3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
42 #include "dquote_static.c"
44 #define new_constant(a,b,c,d,e,f,g) \
45 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
47 #define pl_yylval (PL_parser->yylval)
49 /* XXX temporary backwards compatibility */
50 #define PL_lex_brackets (PL_parser->lex_brackets)
51 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
52 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_preambled (PL_parser->preambled)
70 #define PL_sublex_info (PL_parser->sublex_info)
71 #define PL_linestr (PL_parser->linestr)
72 #define PL_expect (PL_parser->expect)
73 #define PL_copline (PL_parser->copline)
74 #define PL_bufptr (PL_parser->bufptr)
75 #define PL_oldbufptr (PL_parser->oldbufptr)
76 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
77 #define PL_linestart (PL_parser->linestart)
78 #define PL_bufend (PL_parser->bufend)
79 #define PL_last_uni (PL_parser->last_uni)
80 #define PL_last_lop (PL_parser->last_lop)
81 #define PL_last_lop_op (PL_parser->last_lop_op)
82 #define PL_lex_state (PL_parser->lex_state)
83 #define PL_rsfp (PL_parser->rsfp)
84 #define PL_rsfp_filters (PL_parser->rsfp_filters)
85 #define PL_in_my (PL_parser->in_my)
86 #define PL_in_my_stash (PL_parser->in_my_stash)
87 #define PL_tokenbuf (PL_parser->tokenbuf)
88 #define PL_multi_end (PL_parser->multi_end)
89 #define PL_error_count (PL_parser->error_count)
92 # define PL_endwhite (PL_parser->endwhite)
93 # define PL_faketokens (PL_parser->faketokens)
94 # define PL_lasttoke (PL_parser->lasttoke)
95 # define PL_nextwhite (PL_parser->nextwhite)
96 # define PL_realtokenstart (PL_parser->realtokenstart)
97 # define PL_skipwhite (PL_parser->skipwhite)
98 # define PL_thisclose (PL_parser->thisclose)
99 # define PL_thismad (PL_parser->thismad)
100 # define PL_thisopen (PL_parser->thisopen)
101 # define PL_thisstuff (PL_parser->thisstuff)
102 # define PL_thistoken (PL_parser->thistoken)
103 # define PL_thiswhite (PL_parser->thiswhite)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_nexttoke (PL_parser->nexttoke)
106 # define PL_curforce (PL_parser->curforce)
108 # define PL_nexttoke (PL_parser->nexttoke)
109 # define PL_nexttype (PL_parser->nexttype)
110 # define PL_nextval (PL_parser->nextval)
113 static const char* const ident_too_long = "Identifier too long";
116 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
117 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
119 # define CURMAD(slot,sv)
120 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
123 #define XENUMMASK 0x3f
124 #define XFAKEEOF 0x40
125 #define XFAKEBRACK 0x80
127 #ifdef USE_UTF8_SCRIPTS
128 # define UTF (!IN_BYTES)
130 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
133 /* The maximum number of characters preceding the unrecognized one to display */
134 #define UNRECOGNIZED_PRECEDE_COUNT 10
136 /* In variables named $^X, these are the legal values for X.
137 * 1999-02-27 mjd-perl-patch@plover.com */
138 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
140 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
142 /* LEX_* are values for PL_lex_state, the state of the lexer.
143 * They are arranged oddly so that the guard on the switch statement
144 * can get by with a single comparison (if the compiler is smart enough).
146 * These values refer to the various states within a sublex parse,
147 * i.e. within a double quotish string
150 /* #define LEX_NOTPARSING 11 is done in perl.h. */
152 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
153 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
154 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
155 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
156 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
158 /* at end of code, eg "$x" followed by: */
159 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
160 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
162 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
163 string or after \E, $foo, etc */
164 #define LEX_INTERPCONST 2 /* NOT USED */
165 #define LEX_FORMLINE 1 /* expecting a format line */
166 #define LEX_KNOWNEXT 0 /* next token known; just return it */
170 static const char* const lex_state_names[] = {
189 #include "keywords.h"
191 /* CLINE is a macro that ensures PL_copline has a sane value */
196 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
199 # define SKIPSPACE0(s) skipspace0(s)
200 # define SKIPSPACE1(s) skipspace1(s)
201 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
202 # define PEEKSPACE(s) skipspace2(s,0)
204 # define SKIPSPACE0(s) skipspace(s)
205 # define SKIPSPACE1(s) skipspace(s)
206 # define SKIPSPACE2(s,tsv) skipspace(s)
207 # define PEEKSPACE(s) skipspace(s)
211 * Convenience functions to return different tokens and prime the
212 * lexer for the next token. They all take an argument.
214 * TOKEN : generic token (used for '(', DOLSHARP, etc)
215 * OPERATOR : generic operator
216 * AOPERATOR : assignment operator
217 * PREBLOCK : beginning the block after an if, while, foreach, ...
218 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
219 * PREREF : *EXPR where EXPR is not a simple identifier
220 * TERM : expression term
221 * LOOPX : loop exiting command (goto, last, dump, etc)
222 * FTST : file test operator
223 * FUN0 : zero-argument function
224 * FUN0OP : zero-argument function, with its op created in this file
225 * FUN1 : not used, except for not, which isn't a UNIOP
226 * BOop : bitwise or or xor
228 * SHop : shift operator
229 * PWop : power operator
230 * PMop : pattern-matching operator
231 * Aop : addition-level operator
232 * Mop : multiplication-level operator
233 * Eop : equality-testing operator
234 * Rop : relational operator <= != gt
236 * Also see LOP and lop() below.
239 #ifdef DEBUGGING /* Serve -DT. */
240 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
242 # define REPORT(retval) (retval)
245 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
246 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
247 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
248 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
249 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
251 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
252 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
253 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
254 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
255 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI3(f,x,have_x) { \
273 pl_yylval.ival = f; \
274 if (have_x) PL_expect = x; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI3(f,XTERM,1)
284 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
285 #define UNIPROTO(f,optional) { \
286 if (optional) PL_last_uni = PL_oldbufptr; \
290 #define UNIBRACK(f) UNI3(f,0,0)
292 /* grandfather return to old style */
295 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
296 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
297 pl_yylval.ival = (f); \
303 #define COPLINE_INC_WITH_HERELINES \
305 CopLINE_inc(PL_curcop); \
306 if (PL_parser->lex_shared->herelines) \
307 CopLINE(PL_curcop) += PL_parser->lex_shared->herelines, \
308 PL_parser->lex_shared->herelines = 0; \
314 /* how to interpret the pl_yylval associated with the token */
318 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
323 static struct debug_tokens {
325 enum token_type type;
327 } const debug_tokens[] =
329 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
330 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
331 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
332 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
333 { ARROW, TOKENTYPE_NONE, "ARROW" },
334 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
335 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
336 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
337 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
338 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
339 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
340 { DO, TOKENTYPE_NONE, "DO" },
341 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
342 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
343 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
344 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
345 { ELSE, TOKENTYPE_NONE, "ELSE" },
346 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
347 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
348 { FOR, TOKENTYPE_IVAL, "FOR" },
349 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
350 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
351 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
352 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
353 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
354 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
355 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
356 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
357 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
358 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
359 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
360 { IF, TOKENTYPE_IVAL, "IF" },
361 { LABEL, TOKENTYPE_PVAL, "LABEL" },
362 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
363 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
364 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
365 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
366 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
367 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
368 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
369 { MY, TOKENTYPE_IVAL, "MY" },
370 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
371 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
372 { OROP, TOKENTYPE_IVAL, "OROP" },
373 { OROR, TOKENTYPE_NONE, "OROR" },
374 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
375 { PEG, TOKENTYPE_NONE, "PEG" },
376 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
377 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
378 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
379 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
380 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
381 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
382 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
383 { PREINC, TOKENTYPE_NONE, "PREINC" },
384 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
385 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
386 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
387 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
388 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
389 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
390 { SUB, TOKENTYPE_NONE, "SUB" },
391 { THING, TOKENTYPE_OPVAL, "THING" },
392 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
393 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
394 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
395 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
396 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
397 { USE, TOKENTYPE_IVAL, "USE" },
398 { WHEN, TOKENTYPE_IVAL, "WHEN" },
399 { WHILE, TOKENTYPE_IVAL, "WHILE" },
400 { WORD, TOKENTYPE_OPVAL, "WORD" },
401 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
402 { 0, TOKENTYPE_NONE, NULL }
405 /* dump the returned token in rv, plus any optional arg in pl_yylval */
408 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
412 PERL_ARGS_ASSERT_TOKEREPORT;
415 const char *name = NULL;
416 enum token_type type = TOKENTYPE_NONE;
417 const struct debug_tokens *p;
418 SV* const report = newSVpvs("<== ");
420 for (p = debug_tokens; p->token; p++) {
421 if (p->token == (int)rv) {
428 Perl_sv_catpv(aTHX_ report, name);
429 else if ((char)rv > ' ' && (char)rv <= '~')
430 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
432 sv_catpvs(report, "EOF");
434 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
439 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
441 case TOKENTYPE_OPNUM:
442 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
443 PL_op_name[lvalp->ival]);
446 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
448 case TOKENTYPE_OPVAL:
450 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
451 PL_op_name[lvalp->opval->op_type]);
452 if (lvalp->opval->op_type == OP_CONST) {
453 Perl_sv_catpvf(aTHX_ report, " %s",
454 SvPEEK(cSVOPx_sv(lvalp->opval)));
459 sv_catpvs(report, "(opval=null)");
462 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
468 /* print the buffer with suitable escapes */
471 S_printbuf(pTHX_ const char *const fmt, const char *const s)
473 SV* const tmp = newSVpvs("");
475 PERL_ARGS_ASSERT_PRINTBUF;
477 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
484 S_deprecate_commaless_var_list(pTHX) {
486 deprecate("comma-less variable list");
487 return REPORT(','); /* grandfather non-comma-format format */
493 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
494 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
498 S_ao(pTHX_ int toketype)
501 if (*PL_bufptr == '=') {
503 if (toketype == ANDAND)
504 pl_yylval.ival = OP_ANDASSIGN;
505 else if (toketype == OROR)
506 pl_yylval.ival = OP_ORASSIGN;
507 else if (toketype == DORDOR)
508 pl_yylval.ival = OP_DORASSIGN;
516 * When Perl expects an operator and finds something else, no_op
517 * prints the warning. It always prints "<something> found where
518 * operator expected. It prints "Missing semicolon on previous line?"
519 * if the surprise occurs at the start of the line. "do you need to
520 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
521 * where the compiler doesn't know if foo is a method call or a function.
522 * It prints "Missing operator before end of line" if there's nothing
523 * after the missing operator, or "... before <...>" if there is something
524 * after the missing operator.
528 S_no_op(pTHX_ const char *const what, char *s)
531 char * const oldbp = PL_bufptr;
532 const bool is_first = (PL_oldbufptr == PL_linestart);
534 PERL_ARGS_ASSERT_NO_OP;
540 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
541 if (ckWARN_d(WARN_SYNTAX)) {
543 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
544 "\t(Missing semicolon on previous line?)\n");
545 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
547 for (t = PL_oldoldbufptr; (isWORDCHAR_lazy_if(t,UTF) || *t == ':');
548 t += UTF ? UTF8SKIP(t) : 1)
550 if (t < PL_bufptr && isSPACE(*t))
551 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
552 "\t(Do you need to predeclare %"SVf"?)\n",
553 SVfARG(newSVpvn_flags(PL_oldoldbufptr, (STRLEN)(t - PL_oldoldbufptr),
554 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
558 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
559 "\t(Missing operator before %"SVf"?)\n",
560 SVfARG(newSVpvn_flags(oldbp, (STRLEN)(s - oldbp),
561 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
569 * Complain about missing quote/regexp/heredoc terminator.
570 * If it's called with NULL then it cauterizes the line buffer.
571 * If we're in a delimited string and the delimiter is a control
572 * character, it's reformatted into a two-char sequence like ^C.
577 S_missingterm(pTHX_ char *s)
583 char * const nl = strrchr(s,'\n');
587 else if (isCNTRL(PL_multi_close)) {
589 tmpbuf[1] = (char)toCTRL(PL_multi_close);
594 *tmpbuf = (char)PL_multi_close;
598 q = strchr(s,'"') ? '\'' : '"';
599 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
605 * Check whether the named feature is enabled.
608 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
611 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
613 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
615 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
617 if (namelen > MAX_FEATURE_LEN)
619 memcpy(&he_name[8], name, namelen);
621 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
622 REFCOUNTED_HE_EXISTS));
626 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
627 * utf16-to-utf8-reversed.
630 #ifdef PERL_CR_FILTER
634 const char *s = SvPVX_const(sv);
635 const char * const e = s + SvCUR(sv);
637 PERL_ARGS_ASSERT_STRIP_RETURN;
639 /* outer loop optimized to do nothing if there are no CR-LFs */
641 if (*s++ == '\r' && *s == '\n') {
642 /* hit a CR-LF, need to copy the rest */
646 if (*s == '\r' && s[1] == '\n')
657 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
659 const I32 count = FILTER_READ(idx+1, sv, maxlen);
660 if (count > 0 && !maxlen)
667 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
669 Creates and initialises a new lexer/parser state object, supplying
670 a context in which to lex and parse from a new source of Perl code.
671 A pointer to the new state object is placed in L</PL_parser>. An entry
672 is made on the save stack so that upon unwinding the new state object
673 will be destroyed and the former value of L</PL_parser> will be restored.
674 Nothing else need be done to clean up the parsing context.
676 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
677 non-null, provides a string (in SV form) containing code to be parsed.
678 A copy of the string is made, so subsequent modification of I<line>
679 does not affect parsing. I<rsfp>, if non-null, provides an input stream
680 from which code will be read to be parsed. If both are non-null, the
681 code in I<line> comes first and must consist of complete lines of input,
682 and I<rsfp> supplies the remainder of the source.
684 The I<flags> parameter is reserved for future use. Currently it is only
685 used by perl internally, so extensions should always pass zero.
690 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
691 can share filters with the current parser.
692 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
693 caller, hence isn't owned by the parser, so shouldn't be closed on parser
694 destruction. This is used to handle the case of defaulting to reading the
695 script from the standard input because no filename was given on the command
696 line (without getting confused by situation where STDIN has been closed, so
697 the script handle is opened on fd 0) */
700 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
703 const char *s = NULL;
704 yy_parser *parser, *oparser;
705 if (flags && flags & ~LEX_START_FLAGS)
706 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
708 /* create and initialise a parser */
710 Newxz(parser, 1, yy_parser);
711 parser->old_parser = oparser = PL_parser;
714 parser->stack = NULL;
716 parser->stack_size = 0;
718 /* on scope exit, free this parser and restore any outer one */
720 parser->saved_curcop = PL_curcop;
722 /* initialise lexer state */
725 parser->curforce = -1;
727 parser->nexttoke = 0;
729 parser->error_count = oparser ? oparser->error_count : 0;
730 parser->copline = NOLINE;
731 parser->lex_state = LEX_NORMAL;
732 parser->expect = XSTATE;
734 parser->rsfp_filters =
735 !(flags & LEX_START_SAME_FILTER) || !oparser
737 : MUTABLE_AV(SvREFCNT_inc(
738 oparser->rsfp_filters
739 ? oparser->rsfp_filters
740 : (oparser->rsfp_filters = newAV())
743 Newx(parser->lex_brackstack, 120, char);
744 Newx(parser->lex_casestack, 12, char);
745 *parser->lex_casestack = '\0';
746 Newxz(parser->lex_shared, 1, LEXSHARED);
750 s = SvPV_const(line, len);
751 parser->linestr = flags & LEX_START_COPIED
752 ? SvREFCNT_inc_simple_NN(line)
753 : newSVpvn_flags(s, len, SvUTF8(line));
754 sv_catpvs(parser->linestr, "\n;");
756 parser->linestr = newSVpvs("\n;");
758 parser->oldoldbufptr =
761 parser->linestart = SvPVX(parser->linestr);
762 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
763 parser->last_lop = parser->last_uni = NULL;
764 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
765 |LEX_DONT_CLOSE_RSFP);
767 parser->in_pod = parser->filtered = 0;
771 /* delete a parser object */
774 Perl_parser_free(pTHX_ const yy_parser *parser)
776 PERL_ARGS_ASSERT_PARSER_FREE;
778 PL_curcop = parser->saved_curcop;
779 SvREFCNT_dec(parser->linestr);
781 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
782 PerlIO_clearerr(parser->rsfp);
783 else if (parser->rsfp && (!parser->old_parser ||
784 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
785 PerlIO_close(parser->rsfp);
786 SvREFCNT_dec(parser->rsfp_filters);
787 SvREFCNT_dec(parser->lex_stuff);
788 SvREFCNT_dec(parser->sublex_info.repl);
790 Safefree(parser->lex_brackstack);
791 Safefree(parser->lex_casestack);
792 Safefree(parser->lex_shared);
793 PL_parser = parser->old_parser;
798 Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
801 I32 nexttoke = parser->lasttoke;
803 I32 nexttoke = parser->nexttoke;
805 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
808 if (S_is_opval_token(parser->nexttoke[nexttoke].next_type
810 && parser->nexttoke[nexttoke].next_val.opval
811 && parser->nexttoke[nexttoke].next_val.opval->op_slabbed
812 && OpSLAB(parser->nexttoke[nexttoke].next_val.opval) == slab) {
813 op_free(parser->nexttoke[nexttoke].next_val.opval);
814 parser->nexttoke[nexttoke].next_val.opval = NULL;
817 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
818 && parser->nextval[nexttoke].opval
819 && parser->nextval[nexttoke].opval->op_slabbed
820 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
821 op_free(parser->nextval[nexttoke].opval);
822 parser->nextval[nexttoke].opval = NULL;
830 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
832 Buffer scalar containing the chunk currently under consideration of the
833 text currently being lexed. This is always a plain string scalar (for
834 which C<SvPOK> is true). It is not intended to be used as a scalar by
835 normal scalar means; instead refer to the buffer directly by the pointer
836 variables described below.
838 The lexer maintains various C<char*> pointers to things in the
839 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
840 reallocated, all of these pointers must be updated. Don't attempt to
841 do this manually, but rather use L</lex_grow_linestr> if you need to
842 reallocate the buffer.
844 The content of the text chunk in the buffer is commonly exactly one
845 complete line of input, up to and including a newline terminator,
846 but there are situations where it is otherwise. The octets of the
847 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
848 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
849 flag on this scalar, which may disagree with it.
851 For direct examination of the buffer, the variable
852 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
853 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
854 of these pointers is usually preferable to examination of the scalar
855 through normal scalar means.
857 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
859 Direct pointer to the end of the chunk of text currently being lexed, the
860 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
861 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
862 always located at the end of the buffer, and does not count as part of
863 the buffer's contents.
865 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
867 Points to the current position of lexing inside the lexer buffer.
868 Characters around this point may be freely examined, within
869 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
870 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
871 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
873 Lexing code (whether in the Perl core or not) moves this pointer past
874 the characters that it consumes. It is also expected to perform some
875 bookkeeping whenever a newline character is consumed. This movement
876 can be more conveniently performed by the function L</lex_read_to>,
877 which handles newlines appropriately.
879 Interpretation of the buffer's octets can be abstracted out by
880 using the slightly higher-level functions L</lex_peek_unichar> and
881 L</lex_read_unichar>.
883 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
885 Points to the start of the current line inside the lexer buffer.
886 This is useful for indicating at which column an error occurred, and
887 not much else. This must be updated by any lexing code that consumes
888 a newline; the function L</lex_read_to> handles this detail.
894 =for apidoc Amx|bool|lex_bufutf8
896 Indicates whether the octets in the lexer buffer
897 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
898 of Unicode characters. If not, they should be interpreted as Latin-1
899 characters. This is analogous to the C<SvUTF8> flag for scalars.
901 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
902 contains valid UTF-8. Lexing code must be robust in the face of invalid
905 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
906 is significant, but not the whole story regarding the input character
907 encoding. Normally, when a file is being read, the scalar contains octets
908 and its C<SvUTF8> flag is off, but the octets should be interpreted as
909 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
910 however, the scalar may have the C<SvUTF8> flag on, and in this case its
911 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
912 is in effect. This logic may change in the future; use this function
913 instead of implementing the logic yourself.
919 Perl_lex_bufutf8(pTHX)
925 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
927 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
928 at least I<len> octets (including terminating NUL). Returns a
929 pointer to the reallocated buffer. This is necessary before making
930 any direct modification of the buffer that would increase its length.
931 L</lex_stuff_pvn> provides a more convenient way to insert text into
934 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
935 this function updates all of the lexer's variables that point directly
942 Perl_lex_grow_linestr(pTHX_ STRLEN len)
946 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
947 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
948 linestr = PL_parser->linestr;
949 buf = SvPVX(linestr);
950 if (len <= SvLEN(linestr))
952 bufend_pos = PL_parser->bufend - buf;
953 bufptr_pos = PL_parser->bufptr - buf;
954 oldbufptr_pos = PL_parser->oldbufptr - buf;
955 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
956 linestart_pos = PL_parser->linestart - buf;
957 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
958 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
959 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
960 PL_parser->lex_shared->re_eval_start - buf : 0;
962 buf = sv_grow(linestr, len);
964 PL_parser->bufend = buf + bufend_pos;
965 PL_parser->bufptr = buf + bufptr_pos;
966 PL_parser->oldbufptr = buf + oldbufptr_pos;
967 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
968 PL_parser->linestart = buf + linestart_pos;
969 if (PL_parser->last_uni)
970 PL_parser->last_uni = buf + last_uni_pos;
971 if (PL_parser->last_lop)
972 PL_parser->last_lop = buf + last_lop_pos;
973 if (PL_parser->lex_shared->re_eval_start)
974 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
979 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
981 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
982 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
983 reallocating the buffer if necessary. This means that lexing code that
984 runs later will see the characters as if they had appeared in the input.
985 It is not recommended to do this as part of normal parsing, and most
986 uses of this facility run the risk of the inserted characters being
987 interpreted in an unintended manner.
989 The string to be inserted is represented by I<len> octets starting
990 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
991 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
992 The characters are recoded for the lexer buffer, according to how the
993 buffer is currently being interpreted (L</lex_bufutf8>). If a string
994 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
995 function is more convenient.
1001 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
1005 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
1006 if (flags & ~(LEX_STUFF_UTF8))
1007 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
1009 if (flags & LEX_STUFF_UTF8) {
1012 STRLEN highhalf = 0; /* Count of variants */
1013 const char *p, *e = pv+len;
1014 for (p = pv; p != e; p++) {
1015 if (! UTF8_IS_INVARIANT(*p)) {
1021 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1022 bufptr = PL_parser->bufptr;
1023 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
1024 SvCUR_set(PL_parser->linestr,
1025 SvCUR(PL_parser->linestr) + len+highhalf);
1026 PL_parser->bufend += len+highhalf;
1027 for (p = pv; p != e; p++) {
1029 if (! UTF8_IS_INVARIANT(c)) {
1030 *bufptr++ = UTF8_TWO_BYTE_HI(c);
1031 *bufptr++ = UTF8_TWO_BYTE_LO(c);
1033 *bufptr++ = (char)c;
1038 if (flags & LEX_STUFF_UTF8) {
1039 STRLEN highhalf = 0;
1040 const char *p, *e = pv+len;
1041 for (p = pv; p != e; p++) {
1043 if (UTF8_IS_ABOVE_LATIN1(c)) {
1044 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1045 "non-Latin-1 character into Latin-1 input");
1046 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
1049 } else if (! UTF8_IS_INVARIANT(c)) {
1050 /* malformed UTF-8 */
1052 SAVESPTR(PL_warnhook);
1053 PL_warnhook = PERL_WARNHOOK_FATAL;
1054 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1060 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1061 bufptr = PL_parser->bufptr;
1062 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1063 SvCUR_set(PL_parser->linestr,
1064 SvCUR(PL_parser->linestr) + len-highhalf);
1065 PL_parser->bufend += len-highhalf;
1068 if (UTF8_IS_INVARIANT(*p)) {
1074 *bufptr++ = TWO_BYTE_UTF8_TO_UNI(*p, *(p+1));
1080 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1081 bufptr = PL_parser->bufptr;
1082 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1083 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1084 PL_parser->bufend += len;
1085 Copy(pv, bufptr, len, char);
1091 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1093 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1094 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1095 reallocating the buffer if necessary. This means that lexing code that
1096 runs later will see the characters as if they had appeared in the input.
1097 It is not recommended to do this as part of normal parsing, and most
1098 uses of this facility run the risk of the inserted characters being
1099 interpreted in an unintended manner.
1101 The string to be inserted is represented by octets starting at I<pv>
1102 and continuing to the first nul. These octets are interpreted as either
1103 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1104 in I<flags>. The characters are recoded for the lexer buffer, according
1105 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1106 If it is not convenient to nul-terminate a string to be inserted, the
1107 L</lex_stuff_pvn> function is more appropriate.
1113 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1115 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1116 lex_stuff_pvn(pv, strlen(pv), flags);
1120 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1122 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1123 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1124 reallocating the buffer if necessary. This means that lexing code that
1125 runs later will see the characters as if they had appeared in the input.
1126 It is not recommended to do this as part of normal parsing, and most
1127 uses of this facility run the risk of the inserted characters being
1128 interpreted in an unintended manner.
1130 The string to be inserted is the string value of I<sv>. The characters
1131 are recoded for the lexer buffer, according to how the buffer is currently
1132 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1133 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1134 need to construct a scalar.
1140 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1144 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1148 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1152 =for apidoc Amx|void|lex_unstuff|char *ptr
1154 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1155 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1156 This hides the discarded text from any lexing code that runs later,
1157 as if the text had never appeared.
1159 This is not the normal way to consume lexed text. For that, use
1166 Perl_lex_unstuff(pTHX_ char *ptr)
1170 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1171 buf = PL_parser->bufptr;
1173 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1176 bufend = PL_parser->bufend;
1178 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1179 unstuff_len = ptr - buf;
1180 Move(ptr, buf, bufend+1-ptr, char);
1181 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1182 PL_parser->bufend = bufend - unstuff_len;
1186 =for apidoc Amx|void|lex_read_to|char *ptr
1188 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1189 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1190 performing the correct bookkeeping whenever a newline character is passed.
1191 This is the normal way to consume lexed text.
1193 Interpretation of the buffer's octets can be abstracted out by
1194 using the slightly higher-level functions L</lex_peek_unichar> and
1195 L</lex_read_unichar>.
1201 Perl_lex_read_to(pTHX_ char *ptr)
1204 PERL_ARGS_ASSERT_LEX_READ_TO;
1205 s = PL_parser->bufptr;
1206 if (ptr < s || ptr > PL_parser->bufend)
1207 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1208 for (; s != ptr; s++)
1210 COPLINE_INC_WITH_HERELINES;
1211 PL_parser->linestart = s+1;
1213 PL_parser->bufptr = ptr;
1217 =for apidoc Amx|void|lex_discard_to|char *ptr
1219 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1220 up to I<ptr>. The remaining content of the buffer will be moved, and
1221 all pointers into the buffer updated appropriately. I<ptr> must not
1222 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1223 it is not permitted to discard text that has yet to be lexed.
1225 Normally it is not necessarily to do this directly, because it suffices to
1226 use the implicit discarding behaviour of L</lex_next_chunk> and things
1227 based on it. However, if a token stretches across multiple lines,
1228 and the lexing code has kept multiple lines of text in the buffer for
1229 that purpose, then after completion of the token it would be wise to
1230 explicitly discard the now-unneeded earlier lines, to avoid future
1231 multi-line tokens growing the buffer without bound.
1237 Perl_lex_discard_to(pTHX_ char *ptr)
1241 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1242 buf = SvPVX(PL_parser->linestr);
1244 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1247 if (ptr > PL_parser->bufptr)
1248 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1249 discard_len = ptr - buf;
1250 if (PL_parser->oldbufptr < ptr)
1251 PL_parser->oldbufptr = ptr;
1252 if (PL_parser->oldoldbufptr < ptr)
1253 PL_parser->oldoldbufptr = ptr;
1254 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1255 PL_parser->last_uni = NULL;
1256 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1257 PL_parser->last_lop = NULL;
1258 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1259 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1260 PL_parser->bufend -= discard_len;
1261 PL_parser->bufptr -= discard_len;
1262 PL_parser->oldbufptr -= discard_len;
1263 PL_parser->oldoldbufptr -= discard_len;
1264 if (PL_parser->last_uni)
1265 PL_parser->last_uni -= discard_len;
1266 if (PL_parser->last_lop)
1267 PL_parser->last_lop -= discard_len;
1271 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1273 Reads in the next chunk of text to be lexed, appending it to
1274 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1275 looked to the end of the current chunk and wants to know more. It is
1276 usual, but not necessary, for lexing to have consumed the entirety of
1277 the current chunk at this time.
1279 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1280 chunk (i.e., the current chunk has been entirely consumed), normally the
1281 current chunk will be discarded at the same time that the new chunk is
1282 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1283 will not be discarded. If the current chunk has not been entirely
1284 consumed, then it will not be discarded regardless of the flag.
1286 Returns true if some new text was added to the buffer, or false if the
1287 buffer has reached the end of the input text.
1292 #define LEX_FAKE_EOF 0x80000000
1293 #define LEX_NO_TERM 0x40000000
1296 Perl_lex_next_chunk(pTHX_ U32 flags)
1300 STRLEN old_bufend_pos, new_bufend_pos;
1301 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1302 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1303 bool got_some_for_debugger = 0;
1305 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1306 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1307 linestr = PL_parser->linestr;
1308 buf = SvPVX(linestr);
1309 if (!(flags & LEX_KEEP_PREVIOUS) &&
1310 PL_parser->bufptr == PL_parser->bufend) {
1311 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1313 if (PL_parser->last_uni != PL_parser->bufend)
1314 PL_parser->last_uni = NULL;
1315 if (PL_parser->last_lop != PL_parser->bufend)
1316 PL_parser->last_lop = NULL;
1317 last_uni_pos = last_lop_pos = 0;
1321 old_bufend_pos = PL_parser->bufend - buf;
1322 bufptr_pos = PL_parser->bufptr - buf;
1323 oldbufptr_pos = PL_parser->oldbufptr - buf;
1324 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1325 linestart_pos = PL_parser->linestart - buf;
1326 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1327 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1329 if (flags & LEX_FAKE_EOF) {
1331 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1333 } else if (filter_gets(linestr, old_bufend_pos)) {
1335 got_some_for_debugger = 1;
1336 } else if (flags & LEX_NO_TERM) {
1339 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1340 sv_setpvs(linestr, "");
1342 /* End of real input. Close filehandle (unless it was STDIN),
1343 * then add implicit termination.
1345 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1346 PerlIO_clearerr(PL_parser->rsfp);
1347 else if (PL_parser->rsfp)
1348 (void)PerlIO_close(PL_parser->rsfp);
1349 PL_parser->rsfp = NULL;
1350 PL_parser->in_pod = PL_parser->filtered = 0;
1352 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1355 if (!PL_in_eval && PL_minus_p) {
1357 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1358 PL_minus_n = PL_minus_p = 0;
1359 } else if (!PL_in_eval && PL_minus_n) {
1360 sv_catpvs(linestr, /*{*/";}");
1363 sv_catpvs(linestr, ";");
1366 buf = SvPVX(linestr);
1367 new_bufend_pos = SvCUR(linestr);
1368 PL_parser->bufend = buf + new_bufend_pos;
1369 PL_parser->bufptr = buf + bufptr_pos;
1370 PL_parser->oldbufptr = buf + oldbufptr_pos;
1371 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1372 PL_parser->linestart = buf + linestart_pos;
1373 if (PL_parser->last_uni)
1374 PL_parser->last_uni = buf + last_uni_pos;
1375 if (PL_parser->last_lop)
1376 PL_parser->last_lop = buf + last_lop_pos;
1377 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1378 PL_curstash != PL_debstash) {
1379 /* debugger active and we're not compiling the debugger code,
1380 * so store the line into the debugger's array of lines
1382 update_debugger_info(NULL, buf+old_bufend_pos,
1383 new_bufend_pos-old_bufend_pos);
1389 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1391 Looks ahead one (Unicode) character in the text currently being lexed.
1392 Returns the codepoint (unsigned integer value) of the next character,
1393 or -1 if lexing has reached the end of the input text. To consume the
1394 peeked character, use L</lex_read_unichar>.
1396 If the next character is in (or extends into) the next chunk of input
1397 text, the next chunk will be read in. Normally the current chunk will be
1398 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1399 then the current chunk will not be discarded.
1401 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1402 is encountered, an exception is generated.
1408 Perl_lex_peek_unichar(pTHX_ U32 flags)
1412 if (flags & ~(LEX_KEEP_PREVIOUS))
1413 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1414 s = PL_parser->bufptr;
1415 bufend = PL_parser->bufend;
1421 if (!lex_next_chunk(flags))
1423 s = PL_parser->bufptr;
1424 bufend = PL_parser->bufend;
1427 if (UTF8_IS_INVARIANT(head))
1429 if (UTF8_IS_START(head)) {
1430 len = UTF8SKIP(&head);
1431 while ((STRLEN)(bufend-s) < len) {
1432 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1434 s = PL_parser->bufptr;
1435 bufend = PL_parser->bufend;
1438 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1439 if (retlen == (STRLEN)-1) {
1440 /* malformed UTF-8 */
1442 SAVESPTR(PL_warnhook);
1443 PL_warnhook = PERL_WARNHOOK_FATAL;
1444 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1450 if (!lex_next_chunk(flags))
1452 s = PL_parser->bufptr;
1459 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1461 Reads the next (Unicode) character in the text currently being lexed.
1462 Returns the codepoint (unsigned integer value) of the character read,
1463 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1464 if lexing has reached the end of the input text. To non-destructively
1465 examine the next character, use L</lex_peek_unichar> instead.
1467 If the next character is in (or extends into) the next chunk of input
1468 text, the next chunk will be read in. Normally the current chunk will be
1469 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1470 then the current chunk will not be discarded.
1472 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1473 is encountered, an exception is generated.
1479 Perl_lex_read_unichar(pTHX_ U32 flags)
1482 if (flags & ~(LEX_KEEP_PREVIOUS))
1483 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1484 c = lex_peek_unichar(flags);
1487 COPLINE_INC_WITH_HERELINES;
1489 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1491 ++(PL_parser->bufptr);
1497 =for apidoc Amx|void|lex_read_space|U32 flags
1499 Reads optional spaces, in Perl style, in the text currently being
1500 lexed. The spaces may include ordinary whitespace characters and
1501 Perl-style comments. C<#line> directives are processed if encountered.
1502 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1503 at a non-space character (or the end of the input text).
1505 If spaces extend into the next chunk of input text, the next chunk will
1506 be read in. Normally the current chunk will be discarded at the same
1507 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1508 chunk will not be discarded.
1513 #define LEX_NO_NEXT_CHUNK 0x80000000
1516 Perl_lex_read_space(pTHX_ U32 flags)
1519 bool need_incline = 0;
1520 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1521 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1524 sv_free(PL_skipwhite);
1525 PL_skipwhite = NULL;
1528 PL_skipwhite = newSVpvs("");
1529 #endif /* PERL_MAD */
1530 s = PL_parser->bufptr;
1531 bufend = PL_parser->bufend;
1537 } while (!(c == '\n' || (c == 0 && s == bufend)));
1538 } else if (c == '\n') {
1540 PL_parser->linestart = s;
1545 } else if (isSPACE(c)) {
1547 } else if (c == 0 && s == bufend) {
1551 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1552 #endif /* PERL_MAD */
1553 if (flags & LEX_NO_NEXT_CHUNK)
1555 PL_parser->bufptr = s;
1556 COPLINE_INC_WITH_HERELINES;
1557 got_more = lex_next_chunk(flags);
1558 CopLINE_dec(PL_curcop);
1559 s = PL_parser->bufptr;
1560 bufend = PL_parser->bufend;
1563 if (need_incline && PL_parser->rsfp) {
1573 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1574 #endif /* PERL_MAD */
1575 PL_parser->bufptr = s;
1580 * This subroutine has nothing to do with tilting, whether at windmills
1581 * or pinball tables. Its name is short for "increment line". It
1582 * increments the current line number in CopLINE(PL_curcop) and checks
1583 * to see whether the line starts with a comment of the form
1584 * # line 500 "foo.pm"
1585 * If so, it sets the current line number and file to the values in the comment.
1589 S_incline(pTHX_ const char *s)
1597 PERL_ARGS_ASSERT_INCLINE;
1599 COPLINE_INC_WITH_HERELINES;
1600 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1601 && s+1 == PL_bufend && *s == ';') {
1602 /* fake newline in string eval */
1603 CopLINE_dec(PL_curcop);
1608 while (SPACE_OR_TAB(*s))
1610 if (strnEQ(s, "line", 4))
1614 if (SPACE_OR_TAB(*s))
1618 while (SPACE_OR_TAB(*s))
1626 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1628 while (SPACE_OR_TAB(*s))
1630 if (*s == '"' && (t = strchr(s+1, '"'))) {
1636 while (!isSPACE(*t))
1640 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1642 if (*e != '\n' && *e != '\0')
1643 return; /* false alarm */
1645 line_num = atoi(n)-1;
1648 const STRLEN len = t - s;
1649 SV *const temp_sv = CopFILESV(PL_curcop);
1654 cf = SvPVX(temp_sv);
1655 tmplen = SvCUR(temp_sv);
1661 if (!PL_rsfp && !PL_parser->filtered) {
1662 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1663 * to *{"::_<newfilename"} */
1664 /* However, the long form of evals is only turned on by the
1665 debugger - usually they're "(eval %lu)" */
1669 STRLEN tmplen2 = len;
1670 if (tmplen + 2 <= sizeof smallbuf)
1673 Newx(tmpbuf, tmplen + 2, char);
1676 memcpy(tmpbuf + 2, cf, tmplen);
1678 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1683 if (tmplen2 + 2 <= sizeof smallbuf)
1686 Newx(tmpbuf2, tmplen2 + 2, char);
1688 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1689 /* Either they malloc'd it, or we malloc'd it,
1690 so no prefix is present in ours. */
1695 memcpy(tmpbuf2 + 2, s, tmplen2);
1698 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1700 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1701 /* adjust ${"::_<newfilename"} to store the new file name */
1702 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1703 /* The line number may differ. If that is the case,
1704 alias the saved lines that are in the array.
1705 Otherwise alias the whole array. */
1706 if (CopLINE(PL_curcop) == line_num) {
1707 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1708 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1710 else if (GvAV(*gvp)) {
1711 AV * const av = GvAV(*gvp);
1712 const I32 start = CopLINE(PL_curcop)+1;
1713 I32 items = AvFILLp(av) - start;
1715 AV * const av2 = GvAVn(gv2);
1716 SV **svp = AvARRAY(av) + start;
1717 I32 l = (I32)line_num+1;
1719 av_store(av2, l++, SvREFCNT_inc(*svp++));
1724 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1726 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1728 CopFILE_free(PL_curcop);
1729 CopFILE_setn(PL_curcop, s, len);
1731 CopLINE_set(PL_curcop, line_num);
1735 /* skip space before PL_thistoken */
1738 S_skipspace0(pTHX_ char *s)
1740 PERL_ARGS_ASSERT_SKIPSPACE0;
1747 PL_thiswhite = newSVpvs("");
1748 sv_catsv(PL_thiswhite, PL_skipwhite);
1749 sv_free(PL_skipwhite);
1752 PL_realtokenstart = s - SvPVX(PL_linestr);
1756 /* skip space after PL_thistoken */
1759 S_skipspace1(pTHX_ char *s)
1761 const char *start = s;
1762 I32 startoff = start - SvPVX(PL_linestr);
1764 PERL_ARGS_ASSERT_SKIPSPACE1;
1769 start = SvPVX(PL_linestr) + startoff;
1770 if (!PL_thistoken && PL_realtokenstart >= 0) {
1771 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1772 PL_thistoken = newSVpvn(tstart, start - tstart);
1774 PL_realtokenstart = -1;
1777 PL_nextwhite = newSVpvs("");
1778 sv_catsv(PL_nextwhite, PL_skipwhite);
1779 sv_free(PL_skipwhite);
1786 S_skipspace2(pTHX_ char *s, SV **svp)
1789 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1790 const I32 startoff = s - SvPVX(PL_linestr);
1792 PERL_ARGS_ASSERT_SKIPSPACE2;
1795 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1796 if (!PL_madskills || !svp)
1798 start = SvPVX(PL_linestr) + startoff;
1799 if (!PL_thistoken && PL_realtokenstart >= 0) {
1800 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1801 PL_thistoken = newSVpvn(tstart, start - tstart);
1802 PL_realtokenstart = -1;
1806 *svp = newSVpvs("");
1807 sv_setsv(*svp, PL_skipwhite);
1808 sv_free(PL_skipwhite);
1817 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1819 AV *av = CopFILEAVx(PL_curcop);
1821 SV * const sv = newSV_type(SVt_PVMG);
1823 sv_setsv(sv, orig_sv);
1825 sv_setpvn(sv, buf, len);
1828 av_store(av, (I32)CopLINE(PL_curcop), sv);
1834 * Called to gobble the appropriate amount and type of whitespace.
1835 * Skips comments as well.
1839 S_skipspace(pTHX_ char *s)
1843 #endif /* PERL_MAD */
1844 PERL_ARGS_ASSERT_SKIPSPACE;
1847 sv_free(PL_skipwhite);
1848 PL_skipwhite = NULL;
1850 #endif /* PERL_MAD */
1851 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1852 while (s < PL_bufend && SPACE_OR_TAB(*s))
1855 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1857 lex_read_space(LEX_KEEP_PREVIOUS |
1858 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1859 LEX_NO_NEXT_CHUNK : 0));
1861 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1862 if (PL_linestart > PL_bufptr)
1863 PL_bufptr = PL_linestart;
1868 PL_skipwhite = newSVpvn(start, s-start);
1869 #endif /* PERL_MAD */
1875 * Check the unary operators to ensure there's no ambiguity in how they're
1876 * used. An ambiguous piece of code would be:
1878 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1879 * the +5 is its argument.
1889 if (PL_oldoldbufptr != PL_last_uni)
1891 while (isSPACE(*PL_last_uni))
1894 while (isWORDCHAR_lazy_if(s,UTF) || *s == '-')
1896 if ((t = strchr(s, '(')) && t < PL_bufptr)
1899 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1900 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1901 (int)(s - PL_last_uni), PL_last_uni);
1905 * LOP : macro to build a list operator. Its behaviour has been replaced
1906 * with a subroutine, S_lop() for which LOP is just another name.
1909 #define LOP(f,x) return lop(f,x,s)
1913 * Build a list operator (or something that might be one). The rules:
1914 * - if we have a next token, then it's a list operator [why?]
1915 * - if the next thing is an opening paren, then it's a function
1916 * - else it's a list operator
1920 S_lop(pTHX_ I32 f, int x, char *s)
1924 PERL_ARGS_ASSERT_LOP;
1930 PL_last_lop = PL_oldbufptr;
1931 PL_last_lop_op = (OPCODE)f;
1940 return REPORT(FUNC);
1943 return REPORT(FUNC);
1946 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1947 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1948 return REPORT(LSTOP);
1955 * Sets up for an eventual force_next(). start_force(0) basically does
1956 * an unshift, while start_force(-1) does a push. yylex removes items
1961 S_start_force(pTHX_ int where)
1965 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1966 where = PL_lasttoke;
1967 assert(PL_curforce < 0 || PL_curforce == where);
1968 if (PL_curforce != where) {
1969 for (i = PL_lasttoke; i > where; --i) {
1970 PL_nexttoke[i] = PL_nexttoke[i-1];
1974 if (PL_curforce < 0) /* in case of duplicate start_force() */
1975 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1976 PL_curforce = where;
1979 curmad('^', newSVpvs(""));
1980 CURMAD('_', PL_nextwhite);
1985 S_curmad(pTHX_ char slot, SV *sv)
1991 if (PL_curforce < 0)
1992 where = &PL_thismad;
1994 where = &PL_nexttoke[PL_curforce].next_mad;
2000 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
2002 else if (PL_encoding) {
2003 sv_recode_to_utf8(sv, PL_encoding);
2008 /* keep a slot open for the head of the list? */
2009 if (slot != '_' && *where && (*where)->mad_key == '^') {
2010 (*where)->mad_key = slot;
2011 sv_free(MUTABLE_SV(((*where)->mad_val)));
2012 (*where)->mad_val = (void*)sv;
2015 addmad(newMADsv(slot, sv), where, 0);
2018 # define start_force(where) NOOP
2019 # define curmad(slot, sv) NOOP
2024 * When the lexer realizes it knows the next token (for instance,
2025 * it is reordering tokens for the parser) then it can call S_force_next
2026 * to know what token to return the next time the lexer is called. Caller
2027 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
2028 * and possibly PL_expect to ensure the lexer handles the token correctly.
2032 S_force_next(pTHX_ I32 type)
2037 PerlIO_printf(Perl_debug_log, "### forced token:\n");
2038 tokereport(type, &NEXTVAL_NEXTTOKE);
2042 if (PL_curforce < 0)
2043 start_force(PL_lasttoke);
2044 PL_nexttoke[PL_curforce].next_type = type;
2045 if (PL_lex_state != LEX_KNOWNEXT)
2046 PL_lex_defer = PL_lex_state;
2047 PL_lex_state = LEX_KNOWNEXT;
2048 PL_lex_expect = PL_expect;
2051 PL_nexttype[PL_nexttoke] = type;
2053 if (PL_lex_state != LEX_KNOWNEXT) {
2054 PL_lex_defer = PL_lex_state;
2055 PL_lex_expect = PL_expect;
2056 PL_lex_state = LEX_KNOWNEXT;
2064 int yyc = PL_parser->yychar;
2065 if (yyc != YYEMPTY) {
2068 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2069 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2070 PL_lex_allbrackets--;
2072 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2073 } else if (yyc == '('/*)*/) {
2074 PL_lex_allbrackets--;
2079 PL_parser->yychar = YYEMPTY;
2084 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2087 SV * const sv = newSVpvn_utf8(start, len,
2090 && !is_ascii_string((const U8*)start, len)
2091 && is_utf8_string((const U8*)start, len));
2097 * When the lexer knows the next thing is a word (for instance, it has
2098 * just seen -> and it knows that the next char is a word char, then
2099 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2103 * char *start : buffer position (must be within PL_linestr)
2104 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2105 * int check_keyword : if true, Perl checks to make sure the word isn't
2106 * a keyword (do this if the word is a label, e.g. goto FOO)
2107 * int allow_pack : if true, : characters will also be allowed (require,
2108 * use, etc. do this)
2109 * int allow_initial_tick : used by the "sub" lexer only.
2113 S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2119 PERL_ARGS_ASSERT_FORCE_WORD;
2121 start = SKIPSPACE1(start);
2123 if (isIDFIRST_lazy_if(s,UTF) ||
2124 (allow_pack && *s == ':') ||
2125 (allow_initial_tick && *s == '\'') )
2127 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2128 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2130 start_force(PL_curforce);
2132 curmad('X', newSVpvn(start,s-start));
2133 if (token == METHOD) {
2138 PL_expect = XOPERATOR;
2142 curmad('g', newSVpvs( "forced" ));
2143 NEXTVAL_NEXTTOKE.opval
2144 = (OP*)newSVOP(OP_CONST,0,
2145 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2146 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2154 * Called when the lexer wants $foo *foo &foo etc, but the program
2155 * text only contains the "foo" portion. The first argument is a pointer
2156 * to the "foo", and the second argument is the type symbol to prefix.
2157 * Forces the next token to be a "WORD".
2158 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2162 S_force_ident(pTHX_ const char *s, int kind)
2166 PERL_ARGS_ASSERT_FORCE_IDENT;
2169 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
2170 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2171 UTF ? SVf_UTF8 : 0));
2172 start_force(PL_curforce);
2173 NEXTVAL_NEXTTOKE.opval = o;
2176 o->op_private = OPpCONST_ENTERED;
2177 /* XXX see note in pp_entereval() for why we forgo typo
2178 warnings if the symbol must be introduced in an eval.
2180 gv_fetchpvn_flags(s, len,
2181 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2182 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2183 kind == '$' ? SVt_PV :
2184 kind == '@' ? SVt_PVAV :
2185 kind == '%' ? SVt_PVHV :
2193 S_force_ident_maybe_lex(pTHX_ char pit)
2195 start_force(PL_curforce);
2196 NEXTVAL_NEXTTOKE.ival = pit;
2201 Perl_str_to_version(pTHX_ SV *sv)
2206 const char *start = SvPV_const(sv,len);
2207 const char * const end = start + len;
2208 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2210 PERL_ARGS_ASSERT_STR_TO_VERSION;
2212 while (start < end) {
2216 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2221 retval += ((NV)n)/nshift;
2230 * Forces the next token to be a version number.
2231 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2232 * and if "guessing" is TRUE, then no new token is created (and the caller
2233 * must use an alternative parsing method).
2237 S_force_version(pTHX_ char *s, int guessing)
2243 I32 startoff = s - SvPVX(PL_linestr);
2246 PERL_ARGS_ASSERT_FORCE_VERSION;
2254 while (isDIGIT(*d) || *d == '_' || *d == '.')
2258 start_force(PL_curforce);
2259 curmad('X', newSVpvn(s,d-s));
2262 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2264 #ifdef USE_LOCALE_NUMERIC
2265 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2266 setlocale(LC_NUMERIC, "C");
2268 s = scan_num(s, &pl_yylval);
2269 #ifdef USE_LOCALE_NUMERIC
2270 setlocale(LC_NUMERIC, loc);
2273 version = pl_yylval.opval;
2274 ver = cSVOPx(version)->op_sv;
2275 if (SvPOK(ver) && !SvNIOK(ver)) {
2276 SvUPGRADE(ver, SVt_PVNV);
2277 SvNV_set(ver, str_to_version(ver));
2278 SvNOK_on(ver); /* hint that it is a version */
2281 else if (guessing) {
2284 sv_free(PL_nextwhite); /* let next token collect whitespace */
2286 s = SvPVX(PL_linestr) + startoff;
2294 if (PL_madskills && !version) {
2295 sv_free(PL_nextwhite); /* let next token collect whitespace */
2297 s = SvPVX(PL_linestr) + startoff;
2300 /* NOTE: The parser sees the package name and the VERSION swapped */
2301 start_force(PL_curforce);
2302 NEXTVAL_NEXTTOKE.opval = version;
2309 * S_force_strict_version
2310 * Forces the next token to be a version number using strict syntax rules.
2314 S_force_strict_version(pTHX_ char *s)
2319 I32 startoff = s - SvPVX(PL_linestr);
2321 const char *errstr = NULL;
2323 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2325 while (isSPACE(*s)) /* leading whitespace */
2328 if (is_STRICT_VERSION(s,&errstr)) {
2330 s = (char *)scan_version(s, ver, 0);
2331 version = newSVOP(OP_CONST, 0, ver);
2333 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2334 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2338 yyerror(errstr); /* version required */
2343 if (PL_madskills && !version) {
2344 sv_free(PL_nextwhite); /* let next token collect whitespace */
2346 s = SvPVX(PL_linestr) + startoff;
2349 /* NOTE: The parser sees the package name and the VERSION swapped */
2350 start_force(PL_curforce);
2351 NEXTVAL_NEXTTOKE.opval = version;
2359 * Tokenize a quoted string passed in as an SV. It finds the next
2360 * chunk, up to end of string or a backslash. It may make a new
2361 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2366 S_tokeq(pTHX_ SV *sv)
2375 PERL_ARGS_ASSERT_TOKEQ;
2380 s = SvPV_force(sv, len);
2381 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2384 /* This is relying on the SV being "well formed" with a trailing '\0' */
2385 while (s < send && !(*s == '\\' && s[1] == '\\'))
2390 if ( PL_hints & HINT_NEW_STRING ) {
2391 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2395 if (s + 1 < send && (s[1] == '\\'))
2396 s++; /* all that, just for this */
2401 SvCUR_set(sv, d - SvPVX_const(sv));
2403 if ( PL_hints & HINT_NEW_STRING )
2404 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2409 * Now come three functions related to double-quote context,
2410 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2411 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2412 * interact with PL_lex_state, and create fake ( ... ) argument lists
2413 * to handle functions and concatenation.
2417 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2422 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2424 * Pattern matching will set PL_lex_op to the pattern-matching op to
2425 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2427 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2429 * Everything else becomes a FUNC.
2431 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2432 * had an OP_CONST or OP_READLINE). This just sets us up for a
2433 * call to S_sublex_push().
2437 S_sublex_start(pTHX)
2440 const I32 op_type = pl_yylval.ival;
2442 if (op_type == OP_NULL) {
2443 pl_yylval.opval = PL_lex_op;
2447 if (op_type == OP_CONST || op_type == OP_READLINE) {
2448 SV *sv = tokeq(PL_lex_stuff);
2450 if (SvTYPE(sv) == SVt_PVIV) {
2451 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2453 const char * const p = SvPV_const(sv, len);
2454 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2458 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2459 PL_lex_stuff = NULL;
2460 /* Allow <FH> // "foo" */
2461 if (op_type == OP_READLINE)
2462 PL_expect = XTERMORDORDOR;
2465 else if (op_type == OP_BACKTICK && PL_lex_op) {
2466 /* readpipe() vas overriden */
2467 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2468 pl_yylval.opval = PL_lex_op;
2470 PL_lex_stuff = NULL;
2474 PL_sublex_info.super_state = PL_lex_state;
2475 PL_sublex_info.sub_inwhat = (U16)op_type;
2476 PL_sublex_info.sub_op = PL_lex_op;
2477 PL_lex_state = LEX_INTERPPUSH;
2481 pl_yylval.opval = PL_lex_op;
2491 * Create a new scope to save the lexing state. The scope will be
2492 * ended in S_sublex_done. Returns a '(', starting the function arguments
2493 * to the uc, lc, etc. found before.
2494 * Sets PL_lex_state to LEX_INTERPCONCAT.
2504 PL_lex_state = PL_sublex_info.super_state;
2505 SAVEBOOL(PL_lex_dojoin);
2506 SAVEI32(PL_lex_brackets);
2507 SAVEI32(PL_lex_allbrackets);
2508 SAVEI32(PL_lex_formbrack);
2509 SAVEI8(PL_lex_fakeeof);
2510 SAVEI32(PL_lex_casemods);
2511 SAVEI32(PL_lex_starts);
2512 SAVEI8(PL_lex_state);
2513 SAVESPTR(PL_lex_repl);
2514 SAVEVPTR(PL_lex_inpat);
2515 SAVEI16(PL_lex_inwhat);
2516 SAVECOPLINE(PL_curcop);
2517 SAVEPPTR(PL_bufptr);
2518 SAVEPPTR(PL_bufend);
2519 SAVEPPTR(PL_oldbufptr);
2520 SAVEPPTR(PL_oldoldbufptr);
2521 SAVEPPTR(PL_last_lop);
2522 SAVEPPTR(PL_last_uni);
2523 SAVEPPTR(PL_linestart);
2524 SAVESPTR(PL_linestr);
2525 SAVEGENERICPV(PL_lex_brackstack);
2526 SAVEGENERICPV(PL_lex_casestack);
2527 SAVEGENERICPV(PL_parser->lex_shared);
2529 /* The here-doc parser needs to be able to peek into outer lexing
2530 scopes to find the body of the here-doc. So we put PL_linestr and
2531 PL_bufptr into lex_shared, to ‘share’ those values.
2533 PL_parser->lex_shared->ls_linestr = PL_linestr;
2534 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2536 PL_linestr = PL_lex_stuff;
2537 PL_lex_repl = PL_sublex_info.repl;
2538 PL_lex_stuff = NULL;
2539 PL_sublex_info.repl = NULL;
2541 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2542 = SvPVX(PL_linestr);
2543 PL_bufend += SvCUR(PL_linestr);
2544 PL_last_lop = PL_last_uni = NULL;
2545 SAVEFREESV(PL_linestr);
2546 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2548 PL_lex_dojoin = FALSE;
2549 PL_lex_brackets = PL_lex_formbrack = 0;
2550 PL_lex_allbrackets = 0;
2551 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2552 Newx(PL_lex_brackstack, 120, char);
2553 Newx(PL_lex_casestack, 12, char);
2554 PL_lex_casemods = 0;
2555 *PL_lex_casestack = '\0';
2557 PL_lex_state = LEX_INTERPCONCAT;
2558 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2560 Newxz(shared, 1, LEXSHARED);
2561 shared->ls_prev = PL_parser->lex_shared;
2562 PL_parser->lex_shared = shared;
2564 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2565 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2566 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2567 PL_lex_inpat = PL_sublex_info.sub_op;
2569 PL_lex_inpat = NULL;
2576 * Restores lexer state after a S_sublex_push.
2583 if (!PL_lex_starts++) {
2584 SV * const sv = newSVpvs("");
2585 if (SvUTF8(PL_linestr))
2587 PL_expect = XOPERATOR;
2588 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2592 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2593 PL_lex_state = LEX_INTERPCASEMOD;
2597 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2598 assert(PL_lex_inwhat != OP_TRANSR);
2599 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2600 PL_linestr = PL_lex_repl;
2602 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2603 PL_bufend += SvCUR(PL_linestr);
2604 PL_last_lop = PL_last_uni = NULL;
2605 PL_lex_dojoin = FALSE;
2606 PL_lex_brackets = 0;
2607 PL_lex_allbrackets = 0;
2608 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2609 PL_lex_casemods = 0;
2610 *PL_lex_casestack = '\0';
2612 if (SvEVALED(PL_lex_repl)) {
2613 PL_lex_state = LEX_INTERPNORMAL;
2615 /* we don't clear PL_lex_repl here, so that we can check later
2616 whether this is an evalled subst; that means we rely on the
2617 logic to ensure sublex_done() is called again only via the
2618 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2621 PL_lex_state = LEX_INTERPCONCAT;
2631 PL_endwhite = newSVpvs("");
2632 sv_catsv(PL_endwhite, PL_thiswhite);
2636 sv_setpvs(PL_thistoken,"");
2638 PL_realtokenstart = -1;
2642 PL_bufend = SvPVX(PL_linestr);
2643 PL_bufend += SvCUR(PL_linestr);
2644 PL_expect = XOPERATOR;
2645 PL_sublex_info.sub_inwhat = 0;
2650 PERL_STATIC_INLINE SV*
2651 S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2653 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2654 * interior, hence to the "}". Finds what the name resolves to, returning
2655 * an SV* containing it; NULL if no valid one found */
2657 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
2664 const U8* first_bad_char_loc;
2665 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2667 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2669 if (UTF && ! is_utf8_string_loc((U8 *) backslash_ptr,
2671 &first_bad_char_loc))
2673 /* If warnings are on, this will print a more detailed analysis of what
2674 * is wrong than the error message below */
2675 utf8n_to_uvuni(first_bad_char_loc,
2676 e - ((char *) first_bad_char_loc),
2679 /* We deliberately don't try to print the malformed character, which
2680 * might not print very well; it also may be just the first of many
2681 * malformations, so don't print what comes after it */
2682 yyerror(Perl_form(aTHX_
2683 "Malformed UTF-8 character immediately after '%.*s'",
2684 (int) (first_bad_char_loc - (U8 *) backslash_ptr), backslash_ptr));
2688 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2689 /* include the <}> */
2690 e - backslash_ptr + 1);
2692 SvREFCNT_dec_NN(res);
2696 /* See if the charnames handler is the Perl core's, and if so, we can skip
2697 * the validation needed for a user-supplied one, as Perl's does its own
2699 table = GvHV(PL_hintgv); /* ^H */
2700 cvp = hv_fetchs(table, "charnames", FALSE);
2701 if (cvp && (cv = *cvp) && SvROK(cv) && ((rv = SvRV(cv)) != NULL)
2702 && SvTYPE(rv) == SVt_PVCV && ((stash = CvSTASH(rv)) != NULL))
2704 const char * const name = HvNAME(stash);
2705 if strEQ(name, "_charnames") {
2710 /* Here, it isn't Perl's charname handler. We can't rely on a
2711 * user-supplied handler to validate the input name. For non-ut8 input,
2712 * look to see that the first character is legal. Then loop through the
2713 * rest checking that each is a continuation */
2715 /* This code needs to be sync'ed with a regex in _charnames.pm which does
2719 if (! isALPHAU(*s)) {
2724 if (! isCHARNAME_CONT(*s)) {
2727 if (*s == ' ' && *(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2728 Perl_warn(aTHX_ "A sequence of multiple spaces in a charnames alias definition is deprecated");
2732 if (*(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2733 Perl_warn(aTHX_ "Trailing white-space in a charnames alias definition is deprecated");
2737 /* Similarly for utf8. For invariants can check directly; for other
2738 * Latin1, can calculate their code point and check; otherwise use a
2740 if (UTF8_IS_INVARIANT(*s)) {
2741 if (! isALPHAU(*s)) {
2745 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2746 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s, *(s+1))))) {
2752 if (! PL_utf8_charname_begin) {
2753 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2754 PL_utf8_charname_begin = _core_swash_init("utf8",
2755 "_Perl_Charname_Begin",
2757 1, 0, NULL, &flags);
2759 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2766 if (UTF8_IS_INVARIANT(*s)) {
2767 if (! isCHARNAME_CONT(*s)) {
2770 if (*s == ' ' && *(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2771 Perl_warn(aTHX_ "A sequence of multiple spaces in a charnames alias definition is deprecated");
2775 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
2776 if (! isCHARNAME_CONT(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*s,
2784 if (! PL_utf8_charname_continue) {
2785 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2786 PL_utf8_charname_continue = _core_swash_init("utf8",
2787 "_Perl_Charname_Continue",
2789 1, 0, NULL, &flags);
2791 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2797 if (*(s-1) == ' ' && ckWARN(WARN_DEPRECATED)) {
2798 Perl_warn(aTHX_ "Trailing white-space in a charnames alias definition is deprecated");
2802 if (SvUTF8(res)) { /* Don't accept malformed input */
2803 const U8* first_bad_char_loc;
2805 const char* const str = SvPV_const(res, len);
2806 if (! is_utf8_string_loc((U8 *) str, len, &first_bad_char_loc)) {
2807 /* If warnings are on, this will print a more detailed analysis of
2808 * what is wrong than the error message below */
2809 utf8n_to_uvuni(first_bad_char_loc,
2810 (char *) first_bad_char_loc - str,
2813 /* We deliberately don't try to print the malformed character,
2814 * which might not print very well; it also may be just the first
2815 * of many malformations, so don't print what comes after it */
2818 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2819 (int) (e - backslash_ptr + 1), backslash_ptr,
2820 (int) ((char *) first_bad_char_loc - str), str
2830 int bad_char_size = ((UTF) ? UTF8SKIP(s) : 1);
2832 /* The final %.*s makes sure that should the trailing NUL be missing
2833 * that this print won't run off the end of the string */
2836 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2837 (int)(s - backslash_ptr + bad_char_size), backslash_ptr,
2838 (int)(e - s + bad_char_size), s + bad_char_size
2840 UTF ? SVf_UTF8 : 0);
2848 Extracts the next constant part of a pattern, double-quoted string,
2849 or transliteration. This is terrifying code.
2851 For example, in parsing the double-quoted string "ab\x63$d", it would
2852 stop at the '$' and return an OP_CONST containing 'abc'.
2854 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2855 processing a pattern (PL_lex_inpat is true), a transliteration
2856 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2858 Returns a pointer to the character scanned up to. If this is
2859 advanced from the start pointer supplied (i.e. if anything was
2860 successfully parsed), will leave an OP_CONST for the substring scanned
2861 in pl_yylval. Caller must intuit reason for not parsing further
2862 by looking at the next characters herself.
2866 \N{FOO} => \N{U+hex_for_character_FOO}
2867 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
2870 all other \-char, including \N and \N{ apart from \N{ABC}
2873 @ and $ where it appears to be a var, but not for $ as tail anchor
2878 In transliterations:
2879 characters are VERY literal, except for - not at the start or end
2880 of the string, which indicates a range. If the range is in bytes,
2881 scan_const expands the range to the full set of intermediate
2882 characters. If the range is in utf8, the hyphen is replaced with
2883 a certain range mark which will be handled by pmtrans() in op.c.
2885 In double-quoted strings:
2887 double-quoted style: \r and \n
2888 constants: \x31, etc.
2889 deprecated backrefs: \1 (in substitution replacements)
2890 case and quoting: \U \Q \E
2893 scan_const does *not* construct ops to handle interpolated strings.
2894 It stops processing as soon as it finds an embedded $ or @ variable
2895 and leaves it to the caller to work out what's going on.
2897 embedded arrays (whether in pattern or not) could be:
2898 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2900 $ in double-quoted strings must be the symbol of an embedded scalar.
2902 $ in pattern could be $foo or could be tail anchor. Assumption:
2903 it's a tail anchor if $ is the last thing in the string, or if it's
2904 followed by one of "()| \r\n\t"
2906 \1 (backreferences) are turned into $1 in substitutions
2908 The structure of the code is
2909 while (there's a character to process) {
2910 handle transliteration ranges
2911 skip regexp comments /(?#comment)/ and codes /(?{code})/
2912 skip #-initiated comments in //x patterns
2913 check for embedded arrays
2914 check for embedded scalars
2916 deprecate \1 in substitution replacements
2917 handle string-changing backslashes \l \U \Q \E, etc.
2918 switch (what was escaped) {
2919 handle \- in a transliteration (becomes a literal -)
2920 if a pattern and not \N{, go treat as regular character
2921 handle \132 (octal characters)
2922 handle \x15 and \x{1234} (hex characters)
2923 handle \N{name} (named characters, also \N{3,5} in a pattern)
2924 handle \cV (control characters)
2925 handle printf-style backslashes (\f, \r, \n, etc)
2928 } (end if backslash)
2929 handle regular character
2930 } (end while character to read)
2935 S_scan_const(pTHX_ char *start)
2938 char *send = PL_bufend; /* end of the constant */
2939 SV *sv = newSV(send - start); /* sv for the constant. See
2940 note below on sizing. */
2941 char *s = start; /* start of the constant */
2942 char *d = SvPVX(sv); /* destination for copies */
2943 bool dorange = FALSE; /* are we in a translit range? */
2944 bool didrange = FALSE; /* did we just finish a range? */
2945 bool in_charclass = FALSE; /* within /[...]/ */
2946 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2947 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2948 to be UTF8? But, this can
2949 show as true when the source
2950 isn't utf8, as for example
2951 when it is entirely composed
2953 SV *res; /* result from charnames */
2955 /* Note on sizing: The scanned constant is placed into sv, which is
2956 * initialized by newSV() assuming one byte of output for every byte of
2957 * input. This routine expects newSV() to allocate an extra byte for a
2958 * trailing NUL, which this routine will append if it gets to the end of
2959 * the input. There may be more bytes of input than output (eg., \N{LATIN
2960 * CAPITAL LETTER A}), or more output than input if the constant ends up
2961 * recoded to utf8, but each time a construct is found that might increase
2962 * the needed size, SvGROW() is called. Its size parameter each time is
2963 * based on the best guess estimate at the time, namely the length used so
2964 * far, plus the length the current construct will occupy, plus room for
2965 * the trailing NUL, plus one byte for every input byte still unscanned */
2967 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2970 UV literal_endpoint = 0;
2971 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2974 PERL_ARGS_ASSERT_SCAN_CONST;
2976 assert(PL_lex_inwhat != OP_TRANSR);
2977 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2978 /* If we are doing a trans and we know we want UTF8 set expectation */
2979 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2980 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2983 /* Protect sv from errors and fatal warnings. */
2984 ENTER_with_name("scan_const");
2987 while (s < send || dorange) {
2989 /* get transliterations out of the way (they're most literal) */
2990 if (PL_lex_inwhat == OP_TRANS) {
2991 /* expand a range A-Z to the full set of characters. AIE! */
2993 I32 i; /* current expanded character */
2994 I32 min; /* first character in range */
2995 I32 max; /* last character in range */
3006 char * const c = (char*)utf8_hop((U8*)d, -1);
3010 *c = (char)UTF_TO_NATIVE(0xff);
3011 /* mark the range as done, and continue */
3017 i = d - SvPVX_const(sv); /* remember current offset */
3020 SvLEN(sv) + (has_utf8 ?
3021 (512 - UTF_CONTINUATION_MARK +
3024 /* How many two-byte within 0..255: 128 in UTF-8,
3025 * 96 in UTF-8-mod. */
3027 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
3029 d = SvPVX(sv) + i; /* refresh d after realloc */
3033 for (j = 0; j <= 1; j++) {
3034 char * const c = (char*)utf8_hop((U8*)d, -1);
3035 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
3041 max = (U8)0xff; /* only to \xff */
3042 uvmax = uv; /* \x{100} to uvmax */
3044 d = c; /* eat endpoint chars */
3049 d -= 2; /* eat the first char and the - */
3050 min = (U8)*d; /* first char in range */
3051 max = (U8)d[1]; /* last char in range */
3058 "Invalid range \"%c-%c\" in transliteration operator",
3059 (char)min, (char)max);
3063 if (literal_endpoint == 2 &&
3064 ((isLOWER(min) && isLOWER(max)) ||
3065 (isUPPER(min) && isUPPER(max)))) {
3067 for (i = min; i <= max; i++)
3069 *d++ = NATIVE_TO_NEED(has_utf8,i);
3071 for (i = min; i <= max; i++)
3073 *d++ = NATIVE_TO_NEED(has_utf8,i);
3078 for (i = min; i <= max; i++)
3081 const U8 ch = (U8)NATIVE_TO_UTF(i);
3082 if (UNI_IS_INVARIANT(ch))
3085 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
3086 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
3095 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
3097 *d++ = (char)UTF_TO_NATIVE(0xff);
3099 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
3103 /* mark the range as done, and continue */
3107 literal_endpoint = 0;
3112 /* range begins (ignore - as first or last char) */
3113 else if (*s == '-' && s+1 < send && s != start) {
3115 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
3122 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
3132 literal_endpoint = 0;
3133 native_range = TRUE;
3138 /* if we get here, we're not doing a transliteration */
3140 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3143 while (s1 >= start && *s1-- == '\\')
3146 in_charclass = TRUE;
3149 else if (*s == ']' && PL_lex_inpat && in_charclass) {
3152 while (s1 >= start && *s1-- == '\\')
3155 in_charclass = FALSE;
3158 /* skip for regexp comments /(?#comment)/, except for the last
3159 * char, which will be done separately.
3160 * Stop on (?{..}) and friends */
3162 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
3164 while (s+1 < send && *s != ')')
3165 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3167 else if (!PL_lex_casemods && !in_charclass &&
3168 ( s[2] == '{' /* This should match regcomp.c */
3169 || (s[2] == '?' && s[3] == '{')))
3175 /* likewise skip #-initiated comments in //x patterns */
3176 else if (*s == '#' && PL_lex_inpat &&
3177 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
3178 while (s+1 < send && *s != '\n')
3179 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3182 /* no further processing of single-quoted regex */
3183 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3184 goto default_action;
3186 /* check for embedded arrays
3187 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3189 else if (*s == '@' && s[1]) {
3190 if (isWORDCHAR_lazy_if(s+1,UTF))
3192 if (strchr(":'{$", s[1]))
3194 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3195 break; /* in regexp, neither @+ nor @- are interpolated */
3198 /* check for embedded scalars. only stop if we're sure it's a
3201 else if (*s == '$') {
3202 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
3204 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
3206 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3207 "Possible unintended interpolation of $\\ in regex");
3209 break; /* in regexp, $ might be tail anchor */
3213 /* End of else if chain - OP_TRANS rejoin rest */
3216 if (*s == '\\' && s+1 < send) {
3217 char* e; /* Can be used for ending '}', etc. */
3221 /* warn on \1 - \9 in substitution replacements, but note that \11
3222 * is an octal; and \19 is \1 followed by '9' */
3223 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
3224 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
3226 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
3231 /* string-change backslash escapes */
3232 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
3236 /* In a pattern, process \N, but skip any other backslash escapes.
3237 * This is because we don't want to translate an escape sequence
3238 * into a meta symbol and have the regex compiler use the meta
3239 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3240 * in spite of this, we do have to process \N here while the proper
3241 * charnames handler is in scope. See bugs #56444 and #62056.
3242 * There is a complication because \N in a pattern may also stand
3243 * for 'match a non-nl', and not mean a charname, in which case its
3244 * processing should be deferred to the regex compiler. To be a
3245 * charname it must be followed immediately by a '{', and not look
3246 * like \N followed by a curly quantifier, i.e., not something like
3247 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3249 else if (PL_lex_inpat
3252 || regcurly(s + 1, FALSE)))
3254 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3255 goto default_action;
3260 /* quoted - in transliterations */
3262 if (PL_lex_inwhat == OP_TRANS) {
3269 if ((isALPHANUMERIC(*s)))
3270 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3271 "Unrecognized escape \\%c passed through",
3273 /* default action is to copy the quoted character */
3274 goto default_action;
3277 /* eg. \132 indicates the octal constant 0132 */
3278 case '0': case '1': case '2': case '3':
3279 case '4': case '5': case '6': case '7':
3281 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
3283 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3285 if (len < 3 && s < send && isDIGIT(*s)
3286 && ckWARN(WARN_MISC))
3288 Perl_warner(aTHX_ packWARN(WARN_MISC),
3289 "%s", form_short_octal_warning(s, len));
3292 goto NUM_ESCAPE_INSERT;
3294 /* eg. \o{24} indicates the octal constant \024 */
3299 bool valid = grok_bslash_o(&s, &uv, &error,
3300 TRUE, /* Output warning */
3301 FALSE, /* Not strict */
3302 TRUE, /* Output warnings for
3309 goto NUM_ESCAPE_INSERT;
3312 /* eg. \x24 indicates the hex constant 0x24 */
3317 bool valid = grok_bslash_x(&s, &uv, &error,
3318 TRUE, /* Output warning */
3319 FALSE, /* Not strict */
3320 TRUE, /* Output warnings for
3330 /* Insert oct or hex escaped character. There will always be
3331 * enough room in sv since such escapes will be longer than any
3332 * UTF-8 sequence they can end up as, except if they force us
3333 * to recode the rest of the string into utf8 */
3335 /* Here uv is the ordinal of the next character being added in
3336 * unicode (converted from native). */
3337 if (!UNI_IS_INVARIANT(uv)) {
3338 if (!has_utf8 && uv > 255) {
3339 /* Might need to recode whatever we have accumulated so
3340 * far if it contains any chars variant in utf8 or
3343 SvCUR_set(sv, d - SvPVX_const(sv));
3346 /* See Note on sizing above. */
3347 sv_utf8_upgrade_flags_grow(sv,
3348 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3349 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3350 d = SvPVX(sv) + SvCUR(sv);
3355 d = (char*)uvuni_to_utf8((U8*)d, uv);
3356 if (PL_lex_inwhat == OP_TRANS &&
3357 PL_sublex_info.sub_op) {
3358 PL_sublex_info.sub_op->op_private |=
3359 (PL_lex_repl ? OPpTRANS_FROM_UTF
3363 if (uv > 255 && !dorange)
3364 native_range = FALSE;
3377 /* In a non-pattern \N must be a named character, like \N{LATIN
3378 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3379 * mean to match a non-newline. For non-patterns, named
3380 * characters are converted to their string equivalents. In
3381 * patterns, named characters are not converted to their
3382 * ultimate forms for the same reasons that other escapes
3383 * aren't. Instead, they are converted to the \N{U+...} form
3384 * to get the value from the charnames that is in effect right
3385 * now, while preserving the fact that it was a named character
3386 * so that the regex compiler knows this */
3388 /* This section of code doesn't generally use the
3389 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3390 * a close examination of this macro and determined it is a
3391 * no-op except on utfebcdic variant characters. Every
3392 * character generated by this that would normally need to be
3393 * enclosed by this macro is invariant, so the macro is not
3394 * needed, and would complicate use of copy(). XXX There are
3395 * other parts of this file where the macro is used
3396 * inconsistently, but are saved by it being a no-op */
3398 /* The structure of this section of code (besides checking for
3399 * errors and upgrading to utf8) is:
3400 * Further disambiguate between the two meanings of \N, and if
3401 * not a charname, go process it elsewhere
3402 * If of form \N{U+...}, pass it through if a pattern;
3403 * otherwise convert to utf8
3404 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3405 * pattern; otherwise convert to utf8 */
3407 /* Here, s points to the 'N'; the test below is guaranteed to
3408 * succeed if we are being called on a pattern as we already
3409 * know from a test above that the next character is a '{'.
3410 * On a non-pattern \N must mean 'named sequence, which
3411 * requires braces */
3414 yyerror("Missing braces on \\N{}");
3419 /* If there is no matching '}', it is an error. */
3420 if (! (e = strchr(s, '}'))) {
3421 if (! PL_lex_inpat) {
3422 yyerror("Missing right brace on \\N{}");
3424 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3429 /* Here it looks like a named character */
3431 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3432 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3433 | PERL_SCAN_DISALLOW_PREFIX;
3436 /* For \N{U+...}, the '...' is a unicode value even on
3437 * EBCDIC machines */
3438 s += 2; /* Skip to next char after the 'U+' */
3440 uv = grok_hex(s, &len, &flags, NULL);
3441 if (len == 0 || len != (STRLEN)(e - s)) {
3442 yyerror("Invalid hexadecimal number in \\N{U+...}");
3449 /* On non-EBCDIC platforms, pass through to the regex
3450 * compiler unchanged. The reason we evaluated the
3451 * number above is to make sure there wasn't a syntax
3452 * error. But on EBCDIC we convert to native so
3453 * downstream code can continue to assume it's native
3455 s -= 5; /* Include the '\N{U+' */
3457 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3460 (unsigned int) UNI_TO_NATIVE(uv));
3462 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3466 else { /* Not a pattern: convert the hex to string */
3468 /* If destination is not in utf8, unconditionally
3469 * recode it to be so. This is because \N{} implies
3470 * Unicode semantics, and scalars have to be in utf8
3471 * to guarantee those semantics */
3473 SvCUR_set(sv, d - SvPVX_const(sv));
3476 /* See Note on sizing above. */
3477 sv_utf8_upgrade_flags_grow(
3479 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3480 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3481 d = SvPVX(sv) + SvCUR(sv);
3485 /* Add the string to the output */
3486 if (UNI_IS_INVARIANT(uv)) {
3489 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3492 else /* Here is \N{NAME} but not \N{U+...}. */
3493 if ((res = get_and_check_backslash_N_name(s, e)))
3496 const char *str = SvPV_const(res, len);
3499 if (! len) { /* The name resolved to an empty string */
3500 Copy("\\N{}", d, 4, char);
3504 /* In order to not lose information for the regex
3505 * compiler, pass the result in the specially made
3506 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3507 * the code points in hex of each character
3508 * returned by charnames */
3510 const char *str_end = str + len;
3511 const STRLEN off = d - SvPVX_const(sv);
3513 if (! SvUTF8(res)) {
3514 /* For the non-UTF-8 case, we can determine the
3515 * exact length needed without having to parse
3516 * through the string. Each character takes up
3517 * 2 hex digits plus either a trailing dot or
3519 d = off + SvGROW(sv, off
3521 + 6 /* For the "\N{U+", and
3523 + (STRLEN)(send - e));
3524 Copy("\\N{U+", d, 5, char);
3526 while (str < str_end) {
3528 my_snprintf(hex_string, sizeof(hex_string),
3529 "%02X.", (U8) *str);
3530 Copy(hex_string, d, 3, char);
3534 d--; /* We will overwrite below the final
3535 dot with a right brace */
3538 STRLEN char_length; /* cur char's byte length */
3540 /* and the number of bytes after this is
3541 * translated into hex digits */
3542 STRLEN output_length;
3544 /* 2 hex per byte; 2 chars for '\N'; 2 chars
3545 * for max('U+', '.'); and 1 for NUL */
3546 char hex_string[2 * UTF8_MAXBYTES + 5];
3548 /* Get the first character of the result. */
3549 U32 uv = utf8n_to_uvuni((U8 *) str,
3553 /* Convert first code point to hex, including
3554 * the boiler plate before it. For all these,
3555 * we convert to native format so that
3556 * downstream code can continue to assume the
3557 * input is native */
3559 my_snprintf(hex_string, sizeof(hex_string),
3561 (unsigned int) UNI_TO_NATIVE(uv));
3563 /* Make sure there is enough space to hold it */
3564 d = off + SvGROW(sv, off
3566 + (STRLEN)(send - e)
3567 + 2); /* '}' + NUL */
3569 Copy(hex_string, d, output_length, char);
3572 /* For each subsequent character, append dot and
3573 * its ordinal in hex */
3574 while ((str += char_length) < str_end) {
3575 const STRLEN off = d - SvPVX_const(sv);
3576 U32 uv = utf8n_to_uvuni((U8 *) str,
3581 my_snprintf(hex_string,
3584 (unsigned int) UNI_TO_NATIVE(uv));
3586 d = off + SvGROW(sv, off
3588 + (STRLEN)(send - e)
3589 + 2); /* '}' + NUL */
3590 Copy(hex_string, d, output_length, char);
3595 *d++ = '}'; /* Done. Add the trailing brace */
3598 else { /* Here, not in a pattern. Convert the name to a
3601 /* If destination is not in utf8, unconditionally
3602 * recode it to be so. This is because \N{} implies
3603 * Unicode semantics, and scalars have to be in utf8
3604 * to guarantee those semantics */
3606 SvCUR_set(sv, d - SvPVX_const(sv));
3609 /* See Note on sizing above. */
3610 sv_utf8_upgrade_flags_grow(sv,
3611 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3612 len + (STRLEN)(send - s) + 1);
3613 d = SvPVX(sv) + SvCUR(sv);
3615 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3617 /* See Note on sizing above. (NOTE: SvCUR() is not
3618 * set correctly here). */
3619 const STRLEN off = d - SvPVX_const(sv);
3620 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3622 Copy(str, d, len, char);
3628 } /* End \N{NAME} */
3631 native_range = FALSE; /* \N{} is defined to be Unicode */
3633 s = e + 1; /* Point to just after the '}' */
3636 /* \c is a control character */
3640 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3643 yyerror("Missing control char name in \\c");
3647 /* printf-style backslashes, formfeeds, newlines, etc */
3649 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3652 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3655 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3658 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3661 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3664 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3667 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3673 } /* end if (backslash) */
3680 /* If we started with encoded form, or already know we want it,
3681 then encode the next character */
3682 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3686 /* One might think that it is wasted effort in the case of the
3687 * source being utf8 (this_utf8 == TRUE) to take the next character
3688 * in the source, convert it to an unsigned value, and then convert
3689 * it back again. But the source has not been validated here. The
3690 * routine that does the conversion checks for errors like
3693 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3694 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3696 SvCUR_set(sv, d - SvPVX_const(sv));
3699 /* See Note on sizing above. */
3700 sv_utf8_upgrade_flags_grow(sv,
3701 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3702 need + (STRLEN)(send - s) + 1);
3703 d = SvPVX(sv) + SvCUR(sv);
3705 } else if (need > len) {
3706 /* encoded value larger than old, may need extra space (NOTE:
3707 * SvCUR() is not set correctly here). See Note on sizing
3709 const STRLEN off = d - SvPVX_const(sv);
3710 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3714 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3716 if (uv > 255 && !dorange)
3717 native_range = FALSE;
3721 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3723 } /* while loop to process each character */
3725 /* terminate the string and set up the sv */
3727 SvCUR_set(sv, d - SvPVX_const(sv));
3728 if (SvCUR(sv) >= SvLEN(sv))
3729 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3730 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3733 if (PL_encoding && !has_utf8) {
3734 sv_recode_to_utf8(sv, PL_encoding);
3740 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3741 PL_sublex_info.sub_op->op_private |=
3742 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3746 /* shrink the sv if we allocated more than we used */
3747 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3748 SvPV_shrink_to_cur(sv);
3751 /* return the substring (via pl_yylval) only if we parsed anything */
3752 if (s > PL_bufptr) {
3753 SvREFCNT_inc_simple_void_NN(sv);
3754 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3755 const char *const key = PL_lex_inpat ? "qr" : "q";
3756 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3760 if (PL_lex_inwhat == OP_TRANS) {
3763 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3766 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3774 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3777 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3779 LEAVE_with_name("scan_const");
3784 * Returns TRUE if there's more to the expression (e.g., a subscript),
3787 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3789 * ->[ and ->{ return TRUE
3790 * { and [ outside a pattern are always subscripts, so return TRUE
3791 * if we're outside a pattern and it's not { or [, then return FALSE
3792 * if we're in a pattern and the first char is a {
3793 * {4,5} (any digits around the comma) returns FALSE
3794 * if we're in a pattern and the first char is a [
3796 * [SOMETHING] has a funky algorithm to decide whether it's a
3797 * character class or not. It has to deal with things like
3798 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3799 * anything else returns TRUE
3802 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3805 S_intuit_more(pTHX_ char *s)
3809 PERL_ARGS_ASSERT_INTUIT_MORE;
3811 if (PL_lex_brackets)
3813 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3815 if (*s != '{' && *s != '[')
3820 /* In a pattern, so maybe we have {n,m}. */
3822 if (regcurly(s, FALSE)) {
3828 /* On the other hand, maybe we have a character class */
3831 if (*s == ']' || *s == '^')
3834 /* this is terrifying, and it works */
3837 const char * const send = strchr(s,']');
3838 unsigned char un_char, last_un_char;
3839 char tmpbuf[sizeof PL_tokenbuf * 4];
3841 if (!send) /* has to be an expression */
3843 weight = 2; /* let's weigh the evidence */
3847 else if (isDIGIT(*s)) {
3849 if (isDIGIT(s[1]) && s[2] == ']')
3855 Zero(seen,256,char);
3857 for (; s < send; s++) {
3858 last_un_char = un_char;
3859 un_char = (unsigned char)*s;
3864 weight -= seen[un_char] * 10;
3865 if (isWORDCHAR_lazy_if(s+1,UTF)) {
3867 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3868 len = (int)strlen(tmpbuf);
3869 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3870 UTF ? SVf_UTF8 : 0, SVt_PV))
3875 else if (*s == '$' && s[1] &&
3876 strchr("[#!%*<>()-=",s[1])) {
3877 if (/*{*/ strchr("])} =",s[2]))
3886 if (strchr("wds]",s[1]))
3888 else if (seen[(U8)'\''] || seen[(U8)'"'])
3890 else if (strchr("rnftbxcav",s[1]))
3892 else if (isDIGIT(s[1])) {
3894 while (s[1] && isDIGIT(s[1]))
3904 if (strchr("aA01! ",last_un_char))
3906 if (strchr("zZ79~",s[1]))
3908 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3909 weight -= 5; /* cope with negative subscript */
3912 if (!isWORDCHAR(last_un_char)
3913 && !(last_un_char == '$' || last_un_char == '@'
3914 || last_un_char == '&')
3915 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3920 if (keyword(tmpbuf, d - tmpbuf, 0))
3923 if (un_char == last_un_char + 1)
3925 weight -= seen[un_char];
3930 if (weight >= 0) /* probably a character class */
3940 * Does all the checking to disambiguate
3942 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3943 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3945 * First argument is the stuff after the first token, e.g. "bar".
3947 * Not a method if foo is a filehandle.
3948 * Not a method if foo is a subroutine prototyped to take a filehandle.
3949 * Not a method if it's really "Foo $bar"
3950 * Method if it's "foo $bar"
3951 * Not a method if it's really "print foo $bar"
3952 * Method if it's really "foo package::" (interpreted as package->foo)
3953 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3954 * Not a method if bar is a filehandle or package, but is quoted with
3959 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3962 char *s = start + (*start == '$');
3963 char tmpbuf[sizeof PL_tokenbuf];
3970 PERL_ARGS_ASSERT_INTUIT_METHOD;
3972 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3974 if (cv && SvPOK(cv)) {
3975 const char *proto = CvPROTO(cv);
3983 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3984 /* start is the beginning of the possible filehandle/object,
3985 * and s is the end of it
3986 * tmpbuf is a copy of it
3989 if (*start == '$') {
3990 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3991 isUPPER(*PL_tokenbuf))
3994 len = start - SvPVX(PL_linestr);
3998 start = SvPVX(PL_linestr) + len;
4002 return *s == '(' ? FUNCMETH : METHOD;
4004 if (!keyword(tmpbuf, len, 0)) {
4005 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
4009 soff = s - SvPVX(PL_linestr);
4013 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
4014 if (indirgv && GvCVu(indirgv))
4016 /* filehandle or package name makes it a method */
4017 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
4019 soff = s - SvPVX(PL_linestr);
4022 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
4023 return 0; /* no assumptions -- "=>" quotes bareword */
4025 start_force(PL_curforce);
4026 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
4027 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
4028 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
4030 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
4031 ( UTF ? SVf_UTF8 : 0 )));
4036 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
4038 return *s == '(' ? FUNCMETH : METHOD;
4044 /* Encoded script support. filter_add() effectively inserts a
4045 * 'pre-processing' function into the current source input stream.
4046 * Note that the filter function only applies to the current source file
4047 * (e.g., it will not affect files 'require'd or 'use'd by this one).
4049 * The datasv parameter (which may be NULL) can be used to pass
4050 * private data to this instance of the filter. The filter function
4051 * can recover the SV using the FILTER_DATA macro and use it to
4052 * store private buffers and state information.
4054 * The supplied datasv parameter is upgraded to a PVIO type
4055 * and the IoDIRP/IoANY field is used to store the function pointer,
4056 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
4057 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
4058 * private use must be set using malloc'd pointers.
4062 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
4071 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
4072 Perl_croak(aTHX_ "Source filters apply only to byte streams");
4074 if (!PL_rsfp_filters)
4075 PL_rsfp_filters = newAV();
4078 SvUPGRADE(datasv, SVt_PVIO);
4079 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
4080 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
4081 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
4082 FPTR2DPTR(void *, IoANY(datasv)),
4083 SvPV_nolen(datasv)));
4084 av_unshift(PL_rsfp_filters, 1);
4085 av_store(PL_rsfp_filters, 0, datasv) ;
4087 !PL_parser->filtered
4088 && PL_parser->lex_flags & LEX_EVALBYTES
4089 && PL_bufptr < PL_bufend
4091 const char *s = PL_bufptr;
4092 while (s < PL_bufend) {
4094 SV *linestr = PL_parser->linestr;
4095 char *buf = SvPVX(linestr);
4096 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
4097 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
4098 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
4099 STRLEN const linestart_pos = PL_parser->linestart - buf;
4100 STRLEN const last_uni_pos =
4101 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
4102 STRLEN const last_lop_pos =
4103 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
4104 av_push(PL_rsfp_filters, linestr);
4105 PL_parser->linestr =
4106 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
4107 buf = SvPVX(PL_parser->linestr);
4108 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
4109 PL_parser->bufptr = buf + bufptr_pos;
4110 PL_parser->oldbufptr = buf + oldbufptr_pos;
4111 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
4112 PL_parser->linestart = buf + linestart_pos;
4113 if (PL_parser->last_uni)
4114 PL_parser->last_uni = buf + last_uni_pos;
4115 if (PL_parser->last_lop)
4116 PL_parser->last_lop = buf + last_lop_pos;
4117 SvLEN(linestr) = SvCUR(linestr);
4118 SvCUR(linestr) = s-SvPVX(linestr);
4119 PL_parser->filtered = 1;
4129 /* Delete most recently added instance of this filter function. */
4131 Perl_filter_del(pTHX_ filter_t funcp)
4136 PERL_ARGS_ASSERT_FILTER_DEL;
4139 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
4140 FPTR2DPTR(void*, funcp)));
4142 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
4144 /* if filter is on top of stack (usual case) just pop it off */
4145 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
4146 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
4147 sv_free(av_pop(PL_rsfp_filters));
4151 /* we need to search for the correct entry and clear it */
4152 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
4156 /* Invoke the idxth filter function for the current rsfp. */
4157 /* maxlen 0 = read one text line */
4159 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4164 /* This API is bad. It should have been using unsigned int for maxlen.
4165 Not sure if we want to change the API, but if not we should sanity
4166 check the value here. */
4167 unsigned int correct_length
4176 PERL_ARGS_ASSERT_FILTER_READ;
4178 if (!PL_parser || !PL_rsfp_filters)
4180 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4181 /* Provide a default input filter to make life easy. */
4182 /* Note that we append to the line. This is handy. */
4183 DEBUG_P(PerlIO_printf(Perl_debug_log,
4184 "filter_read %d: from rsfp\n", idx));
4185 if (correct_length) {
4188 const int old_len = SvCUR(buf_sv);
4190 /* ensure buf_sv is large enough */
4191 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4192 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4193 correct_length)) <= 0) {
4194 if (PerlIO_error(PL_rsfp))
4195 return -1; /* error */
4197 return 0 ; /* end of file */
4199 SvCUR_set(buf_sv, old_len + len) ;
4200 SvPVX(buf_sv)[old_len + len] = '\0';
4203 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4204 if (PerlIO_error(PL_rsfp))
4205 return -1; /* error */
4207 return 0 ; /* end of file */
4210 return SvCUR(buf_sv);
4212 /* Skip this filter slot if filter has been deleted */
4213 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4214 DEBUG_P(PerlIO_printf(Perl_debug_log,
4215 "filter_read %d: skipped (filter deleted)\n",
4217 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4219 if (SvTYPE(datasv) != SVt_PVIO) {
4220 if (correct_length) {
4222 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4223 if (!remainder) return 0; /* eof */
4224 if (correct_length > remainder) correct_length = remainder;
4225 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4226 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4229 const char *s = SvEND(datasv);
4230 const char *send = SvPVX(datasv) + SvLEN(datasv);
4238 if (s == send) return 0; /* eof */
4239 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4240 SvCUR_set(datasv, s-SvPVX(datasv));
4242 return SvCUR(buf_sv);
4244 /* Get function pointer hidden within datasv */
4245 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4246 DEBUG_P(PerlIO_printf(Perl_debug_log,
4247 "filter_read %d: via function %p (%s)\n",
4248 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4249 /* Call function. The function is expected to */
4250 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4251 /* Return: <0:error, =0:eof, >0:not eof */
4252 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4256 S_filter_gets(pTHX_ SV *sv, STRLEN append)
4260 PERL_ARGS_ASSERT_FILTER_GETS;
4262 #ifdef PERL_CR_FILTER
4263 if (!PL_rsfp_filters) {
4264 filter_add(S_cr_textfilter,NULL);
4267 if (PL_rsfp_filters) {
4269 SvCUR_set(sv, 0); /* start with empty line */
4270 if (FILTER_READ(0, sv, 0) > 0)
4271 return ( SvPVX(sv) ) ;
4276 return (sv_gets(sv, PL_rsfp, append));
4280 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4285 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4287 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4291 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4292 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4294 return GvHV(gv); /* Foo:: */
4297 /* use constant CLASS => 'MyClass' */
4298 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4299 if (gv && GvCV(gv)) {
4300 SV * const sv = cv_const_sv(GvCV(gv));
4302 pkgname = SvPV_const(sv, len);
4305 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4309 * S_readpipe_override
4310 * Check whether readpipe() is overridden, and generates the appropriate
4311 * optree, provided sublex_start() is called afterwards.
4314 S_readpipe_override(pTHX)
4317 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4318 pl_yylval.ival = OP_BACKTICK;
4320 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4322 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4323 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4324 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4326 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4327 op_append_elem(OP_LIST,
4328 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4329 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4336 * The intent of this yylex wrapper is to minimize the changes to the
4337 * tokener when we aren't interested in collecting madprops. It remains
4338 * to be seen how successful this strategy will be...
4345 char *s = PL_bufptr;
4347 /* make sure PL_thiswhite is initialized */
4351 /* previous token ate up our whitespace? */
4352 if (!PL_lasttoke && PL_nextwhite) {
4353 PL_thiswhite = PL_nextwhite;
4357 /* isolate the token, and figure out where it is without whitespace */
4358 PL_realtokenstart = -1;
4362 assert(PL_curforce < 0);
4364 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4365 if (!PL_thistoken) {
4366 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4367 PL_thistoken = newSVpvs("");
4369 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4370 PL_thistoken = newSVpvn(tstart, s - tstart);
4373 if (PL_thismad) /* install head */
4374 CURMAD('X', PL_thistoken);
4377 /* last whitespace of a sublex? */
4378 if (optype == ')' && PL_endwhite) {
4379 CURMAD('X', PL_endwhite);
4384 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4385 if (!PL_thiswhite && !PL_endwhite && !optype) {
4386 sv_free(PL_thistoken);
4391 /* put off final whitespace till peg */
4392 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4393 PL_nextwhite = PL_thiswhite;
4396 else if (PL_thisopen) {
4397 CURMAD('q', PL_thisopen);
4399 sv_free(PL_thistoken);
4403 /* Store actual token text as madprop X */
4404 CURMAD('X', PL_thistoken);
4408 /* add preceding whitespace as madprop _ */
4409 CURMAD('_', PL_thiswhite);
4413 /* add quoted material as madprop = */
4414 CURMAD('=', PL_thisstuff);
4418 /* add terminating quote as madprop Q */
4419 CURMAD('Q', PL_thisclose);
4423 /* special processing based on optype */
4427 /* opval doesn't need a TOKEN since it can already store mp */
4437 if (pl_yylval.opval)
4438 append_madprops(PL_thismad, pl_yylval.opval, 0);
4446 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4459 /* remember any fake bracket that lexer is about to discard */
4460 if (PL_lex_brackets == 1 &&
4461 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4464 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4467 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4468 addmad(newMADsv('#', PL_th