Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | /* pp.c |
79072805 | 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 | |
79072805 | 5 | * |
a0d0e21e 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. | |
79072805 | 8 | * |
a0d0e21e LW |
9 | */ |
10 | ||
11 | /* | |
4ac71550 TC |
12 | * 'It's a big house this, and very peculiar. Always a bit more |
13 | * to discover, and no knowing what you'll find round a corner. | |
14 | * And Elves, sir!' --Samwise Gamgee | |
15 | * | |
16 | * [p.225 of _The Lord of the Rings_, II/i: "Many Meetings"] | |
a0d0e21e | 17 | */ |
79072805 | 18 | |
166f8a29 DM |
19 | /* This file contains general pp ("push/pop") functions that execute the |
20 | * opcodes that make up a perl program. A typical pp function expects to | |
21 | * find its arguments on the stack, and usually pushes its results onto | |
22 | * the stack, hence the 'pp' terminology. Each OP structure contains | |
23 | * a pointer to the relevant pp_foo() function. | |
24 | */ | |
25 | ||
79072805 | 26 | #include "EXTERN.h" |
864dbfa3 | 27 | #define PERL_IN_PP_C |
79072805 | 28 | #include "perl.h" |
77bc9082 | 29 | #include "keywords.h" |
79072805 | 30 | |
a4af207c JH |
31 | #include "reentr.h" |
32 | ||
dfe9444c AD |
33 | /* XXX I can't imagine anyone who doesn't have this actually _needs_ |
34 | it, since pid_t is an integral type. | |
35 | --AD 2/20/1998 | |
36 | */ | |
37 | #ifdef NEED_GETPID_PROTO | |
38 | extern Pid_t getpid (void); | |
8ac85365 NIS |
39 | #endif |
40 | ||
0630166f SP |
41 | /* |
42 | * Some BSDs and Cygwin default to POSIX math instead of IEEE. | |
43 | * This switches them over to IEEE. | |
44 | */ | |
45 | #if defined(LIBM_LIB_VERSION) | |
46 | _LIB_VERSION_TYPE _LIB_VERSION = _IEEE_; | |
47 | #endif | |
48 | ||
13017935 SM |
49 | /* variations on pp_null */ |
50 | ||
93a17b20 LW |
51 | PP(pp_stub) |
52 | { | |
97aff369 | 53 | dVAR; |
39644a26 | 54 | dSP; |
54310121 | 55 | if (GIMME_V == G_SCALAR) |
3280af22 | 56 | XPUSHs(&PL_sv_undef); |
93a17b20 LW |
57 | RETURN; |
58 | } | |
59 | ||
79072805 LW |
60 | /* Pushy stuff. */ |
61 | ||
93a17b20 LW |
62 | PP(pp_padav) |
63 | { | |
97aff369 | 64 | dVAR; dSP; dTARGET; |
13017935 | 65 | I32 gimme; |
e190e9b4 | 66 | assert(SvTYPE(TARG) == SVt_PVAV); |
533c011a | 67 | if (PL_op->op_private & OPpLVAL_INTRO) |
a5911867 RGS |
68 | if (!(PL_op->op_private & OPpPAD_STATE)) |
69 | SAVECLEARSV(PAD_SVl(PL_op->op_targ)); | |
85e6fe83 | 70 | EXTEND(SP, 1); |
533c011a | 71 | if (PL_op->op_flags & OPf_REF) { |
85e6fe83 | 72 | PUSHs(TARG); |
93a17b20 | 73 | RETURN; |
40c94d11 FC |
74 | } else if (PL_op->op_private & OPpMAYBE_LVSUB) { |
75 | const I32 flags = is_lvalue_sub(); | |
76 | if (flags && !(flags & OPpENTERSUB_INARGS)) { | |
78f9721b SM |
77 | if (GIMME == G_SCALAR) |
78 | Perl_croak(aTHX_ "Can't return array to lvalue scalar context"); | |
79 | PUSHs(TARG); | |
80 | RETURN; | |
40c94d11 | 81 | } |
85e6fe83 | 82 | } |
13017935 SM |
83 | gimme = GIMME_V; |
84 | if (gimme == G_ARRAY) { | |
502c6561 | 85 | const I32 maxarg = AvFILL(MUTABLE_AV(TARG)) + 1; |
85e6fe83 | 86 | EXTEND(SP, maxarg); |
93965878 NIS |
87 | if (SvMAGICAL(TARG)) { |
88 | U32 i; | |
eb160463 | 89 | for (i=0; i < (U32)maxarg; i++) { |
502c6561 | 90 | SV * const * const svp = av_fetch(MUTABLE_AV(TARG), i, FALSE); |
3280af22 | 91 | SP[i+1] = (svp) ? *svp : &PL_sv_undef; |
93965878 NIS |
92 | } |
93 | } | |
94 | else { | |
502c6561 | 95 | Copy(AvARRAY((const AV *)TARG), SP+1, maxarg, SV*); |
93965878 | 96 | } |
85e6fe83 LW |
97 | SP += maxarg; |
98 | } | |
13017935 | 99 | else if (gimme == G_SCALAR) { |
1b6737cc | 100 | SV* const sv = sv_newmortal(); |
502c6561 | 101 | const I32 maxarg = AvFILL(MUTABLE_AV(TARG)) + 1; |
85e6fe83 LW |
102 | sv_setiv(sv, maxarg); |
103 | PUSHs(sv); | |
104 | } | |
105 | RETURN; | |
93a17b20 LW |
106 | } |
107 | ||
108 | PP(pp_padhv) | |
109 | { | |
97aff369 | 110 | dVAR; dSP; dTARGET; |
54310121 | 111 | I32 gimme; |
112 | ||
e190e9b4 | 113 | assert(SvTYPE(TARG) == SVt_PVHV); |
93a17b20 | 114 | XPUSHs(TARG); |
533c011a | 115 | if (PL_op->op_private & OPpLVAL_INTRO) |
a5911867 RGS |
116 | if (!(PL_op->op_private & OPpPAD_STATE)) |
117 | SAVECLEARSV(PAD_SVl(PL_op->op_targ)); | |
533c011a | 118 | if (PL_op->op_flags & OPf_REF) |
93a17b20 | 119 | RETURN; |
40c94d11 FC |
120 | else if (PL_op->op_private & OPpMAYBE_LVSUB) { |
121 | const I32 flags = is_lvalue_sub(); | |
122 | if (flags && !(flags & OPpENTERSUB_INARGS)) { | |
78f9721b SM |
123 | if (GIMME == G_SCALAR) |
124 | Perl_croak(aTHX_ "Can't return hash to lvalue scalar context"); | |
125 | RETURN; | |
40c94d11 | 126 | } |
78f9721b | 127 | } |
54310121 | 128 | gimme = GIMME_V; |
129 | if (gimme == G_ARRAY) { | |
981b7185 | 130 | RETURNOP(Perl_do_kv(aTHX)); |
85e6fe83 | 131 | } |
54310121 | 132 | else if (gimme == G_SCALAR) { |
85fbaab2 | 133 | SV* const sv = Perl_hv_scalar(aTHX_ MUTABLE_HV(TARG)); |
85e6fe83 | 134 | SETs(sv); |
85e6fe83 | 135 | } |
54310121 | 136 | RETURN; |
93a17b20 LW |
137 | } |
138 | ||
79072805 LW |
139 | /* Translations. */ |
140 | ||
4bdf8368 | 141 | static const char S_no_symref_sv[] = |
def89bff NC |
142 | "Can't use string (\"%" SVf32 "\"%s) as %s ref while \"strict refs\" in use"; |
143 | ||
6f7909da FC |
144 | /* In some cases this function inspects PL_op. If this function is called |
145 | for new op types, more bool parameters may need to be added in place of | |
146 | the checks. | |
147 | ||
148 | When noinit is true, the absence of a gv will cause a retval of undef. | |
149 | This is unrelated to the cv-to-gv assignment case. | |
8ec5e241 | 150 | |
6f7909da FC |
151 | Make sure to use SPAGAIN after calling this. |
152 | */ | |
153 | ||
154 | static SV * | |
155 | S_rv2gv(pTHX_ SV *sv, const bool vivify_sv, const bool strict, | |
156 | const bool noinit) | |
157 | { | |
14f0f125 | 158 | dVAR; |
f64c9ac5 | 159 | if (!isGV(sv) || SvFAKE(sv)) SvGETMAGIC(sv); |
ed6116ce | 160 | if (SvROK(sv)) { |
93d7320b DM |
161 | if (SvAMAGIC(sv)) { |
162 | sv = amagic_deref_call(sv, to_gv_amg); | |
93d7320b | 163 | } |
e4a1664f | 164 | wasref: |
ed6116ce | 165 | sv = SvRV(sv); |
b1dadf13 | 166 | if (SvTYPE(sv) == SVt_PVIO) { |
159b6efe | 167 | GV * const gv = MUTABLE_GV(sv_newmortal()); |
b1dadf13 | 168 | gv_init(gv, 0, "", 0, 0); |
a45c7426 | 169 | GvIOp(gv) = MUTABLE_IO(sv); |
b37c2d43 | 170 | SvREFCNT_inc_void_NN(sv); |
ad64d0ec | 171 | sv = MUTABLE_SV(gv); |
ef54e1a4 | 172 | } |
6e592b3a | 173 | else if (!isGV_with_GP(sv)) |
6f7909da | 174 | return (SV *)Perl_die(aTHX_ "Not a GLOB reference"); |
79072805 LW |
175 | } |
176 | else { | |
6e592b3a | 177 | if (!isGV_with_GP(sv)) { |
f132ae69 | 178 | if (!SvOK(sv)) { |
b13b2135 | 179 | /* If this is a 'my' scalar and flag is set then vivify |
853846ea | 180 | * NI-S 1999/05/07 |
b13b2135 | 181 | */ |
f132ae69 | 182 | if (vivify_sv && sv != &PL_sv_undef) { |
2c8ac474 | 183 | GV *gv; |
ce74145d FC |
184 | if (SvREADONLY(sv)) |
185 | Perl_croak_no_modify(aTHX); | |
2c8ac474 GS |
186 | if (cUNOP->op_targ) { |
187 | STRLEN len; | |
0bd48802 AL |
188 | SV * const namesv = PAD_SV(cUNOP->op_targ); |
189 | const char * const name = SvPV(namesv, len); | |
159b6efe | 190 | gv = MUTABLE_GV(newSV(0)); |
2c8ac474 GS |
191 | gv_init(gv, CopSTASH(PL_curcop), name, len, 0); |
192 | } | |
193 | else { | |
0bd48802 | 194 | const char * const name = CopSTASHPV(PL_curcop); |
2c8ac474 | 195 | gv = newGVgen(name); |
1d8d4d2a | 196 | } |
43230e26 | 197 | prepare_SV_for_RV(sv); |
ad64d0ec | 198 | SvRV_set(sv, MUTABLE_SV(gv)); |
853846ea | 199 | SvROK_on(sv); |
1d8d4d2a | 200 | SvSETMAGIC(sv); |
853846ea | 201 | goto wasref; |
2c8ac474 | 202 | } |
6f7909da FC |
203 | if (PL_op->op_flags & OPf_REF || strict) |
204 | return (SV *)Perl_die(aTHX_ PL_no_usym, "a symbol"); | |
599cee73 | 205 | if (ckWARN(WARN_UNINITIALIZED)) |
29489e7c | 206 | report_uninit(sv); |
6f7909da | 207 | return &PL_sv_undef; |
a0d0e21e | 208 | } |
6f7909da | 209 | if (noinit) |
35cd451c | 210 | { |
77cb3b01 FC |
211 | if (!(sv = MUTABLE_SV(gv_fetchsv_nomg( |
212 | sv, GV_ADDMG, SVt_PVGV | |
23496c6e | 213 | )))) |
6f7909da | 214 | return &PL_sv_undef; |
35cd451c GS |
215 | } |
216 | else { | |
6f7909da FC |
217 | if (strict) |
218 | return | |
219 | (SV *)Perl_die(aTHX_ | |
220 | S_no_symref_sv, | |
221 | sv, | |
222 | (SvPOK(sv) && SvCUR(sv)>32 ? "..." : ""), | |
223 | "a symbol" | |
224 | ); | |
e26df76a NC |
225 | if ((PL_op->op_private & (OPpLVAL_INTRO|OPpDONT_INIT_GV)) |
226 | == OPpDONT_INIT_GV) { | |
227 | /* We are the target of a coderef assignment. Return | |
228 | the scalar unchanged, and let pp_sasssign deal with | |
229 | things. */ | |
6f7909da | 230 | return sv; |
e26df76a | 231 | } |
77cb3b01 | 232 | sv = MUTABLE_SV(gv_fetchsv_nomg(sv, GV_ADD, SVt_PVGV)); |
35cd451c | 233 | } |
2acc3314 | 234 | /* FAKE globs in the symbol table cause weird bugs (#77810) */ |
96293f45 | 235 | SvFAKE_off(sv); |
93a17b20 | 236 | } |
79072805 | 237 | } |
96293f45 | 238 | if (SvFAKE(sv)) { |
2acc3314 | 239 | SV *newsv = sv_newmortal(); |
5cf4b255 | 240 | sv_setsv_flags(newsv, sv, 0); |
2acc3314 | 241 | SvFAKE_off(newsv); |
d8906c05 | 242 | sv = newsv; |
2acc3314 | 243 | } |
6f7909da FC |
244 | return sv; |
245 | } | |
246 | ||
247 | PP(pp_rv2gv) | |
248 | { | |
249 | dVAR; dSP; dTOPss; | |
250 | ||
251 | sv = S_rv2gv(aTHX_ | |
252 | sv, PL_op->op_private & OPpDEREF, | |
253 | PL_op->op_private & HINT_STRICT_REFS, | |
254 | ((PL_op->op_flags & OPf_SPECIAL) && !(PL_op->op_flags & OPf_MOD)) | |
255 | || PL_op->op_type == OP_READLINE | |
256 | ); | |
257 | SPAGAIN; | |
d8906c05 FC |
258 | if (PL_op->op_private & OPpLVAL_INTRO) |
259 | save_gp(MUTABLE_GV(sv), !(PL_op->op_flags & OPf_SPECIAL)); | |
260 | SETs(sv); | |
79072805 LW |
261 | RETURN; |
262 | } | |
263 | ||
dc3c76f8 NC |
264 | /* Helper function for pp_rv2sv and pp_rv2av */ |
265 | GV * | |
fe9845cc RB |
266 | Perl_softref2xv(pTHX_ SV *const sv, const char *const what, |
267 | const svtype type, SV ***spp) | |
dc3c76f8 NC |
268 | { |
269 | dVAR; | |
270 | GV *gv; | |
271 | ||
7918f24d NC |
272 | PERL_ARGS_ASSERT_SOFTREF2XV; |
273 | ||
dc3c76f8 NC |
274 | if (PL_op->op_private & HINT_STRICT_REFS) { |
275 | if (SvOK(sv)) | |
10b53e54 | 276 | Perl_die(aTHX_ S_no_symref_sv, sv, (SvPOK(sv) && SvCUR(sv)>32 ? "..." : ""), what); |
dc3c76f8 NC |
277 | else |
278 | Perl_die(aTHX_ PL_no_usym, what); | |
279 | } | |
280 | if (!SvOK(sv)) { | |
fd1d9b5c FC |
281 | if ( |
282 | PL_op->op_flags & OPf_REF && | |
283 | PL_op->op_next->op_type != OP_BOOLKEYS | |
284 | ) | |
dc3c76f8 NC |
285 | Perl_die(aTHX_ PL_no_usym, what); |
286 | if (ckWARN(WARN_UNINITIALIZED)) | |
287 | report_uninit(sv); | |
288 | if (type != SVt_PV && GIMME_V == G_ARRAY) { | |
289 | (*spp)--; | |
290 | return NULL; | |
291 | } | |
292 | **spp = &PL_sv_undef; | |
293 | return NULL; | |
294 | } | |
295 | if ((PL_op->op_flags & OPf_SPECIAL) && | |
296 | !(PL_op->op_flags & OPf_MOD)) | |
297 | { | |
77cb3b01 | 298 | if (!(gv = gv_fetchsv_nomg(sv, GV_ADDMG, type))) |
dc3c76f8 NC |
299 | { |
300 | **spp = &PL_sv_undef; | |
301 | return NULL; | |
302 | } | |
303 | } | |
304 | else { | |
77cb3b01 | 305 | gv = gv_fetchsv_nomg(sv, GV_ADD, type); |
dc3c76f8 NC |
306 | } |
307 | return gv; | |
308 | } | |
309 | ||
79072805 LW |
310 | PP(pp_rv2sv) |
311 | { | |
97aff369 | 312 | dVAR; dSP; dTOPss; |
c445ea15 | 313 | GV *gv = NULL; |
79072805 | 314 | |
9026059d | 315 | SvGETMAGIC(sv); |
ed6116ce | 316 | if (SvROK(sv)) { |
93d7320b DM |
317 | if (SvAMAGIC(sv)) { |
318 | sv = amagic_deref_call(sv, to_sv_amg); | |
319 | SPAGAIN; | |
320 | } | |
f5284f61 | 321 | |
ed6116ce | 322 | sv = SvRV(sv); |
79072805 LW |
323 | switch (SvTYPE(sv)) { |
324 | case SVt_PVAV: | |
325 | case SVt_PVHV: | |
326 | case SVt_PVCV: | |
cbae9b9f YST |
327 | case SVt_PVFM: |
328 | case SVt_PVIO: | |
cea2e8a9 | 329 | DIE(aTHX_ "Not a SCALAR reference"); |
42d0e0b7 | 330 | default: NOOP; |
79072805 LW |
331 | } |
332 | } | |
333 | else { | |
159b6efe | 334 | gv = MUTABLE_GV(sv); |
748a9306 | 335 | |
6e592b3a | 336 | if (!isGV_with_GP(gv)) { |
dc3c76f8 NC |
337 | gv = Perl_softref2xv(aTHX_ sv, "a SCALAR", SVt_PV, &sp); |
338 | if (!gv) | |
339 | RETURN; | |
463ee0b2 | 340 | } |
29c711a3 | 341 | sv = GvSVn(gv); |
a0d0e21e | 342 | } |
533c011a | 343 | if (PL_op->op_flags & OPf_MOD) { |
82d03984 RGS |
344 | if (PL_op->op_private & OPpLVAL_INTRO) { |
345 | if (cUNOP->op_first->op_type == OP_NULL) | |
159b6efe | 346 | sv = save_scalar(MUTABLE_GV(TOPs)); |
82d03984 RGS |
347 | else if (gv) |
348 | sv = save_scalar(gv); | |
349 | else | |
f1f66076 | 350 | Perl_croak(aTHX_ "%s", PL_no_localize_ref); |
82d03984 | 351 | } |
533c011a | 352 | else if (PL_op->op_private & OPpDEREF) |
9026059d | 353 | sv = vivify_ref(sv, PL_op->op_private & OPpDEREF); |
79072805 | 354 | } |
a0d0e21e | 355 | SETs(sv); |
79072805 LW |
356 | RETURN; |
357 | } | |
358 | ||
359 | PP(pp_av2arylen) | |
360 | { | |
97aff369 | 361 | dVAR; dSP; |
502c6561 | 362 | AV * const av = MUTABLE_AV(TOPs); |
02d85cc3 EB |
363 | const I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET; |
364 | if (lvalue) { | |
365 | SV ** const sv = Perl_av_arylen_p(aTHX_ MUTABLE_AV(av)); | |
366 | if (!*sv) { | |
367 | *sv = newSV_type(SVt_PVMG); | |
368 | sv_magic(*sv, MUTABLE_SV(av), PERL_MAGIC_arylen, NULL, 0); | |
369 | } | |
370 | SETs(*sv); | |
371 | } else { | |
e1dccc0d | 372 | SETs(sv_2mortal(newSViv(AvFILL(MUTABLE_AV(av))))); |
79072805 | 373 | } |
79072805 LW |
374 | RETURN; |
375 | } | |
376 | ||
a0d0e21e LW |
377 | PP(pp_pos) |
378 | { | |
2154eca7 | 379 | dVAR; dSP; dPOPss; |
8ec5e241 | 380 | |
78f9721b | 381 | if (PL_op->op_flags & OPf_MOD || LVRET) { |
16eb5365 FC |
382 | SV * const ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */ |
383 | sv_magic(ret, NULL, PERL_MAGIC_pos, NULL, 0); | |
384 | LvTYPE(ret) = '.'; | |
385 | LvTARG(ret) = SvREFCNT_inc_simple(sv); | |
2154eca7 | 386 | PUSHs(ret); /* no SvSETMAGIC */ |
a0d0e21e LW |
387 | RETURN; |
388 | } | |
389 | else { | |
a0d0e21e | 390 | if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) { |
1b6737cc | 391 | const MAGIC * const mg = mg_find(sv, PERL_MAGIC_regex_global); |
565764a8 | 392 | if (mg && mg->mg_len >= 0) { |
2154eca7 | 393 | dTARGET; |
a0ed51b3 | 394 | I32 i = mg->mg_len; |
7e2040f0 | 395 | if (DO_UTF8(sv)) |
a0ed51b3 | 396 | sv_pos_b2u(sv, &i); |
e1dccc0d | 397 | PUSHi(i); |
a0d0e21e LW |
398 | RETURN; |
399 | } | |
400 | } | |
401 | RETPUSHUNDEF; | |
402 | } | |
403 | } | |
404 | ||
79072805 LW |
405 | PP(pp_rv2cv) |
406 | { | |
97aff369 | 407 | dVAR; dSP; |
79072805 | 408 | GV *gv; |
1eced8f8 | 409 | HV *stash_unused; |
c445ea15 | 410 | const I32 flags = (PL_op->op_flags & OPf_SPECIAL) |
9da346da | 411 | ? GV_ADDMG |
c445ea15 AL |
412 | : ((PL_op->op_private & (OPpLVAL_INTRO|OPpMAY_RETURN_CONSTANT)) == OPpMAY_RETURN_CONSTANT) |
413 | ? GV_ADD|GV_NOEXPAND | |
414 | : GV_ADD; | |
4633a7c4 LW |
415 | /* We usually try to add a non-existent subroutine in case of AUTOLOAD. */ |
416 | /* (But not in defined().) */ | |
e26df76a | 417 | |
1eced8f8 | 418 | CV *cv = sv_2cv(TOPs, &stash_unused, &gv, flags); |
07055b4c CS |
419 | if (cv) { |
420 | if (CvCLONE(cv)) | |
ad64d0ec | 421 | cv = MUTABLE_CV(sv_2mortal(MUTABLE_SV(cv_clone(cv)))); |
d32f2495 SC |
422 | if ((PL_op->op_private & OPpLVAL_INTRO)) { |
423 | if (gv && GvCV(gv) == cv && (gv = gv_autoload4(GvSTASH(gv), GvNAME(gv), GvNAMELEN(gv), FALSE))) | |
424 | cv = GvCV(gv); | |
425 | if (!CvLVALUE(cv)) | |
426 | DIE(aTHX_ "Can't modify non-lvalue subroutine call"); | |
427 | } | |
07055b4c | 428 | } |
e26df76a | 429 | else if ((flags == (GV_ADD|GV_NOEXPAND)) && gv && SvROK(gv)) { |
ea726b52 | 430 | cv = MUTABLE_CV(gv); |
e26df76a | 431 | } |
07055b4c | 432 | else |
ea726b52 | 433 | cv = MUTABLE_CV(&PL_sv_undef); |
ad64d0ec | 434 | SETs(MUTABLE_SV(cv)); |
79072805 LW |
435 | RETURN; |
436 | } | |
437 | ||
c07a80fd | 438 | PP(pp_prototype) |
439 | { | |
97aff369 | 440 | dVAR; dSP; |
c07a80fd | 441 | CV *cv; |
442 | HV *stash; | |
443 | GV *gv; | |
fabdb6c0 | 444 | SV *ret = &PL_sv_undef; |
c07a80fd | 445 | |
b6c543e3 | 446 | if (SvPOK(TOPs) && SvCUR(TOPs) >= 7) { |
e3f73d4e | 447 | const char * s = SvPVX_const(TOPs); |
b6c543e3 | 448 | if (strnEQ(s, "CORE::", 6)) { |
be1b855b | 449 | const int code = keyword(s + 6, SvCUR(TOPs) - 6, 1); |
b66130dd FC |
450 | if (!code || code == -KEY_CORE) |
451 | DIE(aTHX_ "Can't find an opnumber for \"%s\"", s+6); | |
452 | if (code < 0) { /* Overridable. */ | |
453 | SV * const sv = core_prototype(NULL, s + 6, code, NULL); | |
454 | if (sv) ret = sv; | |
455 | } | |
b8c38f0a | 456 | goto set; |
b6c543e3 IZ |
457 | } |
458 | } | |
f2c0649b | 459 | cv = sv_2cv(TOPs, &stash, &gv, 0); |
5f05dabc | 460 | if (cv && SvPOK(cv)) |
59cd0e26 | 461 | ret = newSVpvn_flags(SvPVX_const(cv), SvCUR(cv), SVs_TEMP); |
b6c543e3 | 462 | set: |
c07a80fd | 463 | SETs(ret); |
464 | RETURN; | |
465 | } | |
466 | ||
a0d0e21e LW |
467 | PP(pp_anoncode) |
468 | { | |
97aff369 | 469 | dVAR; dSP; |
ea726b52 | 470 | CV *cv = MUTABLE_CV(PAD_SV(PL_op->op_targ)); |
a5f75d66 | 471 | if (CvCLONE(cv)) |
ad64d0ec | 472 | cv = MUTABLE_CV(sv_2mortal(MUTABLE_SV(cv_clone(cv)))); |
5f05dabc | 473 | EXTEND(SP,1); |
ad64d0ec | 474 | PUSHs(MUTABLE_SV(cv)); |
a0d0e21e LW |
475 | RETURN; |
476 | } | |
477 | ||
478 | PP(pp_srefgen) | |
79072805 | 479 | { |
97aff369 | 480 | dVAR; dSP; |
71be2cbc | 481 | *SP = refto(*SP); |
79072805 | 482 | RETURN; |
8ec5e241 | 483 | } |
a0d0e21e LW |
484 | |
485 | PP(pp_refgen) | |
486 | { | |
97aff369 | 487 | dVAR; dSP; dMARK; |
a0d0e21e | 488 | if (GIMME != G_ARRAY) { |
5f0b1d4e GS |
489 | if (++MARK <= SP) |
490 | *MARK = *SP; | |
491 | else | |
3280af22 | 492 | *MARK = &PL_sv_undef; |
5f0b1d4e GS |
493 | *MARK = refto(*MARK); |
494 | SP = MARK; | |
495 | RETURN; | |
a0d0e21e | 496 | } |
bbce6d69 | 497 | EXTEND_MORTAL(SP - MARK); |
71be2cbc | 498 | while (++MARK <= SP) |
499 | *MARK = refto(*MARK); | |
a0d0e21e | 500 | RETURN; |
79072805 LW |
501 | } |
502 | ||
76e3520e | 503 | STATIC SV* |
cea2e8a9 | 504 | S_refto(pTHX_ SV *sv) |
71be2cbc | 505 | { |
97aff369 | 506 | dVAR; |
71be2cbc | 507 | SV* rv; |
508 | ||
7918f24d NC |
509 | PERL_ARGS_ASSERT_REFTO; |
510 | ||
71be2cbc | 511 | if (SvTYPE(sv) == SVt_PVLV && LvTYPE(sv) == 'y') { |
512 | if (LvTARGLEN(sv)) | |
68dc0745 | 513 | vivify_defelem(sv); |
514 | if (!(sv = LvTARG(sv))) | |
3280af22 | 515 | sv = &PL_sv_undef; |
0dd88869 | 516 | else |
b37c2d43 | 517 | SvREFCNT_inc_void_NN(sv); |
71be2cbc | 518 | } |
d8b46c1b | 519 | else if (SvTYPE(sv) == SVt_PVAV) { |
502c6561 NC |
520 | if (!AvREAL((const AV *)sv) && AvREIFY((const AV *)sv)) |
521 | av_reify(MUTABLE_AV(sv)); | |
d8b46c1b | 522 | SvTEMP_off(sv); |
b37c2d43 | 523 | SvREFCNT_inc_void_NN(sv); |
d8b46c1b | 524 | } |
f2933f5f DM |
525 | else if (SvPADTMP(sv) && !IS_PADGV(sv)) |
526 | sv = newSVsv(sv); | |
71be2cbc | 527 | else { |
528 | SvTEMP_off(sv); | |
b37c2d43 | 529 | SvREFCNT_inc_void_NN(sv); |
71be2cbc | 530 | } |
531 | rv = sv_newmortal(); | |
4df7f6af | 532 | sv_upgrade(rv, SVt_IV); |
b162af07 | 533 | SvRV_set(rv, sv); |
71be2cbc | 534 | SvROK_on(rv); |
535 | return rv; | |
536 | } | |
537 | ||
79072805 LW |
538 | PP(pp_ref) |
539 | { | |
97aff369 | 540 | dVAR; dSP; dTARGET; |
e1ec3a88 | 541 | const char *pv; |
1b6737cc | 542 | SV * const sv = POPs; |
f12c7020 | 543 | |
5b295bef RD |
544 | if (sv) |
545 | SvGETMAGIC(sv); | |
f12c7020 | 546 | |
a0d0e21e | 547 | if (!sv || !SvROK(sv)) |
4633a7c4 | 548 | RETPUSHNO; |
79072805 | 549 | |
cba0b539 FR |
550 | pv = sv_reftype(SvRV(sv),TRUE); |
551 | PUSHp(pv, strlen(pv)); | |
79072805 LW |
552 | RETURN; |
553 | } | |
554 | ||
555 | PP(pp_bless) | |
556 | { | |
97aff369 | 557 | dVAR; dSP; |
463ee0b2 | 558 | HV *stash; |
79072805 | 559 | |
463ee0b2 | 560 | if (MAXARG == 1) |
c2f922f1 | 561 | curstash: |
11faa288 | 562 | stash = CopSTASH(PL_curcop); |
7b8d334a | 563 | else { |
1b6737cc | 564 | SV * const ssv = POPs; |
7b8d334a | 565 | STRLEN len; |
e1ec3a88 | 566 | const char *ptr; |
81689caa | 567 | |
c2f922f1 FC |
568 | if (!ssv) goto curstash; |
569 | if (!SvGMAGICAL(ssv) && !SvAMAGIC(ssv) && SvROK(ssv)) | |
81689caa | 570 | Perl_croak(aTHX_ "Attempt to bless into a reference"); |
5c144d81 | 571 | ptr = SvPV_const(ssv,len); |
a2a5de95 NC |
572 | if (len == 0) |
573 | Perl_ck_warner(aTHX_ packWARN(WARN_MISC), | |
574 | "Explicit blessing to '' (assuming package main)"); | |
da51bb9b | 575 | stash = gv_stashpvn(ptr, len, GV_ADD); |
7b8d334a | 576 | } |
a0d0e21e | 577 | |
5d3fdfeb | 578 | (void)sv_bless(TOPs, stash); |
79072805 LW |
579 | RETURN; |
580 | } | |
581 | ||
fb73857a | 582 | PP(pp_gelem) |
583 | { | |
97aff369 | 584 | dVAR; dSP; |
b13b2135 | 585 | |
1b6737cc AL |
586 | SV *sv = POPs; |
587 | const char * const elem = SvPV_nolen_const(sv); | |
159b6efe | 588 | GV * const gv = MUTABLE_GV(POPs); |
c445ea15 | 589 | SV * tmpRef = NULL; |
1b6737cc | 590 | |
c445ea15 | 591 | sv = NULL; |
c4ba80c3 NC |
592 | if (elem) { |
593 | /* elem will always be NUL terminated. */ | |
1b6737cc | 594 | const char * const second_letter = elem + 1; |
c4ba80c3 NC |
595 | switch (*elem) { |
596 | case 'A': | |
1b6737cc | 597 | if (strEQ(second_letter, "RRAY")) |
ad64d0ec | 598 | tmpRef = MUTABLE_SV(GvAV(gv)); |
c4ba80c3 NC |
599 | break; |
600 | case 'C': | |
1b6737cc | 601 | if (strEQ(second_letter, "ODE")) |
ad64d0ec | 602 | tmpRef = MUTABLE_SV(GvCVu(gv)); |
c4ba80c3 NC |
603 | break; |
604 | case 'F': | |
1b6737cc | 605 | if (strEQ(second_letter, "ILEHANDLE")) { |
c4ba80c3 NC |
606 | /* finally deprecated in 5.8.0 */ |
607 | deprecate("*glob{FILEHANDLE}"); | |
ad64d0ec | 608 | tmpRef = MUTABLE_SV(GvIOp(gv)); |
c4ba80c3 NC |
609 | } |
610 | else | |
1b6737cc | 611 | if (strEQ(second_letter, "ORMAT")) |
ad64d0ec | 612 | tmpRef = MUTABLE_SV(GvFORM(gv)); |
c4ba80c3 NC |
613 | break; |
614 | case 'G': | |
1b6737cc | 615 | if (strEQ(second_letter, "LOB")) |
ad64d0ec | 616 | tmpRef = MUTABLE_SV(gv); |
c4ba80c3 NC |
617 | break; |
618 | case 'H': | |
1b6737cc | 619 | if (strEQ(second_letter, "ASH")) |
ad64d0ec | 620 | tmpRef = MUTABLE_SV(GvHV(gv)); |
c4ba80c3 NC |
621 | break; |
622 | case 'I': | |
1b6737cc | 623 | if (*second_letter == 'O' && !elem[2]) |
ad64d0ec | 624 | tmpRef = MUTABLE_SV(GvIOp(gv)); |
c4ba80c3 NC |
625 | break; |
626 | case 'N': | |
1b6737cc | 627 | if (strEQ(second_letter, "AME")) |
a663657d | 628 | sv = newSVhek(GvNAME_HEK(gv)); |
c4ba80c3 NC |
629 | break; |
630 | case 'P': | |
1b6737cc | 631 | if (strEQ(second_letter, "ACKAGE")) { |
7fa3a4ab NC |
632 | const HV * const stash = GvSTASH(gv); |
633 | const HEK * const hek = stash ? HvNAME_HEK(stash) : NULL; | |
396482e1 | 634 | sv = hek ? newSVhek(hek) : newSVpvs("__ANON__"); |
c4ba80c3 NC |
635 | } |
636 | break; | |
637 | case 'S': | |
1b6737cc | 638 | if (strEQ(second_letter, "CALAR")) |
f9d52e31 | 639 | tmpRef = GvSVn(gv); |
c4ba80c3 | 640 | break; |
39b99f21 | 641 | } |
fb73857a | 642 | } |
76e3520e GS |
643 | if (tmpRef) |
644 | sv = newRV(tmpRef); | |
fb73857a | 645 | if (sv) |
646 | sv_2mortal(sv); | |
647 | else | |
3280af22 | 648 | sv = &PL_sv_undef; |
fb73857a | 649 | XPUSHs(sv); |
650 | RETURN; | |
651 | } | |
652 | ||
a0d0e21e | 653 | /* Pattern matching */ |
79072805 | 654 | |
a0d0e21e | 655 | PP(pp_study) |
79072805 | 656 | { |
97aff369 | 657 | dVAR; dSP; dPOPss; |
a0d0e21e | 658 | register unsigned char *s; |
72de20cd | 659 | char *sfirst_raw; |
a0d0e21e | 660 | STRLEN len; |
4185c919 | 661 | MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_study) : NULL; |
72de20cd NC |
662 | U8 quanta; |
663 | STRLEN size; | |
4185c919 NC |
664 | |
665 | if (mg && SvSCREAM(sv)) | |
666 | RETPUSHYES; | |
a0d0e21e | 667 | |
a4f4e906 | 668 | s = (unsigned char*)(SvPV(sv, len)); |
bc9a5256 | 669 | if (len == 0 || len > I32_MAX || !SvPOK(sv) || SvUTF8(sv) || SvVALID(sv)) { |
a4f4e906 NC |
670 | /* No point in studying a zero length string, and not safe to study |
671 | anything that doesn't appear to be a simple scalar (and hence might | |
672 | change between now and when the regexp engine runs without our set | |
bd473224 | 673 | magic ever running) such as a reference to an object with overloaded |
bc9a5256 NC |
674 | stringification. Also refuse to study an FBM scalar, as this gives |
675 | more flexibility in SV flag usage. No real-world code would ever | |
676 | end up studying an FBM scalar, so this isn't a real pessimisation. | |
72de20cd NC |
677 | Endemic use of I32 in Perl_screaminstr makes it hard to safely push |
678 | the study length limit from I32_MAX to U32_MAX - 1. | |
bc9a5256 | 679 | */ |
a4f4e906 NC |
680 | RETPUSHNO; |
681 | } | |
682 | ||
72de20cd NC |
683 | if (len < 0xFF) { |
684 | quanta = 1; | |
685 | } else if (len < 0xFFFF) { | |
686 | quanta = 2; | |
687 | } else | |
688 | quanta = 4; | |
a0d0e21e | 689 | |
72de20cd NC |
690 | size = (256 + len) * quanta; |
691 | sfirst_raw = (char *)safemalloc(size); | |
692 | ||
693 | if (!sfirst_raw) | |
cea2e8a9 | 694 | DIE(aTHX_ "do_study: out of memory"); |
a0d0e21e | 695 | |
4185c919 NC |
696 | SvSCREAM_on(sv); |
697 | if (!mg) | |
698 | mg = sv_magicext(sv, NULL, PERL_MAGIC_study, &PL_vtbl_regexp, NULL, 0); | |
72de20cd NC |
699 | mg->mg_ptr = sfirst_raw; |
700 | mg->mg_len = size; | |
701 | mg->mg_private = quanta; | |
702 | ||
703 | memset(sfirst_raw, ~0, 256 * quanta); | |
704 | ||
705 | /* The assumption here is that most studied strings are fairly short, hence | |
706 | the pain of the extra code is worth it, given the memory savings. | |
707 | 80 character string, 336 bytes as U8, down from 1344 as U32 | |
708 | 800 character string, 2112 bytes as U16, down from 4224 as U32 | |
709 | */ | |
710 | ||
711 | if (quanta == 1) { | |
712 | U8 *const sfirst = (U8 *)sfirst_raw; | |
713 | U8 *const snext = sfirst + 256; | |
714 | while (len-- > 0) { | |
715 | const U8 ch = s[len]; | |
716 | snext[len] = sfirst[ch]; | |
717 | sfirst[ch] = len; | |
718 | } | |
719 | } else if (quanta == 2) { | |
720 | U16 *const sfirst = (U16 *)sfirst_raw; | |
721 | U16 *const snext = sfirst + 256; | |
722 | while (len-- > 0) { | |
723 | const U8 ch = s[len]; | |
724 | snext[len] = sfirst[ch]; | |
725 | sfirst[ch] = len; | |
726 | } | |
727 | } else { | |
728 | U32 *const sfirst = (U32 *)sfirst_raw; | |
729 | U32 *const snext = sfirst + 256; | |
730 | while (len-- > 0) { | |
731 | const U8 ch = s[len]; | |
732 | snext[len] = sfirst[ch]; | |
733 | sfirst[ch] = len; | |
734 | } | |
79072805 LW |
735 | } |
736 | ||
1e422769 | 737 | RETPUSHYES; |
79072805 LW |
738 | } |
739 | ||
a0d0e21e | 740 | PP(pp_trans) |
79072805 | 741 | { |
97aff369 | 742 | dVAR; dSP; dTARG; |
a0d0e21e LW |
743 | SV *sv; |
744 | ||
533c011a | 745 | if (PL_op->op_flags & OPf_STACKED) |
a0d0e21e | 746 | sv = POPs; |
59f00321 RGS |
747 | else if (PL_op->op_private & OPpTARGET_MY) |
748 | sv = GETTARGET; | |
79072805 | 749 | else { |
54b9620d | 750 | sv = DEFSV; |
a0d0e21e | 751 | EXTEND(SP,1); |
79072805 | 752 | } |
adbc6bb1 | 753 | TARG = sv_newmortal(); |
bb16bae8 FC |
754 | if(PL_op->op_type == OP_TRANSR) { |
755 | SV * const newsv = newSVsv(sv); | |
756 | do_trans(newsv); | |
757 | mPUSHs(newsv); | |
758 | } | |
759 | else PUSHi(do_trans(sv)); | |
a0d0e21e | 760 | RETURN; |
79072805 LW |
761 | } |
762 | ||
a0d0e21e | 763 | /* Lvalue operators. */ |
79072805 | 764 | |
81745e4e NC |
765 | static void |
766 | S_do_chomp(pTHX_ SV *retval, SV *sv, bool chomping) | |
767 | { | |
768 | dVAR; | |
769 | STRLEN len; | |
770 | char *s; | |
771 | ||
772 | PERL_ARGS_ASSERT_DO_CHOMP; | |
773 | ||
774 | if (chomping && (RsSNARF(PL_rs) || RsRECORD(PL_rs))) | |
775 | return; | |
776 | if (SvTYPE(sv) == SVt_PVAV) { | |
777 | I32 i; | |
778 | AV *const av = MUTABLE_AV(sv); | |
779 | const I32 max = AvFILL(av); | |
780 | ||
781 | for (i = 0; i <= max; i++) { | |
782 | sv = MUTABLE_SV(av_fetch(av, i, FALSE)); | |
783 | if (sv && ((sv = *(SV**)sv), sv != &PL_sv_undef)) | |
784 | do_chomp(retval, sv, chomping); | |
785 | } | |
786 | return; | |
787 | } | |
788 | else if (SvTYPE(sv) == SVt_PVHV) { | |
789 | HV* const hv = MUTABLE_HV(sv); | |
790 | HE* entry; | |
791 | (void)hv_iterinit(hv); | |
792 | while ((entry = hv_iternext(hv))) | |
793 | do_chomp(retval, hv_iterval(hv,entry), chomping); | |
794 | return; | |
795 | } | |
796 | else if (SvREADONLY(sv)) { | |
797 | if (SvFAKE(sv)) { | |
798 | /* SV is copy-on-write */ | |
799 | sv_force_normal_flags(sv, 0); | |
800 | } | |
801 | if (SvREADONLY(sv)) | |
802 | Perl_croak_no_modify(aTHX); | |
803 | } | |
804 | ||
805 | if (PL_encoding) { | |
806 | if (!SvUTF8(sv)) { | |
807 | /* XXX, here sv is utf8-ized as a side-effect! | |
808 | If encoding.pm is used properly, almost string-generating | |
809 | operations, including literal strings, chr(), input data, etc. | |
810 | should have been utf8-ized already, right? | |
811 | */ | |
812 | sv_recode_to_utf8(sv, PL_encoding); | |
813 | } | |
814 | } | |
815 | ||
816 | s = SvPV(sv, len); | |
817 | if (chomping) { | |
818 | char *temp_buffer = NULL; | |
819 | SV *svrecode = NULL; | |
820 | ||
821 | if (s && len) { | |
822 | s += --len; | |
823 | if (RsPARA(PL_rs)) { | |
824 | if (*s != '\n') | |
825 | goto nope; | |
826 | ++SvIVX(retval); | |
827 | while (len && s[-1] == '\n') { | |
828 | --len; | |
829 | --s; | |
830 | ++SvIVX(retval); | |
831 | } | |
832 | } | |
833 | else { | |
834 | STRLEN rslen, rs_charlen; | |
835 | const char *rsptr = SvPV_const(PL_rs, rslen); | |
836 | ||
837 | rs_charlen = SvUTF8(PL_rs) | |
838 | ? sv_len_utf8(PL_rs) | |
839 | : rslen; | |
840 | ||
841 | if (SvUTF8(PL_rs) != SvUTF8(sv)) { | |
842 | /* Assumption is that rs is shorter than the scalar. */ | |
843 | if (SvUTF8(PL_rs)) { | |
844 | /* RS is utf8, scalar is 8 bit. */ | |
845 | bool is_utf8 = TRUE; | |
846 | temp_buffer = (char*)bytes_from_utf8((U8*)rsptr, | |
847 | &rslen, &is_utf8); | |
848 | if (is_utf8) { | |
849 | /* Cannot downgrade, therefore cannot possibly match | |
850 | */ | |
851 | assert (temp_buffer == rsptr); | |
852 | temp_buffer = NULL; | |
853 | goto nope; | |
854 | } | |
855 | rsptr = temp_buffer; | |
856 | } | |
857 | else if (PL_encoding) { | |
858 | /* RS is 8 bit, encoding.pm is used. | |
859 | * Do not recode PL_rs as a side-effect. */ | |
860 | svrecode = newSVpvn(rsptr, rslen); | |
861 | sv_recode_to_utf8(svrecode, PL_encoding); | |
862 | rsptr = SvPV_const(svrecode, rslen); | |
863 | rs_charlen = sv_len_utf8(svrecode); | |
864 | } | |
865 | else { | |
866 | /* RS is 8 bit, scalar is utf8. */ | |
867 | temp_buffer = (char*)bytes_to_utf8((U8*)rsptr, &rslen); | |
868 | rsptr = temp_buffer; | |
869 | } | |
870 | } | |
871 | if (rslen == 1) { | |
872 | if (*s != *rsptr) | |
873 | goto nope; | |
874 | ++SvIVX(retval); | |
875 | } | |
876 | else { | |
877 | if (len < rslen - 1) | |
878 | goto nope; | |
879 | len -= rslen - 1; | |
880 | s -= rslen - 1; | |
881 | if (memNE(s, rsptr, rslen)) | |
882 | goto nope; | |
883 | SvIVX(retval) += rs_charlen; | |
884 | } | |
885 | } | |
886 | s = SvPV_force_nolen(sv); | |
887 | SvCUR_set(sv, len); | |
888 | *SvEND(sv) = '\0'; | |
889 | SvNIOK_off(sv); | |
890 | SvSETMAGIC(sv); | |
891 | } | |
892 | nope: | |
893 | ||
894 | SvREFCNT_dec(svrecode); | |
895 | ||
896 | Safefree(temp_buffer); | |
897 | } else { | |
898 | if (len && !SvPOK(sv)) | |
899 | s = SvPV_force_nomg(sv, len); | |
900 | if (DO_UTF8(sv)) { | |
901 | if (s && len) { | |
902 | char * const send = s + len; | |
903 | char * const start = s; | |
904 | s = send - 1; | |
905 | while (s > start && UTF8_IS_CONTINUATION(*s)) | |
906 | s--; | |
907 | if (is_utf8_string((U8*)s, send - s)) { | |
908 | sv_setpvn(retval, s, send - s); | |
909 | *s = '\0'; | |
910 | SvCUR_set(sv, s - start); | |
911 | SvNIOK_off(sv); | |
912 | SvUTF8_on(retval); | |
913 | } | |
914 | } | |
915 | else | |
916 | sv_setpvs(retval, ""); | |
917 | } | |
918 | else if (s && len) { | |
919 | s += --len; | |
920 | sv_setpvn(retval, s, 1); | |
921 | *s = '\0'; | |
922 | SvCUR_set(sv, len); | |
923 | SvUTF8_off(sv); | |
924 | SvNIOK_off(sv); | |
925 | } | |
926 | else | |
927 | sv_setpvs(retval, ""); | |
928 | SvSETMAGIC(sv); | |
929 | } | |
930 | } | |
931 | ||
a0d0e21e LW |
932 | PP(pp_schop) |
933 | { | |
97aff369 | 934 | dVAR; dSP; dTARGET; |
fa54efae NC |
935 | const bool chomping = PL_op->op_type == OP_SCHOMP; |
936 | ||
937 | if (chomping) | |
938 | sv_setiv(TARG, 0); | |
939 | do_chomp(TARG, TOPs, chomping); | |
a0d0e21e LW |
940 | SETTARG; |
941 | RETURN; | |
79072805 LW |
942 | } |
943 | ||
a0d0e21e | 944 | PP(pp_chop) |
79072805 | 945 | { |
97aff369 | 946 | dVAR; dSP; dMARK; dTARGET; dORIGMARK; |
fa54efae | 947 | const bool chomping = PL_op->op_type == OP_CHOMP; |
8ec5e241 | 948 | |
fa54efae NC |
949 | if (chomping) |
950 | sv_setiv(TARG, 0); | |
20cf1f79 | 951 | while (MARK < SP) |
fa54efae | 952 | do_chomp(TARG, *++MARK, chomping); |
20cf1f79 NC |
953 | SP = ORIGMARK; |
954 | XPUSHTARG; | |
a0d0e21e | 955 | RETURN; |
79072805 LW |
956 | } |
957 | ||
a0d0e21e LW |
958 | PP(pp_undef) |
959 | { | |
97aff369 | 960 | dVAR; dSP; |
a0d0e21e LW |
961 | SV *sv; |
962 | ||
533c011a | 963 | if (!PL_op->op_private) { |
774d564b | 964 | EXTEND(SP, 1); |
a0d0e21e | 965 | RETPUSHUNDEF; |
774d564b | 966 | } |
79072805 | 967 | |
a0d0e21e LW |
968 | sv = POPs; |
969 | if (!sv) | |
970 | RETPUSHUNDEF; | |
85e6fe83 | 971 | |
765f542d | 972 | SV_CHECK_THINKFIRST_COW_DROP(sv); |
85e6fe83 | 973 | |
a0d0e21e LW |
974 | switch (SvTYPE(sv)) { |
975 | case SVt_NULL: | |
976 | break; | |
977 | case SVt_PVAV: | |
502c6561 | 978 | av_undef(MUTABLE_AV(sv)); |
a0d0e21e LW |
979 | break; |
980 | case SVt_PVHV: | |
85fbaab2 | 981 | hv_undef(MUTABLE_HV(sv)); |
a0d0e21e LW |
982 | break; |
983 | case SVt_PVCV: | |
a2a5de95 NC |
984 | if (cv_const_sv((const CV *)sv)) |
985 | Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Constant subroutine %s undefined", | |
986 | CvANON((const CV *)sv) ? "(anonymous)" | |
987 | : GvENAME(CvGV((const CV *)sv))); | |
5f66b61c | 988 | /* FALLTHROUGH */ |
9607fc9c | 989 | case SVt_PVFM: |
6fc92669 GS |
990 | { |
991 | /* let user-undef'd sub keep its identity */ | |
ea726b52 NC |
992 | GV* const gv = CvGV((const CV *)sv); |
993 | cv_undef(MUTABLE_CV(sv)); | |
b3f91e91 | 994 | CvGV_set(MUTABLE_CV(sv), gv); |
6fc92669 | 995 | } |
a0d0e21e | 996 | break; |
8e07c86e | 997 | case SVt_PVGV: |
6e592b3a | 998 | if (SvFAKE(sv)) { |
3280af22 | 999 | SvSetMagicSV(sv, &PL_sv_undef); |
6e592b3a BM |
1000 | break; |
1001 | } | |
1002 | else if (isGV_with_GP(sv)) { | |
20408e3c | 1003 | GP *gp; |
dd69841b BB |
1004 | HV *stash; |
1005 | ||
dd69841b | 1006 | /* undef *Pkg::meth_name ... */ |
e530fb81 FC |
1007 | bool method_changed |
1008 | = GvCVu((const GV *)sv) && (stash = GvSTASH((const GV *)sv)) | |
1009 | && HvENAME_get(stash); | |
1010 | /* undef *Foo:: */ | |
1011 | if((stash = GvHV((const GV *)sv))) { | |
1012 | if(HvENAME_get(stash)) | |
1013 | SvREFCNT_inc_simple_void_NN(sv_2mortal((SV *)stash)); | |
1014 | else stash = NULL; | |
1015 | } | |
dd69841b | 1016 | |
159b6efe | 1017 | gp_free(MUTABLE_GV(sv)); |
a02a5408 | 1018 | Newxz(gp, 1, GP); |
c43ae56f | 1019 | GvGP_set(sv, gp_ref(gp)); |
561b68a9 | 1020 | GvSV(sv) = newSV(0); |
57843af0 | 1021 | GvLINE(sv) = CopLINE(PL_curcop); |
159b6efe | 1022 | GvEGV(sv) = MUTABLE_GV(sv); |
20408e3c | 1023 | GvMULTI_on(sv); |
e530fb81 FC |
1024 | |
1025 | if(stash) | |
afdbe55d | 1026 | mro_package_moved(NULL, stash, (const GV *)sv, 0); |
e530fb81 FC |
1027 | stash = NULL; |
1028 | /* undef *Foo::ISA */ | |
1029 | if( strEQ(GvNAME((const GV *)sv), "ISA") | |
1030 | && (stash = GvSTASH((const GV *)sv)) | |
1031 | && (method_changed || HvENAME(stash)) ) | |
1032 | mro_isa_changed_in(stash); | |
1033 | else if(method_changed) | |
1034 | mro_method_changed_in( | |
da9043f5 | 1035 | GvSTASH((const GV *)sv) |
e530fb81 FC |
1036 | ); |
1037 | ||
6e592b3a | 1038 | break; |
20408e3c | 1039 | } |
6e592b3a | 1040 | /* FALL THROUGH */ |
a0d0e21e | 1041 | default: |
b15aece3 | 1042 | if (SvTYPE(sv) >= SVt_PV && SvPVX_const(sv) && SvLEN(sv)) { |
8bd4d4c5 | 1043 | SvPV_free(sv); |
c445ea15 | 1044 | SvPV_set(sv, NULL); |
4633a7c4 | 1045 | SvLEN_set(sv, 0); |
a0d0e21e | 1046 | } |
0c34ef67 | 1047 | SvOK_off(sv); |
4633a7c4 | 1048 | SvSETMAGIC(sv); |
79072805 | 1049 | } |
a0d0e21e LW |
1050 | |
1051 | RETPUSHUNDEF; | |
79072805 LW |
1052 | } |
1053 | ||
a0d0e21e LW |
1054 | PP(pp_postinc) |
1055 | { | |
97aff369 | 1056 | dVAR; dSP; dTARGET; |
c22c99bc FC |
1057 | const bool inc = |
1058 | PL_op->op_type == OP_POSTINC || PL_op->op_type == OP_I_POSTINC; | |
60092ce4 | 1059 | if (SvTYPE(TOPs) >= SVt_PVAV || (isGV_with_GP(TOPs) && !SvFAKE(TOPs))) |
6ad8f254 | 1060 | Perl_croak_no_modify(aTHX); |
7dcb9b98 DM |
1061 | if (SvROK(TOPs)) |
1062 | TARG = sv_newmortal(); | |
a0d0e21e | 1063 | sv_setsv(TARG, TOPs); |
3510b4a1 | 1064 | if (!SvREADONLY(TOPs) && SvIOK_notUV(TOPs) && !SvNOK(TOPs) && !SvPOK(TOPs) |
c22c99bc | 1065 | && SvIVX(TOPs) != (inc ? IV_MAX : IV_MIN)) |
55497cff | 1066 | { |
c22c99bc | 1067 | SvIV_set(TOPs, SvIVX(TOPs) + (inc ? 1 : -1)); |
55497cff | 1068 | SvFLAGS(TOPs) &= ~(SVp_NOK|SVp_POK); |
748a9306 | 1069 | } |
c22c99bc | 1070 | else if (inc) |
6f1401dc | 1071 | sv_inc_nomg(TOPs); |
c22c99bc | 1072 | else sv_dec_nomg(TOPs); |
a0d0e21e | 1073 | SvSETMAGIC(TOPs); |
1e54a23f | 1074 | /* special case for undef: see thread at 2003-03/msg00536.html in archive */ |
c22c99bc | 1075 | if (inc && !SvOK(TARG)) |
a0d0e21e LW |
1076 | sv_setiv(TARG, 0); |
1077 | SETs(TARG); | |
1078 | return NORMAL; | |
1079 | } | |
79072805 | 1080 | |
a0d0e21e LW |
1081 | /* Ordinary operators. */ |
1082 | ||
1083 | PP(pp_pow) | |
1084 | { | |
800401ee | 1085 | dVAR; dSP; dATARGET; SV *svl, *svr; |
58d76dfd | 1086 | #ifdef PERL_PRESERVE_IVUV |
52a96ae6 HS |
1087 | bool is_int = 0; |
1088 | #endif | |
6f1401dc DM |
1089 | tryAMAGICbin_MG(pow_amg, AMGf_assign|AMGf_numeric); |
1090 | svr = TOPs; | |
1091 | svl = TOPm1s; | |
52a96ae6 HS |
1092 | #ifdef PERL_PRESERVE_IVUV |
1093 | /* For integer to integer power, we do the calculation by hand wherever | |
1094 | we're sure it is safe; otherwise we call pow() and try to convert to | |
1095 | integer afterwards. */ | |
58d76dfd | 1096 | { |
6f1401dc | 1097 | SvIV_please_nomg(svr); |
800401ee | 1098 | if (SvIOK(svr)) { |
6f1401dc | 1099 | SvIV_please_nomg(svl); |
800401ee | 1100 | if (SvIOK(svl)) { |
900658e3 PF |
1101 | UV power; |
1102 | bool baseuok; | |
1103 | UV baseuv; | |
1104 | ||
800401ee JH |
1105 | if (SvUOK(svr)) { |
1106 | power = SvUVX(svr); | |
900658e3 | 1107 | } else { |
800401ee | 1108 | const IV iv = SvIVX(svr); |
900658e3 PF |
1109 | if (iv >= 0) { |
1110 | power = iv; | |
1111 | } else { | |
1112 | goto float_it; /* Can't do negative powers this way. */ | |
1113 | } | |
1114 | } | |
1115 | ||
800401ee | 1116 | baseuok = SvUOK(svl); |
900658e3 | 1117 | if (baseuok) { |
800401ee | 1118 | baseuv = SvUVX(svl); |
900658e3 | 1119 | } else { |
800401ee | 1120 | const IV iv = SvIVX(svl); |
900658e3 PF |
1121 | if (iv >= 0) { |
1122 | baseuv = iv; | |
1123 | baseuok = TRUE; /* effectively it's a UV now */ | |
1124 | } else { | |
1125 | baseuv = -iv; /* abs, baseuok == false records sign */ | |
1126 | } | |
1127 | } | |
52a96ae6 HS |
1128 | /* now we have integer ** positive integer. */ |
1129 | is_int = 1; | |
1130 | ||
1131 | /* foo & (foo - 1) is zero only for a power of 2. */ | |
58d76dfd | 1132 | if (!(baseuv & (baseuv - 1))) { |
52a96ae6 | 1133 | /* We are raising power-of-2 to a positive integer. |
58d76dfd JH |
1134 | The logic here will work for any base (even non-integer |
1135 | bases) but it can be less accurate than | |
1136 | pow (base,power) or exp (power * log (base)) when the | |
1137 | intermediate values start to spill out of the mantissa. | |
1138 | With powers of 2 we know this can't happen. | |
1139 | And powers of 2 are the favourite thing for perl | |
1140 | programmers to notice ** not doing what they mean. */ | |
1141 | NV result = 1.0; | |
1142 | NV base = baseuok ? baseuv : -(NV)baseuv; | |
900658e3 PF |
1143 | |
1144 | if (power & 1) { | |
1145 | result *= base; | |
1146 | } | |
1147 | while (power >>= 1) { | |
1148 | base *= base; | |
1149 | if (power & 1) { | |
1150 | result *= base; | |
1151 | } | |
1152 | } | |
58d76dfd JH |
1153 | SP--; |
1154 | SETn( result ); | |
6f1401dc | 1155 | SvIV_please_nomg(svr); |
58d76dfd | 1156 | RETURN; |
52a96ae6 HS |
1157 | } else { |
1158 | register unsigned int highbit = 8 * sizeof(UV); | |
900658e3 PF |
1159 | register unsigned int diff = 8 * sizeof(UV); |
1160 | while (diff >>= 1) { | |
1161 | highbit -= diff; | |
1162 | if (baseuv >> highbit) { | |
1163 | highbit += diff; | |
1164 | } | |
52a96ae6 HS |
1165 | } |
1166 | /* we now have baseuv < 2 ** highbit */ | |
1167 | if (power * highbit <= 8 * sizeof(UV)) { | |
1168 | /* result will definitely fit in UV, so use UV math | |
1169 | on same algorithm as above */ | |
1170 | register UV result = 1; | |
1171 | register UV base = baseuv; | |
f2338a2e | 1172 | const bool odd_power = cBOOL(power & 1); |
900658e3 PF |
1173 | if (odd_power) { |
1174 | result *= base; | |
1175 | } | |
1176 | while (power >>= 1) { | |
1177 | base *= base; | |
1178 | if (power & 1) { | |
52a96ae6 | 1179 | result *= base; |
52a96ae6 HS |
1180 | } |
1181 | } | |
1182 | SP--; | |
0615a994 | 1183 | if (baseuok || !odd_power) |
52a96ae6 HS |
1184 | /* answer is positive */ |
1185 | SETu( result ); | |
1186 | else if (result <= (UV)IV_MAX) | |
1187 | /* answer negative, fits in IV */ | |
1188 | SETi( -(IV)result ); | |
1189 | else if (result == (UV)IV_MIN) | |
1190 | /* 2's complement assumption: special case IV_MIN */ | |
1191 | SETi( IV_MIN ); | |
1192 | else | |
1193 | /* answer negative, doesn't fit */ | |
1194 | SETn( -(NV)result ); | |
1195 | RETURN; | |
1196 | } | |
1197 | } | |
1198 | } | |
1199 | } | |
58d76dfd | 1200 | } |
52a96ae6 | 1201 | float_it: |
58d76dfd | 1202 | #endif |
a0d0e21e | 1203 | { |
6f1401dc DM |
1204 | NV right = SvNV_nomg(svr); |
1205 | NV left = SvNV_nomg(svl); | |
4efa5a16 | 1206 | (void)POPs; |
3aaeb624 JA |
1207 | |
1208 | #if defined(USE_LONG_DOUBLE) && defined(HAS_AIX_POWL_NEG_BASE_BUG) | |
1209 | /* | |
1210 | We are building perl with long double support and are on an AIX OS | |
1211 | afflicted with a powl() function that wrongly returns NaNQ for any | |
1212 | negative base. This was reported to IBM as PMR #23047-379 on | |
1213 | 03/06/2006. The problem exists in at least the following versions | |
1214 | of AIX and the libm fileset, and no doubt others as well: | |
1215 | ||
1216 | AIX 4.3.3-ML10 bos.adt.libm 4.3.3.50 | |
1217 | AIX 5.1.0-ML04 bos.adt.libm 5.1.0.29 | |
1218 | AIX 5.2.0 bos.adt.libm 5.2.0.85 | |
1219 | ||
1220 | So, until IBM fixes powl(), we provide the following workaround to | |
1221 | handle the problem ourselves. Our logic is as follows: for | |
1222 | negative bases (left), we use fmod(right, 2) to check if the | |
1223 | exponent is an odd or even integer: | |
1224 | ||
1225 | - if odd, powl(left, right) == -powl(-left, right) | |
1226 | - if even, powl(left, right) == powl(-left, right) | |
1227 | ||
1228 | If the exponent is not an integer, the result is rightly NaNQ, so | |
1229 | we just return that (as NV_NAN). | |
1230 | */ | |
1231 | ||
1232 | if (left < 0.0) { | |
1233 | NV mod2 = Perl_fmod( right, 2.0 ); | |
1234 | if (mod2 == 1.0 || mod2 == -1.0) { /* odd integer */ | |
1235 | SETn( -Perl_pow( -left, right) ); | |
1236 | } else if (mod2 == 0.0) { /* even integer */ | |
1237 | SETn( Perl_pow( -left, right) ); | |
1238 | } else { /* fractional power */ | |
1239 | SETn( NV_NAN ); | |
1240 | } | |
1241 | } else { | |
1242 | SETn( Perl_pow( left, right) ); | |
1243 | } | |
1244 | #else | |
52a96ae6 | 1245 | SETn( Perl_pow( left, right) ); |
3aaeb624 JA |
1246 | #endif /* HAS_AIX_POWL_NEG_BASE_BUG */ |
1247 | ||
52a96ae6 HS |
1248 | #ifdef PERL_PRESERVE_IVUV |
1249 | if (is_int) | |
6f1401dc | 1250 | SvIV_please_nomg(svr); |
52a96ae6 HS |
1251 | #endif |
1252 | RETURN; | |
93a17b20 | 1253 | } |
a0d0e21e LW |
1254 | } |
1255 | ||
1256 | PP(pp_multiply) | |
1257 | { | |
800401ee | 1258 | dVAR; dSP; dATARGET; SV *svl, *svr; |
6f1401dc DM |
1259 | tryAMAGICbin_MG(mult_amg, AMGf_assign|AMGf_numeric); |
1260 | svr = TOPs; | |
1261 | svl = TOPm1s; | |
28e5dec8 | 1262 | #ifdef PERL_PRESERVE_IVUV |
6f1401dc | 1263 | SvIV_please_nomg(svr); |
800401ee | 1264 | if (SvIOK(svr)) { |
28e5dec8 JH |
1265 | /* Unless the left argument is integer in range we are going to have to |
1266 | use NV maths. Hence only attempt to coerce the right argument if | |
1267 | we know the left is integer. */ | |
1268 | /* Left operand is defined, so is it IV? */ | |
6f1401dc | 1269 | SvIV_please_nomg(svl); |
800401ee JH |
1270 | if (SvIOK(svl)) { |
1271 | bool auvok = SvUOK(svl); | |
1272 | bool buvok = SvUOK(svr); | |
28e5dec8 JH |
1273 | const UV topmask = (~ (UV)0) << (4 * sizeof (UV)); |
1274 | const UV botmask = ~((~ (UV)0) << (4 * sizeof (UV))); | |
1275 | UV alow; | |
1276 | UV ahigh; | |
1277 | UV blow; | |
1278 | UV bhigh; | |
1279 | ||
1280 | if (auvok) { | |
800401ee | 1281 | alow = SvUVX(svl); |
28e5dec8 | 1282 | } else { |
800401ee | 1283 | const IV aiv = SvIVX(svl); |
28e5dec8 JH |
1284 | if (aiv >= 0) { |
1285 | alow = aiv; | |
1286 | auvok = TRUE; /* effectively it's a UV now */ | |
1287 | } else { | |
1288 | alow = -aiv; /* abs, auvok == false records sign */ | |
1289 | } | |
1290 | } | |
1291 | if (buvok) { | |
800401ee | 1292 | blow = SvUVX(svr); |
28e5dec8 | 1293 | } else { |
800401ee | 1294 | const IV biv = SvIVX(svr); |
28e5dec8 JH |
1295 | if (biv >= 0) { |
1296 | blow = biv; | |
1297 | buvok = TRUE; /* effectively it's a UV now */ | |
1298 | } else { | |
1299 | blow = -biv; /* abs, buvok == false records sign */ | |
1300 | } | |
1301 | } | |
1302 | ||
1303 | /* If this does sign extension on unsigned it's time for plan B */ | |
1304 | ahigh = alow >> (4 * sizeof (UV)); | |
1305 | alow &= botmask; | |
1306 | bhigh = blow >> (4 * sizeof (UV)); | |
1307 | blow &= botmask; | |
1308 | if (ahigh && bhigh) { | |
6f207bd3 | 1309 | NOOP; |
28e5dec8 JH |
1310 | /* eg 32 bit is at least 0x10000 * 0x10000 == 0x100000000 |
1311 | which is overflow. Drop to NVs below. */ | |
1312 | } else if (!ahigh && !bhigh) { | |
1313 | /* eg 32 bit is at most 0xFFFF * 0xFFFF == 0xFFFE0001 | |
1314 | so the unsigned multiply cannot overflow. */ | |
c445ea15 | 1315 | const UV product = alow * blow; |
28e5dec8 JH |
1316 | if (auvok == buvok) { |
1317 | /* -ve * -ve or +ve * +ve gives a +ve result. */ | |
1318 | SP--; | |
1319 | SETu( product ); | |
1320 | RETURN; | |
1321 | } else if (product <= (UV)IV_MIN) { | |
1322 | /* 2s complement assumption that (UV)-IV_MIN is correct. */ | |
1323 | /* -ve result, which could overflow an IV */ | |
1324 | SP--; | |
25716404 | 1325 | SETi( -(IV)product ); |
28e5dec8 JH |
1326 | RETURN; |
1327 | } /* else drop to NVs below. */ | |
1328 | } else { | |
1329 | /* One operand is large, 1 small */ | |
1330 | UV product_middle; | |
1331 | if (bhigh) { | |
1332 | /* swap the operands */ | |
1333 | ahigh = bhigh; | |
1334 | bhigh = blow; /* bhigh now the temp var for the swap */ | |
1335 | blow = alow; | |
1336 | alow = bhigh; | |
1337 | } | |
1338 | /* now, ((ahigh * blow) << half_UV_len) + (alow * blow) | |
1339 | multiplies can't overflow. shift can, add can, -ve can. */ | |
1340 | product_middle = ahigh * blow; | |
1341 | if (!(product_middle & topmask)) { | |
1342 | /* OK, (ahigh * blow) won't lose bits when we shift it. */ | |
1343 | UV product_low; | |
1344 | product_middle <<= (4 * sizeof (UV)); | |
1345 | product_low = alow * blow; | |
1346 | ||
1347 | /* as for pp_add, UV + something mustn't get smaller. | |
1348 | IIRC ANSI mandates this wrapping *behaviour* for | |
1349 | unsigned whatever the actual representation*/ | |
1350 | product_low += product_middle; | |
1351 | if (product_low >= product_middle) { | |
1352 | /* didn't overflow */ | |
1353 | if (auvok == buvok) { | |
1354 | /* -ve * -ve or +ve * +ve gives a +ve result. */ | |
1355 | SP--; | |
1356 | SETu( product_low ); | |
1357 | RETURN; | |
1358 | } else if (product_low <= (UV)IV_MIN) { | |
1359 | /* 2s complement assumption again */ | |
1360 | /* -ve result, which could overflow an IV */ | |
1361 | SP--; | |
25716404 | 1362 | SETi( -(IV)product_low ); |
28e5dec8 JH |
1363 | RETURN; |
1364 | } /* else drop to NVs below. */ | |
1365 | } | |
1366 | } /* product_middle too large */ | |
1367 | } /* ahigh && bhigh */ | |
800401ee JH |
1368 | } /* SvIOK(svl) */ |
1369 | } /* SvIOK(svr) */ | |
28e5dec8 | 1370 | #endif |
a0d0e21e | 1371 | { |
6f1401dc DM |
1372 | NV right = SvNV_nomg(svr); |
1373 | NV left = SvNV_nomg(svl); | |
4efa5a16 | 1374 | (void)POPs; |
a0d0e21e LW |
1375 | SETn( left * right ); |
1376 | RETURN; | |
79072805 | 1377 | } |
a0d0e21e LW |
1378 | } |
1379 | ||
1380 | PP(pp_divide) | |
1381 | { | |
800401ee | 1382 | dVAR; dSP; dATARGET; SV *svl, *svr; |
6f1401dc DM |
1383 | tryAMAGICbin_MG(div_amg, AMGf_assign|AMGf_numeric); |
1384 | svr = TOPs; | |
1385 | svl = TOPm1s; | |
5479d192 | 1386 | /* Only try to do UV divide first |
68795e93 | 1387 | if ((SLOPPYDIVIDE is true) or |
5479d192 NC |
1388 | (PERL_PRESERVE_IVUV is true and one or both SV is a UV too large |
1389 | to preserve)) | |
1390 | The assumption is that it is better to use floating point divide | |
1391 | whenever possible, only doing integer divide first if we can't be sure. | |
1392 | If NV_PRESERVES_UV is true then we know at compile time that no UV | |
1393 | can be too large to preserve, so don't need to compile the code to | |
1394 | test the size of UVs. */ | |
1395 | ||
a0d0e21e | 1396 | #ifdef SLOPPYDIVIDE |
5479d192 NC |
1397 | # define PERL_TRY_UV_DIVIDE |
1398 | /* ensure that 20./5. == 4. */ | |
a0d0e21e | 1399 | #else |
5479d192 NC |
1400 | # ifdef PERL_PRESERVE_IVUV |
1401 | # ifndef NV_PRESERVES_UV | |
1402 | # define PERL_TRY_UV_DIVIDE | |
1403 | # endif | |
1404 | # endif | |
a0d0e21e | 1405 | #endif |
5479d192 NC |
1406 | |
1407 | #ifdef PERL_TRY_UV_DIVIDE | |
6f1401dc | 1408 | SvIV_please_nomg(svr); |
800401ee | 1409 | if (SvIOK(svr)) { |
6f1401dc | 1410 | SvIV_please_nomg(svl); |
800401ee JH |
1411 | if (SvIOK(svl)) { |
1412 | bool left_non_neg = SvUOK(svl); | |
1413 | bool right_non_neg = SvUOK(svr); | |
5479d192 NC |
1414 | UV left; |
1415 | UV right; | |
1416 | ||
1417 | if (right_non_neg) { | |
800401ee | 1418 | right = SvUVX(svr); |
5479d192 NC |
1419 | } |
1420 | else { | |
800401ee | 1421 | const IV biv = SvIVX(svr); |
5479d192 NC |
1422 | if (biv >= 0) { |
1423 | right = biv; | |
1424 | right_non_neg = TRUE; /* effectively it's a UV now */ | |
1425 | } | |
1426 | else { | |
1427 | right = -biv; | |
1428 | } | |
1429 | } | |
1430 | /* historically undef()/0 gives a "Use of uninitialized value" | |
1431 | warning before dieing, hence this test goes here. | |
1432 | If it were immediately before the second SvIV_please, then | |
1433 | DIE() would be invoked before left was even inspected, so | |
486ec47a | 1434 | no inspection would give no warning. */ |
5479d192 NC |
1435 | if (right == 0) |
1436 | DIE(aTHX_ "Illegal division by zero"); | |
1437 | ||
1438 | if (left_non_neg) { | |
800401ee | 1439 | left = SvUVX(svl); |
5479d192 NC |
1440 | } |
1441 | else { | |
800401ee | 1442 | const IV aiv = SvIVX(svl); |
5479d192 NC |
1443 | if (aiv >= 0) { |
1444 | left = aiv; | |
1445 | left_non_neg = TRUE; /* effectively it's a UV now */ | |
1446 | } | |
1447 | else { | |
1448 | left = -aiv; | |
1449 | } | |
1450 | } | |
1451 | ||
1452 | if (left >= right | |
1453 | #ifdef SLOPPYDIVIDE | |
1454 | /* For sloppy divide we always attempt integer division. */ | |
1455 | #else | |
1456 | /* Otherwise we only attempt it if either or both operands | |
1457 | would not be preserved by an NV. If both fit in NVs | |
0c2ee62a NC |
1458 | we fall through to the NV divide code below. However, |
1459 | as left >= right to ensure integer result here, we know that | |
1460 | we can skip the test on the right operand - right big | |
1461 | enough not to be preserved can't get here unless left is | |
1462 | also too big. */ | |
1463 | ||
1464 | && (left > ((UV)1 << NV_PRESERVES_UV_BITS)) | |
5479d192 NC |
1465 | #endif |
1466 | ) { | |
1467 | /* Integer division can't overflow, but it can be imprecise. */ | |
1b6737cc | 1468 | const UV result = left / right; |
5479d192 NC |
1469 | if (result * right == left) { |
1470 | SP--; /* result is valid */ | |
1471 | if (left_non_neg == right_non_neg) { | |
1472 | /* signs identical, result is positive. */ | |
1473 | SETu( result ); | |
1474 | RETURN; | |
1475 | } | |
1476 | /* 2s complement assumption */ | |
1477 | if (result <= (UV)IV_MIN) | |
91f3b821 | 1478 | SETi( -(IV)result ); |
5479d192 NC |
1479 | else { |
1480 | /* It's exact but too negative for IV. */ | |
1481 | SETn( -(NV)result ); | |
1482 | } | |
1483 | RETURN; | |
1484 | } /* tried integer divide but it was not an integer result */ | |
32fdb065 | 1485 | } /* else (PERL_ABS(result) < 1.0) or (both UVs in range for NV) */ |
5479d192 NC |
1486 | } /* left wasn't SvIOK */ |
1487 | } /* right wasn't SvIOK */ | |
1488 | #endif /* PERL_TRY_UV_DIVIDE */ | |
1489 | { | |
6f1401dc DM |
1490 | NV right = SvNV_nomg(svr); |
1491 | NV left = SvNV_nomg(svl); | |
4efa5a16 | 1492 | (void)POPs;(void)POPs; |
ebc6a117 PD |
1493 | #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan) |
1494 | if (! Perl_isnan(right) && right == 0.0) | |
1495 | #else | |
5479d192 | 1496 | if (right == 0.0) |
ebc6a117 | 1497 | #endif |
5479d192 NC |
1498 | DIE(aTHX_ "Illegal division by zero"); |
1499 | PUSHn( left / right ); | |
1500 | RETURN; | |
79072805 | 1501 | } |
a0d0e21e LW |
1502 | } |
1503 | ||
1504 | PP(pp_modulo) | |
1505 | { | |
6f1401dc DM |
1506 | dVAR; dSP; dATARGET; |
1507 | tryAMAGICbin_MG(modulo_amg, AMGf_assign|AMGf_numeric); | |
a0d0e21e | 1508 | { |
9c5ffd7c JH |
1509 | UV left = 0; |
1510 | UV right = 0; | |
dc656993 JH |
1511 | bool left_neg = FALSE; |
1512 | bool right_neg = FALSE; | |
e2c88acc NC |
1513 | bool use_double = FALSE; |
1514 | bool dright_valid = FALSE; | |
9c5ffd7c JH |
1515 | NV dright = 0.0; |
1516 | NV dleft = 0.0; | |
6f1401dc DM |
1517 | SV * const svr = TOPs; |
1518 | SV * const svl = TOPm1s; | |
1519 | SvIV_please_nomg(svr); | |
800401ee JH |
1520 | if (SvIOK(svr)) { |
1521 | right_neg = !SvUOK(svr); | |
e2c88acc | 1522 | if (!right_neg) { |
800401ee | 1523 | right = SvUVX(svr); |
e2c88acc | 1524 | } else { |
800401ee | 1525 | const IV biv = SvIVX(svr); |
e2c88acc NC |
1526 | if (biv >= 0) { |
1527 | right = biv; | |
1528 | right_neg = FALSE; /* effectively it's a UV now */ | |
1529 | } else { | |
1530 | right = -biv; | |
1531 | } | |
1532 | } | |
1533 | } | |
1534 | else { | |
6f1401dc | 1535 | dright = SvNV_nomg(svr); |
787eafbd IZ |
1536 | right_neg = dright < 0; |
1537 | if (right_neg) | |
1538 | dright = -dright; | |
e2c88acc NC |
1539 | if (dright < UV_MAX_P1) { |
1540 | right = U_V(dright); | |
1541 | dright_valid = TRUE; /* In case we need to use double below. */ | |
1542 | } else { | |
1543 | use_double = TRUE; | |
1544 | } | |
787eafbd | 1545 | } |
a0d0e21e | 1546 | |
e2c88acc NC |
1547 | /* At this point use_double is only true if right is out of range for |
1548 | a UV. In range NV has been rounded down to nearest UV and | |
1549 | use_double false. */ | |
6f1401dc | 1550 | SvIV_please_nomg(svl); |
800401ee JH |
1551 | if (!use_double && SvIOK(svl)) { |
1552 | if (SvIOK(svl)) { | |
1553 | left_neg = !SvUOK(svl); | |
e2c88acc | 1554 | if (!left_neg) { |
800401ee | 1555 | left = SvUVX(svl); |
e2c88acc | 1556 | } else { |
800401ee | 1557 | const IV aiv = SvIVX(svl); |
e2c88acc NC |
1558 | if (aiv >= 0) { |
1559 | left = aiv; | |
1560 | left_neg = FALSE; /* effectively it's a UV now */ | |
1561 | } else { | |
1562 | left = -aiv; | |
1563 | } | |
1564 | } | |
1565 | } | |
1566 | } | |
787eafbd | 1567 | else { |
6f1401dc | 1568 | dleft = SvNV_nomg(svl); |
787eafbd IZ |
1569 | left_neg = dleft < 0; |
1570 | if (left_neg) | |
1571 | dleft = -dleft; | |
68dc0745 | 1572 | |
e2c88acc NC |
1573 | /* This should be exactly the 5.6 behaviour - if left and right are |
1574 | both in range for UV then use U_V() rather than floor. */ | |
1575 | if (!use_double) { | |
1576 | if (dleft < UV_MAX_P1) { | |
1577 | /* right was in range, so is dleft, so use UVs not double. | |
1578 | */ | |
1579 | left = U_V(dleft); | |
1580 | } | |
1581 | /* left is out of range for UV, right was in range, so promote | |
1582 | right (back) to double. */ | |
1583 | else { | |
1584 | /* The +0.5 is used in 5.6 even though it is not strictly | |
1585 | consistent with the implicit +0 floor in the U_V() | |
1586 | inside the #if 1. */ | |
1587 | dleft = Perl_floor(dleft + 0.5); | |
1588 | use_double = TRUE; | |
1589 | if (dright_valid) | |
1590 | dright = Perl_floor(dright + 0.5); | |
1591 | else | |
1592 | dright = right; | |
1593 | } | |
1594 | } | |
1595 | } | |
6f1401dc | 1596 | sp -= 2; |
787eafbd | 1597 | if (use_double) { |
65202027 | 1598 | NV dans; |
787eafbd | 1599 | |
787eafbd | 1600 | if (!dright) |
cea2e8a9 | 1601 | DIE(aTHX_ "Illegal modulus zero"); |
787eafbd | 1602 | |
65202027 | 1603 | dans = Perl_fmod(dleft, dright); |
787eafbd IZ |
1604 | if ((left_neg != right_neg) && dans) |
1605 | dans = dright - dans; | |
1606 | if (right_neg) | |
1607 | dans = -dans; | |
1608 | sv_setnv(TARG, dans); | |
1609 | } | |
1610 | else { | |
1611 | UV ans; | |
1612 | ||
787eafbd | 1613 | if (!right) |
cea2e8a9 | 1614 | DIE(aTHX_ "Illegal modulus zero"); |
787eafbd IZ |
1615 | |
1616 | ans = left % right; | |
1617 | if ((left_neg != right_neg) && ans) | |
1618 | ans = right - ans; | |
1619 | if (right_neg) { | |
1620 | /* XXX may warn: unary minus operator applied to unsigned type */ | |
1621 | /* could change -foo to be (~foo)+1 instead */ | |
1622 | if (ans <= ~((UV)IV_MAX)+1) | |
1623 | sv_setiv(TARG, ~ans+1); | |
1624 | else | |
65202027 | 1625 | sv_setnv(TARG, -(NV)ans); |
787eafbd IZ |
1626 | } |
1627 | else | |
1628 | sv_setuv(TARG, ans); | |
1629 | } | |
1630 | PUSHTARG; | |
1631 | RETURN; | |
79072805 | 1632 | } |
a0d0e21e | 1633 | } |
79072805 | 1634 | |
a0d0e21e LW |
1635 | PP(pp_repeat) |
1636 | { | |
6f1401dc | 1637 | dVAR; dSP; dATARGET; |
2b573ace | 1638 | register IV count; |
6f1401dc DM |
1639 | SV *sv; |
1640 | ||
1641 | if (GIMME == G_ARRAY && PL_op->op_private & OPpREPEAT_DOLIST) { | |
1642 | /* TODO: think of some way of doing list-repeat overloading ??? */ | |
1643 | sv = POPs; | |
1644 | SvGETMAGIC(sv); | |
1645 | } | |
1646 | else { | |
1647 | tryAMAGICbin_MG(repeat_amg, AMGf_assign); | |
1648 | sv = POPs; | |
1649 | } | |
1650 | ||
2b573ace JH |
1651 | if (SvIOKp(sv)) { |
1652 | if (SvUOK(sv)) { | |
6f1401dc | 1653 | const UV uv = SvUV_nomg(sv); |
2b573ace JH |
1654 | if (uv > IV_MAX) |
1655 | count = IV_MAX; /* The best we can do? */ | |
1656 | else | |
1657 | count = uv; | |
1658 | } else { | |
6f1401dc | 1659 | const IV iv = SvIV_nomg(sv); |
2b573ace JH |
1660 | if (iv < 0) |
1661 | count = 0; | |
1662 | else | |
1663 | count = iv; | |
1664 | } | |
1665 | } | |
1666 | else if (SvNOKp(sv)) { | |
6f1401dc | 1667 | const NV nv = SvNV_nomg(sv); |
2b573ace JH |
1668 | if (nv < 0.0) |
1669 | count = 0; | |
1670 | else | |
1671 | count = (IV)nv; | |
1672 | } | |
1673 | else | |
6f1401dc DM |
1674 | count = SvIV_nomg(sv); |
1675 | ||
533c011a | 1676 | if (GIMME == G_ARRAY && PL_op->op_private & OPpREPEAT_DOLIST) { |
a0d0e21e | 1677 | dMARK; |
0bd48802 AL |
1678 | static const char oom_list_extend[] = "Out of memory during list extend"; |
1679 | const I32 items = SP - MARK; | |
1680 | const I32 max = items * count; | |
79072805 | 1681 | |
2b573ace JH |
1682 | MEM_WRAP_CHECK_1(max, SV*, oom_list_extend); |
1683 | /* Did the max computation overflow? */ | |
27d5b266 | 1684 | if (items > 0 && max > 0 && (max < items || max < count)) |
2b573ace | 1685 | Perl_croak(aTHX_ oom_list_extend); |
a0d0e21e LW |
1686 | MEXTEND(MARK, max); |
1687 | if (count > 1) { | |
1688 | while (SP > MARK) { | |
976c8a39 JH |
1689 | #if 0 |
1690 | /* This code was intended to fix 20010809.028: | |
1691 | ||
1692 | $x = 'abcd'; | |
1693 | for (($x =~ /./g) x 2) { | |
1694 | print chop; # "abcdabcd" expected as output. | |
1695 | } | |
1696 | ||
1697 | * but that change (#11635) broke this code: | |
1698 | ||
1699 | $x = [("foo")x2]; # only one "foo" ended up in the anonlist. | |
1700 | ||
1701 | * I can't think of a better fix that doesn't introduce | |
1702 | * an efficiency hit by copying the SVs. The stack isn't | |
1703 | * refcounted, and mortalisation obviously doesn't | |
1704 | * Do The Right Thing when the stack has more than | |
1705 | * one pointer to the same mortal value. | |
1706 | * .robin. | |
1707 | */ | |
e30acc16 RH |
1708 | if (*SP) { |
1709 | *SP = sv_2mortal(newSVsv(*SP)); | |
1710 | SvREADONLY_on(*SP); | |
1711 | } | |
976c8a39 JH |
1712 | #else |
1713 | if (*SP) | |
1714 | SvTEMP_off((*SP)); | |
1715 | #endif | |
a0d0e21e | 1716 | SP--; |
79072805 | 1717 | } |
a0d0e21e LW |
1718 | MARK++; |
1719 | repeatcpy((char*)(MARK + items), (char*)MARK, | |
ad64d0ec | 1720 | items * sizeof(const SV *), count - 1); |
a0d0e21e | 1721 | SP += max; |
79072805 | 1722 | } |
a0d0e21e LW |
1723 | else if (count <= 0) |
1724 | SP -= items; | |
79072805 | 1725 | } |
a0d0e21e | 1726 | else { /* Note: mark already snarfed by pp_list */ |
0bd48802 | 1727 | SV * const tmpstr = POPs; |
a0d0e21e | 1728 | STRLEN len; |
9b877dbb | 1729 | bool isutf; |
2b573ace JH |
1730 | static const char oom_string_extend[] = |
1731 | "Out of memory during string extend"; | |
a0d0e21e | 1732 | |
6f1401dc DM |
1733 | if (TARG != tmpstr) |
1734 | sv_setsv_nomg(TARG, tmpstr); | |
1735 | SvPV_force_nomg(TARG, len); | |
9b877dbb | 1736 | isutf = DO_UTF8(TARG); |
8ebc5c01 | 1737 | if (count != 1) { |
1738 | if (count < 1) | |
1739 | SvCUR_set(TARG, 0); | |
1740 | else { | |
c445ea15 | 1741 | const STRLEN max = (UV)count * len; |
19a94d75 | 1742 | if (len > MEM_SIZE_MAX / count) |
2b573ace JH |
1743 | Perl_croak(aTHX_ oom_string_extend); |
1744 | MEM_WRAP_CHECK_1(max, char, oom_string_extend); | |
8569b9dc | 1745 | SvGROW(TARG, max + 1); |
a0d0e21e | 1746 | repeatcpy(SvPVX(TARG) + len, SvPVX(TARG), len, count - 1); |
b162af07 | 1747 | SvCUR_set(TARG, SvCUR(TARG) * count); |
7a4c00b4 | 1748 | } |
a0d0e21e | 1749 | *SvEND(TARG) = '\0'; |
a0d0e21e | 1750 | } |
dfcb284a GS |
1751 | if (isutf) |
1752 | (void)SvPOK_only_UTF8(TARG); | |
1753 | else | |
1754 | (void)SvPOK_only(TARG); | |
b80b6069 RH |
1755 | |
1756 | if (PL_op->op_private & OPpREPEAT_DOLIST) { | |
1757 | /* The parser saw this as a list repeat, and there | |
1758 | are probably several items on the stack. But we're | |
1759 | in scalar context, and there's no pp_list to save us | |
1760 | now. So drop the rest of the items -- robin@kitsite.com | |
1761 | */ | |
1762 | dMARK; | |
1763 | SP = MARK; | |
1764 | } | |
a0d0e21e | 1765 | PUSHTARG; |
79072805 | 1766 | } |
a0d0e21e LW |
1767 | RETURN; |
1768 | } | |
79072805 | 1769 | |
a0d0e21e LW |
1770 | PP(pp_subtract) |
1771 | { | |
800401ee | 1772 | dVAR; dSP; dATARGET; bool useleft; SV *svl, *svr; |
6f1401dc DM |
1773 | tryAMAGICbin_MG(subtr_amg, AMGf_assign|AMGf_numeric); |
1774 | svr = TOPs; | |
1775 | svl = TOPm1s; | |
800401ee | 1776 | useleft = USE_LEFT(svl); |
28e5dec8 | 1777 | #ifdef PERL_PRESERVE_IVUV |
7dca457a NC |
1778 | /* See comments in pp_add (in pp_hot.c) about Overflow, and how |
1779 | "bad things" happen if you rely on signed integers wrapping. */ | |
6f1401dc | 1780 | SvIV_please_nomg(svr); |
800401ee | 1781 | if (SvIOK(svr)) { |
28e5dec8 JH |
1782 | /* Unless the left argument is integer in range we are going to have to |
1783 | use NV maths. Hence only attempt to coerce the right argument if | |
1784 | we know the left is integer. */ | |
9c5ffd7c JH |
1785 | register UV auv = 0; |
1786 | bool auvok = FALSE; | |
7dca457a NC |
1787 | bool a_valid = 0; |
1788 | ||
28e5dec8 | 1789 | if (!useleft) { |
7dca457a NC |
1790 | auv = 0; |
1791 | a_valid = auvok = 1; | |
1792 | /* left operand is undef, treat as zero. */ | |
28e5dec8 JH |
1793 | } else { |
1794 | /* Left operand is defined, so is it IV? */ | |
6f1401dc | 1795 | SvIV_please_nomg(svl); |
800401ee JH |
1796 | if (SvIOK(svl)) { |
1797 | if ((auvok = SvUOK(svl))) | |
1798 | auv = SvUVX(svl); | |
7dca457a | 1799 | else { |
800401ee | 1800 | register const IV aiv = SvIVX(svl); |
7dca457a NC |
1801 | if (aiv >= 0) { |
1802 | auv = aiv; | |
1803 | auvok = 1; /* Now acting as a sign flag. */ | |
1804 | } else { /* 2s complement assumption for IV_MIN */ | |
1805 | auv = (UV)-aiv; | |
28e5dec8 | 1806 | } |
7dca457a NC |
1807 | } |
1808 | a_valid = 1; | |
1809 | } | |
1810 | } | |
1811 | if (a_valid) { | |
1812 | bool result_good = 0; | |
1813 | UV result; | |
1814 | register UV buv; | |
800401ee | 1815 | bool buvok = SvUOK(svr); |
9041c2e3 | 1816 | |
7dca457a | 1817 | if (buvok) |
800401ee | 1818 | buv = SvUVX(svr); |
7dca457a | 1819 | else { |
800401ee | 1820 | register const IV biv = SvIVX(svr); |
7dca457a NC |
1821 | if (biv >= 0) { |
1822 | buv = biv; | |
1823 | buvok = 1; | |
1824 | } else | |
1825 | buv = (UV)-biv; | |
1826 | } | |
1827 | /* ?uvok if value is >= 0. basically, flagged as UV if it's +ve, | |
602f51c4 | 1828 | else "IV" now, independent of how it came in. |
7dca457a NC |
1829 | if a, b represents positive, A, B negative, a maps to -A etc |
1830 | a - b => (a - b) | |
1831 | A - b => -(a + b) | |
1832 | a - B => (a + b) | |
1833 | A - B => -(a - b) | |
1834 | all UV maths. negate result if A negative. | |
1835 | subtract if signs same, add if signs differ. */ | |
1836 | ||
1837 | if (auvok ^ buvok) { | |
1838 | /* Signs differ. */ | |
1839 | result = auv + buv; | |
1840 | if (result >= auv) | |
1841 | result_good = 1; | |
1842 | } else { | |
1843 | /* Signs same */ | |
1844 | if (auv >= buv) { | |
1845 | result = auv - buv; | |
1846 | /* Must get smaller */ | |
1847 | if (result <= auv) | |
1848 | result_good = 1; | |
1849 | } else { | |
1850 | result = buv - auv; | |
1851 | if (result <= buv) { | |
1852 | /* result really should be -(auv-buv). as its negation | |
1853 | of true value, need to swap our result flag */ | |
1854 | auvok = !auvok; | |
1855 | result_good = 1; | |
28e5dec8 | 1856 | } |
28e5dec8 JH |
1857 | } |
1858 | } | |
7dca457a NC |
1859 | if (result_good) { |
1860 | SP--; | |
1861 | if (auvok) | |
1862 | SETu( result ); | |
1863 | else { | |
1864 | /* Negate result */ | |
1865 | if (result <= (UV)IV_MIN) | |
1866 | SETi( -(IV)result ); | |
1867 | else { | |
1868 | /* result valid, but out of range for IV. */ | |
1869 | SETn( -(NV)result ); | |
1870 | } | |
1871 | } | |
1872 | RETURN; | |
1873 | } /* Overflow, drop through to NVs. */ | |
28e5dec8 JH |
1874 | } |
1875 | } | |
1876 | #endif | |
a0d0e21e | 1877 | { |
6f1401dc | 1878 | NV value = SvNV_nomg(svr); |
4efa5a16 RD |
1879 | (void)POPs; |
1880 | ||
28e5dec8 JH |
1881 | if (!useleft) { |
1882 | /* left operand is undef, treat as zero - value */ | |
1883 | SETn(-value); | |
1884 | RETURN; | |
1885 | } | |
6f1401dc | 1886 | SETn( SvNV_nomg(svl) - value ); |
28e5dec8 | 1887 | RETURN; |
79072805 | 1888 | } |
a0d0e21e | 1889 | } |
79072805 | 1890 | |
a0d0e21e LW |
1891 | PP(pp_left_shift) |
1892 | { | |
6f1401dc | 1893 | dVAR; dSP; dATARGET; SV *svl, *svr; |
a42d0242 | 1894 | tryAMAGICbin_MG(lshift_amg, AMGf_assign|AMGf_numeric); |
6f1401dc DM |
1895 | svr = POPs; |
1896 | svl = TOPs; | |
a0d0e21e | 1897 | { |
6f1401dc | 1898 | const IV shift = SvIV_nomg(svr); |
d0ba1bd2 | 1899 | if (PL_op->op_private & HINT_INTEGER) { |
6f1401dc | 1900 | const IV i = SvIV_nomg(svl); |
972b05a9 | 1901 | SETi(i << shift); |
d0ba1bd2 JH |
1902 | } |
1903 | else { | |
6f1401dc | 1904 | const UV u = SvUV_nomg(svl); |
972b05a9 | 1905 | SETu(u << shift); |
d0ba1bd2 | 1906 | } |
55497cff | 1907 | RETURN; |
79072805 | 1908 | } |
a0d0e21e | 1909 | } |
79072805 | 1910 | |
a0d0e21e LW |
1911 | PP(pp_right_shift) |
1912 | { | |
6f1401dc | 1913 | dVAR; dSP; dATARGET; SV *svl, *svr; |
a42d0242 | 1914 | tryAMAGICbin_MG(rshift_amg, AMGf_assign|AMGf_numeric); |
6f1401dc DM |
1915 | svr = POPs; |
1916 | svl = TOPs; | |
a0d0e21e | 1917 | { |
6f1401dc | 1918 | const IV shift = SvIV_nomg(svr); |
d0ba1bd2 | 1919 | if (PL_op->op_private & HINT_INTEGER) { |
6f1401dc | 1920 | const IV i = SvIV_nomg(svl); |
972b05a9 | 1921 | SETi(i >> shift); |
d0ba1bd2 JH |
1922 | } |
1923 | else { | |
6f1401dc | 1924 | const UV u = SvUV_nomg(svl); |
972b05a9 | 1925 | SETu(u >> shift); |
d0ba1bd2 | 1926 | } |
a0d0e21e | 1927 | RETURN; |
93a17b20 | 1928 | } |
79072805 LW |
1929 | } |
1930 | ||
a0d0e21e | 1931 | PP(pp_lt) |
79072805 | 1932 | { |
6f1401dc | 1933 | dVAR; dSP; |
33efebe6 DM |
1934 | SV *left, *right; |
1935 | ||
a42d0242 | 1936 | tryAMAGICbin_MG(lt_amg, AMGf_set|AMGf_numeric); |
33efebe6 DM |
1937 | right = POPs; |
1938 | left = TOPs; | |
1939 | SETs(boolSV( | |
1940 | (SvIOK_notUV(left) && SvIOK_notUV(right)) | |
1941 | ? (SvIVX(left) < SvIVX(right)) | |
1942 | : (do_ncmp(left, right) == -1) | |
1943 | )); | |
1944 | RETURN; | |
a0d0e21e | 1945 | } |
79072805 | 1946 | |
a0d0e21e LW |
1947 | PP(pp_gt) |
1948 | { | |
6f1401dc | 1949 | dVAR; dSP; |
33efebe6 | 1950 | SV *left, *right; |
1b6737cc | 1951 | |
33efebe6 DM |
1952 | tryAMAGICbin_MG(gt_amg, AMGf_set|AMGf_numeric); |
1953 | right = POPs; | |
1954 | left = TOPs; | |
1955 | SETs(boolSV( | |
1956 | (SvIOK_notUV(left) && SvIOK_notUV(right)) | |
1957 | ? (SvIVX(left) > SvIVX(right)) | |
1958 | : (do_ncmp(left, right) == 1) | |
1959 | )); | |
1960 | RETURN; | |
a0d0e21e LW |
1961 | } |
1962 | ||
1963 | PP(pp_le) | |
1964 | { | |
6f1401dc | 1965 | dVAR; dSP; |
33efebe6 | 1966 | SV *left, *right; |
1b6737cc | 1967 | |
33efebe6 DM |
1968 | tryAMAGICbin_MG(le_amg, AMGf_set|AMGf_numeric); |
1969 | right = POPs; | |
1970 | left = TOPs; | |
1971 | SETs(boolSV( | |
1972 | (SvIOK_notUV(left) && SvIOK_notUV(right)) | |
1973 | ? (SvIVX(left) <= SvIVX(right)) | |
1974 | : (do_ncmp(left, right) <= 0) | |
1975 | )); | |
1976 | RETURN; | |
a0d0e21e LW |
1977 | } |
1978 | ||
1979 | PP(pp_ge) | |
1980 | { | |
6f1401dc | 1981 | dVAR; dSP; |
33efebe6 DM |
1982 | SV *left, *right; |
1983 | ||
1984 | tryAMAGICbin_MG(ge_amg, AMGf_set|AMGf_numeric); | |
1985 | right = POPs; | |
1986 | left = TOPs; | |
1987 | SETs(boolSV( | |
1988 | (SvIOK_notUV(left) && SvIOK_notUV(right)) | |
1989 | ? (SvIVX(left) >= SvIVX(right)) | |
1990 | : ( (do_ncmp(left, right) & 2) == 0) | |
1991 | )); | |
1992 | RETURN; | |
1993 | } | |
1b6737cc | 1994 | |
33efebe6 DM |
1995 | PP(pp_ne) |
1996 | { | |
1997 | dVAR; dSP; | |
1998 | SV *left, *right; | |
1999 | ||
2000 | tryAMAGICbin_MG(ne_amg, AMGf_set|AMGf_numeric); | |
2001 | right = POPs; | |
2002 | left = TOPs; | |
2003 | SETs(boolSV( | |
2004 | (SvIOK_notUV(left) && SvIOK_notUV(right)) | |
2005 | ? (SvIVX(left) != SvIVX(right)) | |
2006 | : (do_ncmp(left, right) != 0) | |
2007 | )); | |
2008 | RETURN; | |
2009 | } | |
1b6737cc | 2010 | |
33efebe6 DM |
2011 | /* compare left and right SVs. Returns: |
2012 | * -1: < | |
2013 | * 0: == | |
2014 | * 1: > | |
2015 | * 2: left or right was a NaN | |
2016 | */ | |
2017 | I32 | |
2018 | Perl_do_ncmp(pTHX_ SV* const left, SV * const right) | |
2019 | { | |
2020 | dVAR; | |
1b6737cc | 2021 | |
33efebe6 DM |
2022 | PERL_ARGS_ASSERT_DO_NCMP; |
2023 | #ifdef PERL_PRESERVE_IVUV | |
2024 | SvIV_please_nomg(right); | |
2025 | /* Fortunately it seems NaN isn't IOK */ | |
2026 | if (SvIOK(right)) { | |
2027 | SvIV_please_nomg(left); | |
2028 | if (SvIOK(left)) { | |
2029 | if (!SvUOK(left)) { | |
2030 | const IV leftiv = SvIVX(left); | |
2031 | if (!SvUOK(right)) { | |
2032 | /* ## IV <=> IV ## */ | |
2033 | const IV rightiv = SvIVX(right); | |
2034 | return (leftiv > rightiv) - (leftiv < rightiv); | |
28e5dec8 | 2035 | } |
33efebe6 DM |
2036 | /* ## IV <=> UV ## */ |
2037 | if (leftiv < 0) | |
2038 | /* As (b) is a UV, it's >=0, so it must be < */ | |
2039 | return -1; | |
2040 | { | |
2041 | const UV rightuv = SvUVX(right); | |
2042 | return ((UV)leftiv > rightuv) - ((UV)leftiv < rightuv); | |
28e5dec8 | 2043 | } |
28e5dec8 | 2044 | } |
79072805 | 2045 | |
33efebe6 DM |
2046 | if (SvUOK(right)) { |
2047 | /* ## UV <=> UV ## */ | |
2048 | const UV leftuv = SvUVX(left); | |
2049 | const UV rightuv = SvUVX(right); | |
2050 | return (leftuv > rightuv) - (leftuv < rightuv); | |
28e5dec8 | 2051 | } |
33efebe6 DM |
2052 | /* ## UV <=> IV ## */ |
2053 | { | |
2054 | const IV rightiv = SvIVX(right); | |
2055 | if (rightiv < 0) | |
2056 | /* As (a) is a UV, it's >=0, so it cannot be < */ | |
2057 | return 1; | |
2058 | { | |
2059 | const UV leftuv = SvUVX(left); | |
2060 | return (leftuv > (UV)rightiv) - (leftuv < (UV)rightiv); | |
28e5dec8 | 2061 | } |
28e5dec8 | 2062 | } |
33efebe6 | 2063 | /* NOTREACHED */ |
28e5dec8 JH |
2064 | } |
2065 | } | |
2066 | #endif | |
a0d0e21e | 2067 | { |
33efebe6 DM |
2068 | NV const rnv = SvNV_nomg(right); |
2069 | NV const lnv = SvNV_nomg(left); | |
2070 | ||
cab190d4 | 2071 | #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan) |
33efebe6 DM |
2072 | if (Perl_isnan(lnv) || Perl_isnan(rnv)) { |
2073 | return 2; | |
2074 | } | |
2075 | return (lnv > rnv) - (lnv < rnv); | |
cab190d4 | 2076 | #else |
33efebe6 DM |
2077 | if (lnv < rnv) |
2078 | return -1; | |
2079 | if (lnv > rnv) | |
2080 | return 1; | |
2081 | if (lnv == rnv) | |
2082 | return 0; | |
2083 | return 2; | |
cab190d4 | 2084 | #endif |
a0d0e21e | 2085 | } |
79072805 LW |
2086 | } |
2087 | ||
33efebe6 | 2088 | |
a0d0e21e | 2089 | PP(pp_ncmp) |
79072805 | 2090 | { |
33efebe6 DM |
2091 | dVAR; dSP; |
2092 | SV *left, *right; | |
2093 | I32 value; | |
a42d0242 | 2094 | tryAMAGICbin_MG(ncmp_amg, AMGf_numeric); |
33efebe6 DM |
2095 | right = POPs; |
2096 | left = TOPs; | |
2097 | value = do_ncmp(left, right); | |
2098 | if (value == 2) { | |
3280af22 | 2099 | SETs(&PL_sv_undef); |
79072805 | 2100 | } |
33efebe6 DM |
2101 | else { |
2102 | dTARGET; | |
2103 | SETi(value); | |
2104 | } | |
2105 | RETURN; | |
a0d0e21e | 2106 | } |
79072805 | 2107 | |
afd9910b | 2108 | PP(pp_sle) |
a0d0e21e | 2109 | { |
97aff369 | 2110 | dVAR; dSP; |
79072805 | 2111 | |
afd9910b NC |
2112 | int amg_type = sle_amg; |
2113 | int multiplier = 1; | |
2114 | int rhs = 1; | |
79072805 | 2115 | |
afd9910b NC |
2116 | switch (PL_op->op_type) { |
2117 | case OP_SLT: | |
2118 | amg_type = slt_amg; | |
2119 | /* cmp < 0 */ | |
2120 | rhs = 0; | |
2121 | break; | |
2122 | case OP_SGT: | |
2123 | amg_type = sgt_amg; | |
2124 | /* cmp > 0 */ | |
2125 | multiplier = -1; | |
2126 | rhs = 0; | |
2127 | break; | |
2128 | case OP_SGE: | |
2129 | amg_type = sge_amg; | |
2130 | /* cmp >= 0 */ | |
2131 | multiplier = -1; | |
2132 | break; | |
79072805 | 2133 | } |
79072805 | 2134 | |
6f1401dc | 2135 | tryAMAGICbin_MG(amg_type, AMGf_set); |
a0d0e21e LW |
2136 | { |
2137 | dPOPTOPssrl; | |
1b6737cc | 2138 | const int cmp = (IN_LOCALE_RUNTIME |
078504b2 FC |
2139 | ? sv_cmp_locale_flags(left, right, 0) |
2140 | : sv_cmp_flags(left, right, 0)); | |
afd9910b | 2141 | SETs(boolSV(cmp * multiplier < rhs)); |
a0d0e21e LW |
2142 | RETURN; |
2143 | } | |
2144 | } | |
79072805 | 2145 | |
36477c24 | 2146 | PP(pp_seq) |
2147 | { | |
6f1401dc DM |
2148 | dVAR; dSP; |
2149 | tryAMAGICbin_MG(seq_amg, AMGf_set); | |
36477c24 | 2150 | { |
2151 | dPOPTOPssrl; | |
078504b2 | 2152 | SETs(boolSV(sv_eq_flags(left, right, 0))); |
a0d0e21e LW |
2153 | RETURN; |
2154 | } | |
2155 | } | |
79072805 | 2156 | |
a0d0e21e | 2157 | PP(pp_sne) |
79072805 | 2158 | { |
6f1401dc DM |
2159 | dVAR; dSP; |
2160 | tryAMAGICbin_MG(sne_amg, AMGf_set); | |
a0d0e21e LW |
2161 | { |
2162 | dPOPTOPssrl; | |
078504b2 | 2163 | SETs(boolSV(!sv_eq_flags(left, right, 0))); |
a0d0e21e | 2164 | RETURN; |
463ee0b2 | 2165 | } |
79072805 LW |
2166 | } |
2167 | ||
a0d0e21e | 2168 | PP(pp_scmp) |
79072805 | 2169 | { |
6f1401dc DM |
2170 | dVAR; dSP; dTARGET; |
2171 | tryAMAGICbin_MG(scmp_amg, 0); | |
a0d0e21e LW |
2172 | { |
2173 | dPOPTOPssrl; | |
1b6737cc | 2174 | const int cmp = (IN_LOCALE_RUNTIME |
078504b2 FC |
2175 | ? sv_cmp_locale_flags(left, right, 0) |
2176 | : sv_cmp_flags(left, right, 0)); | |
bbce6d69 | 2177 | SETi( cmp ); |
a0d0e21e LW |
2178 | RETURN; |
2179 | } | |
2180 | } | |
79072805 | 2181 | |
55497cff | 2182 | PP(pp_bit_and) |
2183 | { | |
6f1401dc DM |
2184 | dVAR; dSP; dATARGET; |
2185 | tryAMAGICbin_MG(band_amg, AMGf_assign); | |
a0d0e21e LW |
2186 | { |
2187 | dPOPTOPssrl; | |
4633a7c4 | 2188 | if (SvNIOKp(left) || SvNIOKp(right)) { |
b20c4ee1 FC |
2189 | const bool left_ro_nonnum = !SvNIOKp(left) && SvREADONLY(left); |
2190 | const bool right_ro_nonnum = !SvNIOKp(right) && SvREADONLY(right); | |
d0ba1bd2 | 2191 | if (PL_op->op_private & HINT_INTEGER) { |
1b6737cc | 2192 | const IV i = SvIV_nomg(left) & SvIV_nomg(right); |
972b05a9 | 2193 | SETi(i); |
d0ba1bd2 JH |
2194 | } |
2195 | else { | |
1b6737cc | 2196 | const UV u = SvUV_nomg(left) & SvUV_nomg(right); |
972b05a9 | 2197 | SETu(u); |
d0ba1bd2 | 2198 | } |
b20c4ee1 FC |
2199 | if (left_ro_nonnum) SvNIOK_off(left); |
2200 | if (right_ro_nonnum) SvNIOK_off(right); | |
a0d0e21e LW |
2201 | } |
2202 | else { | |
533c011a | 2203 | do_vop(PL_op->op_type, TARG, left, right); |
a0d0e21e LW |
2204 | SETTARG; |
2205 | } | |
2206 | RETURN; | |
2207 | } | |
2208 | } | |
79072805 | 2209 | |
a0d0e21e LW |
2210 | PP(pp_bit_or) |
2211 | { | |
3658c1f1 NC |
2212 | dVAR; dSP; dATARGET; |
2213 | const int op_type = PL_op->op_type; | |
2214 | ||
6f1401dc | 2215 | tryAMAGICbin_MG((op_type == OP_BIT_OR ? bor_amg : bxor_amg), AMGf_assign); |
a0d0e21e LW |
2216 | { |
2217 | dPOPTOPssrl; | |
4633a7c4 | 2218 | if (SvNIOKp(left) || SvNIOKp(right)) { |
b20c4ee1 FC |
2219 | const bool left_ro_nonnum = !SvNIOKp(left) && SvREADONLY(left); |
2220 | const bool right_ro_nonnum = !SvNIOKp(right) && SvREADONLY(right); | |
d0ba1bd2 | 2221 | if (PL_op->op_private & HINT_INTEGER) { |
3658c1f1 NC |
2222 | const IV l = (USE_LEFT(left) ? SvIV_nomg(left) : 0); |
2223 | const IV r = SvIV_nomg(right); | |
2224 | const IV result = op_type == OP_BIT_OR ? (l | r) : (l ^ r); | |
2225 | SETi(result); | |
d0ba1bd2 JH |
2226 | } |
2227 | else { | |
3658c1f1 NC |
2228 | const UV l = (USE_LEFT(left) ? SvUV_nomg(left) : 0); |
2229 | const UV r = SvUV_nomg(right); | |
2230 | const UV result = op_type == OP_BIT_OR ? (l | r) : (l ^ r); | |
2231 | SETu(result); | |
d0ba1bd2 | 2232 | } |
b20c4ee1 FC |
2233 | if (left_ro_nonnum) SvNIOK_off(left); |
2234 | if (right_ro_nonnum) SvNIOK_off(right); | |
a0d0e21e LW |
2235 | } |
2236 | else { | |
3658c1f1 | 2237 | do_vop(op_type, TARG, left, right); |
a0d0e21e LW |
2238 | SETTARG; |
2239 | } | |
2240 | RETURN; | |
79072805 | 2241 | } |
a0d0e21e | 2242 | } |
79072805 | 2243 | |
a0d0e21e LW |
2244 | PP(pp_negate) |
2245 | { | |
6f1401dc DM |
2246 | dVAR; dSP; dTARGET; |
2247 | tryAMAGICun_MG(neg_amg, AMGf_numeric); | |
a0d0e21e | 2248 | { |
6f1401dc | 2249 | SV * const sv = TOPs; |
1b6737cc | 2250 | const int flags = SvFLAGS(sv); |
a5b92898 | 2251 | |
886a4465 | 2252 | if( !SvNIOK( sv ) && looks_like_number( sv ) ){ |
a5b92898 RB |
2253 | SvIV_please( sv ); |
2254 | } | |
2255 | ||
28e5dec8 JH |
2256 | if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) { |
2257 | /* It's publicly an integer, or privately an integer-not-float */ | |
2258 | oops_its_an_int: | |
9b0e499b GS |
2259 | if (SvIsUV(sv)) { |
2260 | if (SvIVX(sv) == IV_MIN) { | |
28e5dec8 | 2261 | /* 2s complement assumption. */ |
9b0e499b GS |
2262 | SETi(SvIVX(sv)); /* special case: -((UV)IV_MAX+1) == IV_MIN */ |
2263 | RETURN; | |
2264 | } | |
2265 | else if (SvUVX(sv) <= IV_MAX) { | |
beccb14c | 2266 | SETi(-SvIVX(sv)); |
9b0e499b GS |
2267 | RETURN; |
2268 | } | |
2269 | } | |
2270 | else if (SvIVX(sv) != IV_MIN) { | |
2271 | SETi(-SvIVX(sv)); | |
2272 | RETURN; | |
2273 | } | |
28e5dec8 JH |
2274 | #ifdef PERL_PRESERVE_IVUV |
2275 | else { | |
2276 | SETu((UV)IV_MIN); | |
2277 | RETURN; | |
2278 | } | |
2279 | #endif | |
9b0e499b GS |
2280 | } |
2281 | if (SvNIOKp(sv)) | |
6f1401dc | 2282 | SETn(-SvNV_nomg(sv)); |
4633a7c4 | 2283 | else if (SvPOKp(sv)) { |
a0d0e21e | 2284 | STRLEN len; |
6f1401dc | 2285 | const char * const s = SvPV_nomg_const(sv, len); |
bbce6d69 | 2286 | if (isIDFIRST(*s)) { |
76f68e9b | 2287 | sv_setpvs(TARG, "-"); |
a0d0e21e | 2288 | sv_catsv(TARG, sv); |
79072805 | 2289 | } |
a0d0e21e | 2290 | else if (*s == '+' || *s == '-') { |
6f1401dc DM |
2291 | sv_setsv_nomg(TARG, sv); |
2292 | *SvPV_force_nomg(TARG, len) = *s == '-' ? '+' : '-'; | |
79072805 | 2293 | } |
8eb28a70 | 2294 | else if (DO_UTF8(sv)) { |
6f1401dc | 2295 | SvIV_please_nomg(sv); |
8eb28a70 JH |
2296 | if (SvIOK(sv)) |
2297 | goto oops_its_an_int; | |
2298 | if (SvNOK(sv)) | |
6f1401dc | 2299 | sv_setnv(TARG, -SvNV_nomg(sv)); |
8eb28a70 | 2300 | else { |
76f68e9b | 2301 | sv_setpvs(TARG, "-"); |
8eb28a70 JH |
2302 | sv_catsv(TARG, sv); |
2303 | } | |
834a4ddd | 2304 | } |
28e5dec8 | 2305 | else { |
6f1401dc | 2306 | SvIV_please_nomg(sv); |
8eb28a70 JH |
2307 | if (SvIOK(sv)) |
2308 | goto oops_its_an_int; | |
6f1401dc | 2309 | sv_setnv(TARG, -SvNV_nomg(sv)); |
28e5dec8 | 2310 | } |
a0d0e21e | 2311 | SETTARG; |
79072805 | 2312 | } |
4633a7c4 | 2313 | else |
6f1401dc | 2314 | SETn(-SvNV_nomg(sv)); |
79072805 | 2315 | } |
a0d0e21e | 2316 | RETURN; |
79072805 LW |
2317 | } |
2318 | ||
a0d0e21e | 2319 | PP(pp_not) |
79072805 | 2320 | { |
6f1401dc DM |
2321 | dVAR; dSP; |
2322 | tryAMAGICun_MG(not_amg, AMGf_set); | |
06c841cf | 2323 | *PL_stack_sp = boolSV(!SvTRUE_nomg(*PL_stack_sp)); |
a0d0e21e | 2324 | return NORMAL; |
79072805 LW |
2325 | } |
2326 | ||
a0d0e21e | 2327 | PP(pp_complement) |
79072805 | 2328 | { |
6f1401dc | 2329 | dVAR; dSP; dTARGET; |
a42d0242 | 2330 | tryAMAGICun_MG(compl_amg, AMGf_numeric); |
a0d0e21e LW |
2331 | { |
2332 | dTOPss; | |
4633a7c4 | 2333 | if (SvNIOKp(sv)) { |
d0ba1bd2 | 2334 | if (PL_op->op_private & HINT_INTEGER) { |
1b6737cc | 2335 | const IV i = ~SvIV_nomg(sv); |
972b05a9 | 2336 | SETi(i); |
d0ba1bd2 JH |
2337 | } |
2338 | else { | |
1b6737cc | 2339 | const UV u = ~SvUV_nomg(sv); |
972b05a9 | 2340 | SETu(u); |
d0ba1bd2 | 2341 | } |
a0d0e21e LW |
2342 | } |
2343 | else { | |
51723571 | 2344 | register U8 *tmps; |
55497cff | 2345 | register I32 anum; |
a0d0e21e LW |
2346 | STRLEN len; |
2347 | ||
10516c54 | 2348 | (void)SvPV_nomg_const(sv,len); /* force check for uninit var */ |
891f9566 | 2349 | sv_setsv_nomg(TARG, sv); |
6f1401dc | 2350 | tmps = (U8*)SvPV_force_nomg(TARG, len); |
a0d0e21e | 2351 | anum = len; |
1d68d6cd | 2352 | if (SvUTF8(TARG)) { |
a1ca4561 | 2353 | /* Calculate exact length, let's not estimate. */ |
1d68d6cd | 2354 | STRLEN targlen = 0; |
ba210ebe | 2355 | STRLEN l; |
a1ca4561 YST |
2356 | UV nchar = 0; |
2357 | UV nwide = 0; | |
01f6e806 | 2358 | U8 * const send = tmps + len; |
74d49cd0 TS |
2359 | U8 * const origtmps = tmps; |
2360 | const UV utf8flags = UTF8_ALLOW_ANYUV; | |
1d68d6cd | 2361 | |
1d68d6cd | 2362 | while (tmps < send) { |
74d49cd0 TS |
2363 | const UV c = utf8n_to_uvchr(tmps, send-tmps, &l, utf8flags); |
2364 | tmps += l; | |
5bbb0b5a | 2365 | targlen += UNISKIP(~c); |
a1ca4561 YST |
2366 | nchar++; |
2367 | if (c > 0xff) | |
2368 | nwide++; | |
1d68d6cd SC |
2369 | } |
2370 | ||
2371 | /* Now rewind strings and write them. */ | |
74d49cd0 | 2372 | tmps = origtmps; |
a1ca4561 YST |
2373 | |
2374 | if (nwide) { | |
01f6e806 AL |
2375 | U8 *result; |
2376 | U8 *p; | |
2377 | ||
74d49cd0 | 2378 | Newx(result, targlen + 1, U8); |
01f6e806 | 2379 | p = result; |
a1ca4561 | 2380 | while (tmps < send) { |
74d49cd0 TS |
2381 | const UV c = utf8n_to_uvchr(tmps, send-tmps, &l, utf8flags); |
2382 | tmps += l; | |
01f6e806 | 2383 | p = uvchr_to_utf8_flags(p, ~c, UNICODE_ALLOW_ANY); |
a1ca4561 | 2384 | } |
01f6e806 | 2385 | *p = '\0'; |
c1c21316 NC |
2386 | sv_usepvn_flags(TARG, (char*)result, targlen, |
2387 | SV_HAS_TRAILING_NUL); | |
a1ca4561 YST |
2388 | SvUTF8_on(TARG); |
2389 | } | |
2390 | else { | |
01f6e806 AL |
2391 | U8 *result; |
2392 | U8 *p; | |
2393 | ||
74d49cd0 | 2394 | Newx(result, nchar + 1, U8); |
01f6e806 | 2395 | p = result; |
a1ca4561 | 2396 | while (tmps < send) { |
74d49cd0 TS |
2397 | const U8 c = (U8)utf8n_to_uvchr(tmps, send-tmps, &l, utf8flags); |
2398 | tmps += l; | |
01f6e806 | 2399 | *p++ = ~c; |
a1ca4561 | 2400 | } |
01f6e806 | 2401 | *p = '\0'; |
c1c21316 | 2402 | sv_usepvn_flags(TARG, (char*)result, nchar, SV_HAS_TRAILING_NUL); |
d0a21e00 | 2403 | SvUTF8_off(TARG); |
1d68d6cd | 2404 | } |
ec93b65f | 2405 | SETTARG; |
1d68d6cd SC |
2406 | RETURN; |
2407 | } | |
a0d0e21e | 2408 | #ifdef LIBERAL |
51723571 JH |
2409 | { |
2410 | register long *tmpl; | |
2411 | for ( ; anum && (unsigned long)tmps % sizeof(long); anum--, tmps++) | |
2412 | *tmps = ~*tmps; | |
2413 | tmpl = (long*)tmps; | |
bb7a0f54 | 2414 | for ( ; anum >= (I32)sizeof(long); anum -= (I32)sizeof(long), tmpl++) |
51723571 JH |
2415 | *tmpl = ~*tmpl; |
2416 | tmps = (U8*)tmpl; | |
2417 | } | |
a0d0e21e LW |
2418 | #endif |
2419 | for ( ; anum > 0; anum--, tmps++) | |
2420 | *tmps = ~*tmps; | |
ec93b65f | 2421 | SETTARG; |
a0d0e21e LW |
2422 | } |
2423 | RETURN; | |
2424 | } | |
79072805 LW |
2425 | } |
2426 | ||
a0d0e21e LW |
2427 | /* integer versions of some of the above */ |
2428 | ||
a0d0e21e | 2429 | PP(pp_i_multiply) |
79072805 | 2430 | { |
6f1401dc DM |
2431 | dVAR; dSP; dATARGET; |
2432 | tryAMAGICbin_MG(mult_amg, AMGf_assign); | |
a0d0e21e | 2433 | { |
6f1401dc | 2434 | dPOPTOPiirl_nomg; |
a0d0e21e LW |
2435 | SETi( left * right ); |
2436 | RETURN; | |
2437 | } | |
79072805 LW |
2438 | } |
2439 | ||
a0d0e21e | 2440 | PP(pp_i_divide) |
79072805 | 2441 | { |
85935d8e | 2442 | IV num; |
6f1401dc DM |
2443 | dVAR; dSP; dATARGET; |
2444 | tryAMAGICbin_MG(div_amg, AMGf_assign); | |
a0d0e21e | 2445 | { |
6f1401dc | 2446 | dPOPTOPssrl; |
85935d8e | 2447 | IV value = SvIV_nomg(right); |
a0d0e21e | 2448 | if (value == 0) |
ece1bcef | 2449 | DIE(aTHX_ "Illegal division by zero"); |
85935d8e | 2450 | num = SvIV_nomg(left); |
a0cec769 YST |
2451 | |
2452 | /* avoid FPE_INTOVF on some platforms when num is IV_MIN */ | |
2453 | if (value == -1) | |
2454 | value = - num; | |
2455 | else | |
2456 | value = num / value; | |
6f1401dc | 2457 | SETi(value); |
a0d0e21e LW |
2458 | RETURN; |
2459 | } | |
79072805 LW |
2460 | } |
2461 | ||
befad5d1 | 2462 | #if defined(__GLIBC__) && IVSIZE == 8 |
224ec323 JH |
2463 | STATIC |
2464 | PP(pp_i_modulo_0) | |
befad5d1 NC |
2465 | #else |
2466 | PP(pp_i_modulo) | |
2467 | #endif | |
224ec323 JH |
2468 | { |
2469 | /* This is the vanilla old i_modulo. */ | |
6f1401dc DM |
2470 | dVAR; dSP; dATARGET; |
2471 | tryAMAGICbin_MG(modulo_amg, AMGf_assign); | |
224ec323 | 2472 | { |
6f1401dc | 2473 | dPOPTOPiirl_nomg; |
224ec323 JH |
2474 | if (!right) |
2475 | DIE(aTHX_ "Illegal modulus zero"); | |
a0cec769 YST |
2476 | /* avoid FPE_INTOVF on some platforms when left is IV_MIN */ |
2477 | if (right == -1) | |
2478 | SETi( 0 ); | |
2479 | else | |
2480 | SETi( left % right ); | |
224ec323 JH |
2481 | RETURN; |
2482 | } | |
2483 | } | |
2484 | ||
11010fa3 | 2485 | #if defined(__GLIBC__) && IVSIZE == 8 |
224ec323 JH |
2486 | STATIC |
2487 | PP(pp_i_modulo_1) | |
befad5d1 | 2488 | |
224ec323 | 2489 | { |
224ec323 | 2490 | /* This is the i_modulo with the workaround for the _moddi3 bug |
fce2b89e | 2491 | * in (at least) glibc 2.2.5 (the PERL_ABS() the workaround). |
224ec323 | 2492 | * See below for pp_i_modulo. */ |
6f1401dc DM |
2493 | dVAR; dSP; dATARGET; |
2494 | tryAMAGICbin_MG(modulo_amg, AMGf_assign); | |
224ec323 | 2495 | { |
6f1401dc | 2496 | dPOPTOPiirl_nomg; |
224ec323 JH |
2497 | if (!right) |
2498 | DIE(aTHX_ "Illegal modulus zero"); | |
a0cec769 YST |
2499 | /* avoid FPE_INTOVF on some platforms when left is IV_MIN */ |
2500 | if (right == -1) | |
2501 | SETi( 0 ); | |
2502 | else | |
2503 | SETi( left % PERL_ABS(right) ); | |
224ec323 JH |
2504 | RETURN; |
2505 | } | |
224ec323 JH |
2506 | } |
2507 | ||
a0d0e21e | 2508 | PP(pp_i_modulo) |
79072805 | 2509 | { |
6f1401dc DM |
2510 | dVAR; dSP; dATARGET; |
2511 | tryAMAGICbin_MG(modulo_amg, AMGf_assign); | |
224ec323 | 2512 | { |
6f1401dc | 2513 | dPOPTOPiirl_nomg; |
224ec323 JH |
2514 | if (!right) |
2515 | DIE(aTHX_ "Illegal modulus zero"); | |
2516 | /* The assumption is to use hereafter the old vanilla version... */ | |
2517 | PL_op->op_ppaddr = | |
2518 | PL_ppaddr[OP_I_MODULO] = | |
1c127fab | 2519 | Perl_pp_i_modulo_0; |
224ec323 JH |
2520 | /* .. but if we have glibc, we might have a buggy _moddi3 |
2521 | * (at least glicb 2.2.5 is known to have this bug), in other | |
2522 | * words our integer modulus with negative quad as the second | |
2523 | * argument might be broken. Test for this and re-patch the | |
2524 | * opcode dispatch table if that is the case, remembering to | |
2525 | * also apply the workaround so that this first round works | |
2526 | * right, too. See [perl #9402] for more information. */ | |
224ec323 JH |
2527 | { |
2528 | IV l = 3; | |
2529 | IV r = -10; | |
2530 | /* Cannot do this check with inlined IV constants since | |
2531 | * that seems to work correctly even with the buggy glibc. */ | |
2532 | if (l % r == -3) { | |
2533 | /* Yikes, we have the bug. | |
2534 | * Patch in the workaround version. */ | |
2535 | PL_op->op_ppaddr = | |
2536 | PL_ppaddr[OP_I_MODULO] = | |
2537 | &Perl_pp_i_modulo_1; | |
2538 | /* Make certain we work right this time, too. */ | |
32fdb065 | 2539 | right = PERL_ABS(right); |
224ec323 JH |
2540 | } |
2541 | } | |
a0cec769 YST |
2542 | /* avoid FPE_INTOVF on some platforms when left is IV_MIN */ |
2543 | if (right == -1) | |
2544 | SETi( 0 ); | |
2545 | else | |
2546 | SETi( left % right ); | |
224ec323 JH |
2547 | RETURN; |
2548 | } | |
79072805 | 2549 | } |
befad5d1 | 2550 | #endif |
79072805 | 2551 | |
a0d0e21e | 2552 | PP(pp_i_add) |
79072805 | 2553 | { |
6f1401dc DM |
2554 | dVAR; dSP; dATARGET; |
2555 | tryAMAGICbin_MG(add_amg, AMGf_assign); | |
a0d0e21e | 2556 | { |
6f1401dc | 2557 | dPOPTOPiirl_ul_nomg; |
a0d0e21e LW |
2558 | SETi( left + right ); |
2559 | RETURN; | |
79072805 | 2560 | } |
79072805 LW |
2561 | } |
2562 | ||
a0d0e21e | 2563 | PP(pp_i_subtract) |
79072805 | 2564 | { |
6f1401dc DM |
2565 | dVAR; dSP; dATARGET; |
2566 | tryAMAGICbin_MG(subtr_amg, AMGf_assign); | |
a0d0e21e | 2567 | { |
6f1401dc | 2568 | dPOPTOPiirl_ul_nomg; |
a0d0e21e LW |
2569 | SETi( left - right ); |
2570 | RETURN; | |
79072805 | 2571 | } |
79072805 LW |
2572 | } |
2573 | ||
a0d0e21e | 2574 | PP(pp_i_lt) |
79072805 | 2575 | { |
6f1401dc DM |
2576 | dVAR; dSP; |
2577 | tryAMAGICbin_MG(lt_amg, AMGf_set); | |
a0d0e21e | 2578 | { |
96b6b87f | 2579 | dPOPTOPiirl_nomg; |
54310121 | 2580 | SETs(boolSV(left < right)); |
a0d0e21e LW |
2581 | RETURN; |
2582 | } | |
79072805 LW |
2583 | } |
2584 | ||
a0d0e21e | 2585 | PP(pp_i_gt) |
79072805 | 2586 | { |
6f1401dc DM |
2587 | dVAR; dSP; |
2588 | tryAMAGICbin_MG(gt_amg, AMGf_set); | |
a0d0e21e | 2589 | { |
96b6b87f | 2590 | dPOPTOPiirl_nomg; |
54310121 | 2591 | SETs(boolSV(left > right)); |
a0d0e21e LW |
2592 | RETURN; |
2593 | } | |
79072805 LW |
2594 | } |
2595 | ||
a0d0e21e | 2596 | PP(pp_i_le) |
79072805 | 2597 | { |
6f1401dc DM |
2598 | dVAR; dSP; |
2599 | tryAMAGICbin_MG(le_amg, AMGf_set); | |
a0d0e21e | 2600 | { |
96b6b87f | 2601 | dPOPTOPiirl_nomg; |
54310121 | 2602 | SETs(boolSV(left <= right)); |
a0d0e21e | 2603 | RETURN; |
85e6fe83 | 2604 | } |
79072805 LW |
2605 | } |
2606 | ||
a0d0e21e | 2607 | PP(pp_i_ge) |
79072805 | 2608 | { |
6f1401dc DM |
2609 | dVAR; dSP; |
2610 | tryAMAGICbin_MG(ge_amg, AMGf_set); | |
a0d0e21e | 2611 | { |
96b6b87f | 2612 | dPOPTOPiirl_nomg; |
54310121 | 2613 | SETs(boolSV(left >= right)); |
a0d0e21e LW |
2614 | RETURN; |
2615 | } | |
79072805 LW |
2616 | } |
2617 | ||
a0d0e21e | 2618 | PP(pp_i_eq) |
79072805 | 2619 | { |
6f1401dc DM |
2620 | dVAR; dSP; |
2621 | tryAMAGICbin_MG(eq_amg, AMGf_set); | |
a0d0e21e | 2622 | { |
96b6b87f | 2623 | dPOPTOPiirl_nomg; |
54310121 | 2624 | SETs(boolSV(left == right)); |
a0d0e21e LW |
2625 | RETURN; |
2626 | } | |
79072805 LW |
2627 | } |
2628 | ||
a0d0e21e | 2629 | PP(pp_i_ne) |
79072805 | 2630 | { |
6f1401dc DM |
2631 | dVAR; dSP; |
2632 | tryAMAGICbin_MG(ne_amg, AMGf_set); | |
a0d0e21e | 2633 | { |
96b6b87f | 2634 | dPOPTOPiirl_nomg; |
54310121 | 2635 | SETs(boolSV(left != right)); |
a0d0e21e LW |
2636 | RETURN; |
2637 | } | |
79072805 LW |
2638 | } |
2639 | ||
a0d0e21e | 2640 | PP(pp_i_ncmp) |
79072805 | 2641 | { |
6f1401dc DM |
2642 | dVAR; dSP; dTARGET; |
2643 | tryAMAGICbin_MG(ncmp_amg, 0); | |
a0d0e21e | 2644 | { |
96b6b87f | 2645 | dPOPTOPiirl_nomg; |
a0d0e21e | 2646 | I32 value; |
79072805 | 2647 | |
a0d0e21e | 2648 | if (left > right) |
79072805 | 2649 | value = 1; |
a0d0e21e | 2650 | else if (left < right) |
79072805 | 2651 | value = -1; |
a0d0e21e | 2652 | else |
79072805 | 2653 | value = 0; |
a0d0e21e LW |
2654 | SETi(value); |
2655 | RETURN; | |
79072805 | 2656 | } |
85e6fe83 LW |
2657 | } |
2658 | ||
2659 | PP(pp_i_negate) | |
2660 | { | |
6f1401dc DM |
2661 | dVAR; dSP; dTARGET; |
2662 | tryAMAGICun_MG(neg_amg, 0); | |
2663 | { | |
2664 | SV * const sv = TOPs; | |
2665 | IV const i = SvIV_nomg(sv); | |
2666 | SETi(-i); | |
2667 | RETURN; | |
2668 | } | |
85e6fe83 LW |
2669 | } |
2670 | ||
79072805 LW |
2671 | /* High falutin' math. */ |
2672 | ||
2673 | PP(pp_atan2) | |
2674 | { | |
6f1401dc DM |
2675 | dVAR; dSP; dTARGET; |
2676 | tryAMAGICbin_MG(atan2_amg, 0); | |
a0d0e21e | 2677 | { |
096c060c | 2678 | dPOPTOPnnrl_nomg; |
a1021d57 | 2679 | SETn(Perl_atan2(left, right)); |
a0d0e21e LW |
2680 | RETURN; |
2681 | } | |
79072805 LW |
2682 | } |
2683 | ||
2684 | PP(pp_sin) | |
2685 | { | |
71302fe3 NC |
2686 | dVAR; dSP; dTARGET; |
2687 | int amg_type = sin_amg; | |
2688 | const char *neg_report = NULL; | |
bc81784a | 2689 | NV (*func)(NV) = Perl_sin; |
71302fe3 NC |
2690 | const int op_type = PL_op->op_type; |
2691 | ||
2692 | switch (op_type) { | |
2693 | case OP_COS: | |
2694 | amg_type = cos_amg; | |
bc81784a | 2695 | func = Perl_cos; |
71302fe3 NC |
2696 | break; |
2697 | case OP_EXP: | |
2698 | amg_type = exp_amg; | |
bc81784a | 2699 | func = Perl_exp; |
71302fe3 NC |
2700 | break; |
2701 | case OP_LOG: | |
2702 | amg_type = log_amg; | |
bc81784a | 2703 | func = Perl_log; |
71302fe3 NC |
2704 | neg_report = "log"; |
2705 | break; | |
2706 | case OP_SQRT: | |
2707 | amg_type = sqrt_amg; | |
bc81784a | 2708 | func = Perl_sqrt; |
71302fe3 NC |
2709 | neg_report = "sqrt"; |
2710 | break; | |
a0d0e21e | 2711 | } |
79072805 | 2712 | |
6f1401dc DM |
2713 | |
2714 | tryAMAGICun_MG(amg_type, 0); | |
a0d0e21e | 2715 | { |
6f1401dc DM |
2716 | SV * const arg = POPs; |
2717 | const NV value = SvNV_nomg(arg); | |
71302fe3 NC |
2718 | if (neg_report) { |
2719 | if (op_type == OP_LOG ? (value <= 0.0) : (value < 0.0)) { | |
2720 | SET_NUMERIC_STANDARD(); | |
2721 | DIE(aTHX_ "Can't take %s of %"NVgf, neg_report, value); | |
2722 | } | |
2723 | } | |
2724 | XPUSHn(func(value)); | |
a0d0e21e LW |
2725 | RETURN; |
2726 | } | |
79072805 LW |
2727 | } |
2728 | ||
56cb0a1c AD |
2729 | /* Support Configure command-line overrides for rand() functions. |
2730 | After 5.005, perhaps we should replace this by Configure support | |
2731 | for drand48(), random(), or rand(). For 5.005, though, maintain | |
2732 | compatibility by calling rand() but allow the user to override it. | |
2733 | See INSTALL for details. --Andy Dougherty 15 July 1998 | |
2734 | */ | |
85ab1d1d JH |
2735 | /* Now it's after 5.005, and Configure supports drand48() and random(), |
2736 | in addition to rand(). So the overrides should not be needed any more. | |
2737 | --Jarkko Hietaniemi 27 September 1998 | |
2738 | */ | |
2739 | ||
2740 | #ifndef HAS_DRAND48_PROTO | |
20ce7b12 | 2741 | extern double drand48 (void); |
56cb0a1c AD |
2742 | #endif |
2743 | ||
79072805 LW |
2744 | PP(pp_rand) |
2745 | { | |
97aff369 | 2746 | dVAR; dSP; dTARGET; |
65202027 | 2747 | NV value; |
79072805 LW |
2748 | if (MAXARG < 1) |
2749 | value = 1.0; | |
94ec06bc FC |
2750 | else if (!TOPs) { |
2751 | value = 1.0; (void)POPs; | |
2752 | } | |
79072805 LW |
2753 | else |
2754 | value = POPn; | |
2755 | if (value == 0.0) | |
2756 | value = 1.0; | |
80252599 | 2757 | if (!PL_srand_called) { |
85ab1d1d | 2758 | (void)seedDrand01((Rand_seed_t)seed()); |
80252599 | 2759 | PL_srand_called = TRUE; |
93dc8474 | 2760 | } |
85ab1d1d | 2761 | value *= Drand01(); |
79072805 LW |
2762 | XPUSHn(value); |
2763 | RETURN; | |
2764 | } | |
2765 | ||
2766 | PP(pp_srand) | |
2767 | { | |
83832992 | 2768 | dVAR; dSP; dTARGET; |
d22667bf | 2769 | const UV anum = (MAXARG < 1 || (!TOPs && !POPs)) ? seed() : POPu; |
85ab1d1d | 2770 | (void)seedDrand01((Rand_seed_t)anum); |
80252599 | 2771 | PL_srand_called = TRUE; |
da1010ec NC |
2772 | if (anum) |
2773 | XPUSHu(anum); | |
2774 | else { | |
2775 | /* Historically srand always returned true. We can avoid breaking | |
2776 | that like this: */ | |
2777 | sv_setpvs(TARG, "0 but true"); | |
2778 | XPUSHTARG; | |
2779 | } | |
83832992 | 2780 | RETURN; |
79072805 LW |
2781 | } |
2782 | ||
79072805 LW |
2783 | PP(pp_int) |
2784 | { | |
6f1401dc DM |
2785 | dVAR; dSP; dTARGET; |
2786 | tryAMAGICun_MG(int_amg, AMGf_numeric); | |
774d564b | 2787 | { |
6f1401dc DM |
2788 | SV * const sv = TOPs; |
2789 | const IV iv = SvIV_nomg(sv); | |
28e5dec8 JH |
2790 | /* XXX it's arguable that compiler casting to IV might be subtly |
2791 | different from modf (for numbers inside (IV_MIN,UV_MAX)) in which | |
2792 | else preferring IV has introduced a subtle behaviour change bug. OTOH | |
2793 | relying on floating point to be accurate is a bug. */ | |
2794 | ||
c781a409 | 2795 | if (!SvOK(sv)) { |
922c4365 | 2796 | SETu(0); |
c781a409 RD |
2797 | } |
2798 | else if (SvIOK(sv)) { | |
2799 | if (SvIsUV(sv)) | |
6f1401dc | 2800 | SETu(SvUV_nomg(sv)); |
c781a409 | 2801 | else |
28e5dec8 | 2802 | SETi(iv); |
c781a409 | 2803 | } |
c781a409 | 2804 | else { |
6f1401dc | 2805 | const NV value = SvNV_nomg(sv); |
1048ea30 | 2806 | if (value >= 0.0) { |
28e5dec8 JH |
2807 | if (value < (NV)UV_MAX + 0.5) { |
2808 | SETu(U_V(value)); | |
2809 | } else { | |
059a1014 | 2810 | SETn(Perl_floor(value)); |
28e5dec8 | 2811 | } |
1048ea30 | 2812 | } |
28e5dec8 JH |
2813 | else { |
2814 | if (value > (NV)IV_MIN - 0.5) { | |
2815 | SETi(I_V(value)); | |
2816 | } else { | |
1bbae031 | 2817 | SETn(Perl_ceil(value)); |
28e5dec8 JH |
2818 | } |
2819 | } | |
774d564b | 2820 | } |
79072805 | 2821 | } |
79072805 LW |
2822 | RETURN; |
2823 | } | |
2824 | ||
463ee0b2 LW |
2825 | PP(pp_abs) |
2826 | { | |
6f1401dc DM |
2827 | dVAR; dSP; dTARGET; |
2828 | tryAMAGICun_MG(abs_amg, AMGf_numeric); | |
a0d0e21e | 2829 | { |
6f1401dc | 2830 | SV * const sv = TOPs; |
28e5dec8 | 2831 | /* This will cache the NV value if string isn't actually integer */ |
6f1401dc | 2832 | const IV iv = SvIV_nomg(sv); |
a227d84d | 2833 | |
800401ee | 2834 | if (!SvOK(sv)) { |
922c4365 | 2835 | SETu(0); |
800401ee JH |
2836 | } |
2837 | else if (SvIOK(sv)) { | |
28e5dec8 | 2838 | /* IVX is precise */ |
800401ee | 2839 | if (SvIsUV(sv)) { |
6f1401dc | 2840 | SETu(SvUV_nomg(sv)); /* force it to be numeric only */ |
28e5dec8 JH |
2841 | } else { |
2842 | if (iv >= 0) { | |
2843 | SETi(iv); | |
2844 | } else { | |
2845 | if (iv != IV_MIN) { | |
2846 | SETi(-iv); | |
2847 | } else { | |
2848 | /* 2s complement assumption. Also, not really needed as | |
2849 | IV_MIN and -IV_MIN should both be %100...00 and NV-able */ | |
2850 | SETu(IV_MIN); | |
2851 | } | |
a227d84d | 2852 | } |
28e5dec8 JH |
2853 | } |
2854 | } else{ | |
6f1401dc | 2855 | const NV value = SvNV_nomg(sv); |
774d564b | 2856 | if (value < 0.0) |
1b6737cc | 2857 | SETn(-value); |
a4474c9e DD |
2858 | else |
2859 | SETn(value); | |
774d564b | 2860 | } |
a0d0e21e | 2861 | } |
774d564b | 2862 | RETURN; |
463ee0b2 LW |
2863 | } |
2864 | ||
79072805 LW |
2865 | PP(pp_oct) |
2866 | { | |
97aff369 | 2867 | dVAR; dSP; dTARGET; |
5c144d81 | 2868 | const char *tmps; |
53305cf1 | 2869 | I32 flags = PERL_SCAN_ALLOW_UNDERSCORES; |
6f894ead | 2870 | STRLEN len; |
53305cf1 NC |
2871 | NV result_nv; |
2872 | UV result_uv; | |
1b6737cc | 2873 | SV* const sv = POPs; |
79072805 | 2874 | |
349d4f2f | 2875 | tmps = (SvPV_const(sv, len)); |
2bc69dc4 NIS |
2876 | if (DO_UTF8(sv)) { |
2877 | /* If Unicode, try to downgrade | |
2878 | * If not possible, croak. */ | |
1b6737cc | 2879 | SV* const tsv = sv_2mortal(newSVsv(sv)); |
2bc69dc4 NIS |
2880 | |
2881 | SvUTF8_on(tsv); | |
2882 | sv_utf8_downgrade(tsv, FALSE); | |
349d4f2f | 2883 | tmps = SvPV_const(tsv, len); |
2bc69dc4 | 2884 | } |
daa2adfd NC |
2885 | if (PL_op->op_type == OP_HEX) |
2886 | goto hex; | |
2887 | ||
6f894ead | 2888 | while (*tmps && len && isSPACE(*tmps)) |
53305cf1 | 2889 | tmps++, len--; |
9e24b6e2 | 2890 | if (*tmps == '0') |
53305cf1 | 2891 | tmps++, len--; |
a674e8db | 2892 | if (*tmps == 'x' || *tmps == 'X') { |
daa2adfd | 2893 | hex: |
53305cf1 | 2894 | result_uv = grok_hex (tmps, &len, &flags, &result_nv); |
daa2adfd | 2895 | } |
a674e8db | 2896 | else if (*tmps == 'b' || *tmps == 'B') |
53305cf1 | 2897 | result_uv = grok_bin (tmps, &len, &flags, &result_nv); |
464e2e8a | 2898 | else |
53305cf1 NC |
2899 | result_uv = grok_oct (tmps, &len, &flags, &result_nv); |
2900 | ||
2901 | if (flags & PERL_SCAN_GREATER_THAN_UV_MAX) { | |
2902 | XPUSHn(result_nv); | |
2903 | } | |
2904 | else { | |
2905 | XPUSHu(result_uv); | |
2906 | } | |
79072805 LW |
2907 | RETURN; |
2908 | } | |
2909 | ||
2910 | /* String stuff. */ | |
2911 | ||
2912 | PP(pp_length) | |
2913 | { | |
97aff369 | 2914 | dVAR; dSP; dTARGET; |
0bd48802 | 2915 | SV * const sv = TOPs; |
a0ed51b3 | 2916 | |
656266fc | 2917 | if (SvGAMAGIC(sv)) { |
9f621bb0 NC |
2918 | /* For an overloaded or magic scalar, we can't know in advance if |
2919 | it's going to be UTF-8 or not. Also, we can't call sv_len_utf8 as | |
2920 | it likes to cache the length. Maybe that should be a documented | |
2921 | feature of it. | |
92331800 NC |
2922 | */ |
2923 | STRLEN len; | |
9f621bb0 NC |
2924 | const char *const p |
2925 | = sv_2pv_flags(sv, &len, | |
2926 | SV_UNDEF_RETURNS_NULL|SV_CONST_RETURN|SV_GMAGIC); | |
92331800 | 2927 | |
d88e091f | 2928 | if (!p) { |
9407f9c1 DL |
2929 | if (!SvPADTMP(TARG)) { |
2930 | sv_setsv(TARG, &PL_sv_undef); | |
2931 | SETTARG; | |
2932 | } | |
2933 | SETs(&PL_sv_undef); | |
d88e091f | 2934 | } |
9f621bb0 | 2935 | else if (DO_UTF8(sv)) { |
899be101 | 2936 | SETi(utf8_length((U8*)p, (U8*)p + len)); |
92331800 NC |
2937 | } |
2938 | else | |
2939 | SETi(len); | |
656266fc | 2940 | } else if (SvOK(sv)) { |
9f621bb0 NC |
2941 | /* Neither magic nor overloaded. */ |
2942 | if (DO_UTF8(sv)) | |
2943 | SETi(sv_len_utf8(sv)); | |
2944 | else | |
2945 | SETi(sv_len(sv)); | |
656266fc | 2946 | } else { |
9407f9c1 DL |
2947 | if (!SvPADTMP(TARG)) { |
2948 | sv_setsv_nomg(TARG, &PL_sv_undef); | |
2949 | SETTARG; | |
2950 | } | |
2951 | SETs(&PL_sv_undef); | |
92331800 | 2952 | } |
79072805 LW |
2953 | RETURN; |
2954 | } | |
2955 | ||
2956 | PP(pp_substr) | |
2957 | { | |
97aff369 | 2958 | dVAR; dSP; dTARGET; |
79072805 | 2959 | SV *sv; |
463ee0b2 | 2960 | STRLEN curlen; |
9402d6ed | 2961 | STRLEN utf8_curlen; |
777f7c56 EB |
2962 | SV * pos_sv; |
2963 | IV pos1_iv; | |
2964 | int pos1_is_uv; | |
2965 | IV pos2_iv; | |
2966 | int pos2_is_uv; | |
2967 | SV * len_sv; | |
2968 | IV len_iv = 0; | |
2969 | int len_is_uv = 1; | |
050e6362 | 2970 | const I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET; |
e1ec3a88 | 2971 | const char *tmps; |
9402d6ed | 2972 | SV *repl_sv = NULL; |
cbbf8932 | 2973 | const char *repl = NULL; |
7b8d334a | 2974 | STRLEN repl_len; |
7bc95ae1 | 2975 | int num_args = PL_op->op_private & 7; |
13e30c65 | 2976 | bool repl_need_utf8_upgrade = FALSE; |
9402d6ed | 2977 | bool repl_is_utf8 = FALSE; |
79072805 | 2978 | |
78f9721b SM |
2979 | if (num_args > 2) { |
2980 | if (num_args > 3) { | |
7bc95ae1 | 2981 | if((repl_sv = POPs)) { |
83003860 | 2982 | repl = SvPV_const(repl_sv, repl_len); |
9402d6ed | 2983 | repl_is_utf8 = DO_UTF8(repl_sv) && SvCUR(repl_sv); |
7bc95ae1 FC |
2984 | } |
2985 | else num_args--; | |
2986 | } | |
2987 | if ((len_sv = POPs)) { | |
2988 | len_iv = SvIV(len_sv); | |
2989 | len_is_uv = SvIOK_UV(len_sv); | |
7b8d334a | 2990 | } |
7bc95ae1 | 2991 | else num_args--; |
5d82c453 | 2992 | } |
777f7c56 EB |
2993 | pos_sv = POPs; |
2994 | pos1_iv = SvIV(pos_sv); | |
2995 | pos1_is_uv = SvIOK_UV(pos_sv); | |
79072805 | 2996 | sv = POPs; |
849ca7ee | 2997 | PUTBACK; |
9402d6ed JH |
2998 | if (repl_sv) { |
2999 | if (repl_is_utf8) { | |
3000 | if (!DO_UTF8(sv)) | |
3001 | sv_utf8_upgrade(sv); | |
3002 | } | |
13e30c65 JH |
3003 | else if (DO_UTF8(sv)) |
3004 | repl_need_utf8_upgrade = TRUE; | |
9402d6ed | 3005 | } |
5c144d81 | 3006 | tmps = SvPV_const(sv, curlen); |
7e2040f0 | 3007 | if (DO_UTF8(sv)) { |
9402d6ed JH |
3008 | utf8_curlen = sv_len_utf8(sv); |
3009 | if (utf8_curlen == curlen) | |
3010 | utf8_curlen = 0; | |
a0ed51b3 | 3011 | else |
9402d6ed | 3012 | curlen = utf8_curlen; |
a0ed51b3 | 3013 | } |
d1c2b58a | 3014 | else |
9402d6ed | 3015 | utf8_curlen = 0; |
a0ed51b3 | 3016 | |
e1dccc0d Z |
3017 | if (!pos1_is_uv && pos1_iv < 0 && curlen) { |
3018 | pos1_is_uv = curlen-1 > ~(UV)pos1_iv; | |
3019 | pos1_iv += curlen; | |
777f7c56 | 3020 | } |
e1dccc0d Z |
3021 | if ((pos1_is_uv || pos1_iv > 0) && (UV)pos1_iv > curlen) |
3022 | goto bound_fail; | |
777f7c56 EB |
3023 | |
3024 | if (num_args > 2) { | |
3025 | if (!len_is_uv && len_iv < 0) { | |
3026 | pos2_iv = curlen + len_iv; | |
3027 | if (curlen) | |
3028 | pos2_is_uv = curlen-1 > ~(UV)len_iv; | |
3029 | else | |
3030 | pos2_is_uv = 0; | |
3031 | } else { /* len_iv >= 0 */ | |
3032 | if (!pos1_is_uv && pos1_iv < 0) { | |
3033 | pos2_iv = pos1_iv + len_iv; | |
3034 | pos2_is_uv = (UV)len_iv > (UV)IV_MAX; | |
3035 | } else { | |
3036 | if ((UV)len_iv > curlen-(UV)pos1_iv) | |
3037 | pos2_iv = curlen; | |
3038 | else | |
3039 | pos2_iv = pos1_iv+len_iv; | |
3040 | pos2_is_uv = 1; | |
3041 | } | |
5d82c453 | 3042 | } |
2304df62 | 3043 | } |
79072805 | 3044 | else { |
777f7c56 EB |
3045 | pos2_iv = curlen; |
3046 | pos2_is_uv = 1; | |
3047 | } | |
3048 | ||
3049 | if (!pos2_is_uv && pos2_iv < 0) { | |
3050 | if (!pos1_is_uv && pos1_iv < 0) | |
1c900557 | 3051 | goto bound_fail; |
777f7c56 EB |
3052 | pos2_iv = 0; |
3053 | } | |
3054 | else if (!pos1_is_uv && pos1_iv < 0) | |
3055 | pos1_iv = 0; | |
3056 | ||
3057 | if ((UV)pos2_iv < (UV)pos1_iv) | |
3058 | pos2_iv = pos1_iv; | |
3059 | if ((UV)pos2_iv > curlen) | |
3060 | pos2_iv = curlen; | |
3061 | ||
3062 | { | |
3063 | /* pos1_iv and pos2_iv both in 0..curlen, so the cast is safe */ | |
3064 | const STRLEN pos = (STRLEN)( (UV)pos1_iv ); | |
3065 | const STRLEN len = (STRLEN)( (UV)pos2_iv - (UV)pos1_iv ); | |
777f7c56 | 3066 | STRLEN byte_len = len; |
d931b1be NC |
3067 | STRLEN byte_pos = utf8_curlen |
3068 | ? sv_pos_u2b_flags(sv, pos, &byte_len, SV_CONST_RETURN) : pos; | |
3069 | ||
2154eca7 EB |
3070 | if (lvalue && !repl) { |
3071 | SV * ret; | |
3072 | ||
3073 | if (!SvGMAGICAL(sv)) { | |
3074 | if (SvROK(sv)) { | |
3075 | SvPV_force_nolen(sv); | |
3076 | Perl_ck_warner(aTHX_ packWARN(WARN_SUBSTR), | |
3077 | "Attempt to use reference as lvalue in substr"); | |
3078 | } | |
3079 | if (isGV_with_GP(sv)) | |
3080 | SvPV_force_nolen(sv); | |
3081 | else if (SvOK(sv)) /* is it defined ? */ | |
3082 | (void)SvPOK_only_UTF8(sv); | |
3083 | else | |
3084 | sv_setpvs(sv, ""); /* avoid lexical reincarnation */ | |
781e7547 | 3085 | } |
2154eca7 EB |
3086 | |
3087 | ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */ | |
3088 | sv_magic(ret, NULL, PERL_MAGIC_substr, NULL, 0); | |
3089 | LvTYPE(ret) = 'x'; | |
3090 | LvTARG(ret) = SvREFCNT_inc_simple(sv); | |
3091 | LvTARGOFF(ret) = pos; | |
3092 | LvTARGLEN(ret) = len; | |
3093 | ||
3094 | SPAGAIN; | |
3095 | PUSHs(ret); /* avoid SvSETMAGIC here */ | |
3096 | RETURN; | |
781e7547 DM |
3097 | } |
3098 | ||
2154eca7 EB |
3099 | SvTAINTED_off(TARG); /* decontaminate */ |
3100 | SvUTF8_off(TARG); /* decontaminate */ | |
3101 | ||
3102 | tmps += byte_pos; | |
777f7c56 | 3103 | sv_setpvn(TARG, tmps, byte_len); |
12aa1545 | 3104 | #ifdef USE_LOCALE_COLLATE |
14befaf4 | 3105 | sv_unmagic(TARG, PERL_MAGIC_collxfrm); |
12aa1545 | 3106 | #endif |
9402d6ed | 3107 | if (utf8_curlen) |
7f66633b | 3108 | SvUTF8_on(TARG); |
2154eca7 | 3109 | |
f7928d6c | 3110 | if (repl) { |
13e30c65 JH |
3111 | SV* repl_sv_copy = NULL; |
3112 | ||
3113 | if (repl_need_utf8_upgrade) { | |
3114 | repl_sv_copy = newSVsv(repl_sv); | |
3115 | sv_utf8_upgrade(repl_sv_copy); | |
349d4f2f | 3116 | repl = SvPV_const(repl_sv_copy, repl_len); |
13e30c65 JH |
3117 | repl_is_utf8 = DO_UTF8(repl_sv_copy) && SvCUR(sv); |
3118 | } | |
502d9230 VP |
3119 | if (!SvOK(sv)) |
3120 | sv_setpvs(sv, ""); | |
777f7c56 | 3121 | sv_insert_flags(sv, byte_pos, byte_len, repl, repl_len, 0); |
9402d6ed | 3122 | if (repl_is_utf8) |
f7928d6c | 3123 | SvUTF8_on(sv); |
ef8d46e8 | 3124 | SvREFCNT_dec(repl_sv_copy); |
f7928d6c | 3125 | } |
79072805 | 3126 | } |
849ca7ee | 3127 | SPAGAIN; |
e27c778f FC |
3128 | SvSETMAGIC(TARG); |
3129 | PUSHs(TARG); | |
79072805 | 3130 | RETURN; |
777f7c56 | 3131 | |
1c900557 | 3132 | bound_fail: |
777f7c56 EB |
3133 | if (lvalue || repl) |
3134 | Perl_croak(aTHX_ "substr outside of string"); | |
3135 | Perl_ck_warner(aTHX_ packWARN(WARN_SUBSTR), "substr outside of string"); | |
3136 | RETPUSHUNDEF; | |
79072805 LW |
3137 | } |
3138 | ||
3139 | PP(pp_vec) | |
3140 | { | |
2154eca7 | 3141 | dVAR; dSP; |
1b6737cc AL |
3142 | register const IV size = POPi; |
3143 | register const IV offset = POPi; | |
3144 | register SV * const src = POPs; | |
3145 | const I32 lvalue = PL_op->op_flags & OPf_MOD || LVRET; | |
2154eca7 | 3146 | SV * ret; |
a0d0e21e | 3147 | |
81e118e0 | 3148 | if (lvalue) { /* it's an lvalue! */ |
2154eca7 EB |
3149 | ret = sv_2mortal(newSV_type(SVt_PVLV)); /* Not TARG RT#67838 */ |
3150 | sv_magic(ret, NULL, PERL_MAGIC_vec, NULL, 0); | |
3151 | LvTYPE(ret) = 'v'; | |
3152 | LvTARG(ret) = SvREFCNT_inc_simple(src); | |
3153 | LvTARGOFF(ret) = offset; | |
3154 | LvTARGLEN(ret) = size; | |
3155 | } | |
3156 | else { | |
3157 | dTARGET; | |
3158 | SvTAINTED_off(TARG); /* decontaminate */ | |
3159 | ret = TARG; | |
79072805 LW |
3160 | } |
3161 | ||
2154eca7 EB |
3162 | sv_setuv(ret, do_vecget(src, offset, size)); |
3163 | PUSHs(ret); | |
79072805 LW |
3164 | RETURN; |
3165 | } | |
3166 | ||
3167 | PP(pp_index) | |
3168 | { | |
97aff369 | 3169 | dVAR; dSP; dTARGET; |
79072805 LW |
3170 | SV *big; |
3171 | SV *little; | |
c445ea15 | 3172 | SV *temp = NULL; |
ad66a58c | 3173 | STRLEN biglen; |
2723d216 | 3174 | STRLEN llen = 0; |
79072805 LW |
3175 | I32 offset; |
3176 | I32 retval; | |
73ee8be2 NC |
3177 | const char *big_p; |
3178 | const char *little_p; | |
2f040f7f NC |
3179 | bool big_utf8; |
3180 | bool little_utf8; | |
2723d216 | 3181 | const bool is_index = PL_op->op_type == OP_INDEX; |
d3e26383 | 3182 | const bool threeargs = MAXARG >= 3 && (TOPs || ((void)POPs,0)); |
79072805 | 3183 | |
e1dccc0d Z |
3184 | if (threeargs) |
3185 | offset = POPi; | |
79072805 LW |
3186 | little = POPs; |
3187 | big = POPs; | |
73ee8be2 NC |
3188 | big_p = SvPV_const(big, biglen); |
3189 | little_p = SvPV_const(little, llen); | |
3190 | ||
e609e586 NC |
3191 | big_utf8 = DO_UTF8(big); |
3192 | little_utf8 = DO_UTF8(little); | |
3193 | if (big_utf8 ^ little_utf8) { | |
3194 | /* One needs to be upgraded. */ | |
2f040f7f NC |
3195 | if (little_utf8 && !PL_encoding) { |
3196 | /* Well, maybe instead we might be able to downgrade the small | |
3197 | string? */ | |
1eced8f8 | 3198 | char * const pv = (char*)bytes_from_utf8((U8 *)little_p, &llen, |
2f040f7f NC |
3199 | &little_utf8); |
3200 | if (little_utf8) { | |
3201 | /* If the large string is ISO-8859-1, and it's not possible to | |
3202 | convert the small string to ISO-8859-1, then there is no | |
3203 | way that it could be found anywhere by index. */ | |
3204 | retval = -1; | |
3205 | goto fail; | |
3206 | } | |
e609e586 | 3207 | |
2f040f7f NC |
3208 | /* At this point, pv is a malloc()ed string. So donate it to temp |
3209 | to ensure it will get free()d */ | |
3210 | little = temp = newSV(0); | |
73ee8be2 NC |
3211 | sv_usepvn(temp, pv, llen); |
3212 | little_p = SvPVX(little); | |
e609e586 | 3213 | } else { |
73ee8be2 NC |
3214 | temp = little_utf8 |
3215 | ? newSVpvn(big_p, biglen) : newSVpvn(little_p, llen); | |
2f040f7f NC |
3216 | |
3217 | if (PL_encoding) { | |
3218 | sv_recode_to_utf8(temp, PL_encoding); | |
3219 | } else { | |
3220 | sv_utf8_upgrade(temp); | |
3221 | } | |
3222 | if (little_utf8) { | |
3223 | big = temp; | |
3224 | big_utf8 = TRUE; | |
73ee8be2 | 3225 | big_p = SvPV_const(big, biglen); |
2f040f7f NC |
3226 | } else { |
3227 | little = temp; | |
73ee8be2 | 3228 | little_p = SvPV_const(little, llen); |
2f040f7f | 3229 | } |
e609e586 NC |
3230 | } |
3231 | } | |
73ee8be2 NC |
3232 | if (SvGAMAGIC(big)) { |
3233 | /* Life just becomes a lot easier if I use a temporary here. | |
3234 | Otherwise I need to avoid calls to sv_pos_u2b(), which (dangerously) | |
3235 | will trigger magic and overloading again, as will fbm_instr() | |
3236 | */ | |
59cd0e26 NC |
3237 | big = newSVpvn_flags(big_p, biglen, |
3238 | SVs_TEMP | (big_utf8 ? SVf_UTF8 : 0)); | |
73ee8be2 NC |
3239 | big_p = SvPVX(big); |
3240 | } | |
e4e44778 | 3241 | if (SvGAMAGIC(little) || (is_index && !SvOK(little))) { |
73ee8be2 NC |
3242 | /* index && SvOK() is a hack. fbm_instr() calls SvPV_const, which will |
3243 | warn on undef, and we've already triggered a warning with the | |
3244 | SvPV_const some lines above. We can't remove that, as we need to | |
3245 | call some SvPV to trigger overloading early and find out if the | |
3246 | string is UTF-8. | |
3247 | This is all getting to messy. The API isn't quite clean enough, | |
3248 | because data access has side effects. | |
3249 | */ | |
59cd0e26 NC |
3250 | little = newSVpvn_flags(little_p, llen, |
3251 | SVs_TEMP | (little_utf8 ? SVf_UTF8 : 0)); | |
73ee8be2 NC |
3252 | little_p = SvPVX(little); |
3253 | } | |
e609e586 | 3254 | |
d3e26383 | 3255 | if (!threeargs) |
2723d216 | 3256 | offset = is_index ? 0 : biglen; |
a0ed51b3 | 3257 | else { |
ad66a58c | 3258 | if (big_utf8 && offset > 0) |
a0ed51b3 | 3259 | sv_pos_u2b(big, &offset, 0); |
73ee8be2 NC |
3260 | if (!is_index) |
3261 | offset += llen; | |
a0ed51b3 | 3262 | } |
79072805 LW |
3263 | if (offset < 0) |
3264 | offset = 0; | |
ad66a58c NC |
3265 | else if (offset > (I32)biglen) |
3266 | offset = biglen; | |
73ee8be2 NC |
3267 | if (!(little_p = is_index |
3268 | ? fbm_instr((unsigned char*)big_p + offset, | |
3269 | (unsigned char*)big_p + biglen, little, 0) | |
3270 | : rninstr(big_p, big_p + offset, | |
3271 | little_p, little_p + llen))) | |
a0ed51b3 | 3272 | retval = -1; |
ad66a58c | 3273 | else { |
73ee8be2 | 3274 | retval = little_p - big_p; |
ad66a58c NC |
3275 | if (retval > 0 && big_utf8) |
3276 | sv_pos_b2u(big, &retval); | |
3277 | } | |
ef8d46e8 | 3278 | SvREFCNT_dec(temp); |
2723d216 | 3279 | fail: |
e1dccc0d | 3280 | PUSHi(retval); |
79072805 LW |
3281 | RETURN; |
3282 | } | |
3283 | ||
3284 | PP(pp_sprintf) | |
3285 | { | |
97aff369 | 3286 | dVAR; dSP; dMARK; dORIGMARK; dTARGET; |
3e6bd4bf | 3287 | SvTAINTED_off(TARG); |
79072805 | 3288 | do_sprintf(TARG, SP-MARK, MARK+1); |
bbce6d69 | 3289 | TAINT_IF(SvTAINTED(TARG)); |
79072805 LW |
3290 | SP = ORIGMARK; |
3291 | PUSHTARG; | |
3292 | RETURN; | |
3293 | } | |
3294 | ||
79072805 LW |
3295 | PP(pp_ord) |
3296 | { | |
97aff369 | 3297 | dVAR; dSP; dTARGET; |
1eced8f8 | 3298 | |
7df053ec | 3299 | SV *argsv = POPs; |
ba210ebe | 3300 | STRLEN len; |
349d4f2f | 3301 | const U8 *s = (U8*)SvPV_const(argsv, len); |
121910a4 | 3302 | |
799ef3cb | 3303 | if (PL_encoding && SvPOK(argsv) && !DO_UTF8(argsv)) { |
1eced8f8 | 3304 | SV * const tmpsv = sv_2mortal(newSVsv(argsv)); |
799ef3cb | 3305 | s = (U8*)sv_recode_to_utf8(tmpsv, PL_encoding); |
121910a4 JH |
3306 | argsv = tmpsv; |
3307 | } | |
79072805 | 3308 | |
872c91ae | 3309 | XPUSHu(DO_UTF8(argsv) ? |
89ebb4a3 | 3310 | utf8n_to_uvchr(s, UTF8_MAXBYTES, 0, UTF8_ALLOW_ANYUV) : |
5fc32dea | 3311 | (UV)(*s & 0xff)); |
68795e93 | 3312 | |
79072805 LW |
3313 | RETURN; |
3314 | } | |
3315 | ||
463ee0b2 LW |
3316 | PP(pp_chr) |
3317 | { | |
97aff369 | 3318 | dVAR; dSP; dTARGET; |
463ee0b2 | 3319 | char *tmps; |
8a064bd6 JH |
3320 | UV value; |
3321 | ||
3322 | if (((SvIOK_notUV(TOPs) && SvIV(TOPs) < 0) | |
3323 | || | |
3324 | (SvNOK(TOPs) && SvNV(TOPs) < 0.0))) { | |
3325 | if (IN_BYTES) { | |
3326 | value = POPu; /* chr(-1) eq chr(0xff), etc. */ | |
3327 | } else { | |
3328 | (void) POPs; /* Ignore the argument value. */ | |
3329 | value = UNICODE_REPLACEMENT; | |
3330 | } | |
3331 | } else { | |
3332 | value = POPu; | |
3333 | } | |
463ee0b2 | 3334 | |
862a34c6 | 3335 | SvUPGRADE(TARG,SVt_PV); |
a0ed51b3 | 3336 | |
0064a8a9 | 3337 | if (value > 255 && !IN_BYTES) { |
eb160463 | 3338 | SvGROW(TARG, (STRLEN)UNISKIP(value)+1); |
62961d2e | 3339 | tmps = (char*)uvchr_to_utf8_flags((U8*)SvPVX(TARG), value, 0); |
349d4f2f | 3340 | SvCUR_set(TARG, tmps - SvPVX_const(TARG)); |
a0ed51b3 LW |
3341 | *tmps = '\0'; |
3342 | (void)SvPOK_only(TARG); | |
aa6ffa16 | 3343 | SvUTF8_on(TARG); |
a0ed51b3 LW |
3344 | XPUSHs(TARG); |
3345 | RETURN; | |
3346 | } | |
3347 | ||
748a9306 | 3348 | SvGROW(TARG,2); |
463ee0b2 LW |
3349 | SvCUR_set(TARG, 1); |
3350 | tmps = SvPVX(TARG); | |
eb160463 | 3351 | *tmps++ = (char)value; |
748a9306 | 3352 | *tmps = '\0'; |
a0d0e21e | 3353 | (void)SvPOK_only(TARG); |
4c5ed6e2 | 3354 | |
88632417 | 3355 | if (PL_encoding && !IN_BYTES) { |
799ef3cb | 3356 | sv_recode_to_utf8(TARG, PL_encoding); |
88632417 JH |
3357 | tmps = SvPVX(TARG); |
3358 | if (SvCUR(TARG) == 0 || !is_utf8_string((U8*)tmps, SvCUR(TARG)) || | |
4c5ed6e2 TS |
3359 | UNICODE_IS_REPLACEMENT(utf8_to_uvchr((U8*)tmps, NULL))) { |
3360 | SvGROW(TARG, 2); | |
d5a15ac2 | 3361 | tmps = SvPVX(TARG); |
4c5ed6e2 TS |
3362 | SvCUR_set(TARG, 1); |
3363 | *tmps++ = (char)value; | |
88632417 | 3364 | *tmps = '\0'; |
4c5ed6e2 | 3365 | SvUTF8_off(TARG); |
88632417 JH |
3366 | } |
3367 | } | |
4c5ed6e2 | 3368 | |
463ee0b2 LW |
3369 | XPUSHs(TARG); |
3370 | RETURN; | |
3371 | } | |
3372 | ||
79072805 LW |
3373 | PP(pp_crypt) |
3374 | { | |
79072805 | 3375 | #ifdef HAS_CRYPT |
97aff369 | 3376 | dVAR; dSP; dTARGET; |
5f74f29c | 3377 | dPOPTOPssrl; |
85c16d83 | 3378 | STRLEN len; |
10516c54 | 3379 | const char *tmps = SvPV_const(left, len); |
2bc69dc4 | 3380 | |
85c16d83 | 3381 | if (DO_UTF8(left)) { |
2bc69dc4 | 3382 | /* If Unicode, try to downgrade. |
f2791508 JH |
3383 | * If not possible, croak. |
3384 | * Yes, we made this up. */ | |
1b6737cc | 3385 | SV* const tsv = sv_2mortal(newSVsv(left)); |
2bc69dc4 | 3386 | |
f2791508 | 3387 | SvUTF8_on(tsv); |
2bc69dc4 | 3388 | sv_utf8_downgrade(tsv, FALSE); |
349d4f2f | 3389 | tmps = SvPV_const(tsv, len); |
85c16d83 | 3390 | } |
05404ffe JH |
3391 | # ifdef USE_ITHREADS |
3392 | # ifdef HAS_CRYPT_R | |
3393 | if (!PL_reentrant_buffer->_crypt_struct_buffer) { | |
3394 | /* This should be threadsafe because in ithreads there is only | |
3395 | * one thread per interpreter. If this would not be true, | |
3396 | * we would need a mutex to protect this malloc. */ | |
3397 | PL_reentrant_buffer->_crypt_struct_buffer = | |
3398 | (struct crypt_data *)safemalloc(sizeof(struct crypt_data)); | |
3399 | #if defined(__GLIBC__) || defined(__EMX__) | |
3400 | if (PL_reentrant_buffer->_crypt_struct_buffer) { | |
3401 | PL_reentrant_buffer->_crypt_struct_buffer->initialized = 0; | |
3402 | /* work around glibc-2.2.5 bug */ | |
3403 | PL_reentrant_buffer->_crypt_struct_buffer->current_saltbits = 0; | |
3404 | } | |
05404ffe | 3405 | #endif |
6ab58e4d | 3406 | } |
05404ffe JH |
3407 | # endif /* HAS_CRYPT_R */ |
3408 | # endif /* USE_ITHREADS */ | |
5f74f29c | 3409 | # ifdef FCRYPT |
83003860 | 3410 | sv_setpv(TARG, fcrypt(tmps, SvPV_nolen_const(right))); |
5f74f29c | 3411 | # else |
83003860 | 3412 | sv_setpv(TARG, PerlProc_crypt(tmps, SvPV_nolen_const(right))); |
5f74f29c | 3413 | # endif |
ec93b65f | 3414 | SETTARG; |
4808266b | 3415 | RETURN; |
79072805 | 3416 | #else |
b13b2135 | 3417 | DIE(aTHX_ |
79072805 LW |
3418 | "The crypt() function is unimplemented due to excessive paranoia."); |
3419 | #endif | |
79072805 LW |
3420 | } |
3421 | ||
00f254e2 KW |
3422 | /* Generally UTF-8 and UTF-EBCDIC are indistinguishable at this level. So |
3423 | * most comments below say UTF-8, when in fact they mean UTF-EBCDIC as well */ | |
3424 | ||
00f254e2 KW |
3425 | /* Below are several macros that generate code */ |
3426 | /* Generates code to store a unicode codepoint c that is known to occupy | |
3427 | * exactly two UTF-8 and UTF-EBCDIC bytes; it is stored into p and p+1. */ | |
3428 | #define STORE_UNI_TO_UTF8_TWO_BYTE(p, c) \ | |
3429 | STMT_START { \ | |
3430 | *(p) = UTF8_TWO_BYTE_HI(c); \ | |
3431 | *((p)+1) = UTF8_TWO_BYTE_LO(c); \ | |
3432 | } STMT_END | |
3433 | ||
3434 | /* Like STORE_UNI_TO_UTF8_TWO_BYTE, but advances p to point to the next | |
3435 | * available byte after the two bytes */ | |
3436 | #define CAT_UNI_TO_UTF8_TWO_BYTE(p, c) \ | |
3437 | STMT_START { \ | |
3438 | *(p)++ = UTF8_TWO_BYTE_HI(c); \ | |
3439 | *((p)++) = UTF8_TWO_BYTE_LO(c); \ | |
3440 | } STMT_END | |
3441 | ||
3442 | /* Generates code to store the upper case of latin1 character l which is known | |
3443 | * to have its upper case be non-latin1 into the two bytes p and p+1. There | |
3444 | * are only two characters that fit this description, and this macro knows | |
3445 | * about them, and that the upper case values fit into two UTF-8 or UTF-EBCDIC | |
3446 | * bytes */ | |
3447 | #define STORE_NON_LATIN1_UC(p, l) \ | |
3448 | STMT_START { \ | |
3449 | if ((l) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) { \ | |
3450 | STORE_UNI_TO_UTF8_TWO_BYTE((p), LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS); \ | |
3451 | } else { /* Must be the following letter */ \ | |
3452 | STORE_UNI_TO_UTF8_TWO_BYTE((p), GREEK_CAPITAL_LETTER_MU); \ | |
3453 | } \ | |
3454 | } STMT_END | |
3455 | ||
3456 | /* Like STORE_NON_LATIN1_UC, but advances p to point to the next available byte | |
3457 | * after the character stored */ | |
3458 | #define CAT_NON_LATIN1_UC(p, l) \ | |
3459 | STMT_START { \ | |
3460 | if ((l) == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) { \ | |
3461 | CAT_UNI_TO_UTF8_TWO_BYTE((p), LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS); \ | |
3462 | } else { \ | |
3463 | CAT_UNI_TO_UTF8_TWO_BYTE((p), GREEK_CAPITAL_LETTER_MU); \ | |
3464 | } \ | |
3465 | } STMT_END | |
3466 | ||
3467 | /* Generates code to add the two UTF-8 bytes (probably u) that are the upper | |
3468 | * case of l into p and p+1. u must be the result of toUPPER_LATIN1_MOD(l), | |
3469 | * and must require two bytes to store it. Advances p to point to the next | |
3470 | * available position */ | |
3471 | #define CAT_TWO_BYTE_UNI_UPPER_MOD(p, l, u) \ | |
3472 | STMT_START { \ | |
3473 | if ((u) != LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) { \ | |
3474 | CAT_UNI_TO_UTF8_TWO_BYTE((p), (u)); /* not special, just save it */ \ | |
3475 | } else if (l == LATIN_SMALL_LETTER_SHARP_S) { \ | |
3476 | *(p)++ = 'S'; *(p)++ = 'S'; /* upper case is 'SS' */ \ | |
3477 | } else {/* else is one of the other two special cases */ \ | |
3478 | CAT_NON_LATIN1_UC((p), (l)); \ | |
3479 | } \ | |
3480 | } STMT_END | |
3481 | ||
79072805 LW |
3482 | PP(pp_ucfirst) |
3483 | { | |
00f254e2 KW |
3484 | /* Actually is both lcfirst() and ucfirst(). Only the first character |
3485 | * changes. This means that possibly we can change in-place, ie., just | |
3486 | * take the source and change that one character and store it back, but not | |
3487 | * if read-only etc, or if the length changes */ | |
3488 | ||
97aff369 | 3489 | dVAR; |
39644a26 | 3490 | dSP; |
d54190f6 | 3491 | SV *source = TOPs; |
00f254e2 | 3492 | STRLEN slen; /* slen is the byte length of the whole SV. */ |
d54190f6 NC |
3493 | STRLEN need; |
3494 | SV *dest; | |
00f254e2 KW |
3495 | bool inplace; /* ? Convert first char only, in-place */ |
3496 | bool doing_utf8 = FALSE; /* ? using utf8 */ | |
3497 | bool convert_source_to_utf8 = FALSE; /* ? need to convert */ | |
12e9c124 | 3498 | const int op_type = PL_op->op_type; |
d54190f6 NC |
3499 | const U8 *s; |
3500 | U8 *d; | |
3501 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; | |
00f254e2 KW |
3502 | STRLEN ulen; /* ulen is the byte length of the original Unicode character |
3503 | * stored as UTF-8 at s. */ | |
3504 | STRLEN tculen; /* tculen is the byte length of the freshly titlecased (or | |
3505 | * lowercased) character stored in tmpbuf. May be either | |
3506 | * UTF-8 or not, but in either case is the number of bytes */ | |
d54190f6 NC |
3507 | |
3508 | SvGETMAGIC(source); | |
3509 | if (SvOK(source)) { | |
3510 | s = (const U8*)SvPV_nomg_const(source, slen); | |
3511 | } else { | |
0a0ffbce RGS |
3512 | if (ckWARN(WARN_UNINITIALIZED)) |
3513 | report_uninit(source); | |
1eced8f8 | 3514 | s = (const U8*)""; |
d54190f6 NC |
3515 | slen = 0; |
3516 | } | |
a0ed51b3 | 3517 | |
00f254e2 KW |
3518 | /* We may be able to get away with changing only the first character, in |
3519 | * place, but not if read-only, etc. Later we may discover more reasons to | |
3520 | * not convert in-place. */ | |
3521 | inplace = SvPADTMP(source) && !SvREADONLY(source) && SvTEMP(source); | |
3522 | ||
3523 | /* First calculate what the changed first character should be. This affects | |
3524 | * whether we can just swap it out, leaving the rest of the string unchanged, | |
3525 | * or even if have to convert the dest to UTF-8 when the source isn't */ | |
3526 | ||
3527 | if (! slen) { /* If empty */ | |
3528 | need = 1; /* still need a trailing NUL */ | |
3529 | } | |
3530 | else if (DO_UTF8(source)) { /* Is the source utf8? */ | |
d54190f6 | 3531 | doing_utf8 = TRUE; |
00f254e2 | 3532 | |
00f254e2 KW |
3533 | if (UTF8_IS_INVARIANT(*s)) { |
3534 | ||
3535 | /* An invariant source character is either ASCII or, in EBCDIC, an | |
3536 | * ASCII equivalent or a caseless C1 control. In both these cases, | |
3537 | * the lower and upper cases of any character are also invariants | |
3538 | * (and title case is the same as upper case). So it is safe to | |
3539 | * use the simple case change macros which avoid the overhead of | |
3540 | * the general functions. Note that if perl were to be extended to | |
3541 | * do locale handling in UTF-8 strings, this wouldn't be true in, | |
3542 | * for example, Lithuanian or Turkic. */ | |
3543 | *tmpbuf = (op_type == OP_LCFIRST) ? toLOWER(*s) : toUPPER(*s); | |
3544 | tculen = ulen = 1; | |
3545 | need = slen + 1; | |
12e9c124 | 3546 | } |
00f254e2 KW |
3547 | else if (UTF8_IS_DOWNGRADEABLE_START(*s)) { |
3548 | U8 chr; | |
3549 | ||
3550 | /* Similarly, if the source character isn't invariant but is in the | |
3551 | * latin1 range (or EBCDIC equivalent thereof), we have the case | |
3552 | * changes compiled into perl, and can avoid the overhead of the | |
3553 | * general functions. In this range, the characters are stored as | |
3554 | * two UTF-8 bytes, and it so happens that any changed-case version | |
3555 | * is also two bytes (in both ASCIIish and EBCDIC machines). */ | |
3556 | tculen = ulen = 2; | |
3557 | need = slen + 1; | |
3558 | ||
3559 | /* Convert the two source bytes to a single Unicode code point | |
3560 | * value, change case and save for below */ | |
356979f4 | 3561 | chr = TWO_BYTE_UTF8_TO_UNI(*s, *(s+1)); |
00f254e2 KW |
3562 | if (op_type == OP_LCFIRST) { /* lower casing is easy */ |
3563 | U8 lower = toLOWER_LATIN1(chr); | |
3564 | STORE_UNI_TO_UTF8_TWO_BYTE(tmpbuf, lower); | |
3565 | } | |
3566 | else { /* ucfirst */ | |
3567 | U8 upper = toUPPER_LATIN1_MOD(chr); | |
3568 | ||
3569 | /* Most of the latin1 range characters are well-behaved. Their | |
3570 | * title and upper cases are the same, and are also in the | |
3571 | * latin1 range. The macro above returns their upper (hence | |
3572 | * title) case, and all that need be done is to save the result | |
3573 | * for below. However, several characters are problematic, and | |
3574 | * have to be handled specially. The MOD in the macro name | |
3575 | * above means that these tricky characters all get mapped to | |
3576 | * the single character LATIN_SMALL_LETTER_Y_WITH_DIAERESIS. | |
3577 | * This mapping saves some tests for the majority of the | |
3578 | * characters */ | |
3579 | ||
3580 | if (upper != LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) { | |
3581 | ||
3582 | /* Not tricky. Just save it. */ | |
3583 | STORE_UNI_TO_UTF8_TWO_BYTE(tmpbuf, upper); | |
3584 | } | |
3585 | else if (chr == LATIN_SMALL_LETTER_SHARP_S) { | |
3586 | ||
3587 | /* This one is tricky because it is two characters long, | |
3588 | * though the UTF-8 is still two bytes, so the stored | |
3589 | * length doesn't change */ | |
3590 | *tmpbuf = 'S'; /* The UTF-8 is 'Ss' */ | |
3591 | *(tmpbuf + 1) = 's'; | |
3592 | } | |
3593 | else { | |
3594 | ||
3595 | /* The other two have their title and upper cases the same, | |
3596 | * but are tricky because the changed-case characters | |
3597 | * aren't in the latin1 range. They, however, do fit into | |
3598 | * two UTF-8 bytes */ | |
3599 | STORE_NON_LATIN1_UC(tmpbuf, chr); | |
3600 | } | |
3601 | } | |
3602 | } | |
3603 | else { | |
00f254e2 KW |
3604 | |
3605 | /* Here, can't short-cut the general case */ | |
3606 | ||
3607 | utf8_to_uvchr(s, &ulen); | |
3608 | if (op_type == OP_UCFIRST) toTITLE_utf8(s, tmpbuf, &tculen); | |
3609 | else toLOWER_utf8(s, tmpbuf, &tculen); | |
3610 | ||
3611 | /* we can't do in-place if the length changes. */ | |
3612 | if (ulen != tculen) inplace = FALSE; | |
3613 | need = slen + 1 - ulen + tculen; | |
00f254e2 | 3614 | } |
d54190f6 | 3615 | } |
00f254e2 KW |
3616 | else { /* Non-zero length, non-UTF-8, Need to consider locale and if |
3617 | * latin1 is treated as caseless. Note that a locale takes | |
3618 | * precedence */ | |
3619 | tculen = 1; /* Most characters will require one byte, but this will | |
3620 | * need to be overridden for the tricky ones */ | |
3621 | need = slen + 1; | |
3622 | ||
3623 | if (op_type == OP_LCFIRST) { | |
d54190f6 | 3624 | |
00f254e2 KW |
3625 | /* lower case the first letter: no trickiness for any character */ |
3626 | *tmpbuf = (IN_LOCALE_RUNTIME) ? toLOWER_LC(*s) : | |
3627 | ((IN_UNI_8_BIT) ? toLOWER_LATIN1(*s) : toLOWER(*s)); | |
3628 | } | |
3629 | /* is ucfirst() */ | |
3630 | else if (IN_LOCALE_RUNTIME) { | |
3631 | *tmpbuf = toUPPER_LC(*s); /* This would be a bug if any locales | |
3632 | * have upper and title case different | |
3633 | */ | |
3634 | } | |
3635 | else if (! IN_UNI_8_BIT) { | |
3636 | *tmpbuf = toUPPER(*s); /* Returns caseless for non-ascii, or | |
3637 | * on EBCDIC machines whatever the | |
3638 | * native function does */ | |
3639 | } | |
3640 | else { /* is ucfirst non-UTF-8, not in locale, and cased latin1 */ | |
3641 | *tmpbuf = toUPPER_LATIN1_MOD(*s); | |
3642 | ||
3643 | /* tmpbuf now has the correct title case for all latin1 characters | |
3644 | * except for the several ones that have tricky handling. All | |
3645 | * of these are mapped by the MOD to the letter below. */ | |
3646 | if (*tmpbuf == LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) { | |
3647 | ||
3648 | /* The length is going to change, with all three of these, so | |
3649 | * can't replace just the first character */ | |
3650 | inplace = FALSE; | |
3651 | ||
3652 | /* We use the original to distinguish between these tricky | |
3653 | * cases */ | |
3654 | if (*s == LATIN_SMALL_LETTER_SHARP_S) { | |
3655 | /* Two character title case 'Ss', but can remain non-UTF-8 */ | |
3656 | need = slen + 2; | |
3657 | *tmpbuf = 'S'; | |
3658 | *(tmpbuf + 1) = 's'; /* Assert: length(tmpbuf) >= 2 */ | |
3659 | tculen = 2; | |
3660 | } | |
3661 | else { | |
d54190f6 | 3662 | |
00f254e2 KW |
3663 | /* The other two tricky ones have their title case outside |
3664 | * latin1. It is the same as their upper case. */ | |
3665 | doing_utf8 = TRUE; | |
3666 | STORE_NON_LATIN1_UC(tmpbuf, *s); | |
3667 | ||
3668 | /* The UTF-8 and UTF-EBCDIC lengths of both these characters | |
3669 | * and their upper cases is 2. */ | |
3670 | tculen = ulen = 2; | |
3671 | ||
3672 | /* The entire result will have to be in UTF-8. Assume worst | |
3673 | * case sizing in conversion. (all latin1 characters occupy | |
3674 | * at most two bytes in utf8) */ | |
3675 | convert_source_to_utf8 = TRUE; | |
3676 | need = slen * 2 + 1; | |
3677 | } | |
3678 | } /* End of is one of the three special chars */ | |
3679 | } /* End of use Unicode (Latin1) semantics */ | |
3680 | } /* End of changing the case of the first character */ | |
3681 | ||
3682 | /* Here, have the first character's changed case stored in tmpbuf. Ready to | |
3683 | * generate the result */ | |
3684 | if (inplace) { | |
3685 | ||
3686 | /* We can convert in place. This means we change just the first | |
3687 | * character without disturbing the rest; no need to grow */ | |
d54190f6 NC |
3688 | dest = source; |
3689 | s = d = (U8*)SvPV_force_nomg(source, slen); | |
3690 | } else { | |
3691 | dTARGET; | |
3692 | ||
3693 | dest = TARG; | |
3694 | ||
00f254e2 KW |
3695 | /* Here, we can't convert in place; we earlier calculated how much |
3696 | * space we will need, so grow to accommodate that */ | |
d54190f6 | 3697 | SvUPGRADE(dest, SVt_PV); |
3b416f41 | 3698 | d = (U8*)SvGROW(dest, need); |
d54190f6 NC |
3699 | (void)SvPOK_only(dest); |
3700 | ||
3701 | SETs(dest); | |
d54190f6 | 3702 | } |
44bc797b | 3703 | |
d54190f6 | 3704 | if (doing_utf8) { |
00f254e2 KW |
3705 | if (! inplace) { |
3706 | if (! convert_source_to_utf8) { | |
3707 | ||
3708 | /* Here both source and dest are in UTF-8, but have to create | |
3709 | * the entire output. We initialize the result to be the | |
3710 | * title/lower cased first character, and then append the rest | |
3711 | * of the string. */ | |
3712 | sv_setpvn(dest, (char*)tmpbuf, tculen); | |
3713 | if (slen > ulen) { | |
3714 | sv_catpvn(dest, (char*)(s + ulen), slen - ulen); | |
3715 | } | |
3716 | } | |
3717 | else { | |
3718 | const U8 *const send = s + slen; | |
3719 | ||
3720 | /* Here the dest needs to be in UTF-8, but the source isn't, | |
3721 | * except we earlier UTF-8'd the first character of the source | |
3722 | * into tmpbuf. First put that into dest, and then append the | |
3723 | * rest of the source, converting it to UTF-8 as we go. */ | |
3724 | ||
3725 | /* Assert tculen is 2 here because the only two characters that | |
3726 | * get to this part of the code have 2-byte UTF-8 equivalents */ | |
3727 | *d++ = *tmpbuf; | |
3728 | *d++ = *(tmpbuf + 1); | |
3729 | s++; /* We have just processed the 1st char */ | |
3730 | ||
3731 | for (; s < send; s++) { | |
3732 | d = uvchr_to_utf8(d, *s); | |
3733 | } | |
3734 | *d = '\0'; | |
3735 | SvCUR_set(dest, d - (U8*)SvPVX_const(dest)); | |
3736 | } | |
d54190f6 | 3737 | SvUTF8_on(dest); |
a0ed51b3 | 3738 | } |
00f254e2 | 3739 | else { /* in-place UTF-8. Just overwrite the first character */ |
d54190f6 NC |
3740 | Copy(tmpbuf, d, tculen, U8); |
3741 | SvCUR_set(dest, need - 1); | |
a0ed51b3 | 3742 | } |
a0ed51b3 | 3743 | } |
00f254e2 KW |
3744 | else { /* Neither source nor dest are in or need to be UTF-8 */ |
3745 | if (slen) { | |
2de3dbcc | 3746 | if (IN_LOCALE_RUNTIME) { |
31351b04 | 3747 | TAINT; |
d54190f6 | 3748 | SvTAINTED_on(dest); |
31351b04 | 3749 | } |
00f254e2 KW |
3750 | if (inplace) { /* in-place, only need to change the 1st char */ |
3751 | *d = *tmpbuf; | |
3752 | } | |
3753 | else { /* Not in-place */ | |
3754 | ||
3755 | /* Copy the case-changed character(s) from tmpbuf */ | |
3756 | Copy(tmpbuf, d, tculen, U8); | |
3757 | d += tculen - 1; /* Code below expects d to point to final | |
3758 | * character stored */ | |
3759 | } | |
3760 | } | |
3761 | else { /* empty source */ | |
3762 | /* See bug #39028: Don't taint if empty */ | |
d54190f6 NC |
3763 | *d = *s; |
3764 | } | |
3765 | ||
00f254e2 KW |
3766 | /* In a "use bytes" we don't treat the source as UTF-8, but, still want |
3767 | * the destination to retain that flag */ | |
d54190f6 NC |
3768 | if (SvUTF8(source)) |
3769 | SvUTF8_on(dest); | |
3770 | ||
00f254e2 | 3771 | if (!inplace) { /* Finish the rest of the string, unchanged */ |
d54190f6 NC |
3772 | /* This will copy the trailing NUL */ |
3773 | Copy(s + 1, d + 1, slen, U8); | |
3774 | SvCUR_set(dest, need - 1); | |
bbce6d69 | 3775 | } |
bbce6d69 | 3776 | } |
539689e7 FC |
3777 | if (dest != source && SvTAINTED(source)) |
3778 | SvTAINT(dest); | |
d54190f6 | 3779 | SvSETMAGIC(dest); |
79072805 LW |
3780 | RETURN; |
3781 | } | |
3782 | ||
67306194 NC |
3783 | /* There's so much setup/teardown code common between uc and lc, I wonder if |
3784 | it would be worth merging the two, and just having a switch outside each | |
00f254e2 | 3785 | of the three tight loops. There is less and less commonality though */ |
79072805 LW |
3786 | PP(pp_uc) |
3787 | { | |
97aff369 | 3788 | dVAR; |
39644a26 | 3789 | dSP; |
67306194 | 3790 | SV *source = TOPs; |
463ee0b2 | 3791 | STRLEN len; |
67306194 NC |
3792 | STRLEN min; |
3793 | SV *dest; | |
3794 | const U8 *s; | |
3795 | U8 *d; | |
79072805 | 3796 | |
67306194 NC |
3797 | SvGETMAGIC(source); |
3798 | ||
3799 | if (SvPADTMP(source) && !SvREADONLY(source) && !SvAMAGIC(source) | |
00f254e2 KW |
3800 | && SvTEMP(source) && !DO_UTF8(source) |
3801 | && (IN_LOCALE_RUNTIME || ! IN_UNI_8_BIT)) { | |
3802 | ||
3803 | /* We can convert in place. The reason we can't if in UNI_8_BIT is to | |
3804 | * make the loop tight, so we overwrite the source with the dest before | |
3805 | * looking at it, and we need to look at the original source | |
3806 | * afterwards. There would also need to be code added to handle | |
3807 | * switching to not in-place in midstream if we run into characters | |
3808 | * that change the length. | |
3809 | */ | |
67306194 NC |
3810 | dest = source; |
3811 | s = d = (U8*)SvPV_force_nomg(source, len); | |
3812 | min = len + 1; | |
3813 | } else { | |
a0ed51b3 | 3814 | dTARGET; |
a0ed51b3 | 3815 | |
67306194 | 3816 | dest = TARG; |
128c9517 | 3817 | |
67306194 NC |
3818 | /* The old implementation would copy source into TARG at this point. |
3819 | This had the side effect that if source was undef, TARG was now | |
3820 | an undefined SV with PADTMP set, and they don't warn inside | |
3821 | sv_2pv_flags(). However, we're now getting the PV direct from | |
3822 | source, which doesn't have PADTMP set, so it would warn. Hence the | |
3823 | little games. */ | |
3824 | ||
3825 | if (SvOK(source)) { | |
3826 | s = (const U8*)SvPV_nomg_const(source, len); | |
3827 | } else { | |
0a0ffbce RGS |
3828 | if (ckWARN(WARN_UNINITIALIZED)) |
3829 | report_uninit(source); | |
1eced8f8 | 3830 | s = (const U8*)""; |
67306194 | 3831 | len = 0; |
a0ed51b3 | 3832 | } |
67306194 NC |
3833 | min = len + 1; |
3834 | ||
3835 | SvUPGRADE(dest, SVt_PV); | |
3b416f41 | 3836 | d = (U8*)SvGROW(dest, min); |
67306194 NC |
3837 | (void)SvPOK_only(dest); |
3838 | ||
3839 | SETs(dest); | |
a0ed51b3 | 3840 | } |
31351b04 | 3841 | |
67306194 NC |
3842 | /* Overloaded values may have toggled the UTF-8 flag on source, so we need |
3843 | to check DO_UTF8 again here. */ | |
3844 | ||
3845 | if (DO_UTF8(source)) { | |
3846 | const U8 *const send = s + len; | |
3847 | U8 tmpbuf[UTF8_MAXBYTES+1]; | |
3848 | ||
4c8a458a KW |
3849 | /* All occurrences of these are to be moved to follow any other marks. |
3850 | * This is context-dependent. We may not be passed enough context to | |
3851 | * move the iota subscript beyond all of them, but we do the best we can | |
3852 | * with what we're given. The result is always better than if we | |
3853 | * hadn't done this. And, the problem would only arise if we are | |
3854 | * passed a character without all its combining marks, which would be | |
3855 | * the caller's mistake. The information this is based on comes from a | |
3856 | * comment in Unicode SpecialCasing.txt, (and the Standard's text | |
3857 | * itself) and so can't be checked properly to see if it ever gets | |
3858 | * revised. But the likelihood of it changing is remote */ | |
00f254e2 | 3859 | bool in_iota_subscript = FALSE; |
00f254e2 | 3860 | |
67306194 | 3861 | while (s < send) { |
00f254e2 KW |
3862 | if (in_iota_subscript && ! is_utf8_mark(s)) { |
3863 | /* A non-mark. Time to output the iota subscript */ | |
3864 | #define GREEK_CAPITAL_LETTER_IOTA 0x0399 | |
3865 | #define COMBINING_GREEK_YPOGEGRAMMENI 0x0345 | |
3866 | ||
3867 | CAT_UNI_TO_UTF8_TWO_BYTE(d, GREEK_CAPITAL_LETTER_IOTA); | |
3868 | in_iota_subscript = FALSE; | |
3869 | } | |
00f254e2 | 3870 | |
00f254e2 KW |
3871 | /* If the UTF-8 character is invariant, then it is in the range |
3872 | * known by the standard macro; result is only one byte long */ | |
3873 | if (UTF8_IS_INVARIANT(*s)) { | |
3874 | *d++ = toUPPER(*s); | |
3875 | s++; | |
3876 | } | |
3877 | else if (UTF8_IS_DOWNGRADEABLE_START(*s)) { | |
3878 | ||
3879 | /* Likewise, if it fits in a byte, its case change is in our | |
3880 | * table */ | |
e1a8dbf5 | 3881 | U8 orig = TWO_BYTE_UTF8_TO_UNI(*s, *(s+1)); |
00f254e2 KW |
3882 | U8 upper = toUPPER_LATIN1_MOD(orig); |
3883 | CAT_TWO_BYTE_UNI_UPPER_MOD(d, orig, upper); | |
e1a8dbf5 | 3884 | s += 2; |
00f254e2 KW |
3885 | } |
3886 | else { | |
00f254e2 KW |
3887 | |
3888 | /* Otherwise, need the general UTF-8 case. Get the changed | |
3889 | * case value and copy it to the output buffer */ | |
3890 | ||
3891 | const STRLEN u = UTF8SKIP(s); | |
3892 | STRLEN ulen; | |
67306194 | 3893 | |
00f254e2 | 3894 | const UV uv = toUPPER_utf8(s, tmpbuf, &ulen); |
4c8a458a KW |
3895 | if (uv == GREEK_CAPITAL_LETTER_IOTA |
3896 | && utf8_to_uvchr(s, 0) == COMBINING_GREEK_YPOGEGRAMMENI) | |
3897 | { | |
00f254e2 KW |
3898 | in_iota_subscript = TRUE; |
3899 | } | |
3900 | else { | |
00f254e2 KW |
3901 | if (ulen > u && (SvLEN(dest) < (min += ulen - u))) { |
3902 | /* If the eventually required minimum size outgrows | |
3903 | * the available space, we need to grow. */ | |
3904 | const UV o = d - (U8*)SvPVX_const(dest); | |
3905 | ||
3906 | /* If someone uppercases one million U+03B0s we | |
3907 | * SvGROW() one million times. Or we could try | |
3908 | * guessing how much to allocate without allocating too | |
4c8a458a KW |
3909 | * much. Such is life. See corresponding comment in |
3910 | * lc code for another option */ | |
00f254e2 KW |
3911 | SvGROW(dest, min); |
3912 | d = (U8*)SvPVX(dest) + o; | |
3913 | } | |
3914 | Copy(tmpbuf, d, ulen, U8); | |
3915 | d += ulen; | |
00f254e2 | 3916 | } |
00f254e2 | 3917 | s += u; |
67306194 | 3918 | } |
67306194 | 3919 | } |
4c8a458a KW |
3920 | if (in_iota_subscript) { |
3921 | CAT_UNI_TO_UTF8_TWO_BYTE(d, GREEK_CAPITAL_LETTER_IOTA); | |
3922 | } | |
67306194 NC |
3923 | SvUTF8_on(dest); |
3924 | *d = '\0'; | |
3925 | SvCUR_set(dest, d - (U8*)SvPVX_const(dest)); | |
4c8a458a KW |
3926 | } |
3927 | else { /* Not UTF-8 */ | |
67306194 NC |
3928 | if (len) { |
3929 | const U8 *const send = s + len; | |
00f254e2 KW |
3930 | |
3931 | /* Use locale casing if in locale; regular style if not treating | |
3932 | * latin1 as having case; otherwise the latin1 casing. Do the | |
3933 | * whole thing in a tight loop, for speed, */ | |
2de3dbcc | 3934 | if (IN_LOCALE_RUNTIME) { |
31351b04 | 3935 | TAINT; |
67306194 NC |
3936 | SvTAINTED_on(dest); |
3937 | for (; s < send; d++, s++) | |
3938 | *d = toUPPER_LC(*s); | |
31351b04 | 3939 | } |
00f254e2 KW |
3940 | else if (! IN_UNI_8_BIT) { |
3941 | for (; s < send; d++, s++) { | |
67306194 | 3942 | *d = toUPPER(*s); |
00f254e2 | 3943 | } |
31351b04 | 3944 | } |
00f254e2 KW |
3945 | else { |
3946 | for (; s < send; d++, s++) { | |
3947 | *d = toUPPER_LATIN1_MOD(*s); | |
3948 | if (*d != LATIN_SMALL_LETTER_Y_WITH_DIAERESIS) continue; | |
3949 | ||
3950 | /* The mainstream case is the tight loop above. To avoid | |
3951 | * extra tests in that, all three characters that require | |
3952 | * special handling are mapped by the MOD to the one tested | |
3953 | * just above. | |
3954 | * Use the source to distinguish between the three cases */ | |
3955 | ||
3956 | if (*s == LATIN_SMALL_LETTER_SHARP_S) { | |
3957 | ||
3958 | /* uc() of this requires 2 characters, but they are | |
3959 | * ASCII. If not enough room, grow the string */ | |
3960 | if (SvLEN(dest) < ++min) { | |
3961 | const UV o = d - (U8*)SvPVX_const(dest); | |
3962 | SvGROW(dest, min); | |
3963 | d = (U8*)SvPVX(dest) + o; | |
3964 | } | |
3965 | *d++ = 'S'; *d = 'S'; /* upper case is 'SS' */ | |
3966 | continue; /* Back to the tight loop; still in ASCII */ | |
3967 | } | |
3968 | ||
3969 | /* The other two special handling characters have their | |
3970 | * upper cases outside the latin1 range, hence need to be | |
3971 | * in UTF-8, so the whole result needs to be in UTF-8. So, | |
3972 | * here we are somewhere in the middle of processing a | |
3973 | * non-UTF-8 string, and realize that we will have to convert | |
3974 | * the whole thing to UTF-8. What to do? There are | |
3975 | * several possibilities. The simplest to code is to | |
3976 | * convert what we have so far, set a flag, and continue on | |
3977 | * in the loop. The flag would be tested each time through | |
3978 | * the loop, and if set, the next character would be | |
3979 | * converted to UTF-8 and stored. But, I (khw) didn't want | |
3980 | * to slow down the mainstream case at all for this fairly | |
3981 | * rare case, so I didn't want to add a test that didn't | |
3982 | * absolutely have to be there in the loop, besides the | |
3983 | * possibility that it would get too complicated for | |
3984 | * optimizers to deal with. Another possibility is to just | |
3985 | * give up, convert the source to UTF-8, and restart the | |
3986 | * function that way. Another possibility is to convert | |
3987 | * both what has already been processed and what is yet to | |
3988 | * come separately to UTF-8, then jump into the loop that | |
3989 | * handles UTF-8. But the most efficient time-wise of the | |
3990 | * ones I could think of is what follows, and turned out to | |
3991 | * not require much extra code. */ | |
3992 | ||
3993 | /* Convert what we have so far into UTF-8, telling the | |
3994 | * function that we know it should be converted, and to | |
3995 | * allow extra space for what we haven't processed yet. | |
3996 | * Assume the worst case space requirements for converting | |
3997 | * what we haven't processed so far: that it will require | |
3998 | * two bytes for each remaining source character, plus the | |
3999 | * NUL at the end. This may cause the string pointer to | |
4000 | * move, so re-find it. */ | |
4001 | ||
4002 | len = d - (U8*)SvPVX_const(dest); | |
4003 | SvCUR_set(dest, len); | |
4004 | len = sv_utf8_upgrade_flags_grow(dest, | |
4005 | SV_GMAGIC|SV_FORCE_UTF8_UPGRADE, | |
4006 | (send -s) * 2 + 1); | |
4007 | d = (U8*)SvPVX(dest) + len; | |
4008 | ||
4009 | /* And append the current character's upper case in UTF-8 */ | |
4010 | CAT_NON_LATIN1_UC(d, *s); | |
4011 | ||
4012 | /* Now process the remainder of the source, converting to | |
4013 | * upper and UTF-8. If a resulting byte is invariant in | |
4014 | * UTF-8, output it as-is, otherwise convert to UTF-8 and | |
4015 | * append it to the output. */ | |
4016 | ||
4017 | s++; | |
4018 | for (; s < send; s++) { | |
4019 | U8 upper = toUPPER_LATIN1_MOD(*s); | |
4020 | if UTF8_IS_INVARIANT(upper) { | |
4021 | *d++ = upper; | |
4022 | } | |
4023 | else { | |
4024 | CAT_TWO_BYTE_UNI_UPPER_MOD(d, *s, upper); | |
4025 | } | |
4026 | } | |
4027 | ||
4028 | /* Here have processed the whole source; no need to continue | |
4029 | * with the outer loop. Each character has been converted | |
4030 | * to upper case and converted to UTF-8 */ | |
4031 | ||
4032 | break; | |
4033 | } /* End of processing all latin1-style chars */ | |
4034 | } /* End of processing all chars */ | |
4035 | } /* End of source is not empty */ | |
4036 | ||
67306194 | 4037 | if (source != dest) { |
00f254e2 | 4038 | *d = '\0'; /* Here d points to 1 after last char, add NUL */ |
67306194 NC |
4039 | SvCUR_set(dest, d - (U8*)SvPVX_const(dest)); |
4040 | } | |
00f254e2 | 4041 | } /* End of isn't utf8 */ |
539689e7 FC |
4042 | if (dest != source && SvTAINTED(source)) |
4043 | SvTAINT(dest); | |
67306194 | 4044 | SvSETMAGIC(dest); |
79072805 LW |
4045 | RETURN; |
4046 | } | |
4047 | ||
4048 | PP(pp_lc) | |
4049 | { | |
97aff369 | 4050 | dVAR; |
39644a26 | 4051 | dSP; |
ec9af7d4 | 4052 | SV *source = TOPs; |
463ee0b2 | 4053 | STRLEN len; |
ec9af7d4 NC |
4054 | STRLEN min; |
4055 | SV *dest; | |
4056 | const U8 *s; | |
4057 | U8 *d; | |
79072805 | 4058 | |
ec9af7d4 NC |
4059 | SvGETMAGIC(source); |
4060 | ||
4061 | if (SvPADTMP(source) && !SvREADONLY(source) && !SvAMAGIC(source) | |
17fa0776 | 4062 | && SvTEMP(source) && !DO_UTF8(source)) { |
ec9af7d4 | 4063 | |
00f254e2 KW |
4064 | /* We can convert in place, as lowercasing anything in the latin1 range |
4065 | * (or else DO_UTF8 would have been on) doesn't lengthen it */ | |
ec9af7d4 NC |
4066 | dest = source; |
4067 | s = d = (U8*)SvPV_force_nomg(source, len); | |
4068 | min = len + 1; | |
4069 | } else { | |
a0ed51b3 | 4070 | dTARGET; |
a0ed51b3 | 4071 | |
ec9af7d4 NC |
4072 | dest = TARG; |
4073 | ||
4074 | /* The old implementation would copy source into TARG at this point. | |
4075 | This had the side effect that if source was undef, TARG was now | |
4076 | an undefined SV with PADTMP set, and they don't warn inside | |
4077 | sv_2pv_flags(). However, we're now getting the PV direct from | |
4078 | source, which doesn't have PADTMP set, so it would warn. Hence the | |
4079 | little games. */ | |
4080 | ||
4081 | if (SvOK(source)) { | |
4082 | s = (const U8*)SvPV_nomg_const(source, len); | |
4083 | } else { | |
0a0ffbce RGS |
4084 | if (ckWARN(WARN_UNINITIALIZED)) |
4085 | report_uninit(source); | |
1eced8f8 | 4086 | s = (const U8*)""; |
ec9af7d4 | 4087 | len = 0; |
a0ed51b3 | 4088 | } |
ec9af7d4 | 4089 | min = len + 1; |
128c9517 | 4090 | |
ec9af7d4 | 4091 | SvUPGRADE(dest, SVt_PV); |
3b416f41 | 4092 | d = (U8*)SvGROW(dest, min); |
ec9af7d4 NC |
4093 | (void)SvPOK_only(dest); |
4094 | ||
4095 | SETs(dest); | |
4096 | } | |
4097 | ||
4098 | /* Overloaded values may have toggled the UTF-8 flag on source, so we need | |
4099 | to check DO_UTF8 again here. */ | |
4100 | ||
4101 | if (DO_UTF8(source)) { | |
4102 | const U8 *const send = s + len; | |
4103 | U8 tmpbuf[UTF8_MAXBYTES_CASE+1]; | |
4104 | ||
4105 | while (s < send) { | |
00f254e2 | 4106 | if (UTF8_IS_INVARIANT(*s)) { |
89ebb4a3 | 4107 | |
00f254e2 | 4108 | /* Invariant characters use the standard mappings compiled in. |
ec9af7d4 | 4109 | */ |
00f254e2 KW |
4110 | *d++ = toLOWER(*s); |
4111 | s++; | |
ec9af7d4 | 4112 | } |
00f254e2 | 4113 | else if (UTF8_IS_DOWNGRADEABLE_START(*s)) { |
89ebb4a3 | 4114 | |
00f254e2 | 4115 | /* As do the ones in the Latin1 range */ |
e1a8dbf5 | 4116 | U8 lower = toLOWER_LATIN1(TWO_BYTE_UTF8_TO_UNI(*s, *(s+1))); |
00f254e2 | 4117 | CAT_UNI_TO_UTF8_TWO_BYTE(d, lower); |
e1a8dbf5 | 4118 | s += 2; |
a0ed51b3 | 4119 | } |
00f254e2 | 4120 | else { |
00f254e2 KW |
4121 | /* Here, is utf8 not in Latin-1 range, have to go out and get |
4122 | * the mappings from the tables. */ | |
4123 | ||
4124 | const STRLEN u = UTF8SKIP(s); | |
4125 | STRLEN ulen; | |
4126 | ||
00f254e2 KW |
4127 | #ifndef CONTEXT_DEPENDENT_CASING |
4128 | toLOWER_utf8(s, tmpbuf, &ulen); | |
4129 | #else | |
9d2ba0fa KW |
4130 | /* This is ifdefd out because it probably is the wrong thing to do. The right |
4131 | * thing is probably to have an I/O layer that converts final sigma to regular | |
4132 | * on input and vice versa (under the correct circumstances) on output. In | |
4133 | * effect, the final sigma is just a glyph variation when the regular one | |
4134 | * occurs at the end of a word. And we don't really know what's going to be | |
4135 | * the end of the word until it is finally output, as splitting and joining can | |
4136 | * occur at any time and change what once was the word end to be in the middle, | |
4137 | * and vice versa. */ | |
00f254e2 KW |
4138 | |
4139 | const UV uv = toLOWER_utf8(s, tmpbuf, &ulen); | |
4140 | ||
4141 | /* If the lower case is a small sigma, it may be that we need | |
4142 | * to change it to a final sigma. This happens at the end of | |
4143 | * a word that contains more than just this character, and only | |
4144 | * when we started with a capital sigma. */ | |
4145 | if (uv == UNICODE_GREEK_SMALL_LETTER_SIGMA && | |
4146 | s > send - len && /* Makes sure not the first letter */ | |
4147 | utf8_to_uvchr(s, 0) == UNICODE_GREEK_CAPITAL_LETTER_SIGMA | |
4148 | ) { | |
4149 | ||
4150 | /* We use the algorithm in: | |
4151 | * http://www.unicode.org/versions/Unicode5.0.0/ch03.pdf (C | |
4152 | * is a CAPITAL SIGMA): If C is preceded by a sequence | |
4153 | * consisting of a cased letter and a case-ignorable | |
4154 | * sequence, and C is not followed by a sequence consisting | |
4155 | * of a case ignorable sequence and then a cased letter, | |
4156 | * then when lowercasing C, C becomes a final sigma */ | |
4157 | ||
4158 | /* To determine if this is the end of a word, need to peek | |
4159 | * ahead. Look at the next character */ | |
4160 | const U8 *peek = s + u; | |
4161 | ||
4162 | /* Skip any case ignorable characters */ | |
4163 | while (peek < send && is_utf8_case_ignorable(peek)) { | |
4164 | peek += UTF8SKIP(peek); | |
4165 | } | |
4166 | ||
4167 | /* If we reached the end of the string without finding any | |
4168 | * non-case ignorable characters, or if the next such one | |
4169 | * is not-cased, then we have met the conditions for it | |
4170 | * being a final sigma with regards to peek ahead, and so | |
4171 | * must do peek behind for the remaining conditions. (We | |
4172 | * know there is stuff behind to look at since we tested | |
4173 | * above that this isn't the first letter) */ | |
4174 | if (peek >= send || ! is_utf8_cased(peek)) { | |
4175 | peek = utf8_hop(s, -1); | |
4176 | ||
4177 | /* Here are at the beginning of the first character | |
4178 | * before the original upper case sigma. Keep backing | |
4179 | * up, skipping any case ignorable characters */ | |
4180 | while (is_utf8_case_ignorable(peek)) { | |
4181 | peek = utf8_hop(peek, -1); | |
4182 | } | |
4183 | ||
4184 | /* Here peek points to the first byte of the closest | |
4185 | * non-case-ignorable character before the capital | |
4186 | * sigma. If it is cased, then by the Unicode | |
4187 | * algorithm, we should use a small final sigma instead | |
4188 | * of what we have */ | |
4189 | if (is_utf8_cased(peek)) { | |
4190 | STORE_UNI_TO_UTF8_TWO_BYTE(tmpbuf, | |
4191 | UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA); | |
4192 | } | |
4193 | } | |
4194 | } | |
4195 | else { /* Not a context sensitive mapping */ | |
4196 | #endif /* End of commented out context sensitive */ | |
4197 | if (ulen > u && (SvLEN(dest) < (min += ulen - u))) { | |
4198 | ||
4199 | /* If the eventually required minimum size outgrows | |
4200 | * the available space, we need to grow. */ | |
4201 | const UV o = d - (U8*)SvPVX_const(dest); | |
4202 | ||
4203 | /* If someone lowercases one million U+0130s we | |
4204 | * SvGROW() one million times. Or we could try | |
4205 | * guessing how much to allocate without allocating too | |
4206 | * much. Such is life. Another option would be to | |
4207 | * grow an extra byte or two more each time we need to | |
4208 | * grow, which would cut down the million to 500K, with | |
4209 | * little waste */ | |
4210 | SvGROW(dest, min); | |
4211 | d = (U8*)SvPVX(dest) + o; | |
4212 | } | |
4213 | #ifdef CONTEXT_DEPENDENT_CASING | |
4214 | } | |
4215 | #endif | |
4216 | /* Copy the newly lowercased letter to the output buffer we're | |
4217 | * building */ | |
4218 | Copy(tmpbuf, d, ulen, U8); | |
4219 | d += ulen; | |
4220 | s += u; | |
00f254e2 | 4221 | } |
00f254e2 | 4222 | } /* End of looping through the source string */ |
ec9af7d4 NC |
4223 | SvUTF8_on(dest); |
4224 | *d = '\0'; | |
4225 | SvCUR_set(dest, d - (U8*)SvPVX_const(dest)); | |
00f254e2 | 4226 | } else { /* Not utf8 */ |
31351b04 | 4227 | if (len) { |
ec9af7d4 | 4228 | const U8 *const send = s + len; |
00f254e2 KW |
4229 | |
4230 | /* Use locale casing if in locale; regular style if not treating | |
4231 | * latin1 as having case; otherwise the latin1 casing. Do the | |
4232 | * whole thing in a tight loop, for speed, */ | |
2de3dbcc | 4233 | if (IN_LOCALE_RUNTIME) { |
31351b04 | 4234 | TAINT; |
ec9af7d4 NC |
4235 | SvTAINTED_on(dest); |
4236 | for (; s < send; d++, s++) | |
4237 | *d = toLOWER_LC(*s); | |
31351b04 | 4238 | } |
00f254e2 KW |
4239 | else if (! IN_UNI_8_BIT) { |
4240 | for (; s < send; d++, s++) { | |
ec9af7d4 | 4241 | *d = toLOWER(*s); |
00f254e2 KW |
4242 | } |
4243 | } | |
4244 | else { | |
4245 | for (; s < send; d++, s++) { | |
4246 | *d = toLOWER_LATIN1(*s); | |
4247 | } | |
31351b04 | 4248 | } |
bbce6d69 | 4249 | } |
ec9af7d4 NC |
4250 | if (source != dest) { |
4251 | *d = '\0'; | |
4252 | SvCUR_set(dest, d - (U8*)SvPVX_const(dest)); | |
4253 | } | |
79072805 | 4254 | } |
539689e7 FC |
4255 | if (dest != source && SvTAINTED(source)) |
4256 | SvTAINT(dest); | |
ec9af7d4 | 4257 | SvSETMAGIC(dest); |
79072805 LW |
4258 | RETURN; |
4259 | } | |
4260 | ||
a0d0e21e | 4261 | PP(pp_quotemeta) |
79072805 | 4262 | { |
97aff369 | 4263 | dVAR; dSP; dTARGET; |
1b6737cc | 4264 | SV * const sv = TOPs; |
a0d0e21e | 4265 | STRLEN len; |
0d46e09a | 4266 | register const char *s = SvPV_const(sv,len); |
79072805 | 4267 | |
7e2040f0 | 4268 | SvUTF8_off(TARG); /* decontaminate */ |
a0d0e21e | 4269 | if (len) { |
1b6737cc | 4270 | register char *d; |
862a34c6 | 4271 | SvUPGRADE(TARG, SVt_PV); |
c07a80fd | 4272 | SvGROW(TARG, (len * 2) + 1); |
a0d0e21e | 4273 | d = SvPVX(TARG); |
7e2040f0 | 4274 | if (DO_UTF8(sv)) { |
0dd2cdef | 4275 | while (len) { |
fd400ab9 | 4276 | if (UTF8_IS_CONTINUED(*s)) { |
0dd2cdef LW |
4277 | STRLEN ulen = UTF8SKIP(s); |
4278 | if (ulen > len) | |
4279 | ulen = len; | |
4280 | len -= ulen; | |
4281 | while (ulen--) | |
4282 | *d++ = *s++; | |
4283 | } | |
4284 | else { | |
4285 | if (!isALNUM(*s)) | |
4286 | *d++ = '\\'; | |
4287 | *d++ = *s++; | |
4288 | len--; | |
4289 | } | |
4290 | } | |
7e2040f0 | 4291 | SvUTF8_on(TARG); |
0dd2cdef LW |
4292 | } |
4293 | else { | |
4294 | while (len--) { | |
4295 | if (!isALNUM(*s)) | |
4296 | *d++ = '\\'; | |
4297 | *d++ = *s++; | |
4298 | } | |
79072805 | 4299 | } |
a0d0e21e | 4300 | *d = '\0'; |
349d4f2f | 4301 | SvCUR_set(TARG, d - SvPVX_const(TARG)); |
3aa33fe5 | 4302 | (void)SvPOK_only_UTF8(TARG); |
79072805 | 4303 | } |
a0d0e21e LW |
4304 | else |
4305 | sv_setpvn(TARG, s, len); | |
ec93b65f | 4306 | SETTARG; |
79072805 LW |
4307 | RETURN; |
4308 | } | |
4309 | ||
a0d0e21e | 4310 | /* Arrays. */ |
79072805 | 4311 | |
a0d0e21e | 4312 | PP(pp_aslice) |
79072805 | 4313 | { |
97aff369 | 4314 | dVAR; dSP; dMARK; dORIGMARK; |
502c6561 | 4315 | register AV *const av = MUTABLE_AV(POPs); |
1b6737cc | 4316 | register const I32 lval = (PL_op->op_flags & OPf_MOD || LVRET); |
79072805 | 4317 | |
a0d0e21e | 4318 | if (SvTYPE(av) == SVt_PVAV) { |
4ad10a0b VP |
4319 | const bool localizing = PL_op->op_private & OPpLVAL_INTRO; |
4320 | bool can_preserve = FALSE; | |
4321 | ||
4322 | if (localizing) { | |
4323 | MAGIC *mg; | |
4324 | HV *stash; | |
4325 | ||
4326 | can_preserve = SvCANEXISTDELETE(av); | |
4327 | } | |
4328 | ||
4329 | if (lval && localizing) { | |
1b6737cc | 4330 | register SV **svp; |
748a9306 | 4331 | I32 max = -1; |
924508f0 | 4332 | for (svp = MARK + 1; svp <= SP; svp++) { |
4ea561bc | 4333 | const I32 elem = SvIV(*svp); |
748a9306 LW |
4334 | if (elem > max) |
4335 | max = elem; | |
4336 | } | |
4337 | if (max > AvMAX(av)) | |
4338 | av_extend(av, max); | |
4339 | } | |
4ad10a0b | 4340 | |
a0d0e21e | 4341 | while (++MARK <= SP) { |
1b6737cc | 4342 | register SV **svp; |
4ea561bc | 4343 | I32 elem = SvIV(*MARK); |
4ad10a0b | 4344 | bool preeminent = TRUE; |
a0d0e21e | 4345 | |
4ad10a0b VP |
4346 | if (localizing && can_preserve) { |
4347 | /* If we can determine whether the element exist, | |
4348 | * Try to preserve the existenceness of a tied array | |
4349 | * element by using EXISTS and DELETE if possible. | |
4350 | * Fallback to FETCH and STORE otherwise. */ | |
4351 | preeminent = av_exists(av, elem); | |
4352 | } | |
4353 | ||
a0d0e21e LW |
4354 | svp = av_fetch(av, elem, lval); |
4355 | if (lval) { | |
3280af22 | 4356 | if (!svp || *svp == &PL_sv_undef) |
cea2e8a9 | 4357 | DIE(aTHX_ PL_no_aelem, elem); |
4ad10a0b VP |
4358 | if (localizing) { |
4359 | if (preeminent) | |
4360 | save_aelem(av, elem, svp); | |
4361 | else | |
4362 | SAVEADELETE(av, elem); | |
4363 | } | |
79072805 | 4364 | } |
3280af22 | 4365 | *MARK = svp ? *svp : &PL_sv_undef; |
79072805 LW |
4366 | } |
4367 | } | |
748a9306 | 4368 | if (GIMME != G_ARRAY) { |
a0d0e21e | 4369 | MARK = ORIGMARK; |
04ab2c87 | 4370 | *++MARK = SP > ORIGMARK ? *SP : &PL_sv_undef; |
a0d0e21e LW |
4371 | SP = MARK; |
4372 | } | |
79072805 LW |
4373 | RETURN; |
4374 | } | |
4375 | ||
cba5a3b0 DG |
4376 | /* Smart dereferencing for keys, values and each */ |
4377 | PP(pp_rkeys) | |
4378 | { | |
4379 | dVAR; | |
4380 | dSP; | |
4381 | dPOPss; | |
4382 | ||
7ac5715b FC |
4383 | SvGETMAGIC(sv); |
4384 | ||
4385 | if ( | |
4386 | !SvROK(sv) | |
4387 | || (sv = SvRV(sv), | |
4388 | (SvTYPE(sv) != SVt_PVHV && SvTYPE(sv) != SVt_PVAV) | |
4389 | || SvOBJECT(sv) | |
4390 | ) | |
4391 | ) { | |
4392 | DIE(aTHX_ | |
4393 | "Type of argument to %s must be unblessed hashref or arrayref", | |
4c540399 | 4394 | PL_op_desc[PL_op->op_type] ); |
cba5a3b0 DG |
4395 | } |
4396 | ||
d8065907 FC |
4397 | if (PL_op->op_flags & OPf_SPECIAL && SvTYPE(sv) == SVt_PVAV) |
4398 | DIE(aTHX_ | |
4399 | "Can't modify %s in %s", | |
4400 | PL_op_desc[PL_op->op_type], PL_op_desc[PL_op->op_next->op_type] | |
4401 | ); | |
4402 | ||
cba5a3b0 DG |
4403 | /* Delegate to correct function for op type */ |
4404 | PUSHs(sv); | |
4405 | if (PL_op->op_type == OP_RKEYS || PL_op->op_type == OP_RVALUES) { | |
4406 | return (SvTYPE(sv) == SVt_PVHV) ? Perl_do_kv(aTHX) : Perl_pp_akeys(aTHX); | |
4407 | } | |
4408 | else { | |
4409 | return (SvTYPE(sv) == SVt_PVHV) ? Perl_pp_each(aTHX) : Perl_pp_aeach(aTHX); | |
4410 | } | |
4411 | } | |
4412 | ||
878d132a NC |
4413 | PP(pp_aeach) |
4414 | { | |
4415 | dVAR; | |
4416 | dSP; | |
502c6561 | 4417 | AV *array = MUTABLE_AV(POPs); |
878d132a | 4418 | const I32 gimme = GIMME_V; |
453d94a9 | 4419 | IV *iterp = Perl_av_iter_p(aTHX_ array); |
878d132a NC |
4420 | const IV current = (*iterp)++; |
4421 | ||
4422 | if (current > av_len(array)) { | |
4423 | *iterp = 0; | |
4424 | if (gimme == G_SCALAR) | |
4425 | RETPUSHUNDEF; | |
4426 | else | |
4427 | RETURN; | |
4428 | } | |
4429 | ||
4430 | EXTEND(SP, 2); | |
e1dccc0d | 4431 | mPUSHi(current); |
878d132a NC |
4432 | if (gimme == G_ARRAY) { |
4433 | SV **const element = av_fetch(array, current, 0); | |
4434 | PUSHs(element ? *element : &PL_sv_undef); | |
4435 | } | |
4436 | RETURN; | |
4437 | } | |
4438 | ||
4439 | PP(pp_akeys) | |
4440 | { | |
4441 | dVAR; | |
4442 | dSP; | |
502c6561 | 4443 | AV *array = MUTABLE_AV(POPs); |
878d132a NC |
4444 | const I32 gimme = GIMME_V; |
4445 | ||
4446 | *Perl_av_iter_p(aTHX_ array) = 0; | |
4447 | ||
4448 | if (gimme == G_SCALAR) { | |
4449 | dTARGET; | |
4450 | PUSHi(av_len(array) + 1); | |
4451 | } | |
4452 | else if (gimme == G_ARRAY) { | |
4453 | IV n = Perl_av_len(aTHX_ array); | |
e1dccc0d | 4454 | IV i; |
878d132a NC |
4455 | |
4456 | EXTEND(SP, n + 1); | |
4457 | ||
cba5a3b0 | 4458 | if (PL_op->op_type == OP_AKEYS || PL_op->op_type == OP_RKEYS) { |
e1dccc0d | 4459 | for (i = 0; i <= n; i++) { |
878d132a NC |
4460 | mPUSHi(i); |
4461 | } | |
4462 | } | |
4463 | else { | |
4464 | for (i = 0; i <= n; i++) { | |
4465 | SV *const *const elem = Perl_av_fetch(aTHX_ array, i, 0); | |
4466 | PUSHs(elem ? *elem : &PL_sv_undef); | |
4467 | } | |
4468 | } | |
4469 | } | |
4470 | RETURN; | |
4471 | } | |
4472 | ||
79072805 LW |
4473 | /* Associative arrays. */ |
4474 | ||
4475 | PP(pp_each) | |
4476 | { | |
97aff369 | 4477 | dVAR; |
39644a26 | 4478 | dSP; |
85fbaab2 | 4479 | HV * hash = MUTABLE_HV(POPs); |
c07a80fd | 4480 | HE *entry; |
f54cb97a | 4481 | const I32 gimme = GIMME_V; |
8ec5e241 | 4482 | |
c07a80fd | 4483 | PUTBACK; |
c750a3ec | 4484 | /* might clobber stack_sp */ |
6d822dc4 | 4485 | entry = hv_iternext(hash); |
c07a80fd | 4486 | SPAGAIN; |
79072805 | 4487 | |
79072805 LW |
4488 | EXTEND(SP, 2); |
4489 | if (entry) { | |
1b6737cc | 4490 | SV* const sv = hv_iterkeysv(entry); |
574c8022 | 4491 | PUSHs(sv); /* won't clobber stack_sp */ |
54310121 | 4492 | if (gimme == G_ARRAY) { |
59af0135 | 4493 | SV *val; |
c07a80fd | 4494 | PUTBACK; |
c750a3ec | 4495 | /* might clobber stack_sp */ |
6d822dc4 | 4496 | val = hv_iterval(hash, entry); |
c07a80fd | 4497 | SPAGAIN; |
59af0135 | 4498 | PUSHs(val); |
79072805 | 4499 | } |
79072805 | 4500 | } |
54310121 | 4501 | else if (gimme == G_SCALAR) |
79072805 LW |
4502 | RETPUSHUNDEF; |
4503 | ||
4504 | RETURN; | |
4505 | } | |
4506 | ||
7332a6c4 VP |
4507 | STATIC OP * |
4508 | S_do_delete_local(pTHX) | |
79072805 | 4509 | { |
97aff369 | 4510 | dVAR; |
39644a26 | 4511 | dSP; |
f54cb97a | 4512 | const I32 gimme = GIMME_V; |
7332a6c4 VP |
4513 | const MAGIC *mg; |
4514 | HV *stash; | |
4515 | ||
4516 | if (PL_op->op_private & OPpSLICE) { | |
4517 | dMARK; dORIGMARK; | |
4518 | SV * const osv = POPs; | |
4519 | const bool tied = SvRMAGICAL(osv) | |
4520 | && mg_find((const SV *)osv, PERL_MAGIC_tied); | |
4521 | const bool can_preserve = SvCANEXISTDELETE(osv) | |
4522 | || mg_find((const SV *)osv, PERL_MAGIC_env); | |
4523 | const U32 type = SvTYPE(osv); | |
4524 | if (type == SVt_PVHV) { /* hash element */ | |
4525 | HV * const hv = MUTABLE_HV(osv); | |
4526 | while (++MARK <= SP) { | |
4527 | SV * const keysv = *MARK; | |
4528 | SV *sv = NULL; | |
4529 | bool preeminent = TRUE; | |
4530 | if (can_preserve) | |
4531 | preeminent = hv_exists_ent(hv, keysv, 0); | |
4532 | if (tied) { | |
4533 | HE *he = hv_fetch_ent(hv, keysv, 1, 0); | |
4534 | if (he) | |
4535 | sv = HeVAL(he); | |
4536 | else | |
4537 | preeminent = FALSE; | |
4538 | } | |
4539 | else { | |
4540 | sv = hv_delete_ent(hv, keysv, 0, 0); | |
4541 | SvREFCNT_inc_simple_void(sv); /* De-mortalize */ | |
4542 | } | |
4543 | if (preeminent) { | |
4544 | save_helem_flags(hv, keysv, &sv, SAVEf_KEEPOLDELEM); | |
4545 | if (tied) { | |
4546 | *MARK = sv_mortalcopy(sv); | |
4547 | mg_clear(sv); | |
4548 | } else | |
4549 | *MARK = sv; | |
4550 | } | |
4551 | else { | |
4552 | SAVEHDELETE(hv, keysv); | |
4553 | *MARK = &PL_sv_undef; | |
4554 | } | |
4555 | } | |
4556 | } | |
4557 | else if (type == SVt_PVAV) { /* array element */ | |
4558 | if (PL_op->op_flags & OPf_SPECIAL) { | |
4559 | AV * const av = MUTABLE_AV(osv); | |
4560 | while (++MARK <= SP) { | |
4561 | I32 idx = SvIV(*MARK); | |
4562 | SV *sv = NULL; | |
4563 | bool preeminent = TRUE; | |
4564 | if (can_preserve) | |
4565 | preeminent = av_exists(av, idx); | |
4566 | if (tied) { | |
4567 | SV **svp = av_fetch(av, idx, 1); | |
4568 | if (svp) | |
4569 | sv = *svp; | |
4570 | else | |
4571 | preeminent = FALSE; | |
4572 | } | |
4573 | else { | |
4574 | sv = av_delete(av, idx, 0); | |
4575 | SvREFCNT_inc_simple_void(sv); /* De-mortalize */ | |
4576 | } | |
4577 | if (preeminent) { | |
4578 | save_aelem_flags(av, idx, &sv, SAVEf_KEEPOLDELEM); | |
4579 | if (tied) { | |
4580 | *MARK = sv_mortalcopy(sv); | |
4581 | mg_clear(sv); | |
4582 | } else | |
4583 | *MARK = sv; | |
4584 | } | |
4585 | else { | |
4586 | SAVEADELETE(av, idx); | |
4587 | *MARK = &PL_sv_undef; | |
4588 | } | |
4589 | } | |
4590 | } | |
4591 | } | |
4592 | else | |
4593 | DIE(aTHX_ "Not a HASH reference"); | |
4594 | if (gimme == G_VOID) | |
4595 | SP = ORIGMARK; | |
4596 | else if (gimme == G_SCALAR) { | |
4597 | MARK = ORIGMARK; | |
4598 | if (SP > MARK) | |
4599 | *++MARK = *SP; | |
4600 | else | |
4601 | *++MARK = &PL_sv_undef; | |
4602 | SP = MARK; | |
4603 | } | |
4604 | } | |
4605 | else { | |
4606 | SV * const keysv = POPs; | |
4607 | SV * const osv = POPs; | |
4608 | const bool tied = SvRMAGICAL(osv) | |
4609 | && mg_find((const SV *)osv, PERL_MAGIC_tied); | |
4610 | const bool can_preserve = SvCANEXISTDELETE(osv) | |
4611 | || mg_find((const SV *)osv, PERL_MAGIC_env); | |
4612 | const U32 type = SvTYPE(osv); | |
4613 | SV *sv = NULL; | |
4614 | if (type == SVt_PVHV) { | |
4615 | HV * const hv = MUTABLE_HV(osv); | |
4616 | bool preeminent = TRUE; | |
4617 | if (can_preserve) | |
4618 | preeminent = hv_exists_ent(hv, keysv, 0); | |
4619 | if (tied) { | |
4620 | HE *he = hv_fetch_ent(hv, keysv, 1, 0); | |
4621 | if (he) | |
4622 | sv = HeVAL(he); | |
4623 | else | |
4624 | preeminent = FALSE; | |
4625 | } | |
4626 | else { | |
4627 | sv = hv_delete_ent(hv, keysv, 0, 0); | |
4628 | SvREFCNT_inc_simple_void(sv); /* De-mortalize */ | |
4629 | } | |
4630 | if (preeminent) { | |
4631 | save_helem_flags(hv, keysv, &sv, SAVEf_KEEPOLDELEM); | |
4632 | if (tied) { | |
4633 | SV *nsv = sv_mortalcopy(sv); | |
4634 | mg_clear(sv); | |
4635 | sv = nsv; | |
4636 | } | |
4637 | } | |
4638 | else | |
4639 | SAVEHDELETE(hv, keysv); | |
4640 | } | |
4641 | else if (type == SVt_PVAV) { | |
4642 | if (PL_op->op_flags & OPf_SPECIAL) { | |
4643 | AV * const av = MUTABLE_AV(osv); | |
4644 | I32 idx = SvIV(keysv); | |
4645 | bool preeminent = TRUE; | |
4646 | if (can_preserve) | |
4647 | preeminent = av_exists(av, idx); | |
4648 | if (tied) { | |
4649 | SV **svp = av_fetch(av, idx, 1); | |
4650 | if (svp) | |
4651 | sv = *svp; | |
4652 | else | |
4653 | preeminent = FALSE; | |
4654 | } | |
4655 | else { | |
4656 | sv = av_delete(av, idx, 0); | |
4657 | SvREFCNT_inc_simple_void(sv); /* De-mortalize */ | |
4658 | } | |
4659 | if (preeminent) { | |
4660 | save_aelem_flags(av, idx, &sv, SAVEf_KEEPOLDELEM); | |
4661 | if (tied) { | |
4662 | SV *nsv = sv_mortalcopy(sv); | |
4663 | mg_clear(sv); | |
4664 | sv = nsv; | |
4665 | } | |
4666 | } | |
4667 | else | |
4668 | SAVEADELETE(av, idx); | |
4669 | } | |
4670 | else | |
4671 | DIE(aTHX_ "panic: avhv_delete no longer supported"); | |
4672 | } | |
4673 | else | |
4674 | DIE(aTHX_ "Not a HASH reference"); | |
4675 | if (!sv) | |
4676 | sv = &PL_sv_undef; | |
4677 | if (gimme != G_VOID) | |
4678 | PUSHs(sv); | |
4679 | } | |
4680 | ||
4681 | RETURN; | |
4682 | } | |
4683 | ||
4684 | PP(pp_delete) | |
4685 | { | |
4686 | dVAR; | |
4687 | dSP; | |
4688 | I32 gimme; | |
4689 | I32 discard; | |
4690 | ||
4691 | if (PL_op->op_private & OPpLVAL_INTRO) | |
4692 | return do_delete_local(); | |
4693 | ||
4694 | gimme = GIMME_V; | |
4695 | discard = (gimme == G_VOID) ? G_DISCARD : 0; | |
5f05dabc | 4696 | |
533c011a | 4697 | if (PL_op->op_private & OPpSLICE) { |
5f05dabc | 4698 | dMARK; dORIGMARK; |
85fbaab2 | 4699 | HV * const hv = MUTABLE_HV(POPs); |
1b6737cc | 4700 | const U32 hvtype = SvTYPE(hv); |
01020589 GS |
4701 | if (hvtype == SVt_PVHV) { /* hash element */ |
4702 | while (++MARK <= SP) { | |
1b6737cc | 4703 | SV * const sv = hv_delete_ent(hv, *MARK, discard, 0); |
01020589 GS |
4704 | *MARK = sv ? sv : &PL_sv_undef; |
4705 | } | |
5f05dabc | 4706 | } |
6d822dc4 MS |
4707 | else if (hvtype == SVt_PVAV) { /* array element */ |
4708 | if (PL_op->op_flags & OPf_SPECIAL) { | |
4709 | while (++MARK <= SP) { | |
502c6561 | 4710 | SV * const sv = av_delete(MUTABLE_AV(hv), SvIV(*MARK), discard); |
6d822dc4 MS |
4711 | *MARK = sv ? sv : &PL_sv_undef; |
4712 | } | |
4713 | } | |
01020589 GS |
4714 | } |
4715 | else | |
4716 | DIE(aTHX_ "Not a HASH reference"); | |
54310121 | 4717 | if (discard) |
4718 | SP = ORIGMARK; | |
4719 | else if (gimme == G_SCALAR) { | |
5f05dabc | 4720 | MARK = ORIGMARK; |
9111c9c0 DM |
4721 | if (SP > MARK) |
4722 | *++MARK = *SP; | |
4723 | else | |
4724 | *++MARK = &PL_sv_undef; | |
5f05dabc | 4725 | SP = MARK; |
4726 | } | |
4727 | } | |
4728 | else { | |
4729 | SV *keysv = POPs; | |
85fbaab2 | 4730 | HV * const hv = MUTABLE_HV(POPs); |
295d248e | 4731 | SV *sv = NULL; |
97fcbf96 MB |
4732 | if (SvTYPE(hv) == SVt_PVHV) |
4733 | sv = hv_delete_ent(hv, keysv, discard, 0); | |
01020589 GS |
4734 | else if (SvTYPE(hv) == SVt_PVAV) { |
4735 | if (PL_op->op_flags & OPf_SPECIAL) | |
502c6561 | 4736 | sv = av_delete(MUTABLE_AV(hv), SvIV(keysv), discard); |
af288a60 HS |
4737 | else |
4738 | DIE(aTHX_ "panic: avhv_delete no longer supported"); | |
01020589 | 4739 | } |
97fcbf96 | 4740 | else |
cea2e8a9 | 4741 | DIE(aTHX_ "Not a HASH reference"); |
5f05dabc | 4742 | if (!sv) |
3280af22 | 4743 | sv = &PL_sv_undef; |
54310121 | 4744 | if (!discard) |
4745 | PUSHs(sv); | |
79072805 | 4746 | } |
79072805 LW |
4747 | RETURN; |
4748 | } | |
4749 | ||
a0d0e21e | 4750 | PP(pp_exists) |
79072805 | 4751 | { |
97aff369 | 4752 | dVAR; |
39644a26 | 4753 | dSP; |
afebc493 GS |
4754 | SV *tmpsv; |
4755 | HV *hv; | |
4756 | ||
4757 | if (PL_op->op_private & OPpEXISTS_SUB) { | |
4758 | GV *gv; | |
0bd48802 | 4759 | SV * const sv = POPs; |
f2c0649b | 4760 | CV * const cv = sv_2cv(sv, &hv, &gv, 0); |
afebc493 GS |
4761 | if (cv) |
4762 | RETPUSHYES; | |
4763 | if (gv && isGV(gv) && GvCV(gv) && !GvCVGEN(gv)) | |
4764 | RETPUSHYES; | |
4765 | RETPUSHNO; | |
4766 | } | |
4767 | tmpsv = POPs; | |
85fbaab2 | 4768 | hv = MUTABLE_HV(POPs); |
c750a3ec | 4769 | if (SvTYPE(hv) == SVt_PVHV) { |
ae77835f | 4770 | if (hv_exists_ent(hv, tmpsv, 0)) |
c750a3ec | 4771 | RETPUSHYES; |
ef54e1a4 JH |
4772 | } |
4773 | else if (SvTYPE(hv) == SVt_PVAV) { | |
01020589 | 4774 | if (PL_op->op_flags & OPf_SPECIAL) { /* array element */ |
502c6561 | 4775 | if (av_exists(MUTABLE_AV(hv), SvIV(tmpsv))) |
01020589 GS |
4776 | RETPUSHYES; |
4777 | } | |
ef54e1a4 JH |
4778 | } |
4779 | else { | |
cea2e8a9 | 4780 | DIE(aTHX_ "Not a HASH reference"); |
a0d0e21e | 4781 | } |
a0d0e21e LW |
4782 | RETPUSHNO; |
4783 | } | |
79072805 | 4784 | |
a0d0e21e LW |
4785 | PP(pp_hslice) |
4786 | { | |
97aff369 | 4787 | dVAR; dSP; dMARK; dORIGMARK; |
85fbaab2 | 4788 | register HV * const hv = MUTABLE_HV(POPs); |
1b6737cc AL |
4789 | register const I32 lval = (PL_op->op_flags & OPf_MOD || LVRET); |
4790 | const bool localizing = PL_op->op_private & OPpLVAL_INTRO; | |
d30e492c | 4791 | bool can_preserve = FALSE; |
79072805 | 4792 | |
eb85dfd3 DM |
4793 | if (localizing) { |
4794 | MAGIC *mg; | |
4795 | HV *stash; | |
4796 | ||
d30e492c VP |
4797 | if (SvCANEXISTDELETE(hv) || mg_find((const SV *)hv, PERL_MAGIC_env)) |
4798 | can_preserve = TRUE; | |
eb85dfd3 DM |
4799 | } |
4800 | ||
6d822dc4 | 4801 | while (++MARK <= SP) { |
1b6737cc | 4802 | SV * const keysv = *MARK; |
6d822dc4 MS |
4803 | SV **svp; |
4804 | HE *he; | |
d30e492c VP |
4805 | bool preeminent = TRUE; |
4806 | ||
4807 | if (localizing && can_preserve) { | |
4808 | /* If we can determine whether the element exist, | |
4809 | * try to preserve the existenceness of a tied hash | |
4810 | * element by using EXISTS and DELETE if possible. | |
4811 | * Fallback to FETCH and STORE otherwise. */ | |
4812 | preeminent = hv_exists_ent(hv, keysv, 0); | |
6d822dc4 | 4813 | } |
eb85dfd3 | 4814 | |
6d822dc4 | 4815 | he = hv_fetch_ent(hv, keysv, lval, 0); |
fe5bfecd | 4816 | svp = he ? &HeVAL(he) : NULL; |
eb85dfd3 | 4817 | |
6d822dc4 MS |
4818 | if (lval) { |
4819 | if (!svp || *svp == &PL_sv_undef) { | |
be2597df | 4820 | DIE(aTHX_ PL_no_helem_sv, SVfARG(keysv)); |
6d822dc4 MS |
4821 | } |
4822 | if (localizing) { | |
7a2e501a | 4823 | if (HvNAME_get(hv) && isGV(*svp)) |
159b6efe | 4824 | save_gp(MUTABLE_GV(*svp), !(PL_op->op_flags & OPf_SPECIAL)); |
47cfc530 VP |
4825 | else if (preeminent) |
4826 | save_helem_flags(hv, keysv, svp, | |
4827 | (PL_op->op_flags & OPf_SPECIAL) ? 0 : SAVEf_SETMAGIC); | |
4828 | else | |
4829 | SAVEHDELETE(hv, keysv); | |
6d822dc4 MS |
4830 | } |
4831 | } | |
4832 | *MARK = svp ? *svp : &PL_sv_undef; | |
79072805 | 4833 | } |
a0d0e21e LW |
4834 | if (GIMME != G_ARRAY) { |
4835 | MARK = ORIGMARK; | |
04ab2c87 | 4836 | *++MARK = SP > ORIGMARK ? *SP : &PL_sv_undef; |
a0d0e21e | 4837 | SP = MARK; |
79072805 | 4838 | } |
a0d0e21e LW |
4839 | RETURN; |
4840 | } | |
4841 | ||
4842 | /* List operators. */ | |
4843 | ||
4844 | PP(pp_list) | |
4845 | { | |
97aff369 | 4846 | dVAR; dSP; dMARK; |
a0d0e21e LW |
4847 | if (GIMME != G_ARRAY) { |
4848 | if (++MARK <= SP) | |
4849 | *MARK = *SP; /* unwanted list, return last item */ | |
8990e307 | 4850 | else |
3280af22 | 4851 | *MARK = &PL_sv_undef; |
a0d0e21e | 4852 | SP = MARK; |
79072805 | 4853 | } |
a0d0e21e | 4854 | RETURN; |
79072805 LW |
4855 | } |
4856 | ||
a0d0e21e | 4857 | PP(pp_lslice) |
79072805 | 4858 | { |
97aff369 | 4859 | dVAR; |
39644a26 | 4860 | dSP; |
1b6737cc AL |
4861 | SV ** const lastrelem = PL_stack_sp; |
4862 | SV ** const lastlelem = PL_stack_base + POPMARK; | |
4863 | SV ** const firstlelem = PL_stack_base + POPMARK + 1; | |
4864 | register SV ** const firstrelem = lastlelem + 1; | |
42e73ed0 | 4865 | I32 is_something_there = FALSE; |
1b6737cc AL |
4866 | |
4867 | register const I32 max = lastrelem - lastlelem; | |
a0d0e21e | 4868 | register SV **lelem; |
a0d0e21e LW |
4869 | |
4870 | if (GIMME != G_ARRAY) { | |
4ea561bc | 4871 | I32 ix = SvIV(*lastlelem); |
748a9306 LW |
4872 | if (ix < 0) |
4873 | ix += max; | |
a0d0e21e | 4874 | if (ix < 0 || ix >= max) |
3280af22 | 4875 | *firstlelem = &PL_sv_undef; |
a0d0e21e LW |
4876 | else |
4877 | *firstlelem = firstrelem[ix]; | |
4878 | SP = firstlelem; | |
4879 | RETURN; | |
4880 | } | |
4881 | ||
4882 | if (max == 0) { | |
4883 | SP = firstlelem - 1; | |
4884 | RETURN; | |
4885 | } | |
4886 | ||
4887 | for (lelem = firstlelem; lelem <= lastlelem; lelem++) { | |
4ea561bc | 4888 | I32 ix = SvIV(*lelem); |
c73bf8e3 | 4889 | if (ix < 0) |
a0d0e21e | 4890 | ix += max; |
c73bf8e3 HS |
4891 | if (ix < 0 || ix >= max) |
4892 | *lelem = &PL_sv_undef; | |
4893 | else { | |
4894 | is_something_there = TRUE; | |
4895 | if (!(*lelem = firstrelem[ix])) | |
3280af22 | 4896 | *lelem = &PL_sv_undef; |
748a9306 | 4897 | } |
79072805 | 4898 | } |
4633a7c4 LW |
4899 | if (is_something_there) |
4900 | SP = lastlelem; | |
4901 | else | |
4902 | SP = firstlelem - 1; | |
79072805 LW |
4903 | RETURN; |
4904 | } | |
4905 | ||
a0d0e21e LW |
4906 | PP(pp_anonlist) |
4907 | { | |
97aff369 | 4908 | dVAR; dSP; dMARK; dORIGMARK; |
1b6737cc | 4909 | const I32 items = SP - MARK; |
ad64d0ec | 4910 | SV * const av = MUTABLE_SV(av_make(items, MARK+1)); |
44a8e56a | 4911 | SP = ORIGMARK; /* av_make() might realloc stack_sp */ |
6e449a3a MHM |
4912 | mXPUSHs((PL_op->op_flags & OPf_SPECIAL) |
4913 | ? newRV_noinc(av) : av); | |
a0d0e21e LW |
4914 | RETURN; |
4915 | } | |
4916 | ||
4917 | PP(pp_anonhash) | |
79072805 | 4918 | { |
97aff369 | 4919 | dVAR; dSP; dMARK; dORIGMARK; |
78c72037 | 4920 | HV* const hv = newHV(); |
a0d0e21e LW |
4921 | |
4922 | while (MARK < SP) { | |
1b6737cc | 4923 | SV * const key = *++MARK; |
561b68a9 | 4924 | SV * const val = newSV(0); |
a0d0e21e LW |
4925 | if (MARK < SP) |
4926 | sv_setsv(val, *++MARK); | |
a2a5de95 NC |
4927 | else |
4928 | Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Odd number of elements in anonymous hash"); | |
f12c7020 | 4929 | (void)hv_store_ent(hv,key,val,0); |
79072805 | 4930 | } |
a0d0e21e | 4931 | SP = ORIGMARK; |
6e449a3a | 4932 | mXPUSHs((PL_op->op_flags & OPf_SPECIAL) |
ad64d0ec | 4933 | ? newRV_noinc(MUTABLE_SV(hv)) : MUTABLE_SV(hv)); |
79072805 LW |
4934 | RETURN; |
4935 | } | |
4936 | ||
d4fc4415 FC |
4937 | static AV * |
4938 | S_deref_plain_array(pTHX_ AV *ary) | |
4939 | { | |
4940 | if (SvTYPE(ary) == SVt_PVAV) return ary; | |
d2d95e13 | 4941 | SvGETMAGIC((SV *)ary); |
d4fc4415 FC |
4942 | if (!SvROK(ary) || SvTYPE(SvRV(ary)) != SVt_PVAV) |
4943 | Perl_die(aTHX_ "Not an ARRAY reference"); | |
4944 | else if (SvOBJECT(SvRV(ary))) | |
4945 | Perl_die(aTHX_ "Not an unblessed ARRAY reference"); | |
4946 | return (AV *)SvRV(ary); | |
4947 | } | |
4948 | ||
4949 | #if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN) | |
4950 | # define DEREF_PLAIN_ARRAY(ary) \ | |
4951 | ({ \ | |
4952 | AV *aRrRay = ary; \ | |
4953 | SvTYPE(aRrRay) == SVt_PVAV \ | |
4954 | ? aRrRay \ | |
4955 | : S_deref_plain_array(aTHX_ aRrRay); \ | |
4956 | }) | |
4957 | #else | |
4958 | # define DEREF_PLAIN_ARRAY(ary) \ | |
4959 | ( \ | |
3b0f6d32 | 4960 | PL_Sv = (SV *)(ary), \ |
d4fc4415 FC |
4961 | SvTYPE(PL_Sv) == SVt_PVAV \ |
4962 | ? (AV *)PL_Sv \ | |
3b0f6d32 | 4963 | : S_deref_plain_array(aTHX_ (AV *)PL_Sv) \ |
d4fc4415 FC |
4964 | ) |
4965 | #endif | |
4966 | ||
a0d0e21e | 4967 | PP(pp_splice) |
79072805 | 4968 | { |
27da23d5 | 4969 | dVAR; dSP; dMARK; dORIGMARK; |
5cd408a2 | 4970 | int num_args = (SP - MARK); |
d4fc4415 | 4971 | register AV *ary = DEREF_PLAIN_ARRAY(MUTABLE_AV(*++MARK)); |
a0d0e21e LW |
4972 | register SV **src; |
4973 | register SV **dst; | |
4974 | register I32 i; | |
4975 | register I32 offset; | |
4976 | register I32 length; | |
4977 | I32 newlen; | |
4978 | I32 after; | |
4979 | I32 diff; | |
ad64d0ec | 4980 | const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied); |
93965878 | 4981 | |
1b6737cc | 4982 | if (mg) { |
af71faff NC |
4983 | return Perl_tied_method(aTHX_ "SPLICE", mark - 1, MUTABLE_SV(ary), mg, |
4984 | GIMME_V | TIED_METHOD_ARGUMENTS_ON_STACK, | |
4985 | sp - mark); | |
93965878 | 4986 | } |
79072805 | 4987 | |
a0d0e21e | 4988 | SP++; |
79072805 | 4989 | |
a0d0e21e | 4990 | if (++MARK < SP) { |
4ea561bc | 4991 | offset = i = SvIV(*MARK); |
a0d0e21e | 4992 | if (offset < 0) |
93965878 | 4993 | offset += AvFILLp(ary) + 1; |
84902520 | 4994 | if (offset < 0) |
cea2e8a9 | 4995 | DIE(aTHX_ PL_no_aelem, i); |
a0d0e21e LW |
4996 | if (++MARK < SP) { |
4997 | length = SvIVx(*MARK++); | |
48cdf507 GA |
4998 | if (length < 0) { |
4999 | length += AvFILLp(ary) - offset + 1; | |
5000 | if (length < 0) | |
5001 | length = 0; | |
5002 | } | |
79072805 LW |
5003 | } |
5004 | else | |
a0d0e21e | 5005 | length = AvMAX(ary) + 1; /* close enough to infinity */ |
79072805 | 5006 | } |
a0d0e21e LW |
5007 | else { |
5008 | offset = 0; | |
5009 | length = AvMAX(ary) + 1; | |
5010 | } | |
8cbc2e3b | 5011 | if (offset > AvFILLp(ary) + 1) { |
5cd408a2 EB |
5012 | if (num_args > 2) |
5013 | Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "splice() offset past end of array" ); | |
93965878 | 5014 | offset = AvFILLp(ary) + 1; |
8cbc2e3b | 5015 | } |
93965878 | 5016 | after = AvFILLp(ary) + 1 - (offset + length); |
a0d0e21e LW |
5017 | if (after < 0) { /* not that much array */ |
5018 | length += after; /* offset+length now in array */ | |
5019 | after = 0; | |
5020 | if (!AvALLOC(ary)) | |
5021 | av_extend(ary, 0); | |
5022 | } | |
5023 | ||
5024 | /* At this point, MARK .. SP-1 is our new LIST */ | |
5025 | ||
5026 | newlen = SP - MARK; | |
5027 | diff = newlen - length; | |
13d7cbc1 GS |
5028 | if (newlen && !AvREAL(ary) && AvREIFY(ary)) |
5029 | av_reify(ary); | |
a0d0e21e | 5030 | |
50528de0 WL |
5031 | /* make new elements SVs now: avoid problems if they're from the array */ |
5032 | for (dst = MARK, i = newlen; i; i--) { | |
1b6737cc | 5033 | SV * const h = *dst; |
f2b990bf | 5034 | *dst++ = newSVsv(h); |
50528de0 WL |
5035 | } |
5036 | ||
a0d0e21e | 5037 | if (diff < 0) { /* shrinking the area */ |
95b63a38 | 5038 | SV **tmparyval = NULL; |
a0d0e21e | 5039 | if (newlen) { |
a02a5408 | 5040 | Newx(tmparyval, newlen, SV*); /* so remember insertion */ |
a0d0e21e | 5041 | Copy(MARK, tmparyval, newlen, SV*); |
79072805 | 5042 | } |
a0d0e21e LW |
5043 | |
5044 | MARK = ORIGMARK + 1; | |
5045 | if (GIMME == G_ARRAY) { /* copy return vals to stack */ | |
5046 | MEXTEND(MARK, length); | |
5047 | Copy(AvARRAY(ary)+offset, MARK, length, SV*); | |
5048 | if (AvREAL(ary)) { | |
bbce6d69 | 5049 | EXTEND_MORTAL(length); |
36477c24 | 5050 | for (i = length, dst = MARK; i; i--) { |
486ec47a | 5051 | sv_2mortal(*dst); /* free them eventually */ |
36477c24 | 5052 | dst++; |
5053 | } | |
a0d0e21e LW |
5054 | } |
5055 | MARK += length - 1; | |
79072805 | 5056 | } |
a0d0e21e LW |
5057 | else { |
5058 | *MARK = AvARRAY(ary)[offset+length-1]; | |
5059 | if (AvREAL(ary)) { | |
d689ffdd | 5060 | sv_2mortal(*MARK); |
a0d0e21e LW |
5061 | for (i = length - 1, dst = &AvARRAY(ary)[offset]; i > 0; i--) |
5062 | SvREFCNT_dec(*dst++); /* free them now */ | |
79072805 | 5063 | } |
a0d0e21e | 5064 | } |
93965878 | 5065 | AvFILLp(ary) += diff; |
a0d0e21e LW |
5066 | |
5067 | /* pull up or down? */ | |
5068 | ||
5069 | if (offset < after) { /* easier to pull up */ | |
5070 | if (offset) { /* esp. if nothing to pull */ | |
5071 | src = &AvARRAY(ary)[offset-1]; | |
5072 | dst = src - diff; /* diff is negative */ | |
5073 | for (i = offset; i > 0; i--) /* can't trust Copy */ | |
5074 | *dst-- = *src--; | |
79072805 | 5075 | } |
a0d0e21e | 5076 | dst = AvARRAY(ary); |
9c6bc640 | 5077 | AvARRAY(ary) = AvARRAY(ary) - diff; /* diff is negative */ |
a0d0e21e LW |
5078 | AvMAX(ary) += diff; |
5079 | } | |
5080 | else { | |
5081 | if (after) { /* anything to pull down? */ | |
5082 | src = AvARRAY(ary) + offset + length; | |
5083 | dst = src + diff; /* diff is negative */ | |
5084 | Move(src, dst, after, SV*); | |
79072805 | 5085 | } |
93965878 | 5086 | dst = &AvARRAY(ary)[AvFILLp(ary)+1]; |
a0d0e21e LW |
5087 | /* avoid later double free */ |
5088 | } | |
5089 | i = -diff; | |
5090 | while (i) | |
3280af22 | 5091 | dst[--i] = &PL_sv_undef; |
a0d0e21e LW |
5092 | |
5093 | if (newlen) { | |
50528de0 | 5094 | Copy( tmparyval, AvARRAY(ary) + offset, newlen, SV* ); |
a0d0e21e LW |
5095 | Safefree(tmparyval); |
5096 | } | |
5097 | } | |
5098 | else { /* no, expanding (or same) */ | |
d3961450 | 5099 | SV** tmparyval = NULL; |
a0d0e21e | 5100 | if (length) { |
a02a5408 | 5101 | Newx(tmparyval, length, SV*); /* so remember deletion */ |
a0d0e21e LW |
5102 | Copy(AvARRAY(ary)+offset, tmparyval, length, SV*); |
5103 | } | |
5104 | ||
5105 | if (diff > 0) { /* expanding */ | |
a0d0e21e | 5106 | /* push up or down? */ |
a0d0e21e LW |
5107 | if (offset < after && diff <= AvARRAY(ary) - AvALLOC(ary)) { |
5108 | if (offset) { | |
5109 | src = AvARRAY(ary); | |
5110 | dst = src - diff; | |
5111 | Move(src, dst, offset, SV*); | |
79072805 | 5112 | } |
9c6bc640 | 5113 | AvARRAY(ary) = AvARRAY(ary) - diff;/* diff is positive */ |
a0d0e21e | 5114 | AvMAX(ary) += diff; |
93965878 | 5115 | AvFILLp(ary) += diff; |
79072805 LW |
5116 | } |
5117 | else { | |
93965878 NIS |
5118 | if (AvFILLp(ary) + diff >= AvMAX(ary)) /* oh, well */ |
5119 | av_extend(ary, AvFILLp(ary) + diff); | |
5120 | AvFILLp(ary) += diff; | |
a0d0e21e LW |
5121 | |
5122 | if (after) { | |
93965878 | 5123 | dst = AvARRAY(ary) + AvFILLp(ary); |
a0d0e21e LW |
5124 | src = dst - diff; |
5125 | for (i = after; i; i--) { | |
5126 | *dst-- = *src--; | |
5127 | } | |
79072805 LW |
5128 | } |
5129 | } | |
a0d0e21e LW |
5130 | } |
5131 | ||
50528de0 WL |
5132 | if (newlen) { |
5133 | Copy( MARK, AvARRAY(ary) + offset, newlen, SV* ); | |
a0d0e21e | 5134 | } |
50528de0 | 5135 | |
a0d0e21e LW |
5136 | MARK = ORIGMARK + 1; |
5137 | if (GIMME == G_ARRAY) { /* copy return vals to stack */ | |
5138 | if (length) { | |
5139 | Copy(tmparyval, MARK, length, SV*); | |
5140 | if (AvREAL(ary)) { | |
bbce6d69 | 5141 | EXTEND_MORTAL(length); |
36477c24 | 5142 | for (i = length, dst = MARK; i; i--) { |
486ec47a | 5143 | sv_2mortal(*dst); /* free them eventually */ |
36477c24 | 5144 | dst++; |
5145 | } | |
79072805 LW |
5146 | } |
5147 | } | |
a0d0e21e LW |
5148 | MARK += length - 1; |
5149 | } | |
5150 | else if (length--) { | |
5151 | *MARK = tmparyval[length]; | |
5152 | if (AvREAL(ary)) { | |
d689ffdd | 5153 | sv_2mortal(*MARK); |
a0d0e21e LW |
5154 | while (length-- > 0) |
5155 | SvREFCNT_dec(tmparyval[length]); | |
79072805 | 5156 | } |
79072805 | 5157 | } |
a0d0e21e | 5158 | else |
3280af22 | 5159 | *MARK = &PL_sv_undef; |
d3961450 | 5160 | Safefree(tmparyval); |
79072805 | 5161 | } |
474af990 FR |
5162 | |
5163 | if (SvMAGICAL(ary)) | |
5164 | mg_set(MUTABLE_SV(ary)); | |
5165 | ||
a0d0e21e | 5166 | SP = MARK; |
79072805 LW |
5167 | RETURN; |
5168 | } | |
5169 | ||
a0d0e21e | 5170 | PP(pp_push) |
79072805 | 5171 | { |
27da23d5 | 5172 | dVAR; dSP; dMARK; dORIGMARK; dTARGET; |
d4fc4415 | 5173 | register AV * const ary = DEREF_PLAIN_ARRAY(MUTABLE_AV(*++MARK)); |
ad64d0ec | 5174 | const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied); |
79072805 | 5175 | |
1b6737cc | 5176 | if (mg) { |
ad64d0ec | 5177 | *MARK-- = SvTIED_obj(MUTABLE_SV(ary), mg); |
93965878 NIS |
5178 | PUSHMARK(MARK); |
5179 | PUTBACK; | |
d343c3ef | 5180 | ENTER_with_name("call_PUSH"); |
864dbfa3 | 5181 | call_method("PUSH",G_SCALAR|G_DISCARD); |
d343c3ef | 5182 | LEAVE_with_name("call_PUSH"); |
93965878 | 5183 | SPAGAIN; |
93965878 | 5184 | } |
a60c0954 | 5185 | else { |
89c14e2e | 5186 | PL_delaymagic = DM_DELAY; |
a60c0954 | 5187 | for (++MARK; MARK <= SP; MARK++) { |
561b68a9 | 5188 | SV * const sv = newSV(0); |
a60c0954 NIS |
5189 | if (*MARK) |
5190 | sv_setsv(sv, *MARK); | |
0a75904b | 5191 | av_store(ary, AvFILLp(ary)+1, sv); |
a60c0954 | 5192 | } |
354b0578 | 5193 | if (PL_delaymagic & DM_ARRAY_ISA) |
ad64d0ec | 5194 | mg_set(MUTABLE_SV(ary)); |
89c14e2e BB |
5195 | |
5196 | PL_delaymagic = 0; | |
6eeabd23 VP |
5197 | } |
5198 | SP = ORIGMARK; | |
5199 | if (OP_GIMME(PL_op, 0) != G_VOID) { | |
5200 | PUSHi( AvFILL(ary) + 1 ); | |
79072805 | 5201 | } |
79072805 LW |
5202 | RETURN; |
5203 | } | |
5204 | ||
a0d0e21e | 5205 | PP(pp_shift) |
79072805 | 5206 | { |
97aff369 | 5207 | dVAR; |
39644a26 | 5208 | dSP; |
538f5756 | 5209 | AV * const av = PL_op->op_flags & OPf_SPECIAL |
d4fc4415 | 5210 | ? MUTABLE_AV(GvAV(PL_defgv)) : DEREF_PLAIN_ARRAY(MUTABLE_AV(POPs)); |
789b4bc9 | 5211 | SV * const sv = PL_op->op_type == OP_SHIFT ? av_shift(av) : av_pop(av); |
79072805 | 5212 | EXTEND(SP, 1); |
c2b4a044 | 5213 | assert (sv); |
d689ffdd | 5214 | if (AvREAL(av)) |
a0d0e21e LW |
5215 | (void)sv_2mortal(sv); |
5216 | PUSHs(sv); | |
79072805 | 5217 | RETURN; |
79072805 LW |
5218 | } |
5219 | ||
a0d0e21e | 5220 | PP(pp_unshift) |
79072805 | 5221 | { |
27da23d5 | 5222 | dVAR; dSP; dMARK; dORIGMARK; dTARGET; |
d4fc4415 | 5223 | register AV *ary = DEREF_PLAIN_ARRAY(MUTABLE_AV(*++MARK)); |
ad64d0ec | 5224 | const MAGIC * const mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied); |
93965878 | 5225 | |
1b6737cc | 5226 | if (mg) { |
ad64d0ec | 5227 | *MARK-- = SvTIED_obj(MUTABLE_SV(ary), mg); |
7fd66d9d | 5228 | PUSHMARK(MARK); |
93965878 | 5229 | PUTBACK; |
d343c3ef | 5230 | ENTER_with_name("call_UNSHIFT"); |
864dbfa3 | 5231 | call_method("UNSHIFT",G_SCALAR|G_DISCARD); |
d343c3ef | 5232 | LEAVE_with_name("call_UNSHIFT"); |
93965878 | 5233 | SPAGAIN; |
93965878 | 5234 | } |
a60c0954 | 5235 | else { |
1b6737cc | 5236 | register I32 i = 0; |
a60c0954 NIS |
5237 | av_unshift(ary, SP - MARK); |
5238 | while (MARK < SP) { | |
1b6737cc | 5239 | SV * const sv = newSVsv(*++MARK); |
a60c0954 NIS |
5240 | (void)av_store(ary, i++, sv); |
5241 | } | |
79072805 | 5242 | } |
a0d0e21e | 5243 | SP = ORIGMARK; |
6eeabd23 | 5244 | if (OP_GIMME(PL_op, 0) != G_VOID) { |
5658d0a9 LR |
5245 | PUSHi( AvFILL(ary) + 1 ); |
5246 | } | |
79072805 | 5247 | RETURN; |
79072805 LW |
5248 | } |
5249 | ||
a0d0e21e | 5250 | PP(pp_reverse) |
79072805 | 5251 | { |
97aff369 | 5252 | dVAR; dSP; dMARK; |
79072805 | 5253 | |
a0d0e21e | 5254 | if (GIMME == G_ARRAY) { |
484c818f VP |
5255 | if (PL_op->op_private & OPpREVERSE_INPLACE) { |
5256 | AV *av; | |
5257 | ||
5258 | /* See pp_sort() */ | |
5259 | assert( MARK+1 == SP && *SP && SvTYPE(*SP) == SVt_PVAV); | |
5260 | (void)POPMARK; /* remove mark associated with ex-OP_AASSIGN */ | |
5261 | av = MUTABLE_AV((*SP)); | |
5262 | /* In-place reversing only happens in void context for the array | |
5263 | * assignment. We don't need to push anything on the stack. */ | |
5264 | SP = MARK; | |
5265 | ||
5266 | if (SvMAGICAL(av)) { | |
5267 | I32 i, j; | |
5268 | register SV *tmp = sv_newmortal(); | |
5269 | /* For SvCANEXISTDELETE */ | |
5270 | HV *stash; | |
5271 | const MAGIC *mg; | |
5272 | bool can_preserve = SvCANEXISTDELETE(av); | |
5273 | ||
5274 | for (i = 0, j = av_len(av); i < j; ++i, --j) { | |
5275 | register SV *begin, *end; | |
5276 | ||
5277 | if (can_preserve) { | |
5278 | if (!av_exists(av, i)) { | |
5279 | if (av_exists(av, j)) { | |
5280 | register SV *sv = av_delete(av, j, 0); | |
5281 | begin = *av_fetch(av, i, TRUE); | |
5282 | sv_setsv_mg(begin, sv); | |
5283 | } | |
5284 | continue; | |
5285 | } | |
5286 | else if (!av_exists(av, j)) { | |
5287 | register SV *sv = av_delete(av, i, 0); | |
5288 | end = *av_fetch(av, j, TRUE); | |
5289 | sv_setsv_mg(end, sv); | |
5290 | continue; | |
5291 | } | |
5292 | } | |
5293 | ||
5294 | begin = *av_fetch(av, i, TRUE); | |
5295 | end = *av_fetch(av, j, TRUE); | |
5296 | sv_setsv(tmp, begin); | |
5297 | sv_setsv_mg(begin, end); | |
5298 | sv_setsv_mg(end, tmp); | |
5299 | } | |
5300 | } | |
5301 | else { | |
5302 | SV **begin = AvARRAY(av); | |
484c818f | 5303 | |
95a26d8e VP |
5304 | if (begin) { |
5305 | SV **end = begin + AvFILLp(av); | |
5306 | ||
5307 | while (begin < end) { | |
5308 | register SV * const tmp = *begin; | |
5309 | *begin++ = *end; | |
5310 | *end-- = tmp; | |
5311 | } | |
484c818f VP |
5312 | } |
5313 | } | |
5314 | } | |
5315 | else { | |
5316 | SV **oldsp = SP; | |
5317 | MARK++; | |
5318 | while (MARK < SP) { | |
5319 | register SV * const tmp = *MARK; | |
5320 | *MARK++ = *SP; | |
5321 | *SP-- = tmp; | |
5322 | } | |
5323 | /* safe as long as stack cannot get extended in the above */ | |
5324 | SP = oldsp; | |
a0d0e21e | 5325 | } |
79072805 LW |
5326 | } |
5327 | else { | |
a0d0e21e LW |
5328 | register char *up; |
5329 | register char *down; | |
5330 | register I32 tmp; | |
5331 | dTARGET; | |
5332 | STRLEN len; | |
79072805 | 5333 | |
7e2040f0 | 5334 | SvUTF8_off(TARG); /* decontaminate */ |
a0d0e21e | 5335 | if (SP - MARK > 1) |
3280af22 | 5336 | do_join(TARG, &PL_sv_no, MARK, SP); |
1e21d011 | 5337 | else { |
789bd863 | 5338 | sv_setsv(TARG, SP > MARK ? *SP : find_rundefsv()); |
1e21d011 B |
5339 | if (! SvOK(TARG) && ckWARN(WARN_UNINITIALIZED)) |
5340 | report_uninit(TARG); | |
5341 | } | |
5342 | ||
a0d0e21e LW |
5343 | up = SvPV_force(TARG, len); |
5344 | if (len > 1) { | |
7e2040f0 | 5345 | if (DO_UTF8(TARG)) { /* first reverse each character */ |
dfe13c55 | 5346 | U8* s = (U8*)SvPVX(TARG); |
349d4f2f | 5347 | const U8* send = (U8*)(s + len); |
a0ed51b3 | 5348 | while (s < send) { |
d742c382 | 5349 | if (UTF8_IS_INVARIANT(*s)) { |
a0ed51b3 LW |
5350 | s++; |
5351 | continue; | |
5352 | } | |
5353 | else { | |
9041c2e3 | 5354 | if (!utf8_to_uvchr(s, 0)) |
a0dbb045 | 5355 | break; |
dfe13c55 | 5356 | up = (char*)s; |
a0ed51b3 | 5357 | s += UTF8SKIP(s); |
dfe13c55 | 5358 | down = (char*)(s - 1); |
a0dbb045 | 5359 | /* reverse this character */ |
a0ed51b3 LW |
5360 | while (down > up) { |
5361 | tmp = *up; | |
5362 | *up++ = *down; | |
eb160463 | 5363 | *down-- = (char)tmp; |
a0ed51b3 LW |
5364 | } |
5365 | } | |
5366 | } | |
5367 | up = SvPVX(TARG); | |
5368 | } | |
a0d0e21e LW |
5369 | down = SvPVX(TARG) + len - 1; |
5370 | while (down > up) { | |
5371 | tmp = *up; | |
5372 | *up++ = *down; | |
eb160463 | 5373 | *down-- = (char)tmp; |
a0d0e21e | 5374 | } |
3aa33fe5 | 5375 | (void)SvPOK_only_UTF8(TARG); |
79072805 | 5376 | } |
a0d0e21e LW |
5377 | SP = MARK + 1; |
5378 | SETTARG; | |
79072805 | 5379 | } |
a0d0e21e | 5380 | RETURN; |
79072805 LW |
5381 | } |
5382 | ||
a0d0e21e | 5383 | PP(pp_split) |
79072805 | 5384 | { |
27da23d5 | 5385 | dVAR; dSP; dTARG; |
a0d0e21e | 5386 | AV *ary; |
467f0320 | 5387 | register IV limit = POPi; /* note, negative is forever */ |
1b6737cc | 5388 | SV * const sv = POPs; |
a0d0e21e | 5389 | STRLEN len; |
727b7506 | 5390 | register const char *s = SvPV_const(sv, len); |
1b6737cc | 5391 | const bool do_utf8 = DO_UTF8(sv); |
727b7506 | 5392 | const char *strend = s + len; |
44a8e56a | 5393 | register PMOP *pm; |
d9f97599 | 5394 | register REGEXP *rx; |
a0d0e21e | 5395 | register SV *dstr; |
727b7506 | 5396 | register const char *m; |
a0d0e21e | 5397 | I32 iters = 0; |
bb7a0f54 | 5398 | const STRLEN slen = do_utf8 ? utf8_length((U8*)s, (U8*)strend) : (STRLEN)(strend - s); |
792b2c16 | 5399 | I32 maxiters = slen + 10; |
c1a7495a | 5400 | I32 trailing_empty = 0; |
727b7506 | 5401 | const char *orig; |
1b6737cc | 5402 | const I32 origlimit = limit; |
a0d0e21e LW |
5403 | I32 realarray = 0; |
5404 | I32 base; | |
f54cb97a | 5405 | const I32 gimme = GIMME_V; |
941446f6 | 5406 | bool gimme_scalar; |
f54cb97a | 5407 | const I32 oldsave = PL_savestack_ix; |
437d3b4e | 5408 | U32 make_mortal = SVs_TEMP; |
7fba1cd6 | 5409 | bool multiline = 0; |
b37c2d43 | 5410 | MAGIC *mg = NULL; |
79072805 | 5411 | |
44a8e56a | 5412 | #ifdef DEBUGGING |
5413 | Copy(&LvTARGOFF(POPs), &pm, 1, PMOP*); | |
5414 | #else | |
5415 | pm = (PMOP*)POPs; | |
5416 | #endif | |
a0d0e21e | 5417 | if (!pm || !s) |
2269b42e | 5418 | DIE(aTHX_ "panic: pp_split"); |
aaa362c4 | 5419 | rx = PM_GETRE(pm); |
bbce6d69 | 5420 | |
a62b1201 | 5421 | TAINT_IF(get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET && |
07bc277f | 5422 | (RX_EXTFLAGS(rx) & (RXf_WHITE | RXf_SKIPWHITE))); |
bbce6d69 | 5423 | |
a30b2f1f | 5424 | RX_MATCH_UTF8_set(rx, do_utf8); |
d9f424b2 | 5425 | |
971a9dd3 | 5426 | #ifdef USE_ITHREADS |
20e98b0f | 5427 | if (pm->op_pmreplrootu.op_pmtargetoff) { |
159b6efe | 5428 | ary = GvAVn(MUTABLE_GV(PAD_SVl(pm->op_pmreplrootu.op_pmtargetoff))); |
20e98b0f | 5429 | } |
971a9dd3 | 5430 | #else |
20e98b0f NC |
5431 | if (pm->op_pmreplrootu.op_pmtargetgv) { |
5432 | ary = GvAVn(pm->op_pmreplrootu.op_pmtargetgv); | |
971a9dd3 | 5433 | } |
20e98b0f | 5434 | #endif |
79072805 | 5435 | else |
7d49f689 | 5436 | ary = NULL; |
a0d0e21e LW |
5437 | if (ary && (gimme != G_ARRAY || (pm->op_pmflags & PMf_ONCE))) { |
5438 | realarray = 1; | |
8ec5e241 | 5439 | PUTBACK; |
a0d0e21e LW |
5440 | av_extend(ary,0); |
5441 | av_clear(ary); | |
8ec5e241 | 5442 | SPAGAIN; |
ad64d0ec | 5443 | if ((mg = SvTIED_mg((const SV *)ary, PERL_MAGIC_tied))) { |
8ec5e241 | 5444 | PUSHMARK(SP); |
ad64d0ec | 5445 | XPUSHs(SvTIED_obj(MUTABLE_SV(ary), mg)); |
8ec5e241 NIS |
5446 | } |
5447 | else { | |
1c0b011c | 5448 | if (!AvREAL(ary)) { |
1b6737cc | 5449 | I32 i; |
1c0b011c | 5450 | AvREAL_on(ary); |
abff13bb | 5451 | AvREIFY_off(ary); |
1c0b011c | 5452 | for (i = AvFILLp(ary); i >= 0; i--) |
3280af22 | 5453 | AvARRAY(ary)[i] = &PL_sv_undef; /* don't free mere refs */ |
1c0b011c NIS |
5454 | } |
5455 | /* temporarily switch stacks */ | |
8b7059b1 | 5456 | SAVESWITCHSTACK(PL_curstack, ary); |
8ec5e241 | 5457 | make_mortal = 0; |
1c0b011c | 5458 | } |
79072805 | 5459 | } |
3280af22 | 5460 | base = SP - PL_stack_base; |
a0d0e21e | 5461 | orig = s; |
07bc277f | 5462 | if (RX_EXTFLAGS(rx) & RXf_SKIPWHITE) { |
613f191e TS |
5463 | if (do_utf8) { |
5464 | while (*s == ' ' || is_utf8_space((U8*)s)) | |
5465 | s += UTF8SKIP(s); | |
5466 | } | |
a62b1201 | 5467 | else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET) { |
bbce6d69 | 5468 | while (isSPACE_LC(*s)) |
5469 | s++; | |
5470 | } | |
5471 | else { | |
5472 | while (isSPACE(*s)) | |
5473 | s++; | |
5474 | } | |
a0d0e21e | 5475 | } |
73134a2e | 5476 | if (RX_EXTFLAGS(rx) & RXf_PMf_MULTILINE) { |
7fba1cd6 | 5477 | multiline = 1; |
c07a80fd | 5478 | } |
5479 | ||
941446f6 FC |
5480 | gimme_scalar = gimme == G_SCALAR && !ary; |
5481 | ||
a0d0e21e LW |
5482 | if (!limit) |
5483 | limit = maxiters + 2; | |
07bc277f | 5484 | if (RX_EXTFLAGS(rx) & RXf_WHITE) { |
a0d0e21e | 5485 | while (--limit) { |
bbce6d69 | 5486 | m = s; |
8727f688 YO |
5487 | /* this one uses 'm' and is a negative test */ |
5488 | if (do_utf8) { | |
613f191e TS |
5489 | while (m < strend && !( *m == ' ' || is_utf8_space((U8*)m) )) { |
5490 | const int t = UTF8SKIP(m); | |
5491 | /* is_utf8_space returns FALSE for malform utf8 */ | |
5492 | if (strend - m < t) | |
5493 | m = strend; | |
5494 | else | |
5495 | m += t; | |
5496 | } | |
a62b1201 KW |
5497 | } |
5498 | else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET) { | |
8727f688 YO |
5499 | while (m < strend && !isSPACE_LC(*m)) |
5500 | ++m; | |
5501 | } else { | |
5502 | while (m < strend && !isSPACE(*m)) | |
5503 | ++m; | |
5504 | } | |
a0d0e21e LW |
5505 | if (m >= strend) |
5506 | break; | |
bbce6d69 | 5507 | |
c1a7495a BB |
5508 | if (gimme_scalar) { |
5509 | iters++; | |
5510 | if (m-s == 0) | |
5511 | trailing_empty++; | |
5512 | else | |
5513 | trailing_empty = 0; | |
5514 | } else { | |
5515 | dstr = newSVpvn_flags(s, m-s, | |
5516 | (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5517 | XPUSHs(dstr); | |
5518 | } | |
bbce6d69 | 5519 | |
613f191e TS |
5520 | /* skip the whitespace found last */ |
5521 | if (do_utf8) | |
5522 | s = m + UTF8SKIP(m); | |
5523 | else | |
5524 | s = m + 1; | |
5525 | ||
8727f688 YO |
5526 | /* this one uses 's' and is a positive test */ |
5527 | if (do_utf8) { | |
613f191e | 5528 | while (s < strend && ( *s == ' ' || is_utf8_space((U8*)s) )) |
8727f688 | 5529 | s += UTF8SKIP(s); |
a62b1201 KW |
5530 | } |
5531 | else if (get_regex_charset(RX_EXTFLAGS(rx)) == REGEX_LOCALE_CHARSET) { | |
8727f688 YO |
5532 | while (s < strend && isSPACE_LC(*s)) |
5533 | ++s; | |
5534 | } else { | |
5535 | while (s < strend && isSPACE(*s)) | |
5536 | ++s; | |
5537 | } | |
79072805 LW |
5538 | } |
5539 | } | |
07bc277f | 5540 | else if (RX_EXTFLAGS(rx) & RXf_START_ONLY) { |
a0d0e21e | 5541 | while (--limit) { |
a6e20a40 AL |
5542 | for (m = s; m < strend && *m != '\n'; m++) |
5543 | ; | |
a0d0e21e LW |
5544 | m++; |
5545 | if (m >= strend) | |
5546 | break; | |
c1a7495a BB |
5547 | |
5548 | if (gimme_scalar) { | |
5549 | iters++; | |
5550 | if (m-s == 0) | |
5551 | trailing_empty++; | |
5552 | else | |
5553 | trailing_empty = 0; | |
5554 | } else { | |
5555 | dstr = newSVpvn_flags(s, m-s, | |
5556 | (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5557 | XPUSHs(dstr); | |
5558 | } | |
a0d0e21e LW |
5559 | s = m; |
5560 | } | |
5561 | } | |
07bc277f | 5562 | else if (RX_EXTFLAGS(rx) & RXf_NULL && !(s >= strend)) { |
640f820d AB |
5563 | /* |
5564 | Pre-extend the stack, either the number of bytes or | |
5565 | characters in the string or a limited amount, triggered by: | |
5566 | ||
5567 | my ($x, $y) = split //, $str; | |
5568 | or | |
5569 | split //, $str, $i; | |
5570 | */ | |
c1a7495a BB |
5571 | if (!gimme_scalar) { |
5572 | const U32 items = limit - 1; | |
5573 | if (items < slen) | |
5574 | EXTEND(SP, items); | |
5575 | else | |
5576 | EXTEND(SP, slen); | |
5577 | } | |
640f820d | 5578 | |
e9515b0f AB |
5579 | if (do_utf8) { |
5580 | while (--limit) { | |
5581 | /* keep track of how many bytes we skip over */ | |
5582 | m = s; | |
640f820d | 5583 | s += UTF8SKIP(s); |
c1a7495a BB |
5584 | if (gimme_scalar) { |
5585 | iters++; | |
5586 | if (s-m == 0) | |
5587 | trailing_empty++; | |
5588 | else | |
5589 | trailing_empty = 0; | |
5590 | } else { | |
5591 | dstr = newSVpvn_flags(m, s-m, SVf_UTF8 | make_mortal); | |
640f820d | 5592 | |
c1a7495a BB |
5593 | PUSHs(dstr); |
5594 | } | |
640f820d | 5595 | |
e9515b0f AB |
5596 | if (s >= strend) |
5597 | break; | |
5598 | } | |
5599 | } else { | |
5600 | while (--limit) { | |
c1a7495a BB |
5601 | if (gimme_scalar) { |
5602 | iters++; | |
5603 | } else { | |
5604 | dstr = newSVpvn(s, 1); | |
e9515b0f | 5605 | |
e9515b0f | 5606 | |
c1a7495a BB |
5607 | if (make_mortal) |
5608 | sv_2mortal(dstr); | |
640f820d | 5609 | |
c1a7495a BB |
5610 | PUSHs(dstr); |
5611 | } | |
5612 | ||
5613 | s++; | |
e9515b0f AB |
5614 | |
5615 | if (s >= strend) | |
5616 | break; | |
5617 | } | |
640f820d AB |
5618 | } |
5619 | } | |
3c8556c3 | 5620 | else if (do_utf8 == (RX_UTF8(rx) != 0) && |
07bc277f NC |
5621 | (RX_EXTFLAGS(rx) & RXf_USE_INTUIT) && !RX_NPARENS(rx) |
5622 | && (RX_EXTFLAGS(rx) & RXf_CHECK_ALL) | |
5623 | && !(RX_EXTFLAGS(rx) & RXf_ANCH)) { | |
5624 | const int tail = (RX_EXTFLAGS(rx) & RXf_INTUIT_TAIL); | |
f9f4320a | 5625 | SV * const csv = CALLREG_INTUIT_STRING(rx); |
cf93c79d | 5626 | |
07bc277f | 5627 | len = RX_MINLENRET(rx); |
3c8556c3 | 5628 | if (len == 1 && !RX_UTF8(rx) && !tail) { |
1b6737cc | 5629 | const char c = *SvPV_nolen_const(csv); |
a0d0e21e | 5630 | while (--limit) { |
a6e20a40 AL |
5631 | for (m = s; m < strend && *m != c; m++) |
5632 | ; | |
a0d0e21e LW |
5633 | if (m >= strend) |
5634 | break; | |
c1a7495a BB |
5635 | if (gimme_scalar) { |
5636 | iters++; | |
5637 | if (m-s == 0) | |
5638 | trailing_empty++; | |
5639 | else | |
5640 | trailing_empty = 0; | |
5641 | } else { | |
5642 | dstr = newSVpvn_flags(s, m-s, | |
5643 | (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5644 | XPUSHs(dstr); | |
5645 | } | |
93f04dac JH |
5646 | /* The rx->minlen is in characters but we want to step |
5647 | * s ahead by bytes. */ | |
1aa99e6b IH |
5648 | if (do_utf8) |
5649 | s = (char*)utf8_hop((U8*)m, len); | |
5650 | else | |
5651 | s = m + len; /* Fake \n at the end */ | |
a0d0e21e LW |
5652 | } |
5653 | } | |
5654 | else { | |
a0d0e21e | 5655 | while (s < strend && --limit && |
f722798b | 5656 | (m = fbm_instr((unsigned char*)s, (unsigned char*)strend, |
7fba1cd6 | 5657 | csv, multiline ? FBMrf_MULTILINE : 0)) ) |
a0d0e21e | 5658 | { |
c1a7495a BB |
5659 | if (gimme_scalar) { |
5660 | iters++; | |
5661 | if (m-s == 0) | |
5662 | trailing_empty++; | |
5663 | else | |
5664 | trailing_empty = 0; | |
5665 | } else { | |
5666 | dstr = newSVpvn_flags(s, m-s, | |
5667 | (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5668 | XPUSHs(dstr); | |
5669 | } | |
93f04dac JH |
5670 | /* The rx->minlen is in characters but we want to step |
5671 | * s ahead by bytes. */ | |
1aa99e6b IH |
5672 | if (do_utf8) |
5673 | s = (char*)utf8_hop((U8*)m, len); | |
5674 | else | |
5675 | s = m + len; /* Fake \n at the end */ | |
a0d0e21e | 5676 | } |
463ee0b2 | 5677 | } |
463ee0b2 | 5678 | } |
a0d0e21e | 5679 | else { |
07bc277f | 5680 | maxiters += slen * RX_NPARENS(rx); |
080c2dec | 5681 | while (s < strend && --limit) |
bbce6d69 | 5682 | { |
1b6737cc | 5683 | I32 rex_return; |
080c2dec | 5684 | PUTBACK; |
f9f4320a | 5685 | rex_return = CALLREGEXEC(rx, (char*)s, (char*)strend, (char*)orig, 1 , |
bfafcb9a | 5686 | sv, NULL, SvSCREAM(sv) ? REXEC_SCREAM : 0); |
080c2dec | 5687 | SPAGAIN; |
1b6737cc | 5688 | if (rex_return == 0) |
080c2dec | 5689 | break; |
d9f97599 | 5690 | TAINT_IF(RX_MATCH_TAINTED(rx)); |
07bc277f | 5691 | if (RX_MATCH_COPIED(rx) && RX_SUBBEG(rx) != orig) { |
a0d0e21e LW |
5692 | m = s; |
5693 | s = orig; | |
07bc277f | 5694 | orig = RX_SUBBEG(rx); |
a0d0e21e LW |
5695 | s = orig + (m - s); |
5696 | strend = s + (strend - m); | |
5697 | } | |
07bc277f | 5698 | m = RX_OFFS(rx)[0].start + orig; |
c1a7495a BB |
5699 | |
5700 | if (gimme_scalar) { | |
5701 | iters++; | |
5702 | if (m-s == 0) | |
5703 | trailing_empty++; | |
5704 | else | |
5705 | trailing_empty = 0; | |
5706 | } else { | |
5707 | dstr = newSVpvn_flags(s, m-s, | |
5708 | (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5709 | XPUSHs(dstr); | |
5710 | } | |
07bc277f | 5711 | if (RX_NPARENS(rx)) { |
1b6737cc | 5712 | I32 i; |
07bc277f NC |
5713 | for (i = 1; i <= (I32)RX_NPARENS(rx); i++) { |
5714 | s = RX_OFFS(rx)[i].start + orig; | |
5715 | m = RX_OFFS(rx)[i].end + orig; | |
6de67870 JP |
5716 | |
5717 | /* japhy (07/27/01) -- the (m && s) test doesn't catch | |
5718 | parens that didn't match -- they should be set to | |
5719 | undef, not the empty string */ | |
c1a7495a BB |
5720 | if (gimme_scalar) { |
5721 | iters++; | |
5722 | if (m-s == 0) | |
5723 | trailing_empty++; | |
5724 | else | |
5725 | trailing_empty = 0; | |
5726 | } else { | |
5727 | if (m >= orig && s >= orig) { | |
5728 | dstr = newSVpvn_flags(s, m-s, | |
5729 | (do_utf8 ? SVf_UTF8 : 0) | |
5730 | | make_mortal); | |
5731 | } | |
5732 | else | |
5733 | dstr = &PL_sv_undef; /* undef, not "" */ | |
5734 | XPUSHs(dstr); | |
748a9306 | 5735 | } |
c1a7495a | 5736 | |
a0d0e21e LW |
5737 | } |
5738 | } | |
07bc277f | 5739 | s = RX_OFFS(rx)[0].end + orig; |
a0d0e21e | 5740 | } |
79072805 | 5741 | } |
8ec5e241 | 5742 | |
c1a7495a BB |
5743 | if (!gimme_scalar) { |
5744 | iters = (SP - PL_stack_base) - base; | |
5745 | } | |
a0d0e21e | 5746 | if (iters > maxiters) |
cea2e8a9 | 5747 | DIE(aTHX_ "Split loop"); |
8ec5e241 | 5748 | |
a0d0e21e LW |
5749 | /* keep field after final delim? */ |
5750 | if (s < strend || (iters && origlimit)) { | |
c1a7495a BB |
5751 | if (!gimme_scalar) { |
5752 | const STRLEN l = strend - s; | |
5753 | dstr = newSVpvn_flags(s, l, (do_utf8 ? SVf_UTF8 : 0) | make_mortal); | |
5754 | XPUSHs(dstr); | |
5755 | } | |
a0d0e21e | 5756 | iters++; |
79072805 | 5757 | } |
a0d0e21e | 5758 | else if (!origlimit) { |
c1a7495a BB |
5759 | if (gimme_scalar) { |
5760 | iters -= trailing_empty; | |
5761 | } else { | |
5762 | while (iters > 0 && (!TOPs || !SvANY(TOPs) || SvCUR(TOPs) == 0)) { | |
5763 | if (TOPs && !make_mortal) | |
5764 | sv_2mortal(TOPs); | |
5765 | *SP-- = &PL_sv_undef; | |
5766 | iters--; | |
5767 | } | |
89900bd3 | 5768 | } |
a0d0e21e | 5769 | } |
8ec5e241 | 5770 | |
8b7059b1 DM |
5771 | PUTBACK; |
5772 | LEAVE_SCOPE(oldsave); /* may undo an earlier SWITCHSTACK */ | |
5773 | SPAGAIN; | |
a0d0e21e | 5774 | if (realarray) { |
8ec5e241 | 5775 | if (!mg) { |
1c0b011c NIS |
5776 | if (SvSMAGICAL(ary)) { |
5777 | PUTBACK; | |
ad64d0ec | 5778 | mg_set(MUTABLE_SV(ary)); |
1c0b011c NIS |
5779 | SPAGAIN; |
5780 | } | |
5781 | if (gimme == G_ARRAY) { | |
5782 | EXTEND(SP, iters); | |
5783 | Copy(AvARRAY(ary), SP + 1, iters, SV*); | |
5784 | SP += iters; | |
5785 | RETURN; | |
5786 | } | |
8ec5e241 | 5787 | } |
1c0b011c | 5788 | else { |
fb73857a | 5789 | PUTBACK; |
d343c3ef | 5790 | ENTER_with_name("call_PUSH"); |
864dbfa3 | 5791 | call_method("PUSH",G_SCALAR|G_DISCARD); |
d343c3ef | 5792 | LEAVE_with_name("call_PUSH"); |
fb73857a | 5793 | SPAGAIN; |
8ec5e241 | 5794 | if (gimme == G_ARRAY) { |
1b6737cc | 5795 | I32 i; |
8ec5e241 NIS |
5796 | /* EXTEND should not be needed - we just popped them */ |
5797 | EXTEND(SP, iters); | |
5798 | for (i=0; i < iters; i++) { | |
5799 | SV **svp = av_fetch(ary, i, FALSE); | |
3280af22 | 5800 | PUSHs((svp) ? *svp : &PL_sv_undef); |
8ec5e241 | 5801 | } |
1c0b011c NIS |
5802 | RETURN; |
5803 | } | |
a0d0e21e LW |
5804 | } |
5805 | } | |
5806 | else { | |
5807 | if (gimme == G_ARRAY) | |
5808 | RETURN; | |
5809 | } | |
7f18b612 YST |
5810 | |
5811 | GETTARGET; | |
5812 | PUSHi(iters); | |
5813 | RETURN; | |
79072805 | 5814 | } |
85e6fe83 | 5815 | |
c5917253 NC |
5816 | PP(pp_once) |
5817 | { | |
5818 | dSP; | |
5819 | SV *const sv = PAD_SVl(PL_op->op_targ); | |
5820 | ||
5821 | if (SvPADSTALE(sv)) { | |
5822 | /* First time. */ | |
5823 | SvPADSTALE_off(sv); | |
5824 | RETURNOP(cLOGOP->op_other); | |
5825 | } | |
5826 | RETURNOP(cLOGOP->op_next); | |
5827 | } | |
5828 | ||
c0329465 MB |
5829 | PP(pp_lock) |
5830 | { | |
97aff369 | 5831 | dVAR; |
39644a26 | 5832 | dSP; |
c0329465 | 5833 | dTOPss; |
e55aaa0e | 5834 | SV *retsv = sv; |
68795e93 | 5835 | SvLOCK(sv); |
f79aa60b FC |
5836 | if (SvTYPE(retsv) == SVt_PVAV || SvTYPE(retsv) == SVt_PVHV |
5837 | || SvTYPE(retsv) == SVt_PVCV) { | |
e55aaa0e MB |
5838 | retsv = refto(retsv); |
5839 | } | |
5840 | SETs(retsv); | |
c0329465 MB |
5841 | RETURN; |
5842 | } | |
a863c7d1 | 5843 | |
65bca31a NC |
5844 | |
5845 | PP(unimplemented_op) | |
5846 | { | |
97aff369 | 5847 | dVAR; |
361ed549 NC |
5848 | const Optype op_type = PL_op->op_type; |
5849 | /* Using OP_NAME() isn't going to be helpful here. Firstly, it doesn't cope | |
5850 | with out of range op numbers - it only "special" cases op_custom. | |
5851 | Secondly, as the three ops we "panic" on are padmy, mapstart and custom, | |
5852 | if we get here for a custom op then that means that the custom op didn't | |
5853 | have an implementation. Given that OP_NAME() looks up the custom op | |
5854 | by its pp_addr, likely it will return NULL, unless someone (unhelpfully) | |
5855 | registers &PL_unimplemented_op as the address of their custom op. | |
5856 | NULL doesn't generate a useful error message. "custom" does. */ | |
5857 | const char *const name = op_type >= OP_max | |
5858 | ? "[out of range]" : PL_op_name[PL_op->op_type]; | |
7627e6d0 NC |
5859 | if(OP_IS_SOCKET(op_type)) |
5860 | DIE(aTHX_ PL_no_sock_func, name); | |
361ed549 | 5861 | DIE(aTHX_ "panic: unimplemented op %s (#%d) called", name, op_type); |
65bca31a NC |
5862 | } |
5863 | ||
867fa1e2 YO |
5864 | PP(pp_boolkeys) |
5865 | { | |
5866 | dVAR; | |
5867 | dSP; | |
5868 | HV * const hv = (HV*)POPs; | |
5869 | ||
fd1d9b5c FC |
5870 | if (SvTYPE(hv) != SVt_PVHV) { XPUSHs(&PL_sv_no); RETURN; } |
5871 | ||
867fa1e2 YO |
5872 | if (SvRMAGICAL(hv)) { |
5873 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_tied); | |
5874 | if (mg) { | |
5875 | XPUSHs(magic_scalarpack(hv, mg)); | |
5876 | RETURN; | |
5877 | } | |
5878 | } | |
5879 | ||
1b95d04f | 5880 | XPUSHs(boolSV(HvUSEDKEYS(hv) != 0)); |
867fa1e2 YO |
5881 | RETURN; |
5882 | } | |
5883 | ||
deb8a388 FC |
5884 | /* For sorting out arguments passed to a &CORE:: subroutine */ |
5885 | PP(pp_coreargs) | |
5886 | { | |
5887 | dSP; | |
7fa5bd9b | 5888 | int opnum = SvIOK(cSVOP_sv) ? (int)SvUV(cSVOP_sv) : 0; |
19c481f4 | 5889 | int defgv = PL_opargs[opnum] & OA_DEFGV, whicharg = 0; |
7fa5bd9b | 5890 | AV * const at_ = GvAV(PL_defgv); |
46e00a91 | 5891 | SV **svp = AvARRAY(at_); |
19c481f4 | 5892 | I32 minargs = 0, maxargs = 0, numargs = AvFILLp(at_)+1; |
7fa5bd9b | 5893 | I32 oa = opnum ? PL_opargs[opnum] >> OASHIFT : 0; |
46e00a91 | 5894 | bool seen_question = 0; |
7fa5bd9b | 5895 | const char *err = NULL; |
3e6568b4 | 5896 | const bool pushmark = PL_op->op_private & OPpCOREARGS_PUSHMARK; |
7fa5bd9b | 5897 | |
46e00a91 FC |
5898 | /* Count how many args there are first, to get some idea how far to |
5899 | extend the stack. */ | |
7fa5bd9b | 5900 | while (oa) { |
bf0571fd | 5901 | if ((oa & 7) == OA_LIST) { maxargs = I32_MAX; break; } |
7fa5bd9b | 5902 | maxargs++; |
46e00a91 FC |
5903 | if (oa & OA_OPTIONAL) seen_question = 1; |
5904 | if (!seen_question) minargs++; | |
7fa5bd9b FC |
5905 | oa >>= 4; |
5906 | } | |
5907 | ||
5908 | if(numargs < minargs) err = "Not enough"; | |
5909 | else if(numargs > maxargs) err = "Too many"; | |
5910 | if (err) | |
5911 | /* diag_listed_as: Too many arguments for %s */ | |
5912 | Perl_croak(aTHX_ | |
5913 | "%s arguments for %s", err, | |
5914 | opnum ? OP_DESC(PL_op->op_next) : SvPV_nolen_const(cSVOP_sv) | |
5915 | ); | |
5916 | ||
5917 | /* Reset the stack pointer. Without this, we end up returning our own | |
5918 | arguments in list context, in addition to the values we are supposed | |
5919 | to return. nextstate usually does this on sub entry, but we need | |
5920 | to run the next op with the caller’s hints, so we cannot have a | |
5921 | nextstate. */ | |
5922 | SP = PL_stack_base + cxstack[cxstack_ix].blk_oldsp; | |
5923 | ||
46e00a91 FC |
5924 | if(!maxargs) RETURN; |
5925 | ||
bf0571fd FC |
5926 | /* We do this here, rather than with a separate pushmark op, as it has |
5927 | to come in between two things this function does (stack reset and | |
5928 | arg pushing). This seems the easiest way to do it. */ | |
3e6568b4 | 5929 | if (pushmark) { |
bf0571fd FC |
5930 | PUTBACK; |
5931 | (void)Perl_pp_pushmark(aTHX); | |
5932 | } | |
5933 | ||
5934 | EXTEND(SP, maxargs == I32_MAX ? numargs : maxargs); | |
c931b036 | 5935 | PUTBACK; /* The code below can die in various places. */ |
46e00a91 FC |
5936 | |
5937 | oa = PL_opargs[opnum] >> OASHIFT; | |
3e6568b4 | 5938 | for (; oa&&(numargs||!pushmark); (void)(numargs&&(++svp,--numargs))) { |
c931b036 | 5939 | whicharg++; |
46e00a91 FC |
5940 | switch (oa & 7) { |
5941 | case OA_SCALAR: | |
d6d78e19 FC |
5942 | if (!numargs && defgv && whicharg == minargs + 1) { |
5943 | PERL_SI * const oldsi = PL_curstackinfo; | |
5944 | I32 const oldcxix = oldsi->si_cxix; | |
5945 | CV *caller; | |
5946 | if (oldcxix) oldsi->si_cxix--; | |
5947 | else PL_curstackinfo = oldsi->si_prev; | |
5948 | caller = find_runcv(NULL); | |
5949 | PL_curstackinfo = oldsi; | |
5950 | oldsi->si_cxix = oldcxix; | |
5951 | PUSHs(find_rundefsv2( | |
5952 | caller,cxstack[cxstack_ix].blk_oldcop->cop_seq | |
5953 | )); | |
5954 | } | |
5955 | else PUSHs(numargs ? svp && *svp ? *svp : &PL_sv_undef : NULL); | |
46e00a91 | 5956 | break; |
bf0571fd FC |
5957 | case OA_LIST: |
5958 | while (numargs--) { | |
5959 | PUSHs(svp && *svp ? *svp : &PL_sv_undef); | |
5960 | svp++; | |
5961 | } | |
5962 | RETURN; | |
19c481f4 FC |
5963 | case OA_HVREF: |
5964 | if (!svp || !*svp || !SvROK(*svp) | |
5965 | || SvTYPE(SvRV(*svp)) != SVt_PVHV) | |
5966 | DIE(aTHX_ | |
5967 | /* diag_listed_as: Type of arg %d to &CORE::%s must be %s*/ | |
5968 | "Type of arg %d to &CORE::%s must be hash reference", | |
5969 | whicharg, OP_DESC(PL_op->op_next) | |
5970 | ); | |
5971 | PUSHs(SvRV(*svp)); | |
5972 | break; | |
c931b036 | 5973 | case OA_FILEREF: |
30901a8a FC |
5974 | if (!numargs) PUSHs(NULL); |
5975 | else if(svp && *svp && SvROK(*svp) && isGV_with_GP(SvRV(*svp))) | |
c931b036 FC |
5976 | /* no magic here, as the prototype will have added an extra |
5977 | refgen and we just want what was there before that */ | |
5978 | PUSHs(SvRV(*svp)); | |
5979 | else { | |
5980 | const bool constr = PL_op->op_private & whicharg; | |
5981 | PUSHs(S_rv2gv(aTHX_ | |
5982 | svp && *svp ? *svp : &PL_sv_undef, | |
5983 | constr, CopHINTS_get(PL_curcop) & HINT_STRICT_REFS, | |
5984 | !constr | |
5985 | )); | |
5986 | } | |
5987 | break; | |
c72a5629 | 5988 | case OA_SCALARREF: |
17008668 FC |
5989 | { |
5990 | const bool wantscalar = | |
5991 | PL_op->op_private & OPpCOREARGS_SCALARMOD; | |
c72a5629 | 5992 | if (!svp || !*svp || !SvROK(*svp) |
17008668 FC |
5993 | /* We have to permit globrefs even for the \$ proto, as |
5994 | *foo is indistinguishable from ${\*foo}, and the proto- | |
5995 | type permits the latter. */ | |
5996 | || SvTYPE(SvRV(*svp)) > ( | |
efe889ae FC |
5997 | wantscalar ? SVt_PVLV |
5998 | : opnum == OP_LOCK ? SVt_PVCV | |
5999 | : SVt_PVHV | |
17008668 | 6000 | ) |
c72a5629 FC |
6001 | ) |
6002 | DIE(aTHX_ | |
6003 | /* diag_listed_as: Type of arg %d to &CORE::%s must be %s*/ | |
17008668 FC |
6004 | "Type of arg %d to &CORE::%s must be %s", |
6005 | whicharg, OP_DESC(PL_op->op_next), | |
6006 | wantscalar | |
6007 | ? "scalar reference" | |
efe889ae FC |
6008 | : opnum == OP_LOCK |
6009 | ? "reference to one of [$@%&*]" | |
6010 | : "reference to one of [$@%*]" | |
c72a5629 FC |
6011 | ); |
6012 | PUSHs(SvRV(*svp)); | |
6013 | break; | |
17008668 | 6014 | } |
46e00a91 | 6015 | default: |
46e00a91 FC |
6016 | DIE(aTHX_ "panic: unknown OA_*: %x", (unsigned)(oa&7)); |
6017 | } | |
6018 | oa = oa >> 4; | |
6019 | } | |
6020 | ||
deb8a388 FC |
6021 | RETURN; |
6022 | } | |
6023 | ||
e609e586 NC |
6024 | /* |
6025 | * Local variables: | |
6026 | * c-indentation-style: bsd | |
6027 | * c-basic-offset: 4 | |
6028 | * indent-tabs-mode: t | |
6029 | * End: | |
6030 | * | |
37442d52 RGS |
6031 | * ex: set ts=8 sts=4 sw=4 noet: |
6032 | */ |