This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
[perl #114040] Parse here-docs correctly in quoted constructs
[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"
04e98a4d 42#include "dquote_static.c"
378cc40b 43
eb0d8d16
NC
44#define new_constant(a,b,c,d,e,f,g) \
45 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46
6154021b 47#define pl_yylval (PL_parser->yylval)
d3b6f988 48
199e78b7
DM
49/* XXX temporary backwards compatibility */
50#define PL_lex_brackets (PL_parser->lex_brackets)
78cdf107
Z
51#define PL_lex_allbrackets (PL_parser->lex_allbrackets)
52#define PL_lex_fakeeof (PL_parser->lex_fakeeof)
199e78b7
DM
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
a7aaec61
Z
129#define XENUMMASK 0x3f
130#define XFAKEEOF 0x40
131#define XFAKEBRACK 0x80
9059aa12 132
39e02b42
JH
133#ifdef USE_UTF8_SCRIPTS
134# define UTF (!IN_BYTES)
2b9d42f0 135#else
802a15e9 136# define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
2b9d42f0 137#endif
a0ed51b3 138
b1fc3636
CJ
139/* The maximum number of characters preceding the unrecognized one to display */
140#define UNRECOGNIZED_PRECEDE_COUNT 10
141
61f0cdd9 142/* In variables named $^X, these are the legal values for X.
2b92dfce
GS
143 * 1999-02-27 mjd-perl-patch@plover.com */
144#define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
145
bf4acbe4 146#define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
bf4acbe4 147
ffb4593c
NT
148/* LEX_* are values for PL_lex_state, the state of the lexer.
149 * They are arranged oddly so that the guard on the switch statement
79072805 150 * can get by with a single comparison (if the compiler is smart enough).
9da1dd8f
DM
151 *
152 * These values refer to the various states within a sublex parse,
153 * i.e. within a double quotish string
79072805
LW
154 */
155
fb73857a
PP
156/* #define LEX_NOTPARSING 11 is done in perl.h. */
157
b6007c36
DM
158#define LEX_NORMAL 10 /* normal code (ie not within "...") */
159#define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
160#define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
161#define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
162#define LEX_INTERPSTART 6 /* expecting the start of a $var */
163
164 /* at end of code, eg "$x" followed by: */
165#define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
166#define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
167
168#define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
169 string or after \E, $foo, etc */
170#define LEX_INTERPCONST 2 /* NOT USED */
171#define LEX_FORMLINE 1 /* expecting a format line */
172#define LEX_KNOWNEXT 0 /* next token known; just return it */
173
79072805 174
bbf60fe6 175#ifdef DEBUGGING
27da23d5 176static const char* const lex_state_names[] = {
bbf60fe6
DM
177 "KNOWNEXT",
178 "FORMLINE",
179 "INTERPCONST",
180 "INTERPCONCAT",
181 "INTERPENDMAYBE",
182 "INTERPEND",
183 "INTERPSTART",
184 "INTERPPUSH",
185 "INTERPCASEMOD",
186 "INTERPNORMAL",
187 "NORMAL"
188};
189#endif
190
79072805
LW
191#ifdef ff_next
192#undef ff_next
d48672a2
LW
193#endif
194
79072805 195#include "keywords.h"
fe14fcc3 196
ffb4593c
NT
197/* CLINE is a macro that ensures PL_copline has a sane value */
198
ae986130
LW
199#ifdef CLINE
200#undef CLINE
201#endif
57843af0 202#define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
3280af22 203
5db06880 204#ifdef PERL_MAD
29595ff2
NC
205# define SKIPSPACE0(s) skipspace0(s)
206# define SKIPSPACE1(s) skipspace1(s)
207# define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
208# define PEEKSPACE(s) skipspace2(s,0)
209#else
210# define SKIPSPACE0(s) skipspace(s)
211# define SKIPSPACE1(s) skipspace(s)
212# define SKIPSPACE2(s,tsv) skipspace(s)
213# define PEEKSPACE(s) skipspace(s)
214#endif
215
ffb4593c
NT
216/*
217 * Convenience functions to return different tokens and prime the
9cbb5ea2 218 * lexer for the next token. They all take an argument.
ffb4593c
NT
219 *
220 * TOKEN : generic token (used for '(', DOLSHARP, etc)
221 * OPERATOR : generic operator
222 * AOPERATOR : assignment operator
223 * PREBLOCK : beginning the block after an if, while, foreach, ...
224 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
225 * PREREF : *EXPR where EXPR is not a simple identifier
226 * TERM : expression term
227 * LOOPX : loop exiting command (goto, last, dump, etc)
228 * FTST : file test operator
229 * FUN0 : zero-argument function
7eb971ee 230 * FUN0OP : zero-argument function, with its op created in this file
2d2e263d 231 * FUN1 : not used, except for not, which isn't a UNIOP
ffb4593c
NT
232 * BOop : bitwise or or xor
233 * BAop : bitwise and
234 * SHop : shift operator
235 * PWop : power operator
9cbb5ea2 236 * PMop : pattern-matching operator
ffb4593c
NT
237 * Aop : addition-level operator
238 * Mop : multiplication-level operator
239 * Eop : equality-testing operator
e5edeb50 240 * Rop : relational operator <= != gt
ffb4593c
NT
241 *
242 * Also see LOP and lop() below.
243 */
244
998054bd 245#ifdef DEBUGGING /* Serve -DT. */
704d4215 246# define REPORT(retval) tokereport((I32)retval, &pl_yylval)
998054bd 247#else
bbf60fe6 248# define REPORT(retval) (retval)
998054bd
SC
249#endif
250
bbf60fe6
DM
251#define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
252#define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
253#define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
254#define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
255#define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
256#define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
257#define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
6154021b
RGS
258#define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
259#define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
260#define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
7eb971ee 261#define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
6154021b
RGS
262#define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
263#define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
264#define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
265#define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
266#define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
267#define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
268#define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
269#define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
270#define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
271#define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
2f3197b3 272
a687059c
LW
273/* This bit of chicanery makes a unary function followed by
274 * a parenthesis into a function with one argument, highest precedence.
6f33ba73
RGS
275 * The UNIDOR macro is for unary functions that can be followed by the //
276 * operator (such as C<shift // 0>).
a687059c 277 */
d68ce4ac 278#define UNI3(f,x,have_x) { \
6154021b 279 pl_yylval.ival = f; \
d68ce4ac 280 if (have_x) PL_expect = x; \
376fcdbf
AL
281 PL_bufptr = s; \
282 PL_last_uni = PL_oldbufptr; \
283 PL_last_lop_op = f; \
284 if (*s == '(') \
285 return REPORT( (int)FUNC1 ); \
29595ff2 286 s = PEEKSPACE(s); \
376fcdbf
AL
287 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
288 }
d68ce4ac
FC
289#define UNI(f) UNI3(f,XTERM,1)
290#define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
b5fb7ce3
FC
291#define UNIPROTO(f,optional) { \
292 if (optional) PL_last_uni = PL_oldbufptr; \
22393538
MH
293 OPERATOR(f); \
294 }
a687059c 295
d68ce4ac 296#define UNIBRACK(f) UNI3(f,0,0)
79072805 297
9f68db38 298/* grandfather return to old style */
78cdf107
Z
299#define OLDLOP(f) \
300 do { \
301 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
302 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
303 pl_yylval.ival = (f); \
304 PL_expect = XTERM; \
305 PL_bufptr = s; \
306 return (int)LSTOP; \
307 } while(0)
79072805 308
8fa7f367
JH
309#ifdef DEBUGGING
310
6154021b 311/* how to interpret the pl_yylval associated with the token */
bbf60fe6
DM
312enum token_type {
313 TOKENTYPE_NONE,
314 TOKENTYPE_IVAL,
6154021b 315 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
bbf60fe6 316 TOKENTYPE_PVAL,
aeaef349 317 TOKENTYPE_OPVAL
bbf60fe6
DM
318};
319
6d4a66ac
NC
320static struct debug_tokens {
321 const int token;
322 enum token_type type;
323 const char *name;
324} const debug_tokens[] =
9041c2e3 325{
bbf60fe6
DM
326 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
327 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
328 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
329 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
330 { ARROW, TOKENTYPE_NONE, "ARROW" },
331 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
332 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
333 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
334 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
335 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
0d863452 336 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
bbf60fe6
DM
337 { DO, TOKENTYPE_NONE, "DO" },
338 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
339 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
340 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
341 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
342 { ELSE, TOKENTYPE_NONE, "ELSE" },
343 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
344 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
345 { FOR, TOKENTYPE_IVAL, "FOR" },
346 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
705fe0e5
FC
347 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
348 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
bbf60fe6
DM
349 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
350 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
7eb971ee 351 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
bbf60fe6
DM
352 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
353 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
354 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
0d863452 355 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
bbf60fe6
DM
356 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
357 { IF, TOKENTYPE_IVAL, "IF" },
5db1eb8d 358 { LABEL, TOKENTYPE_OPVAL, "LABEL" },
bbf60fe6
DM
359 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
360 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
361 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
362 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
363 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
364 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
365 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
366 { MY, TOKENTYPE_IVAL, "MY" },
367 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
368 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
369 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
370 { OROP, TOKENTYPE_IVAL, "OROP" },
371 { OROR, TOKENTYPE_NONE, "OROR" },
372 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
f3f204dc 373 { PEG, TOKENTYPE_NONE, "PEG" },
88e1f1a2
JV
374 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
375 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
bbf60fe6
DM
376 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
377 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
378 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
379 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
380 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
381 { PREINC, TOKENTYPE_NONE, "PREINC" },
382 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
f3f204dc 383 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
bbf60fe6
DM
384 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
385 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
f3f204dc 386 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
bbf60fe6
DM
387 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
388 { SUB, TOKENTYPE_NONE, "SUB" },
389 { THING, TOKENTYPE_OPVAL, "THING" },
390 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
391 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
392 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
393 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
394 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
395 { USE, TOKENTYPE_IVAL, "USE" },
0d863452 396 { WHEN, TOKENTYPE_IVAL, "WHEN" },
bbf60fe6
DM
397 { WHILE, TOKENTYPE_IVAL, "WHILE" },
398 { WORD, TOKENTYPE_OPVAL, "WORD" },
be25f609 399 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
c35e046a 400 { 0, TOKENTYPE_NONE, NULL }
bbf60fe6
DM
401};
402
6154021b 403/* dump the returned token in rv, plus any optional arg in pl_yylval */
998054bd 404
bbf60fe6 405STATIC int
704d4215 406S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
bbf60fe6 407{
97aff369 408 dVAR;
7918f24d
NC
409
410 PERL_ARGS_ASSERT_TOKEREPORT;
411
bbf60fe6 412 if (DEBUG_T_TEST) {
bd61b366 413 const char *name = NULL;
bbf60fe6 414 enum token_type type = TOKENTYPE_NONE;
f54cb97a 415 const struct debug_tokens *p;
396482e1 416 SV* const report = newSVpvs("<== ");
bbf60fe6 417
f54cb97a 418 for (p = debug_tokens; p->token; p++) {
bbf60fe6
DM
419 if (p->token == (int)rv) {
420 name = p->name;
421 type = p->type;
422 break;
423 }
424 }
425 if (name)
54667de8 426 Perl_sv_catpv(aTHX_ report, name);
bbf60fe6
DM
427 else if ((char)rv > ' ' && (char)rv < '~')
428 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
429 else if (!rv)
396482e1 430 sv_catpvs(report, "EOF");
bbf60fe6
DM
431 else
432 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
433 switch (type) {
434 case TOKENTYPE_NONE:
bbf60fe6
DM
435 break;
436 case TOKENTYPE_IVAL:
704d4215 437 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
bbf60fe6
DM
438 break;
439 case TOKENTYPE_OPNUM:
440 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
704d4215 441 PL_op_name[lvalp->ival]);
bbf60fe6
DM
442 break;
443 case TOKENTYPE_PVAL:
704d4215 444 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
bbf60fe6
DM
445 break;
446 case TOKENTYPE_OPVAL:
704d4215 447 if (lvalp->opval) {
401441c0 448 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
704d4215
GG
449 PL_op_name[lvalp->opval->op_type]);
450 if (lvalp->opval->op_type == OP_CONST) {
b6007c36 451 Perl_sv_catpvf(aTHX_ report, " %s",
704d4215 452 SvPEEK(cSVOPx_sv(lvalp->opval)));
b6007c36
DM
453 }
454
455 }
401441c0 456 else
396482e1 457 sv_catpvs(report, "(opval=null)");
bbf60fe6
DM
458 break;
459 }
b6007c36 460 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
bbf60fe6
DM
461 };
462 return (int)rv;
998054bd
SC
463}
464
b6007c36
DM
465
466/* print the buffer with suitable escapes */
467
468STATIC void
15f169a1 469S_printbuf(pTHX_ const char *const fmt, const char *const s)
b6007c36 470{
396482e1 471 SV* const tmp = newSVpvs("");
7918f24d
NC
472
473 PERL_ARGS_ASSERT_PRINTBUF;
474
b6007c36
DM
475 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
476 SvREFCNT_dec(tmp);
477}
478
8fa7f367
JH
479#endif
480
8290c323
NC
481static int
482S_deprecate_commaless_var_list(pTHX) {
483 PL_expect = XTERM;
484 deprecate("comma-less variable list");
485 return REPORT(','); /* grandfather non-comma-format format */
486}
487
ffb4593c
NT
488/*
489 * S_ao
490 *
c963b151
BD
491 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
492 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
ffb4593c
NT
493 */
494
76e3520e 495STATIC int
cea2e8a9 496S_ao(pTHX_ int toketype)
a0d0e21e 497{
97aff369 498 dVAR;
3280af22
NIS
499 if (*PL_bufptr == '=') {
500 PL_bufptr++;
a0d0e21e 501 if (toketype == ANDAND)
6154021b 502 pl_yylval.ival = OP_ANDASSIGN;
a0d0e21e 503 else if (toketype == OROR)
6154021b 504 pl_yylval.ival = OP_ORASSIGN;
c963b151 505 else if (toketype == DORDOR)
6154021b 506 pl_yylval.ival = OP_DORASSIGN;
a0d0e21e
LW
507 toketype = ASSIGNOP;
508 }
509 return toketype;
510}
511
ffb4593c
NT
512/*
513 * S_no_op
514 * When Perl expects an operator and finds something else, no_op
515 * prints the warning. It always prints "<something> found where
516 * operator expected. It prints "Missing semicolon on previous line?"
517 * if the surprise occurs at the start of the line. "do you need to
518 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
519 * where the compiler doesn't know if foo is a method call or a function.
520 * It prints "Missing operator before end of line" if there's nothing
521 * after the missing operator, or "... before <...>" if there is something
522 * after the missing operator.
523 */
524
76e3520e 525STATIC void
15f169a1 526S_no_op(pTHX_ const char *const what, char *s)
463ee0b2 527{
97aff369 528 dVAR;
9d4ba2ae
AL
529 char * const oldbp = PL_bufptr;
530 const bool is_first = (PL_oldbufptr == PL_linestart);
68dc0745 531
7918f24d
NC
532 PERL_ARGS_ASSERT_NO_OP;
533
1189a94a
GS
534 if (!s)
535 s = oldbp;
07c798fb 536 else
1189a94a 537 PL_bufptr = s;
734ab321 538 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
56da5a46
RGS
539 if (ckWARN_d(WARN_SYNTAX)) {
540 if (is_first)
541 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
542 "\t(Missing semicolon on previous line?)\n");
543 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
f54cb97a 544 const char *t;
734ab321
BF
545 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':');
546 t += UTF ? UTF8SKIP(t) : 1)
c35e046a 547 NOOP;
56da5a46
RGS
548 if (t < PL_bufptr && isSPACE(*t))
549 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
734ab321
BF
550 "\t(Do you need to predeclare %"SVf"?)\n",
551 SVfARG(newSVpvn_flags(PL_oldoldbufptr, (STRLEN)(t - PL_oldoldbufptr),
552 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
56da5a46
RGS
553 }
554 else {
555 assert(s >= oldbp);
556 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
734ab321
BF
557 "\t(Missing operator before %"SVf"?)\n",
558 SVfARG(newSVpvn_flags(oldbp, (STRLEN)(s - oldbp),
559 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
56da5a46 560 }
07c798fb 561 }
3280af22 562 PL_bufptr = oldbp;
8990e307
LW
563}
564
ffb4593c
NT
565/*
566 * S_missingterm
567 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 568 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
569 * If we're in a delimited string and the delimiter is a control
570 * character, it's reformatted into a two-char sequence like ^C.
571 * This is fatal.
572 */
573
76e3520e 574STATIC void
cea2e8a9 575S_missingterm(pTHX_ char *s)
8990e307 576{
97aff369 577 dVAR;
8990e307
LW
578 char tmpbuf[3];
579 char q;
580 if (s) {
9d4ba2ae 581 char * const nl = strrchr(s,'\n');
d2719217 582 if (nl)
8990e307
LW
583 *nl = '\0';
584 }
463559e7 585 else if (isCNTRL(PL_multi_close)) {
8990e307 586 *tmpbuf = '^';
585ec06d 587 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
588 tmpbuf[2] = '\0';
589 s = tmpbuf;
590 }
591 else {
eb160463 592 *tmpbuf = (char)PL_multi_close;
8990e307
LW
593 tmpbuf[1] = '\0';
594 s = tmpbuf;
595 }
596 q = strchr(s,'"') ? '\'' : '"';
cea2e8a9 597 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
463ee0b2 598}
79072805 599
dd0ac2b9
FC
600#include "feature.h"
601
0d863452 602/*
0d863452
RH
603 * Check whether the named feature is enabled.
604 */
26ea9e12 605bool
3fff3427 606Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 607{
97aff369 608 dVAR;
4a731d7b 609 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
610
611 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
ca4d40c4
FC
612
613 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
7918f24d 614
26ea9e12
NC
615 if (namelen > MAX_FEATURE_LEN)
616 return FALSE;
3fff3427 617 memcpy(&he_name[8], name, namelen);
7d69d4a6 618
c8ca97b0
NC
619 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
620 REFCOUNTED_HE_EXISTS));
0d863452
RH
621}
622
ffb4593c 623/*
9cbb5ea2
GS
624 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
625 * utf16-to-utf8-reversed.
ffb4593c
NT
626 */
627
c39cd008
GS
628#ifdef PERL_CR_FILTER
629static void
630strip_return(SV *sv)
631{
eb578fdb
KW
632 const char *s = SvPVX_const(sv);
633 const char * const e = s + SvCUR(sv);
7918f24d
NC
634
635 PERL_ARGS_ASSERT_STRIP_RETURN;
636
c39cd008
GS
637 /* outer loop optimized to do nothing if there are no CR-LFs */
638 while (s < e) {
639 if (*s++ == '\r' && *s == '\n') {
640 /* hit a CR-LF, need to copy the rest */
eb578fdb 641 char *d = s - 1;
c39cd008
GS
642 *d++ = *s++;
643 while (s < e) {
644 if (*s == '\r' && s[1] == '\n')
645 s++;
646 *d++ = *s++;
647 }
648 SvCUR(sv) -= s - d;
649 return;
650 }
651 }
652}
a868473f 653
76e3520e 654STATIC I32
c39cd008 655S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 656{
f54cb97a 657 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
658 if (count > 0 && !maxlen)
659 strip_return(sv);
660 return count;
a868473f
NIS
661}
662#endif
663
ffb4593c 664/*
8eaa0acf
Z
665=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
666
667Creates and initialises a new lexer/parser state object, supplying
668a context in which to lex and parse from a new source of Perl code.
669A pointer to the new state object is placed in L</PL_parser>. An entry
670is made on the save stack so that upon unwinding the new state object
671will be destroyed and the former value of L</PL_parser> will be restored.
672Nothing else need be done to clean up the parsing context.
673
674The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
675non-null, provides a string (in SV form) containing code to be parsed.
676A copy of the string is made, so subsequent modification of I<line>
677does not affect parsing. I<rsfp>, if non-null, provides an input stream
678from which code will be read to be parsed. If both are non-null, the
679code in I<line> comes first and must consist of complete lines of input,
680and I<rsfp> supplies the remainder of the source.
681
e368b3bd
FC
682The I<flags> parameter is reserved for future use. Currently it is only
683used by perl internally, so extensions should always pass zero.
8eaa0acf
Z
684
685=cut
686*/
ffb4593c 687
27fcb6ee 688/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
87606032
NC
689 can share filters with the current parser.
690 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
691 caller, hence isn't owned by the parser, so shouldn't be closed on parser
692 destruction. This is used to handle the case of defaulting to reading the
693 script from the standard input because no filename was given on the command
694 line (without getting confused by situation where STDIN has been closed, so
695 the script handle is opened on fd 0) */
27fcb6ee 696
a0d0e21e 697void
8eaa0acf 698Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
79072805 699{
97aff369 700 dVAR;
6ef55633 701 const char *s = NULL;
5486870f 702 yy_parser *parser, *oparser;
60d63348 703 if (flags && flags & ~LEX_START_FLAGS)
8eaa0acf 704 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
acdf0a21
DM
705
706 /* create and initialise a parser */
707
199e78b7 708 Newxz(parser, 1, yy_parser);
5486870f 709 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
710 PL_parser = parser;
711
28ac2b49
Z
712 parser->stack = NULL;
713 parser->ps = NULL;
714 parser->stack_size = 0;
acdf0a21 715
e3abe207
DM
716 /* on scope exit, free this parser and restore any outer one */
717 SAVEPARSER(parser);
7c4baf47 718 parser->saved_curcop = PL_curcop;
e3abe207 719
acdf0a21 720 /* initialise lexer state */
8990e307 721
fb205e7a
DM
722#ifdef PERL_MAD
723 parser->curforce = -1;
724#else
725 parser->nexttoke = 0;
726#endif
ca4cfd28 727 parser->error_count = oparser ? oparser->error_count : 0;
c2598295 728 parser->copline = NOLINE;
5afb0a62 729 parser->lex_state = LEX_NORMAL;
c2598295 730 parser->expect = XSTATE;
2f9285f8 731 parser->rsfp = rsfp;
27fcb6ee
FC
732 parser->rsfp_filters =
733 !(flags & LEX_START_SAME_FILTER) || !oparser
d3cd8e11
FC
734 ? NULL
735 : MUTABLE_AV(SvREFCNT_inc(
736 oparser->rsfp_filters
737 ? oparser->rsfp_filters
738 : (oparser->rsfp_filters = newAV())
739 ));
2f9285f8 740
199e78b7
DM
741 Newx(parser->lex_brackstack, 120, char);
742 Newx(parser->lex_casestack, 12, char);
743 *parser->lex_casestack = '\0';
02b34bbe 744
10efb74f 745 if (line) {
0528fd32 746 STRLEN len;
10efb74f 747 s = SvPV_const(line, len);
0abcdfa4
FC
748 parser->linestr = flags & LEX_START_COPIED
749 ? SvREFCNT_inc_simple_NN(line)
750 : newSVpvn_flags(s, len, SvUTF8(line));
37555a72 751 if (!len || s[len-1] != ';')
bdc0bf6f 752 sv_catpvs(parser->linestr, "\n;");
0abcdfa4
FC
753 } else {
754 parser->linestr = newSVpvs("\n;");
8990e307 755 }
f06b5848
DM
756 parser->oldoldbufptr =
757 parser->oldbufptr =
758 parser->bufptr =
759 parser->linestart = SvPVX(parser->linestr);
760 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
761 parser->last_lop = parser->last_uni = NULL;
87606032
NC
762 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
763 |LEX_DONT_CLOSE_RSFP);
737c24fc 764
60d63348 765 parser->in_pod = parser->filtered = 0;
79072805 766}
a687059c 767
e3abe207
DM
768
769/* delete a parser object */
770
771void
772Perl_parser_free(pTHX_ const yy_parser *parser)
773{
7918f24d
NC
774 PERL_ARGS_ASSERT_PARSER_FREE;
775
7c4baf47 776 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
777 SvREFCNT_dec(parser->linestr);
778
87606032 779 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
2f9285f8 780 PerlIO_clearerr(parser->rsfp);
799361c3
SH
781 else if (parser->rsfp && (!parser->old_parser ||
782 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 783 PerlIO_close(parser->rsfp);
5486870f 784 SvREFCNT_dec(parser->rsfp_filters);
2f9285f8 785
e3abe207
DM
786 Safefree(parser->lex_brackstack);
787 Safefree(parser->lex_casestack);
788 PL_parser = parser->old_parser;
789 Safefree(parser);
790}
791
792
ffb4593c 793/*
f0e67a1d
Z
794=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
795
796Buffer scalar containing the chunk currently under consideration of the
797text currently being lexed. This is always a plain string scalar (for
798which C<SvPOK> is true). It is not intended to be used as a scalar by
799normal scalar means; instead refer to the buffer directly by the pointer
800variables described below.
801
802The lexer maintains various C<char*> pointers to things in the
803C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
804reallocated, all of these pointers must be updated. Don't attempt to
805do this manually, but rather use L</lex_grow_linestr> if you need to
806reallocate the buffer.
807
808The content of the text chunk in the buffer is commonly exactly one
809complete line of input, up to and including a newline terminator,
810but there are situations where it is otherwise. The octets of the
811buffer may be intended to be interpreted as either UTF-8 or Latin-1.
812The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
813flag on this scalar, which may disagree with it.
814
815For direct examination of the buffer, the variable
816L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
817lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
818of these pointers is usually preferable to examination of the scalar
819through normal scalar means.
820
821=for apidoc AmxU|char *|PL_parser-E<gt>bufend
822
823Direct pointer to the end of the chunk of text currently being lexed, the
824end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
825+ SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
826always located at the end of the buffer, and does not count as part of
827the buffer's contents.
828
829=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
830
831Points to the current position of lexing inside the lexer buffer.
832Characters around this point may be freely examined, within
833the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
834L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
835interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
836
837Lexing code (whether in the Perl core or not) moves this pointer past
838the characters that it consumes. It is also expected to perform some
839bookkeeping whenever a newline character is consumed. This movement
840can be more conveniently performed by the function L</lex_read_to>,
841which handles newlines appropriately.
842
843Interpretation of the buffer's octets can be abstracted out by
844using the slightly higher-level functions L</lex_peek_unichar> and
845L</lex_read_unichar>.
846
847=for apidoc AmxU|char *|PL_parser-E<gt>linestart
848
849Points to the start of the current line inside the lexer buffer.
850This is useful for indicating at which column an error occurred, and
851not much else. This must be updated by any lexing code that consumes
852a newline; the function L</lex_read_to> handles this detail.
853
854=cut
855*/
856
857/*
858=for apidoc Amx|bool|lex_bufutf8
859
860Indicates whether the octets in the lexer buffer
861(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
862of Unicode characters. If not, they should be interpreted as Latin-1
863characters. This is analogous to the C<SvUTF8> flag for scalars.
864
865In UTF-8 mode, it is not guaranteed that the lexer buffer actually
866contains valid UTF-8. Lexing code must be robust in the face of invalid
867encoding.
868
869The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
870is significant, but not the whole story regarding the input character
871encoding. Normally, when a file is being read, the scalar contains octets
872and its C<SvUTF8> flag is off, but the octets should be interpreted as
873UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
874however, the scalar may have the C<SvUTF8> flag on, and in this case its
875octets should be interpreted as UTF-8 unless the C<use bytes> pragma
876is in effect. This logic may change in the future; use this function
877instead of implementing the logic yourself.
878
879=cut
880*/
881
882bool
883Perl_lex_bufutf8(pTHX)
884{
885 return UTF;
886}
887
888/*
889=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
890
891Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
892at least I<len> octets (including terminating NUL). Returns a
893pointer to the reallocated buffer. This is necessary before making
894any direct modification of the buffer that would increase its length.
895L</lex_stuff_pvn> provides a more convenient way to insert text into
896the buffer.
897
898Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
899this function updates all of the lexer's variables that point directly
900into the buffer.
901
902=cut
903*/
904
905char *
906Perl_lex_grow_linestr(pTHX_ STRLEN len)
907{
908 SV *linestr;
909 char *buf;
910 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
c7641931 911 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
f0e67a1d
Z
912 linestr = PL_parser->linestr;
913 buf = SvPVX(linestr);
914 if (len <= SvLEN(linestr))
915 return buf;
916 bufend_pos = PL_parser->bufend - buf;
917 bufptr_pos = PL_parser->bufptr - buf;
918 oldbufptr_pos = PL_parser->oldbufptr - buf;
919 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
920 linestart_pos = PL_parser->linestart - buf;
921 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
922 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
c7641931
DM
923 re_eval_start_pos = PL_sublex_info.re_eval_start ?
924 PL_sublex_info.re_eval_start - buf : 0;
925
f0e67a1d 926 buf = sv_grow(linestr, len);
c7641931 927
f0e67a1d
Z
928 PL_parser->bufend = buf + bufend_pos;
929 PL_parser->bufptr = buf + bufptr_pos;
930 PL_parser->oldbufptr = buf + oldbufptr_pos;
931 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
932 PL_parser->linestart = buf + linestart_pos;
933 if (PL_parser->last_uni)
934 PL_parser->last_uni = buf + last_uni_pos;
935 if (PL_parser->last_lop)
936 PL_parser->last_lop = buf + last_lop_pos;
c7641931
DM
937 if (PL_sublex_info.re_eval_start)
938 PL_sublex_info.re_eval_start = buf + re_eval_start_pos;
f0e67a1d
Z
939 return buf;
940}
941
942/*
83aa740e 943=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
944
945Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
946immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
947reallocating the buffer if necessary. This means that lexing code that
948runs later will see the characters as if they had appeared in the input.
949It is not recommended to do this as part of normal parsing, and most
950uses of this facility run the risk of the inserted characters being
951interpreted in an unintended manner.
952
953The string to be inserted is represented by I<len> octets starting
954at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
955according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
956The characters are recoded for the lexer buffer, according to how the
957buffer is currently being interpreted (L</lex_bufutf8>). If a string
9dcc53ea 958to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
f0e67a1d
Z
959function is more convenient.
960
961=cut
962*/
963
964void
83aa740e 965Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 966{
749123ff 967 dVAR;
f0e67a1d
Z
968 char *bufptr;
969 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
970 if (flags & ~(LEX_STUFF_UTF8))
971 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
972 if (UTF) {
973 if (flags & LEX_STUFF_UTF8) {
974 goto plain_copy;
975 } else {
976 STRLEN highhalf = 0;
83aa740e 977 const char *p, *e = pv+len;
f0e67a1d
Z
978 for (p = pv; p != e; p++)
979 highhalf += !!(((U8)*p) & 0x80);
980 if (!highhalf)
981 goto plain_copy;
982 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
983 bufptr = PL_parser->bufptr;
984 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
985 SvCUR_set(PL_parser->linestr,
986 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
987 PL_parser->bufend += len+highhalf;
988 for (p = pv; p != e; p++) {
989 U8 c = (U8)*p;
990 if (c & 0x80) {
991 *bufptr++ = (char)(0xc0 | (c >> 6));
992 *bufptr++ = (char)(0x80 | (c & 0x3f));
993 } else {
994 *bufptr++ = (char)c;
995 }
996 }
997 }
998 } else {
999 if (flags & LEX_STUFF_UTF8) {
1000 STRLEN highhalf = 0;
83aa740e 1001 const char *p, *e = pv+len;
f0e67a1d
Z
1002 for (p = pv; p != e; p++) {
1003 U8 c = (U8)*p;
1004 if (c >= 0xc4) {
1005 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1006 "non-Latin-1 character into Latin-1 input");
1007 } else if (c >= 0xc2 && p+1 != e &&
1008 (((U8)p[1]) & 0xc0) == 0x80) {
1009 p++;
1010 highhalf++;
1011 } else if (c >= 0x80) {
1012 /* malformed UTF-8 */
1013 ENTER;
1014 SAVESPTR(PL_warnhook);
1015 PL_warnhook = PERL_WARNHOOK_FATAL;
1016 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1017 LEAVE;
1018 }
1019 }
1020 if (!highhalf)
1021 goto plain_copy;
1022 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1023 bufptr = PL_parser->bufptr;
1024 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1025 SvCUR_set(PL_parser->linestr,
1026 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d
Z
1027 PL_parser->bufend += len-highhalf;
1028 for (p = pv; p != e; p++) {
1029 U8 c = (U8)*p;
1030 if (c & 0x80) {
1031 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1032 p++;
1033 } else {
1034 *bufptr++ = (char)c;
1035 }
1036 }
1037 } else {
1038 plain_copy:
1039 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1040 bufptr = PL_parser->bufptr;
1041 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1042 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1043 PL_parser->bufend += len;
1044 Copy(pv, bufptr, len, char);
1045 }
1046 }
1047}
1048
1049/*
9dcc53ea
Z
1050=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1051
1052Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1053immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1054reallocating the buffer if necessary. This means that lexing code that
1055runs later will see the characters as if they had appeared in the input.
1056It is not recommended to do this as part of normal parsing, and most
1057uses of this facility run the risk of the inserted characters being
1058interpreted in an unintended manner.
1059
1060The string to be inserted is represented by octets starting at I<pv>
1061and continuing to the first nul. These octets are interpreted as either
1062UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1063in I<flags>. The characters are recoded for the lexer buffer, according
1064to how the buffer is currently being interpreted (L</lex_bufutf8>).
1065If it is not convenient to nul-terminate a string to be inserted, the
1066L</lex_stuff_pvn> function is more appropriate.
1067
1068=cut
1069*/
1070
1071void
1072Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1073{
1074 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1075 lex_stuff_pvn(pv, strlen(pv), flags);
1076}
1077
1078/*
f0e67a1d
Z
1079=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1080
1081Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1082immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1083reallocating the buffer if necessary. This means that lexing code that
1084runs later will see the characters as if they had appeared in the input.
1085It is not recommended to do this as part of normal parsing, and most
1086uses of this facility run the risk of the inserted characters being
1087interpreted in an unintended manner.
1088
1089The string to be inserted is the string value of I<sv>. The characters
1090are recoded for the lexer buffer, according to how the buffer is currently
9dcc53ea 1091being interpreted (L</lex_bufutf8>). If a string to be inserted is
f0e67a1d
Z
1092not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1093need to construct a scalar.
1094
1095=cut
1096*/
1097
1098void
1099Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1100{
1101 char *pv;
1102 STRLEN len;
1103 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1104 if (flags)
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1106 pv = SvPV(sv, len);
1107 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1108}
1109
1110/*
1111=for apidoc Amx|void|lex_unstuff|char *ptr
1112
1113Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1114I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1115This hides the discarded text from any lexing code that runs later,
1116as if the text had never appeared.
1117
1118This is not the normal way to consume lexed text. For that, use
1119L</lex_read_to>.
1120
1121=cut
1122*/
1123
1124void
1125Perl_lex_unstuff(pTHX_ char *ptr)
1126{
1127 char *buf, *bufend;
1128 STRLEN unstuff_len;
1129 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1130 buf = PL_parser->bufptr;
1131 if (ptr < buf)
1132 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1133 if (ptr == buf)
1134 return;
1135 bufend = PL_parser->bufend;
1136 if (ptr > bufend)
1137 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1138 unstuff_len = ptr - buf;
1139 Move(ptr, buf, bufend+1-ptr, char);
1140 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1141 PL_parser->bufend = bufend - unstuff_len;
1142}
1143
1144/*
1145=for apidoc Amx|void|lex_read_to|char *ptr
1146
1147Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1148to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1149performing the correct bookkeeping whenever a newline character is passed.
1150This is the normal way to consume lexed text.
1151
1152Interpretation of the buffer's octets can be abstracted out by
1153using the slightly higher-level functions L</lex_peek_unichar> and
1154L</lex_read_unichar>.
1155
1156=cut
1157*/
1158
1159void
1160Perl_lex_read_to(pTHX_ char *ptr)
1161{
1162 char *s;
1163 PERL_ARGS_ASSERT_LEX_READ_TO;
1164 s = PL_parser->bufptr;
1165 if (ptr < s || ptr > PL_parser->bufend)
1166 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1167 for (; s != ptr; s++)
1168 if (*s == '\n') {
1169 CopLINE_inc(PL_curcop);
1170 PL_parser->linestart = s+1;
1171 }
1172 PL_parser->bufptr = ptr;
1173}
1174
1175/*
1176=for apidoc Amx|void|lex_discard_to|char *ptr
1177
1178Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1179up to I<ptr>. The remaining content of the buffer will be moved, and
1180all pointers into the buffer updated appropriately. I<ptr> must not
1181be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1182it is not permitted to discard text that has yet to be lexed.
1183
1184Normally it is not necessarily to do this directly, because it suffices to
1185use the implicit discarding behaviour of L</lex_next_chunk> and things
1186based on it. However, if a token stretches across multiple lines,
1f317c95 1187and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1188that purpose, then after completion of the token it would be wise to
1189explicitly discard the now-unneeded earlier lines, to avoid future
1190multi-line tokens growing the buffer without bound.
1191
1192=cut
1193*/
1194
1195void
1196Perl_lex_discard_to(pTHX_ char *ptr)
1197{
1198 char *buf;
1199 STRLEN discard_len;
1200 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1201 buf = SvPVX(PL_parser->linestr);
1202 if (ptr < buf)
1203 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1204 if (ptr == buf)
1205 return;
1206 if (ptr > PL_parser->bufptr)
1207 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1208 discard_len = ptr - buf;
1209 if (PL_parser->oldbufptr < ptr)
1210 PL_parser->oldbufptr = ptr;
1211 if (PL_parser->oldoldbufptr < ptr)
1212 PL_parser->oldoldbufptr = ptr;
1213 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1214 PL_parser->last_uni = NULL;
1215 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1216 PL_parser->last_lop = NULL;
1217 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1218 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1219 PL_parser->bufend -= discard_len;
1220 PL_parser->bufptr -= discard_len;
1221 PL_parser->oldbufptr -= discard_len;
1222 PL_parser->oldoldbufptr -= discard_len;
1223 if (PL_parser->last_uni)
1224 PL_parser->last_uni -= discard_len;
1225 if (PL_parser->last_lop)
1226 PL_parser->last_lop -= discard_len;
1227}
1228
1229/*
1230=for apidoc Amx|bool|lex_next_chunk|U32 flags
1231
1232Reads in the next chunk of text to be lexed, appending it to
1233L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1234looked to the end of the current chunk and wants to know more. It is
1235usual, but not necessary, for lexing to have consumed the entirety of
1236the current chunk at this time.
1237
1238If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1239chunk (i.e., the current chunk has been entirely consumed), normally the
1240current chunk will be discarded at the same time that the new chunk is
1241read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1242will not be discarded. If the current chunk has not been entirely
1243consumed, then it will not be discarded regardless of the flag.
1244
1245Returns true if some new text was added to the buffer, or false if the
1246buffer has reached the end of the input text.
1247
1248=cut
1249*/
1250
1251#define LEX_FAKE_EOF 0x80000000
1252
1253bool
1254Perl_lex_next_chunk(pTHX_ U32 flags)
1255{
1256 SV *linestr;
1257 char *buf;
1258 STRLEN old_bufend_pos, new_bufend_pos;
1259 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1260 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1261 bool got_some_for_debugger = 0;
f0e67a1d
Z
1262 bool got_some;
1263 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1264 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
f0e67a1d
Z
1265 linestr = PL_parser->linestr;
1266 buf = SvPVX(linestr);
1267 if (!(flags & LEX_KEEP_PREVIOUS) &&
1268 PL_parser->bufptr == PL_parser->bufend) {
1269 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1270 linestart_pos = 0;
1271 if (PL_parser->last_uni != PL_parser->bufend)
1272 PL_parser->last_uni = NULL;
1273 if (PL_parser->last_lop != PL_parser->bufend)
1274 PL_parser->last_lop = NULL;
1275 last_uni_pos = last_lop_pos = 0;
1276 *buf = 0;
1277 SvCUR(linestr) = 0;
1278 } else {
1279 old_bufend_pos = PL_parser->bufend - buf;
1280 bufptr_pos = PL_parser->bufptr - buf;
1281 oldbufptr_pos = PL_parser->oldbufptr - buf;
1282 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1283 linestart_pos = PL_parser->linestart - buf;
1284 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1285 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1286 }
1287 if (flags & LEX_FAKE_EOF) {
1288 goto eof;
60d63348 1289 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
f0e67a1d
Z
1290 got_some = 0;
1291 } else if (filter_gets(linestr, old_bufend_pos)) {
1292 got_some = 1;
17cc9359 1293 got_some_for_debugger = 1;
f0e67a1d 1294 } else {
580561a3
Z
1295 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1296 sv_setpvs(linestr, "");
f0e67a1d
Z
1297 eof:
1298 /* End of real input. Close filehandle (unless it was STDIN),
1299 * then add implicit termination.
1300 */
87606032 1301 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
f0e67a1d
Z
1302 PerlIO_clearerr(PL_parser->rsfp);
1303 else if (PL_parser->rsfp)
1304 (void)PerlIO_close(PL_parser->rsfp);
1305 PL_parser->rsfp = NULL;
60d63348 1306 PL_parser->in_pod = PL_parser->filtered = 0;
f0e67a1d
Z
1307#ifdef PERL_MAD
1308 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1309 PL_faketokens = 1;
1310#endif
1311 if (!PL_in_eval && PL_minus_p) {
1312 sv_catpvs(linestr,
1313 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1314 PL_minus_n = PL_minus_p = 0;
1315 } else if (!PL_in_eval && PL_minus_n) {
1316 sv_catpvs(linestr, /*{*/";}");
1317 PL_minus_n = 0;
1318 } else
1319 sv_catpvs(linestr, ";");
1320 got_some = 1;
1321 }
1322 buf = SvPVX(linestr);
1323 new_bufend_pos = SvCUR(linestr);
1324 PL_parser->bufend = buf + new_bufend_pos;
1325 PL_parser->bufptr = buf + bufptr_pos;
1326 PL_parser->oldbufptr = buf + oldbufptr_pos;
1327 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1328 PL_parser->linestart = buf + linestart_pos;
1329 if (PL_parser->last_uni)
1330 PL_parser->last_uni = buf + last_uni_pos;
1331 if (PL_parser->last_lop)
1332 PL_parser->last_lop = buf + last_lop_pos;
17cc9359 1333 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
f0e67a1d
Z
1334 PL_curstash != PL_debstash) {
1335 /* debugger active and we're not compiling the debugger code,
1336 * so store the line into the debugger's array of lines
1337 */
1338 update_debugger_info(NULL, buf+old_bufend_pos,
1339 new_bufend_pos-old_bufend_pos);
1340 }
1341 return got_some;
1342}
1343
1344/*
1345=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1346
1347Looks ahead one (Unicode) character in the text currently being lexed.
1348Returns the codepoint (unsigned integer value) of the next character,
1349or -1 if lexing has reached the end of the input text. To consume the
1350peeked character, use L</lex_read_unichar>.
1351
1352If the next character is in (or extends into) the next chunk of input
1353text, the next chunk will be read in. Normally the current chunk will be
1354discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1355then the current chunk will not be discarded.
1356
1357If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1358is encountered, an exception is generated.
1359
1360=cut
1361*/
1362
1363I32
1364Perl_lex_peek_unichar(pTHX_ U32 flags)
1365{
749123ff 1366 dVAR;
f0e67a1d
Z
1367 char *s, *bufend;
1368 if (flags & ~(LEX_KEEP_PREVIOUS))
1369 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1370 s = PL_parser->bufptr;
1371 bufend = PL_parser->bufend;
1372 if (UTF) {
1373 U8 head;
1374 I32 unichar;
1375 STRLEN len, retlen;
1376 if (s == bufend) {
1377 if (!lex_next_chunk(flags))
1378 return -1;
1379 s = PL_parser->bufptr;
1380 bufend = PL_parser->bufend;
1381 }
1382 head = (U8)*s;
1383 if (!(head & 0x80))
1384 return head;
1385 if (head & 0x40) {
1386 len = PL_utf8skip[head];
1387 while ((STRLEN)(bufend-s) < len) {
1388 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1389 break;
1390 s = PL_parser->bufptr;
1391 bufend = PL_parser->bufend;
1392 }
1393 }
1394 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1395 if (retlen == (STRLEN)-1) {
1396 /* malformed UTF-8 */
1397 ENTER;
1398 SAVESPTR(PL_warnhook);
1399 PL_warnhook = PERL_WARNHOOK_FATAL;
1400 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1401 LEAVE;
1402 }
1403 return unichar;
1404 } else {
1405 if (s == bufend) {
1406 if (!lex_next_chunk(flags))
1407 return -1;
1408 s = PL_parser->bufptr;
1409 }
1410 return (U8)*s;
1411 }
1412}
1413
1414/*
1415=for apidoc Amx|I32|lex_read_unichar|U32 flags
1416
1417Reads the next (Unicode) character in the text currently being lexed.
1418Returns the codepoint (unsigned integer value) of the character read,
1419and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1420if lexing has reached the end of the input text. To non-destructively
1421examine the next character, use L</lex_peek_unichar> instead.
1422
1423If the next character is in (or extends into) the next chunk of input
1424text, the next chunk will be read in. Normally the current chunk will be
1425discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1426then the current chunk will not be discarded.
1427
1428If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1429is encountered, an exception is generated.
1430
1431=cut
1432*/
1433
1434I32
1435Perl_lex_read_unichar(pTHX_ U32 flags)
1436{
1437 I32 c;
1438 if (flags & ~(LEX_KEEP_PREVIOUS))
1439 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1440 c = lex_peek_unichar(flags);
1441 if (c != -1) {
1442 if (c == '\n')
1443 CopLINE_inc(PL_curcop);
d9018cbe
EB
1444 if (UTF)
1445 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1446 else
1447 ++(PL_parser->bufptr);
f0e67a1d
Z
1448 }
1449 return c;
1450}
1451
1452/*
1453=for apidoc Amx|void|lex_read_space|U32 flags
1454
1455Reads optional spaces, in Perl style, in the text currently being
1456lexed. The spaces may include ordinary whitespace characters and
1457Perl-style comments. C<#line> directives are processed if encountered.
1458L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1459at a non-space character (or the end of the input text).
1460
1461If spaces extend into the next chunk of input text, the next chunk will
1462be read in. Normally the current chunk will be discarded at the same
1463time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1464chunk will not be discarded.
1465
1466=cut
1467*/
1468
f0998909
Z
1469#define LEX_NO_NEXT_CHUNK 0x80000000
1470
f0e67a1d
Z
1471void
1472Perl_lex_read_space(pTHX_ U32 flags)
1473{
1474 char *s, *bufend;
1475 bool need_incline = 0;
f0998909 1476 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
f0e67a1d
Z
1477 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1478#ifdef PERL_MAD
1479 if (PL_skipwhite) {
1480 sv_free(PL_skipwhite);
1481 PL_skipwhite = NULL;
1482 }
1483 if (PL_madskills)
1484 PL_skipwhite = newSVpvs("");
1485#endif /* PERL_MAD */
1486 s = PL_parser->bufptr;
1487 bufend = PL_parser->bufend;
1488 while (1) {
1489 char c = *s;
1490 if (c == '#') {
1491 do {
1492 c = *++s;
1493 } while (!(c == '\n' || (c == 0 && s == bufend)));
1494 } else if (c == '\n') {
1495 s++;
1496 PL_parser->linestart = s;
1497 if (s == bufend)
1498 need_incline = 1;
1499 else
1500 incline(s);
1501 } else if (isSPACE(c)) {
1502 s++;
1503 } else if (c == 0 && s == bufend) {
1504 bool got_more;
1505#ifdef PERL_MAD
1506 if (PL_madskills)
1507 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1508#endif /* PERL_MAD */
f0998909
Z
1509 if (flags & LEX_NO_NEXT_CHUNK)
1510 break;
f0e67a1d
Z
1511 PL_parser->bufptr = s;
1512 CopLINE_inc(PL_curcop);
1513 got_more = lex_next_chunk(flags);
1514 CopLINE_dec(PL_curcop);
1515 s = PL_parser->bufptr;
1516 bufend = PL_parser->bufend;
1517 if (!got_more)
1518 break;
1519 if (need_incline && PL_parser->rsfp) {
1520 incline(s);
1521 need_incline = 0;
1522 }
1523 } else {
1524 break;
1525 }
1526 }
1527#ifdef PERL_MAD
1528 if (PL_madskills)
1529 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1530#endif /* PERL_MAD */
1531 PL_parser->bufptr = s;
1532}
1533
1534/*
ffb4593c
NT
1535 * S_incline
1536 * This subroutine has nothing to do with tilting, whether at windmills
1537 * or pinball tables. Its name is short for "increment line". It
57843af0 1538 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1539 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1540 * # line 500 "foo.pm"
1541 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1542 */
1543
76e3520e 1544STATIC void
d9095cec 1545S_incline(pTHX_ const char *s)
463ee0b2 1546{
97aff369 1547 dVAR;
d9095cec
NC
1548 const char *t;
1549 const char *n;
1550 const char *e;
8818d409 1551 line_t line_num;
463ee0b2 1552
7918f24d
NC
1553 PERL_ARGS_ASSERT_INCLINE;
1554
57843af0 1555 CopLINE_inc(PL_curcop);
463ee0b2
LW
1556 if (*s++ != '#')
1557 return;
d4c19fe8
AL
1558 while (SPACE_OR_TAB(*s))
1559 s++;
73659bf1
GS
1560 if (strnEQ(s, "line", 4))
1561 s += 4;
1562 else
1563 return;
084592ab 1564 if (SPACE_OR_TAB(*s))
73659bf1 1565 s++;
4e553d73 1566 else
73659bf1 1567 return;
d4c19fe8
AL
1568 while (SPACE_OR_TAB(*s))
1569 s++;
463ee0b2
LW
1570 if (!isDIGIT(*s))
1571 return;
d4c19fe8 1572
463ee0b2
LW
1573 n = s;
1574 while (isDIGIT(*s))
1575 s++;
07714eb4 1576 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1577 return;
bf4acbe4 1578 while (SPACE_OR_TAB(*s))
463ee0b2 1579 s++;
73659bf1 1580 if (*s == '"' && (t = strchr(s+1, '"'))) {
463ee0b2 1581 s++;
73659bf1
GS
1582 e = t + 1;
1583 }
463ee0b2 1584 else {
c35e046a
AL
1585 t = s;
1586 while (!isSPACE(*t))
1587 t++;
73659bf1 1588 e = t;
463ee0b2 1589 }
bf4acbe4 1590 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1591 e++;
1592 if (*e != '\n' && *e != '\0')
1593 return; /* false alarm */
1594
8818d409
FC
1595 line_num = atoi(n)-1;
1596
f4dd75d9 1597 if (t - s > 0) {
d9095cec 1598 const STRLEN len = t - s;
19bad673
NC
1599 SV *const temp_sv = CopFILESV(PL_curcop);
1600 const char *cf;
1601 STRLEN tmplen;
1602
1603 if (temp_sv) {
1604 cf = SvPVX(temp_sv);
1605 tmplen = SvCUR(temp_sv);
1606 } else {
1607 cf = NULL;
1608 tmplen = 0;
1609 }
1610
d1299d44 1611 if (!PL_rsfp && !PL_parser->filtered) {
e66cf94c
RGS
1612 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1613 * to *{"::_<newfilename"} */
44867030
NC
1614 /* However, the long form of evals is only turned on by the
1615 debugger - usually they're "(eval %lu)" */
1616 char smallbuf[128];
1617 char *tmpbuf;
1618 GV **gvp;
d9095cec 1619 STRLEN tmplen2 = len;
798b63bc 1620 if (tmplen + 2 <= sizeof smallbuf)
e66cf94c
RGS
1621 tmpbuf = smallbuf;
1622 else
2ae0db35 1623 Newx(tmpbuf, tmplen + 2, char);
44867030
NC
1624 tmpbuf[0] = '_';
1625 tmpbuf[1] = '<';
2ae0db35 1626 memcpy(tmpbuf + 2, cf, tmplen);
44867030 1627 tmplen += 2;
8a5ee598
RGS
1628 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1629 if (gvp) {
44867030
NC
1630 char *tmpbuf2;
1631 GV *gv2;
1632
1633 if (tmplen2 + 2 <= sizeof smallbuf)
1634 tmpbuf2 = smallbuf;
1635 else
1636 Newx(tmpbuf2, tmplen2 + 2, char);
1637
1638 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1639 /* Either they malloc'd it, or we malloc'd it,
1640 so no prefix is present in ours. */
1641 tmpbuf2[0] = '_';
1642 tmpbuf2[1] = '<';
1643 }
1644
1645 memcpy(tmpbuf2 + 2, s, tmplen2);
1646 tmplen2 += 2;
1647
8a5ee598 1648 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1649 if (!isGV(gv2)) {
8a5ee598 1650 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1651 /* adjust ${"::_<newfilename"} to store the new file name */
1652 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
8818d409
FC
1653 /* The line number may differ. If that is the case,
1654 alias the saved lines that are in the array.
1655 Otherwise alias the whole array. */
1656 if (CopLINE(PL_curcop) == line_num) {
1657 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1658 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1659 }
1660 else if (GvAV(*gvp)) {
1661 AV * const av = GvAV(*gvp);
1662 const I32 start = CopLINE(PL_curcop)+1;
1663 I32 items = AvFILLp(av) - start;
1664 if (items > 0) {
1665 AV * const av2 = GvAVn(gv2);
1666 SV **svp = AvARRAY(av) + start;
1667 I32 l = (I32)line_num+1;
1668 while (items--)
1669 av_store(av2, l++, SvREFCNT_inc(*svp++));
1670 }
1671 }
e5527e4b 1672 }
44867030
NC
1673
1674 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
8a5ee598 1675 }
e66cf94c 1676 if (tmpbuf != smallbuf) Safefree(tmpbuf);
e66cf94c 1677 }
05ec9bb3 1678 CopFILE_free(PL_curcop);
d9095cec 1679 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1680 }
8818d409 1681 CopLINE_set(PL_curcop, line_num);
463ee0b2
LW
1682}
1683
29595ff2 1684#ifdef PERL_MAD
cd81e915 1685/* skip space before PL_thistoken */
29595ff2
NC
1686
1687STATIC char *
1688S_skipspace0(pTHX_ register char *s)
1689{
7918f24d
NC
1690 PERL_ARGS_ASSERT_SKIPSPACE0;
1691
29595ff2
NC
1692 s = skipspace(s);
1693 if (!PL_madskills)
1694 return s;
cd81e915
NC
1695 if (PL_skipwhite) {
1696 if (!PL_thiswhite)
6b29d1f5 1697 PL_thiswhite = newSVpvs("");
cd81e915
NC
1698 sv_catsv(PL_thiswhite, PL_skipwhite);
1699 sv_free(PL_skipwhite);
1700 PL_skipwhite = 0;
1701 }
1702 PL_realtokenstart = s - SvPVX(PL_linestr);
29595ff2
NC
1703 return s;
1704}
1705
cd81e915 1706/* skip space after PL_thistoken */
29595ff2
NC
1707
1708STATIC char *
1709S_skipspace1(pTHX_ register char *s)
1710{
d4c19fe8 1711 const char *start = s;
29595ff2
NC
1712 I32 startoff = start - SvPVX(PL_linestr);
1713
7918f24d
NC
1714 PERL_ARGS_ASSERT_SKIPSPACE1;
1715
29595ff2
NC
1716 s = skipspace(s);
1717 if (!PL_madskills)
1718 return s;
1719 start = SvPVX(PL_linestr) + startoff;
cd81e915 1720 if (!PL_thistoken && PL_realtokenstart >= 0) {
d4c19fe8 1721 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
cd81e915
NC
1722 PL_thistoken = newSVpvn(tstart, start - tstart);
1723 }
1724 PL_realtokenstart = -1;
1725 if (PL_skipwhite) {
1726 if (!PL_nextwhite)
6b29d1f5 1727 PL_nextwhite = newSVpvs("");
cd81e915
NC
1728 sv_catsv(PL_nextwhite, PL_skipwhite);
1729 sv_free(PL_skipwhite);
1730 PL_skipwhite = 0;
29595ff2
NC
1731 }
1732 return s;
1733}
1734
1735STATIC char *
1736S_skipspace2(pTHX_ register char *s, SV **svp)
1737{
c35e046a
AL
1738 char *start;
1739 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1740 const I32 startoff = s - SvPVX(PL_linestr);
1741
7918f24d
NC
1742 PERL_ARGS_ASSERT_SKIPSPACE2;
1743
29595ff2
NC
1744 s = skipspace(s);
1745 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1746 if (!PL_madskills || !svp)
1747 return s;
1748 start = SvPVX(PL_linestr) + startoff;
cd81e915 1749 if (!PL_thistoken && PL_realtokenstart >= 0) {
d4c19fe8 1750 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
cd81e915
NC
1751 PL_thistoken = newSVpvn(tstart, start - tstart);
1752 PL_realtokenstart = -1;
29595ff2 1753 }
cd81e915 1754 if (PL_skipwhite) {
29595ff2 1755 if (!*svp)
6b29d1f5 1756 *svp = newSVpvs("");
cd81e915
NC
1757 sv_setsv(*svp, PL_skipwhite);
1758 sv_free(PL_skipwhite);
1759 PL_skipwhite = 0;
29595ff2
NC
1760 }
1761
1762 return s;
1763}
1764#endif
1765
80a702cd 1766STATIC void
15f169a1 1767S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1768{
1769 AV *av = CopFILEAVx(PL_curcop);
1770 if (av) {
b9f83d2f 1771 SV * const sv = newSV_type(SVt_PVMG);
5fa550fb
NC
1772 if (orig_sv)
1773 sv_setsv(sv, orig_sv);
1774 else
1775 sv_setpvn(sv, buf, len);
80a702cd
RGS
1776 (void)SvIOK_on(sv);
1777 SvIV_set(sv, 0);
1778 av_store(av, (I32)CopLINE(PL_curcop), sv);
1779 }
1780}
1781
ffb4593c
NT
1782/*
1783 * S_skipspace
1784 * Called to gobble the appropriate amount and type of whitespace.
1785 * Skips comments as well.
1786 */
1787
76e3520e 1788STATIC char *
cea2e8a9 1789S_skipspace(pTHX_ register char *s)
a687059c 1790{
5db06880 1791#ifdef PERL_MAD
f0e67a1d
Z
1792 char *start = s;
1793#endif /* PERL_MAD */
7918f24d 1794 PERL_ARGS_ASSERT_SKIPSPACE;
f0e67a1d 1795#ifdef PERL_MAD
cd81e915
NC
1796 if (PL_skipwhite) {
1797 sv_free(PL_skipwhite);
f0e67a1d 1798 PL_skipwhite = NULL;
5db06880 1799 }
f0e67a1d 1800#endif /* PERL_MAD */
3280af22 1801 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
bf4acbe4 1802 while (s < PL_bufend && SPACE_OR_TAB(*s))
463ee0b2 1803 s++;
f0e67a1d
Z
1804 } else {
1805 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1806 PL_bufptr = s;
f0998909
Z
1807 lex_read_space(LEX_KEEP_PREVIOUS |
1808 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1809 LEX_NO_NEXT_CHUNK : 0));
3280af22 1810 s = PL_bufptr;
f0e67a1d
Z
1811 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1812 if (PL_linestart > PL_bufptr)
1813 PL_bufptr = PL_linestart;
1814 return s;
463ee0b2 1815 }
5db06880 1816#ifdef PERL_MAD
f0e67a1d
Z
1817 if (PL_madskills)
1818 PL_skipwhite = newSVpvn(start, s-start);
1819#endif /* PERL_MAD */
5db06880 1820 return s;
a687059c 1821}
378cc40b 1822
ffb4593c
NT
1823/*
1824 * S_check_uni
1825 * Check the unary operators to ensure there's no ambiguity in how they're
1826 * used. An ambiguous piece of code would be:
1827 * rand + 5
1828 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1829 * the +5 is its argument.
1830 */
1831
76e3520e 1832STATIC void
cea2e8a9 1833S_check_uni(pTHX)
ba106d47 1834{
97aff369 1835 dVAR;
d4c19fe8
AL
1836 const char *s;
1837 const char *t;
2f3197b3 1838
3280af22 1839 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1840 return;
3280af22
NIS
1841 while (isSPACE(*PL_last_uni))
1842 PL_last_uni++;
c35e046a
AL
1843 s = PL_last_uni;
1844 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1845 s++;
3280af22 1846 if ((t = strchr(s, '(')) && t < PL_bufptr)
a0d0e21e 1847 return;
6136c704 1848
9b387841
NC
1849 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1850 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1851 (int)(s - PL_last_uni), PL_last_uni);
2f3197b3
LW
1852}
1853
ffb4593c
NT
1854/*
1855 * LOP : macro to build a list operator. Its behaviour has been replaced
1856 * with a subroutine, S_lop() for which LOP is just another name.
1857 */
1858
a0d0e21e
LW
1859#define LOP(f,x) return lop(f,x,s)
1860
ffb4593c
NT
1861/*
1862 * S_lop
1863 * Build a list operator (or something that might be one). The rules:
1864 * - if we have a next token, then it's a list operator [why?]
1865 * - if the next thing is an opening paren, then it's a function
1866 * - else it's a list operator
1867 */
1868
76e3520e 1869STATIC I32
a0be28da 1870S_lop(pTHX_ I32 f, int x, char *s)
ffed7fef 1871{
97aff369 1872 dVAR;
7918f24d
NC
1873
1874 PERL_ARGS_ASSERT_LOP;
1875
6154021b 1876 pl_yylval.ival = f;
35c8bce7 1877 CLINE;
3280af22
NIS
1878 PL_expect = x;
1879 PL_bufptr = s;
1880 PL_last_lop = PL_oldbufptr;
eb160463 1881 PL_last_lop_op = (OPCODE)f;
5db06880
NC
1882#ifdef PERL_MAD
1883 if (PL_lasttoke)
78cdf107 1884 goto lstop;
5db06880 1885#else
3280af22 1886 if (PL_nexttoke)
78cdf107 1887 goto lstop;
5db06880 1888#endif
79072805 1889 if (*s == '(')
bbf60fe6 1890 return REPORT(FUNC);
29595ff2 1891 s = PEEKSPACE(s);
79072805 1892 if (*s == '(')
bbf60fe6 1893 return REPORT(FUNC);
78cdf107
Z
1894 else {
1895 lstop:
1896 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1897 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
bbf60fe6 1898 return REPORT(LSTOP);
78cdf107 1899 }
79072805
LW
1900}
1901
5db06880
NC
1902#ifdef PERL_MAD
1903 /*
1904 * S_start_force
1905 * Sets up for an eventual force_next(). start_force(0) basically does
1906 * an unshift, while start_force(-1) does a push. yylex removes items
1907 * on the "pop" end.
1908 */
1909
1910STATIC void
1911S_start_force(pTHX_ int where)
1912{
1913 int i;
1914
cd81e915 1915 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
5db06880 1916 where = PL_lasttoke;
cd81e915
NC
1917 assert(PL_curforce < 0 || PL_curforce == where);
1918 if (PL_curforce != where) {
5db06880
NC
1919 for (i = PL_lasttoke; i > where; --i) {
1920 PL_nexttoke[i] = PL_nexttoke[i-1];
1921 }
1922 PL_lasttoke++;
1923 }
cd81e915 1924 if (PL_curforce < 0) /* in case of duplicate start_force() */
5db06880 1925 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
cd81e915
NC
1926 PL_curforce = where;
1927 if (PL_nextwhite) {
5db06880 1928 if (PL_madskills)
6b29d1f5 1929 curmad('^', newSVpvs(""));
cd81e915 1930 CURMAD('_', PL_nextwhite);
5db06880
NC
1931 }
1932}
1933
1934STATIC void
1935S_curmad(pTHX_ char slot, SV *sv)
1936{
1937 MADPROP **where;
1938
1939 if (!sv)
1940 return;
cd81e915
NC
1941 if (PL_curforce < 0)
1942 where = &PL_thismad;
5db06880 1943 else
cd81e915 1944 where = &PL_nexttoke[PL_curforce].next_mad;
5db06880 1945
cd81e915 1946 if (PL_faketokens)
76f68e9b 1947 sv_setpvs(sv, "");
5db06880
NC
1948 else {
1949 if (!IN_BYTES) {
1950 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1951 SvUTF8_on(sv);
1952 else if (PL_encoding) {
1953 sv_recode_to_utf8(sv, PL_encoding);
1954 }
1955 }
1956 }
1957
1958 /* keep a slot open for the head of the list? */
1959 if (slot != '_' && *where && (*where)->mad_key == '^') {
1960 (*where)->mad_key = slot;
daba3364 1961 sv_free(MUTABLE_SV(((*where)->mad_val)));
5db06880
NC
1962 (*where)->mad_val = (void*)sv;
1963 }
1964 else
1965 addmad(newMADsv(slot, sv), where, 0);
1966}
1967#else
b3f24c00
MHM
1968# define start_force(where) NOOP
1969# define curmad(slot, sv) NOOP
5db06880
NC
1970#endif
1971
ffb4593c
NT
1972/*
1973 * S_force_next
9cbb5ea2 1974 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1975 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1976 * to know what token to return the next time the lexer is called. Caller
5db06880
NC
1977 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1978 * and possibly PL_expect to ensure the lexer handles the token correctly.
ffb4593c
NT
1979 */
1980
4e553d73 1981STATIC void
cea2e8a9 1982S_force_next(pTHX_ I32 type)
79072805 1983{
97aff369 1984 dVAR;
704d4215
GG
1985#ifdef DEBUGGING
1986 if (DEBUG_T_TEST) {
1987 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1988 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
1989 }
1990#endif
6c7ae946
FC
1991 /* Don’t let opslab_force_free snatch it */
1992 if (S_is_opval_token(type & 0xffff) && NEXTVAL_NEXTTOKE.opval) {
1993 assert(!NEXTVAL_NEXTTOKE.opval->op_savefree);
1994 NEXTVAL_NEXTTOKE.opval->op_savefree = 1;
1995 }
5db06880 1996#ifdef PERL_MAD
cd81e915 1997 if (PL_curforce < 0)
5db06880 1998 start_force(PL_lasttoke);
cd81e915 1999 PL_nexttoke[PL_curforce].next_type = type;
5db06880
NC
2000 if (PL_lex_state != LEX_KNOWNEXT)
2001 PL_lex_defer = PL_lex_state;
2002 PL_lex_state = LEX_KNOWNEXT;
2003 PL_lex_expect = PL_expect;
cd81e915 2004 PL_curforce = -1;
5db06880 2005#else
3280af22
NIS
2006 PL_nexttype[PL_nexttoke] = type;
2007 PL_nexttoke++;
2008 if (PL_lex_state != LEX_KNOWNEXT) {
2009 PL_lex_defer = PL_lex_state;
2010 PL_lex_expect = PL_expect;
2011 PL_lex_state = LEX_KNOWNEXT;
79072805 2012 }
5db06880 2013#endif
79072805
LW
2014}
2015
28ac2b49
Z
2016void
2017Perl_yyunlex(pTHX)
2018{
a7aaec61
Z
2019 int yyc = PL_parser->yychar;
2020 if (yyc != YYEMPTY) {
2021 if (yyc) {
2022 start_force(-1);
2023 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2024 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
78cdf107 2025 PL_lex_allbrackets--;
a7aaec61 2026 PL_lex_brackets--;
78cdf107
Z
2027 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2028 } else if (yyc == '('/*)*/) {
2029 PL_lex_allbrackets--;
2030 yyc |= (2<<24);
a7aaec61
Z
2031 }
2032 force_next(yyc);
2033 }
28ac2b49
Z
2034 PL_parser->yychar = YYEMPTY;
2035 }
2036}
2037
d0a148a6 2038STATIC SV *
15f169a1 2039S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 2040{
97aff369 2041 dVAR;
740cce10 2042 SV * const sv = newSVpvn_utf8(start, len,
eaf7a4d2
CS
2043 !IN_BYTES
2044 && UTF
2045 && !is_ascii_string((const U8*)start, len)
740cce10 2046 && is_utf8_string((const U8*)start, len));
d0a148a6
NC
2047 return sv;
2048}
2049
ffb4593c
NT
2050/*
2051 * S_force_word
2052 * When the lexer knows the next thing is a word (for instance, it has
2053 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
2054 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2055 * lookahead.
ffb4593c
NT
2056 *
2057 * Arguments:
b1b65b59 2058 * char *start : buffer position (must be within PL_linestr)
02b34bbe 2059 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
ffb4593c
NT
2060 * int check_keyword : if true, Perl checks to make sure the word isn't
2061 * a keyword (do this if the word is a label, e.g. goto FOO)
2062 * int allow_pack : if true, : characters will also be allowed (require,
2063 * use, etc. do this)
9cbb5ea2 2064 * int allow_initial_tick : used by the "sub" lexer only.
ffb4593c
NT
2065 */
2066
76e3520e 2067STATIC char *
cea2e8a9 2068S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
79072805 2069{
97aff369 2070 dVAR;
eb578fdb 2071 char *s;
463ee0b2 2072 STRLEN len;
4e553d73 2073
7918f24d
NC
2074 PERL_ARGS_ASSERT_FORCE_WORD;
2075
29595ff2 2076 start = SKIPSPACE1(start);
463ee0b2 2077 s = start;
7e2040f0 2078 if (isIDFIRST_lazy_if(s,UTF) ||
a0d0e21e 2079 (allow_pack && *s == ':') ||
15f0808c 2080 (allow_initial_tick && *s == '\'') )
a0d0e21e 2081 {
3280af22 2082 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
5458a98a 2083 if (check_keyword && keyword(PL_tokenbuf, len, 0))
463ee0b2 2084 return start;
cd81e915 2085 start_force(PL_curforce);
5db06880
NC
2086 if (PL_madskills)
2087 curmad('X', newSVpvn(start,s-start));
463ee0b2 2088 if (token == METHOD) {
29595ff2 2089 s = SKIPSPACE1(s);
463ee0b2 2090 if (*s == '(')
3280af22 2091 PL_expect = XTERM;
463ee0b2 2092 else {
3280af22 2093 PL_expect = XOPERATOR;
463ee0b2 2094 }
79072805 2095 }
e74e6b3d 2096 if (PL_madskills)
63575281 2097 curmad('g', newSVpvs( "forced" ));
9ded7720 2098 NEXTVAL_NEXTTOKE.opval
d0a148a6
NC
2099 = (OP*)newSVOP(OP_CONST,0,
2100 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 2101 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
2102 force_next(token);
2103 }
2104 return s;
2105}
2106
ffb4593c
NT
2107/*
2108 * S_force_ident
9cbb5ea2 2109 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2110 * text only contains the "foo" portion. The first argument is a pointer
2111 * to the "foo", and the second argument is the type symbol to prefix.
2112 * Forces the next token to be a "WORD".
9cbb5ea2 2113 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2114 */
2115
76e3520e 2116STATIC void
bfed75c6 2117S_force_ident(pTHX_ register const char *s, int kind)
79072805 2118{
97aff369 2119 dVAR;
7918f24d
NC
2120
2121 PERL_ARGS_ASSERT_FORCE_IDENT;
2122
c35e046a 2123 if (*s) {
90e5519e 2124 const STRLEN len = strlen(s);
728847b1
BF
2125 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2126 UTF ? SVf_UTF8 : 0));
cd81e915 2127 start_force(PL_curforce);
9ded7720 2128 NEXTVAL_NEXTTOKE.opval = o;
79072805 2129 force_next(WORD);
748a9306 2130 if (kind) {
11343788 2131 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2132 /* XXX see note in pp_entereval() for why we forgo typo
2133 warnings if the symbol must be introduced in an eval.
2134 GSAR 96-10-12 */
90e5519e 2135 gv_fetchpvn_flags(s, len,
728847b1
BF
2136 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2137 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
90e5519e
NC
2138 kind == '$' ? SVt_PV :
2139 kind == '@' ? SVt_PVAV :
2140 kind == '%' ? SVt_PVHV :
a0d0e21e 2141 SVt_PVGV
90e5519e 2142 );
748a9306 2143 }
79072805
LW
2144 }
2145}
2146
1571675a
GS
2147NV
2148Perl_str_to_version(pTHX_ SV *sv)
2149{
2150 NV retval = 0.0;
2151 NV nshift = 1.0;
2152 STRLEN len;
cfd0369c 2153 const char *start = SvPV_const(sv,len);
9d4ba2ae 2154 const char * const end = start + len;
504618e9 2155 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
7918f24d
NC
2156
2157 PERL_ARGS_ASSERT_STR_TO_VERSION;
2158
1571675a 2159 while (start < end) {
ba210ebe 2160 STRLEN skip;
1571675a
GS
2161 UV n;
2162 if (utf)
9041c2e3 2163 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2164 else {
2165 n = *(U8*)start;
2166 skip = 1;
2167 }
2168 retval += ((NV)n)/nshift;
2169 start += skip;
2170 nshift *= 1000;
2171 }
2172 return retval;
2173}
2174
4e553d73 2175/*
ffb4593c
NT
2176 * S_force_version
2177 * Forces the next token to be a version number.
e759cc13
RGS
2178 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2179 * and if "guessing" is TRUE, then no new token is created (and the caller
2180 * must use an alternative parsing method).
ffb4593c
NT
2181 */
2182
76e3520e 2183STATIC char *
e759cc13 2184S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2185{
97aff369 2186 dVAR;
5f66b61c 2187 OP *version = NULL;
44dcb63b 2188 char *d;
5db06880
NC
2189#ifdef PERL_MAD
2190 I32 startoff = s - SvPVX(PL_linestr);
2191#endif
89bfa8cd 2192
7918f24d
NC
2193 PERL_ARGS_ASSERT_FORCE_VERSION;
2194
29595ff2 2195 s = SKIPSPACE1(s);
89bfa8cd 2196
44dcb63b 2197 d = s;
dd629d5b 2198 if (*d == 'v')
44dcb63b 2199 d++;
44dcb63b 2200 if (isDIGIT(*d)) {
e759cc13
RGS
2201 while (isDIGIT(*d) || *d == '_' || *d == '.')
2202 d++;
5db06880
NC
2203#ifdef PERL_MAD
2204 if (PL_madskills) {
cd81e915 2205 start_force(PL_curforce);
5db06880
NC
2206 curmad('X', newSVpvn(s,d-s));
2207 }
2208#endif
4e4da3ac 2209 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2210 SV *ver;
8d08d9ba 2211#ifdef USE_LOCALE_NUMERIC
909d3787
KW
2212 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2213 setlocale(LC_NUMERIC, "C");
8d08d9ba 2214#endif
6154021b 2215 s = scan_num(s, &pl_yylval);
8d08d9ba
DG
2216#ifdef USE_LOCALE_NUMERIC
2217 setlocale(LC_NUMERIC, loc);
909d3787 2218 Safefree(loc);
8d08d9ba 2219#endif
6154021b 2220 version = pl_yylval.opval;
dd629d5b
GS
2221 ver = cSVOPx(version)->op_sv;
2222 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2223 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2224 SvNV_set(ver, str_to_version(ver));
1571675a 2225 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2226 }
89bfa8cd 2227 }
5db06880
NC
2228 else if (guessing) {
2229#ifdef PERL_MAD
2230 if (PL_madskills) {
cd81e915
NC
2231 sv_free(PL_nextwhite); /* let next token collect whitespace */
2232 PL_nextwhite = 0;
5db06880
NC
2233 s = SvPVX(PL_linestr) + startoff;
2234 }
2235#endif
e759cc13 2236 return s;
5db06880 2237 }
89bfa8cd
PP
2238 }
2239
5db06880
NC
2240#ifdef PERL_MAD
2241 if (PL_madskills && !version) {
cd81e915
NC
2242 sv_free(PL_nextwhite); /* let next token collect whitespace */
2243 PL_nextwhite = 0;
5db06880
NC
2244 s = SvPVX(PL_linestr) + startoff;
2245 }
2246#endif
89bfa8cd 2247 /* NOTE: The parser sees the package name and the VERSION swapped */
cd81e915 2248 start_force(PL_curforce);
9ded7720 2249 NEXTVAL_NEXTTOKE.opval = version;
4e553d73 2250 force_next(WORD);
89bfa8cd 2251
e759cc13 2252 return s;
89bfa8cd
PP
2253}
2254
ffb4593c 2255/*
91152fc1
DG
2256 * S_force_strict_version
2257 * Forces the next token to be a version number using strict syntax rules.
2258 */
2259
2260STATIC char *
2261S_force_strict_version(pTHX_ char *s)
2262{
2263 dVAR;
2264 OP *version = NULL;
2265#ifdef PERL_MAD
2266 I32 startoff = s - SvPVX(PL_linestr);
2267#endif
2268 const char *errstr = NULL;
2269
2270 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2271
2272 while (isSPACE(*s)) /* leading whitespace */
2273 s++;
2274
2275 if (is_STRICT_VERSION(s,&errstr)) {
2276 SV *ver = newSV(0);
2277 s = (char *)scan_version(s, ver, 0);
2278 version = newSVOP(OP_CONST, 0, ver);
2279 }
4e4da3ac
Z
2280 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2281 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2282 {
91152fc1
DG
2283 PL_bufptr = s;
2284 if (errstr)
2285 yyerror(errstr); /* version required */
2286 return s;
2287 }
2288
2289#ifdef PERL_MAD
2290 if (PL_madskills && !version) {
2291 sv_free(PL_nextwhite); /* let next token collect whitespace */
2292 PL_nextwhite = 0;
2293 s = SvPVX(PL_linestr) + startoff;
2294 }
2295#endif
2296 /* NOTE: The parser sees the package name and the VERSION swapped */
2297 start_force(PL_curforce);
2298 NEXTVAL_NEXTTOKE.opval = version;
2299 force_next(WORD);
2300
2301 return s;
2302}
2303
2304/*
ffb4593c
NT
2305 * S_tokeq
2306 * Tokenize a quoted string passed in as an SV. It finds the next
2307 * chunk, up to end of string or a backslash. It may make a new
2308 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2309 * turns \\ into \.
2310 */
2311
76e3520e 2312STATIC SV *
cea2e8a9 2313S_tokeq(pTHX_ SV *sv)
79072805 2314{
97aff369 2315 dVAR;
eb578fdb
KW
2316 char *s;
2317 char *send;
2318 char *d;
b3ac6de7
IZ
2319 STRLEN len = 0;
2320 SV *pv = sv;
79072805 2321
7918f24d
NC
2322 PERL_ARGS_ASSERT_TOKEQ;
2323
79072805 2324 if (!SvLEN(sv))
b3ac6de7 2325 goto finish;
79072805 2326
a0d0e21e 2327 s = SvPV_force(sv, len);
21a311ee 2328 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
b3ac6de7 2329 goto finish;
463ee0b2 2330 send = s + len;
dcb21ed6
NC
2331 /* This is relying on the SV being "well formed" with a trailing '\0' */
2332 while (s < send && !(*s == '\\' && s[1] == '\\'))
79072805
LW
2333 s++;
2334 if (s == send)
b3ac6de7 2335 goto finish;
79072805 2336 d = s;
be4731d2 2337 if ( PL_hints & HINT_NEW_STRING ) {
59cd0e26 2338 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
be4731d2 2339 }
79072805
LW
2340 while (s < send) {
2341 if (*s == '\\') {
a0d0e21e 2342 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2343 s++; /* all that, just for this */
2344 }
2345 *d++ = *s++;
2346 }
2347 *d = '\0';
95a20fc0 2348 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2349 finish:
3280af22 2350 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2351 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2352 return sv;
2353}
2354
ffb4593c
NT
2355/*
2356 * Now come three functions related to double-quote context,
2357 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2358 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2359 * interact with PL_lex_state, and create fake ( ... ) argument lists
2360 * to handle functions and concatenation.
ecd24171
DM
2361 * For example,
2362 * "foo\lbar"
2363 * is tokenised as
2364 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
ffb4593c
NT
2365 */
2366
2367/*
2368 * S_sublex_start
6154021b 2369 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2370 *
2371 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2372 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c
NT
2373 *
2374 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2375 *
2376 * Everything else becomes a FUNC.
2377 *
2378 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2379 * had an OP_CONST or OP_READLINE). This just sets us up for a
2380 * call to S_sublex_push().
2381 */
2382
76e3520e 2383STATIC I32
cea2e8a9 2384S_sublex_start(pTHX)
79072805 2385{
97aff369 2386 dVAR;
eb578fdb 2387 const I32 op_type = pl_yylval.ival;
79072805 2388
5097bf9b
FC
2389 PL_sublex_info.super_bufptr = PL_bufptr;
2390 PL_sublex_info.super_bufend = PL_bufend;
79072805 2391 if (op_type == OP_NULL) {
6154021b 2392 pl_yylval.opval = PL_lex_op;
5f66b61c 2393 PL_lex_op = NULL;
79072805
LW
2394 return THING;
2395 }
2396 if (op_type == OP_CONST || op_type == OP_READLINE) {
3280af22 2397 SV *sv = tokeq(PL_lex_stuff);
b3ac6de7
IZ
2398
2399 if (SvTYPE(sv) == SVt_PVIV) {
2400 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2401 STRLEN len;
96a5add6 2402 const char * const p = SvPV_const(sv, len);
740cce10 2403 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2404 SvREFCNT_dec(sv);
2405 sv = nsv;
4e553d73 2406 }
6154021b 2407 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
a0714e2c 2408 PL_lex_stuff = NULL;
6f33ba73
RGS
2409 /* Allow <FH> // "foo" */
2410 if (op_type == OP_READLINE)
2411 PL_expect = XTERMORDORDOR;
79072805
LW
2412 return THING;
2413 }
e3f73d4e
RGS
2414 else if (op_type == OP_BACKTICK && PL_lex_op) {
2415 /* readpipe() vas overriden */
2416 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
6154021b 2417 pl_yylval.opval = PL_lex_op;
9b201d7d 2418 PL_lex_op = NULL;
e3f73d4e
RGS
2419 PL_lex_stuff = NULL;
2420 return THING;
2421 }
79072805 2422
3280af22 2423 PL_sublex_info.super_state = PL_lex_state;
eac04b2e 2424 PL_sublex_info.sub_inwhat = (U16)op_type;
3280af22
NIS
2425 PL_sublex_info.sub_op = PL_lex_op;
2426 PL_lex_state = LEX_INTERPPUSH;
55497cff 2427
3280af22
NIS
2428 PL_expect = XTERM;
2429 if (PL_lex_op) {
6154021b 2430 pl_yylval.opval = PL_lex_op;
5f66b61c 2431 PL_lex_op = NULL;
55497cff
PP
2432 return PMFUNC;
2433 }
2434 else
2435 return FUNC;
2436}
2437
ffb4593c
NT
2438/*
2439 * S_sublex_push
2440 * Create a new scope to save the lexing state. The scope will be
2441 * ended in S_sublex_done. Returns a '(', starting the function arguments
2442 * to the uc, lc, etc. found before.
2443 * Sets PL_lex_state to LEX_INTERPCONCAT.
2444 */
2445
76e3520e 2446STATIC I32
cea2e8a9 2447S_sublex_push(pTHX)
55497cff 2448{
27da23d5 2449 dVAR;
f46d017c 2450 ENTER;
55497cff 2451
3280af22 2452 PL_lex_state = PL_sublex_info.super_state;
651b5b28 2453 SAVEBOOL(PL_lex_dojoin);
3280af22 2454 SAVEI32(PL_lex_brackets);
78cdf107 2455 SAVEI32(PL_lex_allbrackets);
b27dce25 2456 SAVEI32(PL_lex_formbrack);
78cdf107 2457 SAVEI8(PL_lex_fakeeof);
3280af22
NIS
2458 SAVEI32(PL_lex_casemods);
2459 SAVEI32(PL_lex_starts);
651b5b28 2460 SAVEI8(PL_lex_state);
9da1dd8f 2461 SAVEPPTR(PL_sublex_info.re_eval_start);
7766f137 2462 SAVEVPTR(PL_lex_inpat);
98246f1e 2463 SAVEI16(PL_lex_inwhat);
57843af0 2464 SAVECOPLINE(PL_curcop);
3280af22 2465 SAVEPPTR(PL_bufptr);
8452ff4b 2466 SAVEPPTR(PL_bufend);
3280af22
NIS
2467 SAVEPPTR(PL_oldbufptr);
2468 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2469 SAVEPPTR(PL_last_lop);
2470 SAVEPPTR(PL_last_uni);
3280af22
NIS
2471 SAVEPPTR(PL_linestart);
2472 SAVESPTR(PL_linestr);
8edd5f42
RGS
2473 SAVEGENERICPV(PL_lex_brackstack);
2474 SAVEGENERICPV(PL_lex_casestack);
3280af22
NIS
2475
2476 PL_linestr = PL_lex_stuff;
a0714e2c 2477 PL_lex_stuff = NULL;
9da1dd8f 2478 PL_sublex_info.re_eval_start = NULL;
3280af22 2479
9cbb5ea2
GS
2480 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2481 = SvPVX(PL_linestr);
3280af22 2482 PL_bufend += SvCUR(PL_linestr);
bd61b366 2483 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2484 SAVEFREESV(PL_linestr);
2485
2486 PL_lex_dojoin = FALSE;
b27dce25 2487 PL_lex_brackets = PL_lex_formbrack = 0;
78cdf107
Z
2488 PL_lex_allbrackets = 0;
2489 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
a02a5408
JC
2490 Newx(PL_lex_brackstack, 120, char);
2491 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2492 PL_lex_casemods = 0;
2493 *PL_lex_casestack = '\0';
2494 PL_lex_starts = 0;
2495 PL_lex_state = LEX_INTERPCONCAT;
eb160463 2496 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
3280af22
NIS
2497
2498 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
bb16bae8 2499 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
3280af22
NIS
2500 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2501 PL_lex_inpat = PL_sublex_info.sub_op;
79072805 2502 else
5f66b61c 2503 PL_lex_inpat = NULL;
79072805 2504
55497cff 2505 return '(';
79072805
LW
2506}
2507
ffb4593c
NT
2508/*
2509 * S_sublex_done
2510 * Restores lexer state after a S_sublex_push.
2511 */
2512
76e3520e 2513STATIC I32
cea2e8a9 2514S_sublex_done(pTHX)
79072805 2515{
27da23d5 2516 dVAR;
3280af22 2517 if (!PL_lex_starts++) {
396482e1 2518 SV * const sv = newSVpvs("");
9aa983d2
JH
2519 if (SvUTF8(PL_linestr))
2520 SvUTF8_on(sv);
3280af22 2521 PL_expect = XOPERATOR;
6154021b 2522 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
79072805
LW
2523 return THING;
2524 }
2525
3280af22
NIS
2526 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2527 PL_lex_state = LEX_INTERPCASEMOD;
cea2e8a9 2528 return yylex();
79072805
LW
2529 }
2530
ffb4593c 2531 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
bb16bae8 2532 assert(PL_lex_inwhat != OP_TRANSR);
3280af22
NIS
2533 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2534 PL_linestr = PL_lex_repl;
2535 PL_lex_inpat = 0;
2536 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2537 PL_bufend += SvCUR(PL_linestr);
bd61b366 2538 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
2539 SAVEFREESV(PL_linestr);
2540 PL_lex_dojoin = FALSE;
2541 PL_lex_brackets = 0;
78cdf107
Z
2542 PL_lex_allbrackets = 0;
2543 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
3280af22
NIS
2544 PL_lex_casemods = 0;
2545 *PL_lex_casestack = '\0';
2546 PL_lex_starts = 0;
25da4f38 2547 if (SvEVALED(PL_lex_repl)) {
3280af22
NIS
2548 PL_lex_state = LEX_INTERPNORMAL;
2549 PL_lex_starts++;
e9fa98b2
HS
2550 /* we don't clear PL_lex_repl here, so that we can check later
2551 whether this is an evalled subst; that means we rely on the
2552 logic to ensure sublex_done() is called again only via the
2553 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
79072805 2554 }
e9fa98b2 2555 else {
3280af22 2556 PL_lex_state = LEX_INTERPCONCAT;
a0714e2c 2557 PL_lex_repl = NULL;
e9fa98b2 2558 }
79072805 2559 return ',';
ffed7fef
LW
2560 }
2561 else {
5db06880
NC
2562#ifdef PERL_MAD
2563 if (PL_madskills) {
cd81e915
NC
2564 if (PL_thiswhite) {
2565 if (!PL_endwhite)
6b29d1f5 2566 PL_endwhite = newSVpvs("");
cd81e915
NC
2567 sv_catsv(PL_endwhite, PL_thiswhite);
2568 PL_thiswhite = 0;
2569 }
2570 if (PL_thistoken)
76f68e9b 2571 sv_setpvs(PL_thistoken,"");
5db06880 2572 else
cd81e915 2573 PL_realtokenstart = -1;
5db06880
NC
2574 }
2575#endif
f46d017c 2576 LEAVE;
3280af22
NIS
2577 PL_bufend = SvPVX(PL_linestr);
2578 PL_bufend += SvCUR(PL_linestr);
2579 PL_expect = XOPERATOR;
09bef843 2580 PL_sublex_info.sub_inwhat = 0;
79072805 2581 return ')';
ffed7fef
LW
2582 }
2583}
2584
02aa26ce
NT
2585/*
2586 scan_const
2587
9da1dd8f
DM
2588 Extracts the next constant part of a pattern, double-quoted string,
2589 or transliteration. This is terrifying code.
2590
2591 For example, in parsing the double-quoted string "ab\x63$d", it would
2592 stop at the '$' and return an OP_CONST containing 'abc'.
02aa26ce 2593
94def140 2594 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2595 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2596 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2597
94def140
ST
2598 Returns a pointer to the character scanned up to. If this is
2599 advanced from the start pointer supplied (i.e. if anything was
9da1dd8f 2600 successfully parsed), will leave an OP_CONST for the substring scanned
6154021b 2601 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2602 by looking at the next characters herself.
2603
02aa26ce 2604 In patterns:
9da1dd8f
DM
2605 expand:
2606 \N{ABC} => \N{U+41.42.43}
2607
2608 pass through:
2609 all other \-char, including \N and \N{ apart from \N{ABC}
2610
2611 stops on:
2612 @ and $ where it appears to be a var, but not for $ as tail anchor
2613 \l \L \u \U \Q \E
2614 (?{ or (??{
2615
02aa26ce
NT
2616
2617 In transliterations:
2618 characters are VERY literal, except for - not at the start or end
94def140
ST
2619 of the string, which indicates a range. If the range is in bytes,
2620 scan_const expands the range to the full set of intermediate
2621 characters. If the range is in utf8, the hyphen is replaced with
2622 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2623
2624 In double-quoted strings:
2625 backslashes:
2626 double-quoted style: \r and \n
ff3f963a 2627 constants: \x31, etc.
94def140 2628 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2629 case and quoting: \U \Q \E
2630 stops on @ and $
2631
2632 scan_const does *not* construct ops to handle interpolated strings.
2633 It stops processing as soon as it finds an embedded $ or @ variable
2634 and leaves it to the caller to work out what's going on.
2635
94def140
ST
2636 embedded arrays (whether in pattern or not) could be:
2637 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2638
2639 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2640
2641 $ in pattern could be $foo or could be tail anchor. Assumption:
2642 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2643 followed by one of "()| \r\n\t"
02aa26ce 2644
9da1dd8f 2645 \1 (backreferences) are turned into $1 in substitutions
02aa26ce
NT
2646
2647 The structure of the code is
2648 while (there's a character to process) {
94def140
ST
2649 handle transliteration ranges
2650 skip regexp comments /(?#comment)/ and codes /(?{code})/
2651 skip #-initiated comments in //x patterns
2652 check for embedded arrays
02aa26ce
NT
2653 check for embedded scalars
2654 if (backslash) {
94def140 2655 deprecate \1 in substitution replacements
02aa26ce
NT
2656 handle string-changing backslashes \l \U \Q \E, etc.
2657 switch (what was escaped) {
94def140 2658 handle \- in a transliteration (becomes a literal -)
ff3f963a 2659 if a pattern and not \N{, go treat as regular character
94def140
ST
2660 handle \132 (octal characters)
2661 handle \x15 and \x{1234} (hex characters)
ff3f963a 2662 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2663 handle \cV (control characters)
2664 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2665 } (end switch)
77a135fe 2666 continue
02aa26ce 2667 } (end if backslash)
77a135fe 2668 handle regular character
02aa26ce 2669 } (end while character to read)
4e553d73 2670
02aa26ce
NT
2671*/
2672
76e3520e 2673STATIC char *
cea2e8a9 2674S_scan_const(pTHX_ char *start)
79072805 2675{
97aff369 2676 dVAR;
eb578fdb 2677 char *send = PL_bufend; /* end of the constant */
77a135fe
KW
2678 SV *sv = newSV(send - start); /* sv for the constant. See
2679 note below on sizing. */
eb578fdb
KW
2680 char *s = start; /* start of the constant */
2681 char *d = SvPVX(sv); /* destination for copies */
02aa26ce 2682 bool dorange = FALSE; /* are we in a translit range? */
c2e66d9e 2683 bool didrange = FALSE; /* did we just finish a range? */
2866decb 2684 bool in_charclass = FALSE; /* within /[...]/ */
b953e60c
KW
2685 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2686 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
77a135fe
KW
2687 to be UTF8? But, this can
2688 show as true when the source
2689 isn't utf8, as for example
2690 when it is entirely composed
2691 of hex constants */
2692
2693 /* Note on sizing: The scanned constant is placed into sv, which is
2694 * initialized by newSV() assuming one byte of output for every byte of
2695 * input. This routine expects newSV() to allocate an extra byte for a
2696 * trailing NUL, which this routine will append if it gets to the end of
2697 * the input. There may be more bytes of input than output (eg., \N{LATIN
2698 * CAPITAL LETTER A}), or more output than input if the constant ends up
2699 * recoded to utf8, but each time a construct is found that might increase
2700 * the needed size, SvGROW() is called. Its size parameter each time is
2701 * based on the best guess estimate at the time, namely the length used so
2702 * far, plus the length the current construct will occupy, plus room for
2703 * the trailing NUL, plus one byte for every input byte still unscanned */
2704
012bcf8d 2705 UV uv;
4c3a8340
ST
2706#ifdef EBCDIC
2707 UV literal_endpoint = 0;
e294cc5d 2708 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
4c3a8340 2709#endif
012bcf8d 2710
7918f24d
NC
2711 PERL_ARGS_ASSERT_SCAN_CONST;
2712
bb16bae8 2713 assert(PL_lex_inwhat != OP_TRANSR);
2b9d42f0
NIS
2714 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2715 /* If we are doing a trans and we know we want UTF8 set expectation */
2716 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2717 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2718 }
2719
2720
79072805 2721 while (s < send || dorange) {
ff3f963a 2722
02aa26ce 2723 /* get transliterations out of the way (they're most literal) */
3280af22 2724 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2725 /* expand a range A-Z to the full set of characters. AIE! */
79072805 2726 if (dorange) {
1ba5c669
JH
2727 I32 i; /* current expanded character */
2728 I32 min; /* first character in range */
2729 I32 max; /* last character in range */
02aa26ce 2730
e294cc5d
JH
2731#ifdef EBCDIC
2732 UV uvmax = 0;
2733#endif
2734
2735 if (has_utf8
2736#ifdef EBCDIC
2737 && !native_range
2738#endif
2739 ) {
9d4ba2ae 2740 char * const c = (char*)utf8_hop((U8*)d, -1);
8973db79
JH
2741 char *e = d++;
2742 while (e-- > c)
2743 *(e + 1) = *e;
25716404 2744 *c = (char)UTF_TO_NATIVE(0xff);
8973db79
JH
2745 /* mark the range as done, and continue */
2746 dorange = FALSE;
2747 didrange = TRUE;
2748 continue;
2749 }
2b9d42f0 2750
95a20fc0 2751 i = d - SvPVX_const(sv); /* remember current offset */
e294cc5d
JH
2752#ifdef EBCDIC
2753 SvGROW(sv,
2754 SvLEN(sv) + (has_utf8 ?
2755 (512 - UTF_CONTINUATION_MARK +
2756 UNISKIP(0x100))
2757 : 256));
2758 /* How many two-byte within 0..255: 128 in UTF-8,
2759 * 96 in UTF-8-mod. */
2760#else
9cbb5ea2 2761 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
e294cc5d 2762#endif
9cbb5ea2 2763 d = SvPVX(sv) + i; /* refresh d after realloc */
e294cc5d
JH
2764#ifdef EBCDIC
2765 if (has_utf8) {
2766 int j;
2767 for (j = 0; j <= 1; j++) {
2768 char * const c = (char*)utf8_hop((U8*)d, -1);
2769 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2770 if (j)
2771 min = (U8)uv;
2772 else if (uv < 256)
2773 max = (U8)uv;
2774 else {
2775 max = (U8)0xff; /* only to \xff */
2776 uvmax = uv; /* \x{100} to uvmax */
2777 }
2778 d = c; /* eat endpoint chars */
2779 }
2780 }
2781 else {
2782#endif
2783 d -= 2; /* eat the first char and the - */
2784 min = (U8)*d; /* first char in range */
2785 max = (U8)d[1]; /* last char in range */
2786#ifdef EBCDIC
2787 }
2788#endif
8ada0baa 2789
c2e66d9e 2790 if (min > max) {
01ec43d0 2791 Perl_croak(aTHX_
d1573ac7 2792 "Invalid range \"%c-%c\" in transliteration operator",
1ba5c669 2793 (char)min, (char)max);
c2e66d9e
GS
2794 }
2795
c7f1f016 2796#ifdef EBCDIC
4c3a8340
ST
2797 if (literal_endpoint == 2 &&
2798 ((isLOWER(min) && isLOWER(max)) ||
2799 (isUPPER(min) && isUPPER(max)))) {
8ada0baa
JH
2800 if (isLOWER(min)) {
2801 for (i = min; i <= max; i++)
2802 if (isLOWER(i))
db42d148 2803 *d++ = NATIVE_TO_NEED(has_utf8,i);
8ada0baa
JH
2804 } else {
2805 for (i = min; i <= max; i++)
2806 if (isUPPER(i))
db42d148 2807 *d++ = NATIVE_TO_NEED(has_utf8,i);
8ada0baa
JH
2808 }
2809 }
2810 else
2811#endif
2812 for (i = min; i <= max; i++)
e294cc5d
JH
2813#ifdef EBCDIC
2814 if (has_utf8) {
2815 const U8 ch = (U8)NATIVE_TO_UTF(i);
2816 if (UNI_IS_INVARIANT(ch))
2817 *d++ = (U8)i;
2818 else {
2819 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2820 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2821 }
2822 }
2823 else
2824#endif
2825 *d++ = (char)i;
2826
2827#ifdef EBCDIC
2828 if (uvmax) {
2829 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2830 if (uvmax > 0x101)
2831 *d++ = (char)UTF_TO_NATIVE(0xff);
2832 if (uvmax > 0x100)
2833 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2834 }
2835#endif
02aa26ce
NT
2836
2837 /* mark the range as done, and continue */
79072805 2838 dorange = FALSE;
01ec43d0 2839 didrange = TRUE;
4c3a8340
ST
2840#ifdef EBCDIC
2841 literal_endpoint = 0;
2842#endif
79072805 2843 continue;
4e553d73 2844 }
02aa26ce
NT
2845
2846 /* range begins (ignore - as first or last char) */
79072805 2847 else if (*s == '-' && s+1 < send && s != start) {
4e553d73 2848 if (didrange) {
1fafa243 2849 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
01ec43d0 2850 }
e294cc5d
JH
2851 if (has_utf8
2852#ifdef EBCDIC
2853 && !native_range
2854#endif
2855 ) {
25716404 2856 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
a0ed51b3
LW
2857 s++;
2858 continue;
2859 }
79072805
LW
2860 dorange = TRUE;
2861 s++;
01ec43d0
GS
2862 }
2863 else {
2864 didrange = FALSE;
4c3a8340
ST
2865#ifdef EBCDIC
2866 literal_endpoint = 0;
e294cc5d 2867 native_range = TRUE;
4c3a8340 2868#endif
01ec43d0 2869 }
79072805 2870 }
02aa26ce
NT
2871
2872 /* if we get here, we're not doing a transliteration */
2873
e4a2df84
DM
2874 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
2875 char *s1 = s-1;
2876 int esc = 0;
2877 while (s1 >= start && *s1-- == '\\')
2878 esc = !esc;
2879 if (!esc)
2880 in_charclass = TRUE;
2881 }
2866decb 2882
e4a2df84
DM
2883 else if (*s == ']' && PL_lex_inpat && in_charclass) {
2884 char *s1 = s-1;
2885 int esc = 0;
2886 while (s1 >= start && *s1-- == '\\')
2887 esc = !esc;
2888 if (!esc)
2889 in_charclass = FALSE;
2890 }
2866decb 2891
9da1dd8f
DM
2892 /* skip for regexp comments /(?#comment)/, except for the last
2893 * char, which will be done separately.
2894 * Stop on (?{..}) and friends */
2895
3280af22 2896 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
cc6b7395 2897 if (s[2] == '#') {
e994fd66 2898 while (s+1 < send && *s != ')')
db42d148 2899 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
155aba94 2900 }
2866decb 2901 else if (!PL_lex_casemods && !in_charclass &&
d3cec5e5
DM
2902 ( s[2] == '{' /* This should match regcomp.c */
2903 || (s[2] == '?' && s[3] == '{')))
155aba94 2904 {
9da1dd8f 2905 break;
cc6b7395 2906 }
748a9306 2907 }
02aa26ce
NT
2908
2909 /* likewise skip #-initiated comments in //x patterns */
3280af22 2910 else if (*s == '#' && PL_lex_inpat &&
73134a2e 2911 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
748a9306 2912 while (s+1 < send && *s != '\n')
db42d148 2913 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
748a9306 2914 }
02aa26ce 2915
9da1dd8f
DM
2916 /* no further processing of single-quoted regex */
2917 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
2918 goto default_action;
2919
5d1d4326 2920 /* check for embedded arrays
da6eedaa 2921 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
5d1d4326 2922 */
1749ea0d
ST
2923 else if (*s == '@' && s[1]) {
2924 if (isALNUM_lazy_if(s+1,UTF))
2925 break;
2926 if (strchr(":'{$", s[1]))
2927 break;
2928 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2929 break; /* in regexp, neither @+ nor @- are interpolated */
2930 }
02aa26ce
NT
2931
2932 /* check for embedded scalars. only stop if we're sure it's a
2933 variable.
2934 */
79072805 2935 else if (*s == '$') {
3280af22 2936 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 2937 break;
77772344 2938 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
2939 if (s[1] == '\\') {
2940 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2941 "Possible unintended interpolation of $\\ in regex");
77772344 2942 }
79072805 2943 break; /* in regexp, $ might be tail anchor */
77772344 2944 }
79072805 2945 }
02aa26ce 2946
2b9d42f0
NIS
2947 /* End of else if chain - OP_TRANS rejoin rest */
2948
02aa26ce 2949 /* backslashes */
79072805 2950 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
2951 char* e; /* Can be used for ending '}', etc. */
2952
79072805 2953 s++;
02aa26ce 2954
7d0fc23c
KW
2955 /* warn on \1 - \9 in substitution replacements, but note that \11
2956 * is an octal; and \19 is \1 followed by '9' */
3280af22 2957 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
a0d0e21e 2958 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
79072805 2959 {
a2a5de95 2960 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
2961 *--s = '$';
2962 break;
2963 }
02aa26ce
NT
2964
2965 /* string-change backslash escapes */
838f2281 2966 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
79072805
LW
2967 --s;
2968 break;
2969 }
ff3f963a
KW
2970 /* In a pattern, process \N, but skip any other backslash escapes.
2971 * This is because we don't want to translate an escape sequence
2972 * into a meta symbol and have the regex compiler use the meta
2973 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2974 * in spite of this, we do have to process \N here while the proper
2975 * charnames handler is in scope. See bugs #56444 and #62056.
2976 * There is a complication because \N in a pattern may also stand
2977 * for 'match a non-nl', and not mean a charname, in which case its
2978 * processing should be deferred to the regex compiler. To be a
2979 * charname it must be followed immediately by a '{', and not look
2980 * like \N followed by a curly quantifier, i.e., not something like
2981 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2982 * quantifier */
2983 else if (PL_lex_inpat
2984 && (*s != 'N'
2985 || s[1] != '{'
2986 || regcurly(s + 1)))
2987 {
cc74c5bd
ST
2988 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2989 goto default_action;
2990 }
02aa26ce 2991
79072805 2992 switch (*s) {
02aa26ce
NT
2993
2994 /* quoted - in transliterations */
79072805 2995 case '-':
3280af22 2996 if (PL_lex_inwhat == OP_TRANS) {
79072805
LW
2997 *d++ = *s++;
2998 continue;
2999 }
3000 /* FALL THROUGH */
3001 default:
11b8faa4 3002 {
e4ca4584 3003 if ((isALNUMC(*s)))
a2a5de95
NC
3004 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3005 "Unrecognized escape \\%c passed through",
3006 *s);
11b8faa4 3007 /* default action is to copy the quoted character */
f9a63242 3008 goto default_action;
11b8faa4 3009 }
02aa26ce 3010
632403cc 3011 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
3012 case '0': case '1': case '2': case '3':
3013 case '4': case '5': case '6': case '7':
ba210ebe 3014 {
53305cf1
NC
3015 I32 flags = 0;
3016 STRLEN len = 3;
77a135fe 3017 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
ba210ebe
JH
3018 s += len;
3019 }
012bcf8d 3020 goto NUM_ESCAPE_INSERT;
02aa26ce 3021
f0a2b745
KW
3022 /* eg. \o{24} indicates the octal constant \024 */
3023 case 'o':
3024 {
3025 STRLEN len;
454155d9 3026 const char* error;
f0a2b745 3027
454155d9 3028 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
f0a2b745 3029 s += len;
454155d9 3030 if (! valid) {
f0a2b745
KW
3031 yyerror(error);
3032 continue;
3033 }
3034 goto NUM_ESCAPE_INSERT;
3035 }
3036
77a135fe 3037 /* eg. \x24 indicates the hex constant 0x24 */
79072805 3038 case 'x':
a0481293 3039 {
53305cf1 3040 STRLEN len;
a0481293 3041 const char* error;
355860ce 3042
a0481293
KW
3043 bool valid = grok_bslash_x(s, &uv, &len, &error, 1);
3044 s += len;
3045 if (! valid) {
3046 yyerror(error);
355860ce 3047 continue;
ba210ebe 3048 }
012bcf8d
GS
3049 }
3050
3051 NUM_ESCAPE_INSERT:
ff3f963a
KW
3052 /* Insert oct or hex escaped character. There will always be
3053 * enough room in sv since such escapes will be longer than any
3054 * UTF-8 sequence they can end up as, except if they force us
3055 * to recode the rest of the string into utf8 */
ba7cea30 3056
77a135fe 3057 /* Here uv is the ordinal of the next character being added in
ff3f963a 3058 * unicode (converted from native). */
77a135fe 3059 if (!UNI_IS_INVARIANT(uv)) {
9aa983d2 3060 if (!has_utf8 && uv > 255) {
77a135fe
KW
3061 /* Might need to recode whatever we have accumulated so
3062 * far if it contains any chars variant in utf8 or
3063 * utf-ebcdic. */
3064
3065 SvCUR_set(sv, d - SvPVX_const(sv));
3066 SvPOK_on(sv);
3067 *d = '\0';
77a135fe 3068 /* See Note on sizing above. */
7bf79863
KW
3069 sv_utf8_upgrade_flags_grow(sv,
3070 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3071 UNISKIP(uv) + (STRLEN)(send - s) + 1);
77a135fe
KW
3072 d = SvPVX(sv) + SvCUR(sv);
3073 has_utf8 = TRUE;
012bcf8d
GS
3074 }
3075
77a135fe
KW
3076 if (has_utf8) {
3077 d = (char*)uvuni_to_utf8((U8*)d, uv);
f9a63242
JH
3078 if (PL_lex_inwhat == OP_TRANS &&
3079 PL_sublex_info.sub_op) {
3080 PL_sublex_info.sub_op->op_private |=
3081 (PL_lex_repl ? OPpTRANS_FROM_UTF
3082 : OPpTRANS_TO_UTF);
f9a63242 3083 }
e294cc5d
JH
3084#ifdef EBCDIC
3085 if (uv > 255 && !dorange)
3086 native_range = FALSE;
3087#endif
012bcf8d 3088 }
a0ed51b3 3089 else {
012bcf8d 3090 *d++ = (char)uv;
a0ed51b3 3091 }
012bcf8d
GS
3092 }
3093 else {
c4d5f83a 3094 *d++ = (char) uv;
a0ed51b3 3095 }
79072805 3096 continue;
02aa26ce 3097
4a2d328f 3098 case 'N':
ff3f963a
KW
3099 /* In a non-pattern \N must be a named character, like \N{LATIN
3100 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3101 * mean to match a non-newline. For non-patterns, named
3102 * characters are converted to their string equivalents. In
3103 * patterns, named characters are not converted to their
3104 * ultimate forms for the same reasons that other escapes
3105 * aren't. Instead, they are converted to the \N{U+...} form
3106 * to get the value from the charnames that is in effect right
3107 * now, while preserving the fact that it was a named character
3108 * so that the regex compiler knows this */
3109
3110 /* This section of code doesn't generally use the
3111 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3112 * a close examination of this macro and determined it is a
3113 * no-op except on utfebcdic variant characters. Every
3114 * character generated by this that would normally need to be
3115 * enclosed by this macro is invariant, so the macro is not
7538f724
KW
3116 * needed, and would complicate use of copy(). XXX There are
3117 * other parts of this file where the macro is used
3118 * inconsistently, but are saved by it being a no-op */
ff3f963a
KW
3119
3120 /* The structure of this section of code (besides checking for
3121 * errors and upgrading to utf8) is:
3122 * Further disambiguate between the two meanings of \N, and if
3123 * not a charname, go process it elsewhere
0a96133f
KW
3124 * If of form \N{U+...}, pass it through if a pattern;
3125 * otherwise convert to utf8
3126 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3127 * pattern; otherwise convert to utf8 */
ff3f963a
KW
3128
3129 /* Here, s points to the 'N'; the test below is guaranteed to
3130 * succeed if we are being called on a pattern as we already
3131 * know from a test above that the next character is a '{'.
3132 * On a non-pattern \N must mean 'named sequence, which
3133 * requires braces */
3134 s++;
3135 if (*s != '{') {
3136 yyerror("Missing braces on \\N{}");
3137 continue;
3138 }
3139 s++;
3140
0a96133f 3141 /* If there is no matching '}', it is an error. */
ff3f963a
KW
3142 if (! (e = strchr(s, '}'))) {
3143 if (! PL_lex_inpat) {
5777a3f7 3144 yyerror("Missing right brace on \\N{}");
0a96133f
KW
3145 } else {
3146 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
dbc0d4f2 3147 }
0a96133f 3148 continue;
ff3f963a 3149 }
cddc7ef4 3150
ff3f963a 3151 /* Here it looks like a named character */
cddc7ef4 3152
ff3f963a
KW
3153 if (PL_lex_inpat) {
3154
3155 /* XXX This block is temporary code. \N{} implies that the
3156 * pattern is to have Unicode semantics, and therefore
3157 * currently has to be encoded in utf8. By putting it in
3158 * utf8 now, we save a whole pass in the regular expression
3159 * compiler. Once that code is changed so Unicode
3160 * semantics doesn't necessarily have to be in utf8, this
da3a4baf
KW
3161 * block should be removed. However, the code that parses
3162 * the output of this would have to be changed to not
3163 * necessarily expect utf8 */
ff3f963a 3164 if (!has_utf8) {
77a135fe 3165 SvCUR_set(sv, d - SvPVX_const(sv));
f08d6ad9 3166 SvPOK_on(sv);
e4f3eed8 3167 *d = '\0';
77a135fe 3168 /* See Note on sizing above. */
7bf79863 3169 sv_utf8_upgrade_flags_grow(sv,
ff3f963a
KW
3170 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3171 /* 5 = '\N{' + cur char + NUL */
3172 (STRLEN)(send - s) + 5);
f08d6ad9 3173 d = SvPVX(sv) + SvCUR(sv);
89491803 3174 has_utf8 = TRUE;
ff3f963a
KW
3175 }
3176 }
423cee85 3177
ff3f963a
KW
3178 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3179 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3180 | PERL_SCAN_DISALLOW_PREFIX;
3181 STRLEN len;
3182
3183 /* For \N{U+...}, the '...' is a unicode value even on
3184 * EBCDIC machines */
3185 s += 2; /* Skip to next char after the 'U+' */
3186 len = e - s;
3187 uv = grok_hex(s, &len, &flags, NULL);
3188 if (len == 0 || len != (STRLEN)(e - s)) {
3189 yyerror("Invalid hexadecimal number in \\N{U+...}");
3190 s = e + 1;
3191 continue;
3192 }
3193
3194 if (PL_lex_inpat) {
3195
e2a7e165
KW
3196 /* On non-EBCDIC platforms, pass through to the regex
3197 * compiler unchanged. The reason we evaluated the
3198 * number above is to make sure there wasn't a syntax
3199 * error. But on EBCDIC we convert to native so
3200 * downstream code can continue to assume it's native
3201 */
ff3f963a 3202 s -= 5; /* Include the '\N{U+' */
e2a7e165
KW
3203#ifdef EBCDIC
3204 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3205 and the \0 */
3206 "\\N{U+%X}",
3207 (unsigned int) UNI_TO_NATIVE(uv));
3208#else
ff3f963a
KW
3209 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3210 d += e - s + 1;
e2a7e165 3211#endif
ff3f963a
KW
3212 }
3213 else { /* Not a pattern: convert the hex to string */
3214
3215 /* If destination is not in utf8, unconditionally
3216 * recode it to be so. This is because \N{} implies
3217 * Unicode semantics, and scalars have to be in utf8
3218 * to guarantee those semantics */
3219 if (! has_utf8) {
3220 SvCUR_set(sv, d - SvPVX_const(sv));
3221 SvPOK_on(sv);
3222 *d = '\0';
3223 /* See Note on sizing above. */
3224 sv_utf8_upgrade_flags_grow(
3225 sv,
3226 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3227 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3228 d = SvPVX(sv) + SvCUR(sv);
3229 has_utf8 = TRUE;
3230 }
3231
3232 /* Add the string to the output */
3233 if (UNI_IS_INVARIANT(uv)) {
3234 *d++ = (char) uv;
3235 }
3236 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3237 }
3238 }
3239 else { /* Here is \N{NAME} but not \N{U+...}. */
3240
3241 SV *res; /* result from charnames */
3242 const char *str; /* the string in 'res' */
3243 STRLEN len; /* its length */
3244
3245 /* Get the value for NAME */
3246 res = newSVpvn(s, e - s);
3247 res = new_constant( NULL, 0, "charnames",
3248 /* includes all of: \N{...} */
3249 res, NULL, s - 3, e - s + 4 );
3250
3251 /* Most likely res will be in utf8 already since the
3252 * standard charnames uses pack U, but a custom translator
3253 * can leave it otherwise, so make sure. XXX This can be
3254 * revisited to not have charnames use utf8 for characters
3255 * that don't need it when regexes don't have to be in utf8
3256 * for Unicode semantics. If doing so, remember EBCDIC */
3257 sv_utf8_upgrade(res);
3258 str = SvPV_const(res, len);
3259
3260 /* Don't accept malformed input */
3261 if (! is_utf8_string((U8 *) str, len)) {
3262 yyerror("Malformed UTF-8 returned by \\N");
3263 }
3264 else if (PL_lex_inpat) {
3265
3266 if (! len) { /* The name resolved to an empty string */
3267 Copy("\\N{}", d, 4, char);
3268 d += 4;
3269 }
3270 else {
3271 /* In order to not lose information for the regex
3272 * compiler, pass the result in the specially made
3273 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3274 * the code points in hex of each character
3275 * returned by charnames */
3276
3277 const char *str_end = str + len;
3278 STRLEN char_length; /* cur char's byte length */
3279 STRLEN output_length; /* and the number of bytes
3280 after this is translated
3281 into hex digits */
3282 const STRLEN off = d - SvPVX_const(sv);
3283
3284 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3285 * max('U+', '.'); and 1 for NUL */
3286 char hex_string[2 * UTF8_MAXBYTES + 5];
3287
3288 /* Get the first character of the result. */
3289 U32 uv = utf8n_to_uvuni((U8 *) str,
3290 len,
3291 &char_length,
3292 UTF8_ALLOW_ANYUV);
3293
3294 /* The call to is_utf8_string() above hopefully
3295 * guarantees that there won't be an error. But
3296 * it's easy here to make sure. The function just
3297 * above warns and returns 0 if invalid utf8, but
3298 * it can also return 0 if the input is validly a
3299 * NUL. Disambiguate */
3300 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3301 uv = UNICODE_REPLACEMENT;
3302 }
3303
3304 /* Convert first code point to hex, including the
e2a7e165
KW
3305 * boiler plate before it. For all these, we
3306 * convert to native format so that downstream code
3307 * can continue to assume the input is native */
78c35590 3308 output_length =
3353de27 3309 my_snprintf(hex_string, sizeof(hex_string),
e2a7e165
KW
3310 "\\N{U+%X",
3311 (unsigned int) UNI_TO_NATIVE(uv));
ff3f963a
KW
3312
3313 /* Make sure there is enough space to hold it */
3314 d = off + SvGROW(sv, off
3315 + output_length
3316 + (STRLEN)(send - e)
3317 + 2); /* '}' + NUL */
3318 /* And output it */
3319 Copy(hex_string, d, output_length, char);
3320 d += output_length;
3321
3322 /* For each subsequent character, append dot and
3323 * its ordinal in hex */
3324 while ((str += char_length) < str_end) {
3325 const STRLEN off = d - SvPVX_const(sv);
3326 U32 uv = utf8n_to_uvuni((U8 *) str,
3327 str_end - str,
3328 &char_length,
3329 UTF8_ALLOW_ANYUV);
3330 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3331 uv = UNICODE_REPLACEMENT;
3332 }
3333
78c35590 3334 output_length =
3353de27 3335 my_snprintf(hex_string, sizeof(hex_string),
e2a7e165
KW
3336 ".%X",
3337 (unsigned int) UNI_TO_NATIVE(uv));
ff3f963a
KW
3338
3339 d = off + SvGROW(sv, off
3340 + output_length
3341 + (STRLEN)(send - e)
3342 + 2); /* '}' + NUL */
3343 Copy(hex_string, d, output_length, char);
3344 d += output_length;
3345 }
3346
3347 *d++ = '}'; /* Done. Add the trailing brace */
3348 }
3349 }
3350 else { /* Here, not in a pattern. Convert the name to a
3351 * string. */
3352
3353 /* If destination is not in utf8, unconditionally
3354 * recode it to be so. This is because \N{} implies
3355 * Unicode semantics, and scalars have to be in utf8
3356 * to guarantee those semantics */
3357 if (! has_utf8) {
3358 SvCUR_set(sv, d - SvPVX_const(sv));
3359 SvPOK_on(sv);
3360 *d = '\0';
3361 /* See Note on sizing above. */
3362 sv_utf8_upgrade_flags_grow(sv,
3363 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3364 len + (STRLEN)(send - s) + 1);
3365 d = SvPVX(sv) + SvCUR(sv);
3366 has_utf8 = TRUE;
3367 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3368
3369 /* See Note on sizing above. (NOTE: SvCUR() is not
3370 * set correctly here). */
3371 const STRLEN off = d - SvPVX_const(sv);
3372 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3373 }
3374 Copy(str, d, len, char);
3375 d += len;
423cee85 3376 }
423cee85 3377 SvREFCNT_dec(res);
cb233ae3
KW
3378
3379 /* Deprecate non-approved name syntax */
3380 if (ckWARN_d(WARN_DEPRECATED)) {
3381 bool problematic = FALSE;
3382 char* i = s;
3383
3384 /* For non-ut8 input, look to see that the first
3385 * character is an alpha, then loop through the rest
3386 * checking that each is a continuation */
3387 if (! this_utf8) {
3388 if (! isALPHAU(*i)) problematic = TRUE;
3389 else for (i = s + 1; i < e; i++) {
3390 if (isCHARNAME_CONT(*i)) continue;
3391 problematic = TRUE;
3392 break;
3393 }
3394 }
3395 else {
3396 /* Similarly for utf8. For invariants can check
3397 * directly. We accept anything above the latin1
3398 * range because it is immaterial to Perl if it is
3399 * correct or not, and is expensive to check. But
3400 * it is fairly easy in the latin1 range to convert
3401 * the variants into a single character and check
3402 * those */
3403 if (UTF8_IS_INVARIANT(*i)) {
3404 if (! isALPHAU(*i)) problematic = TRUE;
3405 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
81c14aa2 3406 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
cb233ae3
KW
3407 *(i+1)))))
3408 {
3409 problematic = TRUE;
3410 }
3411 }
3412 if (! problematic) for (i = s + UTF8SKIP(s);
3413 i < e;
3414 i+= UTF8SKIP(i))
3415 {
3416 if (UTF8_IS_INVARIANT(*i)) {
3417 if (isCHARNAME_CONT(*i)) continue;
3418 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3419 continue;
3420 } else if (isCHARNAME_CONT(
3421 UNI_TO_NATIVE(
81c14aa2 3422 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
cb233ae3
KW
3423 {
3424 continue;
3425 }
3426 problematic = TRUE;
3427 break;
3428 }
3429 }
3430 if (problematic) {
6e1bad6c
KW
3431 /* The e-i passed to the final %.*s makes sure that
3432 * should the trailing NUL be missing that this
3433 * print won't run off the end of the string */
cb233ae3 3434 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
b00fc8d4
NC
3435 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3436 (int)(i - s + 1), s, (int)(e - i), i + 1);
cb233ae3
KW
3437 }
3438 }
3439 } /* End \N{NAME} */
ff3f963a
KW
3440#ifdef EBCDIC
3441 if (!dorange)
3442 native_range = FALSE; /* \N{} is defined to be Unicode */
3443#endif
3444 s = e + 1; /* Point to just after the '}' */
423cee85
JH
3445 continue;
3446
02aa26ce 3447 /* \c is a control character */
79072805
LW
3448 case 'c':
3449 s++;
961ce445 3450 if (s < send) {
17a3df4c 3451 *d++ = grok_bslash_c(*s++, has_utf8, 1);
ba210ebe 3452 }
961ce445
RGS
3453 else {
3454 yyerror("Missing control char name in \\c");
3455 }
79072805 3456 continue;
02aa26ce
NT
3457
3458 /* printf-style backslashes, formfeeds, newlines, etc */
79072805 3459 case 'b':
db42d148 3460 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
79072805
LW
3461 break;
3462 case 'n':
db42d148 3463 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
79072805
LW
3464 break;
3465 case 'r':
db42d148 3466 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
79072805
LW
3467 break;
3468 case 'f':
db42d148 3469 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
79072805
LW
3470 break;
3471 case 't':
db42d148 3472 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
79072805 3473 break;
34a3fe2a 3474 case 'e':
db42d148 3475 *d++ = ASCII_TO_NEED(has_utf8,'\033');
34a3fe2a
PP
3476 break;
3477 case 'a':
db42d148 3478 *d++ = ASCII_TO_NEED(has_utf8,'\007');
79072805 3479 break;
02aa26ce
NT
3480 } /* end switch */
3481
79072805
LW
3482 s++;
3483 continue;
02aa26ce 3484 } /* end if (backslash) */
4c3a8340
ST
3485#ifdef EBCDIC
3486 else
3487 literal_endpoint++;
3488#endif
02aa26ce 3489
f9a63242 3490 default_action:
77a135fe
KW
3491 /* If we started with encoded form, or already know we want it,
3492 then encode the next character */
3493 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2b9d42f0 3494 STRLEN len = 1;
77a135fe
KW
3495
3496
3497 /* One might think that it is wasted effort in the case of the
3498 * source being utf8 (this_utf8 == TRUE) to take the next character
3499 * in the source, convert it to an unsigned value, and then convert
3500 * it back again. But the source has not been validated here. The
3501 * routine that does the conversion checks for errors like
3502 * malformed utf8 */
3503
5f66b61c
AL
3504 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3505 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
77a135fe
KW
3506 if (!has_utf8) {
3507 SvCUR_set(sv, d - SvPVX_const(sv));
3508 SvPOK_on(sv);
3509 *d = '\0';
77a135fe 3510 /* See Note on sizing above. */
7bf79863
KW
3511 sv_utf8_upgrade_flags_grow(sv,
3512 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3513 need + (STRLEN)(send - s) + 1);
77a135fe
KW
3514 d = SvPVX(sv) + SvCUR(sv);
3515 has_utf8 = TRUE;
3516 } else if (need > len) {
3517 /* encoded value larger than old, may need extra space (NOTE:
3518 * SvCUR() is not set correctly here). See Note on sizing
3519 * above. */
9d4ba2ae 3520 const STRLEN off = d - SvPVX_const(sv);
77a135fe 3521 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2b9d42f0 3522 }
77a135fe
KW
3523 s += len;
3524
5f66b61c 3525 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
e294cc5d
JH
3526#ifdef EBCDIC
3527 if (uv > 255 && !dorange)
3528 native_range = FALSE;
3529#endif
2b9d42f0
NIS
3530 }
3531 else {
3532 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3533 }
02aa26ce
NT
3534 } /* while loop to process each character */
3535
3536 /* terminate the string and set up the sv */
79072805 3537 *d = '\0';
95a20fc0 3538 SvCUR_set(sv, d - SvPVX_const(sv));
2b9d42f0 3539 if (SvCUR(sv) >= SvLEN(sv))
5637ef5b
NC
3540 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3541 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
2b9d42f0 3542
79072805 3543 SvPOK_on(sv);
9f4817db 3544 if (PL_encoding && !has_utf8) {
d0063567
DK
3545 sv_recode_to_utf8(sv, PL_encoding);
3546 if (SvUTF8(sv))
3547 has_utf8 = TRUE;
9f4817db 3548 }
2b9d42f0 3549 if (has_utf8) {
7e2040f0 3550 SvUTF8_on(sv);
2b9d42f0 3551 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
d0063567 3552 PL_sublex_info.sub_op->op_private |=
2b9d42f0
NIS
3553 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3554 }
3555 }
79072805 3556
02aa26ce 3557 /* shrink the sv if we allocated more than we used */
79072805 3558 if (SvCUR(sv) + 5 < SvLEN(sv)) {
1da4ca5f 3559 SvPV_shrink_to_cur(sv);
79072805 3560 }
02aa26ce 3561
6154021b 3562 /* return the substring (via pl_yylval) only if we parsed anything */
3280af22 3563 if (s > PL_bufptr) {
eb0d8d16
NC
3564 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3565 const char *const key = PL_lex_inpat ? "qr" : "q";
3566 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3567 const char *type;
3568 STRLEN typelen;
3569
3570 if (PL_lex_inwhat == OP_TRANS) {
3571 type = "tr";
3572 typelen = 2;
3573 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3574 type = "s";
3575 typelen = 1;
9da1dd8f
DM
3576 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3577 type = "q";
3578 typelen = 1;
eb0d8d16
NC
3579 } else {
3580 type = "qq";
3581 typelen = 2;
3582 }
3583
3584 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3585 type, typelen);
3586 }
6154021b 3587 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
b3ac6de7 3588 } else
8990e307 3589 SvREFCNT_dec(sv);
79072805
LW
3590 return s;
3591}
3592
ffb4593c
NT
3593/* S_intuit_more
3594 * Returns TRUE if there's more to the expression (e.g., a subscript),
3595 * FALSE otherwise.
ffb4593c
NT
3596 *
3597 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3598 *
3599 * ->[ and ->{ return TRUE
3600 * { and [ outside a pattern are always subscripts, so return TRUE
3601 * if we're outside a pattern and it's not { or [, then return FALSE
3602 * if we're in a pattern and the first char is a {
3603 * {4,5} (any digits around the comma) returns FALSE
3604 * if we're in a pattern and the first char is a [
3605 * [] returns FALSE
3606 * [SOMETHING] has a funky algorithm to decide whether it's a
3607 * character class or not. It has to deal with things like
3608 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3609 * anything else returns TRUE
3610 */
3611
9cbb5ea2
GS
3612/* This is the one truly awful dwimmer necessary to conflate C and sed. */
3613
76e3520e 3614STATIC int
cea2e8a9 3615S_intuit_more(pTHX_ register char *s)
79072805 3616{
97aff369 3617 dVAR;
7918f24d
NC
3618
3619 PERL_ARGS_ASSERT_INTUIT_MORE;
3620
3280af22 3621 if (PL_lex_brackets)
79072805
LW
3622 return TRUE;
3623 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3624 return TRUE;
3625 if (*s != '{' && *s != '[')
3626 return FALSE;
3280af22 3627 if (!PL_lex_inpat)
79072805
LW
3628 return TRUE;
3629
3630 /* In a pattern, so maybe we have {n,m}. */
3631 if (*s == '{') {
b3155d95 3632 if (regcurly(s)) {
79072805 3633 return FALSE;
b3155d95 3634 }
79072805 3635 return TRUE;
79072805
LW
3636 }
3637
3638 /* On the other hand, maybe we have a character class */
3639
3640 s++;
3641 if (*s == ']' || *s == '^')
3642 return FALSE;
3643 else {
ffb4593c 3644 /* this is terrifying, and it works */
79072805
LW
3645 int weight = 2; /* let's weigh the evidence */
3646 char seen[256];
f27ffc4a 3647 unsigned char un_char = 255, last_un_char;
9d4ba2ae 3648 const char * const send = strchr(s,']');
3280af22 3649 char tmpbuf[sizeof PL_tokenbuf * 4];
79072805
LW
3650
3651 if (!send) /* has to be an expression */
3652 return TRUE;
3653
3654 Zero(seen,256,char);
3655 if (*s == '$')
3656 weight -= 3;
3657 else if (isDIGIT(*s)) {
3658 if (s[1] != ']') {
3659 if (isDIGIT(s[1]) && s[2] == ']')
3660 weight -= 10;
3661 }
3662 else
3663 weight -= 100;
3664 }
3665 for (; s < send; s++) {
3666 last_un_char = un_char;
3667 un_char = (unsigned char)*s;
3668 switch (*s) {
3669 case '@':
3670 case '&':
3671 case '$':
3672 weight -= seen[un_char] * 10;
7e2040f0 3673 if (isALNUM_lazy_if(s+1,UTF)) {
90e5519e 3674 int len;
8903cb82 3675 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
90e5519e 3676 len = (int)strlen(tmpbuf);
6fbd0d97
BF
3677 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3678 UTF ? SVf_UTF8 : 0, SVt_PV))
79072805
LW
3679 weight -= 100;
3680 else
3681 weight -= 10;
3682 }
3683 else if (*s == '$' && s[1] &&
93a17b20
LW
3684 strchr("[#!%*<>()-=",s[1])) {
3685 if (/*{*/ strchr("])} =",s[2]))
79072805
LW
3686 weight -= 10;
3687 else
3688 weight -= 1;
3689 }
3690 break;
3691 case '\\':
3692 un_char = 254;
3693 if (s[1]) {
93a17b20 3694 if (strchr("wds]",s[1]))
79072805 3695 weight += 100;
10edeb5d 3696 else if (seen[(U8)'\''] || seen[(U8)'"'])
79072805 3697 weight += 1;
93a17b20 3698 else if (strchr("rnftbxcav",s[1]))
79072805
LW
3699 weight += 40;
3700 else if (isDIGIT(s[1])) {
3701 weight += 40;
3702 while (s[1] && isDIGIT(s[1]))
3703 s++;
3704 }
3705 }
3706 else
3707 weight += 100;
3708 break;
3709 case '-':
3710 if (s[1] == '\\')
3711 weight += 50;
93a17b20 3712 if (strchr("aA01! ",last_un_char))
79072805 3713 weight += 30;
93a17b20 3714 if (strchr("zZ79~",s[1]))
79072805 3715 weight += 30;
f27ffc4a
GS
3716 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3717 weight -= 5; /* cope with negative subscript */
79072805
LW
3718 break;
3719 default:
3792a11b
NC
3720 if (!isALNUM(last_un_char)
3721 && !(last_un_char == '$' || last_un_char == '@'
3722 || last_un_char == '&')
3723 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
79072805
LW
3724 char *d = tmpbuf;
3725 while (isALPHA(*s))
3726 *d++ = *s++;
3727 *d = '\0';
5458a98a 3728 if (keyword(tmpbuf, d - tmpbuf, 0))
79072805
LW
3729 weight -= 150;
3730 }
3731 if (un_char == last_un_char + 1)
3732 weight += 5;
3733 weight -= seen[un_char];
3734 break;
3735 }
3736 seen[un_char]++;
3737 }
3738 if (weight >= 0) /* probably a character class */
3739 return FALSE;
3740 }
3741
3742 return TRUE;
3743}
ffed7fef 3744
ffb4593c
NT
3745/*
3746 * S_intuit_method
3747 *
3748 * Does all the checking to disambiguate
3749 * foo bar
3750 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3751 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3752 *
3753 * First argument is the stuff after the first token, e.g. "bar".
3754 *
a4fd4a89 3755 * Not a method if foo is a filehandle.
ffb4593c
NT
3756 * Not a method if foo is a subroutine prototyped to take a filehandle.
3757 * Not a method if it's really "Foo $bar"
3758 * Method if it's "foo $bar"
3759 * Not a method if it's really "print foo $bar"<