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_preambled (PL_parser->preambled)
70 #define PL_sublex_info (PL_parser->sublex_info)
71 #define PL_linestr (PL_parser->linestr)
72 #define PL_expect (PL_parser->expect)
73 #define PL_copline (PL_parser->copline)
74 #define PL_bufptr (PL_parser->bufptr)
75 #define PL_oldbufptr (PL_parser->oldbufptr)
76 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
77 #define PL_linestart (PL_parser->linestart)
78 #define PL_bufend (PL_parser->bufend)
79 #define PL_last_uni (PL_parser->last_uni)
80 #define PL_last_lop (PL_parser->last_lop)
81 #define PL_last_lop_op (PL_parser->last_lop_op)
82 #define PL_lex_state (PL_parser->lex_state)
83 #define PL_rsfp (PL_parser->rsfp)
84 #define PL_rsfp_filters (PL_parser->rsfp_filters)
85 #define PL_in_my (PL_parser->in_my)
86 #define PL_in_my_stash (PL_parser->in_my_stash)
87 #define PL_tokenbuf (PL_parser->tokenbuf)
88 #define PL_multi_end (PL_parser->multi_end)
89 #define PL_error_count (PL_parser->error_count)
92 # define PL_endwhite (PL_parser->endwhite)
93 # define PL_faketokens (PL_parser->faketokens)
94 # define PL_lasttoke (PL_parser->lasttoke)
95 # define PL_nextwhite (PL_parser->nextwhite)
96 # define PL_realtokenstart (PL_parser->realtokenstart)
97 # define PL_skipwhite (PL_parser->skipwhite)
98 # define PL_thisclose (PL_parser->thisclose)
99 # define PL_thismad (PL_parser->thismad)
100 # define PL_thisopen (PL_parser->thisopen)
101 # define PL_thisstuff (PL_parser->thisstuff)
102 # define PL_thistoken (PL_parser->thistoken)
103 # define PL_thiswhite (PL_parser->thiswhite)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_nexttoke (PL_parser->nexttoke)
106 # define PL_curforce (PL_parser->curforce)
108 # define PL_nexttoke (PL_parser->nexttoke)
109 # define PL_nexttype (PL_parser->nexttype)
110 # define PL_nextval (PL_parser->nextval)
113 static const char* const ident_too_long = "Identifier too long";
116 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
117 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
119 # define CURMAD(slot,sv)
120 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
123 #define XENUMMASK 0x3f
124 #define XFAKEEOF 0x40
125 #define XFAKEBRACK 0x80
127 #ifdef USE_UTF8_SCRIPTS
128 # define UTF (!IN_BYTES)
130 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
133 /* The maximum number of characters preceding the unrecognized one to display */
134 #define UNRECOGNIZED_PRECEDE_COUNT 10
136 /* In variables named $^X, these are the legal values for X.
137 * 1999-02-27 mjd-perl-patch@plover.com */
138 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
140 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
142 /* LEX_* are values for PL_lex_state, the state of the lexer.
143 * They are arranged oddly so that the guard on the switch statement
144 * can get by with a single comparison (if the compiler is smart enough).
146 * These values refer to the various states within a sublex parse,
147 * i.e. within a double quotish string
150 /* #define LEX_NOTPARSING 11 is done in perl.h. */
152 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
153 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
154 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
155 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
156 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
158 /* at end of code, eg "$x" followed by: */
159 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
160 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
162 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
163 string or after \E, $foo, etc */
164 #define LEX_INTERPCONST 2 /* NOT USED */
165 #define LEX_FORMLINE 1 /* expecting a format line */
166 #define LEX_KNOWNEXT 0 /* next token known; just return it */
170 static const char* const lex_state_names[] = {
189 #include "keywords.h"
191 /* CLINE is a macro that ensures PL_copline has a sane value */
196 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
199 # define SKIPSPACE0(s) skipspace0(s)
200 # define SKIPSPACE1(s) skipspace1(s)
201 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
202 # define PEEKSPACE(s) skipspace2(s,0)
204 # define SKIPSPACE0(s) skipspace(s)
205 # define SKIPSPACE1(s) skipspace(s)
206 # define SKIPSPACE2(s,tsv) skipspace(s)
207 # define PEEKSPACE(s) skipspace(s)
211 * Convenience functions to return different tokens and prime the
212 * lexer for the next token. They all take an argument.
214 * TOKEN : generic token (used for '(', DOLSHARP, etc)
215 * OPERATOR : generic operator
216 * AOPERATOR : assignment operator
217 * PREBLOCK : beginning the block after an if, while, foreach, ...
218 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
219 * PREREF : *EXPR where EXPR is not a simple identifier
220 * TERM : expression term
221 * LOOPX : loop exiting command (goto, last, dump, etc)
222 * FTST : file test operator
223 * FUN0 : zero-argument function
224 * FUN0OP : zero-argument function, with its op created in this file
225 * FUN1 : not used, except for not, which isn't a UNIOP
226 * BOop : bitwise or or xor
228 * SHop : shift operator
229 * PWop : power operator
230 * PMop : pattern-matching operator
231 * Aop : addition-level operator
232 * Mop : multiplication-level operator
233 * Eop : equality-testing operator
234 * Rop : relational operator <= != gt
236 * Also see LOP and lop() below.
239 #ifdef DEBUGGING /* Serve -DT. */
240 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
242 # define REPORT(retval) (retval)
245 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
246 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
247 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
248 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
249 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
251 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
252 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
253 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
254 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
255 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI3(f,x,have_x) { \
273 pl_yylval.ival = f; \
274 if (have_x) PL_expect = x; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI3(f,XTERM,1)
284 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
285 #define UNIPROTO(f,optional) { \
286 if (optional) PL_last_uni = PL_oldbufptr; \
290 #define UNIBRACK(f) UNI3(f,0,0)
292 /* grandfather return to old style */
295 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
296 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
297 pl_yylval.ival = (f); \
303 #define COPLINE_INC_WITH_HERELINES \
305 CopLINE_inc(PL_curcop); \
306 if (PL_parser->lex_shared->herelines) \
307 CopLINE(PL_curcop) += PL_parser->lex_shared->herelines, \
308 PL_parser->lex_shared->herelines = 0; \
314 /* how to interpret the pl_yylval associated with the token */
318 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
323 static struct debug_tokens {
325 enum token_type type;
327 } const debug_tokens[] =
329 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
330 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
331 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
332 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
333 { ARROW, TOKENTYPE_NONE, "ARROW" },
334 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
335 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
336 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
337 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
338 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
339 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
340 { DO, TOKENTYPE_NONE, "DO" },
341 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
342 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
343 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
344 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
345 { ELSE, TOKENTYPE_NONE, "ELSE" },
346 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
347 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
348 { FOR, TOKENTYPE_IVAL, "FOR" },
349 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
350 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
351 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
352 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
353 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
354 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
355 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
356 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
357 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
358 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
359 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
360 { IF, TOKENTYPE_IVAL, "IF" },
361 { LABEL, TOKENTYPE_PVAL, "LABEL" },
362 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
363 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
364 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
365 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
366 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
367 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
368 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
369 { MY, TOKENTYPE_IVAL, "MY" },
370 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
371 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
372 { OROP, TOKENTYPE_IVAL, "OROP" },
373 { OROR, TOKENTYPE_NONE, "OROR" },
374 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
375 { PEG, TOKENTYPE_NONE, "PEG" },
376 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
377 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
378 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
379 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
380 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
381 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
382 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
383 { PREINC, TOKENTYPE_NONE, "PREINC" },
384 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
385 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
386 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
387 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
388 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
389 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
390 { SUB, TOKENTYPE_NONE, "SUB" },
391 { THING, TOKENTYPE_OPVAL, "THING" },
392 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
393 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
394 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
395 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
396 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
397 { USE, TOKENTYPE_IVAL, "USE" },
398 { WHEN, TOKENTYPE_IVAL, "WHEN" },
399 { WHILE, TOKENTYPE_IVAL, "WHILE" },
400 { WORD, TOKENTYPE_OPVAL, "WORD" },
401 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
402 { 0, TOKENTYPE_NONE, NULL }
405 /* dump the returned token in rv, plus any optional arg in pl_yylval */
408 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
412 PERL_ARGS_ASSERT_TOKEREPORT;
415 const char *name = NULL;
416 enum token_type type = TOKENTYPE_NONE;
417 const struct debug_tokens *p;
418 SV* const report = newSVpvs("<== ");
420 for (p = debug_tokens; p->token; p++) {
421 if (p->token == (int)rv) {
428 Perl_sv_catpv(aTHX_ report, name);
429 else if ((char)rv > ' ' && (char)rv <= '~')
430 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
432 sv_catpvs(report, "EOF");
434 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
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; (isWORDCHAR_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 const char *s = SvPVX_const(sv);
635 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 */
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';
746 Newxz(parser->lex_shared, 1, LEXSHARED);
750 s = SvPV_const(line, len);
751 parser->linestr = flags & LEX_START_COPIED
752 ? SvREFCNT_inc_simple_NN(line)
753 : newSVpvn_flags(s, len, SvUTF8(line));
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);
787 SvREFCNT_dec(parser->lex_stuff);
788 SvREFCNT_dec(parser->sublex_info.repl);
790 Safefree(parser->lex_brackstack);
791 Safefree(parser->lex_casestack);
792 Safefree(parser->lex_shared);
793 PL_parser = parser->old_parser;
798 Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
801 I32 nexttoke = parser->lasttoke;
803 I32 nexttoke = parser->nexttoke;
805 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
808 if (S_is_opval_token(parser->nexttoke[nexttoke].next_type
810 && parser->nexttoke[nexttoke].next_val.opval
811 && parser->nexttoke[nexttoke].next_val.opval->op_slabbed
812 && OpSLAB(parser->nexttoke[nexttoke].next_val.opval) == slab) {
813 op_free(parser->nexttoke[nexttoke].next_val.opval);
814 parser->nexttoke[nexttoke].next_val.opval = NULL;
817 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
818 && parser->nextval[nexttoke].opval
819 && parser->nextval[nexttoke].opval->op_slabbed
820 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
821 op_free(parser->nextval[nexttoke].opval);
822 parser->nextval[nexttoke].opval = NULL;
830 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
832 Buffer scalar containing the chunk currently under consideration of the
833 text currently being lexed. This is always a plain string scalar (for
834 which C<SvPOK> is true). It is not intended to be used as a scalar by
835 normal scalar means; instead refer to the buffer directly by the pointer
836 variables described below.
838 The lexer maintains various C<char*> pointers to things in the
839 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
840 reallocated, all of these pointers must be updated. Don't attempt to
841 do this manually, but rather use L</lex_grow_linestr> if you need to
842 reallocate the buffer.
844 The content of the text chunk in the buffer is commonly exactly one
845 complete line of input, up to and including a newline terminator,
846 but there are situations where it is otherwise. The octets of the
847 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
848 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
849 flag on this scalar, which may disagree with it.
851 For direct examination of the buffer, the variable
852 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
853 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
854 of these pointers is usually preferable to examination of the scalar
855 through normal scalar means.
857 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
859 Direct pointer to the end of the chunk of text currently being lexed, the
860 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
861 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
862 always located at the end of the buffer, and does not count as part of
863 the buffer's contents.
865 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
867 Points to the current position of lexing inside the lexer buffer.
868 Characters around this point may be freely examined, within
869 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
870 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
871 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
873 Lexing code (whether in the Perl core or not) moves this pointer past
874 the characters that it consumes. It is also expected to perform some
875 bookkeeping whenever a newline character is consumed. This movement
876 can be more conveniently performed by the function L</lex_read_to>,
877 which handles newlines appropriately.
879 Interpretation of the buffer's octets can be abstracted out by
880 using the slightly higher-level functions L</lex_peek_unichar> and
881 L</lex_read_unichar>.
883 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
885 Points to the start of the current line inside the lexer buffer.
886 This is useful for indicating at which column an error occurred, and
887 not much else. This must be updated by any lexing code that consumes
888 a newline; the function L</lex_read_to> handles this detail.
894 =for apidoc Amx|bool|lex_bufutf8
896 Indicates whether the octets in the lexer buffer
897 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
898 of Unicode characters. If not, they should be interpreted as Latin-1
899 characters. This is analogous to the C<SvUTF8> flag for scalars.
901 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
902 contains valid UTF-8. Lexing code must be robust in the face of invalid
905 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
906 is significant, but not the whole story regarding the input character
907 encoding. Normally, when a file is being read, the scalar contains octets
908 and its C<SvUTF8> flag is off, but the octets should be interpreted as
909 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
910 however, the scalar may have the C<SvUTF8> flag on, and in this case its
911 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
912 is in effect. This logic may change in the future; use this function
913 instead of implementing the logic yourself.
919 Perl_lex_bufutf8(pTHX)
925 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
927 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
928 at least I<len> octets (including terminating NUL). Returns a
929 pointer to the reallocated buffer. This is necessary before making
930 any direct modification of the buffer that would increase its length.
931 L</lex_stuff_pvn> provides a more convenient way to insert text into
934 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
935 this function updates all of the lexer's variables that point directly
942 Perl_lex_grow_linestr(pTHX_ STRLEN len)
946 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
947 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
948 linestr = PL_parser->linestr;
949 buf = SvPVX(linestr);
950 if (len <= SvLEN(linestr))
952 bufend_pos = PL_parser->bufend - buf;
953 bufptr_pos = PL_parser->bufptr - buf;
954 oldbufptr_pos = PL_parser->oldbufptr - buf;
955 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
956 linestart_pos = PL_parser->linestart - buf;
957 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
958 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
959 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
960 PL_parser->lex_shared->re_eval_start - buf : 0;
962 buf = sv_grow(linestr, len);
964 PL_parser->bufend = buf + bufend_pos;
965 PL_parser->bufptr = buf + bufptr_pos;
966 PL_parser->oldbufptr = buf + oldbufptr_pos;
967 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
968 PL_parser->linestart = buf + linestart_pos;
969 if (PL_parser->last_uni)
970 PL_parser->last_uni = buf + last_uni_pos;
971 if (PL_parser->last_lop)
972 PL_parser->last_lop = buf + last_lop_pos;
973 if (PL_parser->lex_shared->re_eval_start)
974 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
979 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
981 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
982 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
983 reallocating the buffer if necessary. This means that lexing code that
984 runs later will see the characters as if they had appeared in the input.
985 It is not recommended to do this as part of normal parsing, and most
986 uses of this facility run the risk of the inserted characters being
987 interpreted in an unintended manner.
989 The string to be inserted is represented by I<len> octets starting
990 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
991 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
992 The characters are recoded for the lexer buffer, according to how the
993 buffer is currently being interpreted (L</lex_bufutf8>). If a string
994 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
995 function is more convenient.
1001 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
1005 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
1006 if (flags & ~(LEX_STUFF_UTF8))
1007 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
1009 if (flags & LEX_STUFF_UTF8) {
1012 STRLEN highhalf = 0; /* Count of variants */
1013 const char *p, *e = pv+len;
1014 for (p = pv; p != e; p++) {
1015 if (! UTF8_IS_INVARIANT(*p)) {
1021 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1022 bufptr = PL_parser->bufptr;
1023 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
1024 SvCUR_set(PL_parser->linestr,
1025 SvCUR(PL_parser->linestr) + len+highhalf);
1026 PL_parser->bufend += len+highhalf;
1027 for (p = pv; p != e; p++) {
1029 if (! UTF8_IS_INVARIANT(c)) {
1030 *bufptr++ = UTF8_TWO_BYTE_HI(c);
1031 *bufptr++ = UTF8_TWO_BYTE_LO(c);
1033 *bufptr++ = (char)c;
1038 if (flags & LEX_STUFF_UTF8) {
1039 STRLEN highhalf = 0;
1040 const char *p, *e = pv+len;
1041 for (p = pv; p != e; p++) {
1043 if (UTF8_IS_ABOVE_LATIN1(c)) {
1044 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1045 "non-Latin-1 character into Latin-1 input");
1046 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
1049 } else if (! UTF8_IS_INVARIANT(c)) {
1050 /* malformed UTF-8 */
1052 SAVESPTR(PL_warnhook);
1053 PL_warnhook = PERL_WARNHOOK_FATAL;
1054 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1060 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1061 bufptr = PL_parser->bufptr;
1062 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1063 SvCUR_set(PL_parser->linestr,
1064 SvCUR(PL_parser->linestr) + len-highhalf);
1065 PL_parser->bufend += len-highhalf;
1068 if (UTF8_IS_INVARIANT(*p)) {
1074 *bufptr++ = TWO_BYTE_UTF8_TO_UNI(*p, *(p+1));
1080 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1081 bufptr = PL_parser->bufptr;
1082 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1083 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1084 PL_parser->bufend += len;
1085 Copy(pv, bufptr, len, char);
1091 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1093 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1094 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1095 reallocating the buffer if necessary. This means that lexing code that
1096 runs later will see the characters as if they had appeared in the input.
1097 It is not recommended to do this as part of normal parsing, and most
1098 uses of this facility run the risk of the inserted characters being
1099 interpreted in an unintended manner.
1101 The string to be inserted is represented by octets starting at I<pv>
1102 and continuing to the first nul. These octets are interpreted as either
1103 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1104 in I<flags>. The characters are recoded for the lexer buffer, according
1105 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1106 If it is not convenient to nul-terminate a string to be inserted, the
1107 L</lex_stuff_pvn> function is more appropriate.
1113 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1115 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1116 lex_stuff_pvn(pv, strlen(pv), flags);
1120 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1122 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1123 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1124 reallocating the buffer if necessary. This means that lexing code that
1125 runs later will see the characters as if they had appeared in the input.
1126 It is not recommended to do this as part of normal parsing, and most
1127 uses of this facility run the risk of the inserted characters being
1128 interpreted in an unintended manner.
1130 The string to be inserted is the string value of I<sv>. The characters
1131 are recoded for the lexer buffer, according to how the buffer is currently
1132 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1133 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1134 need to construct a scalar.
1140 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1144 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1148 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1152 =for apidoc Amx|void|lex_unstuff|char *ptr
1154 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1155 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1156 This hides the discarded text from any lexing code that runs later,
1157 as if the text had never appeared.
1159 This is not the normal way to consume lexed text. For that, use
1166 Perl_lex_unstuff(pTHX_ char *ptr)
1170 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1171 buf = PL_parser->bufptr;
1173 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1176 bufend = PL_parser->bufend;
1178 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1179 unstuff_len = ptr - buf;
1180 Move(ptr, buf, bufend+1-ptr, char);
1181 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1182 PL_parser->bufend = bufend - unstuff_len;
1186 =for apidoc Amx|void|lex_read_to|char *ptr
1188 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1189 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1190 performing the correct bookkeeping whenever a newline character is passed.
1191 This is the normal way to consume lexed text.
1193 Interpretation of the buffer's octets can be abstracted out by
1194 using the slightly higher-level functions L</lex_peek_unichar> and
1195 L</lex_read_unichar>.
1201 Perl_lex_read_to(pTHX_ char *ptr)
1204 PERL_ARGS_ASSERT_LEX_READ_TO;
1205 s = PL_parser->bufptr;
1206 if (ptr < s || ptr > PL_parser->bufend)
1207 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1208 for (; s != ptr; s++)
1210 COPLINE_INC_WITH_HERELINES;
1211 PL_parser->linestart = s+1;
1213 PL_parser->bufptr = ptr;
1217 =for apidoc Amx|void|lex_discard_to|char *ptr
1219 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1220 up to I<ptr>. The remaining content of the buffer will be moved, and
1221 all pointers into the buffer updated appropriately. I<ptr> must not
1222 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1223 it is not permitted to discard text that has yet to be lexed.
1225 Normally it is not necessarily to do this directly, because it suffices to
1226 use the implicit discarding behaviour of L</lex_next_chunk> and things
1227 based on it. However, if a token stretches across multiple lines,
1228 and the lexing code has kept multiple lines of text in the buffer for
1229 that purpose, then after completion of the token it would be wise to
1230 explicitly discard the now-unneeded earlier lines, to avoid future
1231 multi-line tokens growing the buffer without bound.
1237 Perl_lex_discard_to(pTHX_ char *ptr)
1241 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1242 buf = SvPVX(PL_parser->linestr);
1244 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1247 if (ptr > PL_parser->bufptr)
1248 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1249 discard_len = ptr - buf;
1250 if (PL_parser->oldbufptr < ptr)
1251 PL_parser->oldbufptr = ptr;
1252 if (PL_parser->oldoldbufptr < ptr)
1253 PL_parser->oldoldbufptr = ptr;
1254 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1255 PL_parser->last_uni = NULL;
1256 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1257 PL_parser->last_lop = NULL;
1258 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1259 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1260 PL_parser->bufend -= discard_len;
1261 PL_parser->bufptr -= discard_len;
1262 PL_parser->oldbufptr -= discard_len;
1263 PL_parser->oldoldbufptr -= discard_len;
1264 if (PL_parser->last_uni)
1265 PL_parser->last_uni -= discard_len;
1266 if (PL_parser->last_lop)
1267 PL_parser->last_lop -= discard_len;
1271 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1273 Reads in the next chunk of text to be lexed, appending it to
1274 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1275 looked to the end of the current chunk and wants to know more. It is
1276 usual, but not necessary, for lexing to have consumed the entirety of
1277 the current chunk at this time.
1279 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1280 chunk (i.e., the current chunk has been entirely consumed), normally the
1281 current chunk will be discarded at the same time that the new chunk is
1282 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1283 will not be discarded. If the current chunk has not been entirely
1284 consumed, then it will not be discarded regardless of the flag.
1286 Returns true if some new text was added to the buffer, or false if the
1287 buffer has reached the end of the input text.
1292 #define LEX_FAKE_EOF 0x80000000
1293 #define LEX_NO_TERM 0x40000000
1296 Perl_lex_next_chunk(pTHX_ U32 flags)
1300 STRLEN old_bufend_pos, new_bufend_pos;
1301 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1302 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1303 bool got_some_for_debugger = 0;
1305 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1306 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1307 linestr = PL_parser->linestr;
1308 buf = SvPVX(linestr);
1309 if (!(flags & LEX_KEEP_PREVIOUS) &&
1310 PL_parser->bufptr == PL_parser->bufend) {
1311 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1313 if (PL_parser->last_uni != PL_parser->bufend)
1314 PL_parser->last_uni = NULL;
1315 if (PL_parser->last_lop != PL_parser->bufend)
1316 PL_parser->last_lop = NULL;
1317 last_uni_pos = last_lop_pos = 0;
1321 old_bufend_pos = PL_parser->bufend - buf;
1322 bufptr_pos = PL_parser->bufptr - buf;
1323 oldbufptr_pos = PL_parser->oldbufptr - buf;
1324 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1325 linestart_pos = PL_parser->linestart - buf;
1326 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1327 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1329 if (flags & LEX_FAKE_EOF) {
1331 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1333 } else if (filter_gets(linestr, old_bufend_pos)) {
1335 got_some_for_debugger = 1;
1336 } else if (flags & LEX_NO_TERM) {
1339 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1340 sv_setpvs(linestr, "");
1342 /* End of real input. Close filehandle (unless it was STDIN),
1343 * then add implicit termination.
1345 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1346 PerlIO_clearerr(PL_parser->rsfp);
1347 else if (PL_parser->rsfp)
1348 (void)PerlIO_close(PL_parser->rsfp);
1349 PL_parser->rsfp = NULL;
1350 PL_parser->in_pod = PL_parser->filtered = 0;
1352 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1355 if (!PL_in_eval && PL_minus_p) {
1357 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1358 PL_minus_n = PL_minus_p = 0;
1359 } else if (!PL_in_eval && PL_minus_n) {
1360 sv_catpvs(linestr, /*{*/";}");
1363 sv_catpvs(linestr, ";");
1366 buf = SvPVX(linestr);
1367 new_bufend_pos = SvCUR(linestr);
1368 PL_parser->bufend = buf + new_bufend_pos;
1369 PL_parser->bufptr = buf + bufptr_pos;
1370 PL_parser->oldbufptr = buf + oldbufptr_pos;
1371 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1372 PL_parser->linestart = buf + linestart_pos;
1373 if (PL_parser->last_uni)
1374 PL_parser->last_uni = buf + last_uni_pos;
1375 if (PL_parser->last_lop)
1376 PL_parser->last_lop = buf + last_lop_pos;
1377 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1378 PL_curstash != PL_debstash) {
1379 /* debugger active and we're not compiling the debugger code,
1380 * so store the line into the debugger's array of lines
1382 update_debugger_info(NULL, buf+old_bufend_pos,
1383 new_bufend_pos-old_bufend_pos);
1389 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1391 Looks ahead one (Unicode) character in the text currently being lexed.
1392 Returns the codepoint (unsigned integer value) of the next character,
1393 or -1 if lexing has reached the end of the input text. To consume the
1394 peeked character, use L</lex_read_unichar>.
1396 If the next character is in (or extends into) the next chunk of input
1397 text, the next chunk will be read in. Normally the current chunk will be
1398 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1399 then the current chunk will not be discarded.
1401 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1402 is encountered, an exception is generated.
1408 Perl_lex_peek_unichar(pTHX_ U32 flags)
1412 if (flags & ~(LEX_KEEP_PREVIOUS))
1413 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1414 s = PL_parser->bufptr;
1415 bufend = PL_parser->bufend;
1421 if (!lex_next_chunk(flags))
1423 s = PL_parser->bufptr;
1424 bufend = PL_parser->bufend;
1427 if (UTF8_IS_INVARIANT(head))
1429 if (UTF8_IS_START(head)) {
1430 len = UTF8SKIP(&head);
1431 while ((STRLEN)(bufend-s) < len) {
1432 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1434 s = PL_parser->bufptr;
1435 bufend = PL_parser->bufend;
1438 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1439 if (retlen == (STRLEN)-1) {
1440 /* malformed UTF-8 */
1442 SAVESPTR(PL_warnhook);
1443 PL_warnhook = PERL_WARNHOOK_FATAL;
1444 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1450 if (!lex_next_chunk(flags))
1452 s = PL_parser->bufptr;
1459 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1461 Reads the next (Unicode) character in the text currently being lexed.
1462 Returns the codepoint (unsigned integer value) of the character read,
1463 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1464 if lexing has reached the end of the input text. To non-destructively
1465 examine the next character, use L</lex_peek_unichar> instead.
1467 If the next character is in (or extends into) the next chunk of input
1468 text, the next chunk will be read in. Normally the current chunk will be
1469 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1470 then the current chunk will not be discarded.
1472 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1473 is encountered, an exception is generated.
1479 Perl_lex_read_unichar(pTHX_ U32 flags)
1482 if (flags & ~(LEX_KEEP_PREVIOUS))
1483 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1484 c = lex_peek_unichar(flags);
1487 COPLINE_INC_WITH_HERELINES;
1489 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1491 ++(PL_parser->bufptr);
1497 =for apidoc Amx|void|lex_read_space|U32 flags
1499 Reads optional spaces, in Perl style, in the text currently being
1500 lexed. The spaces may include ordinary whitespace characters and
1501 Perl-style comments. C<#line> directives are processed if encountered.
1502 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1503 at a non-space character (or the end of the input text).
1505 If spaces extend into the next chunk of input text, the next chunk will
1506 be read in. Normally the current chunk will be discarded at the same
1507 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1508 chunk will not be discarded.
1513 #define LEX_NO_NEXT_CHUNK 0x80000000
1516 Perl_lex_read_space(pTHX_ U32 flags)
1519 bool need_incline = 0;
1520 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1521 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1524 sv_free(PL_skipwhite);
1525 PL_skipwhite = NULL;
1528 PL_skipwhite = newSVpvs("");
1529 #endif /* PERL_MAD */
1530 s = PL_parser->bufptr;
1531 bufend = PL_parser->bufend;
1537 } while (!(c == '\n' || (c == 0 && s == bufend)));
1538 } else if (c == '\n') {
1540 PL_parser->linestart = s;
1545 } else if (isSPACE(c)) {
1547 } else if (c == 0 && s == bufend) {
1551 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1552 #endif /* PERL_MAD */
1553 if (flags & LEX_NO_NEXT_CHUNK)
1555 PL_parser->bufptr = s;
1556 COPLINE_INC_WITH_HERELINES;
1557 got_more = lex_next_chunk(flags);
1558 CopLINE_dec(PL_curcop);
1559 s = PL_parser->bufptr;
1560 bufend = PL_parser->bufend;
1563 if (need_incline && PL_parser->rsfp) {
1573 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1574 #endif /* PERL_MAD */
1575 PL_parser->bufptr = s;
1580 * This subroutine has nothing to do with tilting, whether at windmills
1581 * or pinball tables. Its name is short for "increment line". It
1582 * increments the current line number in CopLINE(PL_curcop) and checks
1583 * to see whether the line starts with a comment of the form
1584 * # line 500 "foo.pm"
1585 * If so, it sets the current line number and file to the values in the comment.
1589 S_incline(pTHX_ const char *s)
1597 PERL_ARGS_ASSERT_INCLINE;
1599 COPLINE_INC_WITH_HERELINES;
1600 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1601 && s+1 == PL_bufend && *s == ';') {
1602 /* fake newline in string eval */
1603 CopLINE_dec(PL_curcop);
1608 while (SPACE_OR_TAB(*s))
1610 if (strnEQ(s, "line", 4))
1614 if (SPACE_OR_TAB(*s))
1618 while (SPACE_OR_TAB(*s))
1626 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1628 while (SPACE_OR_TAB(*s))
1630 if (*s == '"' && (t = strchr(s+1, '"'))) {
1636 while (!isSPACE(*t))
1640 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1642 if (*e != '\n' && *e != '\0')
1643 return; /* false alarm */
1645 line_num = atoi(n)-1;
1648 const STRLEN len = t - s;
1649 SV *const temp_sv = CopFILESV(PL_curcop);
1654 cf = SvPVX(temp_sv);
1655 tmplen = SvCUR(temp_sv);
1661 if (!PL_rsfp && !PL_parser->filtered) {
1662 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1663 * to *{"::_<newfilename"} */
1664 /* However, the long form of evals is only turned on by the
1665 debugger - usually they're "(eval %lu)" */
1669 STRLEN tmplen2 = len;
1670 if (tmplen + 2 <= sizeof smallbuf)
1673 Newx(tmpbuf, tmplen + 2, char);
1676 memcpy(tmpbuf + 2, cf, tmplen);
1678 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1683 if (tmplen2 + 2 <= sizeof smallbuf)
1686 Newx(tmpbuf2, tmplen2 + 2, char);
1688 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1689 /* Either they malloc'd it, or we malloc'd it,
1690 so no prefix is present in ours. */
1695 memcpy(tmpbuf2 + 2, s, tmplen2);
1698 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1700 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1701 /* adjust ${"::_<newfilename"} to store the new file name */
1702 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1703 /* The line number may differ. If that is the case,
1704 alias the saved lines that are in the array.
1705 Otherwise alias the whole array. */
1706 if (CopLINE(PL_curcop) == line_num) {
1707 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1708 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1710 else if (GvAV(*gvp)) {
1711 AV * const av = GvAV(*gvp);
1712 const I32 start = CopLINE(PL_curcop)+1;
1713 I32 items = AvFILLp(av) - start;
1715 AV * const av2 = GvAVn(gv2);
1716 SV **svp = AvARRAY(av) + start;
1717 I32 l = (I32)line_num+1;
1719 av_store(av2, l++, SvREFCNT_inc(*svp++));
1724 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1726 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1728 CopFILE_free(PL_curcop);
1729 CopFILE_setn(PL_curcop, s, len);
1731 CopLINE_set(PL_curcop, line_num);
1735 /* skip space before PL_thistoken */
1738 S_skipspace0(pTHX_ char *s)
1740 PERL_ARGS_ASSERT_SKIPSPACE0;
1747 PL_thiswhite = newSVpvs("");
1748 sv_catsv(PL_thiswhite, PL_skipwhite);
1749 sv_free(PL_skipwhite);
1752 PL_realtokenstart = s - SvPVX(PL_linestr);
1756 /* skip space after PL_thistoken */
1759 S_skipspace1(pTHX_ char *s)
1761 const char *start = s;
1762 I32 startoff = start - SvPVX(PL_linestr);
1764 PERL_ARGS_ASSERT_SKIPSPACE1;
1769 start = SvPVX(PL_linestr) + startoff;
1770 if (!PL_thistoken && PL_realtokenstart >= 0) {
1771 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1772 PL_thistoken = newSVpvn(tstart, start - tstart);
1774 PL_realtokenstart = -1;
1777 PL_nextwhite = newSVpvs("");
1778 sv_catsv(PL_nextwhite, PL_skipwhite);
1779 sv_free(PL_skipwhite);
1786 S_skipspace2(pTHX_ char *s, SV **svp)
1789 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1790 const I32 startoff = s - SvPVX(PL_linestr);
1792 PERL_ARGS_ASSERT_SKIPSPACE2;
1795 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1796 if (!PL_madskills || !svp)
1798 start = SvPVX(PL_linestr) + startoff;
1799 if (!PL_thistoken && PL_realtokenstart >= 0) {
1800 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1801 PL_thistoken = newSVpvn(tstart, start - tstart);
1802 PL_realtokenstart = -1;
1806 *svp = newSVpvs("");
1807 sv_setsv(*svp, PL_skipwhite);
1808 sv_free(PL_skipwhite);
1817 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1819 AV *av = CopFILEAVx(PL_curcop);
1821 SV * const sv = newSV_type(SVt_PVMG);
1823 sv_setsv(sv, orig_sv);
1825 sv_setpvn(sv, buf, len);
1828 av_store(av, (I32)CopLINE(PL_curcop), sv);
1834 * Called to gobble the appropriate amount and type of whitespace.
1835 * Skips comments as well.
1839 S_skipspace(pTHX_ char *s)
1843 #endif /* PERL_MAD */
1844 PERL_ARGS_ASSERT_SKIPSPACE;
1847 sv_free(PL_skipwhite);
1848 PL_skipwhite = NULL;
1850 #endif /* PERL_MAD */
1851 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1852 while (s < PL_bufend && SPACE_OR_TAB(*s))
1855 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1857 lex_read_space(LEX_KEEP_PREVIOUS |
1858 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1859 LEX_NO_NEXT_CHUNK : 0));
1861 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1862 if (PL_linestart > PL_bufptr)
1863 PL_bufptr = PL_linestart;
1868 PL_skipwhite = newSVpvn(start, s-start);
1869 #endif /* PERL_MAD */
1875 * Check the unary operators to ensure there's no ambiguity in how they're
1876 * used. An ambiguous piece of code would be:
1878 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1879 * the +5 is its argument.
1889 if (PL_oldoldbufptr != PL_last_uni)
1891 while (isSPACE(*PL_last_uni))
1894 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1896 if ((t = strchr(s, '(')) && t < PL_bufptr)
1899 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1900 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1901 (int)(s - PL_last_uni), PL_last_uni);
1905 * LOP : macro to build a list operator. Its behaviour has been replaced
1906 * with a subroutine, S_lop() for which LOP is just another name.
1909 #define LOP(f,x) return lop(f,x,s)
1913 * Build a list operator (or something that might be one). The rules:
1914 * - if we have a next token, then it's a list operator [why?]
1915 * - if the next thing is an opening paren, then it's a function
1916 * - else it's a list operator
1920 S_lop(pTHX_ I32 f, int x, char *s)
1924 PERL_ARGS_ASSERT_LOP;
1930 PL_last_lop = PL_oldbufptr;
1931 PL_last_lop_op = (OPCODE)f;
1940 return REPORT(FUNC);
1943 return REPORT(FUNC);
1946 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1947 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1948 return REPORT(LSTOP);
1955 * Sets up for an eventual force_next(). start_force(0) basically does
1956 * an unshift, while start_force(-1) does a push. yylex removes items
1961 S_start_force(pTHX_ int where)
1965 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1966 where = PL_lasttoke;
1967 assert(PL_curforce < 0 || PL_curforce == where);
1968 if (PL_curforce != where) {
1969 for (i = PL_lasttoke; i > where; --i) {
1970 PL_nexttoke[i] = PL_nexttoke[i-1];
1974 if (PL_curforce < 0) /* in case of duplicate start_force() */
1975 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1976 PL_curforce = where;
1979 curmad('^', newSVpvs(""));
1980 CURMAD('_', PL_nextwhite);
1985 S_curmad(pTHX_ char slot, SV *sv)
1991 if (PL_curforce < 0)
1992 where = &PL_thismad;
1994 where = &PL_nexttoke[PL_curforce].next_mad;
2000 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
2002 else if (PL_encoding) {
2003 sv_recode_to_utf8(sv, PL_encoding);
2008 /* keep a slot open for the head of the list? */
2009 if (slot != '_' && *where && (*where)->mad_key == '^') {
2010 (*where)->mad_key = slot;
2011 sv_free(MUTABLE_SV(((*where)->mad_val)));
2012 (*where)->mad_val = (void*)sv;
2015 addmad(newMADsv(slot, sv), where, 0);
2018 # define start_force(where) NOOP
2019 # define curmad(slot, sv) NOOP
2024 * When the lexer realizes it knows the next token (for instance,
2025 * it is reordering tokens for the parser) then it can call S_force_next
2026 * to know what token to return the next time the lexer is called. Caller
2027 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
2028 * and possibly PL_expect to ensure the lexer handles the token correctly.
2032 S_force_next(pTHX_ I32 type)
2037 PerlIO_printf(Perl_debug_log, "### forced token:\n");
2038 tokereport(type, &NEXTVAL_NEXTTOKE);
2042 if (PL_curforce < 0)
2043 start_force(PL_lasttoke);
2044 PL_nexttoke[PL_curforce].next_type = type;
2045 if (PL_lex_state != LEX_KNOWNEXT)
2046 PL_lex_defer = PL_lex_state;
2047 PL_lex_state = LEX_KNOWNEXT;
2048 PL_lex_expect = PL_expect;
2051 PL_nexttype[PL_nexttoke] = type;
2053 if (PL_lex_state != LEX_KNOWNEXT) {
2054 PL_lex_defer = PL_lex_state;
2055 PL_lex_expect = PL_expect;
2056 PL_lex_state = LEX_KNOWNEXT;
2064 int yyc = PL_parser->yychar;
2065 if (yyc != YYEMPTY) {
2068 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2069 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2070 PL_lex_allbrackets--;
2072 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2073 } else if (yyc == '('/*)*/) {
2074 PL_lex_allbrackets--;
2079 PL_parser->yychar = YYEMPTY;
2084 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2087 SV * const sv = newSVpvn_utf8(start, len,
2090 && !is_ascii_string((const U8*)start, len)
2091 && is_utf8_string((const U8*)start, len));
2097 * When the lexer knows the next thing is a word (for instance, it has
2098 * just seen -> and it knows that the next char is a word char, then
2099 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2103 * char *start : buffer position (must be within PL_linestr)
2104 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2105 * int check_keyword : if true, Perl checks to make sure the word isn't
2106 * a keyword (do this if the word is a label, e.g. goto FOO)
2107 * int allow_pack : if true, : characters will also be allowed (require,
2108 * use, etc. do this)
2109 * int allow_initial_tick : used by the "sub" lexer only.
2113 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2119 PERL_ARGS_ASSERT_FORCE_WORD;
2121 start = SKIPSPACE1(start);
2123 if (isIDFIRST_lazy_if(s,UTF) ||
2124 (allow_pack && *s == ':') ||
2125 (allow_initial_tick && *s == '\'') )
2127 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2128 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2130 start_force(PL_curforce);
2132 curmad('X', newSVpvn(start,s-start));
2133 if (token == METHOD) {
2138 PL_expect = XOPERATOR;
2142 curmad('g', newSVpvs( "forced" ));
2143 NEXTVAL_NEXTTOKE.opval
2144 = (OP*)newSVOP(OP_CONST,0,
2145 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2146 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2154 * Called when the lexer wants $foo *foo &foo etc, but the program
2155 * text only contains the "foo" portion. The first argument is a pointer
2156 * to the "foo", and the second argument is the type symbol to prefix.
2157 * Forces the next token to be a "WORD".
2158 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2162 S_force_ident(pTHX_ const char *s, int kind)
2166 PERL_ARGS_ASSERT_FORCE_IDENT;
2169 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2170 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2171 UTF ? SVf_UTF8 : 0));
2172 start_force(PL_curforce);
2173 NEXTVAL_NEXTTOKE.opval = o;
2176 o->op_private = OPpCONST_ENTERED;
2177 /* XXX see note in pp_entereval() for why we forgo typo
2178 warnings if the symbol must be introduced in an eval.
2180 gv_fetchpvn_flags(s, len,
2181 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2182 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2183 kind == '$' ? SVt_PV :
2184 kind == '@' ? SVt_PVAV :
2185 kind == '%' ? SVt_PVHV :
2193 S_force_ident_maybe_lex(pTHX_ char pit)
2195 start_force(PL_curforce);
2196 NEXTVAL_NEXTTOKE.ival = pit;
2201 Perl_str_to_version(pTHX_ SV *sv)
2206 const char *start = SvPV_const(sv,len);
2207 const char * const end = start + len;
2208 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2210 PERL_ARGS_ASSERT_STR_TO_VERSION;
2212 while (start < end) {
2216 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2221 retval += ((NV)n)/nshift;
2230 * Forces the next token to be a version number.
2231 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2232 * and if "guessing" is TRUE, then no new token is created (and the caller
2233 * must use an alternative parsing method).
2237 S_force_version(pTHX_ char *s, int guessing)
2243 I32 startoff = s - SvPVX(PL_linestr);
2246 PERL_ARGS_ASSERT_FORCE_VERSION;
2254 while (isDIGIT(*d) || *d == '_' || *d == '.')
2258 start_force(PL_curforce);
2259 curmad('X', newSVpvn(s,d-s));
2262 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2264 #ifdef USE_LOCALE_NUMERIC
2265 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2266 setlocale(LC_NUMERIC, "C");
2268 s = scan_num(s, &pl_yylval);
2269 #ifdef USE_LOCALE_NUMERIC
2270 setlocale(LC_NUMERIC, loc);
2273 version = pl_yylval.opval;
2274 ver = cSVOPx(version)->op_sv;
2275 if (SvPOK(ver) && !SvNIOK(ver)) {
2276 SvUPGRADE(ver, SVt_PVNV);
2277 SvNV_set(ver, str_to_version(ver));
2278 SvNOK_on(ver); /* hint that it is a version */
2281 else if (guessing) {
2284 sv_free(PL_nextwhite); /* let next token collect whitespace */
2286 s = SvPVX(PL_linestr) + startoff;
2294 if (PL_madskills && !version) {
2295 sv_free(PL_nextwhite); /* let next token collect whitespace */
2297 s = SvPVX(PL_linestr) + startoff;
2300 /* NOTE: The parser sees the package name and the VERSION swapped */
2301 start_force(PL_curforce);
2302 NEXTVAL_NEXTTOKE.opval = version;
2309 * S_force_strict_version
2310 * Forces the next token to be a version number using strict syntax rules.
2314 S_force_strict_version(pTHX_ char *s)
2319 I32 startoff = s - SvPVX(PL_linestr);
2321 const char *errstr = NULL;
2323 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2325 while (isSPACE(*s)) /* leading whitespace */
2328 if (is_STRICT_VERSION(s,&errstr)) {
2330 s = (char *)scan_version(s, ver, 0);
2331 version = newSVOP(OP_CONST, 0, ver);
2333 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2334 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2338 yyerror(errstr); /* version required */
2343 if (PL_madskills && !version) {
2344 sv_free(PL_nextwhite); /* let next token collect whitespace */
2346 s = SvPVX(PL_linestr) + startoff;
2349 /* NOTE: The parser sees the package name and the VERSION swapped */
2350 start_force(PL_curforce);
2351 NEXTVAL_NEXTTOKE.opval = version;
2359 * Tokenize a quoted string passed in as an SV. It finds the next
2360 * chunk, up to end of string or a backslash. It may make a new
2361 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2366 S_tokeq(pTHX_ SV *sv)
2375 PERL_ARGS_ASSERT_TOKEQ;
2380 s = SvPV_force(sv, len);
2381 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2384 /* This is relying on the SV being "well formed" with a trailing '\0' */
2385 while (s < send && !(*s == '\\' && s[1] == '\\'))
2390 if ( PL_hints & HINT_NEW_STRING ) {
2391 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2395 if (s + 1 < send && (s[1] == '\\'))
2396 s++; /* all that, just for this */
2401 SvCUR_set(sv, d - SvPVX_const(sv));
2403 if ( PL_hints & HINT_NEW_STRING )
2404 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2409 * Now come three functions related to double-quote context,
2410 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2411 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2412 * interact with PL_lex_state, and create fake ( ... ) argument lists
2413 * to handle functions and concatenation.
2417 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2422 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2424 * Pattern matching will set PL_lex_op to the pattern-matching op to
2425 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2427 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2429 * Everything else becomes a FUNC.
2431 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2432 * had an OP_CONST or OP_READLINE). This just sets us up for a
2433 * call to S_sublex_push().
2437 S_sublex_start(pTHX)
2440 const I32 op_type = pl_yylval.ival;
2442 if (op_type == OP_NULL) {
2443 pl_yylval.opval = PL_lex_op;
2447 if (op_type == OP_CONST || op_type == OP_READLINE) {
2448 SV *sv = tokeq(PL_lex_stuff);
2450 if (SvTYPE(sv) == SVt_PVIV) {
2451 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2453 const char * const p = SvPV_const(sv, len);
2454 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2458 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2459 PL_lex_stuff = NULL;
2460 /* Allow <FH> // "foo" */
2461 if (op_type == OP_READLINE)
2462 PL_expect = XTERMORDORDOR;
2465 else if (op_type == OP_BACKTICK && PL_lex_op) {
2466 /* readpipe() vas overriden */
2467 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2468 pl_yylval.opval = PL_lex_op;
2470 PL_lex_stuff = NULL;
2474 PL_sublex_info.super_state = PL_lex_state;
2475 PL_sublex_info.sub_inwhat = (U16)op_type;
2476 PL_sublex_info.sub_op = PL_lex_op;
2477 PL_lex_state = LEX_INTERPPUSH;
2481 pl_yylval.opval = PL_lex_op;
2491 * Create a new scope to save the lexing state. The scope will be
2492 * ended in S_sublex_done. Returns a '(', starting the function arguments
2493 * to the uc, lc, etc. found before.
2494 * Sets PL_lex_state to LEX_INTERPCONCAT.
2504 PL_lex_state = PL_sublex_info.super_state;
2505 SAVEBOOL(PL_lex_dojoin);
2506 SAVEI32(PL_lex_brackets);
2507 SAVEI32(PL_lex_allbrackets);
2508 SAVEI32(PL_lex_formbrack);
2509 SAVEI8(PL_lex_fakeeof);
2510 SAVEI32(PL_lex_casemods);
2511 SAVEI32(PL_lex_starts);
2512 SAVEI8(PL_lex_state);
2513 SAVESPTR(PL_lex_repl);
2514 SAVEVPTR(PL_lex_inpat);
2515 SAVEI16(PL_lex_inwhat);
2516 SAVECOPLINE(PL_curcop);
2517 SAVEPPTR(PL_bufptr);
2518 SAVEPPTR(PL_bufend);
2519 SAVEPPTR(PL_oldbufptr);
2520 SAVEPPTR(PL_oldoldbufptr);
2521 SAVEPPTR(PL_last_lop);
2522 SAVEPPTR(PL_last_uni);
2523 SAVEPPTR(PL_linestart);
2524 SAVESPTR(PL_linestr);
2525 SAVEGENERICPV(PL_lex_brackstack);
2526 SAVEGENERICPV(PL_lex_casestack);
2527 SAVEGENERICPV(PL_parser->lex_shared);
2529 /* The here-doc parser needs to be able to peek into outer lexing
2530 scopes to find the body of the here-doc. So we put PL_linestr and
2531 PL_bufptr into lex_shared, to ‘share’ those values.
2533 PL_parser->lex_shared->ls_linestr = PL_linestr;
2534 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2536 PL_linestr = PL_lex_stuff;
2537 PL_lex_repl = PL_sublex_info.repl;
2538 PL_lex_stuff = NULL;
2539 PL_sublex_info.repl = NULL;
2541 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2542 = SvPVX(PL_linestr);
2543 PL_bufend += SvCUR(PL_linestr);
2544 PL_last_lop = PL_last_uni = NULL;
2545 SAVEFREESV(PL_linestr);
2546 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2548 PL_lex_dojoin = FALSE;
2549 PL_lex_brackets = PL_lex_formbrack = 0;
2550 PL_lex_allbrackets = 0;
2551 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2552 Newx(PL_lex_brackstack, 120, char);
2553 Newx(PL_lex_casestack, 12, char);
2554 PL_lex_casemods = 0;
2555 *PL_lex_casestack = '\0';
2557 PL_lex_state = LEX_INTERPCONCAT;
2558 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2560 Newxz(shared, 1, LEXSHARED);
2561 shared->ls_prev = PL_parser->lex_shared;
2562 PL_parser->lex_shared = shared;
2564 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2565 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2566 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2567 PL_lex_inpat = PL_sublex_info.sub_op;
2569 PL_lex_inpat = NULL;
2576 * Restores lexer state after a S_sublex_push.
2583 if (!PL_lex_starts++) {
2584 SV * const sv = newSVpvs("");
2585 if (SvUTF8(PL_linestr))
2587 PL_expect = XOPERATOR;
2588 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2592 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2593 PL_lex_state = LEX_INTERPCASEMOD;
2597 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2598 assert(PL_lex_inwhat != OP_TRANSR);
2599 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2600 PL_linestr = PL_lex_repl;
2602 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2603 PL_bufend += SvCUR(PL_linestr);
2604 PL_last_lop = PL_last_uni = NULL;
2605 PL_lex_dojoin = FALSE;
2606 PL_lex_brackets = 0;
2607 PL_lex_allbrackets = 0;
2608 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2609 PL_lex_casemods = 0;
2610 *PL_lex_casestack = '\0';
2612 if (SvEVALED(PL_lex_repl)) {
2613 PL_lex_state = LEX_INTERPNORMAL;
2615 /* we don't clear PL_lex_repl here, so that we can check later
2616 whether this is an evalled subst; that means we rely on the
2617 logic to ensure sublex_done() is called again only via the
2618 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2621 PL_lex_state = LEX_INTERPCONCAT;
2631 PL_endwhite = newSVpvs("");
2632 sv_catsv(PL_endwhite, PL_thiswhite);
2636 sv_setpvs(PL_thistoken,"");
2638 PL_realtokenstart = -1;
2642 PL_bufend = SvPVX(PL_linestr);
2643 PL_bufend += SvCUR(PL_linestr);
2644 PL_expect = XOPERATOR;
2645 PL_sublex_info.sub_inwhat = 0;
2650 PERL_STATIC_INLINE SV*
2651 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2653 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2654 * interior, hence to the "}". Finds what the name resolves to, returning
2655 * an SV* containing it; NULL if no valid one found */
2657 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2664 const U8* first_bad_char_loc;
2665 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2667 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2669 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2671 &first_bad_char_loc))
2673 /* If warnings are on, this will print a more detailed analysis of what
2674 * is wrong than the error message below */
2675 utf8n_to_uvuni(first_bad_char_loc,
2676 e - ((char *) first_bad_char_loc),
2679 /* We deliberately don't try to print the malformed character, which
2680 * might not print very well; it also may be just the first of many
2681 * malformations, so don't print what comes after it */
2682 yyerror(Perl_form(aTHX_
2683 "Malformed UTF-8 character immediately after '%.*s'",
2684 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr));
2688 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2689 /* include the <}> */
2690 e - backslash_ptr + 1);
2692 SvREFCNT_dec_NN(res);
2696 /* See if the charnames handler is the Perl core's, and if so, we can skip
2697 * the validation needed for a user-supplied one, as Perl's does its own
2699 table = GvHV(PL_hintgv); /* ^H */
2700 cvp = hv_fetchs(table, "charnames", FALSE);
2701 if (cvp && (cv = *cvp) && SvROK(cv) && ((rv = SvRV(cv)) != NULL)
2702 && SvTYPE(rv) == SVt_PVCV && ((stash = CvSTASH(rv)) != NULL))
2704 const char * const name = HvNAME(stash);
2705 if strEQ(name, "_charnames") {
2710 /* Here, it isn't Perl's charname handler. We can't rely on a
2711 * user-supplied handler to validate the input name. For non-ut8 input,
2712 * look to see that the first character is legal. Then loop through the
2713 * rest checking that each is a continuation */
2715 /* This code needs to be sync'ed with a regex in _charnames.pm which does
2719 if (! isALPHAU(*s)) {
2724 if (! isCHARNAME_CONT(*s)) {
2727 if (*s == ' ' && *(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2728 Perl_warn(aTHX_ "A sequence of multiple spaces in a charnames alias definition is deprecated");
2732 if (*(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2733 Perl_warn(aTHX_ "Trailing white-space in a charnames alias definition is deprecated");
2737 /* Similarly for utf8. For invariants can check directly; for other
2738 * Latin1, can calculate their code point and check; otherwise use a
2740 if (UTF8_IS_INVARIANT(*s)) {
2741 if (! isALPHAU(*s)) {
2745 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2746 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s, *(s+1))))) {
2752 if (! PL_utf8_charname_begin) {
2753 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2754 PL_utf8_charname_begin = _core_swash_init("utf8",
2755 "_Perl_Charname_Begin",
2757 1, 0, NULL, &flags);
2759 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2766 if (UTF8_IS_INVARIANT(*s)) {
2767 if (! isCHARNAME_CONT(*s)) {
2770 if (*s == ' ' && *(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2771 Perl_warn(aTHX_ "A sequence of multiple spaces in a charnames alias definition is deprecated");
2775 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2776 if (! isCHARNAME_CONT(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s,
2784 if (! PL_utf8_charname_continue) {
2785 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2786 PL_utf8_charname_continue = _core_swash_init("utf8",
2787 "_Perl_Charname_Continue",
2789 1, 0, NULL, &flags);
2791 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2797 if (*(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2798 Perl_warn(aTHX_ "Trailing white-space in a charnames alias definition is deprecated");
2802 if (SvUTF8(res)) { /* Don't accept malformed input */
2803 const U8* first_bad_char_loc;
2805 const char* const str = SvPV_const(res, len);
2806 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2807 /* If warnings are on, this will print a more detailed analysis of
2808 * what is wrong than the error message below */
2809 utf8n_to_uvuni(first_bad_char_loc,
2810 (char *) first_bad_char_loc - str,
2813 /* We deliberately don't try to print the malformed character,
2814 * which might not print very well; it also may be just the first
2815 * of many malformations, so don't print what comes after it */
2818 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2819 (int) (e - backslash_ptr + 1), backslash_ptr,
2820 (int) ((char *) first_bad_char_loc - str), str
2830 int bad_char_size = ((UTF) ? UTF8SKIP(s) : 1);
2832 /* The final %.*s makes sure that should the trailing NUL be missing
2833 * that this print won't run off the end of the string */
2836 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2837 (int)(s - backslash_ptr + bad_char_size), backslash_ptr,
2838 (int)(e - s + bad_char_size), s + bad_char_size
2840 UTF ? SVf_UTF8 : 0);
2848 Extracts the next constant part of a pattern, double-quoted string,
2849 or transliteration. This is terrifying code.
2851 For example, in parsing the double-quoted string "ab\x63$d", it would
2852 stop at the '$' and return an OP_CONST containing 'abc'.
2854 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2855 processing a pattern (PL_lex_inpat is true), a transliteration
2856 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2858 Returns a pointer to the character scanned up to. If this is
2859 advanced from the start pointer supplied (i.e. if anything was
2860 successfully parsed), will leave an OP_CONST for the substring scanned
2861 in pl_yylval. Caller must intuit reason for not parsing further
2862 by looking at the next characters herself.
2866 \N{FOO} => \N{U+hex_for_character_FOO}
2867 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2870 all other \-char, including \N and \N{ apart from \N{ABC}
2873 @ and $ where it appears to be a var, but not for $ as tail anchor
2878 In transliterations:
2879 characters are VERY literal, except for - not at the start or end
2880 of the string, which indicates a range. If the range is in bytes,
2881 scan_const expands the range to the full set of intermediate
2882 characters. If the range is in utf8, the hyphen is replaced with
2883 a certain range mark which will be handled by pmtrans() in op.c.
2885 In double-quoted strings:
2887 double-quoted style: \r and \n
2888 constants: \x31, etc.
2889 deprecated backrefs: \1 (in substitution replacements)
2890 case and quoting: \U \Q \E
2893 scan_const does *not* construct ops to handle interpolated strings.
2894 It stops processing as soon as it finds an embedded $ or @ variable
2895 and leaves it to the caller to work out what's going on.
2897 embedded arrays (whether in pattern or not) could be:
2898 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2900 $ in double-quoted strings must be the symbol of an embedded scalar.
2902 $ in pattern could be $foo or could be tail anchor. Assumption:
2903 it's a tail anchor if $ is the last thing in the string, or if it's
2904 followed by one of "()| \r\n\t"
2906 \1 (backreferences) are turned into $1 in substitutions
2908 The structure of the code is
2909 while (there's a character to process) {
2910 handle transliteration ranges
2911 skip regexp comments /(?#comment)/ and codes /(?{code})/
2912 skip #-initiated comments in //x patterns
2913 check for embedded arrays
2914 check for embedded scalars
2916 deprecate \1 in substitution replacements
2917 handle string-changing backslashes \l \U \Q \E, etc.
2918 switch (what was escaped) {
2919 handle \- in a transliteration (becomes a literal -)
2920 if a pattern and not \N{, go treat as regular character
2921 handle \132 (octal characters)
2922 handle \x15 and \x{1234} (hex characters)
2923 handle \N{name} (named characters, also \N{3,5} in a pattern)
2924 handle \cV (control characters)
2925 handle printf-style backslashes (\f, \r, \n, etc)
2928 } (end if backslash)
2929 handle regular character
2930 } (end while character to read)
2935 S_scan_const(pTHX_ char *start)
2938 char *send = PL_bufend; /* end of the constant */
2939 SV *sv = newSV(send - start); /* sv for the constant. See
2940 note below on sizing. */
2941 char *s = start; /* start of the constant */
2942 char *d = SvPVX(sv); /* destination for copies */
2943 bool dorange = FALSE; /* are we in a translit range? */
2944 bool didrange = FALSE; /* did we just finish a range? */
2945 bool in_charclass = FALSE; /* within /[...]/ */
2946 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2947 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2948 to be UTF8? But, this can
2949 show as true when the source
2950 isn't utf8, as for example
2951 when it is entirely composed
2953 SV *res; /* result from charnames */
2955 /* Note on sizing: The scanned constant is placed into sv, which is
2956 * initialized by newSV() assuming one byte of output for every byte of
2957 * input. This routine expects newSV() to allocate an extra byte for a
2958 * trailing NUL, which this routine will append if it gets to the end of
2959 * the input. There may be more bytes of input than output (eg., \N{LATIN
2960 * CAPITAL LETTER A}), or more output than input if the constant ends up
2961 * recoded to utf8, but each time a construct is found that might increase
2962 * the needed size, SvGROW() is called. Its size parameter each time is
2963 * based on the best guess estimate at the time, namely the length used so
2964 * far, plus the length the current construct will occupy, plus room for
2965 * the trailing NUL, plus one byte for every input byte still unscanned */
2967 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2970 UV literal_endpoint = 0;
2971 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2974 PERL_ARGS_ASSERT_SCAN_CONST;
2976 assert(PL_lex_inwhat != OP_TRANSR);
2977 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2978 /* If we are doing a trans and we know we want UTF8 set expectation */
2979 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2980 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2983 /* Protect sv from errors and fatal warnings. */
2984 ENTER_with_name("scan_const");
2987 while (s < send || dorange) {
2989 /* get transliterations out of the way (they're most literal) */
2990 if (PL_lex_inwhat == OP_TRANS) {
2991 /* expand a range A-Z to the full set of characters. AIE! */
2993 I32 i; /* current expanded character */
2994 I32 min; /* first character in range */
2995 I32 max; /* last character in range */
3006 char * const c = (char*)utf8_hop((U8*)d, -1);
3010 *c = (char)UTF_TO_NATIVE(0xff);
3011 /* mark the range as done, and continue */
3017 i = d - SvPVX_const(sv); /* remember current offset */
3020 SvLEN(sv) + (has_utf8 ?
3021 (512 - UTF_CONTINUATION_MARK +
3024 /* How many two-byte within 0..255: 128 in UTF-8,
3025 * 96 in UTF-8-mod. */
3027 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
3029 d = SvPVX(sv) + i; /* refresh d after realloc */
3033 for (j = 0; j <= 1; j++) {
3034 char * const c = (char*)utf8_hop((U8*)d, -1);
3035 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
3041 max = (U8)0xff; /* only to \xff */
3042 uvmax = uv; /* \x{100} to uvmax */
3044 d = c; /* eat endpoint chars */
3049 d -= 2; /* eat the first char and the - */
3050 min = (U8)*d; /* first char in range */
3051 max = (U8)d[1]; /* last char in range */
3058 "Invalid range \"%c-%c\" in transliteration operator",
3059 (char)min, (char)max);
3063 if (literal_endpoint == 2 &&
3064 ((isLOWER(min) && isLOWER(max)) ||
3065 (isUPPER(min) && isUPPER(max)))) {
3067 for (i = min; i <= max; i++)
3069 *d++ = NATIVE_TO_NEED(has_utf8,i);
3071 for (i = min; i <= max; i++)
3073 *d++ = NATIVE_TO_NEED(has_utf8,i);
3078 for (i = min; i <= max; i++)
3081 const U8 ch = (U8)NATIVE_TO_UTF(i);
3082 if (UNI_IS_INVARIANT(ch))
3085 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
3086 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
3095 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
3097 *d++ = (char)UTF_TO_NATIVE(0xff);
3099 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
3103 /* mark the range as done, and continue */
3107 literal_endpoint = 0;
3112 /* range begins (ignore - as first or last char) */
3113 else if (*s == '-' && s+1 < send && s != start) {
3115 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
3122 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
3132 literal_endpoint = 0;
3133 native_range = TRUE;
3138 /* if we get here, we're not doing a transliteration */
3140 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3143 while (s1 >= start && *s1-- == '\\')
3146 in_charclass = TRUE;
3149 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3152 while (s1 >= start && *s1-- == '\\')
3155 in_charclass = FALSE;
3158 /* skip for regexp comments /(?#comment)/, except for the last
3159 * char, which will be done separately.
3160 * Stop on (?{..}) and friends */
3162 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
3164 while (s+1 < send && *s != ')')
3165 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3167 else if (!PL_lex_casemods && !in_charclass &&
3168 ( s[2] == '{' /* This should match regcomp.c */
3169 || (s[2] == '?' && s[3] == '{')))
3175 /* likewise skip #-initiated comments in //x patterns */
3176 else if (*s == '#' && PL_lex_inpat &&
3177 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3178 while (s+1 < send && *s != '\n')
3179 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3182 /* no further processing of single-quoted regex */
3183 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3184 goto default_action;
3186 /* check for embedded arrays
3187 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3189 else if (*s == '@' && s[1]) {
3190 if (isWORDCHAR_lazy_if(s+1,UTF))
3192 if (strchr(":'{$", s[1]))
3194 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3195 break; /* in regexp, neither @+ nor @- are interpolated */
3198 /* check for embedded scalars. only stop if we're sure it's a
3201 else if (*s == '$') {
3202 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3204 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3206 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3207 "Possible unintended interpolation of $\\ in regex");
3209 break; /* in regexp, $ might be tail anchor */
3213 /* End of else if chain - OP_TRANS rejoin rest */
3216 if (*s == '\\' && s+1 < send) {
3217 char* e; /* Can be used for ending '}', etc. */
3221 /* warn on \1 - \9 in substitution replacements, but note that \11
3222 * is an octal; and \19 is \1 followed by '9' */
3223 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3224 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3226 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3231 /* string-change backslash escapes */
3232 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3236 /* In a pattern, process \N, but skip any other backslash escapes.
3237 * This is because we don't want to translate an escape sequence
3238 * into a meta symbol and have the regex compiler use the meta
3239 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3240 * in spite of this, we do have to process \N here while the proper
3241 * charnames handler is in scope. See bugs #56444 and #62056.
3242 * There is a complication because \N in a pattern may also stand
3243 * for 'match a non-nl', and not mean a charname, in which case its
3244 * processing should be deferred to the regex compiler. To be a
3245 * charname it must be followed immediately by a '{', and not look
3246 * like \N followed by a curly quantifier, i.e., not something like
3247 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3249 else if (PL_lex_inpat
3252 || regcurly(s + 1, FALSE)))
3254 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3255 goto default_action;
3260 /* quoted - in transliterations */
3262 if (PL_lex_inwhat == OP_TRANS) {
3269 if ((isALPHANUMERIC(*s)))
3270 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3271 "Unrecognized escape \\%c passed through",
3273 /* default action is to copy the quoted character */
3274 goto default_action;
3277 /* eg. \132 indicates the octal constant 0132 */
3278 case '0': case '1': case '2': case '3':
3279 case '4': case '5': case '6': case '7':
3281 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3283 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3285 if (len < 3 && s < send && isDIGIT(*s)
3286 && ckWARN(WARN_MISC))
3288 Perl_warner(aTHX_ packWARN(WARN_MISC),
3289 "%s", form_short_octal_warning(s, len));
3292 goto NUM_ESCAPE_INSERT;
3294 /* eg. \o{24} indicates the octal constant \024 */
3299 bool valid = grok_bslash_o(&s, &uv, &error,
3300 TRUE, /* Output warning */
3301 FALSE, /* Not strict */
3302 TRUE, /* Output warnings for
3309 goto NUM_ESCAPE_INSERT;
3312 /* eg. \x24 indicates the hex constant 0x24 */
3317 bool valid = grok_bslash_x(&s, &uv, &error,
3318 TRUE, /* Output warning */
3319 FALSE, /* Not strict */
3320 TRUE, /* Output warnings for
3330 /* Insert oct or hex escaped character. There will always be
3331 * enough room in sv since such escapes will be longer than any
3332 * UTF-8 sequence they can end up as, except if they force us
3333 * to recode the rest of the string into utf8 */
3335 /* Here uv is the ordinal of the next character being added in
3336 * unicode (converted from native). */
3337 if (!UNI_IS_INVARIANT(uv)) {
3338 if (!has_utf8 && uv > 255) {
3339 /* Might need to recode whatever we have accumulated so
3340 * far if it contains any chars variant in utf8 or
3343 SvCUR_set(sv, d - SvPVX_const(sv));
3346 /* See Note on sizing above. */
3347 sv_utf8_upgrade_flags_grow(sv,
3348 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3349 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3350 d = SvPVX(sv) + SvCUR(sv);
3355 d = (char*)uvuni_to_utf8((U8*)d, uv);
3356 if (PL_lex_inwhat == OP_TRANS &&
3357 PL_sublex_info.sub_op) {
3358 PL_sublex_info.sub_op->op_private |=
3359 (PL_lex_repl ? OPpTRANS_FROM_UTF
3363 if (uv > 255 && !dorange)
3364 native_range = FALSE;
3377 /* In a non-pattern \N must be a named character, like \N{LATIN
3378 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3379 * mean to match a non-newline. For non-patterns, named
3380 * characters are converted to their string equivalents. In
3381 * patterns, named characters are not converted to their
3382 * ultimate forms for the same reasons that other escapes
3383 * aren't. Instead, they are converted to the \N{U+...} form
3384 * to get the value from the charnames that is in effect right
3385 * now, while preserving the fact that it was a named character
3386 * so that the regex compiler knows this */
3388 /* This section of code doesn't generally use the
3389 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3390 * a close examination of this macro and determined it is a
3391 * no-op except on utfebcdic variant characters. Every
3392 * character generated by this that would normally need to be
3393 * enclosed by this macro is invariant, so the macro is not
3394 * needed, and would complicate use of copy(). XXX There are
3395 * other parts of this file where the macro is used
3396 * inconsistently, but are saved by it being a no-op */
3398 /* The structure of this section of code (besides checking for
3399 * errors and upgrading to utf8) is:
3400 * Further disambiguate between the two meanings of \N, and if
3401 * not a charname, go process it elsewhere
3402 * If of form \N{U+...}, pass it through if a pattern;
3403 * otherwise convert to utf8
3404 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3405 * pattern; otherwise convert to utf8 */
3407 /* Here, s points to the 'N'; the test below is guaranteed to
3408 * succeed if we are being called on a pattern as we already
3409 * know from a test above that the next character is a '{'.
3410 * On a non-pattern \N must mean 'named sequence, which
3411 * requires braces */
3414 yyerror("Missing braces on \\N{}");
3419 /* If there is no matching '}', it is an error. */
3420 if (! (e = strchr(s, '}'))) {
3421 if (! PL_lex_inpat) {
3422 yyerror("Missing right brace on \\N{}");
3424 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3429 /* Here it looks like a named character */
3431 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3432 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3433 | PERL_SCAN_DISALLOW_PREFIX;
3436 /* For \N{U+...}, the '...' is a unicode value even on
3437 * EBCDIC machines */
3438 s += 2; /* Skip to next char after the 'U+' */
3440 uv = grok_hex(s, &len, &flags, NULL);
3441 if (len == 0 || len != (STRLEN)(e - s)) {
3442 yyerror("Invalid hexadecimal number in \\N{U+...}");
3449 /* On non-EBCDIC platforms, pass through to the regex
3450 * compiler unchanged. The reason we evaluated the
3451 * number above is to make sure there wasn't a syntax
3452 * error. But on EBCDIC we convert to native so
3453 * downstream code can continue to assume it's native
3455 s -= 5; /* Include the '\N{U+' */
3457 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3460 (unsigned int) UNI_TO_NATIVE(uv));
3462 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3466 else { /* Not a pattern: convert the hex to string */
3468 /* If destination is not in utf8, unconditionally
3469 * recode it to be so. This is because \N{} implies
3470 * Unicode semantics, and scalars have to be in utf8
3471 * to guarantee those semantics */
3473 SvCUR_set(sv, d - SvPVX_const(sv));
3476 /* See Note on sizing above. */
3477 sv_utf8_upgrade_flags_grow(
3479 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3480 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3481 d = SvPVX(sv) + SvCUR(sv);
3485 /* Add the string to the output */
3486 if (UNI_IS_INVARIANT(uv)) {
3489 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3492 else /* Here is \N{NAME} but not \N{U+...}. */
3493 if ((res = get_and_check_backslash_N_name(s, e)))
3496 const char *str = SvPV_const(res, len);
3499 if (! len) { /* The name resolved to an empty string */
3500 Copy("\\N{}", d, 4, char);
3504 /* In order to not lose information for the regex
3505 * compiler, pass the result in the specially made
3506 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3507 * the code points in hex of each character
3508 * returned by charnames */
3510 const char *str_end = str + len;
3511 const STRLEN off = d - SvPVX_const(sv);
3513 if (! SvUTF8(res)) {
3514 /* For the non-UTF-8 case, we can determine the
3515 * exact length needed without having to parse
3516 * through the string. Each character takes up
3517 * 2 hex digits plus either a trailing dot or
3519 d = off + SvGROW(sv, off
3521 + 6 /* For the "\N{U+", and
3523 + (STRLEN)(send - e));
3524 Copy("\\N{U+", d, 5, char);
3526 while (str < str_end) {
3528 my_snprintf(hex_string, sizeof(hex_string),
3529 "%02X.", (U8) *str);
3530 Copy(hex_string, d, 3, char);
3534 d--; /* We will overwrite below the final
3535 dot with a right brace */
3538 STRLEN char_length; /* cur char's byte length */
3540 /* and the number of bytes after this is
3541 * translated into hex digits */
3542 STRLEN output_length;
3544 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3545 * for max('U+', '.'); and 1 for NUL */
3546 char hex_string[2 * UTF8_MAXBYTES + 5];
3548 /* Get the first character of the result. */
3549 U32 uv = utf8n_to_uvuni((U8 *) str,
3553 /* Convert first code point to hex, including
3554 * the boiler plate before it. For all these,
3555 * we convert to native format so that
3556 * downstream code can continue to assume the
3557 * input is native */
3559 my_snprintf(hex_string, sizeof(hex_string),
3561 (unsigned int) UNI_TO_NATIVE(uv));
3563 /* Make sure there is enough space to hold it */
3564 d = off + SvGROW(sv, off
3566 + (STRLEN)(send - e)
3567 + 2); /* '}' + NUL */
3569 Copy(hex_string, d, output_length, char);
3572 /* For each subsequent character, append dot and
3573 * its ordinal in hex */
3574 while ((str += char_length) < str_end) {
3575 const STRLEN off = d - SvPVX_const(sv);
3576 U32 uv = utf8n_to_uvuni((U8 *) str,
3581 my_snprintf(hex_string,
3584 (unsigned int) UNI_TO_NATIVE(uv));
3586 d = off + SvGROW(sv, off
3588 + (STRLEN)(send - e)
3589 + 2); /* '}' + NUL */
3590 Copy(hex_string, d, output_length, char);
3595 *d++ = '}'; /* Done. Add the trailing brace */
3598 else { /* Here, not in a pattern. Convert the name to a
3601 /* If destination is not in utf8, unconditionally
3602 * recode it to be so. This is because \N{} implies
3603 * Unicode semantics, and scalars have to be in utf8
3604 * to guarantee those semantics */
3606 SvCUR_set(sv, d - SvPVX_const(sv));
3609 /* See Note on sizing above. */
3610 sv_utf8_upgrade_flags_grow(sv,
3611 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3612 len + (STRLEN)(send - s) + 1);
3613 d = SvPVX(sv) + SvCUR(sv);
3615 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3617 /* See Note on sizing above. (NOTE: SvCUR() is not
3618 * set correctly here). */
3619 const STRLEN off = d - SvPVX_const(sv);
3620 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3622 Copy(str, d, len, char);
3628 } /* End \N{NAME} */
3631 native_range = FALSE; /* \N{} is defined to be Unicode */
3633 s = e + 1; /* Point to just after the '}' */
3636 /* \c is a control character */
3640 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3643 yyerror("Missing control char name in \\c");
3647 /* printf-style backslashes, formfeeds, newlines, etc */
3649 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3652 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3655 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3658 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3661 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3664 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3667 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3673 } /* end if (backslash) */
3680 /* If we started with encoded form, or already know we want it,
3681 then encode the next character */
3682 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3686 /* One might think that it is wasted effort in the case of the
3687 * source being utf8 (this_utf8 == TRUE) to take the next character
3688 * in the source, convert it to an unsigned value, and then convert
3689 * it back again. But the source has not been validated here. The
3690 * routine that does the conversion checks for errors like
3693 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3694 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3696 SvCUR_set(sv, d - SvPVX_const(sv));
3699 /* See Note on sizing above. */
3700 sv_utf8_upgrade_flags_grow(sv,
3701 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3702 need + (STRLEN)(send - s) + 1);
3703 d = SvPVX(sv) + SvCUR(sv);
3705 } else if (need > len) {
3706 /* encoded value larger than old, may need extra space (NOTE:
3707 * SvCUR() is not set correctly here). See Note on sizing
3709 const STRLEN off = d - SvPVX_const(sv);
3710 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3714 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3716 if (uv > 255 && !dorange)
3717 native_range = FALSE;
3721 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3723 } /* while loop to process each character */
3725 /* terminate the string and set up the sv */
3727 SvCUR_set(sv, d - SvPVX_const(sv));
3728 if (SvCUR(sv) >= SvLEN(sv))
3729 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3730 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3733 if (PL_encoding && !has_utf8) {
3734 sv_recode_to_utf8(sv, PL_encoding);
3740 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3741 PL_sublex_info.sub_op->op_private |=
3742 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3746 /* shrink the sv if we allocated more than we used */
3747 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3748 SvPV_shrink_to_cur(sv);
3751 /* return the substring (via pl_yylval) only if we parsed anything */
3752 if (s > PL_bufptr) {
3753 SvREFCNT_inc_simple_void_NN(sv);
3754 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3755 const char *const key = PL_lex_inpat ? "qr" : "q";
3756 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3760 if (PL_lex_inwhat == OP_TRANS) {
3763 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3766 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3774 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3777 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3779 LEAVE_with_name("scan_const");
3784 * Returns TRUE if there's more to the expression (e.g., a subscript),
3787 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3789 * ->[ and ->{ return TRUE
3790 * { and [ outside a pattern are always subscripts, so return TRUE
3791 * if we're outside a pattern and it's not { or [, then return FALSE
3792 * if we're in a pattern and the first char is a {
3793 * {4,5} (any digits around the comma) returns FALSE
3794 * if we're in a pattern and the first char is a [
3796 * [SOMETHING] has a funky algorithm to decide whether it's a
3797 * character class or not. It has to deal with things like
3798 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3799 * anything else returns TRUE
3802 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3805 S_intuit_more(pTHX_ char *s)
3809 PERL_ARGS_ASSERT_INTUIT_MORE;
3811 if (PL_lex_brackets)
3813 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3815 if (*s != '{' && *s != '[')
3820 /* In a pattern, so maybe we have {n,m}. */
3822 if (regcurly(s, FALSE)) {
3828 /* On the other hand, maybe we have a character class */
3831 if (*s == ']' || *s == '^')
3834 /* this is terrifying, and it works */
3837 const char * const send = strchr(s,']');
3838 unsigned char un_char, last_un_char;
3839 char tmpbuf[sizeof PL_tokenbuf * 4];
3841 if (!send) /* has to be an expression */
3843 weight = 2; /* let's weigh the evidence */
3847 else if (isDIGIT(*s)) {
3849 if (isDIGIT(s[1]) && s[2] == ']')
3855 Zero(seen,256,char);
3857 for (; s < send; s++) {
3858 last_un_char = un_char;
3859 un_char = (unsigned char)*s;
3864 weight -= seen[un_char] * 10;
3865 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3867 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3868 len = (int)strlen(tmpbuf);
3869 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3870 UTF ? SVf_UTF8 : 0, SVt_PV))
3875 else if (*s == '$' && s[1] &&
3876 strchr("[#!%*<>()-=",s[1])) {
3877 if (/*{*/ strchr("])} =",s[2]))
3886 if (strchr("wds]",s[1]))
3888 else if (seen[(U8)'\''] || seen[(U8)'"'])
3890 else if (strchr("rnftbxcav",s[1]))
3892 else if (isDIGIT(s[1])) {
3894 while (s[1] && isDIGIT(s[1]))
3904 if (strchr("aA01! ",last_un_char))
3906 if (strchr("zZ79~",s[1]))
3908 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3909 weight -= 5; /* cope with negative subscript */
3912 if (!isWORDCHAR(last_un_char)
3913 && !(last_un_char == '$' || last_un_char == '@'
3914 || last_un_char == '&')
3915 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3920 if (keyword(tmpbuf, d - tmpbuf, 0))
3923 if (un_char == last_un_char + 1)
3925 weight -= seen[un_char];
3930 if (weight >= 0) /* probably a character class */
3940 * Does all the checking to disambiguate
3942 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3943 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3945 * First argument is the stuff after the first token, e.g. "bar".
3947 * Not a method if foo is a filehandle.
3948 * Not a method if foo is a subroutine prototyped to take a filehandle.
3949 * Not a method if it's really "Foo $bar"
3950 * Method if it's "foo $bar"
3951 * Not a method if it's really "print foo $bar"
3952 * Method if it's really "foo package::" (interpreted as package->foo)
3953 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3954 * Not a method if bar is a filehandle or package, but is quoted with
3959 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3962 char *s = start + (*start == '$');
3963 char tmpbuf[sizeof PL_tokenbuf];
3970 PERL_ARGS_ASSERT_INTUIT_METHOD;
3972 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3974 if (cv && SvPOK(cv)) {
3975 const char *proto = CvPROTO(cv);
3983 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3984 /* start is the beginning of the possible filehandle/object,
3985 * and s is the end of it
3986 * tmpbuf is a copy of it
3989 if (*start == '$') {
3990 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3991 isUPPER(*PL_tokenbuf))
3994 len = start - SvPVX(PL_linestr);
3998 start = SvPVX(PL_linestr) + len;
4002 return *s == '(' ? FUNCMETH : METHOD;
4004 if (!keyword(tmpbuf, len, 0)) {
4005 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
4009 soff = s - SvPVX(PL_linestr);
4013 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
4014 if (indirgv && GvCVu(indirgv))
4016 /* filehandle or package name makes it a method */
4017 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
4019 soff = s - SvPVX(PL_linestr);
4022 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
4023 return 0; /* no assumptions -- "=>" quotes bareword */
4025 start_force(PL_curforce);
4026 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
4027 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
4028 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
4030 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
4031 ( UTF ? SVf_UTF8 : 0 )));
4036 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
4038 return *s == '(' ? FUNCMETH : METHOD;
4044 /* Encoded script support. filter_add() effectively inserts a
4045 * 'pre-processing' function into the current source input stream.
4046 * Note that the filter function only applies to the current source file
4047 * (e.g., it will not affect files 'require'd or 'use'd by this one).
4049 * The datasv parameter (which may be NULL) can be used to pass
4050 * private data to this instance of the filter. The filter function
4051 * can recover the SV using the FILTER_DATA macro and use it to
4052 * store private buffers and state information.
4054 * The supplied datasv parameter is upgraded to a PVIO type
4055 * and the IoDIRP/IoANY field is used to store the function pointer,
4056 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
4057 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
4058 * private use must be set using malloc'd pointers.
4062 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
4071 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
4072 Perl_croak(aTHX_ "Source filters apply only to byte streams");
4074 if (!PL_rsfp_filters)
4075 PL_rsfp_filters = newAV();
4078 SvUPGRADE(datasv, SVt_PVIO);
4079 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
4080 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
4081 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
4082 FPTR2DPTR(void *, IoANY(datasv)),
4083 SvPV_nolen(datasv)));
4084 av_unshift(PL_rsfp_filters, 1);
4085 av_store(PL_rsfp_filters, 0, datasv) ;
4087 !PL_parser->filtered
4088 && PL_parser->lex_flags & LEX_EVALBYTES
4089 && PL_bufptr < PL_bufend
4091 const char *s = PL_bufptr;
4092 while (s < PL_bufend) {
4094 SV *linestr = PL_parser->linestr;
4095 char *buf = SvPVX(linestr);
4096 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
4097 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
4098 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
4099 STRLEN const linestart_pos = PL_parser->linestart - buf;
4100 STRLEN const last_uni_pos =
4101 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
4102 STRLEN const last_lop_pos =
4103 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
4104 av_push(PL_rsfp_filters, linestr);
4105 PL_parser->linestr =
4106 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
4107 buf = SvPVX(PL_parser->linestr);
4108 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4109 PL_parser->bufptr = buf + bufptr_pos;
4110 PL_parser->oldbufptr = buf + oldbufptr_pos;
4111 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4112 PL_parser->linestart = buf + linestart_pos;
4113 if (PL_parser->last_uni)
4114 PL_parser->last_uni = buf + last_uni_pos;
4115 if (PL_parser->last_lop)
4116 PL_parser->last_lop = buf + last_lop_pos;
4117 SvLEN(linestr) = SvCUR(linestr);
4118 SvCUR(linestr) = s-SvPVX(linestr);
4119 PL_parser->filtered = 1;
4129 /* Delete most recently added instance of this filter function. */
4131 Perl_filter_del(pTHX_ filter_t funcp)
4136 PERL_ARGS_ASSERT_FILTER_DEL;
4139 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4140 FPTR2DPTR(void*, funcp)));
4142 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4144 /* if filter is on top of stack (usual case) just pop it off */
4145 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4146 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4147 sv_free(av_pop(PL_rsfp_filters));
4151 /* we need to search for the correct entry and clear it */
4152 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4156 /* Invoke the idxth filter function for the current rsfp. */
4157 /* maxlen 0 = read one text line */
4159 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4164 /* This API is bad. It should have been using unsigned int for maxlen.
4165 Not sure if we want to change the API, but if not we should sanity
4166 check the value here. */
4167 unsigned int correct_length
4176 PERL_ARGS_ASSERT_FILTER_READ;
4178 if (!PL_parser || !PL_rsfp_filters)
4180 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4181 /* Provide a default input filter to make life easy. */
4182 /* Note that we append to the line. This is handy. */
4183 DEBUG_P(PerlIO_printf(Perl_debug_log,
4184 "filter_read %d: from rsfp\n", idx));
4185 if (correct_length) {
4188 const int old_len = SvCUR(buf_sv);
4190 /* ensure buf_sv is large enough */
4191 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4192 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4193 correct_length)) <= 0) {
4194 if (PerlIO_error(PL_rsfp))
4195 return -1; /* error */
4197 return 0 ; /* end of file */
4199 SvCUR_set(buf_sv, old_len + len) ;
4200 SvPVX(buf_sv)[old_len + len] = '\0';
4203 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4204 if (PerlIO_error(PL_rsfp))
4205 return -1; /* error */
4207 return 0 ; /* end of file */
4210 return SvCUR(buf_sv);
4212 /* Skip this filter slot if filter has been deleted */
4213 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4214 DEBUG_P(PerlIO_printf(Perl_debug_log,
4215 "filter_read %d: skipped (filter deleted)\n",
4217 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4219 if (SvTYPE(datasv) != SVt_PVIO) {
4220 if (correct_length) {
4222 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4223 if (!remainder) return 0; /* eof */
4224 if (correct_length > remainder) correct_length = remainder;
4225 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4226 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4229 const char *s = SvEND(datasv);
4230 const char *send = SvPVX(datasv) + SvLEN(datasv);
4238 if (s == send) return 0; /* eof */
4239 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4240 SvCUR_set(datasv, s-SvPVX(datasv));
4242 return SvCUR(buf_sv);
4244 /* Get function pointer hidden within datasv */
4245 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4246 DEBUG_P(PerlIO_printf(Perl_debug_log,
4247 "filter_read %d: via function %p (%s)\n",
4248 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4249 /* Call function. The function is expected to */
4250 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4251 /* Return: <0:error, =0:eof, >0:not eof */
4252 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4256 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4260 PERL_ARGS_ASSERT_FILTER_GETS;
4262 #ifdef PERL_CR_FILTER
4263 if (!PL_rsfp_filters) {
4264 filter_add(S_cr_textfilter,NULL);
4267 if (PL_rsfp_filters) {
4269 SvCUR_set(sv, 0); /* start with empty line */
4270 if (FILTER_READ(0, sv, 0) > 0)
4271 return ( SvPVX(sv) ) ;
4276 return (sv_gets(sv, PL_rsfp, append));
4280 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4285 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4287 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4291 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4292 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4294 return GvHV(gv); /* Foo:: */
4297 /* use constant CLASS => 'MyClass' */
4298 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4299 if (gv && GvCV(gv)) {
4300 SV * const sv = cv_const_sv(GvCV(gv));
4302 pkgname = SvPV_const(sv, len);
4305 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4309 * S_readpipe_override
4310 * Check whether readpipe() is overridden, and generates the appropriate
4311 * optree, provided sublex_start() is called afterwards.
4314 S_readpipe_override(pTHX)
4317 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4318 pl_yylval.ival = OP_BACKTICK;
4320 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4322 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4323 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4324 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4326 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4327 op_append_elem(OP_LIST,
4328 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4329 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4336 * The intent of this yylex wrapper is to minimize the changes to the
4337 * tokener when we aren't interested in collecting madprops. It remains
4338 * to be seen how successful this strategy will be...
4345 char *s = PL_bufptr;
4347 /* make sure PL_thiswhite is initialized */
4351 /* previous token ate up our whitespace? */
4352 if (!PL_lasttoke && PL_nextwhite) {
4353 PL_thiswhite = PL_nextwhite;
4357 /* isolate the token, and figure out where it is without whitespace */
4358 PL_realtokenstart = -1;
4362 assert(PL_curforce < 0);
4364 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4365 if (!PL_thistoken) {
4366 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4367 PL_thistoken = newSVpvs("");
4369 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4370 PL_thistoken = newSVpvn(tstart, s - tstart);
4373 if (PL_thismad) /* install head */
4374 CURMAD('X', PL_thistoken);
4377 /* last whitespace of a sublex? */
4378 if (optype == ')' && PL_endwhite) {
4379 CURMAD('X', PL_endwhite);
4384 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4385 if (!PL_thiswhite && !PL_endwhite && !optype) {
4386 sv_free(PL_thistoken);
4391 /* put off final whitespace till peg */
4392 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4393 PL_nextwhite = PL_thiswhite;
4396 else if (PL_thisopen) {
4397 CURMAD('q', PL_thisopen);
4399 sv_free(PL_thistoken);
4403 /* Store actual token text as madprop X */
4404 CURMAD('X', PL_thistoken);
4408 /* add preceding whitespace as madprop _ */
4409 CURMAD('_', PL_thiswhite);
4413 /* add quoted material as madprop = */
4414 CURMAD('=', PL_thisstuff);
4418 /* add terminating quote as madprop Q */
4419 CURMAD('Q', PL_thisclose);
4423 /* special processing based on optype */
4427 /* opval doesn't need a TOKEN since it can already store mp */
4437 if (pl_yylval.opval)
4438 append_madprops(PL_thismad, pl_yylval.opval, 0);
4446 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4459 /* remember any fake bracket that lexer is about to discard */
4460 if (PL_lex_brackets == 1 &&
4461 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4464 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4467 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4468 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4471 break; /* don't bother looking for trailing comment */
4480 /* attach a trailing comment to its statement instead of next token */
4484 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4486 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4488 if (*s == '\n' || *s == '#') {
4489 while (s < PL_bufend && *s != '\n')
4493 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4494 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4507 /* Create new token struct. Note: opvals return early above. */
4508 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4515 S_tokenize_use(pTHX_ int is_use, char *s) {
4518 PERL_ARGS_ASSERT_TOKENIZE_USE;
4520 if (PL_expect != XSTATE)
4521 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4522 is_use ? "use" : "no"));
4525 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4526 s = force_version(s, TRUE);
4527 if (*s == ';' || *s == '}'
4528 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4529 start_force(PL_curforce);
4530 NEXTVAL_NEXTTOKE.opval = NULL;
4533 else if (*s == 'v') {
4534 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4535 s = force_version(s, FALSE);
4539 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4540 s = force_version(s, FALSE);
4542 pl_yylval.ival = is_use;
4546 static const char* const exp_name[] =
4547 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4548 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4552 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4554 S_word_takes_any_delimeter(char *p, STRLEN len)
4556 return (len == 1 && strchr("msyq", p[0])) ||
4558 (p[0] == 't' && p[1] == 'r') ||
4559 (p[0] == 'q' && strchr("qwxr", p[1]))));
4565 Works out what to call the token just pulled out of the input
4566 stream. The yacc parser takes care of taking the ops we return and
4567 stitching them into a tree.
4570 The type of the next token
4573 Switch based on the current state:
4574 - if we already built the token before, use it
4575 - if we have a case modifier in a string, deal with that
4576 - handle other cases of interpolation inside a string
4577 - scan the next line if we are inside a format
4578 In the normal state switch on the next character:
4580 if alphabetic, go to key lookup
4581 unrecoginized character - croak
4582 - 0/4/26: handle end-of-line or EOF
4583 - cases for whitespace
4584 - \n and #: handle comments and line numbers
4585 - various operators, brackets and sigils
4588 - 'v': vstrings (or go to key lookup)
4589 - 'x' repetition operator (or go to key lookup)
4590 - other ASCII alphanumerics (key lookup begins here):
4593 scan built-in keyword (but do nothing with it yet)
4594 check for statement label
4595 check for lexical subs
4596 goto just_a_word if there is one
4597 see whether built-in keyword is overridden
4598 switch on keyword number:
4599 - default: just_a_word:
4600 not a built-in keyword; handle bareword lookup
4601 disambiguate between method and sub call
4602 fall back to bareword
4603 - cases for built-in keywords
4608 #pragma segment Perl_yylex
4614 char *s = PL_bufptr;
4621 /* orig_keyword, gvp, and gv are initialized here because
4622 * jump to the label just_a_word_zero can bypass their
4623 * initialization later. */
4624 I32 orig_keyword = 0;
4629 SV* tmp = newSVpvs("");
4630 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4631 (IV)CopLINE(PL_curcop),
4632 lex_state_names[PL_lex_state],
4633 exp_name[PL_expect],
4634 pv_display(tmp, s, strlen(s), 0, 60));
4638 switch (PL_lex_state) {
4640 case LEX_NORMAL: /* Some compilers will produce faster */
4641 case LEX_INTERPNORMAL: /* code if we comment these out. */
4645 /* when we've already built the next token, just pull it out of the queue */
4649 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4651 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4652 PL_nexttoke[PL_lasttoke].next_mad = 0;
4653 if (PL_thismad && PL_thismad->mad_key == '_') {
4654 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4655 PL_thismad->mad_val = 0;
4656 mad_free(PL_thismad);
4661 PL_lex_state = PL_lex_defer;
4662 PL_expect = PL_lex_expect;
4663 PL_lex_defer = LEX_NORMAL;
4664 if (!PL_nexttoke[PL_lasttoke].next_type)
4669 pl_yylval = PL_nextval[PL_nexttoke];
4671 PL_lex_state = PL_lex_defer;
4672 PL_expect = PL_lex_expect;
4673 PL_lex_defer = LEX_NORMAL;
4679 next_type = PL_nexttoke[PL_lasttoke].next_type;
4681 next_type = PL_nexttype[PL_nexttoke];
4683 if (next_type & (7<<24)) {
4684 if (next_type & (1<<24)) {
4685 if (PL_lex_brackets > 100)
4686 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4687 PL_lex_brackstack[PL_lex_brackets++] =
4688 (char) ((next_type >> 16) & 0xff);
4690 if (next_type & (2<<24))
4691 PL_lex_allbrackets++;
4692 if (next_type & (4<<24))
4693 PL_lex_allbrackets--;
4694 next_type &= 0xffff;
4696 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4699 /* interpolated case modifiers like \L \U, including \Q and \E.
4700 when we get here, PL_bufptr is at the \
4702 case LEX_INTERPCASEMOD:
4704 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4706 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4707 PL_bufptr, PL_bufend, *PL_bufptr);
4709 /* handle \E or end of string */
4710 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4712 if (PL_lex_casemods) {
4713 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4714 PL_lex_casestack[PL_lex_casemods] = '\0';
4716 if (PL_bufptr != PL_bufend
4717 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4718 || oldmod == 'F')) {
4720 PL_lex_state = LEX_INTERPCONCAT;
4723 PL_thistoken = newSVpvs("\\E");
4726 PL_lex_allbrackets--;
4729 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4730 /* Got an unpaired \E */
4731 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4732 "Useless use of \\E");
4735 while (PL_bufptr != PL_bufend &&
4736 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4739 PL_thiswhite = newSVpvs("");
4740 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4745 if (PL_bufptr != PL_bufend)
4748 PL_lex_state = LEX_INTERPCONCAT;
4752 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4753 "### Saw case modifier\n"); });
4755 if (s[1] == '\\' && s[2] == 'E') {
4759 PL_thiswhite = newSVpvs("");
4760 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4764 PL_lex_state = LEX_INTERPCONCAT;
4769 if (!PL_madskills) /* when just compiling don't need correct */
4770 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4771 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4772 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4773 (strchr(PL_lex_casestack, 'L')
4774 || strchr(PL_lex_casestack, 'U')
4775 || strchr(PL_lex_casestack, 'F'))) {
4776 PL_lex_casestack[--PL_lex_casemods] = '\0';
4777 PL_lex_allbrackets--;
4780 if (PL_lex_casemods > 10)
4781 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4782 PL_lex_casestack[PL_lex_casemods++] = *s;
4783 PL_lex_casestack[PL_lex_casemods] = '\0';
4784 PL_lex_state = LEX_INTERPCONCAT;
4785 start_force(PL_curforce);
4786 NEXTVAL_NEXTTOKE.ival = 0;
4787 force_next((2<<24)|'(');
4788 start_force(PL_curforce);
4790 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4792 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4794 NEXTVAL_NEXTTOKE.ival = OP_LC;
4796 NEXTVAL_NEXTTOKE.ival = OP_UC;
4798 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4800 NEXTVAL_NEXTTOKE.ival = OP_FC;
4802 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4804 SV* const tmpsv = newSVpvs("\\ ");
4805 /* replace the space with the character we want to escape
4807 SvPVX(tmpsv)[1] = *s;
4813 if (PL_lex_starts) {
4819 sv_free(PL_thistoken);
4820 PL_thistoken = newSVpvs("");
4823 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4824 if (PL_lex_casemods == 1 && PL_lex_inpat)
4833 case LEX_INTERPPUSH:
4834 return REPORT(sublex_push());
4836 case LEX_INTERPSTART:
4837 if (PL_bufptr == PL_bufend)
4838 return REPORT(sublex_done());
4839 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4840 "### Interpolated variable\n"); });
4842 PL_lex_dojoin = (*PL_bufptr == '@');
4843 PL_lex_state = LEX_INTERPNORMAL;
4844 if (PL_lex_dojoin) {
4845 start_force(PL_curforce);
4846 NEXTVAL_NEXTTOKE.ival = 0;
4848 start_force(PL_curforce);
4849 force_ident("\"", '$');
4850 start_force(PL_curforce);
4851 NEXTVAL_NEXTTOKE.ival = 0;
4853 start_force(PL_curforce);
4854 NEXTVAL_NEXTTOKE.ival = 0;
4855 force_next((2<<24)|'(');
4856 start_force(PL_curforce);
4857 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4860 /* Convert (?{...}) and friends to 'do {...}' */
4861 if (PL_lex_inpat && *PL_bufptr == '(') {
4862 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4864 if (*PL_bufptr != '{')
4866 start_force(PL_curforce);
4867 /* XXX probably need a CURMAD(something) here */
4868 PL_expect = XTERMBLOCK;
4872 if (PL_lex_starts++) {
4877 sv_free(PL_thistoken);
4878 PL_thistoken = newSVpvs("");
4881 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4882 if (!PL_lex_casemods && PL_lex_inpat)
4889 case LEX_INTERPENDMAYBE:
4890 if (intuit_more(PL_bufptr)) {
4891 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4897 if (PL_lex_dojoin) {
4898 PL_lex_dojoin = FALSE;
4899 PL_lex_state = LEX_INTERPCONCAT;
4903 sv_free(PL_thistoken);
4904 PL_thistoken = newSVpvs("");
4907 PL_lex_allbrackets--;
4910 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4911 && SvEVALED(PL_lex_repl))
4913 if (PL_bufptr != PL_bufend)
4914 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4917 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4918 re_eval_str. If the here-doc body’s length equals the previous
4919 value of re_eval_start, re_eval_start will now be null. So
4920 check re_eval_str as well. */
4921 if (PL_parser->lex_shared->re_eval_start
4922 || PL_parser->lex_shared->re_eval_str) {
4924 if (*PL_bufptr != ')')
4925 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4927 /* having compiled a (?{..}) expression, return the original
4928 * text too, as a const */
4929 if (PL_parser->lex_shared->re_eval_str) {
4930 sv = PL_parser->lex_shared->re_eval_str;
4931 PL_parser->lex_shared->re_eval_str = NULL;
4933 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4934 SvPV_shrink_to_cur(sv);
4936 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4937 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4938 start_force(PL_curforce);
4939 /* XXX probably need a CURMAD(something) here */
4940 NEXTVAL_NEXTTOKE.opval =
4941 (OP*)newSVOP(OP_CONST, 0,
4944 PL_parser->lex_shared->re_eval_start = NULL;
4950 case LEX_INTERPCONCAT:
4952 if (PL_lex_brackets)
4953 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4954 (long) PL_lex_brackets);
4956 if (PL_bufptr == PL_bufend)
4957 return REPORT(sublex_done());
4959 /* m'foo' still needs to be parsed for possible (?{...}) */
4960 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4961 SV *sv = newSVsv(PL_linestr);
4963 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4967 s = scan_const(PL_bufptr);
4969 PL_lex_state = LEX_INTERPCASEMOD;
4971 PL_lex_state = LEX_INTERPSTART;
4974 if (s != PL_bufptr) {
4975 start_force(PL_curforce);
4977 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4979 NEXTVAL_NEXTTOKE = pl_yylval;
4982 if (PL_lex_starts++) {
4986 sv_free(PL_thistoken);
4987 PL_thistoken = newSVpvs("");
4990 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4991 if (!PL_lex_casemods && PL_lex_inpat)
5004 s = scan_formline(PL_bufptr);
5005 if (!PL_lex_formbrack)
5015 PL_oldoldbufptr = PL_oldbufptr;
5021 sv_free(PL_thistoken);
5024 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
5028 if (UTF ? isIDFIRST_utf8((U8*)s) : isALNUMC(*s))
5031 SV *dsv = newSVpvs_flags("", SVs_TEMP);
5032 const char *c = UTF ? savepv(sv_uni_display(dsv, newSVpvn_flags(s,
5034 SVs_TEMP | SVf_UTF8),
5035 10, UNI_DISPLAY_ISPRINT))
5036 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
5037 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
5038 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
5039 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
5047 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"SVf"<-- HERE near column %d", c, SVfARG(dsv), (int) len + 1);
5051 goto fake_eof; /* emulate EOF on ^D or ^Z */
5057 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
5060 if (PL_lex_brackets &&
5061 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
5062 yyerror((const char *)
5064 ? "Format not terminated"
5065 : "Missing right curly or square bracket"));
5067 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5068 "### Tokener got EOF\n");
5072 if (s++ < PL_bufend)
5073 goto retry; /* ignore stray nulls */
5076 if (!PL_in_eval && !PL_preambled) {
5077 PL_preambled = TRUE;
5083 /* Generate a string of Perl code to load the debugger.
5084 * If PERL5DB is set, it will return the contents of that,
5085 * otherwise a compile-time require of perl5db.pl. */
5087 const char * const pdb = PerlEnv_getenv("PERL5DB");
5090 sv_setpv(PL_linestr, pdb);
5091 sv_catpvs(PL_linestr,";");
5093 SETERRNO(0,SS_NORMAL);
5094 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
5097 sv_setpvs(PL_linestr,"");
5098 if (PL_preambleav) {
5099 SV **svp = AvARRAY(PL_preambleav);
5100 SV **const end = svp + AvFILLp(PL_preambleav);
5102 sv_catsv(PL_linestr, *svp);
5104 sv_catpvs(PL_linestr, ";");
5106 sv_free(MUTABLE_SV(PL_preambleav));
5107 PL_preambleav = NULL;
5110 sv_catpvs(PL_linestr,
5111 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
5112 if (PL_minus_n || PL_minus_p) {
5113 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
5115 sv_catpvs(PL_linestr,"chomp;");
5118 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
5119 || *PL_splitstr == '"')
5120 && strchr(PL_splitstr + 1, *PL_splitstr))
5121 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
5123 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
5124 bytes can be used as quoting characters. :-) */
5125 const char *splits = PL_splitstr;
5126 sv_catpvs(PL_linestr, "our @F=split(q\0");
5129 if (*splits == '\\')
5130 sv_catpvn(PL_linestr, splits, 1);
5131 sv_catpvn(PL_linestr, splits, 1);
5132 } while (*splits++);
5133 /* This loop will embed the trailing NUL of
5134 PL_linestr as the last thing it does before
5136 sv_catpvs(PL_linestr, ");");
5140 sv_catpvs(PL_linestr,"our @F=split(' ');");
5143 sv_catpvs(PL_linestr, "\n");
5144 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5145 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5146 PL_last_lop = PL_last_uni = NULL;
5147 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
5148 update_debugger_info(PL_linestr, NULL, 0);
5153 bof = PL_rsfp ? TRUE : FALSE;
5156 fake_eof = LEX_FAKE_EOF;
5158 PL_bufptr = PL_bufend;
5159 COPLINE_INC_WITH_HERELINES;
5160 if (!lex_next_chunk(fake_eof)) {
5161 CopLINE_dec(PL_curcop);
5163 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
5165 CopLINE_dec(PL_curcop);
5168 PL_realtokenstart = -1;
5171 /* If it looks like the start of a BOM or raw UTF-16,
5172 * check if it in fact is. */
5173 if (bof && PL_rsfp &&
5178 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
5179 bof = (offset == (Off_t)SvCUR(PL_linestr));
5180 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
5181 /* offset may include swallowed CR */
5183 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
5186 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5187 s = swallow_bom((U8*)s);
5190 if (PL_parser->in_pod) {
5191 /* Incest with pod. */
5194 sv_catsv(PL_thiswhite, PL_linestr);
5196 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
5197 sv_setpvs(PL_linestr, "");
5198 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5199 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5200 PL_last_lop = PL_last_uni = NULL;
5201 PL_parser->in_pod = 0;
5204 if (PL_rsfp || PL_parser->filtered)
5206 } while (PL_parser->in_pod);
5207 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
5208 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5209 PL_last_lop = PL_last_uni = NULL;
5210 if (CopLINE(PL_curcop) == 1) {
5211 while (s < PL_bufend && isSPACE(*s))
5213 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
5217 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
5221 if (*s == '#' && *(s+1) == '!')
5223 #ifdef ALTERNATE_SHEBANG
5225 static char const as[] = ALTERNATE_SHEBANG;
5226 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5227 d = s + (sizeof(as) - 1);
5229 #endif /* ALTERNATE_SHEBANG */
5238 while (*d && !isSPACE(*d))
5242 #ifdef ARG_ZERO_IS_SCRIPT
5243 if (ipathend > ipath) {
5245 * HP-UX (at least) sets argv[0] to the script name,
5246 * which makes $^X incorrect. And Digital UNIX and Linux,
5247 * at least, set argv[0] to the basename of the Perl
5248 * interpreter. So, having found "#!", we'll set it right.
5250 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5252 assert(SvPOK(x) || SvGMAGICAL(x));
5253 if (sv_eq(x, CopFILESV(PL_curcop))) {
5254 sv_setpvn(x, ipath, ipathend - ipath);
5260 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
5261 const char * const lstart = SvPV_const(x,llen);
5263 bstart += blen - llen;
5264 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5265 sv_setpvn(x, ipath, ipathend - ipath);
5270 TAINT_NOT; /* $^X is always tainted, but that's OK */
5272 #endif /* ARG_ZERO_IS_SCRIPT */
5277 d = instr(s,"perl -");
5279 d = instr(s,"perl");
5281 /* avoid getting into infinite loops when shebang
5282 * line contains "Perl" rather than "perl" */
5284 for (d = ipathend-4; d >= ipath; --d) {
5285 if ((*d == 'p' || *d == 'P')
5286 && !ibcmp(d, "perl", 4))
5296 #ifdef ALTERNATE_SHEBANG
5298 * If the ALTERNATE_SHEBANG on this system starts with a
5299 * character that can be part of a Perl expression, then if
5300 * we see it but not "perl", we're probably looking at the
5301 * start of Perl code, not a request to hand off to some
5302 * other interpreter. Similarly, if "perl" is there, but
5303 * not in the first 'word' of the line, we assume the line
5304 * contains the start of the Perl program.
5306 if (d && *s != '#') {
5307 const char *c = ipath;
5308 while (*c && !strchr("; \t\r\n\f\v#", *c))
5311 d = NULL; /* "perl" not in first word; ignore */
5313 *s = '#'; /* Don't try to parse shebang line */
5315 #endif /* ALTERNATE_SHEBANG */
5320 !instr(s,"indir") &&
5321 instr(PL_origargv[0],"perl"))
5328 while (s < PL_bufend && isSPACE(*s))
5330 if (s < PL_bufend) {
5331 Newx(newargv,PL_origargc+3,char*);
5333 while (s < PL_bufend && !isSPACE(*s))
5336 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5339 newargv = PL_origargv;
5342 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5344 Perl_croak(aTHX_ "Can't exec %s", ipath);
5347 while (*d && !isSPACE(*d))
5349 while (SPACE_OR_TAB(*d))
5353 const bool switches_done = PL_doswitches;
5354 const U32 oldpdb = PL_perldb;
5355 const bool oldn = PL_minus_n;
5356 const bool oldp = PL_minus_p;
5360 bool baduni = FALSE;
5362 const char *d2 = d1 + 1;
5363 if (parse_unicode_opts((const char **)&d2)
5367 if (baduni || *d1 == 'M' || *d1 == 'm') {
5368 const char * const m = d1;
5369 while (*d1 && !isSPACE(*d1))
5371 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5374 d1 = moreswitches(d1);
5376 if (PL_doswitches && !switches_done) {
5377 int argc = PL_origargc;
5378 char **argv = PL_origargv;
5381 } while (argc && argv[0][0] == '-' && argv[0][1]);
5382 init_argv_symbols(argc,argv);
5384 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5385 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5386 /* if we have already added "LINE: while (<>) {",
5387 we must not do it again */
5389 sv_setpvs(PL_linestr, "");
5390 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5391 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5392 PL_last_lop = PL_last_uni = NULL;
5393 PL_preambled = FALSE;
5394 if (PERLDB_LINE || PERLDB_SAVESRC)
5395 (void)gv_fetchfile(PL_origfilename);
5402 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5403 PL_lex_state = LEX_FORMLINE;
5404 start_force(PL_curforce);
5405 NEXTVAL_NEXTTOKE.ival = 0;
5406 force_next(FORMRBRACK);
5411 #ifdef PERL_STRICT_CR
5412 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5414 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5416 case ' ': case '\t': case '\f': case 013:
5418 PL_realtokenstart = -1;
5421 PL_thiswhite = newSVpvs("");
5422 sv_catpvn(PL_thiswhite, s, 1);
5430 PL_realtokenstart = -1;
5434 if (PL_lex_state != LEX_NORMAL ||
5435 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5436 if (*s == '#' && s == PL_linestart && PL_in_eval
5437 && !PL_rsfp && !PL_parser->filtered) {
5438 /* handle eval qq[#line 1 "foo"\n ...] */
5439 CopLINE_dec(PL_curcop);
5442 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5444 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5448 const bool in_comment = *s == '#';
5450 while (d < PL_bufend && *d != '\n')
5454 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5455 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5459 PL_thiswhite = newSVpvn(s, d - s);
5462 if (in_comment && d == PL_bufend
5463 && PL_lex_state == LEX_INTERPNORMAL
5464 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5465 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5468 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5469 PL_lex_state = LEX_FORMLINE;
5470 start_force(PL_curforce);
5471 NEXTVAL_NEXTTOKE.ival = 0;
5472 force_next(FORMRBRACK);
5478 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5479 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5482 TOKEN(PEG); /* make sure any #! line is accessible */
5487 /* if (PL_madskills && PL_lex_formbrack) { */
5489 while (d < PL_bufend && *d != '\n')
5493 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5494 Perl_croak(aTHX_ "panic: input overflow");
5495 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5497 PL_thiswhite = newSVpvs("");
5498 if (CopLINE(PL_curcop) == 1) {
5499 sv_setpvs(PL_thiswhite, "");
5502 sv_catpvn(PL_thiswhite, s, d - s);
5516 if (s[1] && isALPHA(s[1]) && !isWORDCHAR(s[2])) {
5524 while (s < PL_bufend && SPACE_OR_TAB(*s))
5527 if (strnEQ(s,"=>",2)) {
5528 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5529 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5530 OPERATOR('-'); /* unary minus */
5532 PL_last_uni = PL_oldbufptr;
5534 case 'r': ftst = OP_FTEREAD; break;
5535 case 'w': ftst = OP_FTEWRITE; break;
5536 case 'x': ftst = OP_FTEEXEC; break;
5537 case 'o': ftst = OP_FTEOWNED; break;
5538 case 'R': ftst = OP_FTRREAD; break;
5539 case 'W': ftst = OP_FTRWRITE; break;
5540 case 'X': ftst = OP_FTREXEC; break;
5541 case 'O': ftst = OP_FTROWNED; break;
5542 case 'e': ftst = OP_FTIS; break;
5543 case 'z': ftst = OP_FTZERO; break;
5544 case 's': ftst = OP_FTSIZE; break;
5545 case 'f': ftst = OP_FTFILE; break;
5546 case 'd': ftst = OP_FTDIR; break;
5547 case 'l': ftst = OP_FTLINK; break;
5548 case 'p': ftst = OP_FTPIPE; break;
5549 case 'S': ftst = OP_FTSOCK; break;
5550 case 'u': ftst = OP_FTSUID; break;
5551 case 'g': ftst = OP_FTSGID; break;
5552 case 'k': ftst = OP_FTSVTX; break;
5553 case 'b': ftst = OP_FTBLK; break;
5554 case 'c': ftst = OP_FTCHR; break;
5555 case 't': ftst = OP_FTTTY; break;
5556 case 'T': ftst = OP_FTTEXT; break;
5557 case 'B': ftst = OP_FTBINARY; break;
5558 case 'M': case 'A': case 'C':
5559 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5561 case 'M': ftst = OP_FTMTIME; break;
5562 case 'A': ftst = OP_FTATIME; break;
5563 case 'C': ftst = OP_FTCTIME; break;
5571 PL_last_lop_op = (OPCODE)ftst;
5572 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5573 "### Saw file test %c\n", (int)tmp);
5578 /* Assume it was a minus followed by a one-letter named
5579 * subroutine call (or a -bareword), then. */
5580 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5581 "### '-%c' looked like a file test but was not\n",
5588 const char tmp = *s++;
5591 if (PL_expect == XOPERATOR)
5596 else if (*s == '>') {
5599 if (isIDFIRST_lazy_if(s,UTF)) {
5600 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5608 if (PL_expect == XOPERATOR) {
5609 if (*s == '=' && !PL_lex_allbrackets &&
5610 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5617 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5619 OPERATOR('-'); /* unary minus */
5625 const char tmp = *s++;
5628 if (PL_expect == XOPERATOR)
5633 if (PL_expect == XOPERATOR) {
5634 if (*s == '=' && !PL_lex_allbrackets &&
5635 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5642 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5649 if (PL_expect != XOPERATOR) {
5650 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5651 PL_expect = XOPERATOR;
5652 force_ident(PL_tokenbuf, '*');
5660 if (*s == '=' && !PL_lex_allbrackets &&
5661 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5667 if (*s == '=' && !PL_lex_allbrackets &&
5668 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5675 if (PL_expect == XOPERATOR) {
5676 if (s[1] == '=' && !PL_lex_allbrackets &&
5677 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5682 PL_tokenbuf[0] = '%';
5683 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5684 sizeof PL_tokenbuf - 1, FALSE);
5685 if (!PL_tokenbuf[1]) {
5688 PL_expect = XOPERATOR;
5689 force_ident_maybe_lex('%');
5693 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5694 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5699 if (PL_lex_brackets > 100)
5700 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5701 PL_lex_brackstack[PL_lex_brackets++] = 0;
5702 PL_lex_allbrackets++;
5704 const char tmp = *s++;
5709 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5711 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5719 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5726 goto just_a_word_zero_gv;
5729 switch (PL_expect) {
5735 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5737 PL_bufptr = s; /* update in case we back off */
5740 "Use of := for an empty attribute list is not allowed");
5747 PL_expect = XTERMBLOCK;
5750 stuffstart = s - SvPVX(PL_linestr) - 1;
5754 while (isIDFIRST_lazy_if(s,UTF)) {
5757 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5758 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5759 if (tmp < 0) tmp = -tmp;
5774 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5776 d = scan_str(d,TRUE,TRUE,FALSE, FALSE);
5778 /* MUST advance bufptr here to avoid bogus
5779 "at end of line" context messages from yyerror().
5781 PL_bufptr = s + len;
5782 yyerror("Unterminated attribute parameter in attribute list");
5786 return REPORT(0); /* EOF indicator */
5790 sv_catsv(sv, PL_lex_stuff);
5791 attrs = op_append_elem(OP_LIST, attrs,
5792 newSVOP(OP_CONST, 0, sv));
5793 SvREFCNT_dec(PL_lex_stuff);
5794 PL_lex_stuff = NULL;
5797 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5799 if (PL_in_my == KEY_our) {
5800 deprecate(":unique");
5803 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5806 /* NOTE: any CV attrs applied here need to be part of
5807 the CVf_BUILTIN_ATTRS define in cv.h! */
5808 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5810 CvLVALUE_on(PL_compcv);
5812 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5814 deprecate(":locked");
5816 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5818 CvMETHOD_on(PL_compcv);
5820 /* After we've set the flags, it could be argued that
5821 we don't need to do the attributes.pm-based setting
5822 process, and shouldn't bother appending recognized
5823 flags. To experiment with that, uncomment the
5824 following "else". (Note that's already been
5825 uncommented. That keeps the above-applied built-in
5826 attributes from being intercepted (and possibly
5827 rejected) by a package's attribute routines, but is
5828 justified by the performance win for the common case
5829 of applying only built-in attributes.) */
5831 attrs = op_append_elem(OP_LIST, attrs,
5832 newSVOP(OP_CONST, 0,
5836 if (*s == ':' && s[1] != ':')
5839 break; /* require real whitespace or :'s */
5840 /* XXX losing whitespace on sequential attributes here */
5844 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5845 if (*s != ';' && *s != '}' && *s != tmp
5846 && (tmp != '=' || *s != ')')) {
5847 const char q = ((*s == '\'') ? '"' : '\'');
5848 /* If here for an expression, and parsed no attrs, back
5850 if (tmp == '=' && !attrs) {
5854 /* MUST advance bufptr here to avoid bogus "at end of line"
5855 context messages from yyerror().
5858 yyerror( (const char *)
5860 ? Perl_form(aTHX_ "Invalid separator character "
5861 "%c%c%c in attribute list", q, *s, q)
5862 : "Unterminated attribute list" ) );
5870 start_force(PL_curforce);
5871 NEXTVAL_NEXTTOKE.opval = attrs;
5872 CURMAD('_', PL_nextwhite);
5877 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5878 (s - SvPVX(PL_linestr)) - stuffstart);
5883 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5887 PL_lex_allbrackets--;
5891 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5892 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5896 PL_lex_allbrackets++;
5899 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5905 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5908 PL_lex_allbrackets--;
5914 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5917 if (PL_lex_brackets <= 0)
5918 yyerror("Unmatched right square bracket");
5921 PL_lex_allbrackets--;
5922 if (PL_lex_state == LEX_INTERPNORMAL) {
5923 if (PL_lex_brackets == 0) {
5924 if (*s == '-' && s[1] == '>')
5925 PL_lex_state = LEX_INTERPENDMAYBE;
5926 else if (*s != '[' && *s != '{')
5927 PL_lex_state = LEX_INTERPEND;
5934 if (PL_lex_brackets > 100) {
5935 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5937 switch (PL_expect) {
5939 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5940 PL_lex_allbrackets++;
5941 OPERATOR(HASHBRACK);
5943 while (s < PL_bufend && SPACE_OR_TAB(*s))
5946 PL_tokenbuf[0] = '\0';
5947 if (d < PL_bufend && *d == '-') {
5948 PL_tokenbuf[0] = '-';
5950 while (d < PL_bufend && SPACE_OR_TAB(*d))
5953 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5954 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5956 while (d < PL_bufend && SPACE_OR_TAB(*d))
5959 const char minus = (PL_tokenbuf[0] == '-');
5960 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5968 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5969 PL_lex_allbrackets++;
5974 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5975 PL_lex_allbrackets++;
5980 if (PL_oldoldbufptr == PL_last_lop)
5981 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5983 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5984 PL_lex_allbrackets++;
5987 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5989 /* This hack is to get the ${} in the message. */
5991 yyerror("syntax error");
5994 OPERATOR(HASHBRACK);
5996 /* This hack serves to disambiguate a pair of curlies
5997 * as being a block or an anon hash. Normally, expectation
5998 * determines that, but in cases where we're not in a
5999 * position to expect anything in particular (like inside
6000 * eval"") we have to resolve the ambiguity. This code
6001 * covers the case where the first term in the curlies is a
6002 * quoted string. Most other cases need to be explicitly
6003 * disambiguated by prepending a "+" before the opening
6004 * curly in order to force resolution as an anon hash.
6006 * XXX should probably propagate the outer expectation
6007 * into eval"" to rely less on this hack, but that could
6008 * potentially break current behavior of eval"".
6012 if (*s == '\'' || *s == '"' || *s == '`') {
6013 /* common case: get past first string, handling escapes */
6014 for (t++; t < PL_bufend && *t != *s;)
6015 if (*t++ == '\\' && (*t == '\\' || *t == *s))
6019 else if (*s == 'q') {
6022 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
6023 && !isWORDCHAR(*t))))
6025 /* skip q//-like construct */
6027 char open, close, term;
6030 while (t < PL_bufend && isSPACE(*t))
6032 /* check for q => */
6033 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
6034 OPERATOR(HASHBRACK);
6038 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6042 for (t++; t < PL_bufend; t++) {
6043 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
6045 else if (*t == open)
6049 for (t++; t < PL_bufend; t++) {
6050 if (*t == '\\' && t+1 < PL_bufend)
6052 else if (*t == close && --brackets <= 0)
6054 else if (*t == open)
6061 /* skip plain q word */
6062 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
6065 else if (isWORDCHAR_lazy_if(t,UTF)) {
6067 while (t < PL_bufend && isWORDCHAR_lazy_if(t,UTF))
6070 while (t < PL_bufend && isSPACE(*t))
6072 /* if comma follows first term, call it an anon hash */
6073 /* XXX it could be a comma expression with loop modifiers */
6074 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
6075 || (*t == '=' && t[1] == '>')))
6076 OPERATOR(HASHBRACK);
6077 if (PL_expect == XREF)
6080 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
6086 pl_yylval.ival = CopLINE(PL_curcop);
6087 if (isSPACE(*s) || *s == '#')
6088 PL_copline = NOLINE; /* invalidate current command line number */
6089 TOKEN(formbrack ? '=' : '{');
6091 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
6095 if (PL_lex_brackets <= 0)
6096 yyerror("Unmatched right curly bracket");
6098 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
6099 PL_lex_allbrackets--;
6100 if (PL_lex_state == LEX_INTERPNORMAL) {
6101 if (PL_lex_brackets == 0) {
6102 if (PL_expect & XFAKEBRACK) {
6103 PL_expect &= XENUMMASK;
6104 PL_lex_state = LEX_INTERPEND;
6109 PL_thiswhite = newSVpvs("");
6110 sv_catpvs(PL_thiswhite,"}");
6113 return yylex(); /* ignore fake brackets */
6115 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6116 && SvEVALED(PL_lex_repl))
6117 PL_lex_state = LEX_INTERPEND;
6118 else if (*s == '-' && s[1] == '>')
6119 PL_lex_state = LEX_INTERPENDMAYBE;
6120 else if (*s != '[' && *s != '{')
6121 PL_lex_state = LEX_INTERPEND;
6124 if (PL_expect & XFAKEBRACK) {
6125 PL_expect &= XENUMMASK;
6127 return yylex(); /* ignore fake brackets */
6129 start_force(PL_curforce);
6131 curmad('X', newSVpvn(s-1,1));
6132 CURMAD('_', PL_thiswhite);
6134 force_next(formbrack ? '.' : '}');
6135 if (formbrack) LEAVE;
6137 if (PL_madskills && !PL_thistoken)
6138 PL_thistoken = newSVpvs("");
6140 if (formbrack == 2) { /* means . where arguments were expected */
6141 start_force(PL_curforce);
6149 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6150 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6157 if (PL_expect == XOPERATOR) {
6158 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
6159 && isIDFIRST_lazy_if(s,UTF))
6161 CopLINE_dec(PL_curcop);
6162 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6163 CopLINE_inc(PL_curcop);
6165 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6166 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6173 PL_tokenbuf[0] = '&';
6174 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf + 1,
6175 sizeof PL_tokenbuf - 1, TRUE);
6176 if (PL_tokenbuf[1]) {
6177 PL_expect = XOPERATOR;
6178 force_ident_maybe_lex('&');
6182 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
6188 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6189 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6196 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6197 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6205 const char tmp = *s++;
6207 if (!PL_lex_allbrackets &&
6208 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6215 if (!PL_lex_allbrackets &&
6216 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
6224 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6225 && strchr("+-*/%.^&|<",tmp))
6226 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6227 "Reversed %c= operator",(int)tmp);
6229 if (PL_expect == XSTATE && isALPHA(tmp) &&
6230 (s == PL_linestart+1 || s[-2] == '\n') )
6232 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6233 || PL_lex_state != LEX_NORMAL) {
6238 if (strnEQ(s,"=cut",4)) {
6254 PL_thiswhite = newSVpvs("");
6255 sv_catpvn(PL_thiswhite, PL_linestart,
6256 PL_bufend - PL_linestart);
6260 PL_parser->in_pod = 1;
6264 if (PL_expect == XBLOCK) {
6266 #ifdef PERL_STRICT_CR
6267 while (SPACE_OR_TAB(*t))
6269 while (SPACE_OR_TAB(*t) || *t == '\r')
6272 if (*t == '\n' || *t == '#') {
6275 SAVEI8(PL_parser->form_lex_state);
6276 SAVEI32(PL_lex_formbrack);
6277 PL_parser->form_lex_state = PL_lex_state;
6278 PL_lex_formbrack = PL_lex_brackets + 1;
6282 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6291 const char tmp = *s++;
6293 /* was this !=~ where !~ was meant?
6294 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6296 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6297 const char *t = s+1;
6299 while (t < PL_bufend && isSPACE(*t))
6302 if (*t == '/' || *t == '?' ||
6303 ((*t == 'm' || *t == 's' || *t == 'y')
6304 && !isWORDCHAR(t[1])) ||
6305 (*t == 't' && t[1] == 'r' && !isWORDCHAR(t[2])))
6306 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6307 "!=~ should be !~");
6309 if (!PL_lex_allbrackets &&
6310 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6322 if (PL_expect != XOPERATOR) {
6323 if (s[1] != '<' && !strchr(s,'>'))
6326 s = scan_heredoc(s);
6328 s = scan_inputsymbol(s);
6329 PL_expect = XOPERATOR;
6330 TOKEN(sublex_start());
6336 if (*s == '=' && !PL_lex_allbrackets &&
6337 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6341 SHop(OP_LEFT_SHIFT);
6346 if (!PL_lex_allbrackets &&
6347 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6354 if (!PL_lex_allbrackets &&
6355 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6363 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6371 const char tmp = *s++;
6373 if (*s == '=' && !PL_lex_allbrackets &&
6374 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6378 SHop(OP_RIGHT_SHIFT);
6380 else if (tmp == '=') {
6381 if (!PL_lex_allbrackets &&
6382 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6390 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6399 if (PL_expect == XOPERATOR) {
6400 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6401 return deprecate_commaless_var_list();
6405 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6406 PL_tokenbuf[0] = '@';
6407 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6408 sizeof PL_tokenbuf - 1, FALSE);
6409 if (PL_expect == XOPERATOR)
6410 no_op("Array length", s);
6411 if (!PL_tokenbuf[1])
6413 PL_expect = XOPERATOR;
6414 force_ident_maybe_lex('#');
6418 PL_tokenbuf[0] = '$';
6419 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6420 sizeof PL_tokenbuf - 1, FALSE);
6421 if (PL_expect == XOPERATOR)
6423 if (!PL_tokenbuf[1]) {
6425 yyerror("Final $ should be \\$ or $name");
6431 const char tmp = *s;
6432 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6435 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6436 && intuit_more(s)) {
6438 PL_tokenbuf[0] = '@';
6439 if (ckWARN(WARN_SYNTAX)) {
6442 while (isSPACE(*t) || isWORDCHAR_lazy_if(t,UTF) || *t == '$')
6445 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6446 while (t < PL_bufend && *t != ']')
6448 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6449 "Multidimensional syntax %.*s not supported",
6450 (int)((t - PL_bufptr) + 1), PL_bufptr);
6454 else if (*s == '{') {
6456 PL_tokenbuf[0] = '%';
6457 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6458 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6460 char tmpbuf[sizeof PL_tokenbuf];
6463 } while (isSPACE(*t));
6464 if (isIDFIRST_lazy_if(t,UTF)) {
6466 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6471 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6472 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6473 "You need to quote \"%"SVf"\"",
6474 SVfARG(newSVpvn_flags(tmpbuf, len,
6475 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
6481 PL_expect = XOPERATOR;
6482 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6483 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6484 if (!islop || PL_last_lop_op == OP_GREPSTART)
6485 PL_expect = XOPERATOR;
6486 else if (strchr("$@\"'`q", *s))
6487 PL_expect = XTERM; /* e.g. print $fh "foo" */
6488 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6489 PL_expect = XTERM; /* e.g. print $fh &sub */
6490 else if (isIDFIRST_lazy_if(s,UTF)) {
6491 char tmpbuf[sizeof PL_tokenbuf];
6493 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6494 if ((t2 = keyword(tmpbuf, len, 0))) {
6495 /* binary operators exclude handle interpretations */
6507 PL_expect = XTERM; /* e.g. print $fh length() */
6512 PL_expect = XTERM; /* e.g. print $fh subr() */
6515 else if (isDIGIT(*s))
6516 PL_expect = XTERM; /* e.g. print $fh 3 */
6517 else if (*s == '.' && isDIGIT(s[1]))
6518 PL_expect = XTERM; /* e.g. print $fh .3 */
6519 else if ((*s == '?' || *s == '-' || *s == '+')
6520 && !isSPACE(s[1]) && s[1] != '=')
6521 PL_expect = XTERM; /* e.g. print $fh -1 */
6522 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6524 PL_expect = XTERM; /* e.g. print $fh /.../
6525 XXX except DORDOR operator
6527 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6529 PL_expect = XTERM; /* print $fh <<"EOF" */
6532 force_ident_maybe_lex('$');
6536 if (PL_expect == XOPERATOR)
6538 PL_tokenbuf[0] = '@';
6539 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6540 if (!PL_tokenbuf[1]) {
6543 if (PL_lex_state == LEX_NORMAL)
6545 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6547 PL_tokenbuf[0] = '%';
6549 /* Warn about @ where they meant $. */
6550 if (*s == '[' || *s == '{') {
6551 if (ckWARN(WARN_SYNTAX)) {
6552 const char *t = s + 1;
6553 while (*t && (isWORDCHAR_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6554 t += UTF ? UTF8SKIP(t) : 1;
6555 if (*t == '}' || *t == ']') {
6557 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6558 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6559 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6560 "Scalar value %"SVf" better written as $%"SVf,
6561 SVfARG(newSVpvn_flags(PL_bufptr, (STRLEN)(t-PL_bufptr),
6562 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))),
6563 SVfARG(newSVpvn_flags(PL_bufptr+1, (STRLEN)(t-PL_bufptr-1),
6564 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))));
6569 PL_expect = XOPERATOR;
6570 force_ident_maybe_lex('@');
6573 case '/': /* may be division, defined-or, or pattern */
6574 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6575 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6576 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6581 case '?': /* may either be conditional or pattern */
6582 if (PL_expect == XOPERATOR) {
6585 if (!PL_lex_allbrackets &&
6586 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6590 PL_lex_allbrackets++;
6596 /* A // operator. */
6597 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6598 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6599 LEX_FAKEEOF_LOGIC)) {
6607 if (*s == '=' && !PL_lex_allbrackets &&
6608 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6617 /* Disable warning on "study /blah/" */
6618 if (PL_oldoldbufptr == PL_last_uni
6619 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6620 || memNE(PL_last_uni, "study", 5)
6621 || isWORDCHAR_lazy_if(PL_last_uni+5,UTF)
6625 deprecate("?PATTERN? without explicit operator");
6626 s = scan_pat(s,OP_MATCH);
6627 TERM(sublex_start());
6631 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6632 #ifdef PERL_STRICT_CR
6635 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6637 && (s == PL_linestart || s[-1] == '\n') )
6640 formbrack = 2; /* dot seen where arguments expected */
6643 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6647 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6650 if (!PL_lex_allbrackets &&
6651 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6658 pl_yylval.ival = OPf_SPECIAL;
6664 if (*s == '=' && !PL_lex_allbrackets &&
6665 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6672 case '0': case '1': case '2': case '3': case '4':
6673 case '5': case '6': case '7': case '8': case '9':
6674 s = scan_num(s, &pl_yylval);
6675 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6676 if (PL_expect == XOPERATOR)
6681 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
6682 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6683 if (PL_expect == XOPERATOR) {
6684 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6685 return deprecate_commaless_var_list();
6692 pl_yylval.ival = OP_CONST;
6693 TERM(sublex_start());
6696 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
6697 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6698 if (PL_expect == XOPERATOR) {
6699 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6700 return deprecate_commaless_var_list();
6707 pl_yylval.ival = OP_CONST;
6708 /* FIXME. I think that this can be const if char *d is replaced by
6709 more localised variables. */
6710 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6711 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6712 pl_yylval.ival = OP_STRINGIFY;
6716 TERM(sublex_start());
6719 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
6720 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6721 if (PL_expect == XOPERATOR)
6722 no_op("Backticks",s);
6725 readpipe_override();
6726 TERM(sublex_start());
6730 if (PL_lex_inwhat && isDIGIT(*s))
6731 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6733 if (PL_expect == XOPERATOR)
6734 no_op("Backslash",s);
6738 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6739 char *start = s + 2;
6740 while (isDIGIT(*start) || *start == '_')
6742 if (*start == '.' && isDIGIT(start[1])) {
6743 s = scan_num(s, &pl_yylval);
6746 else if ((*start == ':' && start[1] == ':')
6747 || (PL_expect == XSTATE && *start == ':'))
6749 else if (PL_expect == XSTATE) {
6751 while (d < PL_bufend && isSPACE(*d)) d++;
6752 if (*d == ':') goto keylookup;
6754 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6755 if (!isALPHA(*start) && (PL_expect == XTERM
6756 || PL_expect == XREF || PL_expect == XSTATE
6757 || PL_expect == XTERMORDORDOR)) {
6758 GV *const gv = gv_fetchpvn_flags(s, start - s,
6759 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6761 s = scan_num(s, &pl_yylval);
6768 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6821 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6823 /* Some keywords can be followed by any delimiter, including ':' */
6824 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6826 /* x::* is just a word, unless x is "CORE" */
6827 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6831 while (d < PL_bufend && isSPACE(*d))
6832 d++; /* no comments skipped here, or s### is misparsed */
6834 /* Is this a word before a => operator? */
6835 if (*d == '=' && d[1] == '>') {
6838 = (OP*)newSVOP(OP_CONST, 0,
6839 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6840 pl_yylval.opval->op_private = OPpCONST_BARE;
6844 /* Check for plugged-in keyword */
6848 char *saved_bufptr = PL_bufptr;
6850 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6852 if (result == KEYWORD_PLUGIN_DECLINE) {
6853 /* not a plugged-in keyword */
6854 PL_bufptr = saved_bufptr;
6855 } else if (result == KEYWORD_PLUGIN_STMT) {
6856 pl_yylval.opval = o;
6859 return REPORT(PLUGSTMT);
6860 } else if (result == KEYWORD_PLUGIN_EXPR) {
6861 pl_yylval.opval = o;
6863 PL_expect = XOPERATOR;
6864 return REPORT(PLUGEXPR);
6866 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6871 /* Check for built-in keyword */
6872 tmp = keyword(PL_tokenbuf, len, 0);
6874 /* Is this a label? */
6875 if (!anydelim && PL_expect == XSTATE
6876 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6878 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6879 pl_yylval.pval[len] = '\0';
6880 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6885 /* Check for lexical sub */
6886 if (PL_expect != XOPERATOR) {
6887 char tmpbuf[sizeof PL_tokenbuf + 1];
6889 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6890 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6891 if (off != NOT_IN_PAD) {
6892 assert(off); /* we assume this is boolean-true below */
6893 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6894 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6895 HEK * const stashname = HvNAME_HEK(stash);
6896 sv = newSVhek(stashname);
6897 sv_catpvs(sv, "::");
6898 sv_catpvn_flags(sv, PL_tokenbuf, len,
6899 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6900 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6910 rv2cv_op = newOP(OP_PADANY, 0);
6911 rv2cv_op->op_targ = off;
6912 rv2cv_op = (OP*)newCVREF(0, rv2cv_op);
6913 cv = (CV *)PAD_SV(off);
6921 if (tmp < 0) { /* second-class keyword? */
6922 GV *ogv = NULL; /* override (winner) */
6923 GV *hgv = NULL; /* hidden (loser) */
6924 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6926 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6927 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6930 if (GvIMPORTED_CV(gv))
6932 else if (! CvMETHOD(cv))
6936 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6937 UTF ? -(I32)len : (I32)len, FALSE)) &&
6938 (gv = *gvp) && isGV_with_GP(gv) &&
6939 GvCVu(gv) && GvIMPORTED_CV(gv))
6946 tmp = 0; /* overridden by import or by GLOBAL */
6949 && -tmp==KEY_lock /* XXX generalizable kludge */
6952 tmp = 0; /* any sub overrides "weak" keyword */
6954 else { /* no override */
6956 if (tmp == KEY_dump) {
6957 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6958 "dump() better written as CORE::dump()");
6962 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6963 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6964 "Ambiguous call resolved as CORE::%s(), "
6965 "qualify as such or use &",
6973 default: /* not a keyword */
6974 /* Trade off - by using this evil construction we can pull the
6975 variable gv into the block labelled keylookup. If not, then
6976 we have to give it function scope so that the goto from the
6977 earlier ':' case doesn't bypass the initialisation. */
6979 just_a_word_zero_gv:
6991 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6992 const char penultchar =
6993 lastchar && PL_bufptr - 2 >= PL_linestart
6997 SV *nextPL_nextwhite = 0;
7001 /* Get the rest if it looks like a package qualifier */
7003 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
7005 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
7008 Perl_croak(aTHX_ "Bad name after %"SVf"%s",
7009 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7010 (UTF ? SVf_UTF8 : 0) | SVs_TEMP )),
7011 *s == '\'' ? "'" : "::");
7016 if (PL_expect == XOPERATOR) {
7017 if (PL_bufptr == PL_linestart) {
7018 CopLINE_dec(PL_curcop);
7019 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
7020 CopLINE_inc(PL_curcop);
7023 no_op("Bareword",s);
7026 /* Look for a subroutine with this name in current package,
7027 unless this is a lexical sub, or name is "Foo::",
7028 in which case Foo is a bareword
7029 (and a package name). */
7031 if (len > 2 && !PL_madskills &&
7032 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
7034 if (ckWARN(WARN_BAREWORD)
7035 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
7036 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
7037 "Bareword \"%"SVf"\" refers to nonexistent package",
7038 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7039 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7041 PL_tokenbuf[len] = '\0';
7047 /* Mustn't actually add anything to a symbol table.
7048 But also don't want to "initialise" any placeholder
7049 constants that might already be there into full
7050 blown PVGVs with attached PVCV. */
7051 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
7052 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
7058 /* if we saw a global override before, get the right name */
7061 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
7062 len ? len : strlen(PL_tokenbuf));
7064 SV * const tmp_sv = sv;
7065 sv = newSVpvs("CORE::GLOBAL::");
7066 sv_catsv(sv, tmp_sv);
7067 SvREFCNT_dec(tmp_sv);
7071 if (PL_madskills && !PL_thistoken) {
7072 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
7073 PL_thistoken = newSVpvn(start,s - start);
7074 PL_realtokenstart = s - SvPVX(PL_linestr);
7078 /* Presume this is going to be a bareword of some sort. */
7080 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
7081 pl_yylval.opval->op_private = OPpCONST_BARE;
7083 /* And if "Foo::", then that's what it certainly is. */
7089 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
7090 const_op->op_private = OPpCONST_BARE;
7091 rv2cv_op = newCVREF(0, const_op);
7092 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
7095 /* See if it's the indirect object for a list operator. */
7097 if (PL_oldoldbufptr &&
7098 PL_oldoldbufptr < PL_bufptr &&
7099 (PL_oldoldbufptr == PL_last_lop
7100 || PL_oldoldbufptr == PL_last_uni) &&
7101 /* NO SKIPSPACE BEFORE HERE! */
7102 (PL_expect == XREF ||
7103 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
7105 bool immediate_paren = *s == '(';
7107 /* (Now we can afford to cross potential line boundary.) */
7108 s = SKIPSPACE2(s,nextPL_nextwhite);
7110 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
7113 /* Two barewords in a row may indicate method call. */
7115 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
7116 (tmp = intuit_method(s, gv, cv))) {
7118 if (tmp == METHOD && !PL_lex_allbrackets &&
7119 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7120 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7124 /* If not a declared subroutine, it's an indirect object. */
7125 /* (But it's an indir obj regardless for sort.) */
7126 /* Also, if "_" follows a filetest operator, it's a bareword */
7129 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
7131 (PL_last_lop_op != OP_MAPSTART &&
7132 PL_last_lop_op != OP_GREPSTART))))
7133 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
7134 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
7137 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
7142 PL_expect = XOPERATOR;
7145 s = SKIPSPACE2(s,nextPL_nextwhite);
7146 PL_nextwhite = nextPL_nextwhite;
7151 /* Is this a word before a => operator? */
7152 if (*s == '=' && s[1] == '>' && !pkgname) {
7155 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
7156 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
7157 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
7161 /* If followed by a paren, it's certainly a subroutine. */
7166 while (SPACE_OR_TAB(*d))
7168 if (*d == ')' && (sv = cv_const_sv(cv))) {
7175 PL_nextwhite = PL_thiswhite;
7178 start_force(PL_curforce);
7180 NEXTVAL_NEXTTOKE.opval =
7181 off ? rv2cv_op : pl_yylval.opval;
7182 PL_expect = XOPERATOR;
7185 PL_nextwhite = nextPL_nextwhite;
7186 curmad('X', PL_thistoken);
7187 PL_thistoken = newSVpvs("");
7191 op_free(pl_yylval.opval), force_next(PRIVATEREF);
7192 else op_free(rv2cv_op), force_next(WORD);
7197 /* If followed by var or block, call it a method (unless sub) */
7199 if ((*s == '$' || *s == '{') && !cv) {
7201 PL_last_lop = PL_oldbufptr;
7202 PL_last_lop_op = OP_METHOD;
7203 if (!PL_lex_allbrackets &&
7204 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7205 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7209 /* If followed by a bareword, see if it looks like indir obj. */
7212 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7213 && (tmp = intuit_method(s, gv, cv))) {
7215 if (tmp == METHOD && !PL_lex_allbrackets &&
7216 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7217 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7221 /* Not a method, so call it a subroutine (if defined) */
7224 if (lastchar == '-' && penultchar != '-') {
7225 const SV *tmpsv = newSVpvn_flags( PL_tokenbuf, len ? len : strlen(PL_tokenbuf), (UTF ? SVf_UTF8 : 0) | SVs_TEMP );
7226 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7227 "Ambiguous use of -%"SVf" resolved as -&%"SVf"()",
7228 SVfARG(tmpsv), SVfARG(tmpsv));
7230 /* Check for a constant sub */
7231 if ((sv = cv_const_sv(cv))) {
7234 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7235 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7236 pl_yylval.opval->op_private = OPpCONST_FOLDED;
7237 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7241 op_free(pl_yylval.opval);
7242 pl_yylval.opval = rv2cv_op;
7243 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7244 PL_last_lop = PL_oldbufptr;
7245 PL_last_lop_op = OP_ENTERSUB;
7246 /* Is there a prototype? */
7253 STRLEN protolen = CvPROTOLEN(cv);
7254 const char *proto = CvPROTO(cv);
7258 if ((optional = *proto == ';'))
7261 while (*proto == ';');
7265 *proto == '$' || *proto == '_'
7266 || *proto == '*' || *proto == '+'
7271 *proto == '\\' && proto[1] && proto[2] == '\0'
7274 UNIPROTO(UNIOPSUB,optional);
7275 if (*proto == '\\' && proto[1] == '[') {
7276 const char *p = proto + 2;
7277 while(*p && *p != ']')
7279 if(*p == ']' && !p[1])
7280 UNIPROTO(UNIOPSUB,optional);
7282 if (*proto == '&' && *s == '{') {
7284 sv_setpvs(PL_subname, "__ANON__");
7286 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7287 if (!PL_lex_allbrackets &&
7288 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7289 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7296 PL_nextwhite = PL_thiswhite;
7299 start_force(PL_curforce);
7300 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7303 PL_nextwhite = nextPL_nextwhite;
7304 curmad('X', PL_thistoken);
7305 PL_thistoken = newSVpvs("");
7307 force_next(off ? PRIVATEREF : WORD);
7308 if (!PL_lex_allbrackets &&
7309 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7310 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7315 /* Guess harder when madskills require "best effort". */
7316 if (PL_madskills && (!gv || !GvCVu(gv))) {
7317 int probable_sub = 0;
7318 if (strchr("\"'`$@%0123456789!*+{[<", *s))
7320 else if (isALPHA(*s)) {
7324 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
7325 if (!keyword(tmpbuf, tmplen, 0))
7328 while (d < PL_bufend && isSPACE(*d))
7330 if (*d == '=' && d[1] == '>')
7335 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
7337 op_free(pl_yylval.opval);
7338 pl_yylval.opval = rv2cv_op;
7339 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7340 PL_last_lop = PL_oldbufptr;
7341 PL_last_lop_op = OP_ENTERSUB;
7342 PL_nextwhite = PL_thiswhite;
7344 start_force(PL_curforce);
7345 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7347 PL_nextwhite = nextPL_nextwhite;
7348 curmad('X', PL_thistoken);
7349 PL_thistoken = newSVpvs("");
7350 force_next(off ? PRIVATEREF : WORD);
7351 if (!PL_lex_allbrackets &&
7352 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7353 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7357 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7359 force_next(off ? PRIVATEREF : WORD);
7360 if (!PL_lex_allbrackets &&
7361 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7362 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7367 /* Call it a bare word */
7369 if (PL_hints & HINT_STRICT_SUBS)
7370 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7373 /* after "print" and similar functions (corresponding to
7374 * "F? L" in opcode.pl), whatever wasn't already parsed as
7375 * a filehandle should be subject to "strict subs".
7376 * Likewise for the optional indirect-object argument to system
7377 * or exec, which can't be a bareword */
7378 if ((PL_last_lop_op == OP_PRINT
7379 || PL_last_lop_op == OP_PRTF
7380 || PL_last_lop_op == OP_SAY
7381 || PL_last_lop_op == OP_SYSTEM
7382 || PL_last_lop_op == OP_EXEC)
7383 && (PL_hints & HINT_STRICT_SUBS))
7384 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7385 if (lastchar != '-') {
7386 if (ckWARN(WARN_RESERVED)) {
7390 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7391 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7399 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7400 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7401 "Operator or semicolon missing before %c%"SVf,
7402 lastchar, SVfARG(newSVpvn_flags(PL_tokenbuf,
7403 strlen(PL_tokenbuf),
7404 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
7405 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7406 "Ambiguous use of %c resolved as operator %c",
7407 lastchar, lastchar);
7414 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7419 (OP*)newSVOP(OP_CONST, 0,
7420 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7423 case KEY___PACKAGE__:
7425 (OP*)newSVOP(OP_CONST, 0,
7427 ? newSVhek(HvNAME_HEK(PL_curstash))
7434 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7435 const char *pname = "main";
7438 if (PL_tokenbuf[2] == 'D')
7441 PL_curstash ? PL_curstash : PL_defstash;
7442 pname = HvNAME_get(stash);
7443 plen = HvNAMELEN (stash);
7444 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7446 gv = gv_fetchpvn_flags(
7447 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7448 plen+6, GV_ADD|putf8, SVt_PVIO
7452 GvIOp(gv) = newIO();
7453 IoIFP(GvIOp(gv)) = PL_rsfp;
7454 #if defined(HAS_FCNTL) && defined(F_SETFD)
7456 const int fd = PerlIO_fileno(PL_rsfp);
7457 fcntl(fd,F_SETFD,fd >= 3);
7460 /* Mark this internal pseudo-handle as clean */
7461 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7462 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7463 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7465 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7466 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7467 /* if the script was opened in binmode, we need to revert
7468 * it to text mode for compatibility; but only iff it has CRs
7469 * XXX this is a questionable hack at best. */
7470 if (PL_bufend-PL_bufptr > 2
7471 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7474 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7475 loc = PerlIO_tell(PL_rsfp);
7476 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7479 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7481 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7482 #endif /* NETWARE */
7484 PerlIO_seek(PL_rsfp, loc, 0);
7488 #ifdef PERLIO_LAYERS
7491 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7492 else if (PL_encoding) {
7499 XPUSHs(PL_encoding);
7501 call_method("name", G_SCALAR);
7505 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7506 Perl_form(aTHX_ ":encoding(%"SVf")",
7515 if (PL_realtokenstart >= 0) {
7516 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7518 PL_endwhite = newSVpvs("");
7519 sv_catsv(PL_endwhite, PL_thiswhite);
7521 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7522 PL_realtokenstart = -1;
7524 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7534 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7543 if (PL_expect == XSTATE) {
7550 if (*s == ':' && s[1] == ':') {
7554 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7555 if ((*s == ':' && s[1] == ':')
7556 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7560 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7564 Perl_croak(aTHX_ "CORE::%"SVf" is not a keyword",
7565 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7566 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7569 else if (tmp == KEY_require || tmp == KEY_do
7571 /* that's a way to remember we saw "CORE::" */
7584 LOP(OP_ACCEPT,XTERM);
7587 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7592 LOP(OP_ATAN2,XTERM);
7598 LOP(OP_BINMODE,XTERM);
7601 LOP(OP_BLESS,XTERM);
7610 /* We have to disambiguate the two senses of
7611 "continue". If the next token is a '{' then
7612 treat it as the start of a continue block;
7613 otherwise treat it as a control operator.
7623 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7633 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7642 if (!PL_cryptseen) {
7643 PL_cryptseen = TRUE;
7647 LOP(OP_CRYPT,XTERM);
7650 LOP(OP_CHMOD,XTERM);
7653 LOP(OP_CHOWN,XTERM);
7656 LOP(OP_CONNECT,XTERM);
7676 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7678 if (len && !keyword(PL_tokenbuf + 1, len, 0)) {
7681 force_ident_maybe_lex('&');
7686 if (orig_keyword == KEY_do) {
7695 PL_hints |= HINT_BLOCK_SCOPE;
7705 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7706 STR_WITH_LEN("NDBM_File::"),
7707 STR_WITH_LEN("DB_File::"),
7708 STR_WITH_LEN("GDBM_File::"),
7709 STR_WITH_LEN("SDBM_File::"),
7710 STR_WITH_LEN("ODBM_File::"),
7712 LOP(OP_DBMOPEN,XTERM);
7718 PL_expect = XOPERATOR;
7719 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7726 pl_yylval.ival = CopLINE(PL_curcop);
7730 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7744 if (*s == '{') { /* block eval */
7745 PL_expect = XTERMBLOCK;
7746 UNIBRACK(OP_ENTERTRY);
7748 else { /* string eval */
7750 UNIBRACK(OP_ENTEREVAL);
7755 UNIBRACK(-OP_ENTEREVAL);
7769 case KEY_endhostent:
7775 case KEY_endservent:
7778 case KEY_endprotoent:
7789 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7791 pl_yylval.ival = CopLINE(PL_curcop);
7793 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7796 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7799 if ((PL_bufend - p) >= 3 &&
7800 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7802 else if ((PL_bufend - p) >= 4 &&
7803 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7806 if (isIDFIRST_lazy_if(p,UTF)) {
7807 p = scan_ident(p, PL_bufend,
7808 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7812 Perl_croak(aTHX_ "Missing $ on loop variable");
7814 s = SvPVX(PL_linestr) + soff;
7820 LOP(OP_FORMLINE,XTERM);
7829 LOP(OP_FCNTL,XTERM);
7835 LOP(OP_FLOCK,XTERM);
7838 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7843 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7848 LOP(OP_GREPSTART, XREF);
7851 PL_expect = XOPERATOR;
7852 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7867 case KEY_getpriority:
7868 LOP(OP_GETPRIORITY,XTERM);
7870 case KEY_getprotobyname:
7873 case KEY_getprotobynumber:
7874 LOP(OP_GPBYNUMBER,XTERM);
7876 case KEY_getprotoent:
7888 case KEY_getpeername:
7889 UNI(OP_GETPEERNAME);
7891 case KEY_gethostbyname:
7894 case KEY_gethostbyaddr:
7895 LOP(OP_GHBYADDR,XTERM);
7897 case KEY_gethostent:
7900 case KEY_getnetbyname:
7903 case KEY_getnetbyaddr:
7904 LOP(OP_GNBYADDR,XTERM);
7909 case KEY_getservbyname:
7910 LOP(OP_GSBYNAME,XTERM);
7912 case KEY_getservbyport:
7913 LOP(OP_GSBYPORT,XTERM);
7915 case KEY_getservent:
7918 case KEY_getsockname:
7919 UNI(OP_GETSOCKNAME);
7921 case KEY_getsockopt:
7922 LOP(OP_GSOCKOPT,XTERM);
7937 pl_yylval.ival = CopLINE(PL_curcop);
7942 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7950 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7952 pl_yylval.ival = CopLINE(PL_curcop);
7956 LOP(OP_INDEX,XTERM);
7962 LOP(OP_IOCTL,XTERM);
7974 PL_expect = XOPERATOR;
7975 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7992 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7997 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
8011 LOP(OP_LISTEN,XTERM);
8020 s = scan_pat(s,OP_MATCH);
8021 TERM(sublex_start());
8024 LOP(OP_MAPSTART, XREF);
8027 LOP(OP_MKDIR,XTERM);
8030 LOP(OP_MSGCTL,XTERM);
8033 LOP(OP_MSGGET,XTERM);
8036 LOP(OP_MSGRCV,XTERM);
8039 LOP(OP_MSGSND,XTERM);
8044 PL_in_my = (U16)tmp;
8046 if (isIDFIRST_lazy_if(s,UTF)) {
8050 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
8051 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
8053 if (!FEATURE_LEXSUBS_IS_ENABLED)
8055 "Experimental \"%s\" subs not enabled",
8056 tmp == KEY_my ? "my" :
8057 tmp == KEY_state ? "state" : "our");
8058 Perl_ck_warner_d(aTHX_
8059 packWARN(WARN_EXPERIMENTAL__LEXICAL_SUBS),
8060 "The lexical_subs feature is experimental");
8063 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
8064 if (!PL_in_my_stash) {
8067 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
8068 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
8071 if (PL_madskills) { /* just add type to declarator token */
8072 sv_catsv(PL_thistoken, PL_nextwhite);
8074 sv_catpvn(PL_thistoken, start, s - start);
8082 PL_expect = XOPERATOR;
8083 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8087 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
8092 s = tokenize_use(0, s);
8096 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
8099 if (!PL_lex_allbrackets &&
8100 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
8101 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
8107 if (isIDFIRST_lazy_if(s,UTF)) {
8109 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE,
8111 for (t=d; isSPACE(*t);)
8113 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
8115 && !(t[0] == '=' && t[1] == '>')
8116 && !(t[0] == ':' && t[1] == ':')
8117 && !keyword(s, d-s, 0)
8119 SV *tmpsv = newSVpvn_flags(s, (STRLEN)(d-s),
8120 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
8121 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
8122 "Precedence problem: open %"SVf" should be open(%"SVf")",
8123 SVfARG(tmpsv), SVfARG(tmpsv));
8129 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8131 pl_yylval.ival = OP_OR;
8141 LOP(OP_OPEN_DIR,XTERM);
8144 checkcomma(s,PL_tokenbuf,"filehandle");
8148 checkcomma(s,PL_tokenbuf,"filehandle");
8167 s = force_word(s,WORD,FALSE,TRUE,FALSE);
8169 s = force_strict_version(s);
8170 PL_lex_expect = XBLOCK;
8174 LOP(OP_PIPE_OP,XTERM);
8177 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
8180 pl_yylval.ival = OP_CONST;
8181 TERM(sublex_start());
8188 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
8191 PL_expect = XOPERATOR;
8192 if (SvCUR(PL_lex_stuff)) {
8193 int warned_comma = !ckWARN(WARN_QW);
8194 int warned_comment = warned_comma;
8195 d = SvPV_force(PL_lex_stuff, len);
8197 for (; isSPACE(*d) && len; --len, ++d)
8202 if (!warned_comma || !warned_comment) {
8203 for (; !isSPACE(*d) && len; --len, ++d) {
8204 if (!warned_comma && *d == ',') {
8205 Perl_warner(aTHX_ packWARN(WARN_QW),
8206 "Possible attempt to separate words with commas");
8209 else if (!warned_comment && *d == '#') {
8210 Perl_warner(aTHX_ packWARN(WARN_QW),
8211 "Possible attempt to put comments in qw() list");
8217 for (; !isSPACE(*d) && len; --len, ++d)
8220 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
8221 words = op_append_elem(OP_LIST, words,
8222 newSVOP(OP_CONST, 0, tokeq(sv)));
8227 words = newNULLLIST();
8229 SvREFCNT_dec(PL_lex_stuff);
8230 PL_lex_stuff = NULL;
8232 PL_expect = XOPERATOR;
8233 pl_yylval.opval = sawparens(words);
8238 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
8241 pl_yylval.ival = OP_STRINGIFY;
8242 if (SvIVX(PL_lex_stuff) == '\'')
8243 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
8244 TERM(sublex_start());
8247 s = scan_pat(s,OP_QR);
8248 TERM(sublex_start());
8251 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
8254 readpipe_override();
8255 TERM(sublex_start());
8262 PL_expect = XOPERATOR;
8264 s = force_version(s, FALSE);
8266 else if (*s != 'v' || !isDIGIT(s[1])
8267 || (s = force_version(s, TRUE), *s == 'v'))
8269 *PL_tokenbuf = '\0';
8270 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8271 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8272 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8273 GV_ADD | (UTF ? SVf_UTF8 : 0));
8275 yyerror("<> should be quotes");
8277 if (orig_keyword == KEY_require) {
8285 PL_last_uni = PL_oldbufptr;
8286 PL_last_lop_op = OP_REQUIRE;
8288 return REPORT( (int)REQUIRE );
8294 PL_expect = XOPERATOR;
8295 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8299 LOP(OP_RENAME,XTERM);
8308 LOP(OP_RINDEX,XTERM);
8317 UNIDOR(OP_READLINE);
8320 UNIDOR(OP_BACKTICK);
8329 LOP(OP_REVERSE,XTERM);
8332 UNIDOR(OP_READLINK);
8339 if (pl_yylval.opval)
8340 TERM(sublex_start());
8342 TOKEN(1); /* force error */
8345 checkcomma(s,PL_tokenbuf,"filehandle");
8355 LOP(OP_SELECT,XTERM);
8361 LOP(OP_SEMCTL,XTERM);
8364 LOP(OP_SEMGET,XTERM);
8367 LOP(OP_SEMOP,XTERM);
8373 LOP(OP_SETPGRP,XTERM);
8375 case KEY_setpriority:
8376 LOP(OP_SETPRIORITY,XTERM);
8378 case KEY_sethostent:
8384 case KEY_setservent:
8387 case KEY_setprotoent:
8397 LOP(OP_SEEKDIR,XTERM);
8399 case KEY_setsockopt:
8400 LOP(OP_SSOCKOPT,XTERM);
8406 LOP(OP_SHMCTL,XTERM);
8409 LOP(OP_SHMGET,XTERM);
8412 LOP(OP_SHMREAD,XTERM);
8415 LOP(OP_SHMWRITE,XTERM);
8418 LOP(OP_SHUTDOWN,XTERM);
8427 LOP(OP_SOCKET,XTERM);
8429 case KEY_socketpair:
8430 LOP(OP_SOCKPAIR,XTERM);
8433 checkcomma(s,PL_tokenbuf,"subroutine name");
8436 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8440 LOP(OP_SPLIT,XTERM);
8443 LOP(OP_SPRINTF,XTERM);
8446 LOP(OP_SPLICE,XTERM);
8461 LOP(OP_SUBSTR,XTERM);
8467 char * const tmpbuf = PL_tokenbuf + 1;
8468 SSize_t tboffset = 0;
8469 expectation attrful;
8470 bool have_name, have_proto;
8471 const int key = tmp;
8476 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8477 SV *subtoken = PL_madskills
8478 ? newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr))
8483 s = SKIPSPACE2(s,tmpwhite);
8489 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8490 (*s == ':' && s[1] == ':'))
8493 SV *nametoke = NULL;
8497 attrful = XATTRBLOCK;
8498 /* remember buffer pos'n for later force_word */
8499 tboffset = s - PL_oldbufptr;
8500 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8504 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8507 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8509 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
8511 sv_setpvn(PL_subname, tmpbuf, len);
8513 sv_setsv(PL_subname,PL_curstname);
8514 sv_catpvs(PL_subname,"::");
8515 sv_catpvn(PL_subname,tmpbuf,len);
8517 if (SvUTF8(PL_linestr))
8518 SvUTF8_on(PL_subname);
8524 CURMAD('X', nametoke);
8525 CURMAD('_', tmpwhite);
8526 force_ident_maybe_lex('&');
8528 s = SKIPSPACE2(d,tmpwhite);
8534 if (key == KEY_my || key == KEY_our || key==KEY_state)
8537 /* diag_listed_as: Missing name in "%s sub" */
8539 "Missing name in \"%s\"", PL_bufptr);
8541 PL_expect = XTERMBLOCK;
8542 attrful = XATTRTERM;
8543 sv_setpvs(PL_subname,"?");
8547 if (key == KEY_format) {
8549 PL_thistoken = subtoken;
8553 (void) force_word(PL_oldbufptr + tboffset, WORD,
8559 /* Look for a prototype */
8562 bool bad_proto = FALSE;
8563 bool in_brackets = FALSE;
8564 char greedy_proto = ' ';
8565 bool proto_after_greedy_proto = FALSE;
8566 bool must_be_last = FALSE;
8567 bool underscore = FALSE;
8568 bool seen_underscore = FALSE;
8569 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8572 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
8574 Perl_croak(aTHX_ "Prototype not terminated");
8575 /* strip spaces and check for bad characters */
8576 d = SvPV(PL_lex_stuff, tmplen);
8578 for (p = d; tmplen; tmplen--, ++p) {
8582 if (warnillegalproto) {
8584 proto_after_greedy_proto = TRUE;
8585 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8590 if ( !strchr(";@%", *p) )
8597 else if ( *p == ']' ) {
8598 in_brackets = FALSE;
8600 else if ( (*p == '@' || *p == '%') &&
8601 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8603 must_be_last = TRUE;
8606 else if ( *p == '_' ) {
8607 underscore = seen_underscore = TRUE;
8614 if (proto_after_greedy_proto)
8615 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8616 "Prototype after '%c' for %"SVf" : %s",
8617 greedy_proto, SVfARG(PL_subname), d);
8619 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8620 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8621 "Illegal character %sin prototype for %"SVf" : %s",
8622 seen_underscore ? "after '_' " : "",
8624 SvUTF8(PL_lex_stuff)
8625 ? sv_uni_display(dsv,
8626 newSVpvn_flags(d, tmp, SVs_TEMP | SVf_UTF8),
8628 UNI_DISPLAY_ISPRINT)
8629 : pv_pretty(dsv, d, tmp, 60, NULL, NULL,
8630 PERL_PV_ESCAPE_NONASCII));
8632 SvCUR_set(PL_lex_stuff, tmp);
8637 CURMAD('q', PL_thisopen);
8638 CURMAD('_', tmpwhite);
8639 CURMAD('=', PL_thisstuff);
8640 CURMAD('Q', PL_thisclose);
8641 NEXTVAL_NEXTTOKE.opval =
8642 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8643 PL_lex_stuff = NULL;
8646 s = SKIPSPACE2(s,tmpwhite);
8654 if (*s == ':' && s[1] != ':')
8655 PL_expect = attrful;
8656 else if (*s != '{' && key == KEY_sub) {
8658 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8659 else if (*s != ';' && *s != '}')
8660 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8667 curmad('^', newSVpvs(""));
8668 CURMAD('_', tmpwhite);
8672 PL_thistoken = subtoken;
8675 NEXTVAL_NEXTTOKE.opval =
8676 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8677 PL_lex_stuff = NULL;
8683 sv_setpvs(PL_subname, "__ANON__");
8685 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8689 force_ident_maybe_lex('&');
8695 LOP(OP_SYSTEM,XREF);
8698 LOP(OP_SYMLINK,XTERM);
8701 LOP(OP_SYSCALL,XTERM);
8704 LOP(OP_SYSOPEN,XTERM);
8707 LOP(OP_SYSSEEK,XTERM);
8710 LOP(OP_SYSREAD,XTERM);
8713 LOP(OP_SYSWRITE,XTERM);
8718 TERM(sublex_start());
8739 LOP(OP_TRUNCATE,XTERM);
8751 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8753 pl_yylval.ival = CopLINE(PL_curcop);
8757 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8759 pl_yylval.ival = CopLINE(PL_curcop);
8763 LOP(OP_UNLINK,XTERM);
8769 LOP(OP_UNPACK,XTERM);
8772 LOP(OP_UTIME,XTERM);
8778 LOP(OP_UNSHIFT,XTERM);
8781 s = tokenize_use(1, s);
8791 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8793 pl_yylval.ival = CopLINE(PL_curcop);
8797 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8799 pl_yylval.ival = CopLINE(PL_curcop);
8803 PL_hints |= HINT_BLOCK_SCOPE;
8810 LOP(OP_WAITPID,XTERM);
8819 ctl_l[0] = toCTRL('L');
8821 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8824 /* Make sure $^L is defined */
8825 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8830 if (PL_expect == XOPERATOR) {
8831 if (*s == '=' && !PL_lex_allbrackets &&
8832 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8840 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8842 pl_yylval.ival = OP_XOR;
8848 #pragma segment Main
8854 Looks up an identifier in the pad or in a package
8857 PRIVATEREF if this is a lexical name.
8858 WORD if this belongs to a package.
8861 if we're in a my declaration
8862 croak if they tried to say my($foo::bar)
8863 build the ops for a my() declaration
8864 if it's an access to a my() variable
8865 build ops for access to a my() variable
8866 if in a dq string, and they've said @foo and we can't find @foo
8868 build ops for a bareword
8872 S_pending_ident(pTHX)
8876 const char pit = (char)pl_yylval.ival;
8877 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8878 /* All routes through this function want to know if there is a colon. */
8879 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8881 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8882 "### Pending identifier '%s'\n", PL_tokenbuf); });
8884 /* if we're in a my(), we can't allow dynamics here.
8885 $foo'bar has already been turned into $foo::bar, so
8886 just check for colons.
8888 if it's a legal name, the OP is a PADANY.
8891 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8893 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8894 "variable %s in \"our\"",
8895 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8896 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8900 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8901 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8902 UTF ? SVf_UTF8 : 0);
8904 pl_yylval.opval = newOP(OP_PADANY, 0);
8905 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8906 UTF ? SVf_UTF8 : 0);
8912 build the ops for accesses to a my() variable.
8917 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8918 UTF ? SVf_UTF8 : 0);
8919 if (tmp != NOT_IN_PAD) {
8920 /* might be an "our" variable" */
8921 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8922 /* build ops for a bareword */
8923 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8924 HEK * const stashname = HvNAME_HEK(stash);
8925 SV * const sym = newSVhek(stashname);
8926 sv_catpvs(sym, "::");
8927 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8928 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8929 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8933 ? (GV_ADDMULTI | GV_ADDINEVAL)
8936 ((PL_tokenbuf[0] == '$') ? SVt_PV
8937 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8942 pl_yylval.opval = newOP(OP_PADANY, 0);
8943 pl_yylval.opval->op_targ = tmp;
8949 Whine if they've said @foo in a doublequoted string,
8950 and @foo isn't a variable we can find in the symbol
8953 if (ckWARN(WARN_AMBIGUOUS) &&
8954 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8955 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8956 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8957 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8958 /* DO NOT warn for @- and @+ */
8959 && !( PL_tokenbuf[2] == '\0' &&
8960 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8963 /* Downgraded from fatal to warning 20000522 mjd */
8964 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8965 "Possible unintended interpolation of %"SVf" in string",
8966 SVfARG(newSVpvn_flags(PL_tokenbuf, tokenbuf_len,
8967 SVs_TEMP | ( UTF ? SVf_UTF8 : 0 ))));
8971 /* build ops for a bareword */
8972 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8973 newSVpvn_flags(PL_tokenbuf + 1,
8975 UTF ? SVf_UTF8 : 0 ));
8976 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8978 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8979 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8980 | ( UTF ? SVf_UTF8 : 0 ),
8981 ((PL_tokenbuf[0] == '$') ? SVt_PV
8982 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8988 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8992 PERL_ARGS_ASSERT_CHECKCOMMA;
8994 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8995 if (ckWARN(WARN_SYNTAX)) {
8998 for (w = s+2; *w && level; w++) {
9006 /* the list of chars below is for end of statements or
9007 * block / parens, boolean operators (&&, ||, //) and branch
9008 * constructs (or, and, if, until, unless, while, err, for).
9009 * Not a very solid hack... */
9010 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
9011 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
9012 "%s (...) interpreted as function",name);
9015 while (s < PL_bufend && isSPACE(*s))
9019 while (s < PL_bufend && isSPACE(*s))
9021 if (isIDFIRST_lazy_if(s,UTF)) {
9022 const char * const w = s;
9023 s += UTF ? UTF8SKIP(s) : 1;
9024 while (isWORDCHAR_lazy_if(s,UTF))
9025 s += UTF ? UTF8SKIP(s) : 1;
9026 while (s < PL_bufend && isSPACE(*s))
9030 if (keyword(w, s - w, 0))
9033 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
9034 if (gv && GvCVu(gv))
9036 Perl_croak(aTHX_ "No comma allowed after %s", what);
9041 /* Either returns sv, or mortalizes/frees sv and returns a new SV*.
9042 Best used as sv=new_constant(..., sv, ...).
9043 If s, pv are NULL, calls subroutine with one argument,
9044 and <type> is used with error messages only.
9045 <type> is assumed to be well formed UTF-8 */
9048 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
9049 SV *sv, SV *pv, const char *type, STRLEN typelen)
9052 HV * table = GvHV(PL_hintgv); /* ^H */
9057 const char *why1 = "", *why2 = "", *why3 = "";
9059 PERL_ARGS_ASSERT_NEW_CONSTANT;
9060 /* We assume that this is true: */
9061 if (*key == 'c') { assert (strEQ(key, "charnames")); }
9064 /* charnames doesn't work well if there have been errors found */
9065 if (PL_error_count > 0 && *key == 'c')
9067 SvREFCNT_dec_NN(sv);
9068 return &PL_sv_undef;
9071 sv_2mortal(sv); /* Parent created it permanently */
9073 || ! (PL_hints & HINT_LOCALIZE_HH)
9074 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
9079 /* Here haven't found what we're looking for. If it is charnames,
9080 * perhaps it needs to be loaded. Try doing that before giving up */
9082 Perl_load_module(aTHX_
9084 newSVpvs("_charnames"),
9085 /* version parameter; no need to specify it, as if
9086 * we get too early a version, will fail anyway,
9087 * not being able to find '_charnames' */
9093 table = GvHV(PL_hintgv);
9095 && (PL_hints & HINT_LOCALIZE_HH)
9096 && (cvp = hv_fetch(table, key, keylen, FALSE))
9102 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
9103 msg = Perl_form(aTHX_
9104 "Constant(%.*s) unknown",
9105 (int)(type ? typelen : len),
9111 why3 = "} is not defined";
9114 msg = Perl_form(aTHX_
9115 /* The +3 is for '\N{'; -4 for that, plus '}' */
9116 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
9120 msg = Perl_form(aTHX_ "Constant(%.*s): %s%s%s",
9121 (int)(type ? typelen : len),
9122 (type ? type: s), why1, why2, why3);
9125 yyerror_pv(msg, UTF ? SVf_UTF8 : 0);
9126 return SvREFCNT_inc_simple_NN(sv);
9131 pv = newSVpvn_flags(s, len, SVs_TEMP);
9133 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
9135 typesv = &PL_sv_undef;
9137 PUSHSTACKi(PERLSI_OVERLOAD);
9149 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
9153 /* Check the eval first */
9154 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
9156 const char * errstr;
9157 sv_catpvs(errsv, "Propagated");
9158 errstr = SvPV_const(errsv, errlen);
9159 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
9161 res = SvREFCNT_inc_simple_NN(sv);
9165 SvREFCNT_inc_simple_void_NN(res);
9174 why1 = "Call to &{$^H{";
9176 why3 = "}} did not return a defined value";
9178 (void)sv_2mortal(sv);
9185 PERL_STATIC_INLINE void
9186 S_parse_ident(pTHX_ char **s, char **d, char * const e, int allow_package, bool is_utf8) {
9188 PERL_ARGS_ASSERT_PARSE_IDENT;
9192 Perl_croak(aTHX_ "%s", ident_too_long);
9193 if (is_utf8 && isIDFIRST_utf8((U8*)*s)) {
9194 /* The UTF-8 case must come first, otherwise things
9195 * like c\N{COMBINING TILDE} would start failing, as the
9196 * isWORDCHAR_A case below would gobble the 'c' up.
9199 char *t = *s + UTF8SKIP(*s);
9200 while (isIDCONT_utf8((U8*)t))
9202 if (*d + (t - *s) > e)
9203 Perl_croak(aTHX_ "%s", ident_too_long);
9204 Copy(*s, *d, t - *s, char);
9208 else if ( isWORDCHAR_A(**s) ) {
9211 } while isWORDCHAR_A(**s);
9213 else if (allow_package && **s == '\'' && isIDFIRST_lazy_if(*s+1,is_utf8)) {
9218 else if (allow_package && **s == ':' && (*s)[1] == ':'
9219 /* Disallow things like Foo::$bar. For the curious, this is
9220 * the code path that triggers the "Bad name after" warning
9221 * when looking for barewords.
9223 && (*s)[2] != '$') {
9233 /* Returns a NUL terminated string, with the length of the string written to
9237 S_scan_word(pTHX_ char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
9241 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9242 bool is_utf8 = cBOOL(UTF);
9244 PERL_ARGS_ASSERT_SCAN_WORD;
9246 parse_ident(&s, &d, e, allow_package, is_utf8);
9253 S_scan_ident(pTHX_ char *s, const char *send, char *dest, STRLEN destlen, I32 ck_uni)
9256 char *bracket = NULL;
9259 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9260 bool is_utf8 = cBOOL(UTF);
9262 PERL_ARGS_ASSERT_SCAN_IDENT;
9267 while (isDIGIT(*s)) {
9269 Perl_croak(aTHX_ "%s", ident_too_long);
9274 parse_ident(&s, &d, e, 1, is_utf8);
9279 if (PL_lex_state != LEX_NORMAL)
9280 PL_lex_state = LEX_INTERPENDMAYBE;
9283 if (*s == '$' && s[1] &&
9284 (isIDFIRST_lazy_if(s+1,is_utf8)
9285 || isDIGIT_A((U8)s[1])
9288 || strnEQ(s+1,"::",2)) )
9295 while (s < send && SPACE_OR_TAB(*s))
9299 #define VALID_LEN_ONE_IDENT(d, u) (isPUNCT_A((U8)*(d)) \
9300 || isCNTRL_A((U8)*(d)) \
9301 || isDIGIT_A((U8)*(d)) \
9302 || (!(u) && !UTF8_IS_INVARIANT((U8)*(d))))
9304 && (isIDFIRST_lazy_if(s, is_utf8) || VALID_LEN_ONE_IDENT(s, is_utf8)))
9307 const STRLEN skip = UTF8SKIP(s);
9310 for ( i = 0; i < skip; i++ )
9318 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9322 else if (ck_uni && !bracket)
9325 if (isIDFIRST_lazy_if(d,is_utf8)) {
9326 d += is_utf8 ? UTF8SKIP(d) : 1;
9327 parse_ident(&s, &d, e, 1, is_utf8);
9329 while (s < send && SPACE_OR_TAB(*s))
9331 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9332 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9333 const char * const brack =
9335 ((*s == '[') ? "[...]" : "{...}");
9336 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9337 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9338 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9339 funny, dest, brack, funny, dest, brack);
9342 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9343 PL_lex_allbrackets++;
9347 /* Handle extended ${^Foo} variables
9348 * 1999-02-27 mjd-perl-patch@plover.com */
9349 else if (!isWORDCHAR(*d) && !isPRINT(*d) /* isCTRL(d) */
9353 while (isWORDCHAR(*s) && d < e) {
9357 Perl_croak(aTHX_ "%s", ident_too_long);
9361 while (s < send && SPACE_OR_TAB(*s))
9366 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9367 PL_lex_state = LEX_INTERPEND;
9370 if (PL_lex_state == LEX_NORMAL) {
9371 if (ckWARN(WARN_AMBIGUOUS) &&
9372 (keyword(dest, d - dest, 0)
9373 || get_cvn_flags(dest, d - dest, is_utf8 ? SVf_UTF8 : 0)))
9375 SV *tmp = newSVpvn_flags( dest, d - dest,
9376 SVs_TEMP | (is_utf8 ? SVf_UTF8 : 0) );
9379 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9380 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9381 funny, tmp, funny, tmp);
9386 s = bracket; /* let the parser handle it */
9390 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9391 PL_lex_state = LEX_INTERPEND;
9396 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
9398 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
9399 * the parse starting at 's', based on the subset that are valid in this
9400 * context input to this routine in 'valid_flags'. Advances s. Returns
9401 * TRUE if the input should be treated as a valid flag, so the next char
9402 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
9403 * first call on the current regex. This routine will set it to any
9404 * charset modifier found. The caller shouldn't change it. This way,
9405 * another charset modifier encountered in the parse can be detected as an
9406 * error, as we have decided to allow only one */
9409 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9411 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9412 if (isWORDCHAR_lazy_if(*s, UTF)) {
9413 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9414 UTF ? SVf_UTF8 : 0);
9416 /* Pretend that it worked, so will continue processing before
9425 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
9426 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9427 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9428 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9429 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9430 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9431 case LOCALE_PAT_MOD:
9433 goto multiple_charsets;
9435 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9438 case UNICODE_PAT_MOD:
9440 goto multiple_charsets;
9442 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9445 case ASCII_RESTRICT_PAT_MOD:
9447 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9451 /* Error if previous modifier wasn't an 'a', but if it was, see
9452 * if, and accept, a second occurrence (only) */
9454 || get_regex_charset(*pmfl)
9455 != REGEX_ASCII_RESTRICTED_CHARSET)
9457 goto multiple_charsets;
9459 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9463 case DEPENDS_PAT_MOD:
9465 goto multiple_charsets;
9467 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9476 if (*charset != c) {
9477 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9479 else if (c == 'a') {
9480 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9483 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9486 /* Pretend that it worked, so will continue processing before dieing */
9492 S_scan_pat(pTHX_ char *start, I32 type)
9496 char *s = scan_str(start,!!PL_madskills,FALSE, PL_reg_state.re_reparsing,
9497 TRUE /* look for escaped bracketed metas */ );
9498 const char * const valid_flags =
9499 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9500 char charset = '\0'; /* character set modifier */
9505 PERL_ARGS_ASSERT_SCAN_PAT;
9507 /* this was only needed for the initial scan_str; set it to false
9508 * so that any (?{}) code blocks etc are parsed normally */
9509 PL_reg_state.re_reparsing = FALSE;
9511 const char * const delimiter = skipspace(start);
9515 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9516 : "Search pattern not terminated" ));
9519 pm = (PMOP*)newPMOP(type, 0);
9520 if (PL_multi_open == '?') {
9521 /* This is the only point in the code that sets PMf_ONCE: */
9522 pm->op_pmflags |= PMf_ONCE;
9524 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9525 allows us to restrict the list needed by reset to just the ??
9527 assert(type != OP_TRANS);
9529 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9532 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9535 elements = mg->mg_len / sizeof(PMOP**);
9536 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9537 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9538 mg->mg_len = elements * sizeof(PMOP**);
9539 PmopSTASH_set(pm,PL_curstash);
9546 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9547 * anon CV. False positives like qr/[(?{]/ are harmless */
9549 if (type == OP_QR) {
9551 char *e, *p = SvPV(PL_lex_stuff, len);
9553 for (; p < e; p++) {
9554 if (p[0] == '(' && p[1] == '?'
9555 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9557 pm->op_pmflags |= PMf_HAS_CV;
9561 pm->op_pmflags |= PMf_IS_QR;
9564 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9566 if (PL_madskills && modstart != s) {
9567 SV* tmptoken = newSVpvn(modstart, s - modstart);
9568 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9571 /* issue a warning if /c is specified,but /g is not */
9572 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9574 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9575 "Use of /c modifier is meaningless without /g" );
9578 PL_lex_op = (OP*)pm;
9579 pl_yylval.ival = OP_MATCH;
9584 S_scan_subst(pTHX_ char *start)
9591 char charset = '\0'; /* character set modifier */
9596 PERL_ARGS_ASSERT_SCAN_SUBST;
9598 pl_yylval.ival = OP_NULL;
9600 s = scan_str(start,!!PL_madskills,FALSE,FALSE,
9601 TRUE /* look for escaped bracketed metas */ );
9604 Perl_croak(aTHX_ "Substitution pattern not terminated");
9606 if (s[-1] == PL_multi_open)
9610 CURMAD('q', PL_thisopen);
9611 CURMAD('_', PL_thiswhite);
9612 CURMAD('E', PL_thisstuff);
9613 CURMAD('Q', PL_thisclose);
9614 PL_realtokenstart = s - SvPVX(PL_linestr);
9618 first_start = PL_multi_start;
9619 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
9622 SvREFCNT_dec(PL_lex_stuff);
9623 PL_lex_stuff = NULL;
9625 Perl_croak(aTHX_ "Substitution replacement not terminated");
9627 PL_multi_start = first_start; /* so whole substitution is taken together */
9629 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9633 CURMAD('z', PL_thisopen);
9634 CURMAD('R', PL_thisstuff);
9635 CURMAD('Z', PL_thisclose);
9641 if (*s == EXEC_PAT_MOD) {
9645 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9654 curmad('m', newSVpvn(modstart, s - modstart));
9655 append_madprops(PL_thismad, (OP*)pm, 0);
9659 if ((pm->op_pmflags & PMf_CONTINUE)) {
9660 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9664 SV * const repl = newSVpvs("");
9667 pm->op_pmflags |= PMf_EVAL;
9670 sv_catpvs(repl, "eval ");
9672 sv_catpvs(repl, "do ");
9674 sv_catpvs(repl, "{");
9675 sv_catsv(repl, PL_sublex_info.repl);
9676 sv_catpvs(repl, "}");
9678 SvREFCNT_dec(PL_sublex_info.repl);
9679 PL_sublex_info.repl = repl;
9682 PL_lex_op = (OP*)pm;
9683 pl_yylval.ival = OP_SUBST;
9688 S_scan_trans(pTHX_ char *start)
9696 bool nondestruct = 0;
9701 PERL_ARGS_ASSERT_SCAN_TRANS;
9703 pl_yylval.ival = OP_NULL;
9705 s = scan_str(start,!!PL_madskills,FALSE,FALSE, FALSE);
9707 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9709 if (s[-1] == PL_multi_open)
9713 CURMAD('q', PL_thisopen);
9714 CURMAD('_', PL_thiswhite);
9715 CURMAD('E', PL_thisstuff);
9716 CURMAD('Q', PL_thisclose);
9717 PL_realtokenstart = s - SvPVX(PL_linestr);
9721 s = scan_str(s,!!PL_madskills,FALSE,FALSE, FALSE);
9724 SvREFCNT_dec(PL_lex_stuff);
9725 PL_lex_stuff = NULL;
9727 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9730 CURMAD('z', PL_thisopen);
9731 CURMAD('R', PL_thisstuff);
9732 CURMAD('Z', PL_thisclose);
9735 complement = del = squash = 0;
9742 complement = OPpTRANS_COMPLEMENT;
9745 del = OPpTRANS_DELETE;
9748 squash = OPpTRANS_SQUASH;
9760 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9761 o->op_private &= ~OPpTRANS_ALL;
9762 o->op_private |= del|squash|complement|
9763 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9764 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9767 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9772 curmad('m', newSVpvn(modstart, s - modstart));
9773 append_madprops(PL_thismad, o, 0);
9782 Takes a pointer to the first < in <<FOO.
9783 Returns a pointer to the byte following <<FOO.
9785 This function scans a heredoc, which involves different methods
9786 depending on whether we are in a string eval, quoted construct, etc.
9787 This is because PL_linestr could containing a single line of input, or
9788 a whole string being evalled, or the contents of the current quote-
9791 The two basic methods are:
9792 - Steal lines from the input stream
9793 - Scan the heredoc in PL_linestr and remove it therefrom
9795 In a file scope or filtered eval, the first method is used; in a
9796 string eval, the second.
9798 In a quote-like operator, we have to choose between the two,
9799 depending on where we can find a newline. We peek into outer lex-
9800 ing scopes until we find one with a newline in it. If we reach the
9801 outermost lexing scope and it is a file, we use the stream method.
9802 Otherwise it is treated as an eval.
9806 S_scan_heredoc(pTHX_ char *s)
9809 I32 op_type = OP_SCALAR;
9816 const bool infile = PL_rsfp || PL_parser->filtered;
9817 LEXSHARED *shared = PL_parser->lex_shared;
9819 I32 stuffstart = s - SvPVX(PL_linestr);
9822 PL_realtokenstart = -1;
9825 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9828 d = PL_tokenbuf + 1;
9829 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9830 *PL_tokenbuf = '\n';
9832 while (SPACE_OR_TAB(*peek))
9834 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9837 s = delimcpy(d, e, s, PL_bufend, term, &len);
9839 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9845 /* <<\FOO is equivalent to <<'FOO' */
9849 if (!isWORDCHAR_lazy_if(s,UTF))
9850 deprecate("bare << to mean <<\"\"");
9851 for (; isWORDCHAR_lazy_if(s,UTF); s++) {
9856 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9857 Perl_croak(aTHX_ "Delimiter for here document is too long");
9860 len = d - PL_tokenbuf;
9864 tstart = PL_tokenbuf + 1;
9865 PL_thisclose = newSVpvn(tstart, len - 1);
9866 tstart = SvPVX(PL_linestr) + stuffstart;
9867 PL_thisopen = newSVpvn(tstart, s - tstart);
9868 stuffstart = s - SvPVX(PL_linestr);
9871 #ifndef PERL_STRICT_CR
9872 d = strchr(s, '\r');
9874 char * const olds = s;
9876 while (s < PL_bufend) {
9882 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9891 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9897 tstart = SvPVX(PL_linestr) + stuffstart;
9899 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9901 PL_thisstuff = newSVpvn(tstart, s - tstart);
9904 stuffstart = s - SvPVX(PL_linestr);
9907 tmpstr = newSV_type(SVt_PVIV);
9911 SvIV_set(tmpstr, -1);
9913 else if (term == '`') {
9914 op_type = OP_BACKTICK;
9915 SvIV_set(tmpstr, '\\');
9918 PL_multi_start = CopLINE(PL_curcop) + 1;
9919 PL_multi_open = PL_multi_close = '<';
9920 /* inside a string eval or quote-like operator */
9921 if (!infile || PL_lex_inwhat) {
9924 char * const olds = s;
9925 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9926 /* These two fields are not set until an inner lexing scope is
9927 entered. But we need them set here. */
9928 shared->ls_bufptr = s;
9929 shared->ls_linestr = PL_linestr;
9931 /* Look for a newline. If the current buffer does not have one,
9932 peek into the line buffer of the parent lexing scope, going
9933 up as many levels as necessary to find one with a newline
9936 while (!(s = (char *)memchr(
9937 (void *)shared->ls_bufptr, '\n',
9938 SvEND(shared->ls_linestr)-shared->ls_bufptr
9940 shared = shared->ls_prev;
9941 /* shared is only null if we have gone beyond the outermost
9942 lexing scope. In a file, we will have broken out of the
9943 loop in the previous iteration. In an eval, the string buf-
9944 fer ends with "\n;", so the while condition below will have
9945 evaluated to false. So shared can never be null. */
9947 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9948 most lexing scope. In a file, shared->ls_linestr at that
9949 level is just one line, so there is no body to steal. */
9950 if (infile && !shared->ls_prev) {
9956 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9959 linestr = shared->ls_linestr;
9960 bufend = SvEND(linestr);
9962 while (s < bufend &&
9963 (*s != '\n' || memNE(s,PL_tokenbuf,len)) ) {
9965 ++shared->herelines;
9970 sv_setpvn(tmpstr,d+1,s-d);
9974 sv_catpvn(PL_thisstuff, d + 1, s - d);
9976 PL_thisstuff = newSVpvn(d + 1, s - d);
9977 stuffstart = s - SvPVX(PL_linestr);
9981 /* the preceding stmt passes a newline */
9982 shared->herelines++;
9984 /* s now points to the newline after the heredoc terminator.
9985 d points to the newline before the body of the heredoc.
9988 /* We are going to modify linestr in place here, so set
9989 aside copies of the string if necessary for re-evals or
9991 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9992 check shared->re_eval_str. */
9993 if (shared->re_eval_start || shared->re_eval_str) {
9994 /* Set aside the rest of the regexp */
9995 if (!shared->re_eval_str)
9996 shared->re_eval_str =
9997 newSVpvn(shared->re_eval_start,
9998 bufend - shared->re_eval_start);
9999 shared->re_eval_start -= s-d;
10001 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
10002 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
10003 cx->blk_eval.cur_text == linestr)
10005 cx->blk_eval.cur_text = newSVsv(linestr);
10006 SvSCREAM_on(cx->blk_eval.cur_text);
10008 /* Copy everything from s onwards back to d. */
10009 Move(s,d,bufend-s + 1,char);
10010 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
10011 /* Setting PL_bufend only applies when we have not dug deeper
10012 into other scopes, because sublex_done sets PL_bufend to
10013 SvEND(PL_linestr). */
10014 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
10021 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
10022 term = PL_tokenbuf[1];
10024 linestr_save = PL_linestr; /* must restore this afterwards */
10025 d = s; /* and this */
10026 PL_linestr = newSVpvs("");
10027 PL_bufend = SvPVX(PL_linestr);
10030 if (PL_madskills) {
10031 tstart = SvPVX(PL_linestr) + stuffstart;
10033 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10035 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10038 PL_bufptr = PL_bufend;
10039 CopLINE_set(PL_curcop,
10040 PL_multi_start + shared->herelines);
10041 if (!lex_next_chunk(LEX_NO_TERM)
10042 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
10043 SvREFCNT_dec(linestr_save);
10046 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10047 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
10048 lex_grow_linestr(SvCUR(PL_linestr) + 2);
10049 sv_catpvs(PL_linestr, "\n\0");
10053 stuffstart = s - SvPVX(PL_linestr);
10055 shared->herelines++;
10056 PL_last_lop = PL_last_uni = NULL;
10057 #ifndef PERL_STRICT_CR
10058 if (PL_bufend - PL_linestart >= 2) {
10059 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
10060 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
10062 PL_bufend[-2] = '\n';
10064 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10066 else if (PL_bufend[-1] == '\r')
10067 PL_bufend[-1] = '\n';
10069 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
10070 PL_bufend[-1] = '\n';
10072 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
10073 SvREFCNT_dec(PL_linestr);
10074 PL_linestr = linestr_save;
10075 PL_linestart = SvPVX(linestr_save);
10076 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10081 sv_catsv(tmpstr,PL_linestr);
10085 PL_multi_end = CopLINE(PL_curcop);
10086 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
10087 SvPV_shrink_to_cur(tmpstr);
10090 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
10092 else if (PL_encoding)
10093 sv_recode_to_utf8(tmpstr, PL_encoding);
10095 PL_lex_stuff = tmpstr;
10096 pl_yylval.ival = op_type;
10100 SvREFCNT_dec(tmpstr);
10101 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10102 missingterm(PL_tokenbuf + 1);
10105 /* scan_inputsymbol
10106 takes: current position in input buffer
10107 returns: new position in input buffer
10108 side-effects: pl_yylval and lex_op are set.
10113 <FH> read from filehandle
10114 <pkg::FH> read from package qualified filehandle
10115 <pkg'FH> read from package qualified filehandle
10116 <$fh> read from filehandle in $fh
10117 <*.h> filename glob
10122 S_scan_inputsymbol(pTHX_ char *start)
10125 char *s = start; /* current position in buffer */
10128 char *d = PL_tokenbuf; /* start of temp holding space */
10129 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
10131 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
10133 end = strchr(s, '\n');
10136 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
10138 /* die if we didn't have space for the contents of the <>,
10139 or if it didn't end, or if we see a newline
10142 if (len >= (I32)sizeof PL_tokenbuf)
10143 Perl_croak(aTHX_ "Excessively long <> operator");
10145 Perl_croak(aTHX_ "Unterminated <> operator");
10150 Remember, only scalar variables are interpreted as filehandles by
10151 this code. Anything more complex (e.g., <$fh{$num}>) will be
10152 treated as a glob() call.
10153 This code makes use of the fact that except for the $ at the front,
10154 a scalar variable and a filehandle look the same.
10156 if (*d == '$' && d[1]) d++;
10158 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
10159 while (*d && (isWORDCHAR_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
10160 d += UTF ? UTF8SKIP(d) : 1;
10162 /* If we've tried to read what we allow filehandles to look like, and
10163 there's still text left, then it must be a glob() and not a getline.
10164 Use scan_str to pull out the stuff between the <> and treat it
10165 as nothing more than a string.
10168 if (d - PL_tokenbuf != len) {
10169 pl_yylval.ival = OP_GLOB;
10170 s = scan_str(start,!!PL_madskills,FALSE,FALSE, FALSE);
10172 Perl_croak(aTHX_ "Glob not terminated");
10176 bool readline_overriden = FALSE;
10179 /* we're in a filehandle read situation */
10182 /* turn <> into <ARGV> */
10184 Copy("ARGV",d,5,char);
10186 /* Check whether readline() is overriden */
10187 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
10189 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
10191 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
10192 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
10193 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
10194 readline_overriden = TRUE;
10196 /* if <$fh>, create the ops to turn the variable into a
10200 /* try to find it in the pad for this block, otherwise find
10201 add symbol table ops
10203 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
10204 if (tmp != NOT_IN_PAD) {
10205 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
10206 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
10207 HEK * const stashname = HvNAME_HEK(stash);
10208 SV * const sym = sv_2mortal(newSVhek(stashname));
10209 sv_catpvs(sym, "::");
10210 sv_catpv(sym, d+1);
10215 OP * const o = newOP(OP_PADSV, 0);
10217 PL_lex_op = readline_overriden
10218 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10219 op_append_elem(OP_LIST, o,
10220 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
10221 : (OP*)newUNOP(OP_READLINE, 0, o);
10230 ? (GV_ADDMULTI | GV_ADDINEVAL)
10231 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
10233 PL_lex_op = readline_overriden
10234 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10235 op_append_elem(OP_LIST,
10236 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
10237 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10238 : (OP*)newUNOP(OP_READLINE, 0,
10239 newUNOP(OP_RV2SV, 0,
10240 newGVOP(OP_GV, 0, gv)));
10242 if (!readline_overriden)
10243 PL_lex_op->op_flags |= OPf_SPECIAL;
10244 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
10245 pl_yylval.ival = OP_NULL;
10248 /* If it's none of the above, it must be a literal filehandle
10249 (<Foo::BAR> or <FOO>) so build a simple readline OP */
10251 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
10252 PL_lex_op = readline_overriden
10253 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10254 op_append_elem(OP_LIST,
10255 newGVOP(OP_GV, 0, gv),
10256 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10257 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
10258 pl_yylval.ival = OP_NULL;
10267 takes: start position in buffer
10268 keep_quoted preserve \ on the embedded delimiter(s)
10269 keep_delims preserve the delimiters around the string
10270 re_reparse compiling a run-time /(?{})/:
10271 collapse // to /, and skip encoding src
10272 returns: position to continue reading from buffer
10273 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
10274 updates the read buffer.
10276 This subroutine pulls a string out of the input. It is called for:
10277 q single quotes q(literal text)
10278 ' single quotes 'literal text'
10279 qq double quotes qq(interpolate $here please)
10280 " double quotes "interpolate $here please"
10281 qx backticks qx(/bin/ls -l)
10282 ` backticks `/bin/ls -l`
10283 qw quote words @EXPORT_OK = qw( func() $spam )
10284 m// regexp match m/this/
10285 s/// regexp substitute s/this/that/
10286 tr/// string transliterate tr/this/that/
10287 y/// string transliterate y/this/that/
10288 ($*@) sub prototypes sub foo ($)
10289 (stuff) sub attr parameters sub foo : attr(stuff)
10290 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
10292 In most of these cases (all but <>, patterns and transliterate)
10293 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
10294 calls scan_str(). s/// makes yylex() call scan_subst() which calls
10295 scan_str(). tr/// and y/// make yylex() call scan_trans() which
10298 It skips whitespace before the string starts, and treats the first
10299 character as the delimiter. If the delimiter is one of ([{< then
10300 the corresponding "close" character )]}> is used as the closing
10301 delimiter. It allows quoting of delimiters, and if the string has
10302 balanced delimiters ([{<>}]) it allows nesting.
10304 On success, the SV with the resulting string is put into lex_stuff or,
10305 if that is already non-NULL, into lex_repl. The second case occurs only
10306 when parsing the RHS of the special constructs s/// and tr/// (y///).
10307 For convenience, the terminating delimiter character is stuffed into
10312 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims, int re_reparse,
10313 bool deprecate_escaped_meta /* Should we issue a deprecation warning
10314 for certain paired metacharacters that
10315 appear escaped within it */
10319 SV *sv; /* scalar value: string */
10320 const char *tmps; /* temp string, used for delimiter matching */
10321 char *s = start; /* current position in the buffer */
10322 char term; /* terminating character */
10323 char *to; /* current position in the sv's data */
10324 I32 brackets = 1; /* bracket nesting level */
10325 bool has_utf8 = FALSE; /* is there any utf8 content? */
10326 I32 termcode; /* terminating char. code */
10327 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
10328 STRLEN termlen; /* length of terminating string */
10329 int last_off = 0; /* last position for nesting bracket */
10330 char *escaped_open = NULL;
10336 PERL_ARGS_ASSERT_SCAN_STR;
10338 /* skip space before the delimiter */
10344 if (PL_realtokenstart >= 0) {
10345 stuffstart = PL_realtokenstart;
10346 PL_realtokenstart = -1;
10349 stuffstart = start - SvPVX(PL_linestr);
10351 /* mark where we are, in case we need to report errors */
10354 /* after skipping whitespace, the next character is the terminator */
10357 termcode = termstr[0] = term;
10361 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
10362 Copy(s, termstr, termlen, U8);
10363 if (!UTF8_IS_INVARIANT(term))
10367 /* mark where we are */
10368 PL_multi_start = CopLINE(PL_curcop);
10369 PL_multi_open = term;
10371 /* find corresponding closing delimiter */
10372 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
10373 termcode = termstr[0] = term = tmps[5];
10375 PL_multi_close = term;
10377 /* A warning is raised if the input parameter requires it for escaped (by a
10378 * backslash) paired metacharacters {} [] and () when the delimiters are
10379 * those same characters, and the backslash is ineffective. This doesn't
10380 * happen for <>, as they aren't metas. */
10381 if (deprecate_escaped_meta
10382 && (PL_multi_open == PL_multi_close
10383 || ! ckWARN_d(WARN_DEPRECATED)
10384 || PL_multi_open == '<'))
10386 deprecate_escaped_meta = FALSE;
10389 /* create a new SV to hold the contents. 79 is the SV's initial length.
10390 What a random number. */
10391 sv = newSV_type(SVt_PVIV);
10393 SvIV_set(sv, termcode);
10394 (void)SvPOK_only(sv); /* validate pointer */
10396 /* move past delimiter and try to read a complete string */
10398 sv_catpvn(sv, s, termlen);
10401 tstart = SvPVX(PL_linestr) + stuffstart;
10402 if (PL_madskills && !PL_thisopen && !keep_delims) {
10403 PL_thisopen = newSVpvn(tstart, s - tstart);
10404 stuffstart = s - SvPVX(PL_linestr);
10408 if (PL_encoding && !UTF && !re_reparse) {
10412 int offset = s - SvPVX_const(PL_linestr);
10413 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
10414 &offset, (char*)termstr, termlen);
10415 const char * const ns = SvPVX_const(PL_linestr) + offset;
10416 char * const svlast = SvEND(sv) - 1;
10418 for (; s < ns; s++) {
10419 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10420 COPLINE_INC_WITH_HERELINES;
10423 goto read_more_line;
10425 /* handle quoted delimiters */
10426 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10428 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10430 if ((svlast-1 - t) % 2) {
10431 if (!keep_quoted) {
10432 *(svlast-1) = term;
10434 SvCUR_set(sv, SvCUR(sv) - 1);
10439 if (PL_multi_open == PL_multi_close) {
10445 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10446 /* At here, all closes are "was quoted" one,
10447 so we don't check PL_multi_close. */
10449 if (!keep_quoted && *(t+1) == PL_multi_open)
10454 else if (*t == PL_multi_open)
10462 SvCUR_set(sv, w - SvPVX_const(sv));
10464 last_off = w - SvPVX(sv);
10465 if (--brackets <= 0)
10470 if (!keep_delims) {
10471 SvCUR_set(sv, SvCUR(sv) - 1);
10477 /* extend sv if need be */
10478 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10479 /* set 'to' to the next character in the sv's string */
10480 to = SvPVX(sv)+SvCUR(sv);
10482 /* if open delimiter is the close delimiter read unbridle */
10483 if (PL_multi_open == PL_multi_close) {
10484 for (; s < PL_bufend; s++,to++) {
10485 /* embedded newlines increment the current line number */
10486 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10487 COPLINE_INC_WITH_HERELINES;
10488 /* handle quoted delimiters */
10489 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10492 || (re_reparse && s[1] == '\\'))
10495 /* any other quotes are simply copied straight through */
10499 /* terminate when run out of buffer (the for() condition), or
10500 have found the terminator */
10501 else if (*s == term) {
10504 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10507 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10513 /* if the terminator isn't the same as the start character (e.g.,
10514 matched brackets), we have to allow more in the quoting, and
10515 be prepared for nested brackets.
10518 /* read until we run out of string, or we find the terminator */
10519 for (; s < PL_bufend; s++,to++) {
10520 /* embedded newlines increment the line count */
10521 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10522 COPLINE_INC_WITH_HERELINES;
10523 /* backslashes can escape the open or closing characters */
10524 if (*s == '\\' && s+1 < PL_bufend) {
10525 if (!keep_quoted &&
10526 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10530 /* Here, 'deprecate_escaped_meta' is true iff the
10531 * delimiters are paired metacharacters, and 's' points
10532 * to an occurrence of one of them within the string,
10533 * which was preceded by a backslash. If this is a
10534 * context where the delimiter is also a metacharacter,
10535 * the backslash is useless, and deprecated. () and []
10536 * are meta in any context. {} are meta only when
10537 * appearing in a quantifier or in things like '\p{'.
10538 * They also aren't meta unless there is a matching
10539 * closed, escaped char later on within the string.
10540 * If 's' points to an open, set a flag; if to a close,
10541 * test that flag, and raise a warning if it was set */
10543 if (deprecate_escaped_meta) {
10544 if (*s == PL_multi_open) {
10548 else if (regcurly(s,
10549 TRUE /* Look for a closing
10551 || (s - start > 2 /* Look for e.g.
10553 && _generic_isCC(*(s-2), _CC_BACKSLASH_FOO_LBRACE_IS_META)))
10558 else if (escaped_open) {
10559 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
10560 "Useless use of '\\'; doesn't escape metacharacter '%c'", PL_multi_open);
10561 escaped_open = NULL;
10568 /* allow nested opens and closes */
10569 else if (*s == PL_multi_close && --brackets <= 0)
10571 else if (*s == PL_multi_open)
10573 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10578 /* terminate the copied string and update the sv's end-of-string */
10580 SvCUR_set(sv, to - SvPVX_const(sv));
10583 * this next chunk reads more into the buffer if we're not done yet
10587 break; /* handle case where we are done yet :-) */
10589 #ifndef PERL_STRICT_CR
10590 if (to - SvPVX_const(sv) >= 2) {
10591 if ((to[-2] == '\r' && to[-1] == '\n') ||
10592 (to[-2] == '\n' && to[-1] == '\r'))
10596 SvCUR_set(sv, to - SvPVX_const(sv));
10598 else if (to[-1] == '\r')
10601 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10606 /* if we're out of file, or a read fails, bail and reset the current
10607 line marker so we can report where the unterminated string began
10610 if (PL_madskills) {
10611 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10613 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10615 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10618 COPLINE_INC_WITH_HERELINES;
10619 PL_bufptr = PL_bufend;
10620 if (!lex_next_chunk(0)) {
10622 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10631 /* at this point, we have successfully read the delimited string */
10633 if (!PL_encoding || UTF || re_reparse) {
10635 if (PL_madskills) {
10636 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10637 const int len = s - tstart;
10639 sv_catpvn(PL_thisstuff, tstart, len);
10641 PL_thisstuff = newSVpvn(tstart, len);
10642 if (!PL_thisclose && !keep_delims)
10643 PL_thisclose = newSVpvn(s,termlen);
10648 sv_catpvn(sv, s, termlen);
10653 if (PL_madskills) {
10654 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10655 const int len = s - tstart - termlen;
10657 sv_catpvn(PL_thisstuff, tstart, len);
10659 PL_thisstuff = newSVpvn(tstart, len);
10660 if (!PL_thisclose && !keep_delims)
10661 PL_thisclose = newSVpvn(s - termlen,termlen);
10665 if (has_utf8 || (PL_encoding && !re_reparse))
10668 PL_multi_end = CopLINE(PL_curcop);
10670 /* if we allocated too much space, give some back */
10671 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10672 SvLEN_set(sv, SvCUR(sv) + 1);
10673 SvPV_renew(sv, SvLEN(sv));
10676 /* decide whether this is the first or second quoted string we've read
10681 PL_sublex_info.repl = sv;
10689 takes: pointer to position in buffer
10690 returns: pointer to new position in buffer
10691 side-effects: builds ops for the constant in pl_yylval.op
10693 Read a number in any of the formats that Perl accepts:
10695 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10696 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10699 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10701 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10704 If it reads a number without a decimal point or an exponent, it will
10705 try converting the number to an integer and see if it can do so
10706 without loss of precision.
10710 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10713 const char *s = start; /* current position in buffer */
10714 char *d; /* destination in temp buffer */
10715 char *e; /* end of temp buffer */
10716 NV nv; /* number read, as a double */
10717 SV *sv = NULL; /* place to put the converted number */
10718 bool floatit; /* boolean: int or float? */
10719 const char *lastub = NULL; /* position of last underbar */
10720 static const char* const number_too_long = "Number too long";
10722 PERL_ARGS_ASSERT_SCAN_NUM;
10724 /* We use the first character to decide what type of number this is */
10728 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10730 /* if it starts with a 0, it could be an octal number, a decimal in
10731 0.13 disguise, or a hexadecimal number, or a binary number. */
10735 u holds the "number so far"
10736 shift the power of 2 of the base
10737 (hex == 4, octal == 3, binary == 1)
10738 overflowed was the number more than we can hold?
10740 Shift is used when we add a digit. It also serves as an "are
10741 we in octal/hex/binary?" indicator to disallow hex characters
10742 when in octal mode.
10747 bool overflowed = FALSE;
10748 bool just_zero = TRUE; /* just plain 0 or binary number? */
10749 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10750 static const char* const bases[5] =
10751 { "", "binary", "", "octal", "hexadecimal" };
10752 static const char* const Bases[5] =
10753 { "", "Binary", "", "Octal", "Hexadecimal" };
10754 static const char* const maxima[5] =
10756 "0b11111111111111111111111111111111",
10760 const char *base, *Base, *max;
10762 /* check for hex */
10763 if (s[1] == 'x' || s[1] == 'X') {
10767 } else if (s[1] == 'b' || s[1] == 'B') {
10772 /* check for a decimal in disguise */
10773 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10775 /* so it must be octal */
10782 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10783 "Misplaced _ in number");
10787 base = bases[shift];
10788 Base = Bases[shift];
10789 max = maxima[shift];
10791 /* read the rest of the number */
10793 /* x is used in the overflow test,
10794 b is the digit we're adding on. */
10799 /* if we don't mention it, we're done */
10803 /* _ are ignored -- but warned about if consecutive */
10805 if (lastub && s == lastub + 1)
10806 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10807 "Misplaced _ in number");
10811 /* 8 and 9 are not octal */
10812 case '8': case '9':
10814 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10818 case '2': case '3': case '4':
10819 case '5': case '6': case '7':
10821 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10824 case '0': case '1':
10825 b = *s++ & 15; /* ASCII digit -> value of digit */
10829 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10830 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10831 /* make sure they said 0x */
10834 b = (*s++ & 7) + 9;
10836 /* Prepare to put the digit we have onto the end
10837 of the number so far. We check for overflows.
10843 x = u << shift; /* make room for the digit */
10845 if ((x >> shift) != u
10846 && !(PL_hints & HINT_NEW_BINARY)) {
10849 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10850 "Integer overflow in %s number",
10853 u = x | b; /* add the digit to the end */
10856 n *= nvshift[shift];
10857 /* If an NV has not enough bits in its
10858 * mantissa to represent an UV this summing of
10859 * small low-order numbers is a waste of time
10860 * (because the NV cannot preserve the
10861 * low-order bits anyway): we could just
10862 * remember when did we overflow and in the
10863 * end just multiply n by the right
10871 /* if we get here, we had success: make a scalar value from
10876 /* final misplaced underbar check */
10877 if (s[-1] == '_') {
10878 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10882 if (n > 4294967295.0)
10883 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10884 "%s number > %s non-portable",
10890 if (u > 0xffffffff)
10891 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10892 "%s number > %s non-portable",
10897 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10898 sv = new_constant(start, s - start, "integer",
10899 sv, NULL, NULL, 0);
10900 else if (PL_hints & HINT_NEW_BINARY)
10901 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10906 handle decimal numbers.
10907 we're also sent here when we read a 0 as the first digit
10909 case '1': case '2': case '3': case '4': case '5':
10910 case '6': case '7': case '8': case '9': case '.':
10913 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10916 /* read next group of digits and _ and copy into d */
10917 while (isDIGIT(*s) || *s == '_') {
10918 /* skip underscores, checking for misplaced ones
10922 if (lastub && s == lastub + 1)
10923 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10924 "Misplaced _ in number");
10928 /* check for end of fixed-length buffer */
10930 Perl_croak(aTHX_ "%s", number_too_long);
10931 /* if we're ok, copy the character */
10936 /* final misplaced underbar check */
10937 if (lastub && s == lastub + 1) {
10938 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10941 /* read a decimal portion if there is one. avoid
10942 3..5 being interpreted as the number 3. followed
10945 if (*s == '.' && s[1] != '.') {
10950 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10951 "Misplaced _ in number");
10955 /* copy, ignoring underbars, until we run out of digits.
10957 for (; isDIGIT(*s) || *s == '_'; s++) {
10958 /* fixed length buffer check */
10960 Perl_croak(aTHX_ "%s", number_too_long);
10962 if (lastub && s == lastub + 1)
10963 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10964 "Misplaced _ in number");
10970 /* fractional part ending in underbar? */
10971 if (s[-1] == '_') {
10972 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10973 "Misplaced _ in number");
10975 if (*s == '.' && isDIGIT(s[1])) {
10976 /* oops, it's really a v-string, but without the "v" */
10982 /* read exponent part, if present */
10983 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10987 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10988 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10990 /* stray preinitial _ */
10992 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10993 "Misplaced _ in number");
10997 /* allow positive or negative exponent */
10998 if (*s == '+' || *s == '-')
11001 /* stray initial _ */
11003 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
11004 "Misplaced _ in number");
11008 /* read digits of exponent */
11009 while (isDIGIT(*s) || *s == '_') {
11012 Perl_croak(aTHX_ "%s", number_too_long);
11016 if (((lastub && s == lastub + 1) ||
11017 (!isDIGIT(s[1]) && s[1] != '_')))
11018 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
11019 "Misplaced _ in number");
11027 We try to do an integer conversion first if no characters
11028 indicating "float" have been found.
11033 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
11035 if (flags == IS_NUMBER_IN_UV) {
11037 sv = newSViv(uv); /* Prefer IVs over UVs. */
11040 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
11041 if (uv <= (UV) IV_MIN)
11042 sv = newSViv(-(IV)uv);
11049 /* terminate the string */
11051 nv = Atof(PL_tokenbuf);
11056 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
11057 const char *const key = floatit ? "float" : "integer";
11058 const STRLEN keylen = floatit ? 5 : 7;
11059 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
11060 key, keylen, sv, NULL, NULL, 0);
11064 /* if it starts with a v, it could be a v-string */
11067 sv = newSV(5); /* preallocate storage space */
11068 ENTER_with_name("scan_vstring");
11070 s = scan_vstring(s, PL_bufend, sv);
11071 SvREFCNT_inc_simple_void_NN(sv);
11072 LEAVE_with_name("scan_vstring");
11076 /* make the op for the constant and return */
11079 lvalp->opval = newSVOP(OP_CONST, 0, sv);
11081 lvalp->opval = NULL;
11087 S_scan_formline(pTHX_ char *s)
11092 SV * const stuff = newSVpvs("");
11093 bool needargs = FALSE;
11094 bool eofmt = FALSE;
11096 char *tokenstart = s;
11097 SV* savewhite = NULL;
11099 if (PL_madskills) {
11100 savewhite = PL_thiswhite;
11105 PERL_ARGS_ASSERT_SCAN_FORMLINE;
11107 while (!needargs) {
11110 #ifdef PERL_STRICT_CR
11111 while (SPACE_OR_TAB(*t))
11114 while (SPACE_OR_TAB(*t) || *t == '\r')
11117 if (*t == '\n' || t == PL_bufend) {
11122 eol = (char *) memchr(s,'\n',PL_bufend-s);
11126 for (t = s; t < eol; t++) {
11127 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
11129 goto enough; /* ~~ must be first line in formline */
11131 if (*t == '@' || *t == '^')
11135 sv_catpvn(stuff, s, eol-s);
11136 #ifndef PERL_STRICT_CR
11137 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
11138 char *end = SvPVX(stuff) + SvCUR(stuff);
11141 SvCUR_set(stuff, SvCUR(stuff) - 1);
11149 if ((PL_rsfp || PL_parser->filtered)
11150 && PL_parser->form_lex_state == LEX_NORMAL) {
11153 if (PL_madskills) {
11155 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
11157 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
11160 PL_bufptr = PL_bufend;
11161 COPLINE_INC_WITH_HERELINES;
11162 got_some = lex_next_chunk(0);
11163 CopLINE_dec(PL_curcop);
11166 tokenstart = PL_bufptr;
11174 if (!SvCUR(stuff) || needargs)
11175 PL_lex_state = PL_parser->form_lex_state;
11176 if (SvCUR(stuff)) {
11177 PL_expect = XSTATE;
11179 start_force(PL_curforce);
11180 NEXTVAL_NEXTTOKE.ival = 0;
11181 force_next(FORMLBRACK);
11184 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
11186 else if (PL_encoding)
11187 sv_recode_to_utf8(stuff, PL_encoding);
11189 start_force(PL_curforce);
11190 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
11194 SvREFCNT_dec(stuff);
11196 PL_lex_formbrack = 0;
11199 if (PL_madskills) {
11201 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
11203 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
11204 PL_thiswhite = savewhite;
11211 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
11214 const I32 oldsavestack_ix = PL_savestack_ix;
11215 CV* const outsidecv = PL_compcv;
11217 SAVEI32(PL_subline);
11218 save_item(PL_subname);
11219 SAVESPTR(PL_compcv);
11221 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
11222 CvFLAGS(PL_compcv) |= flags;
11224 PL_subline = CopLINE(PL_curcop);
11225 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
11226 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
11227 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
11228 if (outsidecv && CvPADLIST(outsidecv))
11229 CvPADLIST(PL_compcv)->xpadl_outid =
11230 PadlistNAMES(CvPADLIST(outsidecv));
11232 return oldsavestack_ix;
11236 #pragma segment Perl_yylex
11239 S_yywarn(pTHX_ const char *const s, U32 flags)
11243 PERL_ARGS_ASSERT_YYWARN;
11245 PL_in_eval |= EVAL_WARNONLY;
11246 yyerror_pv(s, flags);
11247 PL_in_eval &= ~EVAL_WARNONLY;
11252 Perl_yyerror(pTHX_ const char *const s)
11254 PERL_ARGS_ASSERT_YYERROR;
11255 return yyerror_pvn(s, strlen(s), 0);
11259 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
11261 PERL_ARGS_ASSERT_YYERROR_PV;
11262 return yyerror_pvn(s, strlen(s), flags);
11266 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
11269 const char *context = NULL;
11272 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
11273 int yychar = PL_parser->yychar;
11275 PERL_ARGS_ASSERT_YYERROR_PVN;
11277 if (!yychar || (yychar == ';' && !PL_rsfp))
11278 sv_catpvs(where_sv, "at EOF");
11279 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
11280 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
11281 PL_oldbufptr != PL_bufptr) {
11284 The code below is removed for NetWare because it abends/crashes on NetWare
11285 when the script has error such as not having the closing quotes like:
11286 if ($var eq "value)
11287 Checking of white spaces is anyway done in NetWare code.
11290 while (isSPACE(*PL_oldoldbufptr))
11293 context = PL_oldoldbufptr;
11294 contlen = PL_bufptr - PL_oldoldbufptr;
11296 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
11297 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
11300 The code below is removed for NetWare because it abends/crashes on NetWare
11301 when the script has error such as not having the closing quotes like:
11302 if ($var eq "value)
11303 Checking of white spaces is anyway done in NetWare code.
11306 while (isSPACE(*PL_oldbufptr))
11309 context = PL_oldbufptr;
11310 contlen = PL_bufptr - PL_oldbufptr;
11312 else if (yychar > 255)
11313 sv_catpvs(where_sv, "next token ???");
11314 else if (yychar == -2) { /* YYEMPTY */
11315 if (PL_lex_state == LEX_NORMAL ||
11316 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11317 sv_catpvs(where_sv, "at end of line");
11318 else if (PL_lex_inpat)
11319 sv_catpvs(where_sv, "within pattern");
11321 sv_catpvs(where_sv, "within string");
11324 sv_catpvs(where_sv, "next char ");
11326 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11327 else if (isPRINT_LC(yychar)) {
11328 const char string = yychar;
11329 sv_catpvn(where_sv, &string, 1);
11332 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11334 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
11335 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11336 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
11338 Perl_sv_catpvf(aTHX_ msg, "near \"%"SVf"\"\n",
11339 SVfARG(newSVpvn_flags(context, contlen,
11340 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
11342 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11343 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11344 Perl_sv_catpvf(aTHX_ msg,
11345 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11346 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11349 if (PL_in_eval & EVAL_WARNONLY) {
11350 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11354 if (PL_error_count >= 10) {
11356 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
11357 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11358 SVfARG(errsv), OutCopFILE(PL_curcop));
11360 Perl_croak(aTHX_ "%s has too many errors.\n",
11361 OutCopFILE(PL_curcop));
11364 PL_in_my_stash = NULL;
11368 #pragma segment Main
11372 S_swallow_bom(pTHX_ U8 *s)
11375 const STRLEN slen = SvCUR(PL_linestr);
11377 PERL_ARGS_ASSERT_SWALLOW_BOM;
11381 if (s[1] == 0xFE) {
11382 /* UTF-16 little-endian? (or UTF-32LE?) */
11383 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11384 /* diag_listed_as: Unsupported script encoding %s */
11385 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11386 #ifndef PERL_NO_UTF16_FILTER
11387 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11389 if (PL_bufend > (char*)s) {
11390 s = add_utf16_textfilter(s, TRUE);
11393 /* diag_listed_as: Unsupported script encoding %s */
11394 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11399 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11400 #ifndef PERL_NO_UTF16_FILTER
11401 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11403 if (PL_bufend > (char *)s) {
11404 s = add_utf16_textfilter(s, FALSE);
11407 /* diag_listed_as: Unsupported script encoding %s */
11408 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11413 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
11414 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11415 s += 3; /* UTF-8 */
11421 if (s[2] == 0xFE && s[3] == 0xFF) {
11422 /* UTF-32 big-endian */
11423 /* diag_listed_as: Unsupported script encoding %s */
11424 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11427 else if (s[2] == 0 && s[3] != 0) {
11430 * are a good indicator of UTF-16BE. */
11431 #ifndef PERL_NO_UTF16_FILTER
11432 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11433 s = add_utf16_textfilter(s, FALSE);
11435 /* diag_listed_as: Unsupported script encoding %s */
11436 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11442 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
11443 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11444 s += 4; /* UTF-8 */
11450 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11453 * are a good indicator of UTF-16LE. */
11454 #ifndef PERL_NO_UTF16_FILTER
11455 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11456 s = add_utf16_textfilter(s, TRUE);
11458 /* diag_listed_as: Unsupported script encoding %s */
11459 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11467 #ifndef PERL_NO_UTF16_FILTER
11469 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11472 SV *const filter = FILTER_DATA(idx);
11473 /* We re-use this each time round, throwing the contents away before we
11475 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11476 SV *const utf8_buffer = filter;
11477 IV status = IoPAGE(filter);
11478 const bool reverse = cBOOL(IoLINES(filter));
11481 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11483 /* As we're automatically added, at the lowest level, and hence only called
11484 from this file, we can be sure that we're not called in block mode. Hence
11485 don't bother writing code to deal with block mode. */
11487 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11490 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11492 DEBUG_P(PerlIO_printf(Perl_debug_log,
11493 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11494 FPTR2DPTR(void *, S_utf16_textfilter),
11495 reverse ? 'l' : 'b', idx, maxlen, status,
11496 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11503 /* First, look in our buffer of existing UTF-8 data: */
11504 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11508 } else if (status == 0) {
11510 IoPAGE(filter) = 0;
11511 nl = SvEND(utf8_buffer);
11514 STRLEN got = nl - SvPVX(utf8_buffer);
11515 /* Did we have anything to append? */
11517 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11518 /* Everything else in this code works just fine if SVp_POK isn't
11519 set. This, however, needs it, and we need it to work, else
11520 we loop infinitely because the buffer is never consumed. */
11521 sv_chop(utf8_buffer, nl);
11525 /* OK, not a complete line there, so need to read some more UTF-16.
11526 Read an extra octect if the buffer currently has an odd number. */
11530 if (SvCUR(utf16_buffer) >= 2) {
11531 /* Location of the high octet of the last complete code point.
11532 Gosh, UTF-16 is a pain. All the benefits of variable length,
11533 *coupled* with all the benefits of partial reads and
11535 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11536 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11538 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11542 /* We have the first half of a surrogate. Read more. */
11543 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11546 status = FILTER_READ(idx + 1, utf16_buffer,
11547 160 + (SvCUR(utf16_buffer) & 1));
11548 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11549 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11552 IoPAGE(filter) = status;
11557 chars = SvCUR(utf16_buffer) >> 1;
11558 have = SvCUR(utf8_buffer);
11559 SvGROW(utf8_buffer, have + chars * 3 + 1);
11562 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11563 (U8*)SvPVX_const(utf8_buffer) + have,
11564 chars * 2, &newlen);
11566 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11567 (U8*)SvPVX_const(utf8_buffer) + have,
11568 chars * 2, &newlen);
11570 SvCUR_set(utf8_buffer, have + newlen);
11573 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11574 it's private to us, and utf16_to_utf8{,reversed} take a
11575 (pointer,length) pair, rather than a NUL-terminated string. */
11576 if(SvCUR(utf16_buffer) & 1) {
11577 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11578 SvCUR_set(utf16_buffer, 1);
11580 SvCUR_set(utf16_buffer, 0);
11583 DEBUG_P(PerlIO_printf(Perl_debug_log,
11584 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11586 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11587 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11592 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11594 SV *filter = filter_add(S_utf16_textfilter, NULL);
11596 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11598 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11599 sv_setpvs(filter, "");
11600 IoLINES(filter) = reversed;
11601 IoPAGE(filter) = 1; /* Not EOF */
11603 /* Sadly, we have to return a valid pointer, come what may, so we have to
11604 ignore any error return from this. */
11605 SvCUR_set(PL_linestr, 0);
11606 if (FILTER_READ(0, PL_linestr, 0)) {
11607 SvUTF8_on(PL_linestr);
11609 SvUTF8_on(PL_linestr);
11611 PL_bufend = SvEND(PL_linestr);
11612 return (U8*)SvPVX(PL_linestr);
11617 Returns a pointer to the next character after the parsed
11618 vstring, as well as updating the passed in sv.
11620 Function must be called like
11622 sv = sv_2mortal(newSV(5));
11623 s = scan_vstring(s,e,sv);
11625 where s and e are the start and end of the string.
11626 The sv should already be large enough to store the vstring
11627 passed in, for performance reasons.
11629 This function may croak if fatal warnings are enabled in the
11630 calling scope, hence the sv_2mortal in the example (to prevent
11631 a leak). Make sure to do SvREFCNT_inc afterwards if you use
11637 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11640 const char *pos = s;
11641 const char *start = s;
11643 PERL_ARGS_ASSERT_SCAN_VSTRING;
11645 if (*pos == 'v') pos++; /* get past 'v' */
11646 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11648 if ( *pos != '.') {
11649 /* this may not be a v-string if followed by => */
11650 const char *next = pos;
11651 while (next < e && isSPACE(*next))
11653 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11654 /* return string not v-string */
11655 sv_setpvn(sv,(char *)s,pos-s);
11656 return (char *)pos;
11660 if (!isALPHA(*pos)) {
11661 U8 tmpbuf[UTF8_MAXBYTES+1];
11664 s++; /* get past 'v' */
11669 /* this is atoi() that tolerates underscores */
11672 const char *end = pos;
11674 while (--end >= s) {
11676 const UV orev = rev;
11677 rev += (*end - '0') * mult;
11680 /* diag_listed_as: Integer overflow in %s number */
11681 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11682 "Integer overflow in decimal number");
11686 if (rev > 0x7FFFFFFF)
11687 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11689 /* Append native character for the rev point */
11690 tmpend = uvchr_to_utf8(tmpbuf, rev);
11691 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11692 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11694 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11700 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11704 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11711 Perl_keyword_plugin_standard(pTHX_
11712 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11714 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11715 PERL_UNUSED_CONTEXT;
11716 PERL_UNUSED_ARG(keyword_ptr);
11717 PERL_UNUSED_ARG(keyword_len);
11718 PERL_UNUSED_ARG(op_ptr);
11719 return KEYWORD_PLUGIN_DECLINE;
11722 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11724 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11726 SAVEI32(PL_lex_brackets);
11727 if (PL_lex_brackets > 100)
11728 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11729 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11730 SAVEI32(PL_lex_allbrackets);
11731 PL_lex_allbrackets = 0;
11732 SAVEI8(PL_lex_fakeeof);
11733 PL_lex_fakeeof = (U8)fakeeof;
11734 if(yyparse(gramtype) && !PL_parser->error_count)
11735 qerror(Perl_mess(aTHX_ "Parse error"));
11738 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11740 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11744 SAVEVPTR(PL_eval_root);
11745 PL_eval_root = NULL;
11746 parse_recdescent(gramtype, fakeeof);
11752 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11754 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11757 if (flags & ~PARSE_OPTIONAL)
11758 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11759 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11760 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11761 if (!PL_parser->error_count)
11762 qerror(Perl_mess(aTHX_ "Parse error"));
11763 exprop = newOP(OP_NULL, 0);
11769 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11771 Parse a Perl arithmetic expression. This may contain operators of precedence
11772 down to the bit shift operators. The expression must be followed (and thus
11773 terminated) either by a comparison or lower-precedence operator or by
11774 something that would normally terminate an expression such as semicolon.
11775 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11776 otherwise it is mandatory. It is up to the caller to ensure that the
11777 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11778 the source of the code to be parsed and the lexical context for the
11781 The op tree representing the expression is returned. If an optional
11782 expression is absent, a null pointer is returned, otherwise the pointer
11785 If an error occurs in parsing or compilation, in most cases a valid op
11786 tree is returned anyway. The error is reflected in the parser state,
11787 normally resulting in a single exception at the top level of parsing
11788 which covers all the compilation errors that occurred. Some compilation
11789 errors, however, will throw an exception immediately.
11795 Perl_parse_arithexpr(pTHX_ U32 flags)
11797 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11801 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11803 Parse a Perl term expression. This may contain operators of precedence
11804 down to the assignment operators. The expression must be followed (and thus
11805 terminated) either by a comma or lower-precedence operator or by
11806 something that would normally terminate an expression such as semicolon.
11807 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11808 otherwise it is mandatory. It is up to the caller to ensure that the
11809 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11810 the source of the code to be parsed and the lexical context for the
11813 The op tree representing the expression is returned. If an optional
11814 expression is absent, a null pointer is returned, otherwise the pointer
11817 If an error occurs in parsing or compilation, in most cases a valid op
11818 tree is returned anyway. The error is reflected in the parser state,
11819 normally resulting in a single exception at the top level of parsing
11820 which covers all the compilation errors that occurred. Some compilation
11821 errors, however, will throw an exception immediately.
11827 Perl_parse_termexpr(pTHX_ U32 flags)
11829 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11833 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11835 Parse a Perl list expression. This may contain operators of precedence
11836 down to the comma operator. The expression must be followed (and thus
11837 terminated) either by a low-precedence logic operator such as C<or> or by
11838 something that would normally terminate an expression such as semicolon.
11839 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11840 otherwise it is mandatory. It is up to the caller to ensure that the
11841 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11842 the source of the code to be parsed and the lexical context for the
11845 The op tree representing the expression is returned. If an optional
11846 expression is absent, a null pointer is returned, otherwise the pointer
11849 If an error occurs in parsing or compilation, in most cases a valid op
11850 tree is returned anyway. The error is reflected in the parser state,
11851 normally resulting in a single exception at the top level of parsing
11852 which covers all the compilation errors that occurred. Some compilation
11853 errors, however, will throw an exception immediately.
11859 Perl_parse_listexpr(pTHX_ U32 flags)
11861 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11865 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11867 Parse a single complete Perl expression. This allows the full
11868 expression grammar, including the lowest-precedence operators such
11869 as C<or>. The expression must be followed (and thus terminated) by a
11870 token that an expression would normally be terminated by: end-of-file,
11871 closing bracketing punctuation, semicolon, or one of the keywords that
11872 signals a postfix expression-statement modifier. If I<flags> includes
11873 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11874 mandatory. It is up to the caller to ensure that the dynamic parser
11875 state (L</PL_parser> et al) is correctly set to reflect the source of
11876 the code to be parsed and the lexical context for the expression.
11878 The op tree representing the expression is returned. If an optional
11879 expression is absent, a null pointer is returned, otherwise the pointer
11882 If an error occurs in parsing or compilation, in most cases a valid op
11883 tree is returned anyway. The error is reflected in the parser state,
11884 normally resulting in a single exception at the top level of parsing
11885 which covers all the compilation errors that occurred. Some compilation
11886 errors, however, will throw an exception immediately.
11892 Perl_parse_fullexpr(pTHX_ U32 flags)
11894 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11898 =for apidoc Amx|OP *|parse_block|U32 flags
11900 Parse a single complete Perl code block. This consists of an opening
11901 brace, a sequence of statements, and a closing brace. The block
11902 constitutes a lexical scope, so C<my> variables and various compile-time
11903 effects can be contained within it. It is up to the caller to ensure
11904 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11905 reflect the source of the code to be parsed and the lexical context for
11908 The op tree representing the code block is returned. This is always a
11909 real op, never a null pointer. It will normally be a C<lineseq> list,
11910 including C<nextstate> or equivalent ops. No ops to construct any kind
11911 of runtime scope are included by virtue of it being a block.
11913 If an error occurs in parsing or compilation, in most cases a valid op
11914 tree (most likely null) is returned anyway. The error is reflected in
11915 the parser state, normally resulting in a single exception at the top
11916 level of parsing which covers all the compilation errors that occurred.
11917 Some compilation errors, however, will throw an exception immediately.
11919 The I<flags> parameter is reserved for future use, and must always
11926 Perl_parse_block(pTHX_ U32 flags)
11929 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11930 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11934 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11936 Parse a single unadorned Perl statement. This may be a normal imperative
11937 statement or a declaration that has compile-time effect. It does not
11938 include any label or other affixture. It is up to the caller to ensure
11939 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11940 reflect the source of the code to be parsed and the lexical context for
11943 The op tree representing the statement is returned. This may be a
11944 null pointer if the statement is null, for example if it was actually
11945 a subroutine definition (which has compile-time side effects). If not
11946 null, it will be ops directly implementing the statement, suitable to
11947 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11948 equivalent op (except for those embedded in a scope contained entirely
11949 within the statement).
11951 If an error occurs in parsing or compilation, in most cases a valid op
11952 tree (most likely null) is returned anyway. The error is reflected in
11953 the parser state, normally resulting in a single exception at the top
11954 level of parsing which covers all the compilation errors that occurred.
11955 Some compilation errors, however, will throw an exception immediately.
11957 The I<flags> parameter is reserved for future use, and must always
11964 Perl_parse_barestmt(pTHX_ U32 flags)
11967 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11968 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11972 =for apidoc Amx|SV *|parse_label|U32 flags
11974 Parse a single label, possibly optional, of the type that may prefix a
11975 Perl statement. It is up to the caller to ensure that the dynamic parser
11976 state (L</PL_parser> et al) is correctly set to reflect the source of
11977 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11978 label is optional, otherwise it is mandatory.
11980 The name of the label is returned in the form of a fresh scalar. If an
11981 optional label is absent, a null pointer is returned.
11983 If an error occurs in parsing, which can only occur if the label is
11984 mandatory, a valid label is returned anyway. The error is reflected in
11985 the parser state, normally resulting in a single exception at the top
11986 level of parsing which covers all the compilation errors that occurred.
11992 Perl_parse_label(pTHX_ U32 flags)
11994 if (flags & ~PARSE_OPTIONAL)
11995 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11996 if (PL_lex_state == LEX_KNOWNEXT) {
11997 PL_parser->yychar = yylex();
11998 if (PL_parser->yychar == LABEL) {
11999 char * const lpv = pl_yylval.pval;
12000 STRLEN llen = strlen(lpv);
12001 PL_parser->yychar = YYEMPTY;
12002 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
12009 STRLEN wlen, bufptr_pos;
12012 if (!isIDFIRST_lazy_if(s, UTF))
12014 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
12015 if (word_takes_any_delimeter(s, wlen))
12017 bufptr_pos = s - SvPVX(PL_linestr);
12019 lex_read_space(LEX_KEEP_PREVIOUS);
12021 s = SvPVX(PL_linestr) + bufptr_pos;
12022 if (t[0] == ':' && t[1] != ':') {
12023 PL_oldoldbufptr = PL_oldbufptr;
12026 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
12030 if (flags & PARSE_OPTIONAL) {
12033 qerror(Perl_mess(aTHX_ "Parse error"));
12034 return newSVpvs("x");
12041 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
12043 Parse a single complete Perl statement. This may be a normal imperative
12044 statement or a declaration that has compile-time effect, and may include
12045 optional labels. It is up to the caller to ensure that the dynamic
12046 parser state (L</PL_parser> et al) is correctly set to reflect the source
12047 of the code to be parsed and the lexical context for the statement.
12049 The op tree representing the statement is returned. This may be a
12050 null pointer if the statement is null, for example if it was actually
12051 a subroutine definition (which has compile-time side effects). If not
12052 null, it will be the result of a L</newSTATEOP> call, normally including
12053 a C<nextstate> or equivalent op.
12055 If an error occurs in parsing or compilation, in most cases a valid op
12056 tree (most likely null) is returned anyway. The error is reflected in
12057 the parser state, normally resulting in a single exception at the top
12058 level of parsing which covers all the compilation errors that occurred.
12059 Some compilation errors, however, will throw an exception immediately.
12061 The I<flags> parameter is reserved for future use, and must always
12068 Perl_parse_fullstmt(pTHX_ U32 flags)
12071 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
12072 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
12076 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
12078 Parse a sequence of zero or more Perl statements. These may be normal
12079 imperative statements, including optional labels, or declarations
12080 that have compile-time effect, or any mixture thereof. The statement
12081 sequence ends when a closing brace or end-of-file is encountered in a
12082 place where a new statement could have validly started. It is up to
12083 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
12084 is correctly set to reflect the source of the code to be parsed and the
12085 lexical context for the statements.
12087 The op tree representing the statement sequence is returned. This may
12088 be a null pointer if the statements were all null, for example if there
12089 were no statements or if there were only subroutine definitions (which
12090 have compile-time side effects). If not null, it will be a C<lineseq>
12091 list, normally including C<nextstate> or equivalent ops.
12093 If an error occurs in parsing or compilation, in most cases a valid op
12094 tree is returned anyway. The error is reflected in the parser state,
12095 normally resulting in a single exception at the top level of parsing
12096 which covers all the compilation errors that occurred. Some compilation
12097 errors, however, will throw an exception immediately.
12099 The I<flags> parameter is reserved for future use, and must always
12106 Perl_parse_stmtseq(pTHX_ U32 flags)
12111 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
12112 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
12113 c = lex_peek_unichar(0);
12114 if (c != -1 && c != /*{*/'}')
12115 qerror(Perl_mess(aTHX_ "Parse error"));
12121 * c-indentation-style: bsd
12122 * c-basic-offset: 4
12123 * indent-tabs-mode: nil
12126 * ex: set ts=8 sts=4 sw=4 et: