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_OPVAL, "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);
788 Safefree(parser->lex_brackstack);
789 Safefree(parser->lex_casestack);
790 Safefree(parser->lex_shared);
791 PL_parser = parser->old_parser;
797 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
799 Buffer scalar containing the chunk currently under consideration of the
800 text currently being lexed. This is always a plain string scalar (for
801 which C<SvPOK> is true). It is not intended to be used as a scalar by
802 normal scalar means; instead refer to the buffer directly by the pointer
803 variables described below.
805 The lexer maintains various C<char*> pointers to things in the
806 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
807 reallocated, all of these pointers must be updated. Don't attempt to
808 do this manually, but rather use L</lex_grow_linestr> if you need to
809 reallocate the buffer.
811 The content of the text chunk in the buffer is commonly exactly one
812 complete line of input, up to and including a newline terminator,
813 but there are situations where it is otherwise. The octets of the
814 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
815 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
816 flag on this scalar, which may disagree with it.
818 For direct examination of the buffer, the variable
819 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
820 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
821 of these pointers is usually preferable to examination of the scalar
822 through normal scalar means.
824 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
826 Direct pointer to the end of the chunk of text currently being lexed, the
827 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
828 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
829 always located at the end of the buffer, and does not count as part of
830 the buffer's contents.
832 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
834 Points to the current position of lexing inside the lexer buffer.
835 Characters around this point may be freely examined, within
836 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
837 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
838 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
840 Lexing code (whether in the Perl core or not) moves this pointer past
841 the characters that it consumes. It is also expected to perform some
842 bookkeeping whenever a newline character is consumed. This movement
843 can be more conveniently performed by the function L</lex_read_to>,
844 which handles newlines appropriately.
846 Interpretation of the buffer's octets can be abstracted out by
847 using the slightly higher-level functions L</lex_peek_unichar> and
848 L</lex_read_unichar>.
850 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
852 Points to the start of the current line inside the lexer buffer.
853 This is useful for indicating at which column an error occurred, and
854 not much else. This must be updated by any lexing code that consumes
855 a newline; the function L</lex_read_to> handles this detail.
861 =for apidoc Amx|bool|lex_bufutf8
863 Indicates whether the octets in the lexer buffer
864 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
865 of Unicode characters. If not, they should be interpreted as Latin-1
866 characters. This is analogous to the C<SvUTF8> flag for scalars.
868 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
869 contains valid UTF-8. Lexing code must be robust in the face of invalid
872 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
873 is significant, but not the whole story regarding the input character
874 encoding. Normally, when a file is being read, the scalar contains octets
875 and its C<SvUTF8> flag is off, but the octets should be interpreted as
876 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
877 however, the scalar may have the C<SvUTF8> flag on, and in this case its
878 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
879 is in effect. This logic may change in the future; use this function
880 instead of implementing the logic yourself.
886 Perl_lex_bufutf8(pTHX)
892 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
894 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
895 at least I<len> octets (including terminating NUL). Returns a
896 pointer to the reallocated buffer. This is necessary before making
897 any direct modification of the buffer that would increase its length.
898 L</lex_stuff_pvn> provides a more convenient way to insert text into
901 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
902 this function updates all of the lexer's variables that point directly
909 Perl_lex_grow_linestr(pTHX_ STRLEN len)
913 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
914 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
915 linestr = PL_parser->linestr;
916 buf = SvPVX(linestr);
917 if (len <= SvLEN(linestr))
919 bufend_pos = PL_parser->bufend - buf;
920 bufptr_pos = PL_parser->bufptr - buf;
921 oldbufptr_pos = PL_parser->oldbufptr - buf;
922 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
923 linestart_pos = PL_parser->linestart - buf;
924 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
925 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
926 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
927 PL_parser->lex_shared->re_eval_start - buf : 0;
929 buf = sv_grow(linestr, len);
931 PL_parser->bufend = buf + bufend_pos;
932 PL_parser->bufptr = buf + bufptr_pos;
933 PL_parser->oldbufptr = buf + oldbufptr_pos;
934 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
935 PL_parser->linestart = buf + linestart_pos;
936 if (PL_parser->last_uni)
937 PL_parser->last_uni = buf + last_uni_pos;
938 if (PL_parser->last_lop)
939 PL_parser->last_lop = buf + last_lop_pos;
940 if (PL_parser->lex_shared->re_eval_start)
941 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
946 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
948 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
949 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
950 reallocating the buffer if necessary. This means that lexing code that
951 runs later will see the characters as if they had appeared in the input.
952 It is not recommended to do this as part of normal parsing, and most
953 uses of this facility run the risk of the inserted characters being
954 interpreted in an unintended manner.
956 The string to be inserted is represented by I<len> octets starting
957 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
958 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
959 The characters are recoded for the lexer buffer, according to how the
960 buffer is currently being interpreted (L</lex_bufutf8>). If a string
961 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
962 function is more convenient.
968 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
972 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
973 if (flags & ~(LEX_STUFF_UTF8))
974 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
976 if (flags & LEX_STUFF_UTF8) {
980 const char *p, *e = pv+len;
981 for (p = pv; p != e; p++)
982 highhalf += !!(((U8)*p) & 0x80);
985 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
986 bufptr = PL_parser->bufptr;
987 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
988 SvCUR_set(PL_parser->linestr,
989 SvCUR(PL_parser->linestr) + len+highhalf);
990 PL_parser->bufend += len+highhalf;
991 for (p = pv; p != e; p++) {
994 *bufptr++ = (char)(0xc0 | (c >> 6));
995 *bufptr++ = (char)(0x80 | (c & 0x3f));
1002 if (flags & LEX_STUFF_UTF8) {
1003 STRLEN highhalf = 0;
1004 const char *p, *e = pv+len;
1005 for (p = pv; p != e; p++) {
1008 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1009 "non-Latin-1 character into Latin-1 input");
1010 } else if (c >= 0xc2 && p+1 != e &&
1011 (((U8)p[1]) & 0xc0) == 0x80) {
1014 } else if (c >= 0x80) {
1015 /* malformed UTF-8 */
1017 SAVESPTR(PL_warnhook);
1018 PL_warnhook = PERL_WARNHOOK_FATAL;
1019 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1025 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1026 bufptr = PL_parser->bufptr;
1027 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1028 SvCUR_set(PL_parser->linestr,
1029 SvCUR(PL_parser->linestr) + len-highhalf);
1030 PL_parser->bufend += len-highhalf;
1031 for (p = pv; p != e; p++) {
1034 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1037 *bufptr++ = (char)c;
1042 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1043 bufptr = PL_parser->bufptr;
1044 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1045 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1046 PL_parser->bufend += len;
1047 Copy(pv, bufptr, len, char);
1053 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1055 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1056 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1057 reallocating the buffer if necessary. This means that lexing code that
1058 runs later will see the characters as if they had appeared in the input.
1059 It is not recommended to do this as part of normal parsing, and most
1060 uses of this facility run the risk of the inserted characters being
1061 interpreted in an unintended manner.
1063 The string to be inserted is represented by octets starting at I<pv>
1064 and continuing to the first nul. These octets are interpreted as either
1065 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1066 in I<flags>. The characters are recoded for the lexer buffer, according
1067 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1068 If it is not convenient to nul-terminate a string to be inserted, the
1069 L</lex_stuff_pvn> function is more appropriate.
1075 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1077 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1078 lex_stuff_pvn(pv, strlen(pv), flags);
1082 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1084 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1085 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1086 reallocating the buffer if necessary. This means that lexing code that
1087 runs later will see the characters as if they had appeared in the input.
1088 It is not recommended to do this as part of normal parsing, and most
1089 uses of this facility run the risk of the inserted characters being
1090 interpreted in an unintended manner.
1092 The string to be inserted is the string value of I<sv>. The characters
1093 are recoded for the lexer buffer, according to how the buffer is currently
1094 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1095 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1096 need to construct a scalar.
1102 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1106 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1108 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1110 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1114 =for apidoc Amx|void|lex_unstuff|char *ptr
1116 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1117 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1118 This hides the discarded text from any lexing code that runs later,
1119 as if the text had never appeared.
1121 This is not the normal way to consume lexed text. For that, use
1128 Perl_lex_unstuff(pTHX_ char *ptr)
1132 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1133 buf = PL_parser->bufptr;
1135 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1138 bufend = PL_parser->bufend;
1140 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1141 unstuff_len = ptr - buf;
1142 Move(ptr, buf, bufend+1-ptr, char);
1143 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1144 PL_parser->bufend = bufend - unstuff_len;
1148 =for apidoc Amx|void|lex_read_to|char *ptr
1150 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1151 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1152 performing the correct bookkeeping whenever a newline character is passed.
1153 This is the normal way to consume lexed text.
1155 Interpretation of the buffer's octets can be abstracted out by
1156 using the slightly higher-level functions L</lex_peek_unichar> and
1157 L</lex_read_unichar>.
1163 Perl_lex_read_to(pTHX_ char *ptr)
1166 PERL_ARGS_ASSERT_LEX_READ_TO;
1167 s = PL_parser->bufptr;
1168 if (ptr < s || ptr > PL_parser->bufend)
1169 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1170 for (; s != ptr; s++)
1172 COPLINE_INC_WITH_HERELINES;
1173 PL_parser->linestart = s+1;
1175 PL_parser->bufptr = ptr;
1179 =for apidoc Amx|void|lex_discard_to|char *ptr
1181 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1182 up to I<ptr>. The remaining content of the buffer will be moved, and
1183 all pointers into the buffer updated appropriately. I<ptr> must not
1184 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1185 it is not permitted to discard text that has yet to be lexed.
1187 Normally it is not necessarily to do this directly, because it suffices to
1188 use the implicit discarding behaviour of L</lex_next_chunk> and things
1189 based on it. However, if a token stretches across multiple lines,
1190 and the lexing code has kept multiple lines of text in the buffer for
1191 that purpose, then after completion of the token it would be wise to
1192 explicitly discard the now-unneeded earlier lines, to avoid future
1193 multi-line tokens growing the buffer without bound.
1199 Perl_lex_discard_to(pTHX_ char *ptr)
1203 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1204 buf = SvPVX(PL_parser->linestr);
1206 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1209 if (ptr > PL_parser->bufptr)
1210 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1211 discard_len = ptr - buf;
1212 if (PL_parser->oldbufptr < ptr)
1213 PL_parser->oldbufptr = ptr;
1214 if (PL_parser->oldoldbufptr < ptr)
1215 PL_parser->oldoldbufptr = ptr;
1216 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1217 PL_parser->last_uni = NULL;
1218 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1219 PL_parser->last_lop = NULL;
1220 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1221 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1222 PL_parser->bufend -= discard_len;
1223 PL_parser->bufptr -= discard_len;
1224 PL_parser->oldbufptr -= discard_len;
1225 PL_parser->oldoldbufptr -= discard_len;
1226 if (PL_parser->last_uni)
1227 PL_parser->last_uni -= discard_len;
1228 if (PL_parser->last_lop)
1229 PL_parser->last_lop -= discard_len;
1233 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1235 Reads in the next chunk of text to be lexed, appending it to
1236 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1237 looked to the end of the current chunk and wants to know more. It is
1238 usual, but not necessary, for lexing to have consumed the entirety of
1239 the current chunk at this time.
1241 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1242 chunk (i.e., the current chunk has been entirely consumed), normally the
1243 current chunk will be discarded at the same time that the new chunk is
1244 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1245 will not be discarded. If the current chunk has not been entirely
1246 consumed, then it will not be discarded regardless of the flag.
1248 Returns true if some new text was added to the buffer, or false if the
1249 buffer has reached the end of the input text.
1254 #define LEX_FAKE_EOF 0x80000000
1255 #define LEX_NO_TERM 0x40000000
1258 Perl_lex_next_chunk(pTHX_ U32 flags)
1262 STRLEN old_bufend_pos, new_bufend_pos;
1263 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1264 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1265 bool got_some_for_debugger = 0;
1267 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1268 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1269 linestr = PL_parser->linestr;
1270 buf = SvPVX(linestr);
1271 if (!(flags & LEX_KEEP_PREVIOUS) &&
1272 PL_parser->bufptr == PL_parser->bufend) {
1273 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1275 if (PL_parser->last_uni != PL_parser->bufend)
1276 PL_parser->last_uni = NULL;
1277 if (PL_parser->last_lop != PL_parser->bufend)
1278 PL_parser->last_lop = NULL;
1279 last_uni_pos = last_lop_pos = 0;
1283 old_bufend_pos = PL_parser->bufend - buf;
1284 bufptr_pos = PL_parser->bufptr - buf;
1285 oldbufptr_pos = PL_parser->oldbufptr - buf;
1286 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1287 linestart_pos = PL_parser->linestart - buf;
1288 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1289 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1291 if (flags & LEX_FAKE_EOF) {
1293 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1295 } else if (filter_gets(linestr, old_bufend_pos)) {
1297 got_some_for_debugger = 1;
1298 } else if (flags & LEX_NO_TERM) {
1301 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1302 sv_setpvs(linestr, "");
1304 /* End of real input. Close filehandle (unless it was STDIN),
1305 * then add implicit termination.
1307 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1308 PerlIO_clearerr(PL_parser->rsfp);
1309 else if (PL_parser->rsfp)
1310 (void)PerlIO_close(PL_parser->rsfp);
1311 PL_parser->rsfp = NULL;
1312 PL_parser->in_pod = PL_parser->filtered = 0;
1314 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1317 if (!PL_in_eval && PL_minus_p) {
1319 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1320 PL_minus_n = PL_minus_p = 0;
1321 } else if (!PL_in_eval && PL_minus_n) {
1322 sv_catpvs(linestr, /*{*/";}");
1325 sv_catpvs(linestr, ";");
1328 buf = SvPVX(linestr);
1329 new_bufend_pos = SvCUR(linestr);
1330 PL_parser->bufend = buf + new_bufend_pos;
1331 PL_parser->bufptr = buf + bufptr_pos;
1332 PL_parser->oldbufptr = buf + oldbufptr_pos;
1333 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1334 PL_parser->linestart = buf + linestart_pos;
1335 if (PL_parser->last_uni)
1336 PL_parser->last_uni = buf + last_uni_pos;
1337 if (PL_parser->last_lop)
1338 PL_parser->last_lop = buf + last_lop_pos;
1339 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1340 PL_curstash != PL_debstash) {
1341 /* debugger active and we're not compiling the debugger code,
1342 * so store the line into the debugger's array of lines
1344 update_debugger_info(NULL, buf+old_bufend_pos,
1345 new_bufend_pos-old_bufend_pos);
1351 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1353 Looks ahead one (Unicode) character in the text currently being lexed.
1354 Returns the codepoint (unsigned integer value) of the next character,
1355 or -1 if lexing has reached the end of the input text. To consume the
1356 peeked character, use L</lex_read_unichar>.
1358 If the next character is in (or extends into) the next chunk of input
1359 text, the next chunk will be read in. Normally the current chunk will be
1360 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1361 then the current chunk will not be discarded.
1363 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1364 is encountered, an exception is generated.
1370 Perl_lex_peek_unichar(pTHX_ U32 flags)
1374 if (flags & ~(LEX_KEEP_PREVIOUS))
1375 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1376 s = PL_parser->bufptr;
1377 bufend = PL_parser->bufend;
1383 if (!lex_next_chunk(flags))
1385 s = PL_parser->bufptr;
1386 bufend = PL_parser->bufend;
1392 len = PL_utf8skip[head];
1393 while ((STRLEN)(bufend-s) < len) {
1394 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1396 s = PL_parser->bufptr;
1397 bufend = PL_parser->bufend;
1400 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1401 if (retlen == (STRLEN)-1) {
1402 /* malformed UTF-8 */
1404 SAVESPTR(PL_warnhook);
1405 PL_warnhook = PERL_WARNHOOK_FATAL;
1406 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1412 if (!lex_next_chunk(flags))
1414 s = PL_parser->bufptr;
1421 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1423 Reads the next (Unicode) character in the text currently being lexed.
1424 Returns the codepoint (unsigned integer value) of the character read,
1425 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1426 if lexing has reached the end of the input text. To non-destructively
1427 examine the next character, use L</lex_peek_unichar> instead.
1429 If the next character is in (or extends into) the next chunk of input
1430 text, the next chunk will be read in. Normally the current chunk will be
1431 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1432 then the current chunk will not be discarded.
1434 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1435 is encountered, an exception is generated.
1441 Perl_lex_read_unichar(pTHX_ U32 flags)
1444 if (flags & ~(LEX_KEEP_PREVIOUS))
1445 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1446 c = lex_peek_unichar(flags);
1449 COPLINE_INC_WITH_HERELINES;
1451 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1453 ++(PL_parser->bufptr);
1459 =for apidoc Amx|void|lex_read_space|U32 flags
1461 Reads optional spaces, in Perl style, in the text currently being
1462 lexed. The spaces may include ordinary whitespace characters and
1463 Perl-style comments. C<#line> directives are processed if encountered.
1464 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1465 at a non-space character (or the end of the input text).
1467 If spaces extend into the next chunk of input text, the next chunk will
1468 be read in. Normally the current chunk will be discarded at the same
1469 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1470 chunk will not be discarded.
1475 #define LEX_NO_NEXT_CHUNK 0x80000000
1478 Perl_lex_read_space(pTHX_ U32 flags)
1481 bool need_incline = 0;
1482 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1483 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1486 sv_free(PL_skipwhite);
1487 PL_skipwhite = NULL;
1490 PL_skipwhite = newSVpvs("");
1491 #endif /* PERL_MAD */
1492 s = PL_parser->bufptr;
1493 bufend = PL_parser->bufend;
1499 } while (!(c == '\n' || (c == 0 && s == bufend)));
1500 } else if (c == '\n') {
1502 PL_parser->linestart = s;
1507 } else if (isSPACE(c)) {
1509 } else if (c == 0 && s == bufend) {
1513 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1514 #endif /* PERL_MAD */
1515 if (flags & LEX_NO_NEXT_CHUNK)
1517 PL_parser->bufptr = s;
1518 COPLINE_INC_WITH_HERELINES;
1519 got_more = lex_next_chunk(flags);
1520 CopLINE_dec(PL_curcop);
1521 s = PL_parser->bufptr;
1522 bufend = PL_parser->bufend;
1525 if (need_incline && PL_parser->rsfp) {
1535 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1536 #endif /* PERL_MAD */
1537 PL_parser->bufptr = s;
1542 * This subroutine has nothing to do with tilting, whether at windmills
1543 * or pinball tables. Its name is short for "increment line". It
1544 * increments the current line number in CopLINE(PL_curcop) and checks
1545 * to see whether the line starts with a comment of the form
1546 * # line 500 "foo.pm"
1547 * If so, it sets the current line number and file to the values in the comment.
1551 S_incline(pTHX_ const char *s)
1559 PERL_ARGS_ASSERT_INCLINE;
1561 COPLINE_INC_WITH_HERELINES;
1562 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1563 && s+1 == PL_bufend && *s == ';') {
1564 /* fake newline in string eval */
1565 CopLINE_dec(PL_curcop);
1570 while (SPACE_OR_TAB(*s))
1572 if (strnEQ(s, "line", 4))
1576 if (SPACE_OR_TAB(*s))
1580 while (SPACE_OR_TAB(*s))
1588 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1590 while (SPACE_OR_TAB(*s))
1592 if (*s == '"' && (t = strchr(s+1, '"'))) {
1598 while (!isSPACE(*t))
1602 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1604 if (*e != '\n' && *e != '\0')
1605 return; /* false alarm */
1607 line_num = atoi(n)-1;
1610 const STRLEN len = t - s;
1611 SV *const temp_sv = CopFILESV(PL_curcop);
1616 cf = SvPVX(temp_sv);
1617 tmplen = SvCUR(temp_sv);
1623 if (!PL_rsfp && !PL_parser->filtered) {
1624 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1625 * to *{"::_<newfilename"} */
1626 /* However, the long form of evals is only turned on by the
1627 debugger - usually they're "(eval %lu)" */
1631 STRLEN tmplen2 = len;
1632 if (tmplen + 2 <= sizeof smallbuf)
1635 Newx(tmpbuf, tmplen + 2, char);
1638 memcpy(tmpbuf + 2, cf, tmplen);
1640 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1645 if (tmplen2 + 2 <= sizeof smallbuf)
1648 Newx(tmpbuf2, tmplen2 + 2, char);
1650 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1651 /* Either they malloc'd it, or we malloc'd it,
1652 so no prefix is present in ours. */
1657 memcpy(tmpbuf2 + 2, s, tmplen2);
1660 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1662 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1663 /* adjust ${"::_<newfilename"} to store the new file name */
1664 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1665 /* The line number may differ. If that is the case,
1666 alias the saved lines that are in the array.
1667 Otherwise alias the whole array. */
1668 if (CopLINE(PL_curcop) == line_num) {
1669 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1670 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1672 else if (GvAV(*gvp)) {
1673 AV * const av = GvAV(*gvp);
1674 const I32 start = CopLINE(PL_curcop)+1;
1675 I32 items = AvFILLp(av) - start;
1677 AV * const av2 = GvAVn(gv2);
1678 SV **svp = AvARRAY(av) + start;
1679 I32 l = (I32)line_num+1;
1681 av_store(av2, l++, SvREFCNT_inc(*svp++));
1686 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1688 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1690 CopFILE_free(PL_curcop);
1691 CopFILE_setn(PL_curcop, s, len);
1693 CopLINE_set(PL_curcop, line_num);
1697 /* skip space before PL_thistoken */
1700 S_skipspace0(pTHX_ register char *s)
1702 PERL_ARGS_ASSERT_SKIPSPACE0;
1709 PL_thiswhite = newSVpvs("");
1710 sv_catsv(PL_thiswhite, PL_skipwhite);
1711 sv_free(PL_skipwhite);
1714 PL_realtokenstart = s - SvPVX(PL_linestr);
1718 /* skip space after PL_thistoken */
1721 S_skipspace1(pTHX_ register char *s)
1723 const char *start = s;
1724 I32 startoff = start - SvPVX(PL_linestr);
1726 PERL_ARGS_ASSERT_SKIPSPACE1;
1731 start = SvPVX(PL_linestr) + startoff;
1732 if (!PL_thistoken && PL_realtokenstart >= 0) {
1733 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1734 PL_thistoken = newSVpvn(tstart, start - tstart);
1736 PL_realtokenstart = -1;
1739 PL_nextwhite = newSVpvs("");
1740 sv_catsv(PL_nextwhite, PL_skipwhite);
1741 sv_free(PL_skipwhite);
1748 S_skipspace2(pTHX_ register char *s, SV **svp)
1751 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1752 const I32 startoff = s - SvPVX(PL_linestr);
1754 PERL_ARGS_ASSERT_SKIPSPACE2;
1757 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1758 if (!PL_madskills || !svp)
1760 start = SvPVX(PL_linestr) + startoff;
1761 if (!PL_thistoken && PL_realtokenstart >= 0) {
1762 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1763 PL_thistoken = newSVpvn(tstart, start - tstart);
1764 PL_realtokenstart = -1;
1768 *svp = newSVpvs("");
1769 sv_setsv(*svp, PL_skipwhite);
1770 sv_free(PL_skipwhite);
1779 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1781 AV *av = CopFILEAVx(PL_curcop);
1783 SV * const sv = newSV_type(SVt_PVMG);
1785 sv_setsv(sv, orig_sv);
1787 sv_setpvn(sv, buf, len);
1790 av_store(av, (I32)CopLINE(PL_curcop), sv);
1796 * Called to gobble the appropriate amount and type of whitespace.
1797 * Skips comments as well.
1801 S_skipspace(pTHX_ register char *s)
1805 #endif /* PERL_MAD */
1806 PERL_ARGS_ASSERT_SKIPSPACE;
1809 sv_free(PL_skipwhite);
1810 PL_skipwhite = NULL;
1812 #endif /* PERL_MAD */
1813 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1814 while (s < PL_bufend && SPACE_OR_TAB(*s))
1817 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1819 lex_read_space(LEX_KEEP_PREVIOUS |
1820 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1821 LEX_NO_NEXT_CHUNK : 0));
1823 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1824 if (PL_linestart > PL_bufptr)
1825 PL_bufptr = PL_linestart;
1830 PL_skipwhite = newSVpvn(start, s-start);
1831 #endif /* PERL_MAD */
1837 * Check the unary operators to ensure there's no ambiguity in how they're
1838 * used. An ambiguous piece of code would be:
1840 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1841 * the +5 is its argument.
1851 if (PL_oldoldbufptr != PL_last_uni)
1853 while (isSPACE(*PL_last_uni))
1856 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1858 if ((t = strchr(s, '(')) && t < PL_bufptr)
1861 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1862 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1863 (int)(s - PL_last_uni), PL_last_uni);
1867 * LOP : macro to build a list operator. Its behaviour has been replaced
1868 * with a subroutine, S_lop() for which LOP is just another name.
1871 #define LOP(f,x) return lop(f,x,s)
1875 * Build a list operator (or something that might be one). The rules:
1876 * - if we have a next token, then it's a list operator [why?]
1877 * - if the next thing is an opening paren, then it's a function
1878 * - else it's a list operator
1882 S_lop(pTHX_ I32 f, int x, char *s)
1886 PERL_ARGS_ASSERT_LOP;
1892 PL_last_lop = PL_oldbufptr;
1893 PL_last_lop_op = (OPCODE)f;
1902 return REPORT(FUNC);
1905 return REPORT(FUNC);
1908 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1909 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1910 return REPORT(LSTOP);
1917 * Sets up for an eventual force_next(). start_force(0) basically does
1918 * an unshift, while start_force(-1) does a push. yylex removes items
1923 S_start_force(pTHX_ int where)
1927 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1928 where = PL_lasttoke;
1929 assert(PL_curforce < 0 || PL_curforce == where);
1930 if (PL_curforce != where) {
1931 for (i = PL_lasttoke; i > where; --i) {
1932 PL_nexttoke[i] = PL_nexttoke[i-1];
1936 if (PL_curforce < 0) /* in case of duplicate start_force() */
1937 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1938 PL_curforce = where;
1941 curmad('^', newSVpvs(""));
1942 CURMAD('_', PL_nextwhite);
1947 S_curmad(pTHX_ char slot, SV *sv)
1953 if (PL_curforce < 0)
1954 where = &PL_thismad;
1956 where = &PL_nexttoke[PL_curforce].next_mad;
1962 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1964 else if (PL_encoding) {
1965 sv_recode_to_utf8(sv, PL_encoding);
1970 /* keep a slot open for the head of the list? */
1971 if (slot != '_' && *where && (*where)->mad_key == '^') {
1972 (*where)->mad_key = slot;
1973 sv_free(MUTABLE_SV(((*where)->mad_val)));
1974 (*where)->mad_val = (void*)sv;
1977 addmad(newMADsv(slot, sv), where, 0);
1980 # define start_force(where) NOOP
1981 # define curmad(slot, sv) NOOP
1986 * When the lexer realizes it knows the next token (for instance,
1987 * it is reordering tokens for the parser) then it can call S_force_next
1988 * to know what token to return the next time the lexer is called. Caller
1989 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1990 * and possibly PL_expect to ensure the lexer handles the token correctly.
1994 S_force_next(pTHX_ I32 type)
1999 PerlIO_printf(Perl_debug_log, "### forced token:\n");
2000 tokereport(type, &NEXTVAL_NEXTTOKE);
2003 /* Don’t let opslab_force_free snatch it */
2004 if (S_is_opval_token(type & 0xffff) && NEXTVAL_NEXTTOKE.opval) {
2005 assert(!NEXTVAL_NEXTTOKE.opval->op_savefree);
2006 NEXTVAL_NEXTTOKE.opval->op_savefree = 1;
2009 if (PL_curforce < 0)
2010 start_force(PL_lasttoke);
2011 PL_nexttoke[PL_curforce].next_type = type;
2012 if (PL_lex_state != LEX_KNOWNEXT)
2013 PL_lex_defer = PL_lex_state;
2014 PL_lex_state = LEX_KNOWNEXT;
2015 PL_lex_expect = PL_expect;
2018 PL_nexttype[PL_nexttoke] = type;
2020 if (PL_lex_state != LEX_KNOWNEXT) {
2021 PL_lex_defer = PL_lex_state;
2022 PL_lex_expect = PL_expect;
2023 PL_lex_state = LEX_KNOWNEXT;
2031 int yyc = PL_parser->yychar;
2032 if (yyc != YYEMPTY) {
2035 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2036 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2037 PL_lex_allbrackets--;
2039 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2040 } else if (yyc == '('/*)*/) {
2041 PL_lex_allbrackets--;
2046 PL_parser->yychar = YYEMPTY;
2051 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2054 SV * const sv = newSVpvn_utf8(start, len,
2057 && !is_ascii_string((const U8*)start, len)
2058 && is_utf8_string((const U8*)start, len));
2064 * When the lexer knows the next thing is a word (for instance, it has
2065 * just seen -> and it knows that the next char is a word char, then
2066 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2070 * char *start : buffer position (must be within PL_linestr)
2071 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2072 * int check_keyword : if true, Perl checks to make sure the word isn't
2073 * a keyword (do this if the word is a label, e.g. goto FOO)
2074 * int allow_pack : if true, : characters will also be allowed (require,
2075 * use, etc. do this)
2076 * int allow_initial_tick : used by the "sub" lexer only.
2080 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2086 PERL_ARGS_ASSERT_FORCE_WORD;
2088 start = SKIPSPACE1(start);
2090 if (isIDFIRST_lazy_if(s,UTF) ||
2091 (allow_pack && *s == ':') ||
2092 (allow_initial_tick && *s == '\'') )
2094 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2095 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2097 start_force(PL_curforce);
2099 curmad('X', newSVpvn(start,s-start));
2100 if (token == METHOD) {
2105 PL_expect = XOPERATOR;
2109 curmad('g', newSVpvs( "forced" ));
2110 NEXTVAL_NEXTTOKE.opval
2111 = (OP*)newSVOP(OP_CONST,0,
2112 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2113 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2121 * Called when the lexer wants $foo *foo &foo etc, but the program
2122 * text only contains the "foo" portion. The first argument is a pointer
2123 * to the "foo", and the second argument is the type symbol to prefix.
2124 * Forces the next token to be a "WORD".
2125 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2129 S_force_ident(pTHX_ register const char *s, int kind)
2133 PERL_ARGS_ASSERT_FORCE_IDENT;
2136 const STRLEN len = strlen(s);
2137 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2138 UTF ? SVf_UTF8 : 0));
2139 start_force(PL_curforce);
2140 NEXTVAL_NEXTTOKE.opval = o;
2143 o->op_private = OPpCONST_ENTERED;
2144 /* XXX see note in pp_entereval() for why we forgo typo
2145 warnings if the symbol must be introduced in an eval.
2147 gv_fetchpvn_flags(s, len,
2148 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2149 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2150 kind == '$' ? SVt_PV :
2151 kind == '@' ? SVt_PVAV :
2152 kind == '%' ? SVt_PVHV :
2160 S_force_ident_maybe_lex(pTHX_ char pit)
2162 start_force(PL_curforce);
2163 NEXTVAL_NEXTTOKE.ival = pit;
2168 Perl_str_to_version(pTHX_ SV *sv)
2173 const char *start = SvPV_const(sv,len);
2174 const char * const end = start + len;
2175 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2177 PERL_ARGS_ASSERT_STR_TO_VERSION;
2179 while (start < end) {
2183 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2188 retval += ((NV)n)/nshift;
2197 * Forces the next token to be a version number.
2198 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2199 * and if "guessing" is TRUE, then no new token is created (and the caller
2200 * must use an alternative parsing method).
2204 S_force_version(pTHX_ char *s, int guessing)
2210 I32 startoff = s - SvPVX(PL_linestr);
2213 PERL_ARGS_ASSERT_FORCE_VERSION;
2221 while (isDIGIT(*d) || *d == '_' || *d == '.')
2225 start_force(PL_curforce);
2226 curmad('X', newSVpvn(s,d-s));
2229 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2231 #ifdef USE_LOCALE_NUMERIC
2232 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2233 setlocale(LC_NUMERIC, "C");
2235 s = scan_num(s, &pl_yylval);
2236 #ifdef USE_LOCALE_NUMERIC
2237 setlocale(LC_NUMERIC, loc);
2240 version = pl_yylval.opval;
2241 ver = cSVOPx(version)->op_sv;
2242 if (SvPOK(ver) && !SvNIOK(ver)) {
2243 SvUPGRADE(ver, SVt_PVNV);
2244 SvNV_set(ver, str_to_version(ver));
2245 SvNOK_on(ver); /* hint that it is a version */
2248 else if (guessing) {
2251 sv_free(PL_nextwhite); /* let next token collect whitespace */
2253 s = SvPVX(PL_linestr) + startoff;
2261 if (PL_madskills && !version) {
2262 sv_free(PL_nextwhite); /* let next token collect whitespace */
2264 s = SvPVX(PL_linestr) + startoff;
2267 /* NOTE: The parser sees the package name and the VERSION swapped */
2268 start_force(PL_curforce);
2269 NEXTVAL_NEXTTOKE.opval = version;
2276 * S_force_strict_version
2277 * Forces the next token to be a version number using strict syntax rules.
2281 S_force_strict_version(pTHX_ char *s)
2286 I32 startoff = s - SvPVX(PL_linestr);
2288 const char *errstr = NULL;
2290 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2292 while (isSPACE(*s)) /* leading whitespace */
2295 if (is_STRICT_VERSION(s,&errstr)) {
2297 s = (char *)scan_version(s, ver, 0);
2298 version = newSVOP(OP_CONST, 0, ver);
2300 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2301 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2305 yyerror(errstr); /* version required */
2310 if (PL_madskills && !version) {
2311 sv_free(PL_nextwhite); /* let next token collect whitespace */
2313 s = SvPVX(PL_linestr) + startoff;
2316 /* NOTE: The parser sees the package name and the VERSION swapped */
2317 start_force(PL_curforce);
2318 NEXTVAL_NEXTTOKE.opval = version;
2326 * Tokenize a quoted string passed in as an SV. It finds the next
2327 * chunk, up to end of string or a backslash. It may make a new
2328 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2333 S_tokeq(pTHX_ SV *sv)
2342 PERL_ARGS_ASSERT_TOKEQ;
2347 s = SvPV_force(sv, len);
2348 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2351 /* This is relying on the SV being "well formed" with a trailing '\0' */
2352 while (s < send && !(*s == '\\' && s[1] == '\\'))
2357 if ( PL_hints & HINT_NEW_STRING ) {
2358 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2362 if (s + 1 < send && (s[1] == '\\'))
2363 s++; /* all that, just for this */
2368 SvCUR_set(sv, d - SvPVX_const(sv));
2370 if ( PL_hints & HINT_NEW_STRING )
2371 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2376 * Now come three functions related to double-quote context,
2377 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2378 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2379 * interact with PL_lex_state, and create fake ( ... ) argument lists
2380 * to handle functions and concatenation.
2384 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2389 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2391 * Pattern matching will set PL_lex_op to the pattern-matching op to
2392 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2394 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2396 * Everything else becomes a FUNC.
2398 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2399 * had an OP_CONST or OP_READLINE). This just sets us up for a
2400 * call to S_sublex_push().
2404 S_sublex_start(pTHX)
2407 const I32 op_type = pl_yylval.ival;
2409 if (op_type == OP_NULL) {
2410 pl_yylval.opval = PL_lex_op;
2414 if (op_type == OP_CONST || op_type == OP_READLINE) {
2415 SV *sv = tokeq(PL_lex_stuff);
2417 if (SvTYPE(sv) == SVt_PVIV) {
2418 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2420 const char * const p = SvPV_const(sv, len);
2421 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2425 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2426 PL_lex_stuff = NULL;
2427 /* Allow <FH> // "foo" */
2428 if (op_type == OP_READLINE)
2429 PL_expect = XTERMORDORDOR;
2432 else if (op_type == OP_BACKTICK && PL_lex_op) {
2433 /* readpipe() vas overriden */
2434 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2435 pl_yylval.opval = PL_lex_op;
2437 PL_lex_stuff = NULL;
2441 PL_sublex_info.super_state = PL_lex_state;
2442 PL_sublex_info.sub_inwhat = (U16)op_type;
2443 PL_sublex_info.sub_op = PL_lex_op;
2444 PL_lex_state = LEX_INTERPPUSH;
2448 pl_yylval.opval = PL_lex_op;
2458 * Create a new scope to save the lexing state. The scope will be
2459 * ended in S_sublex_done. Returns a '(', starting the function arguments
2460 * to the uc, lc, etc. found before.
2461 * Sets PL_lex_state to LEX_INTERPCONCAT.
2471 PL_lex_state = PL_sublex_info.super_state;
2472 SAVEBOOL(PL_lex_dojoin);
2473 SAVEI32(PL_lex_brackets);
2474 SAVEI32(PL_lex_allbrackets);
2475 SAVEI32(PL_lex_formbrack);
2476 SAVEI8(PL_lex_fakeeof);
2477 SAVEI32(PL_lex_casemods);
2478 SAVEI32(PL_lex_starts);
2479 SAVEI8(PL_lex_state);
2480 SAVESPTR(PL_lex_repl);
2481 SAVEVPTR(PL_lex_inpat);
2482 SAVEI16(PL_lex_inwhat);
2483 SAVECOPLINE(PL_curcop);
2484 SAVEPPTR(PL_bufptr);
2485 SAVEPPTR(PL_bufend);
2486 SAVEPPTR(PL_oldbufptr);
2487 SAVEPPTR(PL_oldoldbufptr);
2488 SAVEPPTR(PL_last_lop);
2489 SAVEPPTR(PL_last_uni);
2490 SAVEPPTR(PL_linestart);
2491 SAVESPTR(PL_linestr);
2492 SAVEGENERICPV(PL_lex_brackstack);
2493 SAVEGENERICPV(PL_lex_casestack);
2494 SAVEGENERICPV(PL_parser->lex_shared);
2496 /* The here-doc parser needs to be able to peek into outer lexing
2497 scopes to find the body of the here-doc. So we put PL_linestr and
2498 PL_bufptr into lex_shared, to ‘share’ those values.
2500 PL_parser->lex_shared->ls_linestr = PL_linestr;
2501 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2503 PL_linestr = PL_lex_stuff;
2504 PL_lex_repl = PL_sublex_info.repl;
2505 PL_lex_stuff = NULL;
2506 PL_sublex_info.repl = NULL;
2508 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2509 = SvPVX(PL_linestr);
2510 PL_bufend += SvCUR(PL_linestr);
2511 PL_last_lop = PL_last_uni = NULL;
2512 SAVEFREESV(PL_linestr);
2513 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2515 PL_lex_dojoin = FALSE;
2516 PL_lex_brackets = PL_lex_formbrack = 0;
2517 PL_lex_allbrackets = 0;
2518 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2519 Newx(PL_lex_brackstack, 120, char);
2520 Newx(PL_lex_casestack, 12, char);
2521 PL_lex_casemods = 0;
2522 *PL_lex_casestack = '\0';
2524 PL_lex_state = LEX_INTERPCONCAT;
2525 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2527 Newxz(shared, 1, LEXSHARED);
2528 shared->ls_prev = PL_parser->lex_shared;
2529 PL_parser->lex_shared = shared;
2531 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2532 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2533 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2534 PL_lex_inpat = PL_sublex_info.sub_op;
2536 PL_lex_inpat = NULL;
2543 * Restores lexer state after a S_sublex_push.
2550 if (!PL_lex_starts++) {
2551 SV * const sv = newSVpvs("");
2552 if (SvUTF8(PL_linestr))
2554 PL_expect = XOPERATOR;
2555 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2559 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2560 PL_lex_state = LEX_INTERPCASEMOD;
2564 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2565 assert(PL_lex_inwhat != OP_TRANSR);
2566 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2567 PL_linestr = PL_lex_repl;
2569 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2570 PL_bufend += SvCUR(PL_linestr);
2571 PL_last_lop = PL_last_uni = NULL;
2572 PL_lex_dojoin = FALSE;
2573 PL_lex_brackets = 0;
2574 PL_lex_allbrackets = 0;
2575 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2576 PL_lex_casemods = 0;
2577 *PL_lex_casestack = '\0';
2579 if (SvEVALED(PL_lex_repl)) {
2580 PL_lex_state = LEX_INTERPNORMAL;
2582 /* we don't clear PL_lex_repl here, so that we can check later
2583 whether this is an evalled subst; that means we rely on the
2584 logic to ensure sublex_done() is called again only via the
2585 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2588 PL_lex_state = LEX_INTERPCONCAT;
2598 PL_endwhite = newSVpvs("");
2599 sv_catsv(PL_endwhite, PL_thiswhite);
2603 sv_setpvs(PL_thistoken,"");
2605 PL_realtokenstart = -1;
2609 PL_bufend = SvPVX(PL_linestr);
2610 PL_bufend += SvCUR(PL_linestr);
2611 PL_expect = XOPERATOR;
2612 PL_sublex_info.sub_inwhat = 0;
2620 Extracts the next constant part of a pattern, double-quoted string,
2621 or transliteration. This is terrifying code.
2623 For example, in parsing the double-quoted string "ab\x63$d", it would
2624 stop at the '$' and return an OP_CONST containing 'abc'.
2626 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2627 processing a pattern (PL_lex_inpat is true), a transliteration
2628 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2630 Returns a pointer to the character scanned up to. If this is
2631 advanced from the start pointer supplied (i.e. if anything was
2632 successfully parsed), will leave an OP_CONST for the substring scanned
2633 in pl_yylval. Caller must intuit reason for not parsing further
2634 by looking at the next characters herself.
2638 \N{ABC} => \N{U+41.42.43}
2641 all other \-char, including \N and \N{ apart from \N{ABC}
2644 @ and $ where it appears to be a var, but not for $ as tail anchor
2649 In transliterations:
2650 characters are VERY literal, except for - not at the start or end
2651 of the string, which indicates a range. If the range is in bytes,
2652 scan_const expands the range to the full set of intermediate
2653 characters. If the range is in utf8, the hyphen is replaced with
2654 a certain range mark which will be handled by pmtrans() in op.c.
2656 In double-quoted strings:
2658 double-quoted style: \r and \n
2659 constants: \x31, etc.
2660 deprecated backrefs: \1 (in substitution replacements)
2661 case and quoting: \U \Q \E
2664 scan_const does *not* construct ops to handle interpolated strings.
2665 It stops processing as soon as it finds an embedded $ or @ variable
2666 and leaves it to the caller to work out what's going on.
2668 embedded arrays (whether in pattern or not) could be:
2669 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2671 $ in double-quoted strings must be the symbol of an embedded scalar.
2673 $ in pattern could be $foo or could be tail anchor. Assumption:
2674 it's a tail anchor if $ is the last thing in the string, or if it's
2675 followed by one of "()| \r\n\t"
2677 \1 (backreferences) are turned into $1 in substitutions
2679 The structure of the code is
2680 while (there's a character to process) {
2681 handle transliteration ranges
2682 skip regexp comments /(?#comment)/ and codes /(?{code})/
2683 skip #-initiated comments in //x patterns
2684 check for embedded arrays
2685 check for embedded scalars
2687 deprecate \1 in substitution replacements
2688 handle string-changing backslashes \l \U \Q \E, etc.
2689 switch (what was escaped) {
2690 handle \- in a transliteration (becomes a literal -)
2691 if a pattern and not \N{, go treat as regular character
2692 handle \132 (octal characters)
2693 handle \x15 and \x{1234} (hex characters)
2694 handle \N{name} (named characters, also \N{3,5} in a pattern)
2695 handle \cV (control characters)
2696 handle printf-style backslashes (\f, \r, \n, etc)
2699 } (end if backslash)
2700 handle regular character
2701 } (end while character to read)
2706 S_scan_const(pTHX_ char *start)
2709 char *send = PL_bufend; /* end of the constant */
2710 SV *sv = newSV(send - start); /* sv for the constant. See
2711 note below on sizing. */
2712 char *s = start; /* start of the constant */
2713 char *d = SvPVX(sv); /* destination for copies */
2714 bool dorange = FALSE; /* are we in a translit range? */
2715 bool didrange = FALSE; /* did we just finish a range? */
2716 bool in_charclass = FALSE; /* within /[...]/ */
2717 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2718 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2719 to be UTF8? But, this can
2720 show as true when the source
2721 isn't utf8, as for example
2722 when it is entirely composed
2725 /* Note on sizing: The scanned constant is placed into sv, which is
2726 * initialized by newSV() assuming one byte of output for every byte of
2727 * input. This routine expects newSV() to allocate an extra byte for a
2728 * trailing NUL, which this routine will append if it gets to the end of
2729 * the input. There may be more bytes of input than output (eg., \N{LATIN
2730 * CAPITAL LETTER A}), or more output than input if the constant ends up
2731 * recoded to utf8, but each time a construct is found that might increase
2732 * the needed size, SvGROW() is called. Its size parameter each time is
2733 * based on the best guess estimate at the time, namely the length used so
2734 * far, plus the length the current construct will occupy, plus room for
2735 * the trailing NUL, plus one byte for every input byte still unscanned */
2739 UV literal_endpoint = 0;
2740 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2743 PERL_ARGS_ASSERT_SCAN_CONST;
2745 assert(PL_lex_inwhat != OP_TRANSR);
2746 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2747 /* If we are doing a trans and we know we want UTF8 set expectation */
2748 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2749 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2753 while (s < send || dorange) {
2755 /* get transliterations out of the way (they're most literal) */
2756 if (PL_lex_inwhat == OP_TRANS) {
2757 /* expand a range A-Z to the full set of characters. AIE! */
2759 I32 i; /* current expanded character */
2760 I32 min; /* first character in range */
2761 I32 max; /* last character in range */
2772 char * const c = (char*)utf8_hop((U8*)d, -1);
2776 *c = (char)UTF_TO_NATIVE(0xff);
2777 /* mark the range as done, and continue */
2783 i = d - SvPVX_const(sv); /* remember current offset */
2786 SvLEN(sv) + (has_utf8 ?
2787 (512 - UTF_CONTINUATION_MARK +
2790 /* How many two-byte within 0..255: 128 in UTF-8,
2791 * 96 in UTF-8-mod. */
2793 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2795 d = SvPVX(sv) + i; /* refresh d after realloc */
2799 for (j = 0; j <= 1; j++) {
2800 char * const c = (char*)utf8_hop((U8*)d, -1);
2801 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2807 max = (U8)0xff; /* only to \xff */
2808 uvmax = uv; /* \x{100} to uvmax */
2810 d = c; /* eat endpoint chars */
2815 d -= 2; /* eat the first char and the - */
2816 min = (U8)*d; /* first char in range */
2817 max = (U8)d[1]; /* last char in range */
2825 "Invalid range \"%c-%c\" in transliteration operator",
2826 (char)min, (char)max);
2830 if (literal_endpoint == 2 &&
2831 ((isLOWER(min) && isLOWER(max)) ||
2832 (isUPPER(min) && isUPPER(max)))) {
2834 for (i = min; i <= max; i++)
2836 *d++ = NATIVE_TO_NEED(has_utf8,i);
2838 for (i = min; i <= max; i++)
2840 *d++ = NATIVE_TO_NEED(has_utf8,i);
2845 for (i = min; i <= max; i++)
2848 const U8 ch = (U8)NATIVE_TO_UTF(i);
2849 if (UNI_IS_INVARIANT(ch))
2852 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2853 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2862 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2864 *d++ = (char)UTF_TO_NATIVE(0xff);
2866 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2870 /* mark the range as done, and continue */
2874 literal_endpoint = 0;
2879 /* range begins (ignore - as first or last char) */
2880 else if (*s == '-' && s+1 < send && s != start) {
2883 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2890 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2900 literal_endpoint = 0;
2901 native_range = TRUE;
2906 /* if we get here, we're not doing a transliteration */
2908 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
2911 while (s1 >= start && *s1-- == '\\')
2914 in_charclass = TRUE;
2917 else if (*s == ']' && PL_lex_inpat && in_charclass) {
2920 while (s1 >= start && *s1-- == '\\')
2923 in_charclass = FALSE;
2926 /* skip for regexp comments /(?#comment)/, except for the last
2927 * char, which will be done separately.
2928 * Stop on (?{..}) and friends */
2930 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2932 while (s+1 < send && *s != ')')
2933 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2935 else if (!PL_lex_casemods && !in_charclass &&
2936 ( s[2] == '{' /* This should match regcomp.c */
2937 || (s[2] == '?' && s[3] == '{')))
2943 /* likewise skip #-initiated comments in //x patterns */
2944 else if (*s == '#' && PL_lex_inpat &&
2945 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2946 while (s+1 < send && *s != '\n')
2947 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2950 /* no further processing of single-quoted regex */
2951 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
2952 goto default_action;
2954 /* check for embedded arrays
2955 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2957 else if (*s == '@' && s[1]) {
2958 if (isALNUM_lazy_if(s+1,UTF))
2960 if (strchr(":'{$", s[1]))
2962 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2963 break; /* in regexp, neither @+ nor @- are interpolated */
2966 /* check for embedded scalars. only stop if we're sure it's a
2969 else if (*s == '$') {
2970 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2972 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2974 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2975 "Possible unintended interpolation of $\\ in regex");
2977 break; /* in regexp, $ might be tail anchor */
2981 /* End of else if chain - OP_TRANS rejoin rest */
2984 if (*s == '\\' && s+1 < send) {
2985 char* e; /* Can be used for ending '}', etc. */
2989 /* warn on \1 - \9 in substitution replacements, but note that \11
2990 * is an octal; and \19 is \1 followed by '9' */
2991 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2992 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2994 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2999 /* string-change backslash escapes */
3000 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3004 /* In a pattern, process \N, but skip any other backslash escapes.
3005 * This is because we don't want to translate an escape sequence
3006 * into a meta symbol and have the regex compiler use the meta
3007 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3008 * in spite of this, we do have to process \N here while the proper
3009 * charnames handler is in scope. See bugs #56444 and #62056.
3010 * There is a complication because \N in a pattern may also stand
3011 * for 'match a non-nl', and not mean a charname, in which case its
3012 * processing should be deferred to the regex compiler. To be a
3013 * charname it must be followed immediately by a '{', and not look
3014 * like \N followed by a curly quantifier, i.e., not something like
3015 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3017 else if (PL_lex_inpat
3020 || regcurly(s + 1)))
3022 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3023 goto default_action;
3028 /* quoted - in transliterations */
3030 if (PL_lex_inwhat == OP_TRANS) {
3038 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3039 "Unrecognized escape \\%c passed through",
3041 /* default action is to copy the quoted character */
3042 goto default_action;
3045 /* eg. \132 indicates the octal constant 0132 */
3046 case '0': case '1': case '2': case '3':
3047 case '4': case '5': case '6': case '7':
3051 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3054 goto NUM_ESCAPE_INSERT;
3056 /* eg. \o{24} indicates the octal constant \024 */
3062 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
3068 goto NUM_ESCAPE_INSERT;
3071 /* eg. \x24 indicates the hex constant 0x24 */
3077 bool valid = grok_bslash_x(s, &uv, &len, &error, 1);
3086 /* Insert oct or hex escaped character. There will always be
3087 * enough room in sv since such escapes will be longer than any
3088 * UTF-8 sequence they can end up as, except if they force us
3089 * to recode the rest of the string into utf8 */
3091 /* Here uv is the ordinal of the next character being added in
3092 * unicode (converted from native). */
3093 if (!UNI_IS_INVARIANT(uv)) {
3094 if (!has_utf8 && uv > 255) {
3095 /* Might need to recode whatever we have accumulated so
3096 * far if it contains any chars variant in utf8 or
3099 SvCUR_set(sv, d - SvPVX_const(sv));
3102 /* See Note on sizing above. */
3103 sv_utf8_upgrade_flags_grow(sv,
3104 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3105 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3106 d = SvPVX(sv) + SvCUR(sv);
3111 d = (char*)uvuni_to_utf8((U8*)d, uv);
3112 if (PL_lex_inwhat == OP_TRANS &&
3113 PL_sublex_info.sub_op) {
3114 PL_sublex_info.sub_op->op_private |=
3115 (PL_lex_repl ? OPpTRANS_FROM_UTF
3119 if (uv > 255 && !dorange)
3120 native_range = FALSE;
3133 /* In a non-pattern \N must be a named character, like \N{LATIN
3134 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3135 * mean to match a non-newline. For non-patterns, named
3136 * characters are converted to their string equivalents. In
3137 * patterns, named characters are not converted to their
3138 * ultimate forms for the same reasons that other escapes
3139 * aren't. Instead, they are converted to the \N{U+...} form
3140 * to get the value from the charnames that is in effect right
3141 * now, while preserving the fact that it was a named character
3142 * so that the regex compiler knows this */
3144 /* This section of code doesn't generally use the
3145 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3146 * a close examination of this macro and determined it is a
3147 * no-op except on utfebcdic variant characters. Every
3148 * character generated by this that would normally need to be
3149 * enclosed by this macro is invariant, so the macro is not
3150 * needed, and would complicate use of copy(). XXX There are
3151 * other parts of this file where the macro is used
3152 * inconsistently, but are saved by it being a no-op */
3154 /* The structure of this section of code (besides checking for
3155 * errors and upgrading to utf8) is:
3156 * Further disambiguate between the two meanings of \N, and if
3157 * not a charname, go process it elsewhere
3158 * If of form \N{U+...}, pass it through if a pattern;
3159 * otherwise convert to utf8
3160 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3161 * pattern; otherwise convert to utf8 */
3163 /* Here, s points to the 'N'; the test below is guaranteed to
3164 * succeed if we are being called on a pattern as we already
3165 * know from a test above that the next character is a '{'.
3166 * On a non-pattern \N must mean 'named sequence, which
3167 * requires braces */
3170 yyerror("Missing braces on \\N{}");
3175 /* If there is no matching '}', it is an error. */
3176 if (! (e = strchr(s, '}'))) {
3177 if (! PL_lex_inpat) {
3178 yyerror("Missing right brace on \\N{}");
3180 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3185 /* Here it looks like a named character */
3189 /* XXX This block is temporary code. \N{} implies that the
3190 * pattern is to have Unicode semantics, and therefore
3191 * currently has to be encoded in utf8. By putting it in
3192 * utf8 now, we save a whole pass in the regular expression
3193 * compiler. Once that code is changed so Unicode
3194 * semantics doesn't necessarily have to be in utf8, this
3195 * block should be removed. However, the code that parses
3196 * the output of this would have to be changed to not
3197 * necessarily expect utf8 */
3199 SvCUR_set(sv, d - SvPVX_const(sv));
3202 /* See Note on sizing above. */
3203 sv_utf8_upgrade_flags_grow(sv,
3204 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3205 /* 5 = '\N{' + cur char + NUL */
3206 (STRLEN)(send - s) + 5);
3207 d = SvPVX(sv) + SvCUR(sv);
3212 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3213 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3214 | PERL_SCAN_DISALLOW_PREFIX;
3217 /* For \N{U+...}, the '...' is a unicode value even on
3218 * EBCDIC machines */
3219 s += 2; /* Skip to next char after the 'U+' */
3221 uv = grok_hex(s, &len, &flags, NULL);
3222 if (len == 0 || len != (STRLEN)(e - s)) {
3223 yyerror("Invalid hexadecimal number in \\N{U+...}");
3230 /* On non-EBCDIC platforms, pass through to the regex
3231 * compiler unchanged. The reason we evaluated the
3232 * number above is to make sure there wasn't a syntax
3233 * error. But on EBCDIC we convert to native so
3234 * downstream code can continue to assume it's native
3236 s -= 5; /* Include the '\N{U+' */
3238 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3241 (unsigned int) UNI_TO_NATIVE(uv));
3243 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3247 else { /* Not a pattern: convert the hex to string */
3249 /* If destination is not in utf8, unconditionally
3250 * recode it to be so. This is because \N{} implies
3251 * Unicode semantics, and scalars have to be in utf8
3252 * to guarantee those semantics */
3254 SvCUR_set(sv, d - SvPVX_const(sv));
3257 /* See Note on sizing above. */
3258 sv_utf8_upgrade_flags_grow(
3260 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3261 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3262 d = SvPVX(sv) + SvCUR(sv);
3266 /* Add the string to the output */
3267 if (UNI_IS_INVARIANT(uv)) {
3270 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3273 else { /* Here is \N{NAME} but not \N{U+...}. */
3275 SV *res; /* result from charnames */
3276 const char *str; /* the string in 'res' */
3277 STRLEN len; /* its length */
3279 /* Get the value for NAME */
3280 res = newSVpvn(s, e - s);
3281 res = new_constant( NULL, 0, "charnames",
3282 /* includes all of: \N{...} */
3283 res, NULL, s - 3, e - s + 4 );
3285 /* Most likely res will be in utf8 already since the
3286 * standard charnames uses pack U, but a custom translator
3287 * can leave it otherwise, so make sure. XXX This can be
3288 * revisited to not have charnames use utf8 for characters
3289 * that don't need it when regexes don't have to be in utf8
3290 * for Unicode semantics. If doing so, remember EBCDIC */
3292 sv_utf8_upgrade(res);
3293 str = SvPV_const(res, len);
3295 /* Don't accept malformed input */
3296 if (! is_utf8_string((U8 *) str, len)) {
3297 yyerror("Malformed UTF-8 returned by \\N");
3299 else if (PL_lex_inpat) {
3301 if (! len) { /* The name resolved to an empty string */
3302 Copy("\\N{}", d, 4, char);
3306 /* In order to not lose information for the regex
3307 * compiler, pass the result in the specially made
3308 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3309 * the code points in hex of each character
3310 * returned by charnames */
3312 const char *str_end = str + len;
3313 STRLEN char_length; /* cur char's byte length */
3314 STRLEN output_length; /* and the number of bytes
3315 after this is translated
3317 const STRLEN off = d - SvPVX_const(sv);
3319 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3320 * max('U+', '.'); and 1 for NUL */
3321 char hex_string[2 * UTF8_MAXBYTES + 5];
3323 /* Get the first character of the result. */
3324 U32 uv = utf8n_to_uvuni((U8 *) str,
3329 /* The call to is_utf8_string() above hopefully
3330 * guarantees that there won't be an error. But
3331 * it's easy here to make sure. The function just
3332 * above warns and returns 0 if invalid utf8, but
3333 * it can also return 0 if the input is validly a
3334 * NUL. Disambiguate */
3335 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3336 uv = UNICODE_REPLACEMENT;
3339 /* Convert first code point to hex, including the
3340 * boiler plate before it. For all these, we
3341 * convert to native format so that downstream code
3342 * can continue to assume the input is native */
3344 my_snprintf(hex_string, sizeof(hex_string),
3346 (unsigned int) UNI_TO_NATIVE(uv));
3348 /* Make sure there is enough space to hold it */
3349 d = off + SvGROW(sv, off
3351 + (STRLEN)(send - e)
3352 + 2); /* '}' + NUL */
3354 Copy(hex_string, d, output_length, char);
3357 /* For each subsequent character, append dot and
3358 * its ordinal in hex */
3359 while ((str += char_length) < str_end) {
3360 const STRLEN off = d - SvPVX_const(sv);
3361 U32 uv = utf8n_to_uvuni((U8 *) str,
3365 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3366 uv = UNICODE_REPLACEMENT;
3370 my_snprintf(hex_string, sizeof(hex_string),
3372 (unsigned int) UNI_TO_NATIVE(uv));
3374 d = off + SvGROW(sv, off
3376 + (STRLEN)(send - e)
3377 + 2); /* '}' + NUL */
3378 Copy(hex_string, d, output_length, char);
3382 *d++ = '}'; /* Done. Add the trailing brace */
3385 else { /* Here, not in a pattern. Convert the name to a
3388 /* If destination is not in utf8, unconditionally
3389 * recode it to be so. This is because \N{} implies
3390 * Unicode semantics, and scalars have to be in utf8
3391 * to guarantee those semantics */
3393 SvCUR_set(sv, d - SvPVX_const(sv));
3396 /* See Note on sizing above. */
3397 sv_utf8_upgrade_flags_grow(sv,
3398 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3399 len + (STRLEN)(send - s) + 1);
3400 d = SvPVX(sv) + SvCUR(sv);
3402 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3404 /* See Note on sizing above. (NOTE: SvCUR() is not
3405 * set correctly here). */
3406 const STRLEN off = d - SvPVX_const(sv);
3407 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3409 Copy(str, d, len, char);
3414 /* Deprecate non-approved name syntax */
3415 if (ckWARN_d(WARN_DEPRECATED)) {
3416 bool problematic = FALSE;
3419 /* For non-ut8 input, look to see that the first
3420 * character is an alpha, then loop through the rest
3421 * checking that each is a continuation */
3423 if (! isALPHAU(*i)) problematic = TRUE;
3424 else for (i = s + 1; i < e; i++) {
3425 if (isCHARNAME_CONT(*i)) continue;
3431 /* Similarly for utf8. For invariants can check
3432 * directly. We accept anything above the latin1
3433 * range because it is immaterial to Perl if it is
3434 * correct or not, and is expensive to check. But
3435 * it is fairly easy in the latin1 range to convert
3436 * the variants into a single character and check
3438 if (UTF8_IS_INVARIANT(*i)) {
3439 if (! isALPHAU(*i)) problematic = TRUE;
3440 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3441 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3447 if (! problematic) for (i = s + UTF8SKIP(s);
3451 if (UTF8_IS_INVARIANT(*i)) {
3452 if (isCHARNAME_CONT(*i)) continue;
3453 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3455 } else if (isCHARNAME_CONT(
3457 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3466 /* The e-i passed to the final %.*s makes sure that
3467 * should the trailing NUL be missing that this
3468 * print won't run off the end of the string */
3469 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3470 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3471 (int)(i - s + 1), s, (int)(e - i), i + 1);
3475 } /* End \N{NAME} */
3478 native_range = FALSE; /* \N{} is defined to be Unicode */
3480 s = e + 1; /* Point to just after the '}' */
3483 /* \c is a control character */
3487 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3490 yyerror("Missing control char name in \\c");
3494 /* printf-style backslashes, formfeeds, newlines, etc */
3496 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3499 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3502 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3505 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3508 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3511 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3514 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3520 } /* end if (backslash) */
3527 /* If we started with encoded form, or already know we want it,
3528 then encode the next character */
3529 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3533 /* One might think that it is wasted effort in the case of the
3534 * source being utf8 (this_utf8 == TRUE) to take the next character
3535 * in the source, convert it to an unsigned value, and then convert
3536 * it back again. But the source has not been validated here. The
3537 * routine that does the conversion checks for errors like
3540 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3541 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3543 SvCUR_set(sv, d - SvPVX_const(sv));
3546 /* See Note on sizing above. */
3547 sv_utf8_upgrade_flags_grow(sv,
3548 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3549 need + (STRLEN)(send - s) + 1);
3550 d = SvPVX(sv) + SvCUR(sv);
3552 } else if (need > len) {
3553 /* encoded value larger than old, may need extra space (NOTE:
3554 * SvCUR() is not set correctly here). See Note on sizing
3556 const STRLEN off = d - SvPVX_const(sv);
3557 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3561 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3563 if (uv > 255 && !dorange)
3564 native_range = FALSE;
3568 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3570 } /* while loop to process each character */
3572 /* terminate the string and set up the sv */
3574 SvCUR_set(sv, d - SvPVX_const(sv));
3575 if (SvCUR(sv) >= SvLEN(sv))
3576 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3577 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3580 if (PL_encoding && !has_utf8) {
3581 sv_recode_to_utf8(sv, PL_encoding);
3587 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3588 PL_sublex_info.sub_op->op_private |=
3589 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3593 /* shrink the sv if we allocated more than we used */
3594 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3595 SvPV_shrink_to_cur(sv);
3598 /* return the substring (via pl_yylval) only if we parsed anything */
3599 if (s > PL_bufptr) {
3600 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3601 const char *const key = PL_lex_inpat ? "qr" : "q";
3602 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3606 if (PL_lex_inwhat == OP_TRANS) {
3609 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3612 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3620 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3623 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3630 * Returns TRUE if there's more to the expression (e.g., a subscript),
3633 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3635 * ->[ and ->{ return TRUE
3636 * { and [ outside a pattern are always subscripts, so return TRUE
3637 * if we're outside a pattern and it's not { or [, then return FALSE
3638 * if we're in a pattern and the first char is a {
3639 * {4,5} (any digits around the comma) returns FALSE
3640 * if we're in a pattern and the first char is a [
3642 * [SOMETHING] has a funky algorithm to decide whether it's a
3643 * character class or not. It has to deal with things like
3644 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3645 * anything else returns TRUE
3648 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3651 S_intuit_more(pTHX_ register char *s)
3655 PERL_ARGS_ASSERT_INTUIT_MORE;
3657 if (PL_lex_brackets)
3659 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3661 if (*s != '{' && *s != '[')
3666 /* In a pattern, so maybe we have {n,m}. */
3674 /* On the other hand, maybe we have a character class */
3677 if (*s == ']' || *s == '^')
3680 /* this is terrifying, and it works */
3681 int weight = 2; /* let's weigh the evidence */
3683 unsigned char un_char = 255, last_un_char;
3684 const char * const send = strchr(s,']');
3685 char tmpbuf[sizeof PL_tokenbuf * 4];
3687 if (!send) /* has to be an expression */
3690 Zero(seen,256,char);
3693 else if (isDIGIT(*s)) {
3695 if (isDIGIT(s[1]) && s[2] == ']')
3701 for (; s < send; s++) {
3702 last_un_char = un_char;
3703 un_char = (unsigned char)*s;
3708 weight -= seen[un_char] * 10;
3709 if (isALNUM_lazy_if(s+1,UTF)) {
3711 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3712 len = (int)strlen(tmpbuf);
3713 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3714 UTF ? SVf_UTF8 : 0, SVt_PV))
3719 else if (*s == '$' && s[1] &&
3720 strchr("[#!%*<>()-=",s[1])) {
3721 if (/*{*/ strchr("])} =",s[2]))
3730 if (strchr("wds]",s[1]))
3732 else if (seen[(U8)'\''] || seen[(U8)'"'])
3734 else if (strchr("rnftbxcav",s[1]))
3736 else if (isDIGIT(s[1])) {
3738 while (s[1] && isDIGIT(s[1]))
3748 if (strchr("aA01! ",last_un_char))
3750 if (strchr("zZ79~",s[1]))
3752 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3753 weight -= 5; /* cope with negative subscript */
3756 if (!isALNUM(last_un_char)
3757 && !(last_un_char == '$' || last_un_char == '@'
3758 || last_un_char == '&')
3759 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3764 if (keyword(tmpbuf, d - tmpbuf, 0))
3767 if (un_char == last_un_char + 1)
3769 weight -= seen[un_char];
3774 if (weight >= 0) /* probably a character class */
3784 * Does all the checking to disambiguate
3786 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3787 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3789 * First argument is the stuff after the first token, e.g. "bar".
3791 * Not a method if foo is a filehandle.
3792 * Not a method if foo is a subroutine prototyped to take a filehandle.
3793 * Not a method if it's really "Foo $bar"
3794 * Method if it's "foo $bar"
3795 * Not a method if it's really "print foo $bar"
3796 * Method if it's really "foo package::" (interpreted as package->foo)
3797 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3798 * Not a method if bar is a filehandle or package, but is quoted with
3803 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3806 char *s = start + (*start == '$');
3807 char tmpbuf[sizeof PL_tokenbuf];
3814 PERL_ARGS_ASSERT_INTUIT_METHOD;
3816 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3818 if (cv && SvPOK(cv)) {
3819 const char *proto = CvPROTO(cv);
3827 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3828 /* start is the beginning of the possible filehandle/object,
3829 * and s is the end of it
3830 * tmpbuf is a copy of it
3833 if (*start == '$') {
3834 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3835 isUPPER(*PL_tokenbuf))
3838 len = start - SvPVX(PL_linestr);
3842 start = SvPVX(PL_linestr) + len;
3846 return *s == '(' ? FUNCMETH : METHOD;
3848 if (!keyword(tmpbuf, len, 0)) {
3849 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3853 soff = s - SvPVX(PL_linestr);
3857 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3858 if (indirgv && GvCVu(indirgv))
3860 /* filehandle or package name makes it a method */
3861 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3863 soff = s - SvPVX(PL_linestr);
3866 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3867 return 0; /* no assumptions -- "=>" quotes bareword */
3869 start_force(PL_curforce);
3870 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3871 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3872 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3874 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
3875 ( UTF ? SVf_UTF8 : 0 )));
3880 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3882 return *s == '(' ? FUNCMETH : METHOD;
3888 /* Encoded script support. filter_add() effectively inserts a
3889 * 'pre-processing' function into the current source input stream.
3890 * Note that the filter function only applies to the current source file
3891 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3893 * The datasv parameter (which may be NULL) can be used to pass
3894 * private data to this instance of the filter. The filter function
3895 * can recover the SV using the FILTER_DATA macro and use it to
3896 * store private buffers and state information.
3898 * The supplied datasv parameter is upgraded to a PVIO type
3899 * and the IoDIRP/IoANY field is used to store the function pointer,
3900 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3901 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3902 * private use must be set using malloc'd pointers.
3906 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3915 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3916 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3918 if (!PL_rsfp_filters)
3919 PL_rsfp_filters = newAV();
3922 SvUPGRADE(datasv, SVt_PVIO);
3923 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3924 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3925 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3926 FPTR2DPTR(void *, IoANY(datasv)),
3927 SvPV_nolen(datasv)));
3928 av_unshift(PL_rsfp_filters, 1);
3929 av_store(PL_rsfp_filters, 0, datasv) ;
3931 !PL_parser->filtered
3932 && PL_parser->lex_flags & LEX_EVALBYTES
3933 && PL_bufptr < PL_bufend
3935 const char *s = PL_bufptr;
3936 while (s < PL_bufend) {
3938 SV *linestr = PL_parser->linestr;
3939 char *buf = SvPVX(linestr);
3940 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3941 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3942 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3943 STRLEN const linestart_pos = PL_parser->linestart - buf;
3944 STRLEN const last_uni_pos =
3945 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3946 STRLEN const last_lop_pos =
3947 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3948 av_push(PL_rsfp_filters, linestr);
3949 PL_parser->linestr =
3950 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3951 buf = SvPVX(PL_parser->linestr);
3952 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3953 PL_parser->bufptr = buf + bufptr_pos;
3954 PL_parser->oldbufptr = buf + oldbufptr_pos;
3955 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3956 PL_parser->linestart = buf + linestart_pos;
3957 if (PL_parser->last_uni)
3958 PL_parser->last_uni = buf + last_uni_pos;
3959 if (PL_parser->last_lop)
3960 PL_parser->last_lop = buf + last_lop_pos;
3961 SvLEN(linestr) = SvCUR(linestr);
3962 SvCUR(linestr) = s-SvPVX(linestr);
3963 PL_parser->filtered = 1;
3973 /* Delete most recently added instance of this filter function. */
3975 Perl_filter_del(pTHX_ filter_t funcp)
3980 PERL_ARGS_ASSERT_FILTER_DEL;
3983 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3984 FPTR2DPTR(void*, funcp)));
3986 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3988 /* if filter is on top of stack (usual case) just pop it off */
3989 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3990 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3991 sv_free(av_pop(PL_rsfp_filters));
3995 /* we need to search for the correct entry and clear it */
3996 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4000 /* Invoke the idxth filter function for the current rsfp. */
4001 /* maxlen 0 = read one text line */
4003 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4008 /* This API is bad. It should have been using unsigned int for maxlen.
4009 Not sure if we want to change the API, but if not we should sanity
4010 check the value here. */
4011 unsigned int correct_length
4020 PERL_ARGS_ASSERT_FILTER_READ;
4022 if (!PL_parser || !PL_rsfp_filters)
4024 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4025 /* Provide a default input filter to make life easy. */
4026 /* Note that we append to the line. This is handy. */
4027 DEBUG_P(PerlIO_printf(Perl_debug_log,
4028 "filter_read %d: from rsfp\n", idx));
4029 if (correct_length) {
4032 const int old_len = SvCUR(buf_sv);
4034 /* ensure buf_sv is large enough */
4035 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4036 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4037 correct_length)) <= 0) {
4038 if (PerlIO_error(PL_rsfp))
4039 return -1; /* error */
4041 return 0 ; /* end of file */
4043 SvCUR_set(buf_sv, old_len + len) ;
4044 SvPVX(buf_sv)[old_len + len] = '\0';
4047 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4048 if (PerlIO_error(PL_rsfp))
4049 return -1; /* error */
4051 return 0 ; /* end of file */
4054 return SvCUR(buf_sv);
4056 /* Skip this filter slot if filter has been deleted */
4057 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4058 DEBUG_P(PerlIO_printf(Perl_debug_log,
4059 "filter_read %d: skipped (filter deleted)\n",
4061 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4063 if (SvTYPE(datasv) != SVt_PVIO) {
4064 if (correct_length) {
4066 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4067 if (!remainder) return 0; /* eof */
4068 if (correct_length > remainder) correct_length = remainder;
4069 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4070 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4073 const char *s = SvEND(datasv);
4074 const char *send = SvPVX(datasv) + SvLEN(datasv);
4082 if (s == send) return 0; /* eof */
4083 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4084 SvCUR_set(datasv, s-SvPVX(datasv));
4086 return SvCUR(buf_sv);
4088 /* Get function pointer hidden within datasv */
4089 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4090 DEBUG_P(PerlIO_printf(Perl_debug_log,
4091 "filter_read %d: via function %p (%s)\n",
4092 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4093 /* Call function. The function is expected to */
4094 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4095 /* Return: <0:error, =0:eof, >0:not eof */
4096 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4100 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4104 PERL_ARGS_ASSERT_FILTER_GETS;
4106 #ifdef PERL_CR_FILTER
4107 if (!PL_rsfp_filters) {
4108 filter_add(S_cr_textfilter,NULL);
4111 if (PL_rsfp_filters) {
4113 SvCUR_set(sv, 0); /* start with empty line */
4114 if (FILTER_READ(0, sv, 0) > 0)
4115 return ( SvPVX(sv) ) ;
4120 return (sv_gets(sv, PL_rsfp, append));
4124 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4129 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4131 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4135 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4136 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4138 return GvHV(gv); /* Foo:: */
4141 /* use constant CLASS => 'MyClass' */
4142 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4143 if (gv && GvCV(gv)) {
4144 SV * const sv = cv_const_sv(GvCV(gv));
4146 pkgname = SvPV_const(sv, len);
4149 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4153 * S_readpipe_override
4154 * Check whether readpipe() is overridden, and generates the appropriate
4155 * optree, provided sublex_start() is called afterwards.
4158 S_readpipe_override(pTHX)
4161 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4162 pl_yylval.ival = OP_BACKTICK;
4164 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4166 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4167 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4168 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4170 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4171 op_append_elem(OP_LIST,
4172 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4173 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4180 * The intent of this yylex wrapper is to minimize the changes to the
4181 * tokener when we aren't interested in collecting madprops. It remains
4182 * to be seen how successful this strategy will be...
4189 char *s = PL_bufptr;
4191 /* make sure PL_thiswhite is initialized */
4195 /* previous token ate up our whitespace? */
4196 if (!PL_lasttoke && PL_nextwhite) {
4197 PL_thiswhite = PL_nextwhite;
4201 /* isolate the token, and figure out where it is without whitespace */
4202 PL_realtokenstart = -1;
4206 assert(PL_curforce < 0);
4208 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4209 if (!PL_thistoken) {
4210 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4211 PL_thistoken = newSVpvs("");
4213 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4214 PL_thistoken = newSVpvn(tstart, s - tstart);
4217 if (PL_thismad) /* install head */
4218 CURMAD('X', PL_thistoken);
4221 /* last whitespace of a sublex? */
4222 if (optype == ')' && PL_endwhite) {
4223 CURMAD('X', PL_endwhite);
4228 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4229 if (!PL_thiswhite && !PL_endwhite && !optype) {
4230 sv_free(PL_thistoken);
4235 /* put off final whitespace till peg */
4236 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4237 PL_nextwhite = PL_thiswhite;
4240 else if (PL_thisopen) {
4241 CURMAD('q', PL_thisopen);
4243 sv_free(PL_thistoken);
4247 /* Store actual token text as madprop X */
4248 CURMAD('X', PL_thistoken);
4252 /* add preceding whitespace as madprop _ */
4253 CURMAD('_', PL_thiswhite);
4257 /* add quoted material as madprop = */
4258 CURMAD('=', PL_thisstuff);
4262 /* add terminating quote as madprop Q */
4263 CURMAD('Q', PL_thisclose);
4267 /* special processing based on optype */
4271 /* opval doesn't need a TOKEN since it can already store mp */
4282 if (pl_yylval.opval)
4283 append_madprops(PL_thismad, pl_yylval.opval, 0);
4291 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4300 /* remember any fake bracket that lexer is about to discard */
4301 if (PL_lex_brackets == 1 &&
4302 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4305 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4308 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4309 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4312 break; /* don't bother looking for trailing comment */
4321 /* attach a trailing comment to its statement instead of next token */
4325 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4327 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4329 if (*s == '\n' || *s == '#') {
4330 while (s < PL_bufend && *s != '\n')
4334 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4335 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4348 /* Create new token struct. Note: opvals return early above. */
4349 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4356 S_tokenize_use(pTHX_ int is_use, char *s) {
4359 PERL_ARGS_ASSERT_TOKENIZE_USE;
4361 if (PL_expect != XSTATE)
4362 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4363 is_use ? "use" : "no"));
4366 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4367 s = force_version(s, TRUE);
4368 if (*s == ';' || *s == '}'
4369 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4370 start_force(PL_curforce);
4371 NEXTVAL_NEXTTOKE.opval = NULL;
4374 else if (*s == 'v') {
4375 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4376 s = force_version(s, FALSE);
4380 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4381 s = force_version(s, FALSE);
4383 pl_yylval.ival = is_use;
4387 static const char* const exp_name[] =
4388 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4389 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4393 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4395 S_word_takes_any_delimeter(char *p, STRLEN len)
4397 return (len == 1 && strchr("msyq", p[0])) ||
4399 (p[0] == 't' && p[1] == 'r') ||
4400 (p[0] == 'q' && strchr("qwxr", p[1]))));
4406 Works out what to call the token just pulled out of the input
4407 stream. The yacc parser takes care of taking the ops we return and
4408 stitching them into a tree.
4411 The type of the next token
4414 Switch based on the current state:
4415 - if we already built the token before, use it
4416 - if we have a case modifier in a string, deal with that
4417 - handle other cases of interpolation inside a string
4418 - scan the next line if we are inside a format
4419 In the normal state switch on the next character:
4421 if alphabetic, go to key lookup
4422 unrecoginized character - croak
4423 - 0/4/26: handle end-of-line or EOF
4424 - cases for whitespace
4425 - \n and #: handle comments and line numbers
4426 - various operators, brackets and sigils
4429 - 'v': vstrings (or go to key lookup)
4430 - 'x' repetition operator (or go to key lookup)
4431 - other ASCII alphanumerics (key lookup begins here):
4434 scan built-in keyword (but do nothing with it yet)
4435 check for statement label
4436 check for lexical subs
4437 goto just_a_word if there is one
4438 see whether built-in keyword is overridden
4439 switch on keyword number:
4440 - default: just_a_word:
4441 not a built-in keyword; handle bareword lookup
4442 disambiguate between method and sub call
4443 fall back to bareword
4444 - cases for built-in keywords
4449 #pragma segment Perl_yylex
4455 char *s = PL_bufptr;
4462 /* orig_keyword, gvp, and gv are initialized here because
4463 * jump to the label just_a_word_zero can bypass their
4464 * initialization later. */
4465 I32 orig_keyword = 0;
4470 SV* tmp = newSVpvs("");
4471 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4472 (IV)CopLINE(PL_curcop),
4473 lex_state_names[PL_lex_state],
4474 exp_name[PL_expect],
4475 pv_display(tmp, s, strlen(s), 0, 60));
4479 switch (PL_lex_state) {
4481 case LEX_NORMAL: /* Some compilers will produce faster */
4482 case LEX_INTERPNORMAL: /* code if we comment these out. */
4486 /* when we've already built the next token, just pull it out of the queue */
4490 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4492 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4493 PL_nexttoke[PL_lasttoke].next_mad = 0;
4494 if (PL_thismad && PL_thismad->mad_key == '_') {
4495 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4496 PL_thismad->mad_val = 0;
4497 mad_free(PL_thismad);
4502 PL_lex_state = PL_lex_defer;
4503 PL_expect = PL_lex_expect;
4504 PL_lex_defer = LEX_NORMAL;
4505 if (!PL_nexttoke[PL_lasttoke].next_type)
4510 pl_yylval = PL_nextval[PL_nexttoke];
4512 PL_lex_state = PL_lex_defer;
4513 PL_expect = PL_lex_expect;
4514 PL_lex_defer = LEX_NORMAL;
4520 next_type = PL_nexttoke[PL_lasttoke].next_type;
4522 next_type = PL_nexttype[PL_nexttoke];
4524 if (next_type & (7<<24)) {
4525 if (next_type & (1<<24)) {
4526 if (PL_lex_brackets > 100)
4527 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4528 PL_lex_brackstack[PL_lex_brackets++] =
4529 (char) ((next_type >> 16) & 0xff);
4531 if (next_type & (2<<24))
4532 PL_lex_allbrackets++;
4533 if (next_type & (4<<24))
4534 PL_lex_allbrackets--;
4535 next_type &= 0xffff;
4537 if (S_is_opval_token(next_type) && pl_yylval.opval)
4538 pl_yylval.opval->op_savefree = 0; /* release */
4539 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4542 /* interpolated case modifiers like \L \U, including \Q and \E.
4543 when we get here, PL_bufptr is at the \
4545 case LEX_INTERPCASEMOD:
4547 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4549 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4550 PL_bufptr, PL_bufend, *PL_bufptr);
4552 /* handle \E or end of string */
4553 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4555 if (PL_lex_casemods) {
4556 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4557 PL_lex_casestack[PL_lex_casemods] = '\0';
4559 if (PL_bufptr != PL_bufend
4560 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4561 || oldmod == 'F')) {
4563 PL_lex_state = LEX_INTERPCONCAT;
4566 PL_thistoken = newSVpvs("\\E");
4569 PL_lex_allbrackets--;
4572 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4573 /* Got an unpaired \E */
4574 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4575 "Useless use of \\E");
4578 while (PL_bufptr != PL_bufend &&
4579 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4581 PL_thiswhite = newSVpvs("");
4582 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4586 if (PL_bufptr != PL_bufend)
4589 PL_lex_state = LEX_INTERPCONCAT;
4593 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4594 "### Saw case modifier\n"); });
4596 if (s[1] == '\\' && s[2] == 'E') {
4599 PL_thiswhite = newSVpvs("");
4600 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4603 PL_lex_state = LEX_INTERPCONCAT;
4608 if (!PL_madskills) /* when just compiling don't need correct */
4609 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4610 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4611 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4612 (strchr(PL_lex_casestack, 'L')
4613 || strchr(PL_lex_casestack, 'U')
4614 || strchr(PL_lex_casestack, 'F'))) {
4615 PL_lex_casestack[--PL_lex_casemods] = '\0';
4616 PL_lex_allbrackets--;
4619 if (PL_lex_casemods > 10)
4620 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4621 PL_lex_casestack[PL_lex_casemods++] = *s;
4622 PL_lex_casestack[PL_lex_casemods] = '\0';
4623 PL_lex_state = LEX_INTERPCONCAT;
4624 start_force(PL_curforce);
4625 NEXTVAL_NEXTTOKE.ival = 0;
4626 force_next((2<<24)|'(');
4627 start_force(PL_curforce);
4629 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4631 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4633 NEXTVAL_NEXTTOKE.ival = OP_LC;
4635 NEXTVAL_NEXTTOKE.ival = OP_UC;
4637 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4639 NEXTVAL_NEXTTOKE.ival = OP_FC;
4641 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4643 SV* const tmpsv = newSVpvs("\\ ");
4644 /* replace the space with the character we want to escape
4646 SvPVX(tmpsv)[1] = *s;
4652 if (PL_lex_starts) {
4658 sv_free(PL_thistoken);
4659 PL_thistoken = newSVpvs("");
4662 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4663 if (PL_lex_casemods == 1 && PL_lex_inpat)
4672 case LEX_INTERPPUSH:
4673 return REPORT(sublex_push());
4675 case LEX_INTERPSTART:
4676 if (PL_bufptr == PL_bufend)
4677 return REPORT(sublex_done());
4678 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4679 "### Interpolated variable\n"); });
4681 PL_lex_dojoin = (*PL_bufptr == '@');
4682 PL_lex_state = LEX_INTERPNORMAL;
4683 if (PL_lex_dojoin) {
4684 start_force(PL_curforce);
4685 NEXTVAL_NEXTTOKE.ival = 0;
4687 start_force(PL_curforce);
4688 force_ident("\"", '$');
4689 start_force(PL_curforce);
4690 NEXTVAL_NEXTTOKE.ival = 0;
4692 start_force(PL_curforce);
4693 NEXTVAL_NEXTTOKE.ival = 0;
4694 force_next((2<<24)|'(');
4695 start_force(PL_curforce);
4696 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4699 /* Convert (?{...}) and friends to 'do {...}' */
4700 if (PL_lex_inpat && *PL_bufptr == '(') {
4701 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4703 if (*PL_bufptr != '{')
4705 start_force(PL_curforce);
4706 /* XXX probably need a CURMAD(something) here */
4707 PL_expect = XTERMBLOCK;
4711 if (PL_lex_starts++) {
4716 sv_free(PL_thistoken);
4717 PL_thistoken = newSVpvs("");
4720 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4721 if (!PL_lex_casemods && PL_lex_inpat)
4728 case LEX_INTERPENDMAYBE:
4729 if (intuit_more(PL_bufptr)) {
4730 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4736 if (PL_lex_dojoin) {
4737 PL_lex_dojoin = FALSE;
4738 PL_lex_state = LEX_INTERPCONCAT;
4742 sv_free(PL_thistoken);
4743 PL_thistoken = newSVpvs("");
4746 PL_lex_allbrackets--;
4749 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4750 && SvEVALED(PL_lex_repl))
4752 if (PL_bufptr != PL_bufend)
4753 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4756 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4757 re_eval_str. If the here-doc body’s length equals the previous
4758 value of re_eval_start, re_eval_start will now be null. So
4759 check re_eval_str as well. */
4760 if (PL_parser->lex_shared->re_eval_start
4761 || PL_parser->lex_shared->re_eval_str) {
4763 if (*PL_bufptr != ')')
4764 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4766 /* having compiled a (?{..}) expression, return the original
4767 * text too, as a const */
4768 if (PL_parser->lex_shared->re_eval_str) {
4769 sv = PL_parser->lex_shared->re_eval_str;
4770 PL_parser->lex_shared->re_eval_str = NULL;
4772 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4773 SvPV_shrink_to_cur(sv);
4775 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4776 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4777 start_force(PL_curforce);
4778 /* XXX probably need a CURMAD(something) here */
4779 NEXTVAL_NEXTTOKE.opval =
4780 (OP*)newSVOP(OP_CONST, 0,
4783 PL_parser->lex_shared->re_eval_start = NULL;
4789 case LEX_INTERPCONCAT:
4791 if (PL_lex_brackets)
4792 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4793 (long) PL_lex_brackets);
4795 if (PL_bufptr == PL_bufend)
4796 return REPORT(sublex_done());
4798 /* m'foo' still needs to be parsed for possible (?{...}) */
4799 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4800 SV *sv = newSVsv(PL_linestr);
4802 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4806 s = scan_const(PL_bufptr);
4808 PL_lex_state = LEX_INTERPCASEMOD;
4810 PL_lex_state = LEX_INTERPSTART;
4813 if (s != PL_bufptr) {
4814 start_force(PL_curforce);
4816 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4818 NEXTVAL_NEXTTOKE = pl_yylval;
4821 if (PL_lex_starts++) {
4825 sv_free(PL_thistoken);
4826 PL_thistoken = newSVpvs("");
4829 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4830 if (!PL_lex_casemods && PL_lex_inpat)
4843 s = scan_formline(PL_bufptr);
4844 if (!PL_lex_formbrack)
4854 PL_oldoldbufptr = PL_oldbufptr;
4860 sv_free(PL_thistoken);
4863 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4867 if (isIDFIRST_lazy_if(s,UTF))
4870 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4871 const char *c = UTF ? savepv(sv_uni_display(dsv, newSVpvn_flags(s,
4873 SVs_TEMP | SVf_UTF8),
4874 10, UNI_DISPLAY_ISPRINT))
4875 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4876 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4877 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4878 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4886 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"SVf"<-- HERE near column %d", c, SVfARG(dsv), (int) len + 1);
4890 goto fake_eof; /* emulate EOF on ^D or ^Z */
4896 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4899 if (PL_lex_brackets &&
4900 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4901 yyerror((const char *)
4903 ? "Format not terminated"
4904 : "Missing right curly or square bracket"));
4906 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4907 "### Tokener got EOF\n");
4911 if (s++ < PL_bufend)
4912 goto retry; /* ignore stray nulls */
4915 if (!PL_in_eval && !PL_preambled) {
4916 PL_preambled = TRUE;
4922 /* Generate a string of Perl code to load the debugger.
4923 * If PERL5DB is set, it will return the contents of that,
4924 * otherwise a compile-time require of perl5db.pl. */
4926 const char * const pdb = PerlEnv_getenv("PERL5DB");
4929 sv_setpv(PL_linestr, pdb);
4930 sv_catpvs(PL_linestr,";");
4932 SETERRNO(0,SS_NORMAL);
4933 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4936 sv_setpvs(PL_linestr,"");
4937 if (PL_preambleav) {
4938 SV **svp = AvARRAY(PL_preambleav);
4939 SV **const end = svp + AvFILLp(PL_preambleav);
4941 sv_catsv(PL_linestr, *svp);
4943 sv_catpvs(PL_linestr, ";");
4945 sv_free(MUTABLE_SV(PL_preambleav));
4946 PL_preambleav = NULL;
4949 sv_catpvs(PL_linestr,
4950 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4951 if (PL_minus_n || PL_minus_p) {
4952 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4954 sv_catpvs(PL_linestr,"chomp;");
4957 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4958 || *PL_splitstr == '"')
4959 && strchr(PL_splitstr + 1, *PL_splitstr))
4960 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4962 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4963 bytes can be used as quoting characters. :-) */
4964 const char *splits = PL_splitstr;
4965 sv_catpvs(PL_linestr, "our @F=split(q\0");
4968 if (*splits == '\\')
4969 sv_catpvn(PL_linestr, splits, 1);
4970 sv_catpvn(PL_linestr, splits, 1);
4971 } while (*splits++);
4972 /* This loop will embed the trailing NUL of
4973 PL_linestr as the last thing it does before
4975 sv_catpvs(PL_linestr, ");");
4979 sv_catpvs(PL_linestr,"our @F=split(' ');");
4982 sv_catpvs(PL_linestr, "\n");
4983 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4984 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4985 PL_last_lop = PL_last_uni = NULL;
4986 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4987 update_debugger_info(PL_linestr, NULL, 0);
4992 bof = PL_rsfp ? TRUE : FALSE;
4995 fake_eof = LEX_FAKE_EOF;
4997 PL_bufptr = PL_bufend;
4998 COPLINE_INC_WITH_HERELINES;
4999 if (!lex_next_chunk(fake_eof)) {
5000 CopLINE_dec(PL_curcop);
5002 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
5004 CopLINE_dec(PL_curcop);
5007 PL_realtokenstart = -1;
5010 /* If it looks like the start of a BOM or raw UTF-16,
5011 * check if it in fact is. */
5012 if (bof && PL_rsfp &&
5017 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
5018 bof = (offset == (Off_t)SvCUR(PL_linestr));
5019 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
5020 /* offset may include swallowed CR */
5022 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
5025 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5026 s = swallow_bom((U8*)s);
5029 if (PL_parser->in_pod) {
5030 /* Incest with pod. */
5033 sv_catsv(PL_thiswhite, PL_linestr);
5035 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
5036 sv_setpvs(PL_linestr, "");
5037 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5038 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5039 PL_last_lop = PL_last_uni = NULL;
5040 PL_parser->in_pod = 0;
5043 if (PL_rsfp || PL_parser->filtered)
5045 } while (PL_parser->in_pod);
5046 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
5047 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5048 PL_last_lop = PL_last_uni = NULL;
5049 if (CopLINE(PL_curcop) == 1) {
5050 while (s < PL_bufend && isSPACE(*s))
5052 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
5056 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
5060 if (*s == '#' && *(s+1) == '!')
5062 #ifdef ALTERNATE_SHEBANG
5064 static char const as[] = ALTERNATE_SHEBANG;
5065 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5066 d = s + (sizeof(as) - 1);
5068 #endif /* ALTERNATE_SHEBANG */
5077 while (*d && !isSPACE(*d))
5081 #ifdef ARG_ZERO_IS_SCRIPT
5082 if (ipathend > ipath) {
5084 * HP-UX (at least) sets argv[0] to the script name,
5085 * which makes $^X incorrect. And Digital UNIX and Linux,
5086 * at least, set argv[0] to the basename of the Perl
5087 * interpreter. So, having found "#!", we'll set it right.
5089 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5091 assert(SvPOK(x) || SvGMAGICAL(x));
5092 if (sv_eq(x, CopFILESV(PL_curcop))) {
5093 sv_setpvn(x, ipath, ipathend - ipath);
5099 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
5100 const char * const lstart = SvPV_const(x,llen);
5102 bstart += blen - llen;
5103 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5104 sv_setpvn(x, ipath, ipathend - ipath);
5109 TAINT_NOT; /* $^X is always tainted, but that's OK */
5111 #endif /* ARG_ZERO_IS_SCRIPT */
5116 d = instr(s,"perl -");
5118 d = instr(s,"perl");
5120 /* avoid getting into infinite loops when shebang
5121 * line contains "Perl" rather than "perl" */
5123 for (d = ipathend-4; d >= ipath; --d) {
5124 if ((*d == 'p' || *d == 'P')
5125 && !ibcmp(d, "perl", 4))
5135 #ifdef ALTERNATE_SHEBANG
5137 * If the ALTERNATE_SHEBANG on this system starts with a
5138 * character that can be part of a Perl expression, then if
5139 * we see it but not "perl", we're probably looking at the
5140 * start of Perl code, not a request to hand off to some
5141 * other interpreter. Similarly, if "perl" is there, but
5142 * not in the first 'word' of the line, we assume the line
5143 * contains the start of the Perl program.
5145 if (d && *s != '#') {
5146 const char *c = ipath;
5147 while (*c && !strchr("; \t\r\n\f\v#", *c))
5150 d = NULL; /* "perl" not in first word; ignore */
5152 *s = '#'; /* Don't try to parse shebang line */
5154 #endif /* ALTERNATE_SHEBANG */
5159 !instr(s,"indir") &&
5160 instr(PL_origargv[0],"perl"))
5167 while (s < PL_bufend && isSPACE(*s))
5169 if (s < PL_bufend) {
5170 Newx(newargv,PL_origargc+3,char*);
5172 while (s < PL_bufend && !isSPACE(*s))
5175 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5178 newargv = PL_origargv;
5181 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5183 Perl_croak(aTHX_ "Can't exec %s", ipath);
5186 while (*d && !isSPACE(*d))
5188 while (SPACE_OR_TAB(*d))
5192 const bool switches_done = PL_doswitches;
5193 const U32 oldpdb = PL_perldb;
5194 const bool oldn = PL_minus_n;
5195 const bool oldp = PL_minus_p;
5199 bool baduni = FALSE;
5201 const char *d2 = d1 + 1;
5202 if (parse_unicode_opts((const char **)&d2)
5206 if (baduni || *d1 == 'M' || *d1 == 'm') {
5207 const char * const m = d1;
5208 while (*d1 && !isSPACE(*d1))
5210 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5213 d1 = moreswitches(d1);
5215 if (PL_doswitches && !switches_done) {
5216 int argc = PL_origargc;
5217 char **argv = PL_origargv;
5220 } while (argc && argv[0][0] == '-' && argv[0][1]);
5221 init_argv_symbols(argc,argv);
5223 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5224 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5225 /* if we have already added "LINE: while (<>) {",
5226 we must not do it again */
5228 sv_setpvs(PL_linestr, "");
5229 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5230 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5231 PL_last_lop = PL_last_uni = NULL;
5232 PL_preambled = FALSE;
5233 if (PERLDB_LINE || PERLDB_SAVESRC)
5234 (void)gv_fetchfile(PL_origfilename);
5241 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5242 PL_lex_state = LEX_FORMLINE;
5243 start_force(PL_curforce);
5244 NEXTVAL_NEXTTOKE.ival = 0;
5245 force_next(FORMRBRACK);
5250 #ifdef PERL_STRICT_CR
5251 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5253 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5255 case ' ': case '\t': case '\f': case 013:
5257 PL_realtokenstart = -1;
5259 PL_thiswhite = newSVpvs("");
5260 sv_catpvn(PL_thiswhite, s, 1);
5267 PL_realtokenstart = -1;
5271 if (PL_lex_state != LEX_NORMAL ||
5272 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5273 if (*s == '#' && s == PL_linestart && PL_in_eval
5274 && !PL_rsfp && !PL_parser->filtered) {
5275 /* handle eval qq[#line 1 "foo"\n ...] */
5276 CopLINE_dec(PL_curcop);
5279 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5281 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5285 const bool in_comment = *s == '#';
5287 while (d < PL_bufend && *d != '\n')
5291 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5292 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5296 PL_thiswhite = newSVpvn(s, d - s);
5299 if (in_comment && d == PL_bufend
5300 && PL_lex_state == LEX_INTERPNORMAL
5301 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5302 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5305 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5306 PL_lex_state = LEX_FORMLINE;
5307 start_force(PL_curforce);
5308 NEXTVAL_NEXTTOKE.ival = 0;
5309 force_next(FORMRBRACK);
5315 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5316 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5319 TOKEN(PEG); /* make sure any #! line is accessible */
5324 /* if (PL_madskills && PL_lex_formbrack) { */
5326 while (d < PL_bufend && *d != '\n')
5330 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5331 Perl_croak(aTHX_ "panic: input overflow");
5332 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5334 PL_thiswhite = newSVpvs("");
5335 if (CopLINE(PL_curcop) == 1) {
5336 sv_setpvs(PL_thiswhite, "");
5339 sv_catpvn(PL_thiswhite, s, d - s);
5353 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5361 while (s < PL_bufend && SPACE_OR_TAB(*s))
5364 if (strnEQ(s,"=>",2)) {
5365 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5366 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5367 OPERATOR('-'); /* unary minus */
5369 PL_last_uni = PL_oldbufptr;
5371 case 'r': ftst = OP_FTEREAD; break;
5372 case 'w': ftst = OP_FTEWRITE; break;
5373 case 'x': ftst = OP_FTEEXEC; break;
5374 case 'o': ftst = OP_FTEOWNED; break;
5375 case 'R': ftst = OP_FTRREAD; break;
5376 case 'W': ftst = OP_FTRWRITE; break;
5377 case 'X': ftst = OP_FTREXEC; break;
5378 case 'O': ftst = OP_FTROWNED; break;
5379 case 'e': ftst = OP_FTIS; break;
5380 case 'z': ftst = OP_FTZERO; break;
5381 case 's': ftst = OP_FTSIZE; break;
5382 case 'f': ftst = OP_FTFILE; break;
5383 case 'd': ftst = OP_FTDIR; break;
5384 case 'l': ftst = OP_FTLINK; break;
5385 case 'p': ftst = OP_FTPIPE; break;
5386 case 'S': ftst = OP_FTSOCK; break;
5387 case 'u': ftst = OP_FTSUID; break;
5388 case 'g': ftst = OP_FTSGID; break;
5389 case 'k': ftst = OP_FTSVTX; break;
5390 case 'b': ftst = OP_FTBLK; break;
5391 case 'c': ftst = OP_FTCHR; break;
5392 case 't': ftst = OP_FTTTY; break;
5393 case 'T': ftst = OP_FTTEXT; break;
5394 case 'B': ftst = OP_FTBINARY; break;
5395 case 'M': case 'A': case 'C':
5396 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5398 case 'M': ftst = OP_FTMTIME; break;
5399 case 'A': ftst = OP_FTATIME; break;
5400 case 'C': ftst = OP_FTCTIME; break;
5408 PL_last_lop_op = (OPCODE)ftst;
5409 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5410 "### Saw file test %c\n", (int)tmp);
5415 /* Assume it was a minus followed by a one-letter named
5416 * subroutine call (or a -bareword), then. */
5417 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5418 "### '-%c' looked like a file test but was not\n",
5425 const char tmp = *s++;
5428 if (PL_expect == XOPERATOR)
5433 else if (*s == '>') {
5436 if (isIDFIRST_lazy_if(s,UTF)) {
5437 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5445 if (PL_expect == XOPERATOR) {
5446 if (*s == '=' && !PL_lex_allbrackets &&
5447 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5454 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5456 OPERATOR('-'); /* unary minus */
5462 const char tmp = *s++;
5465 if (PL_expect == XOPERATOR)
5470 if (PL_expect == XOPERATOR) {
5471 if (*s == '=' && !PL_lex_allbrackets &&
5472 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5479 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5486 if (PL_expect != XOPERATOR) {
5487 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5488 PL_expect = XOPERATOR;
5489 force_ident(PL_tokenbuf, '*');
5497 if (*s == '=' && !PL_lex_allbrackets &&
5498 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5504 if (*s == '=' && !PL_lex_allbrackets &&
5505 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5512 if (PL_expect == XOPERATOR) {
5513 if (s[1] == '=' && !PL_lex_allbrackets &&
5514 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5519 PL_tokenbuf[0] = '%';
5520 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5521 sizeof PL_tokenbuf - 1, FALSE);
5522 if (!PL_tokenbuf[1]) {
5525 PL_expect = XOPERATOR;
5526 force_ident_maybe_lex('%');
5530 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5531 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5536 if (PL_lex_brackets > 100)
5537 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5538 PL_lex_brackstack[PL_lex_brackets++] = 0;
5539 PL_lex_allbrackets++;
5541 const char tmp = *s++;
5546 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5548 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5556 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5563 goto just_a_word_zero_gv;
5566 switch (PL_expect) {
5572 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5574 PL_bufptr = s; /* update in case we back off */
5577 "Use of := for an empty attribute list is not allowed");
5584 PL_expect = XTERMBLOCK;
5587 stuffstart = s - SvPVX(PL_linestr) - 1;
5591 while (isIDFIRST_lazy_if(s,UTF)) {
5594 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5595 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5596 if (tmp < 0) tmp = -tmp;
5611 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5613 d = scan_str(d,TRUE,TRUE,FALSE);
5615 /* MUST advance bufptr here to avoid bogus
5616 "at end of line" context messages from yyerror().
5618 PL_bufptr = s + len;
5619 yyerror("Unterminated attribute parameter in attribute list");
5623 return REPORT(0); /* EOF indicator */
5627 sv_catsv(sv, PL_lex_stuff);
5628 attrs = op_append_elem(OP_LIST, attrs,
5629 newSVOP(OP_CONST, 0, sv));
5630 SvREFCNT_dec(PL_lex_stuff);
5631 PL_lex_stuff = NULL;
5634 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5636 if (PL_in_my == KEY_our) {
5637 deprecate(":unique");
5640 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5643 /* NOTE: any CV attrs applied here need to be part of
5644 the CVf_BUILTIN_ATTRS define in cv.h! */
5645 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5647 CvLVALUE_on(PL_compcv);
5649 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5651 deprecate(":locked");
5653 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5655 CvMETHOD_on(PL_compcv);
5657 /* After we've set the flags, it could be argued that
5658 we don't need to do the attributes.pm-based setting
5659 process, and shouldn't bother appending recognized
5660 flags. To experiment with that, uncomment the
5661 following "else". (Note that's already been
5662 uncommented. That keeps the above-applied built-in
5663 attributes from being intercepted (and possibly
5664 rejected) by a package's attribute routines, but is
5665 justified by the performance win for the common case
5666 of applying only built-in attributes.) */
5668 attrs = op_append_elem(OP_LIST, attrs,
5669 newSVOP(OP_CONST, 0,
5673 if (*s == ':' && s[1] != ':')
5676 break; /* require real whitespace or :'s */
5677 /* XXX losing whitespace on sequential attributes here */
5681 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5682 if (*s != ';' && *s != '}' && *s != tmp
5683 && (tmp != '=' || *s != ')')) {
5684 const char q = ((*s == '\'') ? '"' : '\'');
5685 /* If here for an expression, and parsed no attrs, back
5687 if (tmp == '=' && !attrs) {
5691 /* MUST advance bufptr here to avoid bogus "at end of line"
5692 context messages from yyerror().
5695 yyerror( (const char *)
5697 ? Perl_form(aTHX_ "Invalid separator character "
5698 "%c%c%c in attribute list", q, *s, q)
5699 : "Unterminated attribute list" ) );
5707 start_force(PL_curforce);
5708 NEXTVAL_NEXTTOKE.opval = attrs;
5709 CURMAD('_', PL_nextwhite);
5714 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5715 (s - SvPVX(PL_linestr)) - stuffstart);
5720 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5724 PL_lex_allbrackets--;
5728 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5729 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5733 PL_lex_allbrackets++;
5736 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5742 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5745 PL_lex_allbrackets--;
5751 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5754 if (PL_lex_brackets <= 0)
5755 yyerror("Unmatched right square bracket");
5758 PL_lex_allbrackets--;
5759 if (PL_lex_state == LEX_INTERPNORMAL) {
5760 if (PL_lex_brackets == 0) {
5761 if (*s == '-' && s[1] == '>')
5762 PL_lex_state = LEX_INTERPENDMAYBE;
5763 else if (*s != '[' && *s != '{')
5764 PL_lex_state = LEX_INTERPEND;
5771 if (PL_lex_brackets > 100) {
5772 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5774 switch (PL_expect) {
5776 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5777 PL_lex_allbrackets++;
5778 OPERATOR(HASHBRACK);
5780 while (s < PL_bufend && SPACE_OR_TAB(*s))
5783 PL_tokenbuf[0] = '\0';
5784 if (d < PL_bufend && *d == '-') {
5785 PL_tokenbuf[0] = '-';
5787 while (d < PL_bufend && SPACE_OR_TAB(*d))
5790 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5791 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5793 while (d < PL_bufend && SPACE_OR_TAB(*d))
5796 const char minus = (PL_tokenbuf[0] == '-');
5797 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5805 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5806 PL_lex_allbrackets++;
5811 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5812 PL_lex_allbrackets++;
5817 if (PL_oldoldbufptr == PL_last_lop)
5818 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5820 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5821 PL_lex_allbrackets++;
5824 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5826 /* This hack is to get the ${} in the message. */
5828 yyerror("syntax error");
5831 OPERATOR(HASHBRACK);
5833 /* This hack serves to disambiguate a pair of curlies
5834 * as being a block or an anon hash. Normally, expectation
5835 * determines that, but in cases where we're not in a
5836 * position to expect anything in particular (like inside
5837 * eval"") we have to resolve the ambiguity. This code
5838 * covers the case where the first term in the curlies is a
5839 * quoted string. Most other cases need to be explicitly
5840 * disambiguated by prepending a "+" before the opening
5841 * curly in order to force resolution as an anon hash.
5843 * XXX should probably propagate the outer expectation
5844 * into eval"" to rely less on this hack, but that could
5845 * potentially break current behavior of eval"".
5849 if (*s == '\'' || *s == '"' || *s == '`') {
5850 /* common case: get past first string, handling escapes */
5851 for (t++; t < PL_bufend && *t != *s;)
5852 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5856 else if (*s == 'q') {
5859 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5862 /* skip q//-like construct */
5864 char open, close, term;
5867 while (t < PL_bufend && isSPACE(*t))
5869 /* check for q => */
5870 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5871 OPERATOR(HASHBRACK);
5875 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5879 for (t++; t < PL_bufend; t++) {
5880 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5882 else if (*t == open)
5886 for (t++; t < PL_bufend; t++) {
5887 if (*t == '\\' && t+1 < PL_bufend)
5889 else if (*t == close && --brackets <= 0)
5891 else if (*t == open)
5898 /* skip plain q word */
5899 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5902 else if (isALNUM_lazy_if(t,UTF)) {
5904 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5907 while (t < PL_bufend && isSPACE(*t))
5909 /* if comma follows first term, call it an anon hash */
5910 /* XXX it could be a comma expression with loop modifiers */
5911 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5912 || (*t == '=' && t[1] == '>')))
5913 OPERATOR(HASHBRACK);
5914 if (PL_expect == XREF)
5917 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5923 pl_yylval.ival = CopLINE(PL_curcop);
5924 if (isSPACE(*s) || *s == '#')
5925 PL_copline = NOLINE; /* invalidate current command line number */
5926 TOKEN(formbrack ? '=' : '{');
5928 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5932 if (PL_lex_brackets <= 0)
5933 yyerror("Unmatched right curly bracket");
5935 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5936 PL_lex_allbrackets--;
5937 if (PL_lex_state == LEX_INTERPNORMAL) {
5938 if (PL_lex_brackets == 0) {
5939 if (PL_expect & XFAKEBRACK) {
5940 PL_expect &= XENUMMASK;
5941 PL_lex_state = LEX_INTERPEND;
5946 PL_thiswhite = newSVpvs("");
5947 sv_catpvs(PL_thiswhite,"}");
5950 return yylex(); /* ignore fake brackets */
5952 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5953 && SvEVALED(PL_lex_repl))
5954 PL_lex_state = LEX_INTERPEND;
5955 else if (*s == '-' && s[1] == '>')
5956 PL_lex_state = LEX_INTERPENDMAYBE;
5957 else if (*s != '[' && *s != '{')
5958 PL_lex_state = LEX_INTERPEND;
5961 if (PL_expect & XFAKEBRACK) {
5962 PL_expect &= XENUMMASK;
5964 return yylex(); /* ignore fake brackets */
5966 start_force(PL_curforce);
5968 curmad('X', newSVpvn(s-1,1));
5969 CURMAD('_', PL_thiswhite);
5971 force_next(formbrack ? '.' : '}');
5972 if (formbrack) LEAVE;
5975 PL_thistoken = newSVpvs("");
5977 if (formbrack == 2) { /* means . where arguments were expected */
5978 start_force(PL_curforce);
5986 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5987 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5994 if (PL_expect == XOPERATOR) {
5995 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5996 && isIDFIRST_lazy_if(s,UTF))
5998 CopLINE_dec(PL_curcop);
5999 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6000 CopLINE_inc(PL_curcop);
6002 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6003 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6010 PL_tokenbuf[0] = '&';
6011 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf + 1,
6012 sizeof PL_tokenbuf - 1, TRUE);
6013 if (PL_tokenbuf[1]) {
6014 PL_expect = XOPERATOR;
6015 force_ident_maybe_lex('&');
6019 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
6025 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6026 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6033 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6034 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6042 const char tmp = *s++;
6044 if (!PL_lex_allbrackets &&
6045 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6052 if (!PL_lex_allbrackets &&
6053 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
6061 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6062 && strchr("+-*/%.^&|<",tmp))
6063 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6064 "Reversed %c= operator",(int)tmp);
6066 if (PL_expect == XSTATE && isALPHA(tmp) &&
6067 (s == PL_linestart+1 || s[-2] == '\n') )
6069 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6070 || PL_lex_state != LEX_NORMAL) {
6075 if (strnEQ(s,"=cut",4)) {
6091 PL_thiswhite = newSVpvs("");
6092 sv_catpvn(PL_thiswhite, PL_linestart,
6093 PL_bufend - PL_linestart);
6097 PL_parser->in_pod = 1;
6101 if (PL_expect == XBLOCK) {
6103 #ifdef PERL_STRICT_CR
6104 while (SPACE_OR_TAB(*t))
6106 while (SPACE_OR_TAB(*t) || *t == '\r')
6109 if (*t == '\n' || *t == '#') {
6112 SAVEI8(PL_parser->form_lex_state);
6113 SAVEI32(PL_lex_formbrack);
6114 PL_parser->form_lex_state = PL_lex_state;
6115 PL_lex_formbrack = PL_lex_brackets + 1;
6119 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6128 const char tmp = *s++;
6130 /* was this !=~ where !~ was meant?
6131 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6133 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6134 const char *t = s+1;
6136 while (t < PL_bufend && isSPACE(*t))
6139 if (*t == '/' || *t == '?' ||
6140 ((*t == 'm' || *t == 's' || *t == 'y')
6141 && !isALNUM(t[1])) ||
6142 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
6143 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6144 "!=~ should be !~");
6146 if (!PL_lex_allbrackets &&
6147 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6159 if (PL_expect != XOPERATOR) {
6160 if (s[1] != '<' && !strchr(s,'>'))
6163 s = scan_heredoc(s);
6165 s = scan_inputsymbol(s);
6166 PL_expect = XOPERATOR;
6167 TOKEN(sublex_start());
6173 if (*s == '=' && !PL_lex_allbrackets &&
6174 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6178 SHop(OP_LEFT_SHIFT);
6183 if (!PL_lex_allbrackets &&
6184 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6191 if (!PL_lex_allbrackets &&
6192 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6200 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6208 const char tmp = *s++;
6210 if (*s == '=' && !PL_lex_allbrackets &&
6211 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6215 SHop(OP_RIGHT_SHIFT);
6217 else if (tmp == '=') {
6218 if (!PL_lex_allbrackets &&
6219 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6227 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6236 if (PL_expect == XOPERATOR) {
6237 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6238 return deprecate_commaless_var_list();
6242 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6243 PL_tokenbuf[0] = '@';
6244 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6245 sizeof PL_tokenbuf - 1, FALSE);
6246 if (PL_expect == XOPERATOR)
6247 no_op("Array length", s);
6248 if (!PL_tokenbuf[1])
6250 PL_expect = XOPERATOR;
6251 force_ident_maybe_lex('#');
6255 PL_tokenbuf[0] = '$';
6256 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6257 sizeof PL_tokenbuf - 1, FALSE);
6258 if (PL_expect == XOPERATOR)
6260 if (!PL_tokenbuf[1]) {
6262 yyerror("Final $ should be \\$ or $name");
6268 const char tmp = *s;
6269 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6272 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6273 && intuit_more(s)) {
6275 PL_tokenbuf[0] = '@';
6276 if (ckWARN(WARN_SYNTAX)) {
6279 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6282 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6283 while (t < PL_bufend && *t != ']')
6285 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6286 "Multidimensional syntax %.*s not supported",
6287 (int)((t - PL_bufptr) + 1), PL_bufptr);
6291 else if (*s == '{') {
6293 PL_tokenbuf[0] = '%';
6294 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6295 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6297 char tmpbuf[sizeof PL_tokenbuf];
6300 } while (isSPACE(*t));
6301 if (isIDFIRST_lazy_if(t,UTF)) {
6303 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6308 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6309 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6310 "You need to quote \"%"SVf"\"",
6311 SVfARG(newSVpvn_flags(tmpbuf, len,
6312 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
6318 PL_expect = XOPERATOR;
6319 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6320 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6321 if (!islop || PL_last_lop_op == OP_GREPSTART)
6322 PL_expect = XOPERATOR;
6323 else if (strchr("$@\"'`q", *s))
6324 PL_expect = XTERM; /* e.g. print $fh "foo" */
6325 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6326 PL_expect = XTERM; /* e.g. print $fh &sub */
6327 else if (isIDFIRST_lazy_if(s,UTF)) {
6328 char tmpbuf[sizeof PL_tokenbuf];
6330 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6331 if ((t2 = keyword(tmpbuf, len, 0))) {
6332 /* binary operators exclude handle interpretations */
6344 PL_expect = XTERM; /* e.g. print $fh length() */
6349 PL_expect = XTERM; /* e.g. print $fh subr() */
6352 else if (isDIGIT(*s))
6353 PL_expect = XTERM; /* e.g. print $fh 3 */
6354 else if (*s == '.' && isDIGIT(s[1]))
6355 PL_expect = XTERM; /* e.g. print $fh .3 */
6356 else if ((*s == '?' || *s == '-' || *s == '+')
6357 && !isSPACE(s[1]) && s[1] != '=')
6358 PL_expect = XTERM; /* e.g. print $fh -1 */
6359 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6361 PL_expect = XTERM; /* e.g. print $fh /.../
6362 XXX except DORDOR operator
6364 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6366 PL_expect = XTERM; /* print $fh <<"EOF" */
6369 force_ident_maybe_lex('$');
6373 if (PL_expect == XOPERATOR)
6375 PL_tokenbuf[0] = '@';
6376 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6377 if (!PL_tokenbuf[1]) {
6380 if (PL_lex_state == LEX_NORMAL)
6382 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6384 PL_tokenbuf[0] = '%';
6386 /* Warn about @ where they meant $. */
6387 if (*s == '[' || *s == '{') {
6388 if (ckWARN(WARN_SYNTAX)) {
6389 const char *t = s + 1;
6390 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6391 t += UTF ? UTF8SKIP(t) : 1;
6392 if (*t == '}' || *t == ']') {
6394 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6395 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6396 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6397 "Scalar value %"SVf" better written as $%"SVf,
6398 SVfARG(newSVpvn_flags(PL_bufptr, (STRLEN)(t-PL_bufptr),
6399 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))),
6400 SVfARG(newSVpvn_flags(PL_bufptr+1, (STRLEN)(t-PL_bufptr-1),
6401 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))));
6406 PL_expect = XOPERATOR;
6407 force_ident_maybe_lex('@');
6410 case '/': /* may be division, defined-or, or pattern */
6411 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6412 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6413 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6418 case '?': /* may either be conditional or pattern */
6419 if (PL_expect == XOPERATOR) {
6422 if (!PL_lex_allbrackets &&
6423 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6427 PL_lex_allbrackets++;
6433 /* A // operator. */
6434 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6435 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6436 LEX_FAKEEOF_LOGIC)) {
6444 if (*s == '=' && !PL_lex_allbrackets &&
6445 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6454 /* Disable warning on "study /blah/" */
6455 if (PL_oldoldbufptr == PL_last_uni
6456 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6457 || memNE(PL_last_uni, "study", 5)
6458 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6462 deprecate("?PATTERN? without explicit operator");
6463 s = scan_pat(s,OP_MATCH);
6464 TERM(sublex_start());
6468 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6469 #ifdef PERL_STRICT_CR
6472 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6474 && (s == PL_linestart || s[-1] == '\n') )
6477 formbrack = 2; /* dot seen where arguments expected */
6480 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6484 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6487 if (!PL_lex_allbrackets &&
6488 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6495 pl_yylval.ival = OPf_SPECIAL;
6501 if (*s == '=' && !PL_lex_allbrackets &&
6502 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6509 case '0': case '1': case '2': case '3': case '4':
6510 case '5': case '6': case '7': case '8': case '9':
6511 s = scan_num(s, &pl_yylval);
6512 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6513 if (PL_expect == XOPERATOR)
6518 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6519 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6520 if (PL_expect == XOPERATOR) {
6521 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6522 return deprecate_commaless_var_list();
6529 pl_yylval.ival = OP_CONST;
6530 TERM(sublex_start());
6533 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6534 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6535 if (PL_expect == XOPERATOR) {
6536 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6537 return deprecate_commaless_var_list();
6544 pl_yylval.ival = OP_CONST;
6545 /* FIXME. I think that this can be const if char *d is replaced by
6546 more localised variables. */
6547 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6548 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6549 pl_yylval.ival = OP_STRINGIFY;
6553 TERM(sublex_start());
6556 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6557 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6558 if (PL_expect == XOPERATOR)
6559 no_op("Backticks",s);
6562 readpipe_override();
6563 TERM(sublex_start());
6567 if (PL_lex_inwhat && isDIGIT(*s))
6568 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6570 if (PL_expect == XOPERATOR)
6571 no_op("Backslash",s);
6575 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6576 char *start = s + 2;
6577 while (isDIGIT(*start) || *start == '_')
6579 if (*start == '.' && isDIGIT(start[1])) {
6580 s = scan_num(s, &pl_yylval);
6583 else if ((*start == ':' && start[1] == ':')
6584 || (PL_expect == XSTATE && *start == ':'))
6586 else if (PL_expect == XSTATE) {
6588 while (d < PL_bufend && isSPACE(*d)) d++;
6589 if (*d == ':') goto keylookup;
6591 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6592 if (!isALPHA(*start) && (PL_expect == XTERM
6593 || PL_expect == XREF || PL_expect == XSTATE
6594 || PL_expect == XTERMORDORDOR)) {
6595 GV *const gv = gv_fetchpvn_flags(s, start - s,
6596 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6598 s = scan_num(s, &pl_yylval);
6605 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6658 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6660 /* Some keywords can be followed by any delimiter, including ':' */
6661 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6663 /* x::* is just a word, unless x is "CORE" */
6664 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6668 while (d < PL_bufend && isSPACE(*d))
6669 d++; /* no comments skipped here, or s### is misparsed */
6671 /* Is this a word before a => operator? */
6672 if (*d == '=' && d[1] == '>') {
6675 = (OP*)newSVOP(OP_CONST, 0,
6676 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6677 pl_yylval.opval->op_private = OPpCONST_BARE;
6681 /* Check for plugged-in keyword */
6685 char *saved_bufptr = PL_bufptr;
6687 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6689 if (result == KEYWORD_PLUGIN_DECLINE) {
6690 /* not a plugged-in keyword */
6691 PL_bufptr = saved_bufptr;
6692 } else if (result == KEYWORD_PLUGIN_STMT) {
6693 pl_yylval.opval = o;
6696 return REPORT(PLUGSTMT);
6697 } else if (result == KEYWORD_PLUGIN_EXPR) {
6698 pl_yylval.opval = o;
6700 PL_expect = XOPERATOR;
6701 return REPORT(PLUGEXPR);
6703 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6708 /* Check for built-in keyword */
6709 tmp = keyword(PL_tokenbuf, len, 0);
6711 /* Is this a label? */
6712 if (!anydelim && PL_expect == XSTATE
6713 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6715 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6716 newSVpvn_flags(PL_tokenbuf,
6717 len, UTF ? SVf_UTF8 : 0));
6722 /* Check for lexical sub */
6723 if (PL_expect != XOPERATOR) {
6724 char tmpbuf[sizeof PL_tokenbuf + 1];
6726 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6727 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6728 if (off != NOT_IN_PAD) {
6729 assert(off); /* we assume this is boolean-true below */
6730 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6731 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6732 HEK * const stashname = HvNAME_HEK(stash);
6733 sv = newSVhek(stashname);
6734 sv_catpvs(sv, "::");
6735 sv_catpvn_flags(sv, PL_tokenbuf, len,
6736 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6737 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6742 rv2cv_op = newOP(OP_PADANY, 0);
6743 rv2cv_op->op_targ = off;
6744 rv2cv_op = (OP*)newCVREF(0, rv2cv_op);
6745 cv = (CV *)PAD_SV(off);
6753 if (tmp < 0) { /* second-class keyword? */
6754 GV *ogv = NULL; /* override (winner) */
6755 GV *hgv = NULL; /* hidden (loser) */
6756 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6758 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6759 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6762 if (GvIMPORTED_CV(gv))
6764 else if (! CvMETHOD(cv))
6768 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6769 UTF ? -(I32)len : (I32)len, FALSE)) &&
6770 (gv = *gvp) && isGV_with_GP(gv) &&
6771 GvCVu(gv) && GvIMPORTED_CV(gv))
6778 tmp = 0; /* overridden by import or by GLOBAL */
6781 && -tmp==KEY_lock /* XXX generalizable kludge */
6784 tmp = 0; /* any sub overrides "weak" keyword */
6786 else { /* no override */
6788 if (tmp == KEY_dump) {
6789 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6790 "dump() better written as CORE::dump()");
6794 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6795 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6796 "Ambiguous call resolved as CORE::%s(), "
6797 "qualify as such or use &",
6805 default: /* not a keyword */
6806 /* Trade off - by using this evil construction we can pull the
6807 variable gv into the block labelled keylookup. If not, then
6808 we have to give it function scope so that the goto from the
6809 earlier ':' case doesn't bypass the initialisation. */
6811 just_a_word_zero_gv:
6823 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6824 const char penultchar =
6825 lastchar && PL_bufptr - 2 >= PL_linestart
6829 SV *nextPL_nextwhite = 0;
6833 /* Get the rest if it looks like a package qualifier */
6835 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6837 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6840 Perl_croak(aTHX_ "Bad name after %"SVf"%s",
6841 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6842 (UTF ? SVf_UTF8 : 0) | SVs_TEMP )),
6843 *s == '\'' ? "'" : "::");
6848 if (PL_expect == XOPERATOR) {
6849 if (PL_bufptr == PL_linestart) {
6850 CopLINE_dec(PL_curcop);
6851 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6852 CopLINE_inc(PL_curcop);
6855 no_op("Bareword",s);
6858 /* Look for a subroutine with this name in current package,
6859 unless this is a lexical sub, or name is "Foo::",
6860 in which case Foo is a bareword
6861 (and a package name). */
6863 if (len > 2 && !PL_madskills &&
6864 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6866 if (ckWARN(WARN_BAREWORD)
6867 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6868 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6869 "Bareword \"%"SVf"\" refers to nonexistent package",
6870 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6871 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
6873 PL_tokenbuf[len] = '\0';
6879 /* Mustn't actually add anything to a symbol table.
6880 But also don't want to "initialise" any placeholder
6881 constants that might already be there into full
6882 blown PVGVs with attached PVCV. */
6883 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6884 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6890 /* if we saw a global override before, get the right name */
6893 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6894 len ? len : strlen(PL_tokenbuf));
6896 SV * const tmp_sv = sv;
6897 sv = newSVpvs("CORE::GLOBAL::");
6898 sv_catsv(sv, tmp_sv);
6899 SvREFCNT_dec(tmp_sv);
6903 if (PL_madskills && !PL_thistoken) {
6904 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6905 PL_thistoken = newSVpvn(start,s - start);
6906 PL_realtokenstart = s - SvPVX(PL_linestr);
6910 /* Presume this is going to be a bareword of some sort. */
6912 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6913 pl_yylval.opval->op_private = OPpCONST_BARE;
6915 /* And if "Foo::", then that's what it certainly is. */
6921 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6922 const_op->op_private = OPpCONST_BARE;
6923 rv2cv_op = newCVREF(0, const_op);
6924 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
6927 /* See if it's the indirect object for a list operator. */
6929 if (PL_oldoldbufptr &&
6930 PL_oldoldbufptr < PL_bufptr &&
6931 (PL_oldoldbufptr == PL_last_lop
6932 || PL_oldoldbufptr == PL_last_uni) &&
6933 /* NO SKIPSPACE BEFORE HERE! */
6934 (PL_expect == XREF ||
6935 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6937 bool immediate_paren = *s == '(';
6939 /* (Now we can afford to cross potential line boundary.) */
6940 s = SKIPSPACE2(s,nextPL_nextwhite);
6942 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6945 /* Two barewords in a row may indicate method call. */
6947 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6948 (tmp = intuit_method(s, gv, cv))) {
6950 if (tmp == METHOD && !PL_lex_allbrackets &&
6951 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6952 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6956 /* If not a declared subroutine, it's an indirect object. */
6957 /* (But it's an indir obj regardless for sort.) */
6958 /* Also, if "_" follows a filetest operator, it's a bareword */
6961 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6963 (PL_last_lop_op != OP_MAPSTART &&
6964 PL_last_lop_op != OP_GREPSTART))))
6965 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6966 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6969 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6974 PL_expect = XOPERATOR;
6977 s = SKIPSPACE2(s,nextPL_nextwhite);
6978 PL_nextwhite = nextPL_nextwhite;
6983 /* Is this a word before a => operator? */
6984 if (*s == '=' && s[1] == '>' && !pkgname) {
6987 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6988 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6989 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6993 /* If followed by a paren, it's certainly a subroutine. */
6998 while (SPACE_OR_TAB(*d))
7000 if (*d == ')' && (sv = cv_const_sv(cv))) {
7007 PL_nextwhite = PL_thiswhite;
7010 start_force(PL_curforce);
7012 NEXTVAL_NEXTTOKE.opval =
7013 off ? rv2cv_op : pl_yylval.opval;
7014 PL_expect = XOPERATOR;
7017 PL_nextwhite = nextPL_nextwhite;
7018 curmad('X', PL_thistoken);
7019 PL_thistoken = newSVpvs("");
7023 op_free(pl_yylval.opval), force_next(PRIVATEREF);
7024 else op_free(rv2cv_op), force_next(WORD);
7029 /* If followed by var or block, call it a method (unless sub) */
7031 if ((*s == '$' || *s == '{') && !cv) {
7033 PL_last_lop = PL_oldbufptr;
7034 PL_last_lop_op = OP_METHOD;
7035 if (!PL_lex_allbrackets &&
7036 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7037 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7041 /* If followed by a bareword, see if it looks like indir obj. */
7044 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7045 && (tmp = intuit_method(s, gv, cv))) {
7047 if (tmp == METHOD && !PL_lex_allbrackets &&
7048 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7049 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7053 /* Not a method, so call it a subroutine (if defined) */
7056 if (lastchar == '-' && penultchar != '-') {
7057 const SV *tmpsv = newSVpvn_flags( PL_tokenbuf, len ? len : strlen(PL_tokenbuf), (UTF ? SVf_UTF8 : 0) | SVs_TEMP );
7058 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7059 "Ambiguous use of -%"SVf" resolved as -&%"SVf"()",
7060 SVfARG(tmpsv), SVfARG(tmpsv));
7062 /* Check for a constant sub */
7063 if ((sv = cv_const_sv(cv))) {
7066 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7067 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7068 pl_yylval.opval->op_private = OPpCONST_FOLDED;
7069 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7073 op_free(pl_yylval.opval);
7074 pl_yylval.opval = rv2cv_op;
7075 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7076 PL_last_lop = PL_oldbufptr;
7077 PL_last_lop_op = OP_ENTERSUB;
7078 /* Is there a prototype? */
7085 STRLEN protolen = CvPROTOLEN(cv);
7086 const char *proto = CvPROTO(cv);
7090 if ((optional = *proto == ';'))
7093 while (*proto == ';');
7097 *proto == '$' || *proto == '_'
7098 || *proto == '*' || *proto == '+'
7103 *proto == '\\' && proto[1] && proto[2] == '\0'
7106 UNIPROTO(UNIOPSUB,optional);
7107 if (*proto == '\\' && proto[1] == '[') {
7108 const char *p = proto + 2;
7109 while(*p && *p != ']')
7111 if(*p == ']' && !p[1])
7112 UNIPROTO(UNIOPSUB,optional);
7114 if (*proto == '&' && *s == '{') {
7116 sv_setpvs(PL_subname, "__ANON__");
7118 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7119 if (!PL_lex_allbrackets &&
7120 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7121 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7128 PL_nextwhite = PL_thiswhite;
7131 start_force(PL_curforce);
7132 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7135 PL_nextwhite = nextPL_nextwhite;
7136 curmad('X', PL_thistoken);
7137 PL_thistoken = newSVpvs("");
7139 force_next(off ? PRIVATEREF : WORD);
7140 if (!PL_lex_allbrackets &&
7141 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7142 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7147 /* Guess harder when madskills require "best effort". */
7148 if (PL_madskills && (!gv || !GvCVu(gv))) {
7149 int probable_sub = 0;
7150 if (strchr("\"'`$@%0123456789!*+{[<", *s))
7152 else if (isALPHA(*s)) {
7156 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
7157 if (!keyword(tmpbuf, tmplen, 0))
7160 while (d < PL_bufend && isSPACE(*d))
7162 if (*d == '=' && d[1] == '>')
7167 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
7169 op_free(pl_yylval.opval);
7170 pl_yylval.opval = rv2cv_op;
7171 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7172 PL_last_lop = PL_oldbufptr;
7173 PL_last_lop_op = OP_ENTERSUB;
7174 PL_nextwhite = PL_thiswhite;
7176 start_force(PL_curforce);
7177 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7179 PL_nextwhite = nextPL_nextwhite;
7180 curmad('X', PL_thistoken);
7181 PL_thistoken = newSVpvs("");
7182 force_next(off ? PRIVATEREF : WORD);
7183 if (!PL_lex_allbrackets &&
7184 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7185 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7189 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7191 force_next(off ? PRIVATEREF : WORD);
7192 if (!PL_lex_allbrackets &&
7193 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7194 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7199 /* Call it a bare word */
7201 if (PL_hints & HINT_STRICT_SUBS)
7202 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7205 /* after "print" and similar functions (corresponding to
7206 * "F? L" in opcode.pl), whatever wasn't already parsed as
7207 * a filehandle should be subject to "strict subs".
7208 * Likewise for the optional indirect-object argument to system
7209 * or exec, which can't be a bareword */
7210 if ((PL_last_lop_op == OP_PRINT
7211 || PL_last_lop_op == OP_PRTF
7212 || PL_last_lop_op == OP_SAY
7213 || PL_last_lop_op == OP_SYSTEM
7214 || PL_last_lop_op == OP_EXEC)
7215 && (PL_hints & HINT_STRICT_SUBS))
7216 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7217 if (lastchar != '-') {
7218 if (ckWARN(WARN_RESERVED)) {
7222 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7223 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7231 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7232 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7233 "Operator or semicolon missing before %c%"SVf,
7234 lastchar, SVfARG(newSVpvn_flags(PL_tokenbuf,
7235 strlen(PL_tokenbuf),
7236 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
7237 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7238 "Ambiguous use of %c resolved as operator %c",
7239 lastchar, lastchar);
7246 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7251 (OP*)newSVOP(OP_CONST, 0,
7252 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7255 case KEY___PACKAGE__:
7257 (OP*)newSVOP(OP_CONST, 0,
7259 ? newSVhek(HvNAME_HEK(PL_curstash))
7266 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7267 const char *pname = "main";
7270 if (PL_tokenbuf[2] == 'D')
7273 PL_curstash ? PL_curstash : PL_defstash;
7274 pname = HvNAME_get(stash);
7275 plen = HvNAMELEN (stash);
7276 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7278 gv = gv_fetchpvn_flags(
7279 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7280 plen+6, GV_ADD|putf8, SVt_PVIO
7284 GvIOp(gv) = newIO();
7285 IoIFP(GvIOp(gv)) = PL_rsfp;
7286 #if defined(HAS_FCNTL) && defined(F_SETFD)
7288 const int fd = PerlIO_fileno(PL_rsfp);
7289 fcntl(fd,F_SETFD,fd >= 3);
7292 /* Mark this internal pseudo-handle as clean */
7293 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7294 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7295 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7297 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7298 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7299 /* if the script was opened in binmode, we need to revert
7300 * it to text mode for compatibility; but only iff it has CRs
7301 * XXX this is a questionable hack at best. */
7302 if (PL_bufend-PL_bufptr > 2
7303 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7306 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7307 loc = PerlIO_tell(PL_rsfp);
7308 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7311 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7313 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7314 #endif /* NETWARE */
7316 PerlIO_seek(PL_rsfp, loc, 0);
7320 #ifdef PERLIO_LAYERS
7323 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7324 else if (PL_encoding) {
7331 XPUSHs(PL_encoding);
7333 call_method("name", G_SCALAR);
7337 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7338 Perl_form(aTHX_ ":encoding(%"SVf")",
7347 if (PL_realtokenstart >= 0) {
7348 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7350 PL_endwhite = newSVpvs("");
7351 sv_catsv(PL_endwhite, PL_thiswhite);
7353 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7354 PL_realtokenstart = -1;
7356 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7366 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7375 if (PL_expect == XSTATE) {
7382 if (*s == ':' && s[1] == ':') {
7386 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7387 if ((*s == ':' && s[1] == ':')
7388 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7392 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7396 Perl_croak(aTHX_ "CORE::%"SVf" is not a keyword",
7397 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7398 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7401 else if (tmp == KEY_require || tmp == KEY_do
7403 /* that's a way to remember we saw "CORE::" */
7416 LOP(OP_ACCEPT,XTERM);
7419 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7424 LOP(OP_ATAN2,XTERM);
7430 LOP(OP_BINMODE,XTERM);
7433 LOP(OP_BLESS,XTERM);
7442 /* We have to disambiguate the two senses of
7443 "continue". If the next token is a '{' then
7444 treat it as the start of a continue block;
7445 otherwise treat it as a control operator.
7455 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7465 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7474 if (!PL_cryptseen) {
7475 PL_cryptseen = TRUE;
7479 LOP(OP_CRYPT,XTERM);
7482 LOP(OP_CHMOD,XTERM);
7485 LOP(OP_CHOWN,XTERM);
7488 LOP(OP_CONNECT,XTERM);
7508 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7510 if (len && !keyword(PL_tokenbuf + 1, len, 0)) {
7513 force_ident_maybe_lex('&');
7518 if (orig_keyword == KEY_do) {
7527 PL_hints |= HINT_BLOCK_SCOPE;
7537 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7538 STR_WITH_LEN("NDBM_File::"),
7539 STR_WITH_LEN("DB_File::"),
7540 STR_WITH_LEN("GDBM_File::"),
7541 STR_WITH_LEN("SDBM_File::"),
7542 STR_WITH_LEN("ODBM_File::"),
7544 LOP(OP_DBMOPEN,XTERM);
7550 PL_expect = XOPERATOR;
7551 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7558 pl_yylval.ival = CopLINE(PL_curcop);
7562 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7576 if (*s == '{') { /* block eval */
7577 PL_expect = XTERMBLOCK;
7578 UNIBRACK(OP_ENTERTRY);
7580 else { /* string eval */
7582 UNIBRACK(OP_ENTEREVAL);
7587 UNIBRACK(-OP_ENTEREVAL);
7601 case KEY_endhostent:
7607 case KEY_endservent:
7610 case KEY_endprotoent:
7621 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7623 pl_yylval.ival = CopLINE(PL_curcop);
7625 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7628 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7631 if ((PL_bufend - p) >= 3 &&
7632 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7634 else if ((PL_bufend - p) >= 4 &&
7635 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7638 if (isIDFIRST_lazy_if(p,UTF)) {
7639 p = scan_ident(p, PL_bufend,
7640 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7644 Perl_croak(aTHX_ "Missing $ on loop variable");
7646 s = SvPVX(PL_linestr) + soff;
7652 LOP(OP_FORMLINE,XTERM);
7661 LOP(OP_FCNTL,XTERM);
7667 LOP(OP_FLOCK,XTERM);
7670 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7675 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7680 LOP(OP_GREPSTART, XREF);
7683 PL_expect = XOPERATOR;
7684 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7699 case KEY_getpriority:
7700 LOP(OP_GETPRIORITY,XTERM);
7702 case KEY_getprotobyname:
7705 case KEY_getprotobynumber:
7706 LOP(OP_GPBYNUMBER,XTERM);
7708 case KEY_getprotoent:
7720 case KEY_getpeername:
7721 UNI(OP_GETPEERNAME);
7723 case KEY_gethostbyname:
7726 case KEY_gethostbyaddr:
7727 LOP(OP_GHBYADDR,XTERM);
7729 case KEY_gethostent:
7732 case KEY_getnetbyname:
7735 case KEY_getnetbyaddr:
7736 LOP(OP_GNBYADDR,XTERM);
7741 case KEY_getservbyname:
7742 LOP(OP_GSBYNAME,XTERM);
7744 case KEY_getservbyport:
7745 LOP(OP_GSBYPORT,XTERM);
7747 case KEY_getservent:
7750 case KEY_getsockname:
7751 UNI(OP_GETSOCKNAME);
7753 case KEY_getsockopt:
7754 LOP(OP_GSOCKOPT,XTERM);
7769 pl_yylval.ival = CopLINE(PL_curcop);
7774 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7782 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7784 pl_yylval.ival = CopLINE(PL_curcop);
7788 LOP(OP_INDEX,XTERM);
7794 LOP(OP_IOCTL,XTERM);
7806 PL_expect = XOPERATOR;
7807 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7824 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7829 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7843 LOP(OP_LISTEN,XTERM);
7852 s = scan_pat(s,OP_MATCH);
7853 TERM(sublex_start());
7856 LOP(OP_MAPSTART, XREF);
7859 LOP(OP_MKDIR,XTERM);
7862 LOP(OP_MSGCTL,XTERM);
7865 LOP(OP_MSGGET,XTERM);
7868 LOP(OP_MSGRCV,XTERM);
7871 LOP(OP_MSGSND,XTERM);
7876 PL_in_my = (U16)tmp;
7878 if (isIDFIRST_lazy_if(s,UTF)) {
7882 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7883 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7885 if (!FEATURE_LEXSUBS_IS_ENABLED)
7887 "Experimental \"%s\" subs not enabled",
7888 tmp == KEY_my ? "my" :
7889 tmp == KEY_state ? "state" : "our");
7892 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7893 if (!PL_in_my_stash) {
7896 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7897 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7900 if (PL_madskills) { /* just add type to declarator token */
7901 sv_catsv(PL_thistoken, PL_nextwhite);
7903 sv_catpvn(PL_thistoken, start, s - start);
7911 PL_expect = XOPERATOR;
7912 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7916 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7921 s = tokenize_use(0, s);
7925 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7928 if (!PL_lex_allbrackets &&
7929 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7930 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7936 if (isIDFIRST_lazy_if(s,UTF)) {
7938 for (d = s; isALNUM_lazy_if(d,UTF);) {
7939 d += UTF ? UTF8SKIP(d) : 1;
7941 while (UTF8_IS_CONTINUED(*d) && is_utf8_mark((U8*)d)) {
7942 d += UTF ? UTF8SKIP(d) : 1;
7946 for (t=d; isSPACE(*t);)
7948 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7950 && !(t[0] == '=' && t[1] == '>')
7951 && !(t[0] == ':' && t[1] == ':')
7952 && !keyword(s, d-s, 0)
7954 SV *tmpsv = newSVpvn_flags(s, (STRLEN)(d-s),
7955 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
7956 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7957 "Precedence problem: open %"SVf" should be open(%"SVf")",
7958 SVfARG(tmpsv), SVfARG(tmpsv));
7964 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7966 pl_yylval.ival = OP_OR;
7976 LOP(OP_OPEN_DIR,XTERM);
7979 checkcomma(s,PL_tokenbuf,"filehandle");
7983 checkcomma(s,PL_tokenbuf,"filehandle");
8002 s = force_word(s,WORD,FALSE,TRUE,FALSE);
8004 s = force_strict_version(s);
8005 PL_lex_expect = XBLOCK;
8009 LOP(OP_PIPE_OP,XTERM);
8012 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8015 pl_yylval.ival = OP_CONST;
8016 TERM(sublex_start());
8023 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8026 PL_expect = XOPERATOR;
8027 if (SvCUR(PL_lex_stuff)) {
8028 int warned_comma = !ckWARN(WARN_QW);
8029 int warned_comment = warned_comma;
8030 d = SvPV_force(PL_lex_stuff, len);
8032 for (; isSPACE(*d) && len; --len, ++d)
8037 if (!warned_comma || !warned_comment) {
8038 for (; !isSPACE(*d) && len; --len, ++d) {
8039 if (!warned_comma && *d == ',') {
8040 Perl_warner(aTHX_ packWARN(WARN_QW),
8041 "Possible attempt to separate words with commas");
8044 else if (!warned_comment && *d == '#') {
8045 Perl_warner(aTHX_ packWARN(WARN_QW),
8046 "Possible attempt to put comments in qw() list");
8052 for (; !isSPACE(*d) && len; --len, ++d)
8055 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
8056 words = op_append_elem(OP_LIST, words,
8057 newSVOP(OP_CONST, 0, tokeq(sv)));
8062 words = newNULLLIST();
8064 SvREFCNT_dec(PL_lex_stuff);
8065 PL_lex_stuff = NULL;
8067 PL_expect = XOPERATOR;
8068 pl_yylval.opval = sawparens(words);
8073 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8076 pl_yylval.ival = OP_STRINGIFY;
8077 if (SvIVX(PL_lex_stuff) == '\'')
8078 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
8079 TERM(sublex_start());
8082 s = scan_pat(s,OP_QR);
8083 TERM(sublex_start());
8086 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8089 readpipe_override();
8090 TERM(sublex_start());
8097 PL_expect = XOPERATOR;
8099 s = force_version(s, FALSE);
8101 else if (*s != 'v' || !isDIGIT(s[1])
8102 || (s = force_version(s, TRUE), *s == 'v'))
8104 *PL_tokenbuf = '\0';
8105 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8106 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8107 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8108 GV_ADD | (UTF ? SVf_UTF8 : 0));
8110 yyerror("<> should be quotes");
8112 if (orig_keyword == KEY_require) {
8120 PL_last_uni = PL_oldbufptr;
8121 PL_last_lop_op = OP_REQUIRE;
8123 return REPORT( (int)REQUIRE );
8129 PL_expect = XOPERATOR;
8130 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8134 LOP(OP_RENAME,XTERM);
8143 LOP(OP_RINDEX,XTERM);
8152 UNIDOR(OP_READLINE);
8155 UNIDOR(OP_BACKTICK);
8164 LOP(OP_REVERSE,XTERM);
8167 UNIDOR(OP_READLINK);
8174 if (pl_yylval.opval)
8175 TERM(sublex_start());
8177 TOKEN(1); /* force error */
8180 checkcomma(s,PL_tokenbuf,"filehandle");
8190 LOP(OP_SELECT,XTERM);
8196 LOP(OP_SEMCTL,XTERM);
8199 LOP(OP_SEMGET,XTERM);
8202 LOP(OP_SEMOP,XTERM);
8208 LOP(OP_SETPGRP,XTERM);
8210 case KEY_setpriority:
8211 LOP(OP_SETPRIORITY,XTERM);
8213 case KEY_sethostent:
8219 case KEY_setservent:
8222 case KEY_setprotoent:
8232 LOP(OP_SEEKDIR,XTERM);
8234 case KEY_setsockopt:
8235 LOP(OP_SSOCKOPT,XTERM);
8241 LOP(OP_SHMCTL,XTERM);
8244 LOP(OP_SHMGET,XTERM);
8247 LOP(OP_SHMREAD,XTERM);
8250 LOP(OP_SHMWRITE,XTERM);
8253 LOP(OP_SHUTDOWN,XTERM);
8262 LOP(OP_SOCKET,XTERM);
8264 case KEY_socketpair:
8265 LOP(OP_SOCKPAIR,XTERM);
8268 checkcomma(s,PL_tokenbuf,"subroutine name");
8271 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8275 LOP(OP_SPLIT,XTERM);
8278 LOP(OP_SPRINTF,XTERM);
8281 LOP(OP_SPLICE,XTERM);
8296 LOP(OP_SUBSTR,XTERM);
8302 char * const tmpbuf = PL_tokenbuf + 1;
8303 SSize_t tboffset = 0;
8304 expectation attrful;
8305 bool have_name, have_proto;
8306 const int key = tmp;
8311 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8312 SV *subtoken = newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr));
8316 s = SKIPSPACE2(s,tmpwhite);
8322 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8323 (*s == ':' && s[1] == ':'))
8326 SV *nametoke = NULL;
8330 attrful = XATTRBLOCK;
8331 /* remember buffer pos'n for later force_word */
8332 tboffset = s - PL_oldbufptr;
8333 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8337 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8340 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8342 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
8344 sv_setpvn(PL_subname, tmpbuf, len);
8346 sv_setsv(PL_subname,PL_curstname);
8347 sv_catpvs(PL_subname,"::");
8348 sv_catpvn(PL_subname,tmpbuf,len);
8350 if (SvUTF8(PL_linestr))
8351 SvUTF8_on(PL_subname);
8357 CURMAD('X', nametoke);
8358 CURMAD('_', tmpwhite);
8359 force_ident_maybe_lex('&');
8361 s = SKIPSPACE2(d,tmpwhite);
8367 if (key == KEY_my || key == KEY_our || key==KEY_state)
8370 /* diag_listed_as: Missing name in "%s sub" */
8372 "Missing name in \"%s\"", PL_bufptr);
8374 PL_expect = XTERMBLOCK;
8375 attrful = XATTRTERM;
8376 sv_setpvs(PL_subname,"?");
8380 if (key == KEY_format) {
8382 PL_thistoken = subtoken;
8386 (void) force_word(PL_oldbufptr + tboffset, WORD,
8392 /* Look for a prototype */
8395 bool bad_proto = FALSE;
8396 bool in_brackets = FALSE;
8397 char greedy_proto = ' ';
8398 bool proto_after_greedy_proto = FALSE;
8399 bool must_be_last = FALSE;
8400 bool underscore = FALSE;
8401 bool seen_underscore = FALSE;
8402 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8405 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8407 Perl_croak(aTHX_ "Prototype not terminated");
8408 /* strip spaces and check for bad characters */
8409 d = SvPV(PL_lex_stuff, tmplen);
8411 for (p = d; tmplen; tmplen--, ++p) {
8415 if (warnillegalproto) {
8417 proto_after_greedy_proto = TRUE;
8418 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8423 if ( !strchr(";@%", *p) )
8430 else if ( *p == ']' ) {
8431 in_brackets = FALSE;
8433 else if ( (*p == '@' || *p == '%') &&
8434 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8436 must_be_last = TRUE;
8439 else if ( *p == '_' ) {
8440 underscore = seen_underscore = TRUE;
8447 if (proto_after_greedy_proto)
8448 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8449 "Prototype after '%c' for %"SVf" : %s",
8450 greedy_proto, SVfARG(PL_subname), d);
8452 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8453 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8454 "Illegal character %sin prototype for %"SVf" : %s",
8455 seen_underscore ? "after '_' " : "",
8457 SvUTF8(PL_lex_stuff)
8458 ? sv_uni_display(dsv,
8459 newSVpvn_flags(d, tmp, SVs_TEMP | SVf_UTF8),
8461 UNI_DISPLAY_ISPRINT)
8462 : pv_pretty(dsv, d, tmp, 60, NULL, NULL,
8463 PERL_PV_ESCAPE_NONASCII));
8465 SvCUR_set(PL_lex_stuff, tmp);
8470 CURMAD('q', PL_thisopen);
8471 CURMAD('_', tmpwhite);
8472 CURMAD('=', PL_thisstuff);
8473 CURMAD('Q', PL_thisclose);
8474 NEXTVAL_NEXTTOKE.opval =
8475 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8476 PL_lex_stuff = NULL;
8479 s = SKIPSPACE2(s,tmpwhite);
8487 if (*s == ':' && s[1] != ':')
8488 PL_expect = attrful;
8489 else if (*s != '{' && key == KEY_sub) {
8491 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8492 else if (*s != ';' && *s != '}')
8493 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8500 curmad('^', newSVpvs(""));
8501 CURMAD('_', tmpwhite);
8505 PL_thistoken = subtoken;
8508 NEXTVAL_NEXTTOKE.opval =
8509 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8510 PL_lex_stuff = NULL;
8516 sv_setpvs(PL_subname, "__ANON__");
8518 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8522 force_ident_maybe_lex('&');
8528 LOP(OP_SYSTEM,XREF);
8531 LOP(OP_SYMLINK,XTERM);
8534 LOP(OP_SYSCALL,XTERM);
8537 LOP(OP_SYSOPEN,XTERM);
8540 LOP(OP_SYSSEEK,XTERM);
8543 LOP(OP_SYSREAD,XTERM);
8546 LOP(OP_SYSWRITE,XTERM);
8551 TERM(sublex_start());
8572 LOP(OP_TRUNCATE,XTERM);
8584 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8586 pl_yylval.ival = CopLINE(PL_curcop);
8590 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8592 pl_yylval.ival = CopLINE(PL_curcop);
8596 LOP(OP_UNLINK,XTERM);
8602 LOP(OP_UNPACK,XTERM);
8605 LOP(OP_UTIME,XTERM);
8611 LOP(OP_UNSHIFT,XTERM);
8614 s = tokenize_use(1, s);
8624 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8626 pl_yylval.ival = CopLINE(PL_curcop);
8630 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8632 pl_yylval.ival = CopLINE(PL_curcop);
8636 PL_hints |= HINT_BLOCK_SCOPE;
8643 LOP(OP_WAITPID,XTERM);
8652 ctl_l[0] = toCTRL('L');
8654 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8657 /* Make sure $^L is defined */
8658 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8663 if (PL_expect == XOPERATOR) {
8664 if (*s == '=' && !PL_lex_allbrackets &&
8665 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8673 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8675 pl_yylval.ival = OP_XOR;
8681 #pragma segment Main
8687 Looks up an identifier in the pad or in a package
8690 PRIVATEREF if this is a lexical name.
8691 WORD if this belongs to a package.
8694 if we're in a my declaration
8695 croak if they tried to say my($foo::bar)
8696 build the ops for a my() declaration
8697 if it's an access to a my() variable
8698 build ops for access to a my() variable
8699 if in a dq string, and they've said @foo and we can't find @foo
8701 build ops for a bareword
8705 S_pending_ident(pTHX)
8709 const char pit = (char)pl_yylval.ival;
8710 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8711 /* All routes through this function want to know if there is a colon. */
8712 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8714 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8715 "### Pending identifier '%s'\n", PL_tokenbuf); });
8717 /* if we're in a my(), we can't allow dynamics here.
8718 $foo'bar has already been turned into $foo::bar, so
8719 just check for colons.
8721 if it's a legal name, the OP is a PADANY.
8724 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8726 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8727 "variable %s in \"our\"",
8728 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8729 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8733 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8734 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8735 UTF ? SVf_UTF8 : 0);
8737 pl_yylval.opval = newOP(OP_PADANY, 0);
8738 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8739 UTF ? SVf_UTF8 : 0);
8745 build the ops for accesses to a my() variable.
8750 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8751 UTF ? SVf_UTF8 : 0);
8752 if (tmp != NOT_IN_PAD) {
8753 /* might be an "our" variable" */
8754 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8755 /* build ops for a bareword */
8756 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8757 HEK * const stashname = HvNAME_HEK(stash);
8758 SV * const sym = newSVhek(stashname);
8759 sv_catpvs(sym, "::");
8760 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8761 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8762 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8766 ? (GV_ADDMULTI | GV_ADDINEVAL)
8769 ((PL_tokenbuf[0] == '$') ? SVt_PV
8770 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8775 pl_yylval.opval = newOP(OP_PADANY, 0);
8776 pl_yylval.opval->op_targ = tmp;
8782 Whine if they've said @foo in a doublequoted string,
8783 and @foo isn't a variable we can find in the symbol
8786 if (ckWARN(WARN_AMBIGUOUS) &&
8787 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8788 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8789 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8790 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8791 /* DO NOT warn for @- and @+ */
8792 && !( PL_tokenbuf[2] == '\0' &&
8793 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8796 /* Downgraded from fatal to warning 20000522 mjd */
8797 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8798 "Possible unintended interpolation of %"SVf" in string",
8799 SVfARG(newSVpvn_flags(PL_tokenbuf, tokenbuf_len,
8800 SVs_TEMP | ( UTF ? SVf_UTF8 : 0 ))));
8804 /* build ops for a bareword */
8805 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8806 newSVpvn_flags(PL_tokenbuf + 1,
8808 UTF ? SVf_UTF8 : 0 ));
8809 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8811 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8812 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8813 | ( UTF ? SVf_UTF8 : 0 ),
8814 ((PL_tokenbuf[0] == '$') ? SVt_PV
8815 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8821 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8825 PERL_ARGS_ASSERT_CHECKCOMMA;
8827 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8828 if (ckWARN(WARN_SYNTAX)) {
8831 for (w = s+2; *w && level; w++) {
8839 /* the list of chars below is for end of statements or
8840 * block / parens, boolean operators (&&, ||, //) and branch
8841 * constructs (or, and, if, until, unless, while, err, for).
8842 * Not a very solid hack... */
8843 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8844 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8845 "%s (...) interpreted as function",name);
8848 while (s < PL_bufend && isSPACE(*s))
8852 while (s < PL_bufend && isSPACE(*s))
8854 if (isIDFIRST_lazy_if(s,UTF)) {
8855 const char * const w = s;
8856 s += UTF ? UTF8SKIP(s) : 1;
8857 while (isALNUM_lazy_if(s,UTF))
8858 s += UTF ? UTF8SKIP(s) : 1;
8859 while (s < PL_bufend && isSPACE(*s))
8863 if (keyword(w, s - w, 0))
8866 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8867 if (gv && GvCVu(gv))
8869 Perl_croak(aTHX_ "No comma allowed after %s", what);
8874 /* Either returns sv, or mortalizes sv and returns a new SV*.
8875 Best used as sv=new_constant(..., sv, ...).
8876 If s, pv are NULL, calls subroutine with one argument,
8877 and type is used with error messages only. */
8880 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8881 SV *sv, SV *pv, const char *type, STRLEN typelen)
8884 HV * table = GvHV(PL_hintgv); /* ^H */
8888 const char *why1 = "", *why2 = "", *why3 = "";
8890 PERL_ARGS_ASSERT_NEW_CONSTANT;
8892 /* charnames doesn't work well if there have been errors found */
8893 if (PL_error_count > 0 && strEQ(key,"charnames"))
8894 return &PL_sv_undef;
8897 || ! (PL_hints & HINT_LOCALIZE_HH)
8898 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8903 /* Here haven't found what we're looking for. If it is charnames,
8904 * perhaps it needs to be loaded. Try doing that before giving up */
8905 if (strEQ(key,"charnames")) {
8906 Perl_load_module(aTHX_
8908 newSVpvs("_charnames"),
8909 /* version parameter; no need to specify it, as if
8910 * we get too early a version, will fail anyway,
8911 * not being able to find '_charnames' */
8917 table = GvHV(PL_hintgv);
8919 && (PL_hints & HINT_LOCALIZE_HH)
8920 && (cvp = hv_fetch(table, key, keylen, FALSE))
8926 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8927 msg = Perl_newSVpvf(aTHX_
8928 "Constant(%s) unknown", (type ? type: "undef"));
8933 why3 = "} is not defined";
8935 if (strEQ(key,"charnames")) {
8936 msg = Perl_newSVpvf(aTHX_
8937 /* The +3 is for '\N{'; -4 for that, plus '}' */
8938 "Unknown charname '%.*s'", (int)typelen - 4, type + 3);
8941 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8942 (type ? type: "undef"), why1, why2, why3);
8945 yyerror(SvPVX_const(msg));
8950 sv_2mortal(sv); /* Parent created it permanently */
8953 pv = newSVpvn_flags(s, len, SVs_TEMP);
8955 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8957 typesv = &PL_sv_undef;
8959 PUSHSTACKi(PERLSI_OVERLOAD);
8971 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8975 /* Check the eval first */
8976 if (!PL_in_eval && SvTRUE(ERRSV)) {
8977 sv_catpvs(ERRSV, "Propagated");
8978 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8980 res = SvREFCNT_inc_simple(sv);
8984 SvREFCNT_inc_simple_void(res);
8993 why1 = "Call to &{$^H{";
8995 why3 = "}} did not return a defined value";
9003 /* Returns a NUL terminated string, with the length of the string written to
9007 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
9011 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9013 PERL_ARGS_ASSERT_SCAN_WORD;
9017 Perl_croak(aTHX_ ident_too_long);
9018 if (isALNUM(*s) || (!UTF && isALNUMC_L1(*s))) /* UTF handled below */
9020 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
9025 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
9029 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9030 char *t = s + UTF8SKIP(s);
9032 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9036 Perl_croak(aTHX_ ident_too_long);
9037 Copy(s, d, len, char);
9050 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
9053 char *bracket = NULL;
9056 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9058 PERL_ARGS_ASSERT_SCAN_IDENT;
9063 while (isDIGIT(*s)) {
9065 Perl_croak(aTHX_ ident_too_long);
9072 Perl_croak(aTHX_ ident_too_long);
9073 if (isALNUM(*s)) /* UTF handled below */
9075 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
9080 else if (*s == ':' && s[1] == ':') {
9084 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9085 char *t = s + UTF8SKIP(s);
9086 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9088 if (d + (t - s) > e)
9089 Perl_croak(aTHX_ ident_too_long);
9090 Copy(s, d, t - s, char);
9101 if (PL_lex_state != LEX_NORMAL)
9102 PL_lex_state = LEX_INTERPENDMAYBE;
9105 if (*s == '$' && s[1] &&
9106 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
9116 const STRLEN skip = UTF8SKIP(s);
9119 for ( i = 0; i < skip; i++ )
9127 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9131 else if (ck_uni && !bracket)
9134 if (isSPACE(s[-1])) {
9136 const char ch = *s++;
9137 if (!SPACE_OR_TAB(ch)) {
9143 if (isIDFIRST_lazy_if(d,UTF)) {
9147 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
9148 end += UTF8SKIP(end);
9149 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
9150 end += UTF8SKIP(end);
9152 Copy(s, d, end - s, char);
9157 while ((isALNUM(*s) || *s == ':') && d < e)
9160 Perl_croak(aTHX_ ident_too_long);
9163 while (s < send && SPACE_OR_TAB(*s))
9165 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9166 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9167 const char * const brack =
9169 ((*s == '[') ? "[...]" : "{...}");
9170 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9171 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9172 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9173 funny, dest, brack, funny, dest, brack);
9176 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9177 PL_lex_allbrackets++;
9181 /* Handle extended ${^Foo} variables
9182 * 1999-02-27 mjd-perl-patch@plover.com */
9183 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
9187 while (isALNUM(*s) && d < e) {
9191 Perl_croak(aTHX_ ident_too_long);
9196 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9197 PL_lex_state = LEX_INTERPEND;
9200 if (PL_lex_state == LEX_NORMAL) {
9201 if (ckWARN(WARN_AMBIGUOUS) &&
9202 (keyword(dest, d - dest, 0)
9203 || get_cvn_flags(dest, d - dest, UTF ? SVf_UTF8 : 0)))
9205 SV *tmp = newSVpvn_flags( dest, d - dest,
9206 SVs_TEMP | (UTF ? SVf_UTF8 : 0) );
9209 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9210 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9211 funny, tmp, funny, tmp);
9216 s = bracket; /* let the parser handle it */
9220 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9221 PL_lex_state = LEX_INTERPEND;
9226 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
9228 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
9229 * the parse starting at 's', based on the subset that are valid in this
9230 * context input to this routine in 'valid_flags'. Advances s. Returns
9231 * TRUE if the input should be treated as a valid flag, so the next char
9232 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
9233 * first call on the current regex. This routine will set it to any
9234 * charset modifier found. The caller shouldn't change it. This way,
9235 * another charset modifier encountered in the parse can be detected as an
9236 * error, as we have decided to allow only one */
9239 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9241 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9242 if (isALNUM_lazy_if(*s, UTF)) {
9243 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9244 UTF ? SVf_UTF8 : 0);
9246 /* Pretend that it worked, so will continue processing before
9255 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
9256 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9257 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9258 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9259 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9260 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9261 case LOCALE_PAT_MOD:
9263 goto multiple_charsets;
9265 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9268 case UNICODE_PAT_MOD:
9270 goto multiple_charsets;
9272 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9275 case ASCII_RESTRICT_PAT_MOD:
9277 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9281 /* Error if previous modifier wasn't an 'a', but if it was, see
9282 * if, and accept, a second occurrence (only) */
9284 || get_regex_charset(*pmfl)
9285 != REGEX_ASCII_RESTRICTED_CHARSET)
9287 goto multiple_charsets;
9289 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9293 case DEPENDS_PAT_MOD:
9295 goto multiple_charsets;
9297 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9306 if (*charset != c) {
9307 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9309 else if (c == 'a') {
9310 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9313 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9316 /* Pretend that it worked, so will continue processing before dieing */
9322 S_scan_pat(pTHX_ char *start, I32 type)
9326 char *s = scan_str(start,!!PL_madskills,FALSE, PL_reg_state.re_reparsing);
9327 const char * const valid_flags =
9328 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9329 char charset = '\0'; /* character set modifier */
9334 PERL_ARGS_ASSERT_SCAN_PAT;
9336 /* this was only needed for the initial scan_str; set it to false
9337 * so that any (?{}) code blocks etc are parsed normally */
9338 PL_reg_state.re_reparsing = FALSE;
9340 const char * const delimiter = skipspace(start);
9344 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9345 : "Search pattern not terminated" ));
9348 pm = (PMOP*)newPMOP(type, 0);
9349 if (PL_multi_open == '?') {
9350 /* This is the only point in the code that sets PMf_ONCE: */
9351 pm->op_pmflags |= PMf_ONCE;
9353 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9354 allows us to restrict the list needed by reset to just the ??
9356 assert(type != OP_TRANS);
9358 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9361 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9364 elements = mg->mg_len / sizeof(PMOP**);
9365 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9366 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9367 mg->mg_len = elements * sizeof(PMOP**);
9368 PmopSTASH_set(pm,PL_curstash);
9375 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9376 * anon CV. False positives like qr/[(?{]/ are harmless */
9378 if (type == OP_QR) {
9380 char *e, *p = SvPV(PL_lex_stuff, len);
9382 for (; p < e; p++) {
9383 if (p[0] == '(' && p[1] == '?'
9384 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9386 pm->op_pmflags |= PMf_HAS_CV;
9390 pm->op_pmflags |= PMf_IS_QR;
9393 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9395 if (PL_madskills && modstart != s) {
9396 SV* tmptoken = newSVpvn(modstart, s - modstart);
9397 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9400 /* issue a warning if /c is specified,but /g is not */
9401 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9403 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9404 "Use of /c modifier is meaningless without /g" );
9407 PL_lex_op = (OP*)pm;
9408 pl_yylval.ival = OP_MATCH;
9413 S_scan_subst(pTHX_ char *start)
9420 char charset = '\0'; /* character set modifier */
9425 PERL_ARGS_ASSERT_SCAN_SUBST;
9427 pl_yylval.ival = OP_NULL;
9429 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9432 Perl_croak(aTHX_ "Substitution pattern not terminated");
9434 if (s[-1] == PL_multi_open)
9438 CURMAD('q', PL_thisopen);
9439 CURMAD('_', PL_thiswhite);
9440 CURMAD('E', PL_thisstuff);
9441 CURMAD('Q', PL_thisclose);
9442 PL_realtokenstart = s - SvPVX(PL_linestr);
9446 first_start = PL_multi_start;
9447 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9450 SvREFCNT_dec(PL_lex_stuff);
9451 PL_lex_stuff = NULL;
9453 Perl_croak(aTHX_ "Substitution replacement not terminated");
9455 PL_multi_start = first_start; /* so whole substitution is taken together */
9457 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9461 CURMAD('z', PL_thisopen);
9462 CURMAD('R', PL_thisstuff);
9463 CURMAD('Z', PL_thisclose);
9469 if (*s == EXEC_PAT_MOD) {
9473 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9482 curmad('m', newSVpvn(modstart, s - modstart));
9483 append_madprops(PL_thismad, (OP*)pm, 0);
9487 if ((pm->op_pmflags & PMf_CONTINUE)) {
9488 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9492 SV * const repl = newSVpvs("");
9495 pm->op_pmflags |= PMf_EVAL;
9498 sv_catpvs(repl, "eval ");
9500 sv_catpvs(repl, "do ");
9502 sv_catpvs(repl, "{");
9503 sv_catsv(repl, PL_sublex_info.repl);
9504 sv_catpvs(repl, "}");
9506 SvREFCNT_dec(PL_sublex_info.repl);
9507 PL_sublex_info.repl = repl;
9510 PL_lex_op = (OP*)pm;
9511 pl_yylval.ival = OP_SUBST;
9516 S_scan_trans(pTHX_ char *start)
9524 bool nondestruct = 0;
9529 PERL_ARGS_ASSERT_SCAN_TRANS;
9531 pl_yylval.ival = OP_NULL;
9533 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9535 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9537 if (s[-1] == PL_multi_open)
9541 CURMAD('q', PL_thisopen);
9542 CURMAD('_', PL_thiswhite);
9543 CURMAD('E', PL_thisstuff);
9544 CURMAD('Q', PL_thisclose);
9545 PL_realtokenstart = s - SvPVX(PL_linestr);
9549 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9552 SvREFCNT_dec(PL_lex_stuff);
9553 PL_lex_stuff = NULL;
9555 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9558 CURMAD('z', PL_thisopen);
9559 CURMAD('R', PL_thisstuff);
9560 CURMAD('Z', PL_thisclose);
9563 complement = del = squash = 0;
9570 complement = OPpTRANS_COMPLEMENT;
9573 del = OPpTRANS_DELETE;
9576 squash = OPpTRANS_SQUASH;
9588 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9589 o->op_private &= ~OPpTRANS_ALL;
9590 o->op_private |= del|squash|complement|
9591 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9592 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9595 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9600 curmad('m', newSVpvn(modstart, s - modstart));
9601 append_madprops(PL_thismad, o, 0);
9610 Takes a pointer to the first < in <<FOO.
9611 Returns a pointer to the byte following <<FOO.
9613 This function scans a heredoc, which involves different methods
9614 depending on whether we are in a string eval, quoted construct, etc.
9615 This is because PL_linestr could containing a single line of input, or
9616 a whole string being evalled, or the contents of the current quote-
9619 The two basic methods are:
9620 - Steal lines from the input stream
9621 - Scan the heredoc in PL_linestr and remove it therefrom
9623 In a file scope or filtered eval, the first method is used; in a
9624 string eval, the second.
9626 In a quote-like operator, we have to choose between the two,
9627 depending on where we can find a newline. We peek into outer lex-
9628 ing scopes until we find one with a newline in it. If we reach the
9629 outermost lexing scope and it is a file, we use the stream method.
9630 Otherwise it is treated as an eval.
9634 S_scan_heredoc(pTHX_ register char *s)
9637 I32 op_type = OP_SCALAR;
9644 const bool infile = PL_rsfp || PL_parser->filtered;
9645 LEXSHARED *shared = PL_parser->lex_shared;
9647 I32 stuffstart = s - SvPVX(PL_linestr);
9650 PL_realtokenstart = -1;
9653 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9656 d = PL_tokenbuf + 1;
9657 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9658 *PL_tokenbuf = '\n';
9660 while (SPACE_OR_TAB(*peek))
9662 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9665 s = delimcpy(d, e, s, PL_bufend, term, &len);
9667 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9673 /* <<\FOO is equivalent to <<'FOO' */
9677 if (!isALNUM_lazy_if(s,UTF))
9678 deprecate("bare << to mean <<\"\"");
9679 for (; isALNUM_lazy_if(s,UTF); s++) {
9684 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9685 Perl_croak(aTHX_ "Delimiter for here document is too long");
9688 len = d - PL_tokenbuf;
9692 tstart = PL_tokenbuf + 1;
9693 PL_thisclose = newSVpvn(tstart, len - 1);
9694 tstart = SvPVX(PL_linestr) + stuffstart;
9695 PL_thisopen = newSVpvn(tstart, s - tstart);
9696 stuffstart = s - SvPVX(PL_linestr);
9699 #ifndef PERL_STRICT_CR
9700 d = strchr(s, '\r');
9702 char * const olds = s;
9704 while (s < PL_bufend) {
9710 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9719 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9725 tstart = SvPVX(PL_linestr) + stuffstart;
9727 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9729 PL_thisstuff = newSVpvn(tstart, s - tstart);
9732 stuffstart = s - SvPVX(PL_linestr);
9735 tmpstr = newSV_type(SVt_PVIV);
9739 SvIV_set(tmpstr, -1);
9741 else if (term == '`') {
9742 op_type = OP_BACKTICK;
9743 SvIV_set(tmpstr, '\\');
9746 PL_multi_start = CopLINE(PL_curcop) + 1;
9747 PL_multi_open = PL_multi_close = '<';
9748 /* inside a string eval or quote-like operator */
9749 if (!infile || PL_lex_inwhat) {
9752 char * const olds = s;
9753 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9754 /* These two fields are not set until an inner lexing scope is
9755 entered. But we need them set here. */
9756 shared->ls_bufptr = s;
9757 shared->ls_linestr = PL_linestr;
9759 /* Look for a newline. If the current buffer does not have one,
9760 peek into the line buffer of the parent lexing scope, going
9761 up as many levels as necessary to find one with a newline
9764 while (!(s = (char *)memchr(
9765 (void *)shared->ls_bufptr, '\n',
9766 SvEND(shared->ls_linestr)-shared->ls_bufptr
9768 shared = shared->ls_prev;
9769 /* shared is only null if we have gone beyond the outermost
9770 lexing scope. In a file, we will have broken out of the
9771 loop in the previous iteration. In an eval, the string buf-
9772 fer ends with "\n;", so the while condition below will have
9773 evaluated to false. So shared can never be null. */
9775 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9776 most lexing scope. In a file, shared->ls_linestr at that
9777 level is just one line, so there is no body to steal. */
9778 if (infile && !shared->ls_prev) {
9784 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9787 linestr = shared->ls_linestr;
9788 bufend = SvEND(linestr);
9790 while (s < bufend &&
9791 (*s != '\n' || memNE(s,PL_tokenbuf,len)) ) {
9793 ++shared->herelines;
9798 sv_setpvn(tmpstr,d+1,s-d);
9802 sv_catpvn(PL_thisstuff, d + 1, s - d);
9804 PL_thisstuff = newSVpvn(d + 1, s - d);
9805 stuffstart = s - SvPVX(PL_linestr);
9809 /* the preceding stmt passes a newline */
9810 shared->herelines++;
9812 /* s now points to the newline after the heredoc terminator.
9813 d points to the newline before the body of the heredoc.
9816 /* We are going to modify linestr in place here, so set
9817 aside copies of the string if necessary for re-evals or
9819 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9820 check shared->re_eval_str. */
9821 if (shared->re_eval_start || shared->re_eval_str) {
9822 /* Set aside the rest of the regexp */
9823 if (!shared->re_eval_str)
9824 shared->re_eval_str =
9825 newSVpvn(shared->re_eval_start,
9826 bufend - shared->re_eval_start);
9827 shared->re_eval_start -= s-d;
9829 if (cxstack_ix >= 0 && CxTYPE(cx) == CXt_EVAL &&
9830 CxOLD_OP_TYPE(cx) == OP_ENTEREVAL &&
9831 cx->blk_eval.cur_text == linestr)
9833 cx->blk_eval.cur_text = newSVsv(linestr);
9834 SvSCREAM_on(cx->blk_eval.cur_text);
9836 /* Copy everything from s onwards back to d. */
9837 Move(s,d,bufend-s + 1,char);
9838 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9839 /* Setting PL_bufend only applies when we have not dug deeper
9840 into other scopes, because sublex_done sets PL_bufend to
9841 SvEND(PL_linestr). */
9842 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9849 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9850 term = PL_tokenbuf[1];
9852 linestr_save = PL_linestr; /* must restore this afterwards */
9853 d = s; /* and this */
9854 PL_linestr = newSVpvs("");
9855 PL_bufend = SvPVX(PL_linestr);
9859 tstart = SvPVX(PL_linestr) + stuffstart;
9861 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9863 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9866 PL_bufptr = PL_bufend;
9867 CopLINE_set(PL_curcop,
9868 PL_multi_start + shared->herelines);
9869 if (!lex_next_chunk(LEX_NO_TERM)
9870 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9871 SvREFCNT_dec(linestr_save);
9874 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
9875 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9876 lex_grow_linestr(SvCUR(PL_linestr) + 2);
9877 sv_catpvs(PL_linestr, "\n\0");
9881 stuffstart = s - SvPVX(PL_linestr);
9883 shared->herelines++;
9884 PL_last_lop = PL_last_uni = NULL;
9885 #ifndef PERL_STRICT_CR
9886 if (PL_bufend - PL_linestart >= 2) {
9887 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9888 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9890 PL_bufend[-2] = '\n';
9892 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9894 else if (PL_bufend[-1] == '\r')
9895 PL_bufend[-1] = '\n';
9897 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9898 PL_bufend[-1] = '\n';
9900 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
9901 SvREFCNT_dec(PL_linestr);
9902 PL_linestr = linestr_save;
9903 PL_linestart = SvPVX(linestr_save);
9904 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9909 sv_catsv(tmpstr,PL_linestr);
9913 PL_multi_end = CopLINE(PL_curcop);
9914 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9915 SvPV_shrink_to_cur(tmpstr);
9918 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9920 else if (PL_encoding)
9921 sv_recode_to_utf8(tmpstr, PL_encoding);
9923 PL_lex_stuff = tmpstr;
9924 pl_yylval.ival = op_type;
9928 SvREFCNT_dec(tmpstr);
9929 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
9930 missingterm(PL_tokenbuf + 1);
9934 takes: current position in input buffer
9935 returns: new position in input buffer
9936 side-effects: pl_yylval and lex_op are set.
9941 <FH> read from filehandle
9942 <pkg::FH> read from package qualified filehandle
9943 <pkg'FH> read from package qualified filehandle
9944 <$fh> read from filehandle in $fh
9950 S_scan_inputsymbol(pTHX_ char *start)
9953 char *s = start; /* current position in buffer */
9956 char *d = PL_tokenbuf; /* start of temp holding space */
9957 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9959 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9961 end = strchr(s, '\n');
9964 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9966 /* die if we didn't have space for the contents of the <>,
9967 or if it didn't end, or if we see a newline
9970 if (len >= (I32)sizeof PL_tokenbuf)
9971 Perl_croak(aTHX_ "Excessively long <> operator");
9973 Perl_croak(aTHX_ "Unterminated <> operator");
9978 Remember, only scalar variables are interpreted as filehandles by
9979 this code. Anything more complex (e.g., <$fh{$num}>) will be
9980 treated as a glob() call.
9981 This code makes use of the fact that except for the $ at the front,
9982 a scalar variable and a filehandle look the same.
9984 if (*d == '$' && d[1]) d++;
9986 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9987 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9988 d += UTF ? UTF8SKIP(d) : 1;
9990 /* If we've tried to read what we allow filehandles to look like, and
9991 there's still text left, then it must be a glob() and not a getline.
9992 Use scan_str to pull out the stuff between the <> and treat it
9993 as nothing more than a string.
9996 if (d - PL_tokenbuf != len) {
9997 pl_yylval.ival = OP_GLOB;
9998 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
10000 Perl_croak(aTHX_ "Glob not terminated");
10004 bool readline_overriden = FALSE;
10007 /* we're in a filehandle read situation */
10010 /* turn <> into <ARGV> */
10012 Copy("ARGV",d,5,char);
10014 /* Check whether readline() is overriden */
10015 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
10017 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
10019 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
10020 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
10021 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
10022 readline_overriden = TRUE;
10024 /* if <$fh>, create the ops to turn the variable into a
10028 /* try to find it in the pad for this block, otherwise find
10029 add symbol table ops
10031 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
10032 if (tmp != NOT_IN_PAD) {
10033 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
10034 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
10035 HEK * const stashname = HvNAME_HEK(stash);
10036 SV * const sym = sv_2mortal(newSVhek(stashname));
10037 sv_catpvs(sym, "::");
10038 sv_catpv(sym, d+1);
10043 OP * const o = newOP(OP_PADSV, 0);
10045 PL_lex_op = readline_overriden
10046 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10047 op_append_elem(OP_LIST, o,
10048 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
10049 : (OP*)newUNOP(OP_READLINE, 0, o);
10058 ? (GV_ADDMULTI | GV_ADDINEVAL)
10059 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
10061 PL_lex_op = readline_overriden
10062 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10063 op_append_elem(OP_LIST,
10064 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
10065 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10066 : (OP*)newUNOP(OP_READLINE, 0,
10067 newUNOP(OP_RV2SV, 0,
10068 newGVOP(OP_GV, 0, gv)));
10070 if (!readline_overriden)
10071 PL_lex_op->op_flags |= OPf_SPECIAL;
10072 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
10073 pl_yylval.ival = OP_NULL;
10076 /* If it's none of the above, it must be a literal filehandle
10077 (<Foo::BAR> or <FOO>) so build a simple readline OP */
10079 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
10080 PL_lex_op = readline_overriden
10081 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10082 op_append_elem(OP_LIST,
10083 newGVOP(OP_GV, 0, gv),
10084 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10085 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
10086 pl_yylval.ival = OP_NULL;
10095 takes: start position in buffer
10096 keep_quoted preserve \ on the embedded delimiter(s)
10097 keep_delims preserve the delimiters around the string
10098 re_reparse compiling a run-time /(?{})/:
10099 collapse // to /, and skip encoding src
10100 returns: position to continue reading from buffer
10101 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
10102 updates the read buffer.
10104 This subroutine pulls a string out of the input. It is called for:
10105 q single quotes q(literal text)
10106 ' single quotes 'literal text'
10107 qq double quotes qq(interpolate $here please)
10108 " double quotes "interpolate $here please"
10109 qx backticks qx(/bin/ls -l)
10110 ` backticks `/bin/ls -l`
10111 qw quote words @EXPORT_OK = qw( func() $spam )
10112 m// regexp match m/this/
10113 s/// regexp substitute s/this/that/
10114 tr/// string transliterate tr/this/that/
10115 y/// string transliterate y/this/that/
10116 ($*@) sub prototypes sub foo ($)
10117 (stuff) sub attr parameters sub foo : attr(stuff)
10118 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
10120 In most of these cases (all but <>, patterns and transliterate)
10121 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
10122 calls scan_str(). s/// makes yylex() call scan_subst() which calls
10123 scan_str(). tr/// and y/// make yylex() call scan_trans() which
10126 It skips whitespace before the string starts, and treats the first
10127 character as the delimiter. If the delimiter is one of ([{< then
10128 the corresponding "close" character )]}> is used as the closing
10129 delimiter. It allows quoting of delimiters, and if the string has
10130 balanced delimiters ([{<>}]) it allows nesting.
10132 On success, the SV with the resulting string is put into lex_stuff or,
10133 if that is already non-NULL, into lex_repl. The second case occurs only
10134 when parsing the RHS of the special constructs s/// and tr/// (y///).
10135 For convenience, the terminating delimiter character is stuffed into
10140 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims, int re_reparse)
10143 SV *sv; /* scalar value: string */
10144 const char *tmps; /* temp string, used for delimiter matching */
10145 char *s = start; /* current position in the buffer */
10146 char term; /* terminating character */
10147 char *to; /* current position in the sv's data */
10148 I32 brackets = 1; /* bracket nesting level */
10149 bool has_utf8 = FALSE; /* is there any utf8 content? */
10150 I32 termcode; /* terminating char. code */
10151 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
10152 STRLEN termlen; /* length of terminating string */
10153 int last_off = 0; /* last position for nesting bracket */
10159 PERL_ARGS_ASSERT_SCAN_STR;
10161 /* skip space before the delimiter */
10167 if (PL_realtokenstart >= 0) {
10168 stuffstart = PL_realtokenstart;
10169 PL_realtokenstart = -1;
10172 stuffstart = start - SvPVX(PL_linestr);
10174 /* mark where we are, in case we need to report errors */
10177 /* after skipping whitespace, the next character is the terminator */
10180 termcode = termstr[0] = term;
10184 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
10185 Copy(s, termstr, termlen, U8);
10186 if (!UTF8_IS_INVARIANT(term))
10190 /* mark where we are */
10191 PL_multi_start = CopLINE(PL_curcop);
10192 PL_multi_open = term;
10194 /* find corresponding closing delimiter */
10195 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
10196 termcode = termstr[0] = term = tmps[5];
10198 PL_multi_close = term;
10200 /* create a new SV to hold the contents. 79 is the SV's initial length.
10201 What a random number. */
10202 sv = newSV_type(SVt_PVIV);
10204 SvIV_set(sv, termcode);
10205 (void)SvPOK_only(sv); /* validate pointer */
10207 /* move past delimiter and try to read a complete string */
10209 sv_catpvn(sv, s, termlen);
10212 tstart = SvPVX(PL_linestr) + stuffstart;
10213 if (!PL_thisopen && !keep_delims) {
10214 PL_thisopen = newSVpvn(tstart, s - tstart);
10215 stuffstart = s - SvPVX(PL_linestr);
10219 if (PL_encoding && !UTF && !re_reparse) {
10223 int offset = s - SvPVX_const(PL_linestr);
10224 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
10225 &offset, (char*)termstr, termlen);
10226 const char * const ns = SvPVX_const(PL_linestr) + offset;
10227 char * const svlast = SvEND(sv) - 1;
10229 for (; s < ns; s++) {
10230 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10231 COPLINE_INC_WITH_HERELINES;
10234 goto read_more_line;
10236 /* handle quoted delimiters */
10237 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10239 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10241 if ((svlast-1 - t) % 2) {
10242 if (!keep_quoted) {
10243 *(svlast-1) = term;
10245 SvCUR_set(sv, SvCUR(sv) - 1);
10250 if (PL_multi_open == PL_multi_close) {
10256 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10257 /* At here, all closes are "was quoted" one,
10258 so we don't check PL_multi_close. */
10260 if (!keep_quoted && *(t+1) == PL_multi_open)
10265 else if (*t == PL_multi_open)
10273 SvCUR_set(sv, w - SvPVX_const(sv));
10275 last_off = w - SvPVX(sv);
10276 if (--brackets <= 0)
10281 if (!keep_delims) {
10282 SvCUR_set(sv, SvCUR(sv) - 1);
10288 /* extend sv if need be */
10289 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10290 /* set 'to' to the next character in the sv's string */
10291 to = SvPVX(sv)+SvCUR(sv);
10293 /* if open delimiter is the close delimiter read unbridle */
10294 if (PL_multi_open == PL_multi_close) {
10295 for (; s < PL_bufend; s++,to++) {
10296 /* embedded newlines increment the current line number */
10297 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10298 COPLINE_INC_WITH_HERELINES;
10299 /* handle quoted delimiters */
10300 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10303 || (re_reparse && s[1] == '\\'))
10306 /* any other quotes are simply copied straight through */
10310 /* terminate when run out of buffer (the for() condition), or
10311 have found the terminator */
10312 else if (*s == term) {
10315 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10318 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10324 /* if the terminator isn't the same as the start character (e.g.,
10325 matched brackets), we have to allow more in the quoting, and
10326 be prepared for nested brackets.
10329 /* read until we run out of string, or we find the terminator */
10330 for (; s < PL_bufend; s++,to++) {
10331 /* embedded newlines increment the line count */
10332 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10333 COPLINE_INC_WITH_HERELINES;
10334 /* backslashes can escape the open or closing characters */
10335 if (*s == '\\' && s+1 < PL_bufend) {
10336 if (!keep_quoted &&
10337 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10342 /* allow nested opens and closes */
10343 else if (*s == PL_multi_close && --brackets <= 0)
10345 else if (*s == PL_multi_open)
10347 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10352 /* terminate the copied string and update the sv's end-of-string */
10354 SvCUR_set(sv, to - SvPVX_const(sv));
10357 * this next chunk reads more into the buffer if we're not done yet
10361 break; /* handle case where we are done yet :-) */
10363 #ifndef PERL_STRICT_CR
10364 if (to - SvPVX_const(sv) >= 2) {
10365 if ((to[-2] == '\r' && to[-1] == '\n') ||
10366 (to[-2] == '\n' && to[-1] == '\r'))
10370 SvCUR_set(sv, to - SvPVX_const(sv));
10372 else if (to[-1] == '\r')
10375 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10380 /* if we're out of file, or a read fails, bail and reset the current
10381 line marker so we can report where the unterminated string began
10384 if (PL_madskills) {
10385 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10387 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10389 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10392 COPLINE_INC_WITH_HERELINES;
10393 PL_bufptr = PL_bufend;
10394 if (!lex_next_chunk(0)) {
10396 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10405 /* at this point, we have successfully read the delimited string */
10407 if (!PL_encoding || UTF || re_reparse) {
10409 if (PL_madskills) {
10410 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10411 const int len = s - tstart;
10413 sv_catpvn(PL_thisstuff, tstart, len);
10415 PL_thisstuff = newSVpvn(tstart, len);
10416 if (!PL_thisclose && !keep_delims)
10417 PL_thisclose = newSVpvn(s,termlen);
10422 sv_catpvn(sv, s, termlen);
10427 if (PL_madskills) {
10428 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10429 const int len = s - tstart - termlen;
10431 sv_catpvn(PL_thisstuff, tstart, len);
10433 PL_thisstuff = newSVpvn(tstart, len);
10434 if (!PL_thisclose && !keep_delims)
10435 PL_thisclose = newSVpvn(s - termlen,termlen);
10439 if (has_utf8 || (PL_encoding && !re_reparse))
10442 PL_multi_end = CopLINE(PL_curcop);
10444 /* if we allocated too much space, give some back */
10445 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10446 SvLEN_set(sv, SvCUR(sv) + 1);
10447 SvPV_renew(sv, SvLEN(sv));
10450 /* decide whether this is the first or second quoted string we've read
10455 PL_sublex_info.repl = sv;
10463 takes: pointer to position in buffer
10464 returns: pointer to new position in buffer
10465 side-effects: builds ops for the constant in pl_yylval.op
10467 Read a number in any of the formats that Perl accepts:
10469 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10470 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10473 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10475 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10478 If it reads a number without a decimal point or an exponent, it will
10479 try converting the number to an integer and see if it can do so
10480 without loss of precision.
10484 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10487 const char *s = start; /* current position in buffer */
10488 char *d; /* destination in temp buffer */
10489 char *e; /* end of temp buffer */
10490 NV nv; /* number read, as a double */
10491 SV *sv = NULL; /* place to put the converted number */
10492 bool floatit; /* boolean: int or float? */
10493 const char *lastub = NULL; /* position of last underbar */
10494 static char const number_too_long[] = "Number too long";
10496 PERL_ARGS_ASSERT_SCAN_NUM;
10498 /* We use the first character to decide what type of number this is */
10502 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10504 /* if it starts with a 0, it could be an octal number, a decimal in
10505 0.13 disguise, or a hexadecimal number, or a binary number. */
10509 u holds the "number so far"
10510 shift the power of 2 of the base
10511 (hex == 4, octal == 3, binary == 1)
10512 overflowed was the number more than we can hold?
10514 Shift is used when we add a digit. It also serves as an "are
10515 we in octal/hex/binary?" indicator to disallow hex characters
10516 when in octal mode.
10521 bool overflowed = FALSE;
10522 bool just_zero = TRUE; /* just plain 0 or binary number? */
10523 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10524 static const char* const bases[5] =
10525 { "", "binary", "", "octal", "hexadecimal" };
10526 static const char* const Bases[5] =
10527 { "", "Binary", "", "Octal", "Hexadecimal" };
10528 static const char* const maxima[5] =
10530 "0b11111111111111111111111111111111",
10534 const char *base, *Base, *max;
10536 /* check for hex */
10537 if (s[1] == 'x' || s[1] == 'X') {
10541 } else if (s[1] == 'b' || s[1] == 'B') {
10546 /* check for a decimal in disguise */
10547 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10549 /* so it must be octal */
10556 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10557 "Misplaced _ in number");
10561 base = bases[shift];
10562 Base = Bases[shift];
10563 max = maxima[shift];
10565 /* read the rest of the number */
10567 /* x is used in the overflow test,
10568 b is the digit we're adding on. */
10573 /* if we don't mention it, we're done */
10577 /* _ are ignored -- but warned about if consecutive */
10579 if (lastub && s == lastub + 1)
10580 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10581 "Misplaced _ in number");
10585 /* 8 and 9 are not octal */
10586 case '8': case '9':
10588 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10592 case '2': case '3': case '4':
10593 case '5': case '6': case '7':
10595 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10598 case '0': case '1':
10599 b = *s++ & 15; /* ASCII digit -> value of digit */
10603 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10604 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10605 /* make sure they said 0x */
10608 b = (*s++ & 7) + 9;
10610 /* Prepare to put the digit we have onto the end
10611 of the number so far. We check for overflows.
10617 x = u << shift; /* make room for the digit */
10619 if ((x >> shift) != u
10620 && !(PL_hints & HINT_NEW_BINARY)) {
10623 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10624 "Integer overflow in %s number",
10627 u = x | b; /* add the digit to the end */
10630 n *= nvshift[shift];
10631 /* If an NV has not enough bits in its
10632 * mantissa to represent an UV this summing of
10633 * small low-order numbers is a waste of time
10634 * (because the NV cannot preserve the
10635 * low-order bits anyway): we could just
10636 * remember when did we overflow and in the
10637 * end just multiply n by the right
10645 /* if we get here, we had success: make a scalar value from
10650 /* final misplaced underbar check */
10651 if (s[-1] == '_') {
10652 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10656 if (n > 4294967295.0)
10657 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10658 "%s number > %s non-portable",
10664 if (u > 0xffffffff)
10665 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10666 "%s number > %s non-portable",
10671 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10672 sv = new_constant(start, s - start, "integer",
10673 sv, NULL, NULL, 0);
10674 else if (PL_hints & HINT_NEW_BINARY)
10675 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10680 handle decimal numbers.
10681 we're also sent here when we read a 0 as the first digit
10683 case '1': case '2': case '3': case '4': case '5':
10684 case '6': case '7': case '8': case '9': case '.':
10687 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10690 /* read next group of digits and _ and copy into d */
10691 while (isDIGIT(*s) || *s == '_') {
10692 /* skip underscores, checking for misplaced ones
10696 if (lastub && s == lastub + 1)
10697 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10698 "Misplaced _ in number");
10702 /* check for end of fixed-length buffer */
10704 Perl_croak(aTHX_ number_too_long);
10705 /* if we're ok, copy the character */
10710 /* final misplaced underbar check */
10711 if (lastub && s == lastub + 1) {
10712 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10715 /* read a decimal portion if there is one. avoid
10716 3..5 being interpreted as the number 3. followed
10719 if (*s == '.' && s[1] != '.') {
10724 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10725 "Misplaced _ in number");
10729 /* copy, ignoring underbars, until we run out of digits.
10731 for (; isDIGIT(*s) || *s == '_'; s++) {
10732 /* fixed length buffer check */
10734 Perl_croak(aTHX_ number_too_long);
10736 if (lastub && s == lastub + 1)
10737 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10738 "Misplaced _ in number");
10744 /* fractional part ending in underbar? */
10745 if (s[-1] == '_') {
10746 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10747 "Misplaced _ in number");
10749 if (*s == '.' && isDIGIT(s[1])) {
10750 /* oops, it's really a v-string, but without the "v" */
10756 /* read exponent part, if present */
10757 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10761 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10762 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10764 /* stray preinitial _ */
10766 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10767 "Misplaced _ in number");
10771 /* allow positive or negative exponent */
10772 if (*s == '+' || *s == '-')
10775 /* stray initial _ */
10777 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10778 "Misplaced _ in number");
10782 /* read digits of exponent */
10783 while (isDIGIT(*s) || *s == '_') {
10786 Perl_croak(aTHX_ number_too_long);
10790 if (((lastub && s == lastub + 1) ||
10791 (!isDIGIT(s[1]) && s[1] != '_')))
10792 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10793 "Misplaced _ in number");
10801 We try to do an integer conversion first if no characters
10802 indicating "float" have been found.
10807 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10809 if (flags == IS_NUMBER_IN_UV) {
10811 sv = newSViv(uv); /* Prefer IVs over UVs. */
10814 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10815 if (uv <= (UV) IV_MIN)
10816 sv = newSViv(-(IV)uv);
10823 /* terminate the string */
10825 nv = Atof(PL_tokenbuf);
10830 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10831 const char *const key = floatit ? "float" : "integer";
10832 const STRLEN keylen = floatit ? 5 : 7;
10833 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10834 key, keylen, sv, NULL, NULL, 0);
10838 /* if it starts with a v, it could be a v-string */
10841 sv = newSV(5); /* preallocate storage space */
10842 s = scan_vstring(s, PL_bufend, sv);
10846 /* make the op for the constant and return */
10849 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10851 lvalp->opval = NULL;
10857 S_scan_formline(pTHX_ register char *s)
10862 SV * const stuff = newSVpvs("");
10863 bool needargs = FALSE;
10864 bool eofmt = FALSE;
10866 char *tokenstart = s;
10867 SV* savewhite = NULL;
10869 if (PL_madskills) {
10870 savewhite = PL_thiswhite;
10875 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10877 while (!needargs) {
10880 #ifdef PERL_STRICT_CR
10881 while (SPACE_OR_TAB(*t))
10884 while (SPACE_OR_TAB(*t) || *t == '\r')
10887 if (*t == '\n' || t == PL_bufend) {
10892 eol = (char *) memchr(s,'\n',PL_bufend-s);
10896 for (t = s; t < eol; t++) {
10897 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10899 goto enough; /* ~~ must be first line in formline */
10901 if (*t == '@' || *t == '^')
10905 sv_catpvn(stuff, s, eol-s);
10906 #ifndef PERL_STRICT_CR
10907 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10908 char *end = SvPVX(stuff) + SvCUR(stuff);
10911 SvCUR_set(stuff, SvCUR(stuff) - 1);
10919 if ((PL_rsfp || PL_parser->filtered)
10920 && PL_parser->form_lex_state == LEX_NORMAL) {
10923 if (PL_madskills) {
10925 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10927 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10930 PL_bufptr = PL_bufend;
10931 COPLINE_INC_WITH_HERELINES;
10932 got_some = lex_next_chunk(0);
10933 CopLINE_dec(PL_curcop);
10936 tokenstart = PL_bufptr;
10944 if (!SvCUR(stuff) || needargs)
10945 PL_lex_state = PL_parser->form_lex_state;
10946 if (SvCUR(stuff)) {
10947 PL_expect = XSTATE;
10949 start_force(PL_curforce);
10950 NEXTVAL_NEXTTOKE.ival = 0;
10951 force_next(FORMLBRACK);
10954 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10956 else if (PL_encoding)
10957 sv_recode_to_utf8(stuff, PL_encoding);
10959 start_force(PL_curforce);
10960 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10964 SvREFCNT_dec(stuff);
10966 PL_lex_formbrack = 0;
10969 if (PL_madskills) {
10971 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10973 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10974 PL_thiswhite = savewhite;
10981 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10984 const I32 oldsavestack_ix = PL_savestack_ix;
10985 CV* const outsidecv = PL_compcv;
10987 SAVEI32(PL_subline);
10988 save_item(PL_subname);
10989 SAVESPTR(PL_compcv);
10991 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10992 CvFLAGS(PL_compcv) |= flags;
10994 PL_subline = CopLINE(PL_curcop);
10995 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10996 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10997 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10998 if (outsidecv && CvPADLIST(outsidecv))
10999 CvPADLIST(PL_compcv)->xpadl_outid =
11000 PadlistNAMES(CvPADLIST(outsidecv));
11002 return oldsavestack_ix;
11006 #pragma segment Perl_yylex
11009 S_yywarn(pTHX_ const char *const s, U32 flags)
11013 PERL_ARGS_ASSERT_YYWARN;
11015 PL_in_eval |= EVAL_WARNONLY;
11016 yyerror_pv(s, flags);
11017 PL_in_eval &= ~EVAL_WARNONLY;
11022 Perl_yyerror(pTHX_ const char *const s)
11024 PERL_ARGS_ASSERT_YYERROR;
11025 return yyerror_pvn(s, strlen(s), 0);
11029 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
11031 PERL_ARGS_ASSERT_YYERROR_PV;
11032 return yyerror_pvn(s, strlen(s), flags);
11036 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
11039 const char *context = NULL;
11042 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
11043 int yychar = PL_parser->yychar;
11044 U32 is_utf8 = flags & SVf_UTF8;
11046 PERL_ARGS_ASSERT_YYERROR_PVN;
11048 if (!yychar || (yychar == ';' && !PL_rsfp))
11049 sv_catpvs(where_sv, "at EOF");
11050 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
11051 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
11052 PL_oldbufptr != PL_bufptr) {
11055 The code below is removed for NetWare because it abends/crashes on NetWare
11056 when the script has error such as not having the closing quotes like:
11057 if ($var eq "value)
11058 Checking of white spaces is anyway done in NetWare code.
11061 while (isSPACE(*PL_oldoldbufptr))
11064 context = PL_oldoldbufptr;
11065 contlen = PL_bufptr - PL_oldoldbufptr;
11067 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
11068 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
11071 The code below is removed for NetWare because it abends/crashes on NetWare
11072 when the script has error such as not having the closing quotes like:
11073 if ($var eq "value)
11074 Checking of white spaces is anyway done in NetWare code.
11077 while (isSPACE(*PL_oldbufptr))
11080 context = PL_oldbufptr;
11081 contlen = PL_bufptr - PL_oldbufptr;
11083 else if (yychar > 255)
11084 sv_catpvs(where_sv, "next token ???");
11085 else if (yychar == -2) { /* YYEMPTY */
11086 if (PL_lex_state == LEX_NORMAL ||
11087 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11088 sv_catpvs(where_sv, "at end of line");
11089 else if (PL_lex_inpat)
11090 sv_catpvs(where_sv, "within pattern");
11092 sv_catpvs(where_sv, "within string");
11095 sv_catpvs(where_sv, "next char ");
11097 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11098 else if (isPRINT_LC(yychar)) {
11099 const char string = yychar;
11100 sv_catpvn(where_sv, &string, 1);
11103 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11105 msg = sv_2mortal(newSVpvn_flags(s, len, is_utf8));
11106 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11107 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
11109 Perl_sv_catpvf(aTHX_ msg, "near \"%"SVf"\"\n",
11110 SVfARG(newSVpvn_flags(context, contlen,
11111 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
11113 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11114 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11115 Perl_sv_catpvf(aTHX_ msg,
11116 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11117 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11120 if (PL_in_eval & EVAL_WARNONLY) {
11121 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11125 if (PL_error_count >= 10) {
11126 if (PL_in_eval && SvCUR(ERRSV))
11127 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11128 SVfARG(ERRSV), OutCopFILE(PL_curcop));
11130 Perl_croak(aTHX_ "%s has too many errors.\n",
11131 OutCopFILE(PL_curcop));
11134 PL_in_my_stash = NULL;
11138 #pragma segment Main
11142 S_swallow_bom(pTHX_ U8 *s)
11145 const STRLEN slen = SvCUR(PL_linestr);
11147 PERL_ARGS_ASSERT_SWALLOW_BOM;
11151 if (s[1] == 0xFE) {
11152 /* UTF-16 little-endian? (or UTF-32LE?) */
11153 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11154 /* diag_listed_as: Unsupported script encoding %s */
11155 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11156 #ifndef PERL_NO_UTF16_FILTER
11157 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11159 if (PL_bufend > (char*)s) {
11160 s = add_utf16_textfilter(s, TRUE);
11163 /* diag_listed_as: Unsupported script encoding %s */
11164 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11169 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11170 #ifndef PERL_NO_UTF16_FILTER
11171 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11173 if (PL_bufend > (char *)s) {
11174 s = add_utf16_textfilter(s, FALSE);
11177 /* diag_listed_as: Unsupported script encoding %s */
11178 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11183 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
11184 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11185 s += 3; /* UTF-8 */
11191 if (s[2] == 0xFE && s[3] == 0xFF) {
11192 /* UTF-32 big-endian */
11193 /* diag_listed_as: Unsupported script encoding %s */
11194 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11197 else if (s[2] == 0 && s[3] != 0) {
11200 * are a good indicator of UTF-16BE. */
11201 #ifndef PERL_NO_UTF16_FILTER
11202 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11203 s = add_utf16_textfilter(s, FALSE);
11205 /* diag_listed_as: Unsupported script encoding %s */
11206 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11212 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
11213 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11214 s += 4; /* UTF-8 */
11220 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11223 * are a good indicator of UTF-16LE. */
11224 #ifndef PERL_NO_UTF16_FILTER
11225 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11226 s = add_utf16_textfilter(s, TRUE);
11228 /* diag_listed_as: Unsupported script encoding %s */
11229 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11237 #ifndef PERL_NO_UTF16_FILTER
11239 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11242 SV *const filter = FILTER_DATA(idx);
11243 /* We re-use this each time round, throwing the contents away before we
11245 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11246 SV *const utf8_buffer = filter;
11247 IV status = IoPAGE(filter);
11248 const bool reverse = cBOOL(IoLINES(filter));
11251 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11253 /* As we're automatically added, at the lowest level, and hence only called
11254 from this file, we can be sure that we're not called in block mode. Hence
11255 don't bother writing code to deal with block mode. */
11257 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11260 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11262 DEBUG_P(PerlIO_printf(Perl_debug_log,
11263 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11264 FPTR2DPTR(void *, S_utf16_textfilter),
11265 reverse ? 'l' : 'b', idx, maxlen, status,
11266 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11273 /* First, look in our buffer of existing UTF-8 data: */
11274 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11278 } else if (status == 0) {
11280 IoPAGE(filter) = 0;
11281 nl = SvEND(utf8_buffer);
11284 STRLEN got = nl - SvPVX(utf8_buffer);
11285 /* Did we have anything to append? */
11287 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11288 /* Everything else in this code works just fine if SVp_POK isn't
11289 set. This, however, needs it, and we need it to work, else
11290 we loop infinitely because the buffer is never consumed. */
11291 sv_chop(utf8_buffer, nl);
11295 /* OK, not a complete line there, so need to read some more UTF-16.
11296 Read an extra octect if the buffer currently has an odd number. */
11300 if (SvCUR(utf16_buffer) >= 2) {
11301 /* Location of the high octet of the last complete code point.
11302 Gosh, UTF-16 is a pain. All the benefits of variable length,
11303 *coupled* with all the benefits of partial reads and
11305 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11306 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11308 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11312 /* We have the first half of a surrogate. Read more. */
11313 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11316 status = FILTER_READ(idx + 1, utf16_buffer,
11317 160 + (SvCUR(utf16_buffer) & 1));
11318 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11319 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11322 IoPAGE(filter) = status;
11327 chars = SvCUR(utf16_buffer) >> 1;
11328 have = SvCUR(utf8_buffer);
11329 SvGROW(utf8_buffer, have + chars * 3 + 1);
11332 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11333 (U8*)SvPVX_const(utf8_buffer) + have,
11334 chars * 2, &newlen);
11336 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11337 (U8*)SvPVX_const(utf8_buffer) + have,
11338 chars * 2, &newlen);
11340 SvCUR_set(utf8_buffer, have + newlen);
11343 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11344 it's private to us, and utf16_to_utf8{,reversed} take a
11345 (pointer,length) pair, rather than a NUL-terminated string. */
11346 if(SvCUR(utf16_buffer) & 1) {
11347 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11348 SvCUR_set(utf16_buffer, 1);
11350 SvCUR_set(utf16_buffer, 0);
11353 DEBUG_P(PerlIO_printf(Perl_debug_log,
11354 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11356 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11357 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11362 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11364 SV *filter = filter_add(S_utf16_textfilter, NULL);
11366 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11368 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11369 sv_setpvs(filter, "");
11370 IoLINES(filter) = reversed;
11371 IoPAGE(filter) = 1; /* Not EOF */
11373 /* Sadly, we have to return a valid pointer, come what may, so we have to
11374 ignore any error return from this. */
11375 SvCUR_set(PL_linestr, 0);
11376 if (FILTER_READ(0, PL_linestr, 0)) {
11377 SvUTF8_on(PL_linestr);
11379 SvUTF8_on(PL_linestr);
11381 PL_bufend = SvEND(PL_linestr);
11382 return (U8*)SvPVX(PL_linestr);
11387 Returns a pointer to the next character after the parsed
11388 vstring, as well as updating the passed in sv.
11390 Function must be called like
11393 s = scan_vstring(s,e,sv);
11395 where s and e are the start and end of the string.
11396 The sv should already be large enough to store the vstring
11397 passed in, for performance reasons.
11402 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11405 const char *pos = s;
11406 const char *start = s;
11408 PERL_ARGS_ASSERT_SCAN_VSTRING;
11410 if (*pos == 'v') pos++; /* get past 'v' */
11411 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11413 if ( *pos != '.') {
11414 /* this may not be a v-string if followed by => */
11415 const char *next = pos;
11416 while (next < e && isSPACE(*next))
11418 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11419 /* return string not v-string */
11420 sv_setpvn(sv,(char *)s,pos-s);
11421 return (char *)pos;
11425 if (!isALPHA(*pos)) {
11426 U8 tmpbuf[UTF8_MAXBYTES+1];
11429 s++; /* get past 'v' */
11434 /* this is atoi() that tolerates underscores */
11437 const char *end = pos;
11439 while (--end >= s) {
11441 const UV orev = rev;
11442 rev += (*end - '0') * mult;
11445 /* diag_listed_as: Integer overflow in %s number */
11446 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11447 "Integer overflow in decimal number");
11451 if (rev > 0x7FFFFFFF)
11452 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11454 /* Append native character for the rev point */
11455 tmpend = uvchr_to_utf8(tmpbuf, rev);
11456 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11457 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11459 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11465 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11469 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11476 Perl_keyword_plugin_standard(pTHX_
11477 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11479 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11480 PERL_UNUSED_CONTEXT;
11481 PERL_UNUSED_ARG(keyword_ptr);
11482 PERL_UNUSED_ARG(keyword_len);
11483 PERL_UNUSED_ARG(op_ptr);
11484 return KEYWORD_PLUGIN_DECLINE;
11487 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11489 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11491 SAVEI32(PL_lex_brackets);
11492 if (PL_lex_brackets > 100)
11493 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11494 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11495 SAVEI32(PL_lex_allbrackets);
11496 PL_lex_allbrackets = 0;
11497 SAVEI8(PL_lex_fakeeof);
11498 PL_lex_fakeeof = (U8)fakeeof;
11499 if(yyparse(gramtype) && !PL_parser->error_count)
11500 qerror(Perl_mess(aTHX_ "Parse error"));
11503 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11505 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11509 SAVEVPTR(PL_eval_root);
11510 PL_eval_root = NULL;
11511 parse_recdescent(gramtype, fakeeof);
11517 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11519 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11522 if (flags & ~PARSE_OPTIONAL)
11523 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11524 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11525 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11526 if (!PL_parser->error_count)
11527 qerror(Perl_mess(aTHX_ "Parse error"));
11528 exprop = newOP(OP_NULL, 0);
11534 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11536 Parse a Perl arithmetic expression. This may contain operators of precedence
11537 down to the bit shift operators. The expression must be followed (and thus
11538 terminated) either by a comparison or lower-precedence operator or by
11539 something that would normally terminate an expression such as semicolon.
11540 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11541 otherwise it is mandatory. It is up to the caller to ensure that the
11542 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11543 the source of the code to be parsed and the lexical context for the
11546 The op tree representing the expression is returned. If an optional
11547 expression is absent, a null pointer is returned, otherwise the pointer
11550 If an error occurs in parsing or compilation, in most cases a valid op
11551 tree is returned anyway. The error is reflected in the parser state,
11552 normally resulting in a single exception at the top level of parsing
11553 which covers all the compilation errors that occurred. Some compilation
11554 errors, however, will throw an exception immediately.
11560 Perl_parse_arithexpr(pTHX_ U32 flags)
11562 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11566 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11568 Parse a Perl term expression. This may contain operators of precedence
11569 down to the assignment operators. The expression must be followed (and thus
11570 terminated) either by a comma or lower-precedence operator or by
11571 something that would normally terminate an expression such as semicolon.
11572 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11573 otherwise it is mandatory. It is up to the caller to ensure that the
11574 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11575 the source of the code to be parsed and the lexical context for the
11578 The op tree representing the expression is returned. If an optional
11579 expression is absent, a null pointer is returned, otherwise the pointer
11582 If an error occurs in parsing or compilation, in most cases a valid op
11583 tree is returned anyway. The error is reflected in the parser state,
11584 normally resulting in a single exception at the top level of parsing
11585 which covers all the compilation errors that occurred. Some compilation
11586 errors, however, will throw an exception immediately.
11592 Perl_parse_termexpr(pTHX_ U32 flags)
11594 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11598 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11600 Parse a Perl list expression. This may contain operators of precedence
11601 down to the comma operator. The expression must be followed (and thus
11602 terminated) either by a low-precedence logic operator such as C<or> or by
11603 something that would normally terminate an expression such as semicolon.
11604 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11605 otherwise it is mandatory. It is up to the caller to ensure that the
11606 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11607 the source of the code to be parsed and the lexical context for the
11610 The op tree representing the expression is returned. If an optional
11611 expression is absent, a null pointer is returned, otherwise the pointer
11614 If an error occurs in parsing or compilation, in most cases a valid op
11615 tree is returned anyway. The error is reflected in the parser state,
11616 normally resulting in a single exception at the top level of parsing
11617 which covers all the compilation errors that occurred. Some compilation
11618 errors, however, will throw an exception immediately.
11624 Perl_parse_listexpr(pTHX_ U32 flags)
11626 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11630 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11632 Parse a single complete Perl expression. This allows the full
11633 expression grammar, including the lowest-precedence operators such
11634 as C<or>. The expression must be followed (and thus terminated) by a
11635 token that an expression would normally be terminated by: end-of-file,
11636 closing bracketing punctuation, semicolon, or one of the keywords that
11637 signals a postfix expression-statement modifier. If I<flags> includes
11638 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11639 mandatory. It is up to the caller to ensure that the dynamic parser
11640 state (L</PL_parser> et al) is correctly set to reflect the source of
11641 the code to be parsed and the lexical context for the expression.
11643 The op tree representing the expression is returned. If an optional
11644 expression is absent, a null pointer is returned, otherwise the pointer
11647 If an error occurs in parsing or compilation, in most cases a valid op
11648 tree is returned anyway. The error is reflected in the parser state,
11649 normally resulting in a single exception at the top level of parsing
11650 which covers all the compilation errors that occurred. Some compilation
11651 errors, however, will throw an exception immediately.
11657 Perl_parse_fullexpr(pTHX_ U32 flags)
11659 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11663 =for apidoc Amx|OP *|parse_block|U32 flags
11665 Parse a single complete Perl code block. This consists of an opening
11666 brace, a sequence of statements, and a closing brace. The block
11667 constitutes a lexical scope, so C<my> variables and various compile-time
11668 effects can be contained within it. It is up to the caller to ensure
11669 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11670 reflect the source of the code to be parsed and the lexical context for
11673 The op tree representing the code block is returned. This is always a
11674 real op, never a null pointer. It will normally be a C<lineseq> list,
11675 including C<nextstate> or equivalent ops. No ops to construct any kind
11676 of runtime scope are included by virtue of it being a block.
11678 If an error occurs in parsing or compilation, in most cases a valid op
11679 tree (most likely null) is returned anyway. The error is reflected in
11680 the parser state, normally resulting in a single exception at the top
11681 level of parsing which covers all the compilation errors that occurred.
11682 Some compilation errors, however, will throw an exception immediately.
11684 The I<flags> parameter is reserved for future use, and must always
11691 Perl_parse_block(pTHX_ U32 flags)
11694 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11695 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11699 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11701 Parse a single unadorned Perl statement. This may be a normal imperative
11702 statement or a declaration that has compile-time effect. It does not
11703 include any label or other affixture. It is up to the caller to ensure
11704 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11705 reflect the source of the code to be parsed and the lexical context for
11708 The op tree representing the statement is returned. This may be a
11709 null pointer if the statement is null, for example if it was actually
11710 a subroutine definition (which has compile-time side effects). If not
11711 null, it will be ops directly implementing the statement, suitable to
11712 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11713 equivalent op (except for those embedded in a scope contained entirely
11714 within the statement).
11716 If an error occurs in parsing or compilation, in most cases a valid op
11717 tree (most likely null) is returned anyway. The error is reflected in
11718 the parser state, normally resulting in a single exception at the top
11719 level of parsing which covers all the compilation errors that occurred.
11720 Some compilation errors, however, will throw an exception immediately.
11722 The I<flags> parameter is reserved for future use, and must always
11729 Perl_parse_barestmt(pTHX_ U32 flags)
11732 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11733 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11737 =for apidoc Amx|SV *|parse_label|U32 flags
11739 Parse a single label, possibly optional, of the type that may prefix a
11740 Perl statement. It is up to the caller to ensure that the dynamic parser
11741 state (L</PL_parser> et al) is correctly set to reflect the source of
11742 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11743 label is optional, otherwise it is mandatory.
11745 The name of the label is returned in the form of a fresh scalar. If an
11746 optional label is absent, a null pointer is returned.
11748 If an error occurs in parsing, which can only occur if the label is
11749 mandatory, a valid label is returned anyway. The error is reflected in
11750 the parser state, normally resulting in a single exception at the top
11751 level of parsing which covers all the compilation errors that occurred.
11757 Perl_parse_label(pTHX_ U32 flags)
11759 if (flags & ~PARSE_OPTIONAL)
11760 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11761 if (PL_lex_state == LEX_KNOWNEXT) {
11762 PL_parser->yychar = yylex();
11763 if (PL_parser->yychar == LABEL) {
11765 PL_parser->yychar = YYEMPTY;
11766 lsv = newSV_type(SVt_PV);
11767 sv_copypv(lsv, cSVOPx(pl_yylval.opval)->op_sv);
11775 STRLEN wlen, bufptr_pos;
11778 if (!isIDFIRST_lazy_if(s, UTF))
11780 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11781 if (word_takes_any_delimeter(s, wlen))
11783 bufptr_pos = s - SvPVX(PL_linestr);
11785 lex_read_space(LEX_KEEP_PREVIOUS);
11787 s = SvPVX(PL_linestr) + bufptr_pos;
11788 if (t[0] == ':' && t[1] != ':') {
11789 PL_oldoldbufptr = PL_oldbufptr;
11792 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11796 if (flags & PARSE_OPTIONAL) {
11799 qerror(Perl_mess(aTHX_ "Parse error"));
11800 return newSVpvs("x");
11807 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11809 Parse a single complete Perl statement. This may be a normal imperative
11810 statement or a declaration that has compile-time effect, and may include
11811 optional labels. It is up to the caller to ensure that the dynamic
11812 parser state (L</PL_parser> et al) is correctly set to reflect the source
11813 of the code to be parsed and the lexical context for the statement.
11815 The op tree representing the statement is returned. This may be a
11816 null pointer if the statement is null, for example if it was actually
11817 a subroutine definition (which has compile-time side effects). If not
11818 null, it will be the result of a L</newSTATEOP> call, normally including
11819 a C<nextstate> or equivalent op.
11821 If an error occurs in parsing or compilation, in most cases a valid op
11822 tree (most likely null) is returned anyway. The error is reflected in
11823 the parser state, normally resulting in a single exception at the top
11824 level of parsing which covers all the compilation errors that occurred.
11825 Some compilation errors, however, will throw an exception immediately.
11827 The I<flags> parameter is reserved for future use, and must always
11834 Perl_parse_fullstmt(pTHX_ U32 flags)
11837 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11838 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11842 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11844 Parse a sequence of zero or more Perl statements. These may be normal
11845 imperative statements, including optional labels, or declarations
11846 that have compile-time effect, or any mixture thereof. The statement
11847 sequence ends when a closing brace or end-of-file is encountered in a
11848 place where a new statement could have validly started. It is up to
11849 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11850 is correctly set to reflect the source of the code to be parsed and the
11851 lexical context for the statements.
11853 The op tree representing the statement sequence is returned. This may
11854 be a null pointer if the statements were all null, for example if there
11855 were no statements or if there were only subroutine definitions (which
11856 have compile-time side effects). If not null, it will be a C<lineseq>
11857 list, normally including C<nextstate> or equivalent ops.
11859 If an error occurs in parsing or compilation, in most cases a valid op
11860 tree is returned anyway. The error is reflected in the parser state,
11861 normally resulting in a single exception at the top level of parsing
11862 which covers all the compilation errors that occurred. Some compilation
11863 errors, however, will throw an exception immediately.
11865 The I<flags> parameter is reserved for future use, and must always
11872 Perl_parse_stmtseq(pTHX_ U32 flags)
11877 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11878 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11879 c = lex_peek_unichar(0);
11880 if (c != -1 && c != /*{*/'}')
11881 qerror(Perl_mess(aTHX_ "Parse error"));
11887 * c-indentation-style: bsd
11888 * c-basic-offset: 4
11889 * indent-tabs-mode: nil
11892 * ex: set ts=8 sts=4 sw=4 et: