3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yychar (*PL_yycharp)
27 #define yylval (*PL_yylvalp)
29 static const char ident_too_long[] = "Identifier too long";
30 static const char commaless_variable_list[] = "comma-less variable list";
32 static void restore_rsfp(pTHX_ void *f);
33 #ifndef PERL_NO_UTF16_FILTER
34 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
35 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
40 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
42 # define CURMAD(slot,sv)
43 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
46 #define XFAKEBRACK 128
49 #ifdef USE_UTF8_SCRIPTS
50 # define UTF (!IN_BYTES)
52 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
55 /* In variables named $^X, these are the legal values for X.
56 * 1999-02-27 mjd-perl-patch@plover.com */
57 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
59 /* On MacOS, respect nonbreaking spaces */
60 #ifdef MACOS_TRADITIONAL
61 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
63 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
66 /* LEX_* are values for PL_lex_state, the state of the lexer.
67 * They are arranged oddly so that the guard on the switch statement
68 * can get by with a single comparison (if the compiler is smart enough).
71 /* #define LEX_NOTPARSING 11 is done in perl.h. */
73 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
74 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
75 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
76 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
77 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
79 /* at end of code, eg "$x" followed by: */
80 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
81 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
83 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
84 string or after \E, $foo, etc */
85 #define LEX_INTERPCONST 2 /* NOT USED */
86 #define LEX_FORMLINE 1 /* expecting a format line */
87 #define LEX_KNOWNEXT 0 /* next token known; just return it */
91 static const char* const lex_state_names[] = {
110 #include "keywords.h"
112 /* CLINE is a macro that ensures PL_copline has a sane value */
117 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
120 # define SKIPSPACE0(s) skipspace0(s)
121 # define SKIPSPACE1(s) skipspace1(s)
122 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
123 # define PEEKSPACE(s) skipspace2(s,0)
125 # define SKIPSPACE0(s) skipspace(s)
126 # define SKIPSPACE1(s) skipspace(s)
127 # define SKIPSPACE2(s,tsv) skipspace(s)
128 # define PEEKSPACE(s) skipspace(s)
132 * Convenience functions to return different tokens and prime the
133 * lexer for the next token. They all take an argument.
135 * TOKEN : generic token (used for '(', DOLSHARP, etc)
136 * OPERATOR : generic operator
137 * AOPERATOR : assignment operator
138 * PREBLOCK : beginning the block after an if, while, foreach, ...
139 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
140 * PREREF : *EXPR where EXPR is not a simple identifier
141 * TERM : expression term
142 * LOOPX : loop exiting command (goto, last, dump, etc)
143 * FTST : file test operator
144 * FUN0 : zero-argument function
145 * FUN1 : not used, except for not, which isn't a UNIOP
146 * BOop : bitwise or or xor
148 * SHop : shift operator
149 * PWop : power operator
150 * PMop : pattern-matching operator
151 * Aop : addition-level operator
152 * Mop : multiplication-level operator
153 * Eop : equality-testing operator
154 * Rop : relational operator <= != gt
156 * Also see LOP and lop() below.
159 #ifdef DEBUGGING /* Serve -DT. */
160 # define REPORT(retval) tokereport((I32)retval)
162 # define REPORT(retval) (retval)
165 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
166 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
167 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
168 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
170 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
171 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
172 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
173 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
174 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
175 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
176 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
177 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
178 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
179 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
180 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
181 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
182 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
183 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
184 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
186 /* This bit of chicanery makes a unary function followed by
187 * a parenthesis into a function with one argument, highest precedence.
188 * The UNIDOR macro is for unary functions that can be followed by the //
189 * operator (such as C<shift // 0>).
191 #define UNI2(f,x) { \
195 PL_last_uni = PL_oldbufptr; \
196 PL_last_lop_op = f; \
198 return REPORT( (int)FUNC1 ); \
200 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
202 #define UNI(f) UNI2(f,XTERM)
203 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
205 #define UNIBRACK(f) { \
208 PL_last_uni = PL_oldbufptr; \
210 return REPORT( (int)FUNC1 ); \
212 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
215 /* grandfather return to old style */
216 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
220 /* how to interpret the yylval associated with the token */
224 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
230 static struct debug_tokens {
232 enum token_type type;
234 } const debug_tokens[] =
236 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
237 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
238 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
239 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
240 { ARROW, TOKENTYPE_NONE, "ARROW" },
241 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
242 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
243 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
244 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
245 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
246 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
247 { DO, TOKENTYPE_NONE, "DO" },
248 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
249 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
250 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
251 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
252 { ELSE, TOKENTYPE_NONE, "ELSE" },
253 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
254 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
255 { FOR, TOKENTYPE_IVAL, "FOR" },
256 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
257 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
258 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
259 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
260 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
261 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
262 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
263 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
264 { IF, TOKENTYPE_IVAL, "IF" },
265 { LABEL, TOKENTYPE_PVAL, "LABEL" },
266 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
267 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
268 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
269 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
270 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
271 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
272 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
273 { MY, TOKENTYPE_IVAL, "MY" },
274 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
275 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
276 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
277 { OROP, TOKENTYPE_IVAL, "OROP" },
278 { OROR, TOKENTYPE_NONE, "OROR" },
279 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
280 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
281 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
282 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
283 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
284 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
285 { PREINC, TOKENTYPE_NONE, "PREINC" },
286 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
287 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
288 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
289 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
290 { SUB, TOKENTYPE_NONE, "SUB" },
291 { THING, TOKENTYPE_OPVAL, "THING" },
292 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
293 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
294 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
295 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
296 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
297 { USE, TOKENTYPE_IVAL, "USE" },
298 { WHEN, TOKENTYPE_IVAL, "WHEN" },
299 { WHILE, TOKENTYPE_IVAL, "WHILE" },
300 { WORD, TOKENTYPE_OPVAL, "WORD" },
301 { 0, TOKENTYPE_NONE, 0 }
304 /* dump the returned token in rv, plus any optional arg in yylval */
307 S_tokereport(pTHX_ I32 rv)
311 const char *name = NULL;
312 enum token_type type = TOKENTYPE_NONE;
313 const struct debug_tokens *p;
314 SV* const report = newSVpvs("<== ");
316 for (p = debug_tokens; p->token; p++) {
317 if (p->token == (int)rv) {
324 Perl_sv_catpv(aTHX_ report, name);
325 else if ((char)rv > ' ' && (char)rv < '~')
326 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
328 sv_catpvs(report, "EOF");
330 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
333 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
336 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
338 case TOKENTYPE_OPNUM:
339 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
340 PL_op_name[yylval.ival]);
343 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
345 case TOKENTYPE_OPVAL:
347 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
348 PL_op_name[yylval.opval->op_type]);
349 if (yylval.opval->op_type == OP_CONST) {
350 Perl_sv_catpvf(aTHX_ report, " %s",
351 SvPEEK(cSVOPx_sv(yylval.opval)));
356 sv_catpvs(report, "(opval=null)");
359 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
365 /* print the buffer with suitable escapes */
368 S_printbuf(pTHX_ const char* fmt, const char* s)
370 SV* const tmp = newSVpvs("");
371 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
380 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
381 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
385 S_ao(pTHX_ int toketype)
388 if (*PL_bufptr == '=') {
390 if (toketype == ANDAND)
391 yylval.ival = OP_ANDASSIGN;
392 else if (toketype == OROR)
393 yylval.ival = OP_ORASSIGN;
394 else if (toketype == DORDOR)
395 yylval.ival = OP_DORASSIGN;
403 * When Perl expects an operator and finds something else, no_op
404 * prints the warning. It always prints "<something> found where
405 * operator expected. It prints "Missing semicolon on previous line?"
406 * if the surprise occurs at the start of the line. "do you need to
407 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
408 * where the compiler doesn't know if foo is a method call or a function.
409 * It prints "Missing operator before end of line" if there's nothing
410 * after the missing operator, or "... before <...>" if there is something
411 * after the missing operator.
415 S_no_op(pTHX_ const char *what, char *s)
418 char * const oldbp = PL_bufptr;
419 const bool is_first = (PL_oldbufptr == PL_linestart);
425 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
426 if (ckWARN_d(WARN_SYNTAX)) {
428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
429 "\t(Missing semicolon on previous line?)\n");
430 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
432 for (t = PL_oldoldbufptr; *t && (isALNUM_lazy_if(t,UTF) || *t == ':'); t++) ;
433 if (t < PL_bufptr && isSPACE(*t))
434 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
435 "\t(Do you need to predeclare %.*s?)\n",
436 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
440 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
441 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
449 * Complain about missing quote/regexp/heredoc terminator.
450 * If it's called with (char *)NULL then it cauterizes the line buffer.
451 * If we're in a delimited string and the delimiter is a control
452 * character, it's reformatted into a two-char sequence like ^C.
457 S_missingterm(pTHX_ char *s)
463 char * const nl = strrchr(s,'\n');
469 iscntrl(PL_multi_close)
471 PL_multi_close < 32 || PL_multi_close == 127
475 tmpbuf[1] = (char)toCTRL(PL_multi_close);
480 *tmpbuf = (char)PL_multi_close;
484 q = strchr(s,'"') ? '\'' : '"';
485 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
488 #define FEATURE_IS_ENABLED(name) \
489 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
490 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
492 * S_feature_is_enabled
493 * Check whether the named feature is enabled.
496 S_feature_is_enabled(pTHX_ char *name, STRLEN namelen)
499 HV * const hinthv = GvHV(PL_hintgv);
500 char he_name[32] = "feature_";
501 (void) strncpy(&he_name[8], name, 24);
503 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
511 Perl_deprecate(pTHX_ const char *s)
513 if (ckWARN(WARN_DEPRECATED))
514 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
518 Perl_deprecate_old(pTHX_ const char *s)
520 /* This function should NOT be called for any new deprecated warnings */
521 /* Use Perl_deprecate instead */
523 /* It is here to maintain backward compatibility with the pre-5.8 */
524 /* warnings category hierarchy. The "deprecated" category used to */
525 /* live under the "syntax" category. It is now a top-level category */
526 /* in its own right. */
528 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
529 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
530 "Use of %s is deprecated", s);
534 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
535 * utf16-to-utf8-reversed.
538 #ifdef PERL_CR_FILTER
542 register const char *s = SvPVX_const(sv);
543 register const char * const e = s + SvCUR(sv);
544 /* outer loop optimized to do nothing if there are no CR-LFs */
546 if (*s++ == '\r' && *s == '\n') {
547 /* hit a CR-LF, need to copy the rest */
548 register char *d = s - 1;
551 if (*s == '\r' && s[1] == '\n')
562 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
564 const I32 count = FILTER_READ(idx+1, sv, maxlen);
565 if (count > 0 && !maxlen)
573 * Initialize variables. Uses the Perl save_stack to save its state (for
574 * recursive calls to the parser).
578 Perl_lex_start(pTHX_ SV *line)
584 SAVEI32(PL_lex_dojoin);
585 SAVEI32(PL_lex_brackets);
586 SAVEI32(PL_lex_casemods);
587 SAVEI32(PL_lex_starts);
588 SAVEI32(PL_lex_state);
589 SAVEVPTR(PL_lex_inpat);
590 SAVEI32(PL_lex_inwhat);
592 if (PL_lex_state == LEX_KNOWNEXT) {
593 I32 toke = PL_lasttoke;
594 while (--toke >= 0) {
595 SAVEI32(PL_nexttoke[toke].next_type);
596 SAVEVPTR(PL_nexttoke[toke].next_val);
598 SAVEVPTR(PL_nexttoke[toke].next_mad);
600 SAVEI32(PL_lasttoke);
603 SAVESPTR(PL_thistoken);
604 SAVESPTR(PL_thiswhite);
605 SAVESPTR(PL_nextwhite);
606 SAVESPTR(PL_thisopen);
607 SAVESPTR(PL_thisclose);
608 SAVESPTR(PL_thisstuff);
609 SAVEVPTR(PL_thismad);
610 SAVEI32(PL_realtokenstart);
611 SAVEI32(PL_faketokens);
613 SAVEI32(PL_curforce);
615 if (PL_lex_state == LEX_KNOWNEXT) {
616 I32 toke = PL_nexttoke;
617 while (--toke >= 0) {
618 SAVEI32(PL_nexttype[toke]);
619 SAVEVPTR(PL_nextval[toke]);
621 SAVEI32(PL_nexttoke);
624 SAVECOPLINE(PL_curcop);
627 SAVEPPTR(PL_oldbufptr);
628 SAVEPPTR(PL_oldoldbufptr);
629 SAVEPPTR(PL_last_lop);
630 SAVEPPTR(PL_last_uni);
631 SAVEPPTR(PL_linestart);
632 SAVESPTR(PL_linestr);
633 SAVEGENERICPV(PL_lex_brackstack);
634 SAVEGENERICPV(PL_lex_casestack);
635 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
636 SAVESPTR(PL_lex_stuff);
637 SAVEI32(PL_lex_defer);
638 SAVEI32(PL_sublex_info.sub_inwhat);
639 SAVESPTR(PL_lex_repl);
641 SAVEINT(PL_lex_expect);
643 PL_lex_state = LEX_NORMAL;
647 Newx(PL_lex_brackstack, 120, char);
648 Newx(PL_lex_casestack, 12, char);
650 *PL_lex_casestack = '\0';
662 PL_sublex_info.sub_inwhat = 0;
664 if (SvREADONLY(PL_linestr))
665 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
666 s = SvPV_const(PL_linestr, len);
667 if (!len || s[len-1] != ';') {
668 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
669 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
670 sv_catpvs(PL_linestr, "\n;");
672 SvTEMP_off(PL_linestr);
673 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
674 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
675 PL_last_lop = PL_last_uni = NULL;
681 * Finalizer for lexing operations. Must be called when the parser is
682 * done with the lexer.
689 PL_doextract = FALSE;
694 * This subroutine has nothing to do with tilting, whether at windmills
695 * or pinball tables. Its name is short for "increment line". It
696 * increments the current line number in CopLINE(PL_curcop) and checks
697 * to see whether the line starts with a comment of the form
698 * # line 500 "foo.pm"
699 * If so, it sets the current line number and file to the values in the comment.
703 S_incline(pTHX_ char *s)
711 CopLINE_inc(PL_curcop);
714 while (SPACE_OR_TAB(*s)) s++;
715 if (strnEQ(s, "line", 4))
719 if (SPACE_OR_TAB(*s))
723 while (SPACE_OR_TAB(*s)) s++;
729 while (SPACE_OR_TAB(*s))
731 if (*s == '"' && (t = strchr(s+1, '"'))) {
736 for (t = s; !isSPACE(*t); t++) ;
739 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
741 if (*e != '\n' && *e != '\0')
742 return; /* false alarm */
748 const char * const cf = CopFILE(PL_curcop);
749 STRLEN tmplen = cf ? strlen(cf) : 0;
750 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
751 /* must copy *{"::_<(eval N)[oldfilename:L]"}
752 * to *{"::_<newfilename"} */
753 char smallbuf[256], smallbuf2[256];
754 char *tmpbuf, *tmpbuf2;
756 STRLEN tmplen2 = strlen(s);
757 if (tmplen + 3 < sizeof smallbuf)
760 Newx(tmpbuf, tmplen + 3, char);
761 if (tmplen2 + 3 < sizeof smallbuf2)
764 Newx(tmpbuf2, tmplen2 + 3, char);
765 tmpbuf[0] = tmpbuf2[0] = '_';
766 tmpbuf[1] = tmpbuf2[1] = '<';
767 memcpy(tmpbuf + 2, cf, ++tmplen);
768 memcpy(tmpbuf2 + 2, s, ++tmplen2);
770 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
772 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
774 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
775 /* adjust ${"::_<newfilename"} to store the new file name */
776 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
777 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
778 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
780 if (tmpbuf != smallbuf) Safefree(tmpbuf);
781 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
784 CopFILE_free(PL_curcop);
785 CopFILE_set(PL_curcop, s);
788 CopLINE_set(PL_curcop, atoi(n)-1);
792 /* skip space before PL_thistoken */
795 S_skipspace0(pTHX_ register char *s)
802 PL_thiswhite = newSVpvn("",0);
803 sv_catsv(PL_thiswhite, PL_skipwhite);
804 sv_free(PL_skipwhite);
807 PL_realtokenstart = s - SvPVX(PL_linestr);
811 /* skip space after PL_thistoken */
814 S_skipspace1(pTHX_ register char *s)
817 I32 startoff = start - SvPVX(PL_linestr);
822 start = SvPVX(PL_linestr) + startoff;
823 if (!PL_thistoken && PL_realtokenstart >= 0) {
824 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
825 PL_thistoken = newSVpvn(tstart, start - tstart);
827 PL_realtokenstart = -1;
830 PL_nextwhite = newSVpvn("",0);
831 sv_catsv(PL_nextwhite, PL_skipwhite);
832 sv_free(PL_skipwhite);
839 S_skipspace2(pTHX_ register char *s, SV **svp)
842 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
843 I32 startoff = start - SvPVX(PL_linestr);
845 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
846 if (!PL_madskills || !svp)
848 start = SvPVX(PL_linestr) + startoff;
849 if (!PL_thistoken && PL_realtokenstart >= 0) {
850 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
851 PL_thistoken = newSVpvn(tstart, start - tstart);
852 PL_realtokenstart = -1;
856 *svp = newSVpvn("",0);
857 sv_setsv(*svp, PL_skipwhite);
858 sv_free(PL_skipwhite);
868 * Called to gobble the appropriate amount and type of whitespace.
869 * Skips comments as well.
873 S_skipspace(pTHX_ register char *s)
878 int startoff = s - SvPVX(PL_linestr);
881 sv_free(PL_skipwhite);
886 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
887 while (s < PL_bufend && SPACE_OR_TAB(*s))
897 SSize_t oldprevlen, oldoldprevlen;
898 SSize_t oldloplen = 0, oldunilen = 0;
899 while (s < PL_bufend && isSPACE(*s)) {
900 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
905 if (s < PL_bufend && *s == '#') {
906 while (s < PL_bufend && *s != '\n')
910 if (PL_in_eval && !PL_rsfp) {
917 /* only continue to recharge the buffer if we're at the end
918 * of the buffer, we're not reading from a source filter, and
919 * we're in normal lexing mode
921 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
922 PL_lex_state == LEX_FORMLINE)
929 /* try to recharge the buffer */
931 curoff = s - SvPVX(PL_linestr);
934 if ((s = filter_gets(PL_linestr, PL_rsfp,
935 (prevlen = SvCUR(PL_linestr)))) == NULL)
938 if (PL_madskills && curoff != startoff) {
940 PL_skipwhite = newSVpvn("",0);
941 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
945 /* mustn't throw out old stuff yet if madpropping */
946 SvCUR(PL_linestr) = curoff;
947 s = SvPVX(PL_linestr) + curoff;
949 if (curoff && s[-1] == '\n')
953 /* end of file. Add on the -p or -n magic */
954 /* XXX these shouldn't really be added here, can't set PL_faketokens */
958 ";}continue{print or die qq(-p destination: $!\\n);}");
961 ";}continue{print or die qq(-p destination: $!\\n);}");
963 PL_minus_n = PL_minus_p = 0;
965 else if (PL_minus_n) {
967 sv_catpvn(PL_linestr, ";}", 2);
969 sv_setpvn(PL_linestr, ";}", 2);
975 sv_catpvn(PL_linestr,";", 1);
977 sv_setpvn(PL_linestr,";", 1);
980 /* reset variables for next time we lex */
981 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
987 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
988 PL_last_lop = PL_last_uni = NULL;
990 /* Close the filehandle. Could be from -P preprocessor,
991 * STDIN, or a regular file. If we were reading code from
992 * STDIN (because the commandline held no -e or filename)
993 * then we don't close it, we reset it so the code can
994 * read from STDIN too.
997 if (PL_preprocess && !PL_in_eval)
998 (void)PerlProc_pclose(PL_rsfp);
999 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1000 PerlIO_clearerr(PL_rsfp);
1002 (void)PerlIO_close(PL_rsfp);
1007 /* not at end of file, so we only read another line */
1008 /* make corresponding updates to old pointers, for yyerror() */
1009 oldprevlen = PL_oldbufptr - PL_bufend;
1010 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1012 oldunilen = PL_last_uni - PL_bufend;
1014 oldloplen = PL_last_lop - PL_bufend;
1015 PL_linestart = PL_bufptr = s + prevlen;
1016 PL_bufend = s + SvCUR(PL_linestr);
1018 PL_oldbufptr = s + oldprevlen;
1019 PL_oldoldbufptr = s + oldoldprevlen;
1021 PL_last_uni = s + oldunilen;
1023 PL_last_lop = s + oldloplen;
1026 /* debugger active and we're not compiling the debugger code,
1027 * so store the line into the debugger's array of lines
1029 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1030 SV * const sv = newSV(0);
1032 sv_upgrade(sv, SVt_PVMG);
1033 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1036 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1044 PL_skipwhite = newSVpvn("",0);
1045 curoff = s - SvPVX(PL_linestr);
1046 if (curoff - startoff)
1047 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1056 * Check the unary operators to ensure there's no ambiguity in how they're
1057 * used. An ambiguous piece of code would be:
1059 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1060 * the +5 is its argument.
1070 if (PL_oldoldbufptr != PL_last_uni)
1072 while (isSPACE(*PL_last_uni))
1074 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++) ;
1075 if ((t = strchr(s, '(')) && t < PL_bufptr)
1078 if (ckWARN_d(WARN_AMBIGUOUS)){
1079 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1080 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1081 (int)(s - PL_last_uni), PL_last_uni);
1086 * LOP : macro to build a list operator. Its behaviour has been replaced
1087 * with a subroutine, S_lop() for which LOP is just another name.
1090 #define LOP(f,x) return lop(f,x,s)
1094 * Build a list operator (or something that might be one). The rules:
1095 * - if we have a next token, then it's a list operator [why?]
1096 * - if the next thing is an opening paren, then it's a function
1097 * - else it's a list operator
1101 S_lop(pTHX_ I32 f, int x, char *s)
1108 PL_last_lop = PL_oldbufptr;
1109 PL_last_lop_op = (OPCODE)f;
1112 return REPORT(LSTOP);
1115 return REPORT(LSTOP);
1118 return REPORT(FUNC);
1121 return REPORT(FUNC);
1123 return REPORT(LSTOP);
1129 * Sets up for an eventual force_next(). start_force(0) basically does
1130 * an unshift, while start_force(-1) does a push. yylex removes items
1135 S_start_force(pTHX_ int where)
1139 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1140 where = PL_lasttoke;
1141 assert(PL_curforce < 0 || PL_curforce == where);
1142 if (PL_curforce != where) {
1143 for (i = PL_lasttoke; i > where; --i) {
1144 PL_nexttoke[i] = PL_nexttoke[i-1];
1148 if (PL_curforce < 0) /* in case of duplicate start_force() */
1149 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1150 PL_curforce = where;
1153 curmad('^', newSVpvn("",0));
1154 CURMAD('_', PL_nextwhite);
1159 S_curmad(pTHX_ char slot, SV *sv)
1165 if (PL_curforce < 0)
1166 where = &PL_thismad;
1168 where = &PL_nexttoke[PL_curforce].next_mad;
1171 sv_setpvn(sv, "", 0);
1174 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1176 else if (PL_encoding) {
1177 sv_recode_to_utf8(sv, PL_encoding);
1182 /* keep a slot open for the head of the list? */
1183 if (slot != '_' && *where && (*where)->mad_key == '^') {
1184 (*where)->mad_key = slot;
1185 sv_free((*where)->mad_val);
1186 (*where)->mad_val = (void*)sv;
1189 addmad(newMADsv(slot, sv), where, 0);
1192 # define start_force(where)
1193 # define curmad(slot, sv)
1198 * When the lexer realizes it knows the next token (for instance,
1199 * it is reordering tokens for the parser) then it can call S_force_next
1200 * to know what token to return the next time the lexer is called. Caller
1201 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1202 * and possibly PL_expect to ensure the lexer handles the token correctly.
1206 S_force_next(pTHX_ I32 type)
1210 if (PL_curforce < 0)
1211 start_force(PL_lasttoke);
1212 PL_nexttoke[PL_curforce].next_type = type;
1213 if (PL_lex_state != LEX_KNOWNEXT)
1214 PL_lex_defer = PL_lex_state;
1215 PL_lex_state = LEX_KNOWNEXT;
1216 PL_lex_expect = PL_expect;
1219 PL_nexttype[PL_nexttoke] = type;
1221 if (PL_lex_state != LEX_KNOWNEXT) {
1222 PL_lex_defer = PL_lex_state;
1223 PL_lex_expect = PL_expect;
1224 PL_lex_state = LEX_KNOWNEXT;
1230 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1233 SV * const sv = newSVpvn(start,len);
1234 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1241 * When the lexer knows the next thing is a word (for instance, it has
1242 * just seen -> and it knows that the next char is a word char, then
1243 * it calls S_force_word to stick the next word into the PL_next lookahead.
1246 * char *start : buffer position (must be within PL_linestr)
1247 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1248 * int check_keyword : if true, Perl checks to make sure the word isn't
1249 * a keyword (do this if the word is a label, e.g. goto FOO)
1250 * int allow_pack : if true, : characters will also be allowed (require,
1251 * use, etc. do this)
1252 * int allow_initial_tick : used by the "sub" lexer only.
1256 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1262 start = SKIPSPACE1(start);
1264 if (isIDFIRST_lazy_if(s,UTF) ||
1265 (allow_pack && *s == ':') ||
1266 (allow_initial_tick && *s == '\'') )
1268 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1269 if (check_keyword && keyword(PL_tokenbuf, len))
1271 start_force(PL_curforce);
1273 curmad('X', newSVpvn(start,s-start));
1274 if (token == METHOD) {
1279 PL_expect = XOPERATOR;
1282 NEXTVAL_NEXTTOKE.opval
1283 = (OP*)newSVOP(OP_CONST,0,
1284 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1285 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1293 * Called when the lexer wants $foo *foo &foo etc, but the program
1294 * text only contains the "foo" portion. The first argument is a pointer
1295 * to the "foo", and the second argument is the type symbol to prefix.
1296 * Forces the next token to be a "WORD".
1297 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1301 S_force_ident(pTHX_ register const char *s, int kind)
1305 const STRLEN len = strlen(s);
1306 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1307 start_force(PL_curforce);
1308 NEXTVAL_NEXTTOKE.opval = o;
1311 o->op_private = OPpCONST_ENTERED;
1312 /* XXX see note in pp_entereval() for why we forgo typo
1313 warnings if the symbol must be introduced in an eval.
1315 gv_fetchpvn_flags(s, len,
1316 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1318 kind == '$' ? SVt_PV :
1319 kind == '@' ? SVt_PVAV :
1320 kind == '%' ? SVt_PVHV :
1328 Perl_str_to_version(pTHX_ SV *sv)
1333 const char *start = SvPV_const(sv,len);
1334 const char * const end = start + len;
1335 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1336 while (start < end) {
1340 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1345 retval += ((NV)n)/nshift;
1354 * Forces the next token to be a version number.
1355 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1356 * and if "guessing" is TRUE, then no new token is created (and the caller
1357 * must use an alternative parsing method).
1361 S_force_version(pTHX_ char *s, int guessing)
1367 I32 startoff = s - SvPVX(PL_linestr);
1376 while (isDIGIT(*d) || *d == '_' || *d == '.')
1380 start_force(PL_curforce);
1381 curmad('X', newSVpvn(s,d-s));
1384 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1386 s = scan_num(s, &yylval);
1387 version = yylval.opval;
1388 ver = cSVOPx(version)->op_sv;
1389 if (SvPOK(ver) && !SvNIOK(ver)) {
1390 SvUPGRADE(ver, SVt_PVNV);
1391 SvNV_set(ver, str_to_version(ver));
1392 SvNOK_on(ver); /* hint that it is a version */
1395 else if (guessing) {
1398 sv_free(PL_nextwhite); /* let next token collect whitespace */
1400 s = SvPVX(PL_linestr) + startoff;
1408 if (PL_madskills && !version) {
1409 sv_free(PL_nextwhite); /* let next token collect whitespace */
1411 s = SvPVX(PL_linestr) + startoff;
1414 /* NOTE: The parser sees the package name and the VERSION swapped */
1415 start_force(PL_curforce);
1416 NEXTVAL_NEXTTOKE.opval = version;
1424 * Tokenize a quoted string passed in as an SV. It finds the next
1425 * chunk, up to end of string or a backslash. It may make a new
1426 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1431 S_tokeq(pTHX_ SV *sv)
1435 register char *send;
1443 s = SvPV_force(sv, len);
1444 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1447 while (s < send && *s != '\\')
1452 if ( PL_hints & HINT_NEW_STRING ) {
1453 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1459 if (s + 1 < send && (s[1] == '\\'))
1460 s++; /* all that, just for this */
1465 SvCUR_set(sv, d - SvPVX_const(sv));
1467 if ( PL_hints & HINT_NEW_STRING )
1468 return new_constant(NULL, 0, "q", sv, pv, "q");
1473 * Now come three functions related to double-quote context,
1474 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1475 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1476 * interact with PL_lex_state, and create fake ( ... ) argument lists
1477 * to handle functions and concatenation.
1478 * They assume that whoever calls them will be setting up a fake
1479 * join call, because each subthing puts a ',' after it. This lets
1482 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1484 * (I'm not sure whether the spurious commas at the end of lcfirst's
1485 * arguments and join's arguments are created or not).
1490 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1492 * Pattern matching will set PL_lex_op to the pattern-matching op to
1493 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1495 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1497 * Everything else becomes a FUNC.
1499 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1500 * had an OP_CONST or OP_READLINE). This just sets us up for a
1501 * call to S_sublex_push().
1505 S_sublex_start(pTHX)
1508 register const I32 op_type = yylval.ival;
1510 if (op_type == OP_NULL) {
1511 yylval.opval = PL_lex_op;
1515 if (op_type == OP_CONST || op_type == OP_READLINE) {
1516 SV *sv = tokeq(PL_lex_stuff);
1518 if (SvTYPE(sv) == SVt_PVIV) {
1519 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1521 const char * const p = SvPV_const(sv, len);
1522 SV * const nsv = newSVpvn(p, len);
1528 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1529 PL_lex_stuff = NULL;
1530 /* Allow <FH> // "foo" */
1531 if (op_type == OP_READLINE)
1532 PL_expect = XTERMORDORDOR;
1536 PL_sublex_info.super_state = PL_lex_state;
1537 PL_sublex_info.sub_inwhat = op_type;
1538 PL_sublex_info.sub_op = PL_lex_op;
1539 PL_lex_state = LEX_INTERPPUSH;
1543 yylval.opval = PL_lex_op;
1553 * Create a new scope to save the lexing state. The scope will be
1554 * ended in S_sublex_done. Returns a '(', starting the function arguments
1555 * to the uc, lc, etc. found before.
1556 * Sets PL_lex_state to LEX_INTERPCONCAT.
1565 PL_lex_state = PL_sublex_info.super_state;
1566 SAVEI32(PL_lex_dojoin);
1567 SAVEI32(PL_lex_brackets);
1568 SAVEI32(PL_lex_casemods);
1569 SAVEI32(PL_lex_starts);
1570 SAVEI32(PL_lex_state);
1571 SAVEVPTR(PL_lex_inpat);
1572 SAVEI32(PL_lex_inwhat);
1573 SAVECOPLINE(PL_curcop);
1574 SAVEPPTR(PL_bufptr);
1575 SAVEPPTR(PL_bufend);
1576 SAVEPPTR(PL_oldbufptr);
1577 SAVEPPTR(PL_oldoldbufptr);
1578 SAVEPPTR(PL_last_lop);
1579 SAVEPPTR(PL_last_uni);
1580 SAVEPPTR(PL_linestart);
1581 SAVESPTR(PL_linestr);
1582 SAVEGENERICPV(PL_lex_brackstack);
1583 SAVEGENERICPV(PL_lex_casestack);
1585 PL_linestr = PL_lex_stuff;
1586 PL_lex_stuff = NULL;
1588 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1589 = SvPVX(PL_linestr);
1590 PL_bufend += SvCUR(PL_linestr);
1591 PL_last_lop = PL_last_uni = NULL;
1592 SAVEFREESV(PL_linestr);
1594 PL_lex_dojoin = FALSE;
1595 PL_lex_brackets = 0;
1596 Newx(PL_lex_brackstack, 120, char);
1597 Newx(PL_lex_casestack, 12, char);
1598 PL_lex_casemods = 0;
1599 *PL_lex_casestack = '\0';
1601 PL_lex_state = LEX_INTERPCONCAT;
1602 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1604 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1605 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1606 PL_lex_inpat = PL_sublex_info.sub_op;
1608 PL_lex_inpat = NULL;
1615 * Restores lexer state after a S_sublex_push.
1622 if (!PL_lex_starts++) {
1623 SV * const sv = newSVpvs("");
1624 if (SvUTF8(PL_linestr))
1626 PL_expect = XOPERATOR;
1627 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1631 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1632 PL_lex_state = LEX_INTERPCASEMOD;
1636 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1637 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1638 PL_linestr = PL_lex_repl;
1640 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1641 PL_bufend += SvCUR(PL_linestr);
1642 PL_last_lop = PL_last_uni = NULL;
1643 SAVEFREESV(PL_linestr);
1644 PL_lex_dojoin = FALSE;
1645 PL_lex_brackets = 0;
1646 PL_lex_casemods = 0;
1647 *PL_lex_casestack = '\0';
1649 if (SvEVALED(PL_lex_repl)) {
1650 PL_lex_state = LEX_INTERPNORMAL;
1652 /* we don't clear PL_lex_repl here, so that we can check later
1653 whether this is an evalled subst; that means we rely on the
1654 logic to ensure sublex_done() is called again only via the
1655 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1658 PL_lex_state = LEX_INTERPCONCAT;
1668 PL_endwhite = newSVpvn("",0);
1669 sv_catsv(PL_endwhite, PL_thiswhite);
1673 sv_setpvn(PL_thistoken,"",0);
1675 PL_realtokenstart = -1;
1679 PL_bufend = SvPVX(PL_linestr);
1680 PL_bufend += SvCUR(PL_linestr);
1681 PL_expect = XOPERATOR;
1682 PL_sublex_info.sub_inwhat = 0;
1690 Extracts a pattern, double-quoted string, or transliteration. This
1693 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1694 processing a pattern (PL_lex_inpat is true), a transliteration
1695 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1697 Returns a pointer to the character scanned up to. Iff this is
1698 advanced from the start pointer supplied (ie if anything was
1699 successfully parsed), will leave an OP for the substring scanned
1700 in yylval. Caller must intuit reason for not parsing further
1701 by looking at the next characters herself.
1705 double-quoted style: \r and \n
1706 regexp special ones: \D \s
1708 backrefs: \1 (deprecated in substitution replacements)
1709 case and quoting: \U \Q \E
1710 stops on @ and $, but not for $ as tail anchor
1712 In transliterations:
1713 characters are VERY literal, except for - not at the start or end
1714 of the string, which indicates a range. scan_const expands the
1715 range to the full set of intermediate characters.
1717 In double-quoted strings:
1719 double-quoted style: \r and \n
1721 backrefs: \1 (deprecated)
1722 case and quoting: \U \Q \E
1725 scan_const does *not* construct ops to handle interpolated strings.
1726 It stops processing as soon as it finds an embedded $ or @ variable
1727 and leaves it to the caller to work out what's going on.
1729 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1731 $ in pattern could be $foo or could be tail anchor. Assumption:
1732 it's a tail anchor if $ is the last thing in the string, or if it's
1733 followed by one of ")| \n\t"
1735 \1 (backreferences) are turned into $1
1737 The structure of the code is
1738 while (there's a character to process) {
1739 handle transliteration ranges
1740 skip regexp comments
1741 skip # initiated comments in //x patterns
1742 check for embedded @foo
1743 check for embedded scalars
1745 leave intact backslashes from leave (below)
1746 deprecate \1 in strings and sub replacements
1747 handle string-changing backslashes \l \U \Q \E, etc.
1748 switch (what was escaped) {
1749 handle - in a transliteration (becomes a literal -)
1750 handle \132 octal characters
1751 handle 0x15 hex characters
1752 handle \cV (control V)
1753 handle printf backslashes (\f, \r, \n, etc)
1755 } (end if backslash)
1756 } (end while character to read)
1761 S_scan_const(pTHX_ char *start)
1764 register char *send = PL_bufend; /* end of the constant */
1765 SV *sv = newSV(send - start); /* sv for the constant */
1766 register char *s = start; /* start of the constant */
1767 register char *d = SvPVX(sv); /* destination for copies */
1768 bool dorange = FALSE; /* are we in a translit range? */
1769 bool didrange = FALSE; /* did we just finish a range? */
1770 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1771 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1774 UV literal_endpoint = 0;
1777 const char *leaveit = /* set of acceptably-backslashed characters */
1779 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1782 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1783 /* If we are doing a trans and we know we want UTF8 set expectation */
1784 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1785 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1789 while (s < send || dorange) {
1790 /* get transliterations out of the way (they're most literal) */
1791 if (PL_lex_inwhat == OP_TRANS) {
1792 /* expand a range A-Z to the full set of characters. AIE! */
1794 I32 i; /* current expanded character */
1795 I32 min; /* first character in range */
1796 I32 max; /* last character in range */
1799 char * const c = (char*)utf8_hop((U8*)d, -1);
1803 *c = (char)UTF_TO_NATIVE(0xff);
1804 /* mark the range as done, and continue */
1810 i = d - SvPVX_const(sv); /* remember current offset */
1811 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1812 d = SvPVX(sv) + i; /* refresh d after realloc */
1813 d -= 2; /* eat the first char and the - */
1815 min = (U8)*d; /* first char in range */
1816 max = (U8)d[1]; /* last char in range */
1820 "Invalid range \"%c-%c\" in transliteration operator",
1821 (char)min, (char)max);
1825 if (literal_endpoint == 2 &&
1826 ((isLOWER(min) && isLOWER(max)) ||
1827 (isUPPER(min) && isUPPER(max)))) {
1829 for (i = min; i <= max; i++)
1831 *d++ = NATIVE_TO_NEED(has_utf8,i);
1833 for (i = min; i <= max; i++)
1835 *d++ = NATIVE_TO_NEED(has_utf8,i);
1840 for (i = min; i <= max; i++)
1843 /* mark the range as done, and continue */
1847 literal_endpoint = 0;
1852 /* range begins (ignore - as first or last char) */
1853 else if (*s == '-' && s+1 < send && s != start) {
1855 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1858 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1868 literal_endpoint = 0;
1873 /* if we get here, we're not doing a transliteration */
1875 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1876 except for the last char, which will be done separately. */
1877 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1879 while (s+1 < send && *s != ')')
1880 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1882 else if (s[2] == '{' /* This should match regcomp.c */
1883 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1886 char *regparse = s + (s[2] == '{' ? 3 : 4);
1889 while (count && (c = *regparse)) {
1890 if (c == '\\' && regparse[1])
1898 if (*regparse != ')')
1899 regparse--; /* Leave one char for continuation. */
1900 while (s < regparse)
1901 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1905 /* likewise skip #-initiated comments in //x patterns */
1906 else if (*s == '#' && PL_lex_inpat &&
1907 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1908 while (s+1 < send && *s != '\n')
1909 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1912 /* check for embedded arrays
1913 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1915 else if (*s == '@' && s[1]
1916 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1919 /* check for embedded scalars. only stop if we're sure it's a
1922 else if (*s == '$') {
1923 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1925 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1926 break; /* in regexp, $ might be tail anchor */
1929 /* End of else if chain - OP_TRANS rejoin rest */
1932 if (*s == '\\' && s+1 < send) {
1935 /* some backslashes we leave behind */
1936 if (*leaveit && *s && strchr(leaveit, *s)) {
1937 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1938 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1942 /* deprecate \1 in strings and substitution replacements */
1943 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1944 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1946 if (ckWARN(WARN_SYNTAX))
1947 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1952 /* string-change backslash escapes */
1953 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1958 /* if we get here, it's either a quoted -, or a digit */
1961 /* quoted - in transliterations */
1963 if (PL_lex_inwhat == OP_TRANS) {
1973 Perl_warner(aTHX_ packWARN(WARN_MISC),
1974 "Unrecognized escape \\%c passed through",
1976 /* default action is to copy the quoted character */
1977 goto default_action;
1980 /* \132 indicates an octal constant */
1981 case '0': case '1': case '2': case '3':
1982 case '4': case '5': case '6': case '7':
1986 uv = grok_oct(s, &len, &flags, NULL);
1989 goto NUM_ESCAPE_INSERT;
1991 /* \x24 indicates a hex constant */
1995 char* const e = strchr(s, '}');
1996 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
1997 PERL_SCAN_DISALLOW_PREFIX;
2002 yyerror("Missing right brace on \\x{}");
2006 uv = grok_hex(s, &len, &flags, NULL);
2012 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2013 uv = grok_hex(s, &len, &flags, NULL);
2019 /* Insert oct or hex escaped character.
2020 * There will always enough room in sv since such
2021 * escapes will be longer than any UTF-8 sequence
2022 * they can end up as. */
2024 /* We need to map to chars to ASCII before doing the tests
2027 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2028 if (!has_utf8 && uv > 255) {
2029 /* Might need to recode whatever we have
2030 * accumulated so far if it contains any
2033 * (Can't we keep track of that and avoid
2034 * this rescan? --jhi)
2038 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2039 if (!NATIVE_IS_INVARIANT(*c)) {
2044 const STRLEN offset = d - SvPVX_const(sv);
2046 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2050 while (src >= (const U8 *)SvPVX_const(sv)) {
2051 if (!NATIVE_IS_INVARIANT(*src)) {
2052 const U8 ch = NATIVE_TO_ASCII(*src);
2053 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2054 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2064 if (has_utf8 || uv > 255) {
2065 d = (char*)uvchr_to_utf8((U8*)d, uv);
2067 if (PL_lex_inwhat == OP_TRANS &&
2068 PL_sublex_info.sub_op) {
2069 PL_sublex_info.sub_op->op_private |=
2070 (PL_lex_repl ? OPpTRANS_FROM_UTF
2083 /* \N{LATIN SMALL LETTER A} is a named character */
2087 char* e = strchr(s, '}');
2093 yyerror("Missing right brace on \\N{}");
2097 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2099 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2100 PERL_SCAN_DISALLOW_PREFIX;
2103 uv = grok_hex(s, &len, &flags, NULL);
2105 goto NUM_ESCAPE_INSERT;
2107 res = newSVpvn(s + 1, e - s - 1);
2108 res = new_constant( NULL, 0, "charnames",
2109 res, NULL, "\\N{...}" );
2111 sv_utf8_upgrade(res);
2112 str = SvPV_const(res,len);
2113 #ifdef EBCDIC_NEVER_MIND
2114 /* charnames uses pack U and that has been
2115 * recently changed to do the below uni->native
2116 * mapping, so this would be redundant (and wrong,
2117 * the code point would be doubly converted).
2118 * But leave this in just in case the pack U change
2119 * gets revoked, but the semantics is still
2120 * desireable for charnames. --jhi */
2122 UV uv = utf8_to_uvchr((const U8*)str, 0);
2125 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2127 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2128 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2129 str = SvPV_const(res, len);
2133 if (!has_utf8 && SvUTF8(res)) {
2134 const char * const ostart = SvPVX_const(sv);
2135 SvCUR_set(sv, d - ostart);
2138 sv_utf8_upgrade(sv);
2139 /* this just broke our allocation above... */
2140 SvGROW(sv, (STRLEN)(send - start));
2141 d = SvPVX(sv) + SvCUR(sv);
2144 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2145 const char * const odest = SvPVX_const(sv);
2147 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2148 d = SvPVX(sv) + (d - odest);
2150 Copy(str, d, len, char);
2157 yyerror("Missing braces on \\N{}");
2160 /* \c is a control character */
2169 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2172 yyerror("Missing control char name in \\c");
2176 /* printf-style backslashes, formfeeds, newlines, etc */
2178 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2181 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2184 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2187 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2190 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2193 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2196 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2202 } /* end if (backslash) */
2209 /* If we started with encoded form, or already know we want it
2210 and then encode the next character */
2211 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2213 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2214 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2217 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2218 const STRLEN off = d - SvPVX_const(sv);
2219 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2221 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2225 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2227 } /* while loop to process each character */
2229 /* terminate the string and set up the sv */
2231 SvCUR_set(sv, d - SvPVX_const(sv));
2232 if (SvCUR(sv) >= SvLEN(sv))
2233 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2236 if (PL_encoding && !has_utf8) {
2237 sv_recode_to_utf8(sv, PL_encoding);
2243 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2244 PL_sublex_info.sub_op->op_private |=
2245 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2249 /* shrink the sv if we allocated more than we used */
2250 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2251 SvPV_shrink_to_cur(sv);
2254 /* return the substring (via yylval) only if we parsed anything */
2255 if (s > PL_bufptr) {
2256 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2257 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2259 ( PL_lex_inwhat == OP_TRANS
2261 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2264 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2271 * Returns TRUE if there's more to the expression (e.g., a subscript),
2274 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2276 * ->[ and ->{ return TRUE
2277 * { and [ outside a pattern are always subscripts, so return TRUE
2278 * if we're outside a pattern and it's not { or [, then return FALSE
2279 * if we're in a pattern and the first char is a {
2280 * {4,5} (any digits around the comma) returns FALSE
2281 * if we're in a pattern and the first char is a [
2283 * [SOMETHING] has a funky algorithm to decide whether it's a
2284 * character class or not. It has to deal with things like
2285 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2286 * anything else returns TRUE
2289 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2292 S_intuit_more(pTHX_ register char *s)
2295 if (PL_lex_brackets)
2297 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2299 if (*s != '{' && *s != '[')
2304 /* In a pattern, so maybe we have {n,m}. */
2321 /* On the other hand, maybe we have a character class */
2324 if (*s == ']' || *s == '^')
2327 /* this is terrifying, and it works */
2328 int weight = 2; /* let's weigh the evidence */
2330 unsigned char un_char = 255, last_un_char;
2331 const char * const send = strchr(s,']');
2332 char tmpbuf[sizeof PL_tokenbuf * 4];
2334 if (!send) /* has to be an expression */
2337 Zero(seen,256,char);
2340 else if (isDIGIT(*s)) {
2342 if (isDIGIT(s[1]) && s[2] == ']')
2348 for (; s < send; s++) {
2349 last_un_char = un_char;
2350 un_char = (unsigned char)*s;
2355 weight -= seen[un_char] * 10;
2356 if (isALNUM_lazy_if(s+1,UTF)) {
2358 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2359 len = (int)strlen(tmpbuf);
2360 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2365 else if (*s == '$' && s[1] &&
2366 strchr("[#!%*<>()-=",s[1])) {
2367 if (/*{*/ strchr("])} =",s[2]))
2376 if (strchr("wds]",s[1]))
2378 else if (seen['\''] || seen['"'])
2380 else if (strchr("rnftbxcav",s[1]))
2382 else if (isDIGIT(s[1])) {
2384 while (s[1] && isDIGIT(s[1]))
2394 if (strchr("aA01! ",last_un_char))
2396 if (strchr("zZ79~",s[1]))
2398 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2399 weight -= 5; /* cope with negative subscript */
2402 if (!isALNUM(last_un_char)
2403 && !(last_un_char == '$' || last_un_char == '@'
2404 || last_un_char == '&')
2405 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2410 if (keyword(tmpbuf, d - tmpbuf))
2413 if (un_char == last_un_char + 1)
2415 weight -= seen[un_char];
2420 if (weight >= 0) /* probably a character class */
2430 * Does all the checking to disambiguate
2432 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2433 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2435 * First argument is the stuff after the first token, e.g. "bar".
2437 * Not a method if bar is a filehandle.
2438 * Not a method if foo is a subroutine prototyped to take a filehandle.
2439 * Not a method if it's really "Foo $bar"
2440 * Method if it's "foo $bar"
2441 * Not a method if it's really "print foo $bar"
2442 * Method if it's really "foo package::" (interpreted as package->foo)
2443 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2444 * Not a method if bar is a filehandle or package, but is quoted with
2449 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2452 char *s = start + (*start == '$');
2453 char tmpbuf[sizeof PL_tokenbuf];
2461 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2465 const char *proto = SvPVX_const(cv);
2476 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2477 /* start is the beginning of the possible filehandle/object,
2478 * and s is the end of it
2479 * tmpbuf is a copy of it
2482 if (*start == '$') {
2483 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2486 len = start - SvPVX(PL_linestr);
2490 start = SvPVX(PL_linestr) + len;
2494 return *s == '(' ? FUNCMETH : METHOD;
2496 if (!keyword(tmpbuf, len)) {
2497 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2501 soff = s - SvPVX(PL_linestr);
2505 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2506 if (indirgv && GvCVu(indirgv))
2508 /* filehandle or package name makes it a method */
2509 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2511 soff = s - SvPVX(PL_linestr);
2514 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2515 return 0; /* no assumptions -- "=>" quotes bearword */
2517 start_force(PL_curforce);
2518 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2519 newSVpvn(tmpbuf,len));
2520 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2522 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2527 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2529 return *s == '(' ? FUNCMETH : METHOD;
2537 * Return a string of Perl code to load the debugger. If PERL5DB
2538 * is set, it will return the contents of that, otherwise a
2539 * compile-time require of perl5db.pl.
2547 const char * const pdb = PerlEnv_getenv("PERL5DB");
2551 SETERRNO(0,SS_NORMAL);
2552 return "BEGIN { require 'perl5db.pl' }";
2558 /* Encoded script support. filter_add() effectively inserts a
2559 * 'pre-processing' function into the current source input stream.
2560 * Note that the filter function only applies to the current source file
2561 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2563 * The datasv parameter (which may be NULL) can be used to pass
2564 * private data to this instance of the filter. The filter function
2565 * can recover the SV using the FILTER_DATA macro and use it to
2566 * store private buffers and state information.
2568 * The supplied datasv parameter is upgraded to a PVIO type
2569 * and the IoDIRP/IoANY field is used to store the function pointer,
2570 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2571 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2572 * private use must be set using malloc'd pointers.
2576 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2582 if (!PL_rsfp_filters)
2583 PL_rsfp_filters = newAV();
2586 SvUPGRADE(datasv, SVt_PVIO);
2587 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2588 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2589 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2590 IoANY(datasv), SvPV_nolen(datasv)));
2591 av_unshift(PL_rsfp_filters, 1);
2592 av_store(PL_rsfp_filters, 0, datasv) ;
2597 /* Delete most recently added instance of this filter function. */
2599 Perl_filter_del(pTHX_ filter_t funcp)
2605 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2607 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2609 /* if filter is on top of stack (usual case) just pop it off */
2610 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2611 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2612 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2613 IoANY(datasv) = (void *)NULL;
2614 sv_free(av_pop(PL_rsfp_filters));
2618 /* we need to search for the correct entry and clear it */
2619 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2623 /* Invoke the idxth filter function for the current rsfp. */
2624 /* maxlen 0 = read one text line */
2626 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2632 if (!PL_rsfp_filters)
2634 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2635 /* Provide a default input filter to make life easy. */
2636 /* Note that we append to the line. This is handy. */
2637 DEBUG_P(PerlIO_printf(Perl_debug_log,
2638 "filter_read %d: from rsfp\n", idx));
2642 const int old_len = SvCUR(buf_sv);
2644 /* ensure buf_sv is large enough */
2645 SvGROW(buf_sv, (STRLEN)(old_len + maxlen)) ;
2646 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len, maxlen)) <= 0){
2647 if (PerlIO_error(PL_rsfp))
2648 return -1; /* error */
2650 return 0 ; /* end of file */
2652 SvCUR_set(buf_sv, old_len + len) ;
2655 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2656 if (PerlIO_error(PL_rsfp))
2657 return -1; /* error */
2659 return 0 ; /* end of file */
2662 return SvCUR(buf_sv);
2664 /* Skip this filter slot if filter has been deleted */
2665 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2666 DEBUG_P(PerlIO_printf(Perl_debug_log,
2667 "filter_read %d: skipped (filter deleted)\n",
2669 return FILTER_READ(idx+1, buf_sv, maxlen); /* recurse */
2671 /* Get function pointer hidden within datasv */
2672 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2673 DEBUG_P(PerlIO_printf(Perl_debug_log,
2674 "filter_read %d: via function %p (%s)\n",
2675 idx, datasv, SvPV_nolen_const(datasv)));
2676 /* Call function. The function is expected to */
2677 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2678 /* Return: <0:error, =0:eof, >0:not eof */
2679 return (*funcp)(aTHX_ idx, buf_sv, maxlen);
2683 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2686 #ifdef PERL_CR_FILTER
2687 if (!PL_rsfp_filters) {
2688 filter_add(S_cr_textfilter,NULL);
2691 if (PL_rsfp_filters) {
2693 SvCUR_set(sv, 0); /* start with empty line */
2694 if (FILTER_READ(0, sv, 0) > 0)
2695 return ( SvPVX(sv) ) ;
2700 return (sv_gets(sv, fp, append));
2704 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2709 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2713 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2714 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2716 return GvHV(gv); /* Foo:: */
2719 /* use constant CLASS => 'MyClass' */
2720 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2722 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2723 pkgname = SvPV_nolen_const(sv);
2727 return gv_stashpv(pkgname, FALSE);
2733 * The intent of this yylex wrapper is to minimize the changes to the
2734 * tokener when we aren't interested in collecting madprops. It remains
2735 * to be seen how successful this strategy will be...
2742 char *s = PL_bufptr;
2744 /* make sure PL_thiswhite is initialized */
2748 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2749 if (PL_pending_ident)
2750 return S_pending_ident(aTHX);
2752 /* previous token ate up our whitespace? */
2753 if (!PL_lasttoke && PL_nextwhite) {
2754 PL_thiswhite = PL_nextwhite;
2758 /* isolate the token, and figure out where it is without whitespace */
2759 PL_realtokenstart = -1;
2763 assert(PL_curforce < 0);
2765 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2766 if (!PL_thistoken) {
2767 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2768 PL_thistoken = newSVpvn("",0);
2770 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2771 PL_thistoken = newSVpvn(tstart, s - tstart);
2774 if (PL_thismad) /* install head */
2775 CURMAD('X', PL_thistoken);
2778 /* last whitespace of a sublex? */
2779 if (optype == ')' && PL_endwhite) {
2780 CURMAD('X', PL_endwhite);
2785 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2786 if (!PL_thiswhite && !PL_endwhite && !optype) {
2787 sv_free(PL_thistoken);
2792 /* put off final whitespace till peg */
2793 if (optype == ';' && !PL_rsfp) {
2794 PL_nextwhite = PL_thiswhite;
2797 else if (PL_thisopen) {
2798 CURMAD('q', PL_thisopen);
2800 sv_free(PL_thistoken);
2804 /* Store actual token text as madprop X */
2805 CURMAD('X', PL_thistoken);
2809 /* add preceding whitespace as madprop _ */
2810 CURMAD('_', PL_thiswhite);
2814 /* add quoted material as madprop = */
2815 CURMAD('=', PL_thisstuff);
2819 /* add terminating quote as madprop Q */
2820 CURMAD('Q', PL_thisclose);
2824 /* special processing based on optype */
2828 /* opval doesn't need a TOKEN since it can already store mp */
2839 append_madprops(PL_thismad, yylval.opval, 0);
2847 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2856 /* remember any fake bracket that lexer is about to discard */
2857 if (PL_lex_brackets == 1 &&
2858 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2861 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2864 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2865 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2868 break; /* don't bother looking for trailing comment */
2877 /* attach a trailing comment to its statement instead of next token */
2881 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2883 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2885 if (*s == '\n' || *s == '#') {
2886 while (s < PL_bufend && *s != '\n')
2890 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2891 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2908 /* Create new token struct. Note: opvals return early above. */
2909 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2916 S_tokenize_use(pTHX_ int is_use, char *s) {
2918 if (PL_expect != XSTATE)
2919 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2920 is_use ? "use" : "no"));
2922 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2923 s = force_version(s, TRUE);
2924 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2925 start_force(PL_curforce);
2926 NEXTVAL_NEXTTOKE.opval = NULL;
2929 else if (*s == 'v') {
2930 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2931 s = force_version(s, FALSE);
2935 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2936 s = force_version(s, FALSE);
2938 yylval.ival = is_use;
2942 static const char* const exp_name[] =
2943 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2944 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2951 Works out what to call the token just pulled out of the input
2952 stream. The yacc parser takes care of taking the ops we return and
2953 stitching them into a tree.
2959 if read an identifier
2960 if we're in a my declaration
2961 croak if they tried to say my($foo::bar)
2962 build the ops for a my() declaration
2963 if it's an access to a my() variable
2964 are we in a sort block?
2965 croak if my($a); $a <=> $b
2966 build ops for access to a my() variable
2967 if in a dq string, and they've said @foo and we can't find @foo
2969 build ops for a bareword
2970 if we already built the token before, use it.
2975 #pragma segment Perl_yylex
2981 register char *s = PL_bufptr;
2987 SV* tmp = newSVpvs("");
2988 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
2989 (IV)CopLINE(PL_curcop),
2990 lex_state_names[PL_lex_state],
2991 exp_name[PL_expect],
2992 pv_display(tmp, s, strlen(s), 0, 60));
2995 /* check if there's an identifier for us to look at */
2996 if (PL_pending_ident)
2997 return REPORT(S_pending_ident(aTHX));
2999 /* no identifier pending identification */
3001 switch (PL_lex_state) {
3003 case LEX_NORMAL: /* Some compilers will produce faster */
3004 case LEX_INTERPNORMAL: /* code if we comment these out. */
3008 /* when we've already built the next token, just pull it out of the queue */
3012 yylval = PL_nexttoke[PL_lasttoke].next_val;
3014 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3015 PL_nexttoke[PL_lasttoke].next_mad = 0;
3016 if (PL_thismad && PL_thismad->mad_key == '_') {
3017 PL_thiswhite = (SV*)PL_thismad->mad_val;
3018 PL_thismad->mad_val = 0;
3019 mad_free(PL_thismad);
3024 PL_lex_state = PL_lex_defer;
3025 PL_expect = PL_lex_expect;
3026 PL_lex_defer = LEX_NORMAL;
3027 if (!PL_nexttoke[PL_lasttoke].next_type)
3032 yylval = PL_nextval[PL_nexttoke];
3034 PL_lex_state = PL_lex_defer;
3035 PL_expect = PL_lex_expect;
3036 PL_lex_defer = LEX_NORMAL;
3040 /* FIXME - can these be merged? */
3041 return(PL_nexttoke[PL_lasttoke].next_type);
3043 return REPORT(PL_nexttype[PL_nexttoke]);
3046 /* interpolated case modifiers like \L \U, including \Q and \E.
3047 when we get here, PL_bufptr is at the \
3049 case LEX_INTERPCASEMOD:
3051 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3052 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3054 /* handle \E or end of string */
3055 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3057 if (PL_lex_casemods) {
3058 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3059 PL_lex_casestack[PL_lex_casemods] = '\0';
3061 if (PL_bufptr != PL_bufend
3062 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3064 PL_lex_state = LEX_INTERPCONCAT;
3067 PL_thistoken = newSVpvn("\\E",2);
3073 while (PL_bufptr != PL_bufend &&
3074 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3076 PL_thiswhite = newSVpvn("",0);
3077 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3081 if (PL_bufptr != PL_bufend)
3084 PL_lex_state = LEX_INTERPCONCAT;
3088 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3089 "### Saw case modifier\n"); });
3091 if (s[1] == '\\' && s[2] == 'E') {
3094 PL_thiswhite = newSVpvn("",0);
3095 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3098 PL_lex_state = LEX_INTERPCONCAT;
3103 if (!PL_madskills) /* when just compiling don't need correct */
3104 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3105 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3106 if ((*s == 'L' || *s == 'U') &&
3107 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3108 PL_lex_casestack[--PL_lex_casemods] = '\0';
3111 if (PL_lex_casemods > 10)
3112 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3113 PL_lex_casestack[PL_lex_casemods++] = *s;
3114 PL_lex_casestack[PL_lex_casemods] = '\0';
3115 PL_lex_state = LEX_INTERPCONCAT;
3116 start_force(PL_curforce);
3117 NEXTVAL_NEXTTOKE.ival = 0;
3119 start_force(PL_curforce);
3121 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3123 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3125 NEXTVAL_NEXTTOKE.ival = OP_LC;
3127 NEXTVAL_NEXTTOKE.ival = OP_UC;
3129 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3131 Perl_croak(aTHX_ "panic: yylex");
3133 SV* tmpsv = newSVpvn("",0);
3134 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3140 if (PL_lex_starts) {
3146 sv_free(PL_thistoken);
3147 PL_thistoken = newSVpvn("",0);
3150 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3151 if (PL_lex_casemods == 1 && PL_lex_inpat)
3160 case LEX_INTERPPUSH:
3161 return REPORT(sublex_push());
3163 case LEX_INTERPSTART:
3164 if (PL_bufptr == PL_bufend)
3165 return REPORT(sublex_done());
3166 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3167 "### Interpolated variable\n"); });
3169 PL_lex_dojoin = (*PL_bufptr == '@');
3170 PL_lex_state = LEX_INTERPNORMAL;
3171 if (PL_lex_dojoin) {
3172 start_force(PL_curforce);
3173 NEXTVAL_NEXTTOKE.ival = 0;
3175 start_force(PL_curforce);
3176 force_ident("\"", '$');
3177 start_force(PL_curforce);
3178 NEXTVAL_NEXTTOKE.ival = 0;
3180 start_force(PL_curforce);
3181 NEXTVAL_NEXTTOKE.ival = 0;
3183 start_force(PL_curforce);
3184 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3187 if (PL_lex_starts++) {
3192 sv_free(PL_thistoken);
3193 PL_thistoken = newSVpvn("",0);
3196 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3197 if (!PL_lex_casemods && PL_lex_inpat)
3204 case LEX_INTERPENDMAYBE:
3205 if (intuit_more(PL_bufptr)) {
3206 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3212 if (PL_lex_dojoin) {
3213 PL_lex_dojoin = FALSE;
3214 PL_lex_state = LEX_INTERPCONCAT;
3218 sv_free(PL_thistoken);
3219 PL_thistoken = newSVpvn("",0);
3224 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3225 && SvEVALED(PL_lex_repl))
3227 if (PL_bufptr != PL_bufend)
3228 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3232 case LEX_INTERPCONCAT:
3234 if (PL_lex_brackets)
3235 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3237 if (PL_bufptr == PL_bufend)
3238 return REPORT(sublex_done());
3240 if (SvIVX(PL_linestr) == '\'') {
3241 SV *sv = newSVsv(PL_linestr);
3244 else if ( PL_hints & HINT_NEW_RE )
3245 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3246 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3250 s = scan_const(PL_bufptr);
3252 PL_lex_state = LEX_INTERPCASEMOD;
3254 PL_lex_state = LEX_INTERPSTART;
3257 if (s != PL_bufptr) {
3258 start_force(PL_curforce);
3260 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3262 NEXTVAL_NEXTTOKE = yylval;
3265 if (PL_lex_starts++) {
3269 sv_free(PL_thistoken);
3270 PL_thistoken = newSVpvn("",0);
3273 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3274 if (!PL_lex_casemods && PL_lex_inpat)
3287 PL_lex_state = LEX_NORMAL;
3288 s = scan_formline(PL_bufptr);
3289 if (!PL_lex_formbrack)
3295 PL_oldoldbufptr = PL_oldbufptr;
3301 sv_free(PL_thistoken);
3304 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3308 if (isIDFIRST_lazy_if(s,UTF))
3310 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3313 goto fake_eof; /* emulate EOF on ^D or ^Z */
3322 if (PL_lex_brackets) {
3323 yyerror(PL_lex_formbrack
3324 ? "Format not terminated"
3325 : "Missing right curly or square bracket");
3327 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3328 "### Tokener got EOF\n");
3332 if (s++ < PL_bufend)
3333 goto retry; /* ignore stray nulls */
3336 if (!PL_in_eval && !PL_preambled) {
3337 PL_preambled = TRUE;
3342 sv_setpv(PL_linestr,incl_perldb());
3343 if (SvCUR(PL_linestr))
3344 sv_catpvs(PL_linestr,";");
3346 while(AvFILLp(PL_preambleav) >= 0) {
3347 SV *tmpsv = av_shift(PL_preambleav);
3348 sv_catsv(PL_linestr, tmpsv);
3349 sv_catpvs(PL_linestr, ";");
3352 sv_free((SV*)PL_preambleav);
3353 PL_preambleav = NULL;
3355 if (PL_minus_n || PL_minus_p) {
3356 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3358 sv_catpvs(PL_linestr,"chomp;");
3361 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3362 || *PL_splitstr == '"')
3363 && strchr(PL_splitstr + 1, *PL_splitstr))
3364 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3366 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3367 bytes can be used as quoting characters. :-) */
3368 const char *splits = PL_splitstr;
3369 sv_catpvs(PL_linestr, "our @F=split(q\0");
3372 if (*splits == '\\')
3373 sv_catpvn(PL_linestr, splits, 1);
3374 sv_catpvn(PL_linestr, splits, 1);
3375 } while (*splits++);
3376 /* This loop will embed the trailing NUL of
3377 PL_linestr as the last thing it does before
3379 sv_catpvs(PL_linestr, ");");
3383 sv_catpvs(PL_linestr,"our @F=split(' ');");
3387 sv_catpvs(PL_linestr,"use feature ':5.10';");
3388 sv_catpvs(PL_linestr, "\n");
3389 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3390 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3391 PL_last_lop = PL_last_uni = NULL;
3392 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3393 SV * const sv = newSV(0);
3395 sv_upgrade(sv, SVt_PVMG);
3396 sv_setsv(sv,PL_linestr);
3399 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3404 bof = PL_rsfp ? TRUE : FALSE;
3405 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3408 PL_realtokenstart = -1;
3411 if (PL_preprocess && !PL_in_eval)
3412 (void)PerlProc_pclose(PL_rsfp);
3413 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3414 PerlIO_clearerr(PL_rsfp);
3416 (void)PerlIO_close(PL_rsfp);
3418 PL_doextract = FALSE;
3420 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3425 sv_setpv(PL_linestr,PL_minus_p
3426 ? ";}continue{print;}" : ";}");
3427 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3428 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3429 PL_last_lop = PL_last_uni = NULL;
3430 PL_minus_n = PL_minus_p = 0;
3433 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3434 PL_last_lop = PL_last_uni = NULL;
3435 sv_setpvn(PL_linestr,"",0);
3436 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3438 /* If it looks like the start of a BOM or raw UTF-16,
3439 * check if it in fact is. */
3445 #ifdef PERLIO_IS_STDIO
3446 # ifdef __GNU_LIBRARY__
3447 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3448 # define FTELL_FOR_PIPE_IS_BROKEN
3452 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3453 # define FTELL_FOR_PIPE_IS_BROKEN
3458 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3459 /* This loses the possibility to detect the bof
3460 * situation on perl -P when the libc5 is being used.
3461 * Workaround? Maybe attach some extra state to PL_rsfp?
3464 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3466 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3469 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3470 s = swallow_bom((U8*)s);
3474 /* Incest with pod. */
3477 sv_catsv(PL_thiswhite, PL_linestr);
3479 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3480 sv_setpvn(PL_linestr, "", 0);
3481 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3482 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3483 PL_last_lop = PL_last_uni = NULL;
3484 PL_doextract = FALSE;
3488 } while (PL_doextract);
3489 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3490 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3491 SV * const sv = newSV(0);
3493 sv_upgrade(sv, SVt_PVMG);
3494 sv_setsv(sv,PL_linestr);
3497 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3499 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3500 PL_last_lop = PL_last_uni = NULL;
3501 if (CopLINE(PL_curcop) == 1) {
3502 while (s < PL_bufend && isSPACE(*s))
3504 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3508 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3512 if (*s == '#' && *(s+1) == '!')
3514 #ifdef ALTERNATE_SHEBANG
3516 static char const as[] = ALTERNATE_SHEBANG;
3517 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3518 d = s + (sizeof(as) - 1);
3520 #endif /* ALTERNATE_SHEBANG */
3529 while (*d && !isSPACE(*d))
3533 #ifdef ARG_ZERO_IS_SCRIPT
3534 if (ipathend > ipath) {
3536 * HP-UX (at least) sets argv[0] to the script name,
3537 * which makes $^X incorrect. And Digital UNIX and Linux,
3538 * at least, set argv[0] to the basename of the Perl
3539 * interpreter. So, having found "#!", we'll set it right.
3541 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3543 assert(SvPOK(x) || SvGMAGICAL(x));
3544 if (sv_eq(x, CopFILESV(PL_curcop))) {
3545 sv_setpvn(x, ipath, ipathend - ipath);
3551 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3552 const char * const lstart = SvPV_const(x,llen);
3554 bstart += blen - llen;
3555 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3556 sv_setpvn(x, ipath, ipathend - ipath);
3561 TAINT_NOT; /* $^X is always tainted, but that's OK */
3563 #endif /* ARG_ZERO_IS_SCRIPT */
3568 d = instr(s,"perl -");
3570 d = instr(s,"perl");
3572 /* avoid getting into infinite loops when shebang
3573 * line contains "Perl" rather than "perl" */
3575 for (d = ipathend-4; d >= ipath; --d) {
3576 if ((*d == 'p' || *d == 'P')
3577 && !ibcmp(d, "perl", 4))
3587 #ifdef ALTERNATE_SHEBANG
3589 * If the ALTERNATE_SHEBANG on this system starts with a
3590 * character that can be part of a Perl expression, then if
3591 * we see it but not "perl", we're probably looking at the
3592 * start of Perl code, not a request to hand off to some
3593 * other interpreter. Similarly, if "perl" is there, but
3594 * not in the first 'word' of the line, we assume the line
3595 * contains the start of the Perl program.
3597 if (d && *s != '#') {
3598 const char *c = ipath;
3599 while (*c && !strchr("; \t\r\n\f\v#", *c))
3602 d = NULL; /* "perl" not in first word; ignore */
3604 *s = '#'; /* Don't try to parse shebang line */
3606 #endif /* ALTERNATE_SHEBANG */
3607 #ifndef MACOS_TRADITIONAL
3612 !instr(s,"indir") &&
3613 instr(PL_origargv[0],"perl"))
3620 while (s < PL_bufend && isSPACE(*s))
3622 if (s < PL_bufend) {
3623 Newxz(newargv,PL_origargc+3,char*);
3625 while (s < PL_bufend && !isSPACE(*s))
3628 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3631 newargv = PL_origargv;
3634 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3636 Perl_croak(aTHX_ "Can't exec %s", ipath);
3640 while (*d && !isSPACE(*d)) d++;
3641 while (SPACE_OR_TAB(*d)) d++;
3644 const bool switches_done = PL_doswitches;
3645 const U32 oldpdb = PL_perldb;
3646 const bool oldn = PL_minus_n;
3647 const bool oldp = PL_minus_p;
3650 if (*d == 'M' || *d == 'm' || *d == 'C') {
3651 const char * const m = d;
3652 while (*d && !isSPACE(*d)) d++;
3653 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3656 d = moreswitches(d);
3658 if (PL_doswitches && !switches_done) {
3659 int argc = PL_origargc;
3660 char **argv = PL_origargv;
3663 } while (argc && argv[0][0] == '-' && argv[0][1]);
3664 init_argv_symbols(argc,argv);
3666 if ((PERLDB_LINE && !oldpdb) ||
3667 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3668 /* if we have already added "LINE: while (<>) {",
3669 we must not do it again */
3671 sv_setpvn(PL_linestr, "", 0);
3672 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3673 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3674 PL_last_lop = PL_last_uni = NULL;
3675 PL_preambled = FALSE;
3677 (void)gv_fetchfile(PL_origfilename);
3684 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3686 PL_lex_state = LEX_FORMLINE;
3691 #ifdef PERL_STRICT_CR
3692 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3694 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3696 case ' ': case '\t': case '\f': case 013:
3697 #ifdef MACOS_TRADITIONAL
3701 PL_realtokenstart = -1;
3710 PL_realtokenstart = -1;
3714 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3715 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3716 /* handle eval qq[#line 1 "foo"\n ...] */
3717 CopLINE_dec(PL_curcop);
3720 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3722 if (!PL_in_eval || PL_rsfp)
3727 while (d < PL_bufend && *d != '\n')
3731 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3732 Perl_croak(aTHX_ "panic: input overflow");
3735 PL_thiswhite = newSVpvn(s, d - s);
3740 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3742 PL_lex_state = LEX_FORMLINE;
3748 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3749 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3752 TOKEN(PEG); /* make sure any #! line is accessible */
3757 /* if (PL_madskills && PL_lex_formbrack) { */
3759 while (d < PL_bufend && *d != '\n')
3763 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3764 Perl_croak(aTHX_ "panic: input overflow");
3765 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3767 PL_thiswhite = newSVpvn("",0);
3768 if (CopLINE(PL_curcop) == 1) {
3769 sv_setpvn(PL_thiswhite, "", 0);
3772 sv_catpvn(PL_thiswhite, s, d - s);
3786 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3794 while (s < PL_bufend && SPACE_OR_TAB(*s))
3797 if (strnEQ(s,"=>",2)) {
3798 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3799 DEBUG_T( { S_printbuf(aTHX_
3800 "### Saw unary minus before =>, forcing word %s\n", s);
3802 OPERATOR('-'); /* unary minus */
3804 PL_last_uni = PL_oldbufptr;
3806 case 'r': ftst = OP_FTEREAD; break;
3807 case 'w': ftst = OP_FTEWRITE; break;
3808 case 'x': ftst = OP_FTEEXEC; break;
3809 case 'o': ftst = OP_FTEOWNED; break;
3810 case 'R': ftst = OP_FTRREAD; break;
3811 case 'W': ftst = OP_FTRWRITE; break;
3812 case 'X': ftst = OP_FTREXEC; break;
3813 case 'O': ftst = OP_FTROWNED; break;
3814 case 'e': ftst = OP_FTIS; break;
3815 case 'z': ftst = OP_FTZERO; break;
3816 case 's': ftst = OP_FTSIZE; break;
3817 case 'f': ftst = OP_FTFILE; break;
3818 case 'd': ftst = OP_FTDIR; break;
3819 case 'l': ftst = OP_FTLINK; break;
3820 case 'p': ftst = OP_FTPIPE; break;
3821 case 'S': ftst = OP_FTSOCK; break;
3822 case 'u': ftst = OP_FTSUID; break;
3823 case 'g': ftst = OP_FTSGID; break;
3824 case 'k': ftst = OP_FTSVTX; break;
3825 case 'b': ftst = OP_FTBLK; break;
3826 case 'c': ftst = OP_FTCHR; break;
3827 case 't': ftst = OP_FTTTY; break;
3828 case 'T': ftst = OP_FTTEXT; break;
3829 case 'B': ftst = OP_FTBINARY; break;
3830 case 'M': case 'A': case 'C':
3831 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3833 case 'M': ftst = OP_FTMTIME; break;
3834 case 'A': ftst = OP_FTATIME; break;
3835 case 'C': ftst = OP_FTCTIME; break;
3843 PL_last_lop_op = (OPCODE)ftst;
3844 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3845 "### Saw file test %c\n", (int)tmp);
3850 /* Assume it was a minus followed by a one-letter named
3851 * subroutine call (or a -bareword), then. */
3852 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3853 "### '-%c' looked like a file test but was not\n",
3860 const char tmp = *s++;
3863 if (PL_expect == XOPERATOR)
3868 else if (*s == '>') {
3871 if (isIDFIRST_lazy_if(s,UTF)) {
3872 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3880 if (PL_expect == XOPERATOR)
3883 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3885 OPERATOR('-'); /* unary minus */
3891 const char tmp = *s++;
3894 if (PL_expect == XOPERATOR)
3899 if (PL_expect == XOPERATOR)
3902 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3909 if (PL_expect != XOPERATOR) {
3910 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3911 PL_expect = XOPERATOR;
3912 force_ident(PL_tokenbuf, '*');
3925 if (PL_expect == XOPERATOR) {
3929 PL_tokenbuf[0] = '%';
3930 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3931 if (!PL_tokenbuf[1]) {
3934 PL_pending_ident = '%';
3945 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3946 && FEATURE_IS_ENABLED("~~"))
3953 const char tmp = *s++;
3959 goto just_a_word_zero_gv;
3962 switch (PL_expect) {
3968 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3970 PL_bufptr = s; /* update in case we back off */
3976 PL_expect = XTERMBLOCK;
3979 stuffstart = s - SvPVX(PL_linestr) - 1;
3983 while (isIDFIRST_lazy_if(s,UTF)) {
3985 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
3986 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
3987 if (tmp < 0) tmp = -tmp;
4003 d = scan_str(d,TRUE,TRUE);
4005 /* MUST advance bufptr here to avoid bogus
4006 "at end of line" context messages from yyerror().
4008 PL_bufptr = s + len;
4009 yyerror("Unterminated attribute parameter in attribute list");
4012 return REPORT(0); /* EOF indicator */
4016 SV *sv = newSVpvn(s, len);
4017 sv_catsv(sv, PL_lex_stuff);
4018 attrs = append_elem(OP_LIST, attrs,
4019 newSVOP(OP_CONST, 0, sv));
4020 SvREFCNT_dec(PL_lex_stuff);
4021 PL_lex_stuff = NULL;
4024 if (len == 6 && strnEQ(s, "unique", len)) {
4025 if (PL_in_my == KEY_our) {
4027 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4029 /* skip to avoid loading attributes.pm */
4031 deprecate(":unique");
4034 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4037 /* NOTE: any CV attrs applied here need to be part of
4038 the CVf_BUILTIN_ATTRS define in cv.h! */
4039 else if (!PL_in_my && len == 6 && strnEQ(s, "lvalue", len))
4040 CvLVALUE_on(PL_compcv);
4041 else if (!PL_in_my && len == 6 && strnEQ(s, "locked", len))
4042 CvLOCKED_on(PL_compcv);
4043 else if (!PL_in_my && len == 6 && strnEQ(s, "method", len))
4044 CvMETHOD_on(PL_compcv);
4045 else if (!PL_in_my && len == 9 && strnEQ(s, "assertion", len))
4046 CvASSERTION_on(PL_compcv);
4047 /* After we've set the flags, it could be argued that
4048 we don't need to do the attributes.pm-based setting
4049 process, and shouldn't bother appending recognized
4050 flags. To experiment with that, uncomment the
4051 following "else". (Note that's already been
4052 uncommented. That keeps the above-applied built-in
4053 attributes from being intercepted (and possibly
4054 rejected) by a package's attribute routines, but is
4055 justified by the performance win for the common case
4056 of applying only built-in attributes.) */
4058 attrs = append_elem(OP_LIST, attrs,
4059 newSVOP(OP_CONST, 0,
4063 if (*s == ':' && s[1] != ':')
4066 break; /* require real whitespace or :'s */
4067 /* XXX losing whitespace on sequential attributes here */
4071 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4072 if (*s != ';' && *s != '}' && *s != tmp
4073 && (tmp != '=' || *s != ')')) {
4074 const char q = ((*s == '\'') ? '"' : '\'');
4075 /* If here for an expression, and parsed no attrs, back
4077 if (tmp == '=' && !attrs) {
4081 /* MUST advance bufptr here to avoid bogus "at end of line"
4082 context messages from yyerror().
4086 ? Perl_form(aTHX_ "Invalid separator character "
4087 "%c%c%c in attribute list", q, *s, q)
4088 : "Unterminated attribute list" );
4096 start_force(PL_curforce);
4097 NEXTVAL_NEXTTOKE.opval = attrs;
4098 CURMAD('_', PL_nextwhite);
4103 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4104 (s - SvPVX(PL_linestr)) - stuffstart);
4112 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4113 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4121 const char tmp = *s++;
4126 const char tmp = *s++;
4134 if (PL_lex_brackets <= 0)
4135 yyerror("Unmatched right square bracket");
4138 if (PL_lex_state == LEX_INTERPNORMAL) {
4139 if (PL_lex_brackets == 0) {
4140 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4141 PL_lex_state = LEX_INTERPEND;
4148 if (PL_lex_brackets > 100) {
4149 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4151 switch (PL_expect) {
4153 if (PL_lex_formbrack) {
4157 if (PL_oldoldbufptr == PL_last_lop)
4158 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4160 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4161 OPERATOR(HASHBRACK);
4163 while (s < PL_bufend && SPACE_OR_TAB(*s))
4166 PL_tokenbuf[0] = '\0';
4167 if (d < PL_bufend && *d == '-') {
4168 PL_tokenbuf[0] = '-';
4170 while (d < PL_bufend && SPACE_OR_TAB(*d))
4173 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4174 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4176 while (d < PL_bufend && SPACE_OR_TAB(*d))
4179 const char minus = (PL_tokenbuf[0] == '-');
4180 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4188 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4193 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4198 if (PL_oldoldbufptr == PL_last_lop)
4199 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4201 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4204 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4206 /* This hack is to get the ${} in the message. */
4208 yyerror("syntax error");
4211 OPERATOR(HASHBRACK);
4213 /* This hack serves to disambiguate a pair of curlies
4214 * as being a block or an anon hash. Normally, expectation
4215 * determines that, but in cases where we're not in a
4216 * position to expect anything in particular (like inside
4217 * eval"") we have to resolve the ambiguity. This code
4218 * covers the case where the first term in the curlies is a
4219 * quoted string. Most other cases need to be explicitly
4220 * disambiguated by prepending a "+" before the opening
4221 * curly in order to force resolution as an anon hash.
4223 * XXX should probably propagate the outer expectation
4224 * into eval"" to rely less on this hack, but that could
4225 * potentially break current behavior of eval"".
4229 if (*s == '\'' || *s == '"' || *s == '`') {
4230 /* common case: get past first string, handling escapes */
4231 for (t++; t < PL_bufend && *t != *s;)
4232 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4236 else if (*s == 'q') {
4239 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4242 /* skip q//-like construct */
4244 char open, close, term;
4247 while (t < PL_bufend && isSPACE(*t))
4249 /* check for q => */
4250 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4251 OPERATOR(HASHBRACK);
4255 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4259 for (t++; t < PL_bufend; t++) {
4260 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4262 else if (*t == open)
4266 for (t++; t < PL_bufend; t++) {
4267 if (*t == '\\' && t+1 < PL_bufend)
4269 else if (*t == close && --brackets <= 0)
4271 else if (*t == open)
4278 /* skip plain q word */
4279 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4282 else if (isALNUM_lazy_if(t,UTF)) {
4284 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4287 while (t < PL_bufend && isSPACE(*t))
4289 /* if comma follows first term, call it an anon hash */
4290 /* XXX it could be a comma expression with loop modifiers */
4291 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4292 || (*t == '=' && t[1] == '>')))
4293 OPERATOR(HASHBRACK);
4294 if (PL_expect == XREF)
4297 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4303 yylval.ival = CopLINE(PL_curcop);
4304 if (isSPACE(*s) || *s == '#')
4305 PL_copline = NOLINE; /* invalidate current command line number */
4310 if (PL_lex_brackets <= 0)
4311 yyerror("Unmatched right curly bracket");
4313 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4314 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4315 PL_lex_formbrack = 0;
4316 if (PL_lex_state == LEX_INTERPNORMAL) {
4317 if (PL_lex_brackets == 0) {
4318 if (PL_expect & XFAKEBRACK) {
4319 PL_expect &= XENUMMASK;
4320 PL_lex_state = LEX_INTERPEND;
4325 PL_thiswhite = newSVpvn("",0);
4326 sv_catpvn(PL_thiswhite,"}",1);
4329 return yylex(); /* ignore fake brackets */
4331 if (*s == '-' && s[1] == '>')
4332 PL_lex_state = LEX_INTERPENDMAYBE;
4333 else if (*s != '[' && *s != '{')
4334 PL_lex_state = LEX_INTERPEND;
4337 if (PL_expect & XFAKEBRACK) {
4338 PL_expect &= XENUMMASK;
4340 return yylex(); /* ignore fake brackets */
4342 start_force(PL_curforce);
4344 curmad('X', newSVpvn(s-1,1));
4345 CURMAD('_', PL_thiswhite);
4350 PL_thistoken = newSVpvn("",0);
4358 if (PL_expect == XOPERATOR) {
4359 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4360 && isIDFIRST_lazy_if(s,UTF))
4362 CopLINE_dec(PL_curcop);
4363 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4364 CopLINE_inc(PL_curcop);
4369 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4371 PL_expect = XOPERATOR;
4372 force_ident(PL_tokenbuf, '&');
4376 yylval.ival = (OPpENTERSUB_AMPER<<8);
4388 const char tmp = *s++;
4395 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4396 && strchr("+-*/%.^&|<",tmp))