This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
yada-yada is a term, not an operator
[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
f0e67a1d
Z
26This is the lower layer of the Perl parser, managing characters and tokens.
27
28=for apidoc AmU|yy_parser *|PL_parser
29
30Pointer to a structure encapsulating the state of the parsing operation
31currently in progress. The pointer can be locally changed to perform
32a nested parse without interfering with the state of an outer parse.
33Individual members of C<PL_parser> have their own documentation.
34
35=cut
36*/
37
378cc40b 38#include "EXTERN.h"
864dbfa3 39#define PERL_IN_TOKE_C
378cc40b 40#include "perl.h"
f7e03a10 41#include "dquote_inline.h"
378cc40b 42
eb0d8d16
NC
43#define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
45
6154021b 46#define pl_yylval (PL_parser->yylval)
d3b6f988 47
199e78b7
DM
48/* XXX temporary backwards compatibility */
49#define PL_lex_brackets (PL_parser->lex_brackets)
78cdf107
Z
50#define PL_lex_allbrackets (PL_parser->lex_allbrackets)
51#define PL_lex_fakeeof (PL_parser->lex_fakeeof)
199e78b7
DM
52#define PL_lex_brackstack (PL_parser->lex_brackstack)
53#define PL_lex_casemods (PL_parser->lex_casemods)
54#define PL_lex_casestack (PL_parser->lex_casestack)
199e78b7 55#define PL_lex_dojoin (PL_parser->lex_dojoin)
199e78b7
DM
56#define PL_lex_formbrack (PL_parser->lex_formbrack)
57#define PL_lex_inpat (PL_parser->lex_inpat)
58#define PL_lex_inwhat (PL_parser->lex_inwhat)
59#define PL_lex_op (PL_parser->lex_op)
60#define PL_lex_repl (PL_parser->lex_repl)
61#define PL_lex_starts (PL_parser->lex_starts)
62#define PL_lex_stuff (PL_parser->lex_stuff)
63#define PL_multi_start (PL_parser->multi_start)
64#define PL_multi_open (PL_parser->multi_open)
65#define PL_multi_close (PL_parser->multi_close)
199e78b7 66#define PL_preambled (PL_parser->preambled)
bdc0bf6f 67#define PL_linestr (PL_parser->linestr)
c2598295
DM
68#define PL_expect (PL_parser->expect)
69#define PL_copline (PL_parser->copline)
f06b5848
DM
70#define PL_bufptr (PL_parser->bufptr)
71#define PL_oldbufptr (PL_parser->oldbufptr)
72#define PL_oldoldbufptr (PL_parser->oldoldbufptr)
73#define PL_linestart (PL_parser->linestart)
74#define PL_bufend (PL_parser->bufend)
75#define PL_last_uni (PL_parser->last_uni)
76#define PL_last_lop (PL_parser->last_lop)
77#define PL_last_lop_op (PL_parser->last_lop_op)
bc177e6b 78#define PL_lex_state (PL_parser->lex_state)
2f9285f8 79#define PL_rsfp (PL_parser->rsfp)
5486870f 80#define PL_rsfp_filters (PL_parser->rsfp_filters)
12bd6ede
DM
81#define PL_in_my (PL_parser->in_my)
82#define PL_in_my_stash (PL_parser->in_my_stash)
14047fc9 83#define PL_tokenbuf (PL_parser->tokenbuf)
670a9cb2 84#define PL_multi_end (PL_parser->multi_end)
13765c85 85#define PL_error_count (PL_parser->error_count)
199e78b7 86
fb205e7a
DM
87# define PL_nexttoke (PL_parser->nexttoke)
88# define PL_nexttype (PL_parser->nexttype)
89# define PL_nextval (PL_parser->nextval)
199e78b7 90
6432a58a
DM
91
92#define SvEVALED(sv) \
93 (SvTYPE(sv) >= SVt_PVNV \
94 && ((XPVIV*)SvANY(sv))->xiv_u.xivu_eval_seen)
95
a1894d81 96static const char* const ident_too_long = "Identifier too long";
8903cb82 97
9ded7720 98# define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
29595ff2 99
a7aaec61
Z
100#define XENUMMASK 0x3f
101#define XFAKEEOF 0x40
102#define XFAKEBRACK 0x80
9059aa12 103
39e02b42 104#ifdef USE_UTF8_SCRIPTS
b3041197 105# define UTF cBOOL(!IN_BYTES)
2b9d42f0 106#else
b3041197 107# define UTF cBOOL((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
2b9d42f0 108#endif
a0ed51b3 109
b1fc3636
CJ
110/* The maximum number of characters preceding the unrecognized one to display */
111#define UNRECOGNIZED_PRECEDE_COUNT 10
112
61f0cdd9 113/* In variables named $^X, these are the legal values for X.
2b92dfce
GS
114 * 1999-02-27 mjd-perl-patch@plover.com */
115#define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
116
14bd96d0 117#define SPACE_OR_TAB(c) isBLANK_A(c)
bf4acbe4 118
9ff909cf
JH
119#define HEXFP_PEEK(s) \
120 (((s[0] == '.') && \
121 (isXDIGIT(s[1]) || isALPHA_FOLD_EQ(s[1], 'p'))) || \
122 isALPHA_FOLD_EQ(s[0], 'p'))
123
ffb4593c
NT
124/* LEX_* are values for PL_lex_state, the state of the lexer.
125 * They are arranged oddly so that the guard on the switch statement
79072805 126 * can get by with a single comparison (if the compiler is smart enough).
9da1dd8f
DM
127 *
128 * These values refer to the various states within a sublex parse,
129 * i.e. within a double quotish string
79072805
LW
130 */
131
fb73857a
PP
132/* #define LEX_NOTPARSING 11 is done in perl.h. */
133
b6007c36
DM
134#define LEX_NORMAL 10 /* normal code (ie not within "...") */
135#define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
136#define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
137#define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
138#define LEX_INTERPSTART 6 /* expecting the start of a $var */
139
140 /* at end of code, eg "$x" followed by: */
141#define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
142#define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
143
144#define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
145 string or after \E, $foo, etc */
146#define LEX_INTERPCONST 2 /* NOT USED */
147#define LEX_FORMLINE 1 /* expecting a format line */
b6007c36 148
79072805 149
bbf60fe6 150#ifdef DEBUGGING
27da23d5 151static const char* const lex_state_names[] = {
bbf60fe6
DM
152 "KNOWNEXT",
153 "FORMLINE",
154 "INTERPCONST",
155 "INTERPCONCAT",
156 "INTERPENDMAYBE",
157 "INTERPEND",
158 "INTERPSTART",
159 "INTERPPUSH",
160 "INTERPCASEMOD",
161 "INTERPNORMAL",
162 "NORMAL"
163};
164#endif
165
79072805 166#include "keywords.h"
fe14fcc3 167
ffb4593c
NT
168/* CLINE is a macro that ensures PL_copline has a sane value */
169
57843af0 170#define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
3280af22 171
ffb4593c
NT
172/*
173 * Convenience functions to return different tokens and prime the
9cbb5ea2 174 * lexer for the next token. They all take an argument.
ffb4593c
NT
175 *
176 * TOKEN : generic token (used for '(', DOLSHARP, etc)
177 * OPERATOR : generic operator
178 * AOPERATOR : assignment operator
179 * PREBLOCK : beginning the block after an if, while, foreach, ...
180 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
181 * PREREF : *EXPR where EXPR is not a simple identifier
182 * TERM : expression term
89f35911 183 * POSTDEREF : postfix dereference (->$* ->@[...] etc.)
ffb4593c
NT
184 * LOOPX : loop exiting command (goto, last, dump, etc)
185 * FTST : file test operator
186 * FUN0 : zero-argument function
7eb971ee 187 * FUN0OP : zero-argument function, with its op created in this file
2d2e263d 188 * FUN1 : not used, except for not, which isn't a UNIOP
ffb4593c
NT
189 * BOop : bitwise or or xor
190 * BAop : bitwise and
8823cb89 191 * BCop : bitwise complement
ffb4593c
NT
192 * SHop : shift operator
193 * PWop : power operator
9cbb5ea2 194 * PMop : pattern-matching operator
ffb4593c 195 * Aop : addition-level operator
e4916dd1 196 * AopNOASSIGN : addition-level operator that is never part of .=
ffb4593c
NT
197 * Mop : multiplication-level operator
198 * Eop : equality-testing operator
e5edeb50 199 * Rop : relational operator <= != gt
ffb4593c
NT
200 *
201 * Also see LOP and lop() below.
202 */
203
998054bd 204#ifdef DEBUGGING /* Serve -DT. */
704d4215 205# define REPORT(retval) tokereport((I32)retval, &pl_yylval)
998054bd 206#else
bbf60fe6 207# define REPORT(retval) (retval)
998054bd
SC
208#endif
209
bbf60fe6
DM
210#define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
211#define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
b1764551 212#define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, retval))
bbf60fe6
DM
213#define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
214#define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
215#define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
216#define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
89f35911 217#define POSTDEREF(f) return (PL_bufptr = s, S_postderef(aTHX_ REPORT(f),s[1]))
185c2e96 218#define LOOPX(f) return (PL_bufptr = force_word(s,BAREWORD,TRUE,FALSE), \
7a61bf3c 219 pl_yylval.ival=f, \
a49203fd 220 PL_expect = PL_nexttoke ? XOPERATOR : XTERM, \
7a61bf3c 221 REPORT((int)LOOPEX))
6154021b
RGS
222#define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
223#define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
7eb971ee 224#define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
6154021b 225#define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
b1764551
FC
226#define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITOROP))
227#define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)BITANDOP))
8823cb89
FC
228#define BCop(f) return pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr = s, \
229 REPORT('~')
b1764551
FC
230#define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)SHIFTOP))
231#define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)POWOP))
6154021b 232#define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
b1764551 233#define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)ADDOP))
e4916dd1 234#define AopNOASSIGN(f) return (pl_yylval.ival=f, PL_bufptr=s, REPORT((int)ADDOP))
b1764551 235#define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, (int)MULOP))
6154021b
RGS
236#define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
237#define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
2f3197b3 238
a687059c
LW
239/* This bit of chicanery makes a unary function followed by
240 * a parenthesis into a function with one argument, highest precedence.
6f33ba73
RGS
241 * The UNIDOR macro is for unary functions that can be followed by the //
242 * operator (such as C<shift // 0>).
a687059c 243 */
d68ce4ac 244#define UNI3(f,x,have_x) { \
6154021b 245 pl_yylval.ival = f; \
d68ce4ac 246 if (have_x) PL_expect = x; \
376fcdbf
AL
247 PL_bufptr = s; \
248 PL_last_uni = PL_oldbufptr; \
0af40c75 249 PL_last_lop_op = (f) < 0 ? -(f) : (f); \
376fcdbf
AL
250 if (*s == '(') \
251 return REPORT( (int)FUNC1 ); \
294a536f 252 s = skipspace(s); \
376fcdbf
AL
253 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
254 }
d68ce4ac
FC
255#define UNI(f) UNI3(f,XTERM,1)
256#define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
b5fb7ce3
FC
257#define UNIPROTO(f,optional) { \
258 if (optional) PL_last_uni = PL_oldbufptr; \
22393538
MH
259 OPERATOR(f); \
260 }
a687059c 261
d68ce4ac 262#define UNIBRACK(f) UNI3(f,0,0)
79072805 263
9f68db38 264/* grandfather return to old style */
78cdf107
Z
265#define OLDLOP(f) \
266 do { \
267 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
268 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
269 pl_yylval.ival = (f); \
270 PL_expect = XTERM; \
271 PL_bufptr = s; \
272 return (int)LSTOP; \
273 } while(0)
79072805 274
83944c01
FC
275#define COPLINE_INC_WITH_HERELINES \
276 STMT_START { \
277 CopLINE_inc(PL_curcop); \
851b527a
FC
278 if (PL_parser->herelines) \
279 CopLINE(PL_curcop) += PL_parser->herelines, \
280 PL_parser->herelines = 0; \
83944c01 281 } STMT_END
ffdb8b16
FC
282/* Called after scan_str to update CopLINE(PL_curcop), but only when there
283 * is no sublex_push to follow. */
284#define COPLINE_SET_FROM_MULTI_END \
285 STMT_START { \
286 CopLINE_set(PL_curcop, PL_multi_end); \
287 if (PL_multi_end != PL_multi_start) \
851b527a 288 PL_parser->herelines = 0; \
ffdb8b16 289 } STMT_END
83944c01
FC
290
291
8fa7f367
JH
292#ifdef DEBUGGING
293
6154021b 294/* how to interpret the pl_yylval associated with the token */
bbf60fe6
DM
295enum token_type {
296 TOKENTYPE_NONE,
297 TOKENTYPE_IVAL,
6154021b 298 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
bbf60fe6 299 TOKENTYPE_PVAL,
aeaef349 300 TOKENTYPE_OPVAL
bbf60fe6
DM
301};
302
6d4a66ac
NC
303static struct debug_tokens {
304 const int token;
305 enum token_type type;
306 const char *name;
307} const debug_tokens[] =
9041c2e3 308{
bbf60fe6
DM
309 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
310 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
311 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
312 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
313 { ARROW, TOKENTYPE_NONE, "ARROW" },
314 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
315 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
316 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
317 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
318 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
0d863452 319 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
bbf60fe6
DM
320 { DO, TOKENTYPE_NONE, "DO" },
321 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
322 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
323 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
324 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
325 { ELSE, TOKENTYPE_NONE, "ELSE" },
326 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
327 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
328 { FOR, TOKENTYPE_IVAL, "FOR" },
329 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
705fe0e5
FC
330 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
331 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
bbf60fe6
DM
332 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
333 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
7eb971ee 334 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
bbf60fe6
DM
335 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
336 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
337 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
0d863452 338 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
bbf60fe6
DM
339 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
340 { IF, TOKENTYPE_IVAL, "IF" },
5504e6cf 341 { LABEL, TOKENTYPE_PVAL, "LABEL" },
bbf60fe6
DM
342 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
343 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
344 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
345 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
346 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
347 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
348 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
349 { MY, TOKENTYPE_IVAL, "MY" },
bbf60fe6
DM
350 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
351 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
352 { OROP, TOKENTYPE_IVAL, "OROP" },
353 { OROR, TOKENTYPE_NONE, "OROR" },
354 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
88e1f1a2
JV
355 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
356 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
bbf60fe6 357 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
cc624add 358 { POSTJOIN, TOKENTYPE_NONE, "POSTJOIN" },
bbf60fe6
DM
359 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
360 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
361 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
362 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
363 { PREINC, TOKENTYPE_NONE, "PREINC" },
364 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
f3f204dc 365 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
bbf60fe6
DM
366 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
367 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
f3f204dc 368 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
bbf60fe6
DM
369 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
370 { SUB, TOKENTYPE_NONE, "SUB" },
371 { THING, TOKENTYPE_OPVAL, "THING" },
372 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
373 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
374 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
375 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
376 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
377 { USE, TOKENTYPE_IVAL, "USE" },
0d863452 378 { WHEN, TOKENTYPE_IVAL, "WHEN" },
bbf60fe6 379 { WHILE, TOKENTYPE_IVAL, "WHILE" },
185c2e96 380 { BAREWORD, TOKENTYPE_OPVAL, "BAREWORD" },
be25f609 381 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
c35e046a 382 { 0, TOKENTYPE_NONE, NULL }
bbf60fe6
DM
383};
384
6154021b 385/* dump the returned token in rv, plus any optional arg in pl_yylval */
998054bd 386
bbf60fe6 387STATIC int
704d4215 388S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
bbf60fe6 389{
7918f24d
NC
390 PERL_ARGS_ASSERT_TOKEREPORT;
391
bbf60fe6 392 if (DEBUG_T_TEST) {
bd61b366 393 const char *name = NULL;
bbf60fe6 394 enum token_type type = TOKENTYPE_NONE;
f54cb97a 395 const struct debug_tokens *p;
396482e1 396 SV* const report = newSVpvs("<== ");
bbf60fe6 397
f54cb97a 398 for (p = debug_tokens; p->token; p++) {
bbf60fe6
DM
399 if (p->token == (int)rv) {
400 name = p->name;
401 type = p->type;
402 break;
403 }
404 }
405 if (name)
54667de8 406 Perl_sv_catpv(aTHX_ report, name);
239f83d5 407 else if (isGRAPH(rv))
4ebc7986 408 {
bbf60fe6 409 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
4ebc7986
FC
410 if ((char)rv == 'p')
411 sv_catpvs(report, " (pending identifier)");
412 }
bbf60fe6 413 else if (!rv)
396482e1 414 sv_catpvs(report, "EOF");
bbf60fe6 415 else
147e3846 416 Perl_sv_catpvf(aTHX_ report, "?? %" IVdf, (IV)rv);
bbf60fe6
DM
417 switch (type) {
418 case TOKENTYPE_NONE:
bbf60fe6
DM
419 break;
420 case TOKENTYPE_IVAL:
147e3846 421 Perl_sv_catpvf(aTHX_ report, "(ival=%" IVdf ")", (IV)lvalp->ival);
bbf60fe6
DM
422 break;
423 case TOKENTYPE_OPNUM:
424 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
704d4215 425 PL_op_name[lvalp->ival]);
bbf60fe6
DM
426 break;
427 case TOKENTYPE_PVAL:
704d4215 428 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
bbf60fe6
DM
429 break;
430 case TOKENTYPE_OPVAL:
704d4215 431 if (lvalp->opval) {
401441c0 432 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
704d4215
GG
433 PL_op_name[lvalp->opval->op_type]);
434 if (lvalp->opval->op_type == OP_CONST) {
b6007c36 435 Perl_sv_catpvf(aTHX_ report, " %s",
704d4215 436 SvPEEK(cSVOPx_sv(lvalp->opval)));
b6007c36
DM
437 }
438
439 }
401441c0 440 else
396482e1 441 sv_catpvs(report, "(opval=null)");
bbf60fe6
DM
442 break;
443 }
b6007c36 444 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
bbf60fe6
DM
445 };
446 return (int)rv;
998054bd
SC
447}
448
b6007c36
DM
449
450/* print the buffer with suitable escapes */
451
452STATIC void
15f169a1 453S_printbuf(pTHX_ const char *const fmt, const char *const s)
b6007c36 454{
396482e1 455 SV* const tmp = newSVpvs("");
7918f24d
NC
456
457 PERL_ARGS_ASSERT_PRINTBUF;
458
5d37acd6 459 GCC_DIAG_IGNORE(-Wformat-nonliteral); /* fmt checked by caller */
b6007c36 460 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
5d37acd6 461 GCC_DIAG_RESTORE;
b6007c36
DM
462 SvREFCNT_dec(tmp);
463}
464
8fa7f367
JH
465#endif
466
ffb4593c
NT
467/*
468 * S_ao
469 *
f393a21a
FC
470 * This subroutine looks for an '=' next to the operator that has just been
471 * parsed and turns it into an ASSIGNOP if it finds one.
ffb4593c
NT
472 */
473
76e3520e 474STATIC int
cea2e8a9 475S_ao(pTHX_ int toketype)
a0d0e21e 476{
3280af22
NIS
477 if (*PL_bufptr == '=') {
478 PL_bufptr++;
a0d0e21e 479 if (toketype == ANDAND)
6154021b 480 pl_yylval.ival = OP_ANDASSIGN;
a0d0e21e 481 else if (toketype == OROR)
6154021b 482 pl_yylval.ival = OP_ORASSIGN;
c963b151 483 else if (toketype == DORDOR)
6154021b 484 pl_yylval.ival = OP_DORASSIGN;
a0d0e21e
LW
485 toketype = ASSIGNOP;
486 }
b1764551 487 return REPORT(toketype);
a0d0e21e
LW
488}
489
ffb4593c
NT
490/*
491 * S_no_op
492 * When Perl expects an operator and finds something else, no_op
493 * prints the warning. It always prints "<something> found where
494 * operator expected. It prints "Missing semicolon on previous line?"
495 * if the surprise occurs at the start of the line. "do you need to
496 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
497 * where the compiler doesn't know if foo is a method call or a function.
498 * It prints "Missing operator before end of line" if there's nothing
499 * after the missing operator, or "... before <...>" if there is something
500 * after the missing operator.
488bc579
FC
501 *
502 * PL_bufptr is expected to point to the start of the thing that was found,
503 * and s after the next token or partial token.
ffb4593c
NT
504 */
505
76e3520e 506STATIC void
15f169a1 507S_no_op(pTHX_ const char *const what, char *s)
463ee0b2 508{
9d4ba2ae
AL
509 char * const oldbp = PL_bufptr;
510 const bool is_first = (PL_oldbufptr == PL_linestart);
68dc0745 511
7918f24d
NC
512 PERL_ARGS_ASSERT_NO_OP;
513
1189a94a
GS
514 if (!s)
515 s = oldbp;
07c798fb 516 else
1189a94a 517 PL_bufptr = s;
734ab321 518 yywarn(Perl_form(aTHX_ "%s found where operator expected", what), UTF ? SVf_UTF8 : 0);
56da5a46
RGS
519 if (ckWARN_d(WARN_SYNTAX)) {
520 if (is_first)
521 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
522 "\t(Missing semicolon on previous line?)\n");
fac0f7a3
KW
523 else if (PL_oldoldbufptr && isIDFIRST_lazy_if_safe(PL_oldoldbufptr,
524 PL_bufend,
525 UTF))
526 {
f54cb97a 527 const char *t;
fac0f7a3
KW
528 for (t = PL_oldoldbufptr;
529 (isWORDCHAR_lazy_if_safe(t, PL_bufend, UTF) || *t == ':');
530 t += UTF ? UTF8SKIP(t) : 1)
531 {
c35e046a 532 NOOP;
fac0f7a3 533 }
56da5a46
RGS
534 if (t < PL_bufptr && isSPACE(*t))
535 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 536 "\t(Do you need to predeclare %" UTF8f "?)\n",
b17a0679 537 UTF8fARG(UTF, t - PL_oldoldbufptr, PL_oldoldbufptr));
56da5a46
RGS
538 }
539 else {
540 assert(s >= oldbp);
541 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
147e3846 542 "\t(Missing operator before %" UTF8f "?)\n",
b17a0679 543 UTF8fARG(UTF, s - oldbp, oldbp));
56da5a46 544 }
07c798fb 545 }
3280af22 546 PL_bufptr = oldbp;
8990e307
LW
547}
548
ffb4593c
NT
549/*
550 * S_missingterm
551 * Complain about missing quote/regexp/heredoc terminator.
d4c19fe8 552 * If it's called with NULL then it cauterizes the line buffer.
ffb4593c
NT
553 * If we're in a delimited string and the delimiter is a control
554 * character, it's reformatted into a two-char sequence like ^C.
555 * This is fatal.
556 */
557
76e3520e 558STATIC void
cea2e8a9 559S_missingterm(pTHX_ char *s)
8990e307 560{
e487ff5e 561 char tmpbuf[UTF8_MAXBYTES + 1];
8990e307 562 char q;
cb650135
FC
563 bool uni = FALSE;
564 SV *sv;
8990e307 565 if (s) {
9d4ba2ae 566 char * const nl = strrchr(s,'\n');
d2719217 567 if (nl)
8990e307 568 *nl = '\0';
7f7f7d08 569 uni = UTF;
8990e307 570 }
cb650135 571 else if (PL_multi_close < 32) {
8990e307 572 *tmpbuf = '^';
585ec06d 573 tmpbuf[1] = (char)toCTRL(PL_multi_close);
8990e307
LW
574 tmpbuf[2] = '\0';
575 s = tmpbuf;
576 }
577 else {
cb650135
FC
578 if (LIKELY(PL_multi_close < 256)) {
579 *tmpbuf = (char)PL_multi_close;
580 tmpbuf[1] = '\0';
581 }
582 else {
583 uni = TRUE;
584 *uvchr_to_utf8((U8 *)tmpbuf, PL_multi_close) = 0;
585 }
8990e307
LW
586 s = tmpbuf;
587 }
588 q = strchr(s,'"') ? '\'' : '"';
cb650135
FC
589 sv = sv_2mortal(newSVpv(s,0));
590 if (uni)
591 SvUTF8_on(sv);
147e3846 592 Perl_croak(aTHX_ "Can't find string terminator %c%" SVf
cb650135 593 "%c anywhere before EOF",q,SVfARG(sv),q);
463ee0b2 594}
79072805 595
dd0ac2b9
FC
596#include "feature.h"
597
0d863452 598/*
0d863452
RH
599 * Check whether the named feature is enabled.
600 */
26ea9e12 601bool
3fff3427 602Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
0d863452 603{
4a731d7b 604 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
7918f24d
NC
605
606 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
ca4d40c4
FC
607
608 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
7918f24d 609
26ea9e12
NC
610 if (namelen > MAX_FEATURE_LEN)
611 return FALSE;
3fff3427 612 memcpy(&he_name[8], name, namelen);
7d69d4a6 613
c8ca97b0
NC
614 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
615 REFCOUNTED_HE_EXISTS));
0d863452
RH
616}
617
ffb4593c 618/*
9cbb5ea2
GS
619 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
620 * utf16-to-utf8-reversed.
ffb4593c
NT
621 */
622
c39cd008
GS
623#ifdef PERL_CR_FILTER
624static void
625strip_return(SV *sv)
626{
eb578fdb
KW
627 const char *s = SvPVX_const(sv);
628 const char * const e = s + SvCUR(sv);
7918f24d
NC
629
630 PERL_ARGS_ASSERT_STRIP_RETURN;
631
c39cd008
GS
632 /* outer loop optimized to do nothing if there are no CR-LFs */
633 while (s < e) {
634 if (*s++ == '\r' && *s == '\n') {
635 /* hit a CR-LF, need to copy the rest */
eb578fdb 636 char *d = s - 1;
c39cd008
GS
637 *d++ = *s++;
638 while (s < e) {
639 if (*s == '\r' && s[1] == '\n')
640 s++;
641 *d++ = *s++;
642 }
643 SvCUR(sv) -= s - d;
644 return;
645 }
646 }
647}
a868473f 648
76e3520e 649STATIC I32
c39cd008 650S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
a868473f 651{
f54cb97a 652 const I32 count = FILTER_READ(idx+1, sv, maxlen);
c39cd008
GS
653 if (count > 0 && !maxlen)
654 strip_return(sv);
655 return count;
a868473f
NIS
656}
657#endif
658
ffb4593c 659/*
8eaa0acf
Z
660=for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
661
662Creates and initialises a new lexer/parser state object, supplying
663a context in which to lex and parse from a new source of Perl code.
664A pointer to the new state object is placed in L</PL_parser>. An entry
c83d5090 665is made on the save stack so that upon unwinding, the new state object
8eaa0acf
Z
666will be destroyed and the former value of L</PL_parser> will be restored.
667Nothing else need be done to clean up the parsing context.
668
2d7f6611 669The code to be parsed comes from C<line> and C<rsfp>. C<line>, if
8eaa0acf 670non-null, provides a string (in SV form) containing code to be parsed.
2d7f6611
KW
671A copy of the string is made, so subsequent modification of C<line>
672does not affect parsing. C<rsfp>, if non-null, provides an input stream
8eaa0acf 673from which code will be read to be parsed. If both are non-null, the
2d7f6611
KW
674code in C<line> comes first and must consist of complete lines of input,
675and C<rsfp> supplies the remainder of the source.
8eaa0acf 676
2d7f6611 677The C<flags> parameter is reserved for future use. Currently it is only
e368b3bd 678used by perl internally, so extensions should always pass zero.
8eaa0acf
Z
679
680=cut
681*/
ffb4593c 682
27fcb6ee 683/* LEX_START_SAME_FILTER indicates that this is not a new file, so it
87606032
NC
684 can share filters with the current parser.
685 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
686 caller, hence isn't owned by the parser, so shouldn't be closed on parser
687 destruction. This is used to handle the case of defaulting to reading the
688 script from the standard input because no filename was given on the command
689 line (without getting confused by situation where STDIN has been closed, so
690 the script handle is opened on fd 0) */
27fcb6ee 691
a0d0e21e 692void
8eaa0acf 693Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
79072805 694{
6ef55633 695 const char *s = NULL;
5486870f 696 yy_parser *parser, *oparser;
90b58c70 697
60d63348 698 if (flags && flags & ~LEX_START_FLAGS)
8eaa0acf 699 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
acdf0a21
DM
700
701 /* create and initialise a parser */
702
199e78b7 703 Newxz(parser, 1, yy_parser);
5486870f 704 parser->old_parser = oparser = PL_parser;
acdf0a21
DM
705 PL_parser = parser;
706
28ac2b49 707 parser->stack = NULL;
df13534a 708 parser->stack_max1 = NULL;
28ac2b49 709 parser->ps = NULL;
acdf0a21 710
e3abe207
DM
711 /* on scope exit, free this parser and restore any outer one */
712 SAVEPARSER(parser);
7c4baf47 713 parser->saved_curcop = PL_curcop;
e3abe207 714
acdf0a21 715 /* initialise lexer state */
8990e307 716
fb205e7a 717 parser->nexttoke = 0;
ca4cfd28 718 parser->error_count = oparser ? oparser->error_count : 0;
7f1c3e8c 719 parser->copline = parser->preambling = NOLINE;
5afb0a62 720 parser->lex_state = LEX_NORMAL;
c2598295 721 parser->expect = XSTATE;
2f9285f8 722 parser->rsfp = rsfp;
efa571ab 723 parser->recheck_utf8_validity = FALSE;
27fcb6ee
FC
724 parser->rsfp_filters =
725 !(flags & LEX_START_SAME_FILTER) || !oparser
d3cd8e11
FC
726 ? NULL
727 : MUTABLE_AV(SvREFCNT_inc(
728 oparser->rsfp_filters
729 ? oparser->rsfp_filters
730 : (oparser->rsfp_filters = newAV())
731 ));
2f9285f8 732
199e78b7
DM
733 Newx(parser->lex_brackstack, 120, char);
734 Newx(parser->lex_casestack, 12, char);
735 *parser->lex_casestack = '\0';
d794b522 736 Newxz(parser->lex_shared, 1, LEXSHARED);
02b34bbe 737
10efb74f 738 if (line) {
0528fd32 739 STRLEN len;
a6909bd2
KW
740 const U8* first_bad_char_loc;
741
10efb74f 742 s = SvPV_const(line, len);
90b58c70 743
0aab20f2
KW
744 if ( SvUTF8(line)
745 && UNLIKELY(! is_utf8_string_loc((U8 *) s,
746 SvCUR(line),
747 &first_bad_char_loc)))
90b58c70
KW
748 {
749 _force_out_malformed_utf8_message(first_bad_char_loc,
750 (U8 *) s + SvCUR(line),
751 0,
752 1 /* 1 means die */ );
753 NOT_REACHED; /* NOTREACHED */
754 }
755
0abcdfa4
FC
756 parser->linestr = flags & LEX_START_COPIED
757 ? SvREFCNT_inc_simple_NN(line)
758 : newSVpvn_flags(s, len, SvUTF8(line));
b3dd0aba
FC
759 if (!rsfp)
760 sv_catpvs(parser->linestr, "\n;");
0abcdfa4 761 } else {
bf1b738b 762 parser->linestr = newSVpvn("\n;", rsfp ? 1 : 2);
8990e307 763 }
218304f9 764
f06b5848
DM
765 parser->oldoldbufptr =
766 parser->oldbufptr =
767 parser->bufptr =
768 parser->linestart = SvPVX(parser->linestr);
769 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
770 parser->last_lop = parser->last_uni = NULL;
b54f893d 771
6d59e610 772 STATIC_ASSERT_STMT(FITS_IN_8_BITS(LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
b54f893d
KW
773 |LEX_DONT_CLOSE_RSFP));
774 parser->lex_flags = (U8) (flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
775 |LEX_DONT_CLOSE_RSFP));
737c24fc 776
60d63348 777 parser->in_pod = parser->filtered = 0;
79072805 778}
a687059c 779
e3abe207
DM
780
781/* delete a parser object */
782
783void
784Perl_parser_free(pTHX_ const yy_parser *parser)
785{
7918f24d
NC
786 PERL_ARGS_ASSERT_PARSER_FREE;
787
7c4baf47 788 PL_curcop = parser->saved_curcop;
bdc0bf6f
DM
789 SvREFCNT_dec(parser->linestr);
790
87606032 791 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
2f9285f8 792 PerlIO_clearerr(parser->rsfp);
407f8cf2
KW
793 else if (parser->rsfp && (!parser->old_parser
794 || (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
2f9285f8 795 PerlIO_close(parser->rsfp);
5486870f 796 SvREFCNT_dec(parser->rsfp_filters);
10002bc1 797 SvREFCNT_dec(parser->lex_stuff);
7ef70b3d 798 SvREFCNT_dec(parser->lex_sub_repl);
3ac7ff8f
FC
799
800 Safefree(parser->lex_brackstack);
801 Safefree(parser->lex_casestack);
802 Safefree(parser->lex_shared);
803 PL_parser = parser->old_parser;
804 Safefree(parser);
805}
806
807void
808Perl_parser_free_nexttoke_ops(pTHX_ yy_parser *parser, OPSLAB *slab)
809{
3ac7ff8f 810 I32 nexttoke = parser->nexttoke;
3ac7ff8f 811 PERL_ARGS_ASSERT_PARSER_FREE_NEXTTOKE_OPS;
3ce3dcd9 812 while (nexttoke--) {
3ac7ff8f
FC
813 if (S_is_opval_token(parser->nexttype[nexttoke] & 0xffff)
814 && parser->nextval[nexttoke].opval
815 && parser->nextval[nexttoke].opval->op_slabbed
816 && OpSLAB(parser->nextval[nexttoke].opval) == slab) {
3ce3dcd9 817 op_free(parser->nextval[nexttoke].opval);
3ac7ff8f
FC
818 parser->nextval[nexttoke].opval = NULL;
819 }
3ce3dcd9 820 }
e3abe207
DM
821}
822
823
ffb4593c 824/*
f0e67a1d
Z
825=for apidoc AmxU|SV *|PL_parser-E<gt>linestr
826
827Buffer scalar containing the chunk currently under consideration of the
828text currently being lexed. This is always a plain string scalar (for
829which C<SvPOK> is true). It is not intended to be used as a scalar by
830normal scalar means; instead refer to the buffer directly by the pointer
831variables described below.
832
833The lexer maintains various C<char*> pointers to things in the
834C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
835reallocated, all of these pointers must be updated. Don't attempt to
836do this manually, but rather use L</lex_grow_linestr> if you need to
837reallocate the buffer.
838
839The content of the text chunk in the buffer is commonly exactly one
840complete line of input, up to and including a newline terminator,
841but there are situations where it is otherwise. The octets of the
842buffer may be intended to be interpreted as either UTF-8 or Latin-1.
843The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
844flag on this scalar, which may disagree with it.
845
846For direct examination of the buffer, the variable
847L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
848lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
849of these pointers is usually preferable to examination of the scalar
850through normal scalar means.
851
852=for apidoc AmxU|char *|PL_parser-E<gt>bufend
853
854Direct pointer to the end of the chunk of text currently being lexed, the
855end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
6602b933 856+ SvCUR(PL_parser-E<gt>linestr)>. A C<NUL> character (zero octet) is
f0e67a1d
Z
857always located at the end of the buffer, and does not count as part of
858the buffer's contents.
859
860=for apidoc AmxU|char *|PL_parser-E<gt>bufptr
861
862Points to the current position of lexing inside the lexer buffer.
863Characters around this point may be freely examined, within
864the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
865L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
866interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
867
868Lexing code (whether in the Perl core or not) moves this pointer past
869the characters that it consumes. It is also expected to perform some
870bookkeeping whenever a newline character is consumed. This movement
871can be more conveniently performed by the function L</lex_read_to>,
872which handles newlines appropriately.
873
874Interpretation of the buffer's octets can be abstracted out by
875using the slightly higher-level functions L</lex_peek_unichar> and
876L</lex_read_unichar>.
877
878=for apidoc AmxU|char *|PL_parser-E<gt>linestart
879
880Points to the start of the current line inside the lexer buffer.
881This is useful for indicating at which column an error occurred, and
882not much else. This must be updated by any lexing code that consumes
883a newline; the function L</lex_read_to> handles this detail.
884
885=cut
886*/
887
888/*
889=for apidoc Amx|bool|lex_bufutf8
890
891Indicates whether the octets in the lexer buffer
892(L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
893of Unicode characters. If not, they should be interpreted as Latin-1
894characters. This is analogous to the C<SvUTF8> flag for scalars.
895
896In UTF-8 mode, it is not guaranteed that the lexer buffer actually
897contains valid UTF-8. Lexing code must be robust in the face of invalid
898encoding.
899
900The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
901is significant, but not the whole story regarding the input character
902encoding. Normally, when a file is being read, the scalar contains octets
903and its C<SvUTF8> flag is off, but the octets should be interpreted as
904UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
905however, the scalar may have the C<SvUTF8> flag on, and in this case its
906octets should be interpreted as UTF-8 unless the C<use bytes> pragma
907is in effect. This logic may change in the future; use this function
908instead of implementing the logic yourself.
909
910=cut
911*/
912
913bool
914Perl_lex_bufutf8(pTHX)
915{
916 return UTF;
917}
918
919/*
920=for apidoc Amx|char *|lex_grow_linestr|STRLEN len
921
922Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
2d7f6611 923at least C<len> octets (including terminating C<NUL>). Returns a
f0e67a1d
Z
924pointer to the reallocated buffer. This is necessary before making
925any direct modification of the buffer that would increase its length.
926L</lex_stuff_pvn> provides a more convenient way to insert text into
927the buffer.
928
929Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
930this function updates all of the lexer's variables that point directly
931into the buffer.
932
933=cut
934*/
935
936char *
937Perl_lex_grow_linestr(pTHX_ STRLEN len)
938{
939 SV *linestr;
940 char *buf;
941 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
c7641931 942 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
98d5e3ef
DM
943 bool current;
944
f0e67a1d
Z
945 linestr = PL_parser->linestr;
946 buf = SvPVX(linestr);
947 if (len <= SvLEN(linestr))
948 return buf;
98d5e3ef
DM
949
950 /* Is the lex_shared linestr SV the same as the current linestr SV?
951 * Only in this case does re_eval_start need adjusting, since it
952 * points within lex_shared->ls_linestr's buffer */
b1b8fb6a
DM
953 current = ( !PL_parser->lex_shared->ls_linestr
954 || linestr == PL_parser->lex_shared->ls_linestr);
98d5e3ef 955
f0e67a1d
Z
956 bufend_pos = PL_parser->bufend - buf;
957 bufptr_pos = PL_parser->bufptr - buf;
958 oldbufptr_pos = PL_parser->oldbufptr - buf;
959 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
960 linestart_pos = PL_parser->linestart - buf;
961 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
962 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
98d5e3ef 963 re_eval_start_pos = (current && PL_parser->lex_shared->re_eval_start) ?
3328ab5a 964 PL_parser->lex_shared->re_eval_start - buf : 0;
c7641931 965
f0e67a1d 966 buf = sv_grow(linestr, len);
c7641931 967
f0e67a1d
Z
968 PL_parser->bufend = buf + bufend_pos;
969 PL_parser->bufptr = buf + bufptr_pos;
970 PL_parser->oldbufptr = buf + oldbufptr_pos;
971 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
972 PL_parser->linestart = buf + linestart_pos;
973 if (PL_parser->last_uni)
974 PL_parser->last_uni = buf + last_uni_pos;
975 if (PL_parser->last_lop)
976 PL_parser->last_lop = buf + last_lop_pos;
98d5e3ef 977 if (current && PL_parser->lex_shared->re_eval_start)
3328ab5a 978 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
f0e67a1d
Z
979 return buf;
980}
981
982/*
83aa740e 983=for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
f0e67a1d
Z
984
985Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
986immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
987reallocating the buffer if necessary. This means that lexing code that
988runs later will see the characters as if they had appeared in the input.
989It is not recommended to do this as part of normal parsing, and most
990uses of this facility run the risk of the inserted characters being
991interpreted in an unintended manner.
992
2d7f6611
KW
993The string to be inserted is represented by C<len> octets starting
994at C<pv>. These octets are interpreted as either UTF-8 or Latin-1,
995according to whether the C<LEX_STUFF_UTF8> flag is set in C<flags>.
f0e67a1d
Z
996The characters are recoded for the lexer buffer, according to how the
997buffer is currently being interpreted (L</lex_bufutf8>). If a string
9dcc53ea 998to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
f0e67a1d
Z
999function is more convenient.
1000
1001=cut
1002*/
1003
1004void
83aa740e 1005Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
f0e67a1d 1006{
749123ff 1007 dVAR;
f0e67a1d
Z
1008 char *bufptr;
1009 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
1010 if (flags & ~(LEX_STUFF_UTF8))
1011 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
1012 if (UTF) {
1013 if (flags & LEX_STUFF_UTF8) {
1014 goto plain_copy;
1015 } else {
54d004e8 1016 STRLEN highhalf = 0; /* Count of variants */
83aa740e 1017 const char *p, *e = pv+len;
54d004e8
KW
1018 for (p = pv; p != e; p++) {
1019 if (! UTF8_IS_INVARIANT(*p)) {
1020 highhalf++;
1021 }
1022 }
f0e67a1d
Z
1023 if (!highhalf)
1024 goto plain_copy;
1025 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
1026 bufptr = PL_parser->bufptr;
1027 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1028 SvCUR_set(PL_parser->linestr,
1029 SvCUR(PL_parser->linestr) + len+highhalf);
f0e67a1d
Z
1030 PL_parser->bufend += len+highhalf;
1031 for (p = pv; p != e; p++) {
1032 U8 c = (U8)*p;
54d004e8
KW
1033 if (! UTF8_IS_INVARIANT(c)) {
1034 *bufptr++ = UTF8_TWO_BYTE_HI(c);
1035 *bufptr++ = UTF8_TWO_BYTE_LO(c);
f0e67a1d
Z
1036 } else {
1037 *bufptr++ = (char)c;
1038 }
1039 }
1040 }
1041 } else {
1042 if (flags & LEX_STUFF_UTF8) {
1043 STRLEN highhalf = 0;
83aa740e 1044 const char *p, *e = pv+len;
f0e67a1d
Z
1045 for (p = pv; p != e; p++) {
1046 U8 c = (U8)*p;
54d004e8 1047 if (UTF8_IS_ABOVE_LATIN1(c)) {
f0e67a1d
Z
1048 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1049 "non-Latin-1 character into Latin-1 input");
54d004e8 1050 } else if (UTF8_IS_NEXT_CHAR_DOWNGRADEABLE(p, e)) {
f0e67a1d
Z
1051 p++;
1052 highhalf++;
f88c6466 1053 } else assert(UTF8_IS_INVARIANT(c));
f0e67a1d
Z
1054 }
1055 if (!highhalf)
1056 goto plain_copy;
1057 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1058 bufptr = PL_parser->bufptr;
1059 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
255fdf19
Z
1060 SvCUR_set(PL_parser->linestr,
1061 SvCUR(PL_parser->linestr) + len-highhalf);
f0e67a1d 1062 PL_parser->bufend += len-highhalf;
54d004e8
KW
1063 p = pv;
1064 while (p < e) {
1065 if (UTF8_IS_INVARIANT(*p)) {
1066 *bufptr++ = *p;
1067 p++;
f0e67a1d 1068 }
54d004e8
KW
1069 else {
1070 assert(p < e -1 );
a62b247b 1071 *bufptr++ = EIGHT_BIT_UTF8_TO_NATIVE(*p, *(p+1));
54d004e8
KW
1072 p += 2;
1073 }
f0e67a1d
Z
1074 }
1075 } else {
54d004e8 1076 plain_copy:
f0e67a1d
Z
1077 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1078 bufptr = PL_parser->bufptr;
1079 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
255fdf19 1080 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
f0e67a1d
Z
1081 PL_parser->bufend += len;
1082 Copy(pv, bufptr, len, char);
1083 }
1084 }
1085}
1086
1087/*
9dcc53ea
Z
1088=for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1089
1090Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1091immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1092reallocating the buffer if necessary. This means that lexing code that
1093runs later will see the characters as if they had appeared in the input.
1094It is not recommended to do this as part of normal parsing, and most
1095uses of this facility run the risk of the inserted characters being
1096interpreted in an unintended manner.
1097
2d7f6611 1098The string to be inserted is represented by octets starting at C<pv>
9dcc53ea
Z
1099and continuing to the first nul. These octets are interpreted as either
1100UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
2d7f6611 1101in C<flags>. The characters are recoded for the lexer buffer, according
9dcc53ea
Z
1102to how the buffer is currently being interpreted (L</lex_bufutf8>).
1103If it is not convenient to nul-terminate a string to be inserted, the
1104L</lex_stuff_pvn> function is more appropriate.
1105
1106=cut
1107*/
1108
1109void
1110Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1111{
1112 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1113 lex_stuff_pvn(pv, strlen(pv), flags);
1114}
1115
1116/*
f0e67a1d
Z
1117=for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1118
1119Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1120immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1121reallocating the buffer if necessary. This means that lexing code that
1122runs later will see the characters as if they had appeared in the input.
1123It is not recommended to do this as part of normal parsing, and most
1124uses of this facility run the risk of the inserted characters being
1125interpreted in an unintended manner.
1126
2d7f6611 1127The string to be inserted is the string value of C<sv>. The characters
f0e67a1d 1128are recoded for the lexer buffer, according to how the buffer is currently
9dcc53ea 1129being interpreted (L</lex_bufutf8>). If a string to be inserted is
f0e67a1d
Z
1130not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1131need to construct a scalar.
1132
1133=cut
1134*/
1135
1136void
1137Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1138{
1139 char *pv;
1140 STRLEN len;
1141 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1142 if (flags)
1143 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1144 pv = SvPV(sv, len);
1145 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1146}
1147
1148/*
1149=for apidoc Amx|void|lex_unstuff|char *ptr
1150
1151Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
2d7f6611 1152C<ptr>. Text following C<ptr> will be moved, and the buffer shortened.
f0e67a1d
Z
1153This hides the discarded text from any lexing code that runs later,
1154as if the text had never appeared.
1155
1156This is not the normal way to consume lexed text. For that, use
1157L</lex_read_to>.
1158
1159=cut
1160*/
1161
1162void
1163Perl_lex_unstuff(pTHX_ char *ptr)
1164{
1165 char *buf, *bufend;
1166 STRLEN unstuff_len;
1167 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1168 buf = PL_parser->bufptr;
1169 if (ptr < buf)
1170 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1171 if (ptr == buf)
1172 return;
1173 bufend = PL_parser->bufend;
1174 if (ptr > bufend)
1175 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1176 unstuff_len = ptr - buf;
1177 Move(ptr, buf, bufend+1-ptr, char);
1178 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1179 PL_parser->bufend = bufend - unstuff_len;
1180}
1181
1182/*
1183=for apidoc Amx|void|lex_read_to|char *ptr
1184
1185Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
2d7f6611 1186to C<ptr>. This advances L</PL_parser-E<gt>bufptr> to match C<ptr>,
f0e67a1d
Z
1187performing the correct bookkeeping whenever a newline character is passed.
1188This is the normal way to consume lexed text.
1189
1190Interpretation of the buffer's octets can be abstracted out by
1191using the slightly higher-level functions L</lex_peek_unichar> and
1192L</lex_read_unichar>.
1193
1194=cut
1195*/
1196
1197void
1198Perl_lex_read_to(pTHX_ char *ptr)
1199{
1200 char *s;
1201 PERL_ARGS_ASSERT_LEX_READ_TO;
1202 s = PL_parser->bufptr;
1203 if (ptr < s || ptr > PL_parser->bufend)
1204 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1205 for (; s != ptr; s++)
1206 if (*s == '\n') {
83944c01 1207 COPLINE_INC_WITH_HERELINES;
f0e67a1d
Z
1208 PL_parser->linestart = s+1;
1209 }
1210 PL_parser->bufptr = ptr;
1211}
1212
1213/*
1214=for apidoc Amx|void|lex_discard_to|char *ptr
1215
1216Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
2d7f6611
KW
1217up to C<ptr>. The remaining content of the buffer will be moved, and
1218all pointers into the buffer updated appropriately. C<ptr> must not
f0e67a1d
Z
1219be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1220it is not permitted to discard text that has yet to be lexed.
1221
1222Normally it is not necessarily to do this directly, because it suffices to
1223use the implicit discarding behaviour of L</lex_next_chunk> and things
1224based on it. However, if a token stretches across multiple lines,
1f317c95 1225and the lexing code has kept multiple lines of text in the buffer for
f0e67a1d
Z
1226that purpose, then after completion of the token it would be wise to
1227explicitly discard the now-unneeded earlier lines, to avoid future
1228multi-line tokens growing the buffer without bound.
1229
1230=cut
1231*/
1232
1233void
1234Perl_lex_discard_to(pTHX_ char *ptr)
1235{
1236 char *buf;
1237 STRLEN discard_len;
1238 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1239 buf = SvPVX(PL_parser->linestr);
1240 if (ptr < buf)
1241 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1242 if (ptr == buf)
1243 return;
1244 if (ptr > PL_parser->bufptr)
1245 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1246 discard_len = ptr - buf;
1247 if (PL_parser->oldbufptr < ptr)
1248 PL_parser->oldbufptr = ptr;
1249 if (PL_parser->oldoldbufptr < ptr)
1250 PL_parser->oldoldbufptr = ptr;
1251 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1252 PL_parser->last_uni = NULL;
1253 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1254 PL_parser->last_lop = NULL;
1255 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1256 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1257 PL_parser->bufend -= discard_len;
1258 PL_parser->bufptr -= discard_len;
1259 PL_parser->oldbufptr -= discard_len;
1260 PL_parser->oldoldbufptr -= discard_len;
1261 if (PL_parser->last_uni)
1262 PL_parser->last_uni -= discard_len;
1263 if (PL_parser->last_lop)
1264 PL_parser->last_lop -= discard_len;
1265}
1266
efa571ab
KW
1267void
1268Perl_notify_parser_that_changed_to_utf8(pTHX)
1269{
1270 /* Called when $^H is changed to indicate that HINT_UTF8 has changed from
1271 * off to on. At compile time, this has the effect of entering a 'use
1272 * utf8' section. This means that any input was not previously checked for
1273 * UTF-8 (because it was off), but now we do need to check it, or our
1274 * assumptions about the input being sane could be wrong, and we could
1275 * segfault. This routine just sets a flag so that the next time we look
1276 * at the input we do the well-formed UTF-8 check. If we aren't in the
1277 * proper phase, there may not be a parser object, but if there is, setting
1278 * the flag is harmless */
1279
1280 if (PL_parser) {
1281 PL_parser->recheck_utf8_validity = TRUE;
1282 }
1283}
1284
f0e67a1d
Z
1285/*
1286=for apidoc Amx|bool|lex_next_chunk|U32 flags
1287
1288Reads in the next chunk of text to be lexed, appending it to
1289L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1290looked to the end of the current chunk and wants to know more. It is
1291usual, but not necessary, for lexing to have consumed the entirety of
1292the current chunk at this time.
1293
1294If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1295chunk (i.e., the current chunk has been entirely consumed), normally the
1296current chunk will be discarded at the same time that the new chunk is
c5608a1f 1297read in. If C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, the current chunk
f0e67a1d
Z
1298will not be discarded. If the current chunk has not been entirely
1299consumed, then it will not be discarded regardless of the flag.
1300
1301Returns true if some new text was added to the buffer, or false if the
1302buffer has reached the end of the input text.
1303
1304=cut
1305*/
1306
1307#define LEX_FAKE_EOF 0x80000000
e47d32dc 1308#define LEX_NO_TERM 0x40000000 /* here-doc */
f0e67a1d
Z
1309
1310bool
1311Perl_lex_next_chunk(pTHX_ U32 flags)
1312{
1313 SV *linestr;
1314 char *buf;
1315 STRLEN old_bufend_pos, new_bufend_pos;
1316 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1317 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
17cc9359 1318 bool got_some_for_debugger = 0;
f0e67a1d 1319 bool got_some;
6cdc5cd8 1320
112d1284 1321 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
f0e67a1d 1322 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
d27f4b91 1323 if (!(flags & LEX_NO_TERM) && PL_lex_inwhat)
e47d32dc 1324 return FALSE;
f0e67a1d
Z
1325 linestr = PL_parser->linestr;
1326 buf = SvPVX(linestr);
407f8cf2
KW
1327 if (!(flags & LEX_KEEP_PREVIOUS)
1328 && PL_parser->bufptr == PL_parser->bufend)
1329 {
f0e67a1d
Z
1330 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1331 linestart_pos = 0;
1332 if (PL_parser->last_uni != PL_parser->bufend)
1333 PL_parser->last_uni = NULL;
1334 if (PL_parser->last_lop != PL_parser->bufend)
1335 PL_parser->last_lop = NULL;
1336 last_uni_pos = last_lop_pos = 0;
1337 *buf = 0;
1338 SvCUR(linestr) = 0;
1339 } else {
1340 old_bufend_pos = PL_parser->bufend - buf;
1341 bufptr_pos = PL_parser->bufptr - buf;
1342 oldbufptr_pos = PL_parser->oldbufptr - buf;
1343 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1344 linestart_pos = PL_parser->linestart - buf;
1345 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1346 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1347 }
1348 if (flags & LEX_FAKE_EOF) {
1349 goto eof;
60d63348 1350 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
f0e67a1d
Z
1351 got_some = 0;
1352 } else if (filter_gets(linestr, old_bufend_pos)) {
1353 got_some = 1;
17cc9359 1354 got_some_for_debugger = 1;
112d1284
FC
1355 } else if (flags & LEX_NO_TERM) {
1356 got_some = 0;
f0e67a1d 1357 } else {
580561a3 1358 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
847cc851 1359 SvPVCLEAR(linestr);
f0e67a1d
Z
1360 eof:
1361 /* End of real input. Close filehandle (unless it was STDIN),
1362 * then add implicit termination.
1363 */
87606032 1364 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
f0e67a1d
Z
1365 PerlIO_clearerr(PL_parser->rsfp);
1366 else if (PL_parser->rsfp)
1367 (void)PerlIO_close(PL_parser->rsfp);
1368 PL_parser->rsfp = NULL;
60d63348 1369 PL_parser->in_pod = PL_parser->filtered = 0;
f0e67a1d
Z
1370 if (!PL_in_eval && PL_minus_p) {
1371 sv_catpvs(linestr,
1372 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1373 PL_minus_n = PL_minus_p = 0;
1374 } else if (!PL_in_eval && PL_minus_n) {
1375 sv_catpvs(linestr, /*{*/";}");
1376 PL_minus_n = 0;
1377 } else
1378 sv_catpvs(linestr, ";");
1379 got_some = 1;
1380 }
1381 buf = SvPVX(linestr);
1382 new_bufend_pos = SvCUR(linestr);
1383 PL_parser->bufend = buf + new_bufend_pos;
1384 PL_parser->bufptr = buf + bufptr_pos;
6cdc5cd8 1385
07337b95
KW
1386 if (UTF) {
1387 const U8* first_bad_char_loc;
1388 if (UNLIKELY(! is_utf8_string_loc(
1389 (U8 *) PL_parser->bufptr,
1390 PL_parser->bufend - PL_parser->bufptr,
1391 &first_bad_char_loc)))
1392 {
1393 _force_out_malformed_utf8_message(first_bad_char_loc,
1394 (U8 *) PL_parser->bufend,
1395 0,
1396 1 /* 1 means die */ );
1397 NOT_REACHED; /* NOTREACHED */
1398 }
6cdc5cd8
KW
1399 }
1400
f0e67a1d
Z
1401 PL_parser->oldbufptr = buf + oldbufptr_pos;
1402 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1403 PL_parser->linestart = buf + linestart_pos;
1404 if (PL_parser->last_uni)
1405 PL_parser->last_uni = buf + last_uni_pos;
1406 if (PL_parser->last_lop)
1407 PL_parser->last_lop = buf + last_lop_pos;
7f1c3e8c
FC
1408 if (PL_parser->preambling != NOLINE) {
1409 CopLINE_set(PL_curcop, PL_parser->preambling + 1);
1410 PL_parser->preambling = NOLINE;
1411 }
407f8cf2
KW
1412 if ( got_some_for_debugger
1413 && PERLDB_LINE_OR_SAVESRC
1414 && PL_curstash != PL_debstash)
1415 {
f0e67a1d
Z
1416 /* debugger active and we're not compiling the debugger code,
1417 * so store the line into the debugger's array of lines
1418 */
1419 update_debugger_info(NULL, buf+old_bufend_pos,
1420 new_bufend_pos-old_bufend_pos);
1421 }
1422 return got_some;
1423}
1424
1425/*
1426=for apidoc Amx|I32|lex_peek_unichar|U32 flags
1427
1428Looks ahead one (Unicode) character in the text currently being lexed.
1429Returns the codepoint (unsigned integer value) of the next character,
1430or -1 if lexing has reached the end of the input text. To consume the
1431peeked character, use L</lex_read_unichar>.
1432
1433If the next character is in (or extends into) the next chunk of input
1434text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1435discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1436bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1437
1438If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1439is encountered, an exception is generated.
1440
1441=cut
1442*/
1443
1444I32
1445Perl_lex_peek_unichar(pTHX_ U32 flags)
1446{
749123ff 1447 dVAR;
f0e67a1d
Z
1448 char *s, *bufend;
1449 if (flags & ~(LEX_KEEP_PREVIOUS))
1450 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1451 s = PL_parser->bufptr;
1452 bufend = PL_parser->bufend;
1453 if (UTF) {
1454 U8 head;
1455 I32 unichar;
1456 STRLEN len, retlen;
1457 if (s == bufend) {
1458 if (!lex_next_chunk(flags))
1459 return -1;
1460 s = PL_parser->bufptr;
1461 bufend = PL_parser->bufend;
1462 }
1463 head = (U8)*s;
54d004e8 1464 if (UTF8_IS_INVARIANT(head))
f0e67a1d 1465 return head;
54d004e8
KW
1466 if (UTF8_IS_START(head)) {
1467 len = UTF8SKIP(&head);
f0e67a1d
Z
1468 while ((STRLEN)(bufend-s) < len) {
1469 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1470 break;
1471 s = PL_parser->bufptr;
1472 bufend = PL_parser->bufend;
1473 }
1474 }
c80e42f3 1475 unichar = utf8n_to_uvchr((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
f0e67a1d 1476 if (retlen == (STRLEN)-1) {
75219bac
KW
1477 _force_out_malformed_utf8_message((U8 *) s,
1478 (U8 *) bufend,
1479 0,
1480 1 /* 1 means die */ );
1481 NOT_REACHED; /* NOTREACHED */
f0e67a1d
Z
1482 }
1483 return unichar;
1484 } else {
1485 if (s == bufend) {
1486 if (!lex_next_chunk(flags))
1487 return -1;
1488 s = PL_parser->bufptr;
1489 }
1490 return (U8)*s;
1491 }
1492}
1493
1494/*
1495=for apidoc Amx|I32|lex_read_unichar|U32 flags
1496
1497Reads the next (Unicode) character in the text currently being lexed.
1498Returns the codepoint (unsigned integer value) of the character read,
1499and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1500if lexing has reached the end of the input text. To non-destructively
1501examine the next character, use L</lex_peek_unichar> instead.
1502
1503If the next character is in (or extends into) the next chunk of input
1504text, the next chunk will be read in. Normally the current chunk will be
c5608a1f
KW
1505discarded at the same time, but if C<flags> has the C<LEX_KEEP_PREVIOUS>
1506bit set, then the current chunk will not be discarded.
f0e67a1d
Z
1507
1508If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1509is encountered, an exception is generated.
1510
1511=cut
1512*/
1513
1514I32
1515Perl_lex_read_unichar(pTHX_ U32 flags)
1516{
1517 I32 c;
1518 if (flags & ~(LEX_KEEP_PREVIOUS))
1519 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1520 c = lex_peek_unichar(flags);
1521 if (c != -1) {
1522 if (c == '\n')
83944c01 1523 COPLINE_INC_WITH_HERELINES;
d9018cbe
EB
1524 if (UTF)
1525 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1526 else
1527 ++(PL_parser->bufptr);
f0e67a1d
Z
1528 }
1529 return c;
1530}
1531
1532/*
1533=for apidoc Amx|void|lex_read_space|U32 flags
1534
1535Reads optional spaces, in Perl style, in the text currently being
1536lexed. The spaces may include ordinary whitespace characters and
1537Perl-style comments. C<#line> directives are processed if encountered.
1538L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1539at a non-space character (or the end of the input text).
1540
1541If spaces extend into the next chunk of input text, the next chunk will
1542be read in. Normally the current chunk will be discarded at the same
c5608a1f 1543time, but if C<flags> has the C<LEX_KEEP_PREVIOUS> bit set, then the current
f0e67a1d
Z
1544chunk will not be discarded.
1545
1546=cut
1547*/
1548
21791330 1549#define LEX_NO_INCLINE 0x40000000
f0998909
Z
1550#define LEX_NO_NEXT_CHUNK 0x80000000
1551
f0e67a1d
Z
1552void
1553Perl_lex_read_space(pTHX_ U32 flags)
1554{
1555 char *s, *bufend;
21791330 1556 const bool can_incline = !(flags & LEX_NO_INCLINE);
f0e67a1d 1557 bool need_incline = 0;
21791330 1558 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK|LEX_NO_INCLINE))
f0e67a1d 1559 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
f0e67a1d
Z
1560 s = PL_parser->bufptr;
1561 bufend = PL_parser->bufend;
1562 while (1) {
1563 char c = *s;
1564 if (c == '#') {
1565 do {
1566 c = *++s;
1567 } while (!(c == '\n' || (c == 0 && s == bufend)));
1568 } else if (c == '\n') {
1569 s++;
21791330
FC
1570 if (can_incline) {
1571 PL_parser->linestart = s;
1572 if (s == bufend)
1573 need_incline = 1;
1574 else
1575 incline(s);
1576 }
f0e67a1d
Z
1577 } else if (isSPACE(c)) {
1578 s++;
1579 } else if (c == 0 && s == bufend) {
1580 bool got_more;
65c68e17 1581 line_t l;
f0998909
Z
1582 if (flags & LEX_NO_NEXT_CHUNK)
1583 break;
f0e67a1d 1584 PL_parser->bufptr = s;
65c68e17 1585 l = CopLINE(PL_curcop);
851b527a 1586 CopLINE(PL_curcop) += PL_parser->herelines + 1;
f0e67a1d 1587 got_more = lex_next_chunk(flags);
65c68e17 1588 CopLINE_set(PL_curcop, l);
f0e67a1d
Z
1589 s = PL_parser->bufptr;
1590 bufend = PL_parser->bufend;
1591 if (!got_more)
1592 break;
21791330 1593 if (can_incline && need_incline && PL_parser->rsfp) {
f0e67a1d
Z
1594 incline(s);
1595 need_incline = 0;
1596 }
3c47da3c
FC
1597 } else if (!c) {
1598 s++;
f0e67a1d
Z
1599 } else {
1600 break;
1601 }
1602 }
f0e67a1d
Z
1603 PL_parser->bufptr = s;
1604}
1605
1606/*
fe788d6b
PM
1607
1608=for apidoc EXMp|bool|validate_proto|SV *name|SV *proto|bool warn
1609
1610This function performs syntax checking on a prototype, C<proto>.
1611If C<warn> is true, any illegal characters or mismatched brackets
1612will trigger illegalproto warnings, declaring that they were
1613detected in the prototype for C<name>.
1614
1615The return value is C<true> if this is a valid prototype, and
1616C<false> if it is not, regardless of whether C<warn> was C<true> or
1617C<false>.
1618
1619Note that C<NULL> is a valid C<proto> and will always return C<true>.
1620
1621=cut
1622
1623 */
1624
1625bool
5783dc51 1626Perl_validate_proto(pTHX_ SV *name, SV *proto, bool warn, bool curstash)
fe788d6b
PM
1627{
1628 STRLEN len, origlen;
11327fa1 1629 char *p;
fe788d6b
PM
1630 bool bad_proto = FALSE;
1631 bool in_brackets = FALSE;
1632 bool after_slash = FALSE;
1633 char greedy_proto = ' ';
1634 bool proto_after_greedy_proto = FALSE;
1635 bool must_be_last = FALSE;
1636 bool underscore = FALSE;
f791a21a 1637 bool bad_proto_after_underscore = FALSE;
fe788d6b
PM
1638
1639 PERL_ARGS_ASSERT_VALIDATE_PROTO;
1640
1641 if (!proto)
1642 return TRUE;
1643
11327fa1 1644 p = SvPV(proto, len);
fe788d6b
PM
1645 origlen = len;
1646 for (; len--; p++) {
1647 if (!isSPACE(*p)) {
1648 if (must_be_last)
1649 proto_after_greedy_proto = TRUE;
f791a21a
PM
1650 if (underscore) {
1651 if (!strchr(";@%", *p))
1652 bad_proto_after_underscore = TRUE;
1653 underscore = FALSE;
1654 }
fe788d6b
PM
1655 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
1656 bad_proto = TRUE;
1657 }
1658 else {
fe788d6b
PM
1659 if (*p == '[')
1660 in_brackets = TRUE;
1661 else if (*p == ']')
1662 in_brackets = FALSE;
407f8cf2
KW
1663 else if ((*p == '@' || *p == '%')
1664 && !after_slash
1665 && !in_brackets )
1666 {
fe788d6b
PM
1667 must_be_last = TRUE;
1668 greedy_proto = *p;
1669 }
1670 else if (*p == '_')
f791a21a 1671 underscore = TRUE;
fe788d6b
PM
1672 }
1673 if (*p == '\\')
1674 after_slash = TRUE;
1675 else
1676 after_slash = FALSE;
1677 }
1678 }
1679
1680 if (warn) {
b54d603d 1681 SV *tmpsv = newSVpvs_flags("", SVs_TEMP);
fe788d6b 1682 p -= origlen;
b54d603d
PM
1683 p = SvUTF8(proto)
1684 ? sv_uni_display(tmpsv, newSVpvn_flags(p, origlen, SVs_TEMP | SVf_UTF8),
1685 origlen, UNI_DISPLAY_ISPRINT)
1686 : pv_pretty(tmpsv, p, origlen, 60, NULL, NULL, PERL_PV_ESCAPE_NONASCII);
1687
cbf83791
FC
1688 if (curstash && !memchr(SvPVX(name), ':', SvCUR(name))) {
1689 SV *name2 = sv_2mortal(newSVsv(PL_curstname));
1690 sv_catpvs(name2, "::");
1691 sv_catsv(name2, (SV *)name);
1692 name = name2;
1693 }
1694
fe788d6b
PM
1695 if (proto_after_greedy_proto)
1696 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1697 "Prototype after '%c' for %" SVf " : %s",
fe788d6b 1698 greedy_proto, SVfARG(name), p);
50278ed0
PM
1699 if (in_brackets)
1700 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1701 "Missing ']' in prototype for %" SVf " : %s",
50278ed0 1702 SVfARG(name), p);
b54d603d 1703 if (bad_proto)
fe788d6b 1704 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1705 "Illegal character in prototype for %" SVf " : %s",
f791a21a
PM
1706 SVfARG(name), p);
1707 if (bad_proto_after_underscore)
1708 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
147e3846 1709 "Illegal character after '_' in prototype for %" SVf " : %s",
f791a21a 1710 SVfARG(name), p);
fe788d6b
PM
1711 }
1712
1713 return (! (proto_after_greedy_proto || bad_proto) );
1714}
1715
1716/*
ffb4593c
NT
1717 * S_incline
1718 * This subroutine has nothing to do with tilting, whether at windmills
1719 * or pinball tables. Its name is short for "increment line". It
57843af0 1720 * increments the current line number in CopLINE(PL_curcop) and checks
ffb4593c 1721 * to see whether the line starts with a comment of the form
9cbb5ea2
GS
1722 * # line 500 "foo.pm"
1723 * If so, it sets the current line number and file to the values in the comment.
ffb4593c
NT
1724 */
1725
76e3520e 1726STATIC void
d9095cec 1727S_incline(pTHX_ const char *s)
463ee0b2 1728{
d9095cec
NC
1729 const char *t;
1730 const char *n;
1731 const char *e;
8818d409 1732 line_t line_num;
22ff3130 1733 UV uv;
463ee0b2 1734
7918f24d
NC
1735 PERL_ARGS_ASSERT_INCLINE;
1736
83944c01 1737 COPLINE_INC_WITH_HERELINES;
451f421f
FC
1738 if (!PL_rsfp && !PL_parser->filtered && PL_lex_state == LEX_NORMAL
1739 && s+1 == PL_bufend && *s == ';') {
1740 /* fake newline in string eval */
1741 CopLINE_dec(PL_curcop);
1742 return;
1743 }
463ee0b2
LW
1744 if (*s++ != '#')
1745 return;
d4c19fe8
AL
1746 while (SPACE_OR_TAB(*s))
1747 s++;
847cc851 1748 if (strEQs(s, "line"))
73659bf1
GS
1749 s += 4;
1750 else
1751 return;
084592ab 1752 if (SPACE_OR_TAB(*s))
73659bf1 1753 s++;
4e553d73 1754 else
73659bf1 1755 return;
d4c19fe8
AL
1756 while (SPACE_OR_TAB(*s))
1757 s++;
463ee0b2
LW
1758 if (!isDIGIT(*s))
1759 return;
d4c19fe8 1760
463ee0b2
LW
1761 n = s;
1762 while (isDIGIT(*s))
1763 s++;
07714eb4 1764 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
26b6dc3f 1765 return;
bf4acbe4 1766 while (SPACE_OR_TAB(*s))
463ee0b2 1767 s++;
73659bf1 1768 if (*s == '"' && (t = strchr(s+1, '"'))) {
463ee0b2 1769 s++;
73659bf1
GS
1770 e = t + 1;
1771 }
463ee0b2 1772 else {
c35e046a 1773 t = s;
1bb1a3d6 1774 while (*t && !isSPACE(*t))
c35e046a 1775 t++;
73659bf1 1776 e = t;
463ee0b2 1777 }
bf4acbe4 1778 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
73659bf1
GS
1779 e++;
1780 if (*e != '\n' && *e != '\0')
1781 return; /* false alarm */
1782
22ff3130
HS
1783 if (!grok_atoUV(n, &uv, &e))
1784 return;
1785 line_num = ((line_t)uv) - 1;
8818d409 1786
f4dd75d9 1787 if (t - s > 0) {
d9095cec 1788 const STRLEN len = t - s;
3df32bda 1789
d36ee5be 1790 if (!PL_rsfp && !PL_parser->filtered) {
e66cf94c
RGS
1791 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1792 * to *{"::_<newfilename"} */
44867030
NC
1793 /* However, the long form of evals is only turned on by the
1794 debugger - usually they're "(eval %lu)" */
d36ee5be
FC
1795 GV * const cfgv = CopFILEGV(PL_curcop);
1796 if (cfgv) {
38bd7ad8
FC
1797 char smallbuf[128];
1798 STRLEN tmplen2 = len;
44867030 1799 char *tmpbuf2;
449dd039 1800 GV *gv2;
44867030
NC
1801
1802 if (tmplen2 + 2 <= sizeof smallbuf)
1803 tmpbuf2 = smallbuf;
1804 else
1805 Newx(tmpbuf2, tmplen2 + 2, char);
1806
38bd7ad8
FC
1807 tmpbuf2[0] = '_';
1808 tmpbuf2[1] = '<';
44867030
NC
1809
1810 memcpy(tmpbuf2 + 2, s, tmplen2);
1811 tmplen2 += 2;
1812
8a5ee598 1813 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
e5527e4b 1814 if (!isGV(gv2)) {
8a5ee598 1815 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
e5527e4b
RGS
1816 /* adjust ${"::_<newfilename"} to store the new file name */
1817 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
8818d409
FC
1818 /* The line number may differ. If that is the case,
1819 alias the saved lines that are in the array.
1820 Otherwise alias the whole array. */
1821 if (CopLINE(PL_curcop) == line_num) {
38bd7ad8
FC
1822 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(cfgv)));
1823 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(cfgv)));
8818d409 1824 }
38bd7ad8
FC
1825 else if (GvAV(cfgv)) {
1826 AV * const av = GvAV(cfgv);
8818d409
FC
1827 const I32 start = CopLINE(PL_curcop)+1;
1828 I32 items = AvFILLp(av) - start;
1829 if (items > 0) {
1830 AV * const av2 = GvAVn(gv2);
1831 SV **svp = AvARRAY(av) + start;
1832 I32 l = (I32)line_num+1;
1833 while (items--)
1834 av_store(av2, l++, SvREFCNT_inc(*svp++));
1835 }
1836 }
e5527e4b 1837 }
44867030
NC
1838
1839 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
d36ee5be 1840 }
e66cf94c 1841 }
05ec9bb3 1842 CopFILE_free(PL_curcop);
449dd039 1843 CopFILE_setn(PL_curcop, s, len);
f4dd75d9 1844 }
8818d409 1845 CopLINE_set(PL_curcop, line_num);
463ee0b2
LW
1846}
1847
80a702cd 1848STATIC void
15f169a1 1849S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
80a702cd
RGS
1850{
1851 AV *av = CopFILEAVx(PL_curcop);
1852 if (av) {
7f1c3e8c
FC
1853 SV * sv;
1854 if (PL_parser->preambling == NOLINE) sv = newSV_type(SVt_PVMG);
1855 else {
1856 sv = *av_fetch(av, 0, 1);
1857 SvUPGRADE(sv, SVt_PVMG);
1858 }
847cc851 1859 if (!SvPOK(sv)) SvPVCLEAR(sv);
5fa550fb 1860 if (orig_sv)
7f1c3e8c 1861 sv_catsv(sv, orig_sv);
5fa550fb 1862 else
7f1c3e8c
FC
1863 sv_catpvn(sv, buf, len);
1864 if (!SvIOK(sv)) {
1865 (void)SvIOK_on(sv);
1866 SvIV_set(sv, 0);
1867 }
1868 if (PL_parser->preambling == NOLINE)
1869 av_store(av, CopLINE(PL_curcop), sv);
80a702cd
RGS
1870 }
1871}
1872
ffb4593c 1873/*
8c6b0c7d 1874 * skipspace
ffb4593c
NT
1875 * Called to gobble the appropriate amount and type of whitespace.
1876 * Skips comments as well.
71fff7cb 1877 * Returns the next character after the whitespace that is skipped.
8c6b0c7d
FC
1878 *
1879 * peekspace
1880 * Same thing, but look ahead without incrementing line numbers or
1881 * adjusting PL_linestart.
ffb4593c
NT
1882 */
1883
3218e223 1884#define skipspace(s) skipspace_flags(s, 0)
8c6b0c7d 1885#define peekspace(s) skipspace_flags(s, LEX_NO_INCLINE)
3218e223 1886
76e3520e 1887STATIC char *
21791330 1888S_skipspace_flags(pTHX_ char *s, U32 flags)
a687059c 1889{
21791330 1890 PERL_ARGS_ASSERT_SKIPSPACE_FLAGS;
3280af22 1891 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3c47da3c 1892 while (s < PL_bufend && (SPACE_OR_TAB(*s) || !*s))
463ee0b2 1893 s++;
f0e67a1d
Z
1894 } else {
1895 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1896 PL_bufptr = s;
21791330 1897 lex_read_space(flags | LEX_KEEP_PREVIOUS |
d27f4b91 1898 (PL_lex_inwhat || PL_lex_state == LEX_FORMLINE ?
f0998909 1899 LEX_NO_NEXT_CHUNK : 0));
3280af22 1900 s = PL_bufptr;
f0e67a1d
Z
1901 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1902 if (PL_linestart > PL_bufptr)
1903 PL_bufptr = PL_linestart;
1904 return s;
463ee0b2 1905 }
5db06880 1906 return s;
a687059c 1907}
378cc40b 1908
ffb4593c
NT
1909/*
1910 * S_check_uni
1911 * Check the unary operators to ensure there's no ambiguity in how they're
1912 * used. An ambiguous piece of code would be:
1913 * rand + 5
1914 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1915 * the +5 is its argument.
1916 */
1917
76e3520e 1918STATIC void
cea2e8a9 1919S_check_uni(pTHX)
ba106d47 1920{
d4c19fe8
AL
1921 const char *s;
1922 const char *t;
2f3197b3 1923
3280af22 1924 if (PL_oldoldbufptr != PL_last_uni)
2f3197b3 1925 return;
3280af22
NIS
1926 while (isSPACE(*PL_last_uni))
1927 PL_last_uni++;
c35e046a 1928 s = PL_last_uni;
fac0f7a3 1929 while (isWORDCHAR_lazy_if_safe(s, PL_bufend, UTF) || *s == '-')
8ce2ba82 1930 s += UTF ? UTF8SKIP(s) : 1;
3280af22 1931 if ((t = strchr(s, '(')) && t < PL_bufptr)
a0d0e21e 1932 return;
6136c704 1933
9b387841 1934 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
147e3846 1935 "Warning: Use of \"%" UTF8f "\" without parentheses is ambiguous",
b59c097b 1936 UTF8fARG(UTF, (int)(s - PL_last_uni), PL_last_uni));
2f3197b3
LW
1937}
1938
ffb4593c
NT
1939/*
1940 * LOP : macro to build a list operator. Its behaviour has been replaced
1941 * with a subroutine, S_lop() for which LOP is just another name.
1942 */
1943
a0d0e21e
LW
1944#define LOP(f,x) return lop(f,x,s)
1945
ffb4593c
NT
1946/*
1947 * S_lop
1948 * Build a list operator (or something that might be one). The rules:
41e8cbf4
FC
1949 * - if we have a next token, then it's a list operator (no parens) for
1950 * which the next token has already been parsed; e.g.,
1951 * sort foo @args
1952 * sort foo (@args)
ffb4593c
NT
1953 * - if the next thing is an opening paren, then it's a function
1954 * - else it's a list operator
1955 */
1956
76e3520e 1957STATIC I32
11288bb3 1958S_lop(pTHX_ I32 f, U8 x, char *s)
ffed7fef 1959{
7918f24d
NC
1960 PERL_ARGS_ASSERT_LOP;
1961
6154021b 1962 pl_yylval.ival = f;
35c8bce7 1963 CLINE;
3280af22
NIS
1964 PL_bufptr = s;
1965 PL_last_lop = PL_oldbufptr;
eb160463 1966 PL_last_lop_op = (OPCODE)f;
3280af22 1967 if (PL_nexttoke)
78cdf107 1968 goto lstop;
19f1898a 1969 PL_expect = x;
79072805 1970 if (*s == '(')
bbf60fe6 1971 return REPORT(FUNC);
294a536f 1972 s = skipspace(s);
79072805 1973 if (*s == '(')
bbf60fe6 1974 return REPORT(FUNC);
78cdf107
Z
1975 else {
1976 lstop:
1977 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1978 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
bbf60fe6 1979 return REPORT(LSTOP);
78cdf107 1980 }
79072805
LW
1981}
1982
ffb4593c
NT
1983/*
1984 * S_force_next
9cbb5ea2 1985 * When the lexer realizes it knows the next token (for instance,
ffb4593c 1986 * it is reordering tokens for the parser) then it can call S_force_next
9cbb5ea2 1987 * to know what token to return the next time the lexer is called. Caller
b5bbe64a
JH
1988 * will need to set PL_nextval[] and possibly PL_expect to ensure
1989 * the lexer handles the token correctly.
ffb4593c
NT
1990 */
1991
4e553d73 1992STATIC void
cea2e8a9 1993S_force_next(pTHX_ I32 type)
79072805 1994{
704d4215
GG
1995#ifdef DEBUGGING
1996 if (DEBUG_T_TEST) {
1997 PerlIO_printf(Perl_debug_log, "### forced token:\n");
f05d7009 1998 tokereport(type, &NEXTVAL_NEXTTOKE);
704d4215
GG
1999 }
2000#endif
1f7c3e7c 2001 assert(PL_nexttoke < C_ARRAY_LENGTH(PL_nexttype));
3280af22
NIS
2002 PL_nexttype[PL_nexttoke] = type;
2003 PL_nexttoke++;
79072805
LW
2004}
2005
89f35911
FC
2006/*
2007 * S_postderef
2008 *
2009 * This subroutine handles postfix deref syntax after the arrow has already
2010 * been emitted. @* $* etc. are emitted as two separate token right here.
2011 * @[ @{ %[ %{ *{ are emitted also as two tokens, but this function emits
2012 * only the first, leaving yylex to find the next.
89f35911
FC
2013 */
2014
2015static int
ff25e5db 2016S_postderef(pTHX_ int const funny, char const next)
89f35911 2017{
ff25e5db 2018 assert(funny == DOLSHARP || strchr("$@%&*", funny));
89f35911
FC
2019 if (next == '*') {
2020 PL_expect = XOPERATOR;
cc624add 2021 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
ff25e5db 2022 assert('@' == funny || '$' == funny || DOLSHARP == funny);
cc624add 2023 PL_lex_state = LEX_INTERPEND;
c3492809
FC
2024 if ('@' == funny)
2025 force_next(POSTJOIN);
cc624add 2026 }
89f35911
FC
2027 force_next(next);
2028 PL_bufptr+=2;
2029 }
2030 else {
760ca746
FC
2031 if ('@' == funny && PL_lex_state == LEX_INTERPNORMAL
2032 && !PL_lex_brackets)
2033 PL_lex_dojoin = 2;
89f35911
FC
2034 PL_expect = XOPERATOR;
2035 PL_bufptr++;
2036 }
2037 return funny;
2038}
2039
28ac2b49
Z
2040void
2041Perl_yyunlex(pTHX)
2042{
a7aaec61
Z
2043 int yyc = PL_parser->yychar;
2044 if (yyc != YYEMPTY) {
2045 if (yyc) {
a7aaec61
Z
2046 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2047 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
78cdf107 2048 PL_lex_allbrackets--;
a7aaec61 2049 PL_lex_brackets--;
78cdf107
Z
2050 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2051 } else if (yyc == '('/*)*/) {
2052 PL_lex_allbrackets--;
2053 yyc |= (2<<24);
a7aaec61
Z
2054 }
2055 force_next(yyc);
2056 }
28ac2b49
Z
2057 PL_parser->yychar = YYEMPTY;
2058 }
2059}
2060
d0a148a6 2061STATIC SV *
15f169a1 2062S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
d0a148a6 2063{
740cce10 2064 SV * const sv = newSVpvn_utf8(start, len,
c5f058df
KW
2065 !IN_BYTES
2066 && UTF
2067 && !is_utf8_invariant_string((const U8*)start, len)
2068 && is_utf8_string((const U8*)start, len));
d0a148a6
NC
2069 return sv;
2070}
2071
ffb4593c
NT
2072/*
2073 * S_force_word
2074 * When the lexer knows the next thing is a word (for instance, it has
2075 * just seen -> and it knows that the next char is a word char, then
02b34bbe
DM
2076 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2077 * lookahead.
ffb4593c
NT
2078 *
2079 * Arguments:
b1b65b59 2080 * char *start : buffer position (must be within PL_linestr)
185c2e96
DM
2081 * int token : PL_next* will be this type of bare word
2082 * (e.g., METHOD,BAREWORD)
ffb4593c
NT
2083 * int check_keyword : if true, Perl checks to make sure the word isn't
2084 * a keyword (do this if the word is a label, e.g. goto FOO)
2085 * int allow_pack : if true, : characters will also be allowed (require,
2086 * use, etc. do this)
ffb4593c
NT
2087 */
2088
76e3520e 2089STATIC char *
345b3785 2090S_force_word(pTHX_ char *start, int token, int check_keyword, int allow_pack)
79072805 2091{
eb578fdb 2092 char *s;
463ee0b2 2093 STRLEN len;
4e553d73 2094
7918f24d
NC
2095 PERL_ARGS_ASSERT_FORCE_WORD;
2096
294a536f 2097 start = skipspace(start);
463ee0b2 2098 s = start;
fac0f7a3 2099 if ( isIDFIRST_lazy_if_safe(s, PL_bufend, UTF)
e7127e21 2100 || (allow_pack && *s == ':' && s[1] == ':') )
a0d0e21e 2101 {
3280af22 2102 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
01b5ef50
FC
2103 if (check_keyword) {
2104 char *s2 = PL_tokenbuf;
487e470d 2105 STRLEN len2 = len;
847cc851 2106 if (allow_pack && len > 6 && strEQs(s2, "CORE::"))
487e470d
FC
2107 s2 += 6, len2 -= 6;
2108 if (keyword(s2, len2, 0))
463ee0b2 2109 return start;
01b5ef50 2110 }
463ee0b2 2111 if (token == METHOD) {
294a536f 2112 s = skipspace(s);
463ee0b2 2113 if (*s == '(')
3280af22 2114 PL_expect = XTERM;
463ee0b2 2115 else {
3280af22 2116 PL_expect = XOPERATOR;
463ee0b2 2117 }
79072805 2118 }
9ded7720 2119 NEXTVAL_NEXTTOKE.opval
275103cd 2120 = newSVOP(OP_CONST,0,
d0a148a6 2121 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
9ded7720 2122 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
79072805
LW
2123 force_next(token);
2124 }
2125 return s;
2126}
2127
ffb4593c
NT
2128/*
2129 * S_force_ident
9cbb5ea2 2130 * Called when the lexer wants $foo *foo &foo etc, but the program
ffb4593c
NT
2131 * text only contains the "foo" portion. The first argument is a pointer
2132 * to the "foo", and the second argument is the type symbol to prefix.
185c2e96 2133 * Forces the next token to be a "BAREWORD".
9cbb5ea2 2134 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
ffb4593c
NT
2135 */
2136
76e3520e 2137STATIC void
5aaab254 2138S_force_ident(pTHX_ const char *s, int kind)
79072805 2139{
7918f24d
NC
2140 PERL_ARGS_ASSERT_FORCE_IDENT;
2141
c9b48522
DD
2142 if (s[0]) {
2143 const STRLEN len = s[1] ? strlen(s) : 1; /* s = "\"" see yylex */
275103cd 2144 OP* const o = newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
728847b1 2145 UTF ? SVf_UTF8 : 0));
9ded7720 2146 NEXTVAL_NEXTTOKE.opval = o;
185c2e96 2147 force_next(BAREWORD);
748a9306 2148 if (kind) {
11343788 2149 o->op_private = OPpCONST_ENTERED;
55497cff
PP
2150 /* XXX see note in pp_entereval() for why we forgo typo
2151 warnings if the symbol must be introduced in an eval.
2152 GSAR 96-10-12 */
90e5519e 2153 gv_fetchpvn_flags(s, len,
4bff32c5 2154 (PL_in_eval ? GV_ADDMULTI
728847b1 2155 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
90e5519e
NC
2156 kind == '$' ? SVt_PV :
2157 kind == '@' ? SVt_PVAV :
2158 kind == '%' ? SVt_PVHV :
a0d0e21e 2159 SVt_PVGV
90e5519e 2160 );
748a9306 2161 }
79072805
LW
2162 }
2163}
2164
3f33d153
FC
2165static void
2166S_force_ident_maybe_lex(pTHX_ char pit)
2167{
3f33d153
FC
2168 NEXTVAL_NEXTTOKE.ival = pit;
2169 force_next('p');
2170}
2171
1571675a
GS
2172NV
2173Perl_str_to_version(pTHX_ SV *sv)
2174{
2175 NV retval = 0.0;
2176 NV nshift = 1.0;
2177 STRLEN len;
cfd0369c 2178 const char *start = SvPV_const(sv,len);
9d4ba2ae 2179 const char * const end = start + len;
8298454c 2180 const bool utf = cBOOL(SvUTF8(sv));
7918f24d
NC
2181
2182 PERL_ARGS_ASSERT_STR_TO_VERSION;
2183
1571675a 2184 while (start < end) {
ba210ebe 2185 STRLEN skip;
1571675a
GS
2186 UV n;
2187 if (utf)
9041c2e3 2188 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1571675a
GS
2189 else {
2190 n = *(U8*)start;
2191 skip = 1;
2192 }
2193 retval += ((NV)n)/nshift;
2194 start += skip;
2195 nshift *= 1000;
2196 }
2197 return retval;
2198}
2199
4e553d73 2200/*
ffb4593c
NT
2201 * S_force_version
2202 * Forces the next token to be a version number.
e759cc13
RGS
2203 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2204 * and if "guessing" is TRUE, then no new token is created (and the caller
2205 * must use an alternative parsing method).
ffb4593c
NT
2206 */
2207
76e3520e 2208STATIC char *
e759cc13 2209S_force_version(pTHX_ char *s, int guessing)
89bfa8cd 2210{
5f66b61c 2211 OP *version = NULL;
44dcb63b 2212 char *d;
89bfa8cd 2213
7918f24d
NC
2214 PERL_ARGS_ASSERT_FORCE_VERSION;
2215
294a536f 2216 s = skipspace(s);
89bfa8cd 2217
44dcb63b 2218 d = s;
dd629d5b 2219 if (*d == 'v')
44dcb63b 2220 d++;
44dcb63b 2221 if (isDIGIT(*d)) {
e759cc13
RGS
2222 while (isDIGIT(*d) || *d == '_' || *d == '.')
2223 d++;
4e4da3ac 2224 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
dd629d5b 2225 SV *ver;
6154021b
RGS
2226 s = scan_num(s, &pl_yylval);
2227 version = pl_yylval.opval;
dd629d5b
GS
2228 ver = cSVOPx(version)->op_sv;
2229 if (SvPOK(ver) && !SvNIOK(ver)) {
862a34c6 2230 SvUPGRADE(ver, SVt_PVNV);
9d6ce603 2231 SvNV_set(ver, str_to_version(ver));
1571675a 2232 SvNOK_on(ver); /* hint that it is a version */
44dcb63b 2233 }
89bfa8cd 2234 }
5db06880 2235 else if (guessing) {
e759cc13 2236 return s;
5db06880 2237 }
89bfa8cd
PP
2238 }
2239
2240 /* NOTE: The parser sees the package name and the VERSION swapped */
9ded7720 2241 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2242 force_next(BAREWORD);
89bfa8cd 2243
e759cc13 2244 return s;
89bfa8cd
PP
2245}
2246
ffb4593c 2247/*
91152fc1
DG
2248 * S_force_strict_version
2249 * Forces the next token to be a version number using strict syntax rules.
2250 */
2251
2252STATIC char *
2253S_force_strict_version(pTHX_ char *s)
2254{
91152fc1 2255 OP *version = NULL;
91152fc1
DG
2256 const char *errstr = NULL;
2257
2258 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2259
2260 while (isSPACE(*s)) /* leading whitespace */
2261 s++;
2262
2263 if (is_STRICT_VERSION(s,&errstr)) {
2264 SV *ver = newSV(0);
2265 s = (char *)scan_version(s, ver, 0);
2266 version = newSVOP(OP_CONST, 0, ver);
2267 }
407f8cf2
KW
2268 else if ((*s != ';' && *s != '{' && *s != '}' )
2269 && (s = skipspace(s), (*s != ';' && *s != '{' && *s != '}' )))
4e4da3ac 2270 {
91152fc1
DG
2271 PL_bufptr = s;
2272 if (errstr)
2273 yyerror(errstr); /* version required */
2274 return s;
2275 }
2276
91152fc1 2277 /* NOTE: The parser sees the package name and the VERSION swapped */
91152fc1 2278 NEXTVAL_NEXTTOKE.opval = version;
185c2e96 2279 force_next(BAREWORD);
91152fc1
DG
2280
2281 return s;
2282}
2283
2284/*
ffb4593c 2285 * S_tokeq
ef3ff34d
KW
2286 * Turns any \\ into \ in a quoted string passed in in 'sv', returning 'sv',
2287 * modified as necessary. However, if HINT_NEW_STRING is on, 'sv' is
2288 * unchanged, and a new SV containing the modified input is returned.
ffb4593c
NT
2289 */
2290
76e3520e 2291STATIC SV *
cea2e8a9 2292S_tokeq(pTHX_ SV *sv)
79072805 2293{
eb578fdb
KW
2294 char *s;
2295 char *send;
2296 char *d;
b3ac6de7 2297 SV *pv = sv;
79072805 2298
7918f24d
NC
2299 PERL_ARGS_ASSERT_TOKEQ;
2300
279b35ad
FC
2301 assert (SvPOK(sv));
2302 assert (SvLEN(sv));
2303 assert (!SvIsCOW(sv));
307ed071 2304 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1) /* <<'heredoc' */
b3ac6de7 2305 goto finish;
279b35ad
FC
2306 s = SvPVX(sv);
2307 send = SvEND(sv);
dcb21ed6
NC
2308 /* This is relying on the SV being "well formed" with a trailing '\0' */
2309 while (s < send && !(*s == '\\' && s[1] == '\\'))
79072805
LW
2310 s++;
2311 if (s == send)
b3ac6de7 2312 goto finish;
79072805 2313 d = s;
be4731d2 2314 if ( PL_hints & HINT_NEW_STRING ) {
279b35ad
FC
2315 pv = newSVpvn_flags(SvPVX_const(pv), SvCUR(sv),
2316 SVs_TEMP | SvUTF8(sv));
be4731d2 2317 }
79072805
LW
2318 while (s < send) {
2319 if (*s == '\\') {
a0d0e21e 2320 if (s + 1 < send && (s[1] == '\\'))
79072805
LW
2321 s++; /* all that, just for this */
2322 }
2323 *d++ = *s++;
2324 }
2325 *d = '\0';
95a20fc0 2326 SvCUR_set(sv, d - SvPVX_const(sv));
b3ac6de7 2327 finish:
3280af22 2328 if ( PL_hints & HINT_NEW_STRING )
eb0d8d16 2329 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
79072805
LW
2330 return sv;
2331}
2332
ffb4593c
NT
2333/*
2334 * Now come three functions related to double-quote context,
2335 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2336 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2337 * interact with PL_lex_state, and create fake ( ... ) argument lists
2338 * to handle functions and concatenation.
ecd24171
DM
2339 * For example,
2340 * "foo\lbar"
2341 * is tokenised as
2342 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
ffb4593c
NT
2343 */
2344
2345/*
2346 * S_sublex_start
6154021b 2347 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
ffb4593c
NT
2348 *
2349 * Pattern matching will set PL_lex_op to the pattern-matching op to
6154021b 2350 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
ffb4593c 2351 *
16d1d8bd 2352 * OP_CONST is easy--just make the new op and return.
ffb4593c
NT
2353 *
2354 * Everything else becomes a FUNC.
2355 *
16d1d8bd
LM
2356 * Sets PL_lex_state to LEX_INTERPPUSH unless ival was OP_NULL or we
2357 * had an OP_CONST. This just sets us up for a
ffb4593c
NT
2358 * call to S_sublex_push().
2359 */
2360
76e3520e 2361STATIC I32
cea2e8a9 2362S_sublex_start(pTHX)
79072805 2363{
eb578fdb 2364 const I32 op_type = pl_yylval.ival;
79072805
LW
2365
2366 if (op_type == OP_NULL) {
6154021b 2367 pl_yylval.opval = PL_lex_op;
5f66b61c 2368 PL_lex_op = NULL;
79072805
LW
2369 return THING;
2370 }
466112bb 2371 if (op_type == OP_CONST) {
67c71cbb
FC
2372 SV *sv = PL_lex_stuff;
2373 PL_lex_stuff = NULL;
2374 sv = tokeq(sv);
b3ac6de7
IZ
2375
2376 if (SvTYPE(sv) == SVt_PVIV) {
2377 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2378 STRLEN len;
96a5add6 2379 const char * const p = SvPV_const(sv, len);
740cce10 2380 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
b3ac6de7
IZ
2381 SvREFCNT_dec(sv);
2382 sv = nsv;
4e553d73 2383 }
275103cd 2384 pl_yylval.opval = newSVOP(op_type, 0, sv);
79072805
LW
2385 return THING;
2386 }
2387
7ef70b3d
FC
2388 PL_parser->lex_super_state = PL_lex_state;
2389 PL_parser->lex_sub_inwhat = (U16)op_type;
2390 PL_parser->lex_sub_op = PL_lex_op;
3280af22 2391 PL_lex_state = LEX_INTERPPUSH;
55497cff 2392
3280af22
NIS
2393 PL_expect = XTERM;
2394 if (PL_lex_op) {
6154021b 2395 pl_yylval.opval = PL_lex_op;
5f66b61c 2396 PL_lex_op = NULL;
55497cff
PP
2397 return PMFUNC;
2398 }
2399 else
2400 return FUNC;
2401}
2402
ffb4593c
NT
2403/*
2404 * S_sublex_push
2405 * Create a new scope to save the lexing state. The scope will be
2406 * ended in S_sublex_done. Returns a '(', starting the function arguments
2407 * to the uc, lc, etc. found before.
2408 * Sets PL_lex_state to LEX_INTERPCONCAT.
2409 */
2410
76e3520e 2411STATIC I32
cea2e8a9 2412S_sublex_push(pTHX)
55497cff 2413{
78a635de 2414 LEXSHARED *shared;
801d32ac 2415 const bool is_heredoc = PL_multi_close == '<';
f46d017c 2416 ENTER;
55497cff 2417
7ef70b3d 2418 PL_lex_state = PL_parser->lex_super_state;
cc624add 2419 SAVEI8(PL_lex_dojoin);
3280af22 2420 SAVEI32(PL_lex_brackets);
78cdf107 2421 SAVEI32(PL_lex_allbrackets);
b27dce25 2422 SAVEI32(PL_lex_formbrack);
78cdf107 2423 SAVEI8(PL_lex_fakeeof);
3280af22
NIS
2424 SAVEI32(PL_lex_casemods);
2425 SAVEI32(PL_lex_starts);
651b5b28 2426 SAVEI8(PL_lex_state);
7cc34111 2427 SAVESPTR(PL_lex_repl);
7766f137 2428 SAVEVPTR(PL_lex_inpat);
98246f1e 2429 SAVEI16(PL_lex_inwhat);
ffdb8b16 2430 if (is_heredoc)
b42366d4 2431 {
ffdb8b16 2432 SAVECOPLINE(PL_curcop);
b42366d4 2433 SAVEI32(PL_multi_end);
851b527a
FC
2434 SAVEI32(PL_parser->herelines);
2435 PL_parser->herelines = 0;
b42366d4 2436 }
2ca4363d 2437 SAVEIV(PL_multi_close);
3280af22 2438 SAVEPPTR(PL_bufptr);
8452ff4b 2439 SAVEPPTR(PL_bufend);
3280af22
NIS
2440 SAVEPPTR(PL_oldbufptr);
2441 SAVEPPTR(PL_oldoldbufptr);
207e3d1a
JH
2442 SAVEPPTR(PL_last_lop);
2443 SAVEPPTR(PL_last_uni);
3280af22
NIS
2444 SAVEPPTR(PL_linestart);
2445 SAVESPTR(PL_linestr);
8edd5f42
RGS
2446 SAVEGENERICPV(PL_lex_brackstack);
2447 SAVEGENERICPV(PL_lex_casestack);
78a635de 2448 SAVEGENERICPV(PL_parser->lex_shared);
3a54fd60 2449 SAVEBOOL(PL_parser->lex_re_reparsing);
ffdb8b16 2450 SAVEI32(PL_copline);
3280af22 2451
99bd9d90 2452 /* The here-doc parser needs to be able to peek into outer lexing
60f40a38
FC
2453 scopes to find the body of the here-doc. So we put PL_linestr and
2454 PL_bufptr into lex_shared, to ‘share’ those values.
99bd9d90 2455 */
60f40a38
FC
2456 PL_parser->lex_shared->ls_linestr = PL_linestr;
2457 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
99bd9d90 2458
3280af22 2459 PL_linestr = PL_lex_stuff;
7ef70b3d 2460 PL_lex_repl = PL_parser->lex_sub_repl;
a0714e2c 2461 PL_lex_stuff = NULL;
7ef70b3d 2462 PL_parser->lex_sub_repl = NULL;
3280af22 2463
eabab8bc
FC
2464 /* Arrange for PL_lex_stuff to be freed on scope exit, in case it gets
2465 set for an inner quote-like operator and then an error causes scope-
2466 popping. We must not have a PL_lex_stuff value left dangling, as
2467 that breaks assumptions elsewhere. See bug #123617. */
2468 SAVEGENERICSV(PL_lex_stuff);
7ef70b3d 2469 SAVEGENERICSV(PL_parser->lex_sub_repl);
eabab8bc 2470
9cbb5ea2
GS
2471 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2472 = SvPVX(PL_linestr);
3280af22 2473 PL_bufend += SvCUR(PL_linestr);
bd61b366 2474 PL_last_lop = PL_last_uni = NULL;
3280af22 2475 SAVEFREESV(PL_linestr);
4dc843bc 2476 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
3280af22
NIS
2477
2478 PL_lex_dojoin = FALSE;
b27dce25 2479 PL_lex_brackets = PL_lex_formbrack = 0;
78cdf107
Z
2480 PL_lex_allbrackets = 0;
2481 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
a02a5408
JC
2482 Newx(PL_lex_brackstack, 120, char);
2483 Newx(PL_lex_casestack, 12, char);
3280af22
NIS
2484 PL_lex_casemods = 0;
2485 *PL_lex_casestack = '\0';
2486 PL_lex_starts = 0;
2487 PL_lex_state = LEX_INTERPCONCAT;
ffdb8b16 2488 if (is_heredoc)
6ddcf93b 2489 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
ffdb8b16 2490 PL_copline = NOLINE;
caae0700 2491
78a635de
FC
2492 Newxz(shared, 1, LEXSHARED);
2493 shared->ls_prev = PL_parser->lex_shared;
2494 PL_parser->lex_shared = shared;
3280af22 2495
7ef70b3d 2496 PL_lex_inwhat = PL_parser->lex_sub_inwhat;
bb16bae8 2497 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
3280af22 2498 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
7ef70b3d 2499 PL_lex_inpat = PL_parser->lex_sub_op;
79072805 2500 else
5f66b61c 2501 PL_lex_inpat = NULL;
79072805 2502
3a54fd60
DM
2503 PL_parser->lex_re_reparsing = cBOOL(PL_in_eval & EVAL_RE_REPARSING);
2504 PL_in_eval &= ~EVAL_RE_REPARSING;
2505
55497cff 2506 return '(';
79072805
LW
2507}
2508
ffb4593c
NT
2509/*
2510 * S_sublex_done
2511 * Restores lexer state after a S_sublex_push.
2512 */
2513
76e3520e 2514STATIC I32
cea2e8a9 2515S_sublex_done(pTHX)
79072805 2516{
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;
275103cd 2522 pl_yylval.opval = 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);
5aa91856
FC
2533 if (PL_lex_repl) {
2534 assert (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS);
3280af22
NIS
2535 PL_linestr = PL_lex_repl;
2536 PL_lex_inpat = 0;
2537 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2538 PL_bufend += SvCUR(PL_linestr);
bd61b366 2539 PL_last_lop = PL_last_uni = NULL;
3280af22
NIS
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 }
ffdb8b16
FC
2559 if (SvTYPE(PL_linestr) >= SVt_PVNV) {
2560 CopLINE(PL_curcop) +=
9420b268 2561 ((XPVNV*)SvANY(PL_linestr))->xnv_u.xnv_lines
851b527a
FC
2562 + PL_parser->herelines;
2563 PL_parser->herelines = 0;
ffdb8b16 2564 }
9b6b7be8 2565 return '/';
ffed7fef
LW
2566 }
2567 else {
b42366d4 2568 const line_t l = CopLINE(PL_curcop);
f46d017c 2569 LEAVE;
b42366d4 2570 if (PL_multi_close == '<')
851b527a 2571 PL_parser->herelines += l - PL_multi_end;
3280af22
NIS
2572 PL_bufend = SvPVX(PL_linestr);
2573 PL_bufend += SvCUR(PL_linestr);
2574 PL_expect = XOPERATOR;
79072805 2575 return ')';
ffed7fef
LW
2576 }
2577}
2578
7a7d14f3 2579STATIC SV*
6f613c73
KW
2580S_get_and_check_backslash_N_name(pTHX_ const char* s, const char* const e)
2581{
140b12ad
KW
2582 /* <s> points to first character of interior of \N{}, <e> to one beyond the
2583 * interior, hence to the "}". Finds what the name resolves to, returning
2584 * an SV* containing it; NULL if no valid one found */
2585
dd2b1b72 2586 SV* res = newSVpvn_flags(s, e - s, UTF ? SVf_UTF8 : 0);
6f613c73 2587
0c415a79
KW
2588 HV * table;
2589 SV **cvp;
2590 SV *cv;
2591 SV *rv;
2592 HV *stash;
0c415a79
KW
2593 const char* backslash_ptr = s - 3; /* Points to the <\> of \N{... */
2594
6f613c73
KW
2595 PERL_ARGS_ASSERT_GET_AND_CHECK_BACKSLASH_N_NAME;
2596
d8d26cac 2597 if (!SvCUR(res)) {
ebcc725e 2598 SvREFCNT_dec_NN(res);
be332ba0
KW
2599 /* diag_listed_as: Unknown charname '%s' */
2600 yyerror("Unknown charname ''");
2601 return NULL;
d8d26cac 2602 }
b7e6151c 2603
107160e2
KW
2604 res = new_constant( NULL, 0, "charnames", res, NULL, backslash_ptr,
2605 /* include the <}> */
2606 e - backslash_ptr + 1);
6f613c73 2607 if (! SvPOK(res)) {
b6407c49 2608 SvREFCNT_dec_NN(res);
6f613c73
KW
2609 return NULL;
2610 }
2611
0c415a79
KW
2612 /* See if the charnames handler is the Perl core's, and if so, we can skip
2613 * the validation needed for a user-supplied one, as Perl's does its own
2614 * validation. */
2615 table = GvHV(PL_hintgv); /* ^H */
2616 cvp = hv_fetchs(table, "charnames", FALSE);
5882ddb3
FC
2617 if (cvp && (cv = *cvp) && SvROK(cv) && (rv = SvRV(cv),
2618 SvTYPE(rv) == SVt_PVCV) && ((stash = CvSTASH(rv)) != NULL))
0c415a79
KW
2619 {
2620 const char * const name = HvNAME(stash);
6a642c21
FC
2621 if (HvNAMELEN(stash) == sizeof("_charnames")-1
2622 && strEQ(name, "_charnames")) {
0c415a79
KW
2623 return res;
2624 }
2625 }
2626
bde9e88d
KW
2627 /* Here, it isn't Perl's charname handler. We can't rely on a
2628 * user-supplied handler to validate the input name. For non-ut8 input,
2629 * look to see that the first character is legal. Then loop through the
2630 * rest checking that each is a continuation */
6f613c73 2631
36897d64
KW
2632 /* This code makes the reasonable assumption that the only Latin1-range
2633 * characters that begin a character name alias are alphabetic, otherwise
2634 * would have to create a isCHARNAME_BEGIN macro */
b6ba1137 2635
b6ba1137 2636 if (! UTF) {
bde9e88d 2637 if (! isALPHAU(*s)) {
b6ba1137
KW
2638 goto bad_charname;
2639 }
bde9e88d
KW
2640 s++;
2641 while (s < e) {
2642 if (! isCHARNAME_CONT(*s)) {
b6ba1137
KW
2643 goto bad_charname;
2644 }
2d8eb851
KW
2645 if (*s == ' ' && *(s-1) == ' ') {
2646 goto multi_spaces;
bd299e29 2647 }
bde9e88d 2648 s++;
b6ba1137
KW
2649 }
2650 }
2651 else {
bde9e88d
KW
2652 /* Similarly for utf8. For invariants can check directly; for other
2653 * Latin1, can calculate their code point and check; otherwise use a
2654 * swash */
2655 if (UTF8_IS_INVARIANT(*s)) {
2656 if (! isALPHAU(*s)) {
140b12ad
KW
2657 goto bad_charname;
2658 }
bde9e88d
KW
2659 s++;
2660 } else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2661 if (! isALPHAU(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1)))) {
b6ba1137 2662 goto bad_charname;
6f613c73 2663 }
bde9e88d 2664 s += 2;
6f613c73 2665 }
bde9e88d
KW
2666 else {
2667 if (! PL_utf8_charname_begin) {
2668 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2669 PL_utf8_charname_begin = _core_swash_init("utf8",
2670 "_Perl_Charname_Begin",
2671 &PL_sv_undef,
2672 1, 0, NULL, &flags);
2673 }
2674 if (! swash_fetch(PL_utf8_charname_begin, (U8 *) s, TRUE)) {
2675 goto bad_charname;
2676 }
2677 s += UTF8SKIP(s);
2678 }
2679
2680 while (s < e) {
2681 if (UTF8_IS_INVARIANT(*s)) {
2682 if (! isCHARNAME_CONT(*s)) {
2683 goto bad_charname;
2684 }
2d8eb851
KW
2685 if (*s == ' ' && *(s-1) == ' ') {
2686 goto multi_spaces;
bd299e29 2687 }
bde9e88d
KW
2688 s++;
2689 }
2690 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
a62b247b 2691 if (! isCHARNAME_CONT(EIGHT_BIT_UTF8_TO_NATIVE(*s, *(s+1))))
bde9e88d
KW
2692 {
2693 goto bad_charname;
2694 }
2695 s += 2;
2696 }
2697 else {
2698 if (! PL_utf8_charname_continue) {
2699 U8 flags = _CORE_SWASH_INIT_ACCEPT_INVLIST;
2700 PL_utf8_charname_continue = _core_swash_init("utf8",
2701 "_Perl_Charname_Continue",
2702 &PL_sv_undef,
2703 1, 0, NULL, &flags);
2704 }
2705 if (! swash_fetch(PL_utf8_charname_continue, (U8 *) s, TRUE)) {
2706 goto bad_charname;
2707 }
2708 s += UTF8SKIP(s);
6f613c73
KW
2709 }
2710 }
2d8eb851
KW
2711 }
2712 if (*(s-1) == ' ') {
8d9d0498
FC
2713 /* diag_listed_as: charnames alias definitions may not contain
2714 trailing white-space; marked by <-- HERE in %s
2715 */
2d8eb851
KW
2716 yyerror_pv(
2717 Perl_form(aTHX_
2718 "charnames alias definitions may not contain trailing "
2719 "white-space; marked by <-- HERE in %.*s<-- HERE %.*s",
2720 (int)(s - backslash_ptr + 1), backslash_ptr,
2721 (int)(e - s + 1), s + 1
2722 ),
2723 UTF ? SVf_UTF8 : 0);
2724 return NULL;
6f613c73
KW
2725 }
2726
94ca1619 2727 if (SvUTF8(res)) { /* Don't accept malformed input */
bde9e88d
KW
2728 const U8* first_bad_char_loc;
2729 STRLEN len;
2730 const char* const str = SvPV_const(res, len);
0aab20f2
KW
2731 if (UNLIKELY(! is_utf8_string_loc((U8 *) str, len,
2732 &first_bad_char_loc)))
2733 {
75219bac
KW
2734 _force_out_malformed_utf8_message(first_bad_char_loc,
2735 (U8 *) PL_parser->bufend,
2736 0,
2737 0 /* 0 means don't die */ );
8d9d0498
FC
2738 /* diag_listed_as: Malformed UTF-8 returned by \N{%s}
2739 immediately after '%s' */
bde9e88d
KW
2740 yyerror_pv(
2741 Perl_form(aTHX_
2742 "Malformed UTF-8 returned by %.*s immediately after '%.*s'",
2743 (int) (e - backslash_ptr + 1), backslash_ptr,
2744 (int) ((char *) first_bad_char_loc - str), str
2745 ),
2746 SVf_UTF8);
2747 return NULL;
2748 }
2749 }
140b12ad 2750
bde9e88d 2751 return res;
140b12ad 2752
bde9e88d 2753 bad_charname: {
bde9e88d
KW
2754
2755 /* The final %.*s makes sure that should the trailing NUL be missing
2756 * that this print won't run off the end of the string */
8d9d0498
FC
2757 /* diag_listed_as: Invalid character in \N{...}; marked by <-- HERE
2758 in \N{%s} */
bde9e88d
KW
2759 yyerror_pv(
2760 Perl_form(aTHX_
2761 "Invalid character in \\N{...}; marked by <-- HERE in %.*s<-- HERE %.*s",
2d8eb851
KW
2762 (int)(s - backslash_ptr + 1), backslash_ptr,
2763 (int)(e - s + 1), s + 1
bde9e88d
KW
2764 ),
2765 UTF ? SVf_UTF8 : 0);
2766 return NULL;
2767 }
2d8eb851
KW
2768
2769 multi_spaces:
8d9d0498
FC
2770 /* diag_listed_as: charnames alias definitions may not contain a
2771 sequence of multiple spaces; marked by <-- HERE
2772 in %s */
2d8eb851
KW
2773 yyerror_pv(
2774 Perl_form(aTHX_
2775 "charnames alias definitions may not contain a sequence of "
2776 "multiple spaces; marked by <-- HERE in %.*s<-- HERE %.*s",
2777 (int)(s - backslash_ptr + 1), backslash_ptr,
2778 (int)(e - s + 1), s + 1
2779 ),
2780 UTF ? SVf_UTF8 : 0);
2781 return NULL;
6f613c73
KW
2782}
2783
02aa26ce
NT
2784/*
2785 scan_const
2786
9da1dd8f
DM
2787 Extracts the next constant part of a pattern, double-quoted string,
2788 or transliteration. This is terrifying code.
2789
2790 For example, in parsing the double-quoted string "ab\x63$d", it would
2791 stop at the '$' and return an OP_CONST containing 'abc'.
02aa26ce 2792
94def140 2793 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
3280af22 2794 processing a pattern (PL_lex_inpat is true), a transliteration
94def140 2795 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
02aa26ce 2796
94def140
ST
2797 Returns a pointer to the character scanned up to. If this is
2798 advanced from the start pointer supplied (i.e. if anything was
9da1dd8f 2799 successfully parsed), will leave an OP_CONST for the substring scanned
6154021b 2800 in pl_yylval. Caller must intuit reason for not parsing further
9b599b2a
GS
2801 by looking at the next characters herself.
2802
02aa26ce 2803 In patterns:
9da1dd8f 2804 expand:
537124e4
KW
2805 \N{FOO} => \N{U+hex_for_character_FOO}
2806 (if FOO expands to multiple characters, expands to \N{U+xx.XX.yy ...})
9da1dd8f
DM
2807
2808 pass through:
2809 all other \-char, including \N and \N{ apart from \N{ABC}
2810
2811 stops on:
2812 @ and $ where it appears to be a var, but not for $ as tail anchor
2813 \l \L \u \U \Q \E
2814 (?{ or (??{
2815
02aa26ce
NT
2816 In transliterations:
2817 characters are VERY literal, except for - not at the start or end
66b09263
KW
2818 of the string, which indicates a range. However some backslash sequences
2819 are recognized: \r, \n, and the like
2820 \007 \o{}, \x{}, \N{}
2821 If all elements in the transliteration are below 256,
94def140
ST
2822 scan_const expands the range to the full set of intermediate
2823 characters. If the range is in utf8, the hyphen is replaced with
2824 a certain range mark which will be handled by pmtrans() in op.c.
02aa26ce
NT
2825
2826 In double-quoted strings:
2827 backslashes:
66b09263 2828 all those recognized in transliterations
94def140 2829 deprecated backrefs: \1 (in substitution replacements)
02aa26ce
NT
2830 case and quoting: \U \Q \E
2831 stops on @ and $
2832
2833 scan_const does *not* construct ops to handle interpolated strings.
2834 It stops processing as soon as it finds an embedded $ or @ variable
2835 and leaves it to the caller to work out what's going on.
2836
94def140
ST
2837 embedded arrays (whether in pattern or not) could be:
2838 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2839
2840 $ in double-quoted strings must be the symbol of an embedded scalar.
02aa26ce
NT
2841
2842 $ in pattern could be $foo or could be tail anchor. Assumption:
2843 it's a tail anchor if $ is the last thing in the string, or if it's
94def140 2844 followed by one of "()| \r\n\t"
02aa26ce 2845
9da1dd8f 2846 \1 (backreferences) are turned into $1 in substitutions
02aa26ce
NT
2847
2848 The structure of the code is
2849 while (there's a character to process) {
94def140
ST
2850 handle transliteration ranges
2851 skip regexp comments /(?#comment)/ and codes /(?{code})/
2852 skip #-initiated comments in //x patterns
2853 check for embedded arrays
02aa26ce
NT
2854 check for embedded scalars
2855 if (backslash) {
94def140 2856 deprecate \1 in substitution replacements
02aa26ce
NT
2857 handle string-changing backslashes \l \U \Q \E, etc.
2858 switch (what was escaped) {
94def140 2859 handle \- in a transliteration (becomes a literal -)
ff3f963a 2860 if a pattern and not \N{, go treat as regular character
94def140
ST
2861 handle \132 (octal characters)
2862 handle \x15 and \x{1234} (hex characters)
ff3f963a 2863 handle \N{name} (named characters, also \N{3,5} in a pattern)
94def140
ST
2864 handle \cV (control characters)
2865 handle printf-style backslashes (\f, \r, \n, etc)
02aa26ce 2866 } (end switch)
77a135fe 2867 continue
02aa26ce 2868 } (end if backslash)
77a135fe 2869 handle regular character
02aa26ce 2870 } (end while character to read)
02cd137d 2871
02aa26ce
NT
2872*/
2873
76e3520e 2874STATIC char *
cea2e8a9 2875S_scan_const(pTHX_ char *start)
79072805 2876{
eb578fdb 2877 char *send = PL_bufend; /* end of the constant */
dc023dbb
KW
2878 SV *sv = newSV(send - start); /* sv for the constant. See note below
2879 on sizing. */
eb578fdb
KW
2880 char *s = start; /* start of the constant */
2881 char *d = SvPVX(sv); /* destination for copies */
dc023dbb
KW
2882 bool dorange = FALSE; /* are we in a translit range? */
2883 bool didrange = FALSE; /* did we just finish a range? */
2884 bool in_charclass = FALSE; /* within /[...]/ */
2885 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2886 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed to be
2887 UTF8? But, this can show as true
2888 when the source isn't utf8, as for
2889 example when it is entirely composed
2890 of hex constants */
af9be36c
KW
2891 STRLEN utf8_variant_count = 0; /* When not in UTF-8, this counts the
2892 number of characters found so far
2893 that will expand (into 2 bytes)
2894 should we have to convert to
2895 UTF-8) */
6f613c73 2896 SV *res; /* result from charnames */
f4240379
KW
2897 STRLEN offset_to_max; /* The offset in the output to where the range
2898 high-end character is temporarily placed */
77a135fe 2899
fe2ba0a2
KW
2900 /* Does something require special handling in tr/// ? This avoids extra
2901 * work in a less likely case. As such, khw didn't feel it was worth
2902 * adding any branches to the more mainline code to handle this, which
2903 * means that this doesn't get set in some circumstances when things like
2904 * \x{100} get expanded out. As a result there needs to be extra testing
2905 * done in the tr code */
2906 bool has_above_latin1 = FALSE;
2907
77a135fe
KW
2908 /* Note on sizing: The scanned constant is placed into sv, which is
2909 * initialized by newSV() assuming one byte of output for every byte of
2910 * input. This routine expects newSV() to allocate an extra byte for a
2911 * trailing NUL, which this routine will append if it gets to the end of
2912 * the input. There may be more bytes of input than output (eg., \N{LATIN
2913 * CAPITAL LETTER A}), or more output than input if the constant ends up
2914 * recoded to utf8, but each time a construct is found that might increase
2915 * the needed size, SvGROW() is called. Its size parameter each time is
2916 * based on the best guess estimate at the time, namely the length used so
2917 * far, plus the length the current construct will occupy, plus room for
caae0700 2918 * the trailing NUL, plus one byte for every input byte still unscanned */
77a135fe 2919
c3320c2a
KW
2920 UV uv = UV_MAX; /* Initialize to weird value to try to catch any uses
2921 before set */
4c3a8340 2922#ifdef EBCDIC
f4240379
KW
2923 int backslash_N = 0; /* ? was the character from \N{} */
2924 int non_portable_endpoint = 0; /* ? In a range is an endpoint
2925 platform-specific like \x65 */
4c3a8340 2926#endif
012bcf8d 2927
7918f24d
NC
2928 PERL_ARGS_ASSERT_SCAN_CONST;
2929
bb16bae8 2930 assert(PL_lex_inwhat != OP_TRANSR);
7ef70b3d 2931 if (PL_lex_inwhat == OP_TRANS && PL_parser->lex_sub_op) {
2b9d42f0 2932 /* If we are doing a trans and we know we want UTF8 set expectation */
7ef70b3d
FC
2933 has_utf8 = PL_parser->lex_sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2934 this_utf8 = PL_parser->lex_sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2b9d42f0
NIS
2935 }
2936
b899e89d
FC
2937 /* Protect sv from errors and fatal warnings. */
2938 ENTER_with_name("scan_const");
2939 SAVEFREESV(sv);
2b9d42f0 2940
f4240379
KW
2941 while (s < send
2942 || dorange /* Handle tr/// range at right edge of input */
2943 ) {
ff3f963a 2944
02aa26ce 2945 /* get transliterations out of the way (they're most literal) */
3280af22 2946 if (PL_lex_inwhat == OP_TRANS) {
02aa26ce 2947
f4240379
KW
2948 /* But there isn't any special handling necessary unless there is a
2949 * range, so for most cases we just drop down and handle the value
2950 * as any other. There are two exceptions.
2951 *
02cd137d
KW
2952 * 1. A hyphen indicates that we are actually going to have a
2953 * range. In this case, skip the '-', set a flag, then drop
f4240379
KW
2954 * down to handle what should be the end range value.
2955 * 2. After we've handled that value, the next time through, that
2956 * flag is set and we fix up the range.
2957 *
2958 * Ranges entirely within Latin1 are expanded out entirely, in
188d22cf
KW
2959 * order to make the transliteration a simple table look-up.
2960 * Ranges that extend above Latin1 have to be done differently, so
2961 * there is no advantage to expanding them here, so they are
2962 * stored here as Min, ILLEGAL_UTF8_BYTE, Max. The illegal byte
2963 * signifies a hyphen without any possible ambiguity. On EBCDIC
2964 * machines, if the range is expressed as Unicode, the Latin1
2965 * portion is expanded out even if the range extends above
2966 * Latin1. This is because each code point in it has to be
2967 * processed here individually to get its native translation */
f4240379
KW
2968
2969 if (! dorange) {
2970
02cd137d
KW
2971 /* Here, we don't think we're in a range. If the new character
2972 * is not a hyphen; or if it is a hyphen, but it's too close to
e8d55f27
TC
2973 * either edge to indicate a range, or if we haven't output any
2974 * characters yet then it's a regular character. */
2975 if (*s != '-' || s >= send - 1 || s == start || d == SvPVX(sv)) {
f4240379
KW
2976
2977 /* A regular character. Process like any other, but first
2978 * clear any flags */
2979 didrange = FALSE;
2980 dorange = FALSE;
e294cc5d 2981#ifdef EBCDIC
f4240379
KW
2982 non_portable_endpoint = 0;
2983 backslash_N = 0;
e294cc5d 2984#endif
02cd137d
KW
2985 /* The tests here for being above Latin1 and similar ones
2986 * in the following 'else' suffice to find all such
2987 * occurences in the constant, except those added by a
fe2ba0a2 2988 * backslash escape sequence, like \x{100}. Mostly, those
02cd137d 2989 * set 'has_above_latin1' as appropriate */
188d22cf
KW
2990 if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
2991 has_above_latin1 = TRUE;
2992 }
2993
f4240379
KW
2994 /* Drops down to generic code to process current byte */
2995 }
02cd137d 2996 else { /* Is a '-' in the context where it means a range */
f4240379 2997 if (didrange) { /* Something like y/A-C-Z// */
02cd137d
KW
2998 Perl_croak(aTHX_ "Ambiguous range in transliteration"
2999 " operator");
f4240379 3000 }
e294cc5d 3001
f4240379 3002 dorange = TRUE;
2b9d42f0 3003
02cd137d 3004 s++; /* Skip past the hyphen */
f4240379
KW
3005
3006 /* d now points to where the end-range character will be
3007 * placed. Save it so won't have to go finding it later,
3008 * and drop down to get that character. (Actually we
3009 * instead save the offset, to handle the case where a
3010 * realloc in the meantime could change the actual
3011 * pointer). We'll finish processing the range the next
3012 * time through the loop */
3013 offset_to_max = d - SvPVX_const(sv);
188d22cf
KW
3014
3015 if (this_utf8 && UTF8_IS_ABOVE_LATIN1(*s)) {
3016 has_above_latin1 = TRUE;
3017 }
02cd137d
KW
3018
3019 /* Drops down to generic code to process current byte */
f4240379
KW
3020 }
3021 } /* End of not a range */
3022 else {
3023 /* Here we have parsed a range. Now must handle it. At this
3024 * point:
3025 * 'sv' is a SV* that contains the output string we are
3026 * constructing. The final two characters in that string
3027 * are the range start and range end, in order.
3028 * 'd' points to just beyond the range end in the 'sv' string,
3029 * where we would next place something
3030 * 'offset_to_max' is the offset in 'sv' at which the character
02cd137d 3031 * (the range's maximum end point) before 'd' begins.
f4240379 3032 */
8efef67c 3033 char * max_ptr = SvPVX(sv) + offset_to_max;
5ca37d54 3034 char * min_ptr;
f4240379
KW
3035 IV range_min;
3036 IV range_max; /* last character in range */
f4240379 3037 STRLEN grow;
5ca37d54
KW
3038 Size_t offset_to_min = 0;
3039 Size_t extras = 0;
11327fa1 3040#ifdef EBCDIC
f4240379
KW
3041 bool convert_unicode;
3042 IV real_range_max = 0;
e294cc5d 3043#endif
02cd137d 3044 /* Get the code point values of the range ends. */
e294cc5d 3045 if (has_utf8) {
f4240379
KW
3046 /* We know the utf8 is valid, because we just constructed
3047 * it ourselves in previous loop iterations */
3048 min_ptr = (char*) utf8_hop( (U8*) max_ptr, -1);
3049 range_min = valid_utf8_to_uvchr( (U8*) min_ptr, NULL);
3050 range_max = valid_utf8_to_uvchr( (U8*) max_ptr, NULL);
fe2ba0a2
KW
3051
3052 /* This compensates for not all code setting
3053 * 'has_above_latin1', so that we don't skip stuff that
3054 * should be executed */
3055 if (range_max > 255) {
3056 has_above_latin1 = TRUE;
3057 }
e294cc5d 3058 }
f4240379
KW
3059 else {
3060 min_ptr = max_ptr - 1;
3061 range_min = * (U8*) min_ptr;
3062 range_max = * (U8*) max_ptr;
3063 }
3064
8efef67c
KW
3065 /* If the range is just a single code point, like tr/a-a/.../,
3066 * that code point is already in the output, twice. We can
3067 * just back up over the second instance and avoid all the rest
3068 * of the work. But if it is a variant character, it's been
218304f9
KW
3069 * counted twice, so decrement. (This unlikely scenario is
3070 * special cased, like the one for a range of 2 code points
3071 * below, only because the main-line code below needs a range
3072 * of 3 or more to work without special casing. Might as well
3073 * get it out of the way now.) */
8efef67c
KW
3074 if (UNLIKELY(range_max == range_min)) {
3075 d = max_ptr;
3076 if (! has_utf8 && ! UVCHR_IS_INVARIANT(range_max)) {
3077 utf8_variant_count--;
3078 }
3079 goto range_done;
3080 }
3081
e294cc5d 3082#ifdef EBCDIC
f4240379
KW
3083 /* On EBCDIC platforms, we may have to deal with portable
3084 * ranges. These happen if at least one range endpoint is a
3085 * Unicode value (\N{...}), or if the range is a subset of
3086 * [A-Z] or [a-z], and both ends are literal characters,
3087 * like 'A', and not like \x{C1} */
87dd6ea7 3088 convert_unicode =
02cd137d
KW
3089 cBOOL(backslash_N) /* \N{} forces Unicode,
3090 hence portable range */
3091 || ( ! non_portable_endpoint
3092 && (( isLOWER_A(range_min) && isLOWER_A(range_max))
3093 || (isUPPER_A(range_min) && isUPPER_A(range_max))));
87dd6ea7 3094 if (convert_unicode) {
f4240379
KW
3095
3096 /* Special handling is needed for these portable ranges.
02cd137d
KW
3097 * They are defined to be in Unicode terms, which includes
3098 * all the Unicode code points between the end points.
f4240379
KW
3099 * Convert to Unicode to get the Unicode range. Later we
3100 * will convert each code point in the range back to
3101 * native. */
3102 range_min = NATIVE_TO_UNI(range_min);
3103 range_max = NATIVE_TO_UNI(range_max);
3104 }
e294cc5d 3105#endif
8ada0baa 3106
f4240379 3107 if (range_min > range_max) {
11327fa1 3108#ifdef EBCDIC
f4240379
KW
3109 if (convert_unicode) {
3110 /* Need to convert back to native for meaningful
3111 * messages for this platform */
3112 range_min = UNI_TO_NATIVE(range_min);
3113 range_max = UNI_TO_NATIVE(range_max);
3114 }
11327fa1 3115#endif
f4240379
KW
3116 /* Use the characters themselves for the error message if
3117 * ASCII printables; otherwise some visible representation
3118 * of them */
3119 if (isPRINT_A(range_min) && isPRINT_A(range_max)) {
3120 Perl_croak(aTHX_
3121 "Invalid range \"%c-%c\" in transliteration operator",
3122 (char)range_min, (char)range_max);
3123 }
11327fa1 3124#ifdef EBCDIC
f4240379 3125 else if (convert_unicode) {
02cd137d 3126 /* diag_listed_as: Invalid range "%s" in transliteration operator */
f4240379 3127 Perl_croak(aTHX_
02cd137d
KW
3128 "Invalid range \"\\N{U+%04" UVXf "}-\\N{U+%04"
3129 UVXf "}\" in transliteration operator",
3130 range_min, range_max);
f4240379 3131 }
11327fa1 3132#endif
f4240379 3133 else {
02cd137d 3134 /* diag_listed_as: Invalid range "%s" in transliteration operator */
f4240379 3135 Perl_croak(aTHX_
02cd137d
KW
3136 "Invalid range \"\\x{%04" UVXf "}-\\x{%04" UVXf "}\""
3137 " in transliteration operator",
3138 range_min, range_max);
f4240379 3139 }
c2e66d9e
GS
3140 }
3141
aca41667
KW
3142 /* If the range is exactly two code points long, they are
3143 * already both in the output */
3144 if (UNLIKELY(range_min + 1 == range_max)) {
3145 goto range_done;
3146 }
3147
3148 /* Here the range contains at least 3 code points */
3149
f4240379
KW
3150 if (has_utf8) {
3151
188d22cf
KW
3152 /* If everything in the transliteration is below 256, we
3153 * can avoid special handling later. A translation table
02cd137d
KW
3154 * for each of those bytes is created by op.c. So we
3155 * expand out all ranges to their constituent code points.
3156 * But if we've encountered something above 255, the
3157 * expanding won't help, so skip doing that. But if it's
3158 * EBCDIC, we may have to look at each character below 256
3159 * if we have to convert to/from Unicode values */
188d22cf 3160 if ( has_above_latin1
c7f1f016 3161#ifdef EBCDIC
f4240379 3162 && (range_min > 255 || ! convert_unicode)
8ada0baa 3163#endif
f4240379
KW
3164 ) {
3165 /* Move the high character one byte to the right; then
3166 * insert between it and the range begin, an illegal
3167 * byte which serves to indicate this is a range (using
02cd137d 3168 * a '-' would be ambiguous). */
f4240379
KW
3169 char *e = d++;
3170 while (e-- > max_ptr) {
3171 *(e + 1) = *e;
e294cc5d 3172 }
f4240379
KW
3173 *(e + 1) = (char) ILLEGAL_UTF8_BYTE;
3174 goto range_done;
3175 }
3176
3177 /* Here, we're going to expand out the range. For EBCDIC
3178 * the range can extend above 255 (not so in ASCII), so
3179 * for EBCDIC, split it into the parts above and below
3180 * 255/256 */
e294cc5d 3181#ifdef EBCDIC
f4240379
KW
3182 if (range_max > 255) {
3183 real_range_max = range_max;
3184 range_max = 255;
3185 }
e294cc5d 3186#endif
f4240379 3187 }
02aa26ce 3188
f4240379 3189 /* Here we need to expand out the string to contain each
5ca37d54
KW
3190 * character in the range. Grow the output to handle this.
3191 * For non-UTF8, we need a byte for each code point in the
3192 * range, minus the three that we've already allocated for: the
3193 * hyphen, the min, and the max. For UTF-8, we need this
3194 * plus an extra byte for each code point that occupies two
3195 * bytes (is variant) when in UTF-8 (except we've already
3196 * allocated for the end points, including if they are
3197 * variants). For ASCII platforms and Unicode ranges on EBCDIC
3198 * platforms, it's easy to calculate a precise number. To
3199 * start, we count the variants in the range, which we need
3200 * elsewhere in this function anyway. (For the case where it
3201 * isn't easy to calculate, 'extras' has been initialized to 0,
3202 * and the calculation is done in a loop further down.) */
3203#ifdef EBCDIC
3204 if (convert_unicode)
3205#endif
3206 {
3207 /* This is executed unconditionally on ASCII, and for
3208 * Unicode ranges on EBCDIC. Under these conditions, all
3209 * code points above a certain value are variant; and none
3210 * under that value are. We just need to find out how much
3211 * of the range is above that value. We don't count the
3212 * end points here, as they will already have been counted
3213 * as they were parsed. */
3214 if (range_min >= UTF_CONTINUATION_MARK) {
3215
3216 /* The whole range is made up of variants */
3217 extras = (range_max - 1) - (range_min + 1) + 1;
3218 }
3219 else if (range_max >= UTF_CONTINUATION_MARK) {
f4240379 3220
5ca37d54
KW
3221 /* Only the higher portion of the range is variants */
3222 extras = (range_max - 1) - UTF_CONTINUATION_MARK + 1;
3223 }
f4240379 3224
5ca37d54
KW
3225 utf8_variant_count += extras;
3226 }
3227
3228 /* The base growth is the number of code points in the range,
3229 * not including the endpoints, which have already been sized
3230 * for (and output). We don't subtract for the hyphen, as it
3231 * has been parsed but not output, and the SvGROW below is
3232 * based only on what's been output plus what's left to parse.
3233 * */
3234 grow = (range_max - 1) - (range_min + 1) + 1;
f4240379 3235
5ca37d54 3236 if (has_utf8) {
4c3a8340 3237#ifdef EBCDIC
5ca37d54
KW
3238 /* In some cases in EBCDIC, we haven't yet calculated a
3239 * precise amount needed for the UTF-8 variants. Just
3240 * assume the worst case, that everything will expand by a
3241 * byte */
3242 if (! convert_unicode) {
f4240379
KW
3243 grow *= 2;
3244 }
5ca37d54 3245 else
4c3a8340 3246#endif
5ca37d54
KW
3247 {
3248 /* Otherwise we know exactly how many variants there
3249 * are in the range. */
3250 grow += extras;
3251 }
f4240379
KW
3252 }
3253
5ca37d54
KW
3254 /* Grow, but position the output to overwrite the range min end
3255 * point, because in some cases we overwrite that */
3256 SvCUR_set(sv, d - SvPVX_const(sv));
3257 offset_to_min = min_ptr - SvPVX_const(sv);
3258
3259 /* See Note on sizing above. */
3260 d = offset_to_min + SvGROW(sv, SvCUR(sv)
3261 + (send - s)
3262 + grow
3263 + 1 /* Trailing NUL */ );
f4240379 3264
5ca37d54 3265 /* Now, we can expand out the range. */
11327fa1 3266#ifdef EBCDIC
f4240379 3267 if (convert_unicode) {
5ca37d54 3268 SSize_t i;
02aa26ce 3269
f4240379
KW
3270 /* Recall that the min and max are now in Unicode terms, so
3271 * we have to convert each character to its native
3272 * equivalent */
3273 if (has_utf8) {
3274 for (i = range_min; i <= range_max; i++) {
02cd137d
KW
3275 append_utf8_from_native_byte(
3276 LATIN1_TO_NATIVE((U8) i),
3277 (U8 **) &d);
f4240379
KW
3278 }
3279 }
3280 else {
3281 for (i = range_min; i <= range_max; i++) {
81324705 3282 *d++ = (char)LATIN1_TO_NATIVE((U8) i);
f4240379
KW
3283 }
3284 }
01ec43d0 3285 }
11327fa1
AL
3286 else
3287#endif
3288 /* Always gets run for ASCII, and sometimes for EBCDIC. */
3289 {
f4240379
KW
3290 /* Here, no conversions are necessary, which means that the
3291 * first character in the range is already in 'd' and
3292 * valid, so we can skip overwriting it */
3293 if (has_utf8) {
19742f39 3294 SSize_t i;
f4240379
KW
3295 d += UTF8SKIP(d);
3296 for (i = range_min + 1; i <= range_max; i++) {
81324705 3297 append_utf8_from_native_byte((U8) i, (U8 **) &d);
f4240379
KW
3298 }
3299 }
3300 else {
19742f39 3301 SSize_t i;
f4240379 3302 d++;
5ca37d54
KW
3303 assert(range_min + 1 <= range_max);
3304 for (i = range_min + 1; i < range_max; i++) {
3305#ifdef EBCDIC
3306 /* In this case on EBCDIC, we haven't calculated
3307 * the variants. Do it here, as we go along */
3308 if (! UVCHR_IS_INVARIANT(i)) {
3309 utf8_variant_count++;
3310 }
3311#endif
f4240379
KW
3312 *d++ = (char)i;
3313 }
5ca37d54
KW
3314
3315 /* The range_max is done outside the loop so as to
3316 * avoid having to special case not incrementing
3317 * 'utf8_variant_count' on EBCDIC (it's already been
3318 * counted when originally parsed) */
3319 *d++ = (char) range_max;
f4240379 3320 }
a0ed51b3 3321 }
02aa26ce 3322
11327fa1 3323#ifdef EBCDIC
02cd137d
KW
3324 /* If the original range extended above 255, add in that
3325 * portion. */
f4240379
KW
3326 if (real_range_max) {
3327 *d++ = (char) UTF8_TWO_BYTE_HI(0x100);
3328 *d++ = (char) UTF8_TWO_BYTE_LO(0x100);
043af6f6
KW
3329 if (real_range_max > 0x100) {
3330 if (real_range_max > 0x101) {
3331 *d++ = (char) ILLEGAL_UTF8_BYTE;
3332 }
f4240379 3333 d = (char*)uvchr_to_utf8((U8*)d, real_range_max);
043af6f6 3334 }
f4240379 3335 }
11327fa1 3336#endif
02aa26ce 3337
f4240379
KW
3338 range_done:
3339 /* mark the range as done, and continue */
3340 didrange = TRUE;
3341 dorange = FALSE;
3342#ifdef EBCDIC
3343 non_portable_endpoint = 0;
3344 backslash_N = 0;
3345#endif
3346 continue;
3347 } /* End of is a range */
3348 } /* End of transliteration. Joins main code after these else's */
e4a2df84
DM
3349 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
3350 char *s1 = s-1;
3351 int esc = 0;
3352 while (s1 >= start && *s1-- == '\\')
3353 esc = !esc;
3354 if (!esc)
3355 in_charclass = TRUE;
3356 }
1e02a175 3357 else if (*s == ']' && PL_lex_inpat && in_charclass) {
e4a2df84
DM
3358 char *s1 = s-1;
3359 int esc = 0;
3360 while (s1 >= start && *s1-- == '\\')
3361 esc = !esc;
3362 if (!esc)
3363 in_charclass = FALSE;
3364 }
02cd137d
KW
3365 /* skip for regexp comments /(?#comment)/, except for the last
3366 * char, which will be done separately. Stop on (?{..}) and
3367 * friends */
c30fc27b 3368 else if (*s == '(' && PL_lex_inpat && s[1] == '?' && !in_charclass) {
cc6b7395 3369 if (s[2] == '#') {
e994fd66 3370 while (s+1 < send && *s != ')')
5ff03569 3371 *d++ = *s++;
155aba94 3372 }
407f8cf2
KW
3373 else if (!PL_lex_casemods
3374 && ( s[2] == '{' /* This should match regcomp.c */
3375 || (s[2] == '?' && s[3] == '{')))
155aba94 3376 {
9da1dd8f 3377 break;
cc6b7395 3378 }
748a9306 3379 }
02cd137d 3380 /* likewise skip #-initiated comments in //x patterns */
407f8cf2
KW
3381 else if (*s == '#'
3382 && PL_lex_inpat
3383 && !in_charclass
3384 && ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED)
3385 {
8faf4f30 3386 while (s < send && *s != '\n')
5ff03569 3387 *d++ = *s++;
748a9306 3388 }
02cd137d 3389 /* no further processing of single-quoted regex */
9da1dd8f
DM
3390 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
3391 goto default_action;
3392
02cd137d
KW
3393 /* check for embedded arrays
3394 * (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
3395 */
1749ea0d 3396 else if (*s == '@' && s[1]) {
fac0f7a3
KW
3397 if (UTF
3398 ? isIDFIRST_utf8_safe(s+1, send)
3399 : isWORDCHAR_A(s[1]))
3400 {
1749ea0d 3401 break;
fac0f7a3 3402 }
1749ea0d
ST
3403 if (strchr(":'{$", s[1]))
3404 break;
3405 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
3406 break; /* in regexp, neither @+ nor @- are interpolated */
3407 }
02cd137d
KW
3408 /* check for embedded scalars. only stop if we're sure it's a
3409 * variable. */
79072805 3410 else if (*s == '$') {
3280af22 3411 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
79072805 3412 break;
77772344 3413 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
a2a5de95
NC
3414 if (s[1] == '\\') {
3415 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
3416 "Possible unintended interpolation of $\\ in regex");
77772344 3417 }
79072805 3418 break; /* in regexp, $ might be tail anchor */
77772344 3419 }
79072805 3420 }
02aa26ce 3421
2b9d42f0
NIS
3422 /* End of else if chain - OP_TRANS rejoin rest */
3423
8faf4f30
HS
3424 if (UNLIKELY(s >= send)) {
3425 assert(s == send);
3426 break;
3427 }
3428
02aa26ce 3429 /* backslashes */
79072805 3430 if (*s == '\\' && s+1 < send) {
ff3f963a
KW
3431 char* e; /* Can be used for ending '}', etc. */
3432
79072805 3433 s++;
02aa26ce 3434
7d0fc23c
KW
3435 /* warn on \1 - \9 in substitution replacements, but note that \11
3436 * is an octal; and \19 is \1 followed by '9' */
407f8cf2
KW
3437 if (PL_lex_inwhat == OP_SUBST
3438 && !PL_lex_inpat
3439 && isDIGIT(*s)
3440 && *s != '0'
3441 && !isDIGIT(s[1]))
79072805 3442 {
c782d7ee 3443 /* diag_listed_as: \%d better written as $%d */
a2a5de95 3444 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
79072805
LW
3445 *--s = '$';
3446 break;
3447 }
02aa26ce
NT
3448
3449 /* string-change backslash escapes */
838f2281 3450 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
79072805
LW
3451 --s;
3452 break;
3453 }
ff3f963a
KW
3454 /* In a pattern, process \N, but skip any other backslash escapes.
3455 * This is because we don't want to translate an escape sequence
3456 * into a meta symbol and have the regex compiler use the meta
3457 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3458 * in spite of this, we do have to process \N here while the proper
3459 * charnames handler is in scope. See bugs #56444 and #62056.
85fba779 3460 *
ff3f963a
KW
3461 * There is a complication because \N in a pattern may also stand
3462 * for 'match a non-nl', and not mean a charname, in which case its
3463 * processing should be deferred to the regex compiler. To be a
3464 * charname it must be followed immediately by a '{', and not look
3465 * like \N followed by a curly quantifier, i.e., not something like
3466 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3467 * quantifier */
3468 else if (PL_lex_inpat
3469 && (*s != 'N'
3470 || s[1] != '{'
412f55bb 3471 || regcurly(s + 1)))
ff3f963a 3472 {
4d73d076 3473 *d++ = '\\';
cc74c5bd
ST
3474 goto default_action;
3475 }
02aa26ce 3476
79072805 3477 switch (*s) {
79072805 3478 default:
11b8faa4 3479 {
15861f94 3480 if ((isALPHANUMERIC(*s)))
a2a5de95
NC
3481 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3482 "Unrecognized escape \\%c passed through",
3483 *s);
11b8faa4 3484 /* default action is to copy the quoted character */
f9a63242 3485 goto default_action;
11b8faa4 3486 }
02aa26ce 3487
632403cc 3488 /* eg. \132 indicates the octal constant 0132 */
79072805
LW
3489 case '0': case '1': case '2': case '3':
3490 case '4': case '5': case '6': case '7':
ba210ebe 3491 {
5e0a247b 3492 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
53305cf1 3493 STRLEN len = 3;
06972766 3494 uv = grok_oct(s, &len, &flags, NULL);
ba210ebe 3495 s += len;
5e0a247b
KW
3496 if (len < 3 && s < send && isDIGIT(*s)
3497 && ckWARN(WARN_MISC))
3498 {
3499 Perl_warner(aTHX_ packWARN(WARN_MISC),
3500 "%s", form_short_octal_warning(s, len));
3501 }
ba210ebe 3502 }
012bcf8d 3503 goto NUM_ESCAPE_INSERT;
02aa26ce 3504
f0a2b745
KW
3505 /* eg. \o{24} indicates the octal constant \024 */
3506 case 'o':
3507 {
454155d9 3508 const char* error;
f0a2b745 3509
00ce5563 3510 bool valid = grok_bslash_o(&s, &uv, &error,
80f4111b
KW
3511 TRUE, /* Output warning */
3512 FALSE, /* Not strict */
17896a85
KW
3513 TRUE, /* Output warnings for
3514 non-portables */
80f4111b 3515 UTF);
454155d9 3516 if (! valid) {
f0a2b745 3517 yyerror(error);
3dd4eaeb 3518 uv = 0; /* drop through to ensure range ends are set */
f0a2b745
KW
3519 }
3520 goto NUM_ESCAPE_INSERT;
3521 }
3522
77a135fe 3523 /* eg. \x24 indicates the hex constant 0x24 */
79072805 3524 case 'x':
a0481293 3525 {
a0481293 3526 const char* error;
355860ce 3527
00ce5563 3528 bool valid = grok_bslash_x(&s, &uv, &error,
80f4111b
KW
3529 TRUE, /* Output warning */
3530 FALSE, /* Not strict */
17896a85
KW
3531 TRUE, /* Output warnings for
3532 non-portables */
80f4111b 3533 UTF);
a0481293
KW
3534 if (! valid) {
3535 yyerror(error);
3dd4eaeb 3536 uv = 0; /* drop through to ensure range ends are set */
ba210ebe 3537 }
012bcf8d
GS
3538 }
3539
3540 NUM_ESCAPE_INSERT:
b67fd2c5 3541 /* Insert oct or hex escaped character. */
02cd137d 3542
06972766 3543 /* Here uv is the ordinal of the next character being added */
f4240379
KW
3544 if (UVCHR_IS_INVARIANT(uv)) {
3545 *d++ = (char) uv;
3546 }
3547 else {
9aa983d2 3548 if (!has_utf8 && uv > 255) {
caae0700 3549
af9be36c
KW
3550 /* Here, 'uv' won't fit unless we convert to UTF-8.
3551 * If we've only seen invariants so far, all we have to
3552 * do is turn on the flag */
3553 if (utf8_variant_count == 0) {
3554 SvUTF8_on(sv);
3555 }
3556 else {
1e02a175
KW
3557 SvCUR_set(sv, d - SvPVX_const(sv));
3558 SvPOK_on(sv);
3559 *d = '\0';