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 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; (isALNUM_lazy_if(t,UTF) || *t == ':');
548 t += UTF ? UTF8SKIP(t) : 1)
550 if (t < PL_bufptr && isSPACE(*t))
551 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
552 "\t(Do you need to predeclare %"SVf"?)\n",
553 SVfARG(newSVpvn_flags(PL_oldoldbufptr, (STRLEN)(t - PL_oldoldbufptr),
554 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
558 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
559 "\t(Missing operator before %"SVf"?)\n",
560 SVfARG(newSVpvn_flags(oldbp, (STRLEN)(s - oldbp),
561 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
569 * Complain about missing quote/regexp/heredoc terminator.
570 * If it's called with NULL then it cauterizes the line buffer.
571 * If we're in a delimited string and the delimiter is a control
572 * character, it's reformatted into a two-char sequence like ^C.
577 S_missingterm(pTHX_ char *s)
583 char * const nl = strrchr(s,'\n');
587 else if (isCNTRL(PL_multi_close)) {
589 tmpbuf[1] = (char)toCTRL(PL_multi_close);
594 *tmpbuf = (char)PL_multi_close;
598 q = strchr(s,'"') ? '\'' : '"';
599 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
605 * Check whether the named feature is enabled.
608 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
611 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
613 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
615 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
617 if (namelen > MAX_FEATURE_LEN)
619 memcpy(&he_name[8], name, namelen);
621 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
622 REFCOUNTED_HE_EXISTS));
626 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
627 * utf16-to-utf8-reversed.
630 #ifdef PERL_CR_FILTER
634 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_ register 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_ register 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_ register 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_ register 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 (isALNUM_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_ register 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 $foo *foo &foo etc, but the program
2155 * text only contains the "foo" portion. The first argument is a pointer
2156 * to the "foo", and the second argument is the type symbol to prefix.
2157 * Forces the next token to be a "WORD".
2158 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2162 S_force_ident(pTHX_ register const char *s, int kind)
2166 PERL_ARGS_ASSERT_FORCE_IDENT;
2169 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2170 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2171 UTF ? SVf_UTF8 : 0));
2172 start_force(PL_curforce);
2173 NEXTVAL_NEXTTOKE.opval = o;
2176 o->op_private = OPpCONST_ENTERED;
2177 /* XXX see note in pp_entereval() for why we forgo typo
2178 warnings if the symbol must be introduced in an eval.
2180 gv_fetchpvn_flags(s, len,
2181 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2182 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2183 kind == '$' ? 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);
2695 /* See if the charnames handler is the Perl core's, and if so, we can skip
2696 * the validation needed for a user-supplied one, as Perl's does its own
2698 table = GvHV(PL_hintgv); /* ^H */
2699 cvp = hv_fetchs(table, "charnames", FALSE);
2701 if (((rv = SvRV(cv)) != NULL)
2702 && ((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)) {
2731 /* Similarly for utf8. For invariants can check directly; for other
2732 * Latin1, can calculate their code point and check; otherwise use a
2734 if (UTF8_IS_INVARIANT(*s)) {
2735 if (! isALPHAU(*s)) {
2739 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2740 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s, *(s+1))))) {
2746 if (! PL_utf8_charname_begin) {
2747 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2748 PL_utf8_charname_begin = _core_swash_init("utf8",
2749 "_Perl_Charname_Begin",
2751 1, 0, NULL, &flags);
2753 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2760 if (UTF8_IS_INVARIANT(*s)) {
2761 if (! isCHARNAME_CONT(*s)) {
2766 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2767 if (! isCHARNAME_CONT(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s,
2775 if (! PL_utf8_charname_continue) {
2776 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2777 PL_utf8_charname_continue = _core_swash_init("utf8",
2778 "_Perl_Charname_Continue",
2780 1, 0, NULL, &flags);
2782 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2790 if (SvUTF8(res)) { /* Don't accept malformed input */
2791 const U8* first_bad_char_loc;
2793 const char* const str = SvPV_const(res, len);
2794 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2795 /* If warnings are on, this will print a more detailed analysis of
2796 * what is wrong than the error message below */
2797 utf8n_to_uvuni(first_bad_char_loc,
2798 (char *) first_bad_char_loc - str,
2801 /* We deliberately don't try to print the malformed character,
2802 * which might not print very well; it also may be just the first
2803 * of many malformations, so don't print what comes after it */
2806 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2807 (int) (e - backslash_ptr + 1), backslash_ptr,
2808 (int) ((char *) first_bad_char_loc - str), str
2818 int bad_char_size = ((UTF) ? UTF8SKIP(s) : 1);
2820 /* The final %.*s makes sure that should the trailing NUL be missing
2821 * that this print won't run off the end of the string */
2824 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2825 (int)(s - backslash_ptr + bad_char_size), backslash_ptr,
2826 (int)(e - s + bad_char_size), s + bad_char_size
2828 UTF ? SVf_UTF8 : 0);
2836 Extracts the next constant part of a pattern, double-quoted string,
2837 or transliteration. This is terrifying code.
2839 For example, in parsing the double-quoted string "ab\x63$d", it would
2840 stop at the '$' and return an OP_CONST containing 'abc'.
2842 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2843 processing a pattern (PL_lex_inpat is true), a transliteration
2844 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2846 Returns a pointer to the character scanned up to. If this is
2847 advanced from the start pointer supplied (i.e. if anything was
2848 successfully parsed), will leave an OP_CONST for the substring scanned
2849 in pl_yylval. Caller must intuit reason for not parsing further
2850 by looking at the next characters herself.
2854 \N{ABC} => \N{U+41.42.43}
2857 all other \-char, including \N and \N{ apart from \N{ABC}
2860 @ and $ where it appears to be a var, but not for $ as tail anchor
2865 In transliterations:
2866 characters are VERY literal, except for - not at the start or end
2867 of the string, which indicates a range. If the range is in bytes,
2868 scan_const expands the range to the full set of intermediate
2869 characters. If the range is in utf8, the hyphen is replaced with
2870 a certain range mark which will be handled by pmtrans() in op.c.
2872 In double-quoted strings:
2874 double-quoted style: \r and \n
2875 constants: \x31, etc.
2876 deprecated backrefs: \1 (in substitution replacements)
2877 case and quoting: \U \Q \E
2880 scan_const does *not* construct ops to handle interpolated strings.
2881 It stops processing as soon as it finds an embedded $ or @ variable
2882 and leaves it to the caller to work out what's going on.
2884 embedded arrays (whether in pattern or not) could be:
2885 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2887 $ in double-quoted strings must be the symbol of an embedded scalar.
2889 $ in pattern could be $foo or could be tail anchor. Assumption:
2890 it's a tail anchor if $ is the last thing in the string, or if it's
2891 followed by one of "()| \r\n\t"
2893 \1 (backreferences) are turned into $1 in substitutions
2895 The structure of the code is
2896 while (there's a character to process) {
2897 handle transliteration ranges
2898 skip regexp comments /(?#comment)/ and codes /(?{code})/
2899 skip #-initiated comments in //x patterns
2900 check for embedded arrays
2901 check for embedded scalars
2903 deprecate \1 in substitution replacements
2904 handle string-changing backslashes \l \U \Q \E, etc.
2905 switch (what was escaped) {
2906 handle \- in a transliteration (becomes a literal -)
2907 if a pattern and not \N{, go treat as regular character
2908 handle \132 (octal characters)
2909 handle \x15 and \x{1234} (hex characters)
2910 handle \N{name} (named characters, also \N{3,5} in a pattern)
2911 handle \cV (control characters)
2912 handle printf-style backslashes (\f, \r, \n, etc)
2915 } (end if backslash)
2916 handle regular character
2917 } (end while character to read)
2922 S_scan_const(pTHX_ char *start)
2925 char *send = PL_bufend; /* end of the constant */
2926 SV *sv = newSV(send - start); /* sv for the constant. See
2927 note below on sizing. */
2928 char *s = start; /* start of the constant */
2929 char *d = SvPVX(sv); /* destination for copies */
2930 bool dorange = FALSE; /* are we in a translit range? */
2931 bool didrange = FALSE; /* did we just finish a range? */
2932 bool in_charclass = FALSE; /* within /[...]/ */
2933 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2934 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2935 to be UTF8? But, this can
2936 show as true when the source
2937 isn't utf8, as for example
2938 when it is entirely composed
2940 SV *res; /* result from charnames */
2942 /* Note on sizing: The scanned constant is placed into sv, which is
2943 * initialized by newSV() assuming one byte of output for every byte of
2944 * input. This routine expects newSV() to allocate an extra byte for a
2945 * trailing NUL, which this routine will append if it gets to the end of
2946 * the input. There may be more bytes of input than output (eg., \N{LATIN
2947 * CAPITAL LETTER A}), or more output than input if the constant ends up
2948 * recoded to utf8, but each time a construct is found that might increase
2949 * the needed size, SvGROW() is called. Its size parameter each time is
2950 * based on the best guess estimate at the time, namely the length used so
2951 * far, plus the length the current construct will occupy, plus room for
2952 * the trailing NUL, plus one byte for every input byte still unscanned */
2956 UV literal_endpoint = 0;
2957 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2960 PERL_ARGS_ASSERT_SCAN_CONST;
2962 assert(PL_lex_inwhat != OP_TRANSR);
2963 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2964 /* If we are doing a trans and we know we want UTF8 set expectation */
2965 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2966 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2969 /* Protect sv from errors and fatal warnings. */
2970 ENTER_with_name("scan_const");
2973 while (s < send || dorange) {
2975 /* get transliterations out of the way (they're most literal) */
2976 if (PL_lex_inwhat == OP_TRANS) {
2977 /* expand a range A-Z to the full set of characters. AIE! */
2979 I32 i; /* current expanded character */
2980 I32 min; /* first character in range */
2981 I32 max; /* last character in range */
2992 char * const c = (char*)utf8_hop((U8*)d, -1);
2996 *c = (char)UTF_TO_NATIVE(0xff);
2997 /* mark the range as done, and continue */
3003 i = d - SvPVX_const(sv); /* remember current offset */
3006 SvLEN(sv) + (has_utf8 ?
3007 (512 - UTF_CONTINUATION_MARK +
3010 /* How many two-byte within 0..255: 128 in UTF-8,
3011 * 96 in UTF-8-mod. */
3013 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
3015 d = SvPVX(sv) + i; /* refresh d after realloc */
3019 for (j = 0; j <= 1; j++) {
3020 char * const c = (char*)utf8_hop((U8*)d, -1);
3021 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
3027 max = (U8)0xff; /* only to \xff */
3028 uvmax = uv; /* \x{100} to uvmax */
3030 d = c; /* eat endpoint chars */
3035 d -= 2; /* eat the first char and the - */
3036 min = (U8)*d; /* first char in range */
3037 max = (U8)d[1]; /* last char in range */
3044 "Invalid range \"%c-%c\" in transliteration operator",
3045 (char)min, (char)max);
3049 if (literal_endpoint == 2 &&
3050 ((isLOWER(min) && isLOWER(max)) ||
3051 (isUPPER(min) && isUPPER(max)))) {
3053 for (i = min; i <= max; i++)
3055 *d++ = NATIVE_TO_NEED(has_utf8,i);
3057 for (i = min; i <= max; i++)
3059 *d++ = NATIVE_TO_NEED(has_utf8,i);
3064 for (i = min; i <= max; i++)
3067 const U8 ch = (U8)NATIVE_TO_UTF(i);
3068 if (UNI_IS_INVARIANT(ch))
3071 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
3072 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
3081 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
3083 *d++ = (char)UTF_TO_NATIVE(0xff);
3085 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
3089 /* mark the range as done, and continue */
3093 literal_endpoint = 0;
3098 /* range begins (ignore - as first or last char) */
3099 else if (*s == '-' && s+1 < send && s != start) {
3101 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
3108 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
3118 literal_endpoint = 0;
3119 native_range = TRUE;
3124 /* if we get here, we're not doing a transliteration */
3126 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3129 while (s1 >= start && *s1-- == '\\')
3132 in_charclass = TRUE;
3135 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3138 while (s1 >= start && *s1-- == '\\')
3141 in_charclass = FALSE;
3144 /* skip for regexp comments /(?#comment)/, except for the last
3145 * char, which will be done separately.
3146 * Stop on (?{..}) and friends */
3148 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
3150 while (s+1 < send && *s != ')')
3151 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3153 else if (!PL_lex_casemods && !in_charclass &&
3154 ( s[2] == '{' /* This should match regcomp.c */
3155 || (s[2] == '?' && s[3] == '{')))
3161 /* likewise skip #-initiated comments in //x patterns */
3162 else if (*s == '#' && PL_lex_inpat &&
3163 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3164 while (s+1 < send && *s != '\n')
3165 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3168 /* no further processing of single-quoted regex */
3169 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3170 goto default_action;
3172 /* check for embedded arrays
3173 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3175 else if (*s == '@' && s[1]) {
3176 if (isALNUM_lazy_if(s+1,UTF))
3178 if (strchr(":'{$", s[1]))
3180 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3181 break; /* in regexp, neither @+ nor @- are interpolated */
3184 /* check for embedded scalars. only stop if we're sure it's a
3187 else if (*s == '$') {
3188 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3190 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3192 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3193 "Possible unintended interpolation of $\\ in regex");
3195 break; /* in regexp, $ might be tail anchor */
3199 /* End of else if chain - OP_TRANS rejoin rest */
3202 if (*s == '\\' && s+1 < send) {
3203 char* e; /* Can be used for ending '}', etc. */
3207 /* warn on \1 - \9 in substitution replacements, but note that \11
3208 * is an octal; and \19 is \1 followed by '9' */
3209 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3210 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3212 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3217 /* string-change backslash escapes */
3218 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3222 /* In a pattern, process \N, but skip any other backslash escapes.
3223 * This is because we don't want to translate an escape sequence
3224 * into a meta symbol and have the regex compiler use the meta
3225 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3226 * in spite of this, we do have to process \N here while the proper
3227 * charnames handler is in scope. See bugs #56444 and #62056.
3228 * There is a complication because \N in a pattern may also stand
3229 * for 'match a non-nl', and not mean a charname, in which case its
3230 * processing should be deferred to the regex compiler. To be a
3231 * charname it must be followed immediately by a '{', and not look
3232 * like \N followed by a curly quantifier, i.e., not something like
3233 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3235 else if (PL_lex_inpat
3238 || regcurly(s + 1)))
3240 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3241 goto default_action;
3246 /* quoted - in transliterations */
3248 if (PL_lex_inwhat == OP_TRANS) {
3256 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3257 "Unrecognized escape \\%c passed through",
3259 /* default action is to copy the quoted character */
3260 goto default_action;
3263 /* eg. \132 indicates the octal constant 0132 */
3264 case '0': case '1': case '2': case '3':
3265 case '4': case '5': case '6': case '7':
3269 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3272 goto NUM_ESCAPE_INSERT;
3274 /* eg. \o{24} indicates the octal constant \024 */
3280 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
3286 goto NUM_ESCAPE_INSERT;
3289 /* eg. \x24 indicates the hex constant 0x24 */
3295 bool valid = grok_bslash_x(s, &uv, &len, &error, 1);
3304 /* Insert oct or hex escaped character. There will always be
3305 * enough room in sv since such escapes will be longer than any
3306 * UTF-8 sequence they can end up as, except if they force us
3307 * to recode the rest of the string into utf8 */
3309 /* Here uv is the ordinal of the next character being added in
3310 * unicode (converted from native). */
3311 if (!UNI_IS_INVARIANT(uv)) {
3312 if (!has_utf8 && uv > 255) {
3313 /* Might need to recode whatever we have accumulated so
3314 * far if it contains any chars variant in utf8 or
3317 SvCUR_set(sv, d - SvPVX_const(sv));
3320 /* See Note on sizing above. */
3321 sv_utf8_upgrade_flags_grow(sv,
3322 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3323 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3324 d = SvPVX(sv) + SvCUR(sv);
3329 d = (char*)uvuni_to_utf8((U8*)d, uv);
3330 if (PL_lex_inwhat == OP_TRANS &&
3331 PL_sublex_info.sub_op) {
3332 PL_sublex_info.sub_op->op_private |=
3333 (PL_lex_repl ? OPpTRANS_FROM_UTF
3337 if (uv > 255 && !dorange)
3338 native_range = FALSE;
3351 /* In a non-pattern \N must be a named character, like \N{LATIN
3352 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3353 * mean to match a non-newline. For non-patterns, named
3354 * characters are converted to their string equivalents. In
3355 * patterns, named characters are not converted to their
3356 * ultimate forms for the same reasons that other escapes
3357 * aren't. Instead, they are converted to the \N{U+...} form
3358 * to get the value from the charnames that is in effect right
3359 * now, while preserving the fact that it was a named character
3360 * so that the regex compiler knows this */
3362 /* This section of code doesn't generally use the
3363 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3364 * a close examination of this macro and determined it is a
3365 * no-op except on utfebcdic variant characters. Every
3366 * character generated by this that would normally need to be
3367 * enclosed by this macro is invariant, so the macro is not
3368 * needed, and would complicate use of copy(). XXX There are
3369 * other parts of this file where the macro is used
3370 * inconsistently, but are saved by it being a no-op */
3372 /* The structure of this section of code (besides checking for
3373 * errors and upgrading to utf8) is:
3374 * Further disambiguate between the two meanings of \N, and if
3375 * not a charname, go process it elsewhere
3376 * If of form \N{U+...}, pass it through if a pattern;
3377 * otherwise convert to utf8
3378 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3379 * pattern; otherwise convert to utf8 */
3381 /* Here, s points to the 'N'; the test below is guaranteed to
3382 * succeed if we are being called on a pattern as we already
3383 * know from a test above that the next character is a '{'.
3384 * On a non-pattern \N must mean 'named sequence, which
3385 * requires braces */
3388 yyerror("Missing braces on \\N{}");
3393 /* If there is no matching '}', it is an error. */
3394 if (! (e = strchr(s, '}'))) {
3395 if (! PL_lex_inpat) {
3396 yyerror("Missing right brace on \\N{}");
3398 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3403 /* Here it looks like a named character */
3405 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3406 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3407 | PERL_SCAN_DISALLOW_PREFIX;
3410 /* For \N{U+...}, the '...' is a unicode value even on
3411 * EBCDIC machines */
3412 s += 2; /* Skip to next char after the 'U+' */
3414 uv = grok_hex(s, &len, &flags, NULL);
3415 if (len == 0 || len != (STRLEN)(e - s)) {
3416 yyerror("Invalid hexadecimal number in \\N{U+...}");
3423 /* On non-EBCDIC platforms, pass through to the regex
3424 * compiler unchanged. The reason we evaluated the
3425 * number above is to make sure there wasn't a syntax
3426 * error. But on EBCDIC we convert to native so
3427 * downstream code can continue to assume it's native
3429 s -= 5; /* Include the '\N{U+' */
3431 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3434 (unsigned int) UNI_TO_NATIVE(uv));
3436 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3440 else { /* Not a pattern: convert the hex to string */
3442 /* If destination is not in utf8, unconditionally
3443 * recode it to be so. This is because \N{} implies
3444 * Unicode semantics, and scalars have to be in utf8
3445 * to guarantee those semantics */
3447 SvCUR_set(sv, d - SvPVX_const(sv));
3450 /* See Note on sizing above. */
3451 sv_utf8_upgrade_flags_grow(
3453 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3454 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3455 d = SvPVX(sv) + SvCUR(sv);
3459 /* Add the string to the output */
3460 if (UNI_IS_INVARIANT(uv)) {
3463 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3466 else /* Here is \N{NAME} but not \N{U+...}. */
3467 if ((res = get_and_check_backslash_N_name(s, e)))
3470 const char *str = SvPV_const(res, len);
3473 if (! len) { /* The name resolved to an empty string */
3474 Copy("\\N{}", d, 4, char);
3478 /* In order to not lose information for the regex
3479 * compiler, pass the result in the specially made
3480 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3481 * the code points in hex of each character
3482 * returned by charnames */
3484 const char *str_end = str + len;
3485 const STRLEN off = d - SvPVX_const(sv);
3487 if (! SvUTF8(res)) {
3488 /* For the non-UTF-8 case, we can determine the
3489 * exact length needed without having to parse
3490 * through the string. Each character takes up
3491 * 2 hex digits plus either a trailing dot or
3493 d = off + SvGROW(sv, off
3495 + 6 /* For the "\N{U+", and
3497 + (STRLEN)(send - e));
3498 Copy("\\N{U+", d, 5, char);
3500 while (str < str_end) {
3502 my_snprintf(hex_string, sizeof(hex_string),
3503 "%02X.", (U8) *str);
3504 Copy(hex_string, d, 3, char);
3508 d--; /* We will overwrite below the final
3509 dot with a right brace */
3512 STRLEN char_length; /* cur char's byte length */
3514 /* and the number of bytes after this is
3515 * translated into hex digits */
3516 STRLEN output_length;
3518 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3519 * for max('U+', '.'); and 1 for NUL */
3520 char hex_string[2 * UTF8_MAXBYTES + 5];
3522 /* Get the first character of the result. */
3523 U32 uv = utf8n_to_uvuni((U8 *) str,
3527 /* Convert first code point to hex, including
3528 * the boiler plate before it. For all these,
3529 * we convert to native format so that
3530 * downstream code can continue to assume the
3531 * input is native */
3533 my_snprintf(hex_string, sizeof(hex_string),
3535 (unsigned int) UNI_TO_NATIVE(uv));
3537 /* Make sure there is enough space to hold it */
3538 d = off + SvGROW(sv, off
3540 + (STRLEN)(send - e)
3541 + 2); /* '}' + NUL */
3543 Copy(hex_string, d, output_length, char);
3546 /* For each subsequent character, append dot and
3547 * its ordinal in hex */
3548 while ((str += char_length) < str_end) {
3549 const STRLEN off = d - SvPVX_const(sv);
3550 U32 uv = utf8n_to_uvuni((U8 *) str,
3555 my_snprintf(hex_string,
3558 (unsigned int) UNI_TO_NATIVE(uv));
3560 d = off + SvGROW(sv, off
3562 + (STRLEN)(send - e)
3563 + 2); /* '}' + NUL */
3564 Copy(hex_string, d, output_length, char);
3569 *d++ = '}'; /* Done. Add the trailing brace */
3572 else { /* Here, not in a pattern. Convert the name to a
3575 /* If destination is not in utf8, unconditionally
3576 * recode it to be so. This is because \N{} implies
3577 * Unicode semantics, and scalars have to be in utf8
3578 * to guarantee those semantics */
3580 SvCUR_set(sv, d - SvPVX_const(sv));
3583 /* See Note on sizing above. */
3584 sv_utf8_upgrade_flags_grow(sv,
3585 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3586 len + (STRLEN)(send - s) + 1);
3587 d = SvPVX(sv) + SvCUR(sv);
3589 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3591 /* See Note on sizing above. (NOTE: SvCUR() is not
3592 * set correctly here). */
3593 const STRLEN off = d - SvPVX_const(sv);
3594 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3596 Copy(str, d, len, char);
3602 } /* End \N{NAME} */
3605 native_range = FALSE; /* \N{} is defined to be Unicode */
3607 s = e + 1; /* Point to just after the '}' */
3610 /* \c is a control character */
3614 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3617 yyerror("Missing control char name in \\c");
3621 /* printf-style backslashes, formfeeds, newlines, etc */
3623 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3626 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3629 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3632 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3635 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3638 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3641 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3647 } /* end if (backslash) */
3654 /* If we started with encoded form, or already know we want it,
3655 then encode the next character */
3656 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3660 /* One might think that it is wasted effort in the case of the
3661 * source being utf8 (this_utf8 == TRUE) to take the next character
3662 * in the source, convert it to an unsigned value, and then convert
3663 * it back again. But the source has not been validated here. The
3664 * routine that does the conversion checks for errors like
3667 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3668 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3670 SvCUR_set(sv, d - SvPVX_const(sv));
3673 /* See Note on sizing above. */
3674 sv_utf8_upgrade_flags_grow(sv,
3675 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3676 need + (STRLEN)(send - s) + 1);
3677 d = SvPVX(sv) + SvCUR(sv);
3679 } else if (need > len) {
3680 /* encoded value larger than old, may need extra space (NOTE:
3681 * SvCUR() is not set correctly here). See Note on sizing
3683 const STRLEN off = d - SvPVX_const(sv);
3684 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3688 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3690 if (uv > 255 && !dorange)
3691 native_range = FALSE;
3695 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3697 } /* while loop to process each character */
3699 /* terminate the string and set up the sv */
3701 SvCUR_set(sv, d - SvPVX_const(sv));
3702 if (SvCUR(sv) >= SvLEN(sv))
3703 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3704 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3707 if (PL_encoding && !has_utf8) {
3708 sv_recode_to_utf8(sv, PL_encoding);
3714 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3715 PL_sublex_info.sub_op->op_private |=
3716 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3720 /* shrink the sv if we allocated more than we used */
3721 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3722 SvPV_shrink_to_cur(sv);
3725 /* return the substring (via pl_yylval) only if we parsed anything */
3726 if (s > PL_bufptr) {
3727 SvREFCNT_inc_simple_void_NN(sv);
3728 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3729 const char *const key = PL_lex_inpat ? "qr" : "q";
3730 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3734 if (PL_lex_inwhat == OP_TRANS) {
3737 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3740 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3748 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3751 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3753 LEAVE_with_name("scan_const");
3758 * Returns TRUE if there's more to the expression (e.g., a subscript),
3761 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3763 * ->[ and ->{ return TRUE
3764 * { and [ outside a pattern are always subscripts, so return TRUE
3765 * if we're outside a pattern and it's not { or [, then return FALSE
3766 * if we're in a pattern and the first char is a {
3767 * {4,5} (any digits around the comma) returns FALSE
3768 * if we're in a pattern and the first char is a [
3770 * [SOMETHING] has a funky algorithm to decide whether it's a
3771 * character class or not. It has to deal with things like
3772 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3773 * anything else returns TRUE
3776 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3779 S_intuit_more(pTHX_ register char *s)
3783 PERL_ARGS_ASSERT_INTUIT_MORE;
3785 if (PL_lex_brackets)
3787 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3789 if (*s != '{' && *s != '[')
3794 /* In a pattern, so maybe we have {n,m}. */
3802 /* On the other hand, maybe we have a character class */
3805 if (*s == ']' || *s == '^')
3808 /* this is terrifying, and it works */
3809 int weight = 2; /* let's weigh the evidence */
3811 unsigned char un_char = 255, last_un_char;
3812 const char * const send = strchr(s,']');
3813 char tmpbuf[sizeof PL_tokenbuf * 4];
3815 if (!send) /* has to be an expression */
3818 Zero(seen,256,char);
3821 else if (isDIGIT(*s)) {
3823 if (isDIGIT(s[1]) && s[2] == ']')
3829 for (; s < send; s++) {
3830 last_un_char = un_char;
3831 un_char = (unsigned char)*s;
3836 weight -= seen[un_char] * 10;
3837 if (isALNUM_lazy_if(s+1,UTF)) {
3839 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3840 len = (int)strlen(tmpbuf);
3841 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3842 UTF ? SVf_UTF8 : 0, SVt_PV))
3847 else if (*s == '$' && s[1] &&
3848 strchr("[#!%*<>()-=",s[1])) {
3849 if (/*{*/ strchr("])} =",s[2]))
3858 if (strchr("wds]",s[1]))
3860 else if (seen[(U8)'\''] || seen[(U8)'"'])
3862 else if (strchr("rnftbxcav",s[1]))
3864 else if (isDIGIT(s[1])) {
3866 while (s[1] && isDIGIT(s[1]))
3876 if (strchr("aA01! ",last_un_char))
3878 if (strchr("zZ79~",s[1]))
3880 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3881 weight -= 5; /* cope with negative subscript */
3884 if (!isALNUM(last_un_char)
3885 && !(last_un_char == '$' || last_un_char == '@'
3886 || last_un_char == '&')
3887 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3892 if (keyword(tmpbuf, d - tmpbuf, 0))
3895 if (un_char == last_un_char + 1)
3897 weight -= seen[un_char];
3902 if (weight >= 0) /* probably a character class */
3912 * Does all the checking to disambiguate
3914 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3915 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3917 * First argument is the stuff after the first token, e.g. "bar".
3919 * Not a method if foo is a filehandle.
3920 * Not a method if foo is a subroutine prototyped to take a filehandle.
3921 * Not a method if it's really "Foo $bar"
3922 * Method if it's "foo $bar"
3923 * Not a method if it's really "print foo $bar"
3924 * Method if it's really "foo package::" (interpreted as package->foo)
3925 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3926 * Not a method if bar is a filehandle or package, but is quoted with
3931 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3934 char *s = start + (*start == '$');
3935 char tmpbuf[sizeof PL_tokenbuf];
3942 PERL_ARGS_ASSERT_INTUIT_METHOD;
3944 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3946 if (cv && SvPOK(cv)) {
3947 const char *proto = CvPROTO(cv);
3955 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3956 /* start is the beginning of the possible filehandle/object,
3957 * and s is the end of it
3958 * tmpbuf is a copy of it
3961 if (*start == '$') {
3962 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3963 isUPPER(*PL_tokenbuf))
3966 len = start - SvPVX(PL_linestr);
3970 start = SvPVX(PL_linestr) + len;
3974 return *s == '(' ? FUNCMETH : METHOD;
3976 if (!keyword(tmpbuf, len, 0)) {
3977 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3981 soff = s - SvPVX(PL_linestr);
3985 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3986 if (indirgv && GvCVu(indirgv))
3988 /* filehandle or package name makes it a method */
3989 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3991 soff = s - SvPVX(PL_linestr);
3994 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3995 return 0; /* no assumptions -- "=>" quotes bareword */
3997 start_force(PL_curforce);
3998 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3999 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
4000 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
4002 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
4003 ( UTF ? SVf_UTF8 : 0 )));
4008 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
4010 return *s == '(' ? FUNCMETH : METHOD;
4016 /* Encoded script support. filter_add() effectively inserts a
4017 * 'pre-processing' function into the current source input stream.
4018 * Note that the filter function only applies to the current source file
4019 * (e.g., it will not affect files 'require'd or 'use'd by this one).
4021 * The datasv parameter (which may be NULL) can be used to pass
4022 * private data to this instance of the filter. The filter function
4023 * can recover the SV using the FILTER_DATA macro and use it to
4024 * store private buffers and state information.
4026 * The supplied datasv parameter is upgraded to a PVIO type
4027 * and the IoDIRP/IoANY field is used to store the function pointer,
4028 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
4029 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
4030 * private use must be set using malloc'd pointers.
4034 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
4043 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
4044 Perl_croak(aTHX_ "Source filters apply only to byte streams");
4046 if (!PL_rsfp_filters)
4047 PL_rsfp_filters = newAV();
4050 SvUPGRADE(datasv, SVt_PVIO);
4051 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
4052 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
4053 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
4054 FPTR2DPTR(void *, IoANY(datasv)),
4055 SvPV_nolen(datasv)));
4056 av_unshift(PL_rsfp_filters, 1);
4057 av_store(PL_rsfp_filters, 0, datasv) ;
4059 !PL_parser->filtered
4060 && PL_parser->lex_flags & LEX_EVALBYTES
4061 && PL_bufptr < PL_bufend
4063 const char *s = PL_bufptr;
4064 while (s < PL_bufend) {
4066 SV *linestr = PL_parser->linestr;
4067 char *buf = SvPVX(linestr);
4068 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
4069 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
4070 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
4071 STRLEN const linestart_pos = PL_parser->linestart - buf;
4072 STRLEN const last_uni_pos =
4073 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
4074 STRLEN const last_lop_pos =
4075 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
4076 av_push(PL_rsfp_filters, linestr);
4077 PL_parser->linestr =
4078 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
4079 buf = SvPVX(PL_parser->linestr);
4080 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4081 PL_parser->bufptr = buf + bufptr_pos;
4082 PL_parser->oldbufptr = buf + oldbufptr_pos;
4083 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4084 PL_parser->linestart = buf + linestart_pos;
4085 if (PL_parser->last_uni)
4086 PL_parser->last_uni = buf + last_uni_pos;
4087 if (PL_parser->last_lop)
4088 PL_parser->last_lop = buf + last_lop_pos;
4089 SvLEN(linestr) = SvCUR(linestr);
4090 SvCUR(linestr) = s-SvPVX(linestr);
4091 PL_parser->filtered = 1;
4101 /* Delete most recently added instance of this filter function. */
4103 Perl_filter_del(pTHX_ filter_t funcp)
4108 PERL_ARGS_ASSERT_FILTER_DEL;
4111 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4112 FPTR2DPTR(void*, funcp)));
4114 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4116 /* if filter is on top of stack (usual case) just pop it off */
4117 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4118 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4119 sv_free(av_pop(PL_rsfp_filters));
4123 /* we need to search for the correct entry and clear it */
4124 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4128 /* Invoke the idxth filter function for the current rsfp. */
4129 /* maxlen 0 = read one text line */
4131 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4136 /* This API is bad. It should have been using unsigned int for maxlen.
4137 Not sure if we want to change the API, but if not we should sanity
4138 check the value here. */
4139 unsigned int correct_length
4148 PERL_ARGS_ASSERT_FILTER_READ;
4150 if (!PL_parser || !PL_rsfp_filters)
4152 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4153 /* Provide a default input filter to make life easy. */
4154 /* Note that we append to the line. This is handy. */
4155 DEBUG_P(PerlIO_printf(Perl_debug_log,
4156 "filter_read %d: from rsfp\n", idx));
4157 if (correct_length) {
4160 const int old_len = SvCUR(buf_sv);
4162 /* ensure buf_sv is large enough */
4163 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4164 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4165 correct_length)) <= 0) {
4166 if (PerlIO_error(PL_rsfp))
4167 return -1; /* error */
4169 return 0 ; /* end of file */
4171 SvCUR_set(buf_sv, old_len + len) ;
4172 SvPVX(buf_sv)[old_len + len] = '\0';
4175 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4176 if (PerlIO_error(PL_rsfp))
4177 return -1; /* error */
4179 return 0 ; /* end of file */
4182 return SvCUR(buf_sv);
4184 /* Skip this filter slot if filter has been deleted */
4185 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4186 DEBUG_P(PerlIO_printf(Perl_debug_log,
4187 "filter_read %d: skipped (filter deleted)\n",
4189 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4191 if (SvTYPE(datasv) != SVt_PVIO) {
4192 if (correct_length) {
4194 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4195 if (!remainder) return 0; /* eof */
4196 if (correct_length > remainder) correct_length = remainder;
4197 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4198 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4201 const char *s = SvEND(datasv);
4202 const char *send = SvPVX(datasv) + SvLEN(datasv);
4210 if (s == send) return 0; /* eof */
4211 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4212 SvCUR_set(datasv, s-SvPVX(datasv));
4214 return SvCUR(buf_sv);
4216 /* Get function pointer hidden within datasv */
4217 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4218 DEBUG_P(PerlIO_printf(Perl_debug_log,
4219 "filter_read %d: via function %p (%s)\n",
4220 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4221 /* Call function. The function is expected to */
4222 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4223 /* Return: <0:error, =0:eof, >0:not eof */
4224 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4228 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4232 PERL_ARGS_ASSERT_FILTER_GETS;
4234 #ifdef PERL_CR_FILTER
4235 if (!PL_rsfp_filters) {
4236 filter_add(S_cr_textfilter,NULL);
4239 if (PL_rsfp_filters) {
4241 SvCUR_set(sv, 0); /* start with empty line */
4242 if (FILTER_READ(0, sv, 0) > 0)
4243 return ( SvPVX(sv) ) ;
4248 return (sv_gets(sv, PL_rsfp, append));
4252 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4257 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4259 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4263 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4264 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4266 return GvHV(gv); /* Foo:: */
4269 /* use constant CLASS => 'MyClass' */
4270 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4271 if (gv && GvCV(gv)) {
4272 SV * const sv = cv_const_sv(GvCV(gv));
4274 pkgname = SvPV_const(sv, len);
4277 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4281 * S_readpipe_override
4282 * Check whether readpipe() is overridden, and generates the appropriate
4283 * optree, provided sublex_start() is called afterwards.
4286 S_readpipe_override(pTHX)
4289 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4290 pl_yylval.ival = OP_BACKTICK;
4292 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4294 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4295 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4296 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4298 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4299 op_append_elem(OP_LIST,
4300 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4301 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4308 * The intent of this yylex wrapper is to minimize the changes to the
4309 * tokener when we aren't interested in collecting madprops. It remains
4310 * to be seen how successful this strategy will be...
4317 char *s = PL_bufptr;
4319 /* make sure PL_thiswhite is initialized */
4323 /* previous token ate up our whitespace? */
4324 if (!PL_lasttoke && PL_nextwhite) {
4325 PL_thiswhite = PL_nextwhite;
4329 /* isolate the token, and figure out where it is without whitespace */
4330 PL_realtokenstart = -1;
4334 assert(PL_curforce < 0);
4336 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4337 if (!PL_thistoken) {
4338 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4339 PL_thistoken = newSVpvs("");
4341 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4342 PL_thistoken = newSVpvn(tstart, s - tstart);
4345 if (PL_thismad) /* install head */
4346 CURMAD('X', PL_thistoken);
4349 /* last whitespace of a sublex? */
4350 if (optype == ')' && PL_endwhite) {
4351 CURMAD('X', PL_endwhite);
4356 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4357 if (!PL_thiswhite && !PL_endwhite && !optype) {
4358 sv_free(PL_thistoken);
4363 /* put off final whitespace till peg */
4364 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4365 PL_nextwhite = PL_thiswhite;
4368 else if (PL_thisopen) {
4369 CURMAD('q', PL_thisopen);
4371 sv_free(PL_thistoken);
4375 /* Store actual token text as madprop X */
4376 CURMAD('X', PL_thistoken);
4380 /* add preceding whitespace as madprop _ */
4381 CURMAD('_', PL_thiswhite);
4385 /* add quoted material as madprop = */
4386 CURMAD('=', PL_thisstuff);
4390 /* add terminating quote as madprop Q */
4391 CURMAD('Q', PL_thisclose);
4395 /* special processing based on optype */
4399 /* opval doesn't need a TOKEN since it can already store mp */
4409 if (pl_yylval.opval)
4410 append_madprops(PL_thismad, pl_yylval.opval, 0);
4418 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4431 /* remember any fake bracket that lexer is about to discard */
4432 if (PL_lex_brackets == 1 &&
4433 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4436 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4439 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4440 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4443 break; /* don't bother looking for trailing comment */
4452 /* attach a trailing comment to its statement instead of next token */
4456 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4458 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4460 if (*s == '\n' || *s == '#') {
4461 while (s < PL_bufend && *s != '\n')
4465 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4466 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4479 /* Create new token struct. Note: opvals return early above. */
4480 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4487 S_tokenize_use(pTHX_ int is_use, char *s) {
4490 PERL_ARGS_ASSERT_TOKENIZE_USE;
4492 if (PL_expect != XSTATE)
4493 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4494 is_use ? "use" : "no"));
4497 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4498 s = force_version(s, TRUE);
4499 if (*s == ';' || *s == '}'
4500 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4501 start_force(PL_curforce);
4502 NEXTVAL_NEXTTOKE.opval = NULL;
4505 else if (*s == 'v') {
4506 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4507 s = force_version(s, FALSE);
4511 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4512 s = force_version(s, FALSE);
4514 pl_yylval.ival = is_use;
4518 static const char* const exp_name[] =
4519 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4520 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4524 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4526 S_word_takes_any_delimeter(char *p, STRLEN len)
4528 return (len == 1 && strchr("msyq", p[0])) ||
4530 (p[0] == 't' && p[1] == 'r') ||
4531 (p[0] == 'q' && strchr("qwxr", p[1]))));
4537 Works out what to call the token just pulled out of the input
4538 stream. The yacc parser takes care of taking the ops we return and
4539 stitching them into a tree.
4542 The type of the next token
4545 Switch based on the current state:
4546 - if we already built the token before, use it
4547 - if we have a case modifier in a string, deal with that
4548 - handle other cases of interpolation inside a string
4549 - scan the next line if we are inside a format
4550 In the normal state switch on the next character:
4552 if alphabetic, go to key lookup
4553 unrecoginized character - croak
4554 - 0/4/26: handle end-of-line or EOF
4555 - cases for whitespace
4556 - \n and #: handle comments and line numbers
4557 - various operators, brackets and sigils
4560 - 'v': vstrings (or go to key lookup)
4561 - 'x' repetition operator (or go to key lookup)
4562 - other ASCII alphanumerics (key lookup begins here):
4565 scan built-in keyword (but do nothing with it yet)
4566 check for statement label
4567 check for lexical subs
4568 goto just_a_word if there is one
4569 see whether built-in keyword is overridden
4570 switch on keyword number:
4571 - default: just_a_word:
4572 not a built-in keyword; handle bareword lookup
4573 disambiguate between method and sub call
4574 fall back to bareword
4575 - cases for built-in keywords
4580 #pragma segment Perl_yylex
4586 char *s = PL_bufptr;
4593 /* orig_keyword, gvp, and gv are initialized here because
4594 * jump to the label just_a_word_zero can bypass their
4595 * initialization later. */
4596 I32 orig_keyword = 0;
4601 SV* tmp = newSVpvs("");
4602 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4603 (IV)CopLINE(PL_curcop),
4604 lex_state_names[PL_lex_state],
4605 exp_name[PL_expect],
4606 pv_display(tmp, s, strlen(s), 0, 60));
4610 switch (PL_lex_state) {
4612 case LEX_NORMAL: /* Some compilers will produce faster */
4613 case LEX_INTERPNORMAL: /* code if we comment these out. */
4617 /* when we've already built the next token, just pull it out of the queue */
4621 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4623 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4624 PL_nexttoke[PL_lasttoke].next_mad = 0;
4625 if (PL_thismad && PL_thismad->mad_key == '_') {
4626 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4627 PL_thismad->mad_val = 0;
4628 mad_free(PL_thismad);
4633 PL_lex_state = PL_lex_defer;
4634 PL_expect = PL_lex_expect;
4635 PL_lex_defer = LEX_NORMAL;
4636 if (!PL_nexttoke[PL_lasttoke].next_type)
4641 pl_yylval = PL_nextval[PL_nexttoke];
4643 PL_lex_state = PL_lex_defer;
4644 PL_expect = PL_lex_expect;
4645 PL_lex_defer = LEX_NORMAL;
4651 next_type = PL_nexttoke[PL_lasttoke].next_type;
4653 next_type = PL_nexttype[PL_nexttoke];
4655 if (next_type & (7<<24)) {
4656 if (next_type & (1<<24)) {
4657 if (PL_lex_brackets > 100)
4658 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4659 PL_lex_brackstack[PL_lex_brackets++] =
4660 (char) ((next_type >> 16) & 0xff);
4662 if (next_type & (2<<24))
4663 PL_lex_allbrackets++;
4664 if (next_type & (4<<24))
4665 PL_lex_allbrackets--;
4666 next_type &= 0xffff;
4668 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4671 /* interpolated case modifiers like \L \U, including \Q and \E.
4672 when we get here, PL_bufptr is at the \
4674 case LEX_INTERPCASEMOD:
4676 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4678 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4679 PL_bufptr, PL_bufend, *PL_bufptr);
4681 /* handle \E or end of string */
4682 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4684 if (PL_lex_casemods) {
4685 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4686 PL_lex_casestack[PL_lex_casemods] = '\0';
4688 if (PL_bufptr != PL_bufend
4689 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4690 || oldmod == 'F')) {
4692 PL_lex_state = LEX_INTERPCONCAT;
4695 PL_thistoken = newSVpvs("\\E");
4698 PL_lex_allbrackets--;
4701 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4702 /* Got an unpaired \E */
4703 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4704 "Useless use of \\E");
4707 while (PL_bufptr != PL_bufend &&
4708 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4711 PL_thiswhite = newSVpvs("");
4712 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4717 if (PL_bufptr != PL_bufend)
4720 PL_lex_state = LEX_INTERPCONCAT;
4724 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4725 "### Saw case modifier\n"); });
4727 if (s[1] == '\\' && s[2] == 'E') {
4731 PL_thiswhite = newSVpvs("");
4732 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4736 PL_lex_state = LEX_INTERPCONCAT;
4741 if (!PL_madskills) /* when just compiling don't need correct */
4742 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4743 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4744 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4745 (strchr(PL_lex_casestack, 'L')
4746 || strchr(PL_lex_casestack, 'U')
4747 || strchr(PL_lex_casestack, 'F'))) {
4748 PL_lex_casestack[--PL_lex_casemods] = '\0';
4749 PL_lex_allbrackets--;
4752 if (PL_lex_casemods > 10)
4753 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4754 PL_lex_casestack[PL_lex_casemods++] = *s;
4755 PL_lex_casestack[PL_lex_casemods] = '\0';
4756 PL_lex_state = LEX_INTERPCONCAT;
4757 start_force(PL_curforce);
4758 NEXTVAL_NEXTTOKE.ival = 0;
4759 force_next((2<<24)|'(');
4760 start_force(PL_curforce);
4762 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4764 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4766 NEXTVAL_NEXTTOKE.ival = OP_LC;
4768 NEXTVAL_NEXTTOKE.ival = OP_UC;
4770 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4772 NEXTVAL_NEXTTOKE.ival = OP_FC;
4774 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4776 SV* const tmpsv = newSVpvs("\\ ");
4777 /* replace the space with the character we want to escape
4779 SvPVX(tmpsv)[1] = *s;
4785 if (PL_lex_starts) {
4791 sv_free(PL_thistoken);
4792 PL_thistoken = newSVpvs("");
4795 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4796 if (PL_lex_casemods == 1 && PL_lex_inpat)
4805 case LEX_INTERPPUSH:
4806 return REPORT(sublex_push());
4808 case LEX_INTERPSTART:
4809 if (PL_bufptr == PL_bufend)
4810 return REPORT(sublex_done());
4811 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4812 "### Interpolated variable\n"); });
4814 PL_lex_dojoin = (*PL_bufptr == '@');
4815 PL_lex_state = LEX_INTERPNORMAL;
4816 if (PL_lex_dojoin) {
4817 start_force(PL_curforce);
4818 NEXTVAL_NEXTTOKE.ival = 0;
4820 start_force(PL_curforce);
4821 force_ident("\"", '$');
4822 start_force(PL_curforce);
4823 NEXTVAL_NEXTTOKE.ival = 0;
4825 start_force(PL_curforce);
4826 NEXTVAL_NEXTTOKE.ival = 0;
4827 force_next((2<<24)|'(');
4828 start_force(PL_curforce);
4829 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4832 /* Convert (?{...}) and friends to 'do {...}' */
4833 if (PL_lex_inpat && *PL_bufptr == '(') {
4834 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4836 if (*PL_bufptr != '{')
4838 start_force(PL_curforce);
4839 /* XXX probably need a CURMAD(something) here */
4840 PL_expect = XTERMBLOCK;
4844 if (PL_lex_starts++) {
4849 sv_free(PL_thistoken);
4850 PL_thistoken = newSVpvs("");
4853 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4854 if (!PL_lex_casemods && PL_lex_inpat)
4861 case LEX_INTERPENDMAYBE:
4862 if (intuit_more(PL_bufptr)) {
4863 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4869 if (PL_lex_dojoin) {
4870 PL_lex_dojoin = FALSE;
4871 PL_lex_state = LEX_INTERPCONCAT;
4875 sv_free(PL_thistoken);
4876 PL_thistoken = newSVpvs("");
4879 PL_lex_allbrackets--;
4882 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4883 && SvEVALED(PL_lex_repl))
4885 if (PL_bufptr != PL_bufend)
4886 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4889 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4890 re_eval_str. If the here-doc body’s length equals the previous
4891 value of re_eval_start, re_eval_start will now be null. So
4892 check re_eval_str as well. */
4893 if (PL_parser->lex_shared->re_eval_start
4894 || PL_parser->lex_shared->re_eval_str) {
4896 if (*PL_bufptr != ')')
4897 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4899 /* having compiled a (?{..}) expression, return the original
4900 * text too, as a const */
4901 if (PL_parser->lex_shared->re_eval_str) {
4902 sv = PL_parser->lex_shared->re_eval_str;
4903 PL_parser->lex_shared->re_eval_str = NULL;
4905 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4906 SvPV_shrink_to_cur(sv);
4908 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4909 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4910 start_force(PL_curforce);
4911 /* XXX probably need a CURMAD(something) here */
4912 NEXTVAL_NEXTTOKE.opval =
4913 (OP*)newSVOP(OP_CONST, 0,
4916 PL_parser->lex_shared->re_eval_start = NULL;
4922 case LEX_INTERPCONCAT:
4924 if (PL_lex_brackets)
4925 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4926 (long) PL_lex_brackets);
4928 if (PL_bufptr == PL_bufend)
4929 return REPORT(sublex_done());
4931 /* m'foo' still needs to be parsed for possible (?{...}) */
4932 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4933 SV *sv = newSVsv(PL_linestr);
4935 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4939 s = scan_const(PL_bufptr);
4941 PL_lex_state = LEX_INTERPCASEMOD;
4943 PL_lex_state = LEX_INTERPSTART;
4946 if (s != PL_bufptr) {
4947 start_force(PL_curforce);
4949 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4951 NEXTVAL_NEXTTOKE = pl_yylval;
4954 if (PL_lex_starts++) {
4958 sv_free(PL_thistoken);
4959 PL_thistoken = newSVpvs("");
4962 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4963 if (!PL_lex_casemods && PL_lex_inpat)
4976 s = scan_formline(PL_bufptr);
4977 if (!PL_lex_formbrack)
4987 PL_oldoldbufptr = PL_oldbufptr;
4993 sv_free(PL_thistoken);
4996 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
5000 if (isIDFIRST_lazy_if(s,UTF))
5003 SV *dsv = newSVpvs_flags("", SVs_TEMP);
5004 const char *c = UTF ? savepv(sv_uni_display(dsv, newSVpvn_flags(s,
5006 SVs_TEMP | SVf_UTF8),
5007 10, UNI_DISPLAY_ISPRINT))
5008 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
5009 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
5010 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
5011 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
5019 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"SVf"<-- HERE near column %d", c, SVfARG(dsv), (int) len + 1);
5023 goto fake_eof; /* emulate EOF on ^D or ^Z */
5029 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
5032 if (PL_lex_brackets &&
5033 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
5034 yyerror((const char *)
5036 ? "Format not terminated"
5037 : "Missing right curly or square bracket"));
5039 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5040 "### Tokener got EOF\n");
5044 if (s++ < PL_bufend)
5045 goto retry; /* ignore stray nulls */
5048 if (!PL_in_eval && !PL_preambled) {
5049 PL_preambled = TRUE;
5055 /* Generate a string of Perl code to load the debugger.
5056 * If PERL5DB is set, it will return the contents of that,
5057 * otherwise a compile-time require of perl5db.pl. */
5059 const char * const pdb = PerlEnv_getenv("PERL5DB");
5062 sv_setpv(PL_linestr, pdb);
5063 sv_catpvs(PL_linestr,";");
5065 SETERRNO(0,SS_NORMAL);
5066 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
5069 sv_setpvs(PL_linestr,"");
5070 if (PL_preambleav) {
5071 SV **svp = AvARRAY(PL_preambleav);
5072 SV **const end = svp + AvFILLp(PL_preambleav);
5074 sv_catsv(PL_linestr, *svp);
5076 sv_catpvs(PL_linestr, ";");
5078 sv_free(MUTABLE_SV(PL_preambleav));
5079 PL_preambleav = NULL;
5082 sv_catpvs(PL_linestr,
5083 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
5084 if (PL_minus_n || PL_minus_p) {
5085 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
5087 sv_catpvs(PL_linestr,"chomp;");
5090 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
5091 || *PL_splitstr == '"')
5092 && strchr(PL_splitstr + 1, *PL_splitstr))
5093 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
5095 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
5096 bytes can be used as quoting characters. :-) */
5097 const char *splits = PL_splitstr;
5098 sv_catpvs(PL_linestr, "our @F=split(q\0");
5101 if (*splits == '\\')
5102 sv_catpvn(PL_linestr, splits, 1);
5103 sv_catpvn(PL_linestr, splits, 1);
5104 } while (*splits++);
5105 /* This loop will embed the trailing NUL of
5106 PL_linestr as the last thing it does before
5108 sv_catpvs(PL_linestr, ");");
5112 sv_catpvs(PL_linestr,"our @F=split(' ');");
5115 sv_catpvs(PL_linestr, "\n");
5116 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5117 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5118 PL_last_lop = PL_last_uni = NULL;
5119 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
5120 update_debugger_info(PL_linestr, NULL, 0);
5125 bof = PL_rsfp ? TRUE : FALSE;
5128 fake_eof = LEX_FAKE_EOF;
5130 PL_bufptr = PL_bufend;
5131 COPLINE_INC_WITH_HERELINES;
5132 if (!lex_next_chunk(fake_eof)) {
5133 CopLINE_dec(PL_curcop);
5135 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
5137 CopLINE_dec(PL_curcop);
5140 PL_realtokenstart = -1;
5143 /* If it looks like the start of a BOM or raw UTF-16,
5144 * check if it in fact is. */
5145 if (bof && PL_rsfp &&
5150 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
5151 bof = (offset == (Off_t)SvCUR(PL_linestr));
5152 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
5153 /* offset may include swallowed CR */
5155 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
5158 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5159 s = swallow_bom((U8*)s);
5162 if (PL_parser->in_pod) {
5163 /* Incest with pod. */
5166 sv_catsv(PL_thiswhite, PL_linestr);
5168 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
5169 sv_setpvs(PL_linestr, "");
5170 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5171 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5172 PL_last_lop = PL_last_uni = NULL;
5173 PL_parser->in_pod = 0;
5176 if (PL_rsfp || PL_parser->filtered)
5178 } while (PL_parser->in_pod);
5179 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
5180 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5181 PL_last_lop = PL_last_uni = NULL;
5182 if (CopLINE(PL_curcop) == 1) {
5183 while (s < PL_bufend && isSPACE(*s))
5185 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
5189 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
5193 if (*s == '#' && *(s+1) == '!')
5195 #ifdef ALTERNATE_SHEBANG
5197 static char const as[] = ALTERNATE_SHEBANG;
5198 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5199 d = s + (sizeof(as) - 1);
5201 #endif /* ALTERNATE_SHEBANG */
5210 while (*d && !isSPACE(*d))
5214 #ifdef ARG_ZERO_IS_SCRIPT
5215 if (ipathend > ipath) {
5217 * HP-UX (at least) sets argv[0] to the script name,
5218 * which makes $^X incorrect. And Digital UNIX and Linux,
5219 * at least, set argv[0] to the basename of the Perl
5220 * interpreter. So, having found "#!", we'll set it right.
5222 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5224 assert(SvPOK(x) || SvGMAGICAL(x));
5225 if (sv_eq(x, CopFILESV(PL_curcop))) {
5226 sv_setpvn(x, ipath, ipathend - ipath);
5232 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
5233 const char * const lstart = SvPV_const(x,llen);
5235 bstart += blen - llen;
5236 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5237 sv_setpvn(x, ipath, ipathend - ipath);
5242 TAINT_NOT; /* $^X is always tainted, but that's OK */
5244 #endif /* ARG_ZERO_IS_SCRIPT */
5249 d = instr(s,"perl -");
5251 d = instr(s,"perl");
5253 /* avoid getting into infinite loops when shebang
5254 * line contains "Perl" rather than "perl" */
5256 for (d = ipathend-4; d >= ipath; --d) {
5257 if ((*d == 'p' || *d == 'P')
5258 && !ibcmp(d, "perl", 4))
5268 #ifdef ALTERNATE_SHEBANG
5270 * If the ALTERNATE_SHEBANG on this system starts with a
5271 * character that can be part of a Perl expression, then if
5272 * we see it but not "perl", we're probably looking at the
5273 * start of Perl code, not a request to hand off to some
5274 * other interpreter. Similarly, if "perl" is there, but
5275 * not in the first 'word' of the line, we assume the line
5276 * contains the start of the Perl program.
5278 if (d && *s != '#') {
5279 const char *c = ipath;
5280 while (*c && !strchr("; \t\r\n\f\v#", *c))
5283 d = NULL; /* "perl" not in first word; ignore */
5285 *s = '#'; /* Don't try to parse shebang line */
5287 #endif /* ALTERNATE_SHEBANG */
5292 !instr(s,"indir") &&
5293 instr(PL_origargv[0],"perl"))
5300 while (s < PL_bufend && isSPACE(*s))
5302 if (s < PL_bufend) {
5303 Newx(newargv,PL_origargc+3,char*);
5305 while (s < PL_bufend && !isSPACE(*s))
5308 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5311 newargv = PL_origargv;
5314 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5316 Perl_croak(aTHX_ "Can't exec %s", ipath);
5319 while (*d && !isSPACE(*d))
5321 while (SPACE_OR_TAB(*d))
5325 const bool switches_done = PL_doswitches;
5326 const U32 oldpdb = PL_perldb;
5327 const bool oldn = PL_minus_n;
5328 const bool oldp = PL_minus_p;
5332 bool baduni = FALSE;
5334 const char *d2 = d1 + 1;
5335 if (parse_unicode_opts((const char **)&d2)
5339 if (baduni || *d1 == 'M' || *d1 == 'm') {
5340 const char * const m = d1;
5341 while (*d1 && !isSPACE(*d1))
5343 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5346 d1 = moreswitches(d1);
5348 if (PL_doswitches && !switches_done) {
5349 int argc = PL_origargc;
5350 char **argv = PL_origargv;
5353 } while (argc && argv[0][0] == '-' && argv[0][1]);
5354 init_argv_symbols(argc,argv);
5356 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5357 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5358 /* if we have already added "LINE: while (<>) {",
5359 we must not do it again */
5361 sv_setpvs(PL_linestr, "");
5362 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5363 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5364 PL_last_lop = PL_last_uni = NULL;
5365 PL_preambled = FALSE;
5366 if (PERLDB_LINE || PERLDB_SAVESRC)
5367 (void)gv_fetchfile(PL_origfilename);
5374 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5375 PL_lex_state = LEX_FORMLINE;
5376 start_force(PL_curforce);
5377 NEXTVAL_NEXTTOKE.ival = 0;
5378 force_next(FORMRBRACK);
5383 #ifdef PERL_STRICT_CR
5384 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5386 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5388 case ' ': case '\t': case '\f': case 013:
5390 PL_realtokenstart = -1;
5393 PL_thiswhite = newSVpvs("");
5394 sv_catpvn(PL_thiswhite, s, 1);
5402 PL_realtokenstart = -1;
5406 if (PL_lex_state != LEX_NORMAL ||
5407 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5408 if (*s == '#' && s == PL_linestart && PL_in_eval
5409 && !PL_rsfp && !PL_parser->filtered) {
5410 /* handle eval qq[#line 1 "foo"\n ...] */
5411 CopLINE_dec(PL_curcop);
5414 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5416 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5420 const bool in_comment = *s == '#';
5422 while (d < PL_bufend && *d != '\n')
5426 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5427 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5431 PL_thiswhite = newSVpvn(s, d - s);
5434 if (in_comment && d == PL_bufend
5435 && PL_lex_state == LEX_INTERPNORMAL
5436 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5437 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5440 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5441 PL_lex_state = LEX_FORMLINE;
5442 start_force(PL_curforce);
5443 NEXTVAL_NEXTTOKE.ival = 0;
5444 force_next(FORMRBRACK);
5450 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5451 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5454 TOKEN(PEG); /* make sure any #! line is accessible */
5459 /* if (PL_madskills && PL_lex_formbrack) { */
5461 while (d < PL_bufend && *d != '\n')
5465 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5466 Perl_croak(aTHX_ "panic: input overflow");
5467 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5469 PL_thiswhite = newSVpvs("");
5470 if (CopLINE(PL_curcop) == 1) {
5471 sv_setpvs(PL_thiswhite, "");
5474 sv_catpvn(PL_thiswhite, s, d - s);
5488 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5496 while (s < PL_bufend && SPACE_OR_TAB(*s))
5499 if (strnEQ(s,"=>",2)) {
5500 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5501 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5502 OPERATOR('-'); /* unary minus */
5504 PL_last_uni = PL_oldbufptr;
5506 case 'r': ftst = OP_FTEREAD; break;
5507 case 'w': ftst = OP_FTEWRITE; break;
5508 case 'x': ftst = OP_FTEEXEC; break;
5509 case 'o': ftst = OP_FTEOWNED; break;
5510 case 'R': ftst = OP_FTRREAD; break;
5511 case 'W': ftst = OP_FTRWRITE; break;
5512 case 'X': ftst = OP_FTREXEC; break;
5513 case 'O': ftst = OP_FTROWNED; break;
5514 case 'e': ftst = OP_FTIS; break;
5515 case 'z': ftst = OP_FTZERO; break;
5516 case 's': ftst = OP_FTSIZE; break;
5517 case 'f': ftst = OP_FTFILE; break;
5518 case 'd': ftst = OP_FTDIR; break;
5519 case 'l': ftst = OP_FTLINK; break;
5520 case 'p': ftst = OP_FTPIPE; break;
5521 case 'S': ftst = OP_FTSOCK; break;
5522 case 'u': ftst = OP_FTSUID; break;
5523 case 'g': ftst = OP_FTSGID; break;
5524 case 'k': ftst = OP_FTSVTX; break;
5525 case 'b': ftst = OP_FTBLK; break;
5526 case 'c': ftst = OP_FTCHR; break;
5527 case 't': ftst = OP_FTTTY; break;
5528 case 'T': ftst = OP_FTTEXT; break;
5529 case 'B': ftst = OP_FTBINARY; break;
5530 case 'M': case 'A': case 'C':
5531 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5533 case 'M': ftst = OP_FTMTIME; break;
5534 case 'A': ftst = OP_FTATIME; break;
5535 case 'C': ftst = OP_FTCTIME; break;
5543 PL_last_lop_op = (OPCODE)ftst;
5544 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5545 "### Saw file test %c\n", (int)tmp);
5550 /* Assume it was a minus followed by a one-letter named
5551 * subroutine call (or a -bareword), then. */
5552 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5553 "### '-%c' looked like a file test but was not\n",
5560 const char tmp = *s++;
5563 if (PL_expect == XOPERATOR)
5568 else if (*s == '>') {
5571 if (isIDFIRST_lazy_if(s,UTF)) {
5572 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5580 if (PL_expect == XOPERATOR) {
5581 if (*s == '=' && !PL_lex_allbrackets &&
5582 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5589 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5591 OPERATOR('-'); /* unary minus */
5597 const char tmp = *s++;
5600 if (PL_expect == XOPERATOR)
5605 if (PL_expect == XOPERATOR) {
5606 if (*s == '=' && !PL_lex_allbrackets &&
5607 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5614 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5621 if (PL_expect != XOPERATOR) {
5622 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5623 PL_expect = XOPERATOR;
5624 force_ident(PL_tokenbuf, '*');
5632 if (*s == '=' && !PL_lex_allbrackets &&
5633 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5639 if (*s == '=' && !PL_lex_allbrackets &&
5640 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5647 if (PL_expect == XOPERATOR) {
5648 if (s[1] == '=' && !PL_lex_allbrackets &&
5649 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5654 PL_tokenbuf[0] = '%';
5655 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5656 sizeof PL_tokenbuf - 1, FALSE);
5657 if (!PL_tokenbuf[1]) {
5660 PL_expect = XOPERATOR;
5661 force_ident_maybe_lex('%');
5665 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5666 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5671 if (PL_lex_brackets > 100)
5672 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5673 PL_lex_brackstack[PL_lex_brackets++] = 0;
5674 PL_lex_allbrackets++;
5676 const char tmp = *s++;
5681 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5683 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5691 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5698 goto just_a_word_zero_gv;
5701 switch (PL_expect) {
5707 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5709 PL_bufptr = s; /* update in case we back off */
5712 "Use of := for an empty attribute list is not allowed");
5719 PL_expect = XTERMBLOCK;
5722 stuffstart = s - SvPVX(PL_linestr) - 1;
5726 while (isIDFIRST_lazy_if(s,UTF)) {
5729 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5730 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5731 if (tmp < 0) tmp = -tmp;
5746 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5748 d = scan_str(d,TRUE,TRUE,FALSE);
5750 /* MUST advance bufptr here to avoid bogus
5751 "at end of line" context messages from yyerror().
5753 PL_bufptr = s + len;
5754 yyerror("Unterminated attribute parameter in attribute list");
5758 return REPORT(0); /* EOF indicator */
5762 sv_catsv(sv, PL_lex_stuff);
5763 attrs = op_append_elem(OP_LIST, attrs,
5764 newSVOP(OP_CONST, 0, sv));
5765 SvREFCNT_dec(PL_lex_stuff);
5766 PL_lex_stuff = NULL;
5769 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5771 if (PL_in_my == KEY_our) {
5772 deprecate(":unique");
5775 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5778 /* NOTE: any CV attrs applied here need to be part of
5779 the CVf_BUILTIN_ATTRS define in cv.h! */
5780 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5782 CvLVALUE_on(PL_compcv);
5784 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5786 deprecate(":locked");
5788 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5790 CvMETHOD_on(PL_compcv);
5792 /* After we've set the flags, it could be argued that
5793 we don't need to do the attributes.pm-based setting
5794 process, and shouldn't bother appending recognized
5795 flags. To experiment with that, uncomment the
5796 following "else". (Note that's already been
5797 uncommented. That keeps the above-applied built-in
5798 attributes from being intercepted (and possibly
5799 rejected) by a package's attribute routines, but is
5800 justified by the performance win for the common case
5801 of applying only built-in attributes.) */
5803 attrs = op_append_elem(OP_LIST, attrs,
5804 newSVOP(OP_CONST, 0,
5808 if (*s == ':' && s[1] != ':')
5811 break; /* require real whitespace or :'s */
5812 /* XXX losing whitespace on sequential attributes here */
5816 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5817 if (*s != ';' && *s != '}' && *s != tmp
5818 && (tmp != '=' || *s != ')')) {
5819 const char q = ((*s == '\'') ? '"' : '\'');
5820 /* If here for an expression, and parsed no attrs, back
5822 if (tmp == '=' && !attrs) {
5826 /* MUST advance bufptr here to avoid bogus "at end of line"
5827 context messages from yyerror().
5830 yyerror( (const char *)
5832 ? Perl_form(aTHX_ "Invalid separator character "
5833 "%c%c%c in attribute list", q, *s, q)
5834 : "Unterminated attribute list" ) );
5842 start_force(PL_curforce);
5843 NEXTVAL_NEXTTOKE.opval = attrs;
5844 CURMAD('_', PL_nextwhite);
5849 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5850 (s - SvPVX(PL_linestr)) - stuffstart);
5855 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5859 PL_lex_allbrackets--;
5863 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5864 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5868 PL_lex_allbrackets++;
5871 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5877 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5880 PL_lex_allbrackets--;
5886 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5889 if (PL_lex_brackets <= 0)
5890 yyerror("Unmatched right square bracket");
5893 PL_lex_allbrackets--;
5894 if (PL_lex_state == LEX_INTERPNORMAL) {
5895 if (PL_lex_brackets == 0) {
5896 if (*s == '-' && s[1] == '>')
5897 PL_lex_state = LEX_INTERPENDMAYBE;
5898 else if (*s != '[' && *s != '{')
5899 PL_lex_state = LEX_INTERPEND;
5906 if (PL_lex_brackets > 100) {
5907 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5909 switch (PL_expect) {
5911 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5912 PL_lex_allbrackets++;
5913 OPERATOR(HASHBRACK);
5915 while (s < PL_bufend && SPACE_OR_TAB(*s))
5918 PL_tokenbuf[0] = '\0';
5919 if (d < PL_bufend && *d == '-') {
5920 PL_tokenbuf[0] = '-';
5922 while (d < PL_bufend && SPACE_OR_TAB(*d))
5925 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5926 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5928 while (d < PL_bufend && SPACE_OR_TAB(*d))
5931 const char minus = (PL_tokenbuf[0] == '-');
5932 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5940 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5941 PL_lex_allbrackets++;
5946 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5947 PL_lex_allbrackets++;
5952 if (PL_oldoldbufptr == PL_last_lop)
5953 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5955 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5956 PL_lex_allbrackets++;
5959 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5961 /* This hack is to get the ${} in the message. */
5963 yyerror("syntax error");
5966 OPERATOR(HASHBRACK);
5968 /* This hack serves to disambiguate a pair of curlies
5969 * as being a block or an anon hash. Normally, expectation
5970 * determines that, but in cases where we're not in a
5971 * position to expect anything in particular (like inside
5972 * eval"") we have to resolve the ambiguity. This code
5973 * covers the case where the first term in the curlies is a
5974 * quoted string. Most other cases need to be explicitly
5975 * disambiguated by prepending a "+" before the opening
5976 * curly in order to force resolution as an anon hash.
5978 * XXX should probably propagate the outer expectation
5979 * into eval"" to rely less on this hack, but that could
5980 * potentially break current behavior of eval"".
5984 if (*s == '\'' || *s == '"' || *s == '`') {
5985 /* common case: get past first string, handling escapes */
5986 for (t++; t < PL_bufend && *t != *s;)
5987 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5991 else if (*s == 'q') {
5994 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5997 /* skip q//-like construct */
5999 char open, close, term;
6002 while (t < PL_bufend && isSPACE(*t))
6004 /* check for q => */
6005 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
6006 OPERATOR(HASHBRACK);
6010 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6014 for (t++; t < PL_bufend; t++) {
6015 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
6017 else if (*t == open)
6021 for (t++; t < PL_bufend; t++) {
6022 if (*t == '\\' && t+1 < PL_bufend)
6024 else if (*t == close && --brackets <= 0)
6026 else if (*t == open)
6033 /* skip plain q word */
6034 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
6037 else if (isALNUM_lazy_if(t,UTF)) {
6039 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
6042 while (t < PL_bufend && isSPACE(*t))
6044 /* if comma follows first term, call it an anon hash */
6045 /* XXX it could be a comma expression with loop modifiers */
6046 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
6047 || (*t == '=' && t[1] == '>')))
6048 OPERATOR(HASHBRACK);
6049 if (PL_expect == XREF)
6052 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
6058 pl_yylval.ival = CopLINE(PL_curcop);
6059 if (isSPACE(*s) || *s == '#')
6060 PL_copline = NOLINE; /* invalidate current command line number */
6061 TOKEN(formbrack ? '=' : '{');
6063 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
6067 if (PL_lex_brackets <= 0)
6068 yyerror("Unmatched right curly bracket");
6070 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
6071 PL_lex_allbrackets--;
6072 if (PL_lex_state == LEX_INTERPNORMAL) {
6073 if (PL_lex_brackets == 0) {
6074 if (PL_expect & XFAKEBRACK) {
6075 PL_expect &= XENUMMASK;
6076 PL_lex_state = LEX_INTERPEND;
6081 PL_thiswhite = newSVpvs("");
6082 sv_catpvs(PL_thiswhite,"}");
6085 return yylex(); /* ignore fake brackets */
6087 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6088 && SvEVALED(PL_lex_repl))
6089 PL_lex_state = LEX_INTERPEND;
6090 else if (*s == '-' && s[1] == '>')
6091 PL_lex_state = LEX_INTERPENDMAYBE;
6092 else if (*s != '[' && *s != '{')
6093 PL_lex_state = LEX_INTERPEND;
6096 if (PL_expect & XFAKEBRACK) {
6097 PL_expect &= XENUMMASK;
6099 return yylex(); /* ignore fake brackets */
6101 start_force(PL_curforce);
6103 curmad('X', newSVpvn(s-1,1));
6104 CURMAD('_', PL_thiswhite);
6106 force_next(formbrack ? '.' : '}');
6107 if (formbrack) LEAVE;
6109 if (PL_madskills && !PL_thistoken)
6110 PL_thistoken = newSVpvs("");
6112 if (formbrack == 2) { /* means . where arguments were expected */
6113 start_force(PL_curforce);
6121 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6122 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6129 if (PL_expect == XOPERATOR) {
6130 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
6131 && isIDFIRST_lazy_if(s,UTF))
6133 CopLINE_dec(PL_curcop);
6134 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6135 CopLINE_inc(PL_curcop);
6137 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6138 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6145 PL_tokenbuf[0] = '&';
6146 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf + 1,
6147 sizeof PL_tokenbuf - 1, TRUE);
6148 if (PL_tokenbuf[1]) {
6149 PL_expect = XOPERATOR;
6150 force_ident_maybe_lex('&');
6154 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
6160 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6161 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6168 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6169 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6177 const char tmp = *s++;
6179 if (!PL_lex_allbrackets &&
6180 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6187 if (!PL_lex_allbrackets &&
6188 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
6196 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6197 && strchr("+-*/%.^&|<",tmp))
6198 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6199 "Reversed %c= operator",(int)tmp);
6201 if (PL_expect == XSTATE && isALPHA(tmp) &&
6202 (s == PL_linestart+1 || s[-2] == '\n') )
6204 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6205 || PL_lex_state != LEX_NORMAL) {
6210 if (strnEQ(s,"=cut",4)) {
6226 PL_thiswhite = newSVpvs("");
6227 sv_catpvn(PL_thiswhite, PL_linestart,
6228 PL_bufend - PL_linestart);
6232 PL_parser->in_pod = 1;
6236 if (PL_expect == XBLOCK) {
6238 #ifdef PERL_STRICT_CR
6239 while (SPACE_OR_TAB(*t))
6241 while (SPACE_OR_TAB(*t) || *t == '\r')
6244 if (*t == '\n' || *t == '#') {
6247 SAVEI8(PL_parser->form_lex_state);
6248 SAVEI32(PL_lex_formbrack);
6249 PL_parser->form_lex_state = PL_lex_state;
6250 PL_lex_formbrack = PL_lex_brackets + 1;
6254 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6263 const char tmp = *s++;
6265 /* was this !=~ where !~ was meant?
6266 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6268 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6269 const char *t = s+1;
6271 while (t < PL_bufend && isSPACE(*t))
6274 if (*t == '/' || *t == '?' ||
6275 ((*t == 'm' || *t == 's' || *t == 'y')
6276 && !isALNUM(t[1])) ||
6277 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
6278 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6279 "!=~ should be !~");
6281 if (!PL_lex_allbrackets &&
6282 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6294 if (PL_expect != XOPERATOR) {
6295 if (s[1] != '<' && !strchr(s,'>'))
6298 s = scan_heredoc(s);
6300 s = scan_inputsymbol(s);
6301 PL_expect = XOPERATOR;
6302 TOKEN(sublex_start());
6308 if (*s == '=' && !PL_lex_allbrackets &&
6309 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6313 SHop(OP_LEFT_SHIFT);
6318 if (!PL_lex_allbrackets &&
6319 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6326 if (!PL_lex_allbrackets &&
6327 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6335 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6343 const char tmp = *s++;
6345 if (*s == '=' && !PL_lex_allbrackets &&
6346 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6350 SHop(OP_RIGHT_SHIFT);
6352 else if (tmp == '=') {
6353 if (!PL_lex_allbrackets &&
6354 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6362 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6371 if (PL_expect == XOPERATOR) {
6372 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6373 return deprecate_commaless_var_list();
6377 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6378 PL_tokenbuf[0] = '@';
6379 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6380 sizeof PL_tokenbuf - 1, FALSE);
6381 if (PL_expect == XOPERATOR)
6382 no_op("Array length", s);
6383 if (!PL_tokenbuf[1])
6385 PL_expect = XOPERATOR;
6386 force_ident_maybe_lex('#');
6390 PL_tokenbuf[0] = '$';
6391 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6392 sizeof PL_tokenbuf - 1, FALSE);
6393 if (PL_expect == XOPERATOR)
6395 if (!PL_tokenbuf[1]) {
6397 yyerror("Final $ should be \\$ or $name");
6403 const char tmp = *s;
6404 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6407 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6408 && intuit_more(s)) {
6410 PL_tokenbuf[0] = '@';
6411 if (ckWARN(WARN_SYNTAX)) {
6414 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6417 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6418 while (t < PL_bufend && *t != ']')
6420 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6421 "Multidimensional syntax %.*s not supported",
6422 (int)((t - PL_bufptr) + 1), PL_bufptr);
6426 else if (*s == '{') {
6428 PL_tokenbuf[0] = '%';
6429 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6430 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6432 char tmpbuf[sizeof PL_tokenbuf];
6435 } while (isSPACE(*t));
6436 if (isIDFIRST_lazy_if(t,UTF)) {
6438 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6443 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6444 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6445 "You need to quote \"%"SVf"\"",
6446 SVfARG(newSVpvn_flags(tmpbuf, len,
6447 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
6453 PL_expect = XOPERATOR;
6454 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6455 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6456 if (!islop || PL_last_lop_op == OP_GREPSTART)
6457 PL_expect = XOPERATOR;
6458 else if (strchr("$@\"'`q", *s))
6459 PL_expect = XTERM; /* e.g. print $fh "foo" */
6460 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6461 PL_expect = XTERM; /* e.g. print $fh &sub */
6462 else if (isIDFIRST_lazy_if(s,UTF)) {
6463 char tmpbuf[sizeof PL_tokenbuf];
6465 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6466 if ((t2 = keyword(tmpbuf, len, 0))) {
6467 /* binary operators exclude handle interpretations */
6479 PL_expect = XTERM; /* e.g. print $fh length() */
6484 PL_expect = XTERM; /* e.g. print $fh subr() */
6487 else if (isDIGIT(*s))
6488 PL_expect = XTERM; /* e.g. print $fh 3 */
6489 else if (*s == '.' && isDIGIT(s[1]))
6490 PL_expect = XTERM; /* e.g. print $fh .3 */
6491 else if ((*s == '?' || *s == '-' || *s == '+')
6492 && !isSPACE(s[1]) && s[1] != '=')
6493 PL_expect = XTERM; /* e.g. print $fh -1 */
6494 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6496 PL_expect = XTERM; /* e.g. print $fh /.../
6497 XXX except DORDOR operator
6499 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6501 PL_expect = XTERM; /* print $fh <<"EOF" */
6504 force_ident_maybe_lex('$');
6508 if (PL_expect == XOPERATOR)
6510 PL_tokenbuf[0] = '@';
6511 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6512 if (!PL_tokenbuf[1]) {
6515 if (PL_lex_state == LEX_NORMAL)
6517 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6519 PL_tokenbuf[0] = '%';
6521 /* Warn about @ where they meant $. */
6522 if (*s == '[' || *s == '{') {
6523 if (ckWARN(WARN_SYNTAX)) {
6524 const char *t = s + 1;
6525 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6526 t += UTF ? UTF8SKIP(t) : 1;
6527 if (*t == '}' || *t == ']') {
6529 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6530 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6531 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6532 "Scalar value %"SVf" better written as $%"SVf,
6533 SVfARG(newSVpvn_flags(PL_bufptr, (STRLEN)(t-PL_bufptr),
6534 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))),
6535 SVfARG(newSVpvn_flags(PL_bufptr+1, (STRLEN)(t-PL_bufptr-1),
6536 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))));
6541 PL_expect = XOPERATOR;
6542 force_ident_maybe_lex('@');
6545 case '/': /* may be division, defined-or, or pattern */
6546 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6547 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6548 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6553 case '?': /* may either be conditional or pattern */
6554 if (PL_expect == XOPERATOR) {
6557 if (!PL_lex_allbrackets &&
6558 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6562 PL_lex_allbrackets++;
6568 /* A // operator. */
6569 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6570 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6571 LEX_FAKEEOF_LOGIC)) {
6579 if (*s == '=' && !PL_lex_allbrackets &&
6580 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6589 /* Disable warning on "study /blah/" */
6590 if (PL_oldoldbufptr == PL_last_uni
6591 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6592 || memNE(PL_last_uni, "study", 5)
6593 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6597 deprecate("?PATTERN? without explicit operator");
6598 s = scan_pat(s,OP_MATCH);
6599 TERM(sublex_start());
6603 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6604 #ifdef PERL_STRICT_CR
6607 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6609 && (s == PL_linestart || s[-1] == '\n') )
6612 formbrack = 2; /* dot seen where arguments expected */
6615 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6619 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6622 if (!PL_lex_allbrackets &&
6623 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6630 pl_yylval.ival = OPf_SPECIAL;
6636 if (*s == '=' && !PL_lex_allbrackets &&
6637 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6644 case '0': case '1': case '2': case '3': case '4':
6645 case '5': case '6': case '7': case '8': case '9':
6646 s = scan_num(s, &pl_yylval);
6647 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6648 if (PL_expect == XOPERATOR)
6653 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6654 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6655 if (PL_expect == XOPERATOR) {
6656 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6657 return deprecate_commaless_var_list();
6664 pl_yylval.ival = OP_CONST;
6665 TERM(sublex_start());
6668 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6669 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6670 if (PL_expect == XOPERATOR) {
6671 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6672 return deprecate_commaless_var_list();
6679 pl_yylval.ival = OP_CONST;
6680 /* FIXME. I think that this can be const if char *d is replaced by
6681 more localised variables. */
6682 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6683 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6684 pl_yylval.ival = OP_STRINGIFY;
6688 TERM(sublex_start());
6691 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6692 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6693 if (PL_expect == XOPERATOR)
6694 no_op("Backticks",s);
6697 readpipe_override();
6698 TERM(sublex_start());
6702 if (PL_lex_inwhat && isDIGIT(*s))
6703 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6705 if (PL_expect == XOPERATOR)
6706 no_op("Backslash",s);
6710 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6711 char *start = s + 2;
6712 while (isDIGIT(*start) || *start == '_')
6714 if (*start == '.' && isDIGIT(start[1])) {
6715 s = scan_num(s, &pl_yylval);
6718 else if ((*start == ':' && start[1] == ':')
6719 || (PL_expect == XSTATE && *start == ':'))
6721 else if (PL_expect == XSTATE) {
6723 while (d < PL_bufend && isSPACE(*d)) d++;
6724 if (*d == ':') goto keylookup;
6726 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6727 if (!isALPHA(*start) && (PL_expect == XTERM
6728 || PL_expect == XREF || PL_expect == XSTATE
6729 || PL_expect == XTERMORDORDOR)) {
6730 GV *const gv = gv_fetchpvn_flags(s, start - s,
6731 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6733 s = scan_num(s, &pl_yylval);
6740 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6793 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6795 /* Some keywords can be followed by any delimiter, including ':' */
6796 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6798 /* x::* is just a word, unless x is "CORE" */
6799 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6803 while (d < PL_bufend && isSPACE(*d))
6804 d++; /* no comments skipped here, or s### is misparsed */
6806 /* Is this a word before a => operator? */
6807 if (*d == '=' && d[1] == '>') {
6810 = (OP*)newSVOP(OP_CONST, 0,
6811 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6812 pl_yylval.opval->op_private = OPpCONST_BARE;
6816 /* Check for plugged-in keyword */
6820 char *saved_bufptr = PL_bufptr;
6822 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6824 if (result == KEYWORD_PLUGIN_DECLINE) {
6825 /* not a plugged-in keyword */
6826 PL_bufptr = saved_bufptr;
6827 } else if (result == KEYWORD_PLUGIN_STMT) {
6828 pl_yylval.opval = o;
6831 return REPORT(PLUGSTMT);
6832 } else if (result == KEYWORD_PLUGIN_EXPR) {
6833 pl_yylval.opval = o;
6835 PL_expect = XOPERATOR;
6836 return REPORT(PLUGEXPR);
6838 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6843 /* Check for built-in keyword */
6844 tmp = keyword(PL_tokenbuf, len, 0);
6846 /* Is this a label? */
6847 if (!anydelim && PL_expect == XSTATE
6848 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6850 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6851 pl_yylval.pval[len] = '\0';
6852 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6857 /* Check for lexical sub */
6858 if (PL_expect != XOPERATOR) {
6859 char tmpbuf[sizeof PL_tokenbuf + 1];
6861 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6862 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6863 if (off != NOT_IN_PAD) {
6864 assert(off); /* we assume this is boolean-true below */
6865 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6866 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6867 HEK * const stashname = HvNAME_HEK(stash);
6868 sv = newSVhek(stashname);
6869 sv_catpvs(sv, "::");
6870 sv_catpvn_flags(sv, PL_tokenbuf, len,
6871 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6872 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6877 rv2cv_op = newOP(OP_PADANY, 0);
6878 rv2cv_op->op_targ = off;
6879 rv2cv_op = (OP*)newCVREF(0, rv2cv_op);
6880 cv = (CV *)PAD_SV(off);
6888 if (tmp < 0) { /* second-class keyword? */
6889 GV *ogv = NULL; /* override (winner) */
6890 GV *hgv = NULL; /* hidden (loser) */
6891 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6893 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6894 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6897 if (GvIMPORTED_CV(gv))
6899 else if (! CvMETHOD(cv))
6903 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6904 UTF ? -(I32)len : (I32)len, FALSE)) &&
6905 (gv = *gvp) && isGV_with_GP(gv) &&
6906 GvCVu(gv) && GvIMPORTED_CV(gv))
6913 tmp = 0; /* overridden by import or by GLOBAL */
6916 && -tmp==KEY_lock /* XXX generalizable kludge */
6919 tmp = 0; /* any sub overrides "weak" keyword */
6921 else { /* no override */
6923 if (tmp == KEY_dump) {
6924 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6925 "dump() better written as CORE::dump()");
6929 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6930 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6931 "Ambiguous call resolved as CORE::%s(), "
6932 "qualify as such or use &",
6940 default: /* not a keyword */
6941 /* Trade off - by using this evil construction we can pull the
6942 variable gv into the block labelled keylookup. If not, then
6943 we have to give it function scope so that the goto from the
6944 earlier ':' case doesn't bypass the initialisation. */
6946 just_a_word_zero_gv:
6958 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6959 const char penultchar =
6960 lastchar && PL_bufptr - 2 >= PL_linestart
6964 SV *nextPL_nextwhite = 0;
6968 /* Get the rest if it looks like a package qualifier */
6970 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6972 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6975 Perl_croak(aTHX_ "Bad name after %"SVf"%s",
6976 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6977 (UTF ? SVf_UTF8 : 0) | SVs_TEMP )),
6978 *s == '\'' ? "'" : "::");
6983 if (PL_expect == XOPERATOR) {
6984 if (PL_bufptr == PL_linestart) {
6985 CopLINE_dec(PL_curcop);
6986 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6987 CopLINE_inc(PL_curcop);
6990 no_op("Bareword",s);
6993 /* Look for a subroutine with this name in current package,
6994 unless this is a lexical sub, or name is "Foo::",
6995 in which case Foo is a bareword
6996 (and a package name). */
6998 if (len > 2 && !PL_madskills &&
6999 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
7001 if (ckWARN(WARN_BAREWORD)
7002 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
7003 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
7004 "Bareword \"%"SVf"\" refers to nonexistent package",
7005 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7006 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7008 PL_tokenbuf[len] = '\0';
7014 /* Mustn't actually add anything to a symbol table.
7015 But also don't want to "initialise" any placeholder
7016 constants that might already be there into full
7017 blown PVGVs with attached PVCV. */
7018 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
7019 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
7025 /* if we saw a global override before, get the right name */
7028 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
7029 len ? len : strlen(PL_tokenbuf));
7031 SV * const tmp_sv = sv;
7032 sv = newSVpvs("CORE::GLOBAL::");
7033 sv_catsv(sv, tmp_sv);
7034 SvREFCNT_dec(tmp_sv);
7038 if (PL_madskills && !PL_thistoken) {
7039 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
7040 PL_thistoken = newSVpvn(start,s - start);
7041 PL_realtokenstart = s - SvPVX(PL_linestr);
7045 /* Presume this is going to be a bareword of some sort. */
7047 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
7048 pl_yylval.opval->op_private = OPpCONST_BARE;
7050 /* And if "Foo::", then that's what it certainly is. */
7056 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
7057 const_op->op_private = OPpCONST_BARE;
7058 rv2cv_op = newCVREF(0, const_op);
7059 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
7062 /* See if it's the indirect object for a list operator. */
7064 if (PL_oldoldbufptr &&
7065 PL_oldoldbufptr < PL_bufptr &&
7066 (PL_oldoldbufptr == PL_last_lop
7067 || PL_oldoldbufptr == PL_last_uni) &&
7068 /* NO SKIPSPACE BEFORE HERE! */
7069 (PL_expect == XREF ||
7070 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
7072 bool immediate_paren = *s == '(';
7074 /* (Now we can afford to cross potential line boundary.) */
7075 s = SKIPSPACE2(s,nextPL_nextwhite);
7077 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
7080 /* Two barewords in a row may indicate method call. */
7082 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
7083 (tmp = intuit_method(s, gv, cv))) {
7085 if (tmp == METHOD && !PL_lex_allbrackets &&
7086 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7087 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7091 /* If not a declared subroutine, it's an indirect object. */
7092 /* (But it's an indir obj regardless for sort.) */
7093 /* Also, if "_" follows a filetest operator, it's a bareword */
7096 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
7098 (PL_last_lop_op != OP_MAPSTART &&
7099 PL_last_lop_op != OP_GREPSTART))))
7100 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
7101 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
7104 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
7109 PL_expect = XOPERATOR;
7112 s = SKIPSPACE2(s,nextPL_nextwhite);
7113 PL_nextwhite = nextPL_nextwhite;
7118 /* Is this a word before a => operator? */
7119 if (*s == '=' && s[1] == '>' && !pkgname) {
7122 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
7123 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
7124 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
7128 /* If followed by a paren, it's certainly a subroutine. */
7133 while (SPACE_OR_TAB(*d))
7135 if (*d == ')' && (sv = cv_const_sv(cv))) {
7142 PL_nextwhite = PL_thiswhite;
7145 start_force(PL_curforce);
7147 NEXTVAL_NEXTTOKE.opval =
7148 off ? rv2cv_op : pl_yylval.opval;
7149 PL_expect = XOPERATOR;
7152 PL_nextwhite = nextPL_nextwhite;
7153 curmad('X', PL_thistoken);
7154 PL_thistoken = newSVpvs("");
7158 op_free(pl_yylval.opval), force_next(PRIVATEREF);
7159 else op_free(rv2cv_op), force_next(WORD);
7164 /* If followed by var or block, call it a method (unless sub) */
7166 if ((*s == '$' || *s == '{') && !cv) {
7168 PL_last_lop = PL_oldbufptr;
7169 PL_last_lop_op = OP_METHOD;
7170 if (!PL_lex_allbrackets &&
7171 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7172 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7176 /* If followed by a bareword, see if it looks like indir obj. */
7179 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7180 && (tmp = intuit_method(s, gv, cv))) {
7182 if (tmp == METHOD && !PL_lex_allbrackets &&
7183 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7184 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7188 /* Not a method, so call it a subroutine (if defined) */
7191 if (lastchar == '-' && penultchar != '-') {
7192 const SV *tmpsv = newSVpvn_flags( PL_tokenbuf, len ? len : strlen(PL_tokenbuf), (UTF ? SVf_UTF8 : 0) | SVs_TEMP );
7193 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7194 "Ambiguous use of -%"SVf" resolved as -&%"SVf"()",
7195 SVfARG(tmpsv), SVfARG(tmpsv));
7197 /* Check for a constant sub */
7198 if ((sv = cv_const_sv(cv))) {
7201 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7202 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7203 pl_yylval.opval->op_private = OPpCONST_FOLDED;
7204 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7208 op_free(pl_yylval.opval);
7209 pl_yylval.opval = rv2cv_op;
7210 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7211 PL_last_lop = PL_oldbufptr;
7212 PL_last_lop_op = OP_ENTERSUB;
7213 /* Is there a prototype? */
7220 STRLEN protolen = CvPROTOLEN(cv);
7221 const char *proto = CvPROTO(cv);
7225 if ((optional = *proto == ';'))
7228 while (*proto == ';');
7232 *proto == '$' || *proto == '_'
7233 || *proto == '*' || *proto == '+'
7238 *proto == '\\' && proto[1] && proto[2] == '\0'
7241 UNIPROTO(UNIOPSUB,optional);
7242 if (*proto == '\\' && proto[1] == '[') {
7243 const char *p = proto + 2;
7244 while(*p && *p != ']')
7246 if(*p == ']' && !p[1])
7247 UNIPROTO(UNIOPSUB,optional);
7249 if (*proto == '&' && *s == '{') {
7251 sv_setpvs(PL_subname, "__ANON__");
7253 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7254 if (!PL_lex_allbrackets &&
7255 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7256 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7263 PL_nextwhite = PL_thiswhite;
7266 start_force(PL_curforce);
7267 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7270 PL_nextwhite = nextPL_nextwhite;
7271 curmad('X', PL_thistoken);
7272 PL_thistoken = newSVpvs("");
7274 force_next(off ? PRIVATEREF : WORD);
7275 if (!PL_lex_allbrackets &&
7276 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7277 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7282 /* Guess harder when madskills require "best effort". */
7283 if (PL_madskills && (!gv || !GvCVu(gv))) {
7284 int probable_sub = 0;
7285 if (strchr("\"'`$@%0123456789!*+{[<", *s))
7287 else if (isALPHA(*s)) {
7291 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
7292 if (!keyword(tmpbuf, tmplen, 0))
7295 while (d < PL_bufend && isSPACE(*d))
7297 if (*d == '=' && d[1] == '>')
7302 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
7304 op_free(pl_yylval.opval);
7305 pl_yylval.opval = rv2cv_op;
7306 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7307 PL_last_lop = PL_oldbufptr;
7308 PL_last_lop_op = OP_ENTERSUB;
7309 PL_nextwhite = PL_thiswhite;
7311 start_force(PL_curforce);
7312 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7314 PL_nextwhite = nextPL_nextwhite;
7315 curmad('X', PL_thistoken);
7316 PL_thistoken = newSVpvs("");
7317 force_next(off ? PRIVATEREF : WORD);
7318 if (!PL_lex_allbrackets &&
7319 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7320 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7324 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7326 force_next(off ? PRIVATEREF : WORD);
7327 if (!PL_lex_allbrackets &&
7328 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7329 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7334 /* Call it a bare word */
7336 if (PL_hints & HINT_STRICT_SUBS)
7337 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7340 /* after "print" and similar functions (corresponding to
7341 * "F? L" in opcode.pl), whatever wasn't already parsed as
7342 * a filehandle should be subject to "strict subs".
7343 * Likewise for the optional indirect-object argument to system
7344 * or exec, which can't be a bareword */
7345 if ((PL_last_lop_op == OP_PRINT
7346 || PL_last_lop_op == OP_PRTF
7347 || PL_last_lop_op == OP_SAY
7348 || PL_last_lop_op == OP_SYSTEM
7349 || PL_last_lop_op == OP_EXEC)
7350 && (PL_hints & HINT_STRICT_SUBS))
7351 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7352 if (lastchar != '-') {
7353 if (ckWARN(WARN_RESERVED)) {
7357 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7358 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7366 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7367 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7368 "Operator or semicolon missing before %c%"SVf,
7369 lastchar, SVfARG(newSVpvn_flags(PL_tokenbuf,
7370 strlen(PL_tokenbuf),
7371 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
7372 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7373 "Ambiguous use of %c resolved as operator %c",
7374 lastchar, lastchar);
7381 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7386 (OP*)newSVOP(OP_CONST, 0,
7387 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7390 case KEY___PACKAGE__:
7392 (OP*)newSVOP(OP_CONST, 0,
7394 ? newSVhek(HvNAME_HEK(PL_curstash))
7401 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7402 const char *pname = "main";
7405 if (PL_tokenbuf[2] == 'D')
7408 PL_curstash ? PL_curstash : PL_defstash;
7409 pname = HvNAME_get(stash);
7410 plen = HvNAMELEN (stash);
7411 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7413 gv = gv_fetchpvn_flags(
7414 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7415 plen+6, GV_ADD|putf8, SVt_PVIO
7419 GvIOp(gv) = newIO();
7420 IoIFP(GvIOp(gv)) = PL_rsfp;
7421 #if defined(HAS_FCNTL) && defined(F_SETFD)
7423 const int fd = PerlIO_fileno(PL_rsfp);
7424 fcntl(fd,F_SETFD,fd >= 3);
7427 /* Mark this internal pseudo-handle as clean */
7428 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7429 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7430 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7432 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7433 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7434 /* if the script was opened in binmode, we need to revert
7435 * it to text mode for compatibility; but only iff it has CRs
7436 * XXX this is a questionable hack at best. */
7437 if (PL_bufend-PL_bufptr > 2
7438 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7441 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7442 loc = PerlIO_tell(PL_rsfp);
7443 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7446 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7448 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7449 #endif /* NETWARE */
7451 PerlIO_seek(PL_rsfp, loc, 0);
7455 #ifdef PERLIO_LAYERS
7458 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7459 else if (PL_encoding) {
7466 XPUSHs(PL_encoding);
7468 call_method("name", G_SCALAR);
7472 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7473 Perl_form(aTHX_ ":encoding(%"SVf")",
7482 if (PL_realtokenstart >= 0) {
7483 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7485 PL_endwhite = newSVpvs("");
7486 sv_catsv(PL_endwhite, PL_thiswhite);
7488 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7489 PL_realtokenstart = -1;
7491 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7501 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7510 if (PL_expect == XSTATE) {
7517 if (*s == ':' && s[1] == ':') {
7521 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7522 if ((*s == ':' && s[1] == ':')
7523 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7527 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7531 Perl_croak(aTHX_ "CORE::%"SVf" is not a keyword",
7532 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7533 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7536 else if (tmp == KEY_require || tmp == KEY_do
7538 /* that's a way to remember we saw "CORE::" */
7551 LOP(OP_ACCEPT,XTERM);
7554 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7559 LOP(OP_ATAN2,XTERM);
7565 LOP(OP_BINMODE,XTERM);
7568 LOP(OP_BLESS,XTERM);
7577 /* We have to disambiguate the two senses of
7578 "continue". If the next token is a '{' then
7579 treat it as the start of a continue block;
7580 otherwise treat it as a control operator.
7590 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7600 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7609 if (!PL_cryptseen) {
7610 PL_cryptseen = TRUE;
7614 LOP(OP_CRYPT,XTERM);
7617 LOP(OP_CHMOD,XTERM);
7620 LOP(OP_CHOWN,XTERM);
7623 LOP(OP_CONNECT,XTERM);
7643 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7645 if (len && !keyword(PL_tokenbuf + 1, len, 0)) {
7648 force_ident_maybe_lex('&');
7653 if (orig_keyword == KEY_do) {
7662 PL_hints |= HINT_BLOCK_SCOPE;
7672 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7673 STR_WITH_LEN("NDBM_File::"),
7674 STR_WITH_LEN("DB_File::"),
7675 STR_WITH_LEN("GDBM_File::"),
7676 STR_WITH_LEN("SDBM_File::"),
7677 STR_WITH_LEN("ODBM_File::"),
7679 LOP(OP_DBMOPEN,XTERM);
7685 PL_expect = XOPERATOR;
7686 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7693 pl_yylval.ival = CopLINE(PL_curcop);
7697 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7711 if (*s == '{') { /* block eval */
7712 PL_expect = XTERMBLOCK;
7713 UNIBRACK(OP_ENTERTRY);
7715 else { /* string eval */
7717 UNIBRACK(OP_ENTEREVAL);
7722 UNIBRACK(-OP_ENTEREVAL);
7736 case KEY_endhostent:
7742 case KEY_endservent:
7745 case KEY_endprotoent:
7756 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7758 pl_yylval.ival = CopLINE(PL_curcop);
7760 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7763 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7766 if ((PL_bufend - p) >= 3 &&
7767 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7769 else if ((PL_bufend - p) >= 4 &&
7770 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7773 if (isIDFIRST_lazy_if(p,UTF)) {
7774 p = scan_ident(p, PL_bufend,
7775 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7779 Perl_croak(aTHX_ "Missing $ on loop variable");
7781 s = SvPVX(PL_linestr) + soff;
7787 LOP(OP_FORMLINE,XTERM);
7796 LOP(OP_FCNTL,XTERM);
7802 LOP(OP_FLOCK,XTERM);
7805 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7810 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7815 LOP(OP_GREPSTART, XREF);
7818 PL_expect = XOPERATOR;
7819 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7834 case KEY_getpriority:
7835 LOP(OP_GETPRIORITY,XTERM);
7837 case KEY_getprotobyname:
7840 case KEY_getprotobynumber:
7841 LOP(OP_GPBYNUMBER,XTERM);
7843 case KEY_getprotoent:
7855 case KEY_getpeername:
7856 UNI(OP_GETPEERNAME);
7858 case KEY_gethostbyname:
7861 case KEY_gethostbyaddr:
7862 LOP(OP_GHBYADDR,XTERM);
7864 case KEY_gethostent:
7867 case KEY_getnetbyname:
7870 case KEY_getnetbyaddr:
7871 LOP(OP_GNBYADDR,XTERM);
7876 case KEY_getservbyname:
7877 LOP(OP_GSBYNAME,XTERM);
7879 case KEY_getservbyport:
7880 LOP(OP_GSBYPORT,XTERM);
7882 case KEY_getservent:
7885 case KEY_getsockname:
7886 UNI(OP_GETSOCKNAME);
7888 case KEY_getsockopt:
7889 LOP(OP_GSOCKOPT,XTERM);
7904 pl_yylval.ival = CopLINE(PL_curcop);
7909 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7917 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7919 pl_yylval.ival = CopLINE(PL_curcop);
7923 LOP(OP_INDEX,XTERM);
7929 LOP(OP_IOCTL,XTERM);
7941 PL_expect = XOPERATOR;
7942 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7959 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7964 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7978 LOP(OP_LISTEN,XTERM);
7987 s = scan_pat(s,OP_MATCH);
7988 TERM(sublex_start());
7991 LOP(OP_MAPSTART, XREF);
7994 LOP(OP_MKDIR,XTERM);
7997 LOP(OP_MSGCTL,XTERM);
8000 LOP(OP_MSGGET,XTERM);
8003 LOP(OP_MSGRCV,XTERM);
8006 LOP(OP_MSGSND,XTERM);
8011 PL_in_my = (U16)tmp;
8013 if (isIDFIRST_lazy_if(s,UTF)) {
8017 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
8018 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
8020 if (!FEATURE_LEXSUBS_IS_ENABLED)
8022 "Experimental \"%s\" subs not enabled",
8023 tmp == KEY_my ? "my" :
8024 tmp == KEY_state ? "state" : "our");
8027 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
8028 if (!PL_in_my_stash) {
8031 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
8032 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
8035 if (PL_madskills) { /* just add type to declarator token */
8036 sv_catsv(PL_thistoken, PL_nextwhite);
8038 sv_catpvn(PL_thistoken, start, s - start);
8046 PL_expect = XOPERATOR;
8047 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8051 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
8056 s = tokenize_use(0, s);
8060 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
8063 if (!PL_lex_allbrackets &&
8064 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
8065 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
8071 if (isIDFIRST_lazy_if(s,UTF)) {
8073 for (d = s; isALNUM_lazy_if(d,UTF);) {
8074 d += UTF ? UTF8SKIP(d) : 1;
8076 while (UTF8_IS_CONTINUED(*d) && is_utf8_mark((U8*)d)) {
8077 d += UTF ? UTF8SKIP(d) : 1;
8081 for (t=d; isSPACE(*t);)
8083 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
8085 && !(t[0] == '=' && t[1] == '>')
8086 && !(t[0] == ':' && t[1] == ':')
8087 && !keyword(s, d-s, 0)
8089 SV *tmpsv = newSVpvn_flags(s, (STRLEN)(d-s),
8090 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
8091 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
8092 "Precedence problem: open %"SVf" should be open(%"SVf")",
8093 SVfARG(tmpsv), SVfARG(tmpsv));
8099 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8101 pl_yylval.ival = OP_OR;
8111 LOP(OP_OPEN_DIR,XTERM);
8114 checkcomma(s,PL_tokenbuf,"filehandle");
8118 checkcomma(s,PL_tokenbuf,"filehandle");
8137 s = force_word(s,WORD,FALSE,TRUE,FALSE);
8139 s = force_strict_version(s);
8140 PL_lex_expect = XBLOCK;
8144 LOP(OP_PIPE_OP,XTERM);
8147 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8150 pl_yylval.ival = OP_CONST;
8151 TERM(sublex_start());
8158 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8161 PL_expect = XOPERATOR;
8162 if (SvCUR(PL_lex_stuff)) {
8163 int warned_comma = !ckWARN(WARN_QW);
8164 int warned_comment = warned_comma;
8165 d = SvPV_force(PL_lex_stuff, len);
8167 for (; isSPACE(*d) && len; --len, ++d)
8172 if (!warned_comma || !warned_comment) {
8173 for (; !isSPACE(*d) && len; --len, ++d) {
8174 if (!warned_comma && *d == ',') {
8175 Perl_warner(aTHX_ packWARN(WARN_QW),
8176 "Possible attempt to separate words with commas");
8179 else if (!warned_comment && *d == '#') {
8180 Perl_warner(aTHX_ packWARN(WARN_QW),
8181 "Possible attempt to put comments in qw() list");
8187 for (; !isSPACE(*d) && len; --len, ++d)
8190 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
8191 words = op_append_elem(OP_LIST, words,
8192 newSVOP(OP_CONST, 0, tokeq(sv)));
8197 words = newNULLLIST();
8199 SvREFCNT_dec(PL_lex_stuff);
8200 PL_lex_stuff = NULL;
8202 PL_expect = XOPERATOR;
8203 pl_yylval.opval = sawparens(words);
8208 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8211 pl_yylval.ival = OP_STRINGIFY;
8212 if (SvIVX(PL_lex_stuff) == '\'')
8213 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
8214 TERM(sublex_start());
8217 s = scan_pat(s,OP_QR);
8218 TERM(sublex_start());
8221 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8224 readpipe_override();
8225 TERM(sublex_start());
8232 PL_expect = XOPERATOR;
8234 s = force_version(s, FALSE);
8236 else if (*s != 'v' || !isDIGIT(s[1])
8237 || (s = force_version(s, TRUE), *s == 'v'))
8239 *PL_tokenbuf = '\0';
8240 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8241 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8242 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8243 GV_ADD | (UTF ? SVf_UTF8 : 0));
8245 yyerror("<> should be quotes");
8247 if (orig_keyword == KEY_require) {
8255 PL_last_uni = PL_oldbufptr;
8256 PL_last_lop_op = OP_REQUIRE;
8258 return REPORT( (int)REQUIRE );
8264 PL_expect = XOPERATOR;
8265 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8269 LOP(OP_RENAME,XTERM);
8278 LOP(OP_RINDEX,XTERM);
8287 UNIDOR(OP_READLINE);
8290 UNIDOR(OP_BACKTICK);
8299 LOP(OP_REVERSE,XTERM);
8302 UNIDOR(OP_READLINK);
8309 if (pl_yylval.opval)
8310 TERM(sublex_start());
8312 TOKEN(1); /* force error */
8315 checkcomma(s,PL_tokenbuf,"filehandle");
8325 LOP(OP_SELECT,XTERM);
8331 LOP(OP_SEMCTL,XTERM);
8334 LOP(OP_SEMGET,XTERM);
8337 LOP(OP_SEMOP,XTERM);
8343 LOP(OP_SETPGRP,XTERM);
8345 case KEY_setpriority:
8346 LOP(OP_SETPRIORITY,XTERM);
8348 case KEY_sethostent:
8354 case KEY_setservent:
8357 case KEY_setprotoent:
8367 LOP(OP_SEEKDIR,XTERM);
8369 case KEY_setsockopt:
8370 LOP(OP_SSOCKOPT,XTERM);
8376 LOP(OP_SHMCTL,XTERM);
8379 LOP(OP_SHMGET,XTERM);
8382 LOP(OP_SHMREAD,XTERM);
8385 LOP(OP_SHMWRITE,XTERM);
8388 LOP(OP_SHUTDOWN,XTERM);
8397 LOP(OP_SOCKET,XTERM);
8399 case KEY_socketpair:
8400 LOP(OP_SOCKPAIR,XTERM);
8403 checkcomma(s,PL_tokenbuf,"subroutine name");
8406 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8410 LOP(OP_SPLIT,XTERM);
8413 LOP(OP_SPRINTF,XTERM);
8416 LOP(OP_SPLICE,XTERM);
8431 LOP(OP_SUBSTR,XTERM);
8437 char * const tmpbuf = PL_tokenbuf + 1;
8438 SSize_t tboffset = 0;
8439 expectation attrful;
8440 bool have_name, have_proto;
8441 const int key = tmp;
8446 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8447 SV *subtoken = PL_madskills
8448 ? newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr))
8453 s = SKIPSPACE2(s,tmpwhite);
8459 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8460 (*s == ':' && s[1] == ':'))
8463 SV *nametoke = NULL;
8467 attrful = XATTRBLOCK;
8468 /* remember buffer pos'n for later force_word */
8469 tboffset = s - PL_oldbufptr;
8470 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8474 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8477 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8479 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
8481 sv_setpvn(PL_subname, tmpbuf, len);
8483 sv_setsv(PL_subname,PL_curstname);
8484 sv_catpvs(PL_subname,"::");
8485 sv_catpvn(PL_subname,tmpbuf,len);
8487 if (SvUTF8(PL_linestr))
8488 SvUTF8_on(PL_subname);
8494 CURMAD('X', nametoke);
8495 CURMAD('_', tmpwhite);
8496 force_ident_maybe_lex('&');
8498 s = SKIPSPACE2(d,tmpwhite);
8504 if (key == KEY_my || key == KEY_our || key==KEY_state)
8507 /* diag_listed_as: Missing name in "%s sub" */
8509 "Missing name in \"%s\"", PL_bufptr);
8511 PL_expect = XTERMBLOCK;
8512 attrful = XATTRTERM;
8513 sv_setpvs(PL_subname,"?");
8517 if (key == KEY_format) {
8519 PL_thistoken = subtoken;
8523 (void) force_word(PL_oldbufptr + tboffset, WORD,
8529 /* Look for a prototype */
8532 bool bad_proto = FALSE;
8533 bool in_brackets = FALSE;
8534 char greedy_proto = ' ';
8535 bool proto_after_greedy_proto = FALSE;
8536 bool must_be_last = FALSE;
8537 bool underscore = FALSE;
8538 bool seen_underscore = FALSE;
8539 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8542 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8544 Perl_croak(aTHX_ "Prototype not terminated");
8545 /* strip spaces and check for bad characters */
8546 d = SvPV(PL_lex_stuff, tmplen);
8548 for (p = d; tmplen; tmplen--, ++p) {
8552 if (warnillegalproto) {
8554 proto_after_greedy_proto = TRUE;
8555 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8560 if ( !strchr(";@%", *p) )
8567 else if ( *p == ']' ) {
8568 in_brackets = FALSE;
8570 else if ( (*p == '@' || *p == '%') &&
8571 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8573 must_be_last = TRUE;
8576 else if ( *p == '_' ) {
8577 underscore = seen_underscore = TRUE;
8584 if (proto_after_greedy_proto)
8585 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8586 "Prototype after '%c' for %"SVf" : %s",
8587 greedy_proto, SVfARG(PL_subname), d);
8589 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8590 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8591 "Illegal character %sin prototype for %"SVf" : %s",
8592 seen_underscore ? "after '_' " : "",
8594 SvUTF8(PL_lex_stuff)
8595 ? sv_uni_display(dsv,
8596 newSVpvn_flags(d, tmp, SVs_TEMP | SVf_UTF8),
8598 UNI_DISPLAY_ISPRINT)
8599 : pv_pretty(dsv, d, tmp, 60, NULL, NULL,
8600 PERL_PV_ESCAPE_NONASCII));
8602 SvCUR_set(PL_lex_stuff, tmp);
8607 CURMAD('q', PL_thisopen);
8608 CURMAD('_', tmpwhite);
8609 CURMAD('=', PL_thisstuff);
8610 CURMAD('Q', PL_thisclose);
8611 NEXTVAL_NEXTTOKE.opval =
8612 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8613 PL_lex_stuff = NULL;
8616 s = SKIPSPACE2(s,tmpwhite);
8624 if (*s == ':' && s[1] != ':')
8625 PL_expect = attrful;
8626 else if (*s != '{' && key == KEY_sub) {
8628 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8629 else if (*s != ';' && *s != '}')
8630 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8637 curmad('^', newSVpvs(""));
8638 CURMAD('_', tmpwhite);
8642 PL_thistoken = subtoken;
8645 NEXTVAL_NEXTTOKE.opval =
8646 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8647 PL_lex_stuff = NULL;
8653 sv_setpvs(PL_subname, "__ANON__");
8655 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8659 force_ident_maybe_lex('&');
8665 LOP(OP_SYSTEM,XREF);
8668 LOP(OP_SYMLINK,XTERM);
8671 LOP(OP_SYSCALL,XTERM);
8674 LOP(OP_SYSOPEN,XTERM);
8677 LOP(OP_SYSSEEK,XTERM);
8680 LOP(OP_SYSREAD,XTERM);
8683 LOP(OP_SYSWRITE,XTERM);
8688 TERM(sublex_start());
8709 LOP(OP_TRUNCATE,XTERM);
8721 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8723 pl_yylval.ival = CopLINE(PL_curcop);
8727 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8729 pl_yylval.ival = CopLINE(PL_curcop);
8733 LOP(OP_UNLINK,XTERM);
8739 LOP(OP_UNPACK,XTERM);
8742 LOP(OP_UTIME,XTERM);
8748 LOP(OP_UNSHIFT,XTERM);
8751 s = tokenize_use(1, s);
8761 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8763 pl_yylval.ival = CopLINE(PL_curcop);
8767 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8769 pl_yylval.ival = CopLINE(PL_curcop);
8773 PL_hints |= HINT_BLOCK_SCOPE;
8780 LOP(OP_WAITPID,XTERM);
8789 ctl_l[0] = toCTRL('L');
8791 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8794 /* Make sure $^L is defined */
8795 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8800 if (PL_expect == XOPERATOR) {
8801 if (*s == '=' && !PL_lex_allbrackets &&
8802 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8810 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8812 pl_yylval.ival = OP_XOR;
8818 #pragma segment Main
8824 Looks up an identifier in the pad or in a package
8827 PRIVATEREF if this is a lexical name.
8828 WORD if this belongs to a package.
8831 if we're in a my declaration
8832 croak if they tried to say my($foo::bar)
8833 build the ops for a my() declaration
8834 if it's an access to a my() variable
8835 build ops for access to a my() variable
8836 if in a dq string, and they've said @foo and we can't find @foo
8838 build ops for a bareword
8842 S_pending_ident(pTHX)
8846 const char pit = (char)pl_yylval.ival;
8847 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8848 /* All routes through this function want to know if there is a colon. */
8849 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8851 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8852 "### Pending identifier '%s'\n", PL_tokenbuf); });
8854 /* if we're in a my(), we can't allow dynamics here.
8855 $foo'bar has already been turned into $foo::bar, so
8856 just check for colons.
8858 if it's a legal name, the OP is a PADANY.
8861 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8863 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8864 "variable %s in \"our\"",
8865 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8866 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8870 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8871 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8872 UTF ? SVf_UTF8 : 0);
8874 pl_yylval.opval = newOP(OP_PADANY, 0);
8875 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8876 UTF ? SVf_UTF8 : 0);
8882 build the ops for accesses to a my() variable.
8887 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8888 UTF ? SVf_UTF8 : 0);
8889 if (tmp != NOT_IN_PAD) {
8890 /* might be an "our" variable" */
8891 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8892 /* build ops for a bareword */
8893 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8894 HEK * const stashname = HvNAME_HEK(stash);
8895 SV * const sym = newSVhek(stashname);
8896 sv_catpvs(sym, "::");
8897 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8898 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8899 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8903 ? (GV_ADDMULTI | GV_ADDINEVAL)
8906 ((PL_tokenbuf[0] == '$') ? SVt_PV
8907 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8912 pl_yylval.opval = newOP(OP_PADANY, 0);
8913 pl_yylval.opval->op_targ = tmp;
8919 Whine if they've said @foo in a doublequoted string,
8920 and @foo isn't a variable we can find in the symbol
8923 if (ckWARN(WARN_AMBIGUOUS) &&
8924 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8925 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8926 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8927 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8928 /* DO NOT warn for @- and @+ */
8929 && !( PL_tokenbuf[2] == '\0' &&
8930 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8933 /* Downgraded from fatal to warning 20000522 mjd */
8934 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8935 "Possible unintended interpolation of %"SVf" in string",
8936 SVfARG(newSVpvn_flags(PL_tokenbuf, tokenbuf_len,
8937 SVs_TEMP | ( UTF ? SVf_UTF8 : 0 ))));
8941 /* build ops for a bareword */
8942 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8943 newSVpvn_flags(PL_tokenbuf + 1,
8945 UTF ? SVf_UTF8 : 0 ));
8946 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8948 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8949 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8950 | ( UTF ? SVf_UTF8 : 0 ),
8951 ((PL_tokenbuf[0] == '$') ? SVt_PV
8952 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8958 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8962 PERL_ARGS_ASSERT_CHECKCOMMA;
8964 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8965 if (ckWARN(WARN_SYNTAX)) {
8968 for (w = s+2; *w && level; w++) {
8976 /* the list of chars below is for end of statements or
8977 * block / parens, boolean operators (&&, ||, //) and branch
8978 * constructs (or, and, if, until, unless, while, err, for).
8979 * Not a very solid hack... */
8980 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8981 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8982 "%s (...) interpreted as function",name);
8985 while (s < PL_bufend && isSPACE(*s))
8989 while (s < PL_bufend && isSPACE(*s))
8991 if (isIDFIRST_lazy_if(s,UTF)) {
8992 const char * const w = s;
8993 s += UTF ? UTF8SKIP(s) : 1;
8994 while (isALNUM_lazy_if(s,UTF))
8995 s += UTF ? UTF8SKIP(s) : 1;
8996 while (s < PL_bufend && isSPACE(*s))
9000 if (keyword(w, s - w, 0))
9003 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
9004 if (gv && GvCVu(gv))
9006 Perl_croak(aTHX_ "No comma allowed after %s", what);
9011 /* Either returns sv, or mortalizes sv and returns a new SV*.
9012 Best used as sv=new_constant(..., sv, ...).
9013 If s, pv are NULL, calls subroutine with one argument,
9014 and <type> is used with error messages only.
9015 <type> is assumed to be well formed UTF-8 */
9018 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
9019 SV *sv, SV *pv, const char *type, STRLEN typelen)
9022 HV * table = GvHV(PL_hintgv); /* ^H */
9027 const char *why1 = "", *why2 = "", *why3 = "";
9029 PERL_ARGS_ASSERT_NEW_CONSTANT;
9031 /* charnames doesn't work well if there have been errors found */
9032 if (PL_error_count > 0 && strEQ(key,"charnames"))
9033 return &PL_sv_undef;
9036 || ! (PL_hints & HINT_LOCALIZE_HH)
9037 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
9042 /* Here haven't found what we're looking for. If it is charnames,
9043 * perhaps it needs to be loaded. Try doing that before giving up */
9044 if (strEQ(key,"charnames")) {
9045 Perl_load_module(aTHX_
9047 newSVpvs("_charnames"),
9048 /* version parameter; no need to specify it, as if
9049 * we get too early a version, will fail anyway,
9050 * not being able to find '_charnames' */
9056 table = GvHV(PL_hintgv);
9058 && (PL_hints & HINT_LOCALIZE_HH)
9059 && (cvp = hv_fetch(table, key, keylen, FALSE))
9065 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
9066 msg = Perl_newSVpvf(aTHX_
9067 "Constant(%s) unknown", (type ? type: "undef"));
9072 why3 = "} is not defined";
9074 if (strEQ(key,"charnames")) {
9075 yyerror_pv(Perl_form(aTHX_
9076 /* The +3 is for '\N{'; -4 for that, plus '}' */
9077 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
9079 UTF ? SVf_UTF8 : 0);
9083 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
9084 (type ? type: "undef"), why1, why2, why3);
9087 yyerror(SvPVX_const(msg));
9092 sv_2mortal(sv); /* Parent created it permanently */
9095 pv = newSVpvn_flags(s, len, SVs_TEMP);
9097 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
9099 typesv = &PL_sv_undef;
9101 PUSHSTACKi(PERLSI_OVERLOAD);
9113 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
9117 /* Check the eval first */
9118 if (!PL_in_eval && ((errsv = ERRSV), SvTRUE_NN(errsv))) {
9120 const char * errstr;
9121 sv_catpvs(errsv, "Propagated");
9122 errstr = SvPV_const(errsv, errlen);
9123 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
9125 res = SvREFCNT_inc_simple(sv);
9129 SvREFCNT_inc_simple_void(res);
9138 why1 = "Call to &{$^H{";
9140 why3 = "}} did not return a defined value";
9148 /* Returns a NUL terminated string, with the length of the string written to
9152 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
9156 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9158 PERL_ARGS_ASSERT_SCAN_WORD;
9162 Perl_croak(aTHX_ ident_too_long);
9163 if (isALNUM(*s) || (!UTF && isALNUMC_L1(*s))) /* UTF handled below */
9165 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
9170 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
9174 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9175 char *t = s + UTF8SKIP(s);
9177 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9181 Perl_croak(aTHX_ ident_too_long);
9182 Copy(s, d, len, char);
9195 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
9198 char *bracket = NULL;
9201 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9203 PERL_ARGS_ASSERT_SCAN_IDENT;
9208 while (isDIGIT(*s)) {
9210 Perl_croak(aTHX_ ident_too_long);
9217 Perl_croak(aTHX_ ident_too_long);
9218 if (isALNUM(*s)) /* UTF handled below */
9220 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
9225 else if (*s == ':' && s[1] == ':') {
9229 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9230 char *t = s + UTF8SKIP(s);
9231 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9233 if (d + (t - s) > e)
9234 Perl_croak(aTHX_ ident_too_long);
9235 Copy(s, d, t - s, char);
9246 if (PL_lex_state != LEX_NORMAL)
9247 PL_lex_state = LEX_INTERPENDMAYBE;
9250 if (*s == '$' && s[1] &&
9251 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
9261 const STRLEN skip = UTF8SKIP(s);
9264 for ( i = 0; i < skip; i++ )
9272 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9276 else if (ck_uni && !bracket)
9279 if (isSPACE(s[-1])) {
9281 const char ch = *s++;
9282 if (!SPACE_OR_TAB(ch)) {
9288 if (isIDFIRST_lazy_if(d,UTF)) {
9292 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
9293 end += UTF8SKIP(end);
9294 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
9295 end += UTF8SKIP(end);
9297 Copy(s, d, end - s, char);
9302 while ((isALNUM(*s) || *s == ':') && d < e)
9305 Perl_croak(aTHX_ ident_too_long);
9308 while (s < send && SPACE_OR_TAB(*s))
9310 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9311 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9312 const char * const brack =
9314 ((*s == '[') ? "[...]" : "{...}");
9315 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9316 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9317 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9318 funny, dest, brack, funny, dest, brack);
9321 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9322 PL_lex_allbrackets++;
9326 /* Handle extended ${^Foo} variables
9327 * 1999-02-27 mjd-perl-patch@plover.com */
9328 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
9332 while (isALNUM(*s) && d < e) {
9336 Perl_croak(aTHX_ ident_too_long);
9341 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9342 PL_lex_state = LEX_INTERPEND;
9345 if (PL_lex_state == LEX_NORMAL) {
9346 if (ckWARN(WARN_AMBIGUOUS) &&
9347 (keyword(dest, d - dest, 0)
9348 || get_cvn_flags(dest, d - dest, UTF ? SVf_UTF8 : 0)))
9350 SV *tmp = newSVpvn_flags( dest, d - dest,
9351 SVs_TEMP | (UTF ? SVf_UTF8 : 0) );
9354 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9355 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9356 funny, tmp, funny, tmp);
9361 s = bracket; /* let the parser handle it */
9365 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9366 PL_lex_state = LEX_INTERPEND;
9371 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
9373 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
9374 * the parse starting at 's', based on the subset that are valid in this
9375 * context input to this routine in 'valid_flags'. Advances s. Returns
9376 * TRUE if the input should be treated as a valid flag, so the next char
9377 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
9378 * first call on the current regex. This routine will set it to any
9379 * charset modifier found. The caller shouldn't change it. This way,
9380 * another charset modifier encountered in the parse can be detected as an
9381 * error, as we have decided to allow only one */
9384 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9386 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9387 if (isALNUM_lazy_if(*s, UTF)) {
9388 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9389 UTF ? SVf_UTF8 : 0);
9391 /* Pretend that it worked, so will continue processing before
9400 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
9401 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9402 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9403 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9404 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9405 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9406 case LOCALE_PAT_MOD:
9408 goto multiple_charsets;
9410 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9413 case UNICODE_PAT_MOD:
9415 goto multiple_charsets;
9417 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9420 case ASCII_RESTRICT_PAT_MOD:
9422 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9426 /* Error if previous modifier wasn't an 'a', but if it was, see
9427 * if, and accept, a second occurrence (only) */
9429 || get_regex_charset(*pmfl)
9430 != REGEX_ASCII_RESTRICTED_CHARSET)
9432 goto multiple_charsets;
9434 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9438 case DEPENDS_PAT_MOD:
9440 goto multiple_charsets;
9442 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9451 if (*charset != c) {
9452 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9454 else if (c == 'a') {
9455 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9458 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9461 /* Pretend that it worked, so will continue processing before dieing */
9467 S_scan_pat(pTHX_ char *start, I32 type)
9471 char *s = scan_str(start,!!PL_madskills,FALSE, PL_reg_state.re_reparsing);
9472 const char * const valid_flags =
9473 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9474 char charset = '\0'; /* character set modifier */
9479 PERL_ARGS_ASSERT_SCAN_PAT;
9481 /* this was only needed for the initial scan_str; set it to false
9482 * so that any (?{}) code blocks etc are parsed normally */
9483 PL_reg_state.re_reparsing = FALSE;
9485 const char * const delimiter = skipspace(start);
9489 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9490 : "Search pattern not terminated" ));
9493 pm = (PMOP*)newPMOP(type, 0);
9494 if (PL_multi_open == '?') {
9495 /* This is the only point in the code that sets PMf_ONCE: */
9496 pm->op_pmflags |= PMf_ONCE;
9498 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9499 allows us to restrict the list needed by reset to just the ??
9501 assert(type != OP_TRANS);
9503 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9506 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9509 elements = mg->mg_len / sizeof(PMOP**);
9510 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9511 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9512 mg->mg_len = elements * sizeof(PMOP**);
9513 PmopSTASH_set(pm,PL_curstash);
9520 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9521 * anon CV. False positives like qr/[(?{]/ are harmless */
9523 if (type == OP_QR) {
9525 char *e, *p = SvPV(PL_lex_stuff, len);
9527 for (; p < e; p++) {
9528 if (p[0] == '(' && p[1] == '?'
9529 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9531 pm->op_pmflags |= PMf_HAS_CV;
9535 pm->op_pmflags |= PMf_IS_QR;
9538 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9540 if (PL_madskills && modstart != s) {
9541 SV* tmptoken = newSVpvn(modstart, s - modstart);
9542 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9545 /* issue a warning if /c is specified,but /g is not */
9546 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9548 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9549 "Use of /c modifier is meaningless without /g" );
9552 PL_lex_op = (OP*)pm;
9553 pl_yylval.ival = OP_MATCH;
9558 S_scan_subst(pTHX_ char *start)
9565 char charset = '\0'; /* character set modifier */
9570 PERL_ARGS_ASSERT_SCAN_SUBST;
9572 pl_yylval.ival = OP_NULL;
9574 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9577 Perl_croak(aTHX_ "Substitution pattern not terminated");
9579 if (s[-1] == PL_multi_open)
9583 CURMAD('q', PL_thisopen);
9584 CURMAD('_', PL_thiswhite);
9585 CURMAD('E', PL_thisstuff);
9586 CURMAD('Q', PL_thisclose);
9587 PL_realtokenstart = s - SvPVX(PL_linestr);
9591 first_start = PL_multi_start;
9592 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9595 SvREFCNT_dec(PL_lex_stuff);
9596 PL_lex_stuff = NULL;
9598 Perl_croak(aTHX_ "Substitution replacement not terminated");
9600 PL_multi_start = first_start; /* so whole substitution is taken together */
9602 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9606 CURMAD('z', PL_thisopen);
9607 CURMAD('R', PL_thisstuff);
9608 CURMAD('Z', PL_thisclose);
9614 if (*s == EXEC_PAT_MOD) {
9618 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9627 curmad('m', newSVpvn(modstart, s - modstart));
9628 append_madprops(PL_thismad, (OP*)pm, 0);
9632 if ((pm->op_pmflags & PMf_CONTINUE)) {
9633 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9637 SV * const repl = newSVpvs("");
9640 pm->op_pmflags |= PMf_EVAL;
9643 sv_catpvs(repl, "eval ");
9645 sv_catpvs(repl, "do ");
9647 sv_catpvs(repl, "{");
9648 sv_catsv(repl, PL_sublex_info.repl);
9649 sv_catpvs(repl, "}");
9651 SvREFCNT_dec(PL_sublex_info.repl);
9652 PL_sublex_info.repl = repl;
9655 PL_lex_op = (OP*)pm;
9656 pl_yylval.ival = OP_SUBST;
9661 S_scan_trans(pTHX_ char *start)
9669 bool nondestruct = 0;
9674 PERL_ARGS_ASSERT_SCAN_TRANS;
9676 pl_yylval.ival = OP_NULL;
9678 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9680 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9682 if (s[-1] == PL_multi_open)
9686 CURMAD('q', PL_thisopen);
9687 CURMAD('_', PL_thiswhite);
9688 CURMAD('E', PL_thisstuff);
9689 CURMAD('Q', PL_thisclose);
9690 PL_realtokenstart = s - SvPVX(PL_linestr);
9694 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9697 SvREFCNT_dec(PL_lex_stuff);
9698 PL_lex_stuff = NULL;
9700 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9703 CURMAD('z', PL_thisopen);
9704 CURMAD('R', PL_thisstuff);
9705 CURMAD('Z', PL_thisclose);
9708 complement = del = squash = 0;
9715 complement = OPpTRANS_COMPLEMENT;
9718 del = OPpTRANS_DELETE;
9721 squash = OPpTRANS_SQUASH;
9733 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9734 o->op_private &= ~OPpTRANS_ALL;
9735 o->op_private |= del|squash|complement|
9736 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9737 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9740 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9745 curmad('m', newSVpvn(modstart, s - modstart));
9746 append_madprops(PL_thismad, o, 0);
9755 Takes a pointer to the first < in <<FOO.
9756 Returns a pointer to the byte following <<FOO.
9758 This function scans a heredoc, which involves different methods
9759 depending on whether we are in a string eval, quoted construct, etc.
9760 This is because PL_linestr could containing a single line of input, or
9761 a whole string being evalled, or the contents of the current quote-
9764 The two basic methods are:
9765 - Steal lines from the input stream
9766 - Scan the heredoc in PL_linestr and remove it therefrom
9768 In a file scope or filtered eval, the first method is used; in a
9769 string eval, the second.
9771 In a quote-like operator, we have to choose between the two,
9772 depending on where we can find a newline. We peek into outer lex-
9773 ing scopes until we find one with a newline in it. If we reach the
9774 outermost lexing scope and it is a file, we use the stream method.
9775 Otherwise it is treated as an eval.
9779 S_scan_heredoc(pTHX_ register char *s)
9782 I32 op_type = OP_SCALAR;
9789 const bool infile = PL_rsfp || PL_parser->filtered;
9790 LEXSHARED *shared = PL_parser->lex_shared;
9792 I32 stuffstart = s - SvPVX(PL_linestr);
9795 PL_realtokenstart = -1;
9798 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9801 d = PL_tokenbuf + 1;
9802 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9803 *PL_tokenbuf = '\n';
9805 while (SPACE_OR_TAB(*peek))
9807 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9810 s = delimcpy(d, e, s, PL_bufend, term, &len);
9812 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9818 /* <<\FOO is equivalent to <<'FOO' */
9822 if (!isALNUM_lazy_if(s,UTF))
9823 deprecate("bare << to mean <<\"\"");
9824 for (; isALNUM_lazy_if(s,UTF); s++) {
9829 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9830 Perl_croak(aTHX_ "Delimiter for here document is too long");
9833 len = d - PL_tokenbuf;
9837 tstart = PL_tokenbuf + 1;
9838 PL_thisclose = newSVpvn(tstart, len - 1);
9839 tstart = SvPVX(PL_linestr) + stuffstart;
9840 PL_thisopen = newSVpvn(tstart, s - tstart);
9841 stuffstart = s - SvPVX(PL_linestr);
9844 #ifndef PERL_STRICT_CR
9845 d = strchr(s, '\r');
9847 char * const olds = s;
9849 while (s < PL_bufend) {
9855 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9864 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9870 tstart = SvPVX(PL_linestr) + stuffstart;
9872 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9874 PL_thisstuff = newSVpvn(tstart, s - tstart);
9877 stuffstart = s - SvPVX(PL_linestr);
9880 tmpstr = newSV_type(SVt_PVIV);
9884 SvIV_set(tmpstr, -1);
9886 else if (term == '`') {
9887 op_type = OP_BACKTICK;
9888 SvIV_set(tmpstr, '\\');
9891 PL_multi_start = CopLINE(PL_curcop) + 1;
9892 PL_multi_open = PL_multi_close = '<';
9893 /* inside a string eval or quote-like operator */
9894 if (!infile || PL_lex_inwhat) {
9897 char * const olds = s;
9898 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9899 /* These two fields are not set until an inner lexing scope is
9900 entered. But we need them set here. */
9901 shared->ls_bufptr = s;
9902 shared->ls_linestr = PL_linestr;
9904 /* Look for a newline. If the current buffer does not have one,
9905 peek into the line buffer of the parent lexing scope, going
9906 up as many levels as necessary to find one with a newline
9909 while (!(s = (char *)memchr(
9910 (void *)shared->ls_bufptr, '\n',
9911 SvEND(shared->ls_linestr)-shared->ls_bufptr
9913 shared = shared->ls_prev;
9914 /* shared is only null if we have gone beyond the outermost
9915 lexing scope. In a file, we will have broken out of the
9916 loop in the previous iteration. In an eval, the string buf-
9917 fer ends with "\n;", so the while condition below will have
9918 evaluated to false. So shared can never be null. */
9920 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9921 most lexing scope. In a file, shared->ls_linestr at that
9922 level is just one line, so there is no body to steal. */
9923 if (infile && !shared->ls_prev) {
9929 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9932 linestr = shared->ls_linestr;
9933 bufend = SvEND(linestr);
9935 while (s < bufend &&
9936 (*s != '\n' || memNE(s,PL_tokenbuf,len)) ) {
9938 ++shared->herelines;
9943 sv_setpvn(tmpstr,d+1,s-d);
9947 sv_catpvn(PL_thisstuff, d + 1, s - d);
9949 PL_thisstuff = newSVpvn(d + 1, s - d);
9950 stuffstart = s - SvPVX(PL_linestr);
9954 /* the preceding stmt passes a newline */
9955 shared->herelines++;
9957 /* s now points to the newline after the heredoc terminator.
9958 d points to the newline before the body of the heredoc.
9961 /* We are going to modify linestr in place here, so set
9962 aside copies of the string if necessary for re-evals or
9964 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9965 check shared->re_eval_str. */
9966 if (shared->re_eval_start || shared->re_eval_str) {
9967 /* Set aside the rest of the regexp */
9968 if (!shared->re_eval_str)
9969 shared->re_eval_str =
9970 newSVpvn(shared->re_eval_start,
9971 bufend - shared->re_eval_start);
9972 shared->re_eval_start -= s-d;
9974 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9975 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9976 cx->blk_eval.cur_text == linestr)
9978 cx->blk_eval.cur_text = newSVsv(linestr);
9979 SvSCREAM_on(cx->blk_eval.cur_text);
9981 /* Copy everything from s onwards back to d. */
9982 Move(s,d,bufend-s + 1,char);
9983 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9984 /* Setting PL_bufend only applies when we have not dug deeper
9985 into other scopes, because sublex_done sets PL_bufend to
9986 SvEND(PL_linestr). */
9987 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9994 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9995 term = PL_tokenbuf[1];
9997 linestr_save = PL_linestr; /* must restore this afterwards */
9998 d = s; /* and this */
9999 PL_linestr = newSVpvs("");
10000 PL_bufend = SvPVX(PL_linestr);
10003 if (PL_madskills) {
10004 tstart = SvPVX(PL_linestr) + stuffstart;
10006 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10008 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10011 PL_bufptr = PL_bufend;
10012 CopLINE_set(PL_curcop,
10013 PL_multi_start + shared->herelines);
10014 if (!lex_next_chunk(LEX_NO_TERM)
10015 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
10016 SvREFCNT_dec(linestr_save);
10019 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10020 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
10021 lex_grow_linestr(SvCUR(PL_linestr) + 2);
10022 sv_catpvs(PL_linestr, "\n\0");
10026 stuffstart = s - SvPVX(PL_linestr);
10028 shared->herelines++;
10029 PL_last_lop = PL_last_uni = NULL;
10030 #ifndef PERL_STRICT_CR
10031 if (PL_bufend - PL_linestart >= 2) {
10032 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
10033 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
10035 PL_bufend[-2] = '\n';
10037 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10039 else if (PL_bufend[-1] == '\r')
10040 PL_bufend[-1] = '\n';
10042 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
10043 PL_bufend[-1] = '\n';
10045 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
10046 SvREFCNT_dec(PL_linestr);
10047 PL_linestr = linestr_save;
10048 PL_linestart = SvPVX(linestr_save);
10049 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10054 sv_catsv(tmpstr,PL_linestr);
10058 PL_multi_end = CopLINE(PL_curcop);
10059 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
10060 SvPV_shrink_to_cur(tmpstr);
10063 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
10065 else if (PL_encoding)
10066 sv_recode_to_utf8(tmpstr, PL_encoding);
10068 PL_lex_stuff = tmpstr;
10069 pl_yylval.ival = op_type;
10073 SvREFCNT_dec(tmpstr);
10074 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10075 missingterm(PL_tokenbuf + 1);
10078 /* scan_inputsymbol
10079 takes: current position in input buffer
10080 returns: new position in input buffer
10081 side-effects: pl_yylval and lex_op are set.
10086 <FH> read from filehandle
10087 <pkg::FH> read from package qualified filehandle
10088 <pkg'FH> read from package qualified filehandle
10089 <$fh> read from filehandle in $fh
10090 <*.h> filename glob
10095 S_scan_inputsymbol(pTHX_ char *start)
10098 char *s = start; /* current position in buffer */
10101 char *d = PL_tokenbuf; /* start of temp holding space */
10102 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
10104 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
10106 end = strchr(s, '\n');
10109 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
10111 /* die if we didn't have space for the contents of the <>,
10112 or if it didn't end, or if we see a newline
10115 if (len >= (I32)sizeof PL_tokenbuf)
10116 Perl_croak(aTHX_ "Excessively long <> operator");
10118 Perl_croak(aTHX_ "Unterminated <> operator");
10123 Remember, only scalar variables are interpreted as filehandles by
10124 this code. Anything more complex (e.g., <$fh{$num}>) will be
10125 treated as a glob() call.
10126 This code makes use of the fact that except for the $ at the front,
10127 a scalar variable and a filehandle look the same.
10129 if (*d == '$' && d[1]) d++;
10131 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
10132 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
10133 d += UTF ? UTF8SKIP(d) : 1;
10135 /* If we've tried to read what we allow filehandles to look like, and
10136 there's still text left, then it must be a glob() and not a getline.
10137 Use scan_str to pull out the stuff between the <> and treat it
10138 as nothing more than a string.
10141 if (d - PL_tokenbuf != len) {
10142 pl_yylval.ival = OP_GLOB;
10143 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
10145 Perl_croak(aTHX_ "Glob not terminated");
10149 bool readline_overriden = FALSE;
10152 /* we're in a filehandle read situation */
10155 /* turn <> into <ARGV> */
10157 Copy("ARGV",d,5,char);
10159 /* Check whether readline() is overriden */
10160 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
10162 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
10164 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
10165 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
10166 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
10167 readline_overriden = TRUE;
10169 /* if <$fh>, create the ops to turn the variable into a
10173 /* try to find it in the pad for this block, otherwise find
10174 add symbol table ops
10176 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
10177 if (tmp != NOT_IN_PAD) {
10178 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
10179 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
10180 HEK * const stashname = HvNAME_HEK(stash);
10181 SV * const sym = sv_2mortal(newSVhek(stashname));
10182 sv_catpvs(sym, "::");
10183 sv_catpv(sym, d+1);
10188 OP * const o = newOP(OP_PADSV, 0);
10190 PL_lex_op = readline_overriden
10191 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10192 op_append_elem(OP_LIST, o,
10193 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
10194 : (OP*)newUNOP(OP_READLINE, 0, o);
10203 ? (GV_ADDMULTI | GV_ADDINEVAL)
10204 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
10206 PL_lex_op = readline_overriden
10207 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10208 op_append_elem(OP_LIST,
10209 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
10210 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10211 : (OP*)newUNOP(OP_READLINE, 0,
10212 newUNOP(OP_RV2SV, 0,
10213 newGVOP(OP_GV, 0, gv)));
10215 if (!readline_overriden)
10216 PL_lex_op->op_flags |= OPf_SPECIAL;
10217 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
10218 pl_yylval.ival = OP_NULL;
10221 /* If it's none of the above, it must be a literal filehandle
10222 (<Foo::BAR> or <FOO>) so build a simple readline OP */
10224 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
10225 PL_lex_op = readline_overriden
10226 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10227 op_append_elem(OP_LIST,
10228 newGVOP(OP_GV, 0, gv),
10229 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10230 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
10231 pl_yylval.ival = OP_NULL;
10240 takes: start position in buffer
10241 keep_quoted preserve \ on the embedded delimiter(s)
10242 keep_delims preserve the delimiters around the string
10243 re_reparse compiling a run-time /(?{})/:
10244 collapse // to /, and skip encoding src
10245 returns: position to continue reading from buffer
10246 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
10247 updates the read buffer.
10249 This subroutine pulls a string out of the input. It is called for:
10250 q single quotes q(literal text)
10251 ' single quotes 'literal text'
10252 qq double quotes qq(interpolate $here please)
10253 " double quotes "interpolate $here please"
10254 qx backticks qx(/bin/ls -l)
10255 ` backticks `/bin/ls -l`
10256 qw quote words @EXPORT_OK = qw( func() $spam )
10257 m// regexp match m/this/
10258 s/// regexp substitute s/this/that/
10259 tr/// string transliterate tr/this/that/
10260 y/// string transliterate y/this/that/
10261 ($*@) sub prototypes sub foo ($)
10262 (stuff) sub attr parameters sub foo : attr(stuff)
10263 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
10265 In most of these cases (all but <>, patterns and transliterate)
10266 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
10267 calls scan_str(). s/// makes yylex() call scan_subst() which calls
10268 scan_str(). tr/// and y/// make yylex() call scan_trans() which
10271 It skips whitespace before the string starts, and treats the first
10272 character as the delimiter. If the delimiter is one of ([{< then
10273 the corresponding "close" character )]}> is used as the closing
10274 delimiter. It allows quoting of delimiters, and if the string has
10275 balanced delimiters ([{<>}]) it allows nesting.
10277 On success, the SV with the resulting string is put into lex_stuff or,
10278 if that is already non-NULL, into lex_repl. The second case occurs only
10279 when parsing the RHS of the special constructs s/// and tr/// (y///).
10280 For convenience, the terminating delimiter character is stuffed into
10285 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims, int re_reparse)
10288 SV *sv; /* scalar value: string */
10289 const char *tmps; /* temp string, used for delimiter matching */
10290 char *s = start; /* current position in the buffer */
10291 char term; /* terminating character */
10292 char *to; /* current position in the sv's data */
10293 I32 brackets = 1; /* bracket nesting level */
10294 bool has_utf8 = FALSE; /* is there any utf8 content? */
10295 I32 termcode; /* terminating char. code */
10296 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
10297 STRLEN termlen; /* length of terminating string */
10298 int last_off = 0; /* last position for nesting bracket */
10304 PERL_ARGS_ASSERT_SCAN_STR;
10306 /* skip space before the delimiter */
10312 if (PL_realtokenstart >= 0) {
10313 stuffstart = PL_realtokenstart;
10314 PL_realtokenstart = -1;
10317 stuffstart = start - SvPVX(PL_linestr);
10319 /* mark where we are, in case we need to report errors */
10322 /* after skipping whitespace, the next character is the terminator */
10325 termcode = termstr[0] = term;
10329 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
10330 Copy(s, termstr, termlen, U8);
10331 if (!UTF8_IS_INVARIANT(term))
10335 /* mark where we are */
10336 PL_multi_start = CopLINE(PL_curcop);
10337 PL_multi_open = term;
10339 /* find corresponding closing delimiter */
10340 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
10341 termcode = termstr[0] = term = tmps[5];
10343 PL_multi_close = term;
10345 /* create a new SV to hold the contents. 79 is the SV's initial length.
10346 What a random number. */
10347 sv = newSV_type(SVt_PVIV);
10349 SvIV_set(sv, termcode);
10350 (void)SvPOK_only(sv); /* validate pointer */
10352 /* move past delimiter and try to read a complete string */
10354 sv_catpvn(sv, s, termlen);
10357 tstart = SvPVX(PL_linestr) + stuffstart;
10358 if (PL_madskills && !PL_thisopen && !keep_delims) {
10359 PL_thisopen = newSVpvn(tstart, s - tstart);
10360 stuffstart = s - SvPVX(PL_linestr);
10364 if (PL_encoding && !UTF && !re_reparse) {
10368 int offset = s - SvPVX_const(PL_linestr);
10369 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
10370 &offset, (char*)termstr, termlen);
10371 const char * const ns = SvPVX_const(PL_linestr) + offset;
10372 char * const svlast = SvEND(sv) - 1;
10374 for (; s < ns; s++) {
10375 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10376 COPLINE_INC_WITH_HERELINES;
10379 goto read_more_line;
10381 /* handle quoted delimiters */
10382 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10384 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10386 if ((svlast-1 - t) % 2) {
10387 if (!keep_quoted) {
10388 *(svlast-1) = term;
10390 SvCUR_set(sv, SvCUR(sv) - 1);
10395 if (PL_multi_open == PL_multi_close) {
10401 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10402 /* At here, all closes are "was quoted" one,
10403 so we don't check PL_multi_close. */
10405 if (!keep_quoted && *(t+1) == PL_multi_open)
10410 else if (*t == PL_multi_open)
10418 SvCUR_set(sv, w - SvPVX_const(sv));
10420 last_off = w - SvPVX(sv);
10421 if (--brackets <= 0)
10426 if (!keep_delims) {
10427 SvCUR_set(sv, SvCUR(sv) - 1);
10433 /* extend sv if need be */
10434 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10435 /* set 'to' to the next character in the sv's string */
10436 to = SvPVX(sv)+SvCUR(sv);
10438 /* if open delimiter is the close delimiter read unbridle */
10439 if (PL_multi_open == PL_multi_close) {
10440 for (; s < PL_bufend; s++,to++) {
10441 /* embedded newlines increment the current line number */
10442 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10443 COPLINE_INC_WITH_HERELINES;
10444 /* handle quoted delimiters */
10445 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10448 || (re_reparse && s[1] == '\\'))
10451 /* any other quotes are simply copied straight through */
10455 /* terminate when run out of buffer (the for() condition), or
10456 have found the terminator */
10457 else if (*s == term) {
10460 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10463 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10469 /* if the terminator isn't the same as the start character (e.g.,
10470 matched brackets), we have to allow more in the quoting, and
10471 be prepared for nested brackets.
10474 /* read until we run out of string, or we find the terminator */
10475 for (; s < PL_bufend; s++,to++) {
10476 /* embedded newlines increment the line count */
10477 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10478 COPLINE_INC_WITH_HERELINES;
10479 /* backslashes can escape the open or closing characters */
10480 if (*s == '\\' && s+1 < PL_bufend) {
10481 if (!keep_quoted &&
10482 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10487 /* allow nested opens and closes */
10488 else if (*s == PL_multi_close && --brackets <= 0)
10490 else if (*s == PL_multi_open)
10492 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10497 /* terminate the copied string and update the sv's end-of-string */
10499 SvCUR_set(sv, to - SvPVX_const(sv));
10502 * this next chunk reads more into the buffer if we're not done yet
10506 break; /* handle case where we are done yet :-) */
10508 #ifndef PERL_STRICT_CR
10509 if (to - SvPVX_const(sv) >= 2) {
10510 if ((to[-2] == '\r' && to[-1] == '\n') ||
10511 (to[-2] == '\n' && to[-1] == '\r'))
10515 SvCUR_set(sv, to - SvPVX_const(sv));
10517 else if (to[-1] == '\r')
10520 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10525 /* if we're out of file, or a read fails, bail and reset the current
10526 line marker so we can report where the unterminated string began
10529 if (PL_madskills) {
10530 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10532 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10534 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10537 COPLINE_INC_WITH_HERELINES;
10538 PL_bufptr = PL_bufend;
10539 if (!lex_next_chunk(0)) {
10541 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10550 /* at this point, we have successfully read the delimited string */
10552 if (!PL_encoding || UTF || re_reparse) {
10554 if (PL_madskills) {
10555 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10556 const int len = s - tstart;
10558 sv_catpvn(PL_thisstuff, tstart, len);
10560 PL_thisstuff = newSVpvn(tstart, len);
10561 if (!PL_thisclose && !keep_delims)
10562 PL_thisclose = newSVpvn(s,termlen);
10567 sv_catpvn(sv, s, termlen);
10572 if (PL_madskills) {
10573 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10574 const int len = s - tstart - termlen;
10576 sv_catpvn(PL_thisstuff, tstart, len);
10578 PL_thisstuff = newSVpvn(tstart, len);
10579 if (!PL_thisclose && !keep_delims)
10580 PL_thisclose = newSVpvn(s - termlen,termlen);
10584 if (has_utf8 || (PL_encoding && !re_reparse))
10587 PL_multi_end = CopLINE(PL_curcop);
10589 /* if we allocated too much space, give some back */
10590 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10591 SvLEN_set(sv, SvCUR(sv) + 1);
10592 SvPV_renew(sv, SvLEN(sv));
10595 /* decide whether this is the first or second quoted string we've read
10600 PL_sublex_info.repl = sv;
10608 takes: pointer to position in buffer
10609 returns: pointer to new position in buffer
10610 side-effects: builds ops for the constant in pl_yylval.op
10612 Read a number in any of the formats that Perl accepts:
10614 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10615 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10618 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10620 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10623 If it reads a number without a decimal point or an exponent, it will
10624 try converting the number to an integer and see if it can do so
10625 without loss of precision.
10629 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10632 const char *s = start; /* current position in buffer */
10633 char *d; /* destination in temp buffer */
10634 char *e; /* end of temp buffer */
10635 NV nv; /* number read, as a double */
10636 SV *sv = NULL; /* place to put the converted number */
10637 bool floatit; /* boolean: int or float? */
10638 const char *lastub = NULL; /* position of last underbar */
10639 static char const number_too_long[] = "Number too long";
10641 PERL_ARGS_ASSERT_SCAN_NUM;
10643 /* We use the first character to decide what type of number this is */
10647 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10649 /* if it starts with a 0, it could be an octal number, a decimal in
10650 0.13 disguise, or a hexadecimal number, or a binary number. */
10654 u holds the "number so far"
10655 shift the power of 2 of the base
10656 (hex == 4, octal == 3, binary == 1)
10657 overflowed was the number more than we can hold?
10659 Shift is used when we add a digit. It also serves as an "are
10660 we in octal/hex/binary?" indicator to disallow hex characters
10661 when in octal mode.
10666 bool overflowed = FALSE;
10667 bool just_zero = TRUE; /* just plain 0 or binary number? */
10668 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10669 static const char* const bases[5] =
10670 { "", "binary", "", "octal", "hexadecimal" };
10671 static const char* const Bases[5] =
10672 { "", "Binary", "", "Octal", "Hexadecimal" };
10673 static const char* const maxima[5] =
10675 "0b11111111111111111111111111111111",
10679 const char *base, *Base, *max;
10681 /* check for hex */
10682 if (s[1] == 'x' || s[1] == 'X') {
10686 } else if (s[1] == 'b' || s[1] == 'B') {
10691 /* check for a decimal in disguise */
10692 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10694 /* so it must be octal */
10701 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10702 "Misplaced _ in number");
10706 base = bases[shift];
10707 Base = Bases[shift];
10708 max = maxima[shift];
10710 /* read the rest of the number */
10712 /* x is used in the overflow test,
10713 b is the digit we're adding on. */
10718 /* if we don't mention it, we're done */
10722 /* _ are ignored -- but warned about if consecutive */
10724 if (lastub && s == lastub + 1)
10725 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10726 "Misplaced _ in number");
10730 /* 8 and 9 are not octal */
10731 case '8': case '9':
10733 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10737 case '2': case '3': case '4':
10738 case '5': case '6': case '7':
10740 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10743 case '0': case '1':
10744 b = *s++ & 15; /* ASCII digit -> value of digit */
10748 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10749 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10750 /* make sure they said 0x */
10753 b = (*s++ & 7) + 9;
10755 /* Prepare to put the digit we have onto the end
10756 of the number so far. We check for overflows.
10762 x = u << shift; /* make room for the digit */
10764 if ((x >> shift) != u
10765 && !(PL_hints & HINT_NEW_BINARY)) {
10768 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10769 "Integer overflow in %s number",
10772 u = x | b; /* add the digit to the end */
10775 n *= nvshift[shift];
10776 /* If an NV has not enough bits in its
10777 * mantissa to represent an UV this summing of
10778 * small low-order numbers is a waste of time
10779 * (because the NV cannot preserve the
10780 * low-order bits anyway): we could just
10781 * remember when did we overflow and in the
10782 * end just multiply n by the right
10790 /* if we get here, we had success: make a scalar value from
10795 /* final misplaced underbar check */
10796 if (s[-1] == '_') {
10797 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10801 if (n > 4294967295.0)
10802 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10803 "%s number > %s non-portable",
10809 if (u > 0xffffffff)
10810 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10811 "%s number > %s non-portable",
10816 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10817 sv = new_constant(start, s - start, "integer",
10818 sv, NULL, NULL, 0);
10819 else if (PL_hints & HINT_NEW_BINARY)
10820 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10825 handle decimal numbers.
10826 we're also sent here when we read a 0 as the first digit
10828 case '1': case '2': case '3': case '4': case '5':
10829 case '6': case '7': case '8': case '9': case '.':
10832 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10835 /* read next group of digits and _ and copy into d */
10836 while (isDIGIT(*s) || *s == '_') {
10837 /* skip underscores, checking for misplaced ones
10841 if (lastub && s == lastub + 1)
10842 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10843 "Misplaced _ in number");
10847 /* check for end of fixed-length buffer */
10849 Perl_croak(aTHX_ number_too_long);
10850 /* if we're ok, copy the character */
10855 /* final misplaced underbar check */
10856 if (lastub && s == lastub + 1) {
10857 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10860 /* read a decimal portion if there is one. avoid
10861 3..5 being interpreted as the number 3. followed
10864 if (*s == '.' && s[1] != '.') {
10869 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10870 "Misplaced _ in number");
10874 /* copy, ignoring underbars, until we run out of digits.
10876 for (; isDIGIT(*s) || *s == '_'; s++) {
10877 /* fixed length buffer check */
10879 Perl_croak(aTHX_ number_too_long);
10881 if (lastub && s == lastub + 1)
10882 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10883 "Misplaced _ in number");
10889 /* fractional part ending in underbar? */
10890 if (s[-1] == '_') {
10891 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10892 "Misplaced _ in number");
10894 if (*s == '.' && isDIGIT(s[1])) {
10895 /* oops, it's really a v-string, but without the "v" */
10901 /* read exponent part, if present */
10902 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10906 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10907 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10909 /* stray preinitial _ */
10911 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10912 "Misplaced _ in number");
10916 /* allow positive or negative exponent */
10917 if (*s == '+' || *s == '-')
10920 /* stray initial _ */
10922 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10923 "Misplaced _ in number");
10927 /* read digits of exponent */
10928 while (isDIGIT(*s) || *s == '_') {
10931 Perl_croak(aTHX_ number_too_long);
10935 if (((lastub && s == lastub + 1) ||
10936 (!isDIGIT(s[1]) && s[1] != '_')))
10937 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10938 "Misplaced _ in number");
10946 We try to do an integer conversion first if no characters
10947 indicating "float" have been found.
10952 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10954 if (flags == IS_NUMBER_IN_UV) {
10956 sv = newSViv(uv); /* Prefer IVs over UVs. */
10959 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10960 if (uv <= (UV) IV_MIN)
10961 sv = newSViv(-(IV)uv);
10968 /* terminate the string */
10970 nv = Atof(PL_tokenbuf);
10975 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10976 const char *const key = floatit ? "float" : "integer";
10977 const STRLEN keylen = floatit ? 5 : 7;
10978 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10979 key, keylen, sv, NULL, NULL, 0);
10983 /* if it starts with a v, it could be a v-string */
10986 sv = newSV(5); /* preallocate storage space */
10987 s = scan_vstring(s, PL_bufend, sv);
10991 /* make the op for the constant and return */
10994 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10996 lvalp->opval = NULL;
11002 S_scan_formline(pTHX_ register char *s)
11007 SV * const stuff = newSVpvs("");
11008 bool needargs = FALSE;
11009 bool eofmt = FALSE;
11011 char *tokenstart = s;
11012 SV* savewhite = NULL;
11014 if (PL_madskills) {
11015 savewhite = PL_thiswhite;
11020 PERL_ARGS_ASSERT_SCAN_FORMLINE;
11022 while (!needargs) {
11025 #ifdef PERL_STRICT_CR
11026 while (SPACE_OR_TAB(*t))
11029 while (SPACE_OR_TAB(*t) || *t == '\r')
11032 if (*t == '\n' || t == PL_bufend) {
11037 eol = (char *) memchr(s,'\n',PL_bufend-s);
11041 for (t = s; t < eol; t++) {
11042 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
11044 goto enough; /* ~~ must be first line in formline */
11046 if (*t == '@' || *t == '^')
11050 sv_catpvn(stuff, s, eol-s);
11051 #ifndef PERL_STRICT_CR
11052 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
11053 char *end = SvPVX(stuff) + SvCUR(stuff);
11056 SvCUR_set(stuff, SvCUR(stuff) - 1);
11064 if ((PL_rsfp || PL_parser->filtered)
11065 && PL_parser->form_lex_state == LEX_NORMAL) {
11068 if (PL_madskills) {
11070 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
11072 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
11075 PL_bufptr = PL_bufend;
11076 COPLINE_INC_WITH_HERELINES;
11077 got_some = lex_next_chunk(0);
11078 CopLINE_dec(PL_curcop);
11081 tokenstart = PL_bufptr;
11089 if (!SvCUR(stuff) || needargs)
11090 PL_lex_state = PL_parser->form_lex_state;
11091 if (SvCUR(stuff)) {
11092 PL_expect = XSTATE;
11094 start_force(PL_curforce);
11095 NEXTVAL_NEXTTOKE.ival = 0;
11096 force_next(FORMLBRACK);
11099 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
11101 else if (PL_encoding)
11102 sv_recode_to_utf8(stuff, PL_encoding);
11104 start_force(PL_curforce);
11105 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
11109 SvREFCNT_dec(stuff);
11111 PL_lex_formbrack = 0;
11114 if (PL_madskills) {
11116 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
11118 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
11119 PL_thiswhite = savewhite;
11126 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
11129 const I32 oldsavestack_ix = PL_savestack_ix;
11130 CV* const outsidecv = PL_compcv;
11132 SAVEI32(PL_subline);
11133 save_item(PL_subname);
11134 SAVESPTR(PL_compcv);
11136 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
11137 CvFLAGS(PL_compcv) |= flags;
11139 PL_subline = CopLINE(PL_curcop);
11140 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
11141 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
11142 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
11143 if (outsidecv && CvPADLIST(outsidecv))
11144 CvPADLIST(PL_compcv)->xpadl_outid =
11145 PadlistNAMES(CvPADLIST(outsidecv));
11147 return oldsavestack_ix;
11151 #pragma segment Perl_yylex
11154 S_yywarn(pTHX_ const char *const s, U32 flags)
11158 PERL_ARGS_ASSERT_YYWARN;
11160 PL_in_eval |= EVAL_WARNONLY;
11161 yyerror_pv(s, flags);
11162 PL_in_eval &= ~EVAL_WARNONLY;
11167 Perl_yyerror(pTHX_ const char *const s)
11169 PERL_ARGS_ASSERT_YYERROR;
11170 return yyerror_pvn(s, strlen(s), 0);
11174 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
11176 PERL_ARGS_ASSERT_YYERROR_PV;
11177 return yyerror_pvn(s, strlen(s), flags);
11181 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
11184 const char *context = NULL;
11187 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
11188 int yychar = PL_parser->yychar;
11190 PERL_ARGS_ASSERT_YYERROR_PVN;
11192 if (!yychar || (yychar == ';' && !PL_rsfp))
11193 sv_catpvs(where_sv, "at EOF");
11194 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
11195 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
11196 PL_oldbufptr != PL_bufptr) {
11199 The code below is removed for NetWare because it abends/crashes on NetWare
11200 when the script has error such as not having the closing quotes like:
11201 if ($var eq "value)
11202 Checking of white spaces is anyway done in NetWare code.
11205 while (isSPACE(*PL_oldoldbufptr))
11208 context = PL_oldoldbufptr;
11209 contlen = PL_bufptr - PL_oldoldbufptr;
11211 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
11212 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
11215 The code below is removed for NetWare because it abends/crashes on NetWare
11216 when the script has error such as not having the closing quotes like:
11217 if ($var eq "value)
11218 Checking of white spaces is anyway done in NetWare code.
11221 while (isSPACE(*PL_oldbufptr))
11224 context = PL_oldbufptr;
11225 contlen = PL_bufptr - PL_oldbufptr;
11227 else if (yychar > 255)
11228 sv_catpvs(where_sv, "next token ???");
11229 else if (yychar == -2) { /* YYEMPTY */
11230 if (PL_lex_state == LEX_NORMAL ||
11231 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11232 sv_catpvs(where_sv, "at end of line");
11233 else if (PL_lex_inpat)
11234 sv_catpvs(where_sv, "within pattern");
11236 sv_catpvs(where_sv, "within string");
11239 sv_catpvs(where_sv, "next char ");
11241 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11242 else if (isPRINT_LC(yychar)) {
11243 const char string = yychar;
11244 sv_catpvn(where_sv, &string, 1);
11247 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11249 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
11250 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11251 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
11253 Perl_sv_catpvf(aTHX_ msg, "near \"%"SVf"\"\n",
11254 SVfARG(newSVpvn_flags(context, contlen,
11255 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
11257 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11258 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11259 Perl_sv_catpvf(aTHX_ msg,
11260 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11261 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11264 if (PL_in_eval & EVAL_WARNONLY) {
11265 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11269 if (PL_error_count >= 10) {
11271 if (PL_in_eval && ((errsv = ERRSV), SvCUR(errsv)))
11272 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11273 SVfARG(errsv), OutCopFILE(PL_curcop));
11275 Perl_croak(aTHX_ "%s has too many errors.\n",
11276 OutCopFILE(PL_curcop));
11279 PL_in_my_stash = NULL;
11283 #pragma segment Main
11287 S_swallow_bom(pTHX_ U8 *s)
11290 const STRLEN slen = SvCUR(PL_linestr);
11292 PERL_ARGS_ASSERT_SWALLOW_BOM;
11296 if (s[1] == 0xFE) {
11297 /* UTF-16 little-endian? (or UTF-32LE?) */
11298 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11299 /* diag_listed_as: Unsupported script encoding %s */
11300 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11301 #ifndef PERL_NO_UTF16_FILTER
11302 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11304 if (PL_bufend > (char*)s) {
11305 s = add_utf16_textfilter(s, TRUE);
11308 /* diag_listed_as: Unsupported script encoding %s */
11309 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11314 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11315 #ifndef PERL_NO_UTF16_FILTER
11316 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11318 if (PL_bufend > (char *)s) {
11319 s = add_utf16_textfilter(s, FALSE);
11322 /* diag_listed_as: Unsupported script encoding %s */
11323 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11328 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
11329 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11330 s += 3; /* UTF-8 */
11336 if (s[2] == 0xFE && s[3] == 0xFF) {
11337 /* UTF-32 big-endian */
11338 /* diag_listed_as: Unsupported script encoding %s */
11339 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11342 else if (s[2] == 0 && s[3] != 0) {
11345 * are a good indicator of UTF-16BE. */
11346 #ifndef PERL_NO_UTF16_FILTER
11347 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11348 s = add_utf16_textfilter(s, FALSE);
11350 /* diag_listed_as: Unsupported script encoding %s */
11351 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11357 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
11358 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11359 s += 4; /* UTF-8 */
11365 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11368 * are a good indicator of UTF-16LE. */
11369 #ifndef PERL_NO_UTF16_FILTER
11370 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11371 s = add_utf16_textfilter(s, TRUE);
11373 /* diag_listed_as: Unsupported script encoding %s */
11374 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11382 #ifndef PERL_NO_UTF16_FILTER
11384 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11387 SV *const filter = FILTER_DATA(idx);
11388 /* We re-use this each time round, throwing the contents away before we
11390 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11391 SV *const utf8_buffer = filter;
11392 IV status = IoPAGE(filter);
11393 const bool reverse = cBOOL(IoLINES(filter));
11396 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11398 /* As we're automatically added, at the lowest level, and hence only called
11399 from this file, we can be sure that we're not called in block mode. Hence
11400 don't bother writing code to deal with block mode. */
11402 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11405 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11407 DEBUG_P(PerlIO_printf(Perl_debug_log,
11408 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11409 FPTR2DPTR(void *, S_utf16_textfilter),
11410 reverse ? 'l' : 'b', idx, maxlen, status,
11411 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11418 /* First, look in our buffer of existing UTF-8 data: */
11419 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11423 } else if (status == 0) {
11425 IoPAGE(filter) = 0;
11426 nl = SvEND(utf8_buffer);
11429 STRLEN got = nl - SvPVX(utf8_buffer);
11430 /* Did we have anything to append? */
11432 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11433 /* Everything else in this code works just fine if SVp_POK isn't
11434 set. This, however, needs it, and we need it to work, else
11435 we loop infinitely because the buffer is never consumed. */
11436 sv_chop(utf8_buffer, nl);
11440 /* OK, not a complete line there, so need to read some more UTF-16.
11441 Read an extra octect if the buffer currently has an odd number. */
11445 if (SvCUR(utf16_buffer) >= 2) {
11446 /* Location of the high octet of the last complete code point.
11447 Gosh, UTF-16 is a pain. All the benefits of variable length,
11448 *coupled* with all the benefits of partial reads and
11450 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11451 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11453 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11457 /* We have the first half of a surrogate. Read more. */
11458 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11461 status = FILTER_READ(idx + 1, utf16_buffer,
11462 160 + (SvCUR(utf16_buffer) & 1));
11463 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11464 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11467 IoPAGE(filter) = status;
11472 chars = SvCUR(utf16_buffer) >> 1;
11473 have = SvCUR(utf8_buffer);
11474 SvGROW(utf8_buffer, have + chars * 3 + 1);
11477 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11478 (U8*)SvPVX_const(utf8_buffer) + have,
11479 chars * 2, &newlen);
11481 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11482 (U8*)SvPVX_const(utf8_buffer) + have,
11483 chars * 2, &newlen);
11485 SvCUR_set(utf8_buffer, have + newlen);
11488 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11489 it's private to us, and utf16_to_utf8{,reversed} take a
11490 (pointer,length) pair, rather than a NUL-terminated string. */
11491 if(SvCUR(utf16_buffer) & 1) {
11492 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11493 SvCUR_set(utf16_buffer, 1);
11495 SvCUR_set(utf16_buffer, 0);
11498 DEBUG_P(PerlIO_printf(Perl_debug_log,
11499 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11501 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11502 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11507 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11509 SV *filter = filter_add(S_utf16_textfilter, NULL);
11511 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11513 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11514 sv_setpvs(filter, "");
11515 IoLINES(filter) = reversed;
11516 IoPAGE(filter) = 1; /* Not EOF */
11518 /* Sadly, we have to return a valid pointer, come what may, so we have to
11519 ignore any error return from this. */
11520 SvCUR_set(PL_linestr, 0);
11521 if (FILTER_READ(0, PL_linestr, 0)) {
11522 SvUTF8_on(PL_linestr);
11524 SvUTF8_on(PL_linestr);
11526 PL_bufend = SvEND(PL_linestr);
11527 return (U8*)SvPVX(PL_linestr);
11532 Returns a pointer to the next character after the parsed
11533 vstring, as well as updating the passed in sv.
11535 Function must be called like
11538 s = scan_vstring(s,e,sv);
11540 where s and e are the start and end of the string.
11541 The sv should already be large enough to store the vstring
11542 passed in, for performance reasons.
11547 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11550 const char *pos = s;
11551 const char *start = s;
11553 PERL_ARGS_ASSERT_SCAN_VSTRING;
11555 if (*pos == 'v') pos++; /* get past 'v' */
11556 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11558 if ( *pos != '.') {
11559 /* this may not be a v-string if followed by => */
11560 const char *next = pos;
11561 while (next < e && isSPACE(*next))
11563 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11564 /* return string not v-string */
11565 sv_setpvn(sv,(char *)s,pos-s);
11566 return (char *)pos;
11570 if (!isALPHA(*pos)) {
11571 U8 tmpbuf[UTF8_MAXBYTES+1];
11574 s++; /* get past 'v' */
11579 /* this is atoi() that tolerates underscores */
11582 const char *end = pos;
11584 while (--end >= s) {
11586 const UV orev = rev;
11587 rev += (*end - '0') * mult;
11590 /* diag_listed_as: Integer overflow in %s number */
11591 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11592 "Integer overflow in decimal number");
11596 if (rev > 0x7FFFFFFF)
11597 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11599 /* Append native character for the rev point */
11600 tmpend = uvchr_to_utf8(tmpbuf, rev);
11601 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11602 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11604 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11610 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11614 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11621 Perl_keyword_plugin_standard(pTHX_
11622 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11624 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11625 PERL_UNUSED_CONTEXT;
11626 PERL_UNUSED_ARG(keyword_ptr);
11627 PERL_UNUSED_ARG(keyword_len);
11628 PERL_UNUSED_ARG(op_ptr);
11629 return KEYWORD_PLUGIN_DECLINE;
11632 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11634 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11636 SAVEI32(PL_lex_brackets);
11637 if (PL_lex_brackets > 100)
11638 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11639 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11640 SAVEI32(PL_lex_allbrackets);
11641 PL_lex_allbrackets = 0;
11642 SAVEI8(PL_lex_fakeeof);
11643 PL_lex_fakeeof = (U8)fakeeof;
11644 if(yyparse(gramtype) && !PL_parser->error_count)
11645 qerror(Perl_mess(aTHX_ "Parse error"));
11648 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11650 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11654 SAVEVPTR(PL_eval_root);
11655 PL_eval_root = NULL;
11656 parse_recdescent(gramtype, fakeeof);
11662 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11664 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11667 if (flags & ~PARSE_OPTIONAL)
11668 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11669 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11670 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11671 if (!PL_parser->error_count)
11672 qerror(Perl_mess(aTHX_ "Parse error"));
11673 exprop = newOP(OP_NULL, 0);
11679 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11681 Parse a Perl arithmetic expression. This may contain operators of precedence
11682 down to the bit shift operators. The expression must be followed (and thus
11683 terminated) either by a comparison or lower-precedence operator or by
11684 something that would normally terminate an expression such as semicolon.
11685 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11686 otherwise it is mandatory. It is up to the caller to ensure that the
11687 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11688 the source of the code to be parsed and the lexical context for the
11691 The op tree representing the expression is returned. If an optional
11692 expression is absent, a null pointer is returned, otherwise the pointer
11695 If an error occurs in parsing or compilation, in most cases a valid op
11696 tree is returned anyway. The error is reflected in the parser state,
11697 normally resulting in a single exception at the top level of parsing
11698 which covers all the compilation errors that occurred. Some compilation
11699 errors, however, will throw an exception immediately.
11705 Perl_parse_arithexpr(pTHX_ U32 flags)
11707 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11711 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11713 Parse a Perl term expression. This may contain operators of precedence
11714 down to the assignment operators. The expression must be followed (and thus
11715 terminated) either by a comma or lower-precedence operator or by
11716 something that would normally terminate an expression such as semicolon.
11717 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11718 otherwise it is mandatory. It is up to the caller to ensure that the
11719 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11720 the source of the code to be parsed and the lexical context for the
11723 The op tree representing the expression is returned. If an optional
11724 expression is absent, a null pointer is returned, otherwise the pointer
11727 If an error occurs in parsing or compilation, in most cases a valid op
11728 tree is returned anyway. The error is reflected in the parser state,
11729 normally resulting in a single exception at the top level of parsing
11730 which covers all the compilation errors that occurred. Some compilation
11731 errors, however, will throw an exception immediately.
11737 Perl_parse_termexpr(pTHX_ U32 flags)
11739 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11743 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11745 Parse a Perl list expression. This may contain operators of precedence
11746 down to the comma operator. The expression must be followed (and thus
11747 terminated) either by a low-precedence logic operator such as C<or> or by
11748 something that would normally terminate an expression such as semicolon.
11749 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11750 otherwise it is mandatory. It is up to the caller to ensure that the
11751 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11752 the source of the code to be parsed and the lexical context for the
11755 The op tree representing the expression is returned. If an optional
11756 expression is absent, a null pointer is returned, otherwise the pointer
11759 If an error occurs in parsing or compilation, in most cases a valid op
11760 tree is returned anyway. The error is reflected in the parser state,
11761 normally resulting in a single exception at the top level of parsing
11762 which covers all the compilation errors that occurred. Some compilation
11763 errors, however, will throw an exception immediately.
11769 Perl_parse_listexpr(pTHX_ U32 flags)
11771 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11775 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11777 Parse a single complete Perl expression. This allows the full
11778 expression grammar, including the lowest-precedence operators such
11779 as C<or>. The expression must be followed (and thus terminated) by a
11780 token that an expression would normally be terminated by: end-of-file,
11781 closing bracketing punctuation, semicolon, or one of the keywords that
11782 signals a postfix expression-statement modifier. If I<flags> includes
11783 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11784 mandatory. It is up to the caller to ensure that the dynamic parser
11785 state (L</PL_parser> et al) is correctly set to reflect the source of
11786 the code to be parsed and the lexical context for the expression.
11788 The op tree representing the expression is returned. If an optional
11789 expression is absent, a null pointer is returned, otherwise the pointer
11792 If an error occurs in parsing or compilation, in most cases a valid op
11793 tree is returned anyway. The error is reflected in the parser state,
11794 normally resulting in a single exception at the top level of parsing
11795 which covers all the compilation errors that occurred. Some compilation
11796 errors, however, will throw an exception immediately.
11802 Perl_parse_fullexpr(pTHX_ U32 flags)
11804 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11808 =for apidoc Amx|OP *|parse_block|U32 flags
11810 Parse a single complete Perl code block. This consists of an opening
11811 brace, a sequence of statements, and a closing brace. The block
11812 constitutes a lexical scope, so C<my> variables and various compile-time
11813 effects can be contained within it. It is up to the caller to ensure
11814 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11815 reflect the source of the code to be parsed and the lexical context for
11818 The op tree representing the code block is returned. This is always a
11819 real op, never a null pointer. It will normally be a C<lineseq> list,
11820 including C<nextstate> or equivalent ops. No ops to construct any kind
11821 of runtime scope are included by virtue of it being a block.
11823 If an error occurs in parsing or compilation, in most cases a valid op
11824 tree (most likely null) is returned anyway. The error is reflected in
11825 the parser state, normally resulting in a single exception at the top
11826 level of parsing which covers all the compilation errors that occurred.
11827 Some compilation errors, however, will throw an exception immediately.
11829 The I<flags> parameter is reserved for future use, and must always
11836 Perl_parse_block(pTHX_ U32 flags)
11839 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11840 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11844 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11846 Parse a single unadorned Perl statement. This may be a normal imperative
11847 statement or a declaration that has compile-time effect. It does not
11848 include any label or other affixture. It is up to the caller to ensure
11849 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11850 reflect the source of the code to be parsed and the lexical context for
11853 The op tree representing the statement is returned. This may be a
11854 null pointer if the statement is null, for example if it was actually
11855 a subroutine definition (which has compile-time side effects). If not
11856 null, it will be ops directly implementing the statement, suitable to
11857 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11858 equivalent op (except for those embedded in a scope contained entirely
11859 within the statement).
11861 If an error occurs in parsing or compilation, in most cases a valid op
11862 tree (most likely null) is returned anyway. The error is reflected in
11863 the parser state, normally resulting in a single exception at the top
11864 level of parsing which covers all the compilation errors that occurred.
11865 Some compilation errors, however, will throw an exception immediately.
11867 The I<flags> parameter is reserved for future use, and must always
11874 Perl_parse_barestmt(pTHX_ U32 flags)
11877 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11878 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11882 =for apidoc Amx|SV *|parse_label|U32 flags
11884 Parse a single label, possibly optional, of the type that may prefix a
11885 Perl statement. It is up to the caller to ensure that the dynamic parser
11886 state (L</PL_parser> et al) is correctly set to reflect the source of
11887 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11888 label is optional, otherwise it is mandatory.
11890 The name of the label is returned in the form of a fresh scalar. If an
11891 optional label is absent, a null pointer is returned.
11893 If an error occurs in parsing, which can only occur if the label is
11894 mandatory, a valid label is returned anyway. The error is reflected in
11895 the parser state, normally resulting in a single exception at the top
11896 level of parsing which covers all the compilation errors that occurred.
11902 Perl_parse_label(pTHX_ U32 flags)
11904 if (flags & ~PARSE_OPTIONAL)
11905 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11906 if (PL_lex_state == LEX_KNOWNEXT) {
11907 PL_parser->yychar = yylex();
11908 if (PL_parser->yychar == LABEL) {
11909 char * const lpv = pl_yylval.pval;
11910 STRLEN llen = strlen(lpv);
11911 PL_parser->yychar = YYEMPTY;
11912 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11919 STRLEN wlen, bufptr_pos;
11922 if (!isIDFIRST_lazy_if(s, UTF))
11924 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11925 if (word_takes_any_delimeter(s, wlen))
11927 bufptr_pos = s - SvPVX(PL_linestr);
11929 lex_read_space(LEX_KEEP_PREVIOUS);
11931 s = SvPVX(PL_linestr) + bufptr_pos;
11932 if (t[0] == ':' && t[1] != ':') {
11933 PL_oldoldbufptr = PL_oldbufptr;
11936 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11940 if (flags & PARSE_OPTIONAL) {
11943 qerror(Perl_mess(aTHX_ "Parse error"));
11944 return newSVpvs("x");
11951 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11953 Parse a single complete Perl statement. This may be a normal imperative
11954 statement or a declaration that has compile-time effect, and may include
11955 optional labels. It is up to the caller to ensure that the dynamic
11956 parser state (L</PL_parser> et al) is correctly set to reflect the source
11957 of the code to be parsed and the lexical context for the statement.
11959 The op tree representing the statement is returned. This may be a
11960 null pointer if the statement is null, for example if it was actually
11961 a subroutine definition (which has compile-time side effects). If not
11962 null, it will be the result of a L</newSTATEOP> call, normally including
11963 a C<nextstate> or equivalent op.
11965 If an error occurs in parsing or compilation, in most cases a valid op
11966 tree (most likely null) is returned anyway. The error is reflected in
11967 the parser state, normally resulting in a single exception at the top
11968 level of parsing which covers all the compilation errors that occurred.
11969 Some compilation errors, however, will throw an exception immediately.
11971 The I<flags> parameter is reserved for future use, and must always
11978 Perl_parse_fullstmt(pTHX_ U32 flags)
11981 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11982 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11986 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11988 Parse a sequence of zero or more Perl statements. These may be normal
11989 imperative statements, including optional labels, or declarations
11990 that have compile-time effect, or any mixture thereof. The statement
11991 sequence ends when a closing brace or end-of-file is encountered in a
11992 place where a new statement could have validly started. It is up to
11993 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11994 is correctly set to reflect the source of the code to be parsed and the
11995 lexical context for the statements.
11997 The op tree representing the statement sequence is returned. This may
11998 be a null pointer if the statements were all null, for example if there
11999 were no statements or if there were only subroutine definitions (which
12000 have compile-time side effects). If not null, it will be a C<lineseq>
12001 list, normally including C<nextstate> or equivalent ops.
12003 If an error occurs in parsing or compilation, in most cases a valid op
12004 tree is returned anyway. The error is reflected in the parser state,
12005 normally resulting in a single exception at the top level of parsing
12006 which covers all the compilation errors that occurred. Some compilation
12007 errors, however, will throw an exception immediately.
12009 The I<flags> parameter is reserved for future use, and must always
12016 Perl_parse_stmtseq(pTHX_ U32 flags)
12021 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
12022 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
12023 c = lex_peek_unichar(0);
12024 if (c != -1 && c != /*{*/'}')
12025 qerror(Perl_mess(aTHX_ "Parse error"));
12031 * c-indentation-style: bsd
12032 * c-basic-offset: 4
12033 * indent-tabs-mode: nil
12036 * ex: set ts=8 sts=4 sw=4 et: