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++)
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, '"'))) {
740 for (t = s; !isSPACE(*t); t++) ;
743 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
745 if (*e != '\n' && *e != '\0')
746 return; /* false alarm */
752 const char * const cf = CopFILE(PL_curcop);
753 STRLEN tmplen = cf ? strlen(cf) : 0;
754 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
755 /* must copy *{"::_<(eval N)[oldfilename:L]"}
756 * to *{"::_<newfilename"} */
757 char smallbuf[256], smallbuf2[256];
758 char *tmpbuf, *tmpbuf2;
760 STRLEN tmplen2 = strlen(s);
761 if (tmplen + 3 < sizeof smallbuf)
764 Newx(tmpbuf, tmplen + 3, char);
765 if (tmplen2 + 3 < sizeof smallbuf2)
768 Newx(tmpbuf2, tmplen2 + 3, char);
769 tmpbuf[0] = tmpbuf2[0] = '_';
770 tmpbuf[1] = tmpbuf2[1] = '<';
771 memcpy(tmpbuf + 2, cf, ++tmplen);
772 memcpy(tmpbuf2 + 2, s, ++tmplen2);
774 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
776 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
778 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
779 /* adjust ${"::_<newfilename"} to store the new file name */
780 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
781 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
782 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
784 if (tmpbuf != smallbuf) Safefree(tmpbuf);
785 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
788 CopFILE_free(PL_curcop);
789 CopFILE_set(PL_curcop, s);
792 CopLINE_set(PL_curcop, atoi(n)-1);
796 /* skip space before PL_thistoken */
799 S_skipspace0(pTHX_ register char *s)
806 PL_thiswhite = newSVpvn("",0);
807 sv_catsv(PL_thiswhite, PL_skipwhite);
808 sv_free(PL_skipwhite);
811 PL_realtokenstart = s - SvPVX(PL_linestr);
815 /* skip space after PL_thistoken */
818 S_skipspace1(pTHX_ register char *s)
820 const char *start = s;
821 I32 startoff = start - SvPVX(PL_linestr);
826 start = SvPVX(PL_linestr) + startoff;
827 if (!PL_thistoken && PL_realtokenstart >= 0) {
828 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
829 PL_thistoken = newSVpvn(tstart, start - tstart);
831 PL_realtokenstart = -1;
834 PL_nextwhite = newSVpvn("",0);
835 sv_catsv(PL_nextwhite, PL_skipwhite);
836 sv_free(PL_skipwhite);
843 S_skipspace2(pTHX_ register char *s, SV **svp)
846 I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
847 I32 startoff = start - SvPVX(PL_linestr);
849 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
850 if (!PL_madskills || !svp)
852 start = SvPVX(PL_linestr) + startoff;
853 if (!PL_thistoken && PL_realtokenstart >= 0) {
854 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
855 PL_thistoken = newSVpvn(tstart, start - tstart);
856 PL_realtokenstart = -1;
860 *svp = newSVpvn("",0);
861 sv_setsv(*svp, PL_skipwhite);
862 sv_free(PL_skipwhite);
872 * Called to gobble the appropriate amount and type of whitespace.
873 * Skips comments as well.
877 S_skipspace(pTHX_ register char *s)
882 int startoff = s - SvPVX(PL_linestr);
885 sv_free(PL_skipwhite);
890 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
891 while (s < PL_bufend && SPACE_OR_TAB(*s))
901 SSize_t oldprevlen, oldoldprevlen;
902 SSize_t oldloplen = 0, oldunilen = 0;
903 while (s < PL_bufend && isSPACE(*s)) {
904 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
909 if (s < PL_bufend && *s == '#') {
910 while (s < PL_bufend && *s != '\n')
914 if (PL_in_eval && !PL_rsfp) {
921 /* only continue to recharge the buffer if we're at the end
922 * of the buffer, we're not reading from a source filter, and
923 * we're in normal lexing mode
925 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
926 PL_lex_state == LEX_FORMLINE)
933 /* try to recharge the buffer */
935 curoff = s - SvPVX(PL_linestr);
938 if ((s = filter_gets(PL_linestr, PL_rsfp,
939 (prevlen = SvCUR(PL_linestr)))) == NULL)
942 if (PL_madskills && curoff != startoff) {
944 PL_skipwhite = newSVpvn("",0);
945 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
949 /* mustn't throw out old stuff yet if madpropping */
950 SvCUR(PL_linestr) = curoff;
951 s = SvPVX(PL_linestr) + curoff;
953 if (curoff && s[-1] == '\n')
957 /* end of file. Add on the -p or -n magic */
958 /* XXX these shouldn't really be added here, can't set PL_faketokens */
962 ";}continue{print or die qq(-p destination: $!\\n);}");
965 ";}continue{print or die qq(-p destination: $!\\n);}");
967 PL_minus_n = PL_minus_p = 0;
969 else if (PL_minus_n) {
971 sv_catpvn(PL_linestr, ";}", 2);
973 sv_setpvn(PL_linestr, ";}", 2);
979 sv_catpvn(PL_linestr,";", 1);
981 sv_setpvn(PL_linestr,";", 1);
984 /* reset variables for next time we lex */
985 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
991 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
992 PL_last_lop = PL_last_uni = NULL;
994 /* Close the filehandle. Could be from -P preprocessor,
995 * STDIN, or a regular file. If we were reading code from
996 * STDIN (because the commandline held no -e or filename)
997 * then we don't close it, we reset it so the code can
998 * read from STDIN too.
1001 if (PL_preprocess && !PL_in_eval)
1002 (void)PerlProc_pclose(PL_rsfp);
1003 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1004 PerlIO_clearerr(PL_rsfp);
1006 (void)PerlIO_close(PL_rsfp);
1011 /* not at end of file, so we only read another line */
1012 /* make corresponding updates to old pointers, for yyerror() */
1013 oldprevlen = PL_oldbufptr - PL_bufend;
1014 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1016 oldunilen = PL_last_uni - PL_bufend;
1018 oldloplen = PL_last_lop - PL_bufend;
1019 PL_linestart = PL_bufptr = s + prevlen;
1020 PL_bufend = s + SvCUR(PL_linestr);
1022 PL_oldbufptr = s + oldprevlen;
1023 PL_oldoldbufptr = s + oldoldprevlen;
1025 PL_last_uni = s + oldunilen;
1027 PL_last_lop = s + oldloplen;
1030 /* debugger active and we're not compiling the debugger code,
1031 * so store the line into the debugger's array of lines
1033 if (PERLDB_LINE && PL_curstash != PL_debstash) {
1034 SV * const sv = newSV(0);
1036 sv_upgrade(sv, SVt_PVMG);
1037 sv_setpvn(sv,PL_bufptr,PL_bufend-PL_bufptr);
1040 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
1048 PL_skipwhite = newSVpvn("",0);
1049 curoff = s - SvPVX(PL_linestr);
1050 if (curoff - startoff)
1051 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1060 * Check the unary operators to ensure there's no ambiguity in how they're
1061 * used. An ambiguous piece of code would be:
1063 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1064 * the +5 is its argument.
1074 if (PL_oldoldbufptr != PL_last_uni)
1076 while (isSPACE(*PL_last_uni))
1078 for (s = PL_last_uni; isALNUM_lazy_if(s,UTF) || *s == '-'; s++)
1080 if ((t = strchr(s, '(')) && t < PL_bufptr)
1083 if (ckWARN_d(WARN_AMBIGUOUS)){
1084 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1085 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1086 (int)(s - PL_last_uni), PL_last_uni);
1091 * LOP : macro to build a list operator. Its behaviour has been replaced
1092 * with a subroutine, S_lop() for which LOP is just another name.
1095 #define LOP(f,x) return lop(f,x,s)
1099 * Build a list operator (or something that might be one). The rules:
1100 * - if we have a next token, then it's a list operator [why?]
1101 * - if the next thing is an opening paren, then it's a function
1102 * - else it's a list operator
1106 S_lop(pTHX_ I32 f, int x, char *s)
1113 PL_last_lop = PL_oldbufptr;
1114 PL_last_lop_op = (OPCODE)f;
1117 return REPORT(LSTOP);
1120 return REPORT(LSTOP);
1123 return REPORT(FUNC);
1126 return REPORT(FUNC);
1128 return REPORT(LSTOP);
1134 * Sets up for an eventual force_next(). start_force(0) basically does
1135 * an unshift, while start_force(-1) does a push. yylex removes items
1140 S_start_force(pTHX_ int where)
1144 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1145 where = PL_lasttoke;
1146 assert(PL_curforce < 0 || PL_curforce == where);
1147 if (PL_curforce != where) {
1148 for (i = PL_lasttoke; i > where; --i) {
1149 PL_nexttoke[i] = PL_nexttoke[i-1];
1153 if (PL_curforce < 0) /* in case of duplicate start_force() */
1154 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1155 PL_curforce = where;
1158 curmad('^', newSVpvn("",0));
1159 CURMAD('_', PL_nextwhite);
1164 S_curmad(pTHX_ char slot, SV *sv)
1170 if (PL_curforce < 0)
1171 where = &PL_thismad;
1173 where = &PL_nexttoke[PL_curforce].next_mad;
1176 sv_setpvn(sv, "", 0);
1179 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1181 else if (PL_encoding) {
1182 sv_recode_to_utf8(sv, PL_encoding);
1187 /* keep a slot open for the head of the list? */
1188 if (slot != '_' && *where && (*where)->mad_key == '^') {
1189 (*where)->mad_key = slot;
1190 sv_free((*where)->mad_val);
1191 (*where)->mad_val = (void*)sv;
1194 addmad(newMADsv(slot, sv), where, 0);
1197 # define start_force(where) NOOP
1198 # define curmad(slot, sv) NOOP
1203 * When the lexer realizes it knows the next token (for instance,
1204 * it is reordering tokens for the parser) then it can call S_force_next
1205 * to know what token to return the next time the lexer is called. Caller
1206 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1207 * and possibly PL_expect to ensure the lexer handles the token correctly.
1211 S_force_next(pTHX_ I32 type)
1215 if (PL_curforce < 0)
1216 start_force(PL_lasttoke);
1217 PL_nexttoke[PL_curforce].next_type = type;
1218 if (PL_lex_state != LEX_KNOWNEXT)
1219 PL_lex_defer = PL_lex_state;
1220 PL_lex_state = LEX_KNOWNEXT;
1221 PL_lex_expect = PL_expect;
1224 PL_nexttype[PL_nexttoke] = type;
1226 if (PL_lex_state != LEX_KNOWNEXT) {
1227 PL_lex_defer = PL_lex_state;
1228 PL_lex_expect = PL_expect;
1229 PL_lex_state = LEX_KNOWNEXT;
1235 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1238 SV * const sv = newSVpvn(start,len);
1239 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1246 * When the lexer knows the next thing is a word (for instance, it has
1247 * just seen -> and it knows that the next char is a word char, then
1248 * it calls S_force_word to stick the next word into the PL_next lookahead.
1251 * char *start : buffer position (must be within PL_linestr)
1252 * int token : PL_next will be this type of bare word (e.g., METHOD,WORD)
1253 * int check_keyword : if true, Perl checks to make sure the word isn't
1254 * a keyword (do this if the word is a label, e.g. goto FOO)
1255 * int allow_pack : if true, : characters will also be allowed (require,
1256 * use, etc. do this)
1257 * int allow_initial_tick : used by the "sub" lexer only.
1261 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1267 start = SKIPSPACE1(start);
1269 if (isIDFIRST_lazy_if(s,UTF) ||
1270 (allow_pack && *s == ':') ||
1271 (allow_initial_tick && *s == '\'') )
1273 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1274 if (check_keyword && keyword(PL_tokenbuf, len))
1276 start_force(PL_curforce);
1278 curmad('X', newSVpvn(start,s-start));
1279 if (token == METHOD) {
1284 PL_expect = XOPERATOR;
1287 NEXTVAL_NEXTTOKE.opval
1288 = (OP*)newSVOP(OP_CONST,0,
1289 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1290 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1298 * Called when the lexer wants $foo *foo &foo etc, but the program
1299 * text only contains the "foo" portion. The first argument is a pointer
1300 * to the "foo", and the second argument is the type symbol to prefix.
1301 * Forces the next token to be a "WORD".
1302 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1306 S_force_ident(pTHX_ register const char *s, int kind)
1310 const STRLEN len = strlen(s);
1311 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1312 start_force(PL_curforce);
1313 NEXTVAL_NEXTTOKE.opval = o;
1316 o->op_private = OPpCONST_ENTERED;
1317 /* XXX see note in pp_entereval() for why we forgo typo
1318 warnings if the symbol must be introduced in an eval.
1320 gv_fetchpvn_flags(s, len,
1321 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1323 kind == '$' ? SVt_PV :
1324 kind == '@' ? SVt_PVAV :
1325 kind == '%' ? SVt_PVHV :
1333 Perl_str_to_version(pTHX_ SV *sv)
1338 const char *start = SvPV_const(sv,len);
1339 const char * const end = start + len;
1340 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1341 while (start < end) {
1345 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1350 retval += ((NV)n)/nshift;
1359 * Forces the next token to be a version number.
1360 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1361 * and if "guessing" is TRUE, then no new token is created (and the caller
1362 * must use an alternative parsing method).
1366 S_force_version(pTHX_ char *s, int guessing)
1372 I32 startoff = s - SvPVX(PL_linestr);
1381 while (isDIGIT(*d) || *d == '_' || *d == '.')
1385 start_force(PL_curforce);
1386 curmad('X', newSVpvn(s,d-s));
1389 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1391 s = scan_num(s, &yylval);
1392 version = yylval.opval;
1393 ver = cSVOPx(version)->op_sv;
1394 if (SvPOK(ver) && !SvNIOK(ver)) {
1395 SvUPGRADE(ver, SVt_PVNV);
1396 SvNV_set(ver, str_to_version(ver));
1397 SvNOK_on(ver); /* hint that it is a version */
1400 else if (guessing) {
1403 sv_free(PL_nextwhite); /* let next token collect whitespace */
1405 s = SvPVX(PL_linestr) + startoff;
1413 if (PL_madskills && !version) {
1414 sv_free(PL_nextwhite); /* let next token collect whitespace */
1416 s = SvPVX(PL_linestr) + startoff;
1419 /* NOTE: The parser sees the package name and the VERSION swapped */
1420 start_force(PL_curforce);
1421 NEXTVAL_NEXTTOKE.opval = version;
1429 * Tokenize a quoted string passed in as an SV. It finds the next
1430 * chunk, up to end of string or a backslash. It may make a new
1431 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1436 S_tokeq(pTHX_ SV *sv)
1440 register char *send;
1448 s = SvPV_force(sv, len);
1449 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1452 while (s < send && *s != '\\')
1457 if ( PL_hints & HINT_NEW_STRING ) {
1458 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1464 if (s + 1 < send && (s[1] == '\\'))
1465 s++; /* all that, just for this */
1470 SvCUR_set(sv, d - SvPVX_const(sv));
1472 if ( PL_hints & HINT_NEW_STRING )
1473 return new_constant(NULL, 0, "q", sv, pv, "q");
1478 * Now come three functions related to double-quote context,
1479 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1480 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1481 * interact with PL_lex_state, and create fake ( ... ) argument lists
1482 * to handle functions and concatenation.
1483 * They assume that whoever calls them will be setting up a fake
1484 * join call, because each subthing puts a ',' after it. This lets
1487 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1489 * (I'm not sure whether the spurious commas at the end of lcfirst's
1490 * arguments and join's arguments are created or not).
1495 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1497 * Pattern matching will set PL_lex_op to the pattern-matching op to
1498 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1500 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1502 * Everything else becomes a FUNC.
1504 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1505 * had an OP_CONST or OP_READLINE). This just sets us up for a
1506 * call to S_sublex_push().
1510 S_sublex_start(pTHX)
1513 register const I32 op_type = yylval.ival;
1515 if (op_type == OP_NULL) {
1516 yylval.opval = PL_lex_op;
1520 if (op_type == OP_CONST || op_type == OP_READLINE) {
1521 SV *sv = tokeq(PL_lex_stuff);
1523 if (SvTYPE(sv) == SVt_PVIV) {
1524 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1526 const char * const p = SvPV_const(sv, len);
1527 SV * const nsv = newSVpvn(p, len);
1533 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1534 PL_lex_stuff = NULL;
1535 /* Allow <FH> // "foo" */
1536 if (op_type == OP_READLINE)
1537 PL_expect = XTERMORDORDOR;
1541 PL_sublex_info.super_state = PL_lex_state;
1542 PL_sublex_info.sub_inwhat = op_type;
1543 PL_sublex_info.sub_op = PL_lex_op;
1544 PL_lex_state = LEX_INTERPPUSH;
1548 yylval.opval = PL_lex_op;
1558 * Create a new scope to save the lexing state. The scope will be
1559 * ended in S_sublex_done. Returns a '(', starting the function arguments
1560 * to the uc, lc, etc. found before.
1561 * Sets PL_lex_state to LEX_INTERPCONCAT.
1570 PL_lex_state = PL_sublex_info.super_state;
1571 SAVEI32(PL_lex_dojoin);
1572 SAVEI32(PL_lex_brackets);
1573 SAVEI32(PL_lex_casemods);
1574 SAVEI32(PL_lex_starts);
1575 SAVEI32(PL_lex_state);
1576 SAVEVPTR(PL_lex_inpat);
1577 SAVEI32(PL_lex_inwhat);
1578 SAVECOPLINE(PL_curcop);
1579 SAVEPPTR(PL_bufptr);
1580 SAVEPPTR(PL_bufend);
1581 SAVEPPTR(PL_oldbufptr);
1582 SAVEPPTR(PL_oldoldbufptr);
1583 SAVEPPTR(PL_last_lop);
1584 SAVEPPTR(PL_last_uni);
1585 SAVEPPTR(PL_linestart);
1586 SAVESPTR(PL_linestr);
1587 SAVEGENERICPV(PL_lex_brackstack);
1588 SAVEGENERICPV(PL_lex_casestack);
1590 PL_linestr = PL_lex_stuff;
1591 PL_lex_stuff = NULL;
1593 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1594 = SvPVX(PL_linestr);
1595 PL_bufend += SvCUR(PL_linestr);
1596 PL_last_lop = PL_last_uni = NULL;
1597 SAVEFREESV(PL_linestr);
1599 PL_lex_dojoin = FALSE;
1600 PL_lex_brackets = 0;
1601 Newx(PL_lex_brackstack, 120, char);
1602 Newx(PL_lex_casestack, 12, char);
1603 PL_lex_casemods = 0;
1604 *PL_lex_casestack = '\0';
1606 PL_lex_state = LEX_INTERPCONCAT;
1607 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1609 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1610 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1611 PL_lex_inpat = PL_sublex_info.sub_op;
1613 PL_lex_inpat = NULL;
1620 * Restores lexer state after a S_sublex_push.
1627 if (!PL_lex_starts++) {
1628 SV * const sv = newSVpvs("");
1629 if (SvUTF8(PL_linestr))
1631 PL_expect = XOPERATOR;
1632 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1636 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1637 PL_lex_state = LEX_INTERPCASEMOD;
1641 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1642 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1643 PL_linestr = PL_lex_repl;
1645 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1646 PL_bufend += SvCUR(PL_linestr);
1647 PL_last_lop = PL_last_uni = NULL;
1648 SAVEFREESV(PL_linestr);
1649 PL_lex_dojoin = FALSE;
1650 PL_lex_brackets = 0;
1651 PL_lex_casemods = 0;
1652 *PL_lex_casestack = '\0';
1654 if (SvEVALED(PL_lex_repl)) {
1655 PL_lex_state = LEX_INTERPNORMAL;
1657 /* we don't clear PL_lex_repl here, so that we can check later
1658 whether this is an evalled subst; that means we rely on the
1659 logic to ensure sublex_done() is called again only via the
1660 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1663 PL_lex_state = LEX_INTERPCONCAT;
1673 PL_endwhite = newSVpvn("",0);
1674 sv_catsv(PL_endwhite, PL_thiswhite);
1678 sv_setpvn(PL_thistoken,"",0);
1680 PL_realtokenstart = -1;
1684 PL_bufend = SvPVX(PL_linestr);
1685 PL_bufend += SvCUR(PL_linestr);
1686 PL_expect = XOPERATOR;
1687 PL_sublex_info.sub_inwhat = 0;
1695 Extracts a pattern, double-quoted string, or transliteration. This
1698 It looks at lex_inwhat and PL_lex_inpat to find out whether it's
1699 processing a pattern (PL_lex_inpat is true), a transliteration
1700 (lex_inwhat & OP_TRANS is true), or a double-quoted string.
1702 Returns a pointer to the character scanned up to. Iff this is
1703 advanced from the start pointer supplied (ie if anything was
1704 successfully parsed), will leave an OP for the substring scanned
1705 in yylval. Caller must intuit reason for not parsing further
1706 by looking at the next characters herself.
1710 double-quoted style: \r and \n
1711 regexp special ones: \D \s
1713 backrefs: \1 (deprecated in substitution replacements)
1714 case and quoting: \U \Q \E
1715 stops on @ and $, but not for $ as tail anchor
1717 In transliterations:
1718 characters are VERY literal, except for - not at the start or end
1719 of the string, which indicates a range. scan_const expands the
1720 range to the full set of intermediate characters.
1722 In double-quoted strings:
1724 double-quoted style: \r and \n
1726 backrefs: \1 (deprecated)
1727 case and quoting: \U \Q \E
1730 scan_const does *not* construct ops to handle interpolated strings.
1731 It stops processing as soon as it finds an embedded $ or @ variable
1732 and leaves it to the caller to work out what's going on.
1734 @ in pattern could be: @foo, @{foo}, @$foo, @'foo, @::foo.
1736 $ in pattern could be $foo or could be tail anchor. Assumption:
1737 it's a tail anchor if $ is the last thing in the string, or if it's
1738 followed by one of ")| \n\t"
1740 \1 (backreferences) are turned into $1
1742 The structure of the code is
1743 while (there's a character to process) {
1744 handle transliteration ranges
1745 skip regexp comments
1746 skip # initiated comments in //x patterns
1747 check for embedded @foo
1748 check for embedded scalars
1750 leave intact backslashes from leave (below)
1751 deprecate \1 in strings and sub replacements
1752 handle string-changing backslashes \l \U \Q \E, etc.
1753 switch (what was escaped) {
1754 handle - in a transliteration (becomes a literal -)
1755 handle \132 octal characters
1756 handle 0x15 hex characters
1757 handle \cV (control V)
1758 handle printf backslashes (\f, \r, \n, etc)
1760 } (end if backslash)
1761 } (end while character to read)
1766 S_scan_const(pTHX_ char *start)
1769 register char *send = PL_bufend; /* end of the constant */
1770 SV *sv = newSV(send - start); /* sv for the constant */
1771 register char *s = start; /* start of the constant */
1772 register char *d = SvPVX(sv); /* destination for copies */
1773 bool dorange = FALSE; /* are we in a translit range? */
1774 bool didrange = FALSE; /* did we just finish a range? */
1775 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1776 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1779 UV literal_endpoint = 0;
1782 const char * const leaveit = /* set of acceptably-backslashed characters */
1784 ? "\\.^$@AGZdDwWsSbBpPXC+*?|()-nrtfeaxz0123456789[{]} \t\n\r\f\v#"
1787 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1788 /* If we are doing a trans and we know we want UTF8 set expectation */
1789 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1790 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1794 while (s < send || dorange) {
1795 /* get transliterations out of the way (they're most literal) */
1796 if (PL_lex_inwhat == OP_TRANS) {
1797 /* expand a range A-Z to the full set of characters. AIE! */
1799 I32 i; /* current expanded character */
1800 I32 min; /* first character in range */
1801 I32 max; /* last character in range */
1804 char * const c = (char*)utf8_hop((U8*)d, -1);
1808 *c = (char)UTF_TO_NATIVE(0xff);
1809 /* mark the range as done, and continue */
1815 i = d - SvPVX_const(sv); /* remember current offset */
1816 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1817 d = SvPVX(sv) + i; /* refresh d after realloc */
1818 d -= 2; /* eat the first char and the - */
1820 min = (U8)*d; /* first char in range */
1821 max = (U8)d[1]; /* last char in range */
1825 "Invalid range \"%c-%c\" in transliteration operator",
1826 (char)min, (char)max);
1830 if (literal_endpoint == 2 &&
1831 ((isLOWER(min) && isLOWER(max)) ||
1832 (isUPPER(min) && isUPPER(max)))) {
1834 for (i = min; i <= max; i++)
1836 *d++ = NATIVE_TO_NEED(has_utf8,i);
1838 for (i = min; i <= max; i++)
1840 *d++ = NATIVE_TO_NEED(has_utf8,i);
1845 for (i = min; i <= max; i++)
1848 /* mark the range as done, and continue */
1852 literal_endpoint = 0;
1857 /* range begins (ignore - as first or last char) */
1858 else if (*s == '-' && s+1 < send && s != start) {
1860 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1863 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1873 literal_endpoint = 0;
1878 /* if we get here, we're not doing a transliteration */
1880 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
1881 except for the last char, which will be done separately. */
1882 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
1884 while (s+1 < send && *s != ')')
1885 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1887 else if (s[2] == '{' /* This should match regcomp.c */
1888 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
1891 char *regparse = s + (s[2] == '{' ? 3 : 4);
1894 while (count && (c = *regparse)) {
1895 if (c == '\\' && regparse[1])
1903 if (*regparse != ')')
1904 regparse--; /* Leave one char for continuation. */
1905 while (s < regparse)
1906 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1910 /* likewise skip #-initiated comments in //x patterns */
1911 else if (*s == '#' && PL_lex_inpat &&
1912 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
1913 while (s+1 < send && *s != '\n')
1914 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1917 /* check for embedded arrays
1918 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
1920 else if (*s == '@' && s[1]
1921 && (isALNUM_lazy_if(s+1,UTF) || strchr(":'{$+-", s[1])))
1924 /* check for embedded scalars. only stop if we're sure it's a
1927 else if (*s == '$') {
1928 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
1930 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
1931 break; /* in regexp, $ might be tail anchor */
1934 /* End of else if chain - OP_TRANS rejoin rest */
1937 if (*s == '\\' && s+1 < send) {
1940 /* some backslashes we leave behind */
1941 if (*leaveit && *s && strchr(leaveit, *s)) {
1942 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
1943 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
1947 /* deprecate \1 in strings and substitution replacements */
1948 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
1949 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
1951 if (ckWARN(WARN_SYNTAX))
1952 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
1957 /* string-change backslash escapes */
1958 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
1963 /* if we get here, it's either a quoted -, or a digit */
1966 /* quoted - in transliterations */
1968 if (PL_lex_inwhat == OP_TRANS) {
1978 Perl_warner(aTHX_ packWARN(WARN_MISC),
1979 "Unrecognized escape \\%c passed through",
1981 /* default action is to copy the quoted character */
1982 goto default_action;
1985 /* \132 indicates an octal constant */
1986 case '0': case '1': case '2': case '3':
1987 case '4': case '5': case '6': case '7':
1991 uv = grok_oct(s, &len, &flags, NULL);
1994 goto NUM_ESCAPE_INSERT;
1996 /* \x24 indicates a hex constant */
2000 char* const e = strchr(s, '}');
2001 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2002 PERL_SCAN_DISALLOW_PREFIX;
2007 yyerror("Missing right brace on \\x{}");
2011 uv = grok_hex(s, &len, &flags, NULL);
2017 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2018 uv = grok_hex(s, &len, &flags, NULL);
2024 /* Insert oct or hex escaped character.
2025 * There will always enough room in sv since such
2026 * escapes will be longer than any UTF-8 sequence
2027 * they can end up as. */
2029 /* We need to map to chars to ASCII before doing the tests
2032 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2033 if (!has_utf8 && uv > 255) {
2034 /* Might need to recode whatever we have
2035 * accumulated so far if it contains any
2038 * (Can't we keep track of that and avoid
2039 * this rescan? --jhi)
2043 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2044 if (!NATIVE_IS_INVARIANT(*c)) {
2049 const STRLEN offset = d - SvPVX_const(sv);
2051 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2055 while (src >= (const U8 *)SvPVX_const(sv)) {
2056 if (!NATIVE_IS_INVARIANT(*src)) {
2057 const U8 ch = NATIVE_TO_ASCII(*src);
2058 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2059 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2069 if (has_utf8 || uv > 255) {
2070 d = (char*)uvchr_to_utf8((U8*)d, uv);
2072 if (PL_lex_inwhat == OP_TRANS &&
2073 PL_sublex_info.sub_op) {
2074 PL_sublex_info.sub_op->op_private |=
2075 (PL_lex_repl ? OPpTRANS_FROM_UTF
2088 /* \N{LATIN SMALL LETTER A} is a named character */
2092 char* e = strchr(s, '}');
2098 yyerror("Missing right brace on \\N{}");
2102 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2104 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2105 PERL_SCAN_DISALLOW_PREFIX;
2108 uv = grok_hex(s, &len, &flags, NULL);
2110 goto NUM_ESCAPE_INSERT;
2112 res = newSVpvn(s + 1, e - s - 1);
2113 res = new_constant( NULL, 0, "charnames",
2114 res, NULL, "\\N{...}" );
2116 sv_utf8_upgrade(res);
2117 str = SvPV_const(res,len);
2118 #ifdef EBCDIC_NEVER_MIND
2119 /* charnames uses pack U and that has been
2120 * recently changed to do the below uni->native
2121 * mapping, so this would be redundant (and wrong,
2122 * the code point would be doubly converted).
2123 * But leave this in just in case the pack U change
2124 * gets revoked, but the semantics is still
2125 * desireable for charnames. --jhi */
2127 UV uv = utf8_to_uvchr((const U8*)str, 0);
2130 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2132 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2133 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2134 str = SvPV_const(res, len);
2138 if (!has_utf8 && SvUTF8(res)) {
2139 const char * const ostart = SvPVX_const(sv);
2140 SvCUR_set(sv, d - ostart);
2143 sv_utf8_upgrade(sv);
2144 /* this just broke our allocation above... */
2145 SvGROW(sv, (STRLEN)(send - start));
2146 d = SvPVX(sv) + SvCUR(sv);
2149 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2150 const char * const odest = SvPVX_const(sv);
2152 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2153 d = SvPVX(sv) + (d - odest);
2155 Copy(str, d, len, char);
2162 yyerror("Missing braces on \\N{}");
2165 /* \c is a control character */
2174 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2177 yyerror("Missing control char name in \\c");
2181 /* printf-style backslashes, formfeeds, newlines, etc */
2183 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2186 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2189 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2192 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2195 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2198 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2201 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2207 } /* end if (backslash) */
2214 /* If we started with encoded form, or already know we want it
2215 and then encode the next character */
2216 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2218 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2219 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2222 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2223 const STRLEN off = d - SvPVX_const(sv);
2224 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2226 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2230 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2232 } /* while loop to process each character */
2234 /* terminate the string and set up the sv */
2236 SvCUR_set(sv, d - SvPVX_const(sv));
2237 if (SvCUR(sv) >= SvLEN(sv))
2238 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2241 if (PL_encoding && !has_utf8) {
2242 sv_recode_to_utf8(sv, PL_encoding);
2248 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2249 PL_sublex_info.sub_op->op_private |=
2250 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2254 /* shrink the sv if we allocated more than we used */
2255 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2256 SvPV_shrink_to_cur(sv);
2259 /* return the substring (via yylval) only if we parsed anything */
2260 if (s > PL_bufptr) {
2261 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2262 sv = new_constant(start, s - start, (PL_lex_inpat ? "qr" : "q"),
2264 ( PL_lex_inwhat == OP_TRANS
2266 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2269 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2276 * Returns TRUE if there's more to the expression (e.g., a subscript),
2279 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2281 * ->[ and ->{ return TRUE
2282 * { and [ outside a pattern are always subscripts, so return TRUE
2283 * if we're outside a pattern and it's not { or [, then return FALSE
2284 * if we're in a pattern and the first char is a {
2285 * {4,5} (any digits around the comma) returns FALSE
2286 * if we're in a pattern and the first char is a [
2288 * [SOMETHING] has a funky algorithm to decide whether it's a
2289 * character class or not. It has to deal with things like
2290 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2291 * anything else returns TRUE
2294 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2297 S_intuit_more(pTHX_ register char *s)
2300 if (PL_lex_brackets)
2302 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2304 if (*s != '{' && *s != '[')
2309 /* In a pattern, so maybe we have {n,m}. */
2326 /* On the other hand, maybe we have a character class */
2329 if (*s == ']' || *s == '^')
2332 /* this is terrifying, and it works */
2333 int weight = 2; /* let's weigh the evidence */
2335 unsigned char un_char = 255, last_un_char;
2336 const char * const send = strchr(s,']');
2337 char tmpbuf[sizeof PL_tokenbuf * 4];
2339 if (!send) /* has to be an expression */
2342 Zero(seen,256,char);
2345 else if (isDIGIT(*s)) {
2347 if (isDIGIT(s[1]) && s[2] == ']')
2353 for (; s < send; s++) {
2354 last_un_char = un_char;
2355 un_char = (unsigned char)*s;
2360 weight -= seen[un_char] * 10;
2361 if (isALNUM_lazy_if(s+1,UTF)) {
2363 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2364 len = (int)strlen(tmpbuf);
2365 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2370 else if (*s == '$' && s[1] &&
2371 strchr("[#!%*<>()-=",s[1])) {
2372 if (/*{*/ strchr("])} =",s[2]))
2381 if (strchr("wds]",s[1]))
2383 else if (seen['\''] || seen['"'])
2385 else if (strchr("rnftbxcav",s[1]))
2387 else if (isDIGIT(s[1])) {
2389 while (s[1] && isDIGIT(s[1]))
2399 if (strchr("aA01! ",last_un_char))
2401 if (strchr("zZ79~",s[1]))
2403 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2404 weight -= 5; /* cope with negative subscript */
2407 if (!isALNUM(last_un_char)
2408 && !(last_un_char == '$' || last_un_char == '@'
2409 || last_un_char == '&')
2410 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2415 if (keyword(tmpbuf, d - tmpbuf))
2418 if (un_char == last_un_char + 1)
2420 weight -= seen[un_char];
2425 if (weight >= 0) /* probably a character class */
2435 * Does all the checking to disambiguate
2437 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2438 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2440 * First argument is the stuff after the first token, e.g. "bar".
2442 * Not a method if bar is a filehandle.
2443 * Not a method if foo is a subroutine prototyped to take a filehandle.
2444 * Not a method if it's really "Foo $bar"
2445 * Method if it's "foo $bar"
2446 * Not a method if it's really "print foo $bar"
2447 * Method if it's really "foo package::" (interpreted as package->foo)
2448 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2449 * Not a method if bar is a filehandle or package, but is quoted with
2454 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2457 char *s = start + (*start == '$');
2458 char tmpbuf[sizeof PL_tokenbuf];
2466 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2470 const char *proto = SvPVX_const(cv);
2481 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2482 /* start is the beginning of the possible filehandle/object,
2483 * and s is the end of it
2484 * tmpbuf is a copy of it
2487 if (*start == '$') {
2488 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2491 len = start - SvPVX(PL_linestr);
2495 start = SvPVX(PL_linestr) + len;
2499 return *s == '(' ? FUNCMETH : METHOD;
2501 if (!keyword(tmpbuf, len)) {
2502 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2506 soff = s - SvPVX(PL_linestr);
2510 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2511 if (indirgv && GvCVu(indirgv))
2513 /* filehandle or package name makes it a method */
2514 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, FALSE)) {
2516 soff = s - SvPVX(PL_linestr);
2519 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2520 return 0; /* no assumptions -- "=>" quotes bearword */
2522 start_force(PL_curforce);
2523 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2524 newSVpvn(tmpbuf,len));
2525 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2527 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2532 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2534 return *s == '(' ? FUNCMETH : METHOD;
2542 * Return a string of Perl code to load the debugger. If PERL5DB
2543 * is set, it will return the contents of that, otherwise a
2544 * compile-time require of perl5db.pl.
2552 const char * const pdb = PerlEnv_getenv("PERL5DB");
2556 SETERRNO(0,SS_NORMAL);
2557 return "BEGIN { require 'perl5db.pl' }";
2563 /* Encoded script support. filter_add() effectively inserts a
2564 * 'pre-processing' function into the current source input stream.
2565 * Note that the filter function only applies to the current source file
2566 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2568 * The datasv parameter (which may be NULL) can be used to pass
2569 * private data to this instance of the filter. The filter function
2570 * can recover the SV using the FILTER_DATA macro and use it to
2571 * store private buffers and state information.
2573 * The supplied datasv parameter is upgraded to a PVIO type
2574 * and the IoDIRP/IoANY field is used to store the function pointer,
2575 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2576 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2577 * private use must be set using malloc'd pointers.
2581 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2587 if (!PL_rsfp_filters)
2588 PL_rsfp_filters = newAV();
2591 SvUPGRADE(datasv, SVt_PVIO);
2592 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2593 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2594 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2595 IoANY(datasv), SvPV_nolen(datasv)));
2596 av_unshift(PL_rsfp_filters, 1);
2597 av_store(PL_rsfp_filters, 0, datasv) ;
2602 /* Delete most recently added instance of this filter function. */
2604 Perl_filter_del(pTHX_ filter_t funcp)
2610 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p", FPTR2DPTR(XPVIO *, funcp)));
2612 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2614 /* if filter is on top of stack (usual case) just pop it off */
2615 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2616 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2617 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2618 IoANY(datasv) = (void *)NULL;
2619 sv_free(av_pop(PL_rsfp_filters));
2623 /* we need to search for the correct entry and clear it */
2624 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2628 /* Invoke the idxth filter function for the current rsfp. */
2629 /* maxlen 0 = read one text line */
2631 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2636 /* This API is bad. It should have been using unsigned int for maxlen.
2637 Not sure if we want to change the API, but if not we should sanity
2638 check the value here. */
2639 const unsigned int correct_length
2648 if (!PL_rsfp_filters)
2650 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2651 /* Provide a default input filter to make life easy. */
2652 /* Note that we append to the line. This is handy. */
2653 DEBUG_P(PerlIO_printf(Perl_debug_log,
2654 "filter_read %d: from rsfp\n", idx));
2655 if (correct_length) {
2658 const int old_len = SvCUR(buf_sv);
2660 /* ensure buf_sv is large enough */
2661 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2662 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2663 correct_length)) <= 0) {
2664 if (PerlIO_error(PL_rsfp))
2665 return -1; /* error */
2667 return 0 ; /* end of file */
2669 SvCUR_set(buf_sv, old_len + len) ;
2672 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2673 if (PerlIO_error(PL_rsfp))
2674 return -1; /* error */
2676 return 0 ; /* end of file */
2679 return SvCUR(buf_sv);
2681 /* Skip this filter slot if filter has been deleted */
2682 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2683 DEBUG_P(PerlIO_printf(Perl_debug_log,
2684 "filter_read %d: skipped (filter deleted)\n",
2686 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2688 /* Get function pointer hidden within datasv */
2689 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2690 DEBUG_P(PerlIO_printf(Perl_debug_log,
2691 "filter_read %d: via function %p (%s)\n",
2692 idx, datasv, SvPV_nolen_const(datasv)));
2693 /* Call function. The function is expected to */
2694 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2695 /* Return: <0:error, =0:eof, >0:not eof */
2696 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2700 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2703 #ifdef PERL_CR_FILTER
2704 if (!PL_rsfp_filters) {
2705 filter_add(S_cr_textfilter,NULL);
2708 if (PL_rsfp_filters) {
2710 SvCUR_set(sv, 0); /* start with empty line */
2711 if (FILTER_READ(0, sv, 0) > 0)
2712 return ( SvPVX(sv) ) ;
2717 return (sv_gets(sv, fp, append));
2721 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2726 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2730 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2731 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2733 return GvHV(gv); /* Foo:: */
2736 /* use constant CLASS => 'MyClass' */
2737 if ((gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV))) {
2739 if (GvCV(gv) && (sv = cv_const_sv(GvCV(gv)))) {
2740 pkgname = SvPV_nolen_const(sv);
2744 return gv_stashpv(pkgname, FALSE);
2750 * The intent of this yylex wrapper is to minimize the changes to the
2751 * tokener when we aren't interested in collecting madprops. It remains
2752 * to be seen how successful this strategy will be...
2759 char *s = PL_bufptr;
2761 /* make sure PL_thiswhite is initialized */
2765 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2766 if (PL_pending_ident)
2767 return S_pending_ident(aTHX);
2769 /* previous token ate up our whitespace? */
2770 if (!PL_lasttoke && PL_nextwhite) {
2771 PL_thiswhite = PL_nextwhite;
2775 /* isolate the token, and figure out where it is without whitespace */
2776 PL_realtokenstart = -1;
2780 assert(PL_curforce < 0);
2782 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2783 if (!PL_thistoken) {
2784 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2785 PL_thistoken = newSVpvn("",0);
2787 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2788 PL_thistoken = newSVpvn(tstart, s - tstart);
2791 if (PL_thismad) /* install head */
2792 CURMAD('X', PL_thistoken);
2795 /* last whitespace of a sublex? */
2796 if (optype == ')' && PL_endwhite) {
2797 CURMAD('X', PL_endwhite);
2802 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2803 if (!PL_thiswhite && !PL_endwhite && !optype) {
2804 sv_free(PL_thistoken);
2809 /* put off final whitespace till peg */
2810 if (optype == ';' && !PL_rsfp) {
2811 PL_nextwhite = PL_thiswhite;
2814 else if (PL_thisopen) {
2815 CURMAD('q', PL_thisopen);
2817 sv_free(PL_thistoken);
2821 /* Store actual token text as madprop X */
2822 CURMAD('X', PL_thistoken);
2826 /* add preceding whitespace as madprop _ */
2827 CURMAD('_', PL_thiswhite);
2831 /* add quoted material as madprop = */
2832 CURMAD('=', PL_thisstuff);
2836 /* add terminating quote as madprop Q */
2837 CURMAD('Q', PL_thisclose);
2841 /* special processing based on optype */
2845 /* opval doesn't need a TOKEN since it can already store mp */
2856 append_madprops(PL_thismad, yylval.opval, 0);
2864 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
2873 /* remember any fake bracket that lexer is about to discard */
2874 if (PL_lex_brackets == 1 &&
2875 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
2878 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2881 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
2882 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2885 break; /* don't bother looking for trailing comment */
2894 /* attach a trailing comment to its statement instead of next token */
2898 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
2900 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
2902 if (*s == '\n' || *s == '#') {
2903 while (s < PL_bufend && *s != '\n')
2907 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
2908 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
2925 /* Create new token struct. Note: opvals return early above. */
2926 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
2933 S_tokenize_use(pTHX_ int is_use, char *s) {
2935 if (PL_expect != XSTATE)
2936 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
2937 is_use ? "use" : "no"));
2939 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
2940 s = force_version(s, TRUE);
2941 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
2942 start_force(PL_curforce);
2943 NEXTVAL_NEXTTOKE.opval = NULL;
2946 else if (*s == 'v') {
2947 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2948 s = force_version(s, FALSE);
2952 s = force_word(s,WORD,FALSE,TRUE,FALSE);
2953 s = force_version(s, FALSE);
2955 yylval.ival = is_use;
2959 static const char* const exp_name[] =
2960 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
2961 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
2968 Works out what to call the token just pulled out of the input
2969 stream. The yacc parser takes care of taking the ops we return and
2970 stitching them into a tree.
2976 if read an identifier
2977 if we're in a my declaration
2978 croak if they tried to say my($foo::bar)
2979 build the ops for a my() declaration
2980 if it's an access to a my() variable
2981 are we in a sort block?
2982 croak if my($a); $a <=> $b
2983 build ops for access to a my() variable
2984 if in a dq string, and they've said @foo and we can't find @foo
2986 build ops for a bareword
2987 if we already built the token before, use it.
2992 #pragma segment Perl_yylex
2998 register char *s = PL_bufptr;
3004 SV* tmp = newSVpvs("");
3005 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3006 (IV)CopLINE(PL_curcop),
3007 lex_state_names[PL_lex_state],
3008 exp_name[PL_expect],
3009 pv_display(tmp, s, strlen(s), 0, 60));
3012 /* check if there's an identifier for us to look at */
3013 if (PL_pending_ident)
3014 return REPORT(S_pending_ident(aTHX));
3016 /* no identifier pending identification */
3018 switch (PL_lex_state) {
3020 case LEX_NORMAL: /* Some compilers will produce faster */
3021 case LEX_INTERPNORMAL: /* code if we comment these out. */
3025 /* when we've already built the next token, just pull it out of the queue */
3029 yylval = PL_nexttoke[PL_lasttoke].next_val;
3031 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3032 PL_nexttoke[PL_lasttoke].next_mad = 0;
3033 if (PL_thismad && PL_thismad->mad_key == '_') {
3034 PL_thiswhite = (SV*)PL_thismad->mad_val;
3035 PL_thismad->mad_val = 0;
3036 mad_free(PL_thismad);
3041 PL_lex_state = PL_lex_defer;
3042 PL_expect = PL_lex_expect;
3043 PL_lex_defer = LEX_NORMAL;
3044 if (!PL_nexttoke[PL_lasttoke].next_type)
3049 yylval = PL_nextval[PL_nexttoke];
3051 PL_lex_state = PL_lex_defer;
3052 PL_expect = PL_lex_expect;
3053 PL_lex_defer = LEX_NORMAL;
3057 /* FIXME - can these be merged? */
3058 return(PL_nexttoke[PL_lasttoke].next_type);
3060 return REPORT(PL_nexttype[PL_nexttoke]);
3063 /* interpolated case modifiers like \L \U, including \Q and \E.
3064 when we get here, PL_bufptr is at the \
3066 case LEX_INTERPCASEMOD:
3068 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3069 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3071 /* handle \E or end of string */
3072 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3074 if (PL_lex_casemods) {
3075 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3076 PL_lex_casestack[PL_lex_casemods] = '\0';
3078 if (PL_bufptr != PL_bufend
3079 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3081 PL_lex_state = LEX_INTERPCONCAT;
3084 PL_thistoken = newSVpvn("\\E",2);
3090 while (PL_bufptr != PL_bufend &&
3091 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3093 PL_thiswhite = newSVpvn("",0);
3094 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3098 if (PL_bufptr != PL_bufend)
3101 PL_lex_state = LEX_INTERPCONCAT;
3105 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3106 "### Saw case modifier\n"); });
3108 if (s[1] == '\\' && s[2] == 'E') {
3111 PL_thiswhite = newSVpvn("",0);
3112 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3115 PL_lex_state = LEX_INTERPCONCAT;
3120 if (!PL_madskills) /* when just compiling don't need correct */
3121 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3122 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3123 if ((*s == 'L' || *s == 'U') &&
3124 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3125 PL_lex_casestack[--PL_lex_casemods] = '\0';
3128 if (PL_lex_casemods > 10)
3129 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3130 PL_lex_casestack[PL_lex_casemods++] = *s;
3131 PL_lex_casestack[PL_lex_casemods] = '\0';
3132 PL_lex_state = LEX_INTERPCONCAT;
3133 start_force(PL_curforce);
3134 NEXTVAL_NEXTTOKE.ival = 0;
3136 start_force(PL_curforce);
3138 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3140 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3142 NEXTVAL_NEXTTOKE.ival = OP_LC;
3144 NEXTVAL_NEXTTOKE.ival = OP_UC;
3146 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3148 Perl_croak(aTHX_ "panic: yylex");
3150 SV* const tmpsv = newSVpvn("",0);
3151 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3157 if (PL_lex_starts) {
3163 sv_free(PL_thistoken);
3164 PL_thistoken = newSVpvn("",0);
3167 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3168 if (PL_lex_casemods == 1 && PL_lex_inpat)
3177 case LEX_INTERPPUSH:
3178 return REPORT(sublex_push());
3180 case LEX_INTERPSTART:
3181 if (PL_bufptr == PL_bufend)
3182 return REPORT(sublex_done());
3183 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3184 "### Interpolated variable\n"); });
3186 PL_lex_dojoin = (*PL_bufptr == '@');
3187 PL_lex_state = LEX_INTERPNORMAL;
3188 if (PL_lex_dojoin) {
3189 start_force(PL_curforce);
3190 NEXTVAL_NEXTTOKE.ival = 0;
3192 start_force(PL_curforce);
3193 force_ident("\"", '$');
3194 start_force(PL_curforce);
3195 NEXTVAL_NEXTTOKE.ival = 0;
3197 start_force(PL_curforce);
3198 NEXTVAL_NEXTTOKE.ival = 0;
3200 start_force(PL_curforce);
3201 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3204 if (PL_lex_starts++) {
3209 sv_free(PL_thistoken);
3210 PL_thistoken = newSVpvn("",0);
3213 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3214 if (!PL_lex_casemods && PL_lex_inpat)
3221 case LEX_INTERPENDMAYBE:
3222 if (intuit_more(PL_bufptr)) {
3223 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3229 if (PL_lex_dojoin) {
3230 PL_lex_dojoin = FALSE;
3231 PL_lex_state = LEX_INTERPCONCAT;
3235 sv_free(PL_thistoken);
3236 PL_thistoken = newSVpvn("",0);
3241 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3242 && SvEVALED(PL_lex_repl))
3244 if (PL_bufptr != PL_bufend)
3245 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3249 case LEX_INTERPCONCAT:
3251 if (PL_lex_brackets)
3252 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3254 if (PL_bufptr == PL_bufend)
3255 return REPORT(sublex_done());
3257 if (SvIVX(PL_linestr) == '\'') {
3258 SV *sv = newSVsv(PL_linestr);
3261 else if ( PL_hints & HINT_NEW_RE )
3262 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3263 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3267 s = scan_const(PL_bufptr);
3269 PL_lex_state = LEX_INTERPCASEMOD;
3271 PL_lex_state = LEX_INTERPSTART;
3274 if (s != PL_bufptr) {
3275 start_force(PL_curforce);
3277 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3279 NEXTVAL_NEXTTOKE = yylval;
3282 if (PL_lex_starts++) {
3286 sv_free(PL_thistoken);
3287 PL_thistoken = newSVpvn("",0);
3290 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3291 if (!PL_lex_casemods && PL_lex_inpat)
3304 PL_lex_state = LEX_NORMAL;
3305 s = scan_formline(PL_bufptr);
3306 if (!PL_lex_formbrack)
3312 PL_oldoldbufptr = PL_oldbufptr;
3318 sv_free(PL_thistoken);
3321 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3325 if (isIDFIRST_lazy_if(s,UTF))
3327 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3330 goto fake_eof; /* emulate EOF on ^D or ^Z */
3339 if (PL_lex_brackets) {
3340 yyerror(PL_lex_formbrack
3341 ? "Format not terminated"
3342 : "Missing right curly or square bracket");
3344 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3345 "### Tokener got EOF\n");
3349 if (s++ < PL_bufend)
3350 goto retry; /* ignore stray nulls */
3353 if (!PL_in_eval && !PL_preambled) {
3354 PL_preambled = TRUE;
3359 sv_setpv(PL_linestr,incl_perldb());
3360 if (SvCUR(PL_linestr))
3361 sv_catpvs(PL_linestr,";");
3363 while(AvFILLp(PL_preambleav) >= 0) {
3364 SV *tmpsv = av_shift(PL_preambleav);
3365 sv_catsv(PL_linestr, tmpsv);
3366 sv_catpvs(PL_linestr, ";");
3369 sv_free((SV*)PL_preambleav);
3370 PL_preambleav = NULL;
3372 if (PL_minus_n || PL_minus_p) {
3373 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3375 sv_catpvs(PL_linestr,"chomp;");
3378 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3379 || *PL_splitstr == '"')
3380 && strchr(PL_splitstr + 1, *PL_splitstr))
3381 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3383 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3384 bytes can be used as quoting characters. :-) */
3385 const char *splits = PL_splitstr;
3386 sv_catpvs(PL_linestr, "our @F=split(q\0");
3389 if (*splits == '\\')
3390 sv_catpvn(PL_linestr, splits, 1);
3391 sv_catpvn(PL_linestr, splits, 1);
3392 } while (*splits++);
3393 /* This loop will embed the trailing NUL of
3394 PL_linestr as the last thing it does before
3396 sv_catpvs(PL_linestr, ");");
3400 sv_catpvs(PL_linestr,"our @F=split(' ');");
3404 sv_catpvs(PL_linestr,"use feature ':5.10';");
3405 sv_catpvs(PL_linestr, "\n");
3406 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3407 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3408 PL_last_lop = PL_last_uni = NULL;
3409 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3410 SV * const sv = newSV(0);
3412 sv_upgrade(sv, SVt_PVMG);
3413 sv_setsv(sv,PL_linestr);
3416 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3421 bof = PL_rsfp ? TRUE : FALSE;
3422 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3425 PL_realtokenstart = -1;
3428 if (PL_preprocess && !PL_in_eval)
3429 (void)PerlProc_pclose(PL_rsfp);
3430 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3431 PerlIO_clearerr(PL_rsfp);
3433 (void)PerlIO_close(PL_rsfp);
3435 PL_doextract = FALSE;
3437 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3442 sv_setpv(PL_linestr,PL_minus_p
3443 ? ";}continue{print;}" : ";}");
3444 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3445 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3446 PL_last_lop = PL_last_uni = NULL;
3447 PL_minus_n = PL_minus_p = 0;
3450 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3451 PL_last_lop = PL_last_uni = NULL;
3452 sv_setpvn(PL_linestr,"",0);
3453 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3455 /* If it looks like the start of a BOM or raw UTF-16,
3456 * check if it in fact is. */
3462 #ifdef PERLIO_IS_STDIO
3463 # ifdef __GNU_LIBRARY__
3464 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3465 # define FTELL_FOR_PIPE_IS_BROKEN
3469 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3470 # define FTELL_FOR_PIPE_IS_BROKEN
3475 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3476 /* This loses the possibility to detect the bof
3477 * situation on perl -P when the libc5 is being used.
3478 * Workaround? Maybe attach some extra state to PL_rsfp?
3481 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3483 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3486 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3487 s = swallow_bom((U8*)s);
3491 /* Incest with pod. */
3494 sv_catsv(PL_thiswhite, PL_linestr);
3496 if (*s == '=' && strnEQ(s, "=cut", 4)) {
3497 sv_setpvn(PL_linestr, "", 0);
3498 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3499 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3500 PL_last_lop = PL_last_uni = NULL;
3501 PL_doextract = FALSE;
3505 } while (PL_doextract);
3506 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3507 if (PERLDB_LINE && PL_curstash != PL_debstash) {
3508 SV * const sv = newSV(0);
3510 sv_upgrade(sv, SVt_PVMG);
3511 sv_setsv(sv,PL_linestr);
3514 av_store(CopFILEAVx(PL_curcop),(I32)CopLINE(PL_curcop),sv);
3516 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3517 PL_last_lop = PL_last_uni = NULL;
3518 if (CopLINE(PL_curcop) == 1) {
3519 while (s < PL_bufend && isSPACE(*s))
3521 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3525 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3529 if (*s == '#' && *(s+1) == '!')
3531 #ifdef ALTERNATE_SHEBANG
3533 static char const as[] = ALTERNATE_SHEBANG;
3534 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3535 d = s + (sizeof(as) - 1);
3537 #endif /* ALTERNATE_SHEBANG */
3546 while (*d && !isSPACE(*d))
3550 #ifdef ARG_ZERO_IS_SCRIPT
3551 if (ipathend > ipath) {
3553 * HP-UX (at least) sets argv[0] to the script name,
3554 * which makes $^X incorrect. And Digital UNIX and Linux,
3555 * at least, set argv[0] to the basename of the Perl
3556 * interpreter. So, having found "#!", we'll set it right.
3558 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3560 assert(SvPOK(x) || SvGMAGICAL(x));
3561 if (sv_eq(x, CopFILESV(PL_curcop))) {
3562 sv_setpvn(x, ipath, ipathend - ipath);
3568 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3569 const char * const lstart = SvPV_const(x,llen);
3571 bstart += blen - llen;
3572 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3573 sv_setpvn(x, ipath, ipathend - ipath);
3578 TAINT_NOT; /* $^X is always tainted, but that's OK */
3580 #endif /* ARG_ZERO_IS_SCRIPT */
3585 d = instr(s,"perl -");
3587 d = instr(s,"perl");
3589 /* avoid getting into infinite loops when shebang
3590 * line contains "Perl" rather than "perl" */
3592 for (d = ipathend-4; d >= ipath; --d) {
3593 if ((*d == 'p' || *d == 'P')
3594 && !ibcmp(d, "perl", 4))
3604 #ifdef ALTERNATE_SHEBANG
3606 * If the ALTERNATE_SHEBANG on this system starts with a
3607 * character that can be part of a Perl expression, then if
3608 * we see it but not "perl", we're probably looking at the
3609 * start of Perl code, not a request to hand off to some
3610 * other interpreter. Similarly, if "perl" is there, but
3611 * not in the first 'word' of the line, we assume the line
3612 * contains the start of the Perl program.
3614 if (d && *s != '#') {
3615 const char *c = ipath;
3616 while (*c && !strchr("; \t\r\n\f\v#", *c))
3619 d = NULL; /* "perl" not in first word; ignore */
3621 *s = '#'; /* Don't try to parse shebang line */
3623 #endif /* ALTERNATE_SHEBANG */
3624 #ifndef MACOS_TRADITIONAL
3629 !instr(s,"indir") &&
3630 instr(PL_origargv[0],"perl"))
3637 while (s < PL_bufend && isSPACE(*s))
3639 if (s < PL_bufend) {
3640 Newxz(newargv,PL_origargc+3,char*);
3642 while (s < PL_bufend && !isSPACE(*s))
3645 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3648 newargv = PL_origargv;
3651 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3653 Perl_croak(aTHX_ "Can't exec %s", ipath);
3657 while (*d && !isSPACE(*d)) d++;
3658 while (SPACE_OR_TAB(*d)) d++;
3661 const bool switches_done = PL_doswitches;
3662 const U32 oldpdb = PL_perldb;
3663 const bool oldn = PL_minus_n;
3664 const bool oldp = PL_minus_p;
3667 if (*d == 'M' || *d == 'm' || *d == 'C') {
3668 const char * const m = d;
3669 while (*d && !isSPACE(*d))
3671 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3674 d = moreswitches(d);
3676 if (PL_doswitches && !switches_done) {
3677 int argc = PL_origargc;
3678 char **argv = PL_origargv;
3681 } while (argc && argv[0][0] == '-' && argv[0][1]);
3682 init_argv_symbols(argc,argv);
3684 if ((PERLDB_LINE && !oldpdb) ||
3685 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3686 /* if we have already added "LINE: while (<>) {",
3687 we must not do it again */
3689 sv_setpvn(PL_linestr, "", 0);
3690 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3691 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3692 PL_last_lop = PL_last_uni = NULL;
3693 PL_preambled = FALSE;
3695 (void)gv_fetchfile(PL_origfilename);
3702 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3704 PL_lex_state = LEX_FORMLINE;
3709 #ifdef PERL_STRICT_CR
3710 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3712 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3714 case ' ': case '\t': case '\f': case 013:
3715 #ifdef MACOS_TRADITIONAL
3719 PL_realtokenstart = -1;
3728 PL_realtokenstart = -1;
3732 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3733 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3734 /* handle eval qq[#line 1 "foo"\n ...] */
3735 CopLINE_dec(PL_curcop);
3738 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3740 if (!PL_in_eval || PL_rsfp)
3745 while (d < PL_bufend && *d != '\n')
3749 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3750 Perl_croak(aTHX_ "panic: input overflow");
3753 PL_thiswhite = newSVpvn(s, d - s);
3758 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3760 PL_lex_state = LEX_FORMLINE;
3766 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3767 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3770 TOKEN(PEG); /* make sure any #! line is accessible */
3775 /* if (PL_madskills && PL_lex_formbrack) { */
3777 while (d < PL_bufend && *d != '\n')
3781 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3782 Perl_croak(aTHX_ "panic: input overflow");
3783 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3785 PL_thiswhite = newSVpvn("",0);
3786 if (CopLINE(PL_curcop) == 1) {
3787 sv_setpvn(PL_thiswhite, "", 0);
3790 sv_catpvn(PL_thiswhite, s, d - s);
3804 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3812 while (s < PL_bufend && SPACE_OR_TAB(*s))
3815 if (strnEQ(s,"=>",2)) {
3816 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3817 DEBUG_T( { S_printbuf(aTHX_
3818 "### Saw unary minus before =>, forcing word %s\n", s);
3820 OPERATOR('-'); /* unary minus */
3822 PL_last_uni = PL_oldbufptr;
3824 case 'r': ftst = OP_FTEREAD; break;
3825 case 'w': ftst = OP_FTEWRITE; break;
3826 case 'x': ftst = OP_FTEEXEC; break;
3827 case 'o': ftst = OP_FTEOWNED; break;
3828 case 'R': ftst = OP_FTRREAD; break;
3829 case 'W': ftst = OP_FTRWRITE; break;
3830 case 'X': ftst = OP_FTREXEC; break;
3831 case 'O': ftst = OP_FTROWNED; break;
3832 case 'e': ftst = OP_FTIS; break;
3833 case 'z': ftst = OP_FTZERO; break;
3834 case 's': ftst = OP_FTSIZE; break;
3835 case 'f': ftst = OP_FTFILE; break;
3836 case 'd': ftst = OP_FTDIR; break;
3837 case 'l': ftst = OP_FTLINK; break;
3838 case 'p': ftst = OP_FTPIPE; break;
3839 case 'S': ftst = OP_FTSOCK; break;
3840 case 'u': ftst = OP_FTSUID; break;
3841 case 'g': ftst = OP_FTSGID; break;
3842 case 'k': ftst = OP_FTSVTX; break;
3843 case 'b': ftst = OP_FTBLK; break;
3844 case 'c': ftst = OP_FTCHR; break;
3845 case 't': ftst = OP_FTTTY; break;
3846 case 'T': ftst = OP_FTTEXT; break;
3847 case 'B': ftst = OP_FTBINARY; break;
3848 case 'M': case 'A': case 'C':
3849 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
3851 case 'M': ftst = OP_FTMTIME; break;
3852 case 'A': ftst = OP_FTATIME; break;
3853 case 'C': ftst = OP_FTCTIME; break;
3861 PL_last_lop_op = (OPCODE)ftst;
3862 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3863 "### Saw file test %c\n", (int)tmp);
3868 /* Assume it was a minus followed by a one-letter named
3869 * subroutine call (or a -bareword), then. */
3870 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3871 "### '-%c' looked like a file test but was not\n",
3878 const char tmp = *s++;
3881 if (PL_expect == XOPERATOR)
3886 else if (*s == '>') {
3889 if (isIDFIRST_lazy_if(s,UTF)) {
3890 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
3898 if (PL_expect == XOPERATOR)
3901 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3903 OPERATOR('-'); /* unary minus */
3909 const char tmp = *s++;
3912 if (PL_expect == XOPERATOR)
3917 if (PL_expect == XOPERATOR)
3920 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
3927 if (PL_expect != XOPERATOR) {
3928 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
3929 PL_expect = XOPERATOR;
3930 force_ident(PL_tokenbuf, '*');
3943 if (PL_expect == XOPERATOR) {
3947 PL_tokenbuf[0] = '%';
3948 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
3949 if (!PL_tokenbuf[1]) {
3952 PL_pending_ident = '%';
3963 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR)
3964 && FEATURE_IS_ENABLED("~~"))
3971 const char tmp = *s++;
3977 goto just_a_word_zero_gv;
3980 switch (PL_expect) {
3986 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
3988 PL_bufptr = s; /* update in case we back off */
3994 PL_expect = XTERMBLOCK;
3997 stuffstart = s - SvPVX(PL_linestr) - 1;
4001 while (isIDFIRST_lazy_if(s,UTF)) {
4004 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4005 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len))) {
4006 if (tmp < 0) tmp = -tmp;
4021 sv = newSVpvn(s, len);
4023 d = scan_str(d,TRUE,TRUE);
4025 /* MUST advance bufptr here to avoid bogus
4026 "at end of line" context messages from yyerror().
4028 PL_bufptr = s + len;
4029 yyerror("Unterminated attribute parameter in attribute list");
4033 return REPORT(0); /* EOF indicator */
4037 sv_catsv(sv, PL_lex_stuff);
4038 attrs = append_elem(OP_LIST, attrs,
4039 newSVOP(OP_CONST, 0, sv));
4040 SvREFCNT_dec(PL_lex_stuff);
4041 PL_lex_stuff = NULL;
4044 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4046 if (PL_in_my == KEY_our) {
4048 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4050 /* skip to avoid loading attributes.pm */
4052 deprecate(":unique");
4055 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4058 /* NOTE: any CV attrs applied here need to be part of
4059 the CVf_BUILTIN_ATTRS define in cv.h! */
4060 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4062 CvLVALUE_on(PL_compcv);
4064 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4066 CvLOCKED_on(PL_compcv);
4068 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4070 CvMETHOD_on(PL_compcv);
4072 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4074 CvASSERTION_on(PL_compcv);
4076 /* After we've set the flags, it could be argued that
4077 we don't need to do the attributes.pm-based setting
4078 process, and shouldn't bother appending recognized
4079 flags. To experiment with that, uncomment the
4080 following "else". (Note that's already been
4081 uncommented. That keeps the above-applied built-in
4082 attributes from being intercepted (and possibly
4083 rejected) by a package's attribute routines, but is
4084 justified by the performance win for the common case
4085 of applying only built-in attributes.) */
4087 attrs = append_elem(OP_LIST, attrs,
4088 newSVOP(OP_CONST, 0,
4092 if (*s == ':' && s[1] != ':')
4095 break; /* require real whitespace or :'s */
4096 /* XXX losing whitespace on sequential attributes here */
4100 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4101 if (*s != ';' && *s != '}' && *s != tmp
4102 && (tmp != '=' || *s != ')')) {
4103 const char q = ((*s == '\'') ? '"' : '\'');
4104 /* If here for an expression, and parsed no attrs, back
4106 if (tmp == '=' && !attrs) {
4110 /* MUST advance bufptr here to avoid bogus "at end of line"
4111 context messages from yyerror().
4115 ? Perl_form(aTHX_ "Invalid separator character "
4116 "%c%c%c in attribute list", q, *s, q)
4117 : "Unterminated attribute list" );
4125 start_force(PL_curforce);
4126 NEXTVAL_NEXTTOKE.opval = attrs;
4127 CURMAD('_', PL_nextwhite);
4132 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4133 (s - SvPVX(PL_linestr)) - stuffstart);
4141 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4142 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4150 const char tmp = *s++;
4155 const char tmp = *s++;
4163 if (PL_lex_brackets <= 0)
4164 yyerror("Unmatched right square bracket");
4167 if (PL_lex_state == LEX_INTERPNORMAL) {
4168 if (PL_lex_brackets == 0) {
4169 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4170 PL_lex_state = LEX_INTERPEND;
4177 if (PL_lex_brackets > 100) {
4178 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4180 switch (PL_expect) {
4182 if (PL_lex_formbrack) {
4186 if (PL_oldoldbufptr == PL_last_lop)
4187 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4189 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4190 OPERATOR(HASHBRACK);
4192 while (s < PL_bufend && SPACE_OR_TAB(*s))
4195 PL_tokenbuf[0] = '\0';
4196 if (d < PL_bufend && *d == '-') {
4197 PL_tokenbuf[0] = '-';
4199 while (d < PL_bufend && SPACE_OR_TAB(*d))
4202 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4203 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4205 while (d < PL_bufend && SPACE_OR_TAB(*d))
4208 const char minus = (PL_tokenbuf[0] == '-');
4209 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4217 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4222 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4227 if (PL_oldoldbufptr == PL_last_lop)
4228 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4230 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4233 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4235 /* This hack is to get the ${} in the message. */
4237 yyerror("syntax error");
4240 OPERATOR(HASHBRACK);
4242 /* This hack serves to disambiguate a pair of curlies
4243 * as being a block or an anon hash. Normally, expectation
4244 * determines that, but in cases where we're not in a
4245 * position to expect anything in particular (like inside
4246 * eval"") we have to resolve the ambiguity. This code
4247 * covers the case where the first term in the curlies is a
4248 * quoted string. Most other cases need to be explicitly
4249 * disambiguated by prepending a "+" before the opening
4250 * curly in order to force resolution as an anon hash.
4252 * XXX should probably propagate the outer expectation
4253 * into eval"" to rely less on this hack, but that could
4254 * potentially break current behavior of eval"".
4258 if (*s == '\'' || *s == '"' || *s == '`') {
4259 /* common case: get past first string, handling escapes */
4260 for (t++; t < PL_bufend && *t != *s;)
4261 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4265 else if (*s == 'q') {
4268 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4271 /* skip q//-like construct */
4273 char open, close, term;
4276 while (t < PL_bufend && isSPACE(*t))
4278 /* check for q => */
4279 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4280 OPERATOR(HASHBRACK);
4284 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4288 for (t++; t < PL_bufend; t++) {
4289 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4291 else if (*t == open)
4295 for (t++; t < PL_bufend; t++) {
4296 if (*t == '\\' && t+1 < PL_bufend)
4298 else if (*t == close && --brackets <= 0)
4300 else if (*t == open)
4307 /* skip plain q word */
4308 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4311 else if (isALNUM_lazy_if(t,UTF)) {
4313 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4316 while (t < PL_bufend && isSPACE(*t))
4318 /* if comma follows first term, call it an anon hash */
4319 /* XXX it could be a comma expression with loop modifiers */
4320 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4321 || (*t == '=' && t[1] == '>')))
4322 OPERATOR(HASHBRACK);
4323 if (PL_expect == XREF)
4326 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4332 yylval.ival = CopLINE(PL_curcop);
4333 if (isSPACE(*s) || *s == '#')
4334 PL_copline = NOLINE; /* invalidate current command line number */
4339 if (PL_lex_brackets <= 0)
4340 yyerror("Unmatched right curly bracket");
4342 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4343 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4344 PL_lex_formbrack = 0;
4345 if (PL_lex_state == LEX_INTERPNORMAL) {
4346 if (PL_lex_brackets == 0) {
4347 if (PL_expect & XFAKEBRACK) {
4348 PL_expect &= XENUMMASK;
4349 PL_lex_state = LEX_INTERPEND;
4354 PL_thiswhite = newSVpvn("",0);
4355 sv_catpvn(PL_thiswhite,"}",1);
4358 return yylex(); /* ignore fake brackets */
4360 if (*s == '-' && s[1] == '>')
4361 PL_lex_state = LEX_INTERPENDMAYBE;
4362 else if (*s != '[' && *s != '{')
4363 PL_lex_state = LEX_INTERPEND;
4366 if (PL_expect & XFAKEBRACK) {
4367 PL_expect &= XENUMMASK;
4369 return yylex(); /* ignore fake brackets */
4371 start_force(PL_curforce);
4373 curmad('X', newSVpvn(s-1,1));
4374 CURMAD('_', PL_thiswhite);
4379 PL_thistoken = newSVpvn("",0);
4387 if (PL_expect == XOPERATOR) {
4388 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4389 && isIDFIRST_lazy_if(s,UTF))
4391 CopLINE_dec(PL_curcop);
4392 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4393 CopLINE_inc(PL_curcop);
<