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);
2970 while (s < send || dorange) {
2972 /* get transliterations out of the way (they're most literal) */
2973 if (PL_lex_inwhat == OP_TRANS) {
2974 /* expand a range A-Z to the full set of characters. AIE! */
2976 I32 i; /* current expanded character */
2977 I32 min; /* first character in range */
2978 I32 max; /* last character in range */
2989 char * const c = (char*)utf8_hop((U8*)d, -1);
2993 *c = (char)UTF_TO_NATIVE(0xff);
2994 /* mark the range as done, and continue */
3000 i = d - SvPVX_const(sv); /* remember current offset */
3003 SvLEN(sv) + (has_utf8 ?
3004 (512 - UTF_CONTINUATION_MARK +
3007 /* How many two-byte within 0..255: 128 in UTF-8,
3008 * 96 in UTF-8-mod. */
3010 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
3012 d = SvPVX(sv) + i; /* refresh d after realloc */
3016 for (j = 0; j <= 1; j++) {
3017 char * const c = (char*)utf8_hop((U8*)d, -1);
3018 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
3024 max = (U8)0xff; /* only to \xff */
3025 uvmax = uv; /* \x{100} to uvmax */
3027 d = c; /* eat endpoint chars */
3032 d -= 2; /* eat the first char and the - */
3033 min = (U8)*d; /* first char in range */
3034 max = (U8)d[1]; /* last char in range */
3042 "Invalid range \"%c-%c\" in transliteration operator",
3043 (char)min, (char)max);
3047 if (literal_endpoint == 2 &&
3048 ((isLOWER(min) && isLOWER(max)) ||
3049 (isUPPER(min) && isUPPER(max)))) {
3051 for (i = min; i <= max; i++)
3053 *d++ = NATIVE_TO_NEED(has_utf8,i);
3055 for (i = min; i <= max; i++)
3057 *d++ = NATIVE_TO_NEED(has_utf8,i);
3062 for (i = min; i <= max; i++)
3065 const U8 ch = (U8)NATIVE_TO_UTF(i);
3066 if (UNI_IS_INVARIANT(ch))
3069 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
3070 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
3079 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
3081 *d++ = (char)UTF_TO_NATIVE(0xff);
3083 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
3087 /* mark the range as done, and continue */
3091 literal_endpoint = 0;
3096 /* range begins (ignore - as first or last char) */
3097 else if (*s == '-' && s+1 < send && s != start) {
3100 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
3107 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
3117 literal_endpoint = 0;
3118 native_range = TRUE;
3123 /* if we get here, we're not doing a transliteration */
3125 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3128 while (s1 >= start && *s1-- == '\\')
3131 in_charclass = TRUE;
3134 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3137 while (s1 >= start && *s1-- == '\\')
3140 in_charclass = FALSE;
3143 /* skip for regexp comments /(?#comment)/, except for the last
3144 * char, which will be done separately.
3145 * Stop on (?{..}) and friends */
3147 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
3149 while (s+1 < send && *s != ')')
3150 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3152 else if (!PL_lex_casemods && !in_charclass &&
3153 ( s[2] == '{' /* This should match regcomp.c */
3154 || (s[2] == '?' && s[3] == '{')))
3160 /* likewise skip #-initiated comments in //x patterns */
3161 else if (*s == '#' && PL_lex_inpat &&
3162 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3163 while (s+1 < send && *s != '\n')
3164 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3167 /* no further processing of single-quoted regex */
3168 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3169 goto default_action;
3171 /* check for embedded arrays
3172 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3174 else if (*s == '@' && s[1]) {
3175 if (isALNUM_lazy_if(s+1,UTF))
3177 if (strchr(":'{$", s[1]))
3179 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3180 break; /* in regexp, neither @+ nor @- are interpolated */
3183 /* check for embedded scalars. only stop if we're sure it's a
3186 else if (*s == '$') {
3187 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3189 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3191 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3192 "Possible unintended interpolation of $\\ in regex");
3194 break; /* in regexp, $ might be tail anchor */
3198 /* End of else if chain - OP_TRANS rejoin rest */
3201 if (*s == '\\' && s+1 < send) {
3202 char* e; /* Can be used for ending '}', etc. */
3206 /* warn on \1 - \9 in substitution replacements, but note that \11
3207 * is an octal; and \19 is \1 followed by '9' */
3208 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3209 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3211 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3216 /* string-change backslash escapes */
3217 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3221 /* In a pattern, process \N, but skip any other backslash escapes.
3222 * This is because we don't want to translate an escape sequence
3223 * into a meta symbol and have the regex compiler use the meta
3224 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3225 * in spite of this, we do have to process \N here while the proper
3226 * charnames handler is in scope. See bugs #56444 and #62056.
3227 * There is a complication because \N in a pattern may also stand
3228 * for 'match a non-nl', and not mean a charname, in which case its
3229 * processing should be deferred to the regex compiler. To be a
3230 * charname it must be followed immediately by a '{', and not look
3231 * like \N followed by a curly quantifier, i.e., not something like
3232 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3234 else if (PL_lex_inpat
3237 || regcurly(s + 1)))
3239 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3240 goto default_action;
3245 /* quoted - in transliterations */
3247 if (PL_lex_inwhat == OP_TRANS) {
3255 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3256 "Unrecognized escape \\%c passed through",
3258 /* default action is to copy the quoted character */
3259 goto default_action;
3262 /* eg. \132 indicates the octal constant 0132 */
3263 case '0': case '1': case '2': case '3':
3264 case '4': case '5': case '6': case '7':
3268 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3271 goto NUM_ESCAPE_INSERT;
3273 /* eg. \o{24} indicates the octal constant \024 */
3279 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
3285 goto NUM_ESCAPE_INSERT;
3288 /* eg. \x24 indicates the hex constant 0x24 */
3294 bool valid = grok_bslash_x(s, &uv, &len, &error, 1);
3303 /* Insert oct or hex escaped character. There will always be
3304 * enough room in sv since such escapes will be longer than any
3305 * UTF-8 sequence they can end up as, except if they force us
3306 * to recode the rest of the string into utf8 */
3308 /* Here uv is the ordinal of the next character being added in
3309 * unicode (converted from native). */
3310 if (!UNI_IS_INVARIANT(uv)) {
3311 if (!has_utf8 && uv > 255) {
3312 /* Might need to recode whatever we have accumulated so
3313 * far if it contains any chars variant in utf8 or
3316 SvCUR_set(sv, d - SvPVX_const(sv));
3319 /* See Note on sizing above. */
3320 sv_utf8_upgrade_flags_grow(sv,
3321 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3322 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3323 d = SvPVX(sv) + SvCUR(sv);
3328 d = (char*)uvuni_to_utf8((U8*)d, uv);
3329 if (PL_lex_inwhat == OP_TRANS &&
3330 PL_sublex_info.sub_op) {
3331 PL_sublex_info.sub_op->op_private |=
3332 (PL_lex_repl ? OPpTRANS_FROM_UTF
3336 if (uv > 255 && !dorange)
3337 native_range = FALSE;
3350 /* In a non-pattern \N must be a named character, like \N{LATIN
3351 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3352 * mean to match a non-newline. For non-patterns, named
3353 * characters are converted to their string equivalents. In
3354 * patterns, named characters are not converted to their
3355 * ultimate forms for the same reasons that other escapes
3356 * aren't. Instead, they are converted to the \N{U+...} form
3357 * to get the value from the charnames that is in effect right
3358 * now, while preserving the fact that it was a named character
3359 * so that the regex compiler knows this */
3361 /* This section of code doesn't generally use the
3362 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3363 * a close examination of this macro and determined it is a
3364 * no-op except on utfebcdic variant characters. Every
3365 * character generated by this that would normally need to be
3366 * enclosed by this macro is invariant, so the macro is not
3367 * needed, and would complicate use of copy(). XXX There are
3368 * other parts of this file where the macro is used
3369 * inconsistently, but are saved by it being a no-op */
3371 /* The structure of this section of code (besides checking for
3372 * errors and upgrading to utf8) is:
3373 * Further disambiguate between the two meanings of \N, and if
3374 * not a charname, go process it elsewhere
3375 * If of form \N{U+...}, pass it through if a pattern;
3376 * otherwise convert to utf8
3377 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3378 * pattern; otherwise convert to utf8 */
3380 /* Here, s points to the 'N'; the test below is guaranteed to
3381 * succeed if we are being called on a pattern as we already
3382 * know from a test above that the next character is a '{'.
3383 * On a non-pattern \N must mean 'named sequence, which
3384 * requires braces */
3387 yyerror("Missing braces on \\N{}");
3392 /* If there is no matching '}', it is an error. */
3393 if (! (e = strchr(s, '}'))) {
3394 if (! PL_lex_inpat) {
3395 yyerror("Missing right brace on \\N{}");
3397 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3402 /* Here it looks like a named character */
3404 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3405 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3406 | PERL_SCAN_DISALLOW_PREFIX;
3409 /* For \N{U+...}, the '...' is a unicode value even on
3410 * EBCDIC machines */
3411 s += 2; /* Skip to next char after the 'U+' */
3413 uv = grok_hex(s, &len, &flags, NULL);
3414 if (len == 0 || len != (STRLEN)(e - s)) {
3415 yyerror("Invalid hexadecimal number in \\N{U+...}");
3422 /* On non-EBCDIC platforms, pass through to the regex
3423 * compiler unchanged. The reason we evaluated the
3424 * number above is to make sure there wasn't a syntax
3425 * error. But on EBCDIC we convert to native so
3426 * downstream code can continue to assume it's native
3428 s -= 5; /* Include the '\N{U+' */
3430 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3433 (unsigned int) UNI_TO_NATIVE(uv));
3435 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3439 else { /* Not a pattern: convert the hex to string */
3441 /* If destination is not in utf8, unconditionally
3442 * recode it to be so. This is because \N{} implies
3443 * Unicode semantics, and scalars have to be in utf8
3444 * to guarantee those semantics */
3446 SvCUR_set(sv, d - SvPVX_const(sv));
3449 /* See Note on sizing above. */
3450 sv_utf8_upgrade_flags_grow(
3452 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3453 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3454 d = SvPVX(sv) + SvCUR(sv);
3458 /* Add the string to the output */
3459 if (UNI_IS_INVARIANT(uv)) {
3462 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3465 else /* Here is \N{NAME} but not \N{U+...}. */
3466 if ((res = get_and_check_backslash_N_name(s, e)))
3469 const char *str = SvPV_const(res, len);
3472 if (! len) { /* The name resolved to an empty string */
3473 Copy("\\N{}", d, 4, char);
3477 /* In order to not lose information for the regex
3478 * compiler, pass the result in the specially made
3479 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3480 * the code points in hex of each character
3481 * returned by charnames */
3483 const char *str_end = str + len;
3484 const STRLEN off = d - SvPVX_const(sv);
3486 if (! SvUTF8(res)) {
3487 /* For the non-UTF-8 case, we can determine the
3488 * exact length needed without having to parse
3489 * through the string. Each character takes up
3490 * 2 hex digits plus either a trailing dot or
3492 d = off + SvGROW(sv, off
3494 + 6 /* For the "\N{U+", and
3496 + (STRLEN)(send - e));
3497 Copy("\\N{U+", d, 5, char);
3499 while (str < str_end) {
3501 my_snprintf(hex_string, sizeof(hex_string),
3502 "%02X.", (U8) *str);
3503 Copy(hex_string, d, 3, char);
3507 d--; /* We will overwrite below the final
3508 dot with a right brace */
3511 STRLEN char_length; /* cur char's byte length */
3513 /* and the number of bytes after this is
3514 * translated into hex digits */
3515 STRLEN output_length;
3517 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3518 * for max('U+', '.'); and 1 for NUL */
3519 char hex_string[2 * UTF8_MAXBYTES + 5];
3521 /* Get the first character of the result. */
3522 U32 uv = utf8n_to_uvuni((U8 *) str,
3526 /* Convert first code point to hex, including
3527 * the boiler plate before it. For all these,
3528 * we convert to native format so that
3529 * downstream code can continue to assume the
3530 * input is native */
3532 my_snprintf(hex_string, sizeof(hex_string),
3534 (unsigned int) UNI_TO_NATIVE(uv));
3536 /* Make sure there is enough space to hold it */
3537 d = off + SvGROW(sv, off
3539 + (STRLEN)(send - e)
3540 + 2); /* '}' + NUL */
3542 Copy(hex_string, d, output_length, char);
3545 /* For each subsequent character, append dot and
3546 * its ordinal in hex */
3547 while ((str += char_length) < str_end) {
3548 const STRLEN off = d - SvPVX_const(sv);
3549 U32 uv = utf8n_to_uvuni((U8 *) str,
3554 my_snprintf(hex_string,
3557 (unsigned int) UNI_TO_NATIVE(uv));
3559 d = off + SvGROW(sv, off
3561 + (STRLEN)(send - e)
3562 + 2); /* '}' + NUL */
3563 Copy(hex_string, d, output_length, char);
3568 *d++ = '}'; /* Done. Add the trailing brace */
3571 else { /* Here, not in a pattern. Convert the name to a
3574 /* If destination is not in utf8, unconditionally
3575 * recode it to be so. This is because \N{} implies
3576 * Unicode semantics, and scalars have to be in utf8
3577 * to guarantee those semantics */
3579 SvCUR_set(sv, d - SvPVX_const(sv));
3582 /* See Note on sizing above. */
3583 sv_utf8_upgrade_flags_grow(sv,
3584 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3585 len + (STRLEN)(send - s) + 1);
3586 d = SvPVX(sv) + SvCUR(sv);
3588 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3590 /* See Note on sizing above. (NOTE: SvCUR() is not
3591 * set correctly here). */
3592 const STRLEN off = d - SvPVX_const(sv);
3593 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3595 Copy(str, d, len, char);
3601 } /* End \N{NAME} */
3604 native_range = FALSE; /* \N{} is defined to be Unicode */
3606 s = e + 1; /* Point to just after the '}' */
3609 /* \c is a control character */
3613 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3616 yyerror("Missing control char name in \\c");
3620 /* printf-style backslashes, formfeeds, newlines, etc */
3622 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3625 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3628 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3631 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3634 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3637 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3640 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3646 } /* end if (backslash) */
3653 /* If we started with encoded form, or already know we want it,
3654 then encode the next character */
3655 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3659 /* One might think that it is wasted effort in the case of the
3660 * source being utf8 (this_utf8 == TRUE) to take the next character
3661 * in the source, convert it to an unsigned value, and then convert
3662 * it back again. But the source has not been validated here. The
3663 * routine that does the conversion checks for errors like
3666 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3667 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3669 SvCUR_set(sv, d - SvPVX_const(sv));
3672 /* See Note on sizing above. */
3673 sv_utf8_upgrade_flags_grow(sv,
3674 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3675 need + (STRLEN)(send - s) + 1);
3676 d = SvPVX(sv) + SvCUR(sv);
3678 } else if (need > len) {
3679 /* encoded value larger than old, may need extra space (NOTE:
3680 * SvCUR() is not set correctly here). See Note on sizing
3682 const STRLEN off = d - SvPVX_const(sv);
3683 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3687 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3689 if (uv > 255 && !dorange)
3690 native_range = FALSE;
3694 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3696 } /* while loop to process each character */
3698 /* terminate the string and set up the sv */
3700 SvCUR_set(sv, d - SvPVX_const(sv));
3701 if (SvCUR(sv) >= SvLEN(sv))
3702 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3703 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3706 if (PL_encoding && !has_utf8) {
3707 sv_recode_to_utf8(sv, PL_encoding);
3713 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3714 PL_sublex_info.sub_op->op_private |=
3715 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3719 /* shrink the sv if we allocated more than we used */
3720 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3721 SvPV_shrink_to_cur(sv);
3724 /* return the substring (via pl_yylval) only if we parsed anything */
3725 if (s > PL_bufptr) {
3726 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3727 const char *const key = PL_lex_inpat ? "qr" : "q";
3728 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3732 if (PL_lex_inwhat == OP_TRANS) {
3735 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3738 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3746 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3749 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3756 * Returns TRUE if there's more to the expression (e.g., a subscript),
3759 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3761 * ->[ and ->{ return TRUE
3762 * { and [ outside a pattern are always subscripts, so return TRUE
3763 * if we're outside a pattern and it's not { or [, then return FALSE
3764 * if we're in a pattern and the first char is a {
3765 * {4,5} (any digits around the comma) returns FALSE
3766 * if we're in a pattern and the first char is a [
3768 * [SOMETHING] has a funky algorithm to decide whether it's a
3769 * character class or not. It has to deal with things like
3770 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3771 * anything else returns TRUE
3774 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3777 S_intuit_more(pTHX_ register char *s)
3781 PERL_ARGS_ASSERT_INTUIT_MORE;
3783 if (PL_lex_brackets)
3785 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3787 if (*s != '{' && *s != '[')
3792 /* In a pattern, so maybe we have {n,m}. */
3800 /* On the other hand, maybe we have a character class */
3803 if (*s == ']' || *s == '^')
3806 /* this is terrifying, and it works */
3807 int weight = 2; /* let's weigh the evidence */
3809 unsigned char un_char = 255, last_un_char;
3810 const char * const send = strchr(s,']');
3811 char tmpbuf[sizeof PL_tokenbuf * 4];
3813 if (!send) /* has to be an expression */
3816 Zero(seen,256,char);
3819 else if (isDIGIT(*s)) {
3821 if (isDIGIT(s[1]) && s[2] == ']')
3827 for (; s < send; s++) {
3828 last_un_char = un_char;
3829 un_char = (unsigned char)*s;
3834 weight -= seen[un_char] * 10;
3835 if (isALNUM_lazy_if(s+1,UTF)) {
3837 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3838 len = (int)strlen(tmpbuf);
3839 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3840 UTF ? SVf_UTF8 : 0, SVt_PV))
3845 else if (*s == '$' && s[1] &&
3846 strchr("[#!%*<>()-=",s[1])) {
3847 if (/*{*/ strchr("])} =",s[2]))
3856 if (strchr("wds]",s[1]))
3858 else if (seen[(U8)'\''] || seen[(U8)'"'])
3860 else if (strchr("rnftbxcav",s[1]))
3862 else if (isDIGIT(s[1])) {
3864 while (s[1] && isDIGIT(s[1]))
3874 if (strchr("aA01! ",last_un_char))
3876 if (strchr("zZ79~",s[1]))
3878 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3879 weight -= 5; /* cope with negative subscript */
3882 if (!isALNUM(last_un_char)
3883 && !(last_un_char == '$' || last_un_char == '@'
3884 || last_un_char == '&')
3885 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3890 if (keyword(tmpbuf, d - tmpbuf, 0))
3893 if (un_char == last_un_char + 1)
3895 weight -= seen[un_char];
3900 if (weight >= 0) /* probably a character class */
3910 * Does all the checking to disambiguate
3912 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3913 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3915 * First argument is the stuff after the first token, e.g. "bar".
3917 * Not a method if foo is a filehandle.
3918 * Not a method if foo is a subroutine prototyped to take a filehandle.
3919 * Not a method if it's really "Foo $bar"
3920 * Method if it's "foo $bar"
3921 * Not a method if it's really "print foo $bar"
3922 * Method if it's really "foo package::" (interpreted as package->foo)
3923 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3924 * Not a method if bar is a filehandle or package, but is quoted with
3929 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3932 char *s = start + (*start == '$');
3933 char tmpbuf[sizeof PL_tokenbuf];
3940 PERL_ARGS_ASSERT_INTUIT_METHOD;
3942 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3944 if (cv && SvPOK(cv)) {
3945 const char *proto = CvPROTO(cv);
3953 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3954 /* start is the beginning of the possible filehandle/object,
3955 * and s is the end of it
3956 * tmpbuf is a copy of it
3959 if (*start == '$') {
3960 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3961 isUPPER(*PL_tokenbuf))
3964 len = start - SvPVX(PL_linestr);
3968 start = SvPVX(PL_linestr) + len;
3972 return *s == '(' ? FUNCMETH : METHOD;
3974 if (!keyword(tmpbuf, len, 0)) {
3975 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3979 soff = s - SvPVX(PL_linestr);
3983 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3984 if (indirgv && GvCVu(indirgv))
3986 /* filehandle or package name makes it a method */
3987 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3989 soff = s - SvPVX(PL_linestr);
3992 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3993 return 0; /* no assumptions -- "=>" quotes bareword */
3995 start_force(PL_curforce);
3996 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3997 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3998 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
4000 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
4001 ( UTF ? SVf_UTF8 : 0 )));
4006 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
4008 return *s == '(' ? FUNCMETH : METHOD;
4014 /* Encoded script support. filter_add() effectively inserts a
4015 * 'pre-processing' function into the current source input stream.
4016 * Note that the filter function only applies to the current source file
4017 * (e.g., it will not affect files 'require'd or 'use'd by this one).
4019 * The datasv parameter (which may be NULL) can be used to pass
4020 * private data to this instance of the filter. The filter function
4021 * can recover the SV using the FILTER_DATA macro and use it to
4022 * store private buffers and state information.
4024 * The supplied datasv parameter is upgraded to a PVIO type
4025 * and the IoDIRP/IoANY field is used to store the function pointer,
4026 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
4027 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
4028 * private use must be set using malloc'd pointers.
4032 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
4041 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
4042 Perl_croak(aTHX_ "Source filters apply only to byte streams");
4044 if (!PL_rsfp_filters)
4045 PL_rsfp_filters = newAV();
4048 SvUPGRADE(datasv, SVt_PVIO);
4049 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
4050 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
4051 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
4052 FPTR2DPTR(void *, IoANY(datasv)),
4053 SvPV_nolen(datasv)));
4054 av_unshift(PL_rsfp_filters, 1);
4055 av_store(PL_rsfp_filters, 0, datasv) ;
4057 !PL_parser->filtered
4058 && PL_parser->lex_flags & LEX_EVALBYTES
4059 && PL_bufptr < PL_bufend
4061 const char *s = PL_bufptr;
4062 while (s < PL_bufend) {
4064 SV *linestr = PL_parser->linestr;
4065 char *buf = SvPVX(linestr);
4066 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
4067 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
4068 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
4069 STRLEN const linestart_pos = PL_parser->linestart - buf;
4070 STRLEN const last_uni_pos =
4071 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
4072 STRLEN const last_lop_pos =
4073 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
4074 av_push(PL_rsfp_filters, linestr);
4075 PL_parser->linestr =
4076 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
4077 buf = SvPVX(PL_parser->linestr);
4078 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4079 PL_parser->bufptr = buf + bufptr_pos;
4080 PL_parser->oldbufptr = buf + oldbufptr_pos;
4081 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4082 PL_parser->linestart = buf + linestart_pos;
4083 if (PL_parser->last_uni)
4084 PL_parser->last_uni = buf + last_uni_pos;
4085 if (PL_parser->last_lop)
4086 PL_parser->last_lop = buf + last_lop_pos;
4087 SvLEN(linestr) = SvCUR(linestr);
4088 SvCUR(linestr) = s-SvPVX(linestr);
4089 PL_parser->filtered = 1;
4099 /* Delete most recently added instance of this filter function. */
4101 Perl_filter_del(pTHX_ filter_t funcp)
4106 PERL_ARGS_ASSERT_FILTER_DEL;
4109 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4110 FPTR2DPTR(void*, funcp)));
4112 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4114 /* if filter is on top of stack (usual case) just pop it off */
4115 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4116 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4117 sv_free(av_pop(PL_rsfp_filters));
4121 /* we need to search for the correct entry and clear it */
4122 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4126 /* Invoke the idxth filter function for the current rsfp. */
4127 /* maxlen 0 = read one text line */
4129 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4134 /* This API is bad. It should have been using unsigned int for maxlen.
4135 Not sure if we want to change the API, but if not we should sanity
4136 check the value here. */
4137 unsigned int correct_length
4146 PERL_ARGS_ASSERT_FILTER_READ;
4148 if (!PL_parser || !PL_rsfp_filters)
4150 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4151 /* Provide a default input filter to make life easy. */
4152 /* Note that we append to the line. This is handy. */
4153 DEBUG_P(PerlIO_printf(Perl_debug_log,
4154 "filter_read %d: from rsfp\n", idx));
4155 if (correct_length) {
4158 const int old_len = SvCUR(buf_sv);
4160 /* ensure buf_sv is large enough */
4161 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4162 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4163 correct_length)) <= 0) {
4164 if (PerlIO_error(PL_rsfp))
4165 return -1; /* error */
4167 return 0 ; /* end of file */
4169 SvCUR_set(buf_sv, old_len + len) ;
4170 SvPVX(buf_sv)[old_len + len] = '\0';
4173 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4174 if (PerlIO_error(PL_rsfp))
4175 return -1; /* error */
4177 return 0 ; /* end of file */
4180 return SvCUR(buf_sv);
4182 /* Skip this filter slot if filter has been deleted */
4183 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4184 DEBUG_P(PerlIO_printf(Perl_debug_log,
4185 "filter_read %d: skipped (filter deleted)\n",
4187 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4189 if (SvTYPE(datasv) != SVt_PVIO) {
4190 if (correct_length) {
4192 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4193 if (!remainder) return 0; /* eof */
4194 if (correct_length > remainder) correct_length = remainder;
4195 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4196 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4199 const char *s = SvEND(datasv);
4200 const char *send = SvPVX(datasv) + SvLEN(datasv);
4208 if (s == send) return 0; /* eof */
4209 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4210 SvCUR_set(datasv, s-SvPVX(datasv));
4212 return SvCUR(buf_sv);
4214 /* Get function pointer hidden within datasv */
4215 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4216 DEBUG_P(PerlIO_printf(Perl_debug_log,
4217 "filter_read %d: via function %p (%s)\n",
4218 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4219 /* Call function. The function is expected to */
4220 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4221 /* Return: <0:error, =0:eof, >0:not eof */
4222 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4226 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4230 PERL_ARGS_ASSERT_FILTER_GETS;
4232 #ifdef PERL_CR_FILTER
4233 if (!PL_rsfp_filters) {
4234 filter_add(S_cr_textfilter,NULL);
4237 if (PL_rsfp_filters) {
4239 SvCUR_set(sv, 0); /* start with empty line */
4240 if (FILTER_READ(0, sv, 0) > 0)
4241 return ( SvPVX(sv) ) ;
4246 return (sv_gets(sv, PL_rsfp, append));
4250 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4255 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4257 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4261 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4262 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4264 return GvHV(gv); /* Foo:: */
4267 /* use constant CLASS => 'MyClass' */
4268 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4269 if (gv && GvCV(gv)) {
4270 SV * const sv = cv_const_sv(GvCV(gv));
4272 pkgname = SvPV_const(sv, len);
4275 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4279 * S_readpipe_override
4280 * Check whether readpipe() is overridden, and generates the appropriate
4281 * optree, provided sublex_start() is called afterwards.
4284 S_readpipe_override(pTHX)
4287 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4288 pl_yylval.ival = OP_BACKTICK;
4290 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4292 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4293 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4294 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4296 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4297 op_append_elem(OP_LIST,
4298 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4299 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4306 * The intent of this yylex wrapper is to minimize the changes to the
4307 * tokener when we aren't interested in collecting madprops. It remains
4308 * to be seen how successful this strategy will be...
4315 char *s = PL_bufptr;
4317 /* make sure PL_thiswhite is initialized */
4321 /* previous token ate up our whitespace? */
4322 if (!PL_lasttoke && PL_nextwhite) {
4323 PL_thiswhite = PL_nextwhite;
4327 /* isolate the token, and figure out where it is without whitespace */
4328 PL_realtokenstart = -1;
4332 assert(PL_curforce < 0);
4334 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4335 if (!PL_thistoken) {
4336 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4337 PL_thistoken = newSVpvs("");
4339 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4340 PL_thistoken = newSVpvn(tstart, s - tstart);
4343 if (PL_thismad) /* install head */
4344 CURMAD('X', PL_thistoken);
4347 /* last whitespace of a sublex? */
4348 if (optype == ')' && PL_endwhite) {
4349 CURMAD('X', PL_endwhite);
4354 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4355 if (!PL_thiswhite && !PL_endwhite && !optype) {
4356 sv_free(PL_thistoken);
4361 /* put off final whitespace till peg */
4362 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4363 PL_nextwhite = PL_thiswhite;
4366 else if (PL_thisopen) {
4367 CURMAD('q', PL_thisopen);
4369 sv_free(PL_thistoken);
4373 /* Store actual token text as madprop X */
4374 CURMAD('X', PL_thistoken);
4378 /* add preceding whitespace as madprop _ */
4379 CURMAD('_', PL_thiswhite);
4383 /* add quoted material as madprop = */
4384 CURMAD('=', PL_thisstuff);
4388 /* add terminating quote as madprop Q */
4389 CURMAD('Q', PL_thisclose);
4393 /* special processing based on optype */
4397 /* opval doesn't need a TOKEN since it can already store mp */
4407 if (pl_yylval.opval)
4408 append_madprops(PL_thismad, pl_yylval.opval, 0);
4416 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4429 /* remember any fake bracket that lexer is about to discard */
4430 if (PL_lex_brackets == 1 &&
4431 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4434 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4437 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4438 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4441 break; /* don't bother looking for trailing comment */
4450 /* attach a trailing comment to its statement instead of next token */
4454 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4456 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4458 if (*s == '\n' || *s == '#') {
4459 while (s < PL_bufend && *s != '\n')
4463 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4464 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4477 /* Create new token struct. Note: opvals return early above. */
4478 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4485 S_tokenize_use(pTHX_ int is_use, char *s) {
4488 PERL_ARGS_ASSERT_TOKENIZE_USE;
4490 if (PL_expect != XSTATE)
4491 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4492 is_use ? "use" : "no"));
4495 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4496 s = force_version(s, TRUE);
4497 if (*s == ';' || *s == '}'
4498 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4499 start_force(PL_curforce);
4500 NEXTVAL_NEXTTOKE.opval = NULL;
4503 else if (*s == 'v') {
4504 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4505 s = force_version(s, FALSE);
4509 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4510 s = force_version(s, FALSE);
4512 pl_yylval.ival = is_use;
4516 static const char* const exp_name[] =
4517 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4518 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4522 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4524 S_word_takes_any_delimeter(char *p, STRLEN len)
4526 return (len == 1 && strchr("msyq", p[0])) ||
4528 (p[0] == 't' && p[1] == 'r') ||
4529 (p[0] == 'q' && strchr("qwxr", p[1]))));
4535 Works out what to call the token just pulled out of the input
4536 stream. The yacc parser takes care of taking the ops we return and
4537 stitching them into a tree.
4540 The type of the next token
4543 Switch based on the current state:
4544 - if we already built the token before, use it
4545 - if we have a case modifier in a string, deal with that
4546 - handle other cases of interpolation inside a string
4547 - scan the next line if we are inside a format
4548 In the normal state switch on the next character:
4550 if alphabetic, go to key lookup
4551 unrecoginized character - croak
4552 - 0/4/26: handle end-of-line or EOF
4553 - cases for whitespace
4554 - \n and #: handle comments and line numbers
4555 - various operators, brackets and sigils
4558 - 'v': vstrings (or go to key lookup)
4559 - 'x' repetition operator (or go to key lookup)
4560 - other ASCII alphanumerics (key lookup begins here):
4563 scan built-in keyword (but do nothing with it yet)
4564 check for statement label
4565 check for lexical subs
4566 goto just_a_word if there is one
4567 see whether built-in keyword is overridden
4568 switch on keyword number:
4569 - default: just_a_word:
4570 not a built-in keyword; handle bareword lookup
4571 disambiguate between method and sub call
4572 fall back to bareword
4573 - cases for built-in keywords
4578 #pragma segment Perl_yylex
4584 char *s = PL_bufptr;
4591 /* orig_keyword, gvp, and gv are initialized here because
4592 * jump to the label just_a_word_zero can bypass their
4593 * initialization later. */
4594 I32 orig_keyword = 0;
4599 SV* tmp = newSVpvs("");
4600 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4601 (IV)CopLINE(PL_curcop),
4602 lex_state_names[PL_lex_state],
4603 exp_name[PL_expect],
4604 pv_display(tmp, s, strlen(s), 0, 60));
4608 switch (PL_lex_state) {
4610 case LEX_NORMAL: /* Some compilers will produce faster */
4611 case LEX_INTERPNORMAL: /* code if we comment these out. */
4615 /* when we've already built the next token, just pull it out of the queue */
4619 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4621 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4622 PL_nexttoke[PL_lasttoke].next_mad = 0;
4623 if (PL_thismad && PL_thismad->mad_key == '_') {
4624 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4625 PL_thismad->mad_val = 0;
4626 mad_free(PL_thismad);
4631 PL_lex_state = PL_lex_defer;
4632 PL_expect = PL_lex_expect;
4633 PL_lex_defer = LEX_NORMAL;
4634 if (!PL_nexttoke[PL_lasttoke].next_type)
4639 pl_yylval = PL_nextval[PL_nexttoke];
4641 PL_lex_state = PL_lex_defer;
4642 PL_expect = PL_lex_expect;
4643 PL_lex_defer = LEX_NORMAL;
4649 next_type = PL_nexttoke[PL_lasttoke].next_type;
4651 next_type = PL_nexttype[PL_nexttoke];
4653 if (next_type & (7<<24)) {
4654 if (next_type & (1<<24)) {
4655 if (PL_lex_brackets > 100)
4656 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4657 PL_lex_brackstack[PL_lex_brackets++] =
4658 (char) ((next_type >> 16) & 0xff);
4660 if (next_type & (2<<24))
4661 PL_lex_allbrackets++;
4662 if (next_type & (4<<24))
4663 PL_lex_allbrackets--;
4664 next_type &= 0xffff;
4666 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4669 /* interpolated case modifiers like \L \U, including \Q and \E.
4670 when we get here, PL_bufptr is at the \
4672 case LEX_INTERPCASEMOD:
4674 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4676 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4677 PL_bufptr, PL_bufend, *PL_bufptr);
4679 /* handle \E or end of string */
4680 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4682 if (PL_lex_casemods) {
4683 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4684 PL_lex_casestack[PL_lex_casemods] = '\0';
4686 if (PL_bufptr != PL_bufend
4687 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4688 || oldmod == 'F')) {
4690 PL_lex_state = LEX_INTERPCONCAT;
4693 PL_thistoken = newSVpvs("\\E");
4696 PL_lex_allbrackets--;
4699 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4700 /* Got an unpaired \E */
4701 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4702 "Useless use of \\E");
4705 while (PL_bufptr != PL_bufend &&
4706 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4709 PL_thiswhite = newSVpvs("");
4710 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4715 if (PL_bufptr != PL_bufend)
4718 PL_lex_state = LEX_INTERPCONCAT;
4722 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4723 "### Saw case modifier\n"); });
4725 if (s[1] == '\\' && s[2] == 'E') {
4729 PL_thiswhite = newSVpvs("");
4730 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4734 PL_lex_state = LEX_INTERPCONCAT;
4739 if (!PL_madskills) /* when just compiling don't need correct */
4740 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4741 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4742 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4743 (strchr(PL_lex_casestack, 'L')
4744 || strchr(PL_lex_casestack, 'U')
4745 || strchr(PL_lex_casestack, 'F'))) {
4746 PL_lex_casestack[--PL_lex_casemods] = '\0';
4747 PL_lex_allbrackets--;
4750 if (PL_lex_casemods > 10)
4751 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4752 PL_lex_casestack[PL_lex_casemods++] = *s;
4753 PL_lex_casestack[PL_lex_casemods] = '\0';
4754 PL_lex_state = LEX_INTERPCONCAT;
4755 start_force(PL_curforce);
4756 NEXTVAL_NEXTTOKE.ival = 0;
4757 force_next((2<<24)|'(');
4758 start_force(PL_curforce);
4760 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4762 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4764 NEXTVAL_NEXTTOKE.ival = OP_LC;
4766 NEXTVAL_NEXTTOKE.ival = OP_UC;
4768 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4770 NEXTVAL_NEXTTOKE.ival = OP_FC;
4772 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4774 SV* const tmpsv = newSVpvs("\\ ");
4775 /* replace the space with the character we want to escape
4777 SvPVX(tmpsv)[1] = *s;
4783 if (PL_lex_starts) {
4789 sv_free(PL_thistoken);
4790 PL_thistoken = newSVpvs("");
4793 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4794 if (PL_lex_casemods == 1 && PL_lex_inpat)
4803 case LEX_INTERPPUSH:
4804 return REPORT(sublex_push());
4806 case LEX_INTERPSTART:
4807 if (PL_bufptr == PL_bufend)
4808 return REPORT(sublex_done());
4809 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4810 "### Interpolated variable\n"); });
4812 PL_lex_dojoin = (*PL_bufptr == '@');
4813 PL_lex_state = LEX_INTERPNORMAL;
4814 if (PL_lex_dojoin) {
4815 start_force(PL_curforce);
4816 NEXTVAL_NEXTTOKE.ival = 0;
4818 start_force(PL_curforce);
4819 force_ident("\"", '$');
4820 start_force(PL_curforce);
4821 NEXTVAL_NEXTTOKE.ival = 0;
4823 start_force(PL_curforce);
4824 NEXTVAL_NEXTTOKE.ival = 0;
4825 force_next((2<<24)|'(');
4826 start_force(PL_curforce);
4827 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4830 /* Convert (?{...}) and friends to 'do {...}' */
4831 if (PL_lex_inpat && *PL_bufptr == '(') {
4832 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4834 if (*PL_bufptr != '{')
4836 start_force(PL_curforce);
4837 /* XXX probably need a CURMAD(something) here */
4838 PL_expect = XTERMBLOCK;
4842 if (PL_lex_starts++) {
4847 sv_free(PL_thistoken);
4848 PL_thistoken = newSVpvs("");
4851 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4852 if (!PL_lex_casemods && PL_lex_inpat)
4859 case LEX_INTERPENDMAYBE:
4860 if (intuit_more(PL_bufptr)) {
4861 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4867 if (PL_lex_dojoin) {
4868 PL_lex_dojoin = FALSE;
4869 PL_lex_state = LEX_INTERPCONCAT;
4873 sv_free(PL_thistoken);
4874 PL_thistoken = newSVpvs("");
4877 PL_lex_allbrackets--;
4880 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4881 && SvEVALED(PL_lex_repl))
4883 if (PL_bufptr != PL_bufend)
4884 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4887 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4888 re_eval_str. If the here-doc body’s length equals the previous
4889 value of re_eval_start, re_eval_start will now be null. So
4890 check re_eval_str as well. */
4891 if (PL_parser->lex_shared->re_eval_start
4892 || PL_parser->lex_shared->re_eval_str) {
4894 if (*PL_bufptr != ')')
4895 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4897 /* having compiled a (?{..}) expression, return the original
4898 * text too, as a const */
4899 if (PL_parser->lex_shared->re_eval_str) {
4900 sv = PL_parser->lex_shared->re_eval_str;
4901 PL_parser->lex_shared->re_eval_str = NULL;
4903 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4904 SvPV_shrink_to_cur(sv);
4906 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4907 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4908 start_force(PL_curforce);
4909 /* XXX probably need a CURMAD(something) here */
4910 NEXTVAL_NEXTTOKE.opval =
4911 (OP*)newSVOP(OP_CONST, 0,
4914 PL_parser->lex_shared->re_eval_start = NULL;
4920 case LEX_INTERPCONCAT:
4922 if (PL_lex_brackets)
4923 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4924 (long) PL_lex_brackets);
4926 if (PL_bufptr == PL_bufend)
4927 return REPORT(sublex_done());
4929 /* m'foo' still needs to be parsed for possible (?{...}) */
4930 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4931 SV *sv = newSVsv(PL_linestr);
4933 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4937 s = scan_const(PL_bufptr);
4939 PL_lex_state = LEX_INTERPCASEMOD;
4941 PL_lex_state = LEX_INTERPSTART;
4944 if (s != PL_bufptr) {
4945 start_force(PL_curforce);
4947 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4949 NEXTVAL_NEXTTOKE = pl_yylval;
4952 if (PL_lex_starts++) {
4956 sv_free(PL_thistoken);
4957 PL_thistoken = newSVpvs("");
4960 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4961 if (!PL_lex_casemods && PL_lex_inpat)
4974 s = scan_formline(PL_bufptr);
4975 if (!PL_lex_formbrack)
4985 PL_oldoldbufptr = PL_oldbufptr;
4991 sv_free(PL_thistoken);
4994 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4998 if (isIDFIRST_lazy_if(s,UTF))
5001 SV *dsv = newSVpvs_flags("", SVs_TEMP);
5002 const char *c = UTF ? savepv(sv_uni_display(dsv, newSVpvn_flags(s,
5004 SVs_TEMP | SVf_UTF8),
5005 10, UNI_DISPLAY_ISPRINT))
5006 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
5007 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
5008 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
5009 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
5017 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"SVf"<-- HERE near column %d", c, SVfARG(dsv), (int) len + 1);
5021 goto fake_eof; /* emulate EOF on ^D or ^Z */
5027 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
5030 if (PL_lex_brackets &&
5031 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
5032 yyerror((const char *)
5034 ? "Format not terminated"
5035 : "Missing right curly or square bracket"));
5037 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5038 "### Tokener got EOF\n");
5042 if (s++ < PL_bufend)
5043 goto retry; /* ignore stray nulls */
5046 if (!PL_in_eval && !PL_preambled) {
5047 PL_preambled = TRUE;
5053 /* Generate a string of Perl code to load the debugger.
5054 * If PERL5DB is set, it will return the contents of that,
5055 * otherwise a compile-time require of perl5db.pl. */
5057 const char * const pdb = PerlEnv_getenv("PERL5DB");
5060 sv_setpv(PL_linestr, pdb);
5061 sv_catpvs(PL_linestr,";");
5063 SETERRNO(0,SS_NORMAL);
5064 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
5067 sv_setpvs(PL_linestr,"");
5068 if (PL_preambleav) {
5069 SV **svp = AvARRAY(PL_preambleav);
5070 SV **const end = svp + AvFILLp(PL_preambleav);
5072 sv_catsv(PL_linestr, *svp);
5074 sv_catpvs(PL_linestr, ";");
5076 sv_free(MUTABLE_SV(PL_preambleav));
5077 PL_preambleav = NULL;
5080 sv_catpvs(PL_linestr,
5081 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
5082 if (PL_minus_n || PL_minus_p) {
5083 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
5085 sv_catpvs(PL_linestr,"chomp;");
5088 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
5089 || *PL_splitstr == '"')
5090 && strchr(PL_splitstr + 1, *PL_splitstr))
5091 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
5093 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
5094 bytes can be used as quoting characters. :-) */
5095 const char *splits = PL_splitstr;
5096 sv_catpvs(PL_linestr, "our @F=split(q\0");
5099 if (*splits == '\\')
5100 sv_catpvn(PL_linestr, splits, 1);
5101 sv_catpvn(PL_linestr, splits, 1);
5102 } while (*splits++);
5103 /* This loop will embed the trailing NUL of
5104 PL_linestr as the last thing it does before
5106 sv_catpvs(PL_linestr, ");");
5110 sv_catpvs(PL_linestr,"our @F=split(' ');");
5113 sv_catpvs(PL_linestr, "\n");
5114 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5115 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5116 PL_last_lop = PL_last_uni = NULL;
5117 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
5118 update_debugger_info(PL_linestr, NULL, 0);
5123 bof = PL_rsfp ? TRUE : FALSE;
5126 fake_eof = LEX_FAKE_EOF;
5128 PL_bufptr = PL_bufend;
5129 COPLINE_INC_WITH_HERELINES;
5130 if (!lex_next_chunk(fake_eof)) {
5131 CopLINE_dec(PL_curcop);
5133 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
5135 CopLINE_dec(PL_curcop);
5138 PL_realtokenstart = -1;
5141 /* If it looks like the start of a BOM or raw UTF-16,
5142 * check if it in fact is. */
5143 if (bof && PL_rsfp &&
5148 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
5149 bof = (offset == (Off_t)SvCUR(PL_linestr));
5150 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
5151 /* offset may include swallowed CR */
5153 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
5156 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5157 s = swallow_bom((U8*)s);
5160 if (PL_parser->in_pod) {
5161 /* Incest with pod. */
5164 sv_catsv(PL_thiswhite, PL_linestr);
5166 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
5167 sv_setpvs(PL_linestr, "");
5168 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5169 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5170 PL_last_lop = PL_last_uni = NULL;
5171 PL_parser->in_pod = 0;
5174 if (PL_rsfp || PL_parser->filtered)
5176 } while (PL_parser->in_pod);
5177 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
5178 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5179 PL_last_lop = PL_last_uni = NULL;
5180 if (CopLINE(PL_curcop) == 1) {
5181 while (s < PL_bufend && isSPACE(*s))
5183 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
5187 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
5191 if (*s == '#' && *(s+1) == '!')
5193 #ifdef ALTERNATE_SHEBANG
5195 static char const as[] = ALTERNATE_SHEBANG;
5196 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5197 d = s + (sizeof(as) - 1);
5199 #endif /* ALTERNATE_SHEBANG */
5208 while (*d && !isSPACE(*d))
5212 #ifdef ARG_ZERO_IS_SCRIPT
5213 if (ipathend > ipath) {
5215 * HP-UX (at least) sets argv[0] to the script name,
5216 * which makes $^X incorrect. And Digital UNIX and Linux,
5217 * at least, set argv[0] to the basename of the Perl
5218 * interpreter. So, having found "#!", we'll set it right.
5220 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5222 assert(SvPOK(x) || SvGMAGICAL(x));
5223 if (sv_eq(x, CopFILESV(PL_curcop))) {
5224 sv_setpvn(x, ipath, ipathend - ipath);
5230 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
5231 const char * const lstart = SvPV_const(x,llen);
5233 bstart += blen - llen;
5234 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5235 sv_setpvn(x, ipath, ipathend - ipath);
5240 TAINT_NOT; /* $^X is always tainted, but that's OK */
5242 #endif /* ARG_ZERO_IS_SCRIPT */
5247 d = instr(s,"perl -");
5249 d = instr(s,"perl");
5251 /* avoid getting into infinite loops when shebang
5252 * line contains "Perl" rather than "perl" */
5254 for (d = ipathend-4; d >= ipath; --d) {
5255 if ((*d == 'p' || *d == 'P')
5256 && !ibcmp(d, "perl", 4))
5266 #ifdef ALTERNATE_SHEBANG
5268 * If the ALTERNATE_SHEBANG on this system starts with a
5269 * character that can be part of a Perl expression, then if
5270 * we see it but not "perl", we're probably looking at the
5271 * start of Perl code, not a request to hand off to some
5272 * other interpreter. Similarly, if "perl" is there, but
5273 * not in the first 'word' of the line, we assume the line
5274 * contains the start of the Perl program.
5276 if (d && *s != '#') {
5277 const char *c = ipath;
5278 while (*c && !strchr("; \t\r\n\f\v#", *c))
5281 d = NULL; /* "perl" not in first word; ignore */
5283 *s = '#'; /* Don't try to parse shebang line */
5285 #endif /* ALTERNATE_SHEBANG */
5290 !instr(s,"indir") &&
5291 instr(PL_origargv[0],"perl"))
5298 while (s < PL_bufend && isSPACE(*s))
5300 if (s < PL_bufend) {
5301 Newx(newargv,PL_origargc+3,char*);
5303 while (s < PL_bufend && !isSPACE(*s))
5306 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5309 newargv = PL_origargv;
5312 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5314 Perl_croak(aTHX_ "Can't exec %s", ipath);
5317 while (*d && !isSPACE(*d))
5319 while (SPACE_OR_TAB(*d))
5323 const bool switches_done = PL_doswitches;
5324 const U32 oldpdb = PL_perldb;
5325 const bool oldn = PL_minus_n;
5326 const bool oldp = PL_minus_p;
5330 bool baduni = FALSE;
5332 const char *d2 = d1 + 1;
5333 if (parse_unicode_opts((const char **)&d2)
5337 if (baduni || *d1 == 'M' || *d1 == 'm') {
5338 const char * const m = d1;
5339 while (*d1 && !isSPACE(*d1))
5341 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5344 d1 = moreswitches(d1);
5346 if (PL_doswitches && !switches_done) {
5347 int argc = PL_origargc;
5348 char **argv = PL_origargv;
5351 } while (argc && argv[0][0] == '-' && argv[0][1]);
5352 init_argv_symbols(argc,argv);
5354 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5355 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5356 /* if we have already added "LINE: while (<>) {",
5357 we must not do it again */
5359 sv_setpvs(PL_linestr, "");
5360 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5361 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5362 PL_last_lop = PL_last_uni = NULL;
5363 PL_preambled = FALSE;
5364 if (PERLDB_LINE || PERLDB_SAVESRC)
5365 (void)gv_fetchfile(PL_origfilename);
5372 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5373 PL_lex_state = LEX_FORMLINE;
5374 start_force(PL_curforce);
5375 NEXTVAL_NEXTTOKE.ival = 0;
5376 force_next(FORMRBRACK);
5381 #ifdef PERL_STRICT_CR
5382 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5384 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5386 case ' ': case '\t': case '\f': case 013:
5388 PL_realtokenstart = -1;
5391 PL_thiswhite = newSVpvs("");
5392 sv_catpvn(PL_thiswhite, s, 1);
5400 PL_realtokenstart = -1;
5404 if (PL_lex_state != LEX_NORMAL ||
5405 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5406 if (*s == '#' && s == PL_linestart && PL_in_eval
5407 && !PL_rsfp && !PL_parser->filtered) {
5408 /* handle eval qq[#line 1 "foo"\n ...] */
5409 CopLINE_dec(PL_curcop);
5412 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5414 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5418 const bool in_comment = *s == '#';
5420 while (d < PL_bufend && *d != '\n')
5424 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5425 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5429 PL_thiswhite = newSVpvn(s, d - s);
5432 if (in_comment && d == PL_bufend
5433 && PL_lex_state == LEX_INTERPNORMAL
5434 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5435 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5438 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5439 PL_lex_state = LEX_FORMLINE;
5440 start_force(PL_curforce);
5441 NEXTVAL_NEXTTOKE.ival = 0;
5442 force_next(FORMRBRACK);
5448 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5449 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5452 TOKEN(PEG); /* make sure any #! line is accessible */
5457 /* if (PL_madskills && PL_lex_formbrack) { */
5459 while (d < PL_bufend && *d != '\n')
5463 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5464 Perl_croak(aTHX_ "panic: input overflow");
5465 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5467 PL_thiswhite = newSVpvs("");
5468 if (CopLINE(PL_curcop) == 1) {
5469 sv_setpvs(PL_thiswhite, "");
5472 sv_catpvn(PL_thiswhite, s, d - s);
5486 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5494 while (s < PL_bufend && SPACE_OR_TAB(*s))
5497 if (strnEQ(s,"=>",2)) {
5498 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5499 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5500 OPERATOR('-'); /* unary minus */
5502 PL_last_uni = PL_oldbufptr;
5504 case 'r': ftst = OP_FTEREAD; break;
5505 case 'w': ftst = OP_FTEWRITE; break;
5506 case 'x': ftst = OP_FTEEXEC; break;
5507 case 'o': ftst = OP_FTEOWNED; break;
5508 case 'R': ftst = OP_FTRREAD; break;
5509 case 'W': ftst = OP_FTRWRITE; break;
5510 case 'X': ftst = OP_FTREXEC; break;
5511 case 'O': ftst = OP_FTROWNED; break;
5512 case 'e': ftst = OP_FTIS; break;
5513 case 'z': ftst = OP_FTZERO; break;
5514 case 's': ftst = OP_FTSIZE; break;
5515 case 'f': ftst = OP_FTFILE; break;
5516 case 'd': ftst = OP_FTDIR; break;
5517 case 'l': ftst = OP_FTLINK; break;
5518 case 'p': ftst = OP_FTPIPE; break;
5519 case 'S': ftst = OP_FTSOCK; break;
5520 case 'u': ftst = OP_FTSUID; break;
5521 case 'g': ftst = OP_FTSGID; break;
5522 case 'k': ftst = OP_FTSVTX; break;
5523 case 'b': ftst = OP_FTBLK; break;
5524 case 'c': ftst = OP_FTCHR; break;
5525 case 't': ftst = OP_FTTTY; break;
5526 case 'T': ftst = OP_FTTEXT; break;
5527 case 'B': ftst = OP_FTBINARY; break;
5528 case 'M': case 'A': case 'C':
5529 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5531 case 'M': ftst = OP_FTMTIME; break;
5532 case 'A': ftst = OP_FTATIME; break;
5533 case 'C': ftst = OP_FTCTIME; break;
5541 PL_last_lop_op = (OPCODE)ftst;
5542 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5543 "### Saw file test %c\n", (int)tmp);
5548 /* Assume it was a minus followed by a one-letter named
5549 * subroutine call (or a -bareword), then. */
5550 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5551 "### '-%c' looked like a file test but was not\n",
5558 const char tmp = *s++;
5561 if (PL_expect == XOPERATOR)
5566 else if (*s == '>') {
5569 if (isIDFIRST_lazy_if(s,UTF)) {
5570 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5578 if (PL_expect == XOPERATOR) {
5579 if (*s == '=' && !PL_lex_allbrackets &&
5580 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5587 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5589 OPERATOR('-'); /* unary minus */
5595 const char tmp = *s++;
5598 if (PL_expect == XOPERATOR)
5603 if (PL_expect == XOPERATOR) {
5604 if (*s == '=' && !PL_lex_allbrackets &&
5605 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5612 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5619 if (PL_expect != XOPERATOR) {
5620 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5621 PL_expect = XOPERATOR;
5622 force_ident(PL_tokenbuf, '*');
5630 if (*s == '=' && !PL_lex_allbrackets &&
5631 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5637 if (*s == '=' && !PL_lex_allbrackets &&
5638 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5645 if (PL_expect == XOPERATOR) {
5646 if (s[1] == '=' && !PL_lex_allbrackets &&
5647 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5652 PL_tokenbuf[0] = '%';
5653 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5654 sizeof PL_tokenbuf - 1, FALSE);
5655 if (!PL_tokenbuf[1]) {
5658 PL_expect = XOPERATOR;
5659 force_ident_maybe_lex('%');
5663 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5664 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5669 if (PL_lex_brackets > 100)
5670 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5671 PL_lex_brackstack[PL_lex_brackets++] = 0;
5672 PL_lex_allbrackets++;
5674 const char tmp = *s++;
5679 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5681 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5689 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5696 goto just_a_word_zero_gv;
5699 switch (PL_expect) {
5705 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5707 PL_bufptr = s; /* update in case we back off */
5710 "Use of := for an empty attribute list is not allowed");
5717 PL_expect = XTERMBLOCK;
5720 stuffstart = s - SvPVX(PL_linestr) - 1;
5724 while (isIDFIRST_lazy_if(s,UTF)) {
5727 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5728 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5729 if (tmp < 0) tmp = -tmp;
5744 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5746 d = scan_str(d,TRUE,TRUE,FALSE);
5748 /* MUST advance bufptr here to avoid bogus
5749 "at end of line" context messages from yyerror().
5751 PL_bufptr = s + len;
5752 yyerror("Unterminated attribute parameter in attribute list");
5756 return REPORT(0); /* EOF indicator */
5760 sv_catsv(sv, PL_lex_stuff);
5761 attrs = op_append_elem(OP_LIST, attrs,
5762 newSVOP(OP_CONST, 0, sv));
5763 SvREFCNT_dec(PL_lex_stuff);
5764 PL_lex_stuff = NULL;
5767 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5769 if (PL_in_my == KEY_our) {
5770 deprecate(":unique");
5773 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5776 /* NOTE: any CV attrs applied here need to be part of
5777 the CVf_BUILTIN_ATTRS define in cv.h! */
5778 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5780 CvLVALUE_on(PL_compcv);
5782 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5784 deprecate(":locked");
5786 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5788 CvMETHOD_on(PL_compcv);
5790 /* After we've set the flags, it could be argued that
5791 we don't need to do the attributes.pm-based setting
5792 process, and shouldn't bother appending recognized
5793 flags. To experiment with that, uncomment the
5794 following "else". (Note that's already been
5795 uncommented. That keeps the above-applied built-in
5796 attributes from being intercepted (and possibly
5797 rejected) by a package's attribute routines, but is
5798 justified by the performance win for the common case
5799 of applying only built-in attributes.) */
5801 attrs = op_append_elem(OP_LIST, attrs,
5802 newSVOP(OP_CONST, 0,
5806 if (*s == ':' && s[1] != ':')
5809 break; /* require real whitespace or :'s */
5810 /* XXX losing whitespace on sequential attributes here */
5814 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5815 if (*s != ';' && *s != '}' && *s != tmp
5816 && (tmp != '=' || *s != ')')) {
5817 const char q = ((*s == '\'') ? '"' : '\'');
5818 /* If here for an expression, and parsed no attrs, back
5820 if (tmp == '=' && !attrs) {
5824 /* MUST advance bufptr here to avoid bogus "at end of line"
5825 context messages from yyerror().
5828 yyerror( (const char *)
5830 ? Perl_form(aTHX_ "Invalid separator character "
5831 "%c%c%c in attribute list", q, *s, q)
5832 : "Unterminated attribute list" ) );
5840 start_force(PL_curforce);
5841 NEXTVAL_NEXTTOKE.opval = attrs;
5842 CURMAD('_', PL_nextwhite);
5847 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5848 (s - SvPVX(PL_linestr)) - stuffstart);
5853 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5857 PL_lex_allbrackets--;
5861 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5862 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5866 PL_lex_allbrackets++;
5869 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5875 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5878 PL_lex_allbrackets--;
5884 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5887 if (PL_lex_brackets <= 0)
5888 yyerror("Unmatched right square bracket");
5891 PL_lex_allbrackets--;
5892 if (PL_lex_state == LEX_INTERPNORMAL) {
5893 if (PL_lex_brackets == 0) {
5894 if (*s == '-' && s[1] == '>')
5895 PL_lex_state = LEX_INTERPENDMAYBE;
5896 else if (*s != '[' && *s != '{')
5897 PL_lex_state = LEX_INTERPEND;
5904 if (PL_lex_brackets > 100) {
5905 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5907 switch (PL_expect) {
5909 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5910 PL_lex_allbrackets++;
5911 OPERATOR(HASHBRACK);
5913 while (s < PL_bufend && SPACE_OR_TAB(*s))
5916 PL_tokenbuf[0] = '\0';
5917 if (d < PL_bufend && *d == '-') {
5918 PL_tokenbuf[0] = '-';
5920 while (d < PL_bufend && SPACE_OR_TAB(*d))
5923 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5924 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5926 while (d < PL_bufend && SPACE_OR_TAB(*d))
5929 const char minus = (PL_tokenbuf[0] == '-');
5930 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5938 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5939 PL_lex_allbrackets++;
5944 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5945 PL_lex_allbrackets++;
5950 if (PL_oldoldbufptr == PL_last_lop)
5951 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5953 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5954 PL_lex_allbrackets++;
5957 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5959 /* This hack is to get the ${} in the message. */
5961 yyerror("syntax error");
5964 OPERATOR(HASHBRACK);
5966 /* This hack serves to disambiguate a pair of curlies
5967 * as being a block or an anon hash. Normally, expectation
5968 * determines that, but in cases where we're not in a
5969 * position to expect anything in particular (like inside
5970 * eval"") we have to resolve the ambiguity. This code
5971 * covers the case where the first term in the curlies is a
5972 * quoted string. Most other cases need to be explicitly
5973 * disambiguated by prepending a "+" before the opening
5974 * curly in order to force resolution as an anon hash.
5976 * XXX should probably propagate the outer expectation
5977 * into eval"" to rely less on this hack, but that could
5978 * potentially break current behavior of eval"".
5982 if (*s == '\'' || *s == '"' || *s == '`') {
5983 /* common case: get past first string, handling escapes */
5984 for (t++; t < PL_bufend && *t != *s;)
5985 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5989 else if (*s == 'q') {
5992 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5995 /* skip q//-like construct */
5997 char open, close, term;
6000 while (t < PL_bufend && isSPACE(*t))
6002 /* check for q => */
6003 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
6004 OPERATOR(HASHBRACK);
6008 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
6012 for (t++; t < PL_bufend; t++) {
6013 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
6015 else if (*t == open)
6019 for (t++; t < PL_bufend; t++) {
6020 if (*t == '\\' && t+1 < PL_bufend)
6022 else if (*t == close && --brackets <= 0)
6024 else if (*t == open)
6031 /* skip plain q word */
6032 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
6035 else if (isALNUM_lazy_if(t,UTF)) {
6037 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
6040 while (t < PL_bufend && isSPACE(*t))
6042 /* if comma follows first term, call it an anon hash */
6043 /* XXX it could be a comma expression with loop modifiers */
6044 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
6045 || (*t == '=' && t[1] == '>')))
6046 OPERATOR(HASHBRACK);
6047 if (PL_expect == XREF)
6050 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
6056 pl_yylval.ival = CopLINE(PL_curcop);
6057 if (isSPACE(*s) || *s == '#')
6058 PL_copline = NOLINE; /* invalidate current command line number */
6059 TOKEN(formbrack ? '=' : '{');
6061 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
6065 if (PL_lex_brackets <= 0)
6066 yyerror("Unmatched right curly bracket");
6068 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
6069 PL_lex_allbrackets--;
6070 if (PL_lex_state == LEX_INTERPNORMAL) {
6071 if (PL_lex_brackets == 0) {
6072 if (PL_expect & XFAKEBRACK) {
6073 PL_expect &= XENUMMASK;
6074 PL_lex_state = LEX_INTERPEND;
6079 PL_thiswhite = newSVpvs("");
6080 sv_catpvs(PL_thiswhite,"}");
6083 return yylex(); /* ignore fake brackets */
6085 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
6086 && SvEVALED(PL_lex_repl))
6087 PL_lex_state = LEX_INTERPEND;
6088 else if (*s == '-' && s[1] == '>')
6089 PL_lex_state = LEX_INTERPENDMAYBE;
6090 else if (*s != '[' && *s != '{')
6091 PL_lex_state = LEX_INTERPEND;
6094 if (PL_expect & XFAKEBRACK) {
6095 PL_expect &= XENUMMASK;
6097 return yylex(); /* ignore fake brackets */
6099 start_force(PL_curforce);
6101 curmad('X', newSVpvn(s-1,1));
6102 CURMAD('_', PL_thiswhite);
6104 force_next(formbrack ? '.' : '}');
6105 if (formbrack) LEAVE;
6107 if (PL_madskills && !PL_thistoken)
6108 PL_thistoken = newSVpvs("");
6110 if (formbrack == 2) { /* means . where arguments were expected */
6111 start_force(PL_curforce);
6119 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6120 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6127 if (PL_expect == XOPERATOR) {
6128 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
6129 && isIDFIRST_lazy_if(s,UTF))
6131 CopLINE_dec(PL_curcop);
6132 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6133 CopLINE_inc(PL_curcop);
6135 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6136 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6143 PL_tokenbuf[0] = '&';
6144 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf + 1,
6145 sizeof PL_tokenbuf - 1, TRUE);
6146 if (PL_tokenbuf[1]) {
6147 PL_expect = XOPERATOR;
6148 force_ident_maybe_lex('&');
6152 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
6158 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6159 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6166 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6167 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6175 const char tmp = *s++;
6177 if (!PL_lex_allbrackets &&
6178 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6185 if (!PL_lex_allbrackets &&
6186 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
6194 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6195 && strchr("+-*/%.^&|<",tmp))
6196 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6197 "Reversed %c= operator",(int)tmp);
6199 if (PL_expect == XSTATE && isALPHA(tmp) &&
6200 (s == PL_linestart+1 || s[-2] == '\n') )
6202 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6203 || PL_lex_state != LEX_NORMAL) {
6208 if (strnEQ(s,"=cut",4)) {
6224 PL_thiswhite = newSVpvs("");
6225 sv_catpvn(PL_thiswhite, PL_linestart,
6226 PL_bufend - PL_linestart);
6230 PL_parser->in_pod = 1;
6234 if (PL_expect == XBLOCK) {
6236 #ifdef PERL_STRICT_CR
6237 while (SPACE_OR_TAB(*t))
6239 while (SPACE_OR_TAB(*t) || *t == '\r')
6242 if (*t == '\n' || *t == '#') {
6245 SAVEI8(PL_parser->form_lex_state);
6246 SAVEI32(PL_lex_formbrack);
6247 PL_parser->form_lex_state = PL_lex_state;
6248 PL_lex_formbrack = PL_lex_brackets + 1;
6252 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6261 const char tmp = *s++;
6263 /* was this !=~ where !~ was meant?
6264 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6266 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6267 const char *t = s+1;
6269 while (t < PL_bufend && isSPACE(*t))
6272 if (*t == '/' || *t == '?' ||
6273 ((*t == 'm' || *t == 's' || *t == 'y')
6274 && !isALNUM(t[1])) ||
6275 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
6276 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6277 "!=~ should be !~");
6279 if (!PL_lex_allbrackets &&
6280 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6292 if (PL_expect != XOPERATOR) {
6293 if (s[1] != '<' && !strchr(s,'>'))
6296 s = scan_heredoc(s);
6298 s = scan_inputsymbol(s);
6299 PL_expect = XOPERATOR;
6300 TOKEN(sublex_start());
6306 if (*s == '=' && !PL_lex_allbrackets &&
6307 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6311 SHop(OP_LEFT_SHIFT);
6316 if (!PL_lex_allbrackets &&
6317 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6324 if (!PL_lex_allbrackets &&
6325 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6333 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6341 const char tmp = *s++;
6343 if (*s == '=' && !PL_lex_allbrackets &&
6344 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6348 SHop(OP_RIGHT_SHIFT);
6350 else if (tmp == '=') {
6351 if (!PL_lex_allbrackets &&
6352 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6360 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6369 if (PL_expect == XOPERATOR) {
6370 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6371 return deprecate_commaless_var_list();
6375 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6376 PL_tokenbuf[0] = '@';
6377 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6378 sizeof PL_tokenbuf - 1, FALSE);
6379 if (PL_expect == XOPERATOR)
6380 no_op("Array length", s);
6381 if (!PL_tokenbuf[1])
6383 PL_expect = XOPERATOR;
6384 force_ident_maybe_lex('#');
6388 PL_tokenbuf[0] = '$';
6389 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6390 sizeof PL_tokenbuf - 1, FALSE);
6391 if (PL_expect == XOPERATOR)
6393 if (!PL_tokenbuf[1]) {
6395 yyerror("Final $ should be \\$ or $name");
6401 const char tmp = *s;
6402 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6405 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6406 && intuit_more(s)) {
6408 PL_tokenbuf[0] = '@';
6409 if (ckWARN(WARN_SYNTAX)) {
6412 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6415 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6416 while (t < PL_bufend && *t != ']')
6418 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6419 "Multidimensional syntax %.*s not supported",
6420 (int)((t - PL_bufptr) + 1), PL_bufptr);
6424 else if (*s == '{') {
6426 PL_tokenbuf[0] = '%';
6427 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6428 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6430 char tmpbuf[sizeof PL_tokenbuf];
6433 } while (isSPACE(*t));
6434 if (isIDFIRST_lazy_if(t,UTF)) {
6436 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6441 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6442 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6443 "You need to quote \"%"SVf"\"",
6444 SVfARG(newSVpvn_flags(tmpbuf, len,
6445 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
6451 PL_expect = XOPERATOR;
6452 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6453 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6454 if (!islop || PL_last_lop_op == OP_GREPSTART)
6455 PL_expect = XOPERATOR;
6456 else if (strchr("$@\"'`q", *s))
6457 PL_expect = XTERM; /* e.g. print $fh "foo" */
6458 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6459 PL_expect = XTERM; /* e.g. print $fh &sub */
6460 else if (isIDFIRST_lazy_if(s,UTF)) {
6461 char tmpbuf[sizeof PL_tokenbuf];
6463 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6464 if ((t2 = keyword(tmpbuf, len, 0))) {
6465 /* binary operators exclude handle interpretations */
6477 PL_expect = XTERM; /* e.g. print $fh length() */
6482 PL_expect = XTERM; /* e.g. print $fh subr() */
6485 else if (isDIGIT(*s))
6486 PL_expect = XTERM; /* e.g. print $fh 3 */
6487 else if (*s == '.' && isDIGIT(s[1]))
6488 PL_expect = XTERM; /* e.g. print $fh .3 */
6489 else if ((*s == '?' || *s == '-' || *s == '+')
6490 && !isSPACE(s[1]) && s[1] != '=')
6491 PL_expect = XTERM; /* e.g. print $fh -1 */
6492 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6494 PL_expect = XTERM; /* e.g. print $fh /.../
6495 XXX except DORDOR operator
6497 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6499 PL_expect = XTERM; /* print $fh <<"EOF" */
6502 force_ident_maybe_lex('$');
6506 if (PL_expect == XOPERATOR)
6508 PL_tokenbuf[0] = '@';
6509 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6510 if (!PL_tokenbuf[1]) {
6513 if (PL_lex_state == LEX_NORMAL)
6515 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6517 PL_tokenbuf[0] = '%';
6519 /* Warn about @ where they meant $. */
6520 if (*s == '[' || *s == '{') {
6521 if (ckWARN(WARN_SYNTAX)) {
6522 const char *t = s + 1;
6523 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6524 t += UTF ? UTF8SKIP(t) : 1;
6525 if (*t == '}' || *t == ']') {
6527 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6528 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6529 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6530 "Scalar value %"SVf" better written as $%"SVf,
6531 SVfARG(newSVpvn_flags(PL_bufptr, (STRLEN)(t-PL_bufptr),
6532 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))),
6533 SVfARG(newSVpvn_flags(PL_bufptr+1, (STRLEN)(t-PL_bufptr-1),
6534 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))));
6539 PL_expect = XOPERATOR;
6540 force_ident_maybe_lex('@');
6543 case '/': /* may be division, defined-or, or pattern */
6544 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6545 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6546 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6551 case '?': /* may either be conditional or pattern */
6552 if (PL_expect == XOPERATOR) {
6555 if (!PL_lex_allbrackets &&
6556 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6560 PL_lex_allbrackets++;
6566 /* A // operator. */
6567 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6568 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6569 LEX_FAKEEOF_LOGIC)) {
6577 if (*s == '=' && !PL_lex_allbrackets &&
6578 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6587 /* Disable warning on "study /blah/" */
6588 if (PL_oldoldbufptr == PL_last_uni
6589 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6590 || memNE(PL_last_uni, "study", 5)
6591 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6595 deprecate("?PATTERN? without explicit operator");
6596 s = scan_pat(s,OP_MATCH);
6597 TERM(sublex_start());
6601 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6602 #ifdef PERL_STRICT_CR
6605 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6607 && (s == PL_linestart || s[-1] == '\n') )
6610 formbrack = 2; /* dot seen where arguments expected */
6613 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6617 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6620 if (!PL_lex_allbrackets &&
6621 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6628 pl_yylval.ival = OPf_SPECIAL;
6634 if (*s == '=' && !PL_lex_allbrackets &&
6635 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6642 case '0': case '1': case '2': case '3': case '4':
6643 case '5': case '6': case '7': case '8': case '9':
6644 s = scan_num(s, &pl_yylval);
6645 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6646 if (PL_expect == XOPERATOR)
6651 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6652 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6653 if (PL_expect == XOPERATOR) {
6654 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6655 return deprecate_commaless_var_list();
6662 pl_yylval.ival = OP_CONST;
6663 TERM(sublex_start());
6666 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6667 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6668 if (PL_expect == XOPERATOR) {
6669 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6670 return deprecate_commaless_var_list();
6677 pl_yylval.ival = OP_CONST;
6678 /* FIXME. I think that this can be const if char *d is replaced by
6679 more localised variables. */
6680 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6681 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6682 pl_yylval.ival = OP_STRINGIFY;
6686 TERM(sublex_start());
6689 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6690 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6691 if (PL_expect == XOPERATOR)
6692 no_op("Backticks",s);
6695 readpipe_override();
6696 TERM(sublex_start());
6700 if (PL_lex_inwhat && isDIGIT(*s))
6701 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6703 if (PL_expect == XOPERATOR)
6704 no_op("Backslash",s);
6708 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6709 char *start = s + 2;
6710 while (isDIGIT(*start) || *start == '_')
6712 if (*start == '.' && isDIGIT(start[1])) {
6713 s = scan_num(s, &pl_yylval);
6716 else if ((*start == ':' && start[1] == ':')
6717 || (PL_expect == XSTATE && *start == ':'))
6719 else if (PL_expect == XSTATE) {
6721 while (d < PL_bufend && isSPACE(*d)) d++;
6722 if (*d == ':') goto keylookup;
6724 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6725 if (!isALPHA(*start) && (PL_expect == XTERM
6726 || PL_expect == XREF || PL_expect == XSTATE
6727 || PL_expect == XTERMORDORDOR)) {
6728 GV *const gv = gv_fetchpvn_flags(s, start - s,
6729 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6731 s = scan_num(s, &pl_yylval);
6738 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6791 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6793 /* Some keywords can be followed by any delimiter, including ':' */
6794 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6796 /* x::* is just a word, unless x is "CORE" */
6797 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6801 while (d < PL_bufend && isSPACE(*d))
6802 d++; /* no comments skipped here, or s### is misparsed */
6804 /* Is this a word before a => operator? */
6805 if (*d == '=' && d[1] == '>') {
6808 = (OP*)newSVOP(OP_CONST, 0,
6809 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6810 pl_yylval.opval->op_private = OPpCONST_BARE;
6814 /* Check for plugged-in keyword */
6818 char *saved_bufptr = PL_bufptr;
6820 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6822 if (result == KEYWORD_PLUGIN_DECLINE) {
6823 /* not a plugged-in keyword */
6824 PL_bufptr = saved_bufptr;
6825 } else if (result == KEYWORD_PLUGIN_STMT) {
6826 pl_yylval.opval = o;
6829 return REPORT(PLUGSTMT);
6830 } else if (result == KEYWORD_PLUGIN_EXPR) {
6831 pl_yylval.opval = o;
6833 PL_expect = XOPERATOR;
6834 return REPORT(PLUGEXPR);
6836 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6841 /* Check for built-in keyword */
6842 tmp = keyword(PL_tokenbuf, len, 0);
6844 /* Is this a label? */
6845 if (!anydelim && PL_expect == XSTATE
6846 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6848 pl_yylval.pval = savepvn(PL_tokenbuf, len+1);
6849 pl_yylval.pval[len] = '\0';
6850 pl_yylval.pval[len+1] = UTF ? 1 : 0;
6855 /* Check for lexical sub */
6856 if (PL_expect != XOPERATOR) {
6857 char tmpbuf[sizeof PL_tokenbuf + 1];
6859 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6860 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6861 if (off != NOT_IN_PAD) {
6862 assert(off); /* we assume this is boolean-true below */
6863 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6864 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6865 HEK * const stashname = HvNAME_HEK(stash);
6866 sv = newSVhek(stashname);
6867 sv_catpvs(sv, "::");
6868 sv_catpvn_flags(sv, PL_tokenbuf, len,
6869 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6870 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6875 rv2cv_op = newOP(OP_PADANY, 0);
6876 rv2cv_op->op_targ = off;
6877 rv2cv_op = (OP*)newCVREF(0, rv2cv_op);
6878 cv = (CV *)PAD_SV(off);
6886 if (tmp < 0) { /* second-class keyword? */
6887 GV *ogv = NULL; /* override (winner) */
6888 GV *hgv = NULL; /* hidden (loser) */
6889 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6891 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6892 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6895 if (GvIMPORTED_CV(gv))
6897 else if (! CvMETHOD(cv))
6901 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6902 UTF ? -(I32)len : (I32)len, FALSE)) &&
6903 (gv = *gvp) && isGV_with_GP(gv) &&
6904 GvCVu(gv) && GvIMPORTED_CV(gv))
6911 tmp = 0; /* overridden by import or by GLOBAL */
6914 && -tmp==KEY_lock /* XXX generalizable kludge */
6917 tmp = 0; /* any sub overrides "weak" keyword */
6919 else { /* no override */
6921 if (tmp == KEY_dump) {
6922 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6923 "dump() better written as CORE::dump()");
6927 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6928 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6929 "Ambiguous call resolved as CORE::%s(), "
6930 "qualify as such or use &",
6938 default: /* not a keyword */
6939 /* Trade off - by using this evil construction we can pull the
6940 variable gv into the block labelled keylookup. If not, then
6941 we have to give it function scope so that the goto from the
6942 earlier ':' case doesn't bypass the initialisation. */
6944 just_a_word_zero_gv:
6956 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6957 const char penultchar =
6958 lastchar && PL_bufptr - 2 >= PL_linestart
6962 SV *nextPL_nextwhite = 0;
6966 /* Get the rest if it looks like a package qualifier */
6968 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6970 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6973 Perl_croak(aTHX_ "Bad name after %"SVf"%s",
6974 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6975 (UTF ? SVf_UTF8 : 0) | SVs_TEMP )),
6976 *s == '\'' ? "'" : "::");
6981 if (PL_expect == XOPERATOR) {
6982 if (PL_bufptr == PL_linestart) {
6983 CopLINE_dec(PL_curcop);
6984 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6985 CopLINE_inc(PL_curcop);
6988 no_op("Bareword",s);
6991 /* Look for a subroutine with this name in current package,
6992 unless this is a lexical sub, or name is "Foo::",
6993 in which case Foo is a bareword
6994 (and a package name). */
6996 if (len > 2 && !PL_madskills &&
6997 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6999 if (ckWARN(WARN_BAREWORD)
7000 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
7001 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
7002 "Bareword \"%"SVf"\" refers to nonexistent package",
7003 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7004 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7006 PL_tokenbuf[len] = '\0';
7012 /* Mustn't actually add anything to a symbol table.
7013 But also don't want to "initialise" any placeholder
7014 constants that might already be there into full
7015 blown PVGVs with attached PVCV. */
7016 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
7017 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
7023 /* if we saw a global override before, get the right name */
7026 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
7027 len ? len : strlen(PL_tokenbuf));
7029 SV * const tmp_sv = sv;
7030 sv = newSVpvs("CORE::GLOBAL::");
7031 sv_catsv(sv, tmp_sv);
7032 SvREFCNT_dec(tmp_sv);
7036 if (PL_madskills && !PL_thistoken) {
7037 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
7038 PL_thistoken = newSVpvn(start,s - start);
7039 PL_realtokenstart = s - SvPVX(PL_linestr);
7043 /* Presume this is going to be a bareword of some sort. */
7045 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
7046 pl_yylval.opval->op_private = OPpCONST_BARE;
7048 /* And if "Foo::", then that's what it certainly is. */
7054 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
7055 const_op->op_private = OPpCONST_BARE;
7056 rv2cv_op = newCVREF(0, const_op);
7057 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
7060 /* See if it's the indirect object for a list operator. */
7062 if (PL_oldoldbufptr &&
7063 PL_oldoldbufptr < PL_bufptr &&
7064 (PL_oldoldbufptr == PL_last_lop
7065 || PL_oldoldbufptr == PL_last_uni) &&
7066 /* NO SKIPSPACE BEFORE HERE! */
7067 (PL_expect == XREF ||
7068 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
7070 bool immediate_paren = *s == '(';
7072 /* (Now we can afford to cross potential line boundary.) */
7073 s = SKIPSPACE2(s,nextPL_nextwhite);
7075 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
7078 /* Two barewords in a row may indicate method call. */
7080 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
7081 (tmp = intuit_method(s, gv, cv))) {
7083 if (tmp == METHOD && !PL_lex_allbrackets &&
7084 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7085 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7089 /* If not a declared subroutine, it's an indirect object. */
7090 /* (But it's an indir obj regardless for sort.) */
7091 /* Also, if "_" follows a filetest operator, it's a bareword */
7094 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
7096 (PL_last_lop_op != OP_MAPSTART &&
7097 PL_last_lop_op != OP_GREPSTART))))
7098 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
7099 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
7102 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
7107 PL_expect = XOPERATOR;
7110 s = SKIPSPACE2(s,nextPL_nextwhite);
7111 PL_nextwhite = nextPL_nextwhite;
7116 /* Is this a word before a => operator? */
7117 if (*s == '=' && s[1] == '>' && !pkgname) {
7120 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
7121 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
7122 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
7126 /* If followed by a paren, it's certainly a subroutine. */
7131 while (SPACE_OR_TAB(*d))
7133 if (*d == ')' && (sv = cv_const_sv(cv))) {
7140 PL_nextwhite = PL_thiswhite;
7143 start_force(PL_curforce);
7145 NEXTVAL_NEXTTOKE.opval =
7146 off ? rv2cv_op : pl_yylval.opval;
7147 PL_expect = XOPERATOR;
7150 PL_nextwhite = nextPL_nextwhite;
7151 curmad('X', PL_thistoken);
7152 PL_thistoken = newSVpvs("");
7156 op_free(pl_yylval.opval), force_next(PRIVATEREF);
7157 else op_free(rv2cv_op), force_next(WORD);
7162 /* If followed by var or block, call it a method (unless sub) */
7164 if ((*s == '$' || *s == '{') && !cv) {
7166 PL_last_lop = PL_oldbufptr;
7167 PL_last_lop_op = OP_METHOD;
7168 if (!PL_lex_allbrackets &&
7169 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7170 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7174 /* If followed by a bareword, see if it looks like indir obj. */
7177 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7178 && (tmp = intuit_method(s, gv, cv))) {
7180 if (tmp == METHOD && !PL_lex_allbrackets &&
7181 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7182 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7186 /* Not a method, so call it a subroutine (if defined) */
7189 if (lastchar == '-' && penultchar != '-') {
7190 const SV *tmpsv = newSVpvn_flags( PL_tokenbuf, len ? len : strlen(PL_tokenbuf), (UTF ? SVf_UTF8 : 0) | SVs_TEMP );
7191 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7192 "Ambiguous use of -%"SVf" resolved as -&%"SVf"()",
7193 SVfARG(tmpsv), SVfARG(tmpsv));
7195 /* Check for a constant sub */
7196 if ((sv = cv_const_sv(cv))) {
7199 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7200 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7201 pl_yylval.opval->op_private = OPpCONST_FOLDED;
7202 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7206 op_free(pl_yylval.opval);
7207 pl_yylval.opval = rv2cv_op;
7208 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7209 PL_last_lop = PL_oldbufptr;
7210 PL_last_lop_op = OP_ENTERSUB;
7211 /* Is there a prototype? */
7218 STRLEN protolen = CvPROTOLEN(cv);
7219 const char *proto = CvPROTO(cv);
7223 if ((optional = *proto == ';'))
7226 while (*proto == ';');
7230 *proto == '$' || *proto == '_'
7231 || *proto == '*' || *proto == '+'
7236 *proto == '\\' && proto[1] && proto[2] == '\0'
7239 UNIPROTO(UNIOPSUB,optional);
7240 if (*proto == '\\' && proto[1] == '[') {
7241 const char *p = proto + 2;
7242 while(*p && *p != ']')
7244 if(*p == ']' && !p[1])
7245 UNIPROTO(UNIOPSUB,optional);
7247 if (*proto == '&' && *s == '{') {
7249 sv_setpvs(PL_subname, "__ANON__");
7251 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7252 if (!PL_lex_allbrackets &&
7253 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7254 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7261 PL_nextwhite = PL_thiswhite;
7264 start_force(PL_curforce);
7265 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7268 PL_nextwhite = nextPL_nextwhite;
7269 curmad('X', PL_thistoken);
7270 PL_thistoken = newSVpvs("");
7272 force_next(off ? PRIVATEREF : WORD);
7273 if (!PL_lex_allbrackets &&
7274 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7275 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7280 /* Guess harder when madskills require "best effort". */
7281 if (PL_madskills && (!gv || !GvCVu(gv))) {
7282 int probable_sub = 0;
7283 if (strchr("\"'`$@%0123456789!*+{[<", *s))
7285 else if (isALPHA(*s)) {
7289 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
7290 if (!keyword(tmpbuf, tmplen, 0))
7293 while (d < PL_bufend && isSPACE(*d))
7295 if (*d == '=' && d[1] == '>')
7300 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
7302 op_free(pl_yylval.opval);
7303 pl_yylval.opval = rv2cv_op;
7304 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7305 PL_last_lop = PL_oldbufptr;
7306 PL_last_lop_op = OP_ENTERSUB;
7307 PL_nextwhite = PL_thiswhite;
7309 start_force(PL_curforce);
7310 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7312 PL_nextwhite = nextPL_nextwhite;
7313 curmad('X', PL_thistoken);
7314 PL_thistoken = newSVpvs("");
7315 force_next(off ? PRIVATEREF : WORD);
7316 if (!PL_lex_allbrackets &&
7317 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7318 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7322 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7324 force_next(off ? PRIVATEREF : WORD);
7325 if (!PL_lex_allbrackets &&
7326 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7327 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7332 /* Call it a bare word */
7334 if (PL_hints & HINT_STRICT_SUBS)
7335 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7338 /* after "print" and similar functions (corresponding to
7339 * "F? L" in opcode.pl), whatever wasn't already parsed as
7340 * a filehandle should be subject to "strict subs".
7341 * Likewise for the optional indirect-object argument to system
7342 * or exec, which can't be a bareword */
7343 if ((PL_last_lop_op == OP_PRINT
7344 || PL_last_lop_op == OP_PRTF
7345 || PL_last_lop_op == OP_SAY
7346 || PL_last_lop_op == OP_SYSTEM
7347 || PL_last_lop_op == OP_EXEC)
7348 && (PL_hints & HINT_STRICT_SUBS))
7349 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7350 if (lastchar != '-') {
7351 if (ckWARN(WARN_RESERVED)) {
7355 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7356 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7364 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7365 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7366 "Operator or semicolon missing before %c%"SVf,
7367 lastchar, SVfARG(newSVpvn_flags(PL_tokenbuf,
7368 strlen(PL_tokenbuf),
7369 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
7370 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7371 "Ambiguous use of %c resolved as operator %c",
7372 lastchar, lastchar);
7379 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7384 (OP*)newSVOP(OP_CONST, 0,
7385 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7388 case KEY___PACKAGE__:
7390 (OP*)newSVOP(OP_CONST, 0,
7392 ? newSVhek(HvNAME_HEK(PL_curstash))
7399 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7400 const char *pname = "main";
7403 if (PL_tokenbuf[2] == 'D')
7406 PL_curstash ? PL_curstash : PL_defstash;
7407 pname = HvNAME_get(stash);
7408 plen = HvNAMELEN (stash);
7409 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7411 gv = gv_fetchpvn_flags(
7412 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7413 plen+6, GV_ADD|putf8, SVt_PVIO
7417 GvIOp(gv) = newIO();
7418 IoIFP(GvIOp(gv)) = PL_rsfp;
7419 #if defined(HAS_FCNTL) && defined(F_SETFD)
7421 const int fd = PerlIO_fileno(PL_rsfp);
7422 fcntl(fd,F_SETFD,fd >= 3);
7425 /* Mark this internal pseudo-handle as clean */
7426 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7427 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7428 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7430 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7431 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7432 /* if the script was opened in binmode, we need to revert
7433 * it to text mode for compatibility; but only iff it has CRs
7434 * XXX this is a questionable hack at best. */
7435 if (PL_bufend-PL_bufptr > 2
7436 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7439 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7440 loc = PerlIO_tell(PL_rsfp);
7441 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7444 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7446 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7447 #endif /* NETWARE */
7449 PerlIO_seek(PL_rsfp, loc, 0);
7453 #ifdef PERLIO_LAYERS
7456 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7457 else if (PL_encoding) {
7464 XPUSHs(PL_encoding);
7466 call_method("name", G_SCALAR);
7470 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7471 Perl_form(aTHX_ ":encoding(%"SVf")",
7480 if (PL_realtokenstart >= 0) {
7481 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7483 PL_endwhite = newSVpvs("");
7484 sv_catsv(PL_endwhite, PL_thiswhite);
7486 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7487 PL_realtokenstart = -1;
7489 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7499 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7508 if (PL_expect == XSTATE) {
7515 if (*s == ':' && s[1] == ':') {
7519 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7520 if ((*s == ':' && s[1] == ':')
7521 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7525 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7529 Perl_croak(aTHX_ "CORE::%"SVf" is not a keyword",
7530 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7531 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7534 else if (tmp == KEY_require || tmp == KEY_do
7536 /* that's a way to remember we saw "CORE::" */
7549 LOP(OP_ACCEPT,XTERM);
7552 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7557 LOP(OP_ATAN2,XTERM);
7563 LOP(OP_BINMODE,XTERM);
7566 LOP(OP_BLESS,XTERM);
7575 /* We have to disambiguate the two senses of
7576 "continue". If the next token is a '{' then
7577 treat it as the start of a continue block;
7578 otherwise treat it as a control operator.
7588 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7598 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7607 if (!PL_cryptseen) {
7608 PL_cryptseen = TRUE;
7612 LOP(OP_CRYPT,XTERM);
7615 LOP(OP_CHMOD,XTERM);
7618 LOP(OP_CHOWN,XTERM);
7621 LOP(OP_CONNECT,XTERM);
7641 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7643 if (len && !keyword(PL_tokenbuf + 1, len, 0)) {
7646 force_ident_maybe_lex('&');
7651 if (orig_keyword == KEY_do) {
7660 PL_hints |= HINT_BLOCK_SCOPE;
7670 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7671 STR_WITH_LEN("NDBM_File::"),
7672 STR_WITH_LEN("DB_File::"),
7673 STR_WITH_LEN("GDBM_File::"),
7674 STR_WITH_LEN("SDBM_File::"),
7675 STR_WITH_LEN("ODBM_File::"),
7677 LOP(OP_DBMOPEN,XTERM);
7683 PL_expect = XOPERATOR;
7684 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7691 pl_yylval.ival = CopLINE(PL_curcop);
7695 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7709 if (*s == '{') { /* block eval */
7710 PL_expect = XTERMBLOCK;
7711 UNIBRACK(OP_ENTERTRY);
7713 else { /* string eval */
7715 UNIBRACK(OP_ENTEREVAL);
7720 UNIBRACK(-OP_ENTEREVAL);
7734 case KEY_endhostent:
7740 case KEY_endservent:
7743 case KEY_endprotoent:
7754 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7756 pl_yylval.ival = CopLINE(PL_curcop);
7758 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7761 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7764 if ((PL_bufend - p) >= 3 &&
7765 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7767 else if ((PL_bufend - p) >= 4 &&
7768 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7771 if (isIDFIRST_lazy_if(p,UTF)) {
7772 p = scan_ident(p, PL_bufend,
7773 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7777 Perl_croak(aTHX_ "Missing $ on loop variable");
7779 s = SvPVX(PL_linestr) + soff;
7785 LOP(OP_FORMLINE,XTERM);
7794 LOP(OP_FCNTL,XTERM);
7800 LOP(OP_FLOCK,XTERM);
7803 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7808 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7813 LOP(OP_GREPSTART, XREF);
7816 PL_expect = XOPERATOR;
7817 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7832 case KEY_getpriority:
7833 LOP(OP_GETPRIORITY,XTERM);
7835 case KEY_getprotobyname:
7838 case KEY_getprotobynumber:
7839 LOP(OP_GPBYNUMBER,XTERM);
7841 case KEY_getprotoent:
7853 case KEY_getpeername:
7854 UNI(OP_GETPEERNAME);
7856 case KEY_gethostbyname:
7859 case KEY_gethostbyaddr:
7860 LOP(OP_GHBYADDR,XTERM);
7862 case KEY_gethostent:
7865 case KEY_getnetbyname:
7868 case KEY_getnetbyaddr:
7869 LOP(OP_GNBYADDR,XTERM);
7874 case KEY_getservbyname:
7875 LOP(OP_GSBYNAME,XTERM);
7877 case KEY_getservbyport:
7878 LOP(OP_GSBYPORT,XTERM);
7880 case KEY_getservent:
7883 case KEY_getsockname:
7884 UNI(OP_GETSOCKNAME);
7886 case KEY_getsockopt:
7887 LOP(OP_GSOCKOPT,XTERM);
7902 pl_yylval.ival = CopLINE(PL_curcop);
7907 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7915 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7917 pl_yylval.ival = CopLINE(PL_curcop);
7921 LOP(OP_INDEX,XTERM);
7927 LOP(OP_IOCTL,XTERM);
7939 PL_expect = XOPERATOR;
7940 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7957 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7962 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7976 LOP(OP_LISTEN,XTERM);
7985 s = scan_pat(s,OP_MATCH);
7986 TERM(sublex_start());
7989 LOP(OP_MAPSTART, XREF);
7992 LOP(OP_MKDIR,XTERM);
7995 LOP(OP_MSGCTL,XTERM);
7998 LOP(OP_MSGGET,XTERM);
8001 LOP(OP_MSGRCV,XTERM);
8004 LOP(OP_MSGSND,XTERM);
8009 PL_in_my = (U16)tmp;
8011 if (isIDFIRST_lazy_if(s,UTF)) {
8015 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
8016 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
8018 if (!FEATURE_LEXSUBS_IS_ENABLED)
8020 "Experimental \"%s\" subs not enabled",
8021 tmp == KEY_my ? "my" :
8022 tmp == KEY_state ? "state" : "our");
8025 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
8026 if (!PL_in_my_stash) {
8029 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
8030 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
8033 if (PL_madskills) { /* just add type to declarator token */
8034 sv_catsv(PL_thistoken, PL_nextwhite);
8036 sv_catpvn(PL_thistoken, start, s - start);
8044 PL_expect = XOPERATOR;
8045 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8049 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
8054 s = tokenize_use(0, s);
8058 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
8061 if (!PL_lex_allbrackets &&
8062 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
8063 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
8069 if (isIDFIRST_lazy_if(s,UTF)) {
8071 for (d = s; isALNUM_lazy_if(d,UTF);) {
8072 d += UTF ? UTF8SKIP(d) : 1;
8074 while (UTF8_IS_CONTINUED(*d) && is_utf8_mark((U8*)d)) {
8075 d += UTF ? UTF8SKIP(d) : 1;
8079 for (t=d; isSPACE(*t);)
8081 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
8083 && !(t[0] == '=' && t[1] == '>')
8084 && !(t[0] == ':' && t[1] == ':')
8085 && !keyword(s, d-s, 0)
8087 SV *tmpsv = newSVpvn_flags(s, (STRLEN)(d-s),
8088 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
8089 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
8090 "Precedence problem: open %"SVf" should be open(%"SVf")",
8091 SVfARG(tmpsv), SVfARG(tmpsv));
8097 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8099 pl_yylval.ival = OP_OR;
8109 LOP(OP_OPEN_DIR,XTERM);
8112 checkcomma(s,PL_tokenbuf,"filehandle");
8116 checkcomma(s,PL_tokenbuf,"filehandle");
8135 s = force_word(s,WORD,FALSE,TRUE,FALSE);
8137 s = force_strict_version(s);
8138 PL_lex_expect = XBLOCK;
8142 LOP(OP_PIPE_OP,XTERM);
8145 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8148 pl_yylval.ival = OP_CONST;
8149 TERM(sublex_start());
8156 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8159 PL_expect = XOPERATOR;
8160 if (SvCUR(PL_lex_stuff)) {
8161 int warned_comma = !ckWARN(WARN_QW);
8162 int warned_comment = warned_comma;
8163 d = SvPV_force(PL_lex_stuff, len);
8165 for (; isSPACE(*d) && len; --len, ++d)
8170 if (!warned_comma || !warned_comment) {
8171 for (; !isSPACE(*d) && len; --len, ++d) {
8172 if (!warned_comma && *d == ',') {
8173 Perl_warner(aTHX_ packWARN(WARN_QW),
8174 "Possible attempt to separate words with commas");
8177 else if (!warned_comment && *d == '#') {
8178 Perl_warner(aTHX_ packWARN(WARN_QW),
8179 "Possible attempt to put comments in qw() list");
8185 for (; !isSPACE(*d) && len; --len, ++d)
8188 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
8189 words = op_append_elem(OP_LIST, words,
8190 newSVOP(OP_CONST, 0, tokeq(sv)));
8195 words = newNULLLIST();
8197 SvREFCNT_dec(PL_lex_stuff);
8198 PL_lex_stuff = NULL;
8200 PL_expect = XOPERATOR;
8201 pl_yylval.opval = sawparens(words);
8206 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8209 pl_yylval.ival = OP_STRINGIFY;
8210 if (SvIVX(PL_lex_stuff) == '\'')
8211 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
8212 TERM(sublex_start());
8215 s = scan_pat(s,OP_QR);
8216 TERM(sublex_start());
8219 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8222 readpipe_override();
8223 TERM(sublex_start());
8230 PL_expect = XOPERATOR;
8232 s = force_version(s, FALSE);
8234 else if (*s != 'v' || !isDIGIT(s[1])
8235 || (s = force_version(s, TRUE), *s == 'v'))
8237 *PL_tokenbuf = '\0';
8238 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8239 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8240 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8241 GV_ADD | (UTF ? SVf_UTF8 : 0));
8243 yyerror("<> should be quotes");
8245 if (orig_keyword == KEY_require) {
8253 PL_last_uni = PL_oldbufptr;
8254 PL_last_lop_op = OP_REQUIRE;
8256 return REPORT( (int)REQUIRE );
8262 PL_expect = XOPERATOR;
8263 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8267 LOP(OP_RENAME,XTERM);
8276 LOP(OP_RINDEX,XTERM);
8285 UNIDOR(OP_READLINE);
8288 UNIDOR(OP_BACKTICK);
8297 LOP(OP_REVERSE,XTERM);
8300 UNIDOR(OP_READLINK);
8307 if (pl_yylval.opval)
8308 TERM(sublex_start());
8310 TOKEN(1); /* force error */
8313 checkcomma(s,PL_tokenbuf,"filehandle");
8323 LOP(OP_SELECT,XTERM);
8329 LOP(OP_SEMCTL,XTERM);
8332 LOP(OP_SEMGET,XTERM);
8335 LOP(OP_SEMOP,XTERM);
8341 LOP(OP_SETPGRP,XTERM);
8343 case KEY_setpriority:
8344 LOP(OP_SETPRIORITY,XTERM);
8346 case KEY_sethostent:
8352 case KEY_setservent:
8355 case KEY_setprotoent:
8365 LOP(OP_SEEKDIR,XTERM);
8367 case KEY_setsockopt:
8368 LOP(OP_SSOCKOPT,XTERM);
8374 LOP(OP_SHMCTL,XTERM);
8377 LOP(OP_SHMGET,XTERM);
8380 LOP(OP_SHMREAD,XTERM);
8383 LOP(OP_SHMWRITE,XTERM);
8386 LOP(OP_SHUTDOWN,XTERM);
8395 LOP(OP_SOCKET,XTERM);
8397 case KEY_socketpair:
8398 LOP(OP_SOCKPAIR,XTERM);
8401 checkcomma(s,PL_tokenbuf,"subroutine name");
8404 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8408 LOP(OP_SPLIT,XTERM);
8411 LOP(OP_SPRINTF,XTERM);
8414 LOP(OP_SPLICE,XTERM);
8429 LOP(OP_SUBSTR,XTERM);
8435 char * const tmpbuf = PL_tokenbuf + 1;
8436 SSize_t tboffset = 0;
8437 expectation attrful;
8438 bool have_name, have_proto;
8439 const int key = tmp;
8444 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8445 SV *subtoken = PL_madskills
8446 ? newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr))
8451 s = SKIPSPACE2(s,tmpwhite);
8457 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8458 (*s == ':' && s[1] == ':'))
8461 SV *nametoke = NULL;
8465 attrful = XATTRBLOCK;
8466 /* remember buffer pos'n for later force_word */
8467 tboffset = s - PL_oldbufptr;
8468 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8472 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8475 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8477 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
8479 sv_setpvn(PL_subname, tmpbuf, len);
8481 sv_setsv(PL_subname,PL_curstname);
8482 sv_catpvs(PL_subname,"::");
8483 sv_catpvn(PL_subname,tmpbuf,len);
8485 if (SvUTF8(PL_linestr))
8486 SvUTF8_on(PL_subname);
8492 CURMAD('X', nametoke);
8493 CURMAD('_', tmpwhite);
8494 force_ident_maybe_lex('&');
8496 s = SKIPSPACE2(d,tmpwhite);
8502 if (key == KEY_my || key == KEY_our || key==KEY_state)
8505 /* diag_listed_as: Missing name in "%s sub" */
8507 "Missing name in \"%s\"", PL_bufptr);
8509 PL_expect = XTERMBLOCK;
8510 attrful = XATTRTERM;
8511 sv_setpvs(PL_subname,"?");
8515 if (key == KEY_format) {
8517 PL_thistoken = subtoken;
8521 (void) force_word(PL_oldbufptr + tboffset, WORD,
8527 /* Look for a prototype */
8530 bool bad_proto = FALSE;
8531 bool in_brackets = FALSE;
8532 char greedy_proto = ' ';
8533 bool proto_after_greedy_proto = FALSE;
8534 bool must_be_last = FALSE;
8535 bool underscore = FALSE;
8536 bool seen_underscore = FALSE;
8537 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8540 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8542 Perl_croak(aTHX_ "Prototype not terminated");
8543 /* strip spaces and check for bad characters */
8544 d = SvPV(PL_lex_stuff, tmplen);
8546 for (p = d; tmplen; tmplen--, ++p) {
8550 if (warnillegalproto) {
8552 proto_after_greedy_proto = TRUE;
8553 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8558 if ( !strchr(";@%", *p) )
8565 else if ( *p == ']' ) {
8566 in_brackets = FALSE;
8568 else if ( (*p == '@' || *p == '%') &&
8569 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8571 must_be_last = TRUE;
8574 else if ( *p == '_' ) {
8575 underscore = seen_underscore = TRUE;
8582 if (proto_after_greedy_proto)
8583 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8584 "Prototype after '%c' for %"SVf" : %s",
8585 greedy_proto, SVfARG(PL_subname), d);
8587 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8588 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8589 "Illegal character %sin prototype for %"SVf" : %s",
8590 seen_underscore ? "after '_' " : "",
8592 SvUTF8(PL_lex_stuff)
8593 ? sv_uni_display(dsv,
8594 newSVpvn_flags(d, tmp, SVs_TEMP | SVf_UTF8),
8596 UNI_DISPLAY_ISPRINT)
8597 : pv_pretty(dsv, d, tmp, 60, NULL, NULL,
8598 PERL_PV_ESCAPE_NONASCII));
8600 SvCUR_set(PL_lex_stuff, tmp);
8605 CURMAD('q', PL_thisopen);
8606 CURMAD('_', tmpwhite);
8607 CURMAD('=', PL_thisstuff);
8608 CURMAD('Q', PL_thisclose);
8609 NEXTVAL_NEXTTOKE.opval =
8610 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8611 PL_lex_stuff = NULL;
8614 s = SKIPSPACE2(s,tmpwhite);
8622 if (*s == ':' && s[1] != ':')
8623 PL_expect = attrful;
8624 else if (*s != '{' && key == KEY_sub) {
8626 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8627 else if (*s != ';' && *s != '}')
8628 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8635 curmad('^', newSVpvs(""));
8636 CURMAD('_', tmpwhite);
8640 PL_thistoken = subtoken;
8643 NEXTVAL_NEXTTOKE.opval =
8644 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8645 PL_lex_stuff = NULL;
8651 sv_setpvs(PL_subname, "__ANON__");
8653 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8657 force_ident_maybe_lex('&');
8663 LOP(OP_SYSTEM,XREF);
8666 LOP(OP_SYMLINK,XTERM);
8669 LOP(OP_SYSCALL,XTERM);
8672 LOP(OP_SYSOPEN,XTERM);
8675 LOP(OP_SYSSEEK,XTERM);
8678 LOP(OP_SYSREAD,XTERM);
8681 LOP(OP_SYSWRITE,XTERM);
8686 TERM(sublex_start());
8707 LOP(OP_TRUNCATE,XTERM);
8719 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8721 pl_yylval.ival = CopLINE(PL_curcop);
8725 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8727 pl_yylval.ival = CopLINE(PL_curcop);
8731 LOP(OP_UNLINK,XTERM);
8737 LOP(OP_UNPACK,XTERM);
8740 LOP(OP_UTIME,XTERM);
8746 LOP(OP_UNSHIFT,XTERM);
8749 s = tokenize_use(1, s);
8759 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8761 pl_yylval.ival = CopLINE(PL_curcop);
8765 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8767 pl_yylval.ival = CopLINE(PL_curcop);
8771 PL_hints |= HINT_BLOCK_SCOPE;
8778 LOP(OP_WAITPID,XTERM);
8787 ctl_l[0] = toCTRL('L');
8789 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8792 /* Make sure $^L is defined */
8793 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8798 if (PL_expect == XOPERATOR) {
8799 if (*s == '=' && !PL_lex_allbrackets &&
8800 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8808 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8810 pl_yylval.ival = OP_XOR;
8816 #pragma segment Main
8822 Looks up an identifier in the pad or in a package
8825 PRIVATEREF if this is a lexical name.
8826 WORD if this belongs to a package.
8829 if we're in a my declaration
8830 croak if they tried to say my($foo::bar)
8831 build the ops for a my() declaration
8832 if it's an access to a my() variable
8833 build ops for access to a my() variable
8834 if in a dq string, and they've said @foo and we can't find @foo
8836 build ops for a bareword
8840 S_pending_ident(pTHX)
8844 const char pit = (char)pl_yylval.ival;
8845 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8846 /* All routes through this function want to know if there is a colon. */
8847 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8849 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8850 "### Pending identifier '%s'\n", PL_tokenbuf); });
8852 /* if we're in a my(), we can't allow dynamics here.
8853 $foo'bar has already been turned into $foo::bar, so
8854 just check for colons.
8856 if it's a legal name, the OP is a PADANY.
8859 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8861 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8862 "variable %s in \"our\"",
8863 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8864 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8868 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8869 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8870 UTF ? SVf_UTF8 : 0);
8872 pl_yylval.opval = newOP(OP_PADANY, 0);
8873 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8874 UTF ? SVf_UTF8 : 0);
8880 build the ops for accesses to a my() variable.
8885 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8886 UTF ? SVf_UTF8 : 0);
8887 if (tmp != NOT_IN_PAD) {
8888 /* might be an "our" variable" */
8889 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8890 /* build ops for a bareword */
8891 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8892 HEK * const stashname = HvNAME_HEK(stash);
8893 SV * const sym = newSVhek(stashname);
8894 sv_catpvs(sym, "::");
8895 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8896 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8897 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8901 ? (GV_ADDMULTI | GV_ADDINEVAL)
8904 ((PL_tokenbuf[0] == '$') ? SVt_PV
8905 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8910 pl_yylval.opval = newOP(OP_PADANY, 0);
8911 pl_yylval.opval->op_targ = tmp;
8917 Whine if they've said @foo in a doublequoted string,
8918 and @foo isn't a variable we can find in the symbol
8921 if (ckWARN(WARN_AMBIGUOUS) &&
8922 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8923 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8924 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8925 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8926 /* DO NOT warn for @- and @+ */
8927 && !( PL_tokenbuf[2] == '\0' &&
8928 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8931 /* Downgraded from fatal to warning 20000522 mjd */
8932 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8933 "Possible unintended interpolation of %"SVf" in string",
8934 SVfARG(newSVpvn_flags(PL_tokenbuf, tokenbuf_len,
8935 SVs_TEMP | ( UTF ? SVf_UTF8 : 0 ))));
8939 /* build ops for a bareword */
8940 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8941 newSVpvn_flags(PL_tokenbuf + 1,
8943 UTF ? SVf_UTF8 : 0 ));
8944 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8946 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8947 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8948 | ( UTF ? SVf_UTF8 : 0 ),
8949 ((PL_tokenbuf[0] == '$') ? SVt_PV
8950 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8956 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8960 PERL_ARGS_ASSERT_CHECKCOMMA;
8962 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8963 if (ckWARN(WARN_SYNTAX)) {
8966 for (w = s+2; *w && level; w++) {
8974 /* the list of chars below is for end of statements or
8975 * block / parens, boolean operators (&&, ||, //) and branch
8976 * constructs (or, and, if, until, unless, while, err, for).
8977 * Not a very solid hack... */
8978 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8979 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8980 "%s (...) interpreted as function",name);
8983 while (s < PL_bufend && isSPACE(*s))
8987 while (s < PL_bufend && isSPACE(*s))
8989 if (isIDFIRST_lazy_if(s,UTF)) {
8990 const char * const w = s;
8991 s += UTF ? UTF8SKIP(s) : 1;
8992 while (isALNUM_lazy_if(s,UTF))
8993 s += UTF ? UTF8SKIP(s) : 1;
8994 while (s < PL_bufend && isSPACE(*s))
8998 if (keyword(w, s - w, 0))
9001 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
9002 if (gv && GvCVu(gv))
9004 Perl_croak(aTHX_ "No comma allowed after %s", what);
9009 /* Either returns sv, or mortalizes sv and returns a new SV*.
9010 Best used as sv=new_constant(..., sv, ...).
9011 If s, pv are NULL, calls subroutine with one argument,
9012 and <type> is used with error messages only.
9013 <type> is assumed to be well formed UTF-8 */
9016 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
9017 SV *sv, SV *pv, const char *type, STRLEN typelen)
9020 HV * table = GvHV(PL_hintgv); /* ^H */
9024 const char *why1 = "", *why2 = "", *why3 = "";
9026 PERL_ARGS_ASSERT_NEW_CONSTANT;
9028 /* charnames doesn't work well if there have been errors found */
9029 if (PL_error_count > 0 && strEQ(key,"charnames"))
9030 return &PL_sv_undef;
9033 || ! (PL_hints & HINT_LOCALIZE_HH)
9034 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
9039 /* Here haven't found what we're looking for. If it is charnames,
9040 * perhaps it needs to be loaded. Try doing that before giving up */
9041 if (strEQ(key,"charnames")) {
9042 Perl_load_module(aTHX_
9044 newSVpvs("_charnames"),
9045 /* version parameter; no need to specify it, as if
9046 * we get too early a version, will fail anyway,
9047 * not being able to find '_charnames' */
9053 table = GvHV(PL_hintgv);
9055 && (PL_hints & HINT_LOCALIZE_HH)
9056 && (cvp = hv_fetch(table, key, keylen, FALSE))
9062 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
9063 msg = Perl_newSVpvf(aTHX_
9064 "Constant(%s) unknown", (type ? type: "undef"));
9069 why3 = "} is not defined";
9071 if (strEQ(key,"charnames")) {
9072 yyerror_pv(Perl_form(aTHX_
9073 /* The +3 is for '\N{'; -4 for that, plus '}' */
9074 "Unknown charname '%.*s'", (int)typelen - 4, type + 3
9076 UTF ? SVf_UTF8 : 0);
9080 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
9081 (type ? type: "undef"), why1, why2, why3);
9084 yyerror(SvPVX_const(msg));
9089 sv_2mortal(sv); /* Parent created it permanently */
9092 pv = newSVpvn_flags(s, len, SVs_TEMP);
9094 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
9096 typesv = &PL_sv_undef;
9098 PUSHSTACKi(PERLSI_OVERLOAD);
9110 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
9114 /* Check the eval first */
9115 if (!PL_in_eval && SvTRUE(ERRSV)) {
9117 const char * errstr;
9118 sv_catpvs(ERRSV, "Propagated");
9119 errstr = SvPV_const(ERRSV, errlen);
9120 yyerror_pvn(errstr, errlen, 0); /* Duplicates the message inside eval */
9122 res = SvREFCNT_inc_simple(sv);
9126 SvREFCNT_inc_simple_void(res);
9135 why1 = "Call to &{$^H{";
9137 why3 = "}} did not return a defined value";
9145 /* Returns a NUL terminated string, with the length of the string written to
9149 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
9153 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9155 PERL_ARGS_ASSERT_SCAN_WORD;
9159 Perl_croak(aTHX_ ident_too_long);
9160 if (isALNUM(*s) || (!UTF && isALNUMC_L1(*s))) /* UTF handled below */
9162 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
9167 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
9171 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9172 char *t = s + UTF8SKIP(s);
9174 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9178 Perl_croak(aTHX_ ident_too_long);
9179 Copy(s, d, len, char);
9192 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
9195 char *bracket = NULL;
9198 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9200 PERL_ARGS_ASSERT_SCAN_IDENT;
9205 while (isDIGIT(*s)) {
9207 Perl_croak(aTHX_ ident_too_long);
9214 Perl_croak(aTHX_ ident_too_long);
9215 if (isALNUM(*s)) /* UTF handled below */
9217 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
9222 else if (*s == ':' && s[1] == ':') {
9226 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9227 char *t = s + UTF8SKIP(s);
9228 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9230 if (d + (t - s) > e)
9231 Perl_croak(aTHX_ ident_too_long);
9232 Copy(s, d, t - s, char);
9243 if (PL_lex_state != LEX_NORMAL)
9244 PL_lex_state = LEX_INTERPENDMAYBE;
9247 if (*s == '$' && s[1] &&
9248 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
9258 const STRLEN skip = UTF8SKIP(s);
9261 for ( i = 0; i < skip; i++ )
9269 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9273 else if (ck_uni && !bracket)
9276 if (isSPACE(s[-1])) {
9278 const char ch = *s++;
9279 if (!SPACE_OR_TAB(ch)) {
9285 if (isIDFIRST_lazy_if(d,UTF)) {
9289 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
9290 end += UTF8SKIP(end);
9291 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
9292 end += UTF8SKIP(end);
9294 Copy(s, d, end - s, char);
9299 while ((isALNUM(*s) || *s == ':') && d < e)
9302 Perl_croak(aTHX_ ident_too_long);
9305 while (s < send && SPACE_OR_TAB(*s))
9307 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9308 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9309 const char * const brack =
9311 ((*s == '[') ? "[...]" : "{...}");
9312 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9313 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9314 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9315 funny, dest, brack, funny, dest, brack);
9318 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9319 PL_lex_allbrackets++;
9323 /* Handle extended ${^Foo} variables
9324 * 1999-02-27 mjd-perl-patch@plover.com */
9325 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
9329 while (isALNUM(*s) && d < e) {
9333 Perl_croak(aTHX_ ident_too_long);
9338 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9339 PL_lex_state = LEX_INTERPEND;
9342 if (PL_lex_state == LEX_NORMAL) {
9343 if (ckWARN(WARN_AMBIGUOUS) &&
9344 (keyword(dest, d - dest, 0)
9345 || get_cvn_flags(dest, d - dest, UTF ? SVf_UTF8 : 0)))
9347 SV *tmp = newSVpvn_flags( dest, d - dest,
9348 SVs_TEMP | (UTF ? SVf_UTF8 : 0) );
9351 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9352 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9353 funny, tmp, funny, tmp);
9358 s = bracket; /* let the parser handle it */
9362 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9363 PL_lex_state = LEX_INTERPEND;
9368 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
9370 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
9371 * the parse starting at 's', based on the subset that are valid in this
9372 * context input to this routine in 'valid_flags'. Advances s. Returns
9373 * TRUE if the input should be treated as a valid flag, so the next char
9374 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
9375 * first call on the current regex. This routine will set it to any
9376 * charset modifier found. The caller shouldn't change it. This way,
9377 * another charset modifier encountered in the parse can be detected as an
9378 * error, as we have decided to allow only one */
9381 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9383 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9384 if (isALNUM_lazy_if(*s, UTF)) {
9385 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9386 UTF ? SVf_UTF8 : 0);
9388 /* Pretend that it worked, so will continue processing before
9397 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
9398 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9399 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9400 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9401 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9402 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9403 case LOCALE_PAT_MOD:
9405 goto multiple_charsets;
9407 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9410 case UNICODE_PAT_MOD:
9412 goto multiple_charsets;
9414 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9417 case ASCII_RESTRICT_PAT_MOD:
9419 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9423 /* Error if previous modifier wasn't an 'a', but if it was, see
9424 * if, and accept, a second occurrence (only) */
9426 || get_regex_charset(*pmfl)
9427 != REGEX_ASCII_RESTRICTED_CHARSET)
9429 goto multiple_charsets;
9431 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9435 case DEPENDS_PAT_MOD:
9437 goto multiple_charsets;
9439 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9448 if (*charset != c) {
9449 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9451 else if (c == 'a') {
9452 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9455 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9458 /* Pretend that it worked, so will continue processing before dieing */
9464 S_scan_pat(pTHX_ char *start, I32 type)
9468 char *s = scan_str(start,!!PL_madskills,FALSE, PL_reg_state.re_reparsing);
9469 const char * const valid_flags =
9470 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9471 char charset = '\0'; /* character set modifier */
9476 PERL_ARGS_ASSERT_SCAN_PAT;
9478 /* this was only needed for the initial scan_str; set it to false
9479 * so that any (?{}) code blocks etc are parsed normally */
9480 PL_reg_state.re_reparsing = FALSE;
9482 const char * const delimiter = skipspace(start);
9486 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9487 : "Search pattern not terminated" ));
9490 pm = (PMOP*)newPMOP(type, 0);
9491 if (PL_multi_open == '?') {
9492 /* This is the only point in the code that sets PMf_ONCE: */
9493 pm->op_pmflags |= PMf_ONCE;
9495 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9496 allows us to restrict the list needed by reset to just the ??
9498 assert(type != OP_TRANS);
9500 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9503 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9506 elements = mg->mg_len / sizeof(PMOP**);
9507 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9508 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9509 mg->mg_len = elements * sizeof(PMOP**);
9510 PmopSTASH_set(pm,PL_curstash);
9517 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9518 * anon CV. False positives like qr/[(?{]/ are harmless */
9520 if (type == OP_QR) {
9522 char *e, *p = SvPV(PL_lex_stuff, len);
9524 for (; p < e; p++) {
9525 if (p[0] == '(' && p[1] == '?'
9526 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9528 pm->op_pmflags |= PMf_HAS_CV;
9532 pm->op_pmflags |= PMf_IS_QR;
9535 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9537 if (PL_madskills && modstart != s) {
9538 SV* tmptoken = newSVpvn(modstart, s - modstart);
9539 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9542 /* issue a warning if /c is specified,but /g is not */
9543 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9545 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9546 "Use of /c modifier is meaningless without /g" );
9549 PL_lex_op = (OP*)pm;
9550 pl_yylval.ival = OP_MATCH;
9555 S_scan_subst(pTHX_ char *start)
9562 char charset = '\0'; /* character set modifier */
9567 PERL_ARGS_ASSERT_SCAN_SUBST;
9569 pl_yylval.ival = OP_NULL;
9571 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9574 Perl_croak(aTHX_ "Substitution pattern not terminated");
9576 if (s[-1] == PL_multi_open)
9580 CURMAD('q', PL_thisopen);
9581 CURMAD('_', PL_thiswhite);
9582 CURMAD('E', PL_thisstuff);
9583 CURMAD('Q', PL_thisclose);
9584 PL_realtokenstart = s - SvPVX(PL_linestr);
9588 first_start = PL_multi_start;
9589 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9592 SvREFCNT_dec(PL_lex_stuff);
9593 PL_lex_stuff = NULL;
9595 Perl_croak(aTHX_ "Substitution replacement not terminated");
9597 PL_multi_start = first_start; /* so whole substitution is taken together */
9599 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9603 CURMAD('z', PL_thisopen);
9604 CURMAD('R', PL_thisstuff);
9605 CURMAD('Z', PL_thisclose);
9611 if (*s == EXEC_PAT_MOD) {
9615 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9624 curmad('m', newSVpvn(modstart, s - modstart));
9625 append_madprops(PL_thismad, (OP*)pm, 0);
9629 if ((pm->op_pmflags & PMf_CONTINUE)) {
9630 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9634 SV * const repl = newSVpvs("");
9637 pm->op_pmflags |= PMf_EVAL;
9640 sv_catpvs(repl, "eval ");
9642 sv_catpvs(repl, "do ");
9644 sv_catpvs(repl, "{");
9645 sv_catsv(repl, PL_sublex_info.repl);
9646 sv_catpvs(repl, "}");
9648 SvREFCNT_dec(PL_sublex_info.repl);
9649 PL_sublex_info.repl = repl;
9652 PL_lex_op = (OP*)pm;
9653 pl_yylval.ival = OP_SUBST;
9658 S_scan_trans(pTHX_ char *start)
9666 bool nondestruct = 0;
9671 PERL_ARGS_ASSERT_SCAN_TRANS;
9673 pl_yylval.ival = OP_NULL;
9675 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9677 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9679 if (s[-1] == PL_multi_open)
9683 CURMAD('q', PL_thisopen);
9684 CURMAD('_', PL_thiswhite);
9685 CURMAD('E', PL_thisstuff);
9686 CURMAD('Q', PL_thisclose);
9687 PL_realtokenstart = s - SvPVX(PL_linestr);
9691 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9694 SvREFCNT_dec(PL_lex_stuff);
9695 PL_lex_stuff = NULL;
9697 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9700 CURMAD('z', PL_thisopen);
9701 CURMAD('R', PL_thisstuff);
9702 CURMAD('Z', PL_thisclose);
9705 complement = del = squash = 0;
9712 complement = OPpTRANS_COMPLEMENT;
9715 del = OPpTRANS_DELETE;
9718 squash = OPpTRANS_SQUASH;
9730 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9731 o->op_private &= ~OPpTRANS_ALL;
9732 o->op_private |= del|squash|complement|
9733 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9734 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9737 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9742 curmad('m', newSVpvn(modstart, s - modstart));
9743 append_madprops(PL_thismad, o, 0);
9752 Takes a pointer to the first < in <<FOO.
9753 Returns a pointer to the byte following <<FOO.
9755 This function scans a heredoc, which involves different methods
9756 depending on whether we are in a string eval, quoted construct, etc.
9757 This is because PL_linestr could containing a single line of input, or
9758 a whole string being evalled, or the contents of the current quote-
9761 The two basic methods are:
9762 - Steal lines from the input stream
9763 - Scan the heredoc in PL_linestr and remove it therefrom
9765 In a file scope or filtered eval, the first method is used; in a
9766 string eval, the second.
9768 In a quote-like operator, we have to choose between the two,
9769 depending on where we can find a newline. We peek into outer lex-
9770 ing scopes until we find one with a newline in it. If we reach the
9771 outermost lexing scope and it is a file, we use the stream method.
9772 Otherwise it is treated as an eval.
9776 S_scan_heredoc(pTHX_ register char *s)
9779 I32 op_type = OP_SCALAR;
9786 const bool infile = PL_rsfp || PL_parser->filtered;
9787 LEXSHARED *shared = PL_parser->lex_shared;
9789 I32 stuffstart = s - SvPVX(PL_linestr);
9792 PL_realtokenstart = -1;
9795 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9798 d = PL_tokenbuf + 1;
9799 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9800 *PL_tokenbuf = '\n';
9802 while (SPACE_OR_TAB(*peek))
9804 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9807 s = delimcpy(d, e, s, PL_bufend, term, &len);
9809 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9815 /* <<\FOO is equivalent to <<'FOO' */
9819 if (!isALNUM_lazy_if(s,UTF))
9820 deprecate("bare << to mean <<\"\"");
9821 for (; isALNUM_lazy_if(s,UTF); s++) {
9826 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9827 Perl_croak(aTHX_ "Delimiter for here document is too long");
9830 len = d - PL_tokenbuf;
9834 tstart = PL_tokenbuf + 1;
9835 PL_thisclose = newSVpvn(tstart, len - 1);
9836 tstart = SvPVX(PL_linestr) + stuffstart;
9837 PL_thisopen = newSVpvn(tstart, s - tstart);
9838 stuffstart = s - SvPVX(PL_linestr);
9841 #ifndef PERL_STRICT_CR
9842 d = strchr(s, '\r');
9844 char * const olds = s;
9846 while (s < PL_bufend) {
9852 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9861 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9867 tstart = SvPVX(PL_linestr) + stuffstart;
9869 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9871 PL_thisstuff = newSVpvn(tstart, s - tstart);
9874 stuffstart = s - SvPVX(PL_linestr);
9877 tmpstr = newSV_type(SVt_PVIV);
9881 SvIV_set(tmpstr, -1);
9883 else if (term == '`') {
9884 op_type = OP_BACKTICK;
9885 SvIV_set(tmpstr, '\\');
9888 PL_multi_start = CopLINE(PL_curcop) + 1;
9889 PL_multi_open = PL_multi_close = '<';
9890 /* inside a string eval or quote-like operator */
9891 if (!infile || PL_lex_inwhat) {
9894 char * const olds = s;
9895 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9896 /* These two fields are not set until an inner lexing scope is
9897 entered. But we need them set here. */
9898 shared->ls_bufptr = s;
9899 shared->ls_linestr = PL_linestr;
9901 /* Look for a newline. If the current buffer does not have one,
9902 peek into the line buffer of the parent lexing scope, going
9903 up as many levels as necessary to find one with a newline
9906 while (!(s = (char *)memchr(
9907 (void *)shared->ls_bufptr, '\n',
9908 SvEND(shared->ls_linestr)-shared->ls_bufptr
9910 shared = shared->ls_prev;
9911 /* shared is only null if we have gone beyond the outermost
9912 lexing scope. In a file, we will have broken out of the
9913 loop in the previous iteration. In an eval, the string buf-
9914 fer ends with "\n;", so the while condition below will have
9915 evaluated to false. So shared can never be null. */
9917 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9918 most lexing scope. In a file, shared->ls_linestr at that
9919 level is just one line, so there is no body to steal. */
9920 if (infile && !shared->ls_prev) {
9926 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9929 linestr = shared->ls_linestr;
9930 bufend = SvEND(linestr);
9932 while (s < bufend &&
9933 (*s != '\n' || memNE(s,PL_tokenbuf,len)) ) {
9935 ++shared->herelines;
9940 sv_setpvn(tmpstr,d+1,s-d);
9944 sv_catpvn(PL_thisstuff, d + 1, s - d);
9946 PL_thisstuff = newSVpvn(d + 1, s - d);
9947 stuffstart = s - SvPVX(PL_linestr);
9951 /* the preceding stmt passes a newline */
9952 shared->herelines++;
9954 /* s now points to the newline after the heredoc terminator.
9955 d points to the newline before the body of the heredoc.
9958 /* We are going to modify linestr in place here, so set
9959 aside copies of the string if necessary for re-evals or
9961 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9962 check shared->re_eval_str. */
9963 if (shared->re_eval_start || shared->re_eval_str) {
9964 /* Set aside the rest of the regexp */
9965 if (!shared->re_eval_str)
9966 shared->re_eval_str =
9967 newSVpvn(shared->re_eval_start,
9968 bufend - shared->re_eval_start);
9969 shared->re_eval_start -= s-d;
9971 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9972 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9973 cx->blk_eval.cur_text == linestr)
9975 cx->blk_eval.cur_text = newSVsv(linestr);
9976 SvSCREAM_on(cx->blk_eval.cur_text);
9978 /* Copy everything from s onwards back to d. */
9979 Move(s,d,bufend-s + 1,char);
9980 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9981 /* Setting PL_bufend only applies when we have not dug deeper
9982 into other scopes, because sublex_done sets PL_bufend to
9983 SvEND(PL_linestr). */
9984 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9991 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9992 term = PL_tokenbuf[1];
9994 linestr_save = PL_linestr; /* must restore this afterwards */
9995 d = s; /* and this */
9996 PL_linestr = newSVpvs("");
9997 PL_bufend = SvPVX(PL_linestr);
10000 if (PL_madskills) {
10001 tstart = SvPVX(PL_linestr) + stuffstart;
10003 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10005 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10008 PL_bufptr = PL_bufend;
10009 CopLINE_set(PL_curcop,
10010 PL_multi_start + shared->herelines);
10011 if (!lex_next_chunk(LEX_NO_TERM)
10012 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
10013 SvREFCNT_dec(linestr_save);
10016 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10017 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
10018 lex_grow_linestr(SvCUR(PL_linestr) + 2);
10019 sv_catpvs(PL_linestr, "\n\0");
10023 stuffstart = s - SvPVX(PL_linestr);
10025 shared->herelines++;
10026 PL_last_lop = PL_last_uni = NULL;
10027 #ifndef PERL_STRICT_CR
10028 if (PL_bufend - PL_linestart >= 2) {
10029 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
10030 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
10032 PL_bufend[-2] = '\n';
10034 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
10036 else if (PL_bufend[-1] == '\r')
10037 PL_bufend[-1] = '\n';
10039 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
10040 PL_bufend[-1] = '\n';
10042 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
10043 SvREFCNT_dec(PL_linestr);
10044 PL_linestr = linestr_save;
10045 PL_linestart = SvPVX(linestr_save);
10046 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10051 sv_catsv(tmpstr,PL_linestr);
10055 PL_multi_end = CopLINE(PL_curcop);
10056 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
10057 SvPV_shrink_to_cur(tmpstr);
10060 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
10062 else if (PL_encoding)
10063 sv_recode_to_utf8(tmpstr, PL_encoding);
10065 PL_lex_stuff = tmpstr;
10066 pl_yylval.ival = op_type;
10070 SvREFCNT_dec(tmpstr);
10071 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
10072 missingterm(PL_tokenbuf + 1);
10075 /* scan_inputsymbol
10076 takes: current position in input buffer
10077 returns: new position in input buffer
10078 side-effects: pl_yylval and lex_op are set.
10083 <FH> read from filehandle
10084 <pkg::FH> read from package qualified filehandle
10085 <pkg'FH> read from package qualified filehandle
10086 <$fh> read from filehandle in $fh
10087 <*.h> filename glob
10092 S_scan_inputsymbol(pTHX_ char *start)
10095 char *s = start; /* current position in buffer */
10098 char *d = PL_tokenbuf; /* start of temp holding space */
10099 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
10101 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
10103 end = strchr(s, '\n');
10106 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
10108 /* die if we didn't have space for the contents of the <>,
10109 or if it didn't end, or if we see a newline
10112 if (len >= (I32)sizeof PL_tokenbuf)
10113 Perl_croak(aTHX_ "Excessively long <> operator");
10115 Perl_croak(aTHX_ "Unterminated <> operator");
10120 Remember, only scalar variables are interpreted as filehandles by
10121 this code. Anything more complex (e.g., <$fh{$num}>) will be
10122 treated as a glob() call.
10123 This code makes use of the fact that except for the $ at the front,
10124 a scalar variable and a filehandle look the same.
10126 if (*d == '$' && d[1]) d++;
10128 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
10129 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
10130 d += UTF ? UTF8SKIP(d) : 1;
10132 /* If we've tried to read what we allow filehandles to look like, and
10133 there's still text left, then it must be a glob() and not a getline.
10134 Use scan_str to pull out the stuff between the <> and treat it
10135 as nothing more than a string.
10138 if (d - PL_tokenbuf != len) {
10139 pl_yylval.ival = OP_GLOB;
10140 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
10142 Perl_croak(aTHX_ "Glob not terminated");
10146 bool readline_overriden = FALSE;
10149 /* we're in a filehandle read situation */
10152 /* turn <> into <ARGV> */
10154 Copy("ARGV",d,5,char);
10156 /* Check whether readline() is overriden */
10157 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
10159 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
10161 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
10162 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
10163 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
10164 readline_overriden = TRUE;
10166 /* if <$fh>, create the ops to turn the variable into a
10170 /* try to find it in the pad for this block, otherwise find
10171 add symbol table ops
10173 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
10174 if (tmp != NOT_IN_PAD) {
10175 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
10176 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
10177 HEK * const stashname = HvNAME_HEK(stash);
10178 SV * const sym = sv_2mortal(newSVhek(stashname));
10179 sv_catpvs(sym, "::");
10180 sv_catpv(sym, d+1);
10185 OP * const o = newOP(OP_PADSV, 0);
10187 PL_lex_op = readline_overriden
10188 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10189 op_append_elem(OP_LIST, o,
10190 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
10191 : (OP*)newUNOP(OP_READLINE, 0, o);
10200 ? (GV_ADDMULTI | GV_ADDINEVAL)
10201 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
10203 PL_lex_op = readline_overriden
10204 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10205 op_append_elem(OP_LIST,
10206 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
10207 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10208 : (OP*)newUNOP(OP_READLINE, 0,
10209 newUNOP(OP_RV2SV, 0,
10210 newGVOP(OP_GV, 0, gv)));
10212 if (!readline_overriden)
10213 PL_lex_op->op_flags |= OPf_SPECIAL;
10214 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
10215 pl_yylval.ival = OP_NULL;
10218 /* If it's none of the above, it must be a literal filehandle
10219 (<Foo::BAR> or <FOO>) so build a simple readline OP */
10221 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
10222 PL_lex_op = readline_overriden
10223 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10224 op_append_elem(OP_LIST,
10225 newGVOP(OP_GV, 0, gv),
10226 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10227 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
10228 pl_yylval.ival = OP_NULL;
10237 takes: start position in buffer
10238 keep_quoted preserve \ on the embedded delimiter(s)
10239 keep_delims preserve the delimiters around the string
10240 re_reparse compiling a run-time /(?{})/:
10241 collapse // to /, and skip encoding src
10242 returns: position to continue reading from buffer
10243 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
10244 updates the read buffer.
10246 This subroutine pulls a string out of the input. It is called for:
10247 q single quotes q(literal text)
10248 ' single quotes 'literal text'
10249 qq double quotes qq(interpolate $here please)
10250 " double quotes "interpolate $here please"
10251 qx backticks qx(/bin/ls -l)
10252 ` backticks `/bin/ls -l`
10253 qw quote words @EXPORT_OK = qw( func() $spam )
10254 m// regexp match m/this/
10255 s/// regexp substitute s/this/that/
10256 tr/// string transliterate tr/this/that/
10257 y/// string transliterate y/this/that/
10258 ($*@) sub prototypes sub foo ($)
10259 (stuff) sub attr parameters sub foo : attr(stuff)
10260 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
10262 In most of these cases (all but <>, patterns and transliterate)
10263 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
10264 calls scan_str(). s/// makes yylex() call scan_subst() which calls
10265 scan_str(). tr/// and y/// make yylex() call scan_trans() which
10268 It skips whitespace before the string starts, and treats the first
10269 character as the delimiter. If the delimiter is one of ([{< then
10270 the corresponding "close" character )]}> is used as the closing
10271 delimiter. It allows quoting of delimiters, and if the string has
10272 balanced delimiters ([{<>}]) it allows nesting.
10274 On success, the SV with the resulting string is put into lex_stuff or,
10275 if that is already non-NULL, into lex_repl. The second case occurs only
10276 when parsing the RHS of the special constructs s/// and tr/// (y///).
10277 For convenience, the terminating delimiter character is stuffed into
10282 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims, int re_reparse)
10285 SV *sv; /* scalar value: string */
10286 const char *tmps; /* temp string, used for delimiter matching */
10287 char *s = start; /* current position in the buffer */
10288 char term; /* terminating character */
10289 char *to; /* current position in the sv's data */
10290 I32 brackets = 1; /* bracket nesting level */
10291 bool has_utf8 = FALSE; /* is there any utf8 content? */
10292 I32 termcode; /* terminating char. code */
10293 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
10294 STRLEN termlen; /* length of terminating string */
10295 int last_off = 0; /* last position for nesting bracket */
10301 PERL_ARGS_ASSERT_SCAN_STR;
10303 /* skip space before the delimiter */
10309 if (PL_realtokenstart >= 0) {
10310 stuffstart = PL_realtokenstart;
10311 PL_realtokenstart = -1;
10314 stuffstart = start - SvPVX(PL_linestr);
10316 /* mark where we are, in case we need to report errors */
10319 /* after skipping whitespace, the next character is the terminator */
10322 termcode = termstr[0] = term;
10326 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
10327 Copy(s, termstr, termlen, U8);
10328 if (!UTF8_IS_INVARIANT(term))
10332 /* mark where we are */
10333 PL_multi_start = CopLINE(PL_curcop);
10334 PL_multi_open = term;
10336 /* find corresponding closing delimiter */
10337 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
10338 termcode = termstr[0] = term = tmps[5];
10340 PL_multi_close = term;
10342 /* create a new SV to hold the contents. 79 is the SV's initial length.
10343 What a random number. */
10344 sv = newSV_type(SVt_PVIV);
10346 SvIV_set(sv, termcode);
10347 (void)SvPOK_only(sv); /* validate pointer */
10349 /* move past delimiter and try to read a complete string */
10351 sv_catpvn(sv, s, termlen);
10354 tstart = SvPVX(PL_linestr) + stuffstart;
10355 if (PL_madskills && !PL_thisopen && !keep_delims) {
10356 PL_thisopen = newSVpvn(tstart, s - tstart);
10357 stuffstart = s - SvPVX(PL_linestr);
10361 if (PL_encoding && !UTF && !re_reparse) {
10365 int offset = s - SvPVX_const(PL_linestr);
10366 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
10367 &offset, (char*)termstr, termlen);
10368 const char * const ns = SvPVX_const(PL_linestr) + offset;
10369 char * const svlast = SvEND(sv) - 1;
10371 for (; s < ns; s++) {
10372 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10373 COPLINE_INC_WITH_HERELINES;
10376 goto read_more_line;
10378 /* handle quoted delimiters */
10379 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10381 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10383 if ((svlast-1 - t) % 2) {
10384 if (!keep_quoted) {
10385 *(svlast-1) = term;
10387 SvCUR_set(sv, SvCUR(sv) - 1);
10392 if (PL_multi_open == PL_multi_close) {
10398 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10399 /* At here, all closes are "was quoted" one,
10400 so we don't check PL_multi_close. */
10402 if (!keep_quoted && *(t+1) == PL_multi_open)
10407 else if (*t == PL_multi_open)
10415 SvCUR_set(sv, w - SvPVX_const(sv));
10417 last_off = w - SvPVX(sv);
10418 if (--brackets <= 0)
10423 if (!keep_delims) {
10424 SvCUR_set(sv, SvCUR(sv) - 1);
10430 /* extend sv if need be */
10431 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10432 /* set 'to' to the next character in the sv's string */
10433 to = SvPVX(sv)+SvCUR(sv);
10435 /* if open delimiter is the close delimiter read unbridle */
10436 if (PL_multi_open == PL_multi_close) {
10437 for (; s < PL_bufend; s++,to++) {
10438 /* embedded newlines increment the current line number */
10439 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10440 COPLINE_INC_WITH_HERELINES;
10441 /* handle quoted delimiters */
10442 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10445 || (re_reparse && s[1] == '\\'))
10448 /* any other quotes are simply copied straight through */
10452 /* terminate when run out of buffer (the for() condition), or
10453 have found the terminator */
10454 else if (*s == term) {
10457 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10460 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10466 /* if the terminator isn't the same as the start character (e.g.,
10467 matched brackets), we have to allow more in the quoting, and
10468 be prepared for nested brackets.
10471 /* read until we run out of string, or we find the terminator */
10472 for (; s < PL_bufend; s++,to++) {
10473 /* embedded newlines increment the line count */
10474 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10475 COPLINE_INC_WITH_HERELINES;
10476 /* backslashes can escape the open or closing characters */
10477 if (*s == '\\' && s+1 < PL_bufend) {
10478 if (!keep_quoted &&
10479 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10484 /* allow nested opens and closes */
10485 else if (*s == PL_multi_close && --brackets <= 0)
10487 else if (*s == PL_multi_open)
10489 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10494 /* terminate the copied string and update the sv's end-of-string */
10496 SvCUR_set(sv, to - SvPVX_const(sv));
10499 * this next chunk reads more into the buffer if we're not done yet
10503 break; /* handle case where we are done yet :-) */
10505 #ifndef PERL_STRICT_CR
10506 if (to - SvPVX_const(sv) >= 2) {
10507 if ((to[-2] == '\r' && to[-1] == '\n') ||
10508 (to[-2] == '\n' && to[-1] == '\r'))
10512 SvCUR_set(sv, to - SvPVX_const(sv));
10514 else if (to[-1] == '\r')
10517 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10522 /* if we're out of file, or a read fails, bail and reset the current
10523 line marker so we can report where the unterminated string began
10526 if (PL_madskills) {
10527 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10529 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10531 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10534 COPLINE_INC_WITH_HERELINES;
10535 PL_bufptr = PL_bufend;
10536 if (!lex_next_chunk(0)) {
10538 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10547 /* at this point, we have successfully read the delimited string */
10549 if (!PL_encoding || UTF || re_reparse) {
10551 if (PL_madskills) {
10552 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10553 const int len = s - tstart;
10555 sv_catpvn(PL_thisstuff, tstart, len);
10557 PL_thisstuff = newSVpvn(tstart, len);
10558 if (!PL_thisclose && !keep_delims)
10559 PL_thisclose = newSVpvn(s,termlen);
10564 sv_catpvn(sv, s, termlen);
10569 if (PL_madskills) {
10570 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10571 const int len = s - tstart - termlen;
10573 sv_catpvn(PL_thisstuff, tstart, len);
10575 PL_thisstuff = newSVpvn(tstart, len);
10576 if (!PL_thisclose && !keep_delims)
10577 PL_thisclose = newSVpvn(s - termlen,termlen);
10581 if (has_utf8 || (PL_encoding && !re_reparse))
10584 PL_multi_end = CopLINE(PL_curcop);
10586 /* if we allocated too much space, give some back */
10587 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10588 SvLEN_set(sv, SvCUR(sv) + 1);
10589 SvPV_renew(sv, SvLEN(sv));
10592 /* decide whether this is the first or second quoted string we've read
10597 PL_sublex_info.repl = sv;
10605 takes: pointer to position in buffer
10606 returns: pointer to new position in buffer
10607 side-effects: builds ops for the constant in pl_yylval.op
10609 Read a number in any of the formats that Perl accepts:
10611 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10612 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10615 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10617 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10620 If it reads a number without a decimal point or an exponent, it will
10621 try converting the number to an integer and see if it can do so
10622 without loss of precision.
10626 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10629 const char *s = start; /* current position in buffer */
10630 char *d; /* destination in temp buffer */
10631 char *e; /* end of temp buffer */
10632 NV nv; /* number read, as a double */
10633 SV *sv = NULL; /* place to put the converted number */
10634 bool floatit; /* boolean: int or float? */
10635 const char *lastub = NULL; /* position of last underbar */
10636 static char const number_too_long[] = "Number too long";
10638 PERL_ARGS_ASSERT_SCAN_NUM;
10640 /* We use the first character to decide what type of number this is */
10644 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10646 /* if it starts with a 0, it could be an octal number, a decimal in
10647 0.13 disguise, or a hexadecimal number, or a binary number. */
10651 u holds the "number so far"
10652 shift the power of 2 of the base
10653 (hex == 4, octal == 3, binary == 1)
10654 overflowed was the number more than we can hold?
10656 Shift is used when we add a digit. It also serves as an "are
10657 we in octal/hex/binary?" indicator to disallow hex characters
10658 when in octal mode.
10663 bool overflowed = FALSE;
10664 bool just_zero = TRUE; /* just plain 0 or binary number? */
10665 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10666 static const char* const bases[5] =
10667 { "", "binary", "", "octal", "hexadecimal" };
10668 static const char* const Bases[5] =
10669 { "", "Binary", "", "Octal", "Hexadecimal" };
10670 static const char* const maxima[5] =
10672 "0b11111111111111111111111111111111",
10676 const char *base, *Base, *max;
10678 /* check for hex */
10679 if (s[1] == 'x' || s[1] == 'X') {
10683 } else if (s[1] == 'b' || s[1] == 'B') {
10688 /* check for a decimal in disguise */
10689 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10691 /* so it must be octal */
10698 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10699 "Misplaced _ in number");
10703 base = bases[shift];
10704 Base = Bases[shift];
10705 max = maxima[shift];
10707 /* read the rest of the number */
10709 /* x is used in the overflow test,
10710 b is the digit we're adding on. */
10715 /* if we don't mention it, we're done */
10719 /* _ are ignored -- but warned about if consecutive */
10721 if (lastub && s == lastub + 1)
10722 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10723 "Misplaced _ in number");
10727 /* 8 and 9 are not octal */
10728 case '8': case '9':
10730 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10734 case '2': case '3': case '4':
10735 case '5': case '6': case '7':
10737 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10740 case '0': case '1':
10741 b = *s++ & 15; /* ASCII digit -> value of digit */
10745 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10746 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10747 /* make sure they said 0x */
10750 b = (*s++ & 7) + 9;
10752 /* Prepare to put the digit we have onto the end
10753 of the number so far. We check for overflows.
10759 x = u << shift; /* make room for the digit */
10761 if ((x >> shift) != u
10762 && !(PL_hints & HINT_NEW_BINARY)) {
10765 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10766 "Integer overflow in %s number",
10769 u = x | b; /* add the digit to the end */
10772 n *= nvshift[shift];
10773 /* If an NV has not enough bits in its
10774 * mantissa to represent an UV this summing of
10775 * small low-order numbers is a waste of time
10776 * (because the NV cannot preserve the
10777 * low-order bits anyway): we could just
10778 * remember when did we overflow and in the
10779 * end just multiply n by the right
10787 /* if we get here, we had success: make a scalar value from
10792 /* final misplaced underbar check */
10793 if (s[-1] == '_') {
10794 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10798 if (n > 4294967295.0)
10799 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10800 "%s number > %s non-portable",
10806 if (u > 0xffffffff)
10807 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10808 "%s number > %s non-portable",
10813 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10814 sv = new_constant(start, s - start, "integer",
10815 sv, NULL, NULL, 0);
10816 else if (PL_hints & HINT_NEW_BINARY)
10817 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10822 handle decimal numbers.
10823 we're also sent here when we read a 0 as the first digit
10825 case '1': case '2': case '3': case '4': case '5':
10826 case '6': case '7': case '8': case '9': case '.':
10829 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10832 /* read next group of digits and _ and copy into d */
10833 while (isDIGIT(*s) || *s == '_') {
10834 /* skip underscores, checking for misplaced ones
10838 if (lastub && s == lastub + 1)
10839 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10840 "Misplaced _ in number");
10844 /* check for end of fixed-length buffer */
10846 Perl_croak(aTHX_ number_too_long);
10847 /* if we're ok, copy the character */
10852 /* final misplaced underbar check */
10853 if (lastub && s == lastub + 1) {
10854 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10857 /* read a decimal portion if there is one. avoid
10858 3..5 being interpreted as the number 3. followed
10861 if (*s == '.' && s[1] != '.') {
10866 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10867 "Misplaced _ in number");
10871 /* copy, ignoring underbars, until we run out of digits.
10873 for (; isDIGIT(*s) || *s == '_'; s++) {
10874 /* fixed length buffer check */
10876 Perl_croak(aTHX_ number_too_long);
10878 if (lastub && s == lastub + 1)
10879 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10880 "Misplaced _ in number");
10886 /* fractional part ending in underbar? */
10887 if (s[-1] == '_') {
10888 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10889 "Misplaced _ in number");
10891 if (*s == '.' && isDIGIT(s[1])) {
10892 /* oops, it's really a v-string, but without the "v" */
10898 /* read exponent part, if present */
10899 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10903 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10904 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10906 /* stray preinitial _ */
10908 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10909 "Misplaced _ in number");
10913 /* allow positive or negative exponent */
10914 if (*s == '+' || *s == '-')
10917 /* stray initial _ */
10919 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10920 "Misplaced _ in number");
10924 /* read digits of exponent */
10925 while (isDIGIT(*s) || *s == '_') {
10928 Perl_croak(aTHX_ number_too_long);
10932 if (((lastub && s == lastub + 1) ||
10933 (!isDIGIT(s[1]) && s[1] != '_')))
10934 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10935 "Misplaced _ in number");
10943 We try to do an integer conversion first if no characters
10944 indicating "float" have been found.
10949 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10951 if (flags == IS_NUMBER_IN_UV) {
10953 sv = newSViv(uv); /* Prefer IVs over UVs. */
10956 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10957 if (uv <= (UV) IV_MIN)
10958 sv = newSViv(-(IV)uv);
10965 /* terminate the string */
10967 nv = Atof(PL_tokenbuf);
10972 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10973 const char *const key = floatit ? "float" : "integer";
10974 const STRLEN keylen = floatit ? 5 : 7;
10975 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10976 key, keylen, sv, NULL, NULL, 0);
10980 /* if it starts with a v, it could be a v-string */
10983 sv = newSV(5); /* preallocate storage space */
10984 s = scan_vstring(s, PL_bufend, sv);
10988 /* make the op for the constant and return */
10991 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10993 lvalp->opval = NULL;
10999 S_scan_formline(pTHX_ register char *s)
11004 SV * const stuff = newSVpvs("");
11005 bool needargs = FALSE;
11006 bool eofmt = FALSE;
11008 char *tokenstart = s;
11009 SV* savewhite = NULL;
11011 if (PL_madskills) {
11012 savewhite = PL_thiswhite;
11017 PERL_ARGS_ASSERT_SCAN_FORMLINE;
11019 while (!needargs) {
11022 #ifdef PERL_STRICT_CR
11023 while (SPACE_OR_TAB(*t))
11026 while (SPACE_OR_TAB(*t) || *t == '\r')
11029 if (*t == '\n' || t == PL_bufend) {
11034 eol = (char *) memchr(s,'\n',PL_bufend-s);
11038 for (t = s; t < eol; t++) {
11039 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
11041 goto enough; /* ~~ must be first line in formline */
11043 if (*t == '@' || *t == '^')
11047 sv_catpvn(stuff, s, eol-s);
11048 #ifndef PERL_STRICT_CR
11049 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
11050 char *end = SvPVX(stuff) + SvCUR(stuff);
11053 SvCUR_set(stuff, SvCUR(stuff) - 1);
11061 if ((PL_rsfp || PL_parser->filtered)
11062 && PL_parser->form_lex_state == LEX_NORMAL) {
11065 if (PL_madskills) {
11067 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
11069 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
11072 PL_bufptr = PL_bufend;
11073 COPLINE_INC_WITH_HERELINES;
11074 got_some = lex_next_chunk(0);
11075 CopLINE_dec(PL_curcop);
11078 tokenstart = PL_bufptr;
11086 if (!SvCUR(stuff) || needargs)
11087 PL_lex_state = PL_parser->form_lex_state;
11088 if (SvCUR(stuff)) {
11089 PL_expect = XSTATE;
11091 start_force(PL_curforce);
11092 NEXTVAL_NEXTTOKE.ival = 0;
11093 force_next(FORMLBRACK);
11096 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
11098 else if (PL_encoding)
11099 sv_recode_to_utf8(stuff, PL_encoding);
11101 start_force(PL_curforce);
11102 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
11106 SvREFCNT_dec(stuff);
11108 PL_lex_formbrack = 0;
11111 if (PL_madskills) {
11113 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
11115 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
11116 PL_thiswhite = savewhite;
11123 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
11126 const I32 oldsavestack_ix = PL_savestack_ix;
11127 CV* const outsidecv = PL_compcv;
11129 SAVEI32(PL_subline);
11130 save_item(PL_subname);
11131 SAVESPTR(PL_compcv);
11133 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
11134 CvFLAGS(PL_compcv) |= flags;
11136 PL_subline = CopLINE(PL_curcop);
11137 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
11138 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
11139 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
11140 if (outsidecv && CvPADLIST(outsidecv))
11141 CvPADLIST(PL_compcv)->xpadl_outid =
11142 PadlistNAMES(CvPADLIST(outsidecv));
11144 return oldsavestack_ix;
11148 #pragma segment Perl_yylex
11151 S_yywarn(pTHX_ const char *const s, U32 flags)
11155 PERL_ARGS_ASSERT_YYWARN;
11157 PL_in_eval |= EVAL_WARNONLY;
11158 yyerror_pv(s, flags);
11159 PL_in_eval &= ~EVAL_WARNONLY;
11164 Perl_yyerror(pTHX_ const char *const s)
11166 PERL_ARGS_ASSERT_YYERROR;
11167 return yyerror_pvn(s, strlen(s), 0);
11171 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
11173 PERL_ARGS_ASSERT_YYERROR_PV;
11174 return yyerror_pvn(s, strlen(s), flags);
11178 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
11181 const char *context = NULL;
11184 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
11185 int yychar = PL_parser->yychar;
11187 PERL_ARGS_ASSERT_YYERROR_PVN;
11189 if (!yychar || (yychar == ';' && !PL_rsfp))
11190 sv_catpvs(where_sv, "at EOF");
11191 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
11192 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
11193 PL_oldbufptr != PL_bufptr) {
11196 The code below is removed for NetWare because it abends/crashes on NetWare
11197 when the script has error such as not having the closing quotes like:
11198 if ($var eq "value)
11199 Checking of white spaces is anyway done in NetWare code.
11202 while (isSPACE(*PL_oldoldbufptr))
11205 context = PL_oldoldbufptr;
11206 contlen = PL_bufptr - PL_oldoldbufptr;
11208 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
11209 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
11212 The code below is removed for NetWare because it abends/crashes on NetWare
11213 when the script has error such as not having the closing quotes like:
11214 if ($var eq "value)
11215 Checking of white spaces is anyway done in NetWare code.
11218 while (isSPACE(*PL_oldbufptr))
11221 context = PL_oldbufptr;
11222 contlen = PL_bufptr - PL_oldbufptr;
11224 else if (yychar > 255)
11225 sv_catpvs(where_sv, "next token ???");
11226 else if (yychar == -2) { /* YYEMPTY */
11227 if (PL_lex_state == LEX_NORMAL ||
11228 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11229 sv_catpvs(where_sv, "at end of line");
11230 else if (PL_lex_inpat)
11231 sv_catpvs(where_sv, "within pattern");
11233 sv_catpvs(where_sv, "within string");
11236 sv_catpvs(where_sv, "next char ");
11238 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11239 else if (isPRINT_LC(yychar)) {
11240 const char string = yychar;
11241 sv_catpvn(where_sv, &string, 1);
11244 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11246 msg = newSVpvn_flags(s, len, (flags & SVf_UTF8) | SVs_TEMP);
11247 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11248 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
11250 Perl_sv_catpvf(aTHX_ msg, "near \"%"SVf"\"\n",
11251 SVfARG(newSVpvn_flags(context, contlen,
11252 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
11254 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11255 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11256 Perl_sv_catpvf(aTHX_ msg,
11257 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11258 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11261 if (PL_in_eval & EVAL_WARNONLY) {
11262 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11266 if (PL_error_count >= 10) {
11267 if (PL_in_eval && SvCUR(ERRSV))
11268 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11269 SVfARG(ERRSV), OutCopFILE(PL_curcop));
11271 Perl_croak(aTHX_ "%s has too many errors.\n",
11272 OutCopFILE(PL_curcop));
11275 PL_in_my_stash = NULL;
11279 #pragma segment Main
11283 S_swallow_bom(pTHX_ U8 *s)
11286 const STRLEN slen = SvCUR(PL_linestr);
11288 PERL_ARGS_ASSERT_SWALLOW_BOM;
11292 if (s[1] == 0xFE) {
11293 /* UTF-16 little-endian? (or UTF-32LE?) */
11294 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11295 /* diag_listed_as: Unsupported script encoding %s */
11296 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11297 #ifndef PERL_NO_UTF16_FILTER
11298 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11300 if (PL_bufend > (char*)s) {
11301 s = add_utf16_textfilter(s, TRUE);
11304 /* diag_listed_as: Unsupported script encoding %s */
11305 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11310 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11311 #ifndef PERL_NO_UTF16_FILTER
11312 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11314 if (PL_bufend > (char *)s) {
11315 s = add_utf16_textfilter(s, FALSE);
11318 /* diag_listed_as: Unsupported script encoding %s */
11319 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11324 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
11325 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11326 s += 3; /* UTF-8 */
11332 if (s[2] == 0xFE && s[3] == 0xFF) {
11333 /* UTF-32 big-endian */
11334 /* diag_listed_as: Unsupported script encoding %s */
11335 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11338 else if (s[2] == 0 && s[3] != 0) {
11341 * are a good indicator of UTF-16BE. */
11342 #ifndef PERL_NO_UTF16_FILTER
11343 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11344 s = add_utf16_textfilter(s, FALSE);
11346 /* diag_listed_as: Unsupported script encoding %s */
11347 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11353 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
11354 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11355 s += 4; /* UTF-8 */
11361 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11364 * are a good indicator of UTF-16LE. */
11365 #ifndef PERL_NO_UTF16_FILTER
11366 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11367 s = add_utf16_textfilter(s, TRUE);
11369 /* diag_listed_as: Unsupported script encoding %s */
11370 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11378 #ifndef PERL_NO_UTF16_FILTER
11380 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11383 SV *const filter = FILTER_DATA(idx);
11384 /* We re-use this each time round, throwing the contents away before we
11386 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11387 SV *const utf8_buffer = filter;
11388 IV status = IoPAGE(filter);
11389 const bool reverse = cBOOL(IoLINES(filter));
11392 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11394 /* As we're automatically added, at the lowest level, and hence only called
11395 from this file, we can be sure that we're not called in block mode. Hence
11396 don't bother writing code to deal with block mode. */
11398 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11401 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11403 DEBUG_P(PerlIO_printf(Perl_debug_log,
11404 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11405 FPTR2DPTR(void *, S_utf16_textfilter),
11406 reverse ? 'l' : 'b', idx, maxlen, status,
11407 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11414 /* First, look in our buffer of existing UTF-8 data: */
11415 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11419 } else if (status == 0) {
11421 IoPAGE(filter) = 0;
11422 nl = SvEND(utf8_buffer);
11425 STRLEN got = nl - SvPVX(utf8_buffer);
11426 /* Did we have anything to append? */
11428 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11429 /* Everything else in this code works just fine if SVp_POK isn't
11430 set. This, however, needs it, and we need it to work, else
11431 we loop infinitely because the buffer is never consumed. */
11432 sv_chop(utf8_buffer, nl);
11436 /* OK, not a complete line there, so need to read some more UTF-16.
11437 Read an extra octect if the buffer currently has an odd number. */
11441 if (SvCUR(utf16_buffer) >= 2) {
11442 /* Location of the high octet of the last complete code point.
11443 Gosh, UTF-16 is a pain. All the benefits of variable length,
11444 *coupled* with all the benefits of partial reads and
11446 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11447 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11449 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11453 /* We have the first half of a surrogate. Read more. */
11454 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11457 status = FILTER_READ(idx + 1, utf16_buffer,
11458 160 + (SvCUR(utf16_buffer) & 1));
11459 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11460 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11463 IoPAGE(filter) = status;
11468 chars = SvCUR(utf16_buffer) >> 1;
11469 have = SvCUR(utf8_buffer);
11470 SvGROW(utf8_buffer, have + chars * 3 + 1);
11473 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11474 (U8*)SvPVX_const(utf8_buffer) + have,
11475 chars * 2, &newlen);
11477 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11478 (U8*)SvPVX_const(utf8_buffer) + have,
11479 chars * 2, &newlen);
11481 SvCUR_set(utf8_buffer, have + newlen);
11484 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11485 it's private to us, and utf16_to_utf8{,reversed} take a
11486 (pointer,length) pair, rather than a NUL-terminated string. */
11487 if(SvCUR(utf16_buffer) & 1) {
11488 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11489 SvCUR_set(utf16_buffer, 1);
11491 SvCUR_set(utf16_buffer, 0);
11494 DEBUG_P(PerlIO_printf(Perl_debug_log,
11495 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11497 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11498 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11503 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11505 SV *filter = filter_add(S_utf16_textfilter, NULL);
11507 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11509 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11510 sv_setpvs(filter, "");
11511 IoLINES(filter) = reversed;
11512 IoPAGE(filter) = 1; /* Not EOF */
11514 /* Sadly, we have to return a valid pointer, come what may, so we have to
11515 ignore any error return from this. */
11516 SvCUR_set(PL_linestr, 0);
11517 if (FILTER_READ(0, PL_linestr, 0)) {
11518 SvUTF8_on(PL_linestr);
11520 SvUTF8_on(PL_linestr);
11522 PL_bufend = SvEND(PL_linestr);
11523 return (U8*)SvPVX(PL_linestr);
11528 Returns a pointer to the next character after the parsed
11529 vstring, as well as updating the passed in sv.
11531 Function must be called like
11534 s = scan_vstring(s,e,sv);
11536 where s and e are the start and end of the string.
11537 The sv should already be large enough to store the vstring
11538 passed in, for performance reasons.
11543 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11546 const char *pos = s;
11547 const char *start = s;
11549 PERL_ARGS_ASSERT_SCAN_VSTRING;
11551 if (*pos == 'v') pos++; /* get past 'v' */
11552 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11554 if ( *pos != '.') {
11555 /* this may not be a v-string if followed by => */
11556 const char *next = pos;
11557 while (next < e && isSPACE(*next))
11559 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11560 /* return string not v-string */
11561 sv_setpvn(sv,(char *)s,pos-s);
11562 return (char *)pos;
11566 if (!isALPHA(*pos)) {
11567 U8 tmpbuf[UTF8_MAXBYTES+1];
11570 s++; /* get past 'v' */
11575 /* this is atoi() that tolerates underscores */
11578 const char *end = pos;
11580 while (--end >= s) {
11582 const UV orev = rev;
11583 rev += (*end - '0') * mult;
11586 /* diag_listed_as: Integer overflow in %s number */
11587 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11588 "Integer overflow in decimal number");
11592 if (rev > 0x7FFFFFFF)
11593 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11595 /* Append native character for the rev point */
11596 tmpend = uvchr_to_utf8(tmpbuf, rev);
11597 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11598 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11600 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11606 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11610 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11617 Perl_keyword_plugin_standard(pTHX_
11618 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11620 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11621 PERL_UNUSED_CONTEXT;
11622 PERL_UNUSED_ARG(keyword_ptr);
11623 PERL_UNUSED_ARG(keyword_len);
11624 PERL_UNUSED_ARG(op_ptr);
11625 return KEYWORD_PLUGIN_DECLINE;
11628 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11630 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11632 SAVEI32(PL_lex_brackets);
11633 if (PL_lex_brackets > 100)
11634 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11635 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11636 SAVEI32(PL_lex_allbrackets);
11637 PL_lex_allbrackets = 0;
11638 SAVEI8(PL_lex_fakeeof);
11639 PL_lex_fakeeof = (U8)fakeeof;
11640 if(yyparse(gramtype) && !PL_parser->error_count)
11641 qerror(Perl_mess(aTHX_ "Parse error"));
11644 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11646 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11650 SAVEVPTR(PL_eval_root);
11651 PL_eval_root = NULL;
11652 parse_recdescent(gramtype, fakeeof);
11658 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11660 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11663 if (flags & ~PARSE_OPTIONAL)
11664 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11665 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11666 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11667 if (!PL_parser->error_count)
11668 qerror(Perl_mess(aTHX_ "Parse error"));
11669 exprop = newOP(OP_NULL, 0);
11675 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11677 Parse a Perl arithmetic expression. This may contain operators of precedence
11678 down to the bit shift operators. The expression must be followed (and thus
11679 terminated) either by a comparison or lower-precedence operator or by
11680 something that would normally terminate an expression such as semicolon.
11681 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11682 otherwise it is mandatory. It is up to the caller to ensure that the
11683 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11684 the source of the code to be parsed and the lexical context for the
11687 The op tree representing the expression is returned. If an optional
11688 expression is absent, a null pointer is returned, otherwise the pointer
11691 If an error occurs in parsing or compilation, in most cases a valid op
11692 tree is returned anyway. The error is reflected in the parser state,
11693 normally resulting in a single exception at the top level of parsing
11694 which covers all the compilation errors that occurred. Some compilation
11695 errors, however, will throw an exception immediately.
11701 Perl_parse_arithexpr(pTHX_ U32 flags)
11703 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11707 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11709 Parse a Perl term expression. This may contain operators of precedence
11710 down to the assignment operators. The expression must be followed (and thus
11711 terminated) either by a comma or lower-precedence operator or by
11712 something that would normally terminate an expression such as semicolon.
11713 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11714 otherwise it is mandatory. It is up to the caller to ensure that the
11715 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11716 the source of the code to be parsed and the lexical context for the
11719 The op tree representing the expression is returned. If an optional
11720 expression is absent, a null pointer is returned, otherwise the pointer
11723 If an error occurs in parsing or compilation, in most cases a valid op
11724 tree is returned anyway. The error is reflected in the parser state,
11725 normally resulting in a single exception at the top level of parsing
11726 which covers all the compilation errors that occurred. Some compilation
11727 errors, however, will throw an exception immediately.
11733 Perl_parse_termexpr(pTHX_ U32 flags)
11735 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11739 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11741 Parse a Perl list expression. This may contain operators of precedence
11742 down to the comma operator. The expression must be followed (and thus
11743 terminated) either by a low-precedence logic operator such as C<or> or by
11744 something that would normally terminate an expression such as semicolon.
11745 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11746 otherwise it is mandatory. It is up to the caller to ensure that the
11747 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11748 the source of the code to be parsed and the lexical context for the
11751 The op tree representing the expression is returned. If an optional
11752 expression is absent, a null pointer is returned, otherwise the pointer
11755 If an error occurs in parsing or compilation, in most cases a valid op
11756 tree is returned anyway. The error is reflected in the parser state,
11757 normally resulting in a single exception at the top level of parsing
11758 which covers all the compilation errors that occurred. Some compilation
11759 errors, however, will throw an exception immediately.
11765 Perl_parse_listexpr(pTHX_ U32 flags)
11767 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11771 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11773 Parse a single complete Perl expression. This allows the full
11774 expression grammar, including the lowest-precedence operators such
11775 as C<or>. The expression must be followed (and thus terminated) by a
11776 token that an expression would normally be terminated by: end-of-file,
11777 closing bracketing punctuation, semicolon, or one of the keywords that
11778 signals a postfix expression-statement modifier. If I<flags> includes
11779 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11780 mandatory. It is up to the caller to ensure that the dynamic parser
11781 state (L</PL_parser> et al) is correctly set to reflect the source of
11782 the code to be parsed and the lexical context for the expression.
11784 The op tree representing the expression is returned. If an optional
11785 expression is absent, a null pointer is returned, otherwise the pointer
11788 If an error occurs in parsing or compilation, in most cases a valid op
11789 tree is returned anyway. The error is reflected in the parser state,
11790 normally resulting in a single exception at the top level of parsing
11791 which covers all the compilation errors that occurred. Some compilation
11792 errors, however, will throw an exception immediately.
11798 Perl_parse_fullexpr(pTHX_ U32 flags)
11800 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11804 =for apidoc Amx|OP *|parse_block|U32 flags
11806 Parse a single complete Perl code block. This consists of an opening
11807 brace, a sequence of statements, and a closing brace. The block
11808 constitutes a lexical scope, so C<my> variables and various compile-time
11809 effects can be contained within it. It is up to the caller to ensure
11810 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11811 reflect the source of the code to be parsed and the lexical context for
11814 The op tree representing the code block is returned. This is always a
11815 real op, never a null pointer. It will normally be a C<lineseq> list,
11816 including C<nextstate> or equivalent ops. No ops to construct any kind
11817 of runtime scope are included by virtue of it being a block.
11819 If an error occurs in parsing or compilation, in most cases a valid op
11820 tree (most likely null) is returned anyway. The error is reflected in
11821 the parser state, normally resulting in a single exception at the top
11822 level of parsing which covers all the compilation errors that occurred.
11823 Some compilation errors, however, will throw an exception immediately.
11825 The I<flags> parameter is reserved for future use, and must always
11832 Perl_parse_block(pTHX_ U32 flags)
11835 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11836 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11840 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11842 Parse a single unadorned Perl statement. This may be a normal imperative
11843 statement or a declaration that has compile-time effect. It does not
11844 include any label or other affixture. It is up to the caller to ensure
11845 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11846 reflect the source of the code to be parsed and the lexical context for
11849 The op tree representing the statement is returned. This may be a
11850 null pointer if the statement is null, for example if it was actually
11851 a subroutine definition (which has compile-time side effects). If not
11852 null, it will be ops directly implementing the statement, suitable to
11853 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11854 equivalent op (except for those embedded in a scope contained entirely
11855 within the statement).
11857 If an error occurs in parsing or compilation, in most cases a valid op
11858 tree (most likely null) is returned anyway. The error is reflected in
11859 the parser state, normally resulting in a single exception at the top
11860 level of parsing which covers all the compilation errors that occurred.
11861 Some compilation errors, however, will throw an exception immediately.
11863 The I<flags> parameter is reserved for future use, and must always
11870 Perl_parse_barestmt(pTHX_ U32 flags)
11873 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11874 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11878 =for apidoc Amx|SV *|parse_label|U32 flags
11880 Parse a single label, possibly optional, of the type that may prefix a
11881 Perl statement. It is up to the caller to ensure that the dynamic parser
11882 state (L</PL_parser> et al) is correctly set to reflect the source of
11883 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11884 label is optional, otherwise it is mandatory.
11886 The name of the label is returned in the form of a fresh scalar. If an
11887 optional label is absent, a null pointer is returned.
11889 If an error occurs in parsing, which can only occur if the label is
11890 mandatory, a valid label is returned anyway. The error is reflected in
11891 the parser state, normally resulting in a single exception at the top
11892 level of parsing which covers all the compilation errors that occurred.
11898 Perl_parse_label(pTHX_ U32 flags)
11900 if (flags & ~PARSE_OPTIONAL)
11901 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11902 if (PL_lex_state == LEX_KNOWNEXT) {
11903 PL_parser->yychar = yylex();
11904 if (PL_parser->yychar == LABEL) {
11905 char * const lpv = pl_yylval.pval;
11906 STRLEN llen = strlen(lpv);
11907 PL_parser->yychar = YYEMPTY;
11908 return newSVpvn_flags(lpv, llen, lpv[llen+1] ? SVf_UTF8 : 0);
11915 STRLEN wlen, bufptr_pos;
11918 if (!isIDFIRST_lazy_if(s, UTF))
11920 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11921 if (word_takes_any_delimeter(s, wlen))
11923 bufptr_pos = s - SvPVX(PL_linestr);
11925 lex_read_space(LEX_KEEP_PREVIOUS);
11927 s = SvPVX(PL_linestr) + bufptr_pos;
11928 if (t[0] == ':' && t[1] != ':') {
11929 PL_oldoldbufptr = PL_oldbufptr;
11932 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11936 if (flags & PARSE_OPTIONAL) {
11939 qerror(Perl_mess(aTHX_ "Parse error"));
11940 return newSVpvs("x");
11947 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11949 Parse a single complete Perl statement. This may be a normal imperative
11950 statement or a declaration that has compile-time effect, and may include
11951 optional labels. It is up to the caller to ensure that the dynamic
11952 parser state (L</PL_parser> et al) is correctly set to reflect the source
11953 of the code to be parsed and the lexical context for the statement.
11955 The op tree representing the statement is returned. This may be a
11956 null pointer if the statement is null, for example if it was actually
11957 a subroutine definition (which has compile-time side effects). If not
11958 null, it will be the result of a L</newSTATEOP> call, normally including
11959 a C<nextstate> or equivalent op.
11961 If an error occurs in parsing or compilation, in most cases a valid op
11962 tree (most likely null) is returned anyway. The error is reflected in
11963 the parser state, normally resulting in a single exception at the top
11964 level of parsing which covers all the compilation errors that occurred.
11965 Some compilation errors, however, will throw an exception immediately.
11967 The I<flags> parameter is reserved for future use, and must always
11974 Perl_parse_fullstmt(pTHX_ U32 flags)
11977 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11978 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11982 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11984 Parse a sequence of zero or more Perl statements. These may be normal
11985 imperative statements, including optional labels, or declarations
11986 that have compile-time effect, or any mixture thereof. The statement
11987 sequence ends when a closing brace or end-of-file is encountered in a
11988 place where a new statement could have validly started. It is up to
11989 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11990 is correctly set to reflect the source of the code to be parsed and the
11991 lexical context for the statements.
11993 The op tree representing the statement sequence is returned. This may
11994 be a null pointer if the statements were all null, for example if there
11995 were no statements or if there were only subroutine definitions (which
11996 have compile-time side effects). If not null, it will be a C<lineseq>
11997 list, normally including C<nextstate> or equivalent ops.
11999 If an error occurs in parsing or compilation, in most cases a valid op
12000 tree is returned anyway. The error is reflected in the parser state,
12001 normally resulting in a single exception at the top level of parsing
12002 which covers all the compilation errors that occurred. Some compilation
12003 errors, however, will throw an exception immediately.
12005 The I<flags> parameter is reserved for future use, and must always
12012 Perl_parse_stmtseq(pTHX_ U32 flags)
12017 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
12018 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
12019 c = lex_peek_unichar(0);
12020 if (c != -1 && c != /*{*/'}')
12021 qerror(Perl_mess(aTHX_ "Parse error"));
12027 * c-indentation-style: bsd
12028 * c-basic-offset: 4
12029 * indent-tabs-mode: nil
12032 * ex: set ts=8 sts=4 sw=4 et: