3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yychar (*PL_yycharp)
27 #define yylval (*PL_yylvalp)
29 static const char ident_too_long[] = "Identifier too long";
30 static const char commaless_variable_list[] = "comma-less variable list";
32 static void restore_rsfp(pTHX_ void *f);
33 #ifndef PERL_NO_UTF16_FILTER
34 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
35 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
39 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
40 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
42 # define CURMAD(slot,sv)
43 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
46 #define XFAKEBRACK 128
49 #ifdef USE_UTF8_SCRIPTS
50 # define UTF (!IN_BYTES)
52 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
55 /* In variables named $^X, these are the legal values for X.
56 * 1999-02-27 mjd-perl-patch@plover.com */
57 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
59 /* On MacOS, respect nonbreaking spaces */
60 #ifdef MACOS_TRADITIONAL
61 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
63 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
66 /* LEX_* are values for PL_lex_state, the state of the lexer.
67 * They are arranged oddly so that the guard on the switch statement
68 * can get by with a single comparison (if the compiler is smart enough).
71 /* #define LEX_NOTPARSING 11 is done in perl.h. */
73 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
74 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
75 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
76 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
77 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
79 /* at end of code, eg "$x" followed by: */
80 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
81 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
83 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
84 string or after \E, $foo, etc */
85 #define LEX_INTERPCONST 2 /* NOT USED */
86 #define LEX_FORMLINE 1 /* expecting a format line */
87 #define LEX_KNOWNEXT 0 /* next token known; just return it */
91 static const char* const lex_state_names[] = {
110 #include "keywords.h"
112 /* CLINE is a macro that ensures PL_copline has a sane value */
117 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
120 # define SKIPSPACE0(s) skipspace0(s)
121 # define SKIPSPACE1(s) skipspace1(s)
122 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
123 # define PEEKSPACE(s) skipspace2(s,0)
125 # define SKIPSPACE0(s) skipspace(s)
126 # define SKIPSPACE1(s) skipspace(s)
127 # define SKIPSPACE2(s,tsv) skipspace(s)
128 # define PEEKSPACE(s) skipspace(s)
132 * Convenience functions to return different tokens and prime the
133 * lexer for the next token. They all take an argument.
135 * TOKEN : generic token (used for '(', DOLSHARP, etc)
136 * OPERATOR : generic operator
137 * AOPERATOR : assignment operator
138 * PREBLOCK : beginning the block after an if, while, foreach, ...
139 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
140 * PREREF : *EXPR where EXPR is not a simple identifier
141 * TERM : expression term
142 * LOOPX : loop exiting command (goto, last, dump, etc)
143 * FTST : file test operator
144 * FUN0 : zero-argument function
145 * FUN1 : not used, except for not, which isn't a UNIOP
146 * BOop : bitwise or or xor
148 * SHop : shift operator
149 * PWop : power operator
150 * PMop : pattern-matching operator
151 * Aop : addition-level operator
152 * Mop : multiplication-level operator
153 * Eop : equality-testing operator
154 * Rop : relational operator <= != gt
156 * Also see LOP and lop() below.
159 #ifdef DEBUGGING /* Serve -DT. */
160 # define REPORT(retval) tokereport((I32)retval)
162 # define REPORT(retval) (retval)
165 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
166 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
167 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
168 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
169 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
170 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
171 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
172 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
173 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
174 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
175 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
176 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
177 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
178 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
179 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
180 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
181 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
182 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
183 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
184 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
186 /* This bit of chicanery makes a unary function followed by
187 * a parenthesis into a function with one argument, highest precedence.
188 * The UNIDOR macro is for unary functions that can be followed by the //
189 * operator (such as C<shift // 0>).
191 #define UNI2(f,x) { \
195 PL_last_uni = PL_oldbufptr; \
196 PL_last_lop_op = f; \
198 return REPORT( (int)FUNC1 ); \
200 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
202 #define UNI(f) UNI2(f,XTERM)
203 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
205 #define UNIBRACK(f) { \
208 PL_last_uni = PL_oldbufptr; \
210 return REPORT( (int)FUNC1 ); \
212 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
215 /* grandfather return to old style */
216 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
220 /* how to interpret the yylval associated with the token */
224 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
230 static struct debug_tokens {
232 enum token_type type;
234 } const debug_tokens[] =
236 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
237 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
238 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
239 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
240 { ARROW, TOKENTYPE_NONE, "ARROW" },
241 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
242 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
243 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
244 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
245 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
246 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
247 { DO, TOKENTYPE_NONE, "DO" },
248 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
249 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
250 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
251 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
252 { ELSE, TOKENTYPE_NONE, "ELSE" },
253 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
254 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
255 { FOR, TOKENTYPE_IVAL, "FOR" },
256 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
257 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
258 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
259 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
260 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
261 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
262 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
263 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
264 { IF, TOKENTYPE_IVAL, "IF" },
265 { LABEL, TOKENTYPE_PVAL, "LABEL" },
266 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
267 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
268 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
269 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
270 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
271 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
272 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
273 { MY, TOKENTYPE_IVAL, "MY" },
274 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
275 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
276 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
277 { OROP, TOKENTYPE_IVAL, "OROP" },
278 { OROR, TOKENTYPE_NONE, "OROR" },
279 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
280 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
281 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
282 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
283 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
284 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
285 { PREINC, TOKENTYPE_NONE, "PREINC" },
286 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
287 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
288 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
289 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
290 { SUB, TOKENTYPE_NONE, "SUB" },
291 { THING, TOKENTYPE_OPVAL, "THING" },
292 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
293 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
294 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
295 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
296 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
297 { USE, TOKENTYPE_IVAL, "USE" },
298 { WHEN, TOKENTYPE_IVAL, "WHEN" },
299 { WHILE, TOKENTYPE_IVAL, "WHILE" },
300 { WORD, TOKENTYPE_OPVAL, "WORD" },
301 { 0, TOKENTYPE_NONE, NULL }
304 /* dump the returned token in rv, plus any optional arg in yylval */
307 S_tokereport(pTHX_ I32 rv)
311 const char *name = NULL;
312 enum token_type type = TOKENTYPE_NONE;
313 const struct debug_tokens *p;
314 SV* const report = newSVpvs("<== ");
316 for (p = debug_tokens; p->token; p++) {
317 if (p->token == (int)rv) {
324 Perl_sv_catpv(aTHX_ report, name);
325 else if ((char)rv > ' ' && (char)rv < '~')
326 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
328 sv_catpvs(report, "EOF");
330 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
333 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
336 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
338 case TOKENTYPE_OPNUM:
339 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
340 PL_op_name[yylval.ival]);
343 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
345 case TOKENTYPE_OPVAL:
347 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
348 PL_op_name[yylval.opval->op_type]);
349 if (yylval.opval->op_type == OP_CONST) {
350 Perl_sv_catpvf(aTHX_ report, " %s",
351 SvPEEK(cSVOPx_sv(yylval.opval)));
356 sv_catpvs(report, "(opval=null)");
359 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
365 /* print the buffer with suitable escapes */
368 S_printbuf(pTHX_ const char* fmt, const char* s)
370 SV* const tmp = newSVpvs("");
371 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
380 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
381 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
385 S_ao(pTHX_ int toketype)
388 if (*PL_bufptr == '=') {
390 if (toketype == ANDAND)
391 yylval.ival = OP_ANDASSIGN;
392 else if (toketype == OROR)
393 yylval.ival = OP_ORASSIGN;
394 else if (toketype == DORDOR)
395 yylval.ival = OP_DORASSIGN;
403 * When Perl expects an operator and finds something else, no_op
404 * prints the warning. It always prints "<something> found where
405 * operator expected. It prints "Missing semicolon on previous line?"
406 * if the surprise occurs at the start of the line. "do you need to
407 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
408 * where the compiler doesn't know if foo is a method call or a function.
409 * It prints "Missing operator before end of line" if there's nothing
410 * after the missing operator, or "... before <...>" if there is something
411 * after the missing operator.
415 S_no_op(pTHX_ const char *what, char *s)
418 char * const oldbp = PL_bufptr;
419 const bool is_first = (PL_oldbufptr == PL_linestart);
425 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
426 if (ckWARN_d(WARN_SYNTAX)) {
428 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
429 "\t(Missing semicolon on previous line?)\n");
430 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
432 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
434 if (t < PL_bufptr && isSPACE(*t))
435 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
436 "\t(Do you need to predeclare %.*s?)\n",
437 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
441 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
442 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
450 * Complain about missing quote/regexp/heredoc terminator.
451 * If it's called with NULL then it cauterizes the line buffer.
452 * If we're in a delimited string and the delimiter is a control
453 * character, it's reformatted into a two-char sequence like ^C.
458 S_missingterm(pTHX_ char *s)
464 char * const nl = strrchr(s,'\n');
470 iscntrl(PL_multi_close)
472 PL_multi_close < 32 || PL_multi_close == 127
476 tmpbuf[1] = (char)toCTRL(PL_multi_close);
481 *tmpbuf = (char)PL_multi_close;
485 q = strchr(s,'"') ? '\'' : '"';
486 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
489 #define FEATURE_IS_ENABLED(name) \
490 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
491 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
493 * S_feature_is_enabled
494 * Check whether the named feature is enabled.
497 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
500 HV * const hinthv = GvHV(PL_hintgv);
501 char he_name[32] = "feature_";
502 (void) strncpy(&he_name[8], name, 24);
504 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
512 Perl_deprecate(pTHX_ const char *s)
514 if (ckWARN(WARN_DEPRECATED))
515 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
519 Perl_deprecate_old(pTHX_ const char *s)
521 /* This function should NOT be called for any new deprecated warnings */
522 /* Use Perl_deprecate instead */
524 /* It is here to maintain backward compatibility with the pre-5.8 */
525 /* warnings category hierarchy. The "deprecated" category used to */
526 /* live under the "syntax" category. It is now a top-level category */
527 /* in its own right. */
529 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
530 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
531 "Use of %s is deprecated", s);
535 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
536 * utf16-to-utf8-reversed.
539 #ifdef PERL_CR_FILTER
543 register const char *s = SvPVX_const(sv);
544 register const char * const e = s + SvCUR(sv);
545 /* outer loop optimized to do nothing if there are no CR-LFs */
547 if (*s++ == '\r' && *s == '\n') {
548 /* hit a CR-LF, need to copy the rest */
549 register char *d = s - 1;
552 if (*s == '\r' && s[1] == '\n')
563 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
565 const I32 count = FILTER_READ(idx+1, sv, maxlen);
566 if (count > 0 && !maxlen)
574 * Initialize variables. Uses the Perl save_stack to save its state (for
575 * recursive calls to the parser).
579 Perl_lex_start(pTHX_ SV *line)
585 SAVEI32(PL_lex_dojoin);
586 SAVEI32(PL_lex_brackets);
587 SAVEI32(PL_lex_casemods);
588 SAVEI32(PL_lex_starts);
589 SAVEI32(PL_lex_state);
590 SAVEVPTR(PL_lex_inpat);
591 SAVEI32(PL_lex_inwhat);
593 if (PL_lex_state == LEX_KNOWNEXT) {
594 I32 toke = PL_lasttoke;
595 while (--toke >= 0) {
596 SAVEI32(PL_nexttoke[toke].next_type);
597 SAVEVPTR(PL_nexttoke[toke].next_val);
599 SAVEVPTR(PL_nexttoke[toke].next_mad);
601 SAVEI32(PL_lasttoke);
604 SAVESPTR(PL_thistoken);
605 SAVESPTR(PL_thiswhite);
606 SAVESPTR(PL_nextwhite);
607 SAVESPTR(PL_thisopen);
608 SAVESPTR(PL_thisclose);
609 SAVESPTR(PL_thisstuff);
610 SAVEVPTR(PL_thismad);
611 SAVEI32(PL_realtokenstart);
612 SAVEI32(PL_faketokens);
614 SAVEI32(PL_curforce);
616 if (PL_lex_state == LEX_KNOWNEXT) {
617 I32 toke = PL_nexttoke;
618 while (--toke >= 0) {
619 SAVEI32(PL_nexttype[toke]);
620 SAVEVPTR(PL_nextval[toke]);
622 SAVEI32(PL_nexttoke);
625 SAVECOPLINE(PL_curcop);
628 SAVEPPTR(PL_oldbufptr);
629 SAVEPPTR(PL_oldoldbufptr);
630 SAVEPPTR(PL_last_lop);
631 SAVEPPTR(PL_last_uni);
632 SAVEPPTR(PL_linestart);
633 SAVESPTR(PL_linestr);
634 SAVEGENERICPV(PL_lex_brackstack);
635 SAVEGENERICPV(PL_lex_casestack);
636 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
637 SAVESPTR(PL_lex_stuff);
638 SAVEI32(PL_lex_defer);
639 SAVEI32(PL_sublex_info.sub_inwhat);
640 SAVESPTR(PL_lex_repl);
642 SAVEINT(PL_lex_expect);
644 PL_lex_state = LEX_NORMAL;
648 Newx(PL_lex_brackstack, 120, char);
649 Newx(PL_lex_casestack, 12, char);
651 *PL_lex_casestack = '\0';
663 PL_sublex_info.sub_inwhat = 0;
665 if (SvREADONLY(PL_linestr))
666 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
667 s = SvPV_const(PL_linestr, len);
668 if (!len || s[len-1] != ';') {
669 if (!(SvFLAGS(PL_linestr) & SVs_TEMP))
670 PL_linestr = sv_2mortal(newSVsv(PL_linestr));
671 sv_catpvs(PL_linestr, "\n;");
673 SvTEMP_off(PL_linestr);
674 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
675 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
676 PL_last_lop = PL_last_uni = NULL;
682 * Finalizer for lexing operations. Must be called when the parser is
683 * done with the lexer.
690 PL_doextract = FALSE;
695 * This subroutine has nothing to do with tilting, whether at windmills
696 * or pinball tables. Its name is short for "increment line". It
697 * increments the current line number in CopLINE(PL_curcop) and checks
698 * to see whether the line starts with a comment of the form
699 * # line 500 "foo.pm"
700 * If so, it sets the current line number and file to the values in the comment.
704 S_incline(pTHX_ char *s)
712 CopLINE_inc(PL_curcop);
715 while (SPACE_OR_TAB(*s))
717 if (strnEQ(s, "line", 4))
721 if (SPACE_OR_TAB(*s))
725 while (SPACE_OR_TAB(*s))
733 while (SPACE_OR_TAB(*s))
735 if (*s == '"' && (t = strchr(s+1, '"'))) {
745 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
747 if (*e != '\n' && *e != '\0')
748 return; /* false alarm */
754 const char * const cf = CopFILE(PL_curcop);
755 STRLEN tmplen = cf ? strlen(cf) : 0;
756 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
757 /* must copy *{"::_<(eval N)[oldfilename:L]"}
758 * to *{"::_<newfilename"} */
759 char smallbuf[256], smallbuf2[256];
760 char *tmpbuf, *tmpbuf2;
762 STRLEN tmplen2 = strlen(s);
763 if (tmplen + 3 < sizeof smallbuf)
766 Newx(tmpbuf, tmplen + 3, char);
767 if (tmplen2 + 3 < sizeof smallbuf2)
770 Newx(tmpbuf2, tmplen2 + 3, char);
771 tmpbuf[0] = tmpbuf2[0] = '_';
772 tmpbuf[1] = tmpbuf2[1] = '<';
773 memcpy(tmpbuf + 2, cf, ++tmplen);
774 memcpy(tmpbuf2 + 2, s, ++tmplen2);
776 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
778 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
780 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
781 /* adjust ${"::_<newfilename"} to store the new file name */
782 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
783 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
784 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
786 if (tmpbuf != smallbuf) Safefree(tmpbuf);
787 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
790 CopFILE_free(PL_curcop);
791 CopFILE_set(PL_curcop, s);
794 CopLINE_set(PL_curcop, atoi(n)-1);
798 /* skip space before PL_thistoken */
801 S_skipspace0(pTHX_ register char *s)
808 PL_thiswhite = newSVpvn("",0);
809 sv_catsv(PL_thiswhite, PL_skipwhite);
810 sv_free(PL_skipwhite);
813 PL_realtokenstart = s - SvPVX(PL_linestr);
817 /* skip space after PL_thistoken */
820 S_skipspace1(pTHX_ register char *s)
822 const char *start = s;
823 I32 startoff = start - SvPVX(PL_linestr);
828 start = SvPVX(PL_linestr) + startoff;
829 if (!PL_thistoken && PL_realtokenstart >= 0) {
830 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
831 PL_thistoken = newSVpvn(tstart, start - tstart);
833 PL_realtokenstart = -1;
836 PL_nextwhite = newSVpvn("",0);
837 sv_catsv(PL_nextwhite, PL_skipwhite);
838 sv_free(PL_skipwhite);
845 S_skipspace2(pTHX_ register char *s, SV **svp)
848 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
849 const I32 startoff = s - SvPVX(PL_linestr);
852 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
853 if (!PL_madskills || !svp)
855 start = SvPVX(PL_linestr) + startoff;
856 if (!PL_thistoken && PL_realtokenstart >= 0) {
857 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
858 PL_thistoken = newSVpvn(tstart, start - tstart);
859 PL_realtokenstart = -1;
863 *svp = newSVpvn("",0);
864 sv_setsv(*svp, PL_skipwhite);
865 sv_free(PL_skipwhite);
875 * Called to gobble the appropriate amount and type of whitespace.
876 * Skips comments as well.
880 S_skipspace(pTHX_ register char *s)
885 int startoff = s - SvPVX(PL_linestr);
888 sv_free(PL_skipwhite);
893 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
894 while (s < PL_bufend && SPACE_OR_TAB(*s))
904 SSize_t oldprevlen, oldoldprevlen;
905 SSize_t oldloplen = 0, oldunilen = 0;
906 while (s < PL_bufend && isSPACE(*s)) {
907 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
912 if (s < PL_bufend && *s == '#') {
913 while (s < PL_bufend && *s != '\n')
917 if (PL_in_eval && !PL_rsfp) {
924 /* only continue to recharge the buffer if we're at the end
925 * of the buffer, we're not reading from a source filter, and
926 * we're in normal lexing mode
928 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
929 PL_lex_state == LEX_FORMLINE)
936 /* try to recharge the buffer */
938 curoff = s - SvPVX(PL_linestr);
941 if ((s = filter_gets(PL_linestr, PL_rsfp,
942 (prevlen = SvCUR(PL_linestr)))) == NULL)
945 if (PL_madskills && curoff != startoff) {
947 PL_skipwhite = newSVpvn("",0);
948 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
952 /* mustn't throw out old stuff yet if madpropping */
953 SvCUR(PL_linestr) = curoff;
954 s = SvPVX(PL_linestr) + curoff;
956 if (curoff && s[-1] == '\n')
960 /* end of file. Add on the -p or -n magic */
961 /* XXX these shouldn't really be added here, can't set PL_faketokens */
965 ";}continue{print or die qq(-p destination: $!\\n);}");
968 ";}continue{print or die qq(-p destination: $!\\n);}");
970 PL_minus_n = PL_minus_p = 0;
972 else if (PL_minus_n) {
974 sv_catpvn(PL_linestr, ";}", 2);
976 sv_setpvn(PL_linestr, ";}", 2);
982 sv_catpvn(PL_linestr,";", 1);
984 sv_setpvn(PL_linestr,";", 1);
987 /* reset variables for next time we lex */
988 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
994 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
995 PL_last_lop = PL_last_uni = NULL;
997 /* Close the filehandle. Could be from -P preprocessor,
998 * STDIN, or a regular file. If we were reading code from
999 * STDIN (because the commandline held no -e or filename)
1000 * then we don't close it, we reset it so the code can
1001 * read from STDIN too.
1004 if (PL_preprocess && !PL_in_eval)
1005 (void)PerlProc_pclose(PL_rsfp);
1006 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1007 PerlIO_clearerr(PL_rsfp);
1009 (void)PerlIO_close(PL_rsfp);
1014 /* not at end of file, so we only read another line */
1015 /* make corresponding updates to old pointers, for yyerror() */
1016 oldprevlen = PL_oldbufptr - PL_bufend;
1017 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1019 oldunilen = PL_last_uni - PL_bufend;
1021 oldloplen = PL_last_lop - PL_bufend;
1022 PL_linestart = PL_bufptr = s + prevlen;
1023 PL_bufend = s + SvCUR(PL_linestr);
1025 PL_oldbufptr = s + oldprevlen;
1026 PL_oldoldbufptr = s + oldoldprevlen;
1028 PL_last_uni = s + oldunilen;
1030 PL_last_lop = s + oldloplen;
1033 /* debugger active and we're not compiling the debugger code,
1034 * so store the line into the debugger's array of lines
1036 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1037 SV * const sv = newSV(0);
1039 sv_upgrade(sv, SVt_PVMG);
1040 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1043 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1051 PL_skipwhite = newSVpvn("",0);
1052 curoff = s - SvPVX(PL_linestr);
1053 if (curoff - startoff)
1054 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1063 * Check the unary operators to ensure there's no ambiguity in how they're
1064 * used. An ambiguous piece of code would be:
1066 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1067 * the +5 is its argument.
1077 if (PL_oldoldbufptr != PL_last_uni)
1079 while (isSPACE(*PL_last_uni))
1082 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1084 if ((t = strchr(s, '(')) && t < PL_bufptr)
1087 if (ckWARN_d(WARN_AMBIGUOUS)){
1088 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1089 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1090 (int)(s - PL_last_uni), PL_last_uni);
1095 * LOP : macro to build a list operator. Its behaviour has been replaced
1096 * with a subroutine, S_lop() for which LOP is just another name.
1099 #define LOP(f,x) return lop(f,x,s)
1103 * Build a list operator (or something that might be one). The rules:
1104 * - if we have a next token, then it's a list operator [why?]
1105 * - if the next thing is an opening paren, then it's a function
1106 * - else it's a list operator
1110 S_lop(pTHX_ I32 f, int x, char *s)
1117 PL_last_lop = PL_oldbufptr;
1118 PL_last_lop_op = (OPCODE)f;
1121 return REPORT(LSTOP);
1124 return REPORT(LSTOP);
1127 return REPORT(FUNC);
1130 return REPORT(FUNC);
1132 return REPORT(LSTOP);
1138 * Sets up for an eventual force_next(). start_force(0) basically does
1139 * an unshift, while start_force(-1) does a push. yylex removes items
1144 S_start_force(pTHX_ int where)
1148 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1149 where = PL_lasttoke;
1150 assert(PL_curforce < 0 || PL_curforce == where);
1151 if (PL_curforce != where) {
1152 for (i = PL_lasttoke; i > where; --i) {
1153 PL_nexttoke[i] = PL_nexttoke[i-1];
1157 if (PL_curforce < 0) /* in case of duplicate start_force() */
1158 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1159 PL_curforce = where;
1162 curmad('^', newSVpvn("",0));
1163 CURMAD('_', PL_nextwhite);
1168 S_curmad(pTHX_ char slot, SV *sv)
1174 if (PL_curforce < 0)
1175 where = &PL_thismad;
1177 where = &PL_nexttoke[PL_curforce].next_mad;
1180 sv_setpvn(sv, "", 0);
1183 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1185 else if (PL_encoding) {
1186 sv_recode_to_utf8(sv, PL_encoding);
1191 /* keep a slot open for the head of the list? */
1192 if (slot != '_' && *where && (*where)->mad_key == '^') {
1193 (*where)->mad_key = slot;
1194 sv_free((*where)->mad_val);
1195 (*where)->mad_val = (void*)sv;
1198 addmad(newMADsv(slot, sv), where, 0);
1201 # define start_force(where) NOOP
1202 # define curmad(slot, sv) NOOP
1207 * When the lexer realizes it knows the next token (for instance,
1208 * it is reordering tokens for the parser) then it can call S_force_next
1209 * to know what token to return the next time the lexer is called. Caller
1210 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1211 * and possibly PL_expect to ensure the lexer handles the token correctly.
1215 S_force_next(pTHX_ I32 type)
1219 if (PL_curforce < 0)
1220 start_force(PL_lasttoke);
1221 PL_nexttoke[PL_curforce].next_type = type;
1222 if (PL_lex_state != LEX_KNOWNEXT)
1223 PL_lex_defer = PL_lex_state;
1224 PL_lex_state = LEX_KNOWNEXT;
1225 PL_lex_expect = PL_expect;
1228 PL_nexttype[PL_nexttoke] = type;
1230 if (PL_lex_state != LEX_KNOWNEXT) {
1231 PL_lex_defer = PL_lex_state;
1232 PL_lex_expect = PL_expect;
1233 PL_lex_state = LEX_KNOWNEXT;
1239 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1242 SV * const sv = newSVpvn(start,len);
1243 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1250 * When the lexer knows the next thing is a word (for instance, it has
1251 * just seen -> and it knows that the next char is a word char, then
1252 * it calls S_force_word to stick the next word into the PL_next lookahead.
1255 * char *start : buffer position (must be within PL_linestr)
1256 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1257 * int check_keyword : if true, Perl checks to make sure the word isn't
1258 * a keyword (do this if the word is a label, e.g. goto FOO)
1259 * int allow_pack : if true, : characters will also be allowed (require,
1260 * use, etc. do this)
1261 * int allow_initial_tick : used by the "sub" lexer only.
1265 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1271 start = SKIPSPACE1(start);
1273 if (isIDFIRST_lazy_if(s,UTF) ||
1274 (allow_pack && *s == ':') ||
1275 (allow_initial_tick && *s == '\'') )
1277 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1278 if (check_keyword && keyword(PL_tokenbuf, len))
1280 start_force(PL_curforce);
1282 curmad('X', newSVpvn(start,s-start));
1283 if (token == METHOD) {
1288 PL_expect = XOPERATOR;
1291 NEXTVAL_NEXTTOKE.opval
1292 = (OP*)newSVOP(OP_CONST,0,
1293 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1294 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1302 * Called when the lexer wants $foo *foo &foo etc, but the program
1303 * text only contains the "foo" portion. The first argument is a pointer
1304 * to the "foo", and the second argument is the type symbol to prefix.
1305 * Forces the next token to be a "WORD".
1306 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1310 S_force_ident(pTHX_ register const char *s, int kind)
1314 const STRLEN len = strlen(s);
1315 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1316 start_force(PL_curforce);
1317 NEXTVAL_NEXTTOKE.opval = o;
1320 o->op_private = OPpCONST_ENTERED;
1321 /* XXX see note in pp_entereval() for why we forgo typo
1322 warnings if the symbol must be introduced in an eval.
1324 gv_fetchpvn_flags(s, len,
1325 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1327 kind == '$' ? SVt_PV :
1328 kind == '@' ? SVt_PVAV :
1329 kind == '%' ? SVt_PVHV :
1337 Perl_str_to_version(pTHX_ SV *sv)
1342 const char *start = SvPV_const(sv,len);
1343 const char * const end = start + len;
1344 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1345 while (start < end) {
1349 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1354 retval += ((NV)n)/nshift;
1363 * Forces the next token to be a version number.
1364 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1365 * and if "guessing" is TRUE, then no new token is created (and the caller
1366 * must use an alternative parsing method).
1370 S_force_version(pTHX_ char *s, int guessing)
1376 I32 startoff = s - SvPVX(PL_linestr);
1385 while (isDIGIT(*d) || *d == '_' || *d == '.')
1389 start_force(PL_curforce);
1390 curmad('X', newSVpvn(s,d-s));
1393 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1395 s = scan_num(s, &yylval);
1396 version = yylval.opval;
1397 ver = cSVOPx(version)->op_sv;
1398 if (SvPOK(ver) && !SvNIOK(ver)) {
1399 SvUPGRADE(ver, SVt_PVNV);
1400 SvNV_set(ver, str_to_version(ver));
1401 SvNOK_on(ver); /* hint that it is a version */
1404 else if (guessing) {
1407 sv_free(PL_nextwhite); /* let next token collect whitespace */
1409 s = SvPVX(PL_linestr) + startoff;
1417 if (PL_madskills && !version) {
1418 sv_free(PL_nextwhite); /* let next token collect whitespace */
1420 s = SvPVX(PL_linestr) + startoff;
1423 /* NOTE: The parser sees the package name and the VERSION swapped */
1424 start_force(PL_curforce);
1425 NEXTVAL_NEXTTOKE.opval = version;
1433 * Tokenize a quoted string passed in as an SV. It finds the next
1434 * chunk, up to end of string or a backslash. It may make a new
1435 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1440 S_tokeq(pTHX_ SV *sv)
1444 register char *send;
1452 s = SvPV_force(sv, len);
1453 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1456 while (s < send && *s != '\\')
1461 if ( PL_hints & HINT_NEW_STRING ) {
1462 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1468 if (s + 1 < send && (s[1] == '\\'))
1469 s++; /* all that, just for this */
1474 SvCUR_set(sv, d - SvPVX_const(sv));
1476 if ( PL_hints & HINT_NEW_STRING )
1477 return new_constant(NULL, 0, "q", sv, pv, "q");
1482 * Now come three functions related to double-quote context,
1483 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1484 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1485 * interact with PL_lex_state, and create fake ( ... ) argument lists
1486 * to handle functions and concatenation.
1487 * They assume that whoever calls them will be setting up a fake
1488 * join call, because each subthing puts a ',' after it. This lets
1491 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1493 * (I'm not sure whether the spurious commas at the end of lcfirst's
1494 * arguments and join's arguments are created or not).
1499 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1501 * Pattern matching will set PL_lex_op to the pattern-matching op to
1502 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1504 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1506 * Everything else becomes a FUNC.
1508 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1509 * had an OP_CONST or OP_READLINE). This just sets us up for a
1510 * call to S_sublex_push().
1514 S_sublex_start(pTHX)
1517 register const I32 op_type = yylval.ival;
1519 if (op_type == OP_NULL) {
1520 yylval.opval = PL_lex_op;
1524 if (op_type == OP_CONST || op_type == OP_READLINE) {
1525 SV *sv = tokeq(PL_lex_stuff);
1527 if (SvTYPE(sv) == SVt_PVIV) {
1528 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1530 const char * const p = SvPV_const(sv, len);
1531 SV * const nsv = newSVpvn(p, len);
1537 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1538 PL_lex_stuff = NULL;
1539 /* Allow <FH> // "foo" */
1540 if (op_type == OP_READLINE)
1541 PL_expect = XTERMORDORDOR;
1545 PL_sublex_info.super_state = PL_lex_state;
1546 PL_sublex_info.sub_inwhat = op_type;
1547 PL_sublex_info.sub_op = PL_lex_op;
1548 PL_lex_state = LEX_INTERPPUSH;
1552 yylval.opval = PL_lex_op;
1562 * Create a new scope to save the lexing state. The scope will be
1563 * ended in S_sublex_done. Returns a '(', starting the function arguments
1564 * to the uc, lc, etc. found before.
1565 * Sets PL_lex_state to LEX_INTERPCONCAT.
1574 PL_lex_state = PL_sublex_info.super_state;
1575 SAVEI32(PL_lex_dojoin);
1576 SAVEI32(PL_lex_brackets);
1577 SAVEI32(PL_lex_casemods);
1578 SAVEI32(PL_lex_starts);
1579 SAVEI32(PL_lex_state);
1580 SAVEVPTR(PL_lex_inpat);
1581 SAVEI32(PL_lex_inwhat);
1582 SAVECOPLINE(PL_curcop);
1583 SAVEPPTR(PL_bufptr);
1584 SAVEPPTR(PL_bufend);
1585 SAVEPPTR(PL_oldbufptr);
1586 SAVEPPTR(PL_oldoldbufptr);
1587 SAVEPPTR(PL_last_lop);
1588 SAVEPPTR(PL_last_uni);
1589 SAVEPPTR(PL_linestart);
1590 SAVESPTR(PL_linestr);
1591 SAVEGENERICPV(PL_lex_brackstack);
1592 SAVEGENERICPV(PL_lex_casestack);
1594 PL_linestr = PL_lex_stuff;
1595 PL_lex_stuff = NULL;
1597 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1598 = SvPVX(PL_linestr);
1599 PL_bufend += SvCUR(PL_linestr);
1600 PL_last_lop = PL_last_uni = NULL;
1601 SAVEFREESV(PL_linestr);
1603 PL_lex_dojoin = FALSE;
1604 PL_lex_brackets = 0;
1605 Newx(PL_lex_brackstack, 120, char);
1606 Newx(PL_lex_casestack, 12, char);
1607 PL_lex_casemods = 0;
1608 *PL_lex_casestack = '\0';
1610 PL_lex_state = LEX_INTERPCONCAT;
1611 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1613 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1614 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1615 PL_lex_inpat = PL_sublex_info.sub_op;
1617 PL_lex_inpat = NULL;
1624 * Restores lexer state after a S_sublex_push.
1631 if (!PL_lex_starts++) {
1632 SV * const sv = newSVpvs("");
1633 if (SvUTF8(PL_linestr))
1635 PL_expect = XOPERATOR;
1636 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1640 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1641 PL_lex_state = LEX_INTERPCASEMOD;
1645 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1646 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1647 PL_linestr = PL_lex_repl;
1649 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1650 PL_bufend += SvCUR(PL_linestr);
1651 PL_last_lop = PL_last_uni = NULL;
1652 SAVEFREESV(PL_linestr);
1653 PL_lex_dojoin = FALSE;
1654 PL_lex_brackets = 0;
1655 PL_lex_casemods = 0;
1656 *PL_lex_casestack = '\0';
1658 if (SvEVALED(PL_lex_repl)) {
1659 PL_lex_state = LEX_INTERPNORMAL;
1661 /* we don't clear PL_lex_repl here, so that we can check later
1662 whether this is an evalled subst; that means we rely on the
1663 logic to ensure sublex_done() is called again only via the
1664 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1667 PL_lex_state = LEX_INTERPCONCAT;
1677 PL_endwhite = newSVpvn("",0);
1678 sv_catsv(PL_endwhite, PL_thiswhite);
1682 sv_setpvn(PL_thistoken,"",0);
1684 PL_realtokenstart = -1;
1688 PL_bufend = SvPVX(PL_linestr);
1689 PL_bufend += SvCUR(PL_linestr);
1690 PL_expect = XOPERATOR;
1691 PL_sublex_info.sub_inwhat = 0;
1699 Extracts a pattern, double-quoted string, or transliteration. This
1702 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1703 processing a pattern (PL_lex_inpat is true), a transliteration
1704 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1706 Returns a pointer to the character scanned up to. Iff this is
1707 advanced from the start pointer supplied (ie if anything was
1708 successfully parsed), will leave an OP for the substring scanned
1709 in yylval. Caller must intuit reason for not parsing further
1710 by looking at the next characters herself.
1714 double-quoted style: \r and \n
1715 regexp special ones: \D \s
1717 backrefs: \1 (deprecated in substitution replacements)
1718 case and quoting: \U \Q \E
1719 stops on @ and $, but not for $ as tail anchor
1721 In transliterations:
1722 characters are VERY literal, except for - not at the start or end
1723 of the string, which indicates a range. scan_const expands the
1724 range to the full set of intermediate characters.
1726 In double-quoted strings:
1728 double-quoted style: \r and \n
1730 backrefs: \1 (deprecated)
1731 case and quoting: \U \Q \E
1734 scan_const does *not* construct ops to handle interpolated strings.
1735 It stops processing as soon as it finds an embedded $ or @ variable
1736 and leaves it to the caller to work out what's going on.
1738 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1740 $ in pattern could be $foo or could be tail anchor. Assumption:
1741 it's a tail anchor if $ is the last thing in the string, or if it's
1742 followed by one of ")| \n\t"
1744 \1 (backreferences) are turned into $1
1746 The structure of the code is
1747 while (there's a character to process) {
1748 handle transliteration ranges
1749 skip regexp comments
1750 skip # initiated comments in //x patterns
1751 check for embedded @foo
1752 check for embedded scalars
1754 leave intact backslashes from leave (below)
1755 deprecate \1 in strings and sub replacements
1756 handle string-changing backslashes \l \U \Q \E, etc.
1757 switch (what was escaped) {
1758 handle - in a transliteration (becomes a literal -)
1759 handle \132 octal characters
1760 handle 0x15 hex characters
1761 handle \cV (control V)
1762 handle printf backslashes (\f, \r, \n, etc)
1764 } (end if backslash)
1765 } (end while character to read)
1770 S_scan_const(pTHX_ char *start)
1773 register char *send = PL_bufend; /* end of the constant */
1774 SV *sv = newSV(send - start); /* sv for the constant */
1775 register char *s = start; /* start of the constant */
1776 register char *d = SvPVX(sv); /* destination for copies */
1777 bool dorange = FALSE; /* are we in a translit range? */
1778 bool didrange = FALSE; /* did we just finish a range? */
1779 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1780 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1783 UV literal_endpoint = 0;
1786 const char * const leaveit = /* set of acceptably-backslashed characters */
1788 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1791 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1792 /* If we are doing a trans and we know we want UTF8 set expectation */
1793 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1794 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1798 while (s < send || dorange) {
1799 /* get transliterations out of the way (they're most literal) */
1800 if (PL_lex_inwhat == OP_TRANS) {
1801 /* expand a range A-Z to the full set of characters. AIE! */
1803 I32 i; /* current expanded character */
1804 I32 min; /* first character in range */
1805 I32 max; /* last character in range */
1808 char * const c = (char*)utf8_hop((U8*)d, -1);
1812 *c = (char)UTF_TO_NATIVE(0xff);
1813 /* mark the range as done, and continue */
1819 i = d - SvPVX_const(sv); /* remember current offset */
1820 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1821 d = SvPVX(sv) + i; /* refresh d after realloc */
1822 d -= 2; /* eat the first char and the - */
1824 min = (U8)*d; /* first char in range */
1825 max = (U8)d[1]; /* last char in range */
1829 "Invalid range \"%c-%c\" in transliteration operator",
1830 (char)min, (char)max);
1834 if (literal_endpoint == 2 &&
1835 ((isLOWER(min) && isLOWER(max)) ||
1836 (isUPPER(min) && isUPPER(max)))) {
1838 for (i = min; i <= max; i++)
1840 *d++ = NATIVE_TO_NEED(has_utf8,i);
1842 for (i = min; i <= max; i++)
1844 *d++ = NATIVE_TO_NEED(has_utf8,i);
1849 for (i = min; i <= max; i++)
1852 /* mark the range as done, and continue */
1856 literal_endpoint = 0;
1861 /* range begins (ignore - as first or last char) */
1862 else if (*s == '-' && s+1 < send && s != start) {
1864 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1867 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1877 literal_endpoint = 0;
1882 /* if we get here, we're not doing a transliteration */
1884 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1885 except for the last char, which will be done separately. */
1886 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1888 while (s+1 < send && *s != ')')
1889 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1891 else if (s[2] == '{' /* This should match regcomp.c */
1892 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1895 char *regparse = s + (s[2] == '{' ? 3 : 4);
1898 while (count && (c = *regparse)) {
1899 if (c == '\\' && regparse[1])
1907 if (*regparse != ')')
1908 regparse--; /* Leave one char for continuation. */
1909 while (s < regparse)
1910 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1914 /* likewise skip #-initiated comments in //x patterns */
1915 else if (*s == '#' && PL_lex_inpat &&
1916 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1917 while (s+1 < send && *s != '\n')
1918 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1921 /* check for embedded arrays
1922 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1924 else if (*s == '@' && s[1]
1925 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1928 /* check for embedded scalars. only stop if we're sure it's a
1931 else if (*s == '$') {
1932 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1934 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1935 break; /* in regexp, $ might be tail anchor */
1938 /* End of else if chain - OP_TRANS rejoin rest */
1941 if (*s == '\\' && s+1 < send) {
1944 /* some backslashes we leave behind */
1945 if (*leaveit && *s && strchr(leaveit, *s)) {
1946 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1947 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1951 /* deprecate \1 in strings and substitution replacements */
1952 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1953 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1955 if (ckWARN(WARN_SYNTAX))
1956 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1961 /* string-change backslash escapes */
1962 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1967 /* if we get here, it's either a quoted -, or a digit */
1970 /* quoted - in transliterations */
1972 if (PL_lex_inwhat == OP_TRANS) {
1982 Perl_warner(aTHX_ packWARN(WARN_MISC),
1983 "Unrecognized escape \\%c passed through",
1985 /* default action is to copy the quoted character */
1986 goto default_action;
1989 /* \132 indicates an octal constant */
1990 case '0': case '1': case '2': case '3':
1991 case '4': case '5': case '6': case '7':
1995 uv = grok_oct(s, &len, &flags, NULL);
1998 goto NUM_ESCAPE_INSERT;
2000 /* \x24 indicates a hex constant */
2004 char* const e = strchr(s, '}');
2005 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2006 PERL_SCAN_DISALLOW_PREFIX;
2011 yyerror("Missing right brace on \\x{}");
2015 uv = grok_hex(s, &len, &flags, NULL);
2021 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2022 uv = grok_hex(s, &len, &flags, NULL);
2028 /* Insert oct or hex escaped character.
2029 * There will always enough room in sv since such
2030 * escapes will be longer than any UTF-8 sequence
2031 * they can end up as. */
2033 /* We need to map to chars to ASCII before doing the tests
2036 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2037 if (!has_utf8 && uv > 255) {
2038 /* Might need to recode whatever we have
2039 * accumulated so far if it contains any
2042 * (Can't we keep track of that and avoid
2043 * this rescan? --jhi)
2047 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2048 if (!NATIVE_IS_INVARIANT(*c)) {
2053 const STRLEN offset = d - SvPVX_const(sv);
2055 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2059 while (src >= (const U8 *)SvPVX_const(sv)) {
2060 if (!NATIVE_IS_INVARIANT(*src)) {
2061 const U8 ch = NATIVE_TO_ASCII(*src);
2062 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2063 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2073 if (has_utf8 || uv > 255) {
2074 d = (char*)uvchr_to_utf8((U8*)d, uv);
2076 if (PL_lex_inwhat == OP_TRANS &&
2077 PL_sublex_info.sub_op) {
2078 PL_sublex_info.sub_op->op_private |=
2079 (PL_lex_repl ? OPpTRANS_FROM_UTF
2092 /* \N{LATIN SMALL LETTER A} is a named character */
2096 char* e = strchr(s, '}');
2102 yyerror("Missing right brace on \\N{}");
2106 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2108 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2109 PERL_SCAN_DISALLOW_PREFIX;
2112 uv = grok_hex(s, &len, &flags, NULL);
2114 goto NUM_ESCAPE_INSERT;
2116 res = newSVpvn(s + 1, e - s - 1);
2117 res = new_constant( NULL, 0, "charnames",
2118 res, NULL, "\\N{...}" );
2120 sv_utf8_upgrade(res);
2121 str = SvPV_const(res,len);
2122 #ifdef EBCDIC_NEVER_MIND
2123 /* charnames uses pack U and that has been
2124 * recently changed to do the below uni->native
2125 * mapping, so this would be redundant (and wrong,
2126 * the code point would be doubly converted).
2127 * But leave this in just in case the pack U change
2128 * gets revoked, but the semantics is still
2129 * desireable for charnames. --jhi */
2131 UV uv = utf8_to_uvchr((const U8*)str, 0);
2134 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2136 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2137 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2138 str = SvPV_const(res, len);
2142 if (!has_utf8 && SvUTF8(res)) {
2143 const char * const ostart = SvPVX_const(sv);
2144 SvCUR_set(sv, d - ostart);
2147 sv_utf8_upgrade(sv);
2148 /* this just broke our allocation above... */
2149 SvGROW(sv, (STRLEN)(send - start));
2150 d = SvPVX(sv) + SvCUR(sv);
2153 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2154 const char * const odest = SvPVX_const(sv);
2156 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2157 d = SvPVX(sv) + (d - odest);
2159 Copy(str, d, len, char);
2166 yyerror("Missing braces on \\N{}");
2169 /* \c is a control character */
2178 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2181 yyerror("Missing control char name in \\c");
2185 /* printf-style backslashes, formfeeds, newlines, etc */
2187 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2190 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2193 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2196 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2199 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2202 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2205 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2211 } /* end if (backslash) */
2218 /* If we started with encoded form, or already know we want it
2219 and then encode the next character */
2220 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2222 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2223 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2226 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2227 const STRLEN off = d - SvPVX_const(sv);
2228 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2230 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2234 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2236 } /* while loop to process each character */
2238 /* terminate the string and set up the sv */
2240 SvCUR_set(sv, d - SvPVX_const(sv));
2241 if (SvCUR(sv) >= SvLEN(sv))
2242 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2245 if (PL_encoding && !has_utf8) {
2246 sv_recode_to_utf8(sv, PL_encoding);
2252 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2253 PL_sublex_info.sub_op->op_private |=
2254 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2258 /* shrink the sv if we allocated more than we used */
2259 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2260 SvPV_shrink_to_cur(sv);
2263 /* return the substring (via yylval) only if we parsed anything */
2264 if (s > PL_bufptr) {
2265 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2266 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2268 ( PL_lex_inwhat == OP_TRANS
2270 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2273 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2280 * Returns TRUE if there's more to the expression (e.g., a subscript),
2283 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2285 * ->[ and ->{ return TRUE
2286 * { and [ outside a pattern are always subscripts, so return TRUE
2287 * if we're outside a pattern and it's not { or [, then return FALSE
2288 * if we're in a pattern and the first char is a {
2289 * {4,5} (any digits around the comma) returns FALSE
2290 * if we're in a pattern and the first char is a [
2292 * [SOMETHING] has a funky algorithm to decide whether it's a
2293 * character class or not. It has to deal with things like
2294 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2295 * anything else returns TRUE
2298 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2301 S_intuit_more(pTHX_ register char *s)
2304 if (PL_lex_brackets)
2306 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2308 if (*s != '{' && *s != '[')
2313 /* In a pattern, so maybe we have {n,m}. */
2330 /* On the other hand, maybe we have a character class */
2333 if (*s == ']' || *s == '^')
2336 /* this is terrifying, and it works */
2337 int weight = 2; /* let's weigh the evidence */
2339 unsigned char un_char = 255, last_un_char;
2340 const char * const send = strchr(s,']');
2341 char tmpbuf[sizeof PL_tokenbuf * 4];
2343 if (!send) /* has to be an expression */
2346 Zero(seen,256,char);
2349 else if (isDIGIT(*s)) {
2351 if (isDIGIT(s[1]) && s[2] == ']')
2357 for (; s < send; s++) {
2358 last_un_char = un_char;
2359 un_char = (unsigned char)*s;
2364 weight -= seen[un_char] * 10;
2365 if (isALNUM_lazy_if(s+1,UTF)) {
2367 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2368 len = (int)strlen(tmpbuf);
2369 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2374 else if (*s == '$' && s[1] &&
2375 strchr("[#!%*<>()-=",s[1])) {
2376 if (/*{*/ strchr("])} =",s[2]))
2385 if (strchr("wds]",s[1]))
2387 else if (seen['\''] || seen['"'])
2389 else if (strchr("rnftbxcav",s[1]))
2391 else if (isDIGIT(s[1])) {
2393 while (s[1] && isDIGIT(s[1]))
2403 if (strchr("aA01! ",last_un_char))
2405 if (strchr("zZ79~",s[1]))
2407 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2408 weight -= 5; /* cope with negative subscript */
2411 if (!isALNUM(last_un_char)
2412 && !(last_un_char == '$' || last_un_char == '@'
2413 || last_un_char == '&')
2414 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2419 if (keyword(tmpbuf, d - tmpbuf))
2422 if (un_char == last_un_char + 1)
2424 weight -= seen[un_char];
2429 if (weight >= 0) /* probably a character class */
2439 * Does all the checking to disambiguate
2441 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2442 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2444 * First argument is the stuff after the first token, e.g. "bar".
2446 * Not a method if bar is a filehandle.
2447 * Not a method if foo is a subroutine prototyped to take a filehandle.
2448 * Not a method if it's really "Foo $bar"
2449 * Method if it's "foo $bar"
2450 * Not a method if it's really "print foo $bar"
2451 * Method if it's really "foo package::" (interpreted as package->foo)
2452 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2453 * Not a method if bar is a filehandle or package, but is quoted with
2458 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2461 char *s = start + (*start == '$');
2462 char tmpbuf[sizeof PL_tokenbuf];
2470 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2474 const char *proto = SvPVX_const(cv);
2485 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2486 /* start is the beginning of the possible filehandle/object,
2487 * and s is the end of it
2488 * tmpbuf is a copy of it
2491 if (*start == '$') {
2492 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2495 len = start - SvPVX(PL_linestr);
2499 start = SvPVX(PL_linestr) + len;
2503 return *s == '(' ? FUNCMETH : METHOD;
2505 if (!keyword(tmpbuf, len)) {
2506 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2510 soff = s - SvPVX(PL_linestr);
2514 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2515 if (indirgv && GvCVu(indirgv))
2517 /* filehandle or package name makes it a method */
2518 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2520 soff = s - SvPVX(PL_linestr);
2523 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2524 return 0; /* no assumptions -- "=>" quotes bearword */
2526 start_force(PL_curforce);
2527 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2528 newSVpvn(tmpbuf,len));
2529 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2531 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2536 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2538 return *s == '(' ? FUNCMETH : METHOD;
2546 * Return a string of Perl code to load the debugger. If PERL5DB
2547 * is set, it will return the contents of that, otherwise a
2548 * compile-time require of perl5db.pl.
2556 const char * const pdb = PerlEnv_getenv("PERL5DB");
2560 SETERRNO(0,SS_NORMAL);
2561 return "BEGIN { require 'perl5db.pl' }";
2567 /* Encoded script support. filter_add() effectively inserts a
2568 * 'pre-processing' function into the current source input stream.
2569 * Note that the filter function only applies to the current source file
2570 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2572 * The datasv parameter (which may be NULL) can be used to pass
2573 * private data to this instance of the filter. The filter function
2574 * can recover the SV using the FILTER_DATA macro and use it to
2575 * store private buffers and state information.
2577 * The supplied datasv parameter is upgraded to a PVIO type
2578 * and the IoDIRP/IoANY field is used to store the function pointer,
2579 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2580 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2581 * private use must be set using malloc'd pointers.
2585 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2591 if (!PL_rsfp_filters)
2592 PL_rsfp_filters = newAV();
2595 SvUPGRADE(datasv, SVt_PVIO);
2596 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2597 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2598 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2599 FPTR2DPTR(void *, IoANY(datasv)),
2600 SvPV_nolen(datasv)));
2601 av_unshift(PL_rsfp_filters, 1);
2602 av_store(PL_rsfp_filters, 0, datasv) ;
2607 /* Delete most recently added instance of this filter function. */
2609 Perl_filter_del(pTHX_ filter_t funcp)
2615 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2616 FPTR2DPTR(void*, funcp)));
2618 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2620 /* if filter is on top of stack (usual case) just pop it off */
2621 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2622 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2623 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2624 IoANY(datasv) = (void *)NULL;
2625 sv_free(av_pop(PL_rsfp_filters));
2629 /* we need to search for the correct entry and clear it */
2630 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2634 /* Invoke the idxth filter function for the current rsfp. */
2635 /* maxlen 0 = read one text line */
2637 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2642 /* This API is bad. It should have been using unsigned int for maxlen.
2643 Not sure if we want to change the API, but if not we should sanity
2644 check the value here. */
2645 const unsigned int correct_length
2654 if (!PL_rsfp_filters)
2656 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2657 /* Provide a default input filter to make life easy. */
2658 /* Note that we append to the line. This is handy. */
2659 DEBUG_P(PerlIO_printf(Perl_debug_log,
2660 "filter_read %d: from rsfp\n", idx));
2661 if (correct_length) {
2664 const int old_len = SvCUR(buf_sv);
2666 /* ensure buf_sv is large enough */
2667 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2668 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2669 correct_length)) <= 0) {
2670 if (PerlIO_error(PL_rsfp))
2671 return -1; /* error */
2673 return 0 ; /* end of file */
2675 SvCUR_set(buf_sv, old_len + len) ;
2678 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2679 if (PerlIO_error(PL_rsfp))
2680 return -1; /* error */
2682 return 0 ; /* end of file */
2685 return SvCUR(buf_sv);
2687 /* Skip this filter slot if filter has been deleted */
2688 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2689 DEBUG_P(PerlIO_printf(Perl_debug_log,
2690 "filter_read %d: skipped (filter deleted)\n",
2692 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2694 /* Get function pointer hidden within datasv */
2695 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2696 DEBUG_P(PerlIO_printf(Perl_debug_log,
2697 "filter_read %d: via function %p (%s)\n",
2698 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2699 /* Call function. The function is expected to */
2700 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2701 /* Return: <0:error, =0:eof, >0:not eof */
2702 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2706 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2709 #ifdef PERL_CR_FILTER
2710 if (!PL_rsfp_filters) {
2711 filter_add(S_cr_textfilter,NULL);
2714 if (PL_rsfp_filters) {
2716 SvCUR_set(sv, 0); /* start with empty line */
2717 if (FILTER_READ(0, sv, 0) > 0)
2718 return ( SvPVX(sv) ) ;
2723 return (sv_gets(sv, fp, append));
2727 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2732 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2736 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2737 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2739 return GvHV(gv); /* Foo:: */
2742 /* use constant CLASS => 'MyClass' */
2743 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2744 if (gv && GvCV(gv)) {
2745 SV * const sv = cv_const_sv(GvCV(gv));
2747 pkgname = SvPV_nolen_const(sv);
2750 return gv_stashpv(pkgname, FALSE);
2756 * The intent of this yylex wrapper is to minimize the changes to the
2757 * tokener when we aren't interested in collecting madprops. It remains
2758 * to be seen how successful this strategy will be...
2765 char *s = PL_bufptr;
2767 /* make sure PL_thiswhite is initialized */
2771 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2772 if (PL_pending_ident)
2773 return S_pending_ident(aTHX);
2775 /* previous token ate up our whitespace? */
2776 if (!PL_lasttoke && PL_nextwhite) {
2777 PL_thiswhite = PL_nextwhite;
2781 /* isolate the token, and figure out where it is without whitespace */
2782 PL_realtokenstart = -1;
2786 assert(PL_curforce < 0);
2788 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2789 if (!PL_thistoken) {
2790 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2791 PL_thistoken = newSVpvn("",0);
2793 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2794 PL_thistoken = newSVpvn(tstart, s - tstart);
2797 if (PL_thismad) /* install head */
2798 CURMAD('X', PL_thistoken);
2801 /* last whitespace of a sublex? */
2802 if (optype == ')' && PL_endwhite) {
2803 CURMAD('X', PL_endwhite);
2808 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2809 if (!PL_thiswhite && !PL_endwhite && !optype) {
2810 sv_free(PL_thistoken);
2815 /* put off final whitespace till peg */
2816 if (optype == ';' && !PL_rsfp) {
2817 PL_nextwhite = PL_thiswhite;
2820 else if (PL_thisopen) {
2821 CURMAD('q', PL_thisopen);
2823 sv_free(PL_thistoken);
2827 /* Store actual token text as madprop X */
2828 CURMAD('X', PL_thistoken);
2832 /* add preceding whitespace as madprop _ */
2833 CURMAD('_', PL_thiswhite);
2837 /* add quoted material as madprop = */
2838 CURMAD('=', PL_thisstuff);
2842 /* add terminating quote as madprop Q */
2843 CURMAD('Q', PL_thisclose);
2847 /* special processing based on optype */
2851 /* opval doesn't need a TOKEN since it can already store mp */
2862 append_madprops(PL_thismad, yylval.opval, 0);
2870 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2879 /* remember any fake bracket that lexer is about to discard */
2880 if (PL_lex_brackets == 1 &&
2881 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2884 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2887 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2888 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2891 break; /* don't bother looking for trailing comment */
2900 /* attach a trailing comment to its statement instead of next token */
2904 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2906 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2908 if (*s == '\n' || *s == '#') {
2909 while (s < PL_bufend && *s != '\n')
2913 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2914 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2931 /* Create new token struct. Note: opvals return early above. */
2932 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2939 S_tokenize_use(pTHX_ int is_use, char *s) {
2941 if (PL_expect != XSTATE)
2942 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2943 is_use ? "use" : "no"));
2945 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2946 s = force_version(s, TRUE);
2947 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2948 start_force(PL_curforce);
2949 NEXTVAL_NEXTTOKE.opval = NULL;
2952 else if (*s == 'v') {
2953 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2954 s = force_version(s, FALSE);
2958 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2959 s = force_version(s, FALSE);
2961 yylval.ival = is_use;
2965 static const char* const exp_name[] =
2966 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2967 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2974 Works out what to call the token just pulled out of the input
2975 stream. The yacc parser takes care of taking the ops we return and
2976 stitching them into a tree.
2982 if read an identifier
2983 if we're in a my declaration
2984 croak if they tried to say my($foo::bar)
2985 build the ops for a my() declaration
2986 if it's an access to a my() variable
2987 are we in a sort block?
2988 croak if my($a); $a <=> $b
2989 build ops for access to a my() variable
2990 if in a dq string, and they've said @foo and we can't find @foo
2992 build ops for a bareword
2993 if we already built the token before, use it.
2998 #pragma segment Perl_yylex
3004 register char *s = PL_bufptr;
3010 SV* tmp = newSVpvs("");
3011 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3012 (IV)CopLINE(PL_curcop),
3013 lex_state_names[PL_lex_state],
3014 exp_name[PL_expect],
3015 pv_display(tmp, s, strlen(s), 0, 60));
3018 /* check if there's an identifier for us to look at */
3019 if (PL_pending_ident)
3020 return REPORT(S_pending_ident(aTHX));
3022 /* no identifier pending identification */
3024 switch (PL_lex_state) {
3026 case LEX_NORMAL: /* Some compilers will produce faster */
3027 case LEX_INTERPNORMAL: /* code if we comment these out. */
3031 /* when we've already built the next token, just pull it out of the queue */
3035 yylval = PL_nexttoke[PL_lasttoke].next_val;
3037 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3038 PL_nexttoke[PL_lasttoke].next_mad = 0;
3039 if (PL_thismad && PL_thismad->mad_key == '_') {
3040 PL_thiswhite = (SV*)PL_thismad->mad_val;
3041 PL_thismad->mad_val = 0;
3042 mad_free(PL_thismad);
3047 PL_lex_state = PL_lex_defer;
3048 PL_expect = PL_lex_expect;
3049 PL_lex_defer = LEX_NORMAL;
3050 if (!PL_nexttoke[PL_lasttoke].next_type)
3055 yylval = PL_nextval[PL_nexttoke];
3057 PL_lex_state = PL_lex_defer;
3058 PL_expect = PL_lex_expect;
3059 PL_lex_defer = LEX_NORMAL;
3063 /* FIXME - can these be merged? */
3064 return(PL_nexttoke[PL_lasttoke].next_type);
3066 return REPORT(PL_nexttype[PL_nexttoke]);
3069 /* interpolated case modifiers like \L \U, including \Q and \E.
3070 when we get here, PL_bufptr is at the \
3072 case LEX_INTERPCASEMOD:
3074 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3075 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3077 /* handle \E or end of string */
3078 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3080 if (PL_lex_casemods) {
3081 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3082 PL_lex_casestack[PL_lex_casemods] = '\0';
3084 if (PL_bufptr != PL_bufend
3085 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3087 PL_lex_state = LEX_INTERPCONCAT;
3090 PL_thistoken = newSVpvn("\\E",2);
3096 while (PL_bufptr != PL_bufend &&
3097 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3099 PL_thiswhite = newSVpvn("",0);
3100 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3104 if (PL_bufptr != PL_bufend)
3107 PL_lex_state = LEX_INTERPCONCAT;
3111 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3112 "### Saw case modifier\n"); });
3114 if (s[1] == '\\' && s[2] == 'E') {
3117 PL_thiswhite = newSVpvn("",0);
3118 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3121 PL_lex_state = LEX_INTERPCONCAT;
3126 if (!PL_madskills) /* when just compiling don't need correct */
3127 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3128 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3129 if ((*s == 'L' || *s == 'U') &&
3130 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3131 PL_lex_casestack[--PL_lex_casemods] = '\0';
3134 if (PL_lex_casemods > 10)
3135 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3136 PL_lex_casestack[PL_lex_casemods++] = *s;
3137 PL_lex_casestack[PL_lex_casemods] = '\0';
3138 PL_lex_state = LEX_INTERPCONCAT;
3139 start_force(PL_curforce);
3140 NEXTVAL_NEXTTOKE.ival = 0;
3142 start_force(PL_curforce);
3144 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3146 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3148 NEXTVAL_NEXTTOKE.ival = OP_LC;
3150 NEXTVAL_NEXTTOKE.ival = OP_UC;
3152 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3154 Perl_croak(aTHX_ "panic: yylex");
3156 SV* const tmpsv = newSVpvn("",0);
3157 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3163 if (PL_lex_starts) {
3169 sv_free(PL_thistoken);
3170 PL_thistoken = newSVpvn("",0);
3173 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3174 if (PL_lex_casemods == 1 && PL_lex_inpat)
3183 case LEX_INTERPPUSH:
3184 return REPORT(sublex_push());
3186 case LEX_INTERPSTART:
3187 if (PL_bufptr == PL_bufend)
3188 return REPORT(sublex_done());
3189 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3190 "### Interpolated variable\n"); });
3192 PL_lex_dojoin = (*PL_bufptr == '@');
3193 PL_lex_state = LEX_INTERPNORMAL;
3194 if (PL_lex_dojoin) {
3195 start_force(PL_curforce);
3196 NEXTVAL_NEXTTOKE.ival = 0;
3198 start_force(PL_curforce);
3199 force_ident("\"", '$');
3200 start_force(PL_curforce);
3201 NEXTVAL_NEXTTOKE.ival = 0;
3203 start_force(PL_curforce);
3204 NEXTVAL_NEXTTOKE.ival = 0;
3206 start_force(PL_curforce);
3207 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3210 if (PL_lex_starts++) {
3215 sv_free(PL_thistoken);
3216 PL_thistoken = newSVpvn("",0);
3219 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3220 if (!PL_lex_casemods && PL_lex_inpat)
3227 case LEX_INTERPENDMAYBE:
3228 if (intuit_more(PL_bufptr)) {
3229 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3235 if (PL_lex_dojoin) {
3236 PL_lex_dojoin = FALSE;
3237 PL_lex_state = LEX_INTERPCONCAT;
3241 sv_free(PL_thistoken);
3242 PL_thistoken = newSVpvn("",0);
3247 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3248 && SvEVALED(PL_lex_repl))
3250 if (PL_bufptr != PL_bufend)
3251 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3255 case LEX_INTERPCONCAT:
3257 if (PL_lex_brackets)
3258 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3260 if (PL_bufptr == PL_bufend)
3261 return REPORT(sublex_done());
3263 if (SvIVX(PL_linestr) == '\'') {
3264 SV *sv = newSVsv(PL_linestr);
3267 else if ( PL_hints & HINT_NEW_RE )
3268 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3269 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3273 s = scan_const(PL_bufptr);
3275 PL_lex_state = LEX_INTERPCASEMOD;
3277 PL_lex_state = LEX_INTERPSTART;
3280 if (s != PL_bufptr) {
3281 start_force(PL_curforce);
3283 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3285 NEXTVAL_NEXTTOKE = yylval;
3288 if (PL_lex_starts++) {
3292 sv_free(PL_thistoken);
3293 PL_thistoken = newSVpvn("",0);
3296 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3297 if (!PL_lex_casemods && PL_lex_inpat)
3310 PL_lex_state = LEX_NORMAL;
3311 s = scan_formline(PL_bufptr);
3312 if (!PL_lex_formbrack)
3318 PL_oldoldbufptr = PL_oldbufptr;
3324 sv_free(PL_thistoken);
3327 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3331 if (isIDFIRST_lazy_if(s,UTF))
3333 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3336 goto fake_eof; /* emulate EOF on ^D or ^Z */
3345 if (PL_lex_brackets) {
3346 yyerror(PL_lex_formbrack
3347 ? "Format not terminated"
3348 : "Missing right curly or square bracket");
3350 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3351 "### Tokener got EOF\n");
3355 if (s++ < PL_bufend)
3356 goto retry; /* ignore stray nulls */
3359 if (!PL_in_eval && !PL_preambled) {
3360 PL_preambled = TRUE;
3365 sv_setpv(PL_linestr,incl_perldb());
3366 if (SvCUR(PL_linestr))
3367 sv_catpvs(PL_linestr,";");
3369 while(AvFILLp(PL_preambleav) >= 0) {
3370 SV *tmpsv = av_shift(PL_preambleav);
3371 sv_catsv(PL_linestr, tmpsv);
3372 sv_catpvs(PL_linestr, ";");
3375 sv_free((SV*)PL_preambleav);
3376 PL_preambleav = NULL;
3378 if (PL_minus_n || PL_minus_p) {
3379 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3381 sv_catpvs(PL_linestr,"chomp;");
3384 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3385 || *PL_splitstr == '"')
3386 && strchr(PL_splitstr + 1, *PL_splitstr))
3387 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3389 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3390 bytes can be used as quoting characters. :-) */
3391 const char *splits = PL_splitstr;
3392 sv_catpvs(PL_linestr, "our @F=split(q\0");
3395 if (*splits == '\\')
3396 sv_catpvn(PL_linestr, splits, 1);
3397 sv_catpvn(PL_linestr, splits, 1);
3398 } while (*splits++);
3399 /* This loop will embed the trailing NUL of
3400 PL_linestr as the last thing it does before
3402 sv_catpvs(PL_linestr, ");");
3406 sv_catpvs(PL_linestr,"our @F=split(' ');");
3410 sv_catpvs(PL_linestr,"use feature ':5.10';");
3411 sv_catpvs(PL_linestr, "\n");
3412 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3413 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3414 PL_last_lop = PL_last_uni = NULL;
3415 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3416 SV * const sv = newSV(0);
3418 sv_upgrade(sv, SVt_PVMG);
3419 sv_setsv(sv,PL_linestr);
3422 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3427 bof = PL_rsfp ? TRUE : FALSE;
3428 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3431 PL_realtokenstart = -1;
3434 if (PL_preprocess && !PL_in_eval)
3435 (void)PerlProc_pclose(PL_rsfp);
3436 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3437 PerlIO_clearerr(PL_rsfp);
3439 (void)PerlIO_close(PL_rsfp);
3441 PL_doextract = FALSE;
3443 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3448 sv_setpv(PL_linestr,PL_minus_p
3449 ? ";}continue{print;}" : ";}");
3450 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3451 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3452 PL_last_lop = PL_last_uni = NULL;
3453 PL_minus_n = PL_minus_p = 0;
3456 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3457 PL_last_lop = PL_last_uni = NULL;
3458 sv_setpvn(PL_linestr,"",0);
3459 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3461 /* If it looks like the start of a BOM or raw UTF-16,
3462 * check if it in fact is. */
3468 #ifdef PERLIO_IS_STDIO
3469 # ifdef __GNU_LIBRARY__
3470 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3471 # define FTELL_FOR_PIPE_IS_BROKEN
3475 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3476 # define FTELL_FOR_PIPE_IS_BROKEN
3481 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3482 /* This loses the possibility to detect the bof
3483 * situation on perl -P when the libc5 is being used.
3484 * Workaround? Maybe attach some extra state to PL_rsfp?
3487 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3489 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3492 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3493 s = swallow_bom((U8*)s);
3497 /* Incest with pod. */
3500 sv_catsv(PL_thiswhite, PL_linestr);
3502 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3503 sv_setpvn(PL_linestr, "", 0);
3504 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3505 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3506 PL_last_lop = PL_last_uni = NULL;
3507 PL_doextract = FALSE;
3511 } while (PL_doextract);
3512 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3513 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3514 SV * const sv = newSV(0);
3516 sv_upgrade(sv, SVt_PVMG);
3517 sv_setsv(sv,PL_linestr);
3520 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3522 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3523 PL_last_lop = PL_last_uni = NULL;
3524 if (CopLINE(PL_curcop) == 1) {
3525 while (s < PL_bufend && isSPACE(*s))
3527 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3531 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3535 if (*s == '#' && *(s+1) == '!')
3537 #ifdef ALTERNATE_SHEBANG
3539 static char const as[] = ALTERNATE_SHEBANG;
3540 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3541 d = s + (sizeof(as) - 1);
3543 #endif /* ALTERNATE_SHEBANG */
3552 while (*d && !isSPACE(*d))
3556 #ifdef ARG_ZERO_IS_SCRIPT
3557 if (ipathend > ipath) {
3559 * HP-UX (at least) sets argv[0] to the script name,
3560 * which makes $^X incorrect. And Digital UNIX and Linux,
3561 * at least, set argv[0] to the basename of the Perl
3562 * interpreter. So, having found "#!", we'll set it right.
3564 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3566 assert(SvPOK(x) || SvGMAGICAL(x));
3567 if (sv_eq(x, CopFILESV(PL_curcop))) {
3568 sv_setpvn(x, ipath, ipathend - ipath);
3574 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3575 const char * const lstart = SvPV_const(x,llen);
3577 bstart += blen - llen;
3578 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3579 sv_setpvn(x, ipath, ipathend - ipath);
3584 TAINT_NOT; /* $^X is always tainted, but that's OK */
3586 #endif /* ARG_ZERO_IS_SCRIPT */
3591 d = instr(s,"perl -");
3593 d = instr(s,"perl");
3595 /* avoid getting into infinite loops when shebang
3596 * line contains "Perl" rather than "perl" */
3598 for (d = ipathend-4; d >= ipath; --d) {
3599 if ((*d == 'p' || *d == 'P')
3600 && !ibcmp(d, "perl", 4))
3610 #ifdef ALTERNATE_SHEBANG
3612 * If the ALTERNATE_SHEBANG on this system starts with a
3613 * character that can be part of a Perl expression, then if
3614 * we see it but not "perl", we're probably looking at the
3615 * start of Perl code, not a request to hand off to some
3616 * other interpreter. Similarly, if "perl" is there, but
3617 * not in the first 'word' of the line, we assume the line
3618 * contains the start of the Perl program.
3620 if (d && *s != '#') {
3621 const char *c = ipath;
3622 while (*c && !strchr("; \t\r\n\f\v#", *c))
3625 d = NULL; /* "perl" not in first word; ignore */
3627 *s = '#'; /* Don't try to parse shebang line */
3629 #endif /* ALTERNATE_SHEBANG */
3630 #ifndef MACOS_TRADITIONAL
3635 !instr(s,"indir") &&
3636 instr(PL_origargv[0],"perl"))
3643 while (s < PL_bufend && isSPACE(*s))
3645 if (s < PL_bufend) {
3646 Newxz(newargv,PL_origargc+3,char*);
3648 while (s < PL_bufend && !isSPACE(*s))
3651 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3654 newargv = PL_origargv;
3657 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3659 Perl_croak(aTHX_ "Can't exec %s", ipath);
3663 while (*d && !isSPACE(*d))
3665 while (SPACE_OR_TAB(*d))
3669 const bool switches_done = PL_doswitches;
3670 const U32 oldpdb = PL_perldb;
3671 const bool oldn = PL_minus_n;
3672 const bool oldp = PL_minus_p;
3675 if (*d == 'M' || *d == 'm' || *d == 'C') {
3676 const char * const m = d;
3677 while (*d && !isSPACE(*d))
3679 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3682 d = moreswitches(d);
3684 if (PL_doswitches && !switches_done) {
3685 int argc = PL_origargc;
3686 char **argv = PL_origargv;
3689 } while (argc && argv[0][0] == '-' && argv[0][1]);
3690 init_argv_symbols(argc,argv);
3692 if ((PERLDB_LINE && !oldpdb) ||
3693 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3694 /* if we have already added "LINE: while (<>) {",
3695 we must not do it again */
3697 sv_setpvn(PL_linestr, "", 0);
3698 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3699 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3700 PL_last_lop = PL_last_uni = NULL;
3701 PL_preambled = FALSE;
3703 (void)gv_fetchfile(PL_origfilename);
3710 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3712 PL_lex_state = LEX_FORMLINE;
3717 #ifdef PERL_STRICT_CR
3718 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3720 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3722 case ' ': case '\t': case '\f': case 013:
3723 #ifdef MACOS_TRADITIONAL
3727 PL_realtokenstart = -1;
3736 PL_realtokenstart = -1;
3740 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3741 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3742 /* handle eval qq[#line 1 "foo"\n ...] */
3743 CopLINE_dec(PL_curcop);
3746 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3748 if (!PL_in_eval || PL_rsfp)
3753 while (d < PL_bufend && *d != '\n')
3757 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3758 Perl_croak(aTHX_ "panic: input overflow");
3761 PL_thiswhite = newSVpvn(s, d - s);
3766 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3768 PL_lex_state = LEX_FORMLINE;
3774 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3775 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3778 TOKEN(PEG); /* make sure any #! line is accessible */
3783 /* if (PL_madskills && PL_lex_formbrack) { */
3785 while (d < PL_bufend && *d != '\n')
3789 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3790 Perl_croak(aTHX_ "panic: input overflow");
3791 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3793 PL_thiswhite = newSVpvn("",0);
3794 if (CopLINE(PL_curcop) == 1) {
3795 sv_setpvn(PL_thiswhite, "", 0);
3798 sv_catpvn(PL_thiswhite, s, d - s);
3812 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3820 while (s < PL_bufend && SPACE_OR_TAB(*s))
3823 if (strnEQ(s,"=>",2)) {
3824 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3825 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3826 OPERATOR('-'); /* unary minus */
3828 PL_last_uni = PL_oldbufptr;
3830 case 'r': ftst = OP_FTEREAD; break;
3831 case 'w': ftst = OP_FTEWRITE; break;
3832 case 'x': ftst = OP_FTEEXEC; break;
3833 case 'o': ftst = OP_FTEOWNED; break;
3834 case 'R': ftst = OP_FTRREAD; break;
3835 case 'W': ftst = OP_FTRWRITE; break;
3836 case 'X': ftst = OP_FTREXEC; break;
3837 case 'O': ftst = OP_FTROWNED; break;
3838 case 'e': ftst = OP_FTIS; break;
3839 case 'z': ftst = OP_FTZERO; break;
3840 case 's': ftst = OP_FTSIZE; break;
3841 case 'f': ftst = OP_FTFILE; break;
3842 case 'd': ftst = OP_FTDIR; break;
3843 case 'l': ftst = OP_FTLINK; break;
3844 case 'p': ftst = OP_FTPIPE; break;
3845 case 'S': ftst = OP_FTSOCK; break;
3846 case 'u': ftst = OP_FTSUID; break;
3847 case 'g': ftst = OP_FTSGID; break;
3848 case 'k': ftst = OP_FTSVTX; break;
3849 case 'b': ftst = OP_FTBLK; break;
3850 case 'c': ftst = OP_FTCHR; break;
3851 case 't': ftst = OP_FTTTY; break;
3852 case 'T': ftst = OP_FTTEXT; break;
3853 case 'B': ftst = OP_FTBINARY; break;
3854 case 'M': case 'A': case 'C':
3855 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3857 case 'M': ftst = OP_FTMTIME; break;
3858 case 'A': ftst = OP_FTATIME; break;
3859 case 'C': ftst = OP_FTCTIME; break;
3867 PL_last_lop_op = (OPCODE)ftst;
3868 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3869 "### Saw file test %c\n", (int)tmp);
3874 /* Assume it was a minus followed by a one-letter named
3875 * subroutine call (or a -bareword), then. */
3876 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3877 "### '-%c' looked like a file test but was not\n",
3884 const char tmp = *s++;
3887 if (PL_expect == XOPERATOR)
3892 else if (*s == '>') {
3895 if (isIDFIRST_lazy_if(s,UTF)) {
3896 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3904 if (PL_expect == XOPERATOR)
3907 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3909 OPERATOR('-'); /* unary minus */
3915 const char tmp = *s++;
3918 if (PL_expect == XOPERATOR)
3923 if (PL_expect == XOPERATOR)
3926 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3933 if (PL_expect != XOPERATOR) {
3934 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3935 PL_expect = XOPERATOR;
3936 force_ident(PL_tokenbuf, '*');
3949 if (PL_expect == XOPERATOR) {
3953 PL_tokenbuf[0] = '%';
3954 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3955 if (!PL_tokenbuf[1]) {
3958 PL_pending_ident = '%';
3969 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3970 && FEATURE_IS_ENABLED("~~"))
3977 const char tmp = *s++;
3983 goto just_a_word_zero_gv;
3986 switch (PL_expect) {
3992 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3994 PL_bufptr = s; /* update in case we back off */
4000 PL_expect = XTERMBLOCK;
4003 stuffstart = s - SvPVX(PL_linestr) - 1;
4007 while (isIDFIRST_lazy_if(s,UTF)) {
4010 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4011 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4012 if (tmp < 0) tmp = -tmp;
4027 sv = newSVpvn(s, len);
4029 d = scan_str(d,TRUE,TRUE);
4031 /* MUST advance bufptr here to avoid bogus
4032 "at end of line" context messages from yyerror().
4034 PL_bufptr = s + len;
4035 yyerror("Unterminated attribute parameter in attribute list");
4039 return REPORT(0); /* EOF indicator */
4043 sv_catsv(sv, PL_lex_stuff);
4044 attrs = append_elem(OP_LIST, attrs,
4045 newSVOP(OP_CONST, 0, sv));
4046 SvREFCNT_dec(PL_lex_stuff);
4047 PL_lex_stuff = NULL;
4050 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4052 if (PL_in_my == KEY_our) {
4054 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4056 /* skip to avoid loading attributes.pm */
4058 deprecate(":unique");
4061 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4064 /* NOTE: any CV attrs applied here need to be part of
4065 the CVf_BUILTIN_ATTRS define in cv.h! */
4066 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4068 CvLVALUE_on(PL_compcv);
4070 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4072 CvLOCKED_on(PL_compcv);
4074 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4076 CvMETHOD_on(PL_compcv);
4078 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4080 CvASSERTION_on(PL_compcv);
4082 /* After we've set the flags, it could be argued that
4083 we don't need to do the attributes.pm-based setting
4084 process, and shouldn't bother appending recognized
4085 flags. To experiment with that, uncomment the
4086 following "else". (Note that's already been
4087 uncommented. That keeps the above-applied built-in
4088 attributes from being intercepted (and possibly
4089 rejected) by a package's attribute routines, but is
4090 justified by the performance win for the common case
4091 of applying only built-in attributes.) */
4093 attrs = append_elem(OP_LIST, attrs,
4094 newSVOP(OP_CONST, 0,
4098 if (*s == ':' && s[1] != ':')
4101 break; /* require real whitespace or :'s */
4102 /* XXX losing whitespace on sequential attributes here */
4106 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4107 if (*s != ';' && *s != '}' && *s != tmp
4108 && (tmp != '=' || *s != ')')) {
4109 const char q = ((*s == '\'') ? '"' : '\'');
4110 /* If here for an expression, and parsed no attrs, back
4112 if (tmp == '=' && !attrs) {
4116 /* MUST advance bufptr here to avoid bogus "at end of line"
4117 context messages from yyerror().
4121 ? Perl_form(aTHX_ "Invalid separator character "
4122 "%c%c%c in attribute list", q, *s, q)
4123 : "Unterminated attribute list" );
4131 start_force(PL_curforce);
4132 NEXTVAL_NEXTTOKE.opval = attrs;
4133 CURMAD('_', PL_nextwhite);
4138 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4139 (s - SvPVX(PL_linestr)) - stuffstart);
4147 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4148 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4156 const char tmp = *s++;
4161 const char tmp = *s++;
4169 if (PL_lex_brackets <= 0)
4170 yyerror("Unmatched right square bracket");
4173 if (PL_lex_state == LEX_INTERPNORMAL) {
4174 if (PL_lex_brackets == 0) {
4175 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4176 PL_lex_state = LEX_INTERPEND;
4183 if (PL_lex_brackets > 100) {
4184 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4186 switch (PL_expect) {
4188 if (PL_lex_formbrack) {
4192 if (PL_oldoldbufptr == PL_last_lop)
4193 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4195 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4196 OPERATOR(HASHBRACK);
4198 while (s < PL_bufend && SPACE_OR_TAB(*s))
4201 PL_tokenbuf[0] = '\0';
4202 if (d < PL_bufend && *d == '-') {
4203 PL_tokenbuf[0] = '-';
4205 while (d < PL_bufend && SPACE_OR_TAB(*d))
4208 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4209 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4211 while (d < PL_bufend && SPACE_OR_TAB(*d))
4214 const char minus = (PL_tokenbuf[0] == '-');
4215 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4223 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4228 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4233 if (PL_oldoldbufptr == PL_last_lop)
4234 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4236 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4239 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4241 /* This hack is to get the ${} in the message. */
4243 yyerror("syntax error");
4246 OPERATOR(HASHBRACK);
4248 /* This hack serves to disambiguate a pair of curlies
4249 * as being a block or an anon hash. Normally, expectation
4250 * determines that, but in cases where we're not in a
4251 * position to expect anything in particular (like inside
4252 * eval"") we have to resolve the ambiguity. This code
4253 * covers the case where the first term in the curlies is a
4254 * quoted string. Most other cases need to be explicitly
4255 * disambiguated by prepending a "+" before the opening
4256 * curly in order to force resolution as an anon hash.
4258 * XXX should probably propagate the outer expectation
4259 * into eval"" to rely less on this hack, but that could
4260 * potentially break current behavior of eval"".
4264 if (*s == '\'' || *s == '"' || *s == '`') {
4265 /* common case: get past first string, handling escapes */
4266 for (t++; t < PL_bufend && *t != *s;)
4267 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4271 else if (*s == 'q') {
4274 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4277 /* skip q//-like construct */
4279 char open, close, term;
4282 while (t < PL_bufend && isSPACE(*t))
4284 /* check for q => */
4285 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4286 OPERATOR(HASHBRACK);
4290 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4294 for (t++; t < PL_bufend; t++) {
4295 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4297 else if (*t == open)
4301 for (t++; t < PL_bufend; t++) {
4302 if (*t == '\\' && t+1 < PL_bufend)
4304 else if (*t == close && --brackets <= 0)
4306 else if (*t == open)
4313 /* skip plain q word */
4314 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4317 else if (isALNUM_lazy_if(t,UTF)) {
4319 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4322 while (t < PL_bufend && isSPACE(*t))
4324 /* if comma follows first term, call it an anon hash */
4325 /* XXX it could be a comma expression with loop modifiers */
4326 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4327 || (*t == '=' && t[1] == '>')))
4328 OPERATOR(HASHBRACK);
4329 if (PL_expect == XREF)
4332 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4338 yylval.ival = CopLINE(PL_curcop);
4339 if (isSPACE(*s) || *s == '#')
4340 PL_copline = NOLINE; /* invalidate current command line number */
4345 if (PL_lex_brackets <= 0)
4346 yyerror("Unmatched right curly bracket");
4348 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4349 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4350 PL_lex_formbrack = 0;
4351 if (PL_lex_state == LEX_INTERPNORMAL) {
4352 if (PL_lex_brackets == 0) {
4353 if (PL_expect & XFAKEBRACK) {
4354 PL_expect &= XENUMMASK;
4355 PL_lex_state = LEX_INTERPEND;
4360 PL_thiswhite = newSVpvn("",0);
4361 sv_catpvn(PL_thiswhite,"}",1);
4364 return yylex(); /* ignore fake brackets */
4366 if (*s == '-' && s[1] == '>')
4367 PL_lex_state = LEX_INTERPENDMAYBE;
4368 else if (*s != '[' && *s != '{')
4369 PL_lex_state = LEX_INTERPEND;
4372 if (PL_expect & XFAKEBRACK) {
4373 PL_expect &= XENUMMASK;
4375 return yylex(); /* ignore fake brackets */
4377 start_force(PL_curforce);
4379 curmad('X', newSVpvn(s-1,1));
4380 CURMAD('_', PL_thiswhite);
4385 PL_thistoken = newSVpvn("",0);
4393 if (PL_expect == XOPERATOR) {
4394 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)