3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 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
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
42 #include "dquote_static.c"
44 #define new_constant(a,b,c,d,e,f,g) \
45 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
47 #define pl_yylval (PL_parser->yylval)
49 /* XXX temporary backwards compatibility */
50 #define PL_lex_brackets (PL_parser->lex_brackets)
51 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
52 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XENUMMASK 0x3f
130 #define XFAKEEOF 0x40
131 #define XFAKEBRACK 0x80
133 #ifdef USE_UTF8_SCRIPTS
134 # define UTF (!IN_BYTES)
136 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
139 /* The maximum number of characters preceding the unrecognized one to display */
140 #define UNRECOGNIZED_PRECEDE_COUNT 10
142 /* In variables named $^X, these are the legal values for X.
143 * 1999-02-27 mjd-perl-patch@plover.com */
144 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
146 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
148 /* LEX_* are values for PL_lex_state, the state of the lexer.
149 * They are arranged oddly so that the guard on the switch statement
150 * can get by with a single comparison (if the compiler is smart enough).
153 /* #define LEX_NOTPARSING 11 is done in perl.h. */
155 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
156 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
157 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
158 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
159 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
161 /* at end of code, eg "$x" followed by: */
162 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
163 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
165 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
166 string or after \E, $foo, etc */
167 #define LEX_INTERPCONST 2 /* NOT USED */
168 #define LEX_FORMLINE 1 /* expecting a format line */
169 #define LEX_KNOWNEXT 0 /* next token known; just return it */
173 static const char* const lex_state_names[] = {
192 #include "keywords.h"
194 /* CLINE is a macro that ensures PL_copline has a sane value */
199 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
202 # define SKIPSPACE0(s) skipspace0(s)
203 # define SKIPSPACE1(s) skipspace1(s)
204 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
205 # define PEEKSPACE(s) skipspace2(s,0)
207 # define SKIPSPACE0(s) skipspace(s)
208 # define SKIPSPACE1(s) skipspace(s)
209 # define SKIPSPACE2(s,tsv) skipspace(s)
210 # define PEEKSPACE(s) skipspace(s)
214 * Convenience functions to return different tokens and prime the
215 * lexer for the next token. They all take an argument.
217 * TOKEN : generic token (used for '(', DOLSHARP, etc)
218 * OPERATOR : generic operator
219 * AOPERATOR : assignment operator
220 * PREBLOCK : beginning the block after an if, while, foreach, ...
221 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
222 * PREREF : *EXPR where EXPR is not a simple identifier
223 * TERM : expression term
224 * LOOPX : loop exiting command (goto, last, dump, etc)
225 * FTST : file test operator
226 * FUN0 : zero-argument function
227 * FUN0OP : zero-argument function, with its op created in this file
228 * FUN1 : not used, except for not, which isn't a UNIOP
229 * BOop : bitwise or or xor
231 * SHop : shift operator
232 * PWop : power operator
233 * PMop : pattern-matching operator
234 * Aop : addition-level operator
235 * Mop : multiplication-level operator
236 * Eop : equality-testing operator
237 * Rop : relational operator <= != gt
239 * Also see LOP and lop() below.
242 #ifdef DEBUGGING /* Serve -DT. */
243 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
245 # define REPORT(retval) (retval)
248 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
249 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
250 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
251 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
252 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
253 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
254 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
255 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
256 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
257 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
258 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
259 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
260 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
261 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
262 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
263 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
264 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
265 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
266 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
267 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
268 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
270 /* This bit of chicanery makes a unary function followed by
271 * a parenthesis into a function with one argument, highest precedence.
272 * The UNIDOR macro is for unary functions that can be followed by the //
273 * operator (such as C<shift // 0>).
275 #define UNI2(f,x) { \
276 pl_yylval.ival = f; \
279 PL_last_uni = PL_oldbufptr; \
280 PL_last_lop_op = f; \
282 return REPORT( (int)FUNC1 ); \
284 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
286 #define UNI(f) UNI2(f,XTERM)
287 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
288 #define UNIPROTO(f,optional) { \
289 if (optional) PL_last_uni = PL_oldbufptr; \
293 #define UNIBRACK(f) { \
294 pl_yylval.ival = f; \
296 PL_last_uni = PL_oldbufptr; \
298 return REPORT( (int)FUNC1 ); \
300 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
303 /* grandfather return to old style */
306 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
307 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
308 pl_yylval.ival = (f); \
316 /* how to interpret the pl_yylval associated with the token */
320 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
326 static struct debug_tokens {
328 enum token_type type;
330 } const debug_tokens[] =
332 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
333 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
334 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
335 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
336 { ARROW, TOKENTYPE_NONE, "ARROW" },
337 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
338 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
339 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
340 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
341 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
342 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
343 { DO, TOKENTYPE_NONE, "DO" },
344 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
345 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
346 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
347 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
348 { ELSE, TOKENTYPE_NONE, "ELSE" },
349 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
350 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
351 { FOR, TOKENTYPE_IVAL, "FOR" },
352 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
353 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
354 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
355 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
356 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
357 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
358 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
359 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
360 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
361 { IF, TOKENTYPE_IVAL, "IF" },
362 { LABEL, TOKENTYPE_OPVAL, "LABEL" },
363 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
364 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
365 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
366 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
367 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
368 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
369 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
370 { MY, TOKENTYPE_IVAL, "MY" },
371 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
372 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
373 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
374 { OROP, TOKENTYPE_IVAL, "OROP" },
375 { OROR, TOKENTYPE_NONE, "OROR" },
376 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
377 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
378 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
379 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
380 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
381 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
382 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
383 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
384 { PREINC, TOKENTYPE_NONE, "PREINC" },
385 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
386 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
387 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
388 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
389 { SUB, TOKENTYPE_NONE, "SUB" },
390 { THING, TOKENTYPE_OPVAL, "THING" },
391 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
392 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
393 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
394 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
395 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
396 { USE, TOKENTYPE_IVAL, "USE" },
397 { WHEN, TOKENTYPE_IVAL, "WHEN" },
398 { WHILE, TOKENTYPE_IVAL, "WHILE" },
399 { WORD, TOKENTYPE_OPVAL, "WORD" },
400 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
401 { 0, TOKENTYPE_NONE, NULL }
404 /* dump the returned token in rv, plus any optional arg in pl_yylval */
407 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
411 PERL_ARGS_ASSERT_TOKEREPORT;
414 const char *name = NULL;
415 enum token_type type = TOKENTYPE_NONE;
416 const struct debug_tokens *p;
417 SV* const report = newSVpvs("<== ");
419 for (p = debug_tokens; p->token; p++) {
420 if (p->token == (int)rv) {
427 Perl_sv_catpv(aTHX_ report, name);
428 else if ((char)rv > ' ' && (char)rv < '~')
429 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
431 sv_catpvs(report, "EOF");
433 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
436 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
439 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
441 case TOKENTYPE_OPNUM:
442 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
443 PL_op_name[lvalp->ival]);
446 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
448 case TOKENTYPE_OPVAL:
450 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
451 PL_op_name[lvalp->opval->op_type]);
452 if (lvalp->opval->op_type == OP_CONST) {
453 Perl_sv_catpvf(aTHX_ report, " %s",
454 SvPEEK(cSVOPx_sv(lvalp->opval)));
459 sv_catpvs(report, "(opval=null)");
462 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
468 /* print the buffer with suitable escapes */
471 S_printbuf(pTHX_ const char *const fmt, const char *const s)
473 SV* const tmp = newSVpvs("");
475 PERL_ARGS_ASSERT_PRINTBUF;
477 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
484 S_deprecate_commaless_var_list(pTHX) {
486 deprecate("comma-less variable list");
487 return REPORT(','); /* grandfather non-comma-format format */
493 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
494 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
498 S_ao(pTHX_ int toketype)
501 if (*PL_bufptr == '=') {
503 if (toketype == ANDAND)
504 pl_yylval.ival = OP_ANDASSIGN;
505 else if (toketype == OROR)
506 pl_yylval.ival = OP_ORASSIGN;
507 else if (toketype == DORDOR)
508 pl_yylval.ival = OP_DORASSIGN;
516 * When Perl expects an operator and finds something else, no_op
517 * prints the warning. It always prints "<something> found where
518 * operator expected. It prints "Missing semicolon on previous line?"
519 * if the surprise occurs at the start of the line. "do you need to
520 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
521 * where the compiler doesn't know if foo is a method call or a function.
522 * It prints "Missing operator before end of line" if there's nothing
523 * after the missing operator, or "... before <...>" if there is something
524 * after the missing operator.
528 S_no_op(pTHX_ const char *const what, char *s)
531 char * const oldbp = PL_bufptr;
532 const bool is_first = (PL_oldbufptr == PL_linestart);
534 PERL_ARGS_ASSERT_NO_OP;
540 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
541 if (ckWARN_d(WARN_SYNTAX)) {
543 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
544 "\t(Missing semicolon on previous line?)\n");
545 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
547 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':');
548 t += UTF ? UTF8SKIP(t) : 1)
550 if (t < PL_bufptr && isSPACE(*t))
551 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
552 "\t(Do you need to predeclare %"SVf"?)\n",
553 SVfARG(newSVpvn_flags(PL_oldoldbufptr, (STRLEN)(t - PL_oldoldbufptr),
554 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
558 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
559 "\t(Missing operator before %"SVf"?)\n",
560 SVfARG(newSVpvn_flags(oldbp, (STRLEN)(s - oldbp),
561 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
569 * Complain about missing quote/regexp/heredoc terminator.
570 * If it's called with NULL then it cauterizes the line buffer.
571 * If we're in a delimited string and the delimiter is a control
572 * character, it's reformatted into a two-char sequence like ^C.
577 S_missingterm(pTHX_ char *s)
583 char * const nl = strrchr(s,'\n');
587 else if (isCNTRL(PL_multi_close)) {
589 tmpbuf[1] = (char)toCTRL(PL_multi_close);
594 *tmpbuf = (char)PL_multi_close;
598 q = strchr(s,'"') ? '\'' : '"';
599 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
605 * Check whether the named feature is enabled.
608 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
611 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
613 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
615 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
617 if (namelen > MAX_FEATURE_LEN)
619 memcpy(&he_name[8], name, namelen);
621 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
622 REFCOUNTED_HE_EXISTS));
626 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
627 * utf16-to-utf8-reversed.
630 #ifdef PERL_CR_FILTER
634 register const char *s = SvPVX_const(sv);
635 register const char * const e = s + SvCUR(sv);
637 PERL_ARGS_ASSERT_STRIP_RETURN;
639 /* outer loop optimized to do nothing if there are no CR-LFs */
641 if (*s++ == '\r' && *s == '\n') {
642 /* hit a CR-LF, need to copy the rest */
643 register char *d = s - 1;
646 if (*s == '\r' && s[1] == '\n')
657 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
659 const I32 count = FILTER_READ(idx+1, sv, maxlen);
660 if (count > 0 && !maxlen)
667 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
669 Creates and initialises a new lexer/parser state object, supplying
670 a context in which to lex and parse from a new source of Perl code.
671 A pointer to the new state object is placed in L</PL_parser>. An entry
672 is made on the save stack so that upon unwinding the new state object
673 will be destroyed and the former value of L</PL_parser> will be restored.
674 Nothing else need be done to clean up the parsing context.
676 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
677 non-null, provides a string (in SV form) containing code to be parsed.
678 A copy of the string is made, so subsequent modification of I<line>
679 does not affect parsing. I<rsfp>, if non-null, provides an input stream
680 from which code will be read to be parsed. If both are non-null, the
681 code in I<line> comes first and must consist of complete lines of input,
682 and I<rsfp> supplies the remainder of the source.
684 The I<flags> parameter is reserved for future use. Currently it is only
685 used by perl internally, so extensions should always pass zero.
690 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
691 can share filters with the current parser.
692 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
693 caller, hence isn't owned by the parser, so shouldn't be closed on parser
694 destruction. This is used to handle the case of defaulting to reading the
695 script from the standard input because no filename was given on the command
696 line (without getting confused by situation where STDIN has been closed, so
697 the script handle is opened on fd 0) */
700 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
703 const char *s = NULL;
704 yy_parser *parser, *oparser;
705 if (flags && flags & ~LEX_START_FLAGS)
706 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
708 /* create and initialise a parser */
710 Newxz(parser, 1, yy_parser);
711 parser->old_parser = oparser = PL_parser;
714 parser->stack = NULL;
716 parser->stack_size = 0;
718 /* on scope exit, free this parser and restore any outer one */
720 parser->saved_curcop = PL_curcop;
722 /* initialise lexer state */
725 parser->curforce = -1;
727 parser->nexttoke = 0;
729 parser->error_count = oparser ? oparser->error_count : 0;
730 parser->copline = NOLINE;
731 parser->lex_state = LEX_NORMAL;
732 parser->expect = XSTATE;
734 parser->rsfp_filters =
735 !(flags & LEX_START_SAME_FILTER) || !oparser
737 : MUTABLE_AV(SvREFCNT_inc(
738 oparser->rsfp_filters
739 ? oparser->rsfp_filters
740 : (oparser->rsfp_filters = newAV())
743 Newx(parser->lex_brackstack, 120, char);
744 Newx(parser->lex_casestack, 12, char);
745 *parser->lex_casestack = '\0';
749 s = SvPV_const(line, len);
750 parser->linestr = flags & LEX_START_COPIED
751 ? SvREFCNT_inc_simple_NN(line)
752 : newSVpvn_flags(s, len, SvUTF8(line));
753 if (!len || s[len-1] != ';')
754 sv_catpvs(parser->linestr, "\n;");
756 parser->linestr = newSVpvs("\n;");
758 parser->oldoldbufptr =
761 parser->linestart = SvPVX(parser->linestr);
762 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
763 parser->last_lop = parser->last_uni = NULL;
764 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
765 |LEX_DONT_CLOSE_RSFP);
767 parser->in_pod = parser->filtered = 0;
771 /* delete a parser object */
774 Perl_parser_free(pTHX_ const yy_parser *parser)
776 PERL_ARGS_ASSERT_PARSER_FREE;
778 PL_curcop = parser->saved_curcop;
779 SvREFCNT_dec(parser->linestr);
781 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
782 PerlIO_clearerr(parser->rsfp);
783 else if (parser->rsfp && (!parser->old_parser ||
784 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
785 PerlIO_close(parser->rsfp);
786 SvREFCNT_dec(parser->rsfp_filters);
788 Safefree(parser->lex_brackstack);
789 Safefree(parser->lex_casestack);
790 PL_parser = parser->old_parser;
796 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
798 Buffer scalar containing the chunk currently under consideration of the
799 text currently being lexed. This is always a plain string scalar (for
800 which C<SvPOK> is true). It is not intended to be used as a scalar by
801 normal scalar means; instead refer to the buffer directly by the pointer
802 variables described below.
804 The lexer maintains various C<char*> pointers to things in the
805 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
806 reallocated, all of these pointers must be updated. Don't attempt to
807 do this manually, but rather use L</lex_grow_linestr> if you need to
808 reallocate the buffer.
810 The content of the text chunk in the buffer is commonly exactly one
811 complete line of input, up to and including a newline terminator,
812 but there are situations where it is otherwise. The octets of the
813 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
814 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
815 flag on this scalar, which may disagree with it.
817 For direct examination of the buffer, the variable
818 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
819 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
820 of these pointers is usually preferable to examination of the scalar
821 through normal scalar means.
823 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
825 Direct pointer to the end of the chunk of text currently being lexed, the
826 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
827 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
828 always located at the end of the buffer, and does not count as part of
829 the buffer's contents.
831 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
833 Points to the current position of lexing inside the lexer buffer.
834 Characters around this point may be freely examined, within
835 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
836 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
837 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
839 Lexing code (whether in the Perl core or not) moves this pointer past
840 the characters that it consumes. It is also expected to perform some
841 bookkeeping whenever a newline character is consumed. This movement
842 can be more conveniently performed by the function L</lex_read_to>,
843 which handles newlines appropriately.
845 Interpretation of the buffer's octets can be abstracted out by
846 using the slightly higher-level functions L</lex_peek_unichar> and
847 L</lex_read_unichar>.
849 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
851 Points to the start of the current line inside the lexer buffer.
852 This is useful for indicating at which column an error occurred, and
853 not much else. This must be updated by any lexing code that consumes
854 a newline; the function L</lex_read_to> handles this detail.
860 =for apidoc Amx|bool|lex_bufutf8
862 Indicates whether the octets in the lexer buffer
863 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
864 of Unicode characters. If not, they should be interpreted as Latin-1
865 characters. This is analogous to the C<SvUTF8> flag for scalars.
867 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
868 contains valid UTF-8. Lexing code must be robust in the face of invalid
871 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
872 is significant, but not the whole story regarding the input character
873 encoding. Normally, when a file is being read, the scalar contains octets
874 and its C<SvUTF8> flag is off, but the octets should be interpreted as
875 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
876 however, the scalar may have the C<SvUTF8> flag on, and in this case its
877 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
878 is in effect. This logic may change in the future; use this function
879 instead of implementing the logic yourself.
885 Perl_lex_bufutf8(pTHX)
891 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
893 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
894 at least I<len> octets (including terminating NUL). Returns a
895 pointer to the reallocated buffer. This is necessary before making
896 any direct modification of the buffer that would increase its length.
897 L</lex_stuff_pvn> provides a more convenient way to insert text into
900 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
901 this function updates all of the lexer's variables that point directly
908 Perl_lex_grow_linestr(pTHX_ STRLEN len)
912 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
913 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
914 linestr = PL_parser->linestr;
915 buf = SvPVX(linestr);
916 if (len <= SvLEN(linestr))
918 bufend_pos = PL_parser->bufend - buf;
919 bufptr_pos = PL_parser->bufptr - buf;
920 oldbufptr_pos = PL_parser->oldbufptr - buf;
921 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
922 linestart_pos = PL_parser->linestart - buf;
923 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
924 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
925 buf = sv_grow(linestr, len);
926 PL_parser->bufend = buf + bufend_pos;
927 PL_parser->bufptr = buf + bufptr_pos;
928 PL_parser->oldbufptr = buf + oldbufptr_pos;
929 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
930 PL_parser->linestart = buf + linestart_pos;
931 if (PL_parser->last_uni)
932 PL_parser->last_uni = buf + last_uni_pos;
933 if (PL_parser->last_lop)
934 PL_parser->last_lop = buf + last_lop_pos;
939 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
941 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
942 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
943 reallocating the buffer if necessary. This means that lexing code that
944 runs later will see the characters as if they had appeared in the input.
945 It is not recommended to do this as part of normal parsing, and most
946 uses of this facility run the risk of the inserted characters being
947 interpreted in an unintended manner.
949 The string to be inserted is represented by I<len> octets starting
950 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
951 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
952 The characters are recoded for the lexer buffer, according to how the
953 buffer is currently being interpreted (L</lex_bufutf8>). If a string
954 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
955 function is more convenient.
961 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
965 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
966 if (flags & ~(LEX_STUFF_UTF8))
967 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
969 if (flags & LEX_STUFF_UTF8) {
973 const char *p, *e = pv+len;
974 for (p = pv; p != e; p++)
975 highhalf += !!(((U8)*p) & 0x80);
978 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
979 bufptr = PL_parser->bufptr;
980 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
981 SvCUR_set(PL_parser->linestr,
982 SvCUR(PL_parser->linestr) + len+highhalf);
983 PL_parser->bufend += len+highhalf;
984 for (p = pv; p != e; p++) {
987 *bufptr++ = (char)(0xc0 | (c >> 6));
988 *bufptr++ = (char)(0x80 | (c & 0x3f));
995 if (flags & LEX_STUFF_UTF8) {
997 const char *p, *e = pv+len;
998 for (p = pv; p != e; p++) {
1001 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1002 "non-Latin-1 character into Latin-1 input");
1003 } else if (c >= 0xc2 && p+1 != e &&
1004 (((U8)p[1]) & 0xc0) == 0x80) {
1007 } else if (c >= 0x80) {
1008 /* malformed UTF-8 */
1010 SAVESPTR(PL_warnhook);
1011 PL_warnhook = PERL_WARNHOOK_FATAL;
1012 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1018 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1019 bufptr = PL_parser->bufptr;
1020 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1021 SvCUR_set(PL_parser->linestr,
1022 SvCUR(PL_parser->linestr) + len-highhalf);
1023 PL_parser->bufend += len-highhalf;
1024 for (p = pv; p != e; p++) {
1027 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1030 *bufptr++ = (char)c;
1035 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1036 bufptr = PL_parser->bufptr;
1037 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1038 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1039 PL_parser->bufend += len;
1040 Copy(pv, bufptr, len, char);
1046 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1048 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1049 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1050 reallocating the buffer if necessary. This means that lexing code that
1051 runs later will see the characters as if they had appeared in the input.
1052 It is not recommended to do this as part of normal parsing, and most
1053 uses of this facility run the risk of the inserted characters being
1054 interpreted in an unintended manner.
1056 The string to be inserted is represented by octets starting at I<pv>
1057 and continuing to the first nul. These octets are interpreted as either
1058 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1059 in I<flags>. The characters are recoded for the lexer buffer, according
1060 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1061 If it is not convenient to nul-terminate a string to be inserted, the
1062 L</lex_stuff_pvn> function is more appropriate.
1068 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1070 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1071 lex_stuff_pvn(pv, strlen(pv), flags);
1075 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1077 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1078 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1079 reallocating the buffer if necessary. This means that lexing code that
1080 runs later will see the characters as if they had appeared in the input.
1081 It is not recommended to do this as part of normal parsing, and most
1082 uses of this facility run the risk of the inserted characters being
1083 interpreted in an unintended manner.
1085 The string to be inserted is the string value of I<sv>. The characters
1086 are recoded for the lexer buffer, according to how the buffer is currently
1087 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1088 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1089 need to construct a scalar.
1095 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1099 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1101 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1103 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1107 =for apidoc Amx|void|lex_unstuff|char *ptr
1109 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1110 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1111 This hides the discarded text from any lexing code that runs later,
1112 as if the text had never appeared.
1114 This is not the normal way to consume lexed text. For that, use
1121 Perl_lex_unstuff(pTHX_ char *ptr)
1125 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1126 buf = PL_parser->bufptr;
1128 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1131 bufend = PL_parser->bufend;
1133 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1134 unstuff_len = ptr - buf;
1135 Move(ptr, buf, bufend+1-ptr, char);
1136 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1137 PL_parser->bufend = bufend - unstuff_len;
1141 =for apidoc Amx|void|lex_read_to|char *ptr
1143 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1144 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1145 performing the correct bookkeeping whenever a newline character is passed.
1146 This is the normal way to consume lexed text.
1148 Interpretation of the buffer's octets can be abstracted out by
1149 using the slightly higher-level functions L</lex_peek_unichar> and
1150 L</lex_read_unichar>.
1156 Perl_lex_read_to(pTHX_ char *ptr)
1159 PERL_ARGS_ASSERT_LEX_READ_TO;
1160 s = PL_parser->bufptr;
1161 if (ptr < s || ptr > PL_parser->bufend)
1162 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1163 for (; s != ptr; s++)
1165 CopLINE_inc(PL_curcop);
1166 PL_parser->linestart = s+1;
1168 PL_parser->bufptr = ptr;
1172 =for apidoc Amx|void|lex_discard_to|char *ptr
1174 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1175 up to I<ptr>. The remaining content of the buffer will be moved, and
1176 all pointers into the buffer updated appropriately. I<ptr> must not
1177 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1178 it is not permitted to discard text that has yet to be lexed.
1180 Normally it is not necessarily to do this directly, because it suffices to
1181 use the implicit discarding behaviour of L</lex_next_chunk> and things
1182 based on it. However, if a token stretches across multiple lines,
1183 and the lexing code has kept multiple lines of text in the buffer for
1184 that purpose, then after completion of the token it would be wise to
1185 explicitly discard the now-unneeded earlier lines, to avoid future
1186 multi-line tokens growing the buffer without bound.
1192 Perl_lex_discard_to(pTHX_ char *ptr)
1196 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1197 buf = SvPVX(PL_parser->linestr);
1199 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1202 if (ptr > PL_parser->bufptr)
1203 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1204 discard_len = ptr - buf;
1205 if (PL_parser->oldbufptr < ptr)
1206 PL_parser->oldbufptr = ptr;
1207 if (PL_parser->oldoldbufptr < ptr)
1208 PL_parser->oldoldbufptr = ptr;
1209 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1210 PL_parser->last_uni = NULL;
1211 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1212 PL_parser->last_lop = NULL;
1213 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1214 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1215 PL_parser->bufend -= discard_len;
1216 PL_parser->bufptr -= discard_len;
1217 PL_parser->oldbufptr -= discard_len;
1218 PL_parser->oldoldbufptr -= discard_len;
1219 if (PL_parser->last_uni)
1220 PL_parser->last_uni -= discard_len;
1221 if (PL_parser->last_lop)
1222 PL_parser->last_lop -= discard_len;
1226 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1228 Reads in the next chunk of text to be lexed, appending it to
1229 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1230 looked to the end of the current chunk and wants to know more. It is
1231 usual, but not necessary, for lexing to have consumed the entirety of
1232 the current chunk at this time.
1234 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1235 chunk (i.e., the current chunk has been entirely consumed), normally the
1236 current chunk will be discarded at the same time that the new chunk is
1237 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1238 will not be discarded. If the current chunk has not been entirely
1239 consumed, then it will not be discarded regardless of the flag.
1241 Returns true if some new text was added to the buffer, or false if the
1242 buffer has reached the end of the input text.
1247 #define LEX_FAKE_EOF 0x80000000
1250 Perl_lex_next_chunk(pTHX_ U32 flags)
1254 STRLEN old_bufend_pos, new_bufend_pos;
1255 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1256 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1257 bool got_some_for_debugger = 0;
1259 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1260 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1261 linestr = PL_parser->linestr;
1262 buf = SvPVX(linestr);
1263 if (!(flags & LEX_KEEP_PREVIOUS) &&
1264 PL_parser->bufptr == PL_parser->bufend) {
1265 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1267 if (PL_parser->last_uni != PL_parser->bufend)
1268 PL_parser->last_uni = NULL;
1269 if (PL_parser->last_lop != PL_parser->bufend)
1270 PL_parser->last_lop = NULL;
1271 last_uni_pos = last_lop_pos = 0;
1275 old_bufend_pos = PL_parser->bufend - buf;
1276 bufptr_pos = PL_parser->bufptr - buf;
1277 oldbufptr_pos = PL_parser->oldbufptr - buf;
1278 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1279 linestart_pos = PL_parser->linestart - buf;
1280 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1281 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1283 if (flags & LEX_FAKE_EOF) {
1285 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1287 } else if (filter_gets(linestr, old_bufend_pos)) {
1289 got_some_for_debugger = 1;
1291 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1292 sv_setpvs(linestr, "");
1294 /* End of real input. Close filehandle (unless it was STDIN),
1295 * then add implicit termination.
1297 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1298 PerlIO_clearerr(PL_parser->rsfp);
1299 else if (PL_parser->rsfp)
1300 (void)PerlIO_close(PL_parser->rsfp);
1301 PL_parser->rsfp = NULL;
1302 PL_parser->in_pod = PL_parser->filtered = 0;
1304 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1307 if (!PL_in_eval && PL_minus_p) {
1309 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1310 PL_minus_n = PL_minus_p = 0;
1311 } else if (!PL_in_eval && PL_minus_n) {
1312 sv_catpvs(linestr, /*{*/";}");
1315 sv_catpvs(linestr, ";");
1318 buf = SvPVX(linestr);
1319 new_bufend_pos = SvCUR(linestr);
1320 PL_parser->bufend = buf + new_bufend_pos;
1321 PL_parser->bufptr = buf + bufptr_pos;
1322 PL_parser->oldbufptr = buf + oldbufptr_pos;
1323 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1324 PL_parser->linestart = buf + linestart_pos;
1325 if (PL_parser->last_uni)
1326 PL_parser->last_uni = buf + last_uni_pos;
1327 if (PL_parser->last_lop)
1328 PL_parser->last_lop = buf + last_lop_pos;
1329 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1330 PL_curstash != PL_debstash) {
1331 /* debugger active and we're not compiling the debugger code,
1332 * so store the line into the debugger's array of lines
1334 update_debugger_info(NULL, buf+old_bufend_pos,
1335 new_bufend_pos-old_bufend_pos);
1341 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1343 Looks ahead one (Unicode) character in the text currently being lexed.
1344 Returns the codepoint (unsigned integer value) of the next character,
1345 or -1 if lexing has reached the end of the input text. To consume the
1346 peeked character, use L</lex_read_unichar>.
1348 If the next character is in (or extends into) the next chunk of input
1349 text, the next chunk will be read in. Normally the current chunk will be
1350 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1351 then the current chunk will not be discarded.
1353 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1354 is encountered, an exception is generated.
1360 Perl_lex_peek_unichar(pTHX_ U32 flags)
1364 if (flags & ~(LEX_KEEP_PREVIOUS))
1365 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1366 s = PL_parser->bufptr;
1367 bufend = PL_parser->bufend;
1373 if (!lex_next_chunk(flags))
1375 s = PL_parser->bufptr;
1376 bufend = PL_parser->bufend;
1382 len = PL_utf8skip[head];
1383 while ((STRLEN)(bufend-s) < len) {
1384 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1386 s = PL_parser->bufptr;
1387 bufend = PL_parser->bufend;
1390 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1391 if (retlen == (STRLEN)-1) {
1392 /* malformed UTF-8 */
1394 SAVESPTR(PL_warnhook);
1395 PL_warnhook = PERL_WARNHOOK_FATAL;
1396 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1402 if (!lex_next_chunk(flags))
1404 s = PL_parser->bufptr;
1411 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1413 Reads the next (Unicode) character in the text currently being lexed.
1414 Returns the codepoint (unsigned integer value) of the character read,
1415 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1416 if lexing has reached the end of the input text. To non-destructively
1417 examine the next character, use L</lex_peek_unichar> instead.
1419 If the next character is in (or extends into) the next chunk of input
1420 text, the next chunk will be read in. Normally the current chunk will be
1421 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1422 then the current chunk will not be discarded.
1424 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1425 is encountered, an exception is generated.
1431 Perl_lex_read_unichar(pTHX_ U32 flags)
1434 if (flags & ~(LEX_KEEP_PREVIOUS))
1435 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1436 c = lex_peek_unichar(flags);
1439 CopLINE_inc(PL_curcop);
1441 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1443 ++(PL_parser->bufptr);
1449 =for apidoc Amx|void|lex_read_space|U32 flags
1451 Reads optional spaces, in Perl style, in the text currently being
1452 lexed. The spaces may include ordinary whitespace characters and
1453 Perl-style comments. C<#line> directives are processed if encountered.
1454 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1455 at a non-space character (or the end of the input text).
1457 If spaces extend into the next chunk of input text, the next chunk will
1458 be read in. Normally the current chunk will be discarded at the same
1459 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1460 chunk will not be discarded.
1465 #define LEX_NO_NEXT_CHUNK 0x80000000
1468 Perl_lex_read_space(pTHX_ U32 flags)
1471 bool need_incline = 0;
1472 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1473 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1476 sv_free(PL_skipwhite);
1477 PL_skipwhite = NULL;
1480 PL_skipwhite = newSVpvs("");
1481 #endif /* PERL_MAD */
1482 s = PL_parser->bufptr;
1483 bufend = PL_parser->bufend;
1489 } while (!(c == '\n' || (c == 0 && s == bufend)));
1490 } else if (c == '\n') {
1492 PL_parser->linestart = s;
1497 } else if (isSPACE(c)) {
1499 } else if (c == 0 && s == bufend) {
1503 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1504 #endif /* PERL_MAD */
1505 if (flags & LEX_NO_NEXT_CHUNK)
1507 PL_parser->bufptr = s;
1508 CopLINE_inc(PL_curcop);
1509 got_more = lex_next_chunk(flags);
1510 CopLINE_dec(PL_curcop);
1511 s = PL_parser->bufptr;
1512 bufend = PL_parser->bufend;
1515 if (need_incline && PL_parser->rsfp) {
1525 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1526 #endif /* PERL_MAD */
1527 PL_parser->bufptr = s;
1532 * This subroutine has nothing to do with tilting, whether at windmills
1533 * or pinball tables. Its name is short for "increment line". It
1534 * increments the current line number in CopLINE(PL_curcop) and checks
1535 * to see whether the line starts with a comment of the form
1536 * # line 500 "foo.pm"
1537 * If so, it sets the current line number and file to the values in the comment.
1541 S_incline(pTHX_ const char *s)
1549 PERL_ARGS_ASSERT_INCLINE;
1551 CopLINE_inc(PL_curcop);
1554 while (SPACE_OR_TAB(*s))
1556 if (strnEQ(s, "line", 4))
1560 if (SPACE_OR_TAB(*s))
1564 while (SPACE_OR_TAB(*s))
1572 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1574 while (SPACE_OR_TAB(*s))
1576 if (*s == '"' && (t = strchr(s+1, '"'))) {
1582 while (!isSPACE(*t))
1586 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1588 if (*e != '\n' && *e != '\0')
1589 return; /* false alarm */
1591 line_num = atoi(n)-1;
1594 const STRLEN len = t - s;
1595 SV *const temp_sv = CopFILESV(PL_curcop);
1600 cf = SvPVX(temp_sv);
1601 tmplen = SvCUR(temp_sv);
1607 if (!PL_rsfp && !PL_parser->filtered) {
1608 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1609 * to *{"::_<newfilename"} */
1610 /* However, the long form of evals is only turned on by the
1611 debugger - usually they're "(eval %lu)" */
1615 STRLEN tmplen2 = len;
1616 if (tmplen + 2 <= sizeof smallbuf)
1619 Newx(tmpbuf, tmplen + 2, char);
1622 memcpy(tmpbuf + 2, cf, tmplen);
1624 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1629 if (tmplen2 + 2 <= sizeof smallbuf)
1632 Newx(tmpbuf2, tmplen2 + 2, char);
1634 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1635 /* Either they malloc'd it, or we malloc'd it,
1636 so no prefix is present in ours. */
1641 memcpy(tmpbuf2 + 2, s, tmplen2);
1644 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1646 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1647 /* adjust ${"::_<newfilename"} to store the new file name */
1648 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1649 /* The line number may differ. If that is the case,
1650 alias the saved lines that are in the array.
1651 Otherwise alias the whole array. */
1652 if (CopLINE(PL_curcop) == line_num) {
1653 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1654 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1656 else if (GvAV(*gvp)) {
1657 AV * const av = GvAV(*gvp);
1658 const I32 start = CopLINE(PL_curcop)+1;
1659 I32 items = AvFILLp(av) - start;
1661 AV * const av2 = GvAVn(gv2);
1662 SV **svp = AvARRAY(av) + start;
1663 I32 l = (I32)line_num+1;
1665 av_store(av2, l++, SvREFCNT_inc(*svp++));
1670 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1672 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1674 CopFILE_free(PL_curcop);
1675 CopFILE_setn(PL_curcop, s, len);
1677 CopLINE_set(PL_curcop, line_num);
1681 /* skip space before PL_thistoken */
1684 S_skipspace0(pTHX_ register char *s)
1686 PERL_ARGS_ASSERT_SKIPSPACE0;
1693 PL_thiswhite = newSVpvs("");
1694 sv_catsv(PL_thiswhite, PL_skipwhite);
1695 sv_free(PL_skipwhite);
1698 PL_realtokenstart = s - SvPVX(PL_linestr);
1702 /* skip space after PL_thistoken */
1705 S_skipspace1(pTHX_ register char *s)
1707 const char *start = s;
1708 I32 startoff = start - SvPVX(PL_linestr);
1710 PERL_ARGS_ASSERT_SKIPSPACE1;
1715 start = SvPVX(PL_linestr) + startoff;
1716 if (!PL_thistoken && PL_realtokenstart >= 0) {
1717 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1718 PL_thistoken = newSVpvn(tstart, start - tstart);
1720 PL_realtokenstart = -1;
1723 PL_nextwhite = newSVpvs("");
1724 sv_catsv(PL_nextwhite, PL_skipwhite);
1725 sv_free(PL_skipwhite);
1732 S_skipspace2(pTHX_ register char *s, SV **svp)
1735 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1736 const I32 startoff = s - SvPVX(PL_linestr);
1738 PERL_ARGS_ASSERT_SKIPSPACE2;
1741 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1742 if (!PL_madskills || !svp)
1744 start = SvPVX(PL_linestr) + startoff;
1745 if (!PL_thistoken && PL_realtokenstart >= 0) {
1746 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1747 PL_thistoken = newSVpvn(tstart, start - tstart);
1748 PL_realtokenstart = -1;
1752 *svp = newSVpvs("");
1753 sv_setsv(*svp, PL_skipwhite);
1754 sv_free(PL_skipwhite);
1763 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1765 AV *av = CopFILEAVx(PL_curcop);
1767 SV * const sv = newSV_type(SVt_PVMG);
1769 sv_setsv(sv, orig_sv);
1771 sv_setpvn(sv, buf, len);
1774 av_store(av, (I32)CopLINE(PL_curcop), sv);
1780 * Called to gobble the appropriate amount and type of whitespace.
1781 * Skips comments as well.
1785 S_skipspace(pTHX_ register char *s)
1789 #endif /* PERL_MAD */
1790 PERL_ARGS_ASSERT_SKIPSPACE;
1793 sv_free(PL_skipwhite);
1794 PL_skipwhite = NULL;
1796 #endif /* PERL_MAD */
1797 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1798 while (s < PL_bufend && SPACE_OR_TAB(*s))
1801 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1803 lex_read_space(LEX_KEEP_PREVIOUS |
1804 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1805 LEX_NO_NEXT_CHUNK : 0));
1807 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1808 if (PL_linestart > PL_bufptr)
1809 PL_bufptr = PL_linestart;
1814 PL_skipwhite = newSVpvn(start, s-start);
1815 #endif /* PERL_MAD */
1821 * Check the unary operators to ensure there's no ambiguity in how they're
1822 * used. An ambiguous piece of code would be:
1824 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1825 * the +5 is its argument.
1835 if (PL_oldoldbufptr != PL_last_uni)
1837 while (isSPACE(*PL_last_uni))
1840 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1842 if ((t = strchr(s, '(')) && t < PL_bufptr)
1845 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1846 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1847 (int)(s - PL_last_uni), PL_last_uni);
1851 * LOP : macro to build a list operator. Its behaviour has been replaced
1852 * with a subroutine, S_lop() for which LOP is just another name.
1855 #define LOP(f,x) return lop(f,x,s)
1859 * Build a list operator (or something that might be one). The rules:
1860 * - if we have a next token, then it's a list operator [why?]
1861 * - if the next thing is an opening paren, then it's a function
1862 * - else it's a list operator
1866 S_lop(pTHX_ I32 f, int x, char *s)
1870 PERL_ARGS_ASSERT_LOP;
1876 PL_last_lop = PL_oldbufptr;
1877 PL_last_lop_op = (OPCODE)f;
1886 return REPORT(FUNC);
1889 return REPORT(FUNC);
1892 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1893 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1894 return REPORT(LSTOP);
1901 * Sets up for an eventual force_next(). start_force(0) basically does
1902 * an unshift, while start_force(-1) does a push. yylex removes items
1907 S_start_force(pTHX_ int where)
1911 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1912 where = PL_lasttoke;
1913 assert(PL_curforce < 0 || PL_curforce == where);
1914 if (PL_curforce != where) {
1915 for (i = PL_lasttoke; i > where; --i) {
1916 PL_nexttoke[i] = PL_nexttoke[i-1];
1920 if (PL_curforce < 0) /* in case of duplicate start_force() */
1921 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1922 PL_curforce = where;
1925 curmad('^', newSVpvs(""));
1926 CURMAD('_', PL_nextwhite);
1931 S_curmad(pTHX_ char slot, SV *sv)
1937 if (PL_curforce < 0)
1938 where = &PL_thismad;
1940 where = &PL_nexttoke[PL_curforce].next_mad;
1946 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1948 else if (PL_encoding) {
1949 sv_recode_to_utf8(sv, PL_encoding);
1954 /* keep a slot open for the head of the list? */
1955 if (slot != '_' && *where && (*where)->mad_key == '^') {
1956 (*where)->mad_key = slot;
1957 sv_free(MUTABLE_SV(((*where)->mad_val)));
1958 (*where)->mad_val = (void*)sv;
1961 addmad(newMADsv(slot, sv), where, 0);
1964 # define start_force(where) NOOP
1965 # define curmad(slot, sv) NOOP
1970 * When the lexer realizes it knows the next token (for instance,
1971 * it is reordering tokens for the parser) then it can call S_force_next
1972 * to know what token to return the next time the lexer is called. Caller
1973 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1974 * and possibly PL_expect to ensure the lexer handles the token correctly.
1978 S_force_next(pTHX_ I32 type)
1983 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1984 tokereport(type, &NEXTVAL_NEXTTOKE);
1988 if (PL_curforce < 0)
1989 start_force(PL_lasttoke);
1990 PL_nexttoke[PL_curforce].next_type = type;
1991 if (PL_lex_state != LEX_KNOWNEXT)
1992 PL_lex_defer = PL_lex_state;
1993 PL_lex_state = LEX_KNOWNEXT;
1994 PL_lex_expect = PL_expect;
1997 PL_nexttype[PL_nexttoke] = type;
1999 if (PL_lex_state != LEX_KNOWNEXT) {
2000 PL_lex_defer = PL_lex_state;
2001 PL_lex_expect = PL_expect;
2002 PL_lex_state = LEX_KNOWNEXT;
2010 int yyc = PL_parser->yychar;
2011 if (yyc != YYEMPTY) {
2014 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2015 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2016 PL_lex_allbrackets--;
2018 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2019 } else if (yyc == '('/*)*/) {
2020 PL_lex_allbrackets--;
2025 PL_parser->yychar = YYEMPTY;
2030 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2033 SV * const sv = newSVpvn_utf8(start, len,
2036 && !is_ascii_string((const U8*)start, len)
2037 && is_utf8_string((const U8*)start, len));
2043 * When the lexer knows the next thing is a word (for instance, it has
2044 * just seen -> and it knows that the next char is a word char, then
2045 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2049 * char *start : buffer position (must be within PL_linestr)
2050 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2051 * int check_keyword : if true, Perl checks to make sure the word isn't
2052 * a keyword (do this if the word is a label, e.g. goto FOO)
2053 * int allow_pack : if true, : characters will also be allowed (require,
2054 * use, etc. do this)
2055 * int allow_initial_tick : used by the "sub" lexer only.
2059 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2065 PERL_ARGS_ASSERT_FORCE_WORD;
2067 start = SKIPSPACE1(start);
2069 if (isIDFIRST_lazy_if(s,UTF) ||
2070 (allow_pack && *s == ':') ||
2071 (allow_initial_tick && *s == '\'') )
2073 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2074 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2076 start_force(PL_curforce);
2078 curmad('X', newSVpvn(start,s-start));
2079 if (token == METHOD) {
2084 PL_expect = XOPERATOR;
2088 curmad('g', newSVpvs( "forced" ));
2089 NEXTVAL_NEXTTOKE.opval
2090 = (OP*)newSVOP(OP_CONST,0,
2091 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2092 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2100 * Called when the lexer wants $foo *foo &foo etc, but the program
2101 * text only contains the "foo" portion. The first argument is a pointer
2102 * to the "foo", and the second argument is the type symbol to prefix.
2103 * Forces the next token to be a "WORD".
2104 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2108 S_force_ident(pTHX_ register const char *s, int kind)
2112 PERL_ARGS_ASSERT_FORCE_IDENT;
2115 const STRLEN len = strlen(s);
2116 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2117 UTF ? SVf_UTF8 : 0));
2118 start_force(PL_curforce);
2119 NEXTVAL_NEXTTOKE.opval = o;
2122 o->op_private = OPpCONST_ENTERED;
2123 /* XXX see note in pp_entereval() for why we forgo typo
2124 warnings if the symbol must be introduced in an eval.
2126 gv_fetchpvn_flags(s, len,
2127 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2128 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2129 kind == '$' ? SVt_PV :
2130 kind == '@' ? SVt_PVAV :
2131 kind == '%' ? SVt_PVHV :
2139 Perl_str_to_version(pTHX_ SV *sv)
2144 const char *start = SvPV_const(sv,len);
2145 const char * const end = start + len;
2146 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2148 PERL_ARGS_ASSERT_STR_TO_VERSION;
2150 while (start < end) {
2154 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2159 retval += ((NV)n)/nshift;
2168 * Forces the next token to be a version number.
2169 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2170 * and if "guessing" is TRUE, then no new token is created (and the caller
2171 * must use an alternative parsing method).
2175 S_force_version(pTHX_ char *s, int guessing)
2181 I32 startoff = s - SvPVX(PL_linestr);
2184 PERL_ARGS_ASSERT_FORCE_VERSION;
2192 while (isDIGIT(*d) || *d == '_' || *d == '.')
2196 start_force(PL_curforce);
2197 curmad('X', newSVpvn(s,d-s));
2200 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2202 #ifdef USE_LOCALE_NUMERIC
2203 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2204 setlocale(LC_NUMERIC, "C");
2206 s = scan_num(s, &pl_yylval);
2207 #ifdef USE_LOCALE_NUMERIC
2208 setlocale(LC_NUMERIC, loc);
2211 version = pl_yylval.opval;
2212 ver = cSVOPx(version)->op_sv;
2213 if (SvPOK(ver) && !SvNIOK(ver)) {
2214 SvUPGRADE(ver, SVt_PVNV);
2215 SvNV_set(ver, str_to_version(ver));
2216 SvNOK_on(ver); /* hint that it is a version */
2219 else if (guessing) {
2222 sv_free(PL_nextwhite); /* let next token collect whitespace */
2224 s = SvPVX(PL_linestr) + startoff;
2232 if (PL_madskills && !version) {
2233 sv_free(PL_nextwhite); /* let next token collect whitespace */
2235 s = SvPVX(PL_linestr) + startoff;
2238 /* NOTE: The parser sees the package name and the VERSION swapped */
2239 start_force(PL_curforce);
2240 NEXTVAL_NEXTTOKE.opval = version;
2247 * S_force_strict_version
2248 * Forces the next token to be a version number using strict syntax rules.
2252 S_force_strict_version(pTHX_ char *s)
2257 I32 startoff = s - SvPVX(PL_linestr);
2259 const char *errstr = NULL;
2261 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2263 while (isSPACE(*s)) /* leading whitespace */
2266 if (is_STRICT_VERSION(s,&errstr)) {
2268 s = (char *)scan_version(s, ver, 0);
2269 version = newSVOP(OP_CONST, 0, ver);
2271 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2272 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2276 yyerror(errstr); /* version required */
2281 if (PL_madskills && !version) {
2282 sv_free(PL_nextwhite); /* let next token collect whitespace */
2284 s = SvPVX(PL_linestr) + startoff;
2287 /* NOTE: The parser sees the package name and the VERSION swapped */
2288 start_force(PL_curforce);
2289 NEXTVAL_NEXTTOKE.opval = version;
2297 * Tokenize a quoted string passed in as an SV. It finds the next
2298 * chunk, up to end of string or a backslash. It may make a new
2299 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2304 S_tokeq(pTHX_ SV *sv)
2308 register char *send;
2313 PERL_ARGS_ASSERT_TOKEQ;
2318 s = SvPV_force(sv, len);
2319 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2322 /* This is relying on the SV being "well formed" with a trailing '\0' */
2323 while (s < send && !(*s == '\\' && s[1] == '\\'))
2328 if ( PL_hints & HINT_NEW_STRING ) {
2329 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2333 if (s + 1 < send && (s[1] == '\\'))
2334 s++; /* all that, just for this */
2339 SvCUR_set(sv, d - SvPVX_const(sv));
2341 if ( PL_hints & HINT_NEW_STRING )
2342 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2347 * Now come three functions related to double-quote context,
2348 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2349 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2350 * interact with PL_lex_state, and create fake ( ... ) argument lists
2351 * to handle functions and concatenation.
2355 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2360 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2362 * Pattern matching will set PL_lex_op to the pattern-matching op to
2363 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2365 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2367 * Everything else becomes a FUNC.
2369 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2370 * had an OP_CONST or OP_READLINE). This just sets us up for a
2371 * call to S_sublex_push().
2375 S_sublex_start(pTHX)
2378 register const I32 op_type = pl_yylval.ival;
2380 if (op_type == OP_NULL) {
2381 pl_yylval.opval = PL_lex_op;
2385 if (op_type == OP_CONST || op_type == OP_READLINE) {
2386 SV *sv = tokeq(PL_lex_stuff);
2388 if (SvTYPE(sv) == SVt_PVIV) {
2389 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2391 const char * const p = SvPV_const(sv, len);
2392 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2396 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2397 PL_lex_stuff = NULL;
2398 /* Allow <FH> // "foo" */
2399 if (op_type == OP_READLINE)
2400 PL_expect = XTERMORDORDOR;
2403 else if (op_type == OP_BACKTICK && PL_lex_op) {
2404 /* readpipe() vas overriden */
2405 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2406 pl_yylval.opval = PL_lex_op;
2408 PL_lex_stuff = NULL;
2412 PL_sublex_info.super_state = PL_lex_state;
2413 PL_sublex_info.sub_inwhat = (U16)op_type;
2414 PL_sublex_info.sub_op = PL_lex_op;
2415 PL_lex_state = LEX_INTERPPUSH;
2419 pl_yylval.opval = PL_lex_op;
2429 * Create a new scope to save the lexing state. The scope will be
2430 * ended in S_sublex_done. Returns a '(', starting the function arguments
2431 * to the uc, lc, etc. found before.
2432 * Sets PL_lex_state to LEX_INTERPCONCAT.
2441 PL_lex_state = PL_sublex_info.super_state;
2442 SAVEBOOL(PL_lex_dojoin);
2443 SAVEI32(PL_lex_brackets);
2444 SAVEI32(PL_lex_allbrackets);
2445 SAVEI8(PL_lex_fakeeof);
2446 SAVEI32(PL_lex_casemods);
2447 SAVEI32(PL_lex_starts);
2448 SAVEI8(PL_lex_state);
2449 SAVEVPTR(PL_lex_inpat);
2450 SAVEI16(PL_lex_inwhat);
2451 SAVECOPLINE(PL_curcop);
2452 SAVEPPTR(PL_bufptr);
2453 SAVEPPTR(PL_bufend);
2454 SAVEPPTR(PL_oldbufptr);
2455 SAVEPPTR(PL_oldoldbufptr);
2456 SAVEPPTR(PL_last_lop);
2457 SAVEPPTR(PL_last_uni);
2458 SAVEPPTR(PL_linestart);
2459 SAVESPTR(PL_linestr);
2460 SAVEGENERICPV(PL_lex_brackstack);
2461 SAVEGENERICPV(PL_lex_casestack);
2463 PL_linestr = PL_lex_stuff;
2464 PL_lex_stuff = NULL;
2466 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2467 = SvPVX(PL_linestr);
2468 PL_bufend += SvCUR(PL_linestr);
2469 PL_last_lop = PL_last_uni = NULL;
2470 SAVEFREESV(PL_linestr);
2472 PL_lex_dojoin = FALSE;
2473 PL_lex_brackets = 0;
2474 PL_lex_allbrackets = 0;
2475 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2476 Newx(PL_lex_brackstack, 120, char);
2477 Newx(PL_lex_casestack, 12, char);
2478 PL_lex_casemods = 0;
2479 *PL_lex_casestack = '\0';
2481 PL_lex_state = LEX_INTERPCONCAT;
2482 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2484 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2485 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2486 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2487 PL_lex_inpat = PL_sublex_info.sub_op;
2489 PL_lex_inpat = NULL;
2496 * Restores lexer state after a S_sublex_push.
2503 if (!PL_lex_starts++) {
2504 SV * const sv = newSVpvs("");
2505 if (SvUTF8(PL_linestr))
2507 PL_expect = XOPERATOR;
2508 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2512 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2513 PL_lex_state = LEX_INTERPCASEMOD;
2517 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2518 assert(PL_lex_inwhat != OP_TRANSR);
2519 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2520 PL_linestr = PL_lex_repl;
2522 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2523 PL_bufend += SvCUR(PL_linestr);
2524 PL_last_lop = PL_last_uni = NULL;
2525 SAVEFREESV(PL_linestr);
2526 PL_lex_dojoin = FALSE;
2527 PL_lex_brackets = 0;
2528 PL_lex_allbrackets = 0;
2529 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2530 PL_lex_casemods = 0;
2531 *PL_lex_casestack = '\0';
2533 if (SvEVALED(PL_lex_repl)) {
2534 PL_lex_state = LEX_INTERPNORMAL;
2536 /* we don't clear PL_lex_repl here, so that we can check later
2537 whether this is an evalled subst; that means we rely on the
2538 logic to ensure sublex_done() is called again only via the
2539 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2542 PL_lex_state = LEX_INTERPCONCAT;
2552 PL_endwhite = newSVpvs("");
2553 sv_catsv(PL_endwhite, PL_thiswhite);
2557 sv_setpvs(PL_thistoken,"");
2559 PL_realtokenstart = -1;
2563 PL_bufend = SvPVX(PL_linestr);
2564 PL_bufend += SvCUR(PL_linestr);
2565 PL_expect = XOPERATOR;
2566 PL_sublex_info.sub_inwhat = 0;
2574 Extracts a pattern, double-quoted string, or transliteration. This
2577 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2578 processing a pattern (PL_lex_inpat is true), a transliteration
2579 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2581 Returns a pointer to the character scanned up to. If this is
2582 advanced from the start pointer supplied (i.e. if anything was
2583 successfully parsed), will leave an OP for the substring scanned
2584 in pl_yylval. Caller must intuit reason for not parsing further
2585 by looking at the next characters herself.
2589 constants: \N{NAME} only
2590 case and quoting: \U \Q \E
2591 stops on @ and $, but not for $ as tail anchor
2593 In transliterations:
2594 characters are VERY literal, except for - not at the start or end
2595 of the string, which indicates a range. If the range is in bytes,
2596 scan_const expands the range to the full set of intermediate
2597 characters. If the range is in utf8, the hyphen is replaced with
2598 a certain range mark which will be handled by pmtrans() in op.c.
2600 In double-quoted strings:
2602 double-quoted style: \r and \n
2603 constants: \x31, etc.
2604 deprecated backrefs: \1 (in substitution replacements)
2605 case and quoting: \U \Q \E
2608 scan_const does *not* construct ops to handle interpolated strings.
2609 It stops processing as soon as it finds an embedded $ or @ variable
2610 and leaves it to the caller to work out what's going on.
2612 embedded arrays (whether in pattern or not) could be:
2613 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2615 $ in double-quoted strings must be the symbol of an embedded scalar.
2617 $ in pattern could be $foo or could be tail anchor. Assumption:
2618 it's a tail anchor if $ is the last thing in the string, or if it's
2619 followed by one of "()| \r\n\t"
2621 \1 (backreferences) are turned into $1
2623 The structure of the code is
2624 while (there's a character to process) {
2625 handle transliteration ranges
2626 skip regexp comments /(?#comment)/ and codes /(?{code})/
2627 skip #-initiated comments in //x patterns
2628 check for embedded arrays
2629 check for embedded scalars
2631 deprecate \1 in substitution replacements
2632 handle string-changing backslashes \l \U \Q \E, etc.
2633 switch (what was escaped) {
2634 handle \- in a transliteration (becomes a literal -)
2635 if a pattern and not \N{, go treat as regular character
2636 handle \132 (octal characters)
2637 handle \x15 and \x{1234} (hex characters)
2638 handle \N{name} (named characters, also \N{3,5} in a pattern)
2639 handle \cV (control characters)
2640 handle printf-style backslashes (\f, \r, \n, etc)
2643 } (end if backslash)
2644 handle regular character
2645 } (end while character to read)
2650 S_scan_const(pTHX_ char *start)
2653 register char *send = PL_bufend; /* end of the constant */
2654 SV *sv = newSV(send - start); /* sv for the constant. See
2655 note below on sizing. */
2656 register char *s = start; /* start of the constant */
2657 register char *d = SvPVX(sv); /* destination for copies */
2658 bool dorange = FALSE; /* are we in a translit range? */
2659 bool didrange = FALSE; /* did we just finish a range? */
2660 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2661 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2662 to be UTF8? But, this can
2663 show as true when the source
2664 isn't utf8, as for example
2665 when it is entirely composed
2668 /* Note on sizing: The scanned constant is placed into sv, which is
2669 * initialized by newSV() assuming one byte of output for every byte of
2670 * input. This routine expects newSV() to allocate an extra byte for a
2671 * trailing NUL, which this routine will append if it gets to the end of
2672 * the input. There may be more bytes of input than output (eg., \N{LATIN
2673 * CAPITAL LETTER A}), or more output than input if the constant ends up
2674 * recoded to utf8, but each time a construct is found that might increase
2675 * the needed size, SvGROW() is called. Its size parameter each time is
2676 * based on the best guess estimate at the time, namely the length used so
2677 * far, plus the length the current construct will occupy, plus room for
2678 * the trailing NUL, plus one byte for every input byte still unscanned */
2682 UV literal_endpoint = 0;
2683 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2686 PERL_ARGS_ASSERT_SCAN_CONST;
2688 assert(PL_lex_inwhat != OP_TRANSR);
2689 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2690 /* If we are doing a trans and we know we want UTF8 set expectation */
2691 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2692 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2696 while (s < send || dorange) {
2698 /* get transliterations out of the way (they're most literal) */
2699 if (PL_lex_inwhat == OP_TRANS) {
2700 /* expand a range A-Z to the full set of characters. AIE! */
2702 I32 i; /* current expanded character */
2703 I32 min; /* first character in range */
2704 I32 max; /* last character in range */
2715 char * const c = (char*)utf8_hop((U8*)d, -1);
2719 *c = (char)UTF_TO_NATIVE(0xff);
2720 /* mark the range as done, and continue */
2726 i = d - SvPVX_const(sv); /* remember current offset */
2729 SvLEN(sv) + (has_utf8 ?
2730 (512 - UTF_CONTINUATION_MARK +
2733 /* How many two-byte within 0..255: 128 in UTF-8,
2734 * 96 in UTF-8-mod. */
2736 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2738 d = SvPVX(sv) + i; /* refresh d after realloc */
2742 for (j = 0; j <= 1; j++) {
2743 char * const c = (char*)utf8_hop((U8*)d, -1);
2744 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2750 max = (U8)0xff; /* only to \xff */
2751 uvmax = uv; /* \x{100} to uvmax */
2753 d = c; /* eat endpoint chars */
2758 d -= 2; /* eat the first char and the - */
2759 min = (U8)*d; /* first char in range */
2760 max = (U8)d[1]; /* last char in range */
2767 "Invalid range \"%c-%c\" in transliteration operator",
2768 (char)min, (char)max);
2772 if (literal_endpoint == 2 &&
2773 ((isLOWER(min) && isLOWER(max)) ||
2774 (isUPPER(min) && isUPPER(max)))) {
2776 for (i = min; i <= max; i++)
2778 *d++ = NATIVE_TO_NEED(has_utf8,i);
2780 for (i = min; i <= max; i++)
2782 *d++ = NATIVE_TO_NEED(has_utf8,i);
2787 for (i = min; i <= max; i++)
2790 const U8 ch = (U8)NATIVE_TO_UTF(i);
2791 if (UNI_IS_INVARIANT(ch))
2794 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2795 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2804 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2806 *d++ = (char)UTF_TO_NATIVE(0xff);
2808 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2812 /* mark the range as done, and continue */
2816 literal_endpoint = 0;
2821 /* range begins (ignore - as first or last char) */
2822 else if (*s == '-' && s+1 < send && s != start) {
2824 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2831 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2841 literal_endpoint = 0;
2842 native_range = TRUE;
2847 /* if we get here, we're not doing a transliteration */
2849 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2850 except for the last char, which will be done separately. */
2851 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2853 while (s+1 < send && *s != ')')
2854 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2856 else if (s[2] == '{' /* This should match regcomp.c */
2857 || (s[2] == '?' && s[3] == '{'))
2860 char *regparse = s + (s[2] == '{' ? 3 : 4);
2863 while (count && (c = *regparse)) {
2864 if (c == '\\' && regparse[1])
2872 if (*regparse != ')')
2873 regparse--; /* Leave one char for continuation. */
2874 while (s < regparse)
2875 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2879 /* likewise skip #-initiated comments in //x patterns */
2880 else if (*s == '#' && PL_lex_inpat &&
2881 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2882 while (s+1 < send && *s != '\n')
2883 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2886 /* check for embedded arrays
2887 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2889 else if (*s == '@' && s[1]) {
2890 if (isALNUM_lazy_if(s+1,UTF))
2892 if (strchr(":'{$", s[1]))
2894 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2895 break; /* in regexp, neither @+ nor @- are interpolated */
2898 /* check for embedded scalars. only stop if we're sure it's a
2901 else if (*s == '$') {
2902 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2904 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2906 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2907 "Possible unintended interpolation of $\\ in regex");
2909 break; /* in regexp, $ might be tail anchor */
2913 /* End of else if chain - OP_TRANS rejoin rest */
2916 if (*s == '\\' && s+1 < send) {
2917 char* e; /* Can be used for ending '}', etc. */
2921 /* warn on \1 - \9 in substitution replacements, but note that \11
2922 * is an octal; and \19 is \1 followed by '9' */
2923 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2924 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2926 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2931 /* string-change backslash escapes */
2932 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
2936 /* In a pattern, process \N, but skip any other backslash escapes.
2937 * This is because we don't want to translate an escape sequence
2938 * into a meta symbol and have the regex compiler use the meta
2939 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2940 * in spite of this, we do have to process \N here while the proper
2941 * charnames handler is in scope. See bugs #56444 and #62056.
2942 * There is a complication because \N in a pattern may also stand
2943 * for 'match a non-nl', and not mean a charname, in which case its
2944 * processing should be deferred to the regex compiler. To be a
2945 * charname it must be followed immediately by a '{', and not look
2946 * like \N followed by a curly quantifier, i.e., not something like
2947 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2949 else if (PL_lex_inpat
2952 || regcurly(s + 1)))
2954 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2955 goto default_action;
2960 /* quoted - in transliterations */
2962 if (PL_lex_inwhat == OP_TRANS) {
2969 if ((isALPHA(*s) || isDIGIT(*s)))
2970 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2971 "Unrecognized escape \\%c passed through",
2973 /* default action is to copy the quoted character */
2974 goto default_action;
2977 /* eg. \132 indicates the octal constant 0132 */
2978 case '0': case '1': case '2': case '3':
2979 case '4': case '5': case '6': case '7':
2983 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2986 goto NUM_ESCAPE_INSERT;
2988 /* eg. \o{24} indicates the octal constant \024 */
2994 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
3000 goto NUM_ESCAPE_INSERT;
3003 /* eg. \x24 indicates the hex constant 0x24 */
3007 char* const e = strchr(s, '}');
3008 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
3009 PERL_SCAN_DISALLOW_PREFIX;
3014 yyerror("Missing right brace on \\x{}");
3018 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3024 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3025 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3031 /* Insert oct or hex escaped character. There will always be
3032 * enough room in sv since such escapes will be longer than any
3033 * UTF-8 sequence they can end up as, except if they force us
3034 * to recode the rest of the string into utf8 */
3036 /* Here uv is the ordinal of the next character being added in
3037 * unicode (converted from native). */
3038 if (!UNI_IS_INVARIANT(uv)) {
3039 if (!has_utf8 && uv > 255) {
3040 /* Might need to recode whatever we have accumulated so
3041 * far if it contains any chars variant in utf8 or
3044 SvCUR_set(sv, d - SvPVX_const(sv));
3047 /* See Note on sizing above. */
3048 sv_utf8_upgrade_flags_grow(sv,
3049 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3050 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3051 d = SvPVX(sv) + SvCUR(sv);
3056 d = (char*)uvuni_to_utf8((U8*)d, uv);
3057 if (PL_lex_inwhat == OP_TRANS &&
3058 PL_sublex_info.sub_op) {
3059 PL_sublex_info.sub_op->op_private |=
3060 (PL_lex_repl ? OPpTRANS_FROM_UTF
3064 if (uv > 255 && !dorange)
3065 native_range = FALSE;
3078 /* In a non-pattern \N must be a named character, like \N{LATIN
3079 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3080 * mean to match a non-newline. For non-patterns, named
3081 * characters are converted to their string equivalents. In
3082 * patterns, named characters are not converted to their
3083 * ultimate forms for the same reasons that other escapes
3084 * aren't. Instead, they are converted to the \N{U+...} form
3085 * to get the value from the charnames that is in effect right
3086 * now, while preserving the fact that it was a named character
3087 * so that the regex compiler knows this */
3089 /* This section of code doesn't generally use the
3090 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3091 * a close examination of this macro and determined it is a
3092 * no-op except on utfebcdic variant characters. Every
3093 * character generated by this that would normally need to be
3094 * enclosed by this macro is invariant, so the macro is not
3095 * needed, and would complicate use of copy(). XXX There are
3096 * other parts of this file where the macro is used
3097 * inconsistently, but are saved by it being a no-op */
3099 /* The structure of this section of code (besides checking for
3100 * errors and upgrading to utf8) is:
3101 * Further disambiguate between the two meanings of \N, and if
3102 * not a charname, go process it elsewhere
3103 * If of form \N{U+...}, pass it through if a pattern;
3104 * otherwise convert to utf8
3105 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3106 * pattern; otherwise convert to utf8 */
3108 /* Here, s points to the 'N'; the test below is guaranteed to
3109 * succeed if we are being called on a pattern as we already
3110 * know from a test above that the next character is a '{'.
3111 * On a non-pattern \N must mean 'named sequence, which
3112 * requires braces */
3115 yyerror("Missing braces on \\N{}");
3120 /* If there is no matching '}', it is an error. */
3121 if (! (e = strchr(s, '}'))) {
3122 if (! PL_lex_inpat) {
3123 yyerror("Missing right brace on \\N{}");
3125 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3130 /* Here it looks like a named character */
3134 /* XXX This block is temporary code. \N{} implies that the
3135 * pattern is to have Unicode semantics, and therefore
3136 * currently has to be encoded in utf8. By putting it in
3137 * utf8 now, we save a whole pass in the regular expression
3138 * compiler. Once that code is changed so Unicode
3139 * semantics doesn't necessarily have to be in utf8, this
3140 * block should be removed. However, the code that parses
3141 * the output of this would have to be changed to not
3142 * necessarily expect utf8 */
3144 SvCUR_set(sv, d - SvPVX_const(sv));
3147 /* See Note on sizing above. */
3148 sv_utf8_upgrade_flags_grow(sv,
3149 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3150 /* 5 = '\N{' + cur char + NUL */
3151 (STRLEN)(send - s) + 5);
3152 d = SvPVX(sv) + SvCUR(sv);
3157 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3158 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3159 | PERL_SCAN_DISALLOW_PREFIX;
3162 /* For \N{U+...}, the '...' is a unicode value even on
3163 * EBCDIC machines */
3164 s += 2; /* Skip to next char after the 'U+' */
3166 uv = grok_hex(s, &len, &flags, NULL);
3167 if (len == 0 || len != (STRLEN)(e - s)) {
3168 yyerror("Invalid hexadecimal number in \\N{U+...}");
3175 /* On non-EBCDIC platforms, pass through to the regex
3176 * compiler unchanged. The reason we evaluated the
3177 * number above is to make sure there wasn't a syntax
3178 * error. But on EBCDIC we convert to native so
3179 * downstream code can continue to assume it's native
3181 s -= 5; /* Include the '\N{U+' */
3183 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3186 (unsigned int) UNI_TO_NATIVE(uv));
3188 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3192 else { /* Not a pattern: convert the hex to string */
3194 /* If destination is not in utf8, unconditionally
3195 * recode it to be so. This is because \N{} implies
3196 * Unicode semantics, and scalars have to be in utf8
3197 * to guarantee those semantics */
3199 SvCUR_set(sv, d - SvPVX_const(sv));
3202 /* See Note on sizing above. */
3203 sv_utf8_upgrade_flags_grow(
3205 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3206 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3207 d = SvPVX(sv) + SvCUR(sv);
3211 /* Add the string to the output */
3212 if (UNI_IS_INVARIANT(uv)) {
3215 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3218 else { /* Here is \N{NAME} but not \N{U+...}. */
3220 SV *res; /* result from charnames */
3221 const char *str; /* the string in 'res' */
3222 STRLEN len; /* its length */
3224 /* Get the value for NAME */
3225 res = newSVpvn(s, e - s);
3226 res = new_constant( NULL, 0, "charnames",
3227 /* includes all of: \N{...} */
3228 res, NULL, s - 3, e - s + 4 );
3230 /* Most likely res will be in utf8 already since the
3231 * standard charnames uses pack U, but a custom translator
3232 * can leave it otherwise, so make sure. XXX This can be
3233 * revisited to not have charnames use utf8 for characters
3234 * that don't need it when regexes don't have to be in utf8
3235 * for Unicode semantics. If doing so, remember EBCDIC */
3236 sv_utf8_upgrade(res);
3237 str = SvPV_const(res, len);
3239 /* Don't accept malformed input */
3240 if (! is_utf8_string((U8 *) str, len)) {
3241 yyerror("Malformed UTF-8 returned by \\N");
3243 else if (PL_lex_inpat) {
3245 if (! len) { /* The name resolved to an empty string */
3246 Copy("\\N{}", d, 4, char);
3250 /* In order to not lose information for the regex
3251 * compiler, pass the result in the specially made
3252 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3253 * the code points in hex of each character
3254 * returned by charnames */
3256 const char *str_end = str + len;
3257 STRLEN char_length; /* cur char's byte length */
3258 STRLEN output_length; /* and the number of bytes
3259 after this is translated
3261 const STRLEN off = d - SvPVX_const(sv);
3263 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3264 * max('U+', '.'); and 1 for NUL */
3265 char hex_string[2 * UTF8_MAXBYTES + 5];
3267 /* Get the first character of the result. */
3268 U32 uv = utf8n_to_uvuni((U8 *) str,
3273 /* The call to is_utf8_string() above hopefully
3274 * guarantees that there won't be an error. But
3275 * it's easy here to make sure. The function just
3276 * above warns and returns 0 if invalid utf8, but
3277 * it can also return 0 if the input is validly a
3278 * NUL. Disambiguate */
3279 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3280 uv = UNICODE_REPLACEMENT;
3283 /* Convert first code point to hex, including the
3284 * boiler plate before it. For all these, we
3285 * convert to native format so that downstream code
3286 * can continue to assume the input is native */
3288 my_snprintf(hex_string, sizeof(hex_string),
3290 (unsigned int) UNI_TO_NATIVE(uv));
3292 /* Make sure there is enough space to hold it */
3293 d = off + SvGROW(sv, off
3295 + (STRLEN)(send - e)
3296 + 2); /* '}' + NUL */
3298 Copy(hex_string, d, output_length, char);
3301 /* For each subsequent character, append dot and
3302 * its ordinal in hex */
3303 while ((str += char_length) < str_end) {
3304 const STRLEN off = d - SvPVX_const(sv);
3305 U32 uv = utf8n_to_uvuni((U8 *) str,
3309 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3310 uv = UNICODE_REPLACEMENT;
3314 my_snprintf(hex_string, sizeof(hex_string),
3316 (unsigned int) UNI_TO_NATIVE(uv));
3318 d = off + SvGROW(sv, off
3320 + (STRLEN)(send - e)
3321 + 2); /* '}' + NUL */
3322 Copy(hex_string, d, output_length, char);
3326 *d++ = '}'; /* Done. Add the trailing brace */
3329 else { /* Here, not in a pattern. Convert the name to a
3332 /* If destination is not in utf8, unconditionally
3333 * recode it to be so. This is because \N{} implies
3334 * Unicode semantics, and scalars have to be in utf8
3335 * to guarantee those semantics */
3337 SvCUR_set(sv, d - SvPVX_const(sv));
3340 /* See Note on sizing above. */
3341 sv_utf8_upgrade_flags_grow(sv,
3342 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3343 len + (STRLEN)(send - s) + 1);
3344 d = SvPVX(sv) + SvCUR(sv);
3346 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3348 /* See Note on sizing above. (NOTE: SvCUR() is not
3349 * set correctly here). */
3350 const STRLEN off = d - SvPVX_const(sv);
3351 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3353 Copy(str, d, len, char);
3358 /* Deprecate non-approved name syntax */
3359 if (ckWARN_d(WARN_DEPRECATED)) {
3360 bool problematic = FALSE;
3363 /* For non-ut8 input, look to see that the first
3364 * character is an alpha, then loop through the rest
3365 * checking that each is a continuation */
3367 if (! isALPHAU(*i)) problematic = TRUE;
3368 else for (i = s + 1; i < e; i++) {
3369 if (isCHARNAME_CONT(*i)) continue;
3375 /* Similarly for utf8. For invariants can check
3376 * directly. We accept anything above the latin1
3377 * range because it is immaterial to Perl if it is
3378 * correct or not, and is expensive to check. But
3379 * it is fairly easy in the latin1 range to convert
3380 * the variants into a single character and check
3382 if (UTF8_IS_INVARIANT(*i)) {
3383 if (! isALPHAU(*i)) problematic = TRUE;
3384 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3385 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3391 if (! problematic) for (i = s + UTF8SKIP(s);
3395 if (UTF8_IS_INVARIANT(*i)) {
3396 if (isCHARNAME_CONT(*i)) continue;
3397 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3399 } else if (isCHARNAME_CONT(
3401 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3410 /* The e-i passed to the final %.*s makes sure that
3411 * should the trailing NUL be missing that this
3412 * print won't run off the end of the string */
3413 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3414 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3415 (int)(i - s + 1), s, (int)(e - i), i + 1);
3418 } /* End \N{NAME} */
3421 native_range = FALSE; /* \N{} is defined to be Unicode */
3423 s = e + 1; /* Point to just after the '}' */
3426 /* \c is a control character */
3430 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3433 yyerror("Missing control char name in \\c");
3437 /* printf-style backslashes, formfeeds, newlines, etc */
3439 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3442 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3445 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3448 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3451 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3454 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3457 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3463 } /* end if (backslash) */
3470 /* If we started with encoded form, or already know we want it,
3471 then encode the next character */
3472 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3476 /* One might think that it is wasted effort in the case of the
3477 * source being utf8 (this_utf8 == TRUE) to take the next character
3478 * in the source, convert it to an unsigned value, and then convert
3479 * it back again. But the source has not been validated here. The
3480 * routine that does the conversion checks for errors like
3483 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3484 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3486 SvCUR_set(sv, d - SvPVX_const(sv));
3489 /* See Note on sizing above. */
3490 sv_utf8_upgrade_flags_grow(sv,
3491 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3492 need + (STRLEN)(send - s) + 1);
3493 d = SvPVX(sv) + SvCUR(sv);
3495 } else if (need > len) {
3496 /* encoded value larger than old, may need extra space (NOTE:
3497 * SvCUR() is not set correctly here). See Note on sizing
3499 const STRLEN off = d - SvPVX_const(sv);
3500 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3504 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3506 if (uv > 255 && !dorange)
3507 native_range = FALSE;
3511 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3513 } /* while loop to process each character */
3515 /* terminate the string and set up the sv */
3517 SvCUR_set(sv, d - SvPVX_const(sv));
3518 if (SvCUR(sv) >= SvLEN(sv))
3519 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3520 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3523 if (PL_encoding && !has_utf8) {
3524 sv_recode_to_utf8(sv, PL_encoding);
3530 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3531 PL_sublex_info.sub_op->op_private |=
3532 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3536 /* shrink the sv if we allocated more than we used */
3537 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3538 SvPV_shrink_to_cur(sv);
3541 /* return the substring (via pl_yylval) only if we parsed anything */
3542 if (s > PL_bufptr) {
3543 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3544 const char *const key = PL_lex_inpat ? "qr" : "q";
3545 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3549 if (PL_lex_inwhat == OP_TRANS) {
3552 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3560 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3563 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3570 * Returns TRUE if there's more to the expression (e.g., a subscript),
3573 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3575 * ->[ and ->{ return TRUE
3576 * { and [ outside a pattern are always subscripts, so return TRUE
3577 * if we're outside a pattern and it's not { or [, then return FALSE
3578 * if we're in a pattern and the first char is a {
3579 * {4,5} (any digits around the comma) returns FALSE
3580 * if we're in a pattern and the first char is a [
3582 * [SOMETHING] has a funky algorithm to decide whether it's a
3583 * character class or not. It has to deal with things like
3584 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3585 * anything else returns TRUE
3588 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3591 S_intuit_more(pTHX_ register char *s)
3595 PERL_ARGS_ASSERT_INTUIT_MORE;
3597 if (PL_lex_brackets)
3599 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3601 if (*s != '{' && *s != '[')
3606 /* In a pattern, so maybe we have {n,m}. */
3614 /* On the other hand, maybe we have a character class */
3617 if (*s == ']' || *s == '^')
3620 /* this is terrifying, and it works */
3621 int weight = 2; /* let's weigh the evidence */
3623 unsigned char un_char = 255, last_un_char;
3624 const char * const send = strchr(s,']');
3625 char tmpbuf[sizeof PL_tokenbuf * 4];
3627 if (!send) /* has to be an expression */
3630 Zero(seen,256,char);
3633 else if (isDIGIT(*s)) {
3635 if (isDIGIT(s[1]) && s[2] == ']')
3641 for (; s < send; s++) {
3642 last_un_char = un_char;
3643 un_char = (unsigned char)*s;
3648 weight -= seen[un_char] * 10;
3649 if (isALNUM_lazy_if(s+1,UTF)) {
3651 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3652 len = (int)strlen(tmpbuf);
3653 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3654 UTF ? SVf_UTF8 : 0, SVt_PV))
3659 else if (*s == '$' && s[1] &&
3660 strchr("[#!%*<>()-=",s[1])) {
3661 if (/*{*/ strchr("])} =",s[2]))
3670 if (strchr("wds]",s[1]))
3672 else if (seen[(U8)'\''] || seen[(U8)'"'])
3674 else if (strchr("rnftbxcav",s[1]))
3676 else if (isDIGIT(s[1])) {
3678 while (s[1] && isDIGIT(s[1]))
3688 if (strchr("aA01! ",last_un_char))
3690 if (strchr("zZ79~",s[1]))
3692 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3693 weight -= 5; /* cope with negative subscript */
3696 if (!isALNUM(last_un_char)
3697 && !(last_un_char == '$' || last_un_char == '@'
3698 || last_un_char == '&')
3699 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3704 if (keyword(tmpbuf, d - tmpbuf, 0))
3707 if (un_char == last_un_char + 1)
3709 weight -= seen[un_char];
3714 if (weight >= 0) /* probably a character class */
3724 * Does all the checking to disambiguate
3726 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3727 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3729 * First argument is the stuff after the first token, e.g. "bar".
3731 * Not a method if bar is a filehandle.
3732 * Not a method if foo is a subroutine prototyped to take a filehandle.
3733 * Not a method if it's really "Foo $bar"
3734 * Method if it's "foo $bar"
3735 * Not a method if it's really "print foo $bar"
3736 * Method if it's really "foo package::" (interpreted as package->foo)
3737 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3738 * Not a method if bar is a filehandle or package, but is quoted with
3743 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3746 char *s = start + (*start == '$');
3747 char tmpbuf[sizeof PL_tokenbuf];
3754 PERL_ARGS_ASSERT_INTUIT_METHOD;
3757 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3761 const char *proto = CvPROTO(cv);
3772 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3773 /* start is the beginning of the possible filehandle/object,
3774 * and s is the end of it
3775 * tmpbuf is a copy of it
3778 if (*start == '$') {
3779 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3780 isUPPER(*PL_tokenbuf))
3783 len = start - SvPVX(PL_linestr);
3787 start = SvPVX(PL_linestr) + len;
3791 return *s == '(' ? FUNCMETH : METHOD;
3793 if (!keyword(tmpbuf, len, 0)) {
3794 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3798 soff = s - SvPVX(PL_linestr);
3802 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3803 if (indirgv && GvCVu(indirgv))
3805 /* filehandle or package name makes it a method */
3806 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3808 soff = s - SvPVX(PL_linestr);
3811 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3812 return 0; /* no assumptions -- "=>" quotes bareword */
3814 start_force(PL_curforce);
3815 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3816 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3817 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3819 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
3820 ( UTF ? SVf_UTF8 : 0 )));
3825 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3827 return *s == '(' ? FUNCMETH : METHOD;
3833 /* Encoded script support. filter_add() effectively inserts a
3834 * 'pre-processing' function into the current source input stream.
3835 * Note that the filter function only applies to the current source file
3836 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3838 * The datasv parameter (which may be NULL) can be used to pass
3839 * private data to this instance of the filter. The filter function
3840 * can recover the SV using the FILTER_DATA macro and use it to
3841 * store private buffers and state information.
3843 * The supplied datasv parameter is upgraded to a PVIO type
3844 * and the IoDIRP/IoANY field is used to store the function pointer,
3845 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3846 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3847 * private use must be set using malloc'd pointers.
3851 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3860 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3861 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3863 if (!PL_rsfp_filters)
3864 PL_rsfp_filters = newAV();
3867 SvUPGRADE(datasv, SVt_PVIO);
3868 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3869 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3870 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3871 FPTR2DPTR(void *, IoANY(datasv)),
3872 SvPV_nolen(datasv)));
3873 av_unshift(PL_rsfp_filters, 1);
3874 av_store(PL_rsfp_filters, 0, datasv) ;
3876 !PL_parser->filtered
3877 && PL_parser->lex_flags & LEX_EVALBYTES
3878 && PL_bufptr < PL_bufend
3880 const char *s = PL_bufptr;
3881 while (s < PL_bufend) {
3883 SV *linestr = PL_parser->linestr;
3884 char *buf = SvPVX(linestr);
3885 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3886 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3887 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3888 STRLEN const linestart_pos = PL_parser->linestart - buf;
3889 STRLEN const last_uni_pos =
3890 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3891 STRLEN const last_lop_pos =
3892 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3893 av_push(PL_rsfp_filters, linestr);
3894 PL_parser->linestr =
3895 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3896 buf = SvPVX(PL_parser->linestr);
3897 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3898 PL_parser->bufptr = buf + bufptr_pos;
3899 PL_parser->oldbufptr = buf + oldbufptr_pos;
3900 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3901 PL_parser->linestart = buf + linestart_pos;
3902 if (PL_parser->last_uni)
3903 PL_parser->last_uni = buf + last_uni_pos;
3904 if (PL_parser->last_lop)
3905 PL_parser->last_lop = buf + last_lop_pos;
3906 SvLEN(linestr) = SvCUR(linestr);
3907 SvCUR(linestr) = s-SvPVX(linestr);
3908 PL_parser->filtered = 1;
3918 /* Delete most recently added instance of this filter function. */
3920 Perl_filter_del(pTHX_ filter_t funcp)
3925 PERL_ARGS_ASSERT_FILTER_DEL;
3928 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3929 FPTR2DPTR(void*, funcp)));
3931 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3933 /* if filter is on top of stack (usual case) just pop it off */
3934 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3935 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3936 sv_free(av_pop(PL_rsfp_filters));
3940 /* we need to search for the correct entry and clear it */
3941 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3945 /* Invoke the idxth filter function for the current rsfp. */
3946 /* maxlen 0 = read one text line */
3948 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3953 /* This API is bad. It should have been using unsigned int for maxlen.
3954 Not sure if we want to change the API, but if not we should sanity
3955 check the value here. */
3956 unsigned int correct_length
3965 PERL_ARGS_ASSERT_FILTER_READ;
3967 if (!PL_parser || !PL_rsfp_filters)
3969 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3970 /* Provide a default input filter to make life easy. */
3971 /* Note that we append to the line. This is handy. */
3972 DEBUG_P(PerlIO_printf(Perl_debug_log,
3973 "filter_read %d: from rsfp\n", idx));
3974 if (correct_length) {
3977 const int old_len = SvCUR(buf_sv);
3979 /* ensure buf_sv is large enough */
3980 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3981 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3982 correct_length)) <= 0) {
3983 if (PerlIO_error(PL_rsfp))
3984 return -1; /* error */
3986 return 0 ; /* end of file */
3988 SvCUR_set(buf_sv, old_len + len) ;
3989 SvPVX(buf_sv)[old_len + len] = '\0';
3992 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3993 if (PerlIO_error(PL_rsfp))
3994 return -1; /* error */
3996 return 0 ; /* end of file */
3999 return SvCUR(buf_sv);
4001 /* Skip this filter slot if filter has been deleted */
4002 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4003 DEBUG_P(PerlIO_printf(Perl_debug_log,
4004 "filter_read %d: skipped (filter deleted)\n",
4006 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4008 if (SvTYPE(datasv) != SVt_PVIO) {
4009 if (correct_length) {
4011 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4012 if (!remainder) return 0; /* eof */
4013 if (correct_length > remainder) correct_length = remainder;
4014 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4015 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4018 const char *s = SvEND(datasv);
4019 const char *send = SvPVX(datasv) + SvLEN(datasv);
4027 if (s == send) return 0; /* eof */
4028 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4029 SvCUR_set(datasv, s-SvPVX(datasv));
4031 return SvCUR(buf_sv);
4033 /* Get function pointer hidden within datasv */
4034 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4035 DEBUG_P(PerlIO_printf(Perl_debug_log,
4036 "filter_read %d: via function %p (%s)\n",
4037 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4038 /* Call function. The function is expected to */
4039 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4040 /* Return: <0:error, =0:eof, >0:not eof */
4041 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4045 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4049 PERL_ARGS_ASSERT_FILTER_GETS;
4051 #ifdef PERL_CR_FILTER
4052 if (!PL_rsfp_filters) {
4053 filter_add(S_cr_textfilter,NULL);
4056 if (PL_rsfp_filters) {
4058 SvCUR_set(sv, 0); /* start with empty line */
4059 if (FILTER_READ(0, sv, 0) > 0)
4060 return ( SvPVX(sv) ) ;
4065 return (sv_gets(sv, PL_rsfp, append));
4069 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4074 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4076 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4080 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4081 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4083 return GvHV(gv); /* Foo:: */
4086 /* use constant CLASS => 'MyClass' */
4087 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4088 if (gv && GvCV(gv)) {
4089 SV * const sv = cv_const_sv(GvCV(gv));
4091 pkgname = SvPV_const(sv, len);
4094 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4098 * S_readpipe_override
4099 * Check whether readpipe() is overridden, and generates the appropriate
4100 * optree, provided sublex_start() is called afterwards.
4103 S_readpipe_override(pTHX)
4106 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4107 pl_yylval.ival = OP_BACKTICK;
4109 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4111 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4112 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4113 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4115 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4116 op_append_elem(OP_LIST,
4117 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4118 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4125 * The intent of this yylex wrapper is to minimize the changes to the
4126 * tokener when we aren't interested in collecting madprops. It remains
4127 * to be seen how successful this strategy will be...
4134 char *s = PL_bufptr;
4136 /* make sure PL_thiswhite is initialized */
4140 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4141 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4142 return S_pending_ident(aTHX);
4144 /* previous token ate up our whitespace? */
4145 if (!PL_lasttoke && PL_nextwhite) {
4146 PL_thiswhite = PL_nextwhite;
4150 /* isolate the token, and figure out where it is without whitespace */
4151 PL_realtokenstart = -1;
4155 assert(PL_curforce < 0);
4157 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4158 if (!PL_thistoken) {
4159 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4160 PL_thistoken = newSVpvs("");
4162 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4163 PL_thistoken = newSVpvn(tstart, s - tstart);
4166 if (PL_thismad) /* install head */
4167 CURMAD('X', PL_thistoken);
4170 /* last whitespace of a sublex? */
4171 if (optype == ')' && PL_endwhite) {
4172 CURMAD('X', PL_endwhite);
4177 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4178 if (!PL_thiswhite && !PL_endwhite && !optype) {
4179 sv_free(PL_thistoken);
4184 /* put off final whitespace till peg */
4185 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4186 PL_nextwhite = PL_thiswhite;
4189 else if (PL_thisopen) {
4190 CURMAD('q', PL_thisopen);
4192 sv_free(PL_thistoken);
4196 /* Store actual token text as madprop X */
4197 CURMAD('X', PL_thistoken);
4201 /* add preceding whitespace as madprop _ */
4202 CURMAD('_', PL_thiswhite);
4206 /* add quoted material as madprop = */
4207 CURMAD('=', PL_thisstuff);
4211 /* add terminating quote as madprop Q */
4212 CURMAD('Q', PL_thisclose);
4216 /* special processing based on optype */
4220 /* opval doesn't need a TOKEN since it can already store mp */
4231 if (pl_yylval.opval)
4232 append_madprops(PL_thismad, pl_yylval.opval, 0);
4240 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4249 /* remember any fake bracket that lexer is about to discard */
4250 if (PL_lex_brackets == 1 &&
4251 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4254 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4257 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4258 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4261 break; /* don't bother looking for trailing comment */
4270 /* attach a trailing comment to its statement instead of next token */
4274 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4276 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4278 if (*s == '\n' || *s == '#') {
4279 while (s < PL_bufend && *s != '\n')
4283 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4284 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4297 /* Create new token struct. Note: opvals return early above. */
4298 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4305 S_tokenize_use(pTHX_ int is_use, char *s) {
4308 PERL_ARGS_ASSERT_TOKENIZE_USE;
4310 if (PL_expect != XSTATE)
4311 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4312 is_use ? "use" : "no"));
4314 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4315 s = force_version(s, TRUE);
4316 if (*s == ';' || *s == '}'
4317 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4318 start_force(PL_curforce);
4319 NEXTVAL_NEXTTOKE.opval = NULL;
4322 else if (*s == 'v') {
4323 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4324 s = force_version(s, FALSE);
4328 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4329 s = force_version(s, FALSE);
4331 pl_yylval.ival = is_use;
4335 static const char* const exp_name[] =
4336 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4337 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4341 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4343 S_word_takes_any_delimeter(char *p, STRLEN len)
4345 return (len == 1 && strchr("msyq", p[0])) ||
4347 (p[0] == 't' && p[1] == 'r') ||
4348 (p[0] == 'q' && strchr("qwxr", p[1]))));
4354 Works out what to call the token just pulled out of the input
4355 stream. The yacc parser takes care of taking the ops we return and
4356 stitching them into a tree.
4362 if read an identifier
4363 if we're in a my declaration
4364 croak if they tried to say my($foo::bar)
4365 build the ops for a my() declaration
4366 if it's an access to a my() variable
4367 are we in a sort block?
4368 croak if my($a); $a <=> $b
4369 build ops for access to a my() variable
4370 if in a dq string, and they've said @foo and we can't find @foo
4372 build ops for a bareword
4373 if we already built the token before, use it.
4378 #pragma segment Perl_yylex
4384 register char *s = PL_bufptr;
4390 /* orig_keyword, gvp, and gv are initialized here because
4391 * jump to the label just_a_word_zero can bypass their
4392 * initialization later. */
4393 I32 orig_keyword = 0;
4398 SV* tmp = newSVpvs("");
4399 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4400 (IV)CopLINE(PL_curcop),
4401 lex_state_names[PL_lex_state],
4402 exp_name[PL_expect],
4403 pv_display(tmp, s, strlen(s), 0, 60));
4406 /* check if there's an identifier for us to look at */
4407 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4408 return REPORT(S_pending_ident(aTHX));
4410 /* no identifier pending identification */
4412 switch (PL_lex_state) {
4414 case LEX_NORMAL: /* Some compilers will produce faster */
4415 case LEX_INTERPNORMAL: /* code if we comment these out. */
4419 /* when we've already built the next token, just pull it out of the queue */
4423 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4425 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4426 PL_nexttoke[PL_lasttoke].next_mad = 0;
4427 if (PL_thismad && PL_thismad->mad_key == '_') {
4428 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4429 PL_thismad->mad_val = 0;
4430 mad_free(PL_thismad);
4435 PL_lex_state = PL_lex_defer;
4436 PL_expect = PL_lex_expect;
4437 PL_lex_defer = LEX_NORMAL;
4438 if (!PL_nexttoke[PL_lasttoke].next_type)
4443 pl_yylval = PL_nextval[PL_nexttoke];
4445 PL_lex_state = PL_lex_defer;
4446 PL_expect = PL_lex_expect;
4447 PL_lex_defer = LEX_NORMAL;
4453 next_type = PL_nexttoke[PL_lasttoke].next_type;
4455 next_type = PL_nexttype[PL_nexttoke];
4457 if (next_type & (7<<24)) {
4458 if (next_type & (1<<24)) {
4459 if (PL_lex_brackets > 100)
4460 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4461 PL_lex_brackstack[PL_lex_brackets++] =