This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Upgrade to Test-Simple-0.97_01
[perl5.git] / toke.c
CommitLineData
a0d0e21e 1/* toke.c
a687059c 2 *
1129b882
NC
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
a687059c 5 *
d48672a2
LW
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.
378cc40b 8 *
a0d0e21e
LW
9 */
10
11/*
4ac71550
TC
12 * 'It all comes from here, the stench and the peril.' --Frodo
13 *
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
378cc40b
LW
15 */
16
9cbb5ea2
GS
17/*
18 * This file is the lexer for Perl. It's closely linked to the
4e553d73 19 * parser, perly.y.
ffb4593c
NT
20 *
21 * The main routine is yylex(), which returns the next token.
22 */
23
f0e67a1d
Z
24/*
25=head1 Lexer interface
26
27This is the lower layer of the Perl parser, managing characters and tokens.
28
29=for apidoc AmU|yy_parser *|PL_parser
30
31Pointer to a structure encapsulating the state of the parsing operation
32currently in progress. The pointer can be locally changed to perform
33a nested parse without interfering with the state of an outer parse.
34Individual members of C<PL_parser> have their own documentation.
35
36=cut
37*/
38
378cc40b 39#include "EXTERN.h"
864dbfa3 40#define PERL_IN_TOKE_C
378cc40b 41#include "perl.h"
378cc40b 42
eb0d8d16
NC
43#define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
45
6154021b 46#define pl_yylval (PL_parser->yylval)
d3b6f988 47
acdf0a21
DM
48/* YYINITDEPTH -- initial size of the parser's stacks. */
49#define YYINITDEPTH 200
50
199e78b7
DM
51/* XXX temporary backwards compatibility */
52#define PL_lex_brackets (PL_parser->lex_brackets)
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_pending_ident (PL_parser->pending_ident)
70#define PL_preambled (PL_parser->preambled)
71#define PL_sublex_info (PL_parser->sublex_info)
bdc0bf6f 72#define PL_linestr (PL_parser->linestr)
c2598295
DM
73#define PL_expect (PL_parser->expect)
74#define PL_copline (PL_parser->copline)
f06b5848
DM
75#define PL_bufptr (PL_parser->bufptr)
76#define PL_oldbufptr (PL_parser->oldbufptr)
77#define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78#define PL_linestart (PL_parser->linestart)
79#define PL_bufend (PL_parser->bufend)
80#define PL_last_uni (PL_parser->last_uni)
81#define PL_last_lop (PL_parser->last_lop)
82#define PL_last_lop_op (PL_parser->last_lop_op)
bc177e6b 83#define PL_lex_state (PL_parser->lex_state)
2f9285f8 84#define PL_rsfp (PL_parser->rsfp)
5486870f 85#define PL_rsfp_filters (PL_parser->rsfp_filters)
12bd6ede
DM
86#define PL_in_my (PL_parser->in_my)
87#define PL_in_my_stash (PL_parser->in_my_stash)
14047fc9 88#define PL_tokenbuf (PL_parser->tokenbuf)
670a9cb2 89#define PL_multi_end (PL_parser->multi_end)
13765c85 90#define PL_error_count (PL_parser->error_count)
199e78b7
DM
91
92#ifdef PERL_MAD
93# define PL_endwhite (PL_parser->endwhite)
94# define PL_faketokens (PL_parser->faketokens)
95# define PL_lasttoke (PL_parser->lasttoke)
96# define PL_nextwhite (PL_parser->nextwhite)
97# define PL_realtokenstart (PL_parser->realtokenstart)
98# define PL_skipwhite (PL_parser->skipwhite)
99# define PL_thisclose (PL_parser->thisclose)
100# define PL_thismad (PL_parser->thismad)
101# define PL_thisopen (PL_parser->thisopen)
102# define PL_thisstuff (PL_parser->thisstuff)
103# define PL_thistoken (PL_parser->thistoken)
104# define PL_thiswhite (PL_parser->thiswhite)
fb205e7a
DM
105# define PL_thiswhite (PL_parser->thiswhite)
106# define PL_nexttoke (PL_parser->nexttoke)
107# define PL_curforce (PL_parser->curforce)
108#else
109# define PL_nexttoke (PL_parser->nexttoke)
110# define PL_nexttype (PL_parser->nexttype)
111# define PL_nextval (PL_parser->nextval)
199e78b7
DM
112#endif
113
16173588
NC
114/* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
3cbf51f5
DM
116static int
117S_pending_ident(pTHX);
199e78b7 118
0bd48802 119static const char ident_too_long[] = "Identifier too long";
8903cb82 120
29595ff2 121#ifdef PERL_MAD
29595ff2 122# define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
cd81e915 123# define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
9ded7720 124#else
5db06880 125# define CURMAD(slot,sv)
9ded7720 126# define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
29595ff2
NC
127#endif
128
9059aa12
LW
129#define XFAKEBRACK 128
130#define XENUMMASK 127
131
39e02b42
JH
132#ifdef USE_UTF8_SCRIPTS
133# define UTF (!IN_BYTES)
2b9d42f0 134#else
746b446a 135# define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
2b9d42f0 136#endif
a0ed51b3 137
b1fc3636
CJ
138/* The maximum number of characters preceding the unrecognized one to display */
139#define UNRECOGNIZED_PRECEDE_COUNT 10
140
61f0cdd9 141/* In variables named $^X, these are the legal values for X.
2b92dfce
GS
142 * 1999-02-27 mjd-perl-patch@plover.com */
143#define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
144
bf4acbe4 145#define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
bf4acbe4 146
ffb4593c
NT
147/* LEX_* are values for PL_lex_state, the state of the lexer.
148 * They are arranged oddly so that the guard on the switch statement
79072805
LW
149 * can get by with a single comparison (if the compiler is smart enough).
150 */
151
fb73857a
PP
152/* #define LEX_NOTPARSING 11 is done in perl.h. */
153
b6007c36
DM
154#define LEX_NORMAL 10 /* normal code (ie not within "...") */
155#define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
156#define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
157#define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
158#define LEX_INTERPSTART 6 /* expecting the start of a $var */
159
160 /* at end of code, eg "$x" followed by: */
161#define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
162#define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
163
164#define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
165 string or after \E, $foo, etc */
166#define LEX_INTERPCONST 2 /* NOT USED */
167#define LEX_FORMLINE 1 /* expecting a format line */
168#define LEX_KNOWNEXT 0 /* next token known; just return it */
169
79072805 170
bbf60fe6 171#ifdef DEBUGGING
27da23d5 172static const char* const lex_state_names[] = {
bbf60fe6
DM
173 "KNOWNEXT",
174 "FORMLINE",
175 "INTERPCONST",
176 "INTERPCONCAT",
177 "INTERPENDMAYBE",
178 "INTERPEND",
179 "INTERPSTART",
180 "INTERPPUSH",
181 "INTERPCASEMOD",
182 "INTERPNORMAL",
183 "NORMAL"
184};
185#endif
186
79072805
LW
187#ifdef ff_next
188#undef ff_next
d48672a2
LW
189#endif
190
79072805 191#include "keywords.h"
fe14fcc3 192
ffb4593c
NT
193/* CLINE is a macro that ensures PL_copline has a sane value */
194
ae986130
LW
195#ifdef CLINE
196#undef CLINE
197#endif
57843af0 198#define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
3280af22 199
5db06880 200#ifdef PERL_MAD
29595ff2
NC
201# define SKIPSPACE0(s) skipspace0(s)
202# define SKIPSPACE1(s) skipspace1(s)
203# define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
204# define PEEKSPACE(s) skipspace2(s,0)
205#else
206# define SKIPSPACE0(s) skipspace(s)
207# define SKIPSPACE1(s) skipspace(s)
208# define SKIPSPACE2(s,tsv) skipspace(s)
209# define PEEKSPACE(s) skipspace(s)
210#endif
211
ffb4593c
NT
212/*
213 * Convenience functions to return different tokens and prime the
9cbb5ea2 214 * lexer for the next token. They all take an argument.
ffb4593c
NT
215 *
216 * TOKEN : generic token (used for '(', DOLSHARP, etc)
217 * OPERATOR : generic operator
218 * AOPERATOR : assignment operator
219 * PREBLOCK : beginning the block after an if, while, foreach, ...
220 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
221 * PREREF : *EXPR where EXPR is not a simple identifier
222 * TERM : expression term
223 * LOOPX : loop exiting command (goto, last, dump, etc)
224 * FTST : file test operator
225 * FUN0 : zero-argument function
2d2e263d 226 * FUN1 : not used, except for not, which isn't a UNIOP
ffb4593c
NT
227 * BOop : bitwise or or xor
228 * BAop : bitwise and
229 * SHop : shift operator
230 * PWop : power operator
9cbb5ea2 231 * PMop : pattern-matching operator
ffb4593c
NT
232 * Aop : addition-level operator
233 * Mop : multiplication-level operator
234 * Eop : equality-testing operator
e5edeb50 235 * Rop : relational operator <= != gt
ffb4593c
NT
236 *
237 * Also see LOP and lop() below.
238 */
239
998054bd 240#ifdef DEBUGGING /* Serve -DT. */
704d4215 241# define REPORT(retval) tokereport((I32)retval, &pl_yylval)
998054bd 242#else
bbf60fe6 243# define REPORT(retval) (retval)
998054bd
SC
244#endif
245
bbf60fe6
DM
246#define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
247#define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
248#define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
249#define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
250#define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
251#define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
252#define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
6154021b
RGS
253#define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
254#define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
255#define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
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))
2f3197b3 266
a687059c
LW
267/* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
6f33ba73
RGS
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
a687059c 271 */
376fcdbf 272#define UNI2(f,x) { \
6154021b 273 pl_yylval.ival = f; \
376fcdbf
AL
274 PL_expect = x; \
275 PL_bufptr = s; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
278 if (*s == '(') \
279 return REPORT( (int)FUNC1 ); \
29595ff2 280 s = PEEKSPACE(s); \
376fcdbf
AL
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
282 }
6f33ba73
RGS
283#define UNI(f) UNI2(f,XTERM)
284#define UNIDOR(f) UNI2(f,XTERMORDORDOR)
a687059c 285
376fcdbf 286#define UNIBRACK(f) { \
6154021b 287 pl_yylval.ival = f; \
376fcdbf
AL
288 PL_bufptr = s; \
289 PL_last_uni = PL_oldbufptr; \
290 if (*s == '(') \
291 return REPORT( (int)FUNC1 ); \
29595ff2 292 s = PEEKSPACE(s); \
376fcdbf
AL
293 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
294 }
79072805 295
9f68db38 296/* grandfather return to old style */
6154021b 297#define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
79072805 298
8fa7f367
JH
299#ifdef DEBUGGING
300
6154021b 301/* how to interpret the pl_yylval associated with the token */
bbf60fe6
DM
302enum token_type {
303 TOKENTYPE_NONE,
304 TOKENTYPE_IVAL,
6154021b 305 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
bbf60fe6
DM
306 TOKENTYPE_PVAL,
307 TOKENTYPE_OPVAL,
308 TOKENTYPE_GVVAL
309};
310
6d4a66ac
NC
311static struct debug_tokens {
312 const int token;
313 enum token_type type;
314 const char *name;
315} const debug_tokens[] =
9041c2e3 316{
bbf60fe6
DM
317 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
318 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
319 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
320 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
321 { ARROW, TOKENTYPE_NONE, "ARROW" },
322 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
323 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
324 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
325 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
326 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
0d863452 327 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
bbf60fe6
DM
328 { DO, TOKENTYPE_NONE, "DO" },
329 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
330 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
331 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
332 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
333 { ELSE, TOKENTYPE_NONE, "ELSE" },
334 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
335 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
336 { FOR, TOKENTYPE_IVAL, "FOR" },
337 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
338 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
339 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
340 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
341 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
342 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
0d863452 343 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
bbf60fe6
DM
344 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
345 { IF, TOKENTYPE_IVAL, "IF" },
346 { LABEL, TOKENTYPE_PVAL, "LABEL" },
347 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
348 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
349 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
350 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
351 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
352 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
353 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
354 { MY, TOKENTYPE_IVAL, "MY" },
355 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
356 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
357 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
358 { OROP, TOKENTYPE_IVAL, "OROP" },
359 { OROR, TOKENTYPE_NONE, "OROR" },
360 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
88e1f1a2
JV
361 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
362 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
bbf60fe6
DM
363 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
364 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
365 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
366 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
367 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
368 { PREINC, TOKENTYPE_NONE, "PREINC" },
369 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
370 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
371 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
372 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
373 { SUB, TOKENTYPE_NONE, "SUB" },
374 { THING, TOKENTYPE_OPVAL, "THING" },
375 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
376 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
377 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
378 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
379 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
380 { USE, TOKENTYPE_IVAL, "USE" },
0d863452 381 { WHEN, TOKENTYPE_IVAL, "WHEN" },
bbf60fe6
DM
382 { WHILE, TOKENTYPE_IVAL, "WHILE" },
383 { WORD, TOKENTYPE_OPVAL, "WORD" },
be25f609 384 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
c35e046a 385 { 0, TOKENTYPE_NONE, NULL }
bbf60fe6
DM
386};
387
6154021b 388/* dump the returned token in rv, plus any optional arg in pl_yylval */
998054bd 389
bbf60fe6 390STATIC int
704d4215 391S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
bbf60fe6 392{
97aff369 393 dVAR;
7918f24d
NC
394
395 PERL_ARGS_ASSERT_TOKEREPORT;
396
bbf60fe6 397 if (DEBUG_T_TEST) {
bd61b366 398 const char *name = NULL;
bbf60fe6 399 enum token_type type = TOKENTYPE_NONE;
f54cb97a 400 const struct debug_tokens *p;
396482e1 401 SV* const report = newSVpvs("<== ");
bbf60fe6 402
f54cb97a 403 for (p = debug_tokens; p->token; p++) {
bbf60fe6
DM
404 if (p->token == (int)rv) {
405 name = p->name;
406 type = p->type;
407 break;
408 }
409 }
410 if (name)
54667de8 411 Perl_sv_catpv(aTHX_ report, name);
bbf60fe6
DM
412 else if ((char)rv > ' ' && (char)rv < '~')
413 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
414 else if (!rv)
396482e1 415 sv_catpvs(report, "EOF");
bbf60fe6
DM
416 else
417 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
418 switch (type) {
419 case TOKENTYPE_NONE:
420 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
421 break;
422 case TOKENTYPE_IVAL:
704d4215 423 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
bbf60fe6
DM
424 break;
425 case TOKENTYPE_OPNUM:
426 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
704d4215 427 PL_op_name[lvalp->ival]);
bbf60fe6
DM
428 break;
429 case TOKENTYPE_PVAL:
704d4215 430 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
bbf60fe6
DM
431 break;
432 case TOKENTYPE_OPVAL:
704d4215 433 if (lvalp->opval) {
401441c0 434 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
704d4215
GG
435 PL_op_name[lvalp->opval->op_type]);
436 if (lvalp->opval->op_type == OP_CONST) {
b6007c36 437 Perl_sv_catpvf(aTHX_ report, " %s",
704d4215 438 SvPEEK(cSVOPx_sv(lvalp->opval)));
b6007c36
DM
439 }
440
441 }
401441c0 442 else
396482e1 443 sv_catpvs(report, "(opval=null)");
bbf60fe6
DM
444 break;
445 }
b6007c36 446 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
bbf60fe6
DM
447 };
448 return (int)rv;
998054bd
SC
449}
450
b6007c36
DM
451
452/* print the buffer with suitable escapes */
453
454STATIC void
15f169a1 455S_printbuf(pTHX_ const char *const fmt, const char *const s)
b6007c36 456{
396482e1 457 SV* const tmp = newSVpvs("");
7918f24d
NC
458
459 PERL_ARGS_ASSERT_PRINTBUF;
460
b6007c36
DM
461 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
462 SvREFCNT_dec(tmp);
463}
464
8fa7f367
JH
465#endif
466
8290c323
NC
467static int
468S_deprecate_commaless_var_list(pTHX) {
469 PL_expect = XTERM;
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
472}
473
ffb4593c
NT
474/*
475 * S_ao
476 *
c963b151
BD
477 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
478 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
ffb4593c
NT
479 */
480
76e3520e 481STATIC int
cea2e8a9 482S_ao(pTHX_ int toketype)
a0d0e21e 483{
97aff369 484 dVAR;
3280af22
NIS
485 if (*PL_bufptr == '=') {
486 PL_bufptr++;
a0d0e21e 487 if (toketype == ANDAND)
6154021b 488 pl_yylval.ival = OP_ANDASSIGN;
a0d0e21e 489 else if (toketype == OROR)
6154021b 490 pl_yylval.ival = OP_ORASSIGN;
c963b151 491 else if (toketype == DORDOR)
6154021b 492 pl_yylval.ival = OP_DORASSIGN;
a0d0e21e
LW
493 toketype = ASSIGNOP;
494 }
495 return toketype;
496}
497
ffb4593c
NT
498/*
499 * S_no_op
500 * When Perl expects an operator and finds something else, no_op
501 * prints the warning. It always prints "<something> found where
502 * operator expected. It prints "Missing semicolon on previous line?"
503 * if the surprise occurs at the start of the line. "do you need to
504 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
505 * where the compiler doesn't know if foo is a method call or a function.
506 * It prints "Missing operator before end of line" if there's nothing
507 * after the missing operator, or "... before <...>" if there is something
508 * after the missing operator.
509 */
510
76e3520e 511STATIC void
15f169a1 512S_no_op(pTHX_ const char *const what, char *s)
463ee0b2 513{
97aff369 514 dVAR;
9d4ba2ae
AL
515 char * const oldbp = PL_bufptr;
516 const bool is_first = (PL_oldbufptr == PL_linestart);
68dc0745 517
7918f24d
NC
518 PERL_ARGS_ASSERT_NO_OP;
519
1189a94a
GS
520 if (!s)
521 s = oldbp;
07c798fb 522 else
1189a94a 523 PL_bufptr = s;
cea2e8a9 524 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
56da5a46
RGS
525 if (ckWARN_d(WARN_SYNTAX)) {
526 if (is_first)
527 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
528 "\t(Missing semicolon on previous line?)\n");
529 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
f54cb97a 530 const char *t;
c35e046a
AL
531 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
532 NOOP;
56da5a46
RGS
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %.*s?)\n",
551405c4 536 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
56da5a46
RGS
537 }
538 else {
539 assert(s >= oldbp);
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
551405c4 541 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
56da5a46 542 }
07c798fb 543 }
3280af22 544 PL_bufptr = oldbp;
8990e307
LW
545}
546
ffb4593c
NT
547/*
548 * S_missingterm
549 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 550 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
553 * This is fatal.
554 */
555
76e3520e 556STATIC void
cea2e8a9 557S_missingterm(pTHX_ char *s)
8990e307 558{
97aff369 559 dVAR;
8990e307
LW
560 char tmpbuf[3];
561 char q;
562 if (s) {
9d4ba2ae 563 char * const nl = strrchr(s,'\n');
d2719217 564 if (nl)
8990e307
LW
565 *nl = '\0';
566 }
463559e7 567 else if (isCNTRL(PL_multi_close)) {
8990e307 568 *tmpbuf = '^';
585ec06d 569 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
570 tmpbuf[2] = '\0';
571 s = tmpbuf;
572 }
573 else {
eb160463 574 *tmpbuf = (char)PL_multi_close;
8990e307
LW
575 tmpbuf[1] = '\0';
576 s = tmpbuf;
577 }
578 q = strchr(s,'"') ? '\'' : '"';
cea2e8a9 579 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
463ee0b2 580}
79072805 581
ef89dcc3 582#define FEATURE_IS_ENABLED(name) \
0d863452 583 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
89529cee 584 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
4a731d7b 585/* The longest string we pass in. */
1863b879 586#define MAX_FEATURE_LEN (sizeof("unicode_strings")-1)
4a731d7b 587
0d863452
RH
588/*
589 * S_feature_is_enabled
590 * Check whether the named feature is enabled.
591 */
592STATIC bool
15f169a1 593S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 594{
97aff369 595 dVAR;
0d863452 596 HV * const hinthv = GvHV(PL_hintgv);
4a731d7b 597 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
598
599 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
600
4a731d7b
NC
601 assert(namelen <= MAX_FEATURE_LEN);
602 memcpy(&he_name[8], name, namelen);
d4c19fe8 603
7b9ef140 604 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
0d863452
RH
605}
606
ffb4593c 607/*
9cbb5ea2
GS
608 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
609 * utf16-to-utf8-reversed.
ffb4593c
NT
610 */
611
c39cd008
GS
612#ifdef PERL_CR_FILTER
613static void
614strip_return(SV *sv)
615{
95a20fc0 616 register const char *s = SvPVX_const(sv);
9d4ba2ae 617 register const char * const e = s + SvCUR(sv);
7918f24d
NC
618
619 PERL_ARGS_ASSERT_STRIP_RETURN;
620
c39cd008
GS
621 /* outer loop optimized to do nothing if there are no CR-LFs */
622 while (s < e) {
623 if (*s++ == '\r' && *s == '\n') {
624 /* hit a CR-LF, need to copy the rest */
625 register char *d = s - 1;
626 *d++ = *s++;
627 while (s < e) {
628 if (*s == '\r' && s[1] == '\n')
629 s++;
630 *d++ = *s++;
631 }
632 SvCUR(sv) -= s - d;
633 return;
634 }
635 }
636}
a868473f 637
76e3520e 638STATIC I32
c39cd008 639S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 640{
f54cb97a 641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
642 if (count > 0 && !maxlen)
643 strip_return(sv);
644 return count;
a868473f
NIS
645}
646#endif
647
199e78b7
DM
648
649
ffb4593c
NT
650/*
651 * Perl_lex_start
5486870f 652 *
e3abe207 653 * Create a parser object and initialise its parser and lexer fields
5486870f
DM
654 *
655 * rsfp is the opened file handle to read from (if any),
656 *
657 * line holds any initial content already read from the file (or in
658 * the case of no file, such as an eval, the whole contents);
659 *
660 * new_filter indicates that this is a new file and it shouldn't inherit
661 * the filters from the current parser (ie require).
ffb4593c
NT
662 */
663
a0d0e21e 664void
5486870f 665Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
79072805 666{
97aff369 667 dVAR;
6ef55633 668 const char *s = NULL;
8990e307 669 STRLEN len;
5486870f 670 yy_parser *parser, *oparser;
acdf0a21
DM
671
672 /* create and initialise a parser */
673
199e78b7 674 Newxz(parser, 1, yy_parser);
5486870f 675 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
676 PL_parser = parser;
677
678 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
679 parser->ps = parser->stack;
680 parser->stack_size = YYINITDEPTH;
681
682 parser->stack->state = 0;
683 parser->yyerrstatus = 0;
684 parser->yychar = YYEMPTY; /* Cause a token to be read. */
685
e3abe207
DM
686 /* on scope exit, free this parser and restore any outer one */
687 SAVEPARSER(parser);
7c4baf47 688 parser->saved_curcop = PL_curcop;
e3abe207 689
acdf0a21 690 /* initialise lexer state */
8990e307 691
fb205e7a
DM
692#ifdef PERL_MAD
693 parser->curforce = -1;
694#else
695 parser->nexttoke = 0;
696#endif
ca4cfd28 697 parser->error_count = oparser ? oparser->error_count : 0;
c2598295 698 parser->copline = NOLINE;
5afb0a62 699 parser->lex_state = LEX_NORMAL;
c2598295 700 parser->expect = XSTATE;
2f9285f8 701 parser->rsfp = rsfp;
56b27c9a 702 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
502c6561 703 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
2f9285f8 704
199e78b7
DM
705 Newx(parser->lex_brackstack, 120, char);
706 Newx(parser->lex_casestack, 12, char);
707 *parser->lex_casestack = '\0';
02b34bbe 708
10efb74f
NC
709 if (line) {
710 s = SvPV_const(line, len);
711 } else {
712 len = 0;
713 }
bdc0bf6f 714
10efb74f 715 if (!len) {
bdc0bf6f 716 parser->linestr = newSVpvs("\n;");
3e5c0189 717 } else if (SvREADONLY(line) || s[len-1] != ';' || !SvPOK(line)) {
719a9bb0
NC
718 /* avoid tie/overload weirdness */
719 parser->linestr = newSVpvn_flags(s, len, SvUTF8(line));
10efb74f 720 if (s[len-1] != ';')
bdc0bf6f 721 sv_catpvs(parser->linestr, "\n;");
6c5ce11d
NC
722 } else {
723 SvTEMP_off(line);
724 SvREFCNT_inc_simple_void_NN(line);
bdc0bf6f 725 parser->linestr = line;
8990e307 726 }
f06b5848
DM
727 parser->oldoldbufptr =
728 parser->oldbufptr =
729 parser->bufptr =
730 parser->linestart = SvPVX(parser->linestr);
731 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
732 parser->last_lop = parser->last_uni = NULL;
79072805 733}
a687059c 734
e3abe207
DM
735
736/* delete a parser object */
737
738void
739Perl_parser_free(pTHX_ const yy_parser *parser)
740{
7918f24d
NC
741 PERL_ARGS_ASSERT_PARSER_FREE;
742
7c4baf47 743 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
744 SvREFCNT_dec(parser->linestr);
745
2f9285f8
DM
746 if (parser->rsfp == PerlIO_stdin())
747 PerlIO_clearerr(parser->rsfp);
799361c3
SH
748 else if (parser->rsfp && (!parser->old_parser ||
749 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 750 PerlIO_close(parser->rsfp);
5486870f 751 SvREFCNT_dec(parser->rsfp_filters);
2f9285f8 752
e3abe207
DM
753 Safefree(parser->stack);
754 Safefree(parser->lex_brackstack);
755 Safefree(parser->lex_casestack);
756 PL_parser = parser->old_parser;
757 Safefree(parser);
758}
759
760
ffb4593c
NT
761/*
762 * Perl_lex_end
9cbb5ea2
GS
763 * Finalizer for lexing operations. Must be called when the parser is
764 * done with the lexer.
ffb4593c
NT
765 */
766
463ee0b2 767void
864dbfa3 768Perl_lex_end(pTHX)
463ee0b2 769{
97aff369 770 dVAR;
3280af22 771 PL_doextract = FALSE;
463ee0b2
LW
772}
773
ffb4593c 774/*
f0e67a1d
Z
775=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
776
777Buffer scalar containing the chunk currently under consideration of the
778text currently being lexed. This is always a plain string scalar (for
779which C<SvPOK> is true). It is not intended to be used as a scalar by
780normal scalar means; instead refer to the buffer directly by the pointer
781variables described below.
782
783The lexer maintains various C<char*> pointers to things in the
784C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
785reallocated, all of these pointers must be updated. Don't attempt to
786do this manually, but rather use L</lex_grow_linestr> if you need to
787reallocate the buffer.
788
789The content of the text chunk in the buffer is commonly exactly one
790complete line of input, up to and including a newline terminator,
791but there are situations where it is otherwise. The octets of the
792buffer may be intended to be interpreted as either UTF-8 or Latin-1.
793The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
794flag on this scalar, which may disagree with it.
795
796For direct examination of the buffer, the variable
797L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
798lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
799of these pointers is usually preferable to examination of the scalar
800through normal scalar means.
801
802=for apidoc AmxU|char *|PL_parser-E<gt>bufend
803
804Direct pointer to the end of the chunk of text currently being lexed, the
805end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
806+ SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
807always located at the end of the buffer, and does not count as part of
808the buffer's contents.
809
810=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
811
812Points to the current position of lexing inside the lexer buffer.
813Characters around this point may be freely examined, within
814the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
815L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
816interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
817
818Lexing code (whether in the Perl core or not) moves this pointer past
819the characters that it consumes. It is also expected to perform some
820bookkeeping whenever a newline character is consumed. This movement
821can be more conveniently performed by the function L</lex_read_to>,
822which handles newlines appropriately.
823
824Interpretation of the buffer's octets can be abstracted out by
825using the slightly higher-level functions L</lex_peek_unichar> and
826L</lex_read_unichar>.
827
828=for apidoc AmxU|char *|PL_parser-E<gt>linestart
829
830Points to the start of the current line inside the lexer buffer.
831This is useful for indicating at which column an error occurred, and
832not much else. This must be updated by any lexing code that consumes
833a newline; the function L</lex_read_to> handles this detail.
834
835=cut
836*/
837
838/*
839=for apidoc Amx|bool|lex_bufutf8
840
841Indicates whether the octets in the lexer buffer
842(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
843of Unicode characters. If not, they should be interpreted as Latin-1
844characters. This is analogous to the C<SvUTF8> flag for scalars.
845
846In UTF-8 mode, it is not guaranteed that the lexer buffer actually
847contains valid UTF-8. Lexing code must be robust in the face of invalid
848encoding.
849
850The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
851is significant, but not the whole story regarding the input character
852encoding. Normally, when a file is being read, the scalar contains octets
853and its C<SvUTF8> flag is off, but the octets should be interpreted as
854UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
855however, the scalar may have the C<SvUTF8> flag on, and in this case its
856octets should be interpreted as UTF-8 unless the C<use bytes> pragma
857is in effect. This logic may change in the future; use this function
858instead of implementing the logic yourself.
859
860=cut
861*/
862
863bool
864Perl_lex_bufutf8(pTHX)
865{
866 return UTF;
867}
868
869/*
870=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
871
872Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
873at least I<len> octets (including terminating NUL). Returns a
874pointer to the reallocated buffer. This is necessary before making
875any direct modification of the buffer that would increase its length.
876L</lex_stuff_pvn> provides a more convenient way to insert text into
877the buffer.
878
879Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
880this function updates all of the lexer's variables that point directly
881into the buffer.
882
883=cut
884*/
885
886char *
887Perl_lex_grow_linestr(pTHX_ STRLEN len)
888{
889 SV *linestr;
890 char *buf;
891 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
892 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
893 linestr = PL_parser->linestr;
894 buf = SvPVX(linestr);
895 if (len <= SvLEN(linestr))
896 return buf;
897 bufend_pos = PL_parser->bufend - buf;
898 bufptr_pos = PL_parser->bufptr - buf;
899 oldbufptr_pos = PL_parser->oldbufptr - buf;
900 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
901 linestart_pos = PL_parser->linestart - buf;
902 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
903 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
904 buf = sv_grow(linestr, len);
905 PL_parser->bufend = buf + bufend_pos;
906 PL_parser->bufptr = buf + bufptr_pos;
907 PL_parser->oldbufptr = buf + oldbufptr_pos;
908 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
909 PL_parser->linestart = buf + linestart_pos;
910 if (PL_parser->last_uni)
911 PL_parser->last_uni = buf + last_uni_pos;
912 if (PL_parser->last_lop)
913 PL_parser->last_lop = buf + last_lop_pos;
914 return buf;
915}
916
917/*
83aa740e 918=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
919
920Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
921immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
922reallocating the buffer if necessary. This means that lexing code that
923runs later will see the characters as if they had appeared in the input.
924It is not recommended to do this as part of normal parsing, and most
925uses of this facility run the risk of the inserted characters being
926interpreted in an unintended manner.
927
928The string to be inserted is represented by I<len> octets starting
929at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
930according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
931The characters are recoded for the lexer buffer, according to how the
932buffer is currently being interpreted (L</lex_bufutf8>). If a string
933to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
934function is more convenient.
935
936=cut
937*/
938
939void
83aa740e 940Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 941{
749123ff 942 dVAR;
f0e67a1d
Z
943 char *bufptr;
944 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
945 if (flags & ~(LEX_STUFF_UTF8))
946 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
947 if (UTF) {
948 if (flags & LEX_STUFF_UTF8) {
949 goto plain_copy;
950 } else {
951 STRLEN highhalf = 0;
83aa740e 952 const char *p, *e = pv+len;
f0e67a1d
Z
953 for (p = pv; p != e; p++)
954 highhalf += !!(((U8)*p) & 0x80);
955 if (!highhalf)
956 goto plain_copy;
957 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
958 bufptr = PL_parser->bufptr;
959 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
960 SvCUR_set(PL_parser->linestr,
961 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
962 PL_parser->bufend += len+highhalf;
963 for (p = pv; p != e; p++) {
964 U8 c = (U8)*p;
965 if (c & 0x80) {
966 *bufptr++ = (char)(0xc0 | (c >> 6));
967 *bufptr++ = (char)(0x80 | (c & 0x3f));
968 } else {
969 *bufptr++ = (char)c;
970 }
971 }
972 }
973 } else {
974 if (flags & LEX_STUFF_UTF8) {
975 STRLEN highhalf = 0;
83aa740e 976 const char *p, *e = pv+len;
f0e67a1d
Z
977 for (p = pv; p != e; p++) {
978 U8 c = (U8)*p;
979 if (c >= 0xc4) {
980 Perl_croak(aTHX_ "Lexing code attempted to stuff "
981 "non-Latin-1 character into Latin-1 input");
982 } else if (c >= 0xc2 && p+1 != e &&
983 (((U8)p[1]) & 0xc0) == 0x80) {
984 p++;
985 highhalf++;
986 } else if (c >= 0x80) {
987 /* malformed UTF-8 */
988 ENTER;
989 SAVESPTR(PL_warnhook);
990 PL_warnhook = PERL_WARNHOOK_FATAL;
991 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
992 LEAVE;
993 }
994 }
995 if (!highhalf)
996 goto plain_copy;
997 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
998 bufptr = PL_parser->bufptr;
999 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1000 SvCUR_set(PL_parser->linestr,
1001 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d
Z
1002 PL_parser->bufend += len-highhalf;
1003 for (p = pv; p != e; p++) {
1004 U8 c = (U8)*p;
1005 if (c & 0x80) {
1006 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1007 p++;
1008 } else {
1009 *bufptr++ = (char)c;
1010 }
1011 }
1012 } else {
1013 plain_copy:
1014 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1015 bufptr = PL_parser->bufptr;
1016 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1017 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1018 PL_parser->bufend += len;
1019 Copy(pv, bufptr, len, char);
1020 }
1021 }
1022}
1023
1024/*
1025=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1026
1027Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1028immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1029reallocating the buffer if necessary. This means that lexing code that
1030runs later will see the characters as if they had appeared in the input.
1031It is not recommended to do this as part of normal parsing, and most
1032uses of this facility run the risk of the inserted characters being
1033interpreted in an unintended manner.
1034
1035The string to be inserted is the string value of I<sv>. The characters
1036are recoded for the lexer buffer, according to how the buffer is currently
1037being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1038not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1039need to construct a scalar.
1040
1041=cut
1042*/
1043
1044void
1045Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1046{
1047 char *pv;
1048 STRLEN len;
1049 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1050 if (flags)
1051 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1052 pv = SvPV(sv, len);
1053 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1054}
1055
1056/*
1057=for apidoc Amx|void|lex_unstuff|char *ptr
1058
1059Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1060I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1061This hides the discarded text from any lexing code that runs later,
1062as if the text had never appeared.
1063
1064This is not the normal way to consume lexed text. For that, use
1065L</lex_read_to>.
1066
1067=cut
1068*/
1069
1070void
1071Perl_lex_unstuff(pTHX_ char *ptr)
1072{
1073 char *buf, *bufend;
1074 STRLEN unstuff_len;
1075 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1076 buf = PL_parser->bufptr;
1077 if (ptr < buf)
1078 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1079 if (ptr == buf)
1080 return;
1081 bufend = PL_parser->bufend;
1082 if (ptr > bufend)
1083 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1084 unstuff_len = ptr - buf;
1085 Move(ptr, buf, bufend+1-ptr, char);
1086 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1087 PL_parser->bufend = bufend - unstuff_len;
1088}
1089
1090/*
1091=for apidoc Amx|void|lex_read_to|char *ptr
1092
1093Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1094to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1095performing the correct bookkeeping whenever a newline character is passed.
1096This is the normal way to consume lexed text.
1097
1098Interpretation of the buffer's octets can be abstracted out by
1099using the slightly higher-level functions L</lex_peek_unichar> and
1100L</lex_read_unichar>.
1101
1102=cut
1103*/
1104
1105void
1106Perl_lex_read_to(pTHX_ char *ptr)
1107{
1108 char *s;
1109 PERL_ARGS_ASSERT_LEX_READ_TO;
1110 s = PL_parser->bufptr;
1111 if (ptr < s || ptr > PL_parser->bufend)
1112 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1113 for (; s != ptr; s++)
1114 if (*s == '\n') {
1115 CopLINE_inc(PL_curcop);
1116 PL_parser->linestart = s+1;
1117 }
1118 PL_parser->bufptr = ptr;
1119}
1120
1121/*
1122=for apidoc Amx|void|lex_discard_to|char *ptr
1123
1124Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1125up to I<ptr>. The remaining content of the buffer will be moved, and
1126all pointers into the buffer updated appropriately. I<ptr> must not
1127be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1128it is not permitted to discard text that has yet to be lexed.
1129
1130Normally it is not necessarily to do this directly, because it suffices to
1131use the implicit discarding behaviour of L</lex_next_chunk> and things
1132based on it. However, if a token stretches across multiple lines,
1f317c95 1133and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1134that purpose, then after completion of the token it would be wise to
1135explicitly discard the now-unneeded earlier lines, to avoid future
1136multi-line tokens growing the buffer without bound.
1137
1138=cut
1139*/
1140
1141void
1142Perl_lex_discard_to(pTHX_ char *ptr)
1143{
1144 char *buf;
1145 STRLEN discard_len;
1146 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1147 buf = SvPVX(PL_parser->linestr);
1148 if (ptr < buf)
1149 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1150 if (ptr == buf)
1151 return;
1152 if (ptr > PL_parser->bufptr)
1153 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1154 discard_len = ptr - buf;
1155 if (PL_parser->oldbufptr < ptr)
1156 PL_parser->oldbufptr = ptr;
1157 if (PL_parser->oldoldbufptr < ptr)
1158 PL_parser->oldoldbufptr = ptr;
1159 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1160 PL_parser->last_uni = NULL;
1161 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1162 PL_parser->last_lop = NULL;
1163 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1164 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1165 PL_parser->bufend -= discard_len;
1166 PL_parser->bufptr -= discard_len;
1167 PL_parser->oldbufptr -= discard_len;
1168 PL_parser->oldoldbufptr -= discard_len;
1169 if (PL_parser->last_uni)
1170 PL_parser->last_uni -= discard_len;
1171 if (PL_parser->last_lop)
1172 PL_parser->last_lop -= discard_len;
1173}
1174
1175/*
1176=for apidoc Amx|bool|lex_next_chunk|U32 flags
1177
1178Reads in the next chunk of text to be lexed, appending it to
1179L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1180looked to the end of the current chunk and wants to know more. It is
1181usual, but not necessary, for lexing to have consumed the entirety of
1182the current chunk at this time.
1183
1184If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1185chunk (i.e., the current chunk has been entirely consumed), normally the
1186current chunk will be discarded at the same time that the new chunk is
1187read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1188will not be discarded. If the current chunk has not been entirely
1189consumed, then it will not be discarded regardless of the flag.
1190
1191Returns true if some new text was added to the buffer, or false if the
1192buffer has reached the end of the input text.
1193
1194=cut
1195*/
1196
1197#define LEX_FAKE_EOF 0x80000000
1198
1199bool
1200Perl_lex_next_chunk(pTHX_ U32 flags)
1201{
1202 SV *linestr;
1203 char *buf;
1204 STRLEN old_bufend_pos, new_bufend_pos;
1205 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1206 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1207 bool got_some_for_debugger = 0;
f0e67a1d
Z
1208 bool got_some;
1209 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1210 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
f0e67a1d
Z
1211 linestr = PL_parser->linestr;
1212 buf = SvPVX(linestr);
1213 if (!(flags & LEX_KEEP_PREVIOUS) &&
1214 PL_parser->bufptr == PL_parser->bufend) {
1215 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1216 linestart_pos = 0;
1217 if (PL_parser->last_uni != PL_parser->bufend)
1218 PL_parser->last_uni = NULL;
1219 if (PL_parser->last_lop != PL_parser->bufend)
1220 PL_parser->last_lop = NULL;
1221 last_uni_pos = last_lop_pos = 0;
1222 *buf = 0;
1223 SvCUR(linestr) = 0;
1224 } else {
1225 old_bufend_pos = PL_parser->bufend - buf;
1226 bufptr_pos = PL_parser->bufptr - buf;
1227 oldbufptr_pos = PL_parser->oldbufptr - buf;
1228 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1229 linestart_pos = PL_parser->linestart - buf;
1230 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1231 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1232 }
1233 if (flags & LEX_FAKE_EOF) {
1234 goto eof;
1235 } else if (!PL_parser->rsfp) {
1236 got_some = 0;
1237 } else if (filter_gets(linestr, old_bufend_pos)) {
1238 got_some = 1;
17cc9359 1239 got_some_for_debugger = 1;
f0e67a1d 1240 } else {
580561a3
Z
1241 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1242 sv_setpvs(linestr, "");
f0e67a1d
Z
1243 eof:
1244 /* End of real input. Close filehandle (unless it was STDIN),
1245 * then add implicit termination.
1246 */
1247 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1248 PerlIO_clearerr(PL_parser->rsfp);
1249 else if (PL_parser->rsfp)
1250 (void)PerlIO_close(PL_parser->rsfp);
1251 PL_parser->rsfp = NULL;
1252 PL_doextract = FALSE;
1253#ifdef PERL_MAD
1254 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1255 PL_faketokens = 1;
1256#endif
1257 if (!PL_in_eval && PL_minus_p) {
1258 sv_catpvs(linestr,
1259 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1260 PL_minus_n = PL_minus_p = 0;
1261 } else if (!PL_in_eval && PL_minus_n) {
1262 sv_catpvs(linestr, /*{*/";}");
1263 PL_minus_n = 0;
1264 } else
1265 sv_catpvs(linestr, ";");
1266 got_some = 1;
1267 }
1268 buf = SvPVX(linestr);
1269 new_bufend_pos = SvCUR(linestr);
1270 PL_parser->bufend = buf + new_bufend_pos;
1271 PL_parser->bufptr = buf + bufptr_pos;
1272 PL_parser->oldbufptr = buf + oldbufptr_pos;
1273 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1274 PL_parser->linestart = buf + linestart_pos;
1275 if (PL_parser->last_uni)
1276 PL_parser->last_uni = buf + last_uni_pos;
1277 if (PL_parser->last_lop)
1278 PL_parser->last_lop = buf + last_lop_pos;
17cc9359 1279 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
f0e67a1d
Z
1280 PL_curstash != PL_debstash) {
1281 /* debugger active and we're not compiling the debugger code,
1282 * so store the line into the debugger's array of lines
1283 */
1284 update_debugger_info(NULL, buf+old_bufend_pos,
1285 new_bufend_pos-old_bufend_pos);
1286 }
1287 return got_some;
1288}
1289
1290/*
1291=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1292
1293Looks ahead one (Unicode) character in the text currently being lexed.
1294Returns the codepoint (unsigned integer value) of the next character,
1295or -1 if lexing has reached the end of the input text. To consume the
1296peeked character, use L</lex_read_unichar>.
1297
1298If the next character is in (or extends into) the next chunk of input
1299text, the next chunk will be read in. Normally the current chunk will be
1300discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1301then the current chunk will not be discarded.
1302
1303If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1304is encountered, an exception is generated.
1305
1306=cut
1307*/
1308
1309I32
1310Perl_lex_peek_unichar(pTHX_ U32 flags)
1311{
749123ff 1312 dVAR;
f0e67a1d
Z
1313 char *s, *bufend;
1314 if (flags & ~(LEX_KEEP_PREVIOUS))
1315 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1316 s = PL_parser->bufptr;
1317 bufend = PL_parser->bufend;
1318 if (UTF) {
1319 U8 head;
1320 I32 unichar;
1321 STRLEN len, retlen;
1322 if (s == bufend) {
1323 if (!lex_next_chunk(flags))
1324 return -1;
1325 s = PL_parser->bufptr;
1326 bufend = PL_parser->bufend;
1327 }
1328 head = (U8)*s;
1329 if (!(head & 0x80))
1330 return head;
1331 if (head & 0x40) {
1332 len = PL_utf8skip[head];
1333 while ((STRLEN)(bufend-s) < len) {
1334 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1335 break;
1336 s = PL_parser->bufptr;
1337 bufend = PL_parser->bufend;
1338 }
1339 }
1340 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1341 if (retlen == (STRLEN)-1) {
1342 /* malformed UTF-8 */
1343 ENTER;
1344 SAVESPTR(PL_warnhook);
1345 PL_warnhook = PERL_WARNHOOK_FATAL;
1346 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1347 LEAVE;
1348 }
1349 return unichar;
1350 } else {
1351 if (s == bufend) {
1352 if (!lex_next_chunk(flags))
1353 return -1;
1354 s = PL_parser->bufptr;
1355 }
1356 return (U8)*s;
1357 }
1358}
1359
1360/*
1361=for apidoc Amx|I32|lex_read_unichar|U32 flags
1362
1363Reads the next (Unicode) character in the text currently being lexed.
1364Returns the codepoint (unsigned integer value) of the character read,
1365and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1366if lexing has reached the end of the input text. To non-destructively
1367examine the next character, use L</lex_peek_unichar> instead.
1368
1369If the next character is in (or extends into) the next chunk of input
1370text, the next chunk will be read in. Normally the current chunk will be
1371discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1372then the current chunk will not be discarded.
1373
1374If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1375is encountered, an exception is generated.
1376
1377=cut
1378*/
1379
1380I32
1381Perl_lex_read_unichar(pTHX_ U32 flags)
1382{
1383 I32 c;
1384 if (flags & ~(LEX_KEEP_PREVIOUS))
1385 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1386 c = lex_peek_unichar(flags);
1387 if (c != -1) {
1388 if (c == '\n')
1389 CopLINE_inc(PL_curcop);
1390 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1391 }
1392 return c;
1393}
1394
1395/*
1396=for apidoc Amx|void|lex_read_space|U32 flags
1397
1398Reads optional spaces, in Perl style, in the text currently being
1399lexed. The spaces may include ordinary whitespace characters and
1400Perl-style comments. C<#line> directives are processed if encountered.
1401L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1402at a non-space character (or the end of the input text).
1403
1404If spaces extend into the next chunk of input text, the next chunk will
1405be read in. Normally the current chunk will be discarded at the same
1406time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1407chunk will not be discarded.
1408
1409=cut
1410*/
1411
f0998909
Z
1412#define LEX_NO_NEXT_CHUNK 0x80000000
1413
f0e67a1d
Z
1414void
1415Perl_lex_read_space(pTHX_ U32 flags)
1416{
1417 char *s, *bufend;
1418 bool need_incline = 0;
f0998909 1419 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
f0e67a1d
Z
1420 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1421#ifdef PERL_MAD
1422 if (PL_skipwhite) {
1423 sv_free(PL_skipwhite);
1424 PL_skipwhite = NULL;
1425 }
1426 if (PL_madskills)
1427 PL_skipwhite = newSVpvs("");
1428#endif /* PERL_MAD */
1429 s = PL_parser->bufptr;
1430 bufend = PL_parser->bufend;
1431 while (1) {
1432 char c = *s;
1433 if (c == '#') {
1434 do {
1435 c = *++s;
1436 } while (!(c == '\n' || (c == 0 && s == bufend)));
1437 } else if (c == '\n') {
1438 s++;
1439 PL_parser->linestart = s;
1440 if (s == bufend)
1441 need_incline = 1;
1442 else
1443 incline(s);
1444 } else if (isSPACE(c)) {
1445 s++;
1446 } else if (c == 0 && s == bufend) {
1447 bool got_more;
1448#ifdef PERL_MAD
1449 if (PL_madskills)
1450 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1451#endif /* PERL_MAD */
f0998909
Z
1452 if (flags & LEX_NO_NEXT_CHUNK)
1453 break;
f0e67a1d
Z
1454 PL_parser->bufptr = s;
1455 CopLINE_inc(PL_curcop);
1456 got_more = lex_next_chunk(flags);
1457 CopLINE_dec(PL_curcop);
1458 s = PL_parser->bufptr;
1459 bufend = PL_parser->bufend;
1460 if (!got_more)
1461 break;
1462 if (need_incline && PL_parser->rsfp) {
1463 incline(s);
1464 need_incline = 0;
1465 }
1466 } else {
1467 break;
1468 }
1469 }
1470#ifdef PERL_MAD
1471 if (PL_madskills)
1472 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1473#endif /* PERL_MAD */
1474 PL_parser->bufptr = s;
1475}
1476
1477/*
ffb4593c
NT
1478 * S_incline
1479 * This subroutine has nothing to do with tilting, whether at windmills
1480 * or pinball tables. Its name is short for "increment line". It
57843af0 1481 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1482 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1483 * # line 500 "foo.pm"
1484 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1485 */
1486
76e3520e 1487STATIC void
d9095cec 1488S_incline(pTHX_ const char *s)
463ee0b2 1489{
97aff369 1490 dVAR;
d9095cec
NC
1491 const char *t;
1492 const char *n;
1493 const char *e;
463ee0b2 1494
7918f24d
NC
1495 PERL_ARGS_ASSERT_INCLINE;
1496
57843af0 1497 CopLINE_inc(PL_curcop);
463ee0b2
LW
1498 if (*s++ != '#')
1499 return;
d4c19fe8
AL
1500 while (SPACE_OR_TAB(*s))
1501 s++;
73659bf1
GS
1502 if (strnEQ(s, "line", 4))
1503 s += 4;
1504 else
1505 return;
084592ab 1506 if (SPACE_OR_TAB(*s))
73659bf1 1507 s++;
4e553d73 1508 else
73659bf1 1509 return;
d4c19fe8
AL
1510 while (SPACE_OR_TAB(*s))
1511 s++;
463ee0b2
LW
1512 if (!isDIGIT(*s))
1513 return;
d4c19fe8 1514
463ee0b2
LW
1515 n = s;
1516 while (isDIGIT(*s))
1517 s++;
07714eb4 1518 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1519 return;
bf4acbe4 1520 while (SPACE_OR_TAB(*s))
463ee0b2 1521 s++;
73659bf1 1522 if (*s == '"' && (t = strchr(s+1, '"'))) {
463ee0b2 1523 s++;
73659bf1
GS
1524 e = t + 1;
1525 }
463ee0b2 1526 else {
c35e046a
AL
1527 t = s;
1528 while (!isSPACE(*t))
1529 t++;
73659bf1 1530 e = t;
463ee0b2 1531 }
bf4acbe4 1532 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1533 e++;
1534 if (*e != '\n' && *e != '\0')
1535 return; /* false alarm */
1536
f4dd75d9 1537 if (t - s > 0) {
d9095cec 1538 const STRLEN len = t - s;
8a5ee598 1539#ifndef USE_ITHREADS
19bad673
NC
1540 SV *const temp_sv = CopFILESV(PL_curcop);
1541 const char *cf;
1542 STRLEN tmplen;
1543
1544 if (temp_sv) {
1545 cf = SvPVX(temp_sv);
1546 tmplen = SvCUR(temp_sv);
1547 } else {
1548 cf = NULL;
1549 tmplen = 0;
1550 }
1551
42d9b98d 1552 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
e66cf94c
RGS
1553 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1554 * to *{"::_<newfilename"} */
44867030
NC
1555 /* However, the long form of evals is only turned on by the
1556 debugger - usually they're "(eval %lu)" */
1557 char smallbuf[128];
1558 char *tmpbuf;
1559 GV **gvp;
d9095cec 1560 STRLEN tmplen2 = len;
798b63bc 1561 if (tmplen + 2 <= sizeof smallbuf)
e66cf94c
RGS
1562 tmpbuf = smallbuf;
1563 else
2ae0db35 1564 Newx(tmpbuf, tmplen + 2, char);
44867030
NC
1565 tmpbuf[0] = '_';
1566 tmpbuf[1] = '<';
2ae0db35 1567 memcpy(tmpbuf + 2, cf, tmplen);
44867030 1568 tmplen += 2;
8a5ee598
RGS
1569 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1570 if (gvp) {
44867030
NC
1571 char *tmpbuf2;
1572 GV *gv2;
1573
1574 if (tmplen2 + 2 <= sizeof smallbuf)
1575 tmpbuf2 = smallbuf;
1576 else
1577 Newx(tmpbuf2, tmplen2 + 2, char);
1578
1579 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1580 /* Either they malloc'd it, or we malloc'd it,
1581 so no prefix is present in ours. */
1582 tmpbuf2[0] = '_';
1583 tmpbuf2[1] = '<';
1584 }
1585
1586 memcpy(tmpbuf2 + 2, s, tmplen2);
1587 tmplen2 += 2;
1588
8a5ee598 1589 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1590 if (!isGV(gv2)) {
8a5ee598 1591 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1592 /* adjust ${"::_<newfilename"} to store the new file name */
1593 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
3cb1dbc6
NC
1594 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1595 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
e5527e4b 1596 }
44867030
NC
1597
1598 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
8a5ee598 1599 }
e66cf94c 1600 if (tmpbuf != smallbuf) Safefree(tmpbuf);
e66cf94c 1601 }
8a5ee598 1602#endif
05ec9bb3 1603 CopFILE_free(PL_curcop);
d9095cec 1604 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1605 }
57843af0 1606 CopLINE_set(PL_curcop, atoi(n)-1);
463ee0b2
LW
1607}
1608
29595ff2 1609#ifdef PERL_MAD
cd81e915 1610/* skip space before PL_thistoken */
29595ff2
NC
1611
1612STATIC char *
1613S_skipspace0(pTHX_ register char *s)
1614{
7918f24d
NC
1615 PERL_ARGS_ASSERT_SKIPSPACE0;
1616
29595ff2
NC
1617 s = skipspace(s);
1618 if (!PL_madskills)
1619 return s;
cd81e915
NC
1620 if (PL_skipwhite) {
1621 if (!PL_thiswhite)
6b29d1f5 1622 PL_thiswhite = newSVpvs("");
cd81e915
NC
1623 sv_catsv(PL_thiswhite, PL_skipwhite);
1624 sv_free(PL_skipwhite);
1625 PL_skipwhite = 0;
1626 }
1627 PL_realtokenstart = s - SvPVX(PL_linestr);
29595ff2
NC
1628 return s;
1629}
1630
cd81e915 1631/* skip space after PL_thistoken */
29595ff2
NC
1632
1633STATIC char *
1634S_skipspace1(pTHX_ register char *s)
1635{
d4c19fe8 1636 const char *start = s;
29595ff2
NC
1637 I32 startoff = start - SvPVX(PL_linestr);
1638
7918f24d
NC
1639 PERL_ARGS_ASSERT_SKIPSPACE1;
1640
29595ff2
NC
1641 s = skipspace(s);
1642 if (!PL_madskills)
1643 return s;
1644 start = SvPVX(PL_linestr) + startoff;
cd81e915 1645 if (!PL_thistoken && PL_realtokenstart >= 0) {
d4c19fe8 1646 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
cd81e915
NC
1647 PL_thistoken = newSVpvn(tstart, start - tstart);
1648 }
1649 PL_realtokenstart = -1;
1650 if (PL_skipwhite) {
1651 if (!PL_nextwhite)
6b29d1f5 1652 PL_nextwhite = newSVpvs("");
cd81e915
NC
1653 sv_catsv(PL_nextwhite, PL_skipwhite);
1654 sv_free(PL_skipwhite);
1655 PL_skipwhite = 0;
29595ff2
NC
1656 }
1657 return s;
1658}
1659
1660STATIC char *
1661S_skipspace2(pTHX_ register char *s, SV **svp)
1662{
c35e046a
AL
1663 char *start;
1664 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1665 const I32 startoff = s - SvPVX(PL_linestr);
1666
7918f24d
NC
1667 PERL_ARGS_ASSERT_SKIPSPACE2;
1668
29595ff2
NC
1669 s = skipspace(s);
1670 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1671 if (!PL_madskills || !svp)
1672 return s;
1673 start = SvPVX(PL_linestr) + startoff;
cd81e915 1674 if (!PL_thistoken && PL_realtokenstart >= 0) {
d4c19fe8 1675 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
cd81e915
NC
1676 PL_thistoken = newSVpvn(tstart, start - tstart);
1677 PL_realtokenstart = -1;
29595ff2 1678 }
cd81e915 1679 if (PL_skipwhite) {
29595ff2 1680 if (!*svp)
6b29d1f5 1681 *svp = newSVpvs("");
cd81e915
NC
1682 sv_setsv(*svp, PL_skipwhite);
1683 sv_free(PL_skipwhite);
1684 PL_skipwhite = 0;
29595ff2
NC
1685 }
1686
1687 return s;
1688}
1689#endif
1690
80a702cd 1691STATIC void
15f169a1 1692S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1693{
1694 AV *av = CopFILEAVx(PL_curcop);
1695 if (av) {
b9f83d2f 1696 SV * const sv = newSV_type(SVt_PVMG);
5fa550fb
NC
1697 if (orig_sv)
1698 sv_setsv(sv, orig_sv);
1699 else
1700 sv_setpvn(sv, buf, len);
80a702cd
RGS
1701 (void)SvIOK_on(sv);
1702 SvIV_set(sv, 0);
1703 av_store(av, (I32)CopLINE(PL_curcop), sv);
1704 }
1705}
1706
ffb4593c
NT
1707/*
1708 * S_skipspace
1709 * Called to gobble the appropriate amount and type of whitespace.
1710 * Skips comments as well.
1711 */
1712
76e3520e 1713STATIC char *
cea2e8a9 1714S_skipspace(pTHX_ register char *s)
a687059c 1715{
5db06880 1716#ifdef PERL_MAD
f0e67a1d
Z
1717 char *start = s;
1718#endif /* PERL_MAD */
7918f24d 1719 PERL_ARGS_ASSERT_SKIPSPACE;
f0e67a1d 1720#ifdef PERL_MAD
cd81e915
NC
1721 if (PL_skipwhite) {
1722 sv_free(PL_skipwhite);
f0e67a1d 1723 PL_skipwhite = NULL;
5db06880 1724 }
f0e67a1d 1725#endif /* PERL_MAD */
3280af22 1726 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
bf4acbe4 1727 while (s < PL_bufend && SPACE_OR_TAB(*s))
463ee0b2 1728 s++;
f0e67a1d
Z
1729 } else {
1730 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1731 PL_bufptr = s;
f0998909
Z
1732 lex_read_space(LEX_KEEP_PREVIOUS |
1733 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1734 LEX_NO_NEXT_CHUNK : 0));
3280af22 1735 s = PL_bufptr;
f0e67a1d
Z
1736 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1737 if (PL_linestart > PL_bufptr)
1738 PL_bufptr = PL_linestart;
1739 return s;
463ee0b2 1740 }
5db06880 1741#ifdef PERL_MAD
f0e67a1d
Z
1742 if (PL_madskills)
1743 PL_skipwhite = newSVpvn(start, s-start);
1744#endif /* PERL_MAD */
5db06880 1745 return s;
a687059c 1746}
378cc40b 1747
ffb4593c
NT
1748/*
1749 * S_check_uni
1750 * Check the unary operators to ensure there's no ambiguity in how they're
1751 * used. An ambiguous piece of code would be:
1752 * rand + 5
1753 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1754 * the +5 is its argument.
1755 */
1756
76e3520e 1757STATIC void
cea2e8a9 1758S_check_uni(pTHX)
ba106d47 1759{
97aff369 1760 dVAR;
d4c19fe8
AL
1761 const char *s;
1762 const char *t;
2f3197b3 1763
3280af22 1764 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1765 return;
3280af22
NIS
1766 while (isSPACE(*PL_last_uni))
1767 PL_last_uni++;
c35e046a
AL
1768 s = PL_last_uni;
1769 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1770 s++;
3280af22 1771 if ((t = strchr(s, '(')) && t < PL_bufptr)
a0d0e21e 1772 return;
6136c704 1773
9b387841
NC
1774 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1775 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1776 (int)(s - PL_last_uni), PL_last_uni);
2f3197b3
LW
1777}
1778
ffb4593c
NT
1779/*
1780 * LOP : macro to build a list operator. Its behaviour has been replaced
1781 * with a subroutine, S_lop() for which LOP is just another name.
1782 */
1783
a0d0e21e
LW
1784#define LOP(f,x) return lop(f,x,s)
1785
ffb4593c
NT
1786/*
1787 * S_lop
1788 * Build a list operator (or something that might be one). The rules:
1789 * - if we have a next token, then it's a list operator [why?]
1790 * - if the next thing is an opening paren, then it's a function
1791 * - else it's a list operator
1792 */
1793
76e3520e 1794STATIC I32
a0be28da 1795S_lop(pTHX_ I32 f, int x, char *s)
ffed7fef 1796{
97aff369 1797 dVAR;
7918f24d
NC
1798
1799 PERL_ARGS_ASSERT_LOP;
1800
6154021b 1801 pl_yylval.ival = f;
35c8bce7 1802 CLINE;
3280af22
NIS
1803 PL_expect = x;
1804 PL_bufptr = s;
1805 PL_last_lop = PL_oldbufptr;
eb160463 1806 PL_last_lop_op = (OPCODE)f;
5db06880
NC
1807#ifdef PERL_MAD
1808 if (PL_lasttoke)
1809 return REPORT(LSTOP);
1810#else
3280af22 1811 if (PL_nexttoke)
bbf60fe6 1812 return REPORT(LSTOP);
5db06880 1813#endif
79072805 1814 if (*s == '(')
bbf60fe6 1815 return REPORT(FUNC);
29595ff2 1816 s = PEEKSPACE(s);
79072805 1817 if (*s == '(')
bbf60fe6 1818 return REPORT(FUNC);
79072805 1819 else
bbf60fe6 1820 return REPORT(LSTOP);
79072805
LW
1821}
1822
5db06880
NC
1823#ifdef PERL_MAD
1824 /*
1825 * S_start_force
1826 * Sets up for an eventual force_next(). start_force(0) basically does
1827 * an unshift, while start_force(-1) does a push. yylex removes items
1828 * on the "pop" end.
1829 */
1830
1831STATIC void
1832S_start_force(pTHX_ int where)
1833{
1834 int i;
1835
cd81e915 1836 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
5db06880 1837 where = PL_lasttoke;
cd81e915
NC
1838 assert(PL_curforce < 0 || PL_curforce == where);
1839 if (PL_curforce != where) {
5db06880
NC
1840 for (i = PL_lasttoke; i > where; --i) {
1841 PL_nexttoke[i] = PL_nexttoke[i-1];
1842 }
1843 PL_lasttoke++;
1844 }
cd81e915 1845 if (PL_curforce < 0) /* in case of duplicate start_force() */
5db06880 1846 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
cd81e915
NC
1847 PL_curforce = where;
1848 if (PL_nextwhite) {
5db06880 1849 if (PL_madskills)
6b29d1f5 1850 curmad('^', newSVpvs(""));
cd81e915 1851 CURMAD('_', PL_nextwhite);
5db06880
NC
1852 }
1853}
1854
1855STATIC void
1856S_curmad(pTHX_ char slot, SV *sv)
1857{
1858 MADPROP **where;
1859
1860 if (!sv)
1861 return;
cd81e915
NC
1862 if (PL_curforce < 0)
1863 where = &PL_thismad;
5db06880 1864 else
cd81e915 1865 where = &PL_nexttoke[PL_curforce].next_mad;
5db06880 1866
cd81e915 1867 if (PL_faketokens)
76f68e9b 1868 sv_setpvs(sv, "");
5db06880
NC
1869 else {
1870 if (!IN_BYTES) {
1871 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1872 SvUTF8_on(sv);
1873 else if (PL_encoding) {
1874 sv_recode_to_utf8(sv, PL_encoding);
1875 }
1876 }
1877 }
1878
1879 /* keep a slot open for the head of the list? */
1880 if (slot != '_' && *where && (*where)->mad_key == '^') {
1881 (*where)->mad_key = slot;
daba3364 1882 sv_free(MUTABLE_SV(((*where)->mad_val)));
5db06880
NC
1883 (*where)->mad_val = (void*)sv;
1884 }
1885 else
1886 addmad(newMADsv(slot, sv), where, 0);
1887}
1888#else
b3f24c00
MHM
1889# define start_force(where) NOOP
1890# define curmad(slot, sv) NOOP
5db06880
NC
1891#endif
1892
ffb4593c
NT
1893/*
1894 * S_force_next
9cbb5ea2 1895 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1896 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1897 * to know what token to return the next time the lexer is called. Caller
5db06880
NC
1898 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1899 * and possibly PL_expect to ensure the lexer handles the token correctly.
ffb4593c
NT
1900 */
1901
4e553d73 1902STATIC void
cea2e8a9 1903S_force_next(pTHX_ I32 type)
79072805 1904{
97aff369 1905 dVAR;
704d4215
GG
1906#ifdef DEBUGGING
1907 if (DEBUG_T_TEST) {
1908 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1909 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
1910 }
1911#endif
5db06880 1912#ifdef PERL_MAD
cd81e915 1913 if (PL_curforce < 0)
5db06880 1914 start_force(PL_lasttoke);
cd81e915 1915 PL_nexttoke[PL_curforce].next_type = type;
5db06880
NC
1916 if (PL_lex_state != LEX_KNOWNEXT)
1917 PL_lex_defer = PL_lex_state;
1918 PL_lex_state = LEX_KNOWNEXT;
1919 PL_lex_expect = PL_expect;
cd81e915 1920 PL_curforce = -1;
5db06880 1921#else
3280af22
NIS
1922 PL_nexttype[PL_nexttoke] = type;
1923 PL_nexttoke++;
1924 if (PL_lex_state != LEX_KNOWNEXT) {
1925 PL_lex_defer = PL_lex_state;
1926 PL_lex_expect = PL_expect;
1927 PL_lex_state = LEX_KNOWNEXT;
79072805 1928 }
5db06880 1929#endif
79072805
LW
1930}
1931
d0a148a6 1932STATIC SV *
15f169a1 1933S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 1934{
97aff369 1935 dVAR;
740cce10 1936 SV * const sv = newSVpvn_utf8(start, len,
eaf7a4d2
CS
1937 !IN_BYTES
1938 && UTF
1939 && !is_ascii_string((const U8*)start, len)
740cce10 1940 && is_utf8_string((const U8*)start, len));
d0a148a6
NC
1941 return sv;
1942}
1943
ffb4593c
NT
1944/*
1945 * S_force_word
1946 * When the lexer knows the next thing is a word (for instance, it has
1947 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
1948 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1949 * lookahead.
ffb4593c
NT
1950 *
1951 * Arguments:
b1b65b59 1952 * char *start : buffer position (must be within PL_linestr)
02b34bbe 1953 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
ffb4593c
NT
1954 * int check_keyword : if true, Perl checks to make sure the word isn't
1955 * a keyword (do this if the word is a label, e.g. goto FOO)
1956 * int allow_pack : if true, : characters will also be allowed (require,
1957 * use, etc. do this)
9cbb5ea2 1958 * int allow_initial_tick : used by the "sub" lexer only.
ffb4593c
NT
1959 */
1960
76e3520e 1961STATIC char *
cea2e8a9 1962S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
79072805 1963{
97aff369 1964 dVAR;
463ee0b2
LW
1965 register char *s;
1966 STRLEN len;
4e553d73 1967
7918f24d
NC
1968 PERL_ARGS_ASSERT_FORCE_WORD;
1969
29595ff2 1970 start = SKIPSPACE1(start);
463ee0b2 1971 s = start;
7e2040f0 1972 if (isIDFIRST_lazy_if(s,UTF) ||
a0d0e21e 1973 (allow_pack && *s == ':') ||
15f0808c 1974 (allow_initial_tick && *s == '\'') )
a0d0e21e 1975 {
3280af22 1976 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
5458a98a 1977 if (check_keyword && keyword(PL_tokenbuf, len, 0))
463ee0b2 1978 return start;
cd81e915 1979 start_force(PL_curforce);
5db06880
NC
1980 if (PL_madskills)
1981 curmad('X', newSVpvn(start,s-start));
463ee0b2 1982 if (token == METHOD) {
29595ff2 1983 s = SKIPSPACE1(s);
463ee0b2 1984 if (*s == '(')
3280af22 1985 PL_expect = XTERM;
463ee0b2 1986 else {
3280af22 1987 PL_expect = XOPERATOR;
463ee0b2 1988 }
79072805 1989 }
e74e6b3d 1990 if (PL_madskills)
63575281 1991 curmad('g', newSVpvs( "forced" ));
9ded7720 1992 NEXTVAL_NEXTTOKE.opval
d0a148a6
NC
1993 = (OP*)newSVOP(OP_CONST,0,
1994 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 1995 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
1996 force_next(token);
1997 }
1998 return s;
1999}
2000
ffb4593c
NT
2001/*
2002 * S_force_ident
9cbb5ea2 2003 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2004 * text only contains the "foo" portion. The first argument is a pointer
2005 * to the "foo", and the second argument is the type symbol to prefix.
2006 * Forces the next token to be a "WORD".
9cbb5ea2 2007 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2008 */
2009
76e3520e 2010STATIC void
bfed75c6 2011S_force_ident(pTHX_ register const char *s, int kind)
79072805 2012{
97aff369 2013 dVAR;
7918f24d
NC
2014
2015 PERL_ARGS_ASSERT_FORCE_IDENT;
2016
c35e046a 2017 if (*s) {
90e5519e
NC
2018 const STRLEN len = strlen(s);
2019 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
cd81e915 2020 start_force(PL_curforce);
9ded7720 2021 NEXTVAL_NEXTTOKE.opval = o;
79072805 2022 force_next(WORD);
748a9306 2023 if (kind) {
11343788 2024 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2025 /* XXX see note in pp_entereval() for why we forgo typo
2026 warnings if the symbol must be introduced in an eval.
2027 GSAR 96-10-12 */
90e5519e
NC
2028 gv_fetchpvn_flags(s, len,
2029 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2030 : GV_ADD,
2031 kind == '$' ? SVt_PV :
2032 kind == '@' ? SVt_PVAV :
2033 kind == '%' ? SVt_PVHV :
a0d0e21e 2034 SVt_PVGV
90e5519e 2035 );
748a9306 2036 }
79072805
LW
2037 }
2038}
2039
1571675a
GS
2040NV
2041Perl_str_to_version(pTHX_ SV *sv)
2042{
2043 NV retval = 0.0;
2044 NV nshift = 1.0;
2045 STRLEN len;
cfd0369c 2046 const char *start = SvPV_const(sv,len);
9d4ba2ae 2047 const char * const end = start + len;
504618e9 2048 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
7918f24d
NC
2049
2050 PERL_ARGS_ASSERT_STR_TO_VERSION;
2051
1571675a 2052 while (start < end) {
ba210ebe 2053 STRLEN skip;
1571675a
GS
2054 UV n;
2055 if (utf)
9041c2e3 2056 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2057 else {
2058 n = *(U8*)start;
2059 skip = 1;
2060 }
2061 retval += ((NV)n)/nshift;
2062 start += skip;
2063 nshift *= 1000;
2064 }
2065 return retval;
2066}
2067
4e553d73 2068/*
ffb4593c
NT
2069 * S_force_version
2070 * Forces the next token to be a version number.
e759cc13
RGS
2071 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2072 * and if "guessing" is TRUE, then no new token is created (and the caller
2073 * must use an alternative parsing method).
ffb4593c
NT
2074 */
2075
76e3520e 2076STATIC char *
e759cc13 2077S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2078{
97aff369 2079 dVAR;
5f66b61c 2080 OP *version = NULL;
44dcb63b 2081 char *d;
5db06880
NC
2082#ifdef PERL_MAD
2083 I32 startoff = s - SvPVX(PL_linestr);
2084#endif
89bfa8cd 2085
7918f24d
NC
2086 PERL_ARGS_ASSERT_FORCE_VERSION;
2087
29595ff2 2088 s = SKIPSPACE1(s);
89bfa8cd 2089
44dcb63b 2090 d = s;
dd629d5b 2091 if (*d == 'v')
44dcb63b 2092 d++;
44dcb63b 2093 if (isDIGIT(*d)) {
e759cc13
RGS
2094 while (isDIGIT(*d) || *d == '_' || *d == '.')
2095 d++;
5db06880
NC
2096#ifdef PERL_MAD
2097 if (PL_madskills) {
cd81e915 2098 start_force(PL_curforce);
5db06880
NC
2099 curmad('X', newSVpvn(s,d-s));
2100 }
2101#endif
4e4da3ac 2102 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2103 SV *ver;
8d08d9ba
DG
2104#ifdef USE_LOCALE_NUMERIC
2105 char *loc = setlocale(LC_NUMERIC, "C");
2106#endif
6154021b 2107 s = scan_num(s, &pl_yylval);
8d08d9ba
DG
2108#ifdef USE_LOCALE_NUMERIC
2109 setlocale(LC_NUMERIC, loc);
2110#endif
6154021b 2111 version = pl_yylval.opval;
dd629d5b
GS
2112 ver = cSVOPx(version)->op_sv;
2113 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2114 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2115 SvNV_set(ver, str_to_version(ver));
1571675a 2116 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2117 }
89bfa8cd 2118 }
5db06880
NC
2119 else if (guessing) {
2120#ifdef PERL_MAD
2121 if (PL_madskills) {
cd81e915
NC
2122 sv_free(PL_nextwhite); /* let next token collect whitespace */
2123 PL_nextwhite = 0;
5db06880
NC
2124 s = SvPVX(PL_linestr) + startoff;
2125 }
2126#endif
e759cc13 2127 return s;
5db06880 2128 }
89bfa8cd
PP
2129 }
2130
5db06880
NC
2131#ifdef PERL_MAD
2132 if (PL_madskills && !version) {
cd81e915
NC
2133 sv_free(PL_nextwhite); /* let next token collect whitespace */
2134 PL_nextwhite = 0;
5db06880
NC
2135 s = SvPVX(PL_linestr) + startoff;
2136 }
2137#endif
89bfa8cd 2138 /* NOTE: The parser sees the package name and the VERSION swapped */
cd81e915 2139 start_force(PL_curforce);
9ded7720 2140 NEXTVAL_NEXTTOKE.opval = version;
4e553d73 2141 force_next(WORD);
89bfa8cd 2142
e759cc13 2143 return s;
89bfa8cd
PP
2144}
2145
ffb4593c 2146/*
91152fc1
DG
2147 * S_force_strict_version
2148 * Forces the next token to be a version number using strict syntax rules.
2149 */
2150
2151STATIC char *
2152S_force_strict_version(pTHX_ char *s)
2153{
2154 dVAR;
2155 OP *version = NULL;
2156#ifdef PERL_MAD
2157 I32 startoff = s - SvPVX(PL_linestr);
2158#endif
2159 const char *errstr = NULL;
2160
2161 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2162
2163 while (isSPACE(*s)) /* leading whitespace */
2164 s++;
2165
2166 if (is_STRICT_VERSION(s,&errstr)) {
2167 SV *ver = newSV(0);
2168 s = (char *)scan_version(s, ver, 0);
2169 version = newSVOP(OP_CONST, 0, ver);
2170 }
4e4da3ac
Z
2171 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2172 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2173 {
91152fc1
DG
2174 PL_bufptr = s;
2175 if (errstr)
2176 yyerror(errstr); /* version required */
2177 return s;
2178 }
2179
2180#ifdef PERL_MAD
2181 if (PL_madskills && !version) {
2182 sv_free(PL_nextwhite); /* let next token collect whitespace */
2183 PL_nextwhite = 0;
2184 s = SvPVX(PL_linestr) + startoff;
2185 }
2186#endif
2187 /* NOTE: The parser sees the package name and the VERSION swapped */
2188 start_force(PL_curforce);
2189 NEXTVAL_NEXTTOKE.opval = version;
2190 force_next(WORD);
2191
2192 return s;
2193}
2194
2195/*
ffb4593c
NT
2196 * S_tokeq
2197 * Tokenize a quoted string passed in as an SV. It finds the next
2198 * chunk, up to end of string or a backslash. It may make a new
2199 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2200 * turns \\ into \.
2201 */
2202
76e3520e 2203STATIC SV *
cea2e8a9 2204S_tokeq(pTHX_ SV *sv)
79072805 2205{
97aff369 2206 dVAR;
79072805
LW
2207 register char *s;
2208 register char *send;
2209 register char *d;
b3ac6de7
IZ
2210 STRLEN len = 0;
2211 SV *pv = sv;
79072805 2212
7918f24d
NC
2213 PERL_ARGS_ASSERT_TOKEQ;
2214
79072805 2215 if (!SvLEN(sv))
b3ac6de7 2216 goto finish;
79072805 2217
a0d0e21e 2218 s = SvPV_force(sv, len);
21a311ee 2219 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
b3ac6de7 2220 goto finish;
463ee0b2 2221 send = s + len;
79072805
LW
2222 while (s < send && *s != '\\')
2223 s++;
2224 if (s == send)
b3ac6de7 2225 goto finish;
79072805 2226 d = s;
be4731d2 2227 if ( PL_hints & HINT_NEW_STRING ) {
59cd0e26 2228 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
be4731d2 2229 }
79072805
LW
2230 while (s < send) {
2231 if (*s == '\\') {
a0d0e21e 2232 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2233 s++; /* all that, just for this */
2234 }
2235 *d++ = *s++;
2236 }
2237 *d = '\0';
95a20fc0 2238 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2239 finish:
3280af22 2240 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2241 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2242 return sv;
2243}
2244
ffb4593c
NT
2245/*
2246 * Now come three functions related to double-quote context,
2247 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2248 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2249 * interact with PL_lex_state, and create fake ( ... ) argument lists
2250 * to handle functions and concatenation.
2251 * They assume that whoever calls them will be setting up a fake
2252 * join call, because each subthing puts a ',' after it. This lets
2253 * "lower \luPpEr"
2254 * become
2255 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2256 *
2257 * (I'm not sure whether the spurious commas at the end of lcfirst's
2258 * arguments and join's arguments are created or not).
2259 */
2260
2261/*
2262 * S_sublex_start
6154021b 2263 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2264 *
2265 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2266 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c
NT
2267 *
2268 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2269 *
2270 * Everything else becomes a FUNC.
2271 *
2272 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2273 * had an OP_CONST or OP_READLINE). This just sets us up for a
2274 * call to S_sublex_push().
2275 */
2276
76e3520e 2277STATIC I32
cea2e8a9 2278S_sublex_start(pTHX)
79072805 2279{
97aff369 2280 dVAR;
6154021b 2281 register const I32 op_type = pl_yylval.ival;
79072805
LW
2282
2283 if (op_type == OP_NULL) {
6154021b 2284 pl_yylval.opval = PL_lex_op;
5f66b61c 2285 PL_lex_op = NULL;
79072805
LW
2286 return THING;
2287 }
2288 if (op_type == OP_CONST || op_type == OP_READLINE) {
3280af22 2289 SV *sv = tokeq(PL_lex_stuff);
b3ac6de7
IZ
2290
2291 if (SvTYPE(sv) == SVt_PVIV) {
2292 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2293 STRLEN len;
96a5add6 2294 const char * const p = SvPV_const(sv, len);
740cce10 2295 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2296 SvREFCNT_dec(sv);
2297 sv = nsv;
4e553d73 2298 }
6154021b 2299 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
a0714e2c 2300 PL_lex_stuff = NULL;
6f33ba73
RGS
2301 /* Allow <FH> // "foo" */
2302 if (op_type == OP_READLINE)
2303 PL_expect = XTERMORDORDOR;
79072805
LW
2304 return THING;
2305 }
e3f73d4e
RGS
2306 else if (op_type == OP_BACKTICK && PL_lex_op) {
2307 /* readpipe() vas overriden */
2308 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
6154021b 2309 pl_yylval.opval = PL_lex_op;
9b201d7d 2310 PL_lex_op = NULL;
e3f73d4e
RGS
2311 PL_lex_stuff = NULL;
2312 return THING;
2313 }
79072805 2314
3280af22 2315 PL_sublex_info.super_state = PL_lex_state;
eac04b2e 2316 PL_sublex_info.sub_inwhat = (U16)op_type;
3280af22
NIS
2317 PL_sublex_info.sub_op = PL_lex_op;
2318 PL_lex_state = LEX_INTERPPUSH;
55497cff 2319
3280af22
NIS
2320 PL_expect = XTERM;
2321 if (PL_lex_op) {
6154021b 2322 pl_yylval.opval = PL_lex_op;
5f66b61c 2323 PL_lex_op = NULL;
55497cff
PP
2324 return PMFUNC;
2325 }
2326 else
2327 return FUNC;
2328}
2329
ffb4593c
NT
2330/*
2331 * S_sublex_push
2332 * Create a new scope to save the lexing state. The scope will be
2333 * ended in S_sublex_done. Returns a '(', starting the function arguments
2334 * to the uc, lc, etc. found before.
2335 * Sets PL_lex_state to LEX_INTERPCONCAT.
2336 */
2337
76e3520e 2338STATIC I32
cea2e8a9 2339S_sublex_push(pTHX)
55497cff 2340{
27da23d5 2341 dVAR;
f46d017c 2342 ENTER;
55497cff 2343
3280af22 2344 PL_lex_state = PL_sublex_info.super_state;
651b5b28 2345 SAVEBOOL(PL_lex_dojoin);
3280af22 2346 SAVEI32(PL_lex_brackets);
3280af22
NIS
2347 SAVEI32(PL_lex_casemods);
2348 SAVEI32(PL_lex_starts);
651b5b28 2349 SAVEI8(PL_lex_state);
7766f137 2350 SAVEVPTR(PL_lex_inpat);
98246f1e 2351 SAVEI16(PL_lex_inwhat);
57843af0 2352 SAVECOPLINE(PL_curcop);
3280af22 2353 SAVEPPTR(PL_bufptr);
8452ff4b 2354 SAVEPPTR(PL_bufend);
3280af22
NIS
2355 SAVEPPTR(PL_oldbufptr);
2356 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2357 SAVEPPTR(PL_last_lop);
2358 SAVEPPTR(PL_last_uni);
3280af22
NIS
2359 SAVEPPTR(PL_linestart);
2360 SAVESPTR(PL_linestr);
8edd5f42
RGS
2361 SAVEGENERICPV(PL_lex_brackstack);
2362 SAVEGENERICPV(PL_lex_casestack);
3280af22
NIS
2363
2364 PL_linestr = PL_lex_stuff;
a0714e2c 2365 PL_lex_stuff = NULL;
3280af22 2366
9cbb5ea2
GS
2367 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2368 = SvPVX(PL_linestr);
3280af22 2369 PL_bufend += SvCUR(PL_linestr);
bd61b366 2370 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2371 SAVEFREESV(PL_linestr);
2372
2373 PL_lex_dojoin = FALSE;
2374 PL_lex_brackets = 0;
a02a5408
JC
2375 Newx(PL_lex_brackstack, 120, char);
2376 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2377 PL_lex_casemods = 0;
2378 *PL_lex_casestack = '\0';
2379 PL_lex_starts = 0;
2380 PL_lex_state = LEX_INTERPCONCAT;
eb160463 2381 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
3280af22
NIS
2382
2383 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2384 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2385 PL_lex_inpat = PL_sublex_info.sub_op;
79072805 2386 else
5f66b61c 2387 PL_lex_inpat = NULL;
79072805 2388
55497cff 2389 return '(';
79072805
LW
2390}
2391
ffb4593c
NT
2392/*
2393 * S_sublex_done
2394 * Restores lexer state after a S_sublex_push.
2395 */
2396
76e3520e 2397STATIC I32
cea2e8a9 2398S_sublex_done(pTHX)
79072805 2399{
27da23d5 2400 dVAR;
3280af22 2401 if (!PL_lex_starts++) {
396482e1 2402 SV * const sv = newSVpvs("");
9aa983d2
JH
2403 if (SvUTF8(PL_linestr))
2404 SvUTF8_on(sv);
3280af22 2405 PL_expect = XOPERATOR;
6154021b 2406 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
79072805
LW
2407 return THING;
2408 }
2409
3280af22
NIS
2410 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2411 PL_lex_state = LEX_INTERPCASEMOD;
cea2e8a9 2412 return yylex();
79072805
LW
2413 }
2414
ffb4593c 2415 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
3280af22
NIS
2416 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2417 PL_linestr = PL_lex_repl;
2418 PL_lex_inpat = 0;
2419 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2420 PL_bufend += SvCUR(PL_linestr);
bd61b366 2421 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2422 SAVEFREESV(PL_linestr);
2423 PL_lex_dojoin = FALSE;
2424 PL_lex_brackets = 0;
3280af22
NIS
2425 PL_lex_casemods = 0;
2426 *PL_lex_casestack = '\0';
2427 PL_lex_starts = 0;
25da4f38 2428 if (SvEVALED(PL_lex_repl)) {
3280af22
NIS
2429 PL_lex_state = LEX_INTERPNORMAL;
2430 PL_lex_starts++;
e9fa98b2
HS
2431 /* we don't clear PL_lex_repl here, so that we can check later
2432 whether this is an evalled subst; that means we rely on the
2433 logic to ensure sublex_done() is called again only via the
2434 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
79072805 2435 }
e9fa98b2 2436 else {
3280af22 2437 PL_lex_state = LEX_INTERPCONCAT;
a0714e2c 2438 PL_lex_repl = NULL;
e9fa98b2 2439 }
79072805 2440 return ',';
ffed7fef
LW
2441 }
2442 else {
5db06880
NC
2443#ifdef PERL_MAD
2444 if (PL_madskills) {
cd81e915
NC
2445 if (PL_thiswhite) {
2446 if (!PL_endwhite)
6b29d1f5 2447 PL_endwhite = newSVpvs("");
cd81e915
NC
2448 sv_catsv(PL_endwhite, PL_thiswhite);
2449 PL_thiswhite = 0;
2450 }
2451 if (PL_thistoken)
76f68e9b 2452 sv_setpvs(PL_thistoken,"");
5db06880 2453 else
cd81e915 2454 PL_realtokenstart = -1;
5db06880
NC
2455 }
2456#endif
f46d017c 2457 LEAVE;
3280af22
NIS
2458 PL_bufend = SvPVX(PL_linestr);
2459 PL_bufend += SvCUR(PL_linestr);
2460 PL_expect = XOPERATOR;
09bef843 2461 PL_sublex_info.sub_inwhat = 0;
79072805 2462 return ')';
ffed7fef
LW
2463 }
2464}
2465
02aa26ce
NT
2466/*
2467 scan_const
2468
2469 Extracts a pattern, double-quoted string, or transliteration. This
2470 is terrifying code.
2471
94def140 2472 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2473 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2474 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2475
94def140
ST
2476 Returns a pointer to the character scanned up to. If this is
2477 advanced from the start pointer supplied (i.e. if anything was
9b599b2a 2478 successfully parsed), will leave an OP for the substring scanned
6154021b 2479 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2480 by looking at the next characters herself.
2481
02aa26ce
NT
2482 In patterns:
2483 backslashes:
ff3f963a 2484 constants: \N{NAME} only
02aa26ce
NT
2485 case and quoting: \U \Q \E
2486 stops on @ and $, but not for $ as tail anchor
2487
2488 In transliterations:
2489 characters are VERY literal, except for - not at the start or end
94def140
ST
2490 of the string, which indicates a range. If the range is in bytes,
2491 scan_const expands the range to the full set of intermediate
2492 characters. If the range is in utf8, the hyphen is replaced with
2493 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2494
2495 In double-quoted strings:
2496 backslashes:
2497 double-quoted style: \r and \n
ff3f963a 2498 constants: \x31, etc.
94def140 2499 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2500 case and quoting: \U \Q \E
2501 stops on @ and $
2502
2503 scan_const does *not* construct ops to handle interpolated strings.
2504 It stops processing as soon as it finds an embedded $ or @ variable
2505 and leaves it to the caller to work out what's going on.
2506
94def140
ST
2507 embedded arrays (whether in pattern or not) could be:
2508 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2509
2510 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2511
2512 $ in pattern could be $foo or could be tail anchor. Assumption:
2513 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2514 followed by one of "()| \r\n\t"
02aa26ce
NT
2515
2516 \1 (backreferences) are turned into $1
2517
2518 The structure of the code is
2519 while (there's a character to process) {
94def140
ST
2520 handle transliteration ranges
2521 skip regexp comments /(?#comment)/ and codes /(?{code})/
2522 skip #-initiated comments in //x patterns
2523 check for embedded arrays
02aa26ce
NT
2524 check for embedded scalars
2525 if (backslash) {
94def140 2526 deprecate \1 in substitution replacements
02aa26ce
NT
2527 handle string-changing backslashes \l \U \Q \E, etc.
2528 switch (what was escaped) {
94def140 2529 handle \- in a transliteration (becomes a literal -)
ff3f963a 2530 if a pattern and not \N{, go treat as regular character
94def140
ST
2531 handle \132 (octal characters)
2532 handle \x15 and \x{1234} (hex characters)
ff3f963a 2533 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2534 handle \cV (control characters)
2535 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2536 } (end switch)
77a135fe 2537 continue
02aa26ce 2538 } (end if backslash)
77a135fe 2539 handle regular character
02aa26ce 2540 } (end while character to read)
4e553d73 2541
02aa26ce
NT
2542*/
2543
76e3520e 2544STATIC char *
cea2e8a9 2545S_scan_const(pTHX_ char *start)
79072805 2546{
97aff369 2547 dVAR;
3280af22 2548 register char *send = PL_bufend; /* end of the constant */
77a135fe
KW
2549 SV *sv = newSV(send - start); /* sv for the constant. See
2550 note below on sizing. */
02aa26ce
NT
2551 register char *s = start; /* start of the constant */
2552 register char *d = SvPVX(sv); /* destination for copies */
2553 bool dorange = FALSE; /* are we in a translit range? */
c2e66d9e 2554 bool didrange = FALSE; /* did we just finish a range? */
2b9d42f0 2555 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
77a135fe
KW
2556 I32 this_utf8 = UTF; /* Is the source string assumed
2557 to be UTF8? But, this can
2558 show as true when the source
2559 isn't utf8, as for example
2560 when it is entirely composed
2561 of hex constants */
2562
2563 /* Note on sizing: The scanned constant is placed into sv, which is
2564 * initialized by newSV() assuming one byte of output for every byte of
2565 * input. This routine expects newSV() to allocate an extra byte for a
2566 * trailing NUL, which this routine will append if it gets to the end of
2567 * the input. There may be more bytes of input than output (eg., \N{LATIN
2568 * CAPITAL LETTER A}), or more output than input if the constant ends up
2569 * recoded to utf8, but each time a construct is found that might increase
2570 * the needed size, SvGROW() is called. Its size parameter each time is
2571 * based on the best guess estimate at the time, namely the length used so
2572 * far, plus the length the current construct will occupy, plus room for
2573 * the trailing NUL, plus one byte for every input byte still unscanned */
2574
012bcf8d 2575 UV uv;
4c3a8340
ST
2576#ifdef EBCDIC
2577 UV literal_endpoint = 0;
e294cc5d 2578 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
4c3a8340 2579#endif
012bcf8d 2580
7918f24d
NC
2581 PERL_ARGS_ASSERT_SCAN_CONST;
2582
2b9d42f0
NIS
2583 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2584 /* If we are doing a trans and we know we want UTF8 set expectation */
2585 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2586 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2587 }
2588
2589
79072805 2590 while (s < send || dorange) {
ff3f963a 2591
02aa26ce 2592 /* get transliterations out of the way (they're most literal) */
3280af22 2593 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2594 /* expand a range A-Z to the full set of characters. AIE! */
79072805 2595 if (dorange) {
1ba5c669
JH
2596 I32 i; /* current expanded character */
2597 I32 min; /* first character in range */
2598 I32 max; /* last character in range */
02aa26ce 2599
e294cc5d
JH
2600#ifdef EBCDIC
2601 UV uvmax = 0;
2602#endif
2603
2604 if (has_utf8
2605#ifdef EBCDIC
2606 && !native_range
2607#endif
2608 ) {
9d4ba2ae 2609 char * const c = (char*)utf8_hop((U8*)d, -1);
8973db79
JH
2610 char *e = d++;
2611 while (e-- > c)
2612 *(e + 1) = *e;
25716404 2613 *c = (char)UTF_TO_NATIVE(0xff);
8973db79
JH
2614 /* mark the range as done, and continue */
2615 dorange = FALSE;
2616 didrange = TRUE;
2617 continue;
2618 }
2b9d42f0 2619
95a20fc0 2620 i = d - SvPVX_const(sv); /* remember current offset */
e294cc5d
JH
2621#ifdef EBCDIC
2622 SvGROW(sv,
2623 SvLEN(sv) + (has_utf8 ?
2624 (512 - UTF_CONTINUATION_MARK +
2625 UNISKIP(0x100))
2626 : 256));
2627 /* How many two-byte within 0..255: 128 in UTF-8,
2628 * 96 in UTF-8-mod. */
2629#else
9cbb5ea2 2630 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
e294cc5d 2631#endif
9cbb5ea2 2632 d = SvPVX(sv) + i; /* refresh d after realloc */
e294cc5d
JH
2633#ifdef EBCDIC
2634 if (has_utf8) {
2635 int j;
2636 for (j = 0; j <= 1; j++) {
2637 char * const c = (char*)utf8_hop((U8*)d, -1);
2638 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2639 if (j)
2640 min = (U8)uv;
2641 else if (uv < 256)
2642 max = (U8)uv;
2643 else {
2644 max = (U8)0xff; /* only to \xff */
2645 uvmax = uv; /* \x{100} to uvmax */
2646 }
2647 d = c; /* eat endpoint chars */
2648 }
2649 }
2650 else {
2651#endif
2652 d -= 2; /* eat the first char and the - */
2653 min = (U8)*d; /* first char in range */
2654 max = (U8)d[1]; /* last char in range */
2655#ifdef EBCDIC
2656 }
2657#endif
8ada0baa 2658
c2e66d9e 2659 if (min > max) {
01ec43d0 2660 Perl_croak(aTHX_
d1573ac7 2661 "Invalid range \"%c-%c\" in transliteration operator",
1ba5c669 2662 (char)min, (char)max);
c2e66d9e
GS
2663 }
2664
c7f1f016 2665#ifdef EBCDIC
4c3a8340
ST
2666 if (literal_endpoint == 2 &&
2667 ((isLOWER(min) && isLOWER(max)) ||
2668 (isUPPER(min) && isUPPER(max)))) {
8ada0baa
JH
2669 if (isLOWER(min)) {
2670 for (i = min; i <= max; i++)
2671 if (isLOWER(i))
db42d148 2672 *d++ = NATIVE_TO_NEED(has_utf8,i);
8ada0baa
JH
2673 } else {
2674 for (i = min; i <= max; i++)
2675 if (isUPPER(i))
db42d148 2676 *d++ = NATIVE_TO_NEED(has_utf8,i);
8ada0baa
JH
2677 }
2678 }
2679 else
2680#endif
2681 for (i = min; i <= max; i++)
e294cc5d
JH
2682#ifdef EBCDIC
2683 if (has_utf8) {
2684 const U8 ch = (U8)NATIVE_TO_UTF(i);
2685 if (UNI_IS_INVARIANT(ch))
2686 *d++ = (U8)i;
2687 else {
2688 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2689 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2690 }
2691 }
2692 else
2693#endif
2694 *d++ = (char)i;
2695
2696#ifdef EBCDIC
2697 if (uvmax) {
2698 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2699 if (uvmax > 0x101)
2700 *d++ = (char)UTF_TO_NATIVE(0xff);
2701 if (uvmax > 0x100)
2702 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2703 }
2704#endif
02aa26ce
NT
2705
2706 /* mark the range as done, and continue */
79072805 2707 dorange = FALSE;
01ec43d0 2708 didrange = TRUE;
4c3a8340
ST
2709#ifdef EBCDIC
2710 literal_endpoint = 0;
2711#endif
79072805 2712 continue;
4e553d73 2713 }
02aa26ce
NT
2714
2715 /* range begins (ignore - as first or last char) */
79072805 2716 else if (*s == '-' && s+1 < send && s != start) {
4e553d73 2717 if (didrange) {
1fafa243 2718 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
01ec43d0 2719 }
e294cc5d
JH
2720 if (has_utf8
2721#ifdef EBCDIC
2722 && !native_range
2723#endif
2724 ) {
25716404 2725 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
a0ed51b3
LW
2726 s++;
2727 continue;
2728 }
79072805
LW
2729 dorange = TRUE;
2730 s++;
01ec43d0
GS
2731 }
2732 else {
2733 didrange = FALSE;
4c3a8340
ST
2734#ifdef EBCDIC
2735 literal_endpoint = 0;
e294cc5d 2736 native_range = TRUE;
4c3a8340 2737#endif
01ec43d0 2738 }
79072805 2739 }
02aa26ce
NT
2740
2741 /* if we get here, we're not doing a transliteration */
2742
0f5d15d6
IZ
2743 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2744 except for the last char, which will be done separately. */
3280af22 2745 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
cc6b7395 2746 if (s[2] == '#') {
e994fd66 2747 while (s+1 < send && *s != ')')
db42d148 2748 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
155aba94
GS
2749 }
2750 else if (s[2] == '{' /* This should match regcomp.c */
67edc0c9 2751 || (s[2] == '?' && s[3] == '{'))
155aba94 2752 {
cc6b7395 2753 I32 count = 1;
0f5d15d6 2754 char *regparse = s + (s[2] == '{' ? 3 : 4);
cc6b7395
IZ
2755 char c;
2756
d9f97599
GS
2757 while (count && (c = *regparse)) {
2758 if (c == '\\' && regparse[1])
2759 regparse++;
4e553d73 2760 else if (c == '{')
cc6b7395 2761 count++;
4e553d73 2762 else if (c == '}')
cc6b7395 2763 count--;
d9f97599 2764 regparse++;
cc6b7395 2765 }
e994fd66 2766 if (*regparse != ')')
5bdf89e7 2767 regparse--; /* Leave one char for continuation. */
0f5d15d6 2768 while (s < regparse)
db42d148 2769 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
cc6b7395 2770 }
748a9306 2771 }
02aa26ce
NT
2772
2773 /* likewise skip #-initiated comments in //x patterns */
3280af22
NIS
2774 else if (*s == '#' && PL_lex_inpat &&
2775 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
748a9306 2776 while (s+1 < send && *s != '\n')
db42d148 2777 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
748a9306 2778 }
02aa26ce 2779
5d1d4326 2780 /* check for embedded arrays
da6eedaa 2781 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
5d1d4326 2782 */
1749ea0d
ST
2783 else if (*s == '@' && s[1]) {
2784 if (isALNUM_lazy_if(s+1,UTF))
2785 break;
2786 if (strchr(":'{$", s[1]))
2787 break;
2788 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2789 break; /* in regexp, neither @+ nor @- are interpolated */
2790 }
02aa26ce
NT
2791
2792 /* check for embedded scalars. only stop if we're sure it's a
2793 variable.
2794 */
79072805 2795 else if (*s == '$') {
3280af22 2796 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 2797 break;
77772344 2798 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
2799 if (s[1] == '\\') {
2800 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2801 "Possible unintended interpolation of $\\ in regex");
77772344 2802 }
79072805 2803 break; /* in regexp, $ might be tail anchor */
77772344 2804 }
79072805 2805 }
02aa26ce 2806
2b9d42f0
NIS
2807 /* End of else if chain - OP_TRANS rejoin rest */
2808
02aa26ce 2809 /* backslashes */
79072805 2810 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
2811 char* e; /* Can be used for ending '}', etc. */
2812
79072805 2813 s++;
02aa26ce 2814
7d0fc23c
KW
2815 /* warn on \1 - \9 in substitution replacements, but note that \11
2816 * is an octal; and \19 is \1 followed by '9' */
3280af22 2817 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
a0d0e21e 2818 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
79072805 2819 {
a2a5de95 2820 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
2821 *--s = '$';
2822 break;
2823 }
02aa26ce
NT
2824
2825 /* string-change backslash escapes */
3280af22 2826 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
79072805
LW
2827 --s;
2828 break;
2829 }
ff3f963a
KW
2830 /* In a pattern, process \N, but skip any other backslash escapes.
2831 * This is because we don't want to translate an escape sequence
2832 * into a meta symbol and have the regex compiler use the meta
2833 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2834 * in spite of this, we do have to process \N here while the proper
2835 * charnames handler is in scope. See bugs #56444 and #62056.
2836 * There is a complication because \N in a pattern may also stand
2837 * for 'match a non-nl', and not mean a charname, in which case its
2838 * processing should be deferred to the regex compiler. To be a
2839 * charname it must be followed immediately by a '{', and not look
2840 * like \N followed by a curly quantifier, i.e., not something like
2841 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2842 * quantifier */
2843 else if (PL_lex_inpat
2844 && (*s != 'N'
2845 || s[1] != '{'
2846 || regcurly(s + 1)))
2847 {
cc74c5bd
ST
2848 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2849 goto default_action;
2850 }
02aa26ce 2851
79072805 2852 switch (*s) {
02aa26ce
NT
2853
2854 /* quoted - in transliterations */
79072805 2855 case '-':
3280af22 2856 if (PL_lex_inwhat == OP_TRANS) {
79072805
LW
2857 *d++ = *s++;
2858 continue;
2859 }
2860 /* FALL THROUGH */
2861 default:
11b8faa4 2862 {
a2a5de95
NC
2863 if ((isALPHA(*s) || isDIGIT(*s)))
2864 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2865 "Unrecognized escape \\%c passed through",
2866 *s);
11b8faa4 2867 /* default action is to copy the quoted character */
f9a63242 2868 goto default_action;
11b8faa4 2869 }
02aa26ce 2870
632403cc 2871 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
2872 case '0': case '1': case '2': case '3':
2873 case '4': case '5': case '6': case '7':
ba210ebe 2874 {
53305cf1
NC
2875 I32 flags = 0;
2876 STRLEN len = 3;
77a135fe 2877 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
ba210ebe
JH
2878 s += len;
2879 }
012bcf8d 2880 goto NUM_ESCAPE_INSERT;
02aa26ce 2881
f0a2b745
KW
2882 /* eg. \o{24} indicates the octal constant \024 */
2883 case 'o':
2884 {
2885 STRLEN len;
454155d9 2886 const char* error;
f0a2b745 2887
454155d9 2888 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
f0a2b745 2889 s += len;
454155d9 2890 if (! valid) {
f0a2b745
KW
2891 yyerror(error);
2892 continue;
2893 }
2894 goto NUM_ESCAPE_INSERT;
2895 }
2896
77a135fe 2897 /* eg. \x24 indicates the hex constant 0x24 */
79072805 2898 case 'x':
a0ed51b3
LW
2899 ++s;
2900 if (*s == '{') {
9d4ba2ae 2901 char* const e = strchr(s, '}');
a4c04bdc
NC
2902 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2903 PERL_SCAN_DISALLOW_PREFIX;
53305cf1 2904 STRLEN len;
355860ce 2905
53305cf1 2906 ++s;
adaeee49 2907 if (!e) {
a0ed51b3 2908 yyerror("Missing right brace on \\x{}");
355860ce 2909 continue;
ba210ebe 2910 }
53305cf1 2911 len = e - s;
77a135fe 2912 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
ba210ebe 2913 s = e + 1;
a0ed51b3
LW
2914 }
2915 else {
ba210ebe 2916 {
53305cf1 2917 STRLEN len = 2;
a4c04bdc 2918 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
77a135fe 2919 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
ba210ebe
JH
2920 s += len;
2921 }
012bcf8d
GS
2922 }
2923
2924 NUM_ESCAPE_INSERT:
ff3f963a
KW
2925 /* Insert oct or hex escaped character. There will always be
2926 * enough room in sv since such escapes will be longer than any
2927 * UTF-8 sequence they can end up as, except if they force us
2928 * to recode the rest of the string into utf8 */
ba7cea30 2929
77a135fe 2930 /* Here uv is the ordinal of the next character being added in
ff3f963a 2931 * unicode (converted from native). */
77a135fe 2932 if (!UNI_IS_INVARIANT(uv)) {
9aa983d2 2933 if (!has_utf8 && uv > 255) {
77a135fe
KW
2934 /* Might need to recode whatever we have accumulated so
2935 * far if it contains any chars variant in utf8 or
2936 * utf-ebcdic. */
2937
2938 SvCUR_set(sv, d - SvPVX_const(sv));
2939 SvPOK_on(sv);
2940 *d = '\0';
77a135fe 2941 /* See Note on sizing above. */
7bf79863
KW
2942 sv_utf8_upgrade_flags_grow(sv,
2943 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2944 UNISKIP(uv) + (STRLEN)(send - s) + 1);
77a135fe
KW
2945 d = SvPVX(sv) + SvCUR(sv);
2946 has_utf8 = TRUE;
012bcf8d
GS
2947 }
2948
77a135fe
KW
2949 if (has_utf8) {
2950 d = (char*)uvuni_to_utf8((U8*)d, uv);
f9a63242
JH
2951 if (PL_lex_inwhat == OP_TRANS &&
2952 PL_sublex_info.sub_op) {
2953 PL_sublex_info.sub_op->op_private |=
2954 (PL_lex_repl ? OPpTRANS_FROM_UTF
2955 : OPpTRANS_TO_UTF);
f9a63242 2956 }
e294cc5d
JH
2957#ifdef EBCDIC
2958 if (uv > 255 && !dorange)
2959 native_range = FALSE;
2960#endif
012bcf8d 2961 }
a0ed51b3 2962 else {
012bcf8d 2963 *d++ = (char)uv;
a0ed51b3 2964 }
012bcf8d
GS
2965 }
2966 else {
c4d5f83a 2967 *d++ = (char) uv;
a0ed51b3 2968 }
79072805 2969 continue;
02aa26ce 2970
4a2d328f 2971 case 'N':
ff3f963a
KW
2972 /* In a non-pattern \N must be a named character, like \N{LATIN
2973 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2974 * mean to match a non-newline. For non-patterns, named
2975 * characters are converted to their string equivalents. In
2976 * patterns, named characters are not converted to their
2977 * ultimate forms for the same reasons that other escapes
2978 * aren't. Instead, they are converted to the \N{U+...} form
2979 * to get the value from the charnames that is in effect right
2980 * now, while preserving the fact that it was a named character
2981 * so that the regex compiler knows this */
2982
2983 /* This section of code doesn't generally use the
2984 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2985 * a close examination of this macro and determined it is a
2986 * no-op except on utfebcdic variant characters. Every
2987 * character generated by this that would normally need to be
2988 * enclosed by this macro is invariant, so the macro is not
2989 * needed, and would complicate use of copy(). There are other
2990 * parts of this file where the macro is used inconsistently,
2991 * but are saved by it being a no-op */
2992
2993 /* The structure of this section of code (besides checking for
2994 * errors and upgrading to utf8) is:
2995 * Further disambiguate between the two meanings of \N, and if
2996 * not a charname, go process it elsewhere
0a96133f
KW
2997 * If of form \N{U+...}, pass it through if a pattern;
2998 * otherwise convert to utf8
2999 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3000 * pattern; otherwise convert to utf8 */
ff3f963a
KW
3001
3002 /* Here, s points to the 'N'; the test below is guaranteed to
3003 * succeed if we are being called on a pattern as we already
3004 * know from a test above that the next character is a '{'.
3005 * On a non-pattern \N must mean 'named sequence, which
3006 * requires braces */
3007 s++;
3008 if (*s != '{') {
3009 yyerror("Missing braces on \\N{}");
3010 continue;
3011 }
3012 s++;
3013
0a96133f 3014 /* If there is no matching '}', it is an error. */
ff3f963a
KW
3015 if (! (e = strchr(s, '}'))) {
3016 if (! PL_lex_inpat) {
5777a3f7 3017 yyerror("Missing right brace on \\N{}");
0a96133f
KW
3018 } else {
3019 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
dbc0d4f2 3020 }
0a96133f 3021 continue;
ff3f963a 3022 }
cddc7ef4 3023
ff3f963a 3024 /* Here it looks like a named character */
cddc7ef4 3025
ff3f963a
KW
3026 if (PL_lex_inpat) {
3027
3028 /* XXX This block is temporary code. \N{} implies that the
3029 * pattern is to have Unicode semantics, and therefore
3030 * currently has to be encoded in utf8. By putting it in
3031 * utf8 now, we save a whole pass in the regular expression
3032 * compiler. Once that code is changed so Unicode
3033 * semantics doesn't necessarily have to be in utf8, this
3034 * block should be removed */
3035 if (!has_utf8) {
77a135fe 3036 SvCUR_set(sv, d - SvPVX_const(sv));
f08d6ad9 3037 SvPOK_on(sv);
e4f3eed8 3038 *d = '\0';
77a135fe 3039 /* See Note on sizing above. */
7bf79863 3040 sv_utf8_upgrade_flags_grow(sv,
ff3f963a
KW
3041 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3042 /* 5 = '\N{' + cur char + NUL */
3043 (STRLEN)(send - s) + 5);
f08d6ad9 3044 d = SvPVX(sv) + SvCUR(sv);
89491803 3045 has_utf8 = TRUE;
ff3f963a
KW
3046 }
3047 }
423cee85 3048
ff3f963a
KW
3049 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3050 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3051 | PERL_SCAN_DISALLOW_PREFIX;
3052 STRLEN len;
3053
3054 /* For \N{U+...}, the '...' is a unicode value even on
3055 * EBCDIC machines */
3056 s += 2; /* Skip to next char after the 'U+' */
3057 len = e - s;
3058 uv = grok_hex(s, &len, &flags, NULL);
3059 if (len == 0 || len != (STRLEN)(e - s)) {
3060 yyerror("Invalid hexadecimal number in \\N{U+...}");
3061 s = e + 1;
3062 continue;
3063 }
3064
3065 if (PL_lex_inpat) {
3066
3067 /* Pass through to the regex compiler unchanged. The
3068 * reason we evaluated the number above is to make sure
0a96133f 3069 * there wasn't a syntax error. */
ff3f963a
KW
3070 s -= 5; /* Include the '\N{U+' */
3071 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3072 d += e - s + 1;
3073 }
3074 else { /* Not a pattern: convert the hex to string */
3075
3076 /* If destination is not in utf8, unconditionally
3077 * recode it to be so. This is because \N{} implies
3078 * Unicode semantics, and scalars have to be in utf8
3079 * to guarantee those semantics */
3080 if (! has_utf8) {
3081 SvCUR_set(sv, d - SvPVX_const(sv));
3082 SvPOK_on(sv);
3083 *d = '\0';
3084 /* See Note on sizing above. */
3085 sv_utf8_upgrade_flags_grow(
3086 sv,
3087 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3088 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3089 d = SvPVX(sv) + SvCUR(sv);
3090 has_utf8 = TRUE;
3091 }
3092
3093 /* Add the string to the output */
3094 if (UNI_IS_INVARIANT(uv)) {
3095 *d++ = (char) uv;
3096 }
3097 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3098 }
3099 }
3100 else { /* Here is \N{NAME} but not \N{U+...}. */
3101
3102 SV *res; /* result from charnames */
3103 const char *str; /* the string in 'res' */
3104 STRLEN len; /* its length */
3105
3106 /* Get the value for NAME */
3107 res = newSVpvn(s, e - s);
3108 res = new_constant( NULL, 0, "charnames",
3109 /* includes all of: \N{...} */
3110 res, NULL, s - 3, e - s + 4 );
3111
3112 /* Most likely res will be in utf8 already since the
3113 * standard charnames uses pack U, but a custom translator
3114 * can leave it otherwise, so make sure. XXX This can be
3115 * revisited to not have charnames use utf8 for characters
3116 * that don't need it when regexes don't have to be in utf8
3117 * for Unicode semantics. If doing so, remember EBCDIC */
3118 sv_utf8_upgrade(res);
3119 str = SvPV_const(res, len);
3120
3121 /* Don't accept malformed input */
3122 if (! is_utf8_string((U8 *) str, len)) {
3123 yyerror("Malformed UTF-8 returned by \\N");
3124 }
3125 else if (PL_lex_inpat) {
3126
3127 if (! len) { /* The name resolved to an empty string */
3128 Copy("\\N{}", d, 4, char);
3129 d += 4;
3130 }
3131 else {
3132 /* In order to not lose information for the regex
3133 * compiler, pass the result in the specially made
3134 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3135 * the code points in hex of each character
3136 * returned by charnames */
3137
3138 const char *str_end = str + len;
3139 STRLEN char_length; /* cur char's byte length */
3140 STRLEN output_length; /* and the number of bytes
3141 after this is translated
3142 into hex digits */
3143 const STRLEN off = d - SvPVX_const(sv);
3144
3145 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3146 * max('U+', '.'); and 1 for NUL */
3147 char hex_string[2 * UTF8_MAXBYTES + 5];
3148
3149 /* Get the first character of the result. */
3150 U32 uv = utf8n_to_uvuni((U8 *) str,
3151 len,
3152 &char_length,
3153 UTF8_ALLOW_ANYUV);
3154
3155 /* The call to is_utf8_string() above hopefully
3156 * guarantees that there won't be an error. But
3157 * it's easy here to make sure. The function just
3158 * above warns and returns 0 if invalid utf8, but
3159 * it can also return 0 if the input is validly a
3160 * NUL. Disambiguate */
3161 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3162 uv = UNICODE_REPLACEMENT;
3163 }
3164
3165 /* Convert first code point to hex, including the
3166 * boiler plate before it */
3167 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3168 output_length = strlen(hex_string);
3169
3170 /* Make sure there is enough space to hold it */
3171 d = off + SvGROW(sv, off
3172 + output_length
3173 + (STRLEN)(send - e)
3174 + 2); /* '}' + NUL */
3175 /* And output it */
3176 Copy(hex_string, d, output_length, char);
3177 d += output_length;
3178
3179 /* For each subsequent character, append dot and
3180 * its ordinal in hex */
3181 while ((str += char_length) < str_end) {
3182 const STRLEN off = d - SvPVX_const(sv);
3183 U32 uv = utf8n_to_uvuni((U8 *) str,
3184 str_end - str,
3185 &char_length,
3186 UTF8_ALLOW_ANYUV);
3187 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3188 uv = UNICODE_REPLACEMENT;
3189 }
3190
3191 sprintf(hex_string, ".%X", (unsigned int) uv);
3192 output_length = strlen(hex_string);
3193
3194 d = off + SvGROW(sv, off
3195 + output_length
3196 + (STRLEN)(send - e)
3197 + 2); /* '}' + NUL */
3198 Copy(hex_string, d, output_length, char);
3199 d += output_length;
3200 }
3201
3202 *d++ = '}'; /* Done. Add the trailing brace */
3203 }
3204 }
3205 else { /* Here, not in a pattern. Convert the name to a
3206 * string. */
3207
3208 /* If destination is not in utf8, unconditionally
3209 * recode it to be so. This is because \N{} implies
3210 * Unicode semantics, and scalars have to be in utf8
3211 * to guarantee those semantics */
3212 if (! has_utf8) {
3213 SvCUR_set(sv, d - SvPVX_const(sv));
3214 SvPOK_on(sv);
3215 *d = '\0';
3216 /* See Note on sizing above. */
3217 sv_utf8_upgrade_flags_grow(sv,
3218 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3219 len + (STRLEN)(send - s) + 1);
3220 d = SvPVX(sv) + SvCUR(sv);
3221 has_utf8 = TRUE;
3222 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3223
3224 /* See Note on sizing above. (NOTE: SvCUR() is not
3225 * set correctly here). */
3226 const STRLEN off = d - SvPVX_const(sv);
3227 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3228 }
3229 Copy(str, d, len, char);
3230 d += len;
423cee85 3231 }
423cee85 3232 SvREFCNT_dec(res);
cb233ae3
KW
3233
3234 /* Deprecate non-approved name syntax */
3235 if (ckWARN_d(WARN_DEPRECATED)) {
3236 bool problematic = FALSE;
3237 char* i = s;
3238
3239 /* For non-ut8 input, look to see that the first
3240 * character is an alpha, then loop through the rest
3241 * checking that each is a continuation */
3242 if (! this_utf8) {
3243 if (! isALPHAU(*i)) problematic = TRUE;
3244 else for (i = s + 1; i < e; i++) {
3245 if (isCHARNAME_CONT(*i)) continue;
3246 problematic = TRUE;
3247 break;
3248 }
3249 }
3250 else {
3251 /* Similarly for utf8. For invariants can check
3252 * directly. We accept anything above the latin1
3253 * range because it is immaterial to Perl if it is
3254 * correct or not, and is expensive to check. But
3255 * it is fairly easy in the latin1 range to convert
3256 * the variants into a single character and check
3257 * those */
3258 if (UTF8_IS_INVARIANT(*i)) {
3259 if (! isALPHAU(*i)) problematic = TRUE;
3260 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3261 if (! isALPHAU(UNI_TO_NATIVE(UTF8_ACCUMULATE(*i,
3262 *(i+1)))))
3263 {
3264 problematic = TRUE;
3265 }
3266 }
3267 if (! problematic) for (i = s + UTF8SKIP(s);
3268 i < e;
3269 i+= UTF8SKIP(i))
3270 {
3271 if (UTF8_IS_INVARIANT(*i)) {
3272 if (isCHARNAME_CONT(*i)) continue;
3273 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3274 continue;
3275 } else if (isCHARNAME_CONT(
3276 UNI_TO_NATIVE(
3277 UTF8_ACCUMULATE(*i, *(i+1)))))
3278 {
3279 continue;
3280 }
3281 problematic = TRUE;
3282 break;
3283 }
3284 }
3285 if (problematic) {
6e1bad6c
KW
3286 /* The e-i passed to the final %.*s makes sure that
3287 * should the trailing NUL be missing that this
3288 * print won't run off the end of the string */
cb233ae3 3289 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
b00fc8d4
NC
3290 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3291 (int)(i - s + 1), s, (int)(e - i), i + 1);
cb233ae3
KW
3292 }
3293 }
3294 } /* End \N{NAME} */
ff3f963a
KW
3295#ifdef EBCDIC
3296 if (!dorange)
3297 native_range = FALSE; /* \N{} is defined to be Unicode */
3298#endif
3299 s = e + 1; /* Point to just after the '}' */
423cee85
JH
3300 continue;
3301
02aa26ce 3302 /* \c is a control character */
79072805
LW
3303 case 'c':
3304 s++;
961ce445 3305 if (s < send) {
f9d13529 3306 *d++ = grok_bslash_c(*s++, 1);
ba210ebe 3307 }
961ce445
RGS
3308 else {
3309 yyerror("Missing control char name in \\c");
3310 }
79072805 3311 continue;
02aa26ce
NT
3312
3313 /* printf-style backslashes, formfeeds, newlines, etc */
79072805 3314 case 'b':
db42d148 3315 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
79072805
LW
3316 break;
3317 case 'n':
db42d148 3318 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
79072805
LW
3319 break;
3320 case 'r':
db42d148 3321 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
79072805
LW
3322 break;
3323 case 'f':
db42d148 3324 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
79072805
LW
3325 break;
3326 case 't':
db42d148 3327 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
79072805 3328 break;
34a3fe2a 3329 case 'e':
db42d148 3330 *d++ = ASCII_TO_NEED(has_utf8,'\033');
34a3fe2a
PP
3331 break;
3332 case 'a':
db42d148 3333 *d++ = ASCII_TO_NEED(has_utf8,'\007');
79072805 3334 break;
02aa26ce
NT
3335 } /* end switch */
3336
79072805
LW
3337 s++;
3338 continue;
02aa26ce 3339 } /* end if (backslash) */
4c3a8340
ST
3340#ifdef EBCDIC
3341 else
3342 literal_endpoint++;
3343#endif
02aa26ce 3344
f9a63242 3345 default_action:
77a135fe
KW
3346 /* If we started with encoded form, or already know we want it,
3347 then encode the next character */
3348 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2b9d42f0 3349 STRLEN len = 1;
77a135fe
KW
3350
3351
3352 /* One might think that it is wasted effort in the case of the
3353 * source being utf8 (this_utf8 == TRUE) to take the next character
3354 * in the source, convert it to an unsigned value, and then convert
3355 * it back again. But the source has not been validated here. The
3356 * routine that does the conversion checks for errors like
3357 * malformed utf8 */
3358
5f66b61c
AL
3359 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3360 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
77a135fe
KW
3361 if (!has_utf8) {
3362 SvCUR_set(sv, d - SvPVX_const(sv));
3363 SvPOK_on(sv);
3364 *d = '\0';
77a135fe 3365 /* See Note on sizing above. */
7bf79863
KW
3366 sv_utf8_upgrade_flags_grow(sv,
3367 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3368 need + (STRLEN)(send - s) + 1);
77a135fe
KW
3369 d = SvPVX(sv) + SvCUR(sv);
3370 has_utf8 = TRUE;
3371 } else if (need > len) {
3372 /* encoded value larger than old, may need extra space (NOTE:
3373 * SvCUR() is not set correctly here). See Note on sizing
3374 * above. */
9d4ba2ae 3375 const STRLEN off = d - SvPVX_const(sv);
77a135fe 3376 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2b9d42f0 3377 }
77a135fe
KW
3378 s += len;
3379
5f66b61c 3380 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
e294cc5d
JH
3381#ifdef EBCDIC
3382 if (uv > 255 && !dorange)
3383 native_range = FALSE;
3384#endif
2b9d42f0
NIS
3385 }
3386 else {
3387 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3388 }
02aa26ce
NT
3389 } /* while loop to process each character */
3390
3391 /* terminate the string and set up the sv */
79072805 3392 *d = '\0';
95a20fc0 3393 SvCUR_set(sv, d - SvPVX_const(sv));
2b9d42f0 3394 if (SvCUR(sv) >= SvLEN(sv))
d0063567 3395 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2b9d42f0 3396
79072805 3397 SvPOK_on(sv);
9f4817db 3398 if (PL_encoding && !has_utf8) {
d0063567
DK
3399 sv_recode_to_utf8(sv, PL_encoding);
3400 if (SvUTF8(sv))
3401 has_utf8 = TRUE;
9f4817db 3402 }
2b9d42f0 3403 if (has_utf8) {
7e2040f0 3404 SvUTF8_on(sv);
2b9d42f0 3405 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
d0063567 3406 PL_sublex_info.sub_op->op_private |=
2b9d42f0
NIS
3407 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3408 }
3409 }
79072805 3410
02aa26ce 3411 /* shrink the sv if we allocated more than we used */
79072805 3412 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1da4ca5f 3413 SvPV_shrink_to_cur(sv);
79072805 3414 }
02aa26ce 3415
6154021b 3416 /* return the substring (via pl_yylval) only if we parsed anything */
3280af22 3417 if (s > PL_bufptr) {
eb0d8d16
NC
3418 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3419 const char *const key = PL_lex_inpat ? "qr" : "q";
3420 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3421 const char *type;
3422 STRLEN typelen;
3423
3424 if (PL_lex_inwhat == OP_TRANS) {
3425 type = "tr";
3426 typelen = 2;
3427 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3428 type = "s";
3429 typelen = 1;
3430 } else {
3431 type = "qq";
3432 typelen = 2;
3433 }
3434
3435 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3436 type, typelen);
3437 }
6154021b 3438 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
b3ac6de7 3439 } else
8990e307 3440 SvREFCNT_dec(sv);
79072805
LW
3441 return s;
3442}
3443
ffb4593c
NT
3444/* S_intuit_more
3445 * Returns TRUE if there's more to the expression (e.g., a subscript),
3446 * FALSE otherwise.
ffb4593c
NT
3447 *
3448 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3449 *
3450 * ->[ and ->{ return TRUE
3451 * { and [ outside a pattern are always subscripts, so return TRUE
3452 * if we're outside a pattern and it's not { or [, then return FALSE
3453 * if we're in a pattern and the first char is a {
3454 * {4,5} (any digits around the comma) returns FALSE
3455 * if we're in a pattern and the first char is a [
3456 * [] returns FALSE
3457 * [SOMETHING] has a funky algorithm to decide whether it's a
3458 * character class or not. It has to deal with things like
3459 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3460 * anything else returns TRUE
3461 */
3462
9cbb5ea2
GS
3463/* This is the one truly awful dwimmer necessary to conflate C and sed. */
3464
76e3520e 3465STATIC int
cea2e8a9 3466S_intuit_more(pTHX_ register char *s)
79072805 3467{
97aff369 3468 dVAR;
7918f24d
NC
3469
3470 PERL_ARGS_ASSERT_INTUIT_MORE;
3471
3280af22 3472 if (PL_lex_brackets)
79072805
LW
3473 return TRUE;
3474 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3475 return TRUE;
3476 if (*s != '{' && *s != '[')
3477 return FALSE;
3280af22 3478 if (!PL_lex_inpat)
79072805
LW
3479 return TRUE;
3480
3481 /* In a pattern, so maybe we have {n,m}. */
3482 if (*s == '{') {
3483 s++;
3484 if (!isDIGIT(*s))
3485 return TRUE;
3486 while (isDIGIT(*s))
3487 s++;
3488 if (*s == ',')
3489 s++;
3490 while (isDIGIT(*s))
3491 s++;
3492 if (*s == '}')
3493 return FALSE;
3494 return TRUE;
3495
3496 }
3497
3498 /* On the other hand, maybe we have a character class */
3499
3500 s++;
3501 if (*s == ']' || *s == '^')
3502 return FALSE;
3503 else {
ffb4593c 3504 /* this is terrifying, and it works */
79072805
LW
3505 int weight = 2; /* let's weigh the evidence */
3506 char seen[256];
f27ffc4a 3507 unsigned char un_char = 255, last_un_char;
9d4ba2ae 3508 const char * const send = strchr(s,']');
3280af22 3509 char tmpbuf[sizeof PL_tokenbuf * 4];
79072805
LW
3510
3511 if (!send) /* has to be an expression */
3512 return TRUE;
3513
3514 Zero(seen,256,char);
3515 if (*s == '$')
3516 weight -= 3;
3517 else if (isDIGIT(*s)) {
3518 if (s[1] != ']') {
3519 if (isDIGIT(s[1]) && s[2] == ']')
3520 weight -= 10;
3521 }
3522 else
3523 weight -= 100;
3524 }
3525 for (; s < send; s++) {
3526 last_un_char = un_char;
3527 un_char = (unsigned char)*s;
3528 switch (*s) {
3529 case '@':
3530 case '&':
3531 case '$':
3532 weight -= seen[un_char] * 10;
7e2040f0 3533 if (isALNUM_lazy_if(s+1,UTF)) {
90e5519e 3534 int len;
8903cb82 3535 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
90e5519e
NC
3536 len = (int)strlen(tmpbuf);
3537 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
79072805
LW
3538 weight -= 100;
3539 else
3540 weight -= 10;
3541 }
3542 else if (*s == '$' && s[1] &&
93a17b20
LW
3543 strchr("[#!%*<>()-=",s[1])) {
3544 if (/*{*/ strchr("])} =",s[2]))
79072805
LW
3545 weight -= 10;
3546 else
3547 weight -= 1;
3548 }
3549 break;
3550 case '\\':
3551 un_char = 254;
3552 if (s[1]) {
93a17b20 3553 if (strchr("wds]",s[1]))
79072805 3554 weight += 100;
10edeb5d 3555 else if (seen[(U8)'\''] || seen[(U8)'"'])
79072805 3556 weight += 1;
93a17b20 3557 else if (strchr("rnftbxcav",s[1]))
79072805
LW
3558 weight += 40;
3559 else if (isDIGIT(s[1])) {
3560 weight += 40;
3561 while (s[1] && isDIGIT(s[1]))
3562 s++;
3563 }
3564 }
3565 else
3566 weight += 100;
3567 break;
3568 case '-':
3569 if (s[1] == '\\')
3570 weight += 50;
93a17b20 3571 if (strchr("aA01! ",last_un_char))
79072805 3572 weight += 30;
93a17b20 3573 if (strchr("zZ79~",s[1]))
79072805 3574 weight += 30;
f27ffc4a
GS
3575 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3576 weight -= 5; /* cope with negative subscript */
79072805
LW
3577 break;
3578 default:
3792a11b
NC
3579 if (!isALNUM(last_un_char)
3580 && !(last_un_char == '$' || last_un_char == '@'
3581 || last_un_char == '&')
3582 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
79072805
LW
3583 char *d = tmpbuf;
3584 while (isALPHA(*s))
3585 *d++ = *s++;
3586 *d = '\0';
5458a98a 3587 if (keyword(tmpbuf, d - tmpbuf, 0))
79072805
LW
3588 weight -= 150;
3589 }
3590 if (un_char == last_un_char + 1)
3591 weight += 5;
3592 weight -= seen[un_char];
3593 break;
3594 }
3595 seen[un_char]++;
3596 }
3597 if (weight >= 0) /* probably a character class */
3598 return FALSE;
3599 }
3600
3601 return TRUE;
3602}
ffed7fef 3603
ffb4593c
NT
3604/*
3605 * S_intuit_method
3606 *
3607 * Does all the checking to disambiguate
3608 * foo bar
3609 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3610 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3611 *
3612 * First argument is the stuff after the first token, e.g. "bar".
3613 *
3614 * Not a method if bar is a filehandle.
3615 * Not a method if foo is a subroutine prototyped to take a filehandle.
3616 * Not a method if it's really "Foo $bar"
3617 * Method if it's "foo $bar"
3618 * Not a method if it's really "print foo $bar"
3619 * Method if it's really "foo package::" (interpreted as package->foo)
8f8cf39c 3620 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3cb0bbe5 3621 * Not a method if bar is a filehandle or package, but is quoted with
ffb4593c
NT
3622 * =>
3623 */
3624
76e3520e 3625STATIC int
62d55b22 3626S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
a0d0e21e 3627{
97aff369 3628 dVAR;
a0d0e21e 3629 char *s = start + (*start == '$');
3280af22 3630 char tmpbuf[sizeof PL_tokenbuf];
a0d0e21e
LW
3631 STRLEN len;
3632 GV* indirgv;
5db06880
NC
3633#ifdef PERL_MAD
3634 int soff;
3635#endif
a0d0e21e 3636
7918f24d
NC
3637 PERL_ARGS_ASSERT_INTUIT_METHOD;
3638
a0d0e21e 3639 if (gv) {
62d55b22 3640 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
a0d0e21e 3641 return 0;
62d55b22
NC
3642 if (cv) {
3643 if (SvPOK(cv)) {
3644 const char *proto = SvPVX_const(cv);
3645 if (proto) {
3646 if (*proto == ';')
3647 proto++;
3648 if (*proto == '*')
3649 return 0;
3650 }
b6c543e3
IZ
3651 }
3652 } else
c35e046a 3653 gv = NULL;
a0d0e21e 3654 }
8903cb82 3655 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
ffb4593c
NT
3656 /* start is the beginning of the possible filehandle/object,
3657 * and s is the end of it
3658 * tmpbuf is a copy of it
3659 */
3660
a0d0e21e 3661 if (*start == '$') {
3ef1310e
RGS
3662 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3663 isUPPER(*PL_tokenbuf))
a0d0e21e 3664 return 0;
5db06880
NC
3665#ifdef PERL_MAD
3666 len = start - SvPVX(PL_linestr);
3667#endif
29595ff2 3668 s = PEEKSPACE(s);
f0092767 3669#ifdef PERL_MAD
5db06880
NC
3670 start = SvPVX(PL_linestr) + len;
3671#endif
3280af22
NIS
3672 PL_bufptr = start;
3673 PL_expect = XREF;
a0d0e21e
LW
3674 return *s == '(' ? FUNCMETH : METHOD;
3675 }
5458a98a 3676 if (!keyword(tmpbuf, len, 0)) {
c3e0f903
GS
3677 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3678 len -= 2;
3679 tmpbuf[len] = '\0';
5db06880
NC
3680#ifdef PERL_MAD
3681 soff = s - SvPVX(PL_linestr);
3682#endif
c3e0f903
GS
3683 goto bare_package;
3684 }
90e5519e 3685 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
8ebc5c01 3686 if (indirgv && GvCVu(indirgv))
a0d0e21e
LW
3687 return 0;
3688 /* filehandle or package name makes it a method */
da51bb9b 3689 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
5db06880
NC
3690#ifdef PERL_MAD
3691 soff = s - SvPVX(PL_linestr);
3692#endif
29595ff2 3693 s = PEEKSPACE(s);
3280af22 3694 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
55497cff 3695 return 0; /* no assumptions -- "=>" quotes bearword */
c3e0f903 3696 bare_package:
cd81e915 3697 start_force(PL_curforce);
9ded7720 3698 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
64142370 3699 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
9ded7720 3700 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
5db06880
NC
3701 if (PL_madskills)
3702 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3280af22 3703 PL_expect = XTERM;
a0d0e21e 3704 force_next(WORD);
3280af22 3705 PL_bufptr = s;
5db06880
NC
3706#ifdef PERL_MAD
3707 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3708#endif
a0d0e21e
LW
3709 return *s == '(' ? FUNCMETH : METHOD;
3710 }
3711 }
3712 return 0;
3713}
3714
16d20bd9 3715/* Encoded script support. filter_add() effectively inserts a
4e553d73 3716 * 'pre-processing' function into the current source input stream.
16d20bd9
AD
3717 * Note that the filter function only applies to the current source file
3718 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3719 *
3720 * The datasv parameter (which may be NULL) can be used to pass
3721 * private data to this instance of the filter. The filter function
3722 * can recover the SV using the FILTER_DATA macro and use it to
3723 * store private buffers and state information.
3724 *
3725 * The supplied datasv parameter is upgraded to a PVIO type
4755096e 3726 * and the IoDIRP/IoANY field is used to store the function pointer,
e0c19803 3727 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
16d20bd9
AD
3728 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3729 * private use must be set using malloc'd pointers.
3730 */
16d20bd9
AD
3731
3732SV *
864dbfa3 3733Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
16d20bd9 3734{
97aff369 3735 dVAR;
f4c556ac 3736 if (!funcp)
a0714e2c 3737 return NULL;
f4c556ac 3738
5486870f
DM
3739 if (!PL_parser)
3740 return NULL;
3741
3280af22
NIS
3742 if (!PL_rsfp_filters)
3743 PL_rsfp_filters = newAV();
16d20bd9 3744 if (!datasv)
561b68a9 3745 datasv = newSV(0);
862a34c6 3746 SvUPGRADE(datasv, SVt_PVIO);
8141890a 3747 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
e0c19803 3748 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
f4c556ac 3749 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
55662e27
JH
3750 FPTR2DPTR(void *, IoANY(datasv)),
3751 SvPV_nolen(datasv)));
3280af22
NIS
3752 av_unshift(PL_rsfp_filters, 1);
3753 av_store(PL_rsfp_filters, 0, datasv) ;
16d20bd9
AD
3754 return(datasv);
3755}
4e553d73 3756
16d20bd9
AD
3757
3758/* Delete most recently added instance of this filter function. */
a0d0e21e 3759void
864dbfa3 3760Perl_filter_del(pTHX_ filter_t funcp)
16d20bd9 3761{
97aff369 3762 dVAR;
e0c19803 3763 SV *datasv;
24801a4b 3764
7918f24d
NC
3765 PERL_ARGS_ASSERT_FILTER_DEL;
3766
33073adb 3767#ifdef DEBUGGING
55662e27
JH
3768 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3769 FPTR2DPTR(void*, funcp)));
33073adb 3770#endif
5486870f 3771 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
16d20bd9
AD
3772 return;
3773 /* if filter is on top of stack (usual case) just pop it off */
e0c19803 3774</