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 /* XXX Things like this are just so nasty. We shouldn't be modifying
1093 source code, even if we realquick set it back. */
1094 if (ckWARN_d(WARN_AMBIGUOUS)){
1097 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1098 "Warning: Use of \"%s\" without parentheses is ambiguous",
1105 * LOP : macro to build a list operator. Its behaviour has been replaced
1106 * with a subroutine, S_lop() for which LOP is just another name.
1109 #define LOP(f,x) return lop(f,x,s)
1113 * Build a list operator (or something that might be one). The rules:
1114 * - if we have a next token, then it's a list operator [why?]
1115 * - if the next thing is an opening paren, then it's a function
1116 * - else it's a list operator
1120 S_lop(pTHX_ I32 f, int x, char *s)
1127 PL_last_lop = PL_oldbufptr;
1128 PL_last_lop_op = (OPCODE)f;
1131 return REPORT(LSTOP);
1134 return REPORT(LSTOP);
1137 return REPORT(FUNC);
1140 return REPORT(FUNC);
1142 return REPORT(LSTOP);
1148 * Sets up for an eventual force_next(). start_force(0) basically does
1149 * an unshift, while start_force(-1) does a push. yylex removes items
1154 S_start_force(pTHX_ int where)
1158 if (where < 0) /* so people can duplicate start_force(curforce) */
1159 where = PL_lasttoke;
1160 assert(curforce < 0 || curforce == where);
1161 if (curforce != where) {
1162 for (i = PL_lasttoke; i > where; --i) {
1163 PL_nexttoke[i] = PL_nexttoke[i-1];
1167 if (curforce < 0) /* in case of duplicate start_force() */
1168 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1172 curmad('^', newSVpvn("",0));
1173 CURMAD('_', nextwhite);
1178 S_curmad(pTHX_ char slot, SV *sv)
1187 where = &PL_nexttoke[curforce].next_mad;
1190 sv_setpvn(sv, "", 0);
1193 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1195 else if (PL_encoding) {
1196 sv_recode_to_utf8(sv, PL_encoding);
1201 /* keep a slot open for the head of the list? */
1202 if (slot != '_' && *where && (*where)->mad_key == '^') {
1203 (*where)->mad_key = slot;
1204 sv_free((*where)->mad_val);
1205 (*where)->mad_val = (void*)sv;
1208 addmad(newMADsv(slot, sv), where, 0);
1211 # define start_force(where)
1212 # define curmad(slot, sv)
1217 * When the lexer realizes it knows the next token (for instance,
1218 * it is reordering tokens for the parser) then it can call S_force_next
1219 * to know what token to return the next time the lexer is called. Caller
1220 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1221 * and possibly PL_expect to ensure the lexer handles the token correctly.
1225 S_force_next(pTHX_ I32 type)
1230 start_force(PL_lasttoke);
1231 PL_nexttoke[curforce].next_type = type;
1232 if (PL_lex_state != LEX_KNOWNEXT)
1233 PL_lex_defer = PL_lex_state;
1234 PL_lex_state = LEX_KNOWNEXT;
1235 PL_lex_expect = PL_expect;
1238 PL_nexttype[PL_nexttoke] = type;
1240 if (PL_lex_state != LEX_KNOWNEXT) {
1241 PL_lex_defer = PL_lex_state;
1242 PL_lex_expect = PL_expect;
1243 PL_lex_state = LEX_KNOWNEXT;
1249 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1252 SV * const sv = newSVpvn(start,len);
1253 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1260 * When the lexer knows the next thing is a word (for instance, it has
1261 * just seen -> and it knows that the next char is a word char, then
1262 * it calls S_force_word to stick the next word into the PL_next lookahead.
1265 * char *start : buffer position (must be within PL_linestr)
1266 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1267 * int check_keyword : if true, Perl checks to make sure the word isn't
1268 * a keyword (do this if the word is a label, e.g. goto FOO)
1269 * int allow_pack : if true, : characters will also be allowed (require,
1270 * use, etc. do this)
1271 * int allow_initial_tick : used by the "sub" lexer only.
1275 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1281 start = SKIPSPACE1(start);
1283 if (isIDFIRST_lazy_if(s,UTF) ||
1284 (allow_pack && *s == ':') ||
1285 (allow_initial_tick && *s == '\'') )
1287 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1288 if (check_keyword && keyword(PL_tokenbuf, len))
1290 start_force(curforce);
1292 curmad('X', newSVpvn(start,s-start));
1293 if (token == METHOD) {
1298 PL_expect = XOPERATOR;
1301 NEXTVAL_NEXTTOKE.opval
1302 = (OP*)newSVOP(OP_CONST,0,
1303 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1304 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1312 * Called when the lexer wants $foo *foo &foo etc, but the program
1313 * text only contains the "foo" portion. The first argument is a pointer
1314 * to the "foo", and the second argument is the type symbol to prefix.
1315 * Forces the next token to be a "WORD".
1316 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1320 S_force_ident(pTHX_ register const char *s, int kind)
1324 const STRLEN len = strlen(s);
1325 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1326 start_force(curforce);
1327 NEXTVAL_NEXTTOKE.opval = o;
1330 o->op_private = OPpCONST_ENTERED;
1331 /* XXX see note in pp_entereval() for why we forgo typo
1332 warnings if the symbol must be introduced in an eval.
1334 gv_fetchpvn_flags(s, len,
1335 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1337 kind == '$' ? SVt_PV :
1338 kind == '@' ? SVt_PVAV :
1339 kind == '%' ? SVt_PVHV :
1347 Perl_str_to_version(pTHX_ SV *sv)
1352 const char *start = SvPV_const(sv,len);
1353 const char * const end = start + len;
1354 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1355 while (start < end) {
1359 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1364 retval += ((NV)n)/nshift;
1373 * Forces the next token to be a version number.
1374 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1375 * and if "guessing" is TRUE, then no new token is created (and the caller
1376 * must use an alternative parsing method).
1380 S_force_version(pTHX_ char *s, int guessing)
1386 I32 startoff = s - SvPVX(PL_linestr);
1395 while (isDIGIT(*d) || *d == '_' || *d == '.')
1399 start_force(curforce);
1400 curmad('X', newSVpvn(s,d-s));
1403 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1405 s = scan_num(s, &yylval);
1406 version = yylval.opval;
1407 ver = cSVOPx(version)->op_sv;
1408 if (SvPOK(ver) && !SvNIOK(ver)) {
1409 SvUPGRADE(ver, SVt_PVNV);
1410 SvNV_set(ver, str_to_version(ver));
1411 SvNOK_on(ver); /* hint that it is a version */
1414 else if (guessing) {
1417 sv_free(nextwhite); /* let next token collect whitespace */
1419 s = SvPVX(PL_linestr) + startoff;
1427 if (PL_madskills && !version) {
1428 sv_free(nextwhite); /* let next token collect whitespace */
1430 s = SvPVX(PL_linestr) + startoff;
1433 /* NOTE: The parser sees the package name and the VERSION swapped */
1434 start_force(curforce);
1435 NEXTVAL_NEXTTOKE.opval = version;
1443 * Tokenize a quoted string passed in as an SV. It finds the next
1444 * chunk, up to end of string or a backslash. It may make a new
1445 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1450 S_tokeq(pTHX_ SV *sv)
1454 register char *send;
1462 s = SvPV_force(sv, len);
1463 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1466 while (s < send && *s != '\\')
1471 if ( PL_hints & HINT_NEW_STRING ) {
1472 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1478 if (s + 1 < send && (s[1] == '\\'))
1479 s++; /* all that, just for this */
1484 SvCUR_set(sv, d - SvPVX_const(sv));
1486 if ( PL_hints & HINT_NEW_STRING )
1487 return new_constant(NULL, 0, "q", sv, pv, "q");
1492 * Now come three functions related to double-quote context,
1493 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1494 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1495 * interact with PL_lex_state, and create fake ( ... ) argument lists
1496 * to handle functions and concatenation.
1497 * They assume that whoever calls them will be setting up a fake
1498 * join call, because each subthing puts a ',' after it. This lets
1501 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1503 * (I'm not sure whether the spurious commas at the end of lcfirst's
1504 * arguments and join's arguments are created or not).
1509 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1511 * Pattern matching will set PL_lex_op to the pattern-matching op to
1512 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1514 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1516 * Everything else becomes a FUNC.
1518 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1519 * had an OP_CONST or OP_READLINE). This just sets us up for a
1520 * call to S_sublex_push().
1524 S_sublex_start(pTHX)
1527 register const I32 op_type = yylval.ival;
1529 if (op_type == OP_NULL) {
1530 yylval.opval = PL_lex_op;
1534 if (op_type == OP_CONST || op_type == OP_READLINE) {
1535 SV *sv = tokeq(PL_lex_stuff);
1537 if (SvTYPE(sv) == SVt_PVIV) {
1538 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1540 const char * const p = SvPV_const(sv, len);
1541 SV * const nsv = newSVpvn(p, len);
1547 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1548 PL_lex_stuff = NULL;
1549 /* Allow <FH> // "foo" */
1550 if (op_type == OP_READLINE)
1551 PL_expect = XTERMORDORDOR;
1555 PL_sublex_info.super_state = PL_lex_state;
1556 PL_sublex_info.sub_inwhat = op_type;
1557 PL_sublex_info.sub_op = PL_lex_op;
1558 PL_lex_state = LEX_INTERPPUSH;
1562 yylval.opval = PL_lex_op;
1572 * Create a new scope to save the lexing state. The scope will be
1573 * ended in S_sublex_done. Returns a '(', starting the function arguments
1574 * to the uc, lc, etc. found before.
1575 * Sets PL_lex_state to LEX_INTERPCONCAT.
1584 PL_lex_state = PL_sublex_info.super_state;
1585 SAVEI32(PL_lex_dojoin);
1586 SAVEI32(PL_lex_brackets);
1587 SAVEI32(PL_lex_casemods);
1588 SAVEI32(PL_lex_starts);
1589 SAVEI32(PL_lex_state);
1590 SAVEVPTR(PL_lex_inpat);
1591 SAVEI32(PL_lex_inwhat);
1592 SAVECOPLINE(PL_curcop);
1593 SAVEPPTR(PL_bufptr);
1594 SAVEPPTR(PL_bufend);
1595 SAVEPPTR(PL_oldbufptr);
1596 SAVEPPTR(PL_oldoldbufptr);
1597 SAVEPPTR(PL_last_lop);
1598 SAVEPPTR(PL_last_uni);
1599 SAVEPPTR(PL_linestart);
1600 SAVESPTR(PL_linestr);
1601 SAVEGENERICPV(PL_lex_brackstack);
1602 SAVEGENERICPV(PL_lex_casestack);
1604 PL_linestr = PL_lex_stuff;
1605 PL_lex_stuff = NULL;
1607 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1608 = SvPVX(PL_linestr);
1609 PL_bufend += SvCUR(PL_linestr);
1610 PL_last_lop = PL_last_uni = NULL;
1611 SAVEFREESV(PL_linestr);
1613 PL_lex_dojoin = FALSE;
1614 PL_lex_brackets = 0;
1615 Newx(PL_lex_brackstack, 120, char);
1616 Newx(PL_lex_casestack, 12, char);
1617 PL_lex_casemods = 0;
1618 *PL_lex_casestack = '\0';
1620 PL_lex_state = LEX_INTERPCONCAT;
1621 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1623 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1624 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1625 PL_lex_inpat = PL_sublex_info.sub_op;
1627 PL_lex_inpat = NULL;
1634 * Restores lexer state after a S_sublex_push.
1641 if (!PL_lex_starts++) {
1642 SV * const sv = newSVpvs("");
1643 if (SvUTF8(PL_linestr))
1645 PL_expect = XOPERATOR;
1646 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1650 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1651 PL_lex_state = LEX_INTERPCASEMOD;
1655 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1656 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1657 PL_linestr = PL_lex_repl;
1659 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1660 PL_bufend += SvCUR(PL_linestr);
1661 PL_last_lop = PL_last_uni = NULL;
1662 SAVEFREESV(PL_linestr);
1663 PL_lex_dojoin = FALSE;
1664 PL_lex_brackets = 0;
1665 PL_lex_casemods = 0;
1666 *PL_lex_casestack = '\0';
1668 if (SvEVALED(PL_lex_repl)) {
1669 PL_lex_state = LEX_INTERPNORMAL;
1671 /* we don't clear PL_lex_repl here, so that we can check later
1672 whether this is an evalled subst; that means we rely on the
1673 logic to ensure sublex_done() is called again only via the
1674 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1677 PL_lex_state = LEX_INTERPCONCAT;
1687 endwhite = newSVpvn("",0);
1688 sv_catsv(endwhite, thiswhite);
1692 sv_setpvn(thistoken,"",0);
1694 realtokenstart = -1;
1698 PL_bufend = SvPVX(PL_linestr);
1699 PL_bufend += SvCUR(PL_linestr);
1700 PL_expect = XOPERATOR;
1701 PL_sublex_info.sub_inwhat = 0;
1709 Extracts a pattern, double-quoted string, or transliteration. This
1712 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1713 processing a pattern (PL_lex_inpat is true), a transliteration
1714 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1716 Returns a pointer to the character scanned up to. Iff this is
1717 advanced from the start pointer supplied (ie if anything was
1718 successfully parsed), will leave an OP for the substring scanned
1719 in yylval. Caller must intuit reason for not parsing further
1720 by looking at the next characters herself.
1724 double-quoted style: \r and \n
1725 regexp special ones: \D \s
1727 backrefs: \1 (deprecated in substitution replacements)
1728 case and quoting: \U \Q \E
1729 stops on @ and $, but not for $ as tail anchor
1731 In transliterations:
1732 characters are VERY literal, except for - not at the start or end
1733 of the string, which indicates a range. scan_const expands the
1734 range to the full set of intermediate characters.
1736 In double-quoted strings:
1738 double-quoted style: \r and \n
1740 backrefs: \1 (deprecated)
1741 case and quoting: \U \Q \E
1744 scan_const does *not* construct ops to handle interpolated strings.
1745 It stops processing as soon as it finds an embedded $ or @ variable
1746 and leaves it to the caller to work out what's going on.
1748 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1750 $ in pattern could be $foo or could be tail anchor. Assumption:
1751 it's a tail anchor if $ is the last thing in the string, or if it's
1752 followed by one of ")| \n\t"
1754 \1 (backreferences) are turned into $1
1756 The structure of the code is
1757 while (there's a character to process) {
1758 handle transliteration ranges
1759 skip regexp comments
1760 skip # initiated comments in //x patterns
1761 check for embedded @foo
1762 check for embedded scalars
1764 leave intact backslashes from leave (below)
1765 deprecate \1 in strings and sub replacements
1766 handle string-changing backslashes \l \U \Q \E, etc.
1767 switch (what was escaped) {
1768 handle - in a transliteration (becomes a literal -)
1769 handle \132 octal characters
1770 handle 0x15 hex characters
1771 handle \cV (control V)
1772 handle printf backslashes (\f, \r, \n, etc)
1774 } (end if backslash)
1775 } (end while character to read)
1780 S_scan_const(pTHX_ char *start)
1783 register char *send = PL_bufend; /* end of the constant */
1784 SV *sv = newSV(send - start); /* sv for the constant */
1785 register char *s = start; /* start of the constant */
1786 register char *d = SvPVX(sv); /* destination for copies */
1787 bool dorange = FALSE; /* are we in a translit range? */
1788 bool didrange = FALSE; /* did we just finish a range? */
1789 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1790 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1793 UV literal_endpoint = 0;
1796 const char *leaveit = /* set of acceptably-backslashed characters */
1798 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1801 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1802 /* If we are doing a trans and we know we want UTF8 set expectation */
1803 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1804 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1808 while (s < send || dorange) {
1809 /* get transliterations out of the way (they're most literal) */
1810 if (PL_lex_inwhat == OP_TRANS) {
1811 /* expand a range A-Z to the full set of characters. AIE! */
1813 I32 i; /* current expanded character */
1814 I32 min; /* first character in range */
1815 I32 max; /* last character in range */
1818 char * const c = (char*)utf8_hop((U8*)d, -1);
1822 *c = (char)UTF_TO_NATIVE(0xff);
1823 /* mark the range as done, and continue */
1829 i = d - SvPVX_const(sv); /* remember current offset */
1830 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1831 d = SvPVX(sv) + i; /* refresh d after realloc */
1832 d -= 2; /* eat the first char and the - */
1834 min = (U8)*d; /* first char in range */
1835 max = (U8)d[1]; /* last char in range */
1839 "Invalid range \"%c-%c\" in transliteration operator",
1840 (char)min, (char)max);
1844 if (literal_endpoint == 2 &&
1845 ((isLOWER(min) && isLOWER(max)) ||
1846 (isUPPER(min) && isUPPER(max)))) {
1848 for (i = min; i <= max; i++)
1850 *d++ = NATIVE_TO_NEED(has_utf8,i);
1852 for (i = min; i <= max; i++)
1854 *d++ = NATIVE_TO_NEED(has_utf8,i);
1859 for (i = min; i <= max; i++)
1862 /* mark the range as done, and continue */
1866 literal_endpoint = 0;
1871 /* range begins (ignore - as first or last char) */
1872 else if (*s == '-' && s+1 < send && s != start) {
1874 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1877 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1887 literal_endpoint = 0;
1892 /* if we get here, we're not doing a transliteration */
1894 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1895 except for the last char, which will be done separately. */
1896 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1898 while (s+1 < send && *s != ')')
1899 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1901 else if (s[2] == '{' /* This should match regcomp.c */
1902 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1905 char *regparse = s + (s[2] == '{' ? 3 : 4);
1908 while (count && (c = *regparse)) {
1909 if (c == '\\' && regparse[1])
1917 if (*regparse != ')')
1918 regparse--; /* Leave one char for continuation. */
1919 while (s < regparse)
1920 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1924 /* likewise skip #-initiated comments in //x patterns */
1925 else if (*s == '#' && PL_lex_inpat &&
1926 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1927 while (s+1 < send && *s != '\n')
1928 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1931 /* check for embedded arrays
1932 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1934 else if (*s == '@' && s[1]
1935 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1938 /* check for embedded scalars. only stop if we're sure it's a
1941 else if (*s == '$') {
1942 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1944 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1945 break; /* in regexp, $ might be tail anchor */
1948 /* End of else if chain - OP_TRANS rejoin rest */
1951 if (*s == '\\' && s+1 < send) {
1954 /* some backslashes we leave behind */
1955 if (*leaveit && *s && strchr(leaveit, *s)) {
1956 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1957 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1961 /* deprecate \1 in strings and substitution replacements */
1962 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1963 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1965 if (ckWARN(WARN_SYNTAX))
1966 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1971 /* string-change backslash escapes */
1972 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1977 /* if we get here, it's either a quoted -, or a digit */
1980 /* quoted - in transliterations */
1982 if (PL_lex_inwhat == OP_TRANS) {
1992 Perl_warner(aTHX_ packWARN(WARN_MISC),
1993 "Unrecognized escape \\%c passed through",
1995 /* default action is to copy the quoted character */
1996 goto default_action;
1999 /* \132 indicates an octal constant */
2000 case '0': case '1': case '2': case '3':
2001 case '4': case '5': case '6': case '7':
2005 uv = grok_oct(s, &len, &flags, NULL);
2008 goto NUM_ESCAPE_INSERT;
2010 /* \x24 indicates a hex constant */
2014 char* const e = strchr(s, '}');
2015 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2016 PERL_SCAN_DISALLOW_PREFIX;
2021 yyerror("Missing right brace on \\x{}");
2025 uv = grok_hex(s, &len, &flags, NULL);
2031 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2032 uv = grok_hex(s, &len, &flags, NULL);
2038 /* Insert oct or hex escaped character.
2039 * There will always enough room in sv since such
2040 * escapes will be longer than any UTF-8 sequence
2041 * they can end up as. */
2043 /* We need to map to chars to ASCII before doing the tests
2046 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2047 if (!has_utf8 && uv > 255) {
2048 /* Might need to recode whatever we have
2049 * accumulated so far if it contains any
2052 * (Can't we keep track of that and avoid
2053 * this rescan? --jhi)
2057 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2058 if (!NATIVE_IS_INVARIANT(*c)) {
2063 const STRLEN offset = d - SvPVX_const(sv);
2065 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2069 while (src >= (const U8 *)SvPVX_const(sv)) {
2070 if (!NATIVE_IS_INVARIANT(*src)) {
2071 const U8 ch = NATIVE_TO_ASCII(*src);
2072 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2073 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2083 if (has_utf8 || uv > 255) {
2084 d = (char*)uvchr_to_utf8((U8*)d, uv);
2086 if (PL_lex_inwhat == OP_TRANS &&
2087 PL_sublex_info.sub_op) {
2088 PL_sublex_info.sub_op->op_private |=
2089 (PL_lex_repl ? OPpTRANS_FROM_UTF
2102 /* \N{LATIN SMALL LETTER A} is a named character */
2106 char* e = strchr(s, '}');
2112 yyerror("Missing right brace on \\N{}");
2116 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2118 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2119 PERL_SCAN_DISALLOW_PREFIX;
2122 uv = grok_hex(s, &len, &flags, NULL);
2124 goto NUM_ESCAPE_INSERT;
2126 res = newSVpvn(s + 1, e - s - 1);
2127 res = new_constant( NULL, 0, "charnames",
2128 res, NULL, "\\N{...}" );
2130 sv_utf8_upgrade(res);
2131 str = SvPV_const(res,len);
2132 #ifdef EBCDIC_NEVER_MIND
2133 /* charnames uses pack U and that has been
2134 * recently changed to do the below uni->native
2135 * mapping, so this would be redundant (and wrong,
2136 * the code point would be doubly converted).
2137 * But leave this in just in case the pack U change
2138 * gets revoked, but the semantics is still
2139 * desireable for charnames. --jhi */
2141 UV uv = utf8_to_uvchr((const U8*)str, 0);
2144 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2146 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2147 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2148 str = SvPV_const(res, len);
2152 if (!has_utf8 && SvUTF8(res)) {
2153 const char * const ostart = SvPVX_const(sv);
2154 SvCUR_set(sv, d - ostart);
2157 sv_utf8_upgrade(sv);
2158 /* this just broke our allocation above... */
2159 SvGROW(sv, (STRLEN)(send - start));
2160 d = SvPVX(sv) + SvCUR(sv);
2163 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2164 const char * const odest = SvPVX_const(sv);
2166 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2167 d = SvPVX(sv) + (d - odest);
2169 Copy(str, d, len, char);
2176 yyerror("Missing braces on \\N{}");
2179 /* \c is a control character */
2188 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2191 yyerror("Missing control char name in \\c");
2195 /* printf-style backslashes, formfeeds, newlines, etc */
2197 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2200 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2203 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2206 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2209 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2212 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2215 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2221 } /* end if (backslash) */
2228 /* If we started with encoded form, or already know we want it
2229 and then encode the next character */
2230 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2232 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2233 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2236 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2237 const STRLEN off = d - SvPVX_const(sv);
2238 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2240 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2244 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2246 } /* while loop to process each character */
2248 /* terminate the string and set up the sv */
2250 SvCUR_set(sv, d - SvPVX_const(sv));
2251 if (SvCUR(sv) >= SvLEN(sv))
2252 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2255 if (PL_encoding && !has_utf8) {
2256 sv_recode_to_utf8(sv, PL_encoding);
2262 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2263 PL_sublex_info.sub_op->op_private |=
2264 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2268 /* shrink the sv if we allocated more than we used */
2269 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2270 SvPV_shrink_to_cur(sv);
2273 /* return the substring (via yylval) only if we parsed anything */
2274 if (s > PL_bufptr) {
2275 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2276 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2278 ( PL_lex_inwhat == OP_TRANS
2280 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2283 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2290 * Returns TRUE if there's more to the expression (e.g., a subscript),
2293 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2295 * ->[ and ->{ return TRUE
2296 * { and [ outside a pattern are always subscripts, so return TRUE
2297 * if we're outside a pattern and it's not { or [, then return FALSE
2298 * if we're in a pattern and the first char is a {
2299 * {4,5} (any digits around the comma) returns FALSE
2300 * if we're in a pattern and the first char is a [
2302 * [SOMETHING] has a funky algorithm to decide whether it's a
2303 * character class or not. It has to deal with things like
2304 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2305 * anything else returns TRUE
2308 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2311 S_intuit_more(pTHX_ register char *s)
2314 if (PL_lex_brackets)
2316 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2318 if (*s != '{' && *s != '[')
2323 /* In a pattern, so maybe we have {n,m}. */
2340 /* On the other hand, maybe we have a character class */
2343 if (*s == ']' || *s == '^')
2346 /* this is terrifying, and it works */
2347 int weight = 2; /* let's weigh the evidence */
2349 unsigned char un_char = 255, last_un_char;
2350 const char * const send = strchr(s,']');
2351 char tmpbuf[sizeof PL_tokenbuf * 4];
2353 if (!send) /* has to be an expression */
2356 Zero(seen,256,char);
2359 else if (isDIGIT(*s)) {
2361 if (isDIGIT(s[1]) && s[2] == ']')
2367 for (; s < send; s++) {
2368 last_un_char = un_char;
2369 un_char = (unsigned char)*s;
2374 weight -= seen[un_char] * 10;
2375 if (isALNUM_lazy_if(s+1,UTF)) {
2377 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2378 len = (int)strlen(tmpbuf);
2379 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2384 else if (*s == '$' && s[1] &&
2385 strchr("[#!%*<>()-=",s[1])) {
2386 if (/*{*/ strchr("])} =",s[2]))
2395 if (strchr("wds]",s[1]))
2397 else if (seen['\''] || seen['"'])
2399 else if (strchr("rnftbxcav",s[1]))
2401 else if (isDIGIT(s[1])) {
2403 while (s[1] && isDIGIT(s[1]))
2413 if (strchr("aA01! ",last_un_char))
2415 if (strchr("zZ79~",s[1]))
2417 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2418 weight -= 5; /* cope with negative subscript */
2421 if (!isALNUM(last_un_char)
2422 && !(last_un_char == '$' || last_un_char == '@'
2423 || last_un_char == '&')
2424 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2429 if (keyword(tmpbuf, d - tmpbuf))
2432 if (un_char == last_un_char + 1)
2434 weight -= seen[un_char];
2439 if (weight >= 0) /* probably a character class */
2449 * Does all the checking to disambiguate
2451 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2452 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2454 * First argument is the stuff after the first token, e.g. "bar".
2456 * Not a method if bar is a filehandle.
2457 * Not a method if foo is a subroutine prototyped to take a filehandle.
2458 * Not a method if it's really "Foo $bar"
2459 * Method if it's "foo $bar"
2460 * Not a method if it's really "print foo $bar"
2461 * Method if it's really "foo package::" (interpreted as package->foo)
2462 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2463 * Not a method if bar is a filehandle or package, but is quoted with
2468 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2471 char *s = start + (*start == '$');
2472 char tmpbuf[sizeof PL_tokenbuf];
2480 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2484 const char *proto = SvPVX_const(cv);
2495 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2496 /* start is the beginning of the possible filehandle/object,
2497 * and s is the end of it
2498 * tmpbuf is a copy of it
2501 if (*start == '$') {
2502 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2505 len = start - SvPVX(PL_linestr);
2509 start = SvPVX(PL_linestr) + len;
2513 return *s == '(' ? FUNCMETH : METHOD;
2515 if (!keyword(tmpbuf, len)) {
2516 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2520 soff = s - SvPVX(PL_linestr);
2524 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2525 if (indirgv && GvCVu(indirgv))
2527 /* filehandle or package name makes it a method */
2528 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2530 soff = s - SvPVX(PL_linestr);
2533 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2534 return 0; /* no assumptions -- "=>" quotes bearword */
2536 start_force(curforce);
2537 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2538 newSVpvn(tmpbuf,len));
2539 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2541 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2546 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2548 return *s == '(' ? FUNCMETH : METHOD;
2556 * Return a string of Perl code to load the debugger. If PERL5DB
2557 * is set, it will return the contents of that, otherwise a
2558 * compile-time require of perl5db.pl.
2566 const char * const pdb = PerlEnv_getenv("PERL5DB");
2570 SETERRNO(0,SS_NORMAL);
2571 return "BEGIN { require 'perl5db.pl' }";
2577 /* Encoded script support. filter_add() effectively inserts a
2578 * 'pre-processing' function into the current source input stream.
2579 * Note that the filter function only applies to the current source file
2580 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2582 * The datasv parameter (which may be NULL) can be used to pass
2583 * private data to this instance of the filter. The filter function
2584 * can recover the SV using the FILTER_DATA macro and use it to
2585 * store private buffers and state information.
2587 * The supplied datasv parameter is upgraded to a PVIO type
2588 * and the IoDIRP/IoANY field is used to store the function pointer,
2589 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2590 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2591 * private use must be set using malloc'd pointers.
2595 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2601 if (!PL_rsfp_filters)
2602 PL_rsfp_filters = newAV();
2605 SvUPGRADE(datasv, SVt_PVIO);
2606 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2607 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2608 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2609 IoANY(datasv), SvPV_nolen(datasv)));
2610 av_unshift(PL_rsfp_filters, 1);
2611 av_store(PL_rsfp_filters, 0, datasv) ;
2616 /* Delete most recently added instance of this filter function. */
2618 Perl_filter_del(pTHX_ filter_t funcp)
2624 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2626 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2628 /* if filter is on top of stack (usual case) just pop it off */
2629 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2630 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2631 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2632 IoANY(datasv) = (void *)NULL;
2633 sv_free(av_pop(PL_rsfp_filters));
2637 /* we need to search for the correct entry and clear it */
2638 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2642 /* Invoke the idxth filter function for the current rsfp. */
2643 /* maxlen 0 = read one text line */
2645 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2651 if (!PL_rsfp_filters)
2653 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2654 /* Provide a default input filter to make life easy. */
2655 /* Note that we append to the line. This is handy. */
2656 DEBUG_P(PerlIO_printf(Perl_debug_log,
2657 "filter_read %d: from rsfp\n", idx));
2661 const int old_len = SvCUR(buf_sv);
2663 /* ensure buf_sv is large enough */
2664 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2665 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2666 if (PerlIO_error(PL_rsfp))
2667 return -1; /* error */
2669 return 0 ; /* end of file */
2671 SvCUR_set(buf_sv, old_len + len) ;
2674 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2675 if (PerlIO_error(PL_rsfp))
2676 return -1; /* error */
2678 return 0 ; /* end of file */
2681 return SvCUR(buf_sv);
2683 /* Skip this filter slot if filter has been deleted */
2684 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2685 DEBUG_P(PerlIO_printf(Perl_debug_log,
2686 "filter_read %d: skipped (filter deleted)\n",
2688 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2690 /* Get function pointer hidden within datasv */
2691 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2692 DEBUG_P(PerlIO_printf(Perl_debug_log,
2693 "filter_read %d: via function %p (%s)\n",
2694 idx, datasv, SvPV_nolen_const(datasv)));
2695 /* Call function. The function is expected to */
2696 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2697 /* Return: <0:error, =0:eof, >0:not eof */
2698 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2702 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2705 #ifdef PERL_CR_FILTER
2706 if (!PL_rsfp_filters) {
2707 filter_add(S_cr_textfilter,NULL);
2710 if (PL_rsfp_filters) {
2712 SvCUR_set(sv, 0); /* start with empty line */
2713 if (FILTER_READ(0, sv, 0) > 0)
2714 return ( SvPVX(sv) ) ;
2719 return (sv_gets(sv, fp, append));
2723 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2728 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2732 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2733 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2735 return GvHV(gv); /* Foo:: */
2738 /* use constant CLASS => 'MyClass' */
2739 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2741 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2742 pkgname = SvPV_nolen_const(sv);
2746 return gv_stashpv(pkgname, FALSE);
2752 * The intent of this yylex wrapper is to minimize the changes to the
2753 * tokener when we aren't interested in collecting madprops. It remains
2754 * to be seen how successful this strategy will be...
2761 char *s = PL_bufptr;
2763 /* make sure thiswhite is initialized */
2767 /* just do what yylex would do on pending identifier; leave thiswhite alone */
2768 if (PL_pending_ident)
2769 return S_pending_ident(aTHX);
2771 /* previous token ate up our whitespace? */
2772 if (!PL_lasttoke && nextwhite) {
2773 thiswhite = nextwhite;
2777 /* isolate the token, and figure out where it is without whitespace */
2778 realtokenstart = -1;
2782 assert(curforce < 0);
2784 if (!thismad || thismad->mad_key == '^') { /* not forced already? */
2786 if (realtokenstart < 0 || !CopLINE(PL_curcop))
2787 thistoken = newSVpvn("",0);
2789 char *tstart = SvPVX(PL_linestr) + realtokenstart;
2790 thistoken = newSVpvn(tstart, s - tstart);
2793 if (thismad) /* install head */
2794 CURMAD('X', thistoken);
2797 /* last whitespace of a sublex? */
2798 if (optype == ')' && endwhite) {
2799 CURMAD('X', endwhite);
2804 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2805 if (!thiswhite && !endwhite && !optype) {
2811 /* put off final whitespace till peg */
2812 if (optype == ';' && !PL_rsfp) {
2813 nextwhite = thiswhite;
2816 else if (thisopen) {
2817 CURMAD('q', thisopen);
2823 /* Store actual token text as madprop X */
2824 CURMAD('X', thistoken);
2828 /* add preceding whitespace as madprop _ */
2829 CURMAD('_', thiswhite);
2833 /* add quoted material as madprop = */
2834 CURMAD('=', thisstuff);
2838 /* add terminating quote as madprop Q */
2839 CURMAD('Q', thisclose);
2843 /* special processing based on optype */
2847 /* opval doesn't need a TOKEN since it can already store mp */
2858 append_madprops(thismad, yylval.opval, 0);
2866 addmad(newMADsv('p', endwhite), &thismad, 0);
2875 /* remember any fake bracket that lexer is about to discard */
2876 if (PL_lex_brackets == 1 &&
2877 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2880 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2883 thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2884 addmad(newMADsv('#', thiswhite), &thismad, 0);
2887 break; /* don't bother looking for trailing comment */
2896 /* attach a trailing comment to its statement instead of next token */
2900 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2902 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2904 if (*s == '\n' || *s == '#') {
2905 while (s < PL_bufend && *s != '\n')
2909 thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2910 addmad(newMADsv('#', thiswhite), &thismad, 0);
2927 /* Create new token struct. Note: opvals return early above. */
2928 yylval.tkval = newTOKEN(optype, yylval, thismad);
2935 S_tokenize_use(pTHX_ int is_use, char *s) {
2937 if (PL_expect != XSTATE)
2938 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2939 is_use ? "use" : "no"));
2941 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2942 s = force_version(s, TRUE);
2943 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2944 start_force(curforce);
2945 NEXTVAL_NEXTTOKE.opval = NULL;
2948 else if (*s == 'v') {
2949 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2950 s = force_version(s, FALSE);
2954 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2955 s = force_version(s, FALSE);
2957 yylval.ival = is_use;
2961 static const char* const exp_name[] =
2962 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2963 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2970 Works out what to call the token just pulled out of the input
2971 stream. The yacc parser takes care of taking the ops we return and
2972 stitching them into a tree.
2978 if read an identifier
2979 if we're in a my declaration
2980 croak if they tried to say my($foo::bar)
2981 build the ops for a my() declaration
2982 if it's an access to a my() variable
2983 are we in a sort block?
2984 croak if my($a); $a <=> $b
2985 build ops for access to a my() variable
2986 if in a dq string, and they've said @foo and we can't find @foo
2988 build ops for a bareword
2989 if we already built the token before, use it.
2994 #pragma segment Perl_yylex
3000 register char *s = PL_bufptr;
3006 SV* tmp = newSVpvs("");
3007 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3008 (IV)CopLINE(PL_curcop),
3009 lex_state_names[PL_lex_state],
3010 exp_name[PL_expect],
3011 pv_display(tmp, s, strlen(s), 0, 60));
3014 /* check if there's an identifier for us to look at */
3015 if (PL_pending_ident)
3016 return REPORT(S_pending_ident(aTHX));
3018 /* no identifier pending identification */
3020 switch (PL_lex_state) {
3022 case LEX_NORMAL: /* Some compilers will produce faster */
3023 case LEX_INTERPNORMAL: /* code if we comment these out. */
3027 /* when we've already built the next token, just pull it out of the queue */
3031 yylval = PL_nexttoke[PL_lasttoke].next_val;
3033 thismad = PL_nexttoke[PL_lasttoke].next_mad;
3034 PL_nexttoke[PL_lasttoke].next_mad = 0;
3035 if (thismad && thismad->mad_key == '_') {
3036 thiswhite = (SV*)thismad->mad_val;
3037 thismad->mad_val = 0;
3043 PL_lex_state = PL_lex_defer;
3044 PL_expect = PL_lex_expect;
3045 PL_lex_defer = LEX_NORMAL;
3046 if (!PL_nexttoke[PL_lasttoke].next_type)
3051 yylval = PL_nextval[PL_nexttoke];
3053 PL_lex_state = PL_lex_defer;
3054 PL_expect = PL_lex_expect;
3055 PL_lex_defer = LEX_NORMAL;
3059 /* FIXME - can these be merged? */
3060 return(PL_nexttoke[PL_lasttoke].next_type);
3062 return REPORT(PL_nexttype[PL_nexttoke]);
3065 /* interpolated case modifiers like \L \U, including \Q and \E.
3066 when we get here, PL_bufptr is at the \
3068 case LEX_INTERPCASEMOD:
3070 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3071 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3073 /* handle \E or end of string */
3074 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3076 if (PL_lex_casemods) {
3077 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3078 PL_lex_casestack[PL_lex_casemods] = '\0';
3080 if (PL_bufptr != PL_bufend
3081 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3083 PL_lex_state = LEX_INTERPCONCAT;
3086 thistoken = newSVpvn("\\E",2);
3092 while (PL_bufptr != PL_bufend &&
3093 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3095 thiswhite = newSVpvn("",0);
3096 sv_catpvn(thiswhite, PL_bufptr, 2);
3100 if (PL_bufptr != PL_bufend)
3103 PL_lex_state = LEX_INTERPCONCAT;
3107 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3108 "### Saw case modifier\n"); });
3110 if (s[1] == '\\' && s[2] == 'E') {
3113 thiswhite = newSVpvn("",0);
3114 sv_catpvn(thiswhite, PL_bufptr, 4);
3117 PL_lex_state = LEX_INTERPCONCAT;
3122 if (!PL_madskills) /* when just compiling don't need correct */
3123 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3124 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3125 if ((*s == 'L' || *s == 'U') &&
3126 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3127 PL_lex_casestack[--PL_lex_casemods] = '\0';
3130 if (PL_lex_casemods > 10)
3131 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3132 PL_lex_casestack[PL_lex_casemods++] = *s;
3133 PL_lex_casestack[PL_lex_casemods] = '\0';
3134 PL_lex_state = LEX_INTERPCONCAT;
3135 start_force(curforce);
3136 NEXTVAL_NEXTTOKE.ival = 0;
3138 start_force(curforce);
3140 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3142 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3144 NEXTVAL_NEXTTOKE.ival = OP_LC;
3146 NEXTVAL_NEXTTOKE.ival = OP_UC;
3148 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3150 Perl_croak(aTHX_ "panic: yylex");
3152 SV* tmpsv = newSVpvn("",0);
3153 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3159 if (PL_lex_starts) {
3166 thistoken = newSVpvn("",0);
3169 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3170 if (PL_lex_casemods == 1 && PL_lex_inpat)
3179 case LEX_INTERPPUSH:
3180 return REPORT(sublex_push());
3182 case LEX_INTERPSTART:
3183 if (PL_bufptr == PL_bufend)
3184 return REPORT(sublex_done());
3185 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3186 "### Interpolated variable\n"); });
3188 PL_lex_dojoin = (*PL_bufptr == '@');
3189 PL_lex_state = LEX_INTERPNORMAL;
3190 if (PL_lex_dojoin) {
3191 start_force(curforce);
3192 NEXTVAL_NEXTTOKE.ival = 0;
3194 start_force(curforce);
3195 force_ident("\"", '$');
3196 start_force(curforce);
3197 NEXTVAL_NEXTTOKE.ival = 0;
3199 start_force(curforce);
3200 NEXTVAL_NEXTTOKE.ival = 0;
3202 start_force(curforce);
3203 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3206 if (PL_lex_starts++) {
3212 thistoken = newSVpvn("",0);
3215 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3216 if (!PL_lex_casemods && PL_lex_inpat)
3223 case LEX_INTERPENDMAYBE:
3224 if (intuit_more(PL_bufptr)) {
3225 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3231 if (PL_lex_dojoin) {
3232 PL_lex_dojoin = FALSE;
3233 PL_lex_state = LEX_INTERPCONCAT;
3238 thistoken = newSVpvn("",0);
3243 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3244 && SvEVALED(PL_lex_repl))
3246 if (PL_bufptr != PL_bufend)
3247 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3251 case LEX_INTERPCONCAT:
3253 if (PL_lex_brackets)
3254 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3256 if (PL_bufptr == PL_bufend)
3257 return REPORT(sublex_done());
3259 if (SvIVX(PL_linestr) == '\'') {
3260 SV *sv = newSVsv(PL_linestr);
3263 else if ( PL_hints & HINT_NEW_RE )
3264 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3265 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3269 s = scan_const(PL_bufptr);
3271 PL_lex_state = LEX_INTERPCASEMOD;
3273 PL_lex_state = LEX_INTERPSTART;
3276 if (s != PL_bufptr) {
3277 start_force(curforce);
3279 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3281 NEXTVAL_NEXTTOKE = yylval;
3284 if (PL_lex_starts++) {
3289 thistoken = newSVpvn("",0);
3292 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3293 if (!PL_lex_casemods && PL_lex_inpat)
3306 PL_lex_state = LEX_NORMAL;
3307 s = scan_formline(PL_bufptr);
3308 if (!PL_lex_formbrack)
3314 PL_oldoldbufptr = PL_oldbufptr;
3323 realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3327 if (isIDFIRST_lazy_if(s,UTF))
3329 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3332 goto fake_eof; /* emulate EOF on ^D or ^Z */
3341 if (PL_lex_brackets) {
3342 yyerror(PL_lex_formbrack
3343 ? "Format not terminated"
3344 : "Missing right curly or square bracket");
3346 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3347 "### Tokener got EOF\n");
3351 if (s++ < PL_bufend)
3352 goto retry; /* ignore stray nulls */
3355 if (!PL_in_eval && !PL_preambled) {
3356 PL_preambled = TRUE;
3361 sv_setpv(PL_linestr,incl_perldb());
3362 if (SvCUR(PL_linestr))
3363 sv_catpvs(PL_linestr,";");
3365 while(AvFILLp(PL_preambleav) >= 0) {
3366 SV *tmpsv = av_shift(PL_preambleav);
3367 sv_catsv(PL_linestr, tmpsv);
3368 sv_catpvs(PL_linestr, ";");
3371 sv_free((SV*)PL_preambleav);
3372 PL_preambleav = NULL;
3374 if (PL_minus_n || PL_minus_p) {
3375 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3377 sv_catpvs(PL_linestr,"chomp;");
3380 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3381 || *PL_splitstr == '"')
3382 && strchr(PL_splitstr + 1, *PL_splitstr))
3383 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3385 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3386 bytes can be used as quoting characters. :-) */
3387 const char *splits = PL_splitstr;
3388 sv_catpvs(PL_linestr, "our @F=split(q\0");
3391 if (*splits == '\\')
3392 sv_catpvn(PL_linestr, splits, 1);
3393 sv_catpvn(PL_linestr, splits, 1);
3394 } while (*splits++);
3395 /* This loop will embed the trailing NUL of
3396 PL_linestr as the last thing it does before
3398 sv_catpvs(PL_linestr, ");");
3402 sv_catpvs(PL_linestr,"our @F=split(' ');");
3406 sv_catpvs(PL_linestr,"use feature ':5.10';");
3407 sv_catpvs(PL_linestr, "\n");
3408 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3409 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3410 PL_last_lop = PL_last_uni = NULL;
3411 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3412 SV * const sv = newSV(0);
3414 sv_upgrade(sv, SVt_PVMG);
3415 sv_setsv(sv,PL_linestr);
3418 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3423 bof = PL_rsfp ? TRUE : FALSE;
3424 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3427 realtokenstart = -1;
3430 if (PL_preprocess && !PL_in_eval)
3431 (void)PerlProc_pclose(PL_rsfp);
3432 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3433 PerlIO_clearerr(PL_rsfp);
3435 (void)PerlIO_close(PL_rsfp);
3437 PL_doextract = FALSE;
3439 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3444 sv_setpv(PL_linestr,PL_minus_p
3445 ? ";}continue{print;}" : ";}");
3446 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3447 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3448 PL_last_lop = PL_last_uni = NULL;
3449 PL_minus_n = PL_minus_p = 0;
3452 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3453 PL_last_lop = PL_last_uni = NULL;
3454 sv_setpvn(PL_linestr,"",0);
3455 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3457 /* If it looks like the start of a BOM or raw UTF-16,
3458 * check if it in fact is. */
3464 #ifdef PERLIO_IS_STDIO
3465 # ifdef __GNU_LIBRARY__
3466 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3467 # define FTELL_FOR_PIPE_IS_BROKEN
3471 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3472 # define FTELL_FOR_PIPE_IS_BROKEN
3477 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3478 /* This loses the possibility to detect the bof
3479 * situation on perl -P when the libc5 is being used.
3480 * Workaround? Maybe attach some extra state to PL_rsfp?
3483 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3485 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3488 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3489 s = swallow_bom((U8*)s);
3493 /* Incest with pod. */
3496 sv_catsv(thiswhite, PL_linestr);
3498 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3499 sv_setpvn(PL_linestr, "", 0);
3500 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3501 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3502 PL_last_lop = PL_last_uni = NULL;
3503 PL_doextract = FALSE;
3507 } while (PL_doextract);
3508 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3509 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3510 SV * const sv = newSV(0);
3512 sv_upgrade(sv, SVt_PVMG);
3513 sv_setsv(sv,PL_linestr);
3516 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3518 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3519 PL_last_lop = PL_last_uni = NULL;
3520 if (CopLINE(PL_curcop) == 1) {
3521 while (s < PL_bufend && isSPACE(*s))
3523 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3527 thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3531 if (*s == '#' && *(s+1) == '!')
3533 #ifdef ALTERNATE_SHEBANG
3535 static char const as[] = ALTERNATE_SHEBANG;
3536 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3537 d = s + (sizeof(as) - 1);
3539 #endif /* ALTERNATE_SHEBANG */
3548 while (*d && !isSPACE(*d))
3552 #ifdef ARG_ZERO_IS_SCRIPT
3553 if (ipathend > ipath) {
3555 * HP-UX (at least) sets argv[0] to the script name,
3556 * which makes $^X incorrect. And Digital UNIX and Linux,
3557 * at least, set argv[0] to the basename of the Perl
3558 * interpreter. So, having found "#!", we'll set it right.
3560 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3562 assert(SvPOK(x) || SvGMAGICAL(x));
3563 if (sv_eq(x, CopFILESV(PL_curcop))) {
3564 sv_setpvn(x, ipath, ipathend - ipath);
3570 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3571 const char * const lstart = SvPV_const(x,llen);
3573 bstart += blen - llen;
3574 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3575 sv_setpvn(x, ipath, ipathend - ipath);
3580 TAINT_NOT; /* $^X is always tainted, but that's OK */
3582 #endif /* ARG_ZERO_IS_SCRIPT */
3587 d = instr(s,"perl -");
3589 d = instr(s,"perl");
3591 /* avoid getting into infinite loops when shebang
3592 * line contains "Perl" rather than "perl" */
3594 for (d = ipathend-4; d >= ipath; --d) {
3595 if ((*d == 'p' || *d == 'P')
3596 && !ibcmp(d, "perl", 4))
3606 #ifdef ALTERNATE_SHEBANG
3608 * If the ALTERNATE_SHEBANG on this system starts with a
3609 * character that can be part of a Perl expression, then if
3610 * we see it but not "perl", we're probably looking at the
3611 * start of Perl code, not a request to hand off to some
3612 * other interpreter. Similarly, if "perl" is there, but
3613 * not in the first 'word' of the line, we assume the line
3614 * contains the start of the Perl program.
3616 if (d && *s != '#') {
3617 const char *c = ipath;
3618 while (*c && !strchr("; \t\r\n\f\v#", *c))
3621 d = NULL; /* "perl" not in first word; ignore */
3623 *s = '#'; /* Don't try to parse shebang line */
3625 #endif /* ALTERNATE_SHEBANG */
3626 #ifndef MACOS_TRADITIONAL
3631 !instr(s,"indir") &&
3632 instr(PL_origargv[0],"perl"))
3639 while (s < PL_bufend && isSPACE(*s))
3641 if (s < PL_bufend) {
3642 Newxz(newargv,PL_origargc+3,char*);
3644 while (s < PL_bufend && !isSPACE(*s))
3647 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3650 newargv = PL_origargv;
3653 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3655 Perl_croak(aTHX_ "Can't exec %s", ipath);
3659 while (*d && !isSPACE(*d)) d++;
3660 while (SPACE_OR_TAB(*d)) d++;
3663 const bool switches_done = PL_doswitches;
3664 const U32 oldpdb = PL_perldb;
3665 const bool oldn = PL_minus_n;
3666 const bool oldp = PL_minus_p;
3669 if (*d == 'M' || *d == 'm' || *d == 'C') {
3670 const char * const m = d;
3671 while (*d && !isSPACE(*d)) d++;
3672 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3675 d = moreswitches(d);
3677 if (PL_doswitches && !switches_done) {
3678 int argc = PL_origargc;
3679 char **argv = PL_origargv;
3682 } while (argc && argv[0][0] == '-' && argv[0][1]);
3683 init_argv_symbols(argc,argv);
3685 if ((PERLDB_LINE && !oldpdb) ||
3686 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3687 /* if we have already added "LINE: while (<>) {",
3688 we must not do it again */
3690 sv_setpvn(PL_linestr, "", 0);
3691 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3692 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3693 PL_last_lop = PL_last_uni = NULL;
3694 PL_preambled = FALSE;
3696 (void)gv_fetchfile(PL_origfilename);
3703 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3705 PL_lex_state = LEX_FORMLINE;
3710 #ifdef PERL_STRICT_CR
3711 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3713 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3715 case ' ': case '\t': case '\f': case 013:
3716 #ifdef MACOS_TRADITIONAL
3720 realtokenstart = -1;
3729 realtokenstart = -1;
3733 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3734 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3735 /* handle eval qq[#line 1 "foo"\n ...] */
3736 CopLINE_dec(PL_curcop);
3739 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3741 if (!PL_in_eval || PL_rsfp)
3746 while (d < PL_bufend && *d != '\n')
3750 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3751 Perl_croak(aTHX_ "panic: input overflow");
3754 thiswhite = newSVpvn(s, d - s);
3759 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3761 PL_lex_state = LEX_FORMLINE;
3767 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3768 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3771 TOKEN(PEG); /* make sure any #! line is accessible */
3776 /* if (PL_madskills && PL_lex_formbrack) { */
3778 while (d < PL_bufend && *d != '\n')
3782 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3783 Perl_croak(aTHX_ "panic: input overflow");
3784 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3786 thiswhite = newSVpvn("",0);
3787 if (CopLINE(PL_curcop) == 1) {
3788 sv_setpvn(thiswhite, "", 0);
3791 sv_catpvn(thiswhite, s, d - s);
3805 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3813 while (s < PL_bufend && SPACE_OR_TAB(*s))
3816 if (strnEQ(s,"=>",2)) {
3817 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3818 DEBUG_T( { S_printbuf(aTHX_
3819 "### Saw unary minus before =>, forcing word %s\n", s);
3821 OPERATOR('-'); /* unary minus */
3823 PL_last_uni = PL_oldbufptr;
3825 case 'r': ftst = OP_FTEREAD; break;
3826 case 'w': ftst = OP_FTEWRITE; break;
3827 case 'x': ftst = OP_FTEEXEC; break;
3828 case 'o': ftst = OP_FTEOWNED; break;
3829 case 'R': ftst = OP_FTRREAD; break;
3830 case 'W': ftst = OP_FTRWRITE; break;
3831 case 'X': ftst = OP_FTREXEC; break;
3832 case 'O': ftst = OP_FTROWNED; break;
3833 case 'e': ftst = OP_FTIS; break;
3834 case 'z': ftst = OP_FTZERO; break;
3835 case 's': ftst = OP_FTSIZE; break;
3836 case 'f': ftst = OP_FTFILE; break;
3837 case 'd': ftst = OP_FTDIR; break;
3838 case 'l': ftst = OP_FTLINK; break;
3839 case 'p': ftst = OP_FTPIPE; break;
3840 case 'S': ftst = OP_FTSOCK; break;
3841 case 'u': ftst = OP_FTSUID; break;
3842 case 'g': ftst = OP_FTSGID; break;
3843 case 'k': ftst = OP_FTSVTX; break;
3844 case 'b': ftst = OP_FTBLK; break;
3845 case 'c': ftst = OP_FTCHR; break;
3846 case 't': ftst = OP_FTTTY; break;
3847 case 'T': ftst = OP_FTTEXT; break;
3848 case 'B': ftst = OP_FTBINARY; break;
3849 case 'M': case 'A': case 'C':
3850 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3852 case 'M': ftst = OP_FTMTIME; break;
3853 case 'A': ftst = OP_FTATIME; break;
3854 case 'C': ftst = OP_FTCTIME; break;
3862 PL_last_lop_op = (OPCODE)ftst;
3863 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3864 "### Saw file test %c\n", (int)tmp);
3869 /* Assume it was a minus followed by a one-letter named
3870 * subroutine call (or a -bareword), then. */
3871 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3872 "### '-%c' looked like a file test but was not\n",
3879 const char tmp = *s++;
3882 if (PL_expect == XOPERATOR)
3887 else if (*s == '>') {
3890 if (isIDFIRST_lazy_if(s,UTF)) {
3891 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3899 if (PL_expect == XOPERATOR)
3902 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3904 OPERATOR('-'); /* unary minus */
3910 const char tmp = *s++;
3913 if (PL_expect == XOPERATOR)
3918 if (PL_expect == XOPERATOR)
3921 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3928 if (PL_expect != XOPERATOR) {
3929 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3930 PL_expect = XOPERATOR;
3931 force_ident(PL_tokenbuf, '*');
3944 if (PL_expect == XOPERATOR) {
3948 PL_tokenbuf[0] = '%';
3949 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3950 if (!PL_tokenbuf[1]) {
3953 PL_pending_ident = '%';
3964 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3965 && FEATURE_IS_ENABLED("~~"))
3972 const char tmp = *s++;
3978 goto just_a_word_zero_gv;
3981 switch (PL_expect) {
3987 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3989 PL_bufptr = s; /* update in case we back off */
3995 PL_expect = XTERMBLOCK;
3998 stuffstart = s - SvPVX(PL_linestr) - 1;
4002 while (isIDFIRST_lazy_if(s,UTF)) {
4004 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4005 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4006 if (tmp < 0) tmp = -tmp;
4022 d = scan_str(d,TRUE,TRUE);
4024 /* MUST advance bufptr here to avoid bogus
4025 "at end of line" context messages from yyerror().
4027 PL_bufptr = s + len;
4028 yyerror("Unterminated attribute parameter in attribute list");
4031 return REPORT(0); /* EOF indicator */
4035 SV *sv = newSVpvn(s, len);
4036 sv_catsv(sv, PL_lex_stuff);
4037 attrs = append_elem(OP_LIST, attrs,
4038 newSVOP(OP_CONST, 0, sv));
4039 SvREFCNT_dec(PL_lex_stuff);
4040 PL_lex_stuff = NULL;
4043 if (len == 6 && strnEQ(s, "unique", len)) {
4044 if (PL_in_my == KEY_our)
4046 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4048 /*EMPTY*/; /* skip to avoid loading attributes.pm */
4051 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4054 /* NOTE: any CV attrs applied here need to be part of
4055 the CVf_BUILTIN_ATTRS define in cv.h! */
4056 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
4057 CvLVALUE_on(PL_compcv);
4058 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
4059 CvLOCKED_on(PL_compcv);
4060 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
4061 CvMETHOD_on(PL_compcv);
4062 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
4063 CvASSERTION_on(PL_compcv);
4064 /* After we've set the flags, it could be argued that
4065 we don't need to do the attributes.pm-based setting
4066 process, and shouldn't bother appending recognized
4067 flags. To experiment with that, uncomment the
4068 following "else". (Note that's already been
4069 uncommented. That keeps the above-applied built-in
4070 attributes from being intercepted (and possibly
4071 rejected) by a package's attribute routines, but is
4072 justified by the performance win for the common case
4073 of applying only built-in attributes.) */
4075 attrs = append_elem(OP_LIST, attrs,
4076 newSVOP(OP_CONST, 0,
4080 if (*s == ':' && s[1] != ':')
4083 break; /* require real whitespace or :'s */
4084 /* XXX losing whitespace on sequential attributes here */
4088 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4089 if (*s != ';' && *s != '}' && *s != tmp
4090 && (tmp != '=' || *s != ')')) {
4091 const char q = ((*s == '\'') ? '"' : '\'');
4092 /* If here for an expression, and parsed no attrs, back
4094 if (tmp == '=' && !attrs) {
4098 /* MUST advance bufptr here to avoid bogus "at end of line"
4099 context messages from yyerror().
4103 ? Perl_form(aTHX_ "Invalid separator character "
4104 "%c%c%c in attribute list", q, *s, q)
4105 : "Unterminated attribute list" );
4113 start_force(curforce);
4114 NEXTVAL_NEXTTOKE.opval = attrs;
4115 CURMAD('_', nextwhite);
4120 thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4121 (s - SvPVX(PL_linestr)) - stuffstart);
4129 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4130 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4138 const char tmp = *s++;
4143 const char tmp = *s++;
4151 if (PL_lex_brackets <= 0)
4152 yyerror("Unmatched right square bracket");
4155 if (PL_lex_state == LEX_INTERPNORMAL) {
4156 if (PL_lex_brackets == 0) {
4157 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4158 PL_lex_state = LEX_INTERPEND;
4165 if (PL_lex_brackets > 100) {
4166 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4168 switch (PL_expect) {
4170 if (PL_lex_formbrack) {
4174 if (PL_oldoldbufptr == PL_last_lop)
4175 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4177 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4178 OPERATOR(HASHBRACK);
4180 while (s < PL_bufend && SPACE_OR_TAB(*s))
4183 PL_tokenbuf[0] = '\0';
4184 if (d < PL_bufend && *d == '-') {
4185 PL_tokenbuf[0] = '-';
4187 while (d < PL_bufend && SPACE_OR_TAB(*d))
4190 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4191 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4193 while (d < PL_bufend && SPACE_OR_TAB(*d))
4196 const char minus = (PL_tokenbuf[0] == '-');
4197 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4205 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4210 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4215 if (PL_oldoldbufptr == PL_last_lop)
4216 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4218 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4221 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4223 /* This hack is to get the ${} in the message. */
4225 yyerror("syntax error");
4228 OPERATOR(HASHBRACK);
4230 /* This hack serves to disambiguate a pair of curlies
4231 * as being a block or an anon hash. Normally, expectation
4232 * determines that, but in cases where we're not in a
4233 * position to expect anything in particular (like inside
4234 * eval"") we have to resolve the ambiguity. This code
4235 * covers the case where the first term in the curlies is a
4236 * quoted string. Most other cases need to be explicitly
4237 * disambiguated by prepending a "+" before the opening
4238 * curly in order to force resolution as an anon hash.
4240 * XXX should probably propagate the outer expectation
4241 * into eval"" to rely less on this hack, but that could
4242 * potentially break current behavior of eval"".
4246 if (*s == '\'' || *s == '"' || *s == '`') {
4247 /* common case: get past first string, handling escapes */
4248 for (t++; t < PL_bufend && *t != *s;)
4249 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4253 else if (*s == 'q') {
4256 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4259 /* skip q//-like construct */
4261 char open, close, term;
4264 while (t < PL_bufend && isSPACE(*t))
4266 /* check for q => */
4267 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4268 OPERATOR(HASHBRACK);
4272 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4276 for (t++; t < PL_bufend; t++) {
4277 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4279 else if (*t == open)
4283 for (t++; t < PL_bufend; t++) {
4284 if (*t == '\\' && t+1 < PL_bufend)
4286 else if (*t == close && --brackets <= 0)
4288 else if (*t == open)
4295 /* skip plain q word */
4296 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4299 else if (isALNUM_lazy_if(t,UTF)) {
4301 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4304 while (t < PL_bufend && isSPACE(*t))
4306 /* if comma follows first term, call it an anon hash */
4307 /* XXX it could be a comma expression with loop modifiers */
4308 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4309 || (*t == '=' && t[1] == '>')))
4310 OPERATOR(HASHBRACK);
4311 if (PL_expect == XREF)
4314 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4320 yylval.ival = CopLINE(PL_curcop);
4321 if (isSPACE(*s) || *s == '#')
4322 PL_copline = NOLINE; /* invalidate current command line number */
4327 if (PL_lex_brackets <= 0)
4328 yyerror("Unmatched right curly bracket");
4330 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4331 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4332 PL_lex_formbrack = 0;
4333 if (PL_lex_state == LEX_INTERPNORMAL) {
4334 if (PL_lex_brackets == 0) {
4335 if (PL_expect & XFAKEBRACK) {
4336 PL_expect &= XENUMMASK;
4337 PL_lex_state = LEX_INTERPEND;
4342 thiswhite = newSVpvn("",0);
4343 sv_catpvn(thiswhite,"}",1);
4346 return yylex(); /* ignore fake brackets */
4348 if (*s == '-' && s[1] == '>')
4349 PL_lex_state = LEX_INTERPENDMAYBE;
4350 else if (*s != '[' && *s != '{')
4351 PL_lex_state = LEX_INTERPEND;
4354 if (PL_expect & XFAKEBRACK) {
4355 PL_expect &= XENUMMASK;
4357 return yylex(); /* ignore fake brackets */
4359 start_force(curforce);
4361 curmad('X', newSVpvn(s-1,1));
4362 CURMAD('_', thiswhite);
4367 thistoken = newSVpvn("",0);
4375 if (PL_expect == XOPERATOR) {
4376 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4377 && isIDFIRST_lazy_if(s,UTF))
4379 CopLINE_dec(PL_curcop);
4380 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4381 CopLINE_inc(PL_curcop);
4386 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4388 PL_expect = XOPERATOR;
4389 force_ident(PL_tokenbuf, '&');
4393 yylval.ival = (OPpENTERSUB_AMPER<<8);