 oldprevlen = PL_oldbufptr - PL_bufend;
1024 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1026 oldunilen = PL_last_uni - PL_bufend;
1028 oldloplen = PL_last_lop - PL_bufend;
1029 PL_linestart = PL_bufptr = s + prevlen;
1030 PL_bufend = s + SvCUR(PL_linestr);
1032 PL_oldbufptr = s + oldprevlen;
1033 PL_oldoldbufptr = s + oldoldprevlen;
1035 PL_last_uni = s + oldunilen;
1037 PL_last_lop = s + oldloplen;
1040 /* debugger active and we're not compiling the debugger code,
1041 * so store the line into the debugger's array of lines
1043 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1044 SV * const sv = newSV(0);
1046 sv_upgrade(sv, SVt_PVMG);
1047 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1050 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1058 skipwhite = newSVpvn("",0);
1059 curoff = s - SvPVX(PL_linestr);
1060 if (curoff - startoff)
1061 sv_catpvn(skipwhite, SvPVX(PL_linestr) + startoff,
1070 * Check the unary operators to ensure there's no ambiguity in how they're
1071 * used. An ambiguous piece of code would be:
1073 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1074 * the +5 is its argument.
1084 if (PL_oldoldbufptr != PL_last_uni)
1086 while (isSPACE(*PL_last_uni))
1088 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
1089 if ((t = strchr(s, '(')) && t < PL_bufptr)
1092 if (ckWARN_d(WARN_AMBIGUOUS)){
1093 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1094 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1095 (int)(s - PL_last_uni), PL_last_uni);
1100 * LOP : macro to build a list operator. Its behaviour has been replaced
1101 * with a subroutine, S_lop() for which LOP is just another name.
1104 #define LOP(f,x) return lop(f,x,s)
1108 * Build a list operator (or something that might be one). The rules:
1109 * - if we have a next token, then it's a list operator [why?]
1110 * - if the next thing is an opening paren, then it's a function
1111 * - else it's a list operator
1115 S_lop(pTHX_ I32 f, int x, char *s)
1122 PL_last_lop = PL_oldbufptr;
1123 PL_last_lop_op = (OPCODE)f;
1126 return REPORT(LSTOP);
1129 return REPORT(LSTOP);
1132 return REPORT(FUNC);
1135 return REPORT(FUNC);
1137 return REPORT(LSTOP);
1143 * Sets up for an eventual force_next(). start_force(0) basically does
1144 * an unshift, while start_force(-1) does a push. yylex removes items
1149 S_start_force(pTHX_ int where)
1153 if (where < 0) /* so people can duplicate start_force(curforce) */
1154 where = PL_lasttoke;
1155 assert(curforce < 0 || curforce == where);
1156 if (curforce != where) {
1157 for (i = PL_lasttoke; i > where; --i) {
1158 PL_nexttoke[i] = PL_nexttoke[i-1];
1162 if (curforce < 0) /* in case of duplicate start_force() */
1163 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1167 curmad('^', newSVpvn("",0));
1168 CURMAD('_', nextwhite);
1173 S_curmad(pTHX_ char slot, SV *sv)
1182 where = &PL_nexttoke[curforce].next_mad;
1185 sv_setpvn(sv, "", 0);
1188 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1190 else if (PL_encoding) {
1191 sv_recode_to_utf8(sv, PL_encoding);
1196 /* keep a slot open for the head of the list? */
1197 if (slot != '_' && *where && (*where)->mad_key == '^') {
1198 (*where)->mad_key = slot;
1199 sv_free((*where)->mad_val);
1200 (*where)->mad_val = (void*)sv;
1203 addmad(newMADsv(slot, sv), where, 0);
1206 # define start_force(where)
1207 # define curmad(slot, sv)
1212 * When the lexer realizes it knows the next token (for instance,
1213 * it is reordering tokens for the parser) then it can call S_force_next
1214 * to know what token to return the next time the lexer is called. Caller
1215 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1216 * and possibly PL_expect to ensure the lexer handles the token correctly.
1220 S_force_next(pTHX_ I32 type)
1225 start_force(PL_lasttoke);
1226 PL_nexttoke[curforce].next_type = type;
1227 if (PL_lex_state != LEX_KNOWNEXT)
1228 PL_lex_defer = PL_lex_state;
1229 PL_lex_state = LEX_KNOWNEXT;
1230 PL_lex_expect = PL_expect;
1233 PL_nexttype[PL_nexttoke] = type;
1235 if (PL_lex_state != LEX_KNOWNEXT) {
1236 PL_lex_defer = PL_lex_state;
1237 PL_lex_expect = PL_expect;
1238 PL_lex_state = LEX_KNOWNEXT;
1244 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1247 SV * const sv = newSVpvn(start,len);
1248 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1255 * When the lexer knows the next thing is a word (for instance, it has
1256 * just seen -> and it knows that the next char is a word char, then
1257 * it calls S_force_word to stick the next word into the PL_next lookahead.
1260 * char *start : buffer position (must be within PL_linestr)
1261 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1262 * int check_keyword : if true, Perl checks to make sure the word isn't
1263 * a keyword (do this if the word is a label, e.g. goto FOO)
1264 * int allow_pack : if true, : characters will also be allowed (require,
1265 * use, etc. do this)
1266 * int allow_initial_tick : used by the "sub" lexer only.
1270 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1276 start = SKIPSPACE1(start);
1278 if (isIDFIRST_lazy_if(s,UTF) ||
1279 (allow_pack && *s == ':') ||
1280 (allow_initial_tick && *s == '\'') )
1282 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1283 if (check_keyword && keyword(PL_tokenbuf, len))
1285 start_force(curforce);
1287 curmad('X', newSVpvn(start,s-start));
1288 if (token == METHOD) {
1293 PL_expect = XOPERATOR;
1296 NEXTVAL_NEXTTOKE.opval
1297 = (OP*)newSVOP(OP_CONST,0,
1298 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1299 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1307 * Called when the lexer wants $foo *foo &foo etc, but the program
1308 * text only contains the "foo" portion. The first argument is a pointer
1309 * to the "foo", and the second argument is the type symbol to prefix.
1310 * Forces the next token to be a "WORD".
1311 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1315 S_force_ident(pTHX_ register const char *s, int kind)
1319 const STRLEN len = strlen(s);
1320 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1321 start_force(curforce);
1322 NEXTVAL_NEXTTOKE.opval = o;
1325 o->op_private = OPpCONST_ENTERED;
1326 /* XXX see note in pp_entereval() for why we forgo typo
1327 warnings if the symbol must be introduced in an eval.
1329 gv_fetchpvn_flags(s, len,
1330 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1332 kind == '$' ? SVt_PV :
1333 kind == '@' ? SVt_PVAV :
1334 kind == '%' ? SVt_PVHV :
1342 Perl_str_to_version(pTHX_ SV *sv)
1347 const char *start = SvPV_const(sv,len);
1348 const char * const end = start + len;
1349 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1350 while (start < end) {
1354 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1359 retval += ((NV)n)/nshift;
1368 * Forces the next token to be a version number.
1369 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1370 * and if "guessing" is TRUE, then no new token is created (and the caller
1371 * must use an alternative parsing method).
1375 S_force_version(pTHX_ char *s, int guessing)
1381 I32 startoff = s - SvPVX(PL_linestr);
1390 while (isDIGIT(*d) || *d == '_' || *d == '.')
1394 start_force(curforce);
1395 curmad('X', newSVpvn(s,d-s));
1398 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1400 s = scan_num(s, &yylval);
1401 version = yylval.opval;
1402 ver = cSVOPx(version)->op_sv;
1403 if (SvPOK(ver) && !SvNIOK(ver)) {
1404 SvUPGRADE(ver, SVt_PVNV);
1405 SvNV_set(ver, str_to_version(ver));
1406 SvNOK_on(ver); /* hint that it is a version */
1409 else if (guessing) {
1412 sv_free(nextwhite); /* let next token collect whitespace */
1414 s = SvPVX(PL_linestr) + startoff;
1422 if (PL_madskills && !version) {
1423 sv_free(nextwhite); /* let next token collect whitespace */
1425 s = SvPVX(PL_linestr) + startoff;
1428 /* NOTE: The parser sees the package name and the VERSION swapped */
1429 start_force(curforce);
1430 NEXTVAL_NEXTTOKE.opval = version;
1438 * Tokenize a quoted string passed in as an SV. It finds the next
1439 * chunk, up to end of string or a backslash. It may make a new
1440 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1445 S_tokeq(pTHX_ SV *sv)
1449 register char *send;
1457 s = SvPV_force(sv, len);
1458 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1461 while (s < send && *s != '\\')
1466 if ( PL_hints & HINT_NEW_STRING ) {
1467 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1473 if (s + 1 < send && (s[1] == '\\'))
1474 s++; /* all that, just for this */
1479 SvCUR_set(sv, d - SvPVX_const(sv));
1481 if ( PL_hints & HINT_NEW_STRING )
1482 return new_constant(NULL, 0, "q", sv, pv, "q");
1487 * Now come three functions related to double-quote context,
1488 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1489 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1490 * interact with PL_lex_state, and create fake ( ... ) argument lists
1491 * to handle functions and concatenation.
1492 * They assume that whoever calls them will be setting up a fake
1493 * join call, because each subthing puts a ',' after it. This lets
1496 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1498 * (I'm not sure whether the spurious commas at the end of lcfirst's
1499 * arguments and join's arguments are created or not).
1504 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1506 * Pattern matching will set PL_lex_op to the pattern-matching op to
1507 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1509 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1511 * Everything else becomes a FUNC.
1513 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1514 * had an OP_CONST or OP_READLINE). This just sets us up for a
1515 * call to S_sublex_push().
1519 S_sublex_start(pTHX)
1522 register const I32 op_type = yylval.ival;
1524 if (op_type == OP_NULL) {
1525 yylval.opval = PL_lex_op;
1529 if (op_type == OP_CONST || op_type == OP_READLINE) {
1530 SV *sv = tokeq(PL_lex_stuff);
1532 if (SvTYPE(sv) == SVt_PVIV) {
1533 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1535 const char * const p = SvPV_const(sv, len);
1536 SV * const nsv = newSVpvn(p, len);
1542 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1543 PL_lex_stuff = NULL;
1544 /* Allow <FH> // "foo" */
1545 if (op_type == OP_READLINE)
1546 PL_expect = XTERMORDORDOR;
1550 PL_sublex_info.super_state = PL_lex_state;
1551 PL_sublex_info.sub_inwhat = op_type;
1552 PL_sublex_info.sub_op = PL_lex_op;
1553 PL_lex_state = LEX_INTERPPUSH;
1557 yylval.opval = PL_lex_op;
1567 * Create a new scope to save the lexing state. The scope will be
1568 * ended in S_sublex_done. Returns a '(', starting the function arguments
1569 * to the uc, lc, etc. found before.
1570 * Sets PL_lex_state to LEX_INTERPCONCAT.
1579 PL_lex_state = PL_sublex_info.super_state;
1580 SAVEI32(PL_lex_dojoin);
1581 SAVEI32(PL_lex_brackets);
1582 SAVEI32(PL_lex_casemods);
1583 SAVEI32(PL_lex_starts);
1584 SAVEI32(PL_lex_state);
1585 SAVEVPTR(PL_lex_inpat);
1586 SAVEI32(PL_lex_inwhat);
1587 SAVECOPLINE(PL_curcop);
1588 SAVEPPTR(PL_bufptr);
1589 SAVEPPTR(PL_bufend);
1590 SAVEPPTR(PL_oldbufptr);
1591 SAVEPPTR(PL_oldoldbufptr);
1592 SAVEPPTR(PL_last_lop);
1593 SAVEPPTR(PL_last_uni);
1594 SAVEPPTR(PL_linestart);
1595 SAVESPTR(PL_linestr);
1596 SAVEGENERICPV(PL_lex_brackstack);
1597 SAVEGENERICPV(PL_lex_casestack);
1599 PL_linestr = PL_lex_stuff;
1600 PL_lex_stuff = NULL;
1602 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1603 = SvPVX(PL_linestr);
1604 PL_bufend += SvCUR(PL_linestr);
1605 PL_last_lop = PL_last_uni = NULL;
1606 SAVEFREESV(PL_linestr);
1608 PL_lex_dojoin = FALSE;
1609 PL_lex_brackets = 0;
1610 Newx(PL_lex_brackstack, 120, char);
1611 Newx(PL_lex_casestack, 12, char);
1612 PL_lex_casemods = 0;
1613 *PL_lex_casestack = '\0';
1615 PL_lex_state = LEX_INTERPCONCAT;
1616 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1618 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1619 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1620 PL_lex_inpat = PL_sublex_info.sub_op;
1622 PL_lex_inpat = NULL;
1629 * Restores lexer state after a S_sublex_push.
1636 if (!PL_lex_starts++) {
1637 SV * const sv = newSVpvs("");
1638 if (SvUTF8(PL_linestr))
1640 PL_expect = XOPERATOR;
1641 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1645 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1646 PL_lex_state = LEX_INTERPCASEMOD;
1650 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1651 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1652 PL_linestr = PL_lex_repl;
1654 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1655 PL_bufend += SvCUR(PL_linestr);
1656 PL_last_lop = PL_last_uni = NULL;
1657 SAVEFREESV(PL_linestr);
1658 PL_lex_dojoin = FALSE;
1659 PL_lex_brackets = 0;
1660 PL_lex_casemods = 0;
1661 *PL_lex_casestack = '\0';
1663 if (SvEVALED(PL_lex_repl)) {
1664 PL_lex_state = LEX_INTERPNORMAL;
1666 /* we don't clear PL_lex_repl here, so that we can check later
1667 whether this is an evalled subst; that means we rely on the
1668 logic to ensure sublex_done() is called again only via the
1669 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1672 PL_lex_state = LEX_INTERPCONCAT;
1682 endwhite = newSVpvn("",0);
1683 sv_catsv(endwhite, thiswhite);
1687 sv_setpvn(thistoken,"",0);
1689 realtokenstart = -1;
1693 PL_bufend = SvPVX(PL_linestr);
1694 PL_bufend += SvCUR(PL_linestr);
1695 PL_expect = XOPERATOR;
1696 PL_sublex_info.sub_inwhat = 0;
1704 Extracts a pattern, double-quoted string, or transliteration. This
1707 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1708 processing a pattern (PL_lex_inpat is true), a transliteration
1709 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1711 Returns a pointer to the character scanned up to. Iff this is
1712 advanced from the start pointer supplied (ie if anything was
1713 successfully parsed), will leave an OP for the substring scanned
1714 in yylval. Caller must intuit reason for not parsing further
1715 by looking at the next characters herself.
1719 double-quoted style: \r and \n
1720 regexp special ones: \D \s
1722 backrefs: \1 (deprecated in substitution replacements)
1723 case and quoting: \U \Q \E
1724 stops on @ and $, but not for $ as tail anchor
1726 In transliterations:
1727 characters are VERY literal, except for - not at the start or end
1728 of the string, which indicates a range. scan_const expands the
1729 range to the full set of intermediate characters.
1731 In double-quoted strings:
1733 double-quoted style: \r and \n
1735 backrefs: \1 (deprecated)
1736 case and quoting: \U \Q \E
1739 scan_const does *not* construct ops to handle interpolated strings.
1740 It stops processing as soon as it finds an embedded $ or @ variable
1741 and leaves it to the caller to work out what's going on.
1743 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1745 $ in pattern could be $foo or could be tail anchor. Assumption:
1746 it's a tail anchor if $ is the last thing in the string, or if it's
1747 followed by one of ")| \n\t"
1749 \1 (backreferences) are turned into $1
1751 The structure of the code is
1752 while (there's a character to process) {
1753 handle transliteration ranges
1754 skip regexp comments
1755 skip # initiated comments in //x patterns
1756 check for embedded @foo
1757 check for embedded scalars
1759 leave intact backslashes from leave (below)
1760 deprecate \1 in strings and sub replacements
1761 handle string-changing backslashes \l \U \Q \E, etc.
1762 switch (what was escaped) {
1763 handle - in a transliteration (becomes a literal -)
1764 handle \132 octal characters
1765 handle 0x15 hex characters
1766 handle \cV (control V)
1767 handle printf backslashes (\f, \r, \n, etc)
1769 } (end if backslash)
1770 } (end while character to read)
1775 S_scan_const(pTHX_ char *start)
1778 register char *send = PL_bufend; /* end of the constant */
1779 SV *sv = newSV(send - start); /* sv for the constant */
1780 register char *s = start; /* start of the constant */
1781 register char *d = SvPVX(sv); /* destination for copies */
1782 bool dorange = FALSE; /* are we in a translit range? */
1783 bool didrange = FALSE; /* did we just finish a range? */
1784 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1785 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1788 UV literal_endpoint = 0;
1791 const char *leaveit = /* set of acceptably-backslashed characters */
1793 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1796 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1797 /* If we are doing a trans and we know we want UTF8 set expectation */
1798 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1799 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1803 while (s < send || dorange) {
1804 /* get transliterations out of the way (they're most literal) */
1805 if (PL_lex_inwhat == OP_TRANS) {
1806 /* expand a range A-Z to the full set of characters. AIE! */
1808 I32 i; /* current expanded character */
1809 I32 min; /* first character in range */
1810 I32 max; /* last character in range */
1813 char * const c = (char*)utf8_hop((U8*)d, -1);
1817 *c = (char)UTF_TO_NATIVE(0xff);
1818 /* mark the range as done, and continue */
1824 i = d - SvPVX_const(sv); /* remember current offset */
1825 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1826 d = SvPVX(sv) + i; /* refresh d after realloc */
1827 d -= 2; /* eat the first char and the - */
1829 min = (U8)*d; /* first char in range */
1830 max = (U8)d[1]; /* last char in range */
1834 "Invalid range \"%c-%c\" in transliteration operator",
1835 (char)min, (char)max);
1839 if (literal_endpoint == 2 &&
1840 ((isLOWER(min) && isLOWER(max)) ||
1841 (isUPPER(min) && isUPPER(max)))) {
1843 for (i = min; i <= max; i++)
1845 *d++ = NATIVE_TO_NEED(has_utf8,i);
1847 for (i = min; i <= max; i++)
1849 *d++ = NATIVE_TO_NEED(has_utf8,i);
1854 for (i = min; i <= max; i++)
1857 /* mark the range as done, and continue */
1861 literal_endpoint = 0;
1866 /* range begins (ignore - as first or last char) */
1867 else if (*s == '-' && s+1 < send && s != start) {
1869 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1872 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1882 literal_endpoint = 0;
1887 /* if we get here, we're not doing a transliteration */
1889 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1890 except for the last char, which will be done separately. */
1891 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1893 while (s+1 < send && *s != ')')
1894 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1896 else if (s[2] == '{' /* This should match regcomp.c */
1897 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1900 char *regparse = s + (s[2] == '{' ? 3 : 4);
1903 while (count && (c = *regparse)) {
1904 if (c == '\\' && regparse[1])
1912 if (*regparse != ')')
1913 regparse--; /* Leave one char for continuation. */
1914 while (s < regparse)
1915 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1919 /* likewise skip #-initiated comments in //x patterns */
1920 else if (*s == '#' && PL_lex_inpat &&
1921 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1922 while (s+1 < send && *s != '\n')
1923 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1926 /* check for embedded arrays
1927 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1929 else if (*s == '@' && s[1]
1930 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1933 /* check for embedded scalars. only stop if we're sure it's a
1936 else if (*s == '$') {
1937 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1939 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1940 break; /* in regexp, $ might be tail anchor */
1943 /* End of else if chain - OP_TRANS rejoin rest */
1946 if (*s == '\\' && s+1 < send) {
1949 /* some backslashes we leave behind */
1950 if (*leaveit && *s && strchr(leaveit, *s)) {
1951 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1952 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1956 /* deprecate \1 in strings and substitution replacements */
1957 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1958 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1960 if (ckWARN(WARN_SYNTAX))
1961 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1966 /* string-change backslash escapes */
1967 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1972 /* if we get here, it's either a quoted -, or a digit */
1975 /* quoted - in transliterations */
1977 if (PL_lex_inwhat == OP_TRANS) {
1987 Perl_warner(aTHX_ packWARN(WARN_MISC),
1988 "Unrecognized escape \\%c passed through",
1990 /* default action is to copy the quoted character */
1991 goto default_action;
1994 /* \132 indicates an octal constant */
1995 case '0': case '1': case '2': case '3':
1996 case '4': case '5': case '6': case '7':
2000 uv = grok_oct(s, &len, &flags, NULL);
2003 goto NUM_ESCAPE_INSERT;
2005 /* \x24 indicates a hex constant */
2009 char* const e = strchr(s, '}');
2010 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2011 PERL_SCAN_DISALLOW_PREFIX;
2016 yyerror("Missing right brace on \\x{}");
2020 uv = grok_hex(s, &len, &flags, NULL);
2026 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2027 uv = grok_hex(s, &len, &flags, NULL);
2033 /* Insert oct or hex escaped character.
2034 * There will always enough room in sv since such
2035 * escapes will be longer than any UTF-8 sequence
2036 * they can end up as. */
2038 /* We need to map to chars to ASCII before doing the tests
2041 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2042 if (!has_utf8 && uv > 255) {
2043 /* Might need to recode whatever we have
2044 * accumulated so far if it contains any
2047 * (Can't we keep track of that and avoid
2048 * this rescan? --jhi)
2052 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2053 if (!NATIVE_IS_INVARIANT(*c)) {
2058 const STRLEN offset = d - SvPVX_const(sv);
2060 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2064 while (src >= (const U8 *)SvPVX_const(sv)) {
2065 if (!NATIVE_IS_INVARIANT(*src)) {
2066 const U8 ch = NATIVE_TO_ASCII(*src);
2067 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2068 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2078 if (has_utf8 || uv > 255) {
2079 d = (char*)uvchr_to_utf8((U8*)d, uv);
2081 if (PL_lex_inwhat == OP_TRANS &&
2082 PL_sublex_info.sub_op) {
2083 PL_sublex_info.sub_op->op_private |=
2084 (PL_lex_repl ? OPpTRANS_FROM_UTF
2097 /* \N{LATIN SMALL LETTER A} is a named character */
2101 char* e = strchr(s, '}');
2107 yyerror("Missing right brace on \\N{}");
2111 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2113 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2114 PERL_SCAN_DISALLOW_PREFIX;
2117 uv = grok_hex(s, &len, &flags, NULL);
2119 goto NUM_ESCAPE_INSERT;
2121 res = newSVpvn(s + 1, e - s - 1);
2122 res = new_constant( NULL, 0, "charnames",
2123 res, NULL, "\\N{...}" );
2125 sv_utf8_upgrade(res);
2126 str = SvPV_const(res,len);
2127 #ifdef EBCDIC_NEVER_MIND
2128 /* charnames uses pack U and that has been
2129 * recently changed to do the below uni->native
2130 * mapping, so this would be redundant (and wrong,
2131 * the code point would be doubly converted).
2132 * But leave this in just in case the pack U change
2133 * gets revoked, but the semantics is still
2134 * desireable for charnames. --jhi */
2136 UV uv = utf8_to_uvchr((const U8*)str, 0);
2139 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2141 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2142 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2143 str = SvPV_const(res, len);
2147 if (!has_utf8 && SvUTF8(res)) {
2148 const char * const ostart = SvPVX_const(sv);
2149 SvCUR_set(sv, d - ostart);
2152 sv_utf8_upgrade(sv);
2153 /* this just broke our allocation above... */
2154 SvGROW(sv, (STRLEN)(send - start));
2155 d = SvPVX(sv) + SvCUR(sv);
2158 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2159 const char * const odest = SvPVX_const(sv);
2161 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2162 d = SvPVX(sv) + (d - odest);
2164 Copy(str, d, len, char);
2171 yyerror("Missing braces on \\N{}");
2174 /* \c is a control character */
2183 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2186 yyerror("Missing control char name in \\c");
2190 /* printf-style backslashes, formfeeds, newlines, etc */
2192 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2195 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2198 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2201 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2204 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2207 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2210 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2216 } /* end if (backslash) */
2223 /* If we started with encoded form, or already know we want it
2224 and then encode the next character */
2225 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2227 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2228 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2231 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2232 const STRLEN off = d - SvPVX_const(sv);
2233 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2235 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2239 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2241 } /* while loop to process each character */
2243 /* terminate the string and set up the sv */
2245 SvCUR_set(sv, d - SvPVX_const(sv));
2246 if (SvCUR(sv) >= SvLEN(sv))
2247 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2250 if (PL_encoding && !has_utf8) {
2251 sv_recode_to_utf8(sv, PL_encoding);
2257 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2258 PL_sublex_info.sub_op->op_private |=
2259 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2263 /* shrink the sv if we allocated more than we used */
2264 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2265 SvPV_shrink_to_cur(sv);
2268 /* return the substring (via yylval) only if we parsed anything */
2269 if (s > PL_bufptr) {
2270 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2271 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2273 ( PL_lex_inwhat == OP_TRANS
2275 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2278 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2285 * Returns TRUE if there's more to the expression (e.g., a subscript),
2288 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2290 * ->[ and ->{ return TRUE
2291 * { and [ outside a pattern are always subscripts, so return TRUE
2292 * if we're outside a pattern and it's not { or [, then return FALSE
2293 * if we're in a pattern and the first char is a {
2294 * {4,5} (any digits around the comma) returns FALSE
2295 * if we're in a pattern and the first char is a [
2297 * [SOMETHING] has a funky algorithm to decide whether it's a
2298 * character class or not. It has to deal with things like
2299 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2300 * anything else returns TRUE
2303 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2306 S_intuit_more(pTHX_ register char *s)
2309 if (PL_lex_brackets)
2311 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2313 if (*s != '{' && *s != '[')
2318 /* In a pattern, so maybe we have {n,m}. */
2335 /* On the other hand, maybe we have a character class */
2338 if (*s == ']' || *s == '^')
2341 /* this is terrifying, and it works */
2342 int weight = 2; /* let's weigh the evidence */
2344 unsigned char un_char = 255, last_un_char;
2345 const char * const send = strchr(s,']');
2346 char tmpbuf[sizeof PL_tokenbuf * 4];
2348 if (!send) /* has to be an expression */
2351 Zero(seen,256,char);
2354 else if (isDIGIT(*s)) {
2356 if (isDIGIT(s[1]) && s[2] == ']')
2362 for (; s < send; s++) {
2363 last_un_char = un_char;
2364 un_char = (unsigned char)*s;
2369 weight -= seen[un_char] * 10;
2370 if (isALNUM_lazy_if(s+1,UTF)) {
2372 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2373 len = (int)strlen(tmpbuf);
2374 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2379 else if (*s == '$' && s[1] &&
2380 strchr("[#!%*<>()-=",s[1])) {
2381 if (/*{*/ strchr("])} =",s[2]))
2390 if (strchr("wds]",s[1]))
2392 else if (seen['\''] || seen['"'])
2394 else if (strchr("rnftbxcav",s[1]))
2396 else if (isDIGIT(s[1])) {
2398 while (s[1] && isDIGIT(s[1]))
2408 if (strchr("aA01! ",last_un_char))
2410 if (strchr("zZ79~",s[1]))
2412 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2413 weight -= 5; /* cope with negative subscript */
2416 if (!isALNUM(last_un_char)
2417 && !(last_un_char == '$' || last_un_char == '@'
2418 || last_un_char == '&')
2419 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2424 if (keyword(tmpbuf, d - tmpbuf))
2427 if (un_char == last_un_char + 1)
2429 weight -= seen[un_char];
2434 if (weight >= 0) /* probably a character class */
2444 * Does all the checking to disambiguate
2446 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2447 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2449 * First argument is the stuff after the first token, e.g. "bar".
2451 * Not a method if bar is a filehandle.
2452 * Not a method if foo is a subroutine prototyped to take a filehandle.
2453 * Not a method if it's really "Foo $bar"
2454 * Method if it's "foo $bar"
2455 * Not a method if it's really "print foo $bar"
2456 * Method if it's really "foo package::" (interpreted as package->foo)
2457 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2458 * Not a method if bar is a filehandle or package, but is quoted with
2463 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2466 char *s = start + (*start == '$');
2467 char tmpbuf[sizeof PL_tokenbuf];
2475 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2479 const char *proto = SvPVX_const(cv);
2490 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2491 /* start is the beginning of the possible filehandle/object,
2492 * and s is the end of it
2493 * tmpbuf is a copy of it
2496 if (*start == '$') {
2497 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2500 len = start - SvPVX(PL_linestr);
2504 start = SvPVX(PL_linestr) + len;
2508 return *s == '(' ? FUNCMETH : METHOD;
2510 if (!keyword(tmpbuf, len)) {
2511 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2515 soff = s - SvPVX(PL_linestr);
2519 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2520 if (indirgv && GvCVu(indirgv))
2522 /* filehandle or package name makes it a method */
2523 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2525 soff = s - SvPVX(PL_linestr);
2528 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2529 return 0; /* no assumptions -- "=>" quotes bearword */
2531 start_force(curforce);
2532 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2533 newSVpvn(tmpbuf,len));
2534 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2536 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2541 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2543 return *s == '(' ? FUNCMETH : METHOD;
2551 * Return a string of Perl code to load the debugger. If PERL5DB
2552 * is set, it will return the contents of that, otherwise a
2553 * compile-time require of perl5db.pl.
2561 const char * const pdb = PerlEnv_getenv("PERL5DB");
2565 SETERRNO(0,SS_NORMAL);
2566 return "BEGIN { require 'perl5db.pl' }";
2572 /* Encoded script support. filter_add() effectively inserts a
2573 * 'pre-processing' function into the current source input stream.
2574 * Note that the filter function only applies to the current source file
2575 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2577 * The datasv parameter (which may be NULL) can be used to pass
2578 * private data to this instance of the filter. The filter function
2579 * can recover the SV using the FILTER_DATA macro and use it to
2580 * store private buffers and state information.
2582 * The supplied datasv parameter is upgraded to a PVIO type
2583 * and the IoDIRP/IoANY field is used to store the function pointer,
2584 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2585 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2586 * private use must be set using malloc'd pointers.
2590 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2596 if (!PL_rsfp_filters)
2597 PL_rsfp_filters = newAV();
2600 SvUPGRADE(datasv, SVt_PVIO);
2601 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2602 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2603 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2604 IoANY(datasv), SvPV_nolen(datasv)));
2605 av_unshift(PL_rsfp_filters, 1);
2606 av_store(PL_rsfp_filters, 0, datasv) ;
2611 /* Delete most recently added instance of this filter function. */
2613 Perl_filter_del(pTHX_ filter_t funcp)
2619 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2621 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2623 /* if filter is on top of stack (usual case) just pop it off */
2624 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2625 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2626 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2627 IoANY(datasv) = (void *)NULL;
2628 sv_free(av_pop(PL_rsfp_filters));
2632 /* we need to search for the correct entry and clear it */
2633 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2637 /* Invoke the idxth filter function for the current rsfp. */
2638 /* maxlen 0 = read one text line */
2640 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2646 if (!PL_rsfp_filters)
2648 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2649 /* Provide a default input filter to make life easy. */
2650 /* Note that we append to the line. This is handy. */
2651 DEBUG_P(PerlIO_printf(Perl_debug_log,
2652 "filter_read %d: from rsfp\n", idx));
2656 const int old_len = SvCUR(buf_sv);
2658 /* ensure buf_sv is large enough */
2659 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2660 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2661 if (PerlIO_error(PL_rsfp))
2662 return -1; /* error */
2664 return 0 ; /* end of file */
2666 SvCUR_set(buf_sv, old_len + len) ;
2669 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2670 if (PerlIO_error(PL_rsfp))
2671 return -1; /* error */
2673 return 0 ; /* end of file */
2676 return SvCUR(buf_sv);
2678 /* Skip this filter slot if filter has been deleted */
2679 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2680 DEBUG_P(PerlIO_printf(Perl_debug_log,
2681 "filter_read %d: skipped (filter deleted)\n",
2683 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2685 /* Get function pointer hidden within datasv */
2686 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2687 DEBUG_P(PerlIO_printf(Perl_debug_log,
2688 "filter_read %d: via function %p (%s)\n",
2689 idx, datasv, SvPV_nolen_const(datasv)));
2690 /* Call function. The function is expected to */
2691 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2692 /* Return: <0:error, =0:eof, >0:not eof */
2693 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2697 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2700 #ifdef PERL_CR_FILTER
2701 if (!PL_rsfp_filters) {
2702 filter_add(S_cr_textfilter,NULL);
2705 if (PL_rsfp_filters) {
2707 SvCUR_set(sv, 0); /* start with empty line */
2708 if (FILTER_READ(0, sv, 0) > 0)
2709 return ( SvPVX(sv) ) ;
2714 return (sv_gets(sv, fp, append));
2718 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2723 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2727 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2728 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2730 return GvHV(gv); /* Foo:: */
2733 /* use constant CLASS => 'MyClass' */
2734 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2736 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2737 pkgname = SvPV_nolen_const(sv);
2741 return gv_stashpv(pkgname, FALSE);
2747 * The intent of this yylex wrapper is to minimize the changes to the
2748 * tokener when we aren't interested in collecting madprops. It remains
2749 * to be seen how successful this strategy will be...
2756 char *s = PL_bufptr;
2758 /* make sure thiswhite is initialized */
2762 /* just do what yylex would do on pending identifier; leave thiswhite alone */
2763 if (PL_pending_ident)
2764 return S_pending_ident(aTHX);
2766 /* previous token ate up our whitespace? */
2767 if (!PL_lasttoke && nextwhite) {
2768 thiswhite = nextwhite;
2772 /* isolate the token, and figure out where it is without whitespace */
2773 realtokenstart = -1;
2777 assert(curforce < 0);
2779 if (!thismad || thismad->mad_key == '^') { /* not forced already? */
2781 if (realtokenstart < 0 || !CopLINE(PL_curcop))
2782 thistoken = newSVpvn("",0);
2784 char *tstart = SvPVX(PL_linestr) + realtokenstart;
2785 thistoken = newSVpvn(tstart, s - tstart);
2788 if (thismad) /* install head */
2789 CURMAD('X', thistoken);
2792 /* last whitespace of a sublex? */
2793 if (optype == ')' && endwhite) {
2794 CURMAD('X', endwhite);
2799 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2800 if (!thiswhite && !endwhite && !optype) {
2806 /* put off final whitespace till peg */
2807 if (optype == ';' && !PL_rsfp) {
2808 nextwhite = thiswhite;
2811 else if (thisopen) {
2812 CURMAD('q', thisopen);
2818 /* Store actual token text as madprop X */
2819 CURMAD('X', thistoken);
2823 /* add preceding whitespace as madprop _ */
2824 CURMAD('_', thiswhite);
2828 /* add quoted material as madprop = */
2829 CURMAD('=', thisstuff);
2833 /* add terminating quote as madprop Q */
2834 CURMAD('Q', thisclose);
2838 /* special processing based on optype */
2842 /* opval doesn't need a TOKEN since it can already store mp */
2853 append_madprops(thismad, yylval.opval, 0);
2861 addmad(newMADsv('p', endwhite), &thismad, 0);
2870 /* remember any fake bracket that lexer is about to discard */
2871 if (PL_lex_brackets == 1 &&
2872 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2875 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2878 thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2879 addmad(newMADsv('#', thiswhite), &thismad, 0);
2882 break; /* don't bother looking for trailing comment */
2891 /* attach a trailing comment to its statement instead of next token */
2895 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2897 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2899 if (*s == '\n' || *s == '#') {
2900 while (s < PL_bufend && *s != '\n')
2904 thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2905 addmad(newMADsv('#', thiswhite), &thismad, 0);
2922 /* Create new token struct. Note: opvals return early above. */
2923 yylval.tkval = newTOKEN(optype, yylval, thismad);
2930 S_tokenize_use(pTHX_ int is_use, char *s) {
2932 if (PL_expect != XSTATE)
2933 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2934 is_use ? "use" : "no"));
2936 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2937 s = force_version(s, TRUE);
2938 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2939 start_force(curforce);
2940 NEXTVAL_NEXTTOKE.opval = NULL;
2943 else if (*s == 'v') {
2944 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2945 s = force_version(s, FALSE);
2949 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2950 s = force_version(s, FALSE);
2952 yylval.ival = is_use;
2956 static const char* const exp_name[] =
2957 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2958 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2965 Works out what to call the token just pulled out of the input
2966 stream. The yacc parser takes care of taking the ops we return and
2967 stitching them into a tree.
2973 if read an identifier
2974 if we're in a my declaration
2975 croak if they tried to say my($foo::bar)
2976 build the ops for a my() declaration
2977 if it's an access to a my() variable
2978 are we in a sort block?
2979 croak if my($a); $a <=> $b
2980 build ops for access to a my() variable
2981 if in a dq string, and they've said @foo and we can't find @foo
2983 build ops for a bareword
2984 if we already built the token before, use it.
2989 #pragma segment Perl_yylex
2995 register char *s = PL_bufptr;
3001 SV* tmp = newSVpvs("");
3002 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3003 (IV)CopLINE(PL_curcop),
3004 lex_state_names[PL_lex_state],
3005 exp_name[PL_expect],
3006 pv_display(tmp, s, strlen(s), 0, 60));
3009 /* check if there's an identifier for us to look at */
3010 if (PL_pending_ident)
3011 return REPORT(S_pending_ident(aTHX));
3013 /* no identifier pending identification */
3015 switch (PL_lex_state) {
3017 case LEX_NORMAL: /* Some compilers will produce faster */
3018 case LEX_INTERPNORMAL: /* code if we comment these out. */
3022 /* when we've already built the next token, just pull it out of the queue */
3026 yylval = PL_nexttoke[PL_lasttoke].next_val;
3028 thismad = PL_nexttoke[PL_lasttoke].next_mad;
3029 PL_nexttoke[PL_lasttoke].next_mad = 0;
3030 if (thismad && thismad->mad_key == '_') {
3031 thiswhite = (SV*)thismad->mad_val;
3032 thismad->mad_val = 0;
3038 PL_lex_state = PL_lex_defer;
3039 PL_expect = PL_lex_expect;
3040 PL_lex_defer = LEX_NORMAL;
3041 if (!PL_nexttoke[PL_lasttoke].next_type)
3046 yylval = PL_nextval[PL_nexttoke];
3048 PL_lex_state = PL_lex_defer;
3049 PL_expect = PL_lex_expect;
3050 PL_lex_defer = LEX_NORMAL;
3054 /* FIXME - can these be merged? */
3055 return(PL_nexttoke[PL_lasttoke].next_type);
3057 return REPORT(PL_nexttype[PL_nexttoke]);
3060 /* interpolated case modifiers like \L \U, including \Q and \E.
3061 when we get here, PL_bufptr is at the \
3063 case LEX_INTERPCASEMOD:
3065 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3066 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3068 /* handle \E or end of string */
3069 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3071 if (PL_lex_casemods) {
3072 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3073 PL_lex_casestack[PL_lex_casemods] = '\0';
3075 if (PL_bufptr != PL_bufend
3076 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3078 PL_lex_state = LEX_INTERPCONCAT;
3081 thistoken = newSVpvn("\\E",2);
3087 while (PL_bufptr != PL_bufend &&
3088 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3090 thiswhite = newSVpvn("",0);
3091 sv_catpvn(thiswhite, PL_bufptr, 2);
3095 if (PL_bufptr != PL_bufend)
3098 PL_lex_state = LEX_INTERPCONCAT;
3102 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3103 "### Saw case modifier\n"); });
3105 if (s[1] == '\\' && s[2] == 'E') {
3108 thiswhite = newSVpvn("",0);
3109 sv_catpvn(thiswhite, PL_bufptr, 4);
3112 PL_lex_state = LEX_INTERPCONCAT;
3117 if (!PL_madskills) /* when just compiling don't need correct */
3118 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3119 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3120 if ((*s == 'L' || *s == 'U') &&
3121 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3122 PL_lex_casestack[--PL_lex_casemods] = '\0';
3125 if (PL_lex_casemods > 10)
3126 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3127 PL_lex_casestack[PL_lex_casemods++] = *s;
3128 PL_lex_casestack[PL_lex_casemods] = '\0';
3129 PL_lex_state = LEX_INTERPCONCAT;
3130 start_force(curforce);
3131 NEXTVAL_NEXTTOKE.ival = 0;
3133 start_force(curforce);
3135 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3137 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3139 NEXTVAL_NEXTTOKE.ival = OP_LC;
3141 NEXTVAL_NEXTTOKE.ival = OP_UC;
3143 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3145 Perl_croak(aTHX_ "panic: yylex");
3147 SV* tmpsv = newSVpvn("",0);
3148 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3154 if (PL_lex_starts) {
3161 thistoken = newSVpvn("",0);
3164 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3165 if (PL_lex_casemods == 1 && PL_lex_inpat)
3174 case LEX_INTERPPUSH:
3175 return REPORT(sublex_push());
3177 case LEX_INTERPSTART:
3178 if (PL_bufptr == PL_bufend)
3179 return REPORT(sublex_done());
3180 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3181 "### Interpolated variable\n"); });
3183 PL_lex_dojoin = (*PL_bufptr == '@');
3184 PL_lex_state = LEX_INTERPNORMAL;
3185 if (PL_lex_dojoin) {
3186 start_force(curforce);
3187 NEXTVAL_NEXTTOKE.ival = 0;
3189 start_force(curforce);
3190 force_ident("\"", '$');
3191 start_force(curforce);
3192 NEXTVAL_NEXTTOKE.ival = 0;
3194 start_force(curforce);
3195 NEXTVAL_NEXTTOKE.ival = 0;
3197 start_force(curforce);
3198 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3201 if (PL_lex_starts++) {
3207 thistoken = newSVpvn("",0);
3210 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3211 if (!PL_lex_casemods && PL_lex_inpat)
3218 case LEX_INTERPENDMAYBE:
3219 if (intuit_more(PL_bufptr)) {
3220 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3226 if (PL_lex_dojoin) {
3227 PL_lex_dojoin = FALSE;
3228 PL_lex_state = LEX_INTERPCONCAT;
3233 thistoken = newSVpvn("",0);
3238 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3239 && SvEVALED(PL_lex_repl))
3241 if (PL_bufptr != PL_bufend)
3242 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3246 case LEX_INTERPCONCAT:
3248 if (PL_lex_brackets)
3249 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3251 if (PL_bufptr == PL_bufend)
3252 return REPORT(sublex_done());
3254 if (SvIVX(PL_linestr) == '\'') {
3255 SV *sv = newSVsv(PL_linestr);
3258 else if ( PL_hints & HINT_NEW_RE )
3259 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3260 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3264 s = scan_const(PL_bufptr);
3266 PL_lex_state = LEX_INTERPCASEMOD;
3268 PL_lex_state = LEX_INTERPSTART;
3271 if (s != PL_bufptr) {
3272 start_force(curforce);
3274 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3276 NEXTVAL_NEXTTOKE = yylval;
3279 if (PL_lex_starts++) {
3284 thistoken = newSVpvn("",0);
3287 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3288 if (!PL_lex_casemods && PL_lex_inpat)
3301 PL_lex_state = LEX_NORMAL;
3302 s = scan_formline(PL_bufptr);
3303 if (!PL_lex_formbrack)
3309 PL_oldoldbufptr = PL_oldbufptr;
3318 realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3322 if (isIDFIRST_lazy_if(s,UTF))
3324 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3327 goto fake_eof; /* emulate EOF on ^D or ^Z */
3336 if (PL_lex_brackets) {
3337 yyerror(PL_lex_formbrack
3338 ? "Format not terminated"
3339 : "Missing right curly or square bracket");
3341 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3342 "### Tokener got EOF\n");
3346 if (s++ < PL_bufend)
3347 goto retry; /* ignore stray nulls */
3350 if (!PL_in_eval && !PL_preambled) {
3351 PL_preambled = TRUE;
3356 sv_setpv(PL_linestr,incl_perldb());
3357 if (SvCUR(PL_linestr))
3358 sv_catpvs(PL_linestr,";");
3360 while(AvFILLp(PL_preambleav) >= 0) {
3361 SV *tmpsv = av_shift(PL_preambleav);
3362 sv_catsv(PL_linestr, tmpsv);
3363 sv_catpvs(PL_linestr, ";");
3366 sv_free((SV*)PL_preambleav);
3367 PL_preambleav = NULL;
3369 if (PL_minus_n || PL_minus_p) {
3370 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3372 sv_catpvs(PL_linestr,"chomp;");
3375 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3376 || *PL_splitstr == '"')
3377 && strchr(PL_splitstr + 1, *PL_splitstr))
3378 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3380 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3381 bytes can be used as quoting characters. :-) */
3382 const char *splits = PL_splitstr;
3383 sv_catpvs(PL_linestr, "our @F=split(q\0");
3386 if (*splits == '\\')
3387 sv_catpvn(PL_linestr, splits, 1);
3388 sv_catpvn(PL_linestr, splits, 1);
3389 } while (*splits++);
3390 /* This loop will embed the trailing NUL of
3391 PL_linestr as the last thing it does before
3393 sv_catpvs(PL_linestr, ");");
3397 sv_catpvs(PL_linestr,"our @F=split(' ');");
3401 sv_catpvs(PL_linestr,"use feature ':5.10';");
3402 sv_catpvs(PL_linestr, "\n");
3403 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3404 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3405 PL_last_lop = PL_last_uni = NULL;
3406 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3407 SV * const sv = newSV(0);
3409 sv_upgrade(sv, SVt_PVMG);
3410 sv_setsv(sv,PL_linestr);
3413 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3418 bof = PL_rsfp ? TRUE : FALSE;
3419 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3422 realtokenstart = -1;
3425 if (PL_preprocess && !PL_in_eval)
3426 (void)PerlProc_pclose(PL_rsfp);
3427 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3428 PerlIO_clearerr(PL_rsfp);
3430 (void)PerlIO_close(PL_rsfp);
3432 PL_doextract = FALSE;
3434 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3439 sv_setpv(PL_linestr,PL_minus_p
3440 ? ";}continue{print;}" : ";}");
3441 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3442 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3443 PL_last_lop = PL_last_uni = NULL;
3444 PL_minus_n = PL_minus_p = 0;
3447 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3448 PL_last_lop = PL_last_uni = NULL;
3449 sv_setpvn(PL_linestr,"",0);
3450 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3452 /* If it looks like the start of a BOM or raw UTF-16,
3453 * check if it in fact is. */
3459 #ifdef PERLIO_IS_STDIO
3460 # ifdef __GNU_LIBRARY__
3461 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3462 # define FTELL_FOR_PIPE_IS_BROKEN
3466 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3467 # define FTELL_FOR_PIPE_IS_BROKEN
3472 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3473 /* This loses the possibility to detect the bof
3474 * situation on perl -P when the libc5 is being used.
3475 * Workaround? Maybe attach some extra state to PL_rsfp?
3478 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3480 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3483 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3484 s = swallow_bom((U8*)s);
3488 /* Incest with pod. */
3491 sv_catsv(thiswhite, PL_linestr);
3493 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3494 sv_setpvn(PL_linestr, "", 0);
3495 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3496 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3497 PL_last_lop = PL_last_uni = NULL;
3498 PL_doextract = FALSE;
3502 } while (PL_doextract);
3503 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3504 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3505 SV * const sv = newSV(0);
3507 sv_upgrade(sv, SVt_PVMG);
3508 sv_setsv(sv,PL_linestr);
3511 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3513 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3514 PL_last_lop = PL_last_uni = NULL;
3515 if (CopLINE(PL_curcop) == 1) {
3516 while (s < PL_bufend && isSPACE(*s))
3518 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3522 thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3526 if (*s == '#' && *(s+1) == '!')
3528 #ifdef ALTERNATE_SHEBANG
3530 static char const as[] = ALTERNATE_SHEBANG;
3531 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3532 d = s + (sizeof(as) - 1);
3534 #endif /* ALTERNATE_SHEBANG */
3543 while (*d && !isSPACE(*d))
3547 #ifdef ARG_ZERO_IS_SCRIPT
3548 if (ipathend > ipath) {
3550 * HP-UX (at least) sets argv[0] to the script name,
3551 * which makes $^X incorrect. And Digital UNIX and Linux,
3552 * at least, set argv[0] to the basename of the Perl
3553 * interpreter. So, having found "#!", we'll set it right.
3555 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3557 assert(SvPOK(x) || SvGMAGICAL(x));
3558 if (sv_eq(x, CopFILESV(PL_curcop))) {
3559 sv_setpvn(x, ipath, ipathend - ipath);
3565 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3566 const char * const lstart = SvPV_const(x,llen);
3568 bstart += blen - llen;
3569 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3570 sv_setpvn(x, ipath, ipathend - ipath);
3575 TAINT_NOT; /* $^X is always tainted, but that's OK */
3577 #endif /* ARG_ZERO_IS_SCRIPT */
3582 d = instr(s,"perl -");
3584 d = instr(s,"perl");
3586 /* avoid getting into infinite loops when shebang
3587 * line contains "Perl" rather than "perl" */
3589 for (d = ipathend-4; d >= ipath; --d) {
3590 if ((*d == 'p' || *d == 'P')
3591 && !ibcmp(d, "perl", 4))
3601 #ifdef ALTERNATE_SHEBANG
3603 * If the ALTERNATE_SHEBANG on this system starts with a
3604 * character that can be part of a Perl expression, then if
3605 * we see it but not "perl", we're probably looking at the
3606 * start of Perl code, not a request to hand off to some
3607 * other interpreter. Similarly, if "perl" is there, but
3608 * not in the first 'word' of the line, we assume the line
3609 * contains the start of the Perl program.
3611 if (d && *s != '#') {
3612 const char *c = ipath;
3613 while (*c && !strchr("; \t\r\n\f\v#", *c))
3616 d = NULL; /* "perl" not in first word; ignore */
3618 *s = '#'; /* Don't try to parse shebang line */
3620 #endif /* ALTERNATE_SHEBANG */
3621 #ifndef MACOS_TRADITIONAL
3626 !instr(s,"indir") &&
3627 instr(PL_origargv[0],"perl"))
3634 while (s < PL_bufend && isSPACE(*s))
3636 if (s < PL_bufend) {
3637 Newxz(newargv,PL_origargc+3,char*);
3639 while (s < PL_bufend && !isSPACE(*s))
3642 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3645 newargv = PL_origargv;
3648 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3650 Perl_croak(aTHX_ "Can't exec %s", ipath);
3654 while (*d && !isSPACE(*d)) d++;
3655 while (SPACE_OR_TAB(*d)) d++;
3658 const bool switches_done = PL_doswitches;
3659 const U32 oldpdb = PL_perldb;
3660 const bool oldn = PL_minus_n;
3661 const bool oldp = PL_minus_p;
3664 if (*d == 'M' || *d == 'm' || *d == 'C') {
3665 const char * const m = d;
3666 while (*d && !isSPACE(*d)) d++;
3667 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3670 d = moreswitches(d);
3672 if (PL_doswitches && !switches_done) {
3673 int argc = PL_origargc;
3674 char **argv = PL_origargv;
3677 } while (argc && argv[0][0] == '-' && argv[0][1]);
3678 init_argv_symbols(argc,argv);
3680 if ((PERLDB_LINE && !oldpdb) ||
3681 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3682 /* if we have already added "LINE: while (<>) {",
3683 we must not do it again */
3685 sv_setpvn(PL_linestr, "", 0);
3686 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3687 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3688 PL_last_lop = PL_last_uni = NULL;
3689 PL_preambled = FALSE;
3691 (void)gv_fetchfile(PL_origfilename);
3698 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3700 PL_lex_state = LEX_FORMLINE;
3705 #ifdef PERL_STRICT_CR
3706 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3708 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3710 case ' ': case '\t': case '\f': case 013:
3711 #ifdef MACOS_TRADITIONAL
3715 realtokenstart = -1;
3724 realtokenstart = -1;
3728 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3729 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3730 /* handle eval qq[#line 1 "foo"\n ...] */
3731 CopLINE_dec(PL_curcop);
3734 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3736 if (!PL_in_eval || PL_rsfp)
3741 while (d < PL_bufend && *d != '\n')
3745 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3746 Perl_croak(aTHX_ "panic: input overflow");
3749 thiswhite = newSVpvn(s, d - s);
3754 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3756 PL_lex_state = LEX_FORMLINE;
3762 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3763 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3766 TOKEN(PEG); /* make sure any #! line is accessible */
3771 /* if (PL_madskills && PL_lex_formbrack) { */
3773 while (d < PL_bufend && *d != '\n')
3777 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3778 Perl_croak(aTHX_ "panic: input overflow");
3779 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3781 thiswhite = newSVpvn("",0);
3782 if (CopLINE(PL_curcop) == 1) {
3783 sv_setpvn(thiswhite, "", 0);
3786 sv_catpvn(thiswhite, s, d - s);
3800 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3808 while (s < PL_bufend && SPACE_OR_TAB(*s))
3811 if (strnEQ(s,"=>",2)) {
3812 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3813 DEBUG_T( { S_printbuf(aTHX_
3814 "### Saw unary minus before =>, forcing word %s\n", s);
3816 OPERATOR('-'); /* unary minus */
3818 PL_last_uni = PL_oldbufptr;
3820 case 'r': ftst = OP_FTEREAD; break;
3821 case 'w': ftst = OP_FTEWRITE; break;
3822 case 'x': ftst = OP_FTEEXEC; break;
3823 case 'o': ftst = OP_FTEOWNED; break;
3824 case 'R': ftst = OP_FTRREAD; break;
3825 case 'W': ftst = OP_FTRWRITE; break;
3826 case 'X': ftst = OP_FTREXEC; break;
3827 case 'O': ftst = OP_FTROWNED; break;
3828 case 'e': ftst = OP_FTIS; break;
3829 case 'z': ftst = OP_FTZERO; break;
3830 case 's': ftst = OP_FTSIZE; break;
3831 case 'f': ftst = OP_FTFILE; break;
3832 case 'd': ftst = OP_FTDIR; break;
3833 case 'l': ftst = OP_FTLINK; break;
3834 case 'p': ftst = OP_FTPIPE; break;
3835 case 'S': ftst = OP_FTSOCK; break;
3836 case 'u': ftst = OP_FTSUID; break;
3837 case 'g': ftst = OP_FTSGID; break;
3838 case 'k': ftst = OP_FTSVTX; break;
3839 case 'b': ftst = OP_FTBLK; break;
3840 case 'c': ftst = OP_FTCHR; break;
3841 case 't': ftst = OP_FTTTY; break;
3842 case 'T': ftst = OP_FTTEXT; break;
3843 case 'B': ftst = OP_FTBINARY; break;
3844 case 'M': case 'A': case 'C':
3845 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3847 case 'M': ftst = OP_FTMTIME; break;
3848 case 'A': ftst = OP_FTATIME; break;
3849 case 'C': ftst = OP_FTCTIME; break;
3857 PL_last_lop_op = (OPCODE)ftst;
3858 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3859 "### Saw file test %c\n", (int)tmp);
3864 /* Assume it was a minus followed by a one-letter named
3865 * subroutine call (or a -bareword), then. */
3866 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3867 "### '-%c' looked like a file test but was not\n",
3874 const char tmp = *s++;
3877 if (PL_expect == XOPERATOR)
3882 else if (*s == '>') {
3885 if (isIDFIRST_lazy_if(s,UTF)) {
3886 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3894 if (PL_expect == XOPERATOR)
3897 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3899 OPERATOR('-'); /* unary minus */
3905 const char tmp = *s++;
3908 if (PL_expect == XOPERATOR)
3913 if (PL_expect == XOPERATOR)
3916 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3923 if (PL_expect != XOPERATOR) {
3924 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3925 PL_expect = XOPERATOR;
3926 force_ident(PL_tokenbuf, '*');
3939 if (PL_expect == XOPERATOR) {
3943 PL_tokenbuf[0] = '%';
3944 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3945 if (!PL_tokenbuf[1]) {
3948 PL_pending_ident = '%';
3959 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3960 && FEATURE_IS_ENABLED("~~"))
3967 const char tmp = *s++;
3973 goto just_a_word_zero_gv;
3976 switch (PL_expect) {
3982 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3984 PL_bufptr = s; /* update in case we back off */
3990 PL_expect = XTERMBLOCK;
3993 stuffstart = s - SvPVX(PL_linestr) - 1;
3997 while (isIDFIRST_lazy_if(s,UTF)) {
3999 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4000 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4001 if (tmp < 0) tmp = -tmp;
4017 d = scan_str(d,TRUE,TRUE);
4019 /* MUST advance bufptr here to avoid bogus
4020 "at end of line" context messages from yyerror().
4022 PL_bufptr = s + len;
4023 yyerror("Unterminated attribute parameter in attribute list");
4026 return REPORT(0); /* EOF indicator */
4030 SV *sv = newSVpvn(s, len);
4031 sv_catsv(sv, PL_lex_stuff);
4032 attrs = append_elem(OP_LIST, attrs,
4033 newSVOP(OP_CONST, 0, sv));
4034 SvREFCNT_dec(PL_lex_stuff);
4035 PL_lex_stuff = NULL;
4038 if (len == 6 && strnEQ(s, "unique", len)) {
4039 if (PL_in_my == KEY_our)
4041 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4043 /*EMPTY*/; /* skip to avoid loading attributes.pm */
4046 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4049 /* NOTE: any CV attrs applied here need to be part of
4050 the CVf_BUILTIN_ATTRS define in cv.h! */
4051 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
4052 CvLVALUE_on(PL_compcv);
4053 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
4054 CvLOCKED_on(PL_compcv);
4055 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
4056 CvMETHOD_on(PL_compcv);
4057 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
4058 CvASSERTION_on(PL_compcv);
4059 /* After we've set the flags, it could be argued that
4060 we don't need to do the attributes.pm-based setting
4061 process, and shouldn't bother appending recognized
4062 flags. To experiment with that, uncomment the
4063 following "else". (Note that's already been
4064 uncommented. That keeps the above-applied built-in
4065 attributes from being intercepted (and possibly
4066 rejected) by a package's attribute routines, but is
4067 justified by the performance win for the common case
4068 of applying only built-in attributes.) */
4070 attrs = append_elem(OP_LIST, attrs,
4071 newSVOP(OP_CONST, 0,
4075 if (*s == ':' && s[1] != ':')
4078 break; /* require real whitespace or :'s */
4079 /* XXX losing whitespace on sequential attributes here */
4083 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4084 if (*s != ';' && *s != '}' && *s != tmp
4085 && (tmp != '=' || *s != ')')) {
4086 const char q = ((*s == '\'') ? '"' : '\'');
4087 /* If here for an expression, and parsed no attrs, back
4089 if (tmp == '=' && !attrs) {
4093 /* MUST advance bufptr here to avoid bogus "at end of line"
4094 context messages from yyerror().
4098 ? Perl_form(aTHX_ "Invalid separator character "
4099 "%c%c%c in attribute list", q, *s, q)
4100 : "Unterminated attribute list" );
4108 start_force(curforce);
4109 NEXTVAL_NEXTTOKE.opval = attrs;
4110 CURMAD('_', nextwhite);
4115 thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4116 (s - SvPVX(PL_linestr)) - stuffstart);
4124 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4125 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4133 const char tmp = *s++;
4138 const char tmp = *s++;
4146 if (PL_lex_brackets <= 0)
4147 yyerror("Unmatched right square bracket");
4150 if (PL_lex_state == LEX_INTERPNORMAL) {
4151 if (PL_lex_brackets == 0) {
4152 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4153 PL_lex_state = LEX_INTERPEND;
4160 if (PL_lex_brackets > 100) {
4161 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4163 switch (PL_expect) {
4165 if (PL_lex_formbrack) {
4169 if (PL_oldoldbufptr == PL_last_lop)
4170 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4172 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4173 OPERATOR(HASHBRACK);
4175 while (s < PL_bufend && SPACE_OR_TAB(*s))
4178 PL_tokenbuf[0] = '\0';
4179 if (d < PL_bufend && *d == '-') {
4180 PL_tokenbuf[0] = '-';
4182 while (d < PL_bufend && SPACE_OR_TAB(*d))
4185 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4186 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4188 while (d < PL_bufend && SPACE_OR_TAB(*d))
4191 const char minus = (PL_tokenbuf[0] == '-');
4192 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4200 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4205 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4210 if (PL_oldoldbufptr == PL_last_lop)
4211 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4213 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4216 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4218 /* This hack is to get the ${} in the message. */
4220 yyerror("syntax error");
4223 OPERATOR(HASHBRACK);
4225 /* This hack serves to disambiguate a pair of curlies
4226 * as being a block or an anon hash. Normally, expectation
4227 * determines that, but in cases where we're not in a
4228 * position to expect anything in particular (like inside
4229 * eval"") we have to resolve the ambiguity. This code
4230 * covers the case where the first term in the curlies is a
4231 * quoted string. Most other cases need to be explicitly
4232 * disambiguated by prepending a "+" before the opening
4233 * curly in order to force resolution as an anon hash.
4235 * XXX should probably propagate the outer expectation
4236 * into eval"" to rely less on this hack, but that could
4237 * potentially break current behavior of eval"".
4241 if (*s == '\'' || *s == '"' || *s == '`') {
4242 /* common case: get past first string, handling escapes */
4243 for (t++; t < PL_bufend && *t != *s;)
4244 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4248 else if (*s == 'q') {
4251 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4254 /* skip q//-like construct */
4256 char open, close, term;
4259 while (t < PL_bufend && isSPACE(*t))
4261 /* check for q => */
4262 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4263 OPERATOR(HASHBRACK);
4267 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4271 for (t++; t < PL_bufend; t++) {
4272 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4274 else if (*t == open)
4278 for (t++; t < PL_bufend; t++) {
4279 if (*t == '\\' && t+1 < PL_bufend)
4281 else if (*t == close && --brackets <= 0)
4283 else if (*t == open)
4290 /* skip plain q word */
4291 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4294 else if (isALNUM_lazy_if(t,UTF)) {
4296 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4299 while (t < PL_bufend && isSPACE(*t))
4301 /* if comma follows first term, call it an anon hash */
4302 /* XXX it could be a comma expression with loop modifiers */
4303 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4304 || (*t == '=' && t[1] == '>')))
4305 OPERATOR(HASHBRACK);
4306 if (PL_expect == XREF)
4309 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4315 yylval.ival = CopLINE(PL_curcop);
4316 if (isSPACE(*s) || *s == '#')
4317 PL_copline = NOLINE; /* invalidate current command line number */
4322 if (PL_lex_brackets <= 0)
4323 yyerror("Unmatched right curly bracket");
4325 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4326 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4327 PL_lex_formbrack = 0;
4328 if (PL_lex_state == LEX_INTERPNORMAL) {
4329 if (PL_lex_brackets == 0) {
4330 if (PL_expect & XFAKEBRACK) {
4331 PL_expect &= XENUMMASK;
4332 PL_lex_state = LEX_INTERPEND;
4337 thiswhite = newSVpvn("",0);
4338 sv_catpvn(thiswhite,"}",1);
4341 return yylex(); /* ignore fake brackets */
4343 if (*s == '-' && s[1] == '>')
4344 PL_lex_state = LEX_INTERPENDMAYBE;
4345 else if (*s != '[' && *s != '{')
4346 PL_lex_state = LEX_INTERPEND;
4349 if (PL_expect & XFAKEBRACK) {
4350 PL_expect &= XENUMMASK;
4352 return yylex(); /* ignore fake brackets */
4354 start_force(curforce);
4356 curmad('X', newSVpvn(s-1,1));
4357 CURMAD('_', thiswhite);
4362 thistoken = newSVpvn("",0);
4370 if (PL_expect == XOPERATOR) {
4371 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4372 && isIDFIRST_lazy_if(s,UTF))
4374 CopLINE_dec(PL_curcop);
4375 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4376 CopLINE_inc(PL_curcop);
4381 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4383 PL_expect = XOPERATOR;
4384 force_ident(PL_tokenbuf, '&');
4388 yylval.ival = (OPpENTERSUB_AMPER<<8);
4400 const char tmp = *s++;