This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Inline asIV and asUV, as each is only used once, and it distracts from
[perl5.git] / sv.c
CommitLineData
a0d0e21e 1/* sv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
241d1a3b 4 * 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others
79072805
LW
5 *
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.
8 *
a0d0e21e 9 * "I wonder what the Entish is for 'yes' and 'no'," he thought.
645c22ef
DM
10 *
11 *
5e045b90
AMS
12 * This file contains the code that creates, manipulates and destroys
13 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
14 * structure of an SV, so their creation and destruction is handled
15 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
16 * level functions (eg. substr, split, join) for each of the types are
17 * in the pp*.c files.
79072805
LW
18 */
19
20#include "EXTERN.h"
864dbfa3 21#define PERL_IN_SV_C
79072805 22#include "perl.h"
d2f185dc 23#include "regcomp.h"
79072805 24
51371543 25#define FCALL *f
2c5424a7 26
2f8ed50e
OS
27#ifdef __Lynx__
28/* Missing proto on LynxOS */
29 char *gconvert(double, int, int, char *);
30#endif
31
e23c8137
JH
32#ifdef PERL_UTF8_CACHE_ASSERT
33/* The cache element 0 is the Unicode offset;
34 * the cache element 1 is the byte offset of the element 0;
35 * the cache element 2 is the Unicode length of the substring;
36 * the cache element 3 is the byte length of the substring;
37 * The checking of the substring side would be good
38 * but substr() has enough code paths to make my head spin;
39 * if adding more checks watch out for the following tests:
40 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
41 * lib/utf8.t lib/Unicode/Collate/t/index.t
42 * --jhi
43 */
44#define ASSERT_UTF8_CACHE(cache) \
45 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); } } STMT_END
46#else
47#define ASSERT_UTF8_CACHE(cache) NOOP
48#endif
49
f8c7b90f 50#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 51#define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
607fa7f2 52#define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
b5ccf5f2 53/* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
765f542d 54 on-write. */
765f542d 55#endif
645c22ef
DM
56
57/* ============================================================================
58
59=head1 Allocation and deallocation of SVs.
60
5e045b90
AMS
61An SV (or AV, HV, etc.) is allocated in two parts: the head (struct sv,
62av, hv...) contains type and reference count information, as well as a
63pointer to the body (struct xrv, xpv, xpviv...), which contains fields
64specific to each type.
65
93e68bfb
JC
66In all but the most memory-paranoid configuations (ex: PURIFY), this
67allocation is done using arenas, which by default are approximately 4K
68chunks of memory parcelled up into N heads or bodies (of same size).
69Sv-bodies are allocated by their sv-type, guaranteeing size
70consistency needed to allocate safely from arrays.
71
72The first slot in each arena is reserved, and is used to hold a link
73to the next arena. In the case of heads, the unused first slot also
74contains some flags and a note of the number of slots. Snaked through
75each arena chain is a linked list of free items; when this becomes
76empty, an extra arena is allocated and divided up into N items which
77are threaded into the free list.
645c22ef
DM
78
79The following global variables are associated with arenas:
80
81 PL_sv_arenaroot pointer to list of SV arenas
82 PL_sv_root pointer to list of free SV structures
83
93e68bfb
JC
84 PL_body_arenaroots[] array of pointers to list of arenas, 1 per svtype
85 PL_body_roots[] array of pointers to list of free bodies of svtype
86 arrays are indexed by the svtype needed
645c22ef 87
93e68bfb
JC
88Note that some of the larger and more rarely used body types (eg
89xpvio) are not allocated using arenas, but are instead just
90malloc()/free()ed as required.
91
92In addition, a few SV heads are not allocated from an arena, but are
93instead directly created as static or auto variables, eg PL_sv_undef.
94The size of arenas can be changed from the default by setting
95PERL_ARENA_SIZE appropriately at compile time.
645c22ef
DM
96
97The SV arena serves the secondary purpose of allowing still-live SVs
98to be located and destroyed during final cleanup.
99
100At the lowest level, the macros new_SV() and del_SV() grab and free
101an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
102to return the SV to the free list with error checking.) new_SV() calls
103more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
104SVs in the free list have their SvTYPE field set to all ones.
105
106Similarly, there are macros new_XIV()/del_XIV(), new_XNV()/del_XNV() etc
107that allocate and return individual body types. Normally these are mapped
ff276b08
RG
108to the arena-manipulating functions new_xiv()/del_xiv() etc, but may be
109instead mapped directly to malloc()/free() if PURIFY is defined. The
645c22ef
DM
110new/del functions remove from, or add to, the appropriate PL_foo_root
111list, and call more_xiv() etc to add a new arena if the list is empty.
112
ff276b08 113At the time of very final cleanup, sv_free_arenas() is called from
645c22ef 114perl_destruct() to physically free all the arenas allocated since the
6a93a7e5 115start of the interpreter.
645c22ef
DM
116
117Manipulation of any of the PL_*root pointers is protected by enclosing
118LOCK_SV_MUTEX; ... UNLOCK_SV_MUTEX calls which should Do the Right Thing
119if threads are enabled.
120
121The function visit() scans the SV arenas list, and calls a specified
122function for each SV it finds which is still live - ie which has an SvTYPE
123other than all 1's, and a non-zero SvREFCNT. visit() is used by the
124following functions (specified as [function that calls visit()] / [function
125called by visit() for each SV]):
126
127 sv_report_used() / do_report_used()
128 dump all remaining SVs (debugging aid)
129
130 sv_clean_objs() / do_clean_objs(),do_clean_named_objs()
131 Attempt to free all objects pointed to by RVs,
132 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
133 try to do the same for all objects indirectly
134 referenced by typeglobs too. Called once from
135 perl_destruct(), prior to calling sv_clean_all()
136 below.
137
138 sv_clean_all() / do_clean_all()
139 SvREFCNT_dec(sv) each remaining SV, possibly
140 triggering an sv_free(). It also sets the
141 SVf_BREAK flag on the SV to indicate that the
142 refcnt has been artificially lowered, and thus
143 stopping sv_free() from giving spurious warnings
144 about SVs which unexpectedly have a refcnt
145 of zero. called repeatedly from perl_destruct()
146 until there are no SVs left.
147
93e68bfb 148=head2 Arena allocator API Summary
645c22ef
DM
149
150Private API to rest of sv.c
151
152 new_SV(), del_SV(),
153
154 new_XIV(), del_XIV(),
155 new_XNV(), del_XNV(),
156 etc
157
158Public API:
159
8cf8f3d1 160 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
645c22ef
DM
161
162
163=cut
164
165============================================================================ */
166
167
51371543 168
4561caa4
CS
169/*
170 * "A time to plant, and a time to uproot what was planted..."
171 */
172
77354fb4
NC
173/*
174 * nice_chunk and nice_chunk size need to be set
175 * and queried under the protection of sv_mutex
176 */
177void
178Perl_offer_nice_chunk(pTHX_ void *chunk, U32 chunk_size)
179{
180 void *new_chunk;
181 U32 new_chunk_size;
182 LOCK_SV_MUTEX;
183 new_chunk = (void *)(chunk);
184 new_chunk_size = (chunk_size);
185 if (new_chunk_size > PL_nice_chunk_size) {
186 Safefree(PL_nice_chunk);
187 PL_nice_chunk = (char *) new_chunk;
188 PL_nice_chunk_size = new_chunk_size;
189 } else {
190 Safefree(chunk);
191 }
192 UNLOCK_SV_MUTEX;
193}
cac9b346 194
fd0854ff 195#ifdef DEBUG_LEAKING_SCALARS
22162ca8 196# define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
fd0854ff
DM
197#else
198# define FREE_SV_DEBUG_FILE(sv)
199#endif
200
48614a46
NC
201#ifdef PERL_POISON
202# define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
203/* Whilst I'd love to do this, it seems that things like to check on
204 unreferenced scalars
205# define POSION_SV_HEAD(sv) Poison(sv, 1, struct STRUCT_SV)
206*/
207# define POSION_SV_HEAD(sv) Poison(&SvANY(sv), 1, void *), \
208 Poison(&SvREFCNT(sv), 1, U32)
209#else
210# define SvARENA_CHAIN(sv) SvANY(sv)
211# define POSION_SV_HEAD(sv)
212#endif
213
053fc874
GS
214#define plant_SV(p) \
215 STMT_START { \
fd0854ff 216 FREE_SV_DEBUG_FILE(p); \
48614a46
NC
217 POSION_SV_HEAD(p); \
218 SvARENA_CHAIN(p) = (void *)PL_sv_root; \
053fc874
GS
219 SvFLAGS(p) = SVTYPEMASK; \
220 PL_sv_root = (p); \
221 --PL_sv_count; \
222 } STMT_END
a0d0e21e 223
fba3b22e 224/* sv_mutex must be held while calling uproot_SV() */
053fc874
GS
225#define uproot_SV(p) \
226 STMT_START { \
227 (p) = PL_sv_root; \
48614a46 228 PL_sv_root = (SV*)SvARENA_CHAIN(p); \
053fc874
GS
229 ++PL_sv_count; \
230 } STMT_END
231
645c22ef 232
cac9b346
NC
233/* make some more SVs by adding another arena */
234
235/* sv_mutex must be held while calling more_sv() */
236STATIC SV*
237S_more_sv(pTHX)
238{
239 SV* sv;
240
241 if (PL_nice_chunk) {
242 sv_add_arena(PL_nice_chunk, PL_nice_chunk_size, 0);
243 PL_nice_chunk = Nullch;
244 PL_nice_chunk_size = 0;
245 }
246 else {
247 char *chunk; /* must use New here to match call to */
a02a5408 248 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
2e7ed132 249 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
cac9b346
NC
250 }
251 uproot_SV(sv);
252 return sv;
253}
254
645c22ef
DM
255/* new_SV(): return a new, empty SV head */
256
eba0f806
DM
257#ifdef DEBUG_LEAKING_SCALARS
258/* provide a real function for a debugger to play with */
259STATIC SV*
260S_new_SV(pTHX)
261{
262 SV* sv;
263
264 LOCK_SV_MUTEX;
265 if (PL_sv_root)
266 uproot_SV(sv);
267 else
cac9b346 268 sv = S_more_sv(aTHX);
eba0f806
DM
269 UNLOCK_SV_MUTEX;
270 SvANY(sv) = 0;
271 SvREFCNT(sv) = 1;
272 SvFLAGS(sv) = 0;
fd0854ff
DM
273 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
274 sv->sv_debug_line = (U16) ((PL_copline == NOLINE) ?
275 (PL_curcop ? CopLINE(PL_curcop) : 0) : PL_copline);
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_cloned = 0;
fd0854ff 278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
fd0854ff 279
eba0f806
DM
280 return sv;
281}
282# define new_SV(p) (p)=S_new_SV(aTHX)
283
284#else
285# define new_SV(p) \
053fc874
GS
286 STMT_START { \
287 LOCK_SV_MUTEX; \
288 if (PL_sv_root) \
289 uproot_SV(p); \
290 else \
cac9b346 291 (p) = S_more_sv(aTHX); \
053fc874
GS
292 UNLOCK_SV_MUTEX; \
293 SvANY(p) = 0; \
294 SvREFCNT(p) = 1; \
295 SvFLAGS(p) = 0; \
296 } STMT_END
eba0f806 297#endif
463ee0b2 298
645c22ef
DM
299
300/* del_SV(): return an empty SV head to the free list */
301
a0d0e21e 302#ifdef DEBUGGING
4561caa4 303
053fc874
GS
304#define del_SV(p) \
305 STMT_START { \
306 LOCK_SV_MUTEX; \
aea4f609 307 if (DEBUG_D_TEST) \
053fc874
GS
308 del_sv(p); \
309 else \
310 plant_SV(p); \
311 UNLOCK_SV_MUTEX; \
312 } STMT_END
a0d0e21e 313
76e3520e 314STATIC void
cea2e8a9 315S_del_sv(pTHX_ SV *p)
463ee0b2 316{
aea4f609 317 if (DEBUG_D_TEST) {
4633a7c4 318 SV* sva;
a3b680e6 319 bool ok = 0;
3280af22 320 for (sva = PL_sv_arenaroot; sva; sva = (SV *) SvANY(sva)) {
53c1dcc0
AL
321 const SV * const sv = sva + 1;
322 const SV * const svend = &sva[SvREFCNT(sva)];
c0ff570e 323 if (p >= sv && p < svend) {
a0d0e21e 324 ok = 1;
c0ff570e
NC
325 break;
326 }
a0d0e21e
LW
327 }
328 if (!ok) {
0453d815 329 if (ckWARN_d(WARN_INTERNAL))
9014280d 330 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
331 "Attempt to free non-arena SV: 0x%"UVxf
332 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
a0d0e21e
LW
333 return;
334 }
335 }
4561caa4 336 plant_SV(p);
463ee0b2 337}
a0d0e21e 338
4561caa4
CS
339#else /* ! DEBUGGING */
340
341#define del_SV(p) plant_SV(p)
342
343#endif /* DEBUGGING */
463ee0b2 344
645c22ef
DM
345
346/*
ccfc67b7
JH
347=head1 SV Manipulation Functions
348
645c22ef
DM
349=for apidoc sv_add_arena
350
351Given a chunk of memory, link it to the head of the list of arenas,
352and split it into a list of free SVs.
353
354=cut
355*/
356
4633a7c4 357void
864dbfa3 358Perl_sv_add_arena(pTHX_ char *ptr, U32 size, U32 flags)
463ee0b2 359{
0bd48802 360 SV* const sva = (SV*)ptr;
463ee0b2
LW
361 register SV* sv;
362 register SV* svend;
4633a7c4
LW
363
364 /* The first SV in an arena isn't an SV. */
3280af22 365 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
4633a7c4
LW
366 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
367 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
368
3280af22
NIS
369 PL_sv_arenaroot = sva;
370 PL_sv_root = sva + 1;
4633a7c4
LW
371
372 svend = &sva[SvREFCNT(sva) - 1];
373 sv = sva + 1;
463ee0b2 374 while (sv < svend) {
48614a46 375 SvARENA_CHAIN(sv) = (void *)(SV*)(sv + 1);
03e36789 376#ifdef DEBUGGING
978b032e 377 SvREFCNT(sv) = 0;
03e36789
NC
378#endif
379 /* Must always set typemask because it's awlays checked in on cleanup
380 when the arenas are walked looking for objects. */
8990e307 381 SvFLAGS(sv) = SVTYPEMASK;
463ee0b2
LW
382 sv++;
383 }
48614a46 384 SvARENA_CHAIN(sv) = 0;
03e36789
NC
385#ifdef DEBUGGING
386 SvREFCNT(sv) = 0;
387#endif
4633a7c4
LW
388 SvFLAGS(sv) = SVTYPEMASK;
389}
390
055972dc
DM
391/* visit(): call the named function for each non-free SV in the arenas
392 * whose flags field matches the flags/mask args. */
645c22ef 393
5226ed68 394STATIC I32
055972dc 395S_visit(pTHX_ SVFUNC_t f, U32 flags, U32 mask)
8990e307 396{
4633a7c4 397 SV* sva;
5226ed68 398 I32 visited = 0;
8990e307 399
3280af22 400 for (sva = PL_sv_arenaroot; sva; sva = (SV*)SvANY(sva)) {
53c1dcc0 401 register const SV * const svend = &sva[SvREFCNT(sva)];
a3b680e6 402 register SV* sv;
4561caa4 403 for (sv = sva + 1; sv < svend; ++sv) {
055972dc
DM
404 if (SvTYPE(sv) != SVTYPEMASK
405 && (sv->sv_flags & mask) == flags
406 && SvREFCNT(sv))
407 {
acfe0abc 408 (FCALL)(aTHX_ sv);
5226ed68
JH
409 ++visited;
410 }
8990e307
LW
411 }
412 }
5226ed68 413 return visited;
8990e307
LW
414}
415
758a08c3
JH
416#ifdef DEBUGGING
417
645c22ef
DM
418/* called by sv_report_used() for each live SV */
419
420static void
acfe0abc 421do_report_used(pTHX_ SV *sv)
645c22ef
DM
422{
423 if (SvTYPE(sv) != SVTYPEMASK) {
424 PerlIO_printf(Perl_debug_log, "****\n");
425 sv_dump(sv);
426 }
427}
758a08c3 428#endif
645c22ef
DM
429
430/*
431=for apidoc sv_report_used
432
433Dump the contents of all SVs not yet freed. (Debugging aid).
434
435=cut
436*/
437
8990e307 438void
864dbfa3 439Perl_sv_report_used(pTHX)
4561caa4 440{
ff270d3a 441#ifdef DEBUGGING
055972dc 442 visit(do_report_used, 0, 0);
ff270d3a 443#endif
4561caa4
CS
444}
445
645c22ef
DM
446/* called by sv_clean_objs() for each live SV */
447
448static void
e15faf7d 449do_clean_objs(pTHX_ SV *ref)
645c22ef 450{
823a54a3
AL
451 if (SvROK(ref)) {
452 SV * const target = SvRV(ref);
453 if (SvOBJECT(target)) {
454 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
455 if (SvWEAKREF(ref)) {
456 sv_del_backref(target, ref);
457 SvWEAKREF_off(ref);
458 SvRV_set(ref, NULL);
459 } else {
460 SvROK_off(ref);
461 SvRV_set(ref, NULL);
462 SvREFCNT_dec(target);
463 }
645c22ef
DM
464 }
465 }
466
467 /* XXX Might want to check arrays, etc. */
468}
469
470/* called by sv_clean_objs() for each live SV */
471
472#ifndef DISABLE_DESTRUCTOR_KLUDGE
473static void
acfe0abc 474do_clean_named_objs(pTHX_ SV *sv)
645c22ef
DM
475{
476 if (SvTYPE(sv) == SVt_PVGV && GvGP(sv)) {
c69033f2
NC
477 if ((
478#ifdef PERL_DONT_CREATE_GVSV
479 GvSV(sv) &&
480#endif
481 SvOBJECT(GvSV(sv))) ||
645c22ef
DM
482 (GvAV(sv) && SvOBJECT(GvAV(sv))) ||
483 (GvHV(sv) && SvOBJECT(GvHV(sv))) ||
484 (GvIO(sv) && SvOBJECT(GvIO(sv))) ||
485 (GvCV(sv) && SvOBJECT(GvCV(sv))) )
486 {
487 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning named glob object:\n "), sv_dump(sv)));
ec5f3c78 488 SvFLAGS(sv) |= SVf_BREAK;
645c22ef
DM
489 SvREFCNT_dec(sv);
490 }
491 }
492}
493#endif
494
495/*
496=for apidoc sv_clean_objs
497
498Attempt to destroy all objects not yet freed
499
500=cut
501*/
502
4561caa4 503void
864dbfa3 504Perl_sv_clean_objs(pTHX)
4561caa4 505{
3280af22 506 PL_in_clean_objs = TRUE;
055972dc 507 visit(do_clean_objs, SVf_ROK, SVf_ROK);
4561caa4 508#ifndef DISABLE_DESTRUCTOR_KLUDGE
2d0f3c12 509 /* some barnacles may yet remain, clinging to typeglobs */
055972dc 510 visit(do_clean_named_objs, SVt_PVGV, SVTYPEMASK);
4561caa4 511#endif
3280af22 512 PL_in_clean_objs = FALSE;
4561caa4
CS
513}
514
645c22ef
DM
515/* called by sv_clean_all() for each live SV */
516
517static void
acfe0abc 518do_clean_all(pTHX_ SV *sv)
645c22ef
DM
519{
520 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
521 SvFLAGS(sv) |= SVf_BREAK;
0e705b3b
DM
522 if (PL_comppad == (AV*)sv) {
523 PL_comppad = Nullav;
524 PL_curpad = Null(SV**);
525 }
645c22ef
DM
526 SvREFCNT_dec(sv);
527}
528
529/*
530=for apidoc sv_clean_all
531
532Decrement the refcnt of each remaining SV, possibly triggering a
533cleanup. This function may have to be called multiple times to free
ff276b08 534SVs which are in complex self-referential hierarchies.
645c22ef
DM
535
536=cut
537*/
538
5226ed68 539I32
864dbfa3 540Perl_sv_clean_all(pTHX)
8990e307 541{
5226ed68 542 I32 cleaned;
3280af22 543 PL_in_clean_all = TRUE;
055972dc 544 cleaned = visit(do_clean_all, 0,0);
3280af22 545 PL_in_clean_all = FALSE;
5226ed68 546 return cleaned;
8990e307 547}
463ee0b2 548
7cfef17e
NC
549static void
550S_free_arena(pTHX_ void **root) {
551 while (root) {
1b6737cc 552 void ** const next = *(void **)root;
7cfef17e
NC
553 Safefree(root);
554 root = next;
555 }
556}
557
645c22ef
DM
558/*
559=for apidoc sv_free_arenas
560
561Deallocate the memory used by all arenas. Note that all the individual SV
562heads and bodies within the arenas must already have been freed.
563
564=cut
565*/
7cfef17e
NC
566#define free_arena(name) \
567 STMT_START { \
568 S_free_arena(aTHX_ (void**) PL_ ## name ## _arenaroot); \
569 PL_ ## name ## _arenaroot = 0; \
570 PL_ ## name ## _root = 0; \
571 } STMT_END
572
4633a7c4 573void
864dbfa3 574Perl_sv_free_arenas(pTHX)
4633a7c4
LW
575{
576 SV* sva;
577 SV* svanext;
93e68bfb 578 int i;
4633a7c4
LW
579
580 /* Free arenas here, but be careful about fake ones. (We assume
581 contiguity of the fake ones with the corresponding real ones.) */
582
3280af22 583 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
4633a7c4
LW
584 svanext = (SV*) SvANY(sva);
585 while (svanext && SvFAKE(svanext))
586 svanext = (SV*) SvANY(svanext);
587
588 if (!SvFAKE(sva))
1df70142 589 Safefree(sva);
4633a7c4 590 }
93e68bfb
JC
591
592 for (i=0; i<SVt_LAST; i++) {
593 S_free_arena(aTHX_ (void**) PL_body_arenaroots[i]);
594 PL_body_arenaroots[i] = 0;
595 PL_body_roots[i] = 0;
596 }
597
43c5f42d 598 Safefree(PL_nice_chunk);
3280af22
NIS
599 PL_nice_chunk = Nullch;
600 PL_nice_chunk_size = 0;
601 PL_sv_arenaroot = 0;
602 PL_sv_root = 0;
4633a7c4
LW
603}
604
bd81e77b
NC
605/*
606 Here are mid-level routines that manage the allocation of bodies out
607 of the various arenas. There are 5 kinds of arenas:
29489e7c 608
bd81e77b
NC
609 1. SV-head arenas, which are discussed and handled above
610 2. regular body arenas
611 3. arenas for reduced-size bodies
612 4. Hash-Entry arenas
613 5. pte arenas (thread related)
29489e7c 614
bd81e77b
NC
615 Arena types 2 & 3 are chained by body-type off an array of
616 arena-root pointers, which is indexed by svtype. Some of the
617 larger/less used body types are malloced singly, since a large
618 unused block of them is wasteful. Also, several svtypes dont have
619 bodies; the data fits into the sv-head itself. The arena-root
620 pointer thus has a few unused root-pointers (which may be hijacked
621 later for arena types 4,5)
29489e7c 622
bd81e77b
NC
623 3 differs from 2 as an optimization; some body types have several
624 unused fields in the front of the structure (which are kept in-place
625 for consistency). These bodies can be allocated in smaller chunks,
626 because the leading fields arent accessed. Pointers to such bodies
627 are decremented to point at the unused 'ghost' memory, knowing that
628 the pointers are used with offsets to the real memory.
29489e7c 629
bd81e77b
NC
630 HE, HEK arenas are managed separately, with separate code, but may
631 be merge-able later..
632
633 PTE arenas are not sv-bodies, but they share these mid-level
634 mechanics, so are considered here. The new mid-level mechanics rely
635 on the sv_type of the body being allocated, so we just reserve one
636 of the unused body-slots for PTEs, then use it in those (2) PTE
637 contexts below (line ~10k)
638*/
639
640STATIC void *
641S_more_bodies (pTHX_ size_t size, svtype sv_type)
29489e7c 642{
00b6aa41
AL
643 void ** const arena_root = &PL_body_arenaroots[sv_type];
644 void ** const root = &PL_body_roots[sv_type];
bd81e77b
NC
645 char *start;
646 const char *end;
647 const size_t count = PERL_ARENA_SIZE / size;
29489e7c 648
bd81e77b
NC
649 Newx(start, count*size, char);
650 *((void **) start) = *arena_root;
651 *arena_root = (void *)start;
29489e7c 652
bd81e77b 653 end = start + (count-1) * size;
29489e7c 654
bd81e77b
NC
655 /* The initial slot is used to link the arenas together, so it isn't to be
656 linked into the list of ready-to-use bodies. */
29489e7c 657
bd81e77b 658 start += size;
29489e7c 659
bd81e77b 660 *root = (void *)start;
29489e7c 661
bd81e77b
NC
662 while (start < end) {
663 char * const next = start + size;
664 *(void**) start = (void *)next;
665 start = next;
29489e7c 666 }
bd81e77b
NC
667 *(void **)start = 0;
668
669 return *root;
29489e7c
DM
670}
671
bd81e77b 672/* grab a new thing from the free list, allocating more if necessary */
29489e7c 673
bd81e77b 674/* 1st, the inline version */
29489e7c 675
bd81e77b
NC
676#define new_body_inline(xpv, size, sv_type) \
677 STMT_START { \
00b6aa41 678 void ** const r3wt = &PL_body_roots[sv_type]; \
bd81e77b
NC
679 LOCK_SV_MUTEX; \
680 xpv = *((void **)(r3wt)) \
681 ? *((void **)(r3wt)) : S_more_bodies(aTHX_ size, sv_type); \
682 *(r3wt) = *(void**)(xpv); \
683 UNLOCK_SV_MUTEX; \
684 } STMT_END
29489e7c 685
bd81e77b 686/* now use the inline version in the proper function */
29489e7c 687
bd81e77b 688#ifndef PURIFY
9393da09 689
bd81e77b
NC
690/* This isn't being used with -DPURIFY, so don't declare it. Otherwise
691 compilers issue warnings. */
9393da09 692
bd81e77b
NC
693STATIC void *
694S_new_body(pTHX_ size_t size, svtype sv_type)
695{
696 void *xpv;
697 new_body_inline(xpv, size, sv_type);
698 return xpv;
699}
9393da09 700
bd81e77b 701#endif
53c1dcc0 702
bd81e77b 703/* return a thing to the free list */
29489e7c 704
bd81e77b
NC
705#define del_body(thing, root) \
706 STMT_START { \
00b6aa41 707 void ** const thing_copy = (void **)thing;\
bd81e77b
NC
708 LOCK_SV_MUTEX; \
709 *thing_copy = *root; \
710 *root = (void*)thing_copy; \
711 UNLOCK_SV_MUTEX; \
712 } STMT_END
29489e7c 713
bd81e77b
NC
714/*
715 Revisiting type 3 arenas, there are 4 body-types which have some
716 members that are never accessed. They are XPV, XPVIV, XPVAV,
717 XPVHV, which have corresponding types: xpv_allocated,
718 xpviv_allocated, xpvav_allocated, xpvhv_allocated,
29489e7c 719
bd81e77b
NC
720 For these types, the arenas are carved up into *_allocated size
721 chunks, we thus avoid wasted memory for those unaccessed members.
722 When bodies are allocated, we adjust the pointer back in memory by
723 the size of the bit not allocated, so it's as if we allocated the
724 full structure. (But things will all go boom if you write to the
725 part that is "not there", because you'll be overwriting the last
726 members of the preceding structure in memory.)
29489e7c 727
bd81e77b
NC
728 We calculate the correction using the STRUCT_OFFSET macro. For example, if
729 xpv_allocated is the same structure as XPV then the two OFFSETs sum to zero,
730 and the pointer is unchanged. If the allocated structure is smaller (no
731 initial NV actually allocated) then the net effect is to subtract the size
732 of the NV from the pointer, to return a new pointer as if an initial NV were
733 actually allocated.
29489e7c 734
bd81e77b
NC
735 This is the same trick as was used for NV and IV bodies. Ironically it
736 doesn't need to be used for NV bodies any more, because NV is now at the
737 start of the structure. IV bodies don't need it either, because they are
738 no longer allocated. */
29489e7c 739
bd81e77b
NC
740/* The following 2 arrays hide the above details in a pair of
741 lookup-tables, allowing us to be body-type agnostic.
29489e7c 742
bd81e77b
NC
743 size maps svtype to its body's allocated size.
744 offset maps svtype to the body-pointer adjustment needed
29489e7c 745
bd81e77b
NC
746 NB: elements in latter are 0 or <0, and are added during
747 allocation, and subtracted during deallocation. It may be clearer
748 to invert the values, and call it shrinkage_by_svtype.
29489e7c
DM
749*/
750
bd81e77b
NC
751struct body_details {
752 size_t size; /* Size to allocate */
753 size_t copy; /* Size of structure to copy (may be shorter) */
754 size_t offset;
755 bool cant_upgrade; /* Can upgrade this type */
756 bool zero_nv; /* zero the NV when upgrading from this */
757 bool arena; /* Allocated from an arena */
758};
29489e7c 759
bd81e77b
NC
760#define HADNV FALSE
761#define NONV TRUE
29489e7c 762
bd81e77b
NC
763#ifdef PURIFY
764/* With -DPURFIY we allocate everything directly, and don't use arenas.
765 This seems a rather elegant way to simplify some of the code below. */
766#define HASARENA FALSE
767#else
768#define HASARENA TRUE
769#endif
770#define NOARENA FALSE
29489e7c 771
bd81e77b 772/* A macro to work out the offset needed to subtract from a pointer to (say)
29489e7c 773
bd81e77b
NC
774typedef struct {
775 STRLEN xpv_cur;
776 STRLEN xpv_len;
777} xpv_allocated;
29489e7c 778
bd81e77b 779to make its members accessible via a pointer to (say)
29489e7c 780
bd81e77b
NC
781struct xpv {
782 NV xnv_nv;
783 STRLEN xpv_cur;
784 STRLEN xpv_len;
785};
29489e7c 786
bd81e77b 787*/
29489e7c 788
bd81e77b
NC
789#define relative_STRUCT_OFFSET(longer, shorter, member) \
790 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
29489e7c 791
bd81e77b
NC
792/* Calculate the length to copy. Specifically work out the length less any
793 final padding the compiler needed to add. See the comment in sv_upgrade
794 for why copying the padding proved to be a bug. */
29489e7c 795
bd81e77b
NC
796#define copy_length(type, last_member) \
797 STRUCT_OFFSET(type, last_member) \
798 + sizeof (((type*)SvANY((SV*)0))->last_member)
29489e7c 799
bd81e77b
NC
800static const struct body_details bodies_by_type[] = {
801 {0, 0, 0, FALSE, NONV, NOARENA},
802 /* IVs are in the head, so the allocation size is 0 */
803 {0, sizeof(IV), STRUCT_OFFSET(XPVIV, xiv_iv), FALSE, NONV, NOARENA},
804 /* 8 bytes on most ILP32 with IEEE doubles */
805 {sizeof(NV), sizeof(NV), 0, FALSE, HADNV, HASARENA},
806 /* RVs are in the head now */
807 /* However, this slot is overloaded and used by the pte */
808 {0, 0, 0, FALSE, NONV, NOARENA},
809 /* 8 bytes on most ILP32 with IEEE doubles */
810 {sizeof(xpv_allocated),
811 copy_length(XPV, xpv_len)
812 + relative_STRUCT_OFFSET(XPV, xpv_allocated, xpv_cur),
813 - relative_STRUCT_OFFSET(XPV, xpv_allocated, xpv_cur),
814 FALSE, NONV, HASARENA},
815 /* 12 */
816 {sizeof(xpviv_allocated),
817 copy_length(XPVIV, xiv_u)
818 + relative_STRUCT_OFFSET(XPVIV, xpviv_allocated, xpv_cur),
819 - relative_STRUCT_OFFSET(XPVIV, xpviv_allocated, xpv_cur),
820 FALSE, NONV, HASARENA},
821 /* 20 */
822 {sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, FALSE, HADNV, HASARENA},
823 /* 28 */
824 {sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, FALSE, HADNV, HASARENA},
825 /* 36 */
826 {sizeof(XPVBM), sizeof(XPVBM), 0, TRUE, HADNV, HASARENA},
827 /* 48 */
828 {sizeof(XPVGV), sizeof(XPVGV), 0, TRUE, HADNV, HASARENA},
829 /* 64 */
830 {sizeof(XPVLV), sizeof(XPVLV), 0, TRUE, HADNV, HASARENA},
831 /* 20 */
832 {sizeof(xpvav_allocated),
833 copy_length(XPVAV, xmg_stash)
834 + relative_STRUCT_OFFSET(XPVAV, xpvav_allocated, xav_fill),
835 - relative_STRUCT_OFFSET(XPVAV, xpvav_allocated, xav_fill),
836 TRUE, HADNV, HASARENA},
837 /* 20 */
838 {sizeof(xpvhv_allocated),
839 copy_length(XPVHV, xmg_stash)
840 + relative_STRUCT_OFFSET(XPVHV, xpvhv_allocated, xhv_fill),
841 - relative_STRUCT_OFFSET(XPVHV, xpvhv_allocated, xhv_fill),
842 TRUE, HADNV, HASARENA},
843 /* 76 */
844 {sizeof(XPVCV), sizeof(XPVCV), 0, TRUE, HADNV, HASARENA},
845 /* 80 */
846 {sizeof(XPVFM), sizeof(XPVFM), 0, TRUE, HADNV, NOARENA},
847 /* 84 */
848 {sizeof(XPVIO), sizeof(XPVIO), 0, TRUE, HADNV, NOARENA}
849};
29489e7c 850
bd81e77b
NC
851#define new_body_type(sv_type) \
852 (void *)((char *)S_new_body(aTHX_ bodies_by_type[sv_type].size, sv_type)\
853 - bodies_by_type[sv_type].offset)
29489e7c 854
bd81e77b
NC
855#define del_body_type(p, sv_type) \
856 del_body(p, &PL_body_roots[sv_type])
29489e7c 857
29489e7c 858
bd81e77b
NC
859#define new_body_allocated(sv_type) \
860 (void *)((char *)S_new_body(aTHX_ bodies_by_type[sv_type].size, sv_type)\
861 - bodies_by_type[sv_type].offset)
29489e7c 862
bd81e77b
NC
863#define del_body_allocated(p, sv_type) \
864 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
29489e7c 865
29489e7c 866
bd81e77b
NC
867#define my_safemalloc(s) (void*)safemalloc(s)
868#define my_safecalloc(s) (void*)safecalloc(s, 1)
869#define my_safefree(p) safefree((char*)p)
29489e7c 870
bd81e77b 871#ifdef PURIFY
29489e7c 872
bd81e77b
NC
873#define new_XNV() my_safemalloc(sizeof(XPVNV))
874#define del_XNV(p) my_safefree(p)
29489e7c 875
bd81e77b
NC
876#define new_XPVNV() my_safemalloc(sizeof(XPVNV))
877#define del_XPVNV(p) my_safefree(p)
29489e7c 878
bd81e77b
NC
879#define new_XPVAV() my_safemalloc(sizeof(XPVAV))
880#define del_XPVAV(p) my_safefree(p)
29489e7c 881
bd81e77b
NC
882#define new_XPVHV() my_safemalloc(sizeof(XPVHV))
883#define del_XPVHV(p) my_safefree(p)
29489e7c 884
bd81e77b
NC
885#define new_XPVMG() my_safemalloc(sizeof(XPVMG))
886#define del_XPVMG(p) my_safefree(p)
29489e7c 887
bd81e77b
NC
888#define new_XPVGV() my_safemalloc(sizeof(XPVGV))
889#define del_XPVGV(p) my_safefree(p)
29489e7c 890
bd81e77b 891#else /* !PURIFY */
29489e7c 892
bd81e77b
NC
893#define new_XNV() new_body_type(SVt_NV)
894#define del_XNV(p) del_body_type(p, SVt_NV)
29489e7c 895
bd81e77b
NC
896#define new_XPVNV() new_body_type(SVt_PVNV)
897#define del_XPVNV(p) del_body_type(p, SVt_PVNV)
29489e7c 898
bd81e77b
NC
899#define new_XPVAV() new_body_allocated(SVt_PVAV)
900#define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
645c22ef 901
bd81e77b
NC
902#define new_XPVHV() new_body_allocated(SVt_PVHV)
903#define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
645c22ef 904
bd81e77b
NC
905#define new_XPVMG() new_body_type(SVt_PVMG)
906#define del_XPVMG(p) del_body_type(p, SVt_PVMG)
645c22ef 907
bd81e77b
NC
908#define new_XPVGV() new_body_type(SVt_PVGV)
909#define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1d7c1841 910
bd81e77b 911#endif /* PURIFY */
93e68bfb 912
bd81e77b 913/* no arena for you! */
93e68bfb 914
bd81e77b
NC
915#define new_NOARENA(details) \
916 my_safemalloc((details)->size + (details)->offset)
917#define new_NOARENAZ(details) \
918 my_safecalloc((details)->size + (details)->offset)
93e68bfb 919
bd81e77b
NC
920/*
921=for apidoc sv_upgrade
93e68bfb 922
bd81e77b
NC
923Upgrade an SV to a more complex form. Generally adds a new body type to the
924SV, then copies across as much information as possible from the old body.
925You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
93e68bfb 926
bd81e77b 927=cut
93e68bfb 928*/
93e68bfb 929
bd81e77b
NC
930void
931Perl_sv_upgrade(pTHX_ register SV *sv, U32 new_type)
cac9b346 932{
bd81e77b
NC
933 void* old_body;
934 void* new_body;
935 const U32 old_type = SvTYPE(sv);
936 const struct body_details *const old_type_details
937 = bodies_by_type + old_type;
938 const struct body_details *new_type_details = bodies_by_type + new_type;
cac9b346 939
bd81e77b
NC
940 if (new_type != SVt_PV && SvIsCOW(sv)) {
941 sv_force_normal_flags(sv, 0);
942 }
cac9b346 943
bd81e77b
NC
944 if (old_type == new_type)
945 return;
cac9b346 946
bd81e77b
NC
947 if (old_type > new_type)
948 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
949 (int)old_type, (int)new_type);
cac9b346 950
cac9b346 951
bd81e77b 952 old_body = SvANY(sv);
de042e1d 953
bd81e77b
NC
954 /* Copying structures onto other structures that have been neatly zeroed
955 has a subtle gotcha. Consider XPVMG
cac9b346 956
bd81e77b
NC
957 +------+------+------+------+------+-------+-------+
958 | NV | CUR | LEN | IV | MAGIC | STASH |
959 +------+------+------+------+------+-------+-------+
960 0 4 8 12 16 20 24 28
645c22ef 961
bd81e77b
NC
962 where NVs are aligned to 8 bytes, so that sizeof that structure is
963 actually 32 bytes long, with 4 bytes of padding at the end:
08742458 964
bd81e77b
NC
965 +------+------+------+------+------+-------+-------+------+
966 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
967 +------+------+------+------+------+-------+-------+------+
968 0 4 8 12 16 20 24 28 32
08742458 969
bd81e77b 970 so what happens if you allocate memory for this structure:
30f9da9e 971
bd81e77b
NC
972 +------+------+------+------+------+-------+-------+------+------+...
973 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
974 +------+------+------+------+------+-------+-------+------+------+...
975 0 4 8 12 16 20 24 28 32 36
bfc44f79 976
bd81e77b
NC
977 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
978 expect, because you copy the area marked ??? onto GP. Now, ??? may have
979 started out as zero once, but it's quite possible that it isn't. So now,
980 rather than a nicely zeroed GP, you have it pointing somewhere random.
981 Bugs ensue.
bfc44f79 982
bd81e77b
NC
983 (In fact, GP ends up pointing at a previous GP structure, because the
984 principle cause of the padding in XPVMG getting garbage is a copy of
985 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob)
30f9da9e 986
bd81e77b
NC
987 So we are careful and work out the size of used parts of all the
988 structures. */
bfc44f79 989
bd81e77b
NC
990 switch (old_type) {
991 case SVt_NULL:
992 break;
993 case SVt_IV:
994 if (new_type < SVt_PVIV) {
995 new_type = (new_type == SVt_NV)
996 ? SVt_PVNV : SVt_PVIV;
997 new_type_details = bodies_by_type + new_type;
998 }
999 break;
1000 case SVt_NV:
1001 if (new_type < SVt_PVNV) {
1002 new_type = SVt_PVNV;
1003 new_type_details = bodies_by_type + new_type;
1004 }
1005 break;
1006 case SVt_RV:
1007 break;
1008 case SVt_PV:
1009 assert(new_type > SVt_PV);
1010 assert(SVt_IV < SVt_PV);
1011 assert(SVt_NV < SVt_PV);
1012 break;
1013 case SVt_PVIV:
1014 break;
1015 case SVt_PVNV:
1016 break;
1017 case SVt_PVMG:
1018 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1019 there's no way that it can be safely upgraded, because perl.c
1020 expects to Safefree(SvANY(PL_mess_sv)) */
1021 assert(sv != PL_mess_sv);
1022 /* This flag bit is used to mean other things in other scalar types.
1023 Given that it only has meaning inside the pad, it shouldn't be set
1024 on anything that can get upgraded. */
1025 assert((SvFLAGS(sv) & SVpad_TYPED) == 0);
1026 break;
1027 default:
1028 if (old_type_details->cant_upgrade)
1029 Perl_croak(aTHX_ "Can't upgrade that kind of scalar");
1030 }
645c22ef 1031
bd81e77b
NC
1032 SvFLAGS(sv) &= ~SVTYPEMASK;
1033 SvFLAGS(sv) |= new_type;
932e9ff9 1034
bd81e77b
NC
1035 switch (new_type) {
1036 case SVt_NULL:
1037 Perl_croak(aTHX_ "Can't upgrade to undef");
1038 case SVt_IV:
1039 assert(old_type == SVt_NULL);
1040 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1041 SvIV_set(sv, 0);
1042 return;
1043 case SVt_NV:
1044 assert(old_type == SVt_NULL);
1045 SvANY(sv) = new_XNV();
1046 SvNV_set(sv, 0);
1047 return;
1048 case SVt_RV:
1049 assert(old_type == SVt_NULL);
1050 SvANY(sv) = &sv->sv_u.svu_rv;
1051 SvRV_set(sv, 0);
1052 return;
1053 case SVt_PVHV:
1054 SvANY(sv) = new_XPVHV();
1055 HvFILL(sv) = 0;
1056 HvMAX(sv) = 0;
1057 HvTOTALKEYS(sv) = 0;
645c22ef 1058
bd81e77b 1059 goto hv_av_common;
aeb18a1e 1060
bd81e77b
NC
1061 case SVt_PVAV:
1062 SvANY(sv) = new_XPVAV();
1063 AvMAX(sv) = -1;
1064 AvFILLp(sv) = -1;
1065 AvALLOC(sv) = 0;
1066 AvREAL_only(sv);
aeb18a1e 1067
bd81e77b
NC
1068 hv_av_common:
1069 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1070 The target created by newSVrv also is, and it can have magic.
1071 However, it never has SvPVX set.
1072 */
1073 if (old_type >= SVt_RV) {
1074 assert(SvPVX_const(sv) == 0);
1075 }
aeb18a1e 1076
bd81e77b
NC
1077 /* Could put this in the else clause below, as PVMG must have SvPVX
1078 0 already (the assertion above) */
1079 SvPV_set(sv, (char*)0);
93e68bfb 1080
bd81e77b
NC
1081 if (old_type >= SVt_PVMG) {
1082 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_magic);
1083 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1084 } else {
1085 SvMAGIC_set(sv, 0);
1086 SvSTASH_set(sv, 0);
1087 }
1088 break;
93e68bfb 1089
93e68bfb 1090
bd81e77b
NC
1091 case SVt_PVIV:
1092 /* XXX Is this still needed? Was it ever needed? Surely as there is
1093 no route from NV to PVIV, NOK can never be true */
1094 assert(!SvNOKp(sv));
1095 assert(!SvNOK(sv));
1096 case SVt_PVIO:
1097 case SVt_PVFM:
1098 case SVt_PVBM:
1099 case SVt_PVGV:
1100 case SVt_PVCV:
1101 case SVt_PVLV:
1102 case SVt_PVMG:
1103 case SVt_PVNV:
1104 case SVt_PV:
93e68bfb 1105
bd81e77b
NC
1106 assert(new_type_details->size);
1107 /* We always allocated the full length item with PURIFY. To do this
1108 we fake things so that arena is false for all 16 types.. */
1109 if(new_type_details->arena) {
1110 /* This points to the start of the allocated area. */
1111 new_body_inline(new_body, new_type_details->size, new_type);
1112 Zero(new_body, new_type_details->size, char);
1113 new_body = ((char *)new_body) - new_type_details->offset;
1114 } else {
1115 new_body = new_NOARENAZ(new_type_details);
1116 }
1117 SvANY(sv) = new_body;
5e2fc214 1118
bd81e77b
NC
1119 if (old_type_details->copy) {
1120 Copy((char *)old_body + old_type_details->offset,
1121 (char *)new_body + old_type_details->offset,
1122 old_type_details->copy, char);
1123 }
1124
1125#ifndef NV_ZERO_IS_ALLBITS_ZERO
1126 /* If NV 0.0 is store as all bits 0 then Zero() already creates a correct
1127 0.0 for us. */
1128 if (old_type_details->zero_nv)
1129 SvNV_set(sv, 0);
82048762 1130#endif
5e2fc214 1131
bd81e77b
NC
1132 if (new_type == SVt_PVIO)
1133 IoPAGE_LEN(sv) = 60;
1134 if (old_type < SVt_RV)
1135 SvPV_set(sv, 0);
1136 break;
1137 default:
1138 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu", new_type);
1139 }
73171d91 1140
bd81e77b
NC
1141 if (old_type_details->size) {
1142 /* If the old body had an allocated size, then we need to free it. */
1143#ifdef PURIFY
1144 my_safefree(old_body);
1145#else
1146 del_body((void*)((char*)old_body + old_type_details->offset),
1147 &PL_body_roots[old_type]);
1148#endif
1149 }
1150}
73171d91 1151
bd81e77b
NC
1152/*
1153=for apidoc sv_backoff
73171d91 1154
bd81e77b
NC
1155Remove any string offset. You should normally use the C<SvOOK_off> macro
1156wrapper instead.
73171d91 1157
bd81e77b 1158=cut
73171d91
NC
1159*/
1160
bd81e77b
NC
1161int
1162Perl_sv_backoff(pTHX_ register SV *sv)
1163{
1164 assert(SvOOK(sv));
1165 assert(SvTYPE(sv) != SVt_PVHV);
1166 assert(SvTYPE(sv) != SVt_PVAV);
1167 if (SvIVX(sv)) {
1168 const char * const s = SvPVX_const(sv);
1169 SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
1170 SvPV_set(sv, SvPVX(sv) - SvIVX(sv));
1171 SvIV_set(sv, 0);
1172 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1173 }
1174 SvFLAGS(sv) &= ~SVf_OOK;
1175 return 0;
1176}
73171d91 1177
bd81e77b
NC
1178/*
1179=for apidoc sv_grow
73171d91 1180
bd81e77b
NC
1181Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1182upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1183Use the C<SvGROW> wrapper instead.
93e68bfb 1184
bd81e77b
NC
1185=cut
1186*/
93e68bfb 1187
bd81e77b
NC
1188char *
1189Perl_sv_grow(pTHX_ register SV *sv, register STRLEN newlen)
1190{
1191 register char *s;
93e68bfb 1192
bd81e77b
NC
1193#ifdef HAS_64K_LIMIT
1194 if (newlen >= 0x10000) {
1195 PerlIO_printf(Perl_debug_log,
1196 "Allocation too large: %"UVxf"\n", (UV)newlen);
1197 my_exit(1);
1198 }
1199#endif /* HAS_64K_LIMIT */
1200 if (SvROK(sv))
1201 sv_unref(sv);
1202 if (SvTYPE(sv) < SVt_PV) {
1203 sv_upgrade(sv, SVt_PV);
1204 s = SvPVX_mutable(sv);
1205 }
1206 else if (SvOOK(sv)) { /* pv is offset? */
1207 sv_backoff(sv);
1208 s = SvPVX_mutable(sv);
1209 if (newlen > SvLEN(sv))
1210 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1211#ifdef HAS_64K_LIMIT
1212 if (newlen >= 0x10000)
1213 newlen = 0xFFFF;
1214#endif
1215 }
1216 else
1217 s = SvPVX_mutable(sv);
aeb18a1e 1218
bd81e77b
NC
1219 if (newlen > SvLEN(sv)) { /* need more room? */
1220 newlen = PERL_STRLEN_ROUNDUP(newlen);
1221 if (SvLEN(sv) && s) {
1222#ifdef MYMALLOC
1223 const STRLEN l = malloced_size((void*)SvPVX_const(sv));
1224 if (newlen <= l) {
1225 SvLEN_set(sv, l);
1226 return s;
1227 } else
1228#endif
1229 s = saferealloc(s, newlen);
1230 }
1231 else {
1232 s = safemalloc(newlen);
1233 if (SvPVX_const(sv) && SvCUR(sv)) {
1234 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1235 }
1236 }
1237 SvPV_set(sv, s);
1238 SvLEN_set(sv, newlen);
1239 }
1240 return s;
1241}
aeb18a1e 1242
bd81e77b
NC
1243/*
1244=for apidoc sv_setiv
932e9ff9 1245
bd81e77b
NC
1246Copies an integer into the given SV, upgrading first if necessary.
1247Does not handle 'set' magic. See also C<sv_setiv_mg>.
463ee0b2 1248
bd81e77b
NC
1249=cut
1250*/
463ee0b2 1251
bd81e77b
NC
1252void
1253Perl_sv_setiv(pTHX_ register SV *sv, IV i)
1254{
1255 SV_CHECK_THINKFIRST_COW_DROP(sv);
1256 switch (SvTYPE(sv)) {
1257 case SVt_NULL:
1258 sv_upgrade(sv, SVt_IV);
1259 break;
1260 case SVt_NV:
1261 sv_upgrade(sv, SVt_PVNV);
1262 break;
1263 case SVt_RV:
1264 case SVt_PV:
1265 sv_upgrade(sv, SVt_PVIV);
1266 break;
463ee0b2 1267
bd81e77b
NC
1268 case SVt_PVGV:
1269 case SVt_PVAV:
1270 case SVt_PVHV:
1271 case SVt_PVCV:
1272 case SVt_PVFM:
1273 case SVt_PVIO:
1274 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1275 OP_DESC(PL_op));
1276 }
1277 (void)SvIOK_only(sv); /* validate number */
1278 SvIV_set(sv, i);
1279 SvTAINT(sv);
1280}
932e9ff9 1281
bd81e77b
NC
1282/*
1283=for apidoc sv_setiv_mg
d33b2eba 1284
bd81e77b 1285Like C<sv_setiv>, but also handles 'set' magic.
1c846c1f 1286
bd81e77b
NC
1287=cut
1288*/
d33b2eba 1289
bd81e77b
NC
1290void
1291Perl_sv_setiv_mg(pTHX_ register SV *sv, IV i)
1292{
1293 sv_setiv(sv,i);
1294 SvSETMAGIC(sv);
1295}
727879eb 1296
bd81e77b
NC
1297/*
1298=for apidoc sv_setuv
d33b2eba 1299
bd81e77b
NC
1300Copies an unsigned integer into the given SV, upgrading first if necessary.
1301Does not handle 'set' magic. See also C<sv_setuv_mg>.
9b94d1dd 1302
bd81e77b
NC
1303=cut
1304*/
d33b2eba 1305
bd81e77b
NC
1306void
1307Perl_sv_setuv(pTHX_ register SV *sv, UV u)
1308{
1309 /* With these two if statements:
1310 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
d33b2eba 1311
bd81e77b
NC
1312 without
1313 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1c846c1f 1314
bd81e77b
NC
1315 If you wish to remove them, please benchmark to see what the effect is
1316 */
1317 if (u <= (UV)IV_MAX) {
1318 sv_setiv(sv, (IV)u);
1319 return;
1320 }
1321 sv_setiv(sv, 0);
1322 SvIsUV_on(sv);
1323 SvUV_set(sv, u);
1324}
d33b2eba 1325
bd81e77b
NC
1326/*
1327=for apidoc sv_setuv_mg
727879eb 1328
bd81e77b 1329Like C<sv_setuv>, but also handles 'set' magic.
9b94d1dd 1330
bd81e77b
NC
1331=cut
1332*/
5e2fc214 1333
bd81e77b
NC
1334void
1335Perl_sv_setuv_mg(pTHX_ register SV *sv, UV u)
1336{
1337 sv_setiv(sv, 0);
1338 SvIsUV_on(sv);
1339 sv_setuv(sv,u);
1340 SvSETMAGIC(sv);
1341}
5e2fc214 1342
954c1994 1343/*
bd81e77b 1344=for apidoc sv_setnv
954c1994 1345
bd81e77b
NC
1346Copies a double into the given SV, upgrading first if necessary.
1347Does not handle 'set' magic. See also C<sv_setnv_mg>.
954c1994
GS
1348
1349=cut
1350*/
1351
63f97190 1352void
bd81e77b 1353Perl_sv_setnv(pTHX_ register SV *sv, NV num)
79072805 1354{
bd81e77b
NC
1355 SV_CHECK_THINKFIRST_COW_DROP(sv);
1356 switch (SvTYPE(sv)) {
79072805 1357 case SVt_NULL:
79072805 1358 case SVt_IV:
bd81e77b 1359 sv_upgrade(sv, SVt_NV);
79072805 1360 break;
ed6116ce 1361 case SVt_RV:
79072805 1362 case SVt_PV:
79072805 1363 case SVt_PVIV:
bd81e77b 1364 sv_upgrade(sv, SVt_PVNV);
79072805 1365 break;
bd4b1eb5 1366
bd4b1eb5 1367 case SVt_PVGV:
bd81e77b
NC
1368 case SVt_PVAV:
1369 case SVt_PVHV:
79072805 1370 case SVt_PVCV:
bd81e77b
NC
1371 case SVt_PVFM:
1372 case SVt_PVIO:
1373 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1374 OP_NAME(PL_op));
2068cd4d 1375 }
bd81e77b
NC
1376 SvNV_set(sv, num);
1377 (void)SvNOK_only(sv); /* validate number */
1378 SvTAINT(sv);
79072805
LW
1379}
1380
645c22ef 1381/*
bd81e77b 1382=for apidoc sv_setnv_mg
645c22ef 1383
bd81e77b 1384Like C<sv_setnv>, but also handles 'set' magic.
645c22ef
DM
1385
1386=cut
1387*/
1388
bd81e77b
NC
1389void
1390Perl_sv_setnv_mg(pTHX_ register SV *sv, NV num)
79072805 1391{
bd81e77b
NC
1392 sv_setnv(sv,num);
1393 SvSETMAGIC(sv);
79072805
LW
1394}
1395
bd81e77b
NC
1396/* Print an "isn't numeric" warning, using a cleaned-up,
1397 * printable version of the offending string
1398 */
954c1994 1399
bd81e77b
NC
1400STATIC void
1401S_not_a_number(pTHX_ SV *sv)
79072805 1402{
bd81e77b
NC
1403 SV *dsv;
1404 char tmpbuf[64];
1405 const char *pv;
94463019
JH
1406
1407 if (DO_UTF8(sv)) {
d0043bd1 1408 dsv = sv_2mortal(newSVpvn("", 0));
94463019
JH
1409 pv = sv_uni_display(dsv, sv, 10, 0);
1410 } else {
1411 char *d = tmpbuf;
551405c4 1412 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
94463019
JH
1413 /* each *s can expand to 4 chars + "...\0",
1414 i.e. need room for 8 chars */
ecdeb87c 1415
00b6aa41
AL
1416 const char *s = SvPVX_const(sv);
1417 const char * const end = s + SvCUR(sv);
1418 for ( ; s < end && d < limit; s++ ) {
94463019
JH
1419 int ch = *s & 0xFF;
1420 if (ch & 128 && !isPRINT_LC(ch)) {
1421 *d++ = 'M';
1422 *d++ = '-';
1423 ch &= 127;
1424 }
1425 if (ch == '\n') {
1426 *d++ = '\\';
1427 *d++ = 'n';
1428 }
1429 else if (ch == '\r') {
1430 *d++ = '\\';
1431 *d++ = 'r';
1432 }
1433 else if (ch == '\f') {
1434 *d++ = '\\';
1435 *d++ = 'f';
1436 }
1437 else if (ch == '\\') {
1438 *d++ = '\\';
1439 *d++ = '\\';
1440 }
1441 else if (ch == '\0') {
1442 *d++ = '\\';
1443 *d++ = '0';
1444 }
1445 else if (isPRINT_LC(ch))
1446 *d++ = ch;
1447 else {
1448 *d++ = '^';
1449 *d++ = toCTRL(ch);
1450 }
1451 }
1452 if (s < end) {
1453 *d++ = '.';
1454 *d++ = '.';
1455 *d++ = '.';
1456 }
1457 *d = '\0';
1458 pv = tmpbuf;
a0d0e21e 1459 }
a0d0e21e 1460
533c011a 1461 if (PL_op)
9014280d 1462 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019
JH
1463 "Argument \"%s\" isn't numeric in %s", pv,
1464 OP_DESC(PL_op));
a0d0e21e 1465 else
9014280d 1466 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
94463019 1467 "Argument \"%s\" isn't numeric", pv);
a0d0e21e
LW
1468}
1469
c2988b20
NC
1470/*
1471=for apidoc looks_like_number
1472
645c22ef
DM
1473Test if the content of an SV looks like a number (or is a number).
1474C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1475non-numeric warning), even if your atof() doesn't grok them.
c2988b20
NC
1476
1477=cut
1478*/
1479
1480I32
1481Perl_looks_like_number(pTHX_ SV *sv)
1482{
a3b680e6 1483 register const char *sbegin;
c2988b20
NC
1484 STRLEN len;
1485
1486 if (SvPOK(sv)) {
3f7c398e 1487 sbegin = SvPVX_const(sv);
c2988b20
NC
1488 len = SvCUR(sv);
1489 }
1490 else if (SvPOKp(sv))
83003860 1491 sbegin = SvPV_const(sv, len);
c2988b20 1492 else
e0ab1c0e 1493 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
c2988b20
NC
1494 return grok_number(sbegin, len, NULL);
1495}
25da4f38
IZ
1496
1497/* Actually, ISO C leaves conversion of UV to IV undefined, but
1498 until proven guilty, assume that things are not that bad... */
1499
645c22ef
DM
1500/*
1501 NV_PRESERVES_UV:
1502
1503 As 64 bit platforms often have an NV that doesn't preserve all bits of
28e5dec8
JH
1504 an IV (an assumption perl has been based on to date) it becomes necessary
1505 to remove the assumption that the NV always carries enough precision to
1506 recreate the IV whenever needed, and that the NV is the canonical form.
1507 Instead, IV/UV and NV need to be given equal rights. So as to not lose
645c22ef 1508 precision as a side effect of conversion (which would lead to insanity
28e5dec8
JH
1509 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1510 1) to distinguish between IV/UV/NV slots that have cached a valid
1511 conversion where precision was lost and IV/UV/NV slots that have a
1512 valid conversion which has lost no precision
645c22ef 1513 2) to ensure that if a numeric conversion to one form is requested that
28e5dec8
JH
1514 would lose precision, the precise conversion (or differently
1515 imprecise conversion) is also performed and cached, to prevent
1516 requests for different numeric formats on the same SV causing
1517 lossy conversion chains. (lossless conversion chains are perfectly
1518 acceptable (still))
1519
1520
1521 flags are used:
1522 SvIOKp is true if the IV slot contains a valid value
1523 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1524 SvNOKp is true if the NV slot contains a valid value
1525 SvNOK is true only if the NV value is accurate
1526
1527 so
645c22ef 1528 while converting from PV to NV, check to see if converting that NV to an
28e5dec8
JH
1529 IV(or UV) would lose accuracy over a direct conversion from PV to
1530 IV(or UV). If it would, cache both conversions, return NV, but mark
1531 SV as IOK NOKp (ie not NOK).
1532
645c22ef 1533 While converting from PV to IV, check to see if converting that IV to an
28e5dec8
JH
1534 NV would lose accuracy over a direct conversion from PV to NV. If it
1535 would, cache both conversions, flag similarly.
1536
1537 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1538 correctly because if IV & NV were set NV *always* overruled.
645c22ef
DM
1539 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1540 changes - now IV and NV together means that the two are interchangeable:
28e5dec8 1541 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
d460ef45 1542
645c22ef
DM
1543 The benefit of this is that operations such as pp_add know that if
1544 SvIOK is true for both left and right operands, then integer addition
1545 can be used instead of floating point (for cases where the result won't
1546 overflow). Before, floating point was always used, which could lead to
28e5dec8
JH
1547 loss of precision compared with integer addition.
1548
1549 * making IV and NV equal status should make maths accurate on 64 bit
1550 platforms
1551 * may speed up maths somewhat if pp_add and friends start to use
645c22ef 1552 integers when possible instead of fp. (Hopefully the overhead in
28e5dec8
JH
1553 looking for SvIOK and checking for overflow will not outweigh the
1554 fp to integer speedup)
1555 * will slow down integer operations (callers of SvIV) on "inaccurate"
1556 values, as the change from SvIOK to SvIOKp will cause a call into
1557 sv_2iv each time rather than a macro access direct to the IV slot
1558 * should speed up number->string conversion on integers as IV is
645c22ef 1559 favoured when IV and NV are equally accurate
28e5dec8
JH
1560
1561 ####################################################################
645c22ef
DM
1562 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1563 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1564 On the other hand, SvUOK is true iff UV.
28e5dec8
JH
1565 ####################################################################
1566
645c22ef 1567 Your mileage will vary depending your CPU's relative fp to integer
28e5dec8
JH
1568 performance ratio.
1569*/
1570
1571#ifndef NV_PRESERVES_UV
645c22ef
DM
1572# define IS_NUMBER_UNDERFLOW_IV 1
1573# define IS_NUMBER_UNDERFLOW_UV 2
1574# define IS_NUMBER_IV_AND_UV 2
1575# define IS_NUMBER_OVERFLOW_IV 4
1576# define IS_NUMBER_OVERFLOW_UV 5
1577
1578/* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
28e5dec8
JH
1579
1580/* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1581STATIC int
645c22ef 1582S_sv_2iuv_non_preserve(pTHX_ register SV *sv, I32 numtype)
28e5dec8 1583{
3f7c398e 1584 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
28e5dec8
JH
1585 if (SvNVX(sv) < (NV)IV_MIN) {
1586 (void)SvIOKp_on(sv);
1587 (void)SvNOK_on(sv);
45977657 1588 SvIV_set(sv, IV_MIN);
28e5dec8
JH
1589 return IS_NUMBER_UNDERFLOW_IV;
1590 }
1591 if (SvNVX(sv) > (NV)UV_MAX) {
1592 (void)SvIOKp_on(sv);
1593 (void)SvNOK_on(sv);
1594 SvIsUV_on(sv);
607fa7f2 1595 SvUV_set(sv, UV_MAX);
28e5dec8
JH
1596 return IS_NUMBER_OVERFLOW_UV;
1597 }
c2988b20
NC
1598 (void)SvIOKp_on(sv);
1599 (void)SvNOK_on(sv);
1600 /* Can't use strtol etc to convert this string. (See truth table in
1601 sv_2iv */
1602 if (SvNVX(sv) <= (UV)IV_MAX) {
45977657 1603 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1604 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1605 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1606 } else {
1607 /* Integer is imprecise. NOK, IOKp */
1608 }
1609 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1610 }
1611 SvIsUV_on(sv);
607fa7f2 1612 SvUV_set(sv, U_V(SvNVX(sv)));
c2988b20
NC
1613 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1614 if (SvUVX(sv) == UV_MAX) {
1615 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1616 possibly be preserved by NV. Hence, it must be overflow.
1617 NOK, IOKp */
1618 return IS_NUMBER_OVERFLOW_UV;
1619 }
1620 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1621 } else {
1622 /* Integer is imprecise. NOK, IOKp */
28e5dec8 1623 }
c2988b20 1624 return IS_NUMBER_OVERFLOW_IV;
28e5dec8 1625}
645c22ef
DM
1626#endif /* !NV_PRESERVES_UV*/
1627
af359546
NC
1628STATIC bool
1629S_sv_2iuv_common(pTHX_ SV *sv) {
1630 if (SvNOKp(sv)) {
28e5dec8
JH
1631 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1632 * without also getting a cached IV/UV from it at the same time
1633 * (ie PV->NV conversion should detect loss of accuracy and cache
af359546
NC
1634 * IV or UV at same time to avoid this. */
1635 /* IV-over-UV optimisation - choose to cache IV if possible */
25da4f38
IZ
1636
1637 if (SvTYPE(sv) == SVt_NV)
1638 sv_upgrade(sv, SVt_PVNV);
1639
28e5dec8
JH
1640 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1641 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1642 certainly cast into the IV range at IV_MAX, whereas the correct
1643 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1644 cases go to UV */
1645 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
45977657 1646 SvIV_set(sv, I_V(SvNVX(sv)));
28e5dec8
JH
1647 if (SvNVX(sv) == (NV) SvIVX(sv)
1648#ifndef NV_PRESERVES_UV
1649 && (((UV)1 << NV_PRESERVES_UV_BITS) >
1650 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
1651 /* Don't flag it as "accurately an integer" if the number
1652 came from a (by definition imprecise) NV operation, and
1653 we're outside the range of NV integer precision */
1654#endif
1655 ) {
1656 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
1657 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1658 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
28e5dec8
JH
1659 PTR2UV(sv),
1660 SvNVX(sv),
1661 SvIVX(sv)));
1662
1663 } else {
1664 /* IV not precise. No need to convert from PV, as NV
1665 conversion would already have cached IV if it detected
1666 that PV->IV would be better than PV->NV->IV
1667 flags already correct - don't set public IOK. */
1668 DEBUG_c(PerlIO_printf(Perl_debug_log,
7234c960 1669 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
28e5dec8
JH
1670 PTR2UV(sv),
1671 SvNVX(sv),
1672 SvIVX(sv)));
1673 }
1674 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
1675 but the cast (NV)IV_MIN rounds to a the value less (more
1676 negative) than IV_MIN which happens to be equal to SvNVX ??
1677 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
1678 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
1679 (NV)UVX == NVX are both true, but the values differ. :-(
1680 Hopefully for 2s complement IV_MIN is something like
1681 0x8000000000000000 which will be exact. NWC */
d460ef45 1682 }
25da4f38 1683 else {
607fa7f2 1684 SvUV_set(sv, U_V(SvNVX(sv)));
28e5dec8
JH
1685 if (
1686 (SvNVX(sv) == (NV) SvUVX(sv))
1687#ifndef NV_PRESERVES_UV
1688 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
1689 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
1690 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
1691 /* Don't flag it as "accurately an integer" if the number
1692 came from a (by definition imprecise) NV operation, and
1693 we're outside the range of NV integer precision */
1694#endif
1695 )
1696 SvIOK_on(sv);
25da4f38 1697 SvIsUV_on(sv);
1c846c1f 1698 DEBUG_c(PerlIO_printf(Perl_debug_log,
57def98f 1699 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
56431972 1700 PTR2UV(sv),
57def98f
JH
1701 SvUVX(sv),
1702 SvUVX(sv)));
25da4f38 1703 }
748a9306
LW
1704 }
1705 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 1706 UV value;
504618e9 1707 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
af359546 1708 /* We want to avoid a possible problem when we cache an IV/ a UV which
25da4f38 1709 may be later translated to an NV, and the resulting NV is not
c2988b20
NC
1710 the same as the direct translation of the initial string
1711 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
1712 be careful to ensure that the value with the .456 is around if the
1713 NV value is requested in the future).
1c846c1f 1714
af359546 1715 This means that if we cache such an IV/a UV, we need to cache the
25da4f38 1716 NV as well. Moreover, we trade speed for space, and do not
28e5dec8 1717 cache the NV if we are sure it's not needed.
25da4f38 1718 */
16b7a9a4 1719
c2988b20
NC
1720 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
1721 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1722 == IS_NUMBER_IN_UV) {
5e045b90 1723 /* It's definitely an integer, only upgrade to PVIV */
28e5dec8
JH
1724 if (SvTYPE(sv) < SVt_PVIV)
1725 sv_upgrade(sv, SVt_PVIV);
f7bbb42a 1726 (void)SvIOK_on(sv);
c2988b20
NC
1727 } else if (SvTYPE(sv) < SVt_PVNV)
1728 sv_upgrade(sv, SVt_PVNV);
28e5dec8 1729
c2988b20
NC
1730 /* If NV preserves UV then we only use the UV value if we know that
1731 we aren't going to call atof() below. If NVs don't preserve UVs
1732 then the value returned may have more precision than atof() will
1733 return, even though value isn't perfectly accurate. */
1734 if ((numtype & (IS_NUMBER_IN_UV
1735#ifdef NV_PRESERVES_UV
1736 | IS_NUMBER_NOT_INT
1737#endif
1738 )) == IS_NUMBER_IN_UV) {
1739 /* This won't turn off the public IOK flag if it was set above */
1740 (void)SvIOKp_on(sv);
1741
1742 if (!(numtype & IS_NUMBER_NEG)) {
1743 /* positive */;
1744 if (value <= (UV)IV_MAX) {
45977657 1745 SvIV_set(sv, (IV)value);
c2988b20 1746 } else {
af359546 1747 /* it didn't overflow, and it was positive. */
607fa7f2 1748 SvUV_set(sv, value);
c2988b20
NC
1749 SvIsUV_on(sv);
1750 }
1751 } else {
1752 /* 2s complement assumption */
1753 if (value <= (UV)IV_MIN) {
45977657 1754 SvIV_set(sv, -(IV)value);
c2988b20
NC
1755 } else {
1756 /* Too negative for an IV. This is a double upgrade, but
d1be9408 1757 I'm assuming it will be rare. */
c2988b20
NC
1758 if (SvTYPE(sv) < SVt_PVNV)
1759 sv_upgrade(sv, SVt_PVNV);
1760 SvNOK_on(sv);
1761 SvIOK_off(sv);
1762 SvIOKp_on(sv);
9d6ce603 1763 SvNV_set(sv, -(NV)value);
45977657 1764 SvIV_set(sv, IV_MIN);
c2988b20
NC
1765 }
1766 }
1767 }
1768 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
1769 will be in the previous block to set the IV slot, and the next
1770 block to set the NV slot. So no else here. */
1771
1772 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1773 != IS_NUMBER_IN_UV) {
1774 /* It wasn't an (integer that doesn't overflow the UV). */
3f7c398e 1775 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8 1776
c2988b20
NC
1777 if (! numtype && ckWARN(WARN_NUMERIC))
1778 not_a_number(sv);
28e5dec8 1779
65202027 1780#if defined(USE_LONG_DOUBLE)
c2988b20
NC
1781 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
1782 PTR2UV(sv), SvNVX(sv)));
65202027 1783#else
1779d84d 1784 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
c2988b20 1785 PTR2UV(sv), SvNVX(sv)));
65202027 1786#endif
28e5dec8 1787
28e5dec8 1788#ifdef NV_PRESERVES_UV
af359546
NC
1789 (void)SvIOKp_on(sv);
1790 (void)SvNOK_on(sv);
1791 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
1792 SvIV_set(sv, I_V(SvNVX(sv)));
1793 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1794 SvIOK_on(sv);
1795 } else {
1796 /* Integer is imprecise. NOK, IOKp */
1797 }
1798 /* UV will not work better than IV */
1799 } else {
1800 if (SvNVX(sv) > (NV)UV_MAX) {
1801 SvIsUV_on(sv);
1802 /* Integer is inaccurate. NOK, IOKp, is UV */
1803 SvUV_set(sv, UV_MAX);
af359546
NC
1804 } else {
1805 SvUV_set(sv, U_V(SvNVX(sv)));
1806 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
1807 NV preservse UV so can do correct comparison. */
1808 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1809 SvIOK_on(sv);
af359546
NC
1810 } else {
1811 /* Integer is imprecise. NOK, IOKp, is UV */
af359546
NC
1812 }
1813 }
4b0c9573 1814 SvIsUV_on(sv);
af359546 1815 }
28e5dec8 1816#else /* NV_PRESERVES_UV */
c2988b20
NC
1817 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1818 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
af359546 1819 /* The IV/UV slot will have been set from value returned by
c2988b20
NC
1820 grok_number above. The NV slot has just been set using
1821 Atof. */
560b0c46 1822 SvNOK_on(sv);
c2988b20
NC
1823 assert (SvIOKp(sv));
1824 } else {
1825 if (((UV)1 << NV_PRESERVES_UV_BITS) >
1826 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
1827 /* Small enough to preserve all bits. */
1828 (void)SvIOKp_on(sv);
1829 SvNOK_on(sv);
45977657 1830 SvIV_set(sv, I_V(SvNVX(sv)));
c2988b20
NC
1831 if ((NV)(SvIVX(sv)) == SvNVX(sv))
1832 SvIOK_on(sv);
1833 /* Assumption: first non-preserved integer is < IV_MAX,
1834 this NV is in the preserved range, therefore: */
1835 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
1836 < (UV)IV_MAX)) {
32fdb065 1837 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
c2988b20
NC
1838 }
1839 } else {
1840 /* IN_UV NOT_INT
1841 0 0 already failed to read UV.
1842 0 1 already failed to read UV.
1843 1 0 you won't get here in this case. IV/UV
1844 slot set, public IOK, Atof() unneeded.
1845 1 1 already read UV.
1846 so there's no point in sv_2iuv_non_preserve() attempting
1847 to use atol, strtol, strtoul etc. */
40a17c4c 1848 sv_2iuv_non_preserve (sv, numtype);
c2988b20
NC
1849 }
1850 }
28e5dec8 1851#endif /* NV_PRESERVES_UV */
25da4f38 1852 }
af359546
NC
1853 }
1854 else {
1855 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
1856 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
1857 report_uninit(sv);
1858 }
25da4f38
IZ
1859 if (SvTYPE(sv) < SVt_IV)
1860 /* Typically the caller expects that sv_any is not NULL now. */
1861 sv_upgrade(sv, SVt_IV);
af359546
NC
1862 /* Return 0 from the caller. */
1863 return TRUE;
1864 }
1865 return FALSE;
1866}
1867
1868/*
1869=for apidoc sv_2iv_flags
1870
1871Return the integer value of an SV, doing any necessary string
1872conversion. If flags includes SV_GMAGIC, does an mg_get() first.
1873Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
1874
1875=cut
1876*/
1877
1878IV
1879Perl_sv_2iv_flags(pTHX_ register SV *sv, I32 flags)
1880{
1881 if (!sv)
a0d0e21e 1882 return 0;
af359546
NC
1883 if (SvGMAGICAL(sv)) {
1884 if (flags & SV_GMAGIC)
1885 mg_get(sv);
1886 if (SvIOKp(sv))
1887 return SvIVX(sv);
1888 if (SvNOKp(sv)) {
1889 return I_V(SvNVX(sv));
1890 }
71c558c3
NC
1891 if (SvPOKp(sv) && SvLEN(sv)) {
1892 UV value;
1893 const int numtype
1894 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
1895
1896 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1897 == IS_NUMBER_IN_UV) {
1898 /* It's definitely an integer */
1899 if (numtype & IS_NUMBER_NEG) {
1900 if (value < (UV)IV_MIN)
1901 return -(IV)value;
1902 } else {
1903 if (value < (UV)IV_MAX)
1904 return (IV)value;
1905 }
1906 }
1907 if (!numtype) {
1908 if (ckWARN(WARN_NUMERIC))
1909 not_a_number(sv);
1910 }
1911 return I_V(Atof(SvPVX_const(sv)));
1912 }
af359546
NC
1913 if (!SvROK(sv)) {
1914 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
1915 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
1916 report_uninit(sv);
1917 }
1918 return 0;
1919 }
71c558c3
NC
1920 /* Else this will drop through into the SvROK case just below, which
1921 will return within the {} for all code paths. */
af359546
NC
1922 }
1923 if (SvTHINKFIRST(sv)) {
1924 if (SvROK(sv)) {
1925 if (SvAMAGIC(sv)) {
1926 SV * const tmpstr=AMG_CALLun(sv,numer);
1927 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
1928 return SvIV(tmpstr);
1929 }
1930 }
1931 return PTR2IV(SvRV(sv));
1932 }
1933 if (SvIsCOW(sv)) {
1934 sv_force_normal_flags(sv, 0);
1935 }
1936 if (SvREADONLY(sv) && !SvOK(sv)) {
1937 if (ckWARN(WARN_UNINITIALIZED))
1938 report_uninit(sv);
1939 return 0;
1940 }
1941 }
1942 if (!SvIOKp(sv)) {
1943 if (S_sv_2iuv_common(aTHX_ sv))
1944 return 0;
79072805 1945 }
1d7c1841
GS
1946 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
1947 PTR2UV(sv),SvIVX(sv)));
25da4f38 1948 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
79072805
LW
1949}
1950
645c22ef 1951/*
891f9566 1952=for apidoc sv_2uv_flags
645c22ef
DM
1953
1954Return the unsigned integer value of an SV, doing any necessary string
891f9566
YST
1955conversion. If flags includes SV_GMAGIC, does an mg_get() first.
1956Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
645c22ef
DM
1957
1958=cut
1959*/
1960
ff68c719 1961UV
891f9566 1962Perl_sv_2uv_flags(pTHX_ register SV *sv, I32 flags)
ff68c719
PP
1963{
1964 if (!sv)
1965 return 0;
1966 if (SvGMAGICAL(sv)) {
891f9566
YST
1967 if (flags & SV_GMAGIC)
1968 mg_get(sv);
ff68c719
PP
1969 if (SvIOKp(sv))
1970 return SvUVX(sv);
1971 if (SvNOKp(sv))
1972 return U_V(SvNVX(sv));
71c558c3
NC
1973 if (SvPOKp(sv) && SvLEN(sv)) {
1974 UV value;
1975 const int numtype
1976 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
1977
1978 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
1979 == IS_NUMBER_IN_UV) {
1980 /* It's definitely an integer */
1981 if (!(numtype & IS_NUMBER_NEG))
1982 return value;
1983 }
1984 if (!numtype) {
1985 if (ckWARN(WARN_NUMERIC))
1986 not_a_number(sv);
1987 }
1988 return U_V(Atof(SvPVX_const(sv)));
1989 }
3fe9a6f1 1990 if (!SvROK(sv)) {
d008e5eb 1991 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 1992 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 1993 report_uninit(sv);
c6ee37c5 1994 }
36477c24 1995 return 0;
3fe9a6f1 1996 }
71c558c3
NC
1997 /* Else this will drop through into the SvROK case just below, which
1998 will return within the {} for all code paths. */
ff68c719
PP
1999 }
2000 if (SvTHINKFIRST(sv)) {
2001 if (SvROK(sv)) {
ff68c719 2002 SV* tmpstr;
1554e226 2003 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,numer)) &&
b4b9a328 2004 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv))))
9e7bc3e8 2005 return SvUV(tmpstr);
56431972 2006 return PTR2UV(SvRV(sv));
ff68c719 2007 }
765f542d
NC
2008 if (SvIsCOW(sv)) {
2009 sv_force_normal_flags(sv, 0);
8a818333 2010 }
0336b60e 2011 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2012 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2013 report_uninit(sv);
ff68c719
PP
2014 return 0;
2015 }
2016 }
af359546
NC
2017 if (!SvIOKp(sv)) {
2018 if (S_sv_2iuv_common(aTHX_ sv))
2019 return 0;
ff68c719 2020 }
25da4f38 2021
1d7c1841
GS
2022 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2023 PTR2UV(sv),SvUVX(sv)));
25da4f38 2024 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
ff68c719
PP
2025}
2026
645c22ef
DM
2027/*
2028=for apidoc sv_2nv
2029
2030Return the num value of an SV, doing any necessary string or integer
2031conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2032macros.
2033
2034=cut
2035*/
2036
65202027 2037NV
864dbfa3 2038Perl_sv_2nv(pTHX_ register SV *sv)
79072805
LW
2039{
2040 if (!sv)
2041 return 0.0;
8990e307 2042 if (SvGMAGICAL(sv)) {
463ee0b2
LW
2043 mg_get(sv);
2044 if (SvNOKp(sv))
2045 return SvNVX(sv);
a0d0e21e 2046 if (SvPOKp(sv) && SvLEN(sv)) {
041457d9 2047 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
504618e9 2048 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
a0d0e21e 2049 not_a_number(sv);
3f7c398e 2050 return Atof(SvPVX_const(sv));
a0d0e21e 2051 }
25da4f38 2052 if (SvIOKp(sv)) {
1c846c1f 2053 if (SvIsUV(sv))
65202027 2054 return (NV)SvUVX(sv);
25da4f38 2055 else
65202027 2056 return (NV)SvIVX(sv);
25da4f38 2057 }
16d20bd9 2058 if (!SvROK(sv)) {
d008e5eb 2059 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 2060 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 2061 report_uninit(sv);
c6ee37c5 2062 }
66a1b24b 2063 return (NV)0;
16d20bd9 2064 }
71c558c3
NC
2065 /* Else this will drop through into the SvROK case just below, which
2066 will return within the {} for all code paths. */
463ee0b2 2067 }
ed6116ce 2068 if (SvTHINKFIRST(sv)) {
a0d0e21e 2069 if (SvROK(sv)) {
a0d0e21e 2070 SV* tmpstr;
1554e226 2071 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,numer)) &&
b4b9a328 2072 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv))))
9e7bc3e8 2073 return SvNV(tmpstr);
56431972 2074 return PTR2NV(SvRV(sv));
a0d0e21e 2075 }
765f542d
NC
2076 if (SvIsCOW(sv)) {
2077 sv_force_normal_flags(sv, 0);
8a818333 2078 }
0336b60e 2079 if (SvREADONLY(sv) && !SvOK(sv)) {
599cee73 2080 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2081 report_uninit(sv);
ed6116ce
LW
2082 return 0.0;
2083 }
79072805
LW
2084 }
2085 if (SvTYPE(sv) < SVt_NV) {
463ee0b2
LW
2086 if (SvTYPE(sv) == SVt_IV)
2087 sv_upgrade(sv, SVt_PVNV);
2088 else
2089 sv_upgrade(sv, SVt_NV);
906f284f 2090#ifdef USE_LONG_DOUBLE
097ee67d 2091 DEBUG_c({
f93f4e46 2092 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2093 PerlIO_printf(Perl_debug_log,
2094 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2095 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2096 RESTORE_NUMERIC_LOCAL();
2097 });
65202027 2098#else
572bbb43 2099 DEBUG_c({
f93f4e46 2100 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2101 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
1d7c1841 2102 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2103 RESTORE_NUMERIC_LOCAL();
2104 });
572bbb43 2105#endif
79072805
LW
2106 }
2107 else if (SvTYPE(sv) < SVt_PVNV)
2108 sv_upgrade(sv, SVt_PVNV);
59d8ce62
NC
2109 if (SvNOKp(sv)) {
2110 return SvNVX(sv);
61604483 2111 }
59d8ce62 2112 if (SvIOKp(sv)) {
9d6ce603 2113 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
28e5dec8
JH
2114#ifdef NV_PRESERVES_UV
2115 SvNOK_on(sv);
2116#else
2117 /* Only set the public NV OK flag if this NV preserves the IV */
2118 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2119 if (SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2120 : (SvIVX(sv) == I_V(SvNVX(sv))))
2121 SvNOK_on(sv);
2122 else
2123 SvNOKp_on(sv);
2124#endif
93a17b20 2125 }
748a9306 2126 else if (SvPOKp(sv) && SvLEN(sv)) {
c2988b20 2127 UV value;
3f7c398e 2128 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
041457d9 2129 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
a0d0e21e 2130 not_a_number(sv);
28e5dec8 2131#ifdef NV_PRESERVES_UV
c2988b20
NC
2132 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2133 == IS_NUMBER_IN_UV) {
5e045b90 2134 /* It's definitely an integer */
9d6ce603 2135 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
c2988b20 2136 } else
3f7c398e 2137 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2138 SvNOK_on(sv);
2139#else
3f7c398e 2140 SvNV_set(sv, Atof(SvPVX_const(sv)));
28e5dec8
JH
2141 /* Only set the public NV OK flag if this NV preserves the value in
2142 the PV at least as well as an IV/UV would.
2143 Not sure how to do this 100% reliably. */
2144 /* if that shift count is out of range then Configure's test is
2145 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2146 UV_BITS */
2147 if (((UV)1 << NV_PRESERVES_UV_BITS) >
c2988b20 2148 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
28e5dec8 2149 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
c2988b20
NC
2150 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2151 /* Can't use strtol etc to convert this string, so don't try.
2152 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2153 SvNOK_on(sv);
2154 } else {
2155 /* value has been set. It may not be precise. */
2156 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2157 /* 2s complement assumption for (UV)IV_MIN */
2158 SvNOK_on(sv); /* Integer is too negative. */
2159 } else {
2160 SvNOKp_on(sv);
2161 SvIOKp_on(sv);
6fa402ec 2162
c2988b20 2163 if (numtype & IS_NUMBER_NEG) {
45977657 2164 SvIV_set(sv, -(IV)value);
c2988b20 2165 } else if (value <= (UV)IV_MAX) {
45977657 2166 SvIV_set(sv, (IV)value);
c2988b20 2167 } else {
607fa7f2 2168 SvUV_set(sv, value);
c2988b20
NC
2169 SvIsUV_on(sv);
2170 }
2171
2172 if (numtype & IS_NUMBER_NOT_INT) {
2173 /* I believe that even if the original PV had decimals,
2174 they are lost beyond the limit of the FP precision.
2175 However, neither is canonical, so both only get p
2176 flags. NWC, 2000/11/25 */
2177 /* Both already have p flags, so do nothing */
2178 } else {
66a1b24b 2179 const NV nv = SvNVX(sv);
c2988b20
NC
2180 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2181 if (SvIVX(sv) == I_V(nv)) {
2182 SvNOK_on(sv);
c2988b20 2183 } else {
c2988b20
NC
2184 /* It had no "." so it must be integer. */
2185 }
00b6aa41 2186 SvIOK_on(sv);
c2988b20
NC
2187 } else {
2188 /* between IV_MAX and NV(UV_MAX).
2189 Could be slightly > UV_MAX */
6fa402ec 2190
c2988b20
NC
2191 if (numtype & IS_NUMBER_NOT_INT) {
2192 /* UV and NV both imprecise. */
2193 } else {
66a1b24b 2194 const UV nv_as_uv = U_V(nv);
c2988b20
NC
2195
2196 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2197 SvNOK_on(sv);
c2988b20 2198 }
00b6aa41 2199 SvIOK_on(sv);
c2988b20
NC
2200 }
2201 }
2202 }
2203 }
2204 }
28e5dec8 2205#endif /* NV_PRESERVES_UV */
93a17b20 2206 }
79072805 2207 else {
041457d9 2208 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2209 report_uninit(sv);
25da4f38
IZ
2210 if (SvTYPE(sv) < SVt_NV)
2211 /* Typically the caller expects that sv_any is not NULL now. */
28e5dec8
JH
2212 /* XXX Ilya implies that this is a bug in callers that assume this
2213 and ideally should be fixed. */
25da4f38 2214 sv_upgrade(sv, SVt_NV);
a0d0e21e 2215 return 0.0;
79072805 2216 }
572bbb43 2217#if defined(USE_LONG_DOUBLE)
097ee67d 2218 DEBUG_c({
f93f4e46 2219 STORE_NUMERIC_LOCAL_SET_STANDARD();
1d7c1841
GS
2220 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2221 PTR2UV(sv), SvNVX(sv));
572bbb43
GS
2222 RESTORE_NUMERIC_LOCAL();
2223 });
65202027 2224#else
572bbb43 2225 DEBUG_c({
f93f4e46 2226 STORE_NUMERIC_LOCAL_SET_STANDARD();
1779d84d 2227 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
1d7c1841 2228 PTR2UV(sv), SvNVX(sv));
097ee67d
JH
2229 RESTORE_NUMERIC_LOCAL();
2230 });
572bbb43 2231#endif
463ee0b2 2232 return SvNVX(sv);
79072805
LW
2233}
2234
645c22ef
DM
2235/* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2236 * UV as a string towards the end of buf, and return pointers to start and
2237 * end of it.
2238 *
2239 * We assume that buf is at least TYPE_CHARS(UV) long.
2240 */
2241
864dbfa3 2242static char *
aec46f14 2243S_uiv_2buf(char *buf, IV iv, UV uv, int is_uv, char **peob)
25da4f38 2244{
25da4f38 2245 char *ptr = buf + TYPE_CHARS(UV);
823a54a3 2246 char * const ebuf = ptr;
25da4f38 2247 int sign;
25da4f38
IZ
2248
2249 if (is_uv)
2250 sign = 0;
2251 else if (iv >= 0) {
2252 uv = iv;
2253 sign = 0;
2254 } else {
2255 uv = -iv;
2256 sign = 1;
2257 }
2258 do {
eb160463 2259 *--ptr = '0' + (char)(uv % 10);
25da4f38
IZ
2260 } while (uv /= 10);
2261 if (sign)
2262 *--ptr = '-';
2263 *peob = ebuf;
2264 return ptr;
2265}
2266
9af30d34
NC
2267/* stringify_regexp(): private routine for use by sv_2pv_flags(): converts
2268 * a regexp to its stringified form.
2269 */
2270
2271static char *
2272S_stringify_regexp(pTHX_ SV *sv, MAGIC *mg, STRLEN *lp) {
00b6aa41 2273 const regexp * const re = (regexp *)mg->mg_obj;
9af30d34
NC
2274
2275 if (!mg->mg_ptr) {
2276 const char *fptr = "msix";
2277 char reflags[6];
2278 char ch;
2279 int left = 0;
2280 int right = 4;
00b6aa41 2281 bool need_newline = 0;
9af30d34
NC
2282 U16 reganch = (U16)((re->reganch & PMf_COMPILETIME) >> 12);
2283
2284 while((ch = *fptr++)) {
2285 if(reganch & 1) {
2286 reflags[left++] = ch;
2287 }
2288 else {
2289 reflags[right--] = ch;
2290 }
2291 reganch >>= 1;
2292 }
2293 if(left != 4) {
2294 reflags[left] = '-';
2295 left = 5;
2296 }
2297
2298 mg->mg_len = re->prelen + 4 + left;
2299 /*
2300 * If /x was used, we have to worry about a regex ending with a
2301 * comment later being embedded within another regex. If so, we don't
2302 * want this regex's "commentization" to leak out to the right part of
2303 * the enclosing regex, we must cap it with a newline.
2304 *
2305 * So, if /x was used, we scan backwards from the end of the regex. If
2306 * we find a '#' before we find a newline, we need to add a newline
2307 * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
2308 * we don't need to add anything. -jfriedl
2309 */
2310 if (PMf_EXTENDED & re->reganch) {
2311 const char *endptr = re->precomp + re->prelen;
2312 while (endptr >= re->precomp) {
2313 const char c = *(endptr--);
2314 if (c == '\n')
2315 break; /* don't need another */
2316 if (c == '#') {
2317 /* we end while in a comment, so we need a newline */
2318 mg->mg_len++; /* save space for it */
2319 need_newline = 1; /* note to add it */
2320 break;
2321 }
2322 }
2323 }
2324
2325 Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
2326 mg->mg_ptr[0] = '(';
2327 mg->mg_ptr[1] = '?';
2328 Copy(reflags, mg->mg_ptr+2, left, char);
2329 *(mg->mg_ptr+left+2) = ':';
2330 Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
2331 if (need_newline)
2332 mg->mg_ptr[mg->mg_len - 2] = '\n';
2333 mg->mg_ptr[mg->mg_len - 1] = ')';
2334 mg->mg_ptr[mg->mg_len] = 0;
2335 }
2336 PL_reginterp_cnt += re->program[0].next_off;
2337
2338 if (re->reganch & ROPT_UTF8)
2339 SvUTF8_on(sv);
2340 else
2341 SvUTF8_off(sv);
2342 if (lp)
2343 *lp = mg->mg_len;
2344 return mg->mg_ptr;
2345}
2346
645c22ef
DM
2347/*
2348=for apidoc sv_2pv_flags
2349
ff276b08 2350Returns a pointer to the string value of an SV, and sets *lp to its length.
645c22ef
DM
2351If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2352if necessary.
2353Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2354usually end up here too.
2355
2356=cut
2357*/
2358
8d6d96c1
HS
2359char *
2360Perl_sv_2pv_flags(pTHX_ register SV *sv, STRLEN *lp, I32 flags)
2361{
79072805
LW
2362 register char *s;
2363 int olderrno;
2364
463ee0b2 2365 if (!sv) {
cdb061a3
NC
2366 if (lp)
2367 *lp = 0;
73d840c0 2368 return (char *)"";
463ee0b2 2369 }
8990e307 2370 if (SvGMAGICAL(sv)) {
8d6d96c1
HS
2371 if (flags & SV_GMAGIC)
2372 mg_get(sv);
463ee0b2 2373 if (SvPOKp(sv)) {
cdb061a3
NC
2374 if (lp)
2375 *lp = SvCUR(sv);
10516c54
NC
2376 if (flags & SV_MUTABLE_RETURN)
2377 return SvPVX_mutable(sv);
4d84ee25
NC
2378 if (flags & SV_CONST_RETURN)
2379 return (char *)SvPVX_const(sv);
463ee0b2
LW
2380 return SvPVX(sv);
2381 }
75dfc8ec
NC
2382 if (SvIOKp(sv) || SvNOKp(sv)) {
2383 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
75dfc8ec
NC
2384 STRLEN len;
2385
2386 if (SvIOKp(sv)) {
e8ada2d0
NC
2387 len = SvIsUV(sv) ? my_sprintf(tbuf,"%"UVuf, (UV)SvUVX(sv))
2388 : my_sprintf(tbuf,"%"IVdf, (IV)SvIVX(sv));
75dfc8ec 2389 } else {
e8ada2d0
NC
2390 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2391 len = strlen(tbuf);
75dfc8ec
NC
2392 }
2393 if (SvROK(sv)) { /* XXX Skip this when sv_pvn_force calls */
2394 /* Sneaky stuff here */
00b6aa41 2395 SV * const tsv = newSVpvn(tbuf, len);
75dfc8ec
NC
2396
2397 sv_2mortal(tsv);
2398 if (lp)
2399 *lp = SvCUR(tsv);
2400 return SvPVX(tsv);
2401 }
2402 else {
2403 dVAR;
2404
2405#ifdef FIXNEGATIVEZERO
e8ada2d0
NC
2406 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2407 tbuf[0] = '0';
2408 tbuf[1] = 0;
75dfc8ec
NC
2409 len = 1;
2410 }
2411#endif
2412 SvUPGRADE(sv, SVt_PV);
2413 if (lp)
2414 *lp = len;
2415 s = SvGROW_mutable(sv, len + 1);
2416 SvCUR_set(sv, len);
2417 SvPOKp_on(sv);
e8ada2d0 2418 return memcpy(s, tbuf, len + 1);
75dfc8ec 2419 }
463ee0b2 2420 }
16d20bd9 2421 if (!SvROK(sv)) {
d008e5eb 2422 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
041457d9 2423 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
29489e7c 2424 report_uninit(sv);
c6ee37c5 2425 }
cdb061a3
NC
2426 if (lp)
2427 *lp = 0;
73d840c0 2428 return (char *)"";
16d20bd9 2429 }
71c558c3
NC
2430 /* Else this will drop through into the SvROK case just below, which
2431 will return within the {} for all code paths. */
463ee0b2 2432 }
ed6116ce
LW
2433 if (SvTHINKFIRST(sv)) {
2434 if (SvROK(sv)) {
a0d0e21e 2435 SV* tmpstr;
d8eae41e 2436
1554e226 2437 if (SvAMAGIC(sv) && (tmpstr=AMG_CALLun(sv,string)) &&
b4b9a328 2438 (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
50adf7d2
NC
2439 /* Unwrap this: */
2440 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr); */
2441
2442 char *pv;
2443 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2444 if (flags & SV_CONST_RETURN) {
2445 pv = (char *) SvPVX_const(tmpstr);
2446 } else {
2447 pv = (flags & SV_MUTABLE_RETURN)
2448 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2449 }
2450 if (lp)
2451 *lp = SvCUR(tmpstr);
2452 } else {
2453 pv = sv_2pv_flags(tmpstr, lp, flags);
2454 }
446eaa42
YST
2455 if (SvUTF8(tmpstr))
2456 SvUTF8_on(sv);
2457 else
2458 SvUTF8_off(sv);
2459 return pv;
d8eae41e 2460 } else {
75dfc8ec 2461 SV *tsv;
f9277f47 2462 MAGIC *mg;
d8eae41e
NC
2463 const SV *const referent = (SV*)SvRV(sv);
2464
2465 if (!referent) {
042dae7a
NC
2466 tsv = sv_2mortal(newSVpvn("NULLREF", 7));
2467 } else if (SvTYPE(referent) == SVt_PVMG
2468 && ((SvFLAGS(referent) &
2469 (SVs_OBJECT|SVf_OK|SVs_GMG|SVs_SMG|SVs_RMG))
2470 == (SVs_OBJECT|SVs_SMG))
2471 && (mg = mg_find(referent, PERL_MAGIC_qr))) {
9af30d34 2472 return S_stringify_regexp(aTHX_ sv, mg, lp);
d8eae41e
NC
2473 } else {
2474 const char *const typestr = sv_reftype(referent, 0);
2475
2476 tsv = sv_newmortal();
2477 if (SvOBJECT(referent)) {
2478 const char *const name = HvNAME_get(SvSTASH(referent));
2479 Perl_sv_setpvf(aTHX_ tsv, "%s=%s(0x%"UVxf")",
2480 name ? name : "__ANON__" , typestr,
2481 PTR2UV(referent));
2482 }
2483 else
2484 Perl_sv_setpvf(aTHX_ tsv, "%s(0x%"UVxf")", typestr,
2485 PTR2UV(referent));
c080367d 2486 }
042dae7a
NC
2487 if (lp)
2488 *lp = SvCUR(tsv);
2489 return SvPVX(tsv);
463ee0b2 2490 }
79072805 2491 }
0336b60e 2492 if (SvREADONLY(sv) && !SvOK(sv)) {
0336b60e 2493 if (ckWARN(WARN_UNINITIALIZED))
29489e7c 2494 report_uninit(sv);
cdb061a3
NC
2495 if (lp)
2496 *lp = 0;
73d840c0 2497 return (char *)"";
79072805 2498 }
79072805 2499 }
28e5dec8
JH
2500 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2501 /* I'm assuming that if both IV and NV are equally valid then
2502 converting the IV is going to be more efficient */
e1ec3a88
AL
2503 const U32 isIOK = SvIOK(sv);
2504 const U32 isUIOK = SvIsUV(sv);
28e5dec8
JH
2505 char buf[TYPE_CHARS(UV)];
2506 char *ebuf, *ptr;
2507
2508 if (SvTYPE(sv) < SVt_PVIV)
2509 sv_upgrade(sv, SVt_PVIV);
2510 if (isUIOK)
2511 ptr = uiv_2buf(buf, 0, SvUVX(sv), 1, &ebuf);
2512 else
2513 ptr = uiv_2buf(buf, SvIVX(sv), 0, 0, &ebuf);
5902b6a9
NC
2514 /* inlined from sv_setpvn */
2515 SvGROW_mutable(sv, (STRLEN)(ebuf - ptr + 1));
4d84ee25 2516 Move(ptr,SvPVX_mutable(sv),ebuf - ptr,char);
28e5dec8
JH
2517 SvCUR_set(sv, ebuf - ptr);
2518 s = SvEND(sv);
2519 *s = '\0';
2520 if (isIOK)
2521 SvIOK_on(sv);
2522 else
2523 SvIOKp_on(sv);
2524 if (isUIOK)
2525 SvIsUV_on(sv);
2526 }
2527 else if (SvNOKp(sv)) {
79072805
LW
2528 if (SvTYPE(sv) < SVt_PVNV)
2529 sv_upgrade(sv, SVt_PVNV);
1c846c1f 2530 /* The +20 is pure guesswork. Configure test needed. --jhi */
5902b6a9 2531 s = SvGROW_mutable(sv, NV_DIG + 20);
79072805 2532 olderrno = errno; /* some Xenix systems wipe out errno here */
79072805 2533#ifdef apollo
463ee0b2 2534 if (SvNVX(sv) == 0.0)
79072805
LW
2535 (void)strcpy(s,"0");
2536 else
2537#endif /*apollo*/
bbce6d69 2538 {
2d4389e4 2539 Gconvert(SvNVX(sv), NV_DIG, 0, s);
bbce6d69 2540 }
79072805 2541 errno = olderrno;
a0d0e21e
LW
2542#ifdef FIXNEGATIVEZERO
2543 if (*s == '-' && s[1] == '0' && !s[2])
2544 strcpy(s,"0");
2545#endif
79072805
LW
2546 while (*s) s++;
2547#ifdef hcx
2548 if (s[-1] == '.')
46fc3d4c 2549 *--s = '\0';
79072805
LW
2550#endif
2551 }
79072805 2552 else {
041457d9 2553 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
29489e7c 2554 report_uninit(sv);
cdb061a3 2555 if (lp)
00b6aa41 2556 *lp = 0;
25da4f38
IZ
2557 if (SvTYPE(sv) < SVt_PV)
2558 /* Typically the caller expects that sv_any is not NULL now. */
2559 sv_upgrade(sv, SVt_PV);
73d840c0 2560 return (char *)"";
79072805 2561 }
cdb061a3 2562 {
823a54a3 2563 const STRLEN len = s - SvPVX_const(sv);
cdb061a3
NC
2564 if (lp)
2565 *lp = len;
2566 SvCUR_set(sv, len);
2567 }
79072805 2568 SvPOK_on(sv);
1d7c1841 2569 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3f7c398e 2570 PTR2UV(sv),SvPVX_const(sv)));
4d84ee25
NC
2571 if (flags & SV_CONST_RETURN)
2572 return (char *)SvPVX_const(sv);
10516c54
NC
2573 if (flags & SV_MUTABLE_RETURN)
2574 return SvPVX_mutable(sv);
463ee0b2
LW
2575 return SvPVX(sv);
2576}
2577
645c22ef 2578/*
6050d10e
JP
2579=for apidoc sv_copypv
2580
2581Copies a stringified representation of the source SV into the
2582destination SV. Automatically performs any necessary mg_get and
54f0641b 2583coercion of numeric values into strings. Guaranteed to preserve
6050d10e 2584UTF-8 flag even from overloaded objects. Similar in nature to
54f0641b
NIS
2585sv_2pv[_flags] but operates directly on an SV instead of just the
2586string. Mostly uses sv_2pv_flags to do its work, except when that
6050d10e
JP
2587would lose the UTF-8'ness of the PV.
2588
2589=cut
2590*/
2591
2592void
2593Perl_sv_copypv(pTHX_ SV *dsv, register SV *ssv)
2594{
446eaa42 2595 STRLEN len;
53c1dcc0 2596 const char * const s = SvPV_const(ssv,len);
cb50f42d 2597 sv_setpvn(dsv,s,len);
446eaa42 2598 if (SvUTF8(ssv))
cb50f42d 2599 SvUTF8_on(dsv);
446eaa42 2600 else
cb50f42d 2601 SvUTF8_off(dsv);
6050d10e
JP
2602}
2603
2604/*
645c22ef
DM
2605=for apidoc sv_2pvbyte
2606
2607Return a pointer to the byte-encoded representation of the SV, and set *lp
1e54db1a 2608to its length. May cause the SV to be downgraded from UTF-8 as a
645c22ef
DM
2609side-effect.
2610
2611Usually accessed via the C<SvPVbyte> macro.
2612
2613=cut
2614*/
2615
7340a771
GS
2616char *
2617Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *lp)
2618{
0875d2fe 2619 sv_utf8_downgrade(sv,0);
97972285 2620 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771
GS
2621}
2622
645c22ef 2623/*
035cbb0e
RGS
2624=for apidoc sv_2pvutf8
2625
2626Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
2627to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
2628
2629Usually accessed via the C<SvPVutf8> macro.
2630
2631=cut
2632*/
645c22ef 2633
7340a771
GS
2634char *
2635Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *lp)
2636{
035cbb0e
RGS
2637 sv_utf8_upgrade(sv);
2638 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
7340a771 2639}
1c846c1f 2640
7ee2227d 2641
645c22ef
DM
2642/*
2643=for apidoc sv_2bool
2644
2645This function is only called on magical items, and is only used by
8cf8f3d1 2646sv_true() or its macro equivalent.
645c22ef
DM
2647
2648=cut
2649*/
2650
463ee0b2 2651bool
864dbfa3 2652Perl_sv_2bool(pTHX_ register SV *sv)
463ee0b2 2653{
5b295bef 2654 SvGETMAGIC(sv);
463ee0b2 2655
a0d0e21e
LW
2656 if (!SvOK(sv))
2657 return 0;
2658 if (SvROK(sv)) {
a0d0e21e 2659 SV* tmpsv;
1554e226 2660 if (SvAMAGIC(sv) && (tmpsv=AMG_CALLun(sv,bool_)) &&
9e3013b1 2661 (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
8a31060d 2662 return (bool)SvTRUE(tmpsv);
a0d0e21e
LW
2663 return SvRV(sv) != 0;
2664 }
463ee0b2 2665 if (SvPOKp(sv)) {
53c1dcc0
AL
2666 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
2667 if (Xpvtmp &&
339049b0 2668 (*sv->sv_u.svu_pv > '0' ||
11343788 2669 Xpvtmp->xpv_cur > 1 ||
339049b0 2670 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
463ee0b2
LW
2671 return 1;
2672 else
2673 return 0;
2674 }
2675 else {
2676 if (SvIOKp(sv))
2677 return SvIVX(sv) != 0;
2678 else {
2679 if (SvNOKp(sv))
2680 return SvNVX(sv) != 0.0;
2681 else
2682 return FALSE;
2683 }
2684 }
79072805
LW
2685}
2686
c461cf8f
JH
2687/*
2688=for apidoc sv_utf8_upgrade
2689
78ea37eb 2690Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2691Forces the SV to string form if it is not already.
4411f3b6
NIS
2692Always sets the SvUTF8 flag to avoid future validity checks even
2693if all the bytes have hibit clear.
c461cf8f 2694
13a6c0e0
JH
2695This is not as a general purpose byte encoding to Unicode interface:
2696use the Encode extension for that.
2697
8d6d96c1
HS
2698=for apidoc sv_utf8_upgrade_flags
2699
78ea37eb 2700Converts the PV of an SV to its UTF-8-encoded form.
645c22ef 2701Forces the SV to string form if it is not already.
8d6d96c1
HS
2702Always sets the SvUTF8 flag to avoid future validity checks even
2703if all the bytes have hibit clear. If C<flags> has C<SV_GMAGIC> bit set,
2704will C<mg_get> on C<sv> if appropriate, else not. C<sv_utf8_upgrade> and
2705C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
2706
13a6c0e0
JH
2707This is not as a general purpose byte encoding to Unicode interface:
2708use the Encode extension for that.
2709
8d6d96c1
HS
2710=cut
2711*/
2712
2713STRLEN
2714Perl_sv_utf8_upgrade_flags(pTHX_ register SV *sv, I32 flags)
2715{
808c356f
RGS
2716 if (sv == &PL_sv_undef)
2717 return 0;
e0e62c2a
NIS
2718 if (!SvPOK(sv)) {
2719 STRLEN len = 0;
d52b7888
NC
2720 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
2721 (void) sv_2pv_flags(sv,&len, flags);
2722 if (SvUTF8(sv))
2723 return len;
2724 } else {
2725 (void) SvPV_force(sv,len);
2726 }
e0e62c2a 2727 }
4411f3b6 2728
f5cee72b 2729 if (SvUTF8(sv)) {
5fec3b1d 2730 return SvCUR(sv);
f5cee72b 2731 }
5fec3b1d 2732
765f542d
NC
2733 if (SvIsCOW(sv)) {
2734 sv_force_normal_flags(sv, 0);
db42d148
NIS
2735 }
2736
88632417 2737 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING))
799ef3cb 2738 sv_recode_to_utf8(sv, PL_encoding);
9f4817db 2739 else { /* Assume Latin-1/EBCDIC */
c4e7c712
NC
2740 /* This function could be much more efficient if we
2741 * had a FLAG in SVs to signal if there are any hibit
2742 * chars in the PV. Given that there isn't such a flag
2743 * make the loop as fast as possible. */
00b6aa41 2744 const U8 * const s = (U8 *) SvPVX_const(sv);
c4420975 2745 const U8 * const e = (U8 *) SvEND(sv);
93524f2b 2746 const U8 *t = s;
c4e7c712
NC
2747
2748 while (t < e) {
53c1dcc0 2749 const U8 ch = *t++;
00b6aa41
AL
2750 /* Check for hi bit */
2751 if (!NATIVE_IS_INVARIANT(ch)) {
2752 STRLEN len = SvCUR(sv) + 1; /* Plus the \0 */
2753 U8 * const recoded = bytes_to_utf8((U8*)s, &len);
2754
2755 SvPV_free(sv); /* No longer using what was there before. */
2756 SvPV_set(sv, (char*)recoded);
2757 SvCUR_set(sv, len - 1);
2758 SvLEN_set(sv, len); /* No longer know the real size. */
c4e7c712 2759 break;
00b6aa41 2760 }
c4e7c712
NC
2761 }
2762 /* Mark as UTF-8 even if no hibit - saves scanning loop */
2763 SvUTF8_on(sv);
560a288e 2764 }
4411f3b6 2765 return SvCUR(sv);
560a288e
GS
2766}
2767
c461cf8f
JH
2768/*
2769=for apidoc sv_utf8_downgrade
2770
78ea37eb
ST
2771Attempts to convert the PV of an SV from characters to bytes.
2772If the PV contains a character beyond byte, this conversion will fail;
2773in this case, either returns false or, if C<fail_ok> is not
c461cf8f
JH
2774true, croaks.
2775
13a6c0e0
JH
2776This is not as a general purpose Unicode to byte encoding interface:
2777use the Encode extension for that.
2778
c461cf8f
JH
2779=cut
2780*/
2781
560a288e
GS
2782bool
2783Perl_sv_utf8_downgrade(pTHX_ register SV* sv, bool fail_ok)
2784{
78ea37eb 2785 if (SvPOKp(sv) && SvUTF8(sv)) {
fa301091 2786 if (SvCUR(sv)) {
03cfe0ae 2787 U8 *s;
652088fc 2788 STRLEN len;
fa301091 2789
765f542d
NC
2790 if (SvIsCOW(sv)) {
2791 sv_force_normal_flags(sv, 0);
2792 }
03cfe0ae
NIS
2793 s = (U8 *) SvPV(sv, len);
2794 if (!utf8_to_bytes(s, &len)) {
fa301091
JH
2795 if (fail_ok)
2796 return FALSE;
2797 else {
2798 if (PL_op)
2799 Perl_croak(aTHX_ "Wide character in %s",
53e06cf0 2800 OP_DESC(PL_op));
fa301091
JH
2801 else
2802 Perl_croak(aTHX_ "Wide character");
2803 }
4b3603a4 2804 }
b162af07 2805 SvCUR_set(sv, len);
67e989fb 2806 }
560a288e 2807 }
ffebcc3e 2808 SvUTF8_off(sv);
560a288e
GS
2809 return TRUE;
2810}
2811
c461cf8f
JH
2812/*
2813=for apidoc sv_utf8_encode
2814
78ea37eb
ST
2815Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
2816flag off so that it looks like octets again.
c461cf8f
JH
2817
2818=cut
2819*/
2820
560a288e
GS
2821void
2822Perl_sv_utf8_encode(pTHX_ register SV *sv)
2823{
4411f3b6 2824 (void) sv_utf8_upgrade(sv);
4c94c214
NC
2825 if (SvIsCOW(sv)) {
2826 sv_force_normal_flags(sv, 0);
2827 }
2828 if (SvREADONLY(sv)) {
2829 Perl_croak(aTHX_ PL_no_modify);
2830 }
560a288e
GS
2831 SvUTF8_off(sv);
2832}
2833
4411f3b6
NIS
2834/*
2835=for apidoc sv_utf8_decode
2836
78ea37eb
ST
2837If the PV of the SV is an octet sequence in UTF-8
2838and contains a multiple-byte character, the C<SvUTF8> flag is turned on
2839so that it looks like a character. If the PV contains only single-byte
2840characters, the C<SvUTF8> flag stays being off.
2841Scans PV for validity and returns false if the PV is invalid UTF-8.
4411f3b6
NIS
2842
2843=cut
2844*/
2845
560a288e
GS
2846bool
2847Perl_sv_utf8_decode(pTHX_ register SV *sv)
2848{
78ea37eb 2849 if (SvPOKp(sv)) {
93524f2b
NC
2850 const U8 *c;
2851 const U8 *e;
9cbac4c7 2852
645c22ef
DM
2853 /* The octets may have got themselves encoded - get them back as
2854 * bytes
2855 */
2856 if (!sv_utf8_downgrade(sv, TRUE))
560a288e
GS
2857 return FALSE;
2858
2859 /* it is actually just a matter of turning the utf8 flag on, but
2860 * we want to make sure everything inside is valid utf8 first.
2861 */
93524f2b 2862 c = (const U8 *) SvPVX_const(sv);
63cd0674 2863 if (!is_utf8_string(c, SvCUR(sv)+1))
67e989fb 2864 return FALSE;
93524f2b 2865 e = (const U8 *) SvEND(sv);
511c2ff0 2866 while (c < e) {
b64e5050 2867 const U8 ch = *c++;
c4d5f83a 2868 if (!UTF8_IS_INVARIANT(ch)) {
67e989fb
JH
2869 SvUTF8_on(sv);
2870 break;
2871 }
560a288e 2872 }
560a288e
GS
2873 }
2874 return TRUE;
2875}
2876
954c1994
GS
2877/*
2878=for apidoc sv_setsv
2879
645c22ef
DM
2880Copies the contents of the source SV C<ssv> into the destination SV
2881C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
2882function if the source SV needs to be reused. Does not handle 'set' magic.
2883Loosely speaking, it performs a copy-by-value, obliterating any previous
2884content of the destination.
2885
2886You probably want to use one of the assortment of wrappers, such as
2887C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
2888C<SvSetMagicSV_nosteal>.
2889
8d6d96c1
HS
2890=for apidoc sv_setsv_flags
2891
645c22ef
DM
2892Copies the contents of the source SV C<ssv> into the destination SV
2893C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
2894function if the source SV needs to be reused. Does not handle 'set' magic.
2895Loosely speaking, it performs a copy-by-value, obliterating any previous
2896content of the destination.
2897If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
5fcdf167
NC
2898C<ssv> if appropriate, else not. If the C<flags> parameter has the
2899C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
2900and C<sv_setsv_nomg> are implemented in terms of this function.
645c22ef
DM
2901
2902You probably want to use one of the assortment of wrappers, such as
2903C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
2904C<SvSetMagicSV_nosteal>.
2905
2906This is the primary function for copying scalars, and most other
2907copy-ish functions and macros use this underneath.
8d6d96c1
HS
2908
2909=cut
2910*/
2911
2912void
2913Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV *sstr, I32 flags)
2914{
8990e307
LW
2915 register U32 sflags;
2916 register int dtype;
2917 register int stype;
463ee0b2 2918
79072805
LW
2919 if (sstr == dstr)
2920 return;
765f542d 2921 SV_CHECK_THINKFIRST_COW_DROP(dstr);
79072805 2922 if (!sstr)
3280af22 2923 sstr = &PL_sv_undef;
8990e307
LW
2924 stype = SvTYPE(sstr);
2925 dtype = SvTYPE(dstr);
79072805 2926
a0d0e21e 2927 SvAMAGIC_off(dstr);
7a5fa8a2 2928 if ( SvVOK(dstr) )
ece467f9
JP
2929 {
2930 /* need to nuke the magic */
2931 mg_free(dstr);
2932 SvRMAGICAL_off(dstr);
2933 }
9e7bc3e8 2934
463ee0b2 2935 /* There's a lot of redundancy below but we're going for speed here */
79072805 2936
8990e307 2937 switch (stype) {
79072805 2938 case SVt_NULL:
aece5585 2939 undef_sstr:
20408e3c
GS
2940 if (dtype != SVt_PVGV) {
2941 (void)SvOK_off(dstr);
2942 return;
2943 }
2944 break;
463ee0b2 2945 case SVt_IV:
aece5585
GA
2946 if (SvIOK(sstr)) {
2947 switch (dtype) {
2948 case SVt_NULL:
8990e307 2949 sv_upgrade(dstr, SVt_IV);
aece5585
GA
2950 break;
2951 case SVt_NV:
8990e307 2952 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
2953 break;
2954 case SVt_RV:
2955 case SVt_PV:
a0d0e21e 2956 sv_upgrade(dstr, SVt_PVIV);
aece5585
GA
2957 break;
2958 }
2959 (void)SvIOK_only(dstr);
45977657 2960 SvIV_set(dstr, SvIVX(sstr));
25da4f38
IZ
2961 if (SvIsUV(sstr))
2962 SvIsUV_on(dstr);
27c9684d
AP
2963 if (SvTAINTED(sstr))
2964 SvTAINT(dstr);
aece5585 2965 return;
8990e307 2966 }
aece5585
GA
2967 goto undef_sstr;
2968
463ee0b2 2969 case SVt_NV:
aece5585
GA
2970 if (SvNOK(sstr)) {
2971 switch (dtype) {
2972 case SVt_NULL:
2973 case SVt_IV:
8990e307 2974 sv_upgrade(dstr, SVt_NV);
aece5585
GA
2975 break;
2976 case SVt_RV:
2977 case SVt_PV:
2978 case SVt_PVIV:
a0d0e21e 2979 sv_upgrade(dstr, SVt_PVNV);
aece5585
GA
2980 break;
2981 }
9d6ce603 2982 SvNV_set(dstr, SvNVX(sstr));
aece5585 2983 (void)SvNOK_only(dstr);
27c9684d
AP
2984 if (SvTAINTED(sstr))
2985 SvTAINT(dstr);
aece5585 2986 return;
8990e307 2987 }
aece5585
GA
2988 goto undef_sstr;
2989
ed6116ce 2990 case SVt_RV:
8990e307 2991 if (dtype < SVt_RV)
ed6116ce 2992 sv_upgrade(dstr, SVt_RV);
c07a80fd 2993 else if (dtype == SVt_PVGV &&
23bb1b96 2994 SvROK(sstr) && SvTYPE(SvRV(sstr)) == SVt_PVGV) {
c07a80fd 2995 sstr = SvRV(sstr);
a5f75d66 2996 if (sstr == dstr) {
1d7c1841
GS
2997 if (GvIMPORTED(dstr) != GVf_IMPORTED
2998 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
2999 {
a5f75d66 3000 GvIMPORTED_on(dstr);
1d7c1841 3001 }
a5f75d66
AD
3002 GvMULTI_on(dstr);
3003 return;
3004 }
c07a80fd
PP
3005 goto glob_assign;
3006 }
ed6116ce 3007 break;
fc36a67e 3008 case SVt_PVFM:
f8c7b90f 3009#ifdef PERL_OLD_COPY_ON_WRITE
d89fc664
NC
3010 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3011 if (dtype < SVt_PVIV)
3012 sv_upgrade(dstr, SVt_PVIV);
3013 break;
3014 }
3015 /* Fall through */
3016#endif
3017 case SVt_PV:
8990e307 3018 if (dtype < SVt_PV)
463ee0b2 3019 sv_upgrade(dstr, SVt_PV);
463ee0b2
LW
3020 break;
3021 case SVt_PVIV:
8990e307 3022 if (dtype < SVt_PVIV)
463ee0b2 3023 sv_upgrade(dstr, SVt_PVIV);
463ee0b2
LW
3024 break;
3025 case SVt_PVNV:
8990e307 3026 if (dtype < SVt_PVNV)
463ee0b2 3027 sv_upgrade(dstr, SVt_PVNV);
463ee0b2 3028 break;
4633a7c4
LW
3029 case SVt_PVAV:
3030 case SVt_PVHV:
3031 case SVt_PVCV:
4633a7c4 3032 case SVt_PVIO:
a3b680e6
AL
3033 {
3034 const char * const type = sv_reftype(sstr,0);
533c011a 3035 if (PL_op)
a3b680e6 3036 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
4633a7c4 3037 else
a3b680e6
AL
3038 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3039 }
4633a7c4
LW
3040 break;
3041
79072805 3042 case SVt_PVGV:
8990e307 3043 if (dtype <= SVt_PVGV) {
c07a80fd 3044 glob_assign:
a5f75d66 3045 if (dtype != SVt_PVGV) {
a3b680e6
AL
3046 const char * const name = GvNAME(sstr);
3047 const STRLEN len = GvNAMELEN(sstr);
b76195c2
DM
3048 /* don't upgrade SVt_PVLV: it can hold a glob */
3049 if (dtype != SVt_PVLV)
3050 sv_upgrade(dstr, SVt_PVGV);
14befaf4 3051 sv_magic(dstr, dstr, PERL_MAGIC_glob, Nullch, 0);
e15faf7d
NC
3052 GvSTASH(dstr) = GvSTASH(sstr);
3053 if (GvSTASH(dstr))
3054 Perl_sv_add_backref(aTHX_ (SV*)GvSTASH(dstr), dstr);
a0d0e21e
LW
3055 GvNAME(dstr) = savepvn(name, len);
3056 GvNAMELEN(dstr) = len;
3057 SvFAKE_on(dstr); /* can coerce to non-glob */
3058 }
5bd07a3d 3059
7fb37951
AMS
3060#ifdef GV_UNIQUE_CHECK
3061 if (GvUNIQUE((GV*)dstr)) {
5bd07a3d
DM
3062 Perl_croak(aTHX_ PL_no_modify);
3063 }
3064#endif
3065
a0d0e21e 3066 (void)SvOK_off(dstr);
a5f75d66 3067 GvINTRO_off(dstr); /* one-shot flag */
1edc1566 3068 gp_free((GV*)dstr);
79072805 3069 GvGP(dstr) = gp_ref(GvGP(sstr));
27c9684d
AP
3070 if (SvTAINTED(sstr))
3071 SvTAINT(dstr);
1d7c1841
GS
3072 if (GvIMPORTED(dstr) != GVf_IMPORTED
3073 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3074 {
a5f75d66 3075 GvIMPORTED_on(dstr);
1d7c1841 3076 }
a5f75d66 3077 GvMULTI_on(dstr);
79072805
LW
3078 return;
3079 }
3080 /* FALL THROUGH */
3081
3082 default:
8d6d96c1 3083 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
973f89ab 3084 mg_get(sstr);
eb160463 3085 if ((int)SvTYPE(sstr) != stype) {
973f89ab
CS
3086 stype = SvTYPE(sstr);
3087 if (stype == SVt_PVGV && dtype <= SVt_PVGV)
3088 goto glob_assign;
3089 }
3090 }
ded42b9f 3091 if (stype == SVt_PVLV)
862a34c6 3092 SvUPGRADE(dstr, SVt_PVNV);
ded42b9f 3093 else
862a34c6 3094 SvUPGRADE(dstr, (U32)stype);
79072805
LW
3095 }
3096
8990e307
LW
3097 sflags = SvFLAGS(sstr);
3098
3099 if (sflags & SVf_ROK) {
3100 if (dtype >= SVt_PV) {
3101 if (dtype == SVt_PVGV) {
823a54a3 3102 SV * const sref = SvREFCNT_inc(SvRV(sstr));
8990e307 3103 SV *dref = 0;
a3b680e6 3104 const int intro = GvINTRO(dstr);
a0d0e21e 3105
7fb37951
AMS
3106#ifdef GV_UNIQUE_CHECK
3107 if (GvUNIQUE((GV*)dstr)) {
5bd07a3d
DM
3108 Perl_croak(aTHX_ PL_no_modify);
3109 }
3110#endif
3111
a0d0e21e 3112 if (intro) {
a5f75d66 3113 GvINTRO_off(dstr); /* one-shot flag */
1d7c1841 3114 GvLINE(dstr) = CopLINE(PL_curcop);
1edc1566 3115 GvEGV(dstr) = (GV*)dstr;
a0d0e21e 3116 }
a5f75d66 3117 GvMULTI_on(dstr);
8990e307
LW
3118 switch (SvTYPE(sref)) {
3119 case SVt_PVAV:
a0d0e21e 3120 if (intro)
890ed176 3121 SAVEGENERICSV(GvAV(dstr));
a0d0e21e
LW
3122 else
3123 dref = (SV*)GvAV(dstr);
8990e307 3124 GvAV(dstr) = (AV*)sref;
39bac7f7 3125 if (!GvIMPORTED_AV(dstr)
1d7c1841
GS
3126 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3127 {
a5f75d66 3128 GvIMPORTED_AV_on(dstr);
1d7c1841 3129 }
8990e307
LW
3130 break;
3131 case SVt_PVHV:
a0d0e21e 3132 if (intro)
890ed176 3133 SAVEGENERICSV(GvHV(dstr));
a0d0e21e
LW
3134 else
3135 dref = (SV*)GvHV(dstr);
8990e307 3136 GvHV(dstr) = (HV*)sref;
39bac7f7 3137 if (!GvIMPORTED_HV(dstr)
1d7c1841
GS
3138 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3139 {
a5f75d66 3140 GvIMPORTED_HV_on(dstr);
1d7c1841 3141 }
8990e307
LW
3142 break;
3143 case SVt_PVCV:
8ebc5c01
PP
3144 if (intro) {
3145 if (GvCVGEN(dstr) && GvCV(dstr) != (CV*)sref) {
3146 SvREFCNT_dec(GvCV(dstr));
3147 GvCV(dstr) = Nullcv;
68dc0745 3148 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3280af22 3149 PL_sub_generation++;
8ebc5c01 3150 }
890ed176 3151 SAVEGENERICSV(GvCV(dstr));
8ebc5c01 3152 }
68dc0745
PP
3153 else
3154 dref = (SV*)GvCV(dstr);
3155 if (GvCV(dstr) != (CV*)sref) {
823a54a3 3156 CV* const cv = GvCV(dstr);
4633a7c4 3157 if (cv) {
68dc0745
PP
3158 if (!GvCVGEN((GV*)dstr) &&
3159 (CvROOT(cv) || CvXSUB(cv)))
3160 {
beab0874
JT
3161 /* Redefining a sub - warning is mandatory if
3162 it was a const and its value changed. */
3163 if (ckWARN(WARN_REDEFINE)
3164 || (CvCONST(cv)
3165 && (!CvCONST((CV*)sref)
3166 || sv_cmp(cv_const_sv(cv),
3167 cv_const_sv((CV*)sref)))))
3168 {
9014280d 3169 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
beab0874 3170 CvCONST(cv)
910764e6
RGS
3171 ? "Constant subroutine %s::%s redefined"
3172 : "Subroutine %s::%s redefined",
bfcb3514 3173 HvNAME_get(GvSTASH((GV*)dstr)),
beab0874
JT
3174 GvENAME((GV*)dstr));
3175 }
9607fc9c 3176 }
fb24441d
RGS
3177 if (!intro)
3178 cv_ckproto(cv, (GV*)dstr,
93524f2b
NC
3179 SvPOK(sref)
3180 ? SvPVX_const(sref) : Nullch);
4633a7c4 3181 }
a5f75d66 3182 GvCV(dstr) = (CV*)sref;
7a4c00b4 3183 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
a5f75d66 3184 GvASSUMECV_on(dstr);
3280af22 3185 PL_sub_generation++;
a5f75d66 3186 }
39bac7f7 3187 if (!GvIMPORTED_CV(dstr)
1d7c1841
GS
3188 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3189 {
a5f75d66 3190 GvIMPORTED_CV_on(dstr);
1d7c1841 3191 }
8990e307 3192 break;
91bba347
LW
3193 case SVt_PVIO:
3194 if (intro)
890ed176 3195 SAVEGENERICSV(GvIOp(dstr));
91bba347
LW
3196 else
3197 dref = (SV*)GvIOp(dstr);
3198 GvIOp(dstr) = (IO*)sref;
3199 break;
f4d13ee9
JH
3200 case SVt_PVFM:
3201 if (intro)
890ed176 3202 SAVEGENERICSV(GvFORM(dstr));
f4d13ee9
JH
3203 else
3204 dref = (SV*)GvFORM(dstr);
3205 GvFORM(dstr) = (CV*)sref;
3206 break;
8990e307 3207 default:
a0d0e21e 3208 if (intro)
890ed176 3209 SAVEGENERICSV(GvSV(dstr));
a0d0e21e
LW
3210 else
3211 dref = (SV*)GvSV(dstr);
8990e307 3212 GvSV(dstr) = sref;
39bac7f7 3213 if (!GvIMPORTED_SV(dstr)
1d7c1841
GS
3214 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3215 {
a5f75d66 3216 GvIMPORTED_SV_on(dstr);
1d7c1841 3217 }
8990e307
LW
3218 break;
3219 }
3220 if (dref)
3221 SvREFCNT_dec(dref);
27c9684d
AP
3222 if (SvTAINTED(sstr))
3223 SvTAINT(dstr);
8990e307
LW
3224 return;
3225 }
3f7c398e 3226 if (SvPVX_const(dstr)) {
8bd4d4c5 3227 SvPV_free(dstr);
b162af07
SP
3228 SvLEN_set(dstr, 0);
3229 SvCUR_set(dstr, 0);
a0d0e21e 3230 }
8990e307 3231 }
a0d0e21e 3232 (void)SvOK_off(dstr);
b162af07 3233 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
ed6116ce 3234 SvROK_on(dstr);
8990e307 3235 if (sflags & SVp_NOK) {
3332b3c1
JH
3236 SvNOKp_on(dstr);
3237 /* Only set the public OK flag if the source has public OK. */
3238 if (sflags & SVf_NOK)
3239 SvFLAGS(dstr) |= SVf_NOK;
9d6ce603 3240 SvNV_set(dstr, SvNVX(sstr));
ed6116ce 3241 }
8990e307 3242 if (sflags & SVp_IOK) {
3332b3c1
JH
3243 (void)SvIOKp_on(dstr);
3244 if (sflags & SVf_IOK)
3245 SvFLAGS(dstr) |= SVf_IOK;
2b1c7e3e 3246 if (sflags & SVf_IVisUV)
25da4f38 3247 SvIsUV_on(dstr);
45977657 3248 SvIV_set(dstr, SvIVX(sstr));
ed6116ce 3249 }
a0d0e21e
LW
3250 if (SvAMAGIC(sstr)) {
3251 SvAMAGIC_on(dstr);
3252 }
ed6116ce 3253 }
8990e307 3254 else if (sflags & SVp_POK) {
765f542d 3255 bool isSwipe = 0;
79072805
LW
3256
3257 /*
3258 * Check to see if we can just swipe the string. If so, it's a
3259 * possible small lose on short strings, but a big win on long ones.
3f7c398e
SP
3260 * It might even be a win on short strings if SvPVX_const(dstr)
3261 * has to be allocated and SvPVX_const(sstr) has to be freed.
79072805
LW
3262 */
3263
120fac95
NC
3264 /* Whichever path we take through the next code, we want this true,
3265 and doing it now facilitates the COW check. */
3266 (void)SvPOK_only(dstr);
3267
765f542d 3268 if (
b8f9541a
NC
3269 /* We're not already COW */
3270 ((sflags & (SVf_FAKE | SVf_READONLY)) != (SVf_FAKE | SVf_READONLY)
f8c7b90f 3271#ifndef PERL_OLD_COPY_ON_WRITE
b8f9541a
NC
3272 /* or we are, but dstr isn't a suitable target. */
3273 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
3274#endif
3275 )
765f542d 3276 &&
765f542d
NC
3277 !(isSwipe =
3278 (sflags & SVs_TEMP) && /* slated for free anyway? */
3279 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
5fcdf167
NC
3280 (!(flags & SV_NOSTEAL)) &&
3281 /* and we're allowed to steal temps */
765f542d
NC
3282 SvREFCNT(sstr) == 1 && /* and no other references to it? */
3283 SvLEN(sstr) && /* and really is a string */
645c22ef 3284 /* and won't be needed again, potentially */
765f542d 3285 !(PL_op && PL_op->op_type == OP_AASSIGN))
f8c7b90f 3286#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3287 && !((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
120fac95 3288 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
765f542d
NC
3289 && SvTYPE(sstr) >= SVt_PVIV)
3290#endif
3291 ) {
3292 /* Failed the swipe test, and it's not a shared hash key either.
3293 Have to copy the string. */
3294 STRLEN len = SvCUR(sstr);
3295 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
3f7c398e 3296 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
765f542d
NC
3297 SvCUR_set(dstr, len);
3298 *SvEND(dstr) = '\0';
765f542d 3299 } else {
f8c7b90f 3300 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
765f542d 3301 be true in here. */
765f542d
NC
3302 /* Either it's a shared hash key, or it's suitable for
3303 copy-on-write or we can swipe the string. */
46187eeb 3304 if (DEBUG_C_TEST) {
ed252734 3305 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
e419cbc5
NC
3306 sv_dump(sstr);
3307 sv_dump(dstr);
46187eeb 3308 }
f8c7b90f 3309#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3310 if (!isSwipe) {
3311 /* I believe I should acquire a global SV mutex if
3312 it's a COW sv (not a shared hash key) to stop
3313 it going un copy-on-write.
3314 If the source SV has gone un copy on write between up there
3315 and down here, then (assert() that) it is of the correct
3316 form to make it copy on write again */
3317 if ((sflags & (SVf_FAKE | SVf_READONLY))
3318 != (SVf_FAKE | SVf_READONLY)) {
3319 SvREADONLY_on(sstr);
3320 SvFAKE_on(sstr);
3321 /* Make the source SV into a loop of 1.
3322 (about to become 2) */
a29f6d03 3323 SV_COW_NEXT_SV_SET(sstr, sstr);
765f542d
NC
3324 }
3325 }
3326#endif
3327 /* Initial code is common. */
94010e71
NC
3328 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
3329 SvPV_free(dstr);
79072805 3330 }
765f542d 3331
765f542d
NC
3332 if (!isSwipe) {
3333 /* making another shared SV. */
3334 STRLEN cur = SvCUR(sstr);
3335 STRLEN len = SvLEN(sstr);
f8c7b90f 3336#ifdef PERL_OLD_COPY_ON_WRITE
765f542d 3337 if (len) {
b8f9541a 3338 assert (SvTYPE(dstr) >= SVt_PVIV);
765f542d
NC
3339 /* SvIsCOW_normal */
3340 /* splice us in between source and next-after-source. */
a29f6d03
NC
3341 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3342 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3343 SvPV_set(dstr, SvPVX_mutable(sstr));
a604c751
NC
3344 } else
3345#endif
3346 {
765f542d 3347 /* SvIsCOW_shared_hash */
46187eeb
NC
3348 DEBUG_C(PerlIO_printf(Perl_debug_log,
3349 "Copy on write: Sharing hash\n"));
b8f9541a 3350
bdd68bc3 3351 assert (SvTYPE(dstr) >= SVt_PV);
765f542d 3352 SvPV_set(dstr,
d1db91c6 3353 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
bdd68bc3 3354 }
87a1ef3d
SP
3355 SvLEN_set(dstr, len);
3356 SvCUR_set(dstr, cur);
765f542d
NC
3357 SvREADONLY_on(dstr);
3358 SvFAKE_on(dstr);
3359 /* Relesase a global SV mutex. */
3360 }
3361 else
765f542d 3362 { /* Passes the swipe test. */
78d1e721 3363 SvPV_set(dstr, SvPVX_mutable(sstr));
765f542d
NC
3364 SvLEN_set(dstr, SvLEN(sstr));
3365 SvCUR_set(dstr, SvCUR(sstr));
3366
3367 SvTEMP_off(dstr);
3368 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
3369 SvPV_set(sstr, Nullch);
3370 SvLEN_set(sstr, 0);
3371 SvCUR_set(sstr, 0);
3372 SvTEMP_off(sstr);
3373 }
3374 }
9aa983d2 3375 if (sflags & SVf_UTF8)
a7cb1f99 3376 SvUTF8_on(dstr);
8990e307 3377 if (sflags & SVp_NOK) {
3332b3c1
JH
3378 SvNOKp_on(dstr);
3379 if (sflags & SVf_NOK)
3380 SvFLAGS(dstr) |= SVf_NOK;
9d6ce603 3381 SvNV_set(dstr, SvNVX(sstr));
79072805 3382 }
8990e307 3383 if (sflags & SVp_IOK) {
3332b3c1
JH
3384 (void)SvIOKp_on(dstr);
3385 if (sflags & SVf_IOK)
3386 SvFLAGS(dstr) |= SVf_IOK;
2b1c7e3e 3387 if (sflags & SVf_IVisUV)
25da4f38 3388 SvIsUV_on(dstr);
45977657 3389 SvIV_set(dstr, SvIVX(sstr));
79072805 3390 }
92f0c265 3391 if (SvVOK(sstr)) {
00b6aa41 3392 const MAGIC * const smg = mg_find(sstr,PERL_MAGIC_vstring);
ece467f9
JP
3393 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
3394 smg->mg_ptr, smg->mg_len);
439cb1c4 3395 SvRMAGICAL_on(dstr);
7a5fa8a2 3396 }
79072805 3397 }
8990e307 3398 else if (sflags & SVp_IOK) {
3332b3c1
JH
3399 if (sflags & SVf_IOK)
3400 (void)SvIOK_only(dstr);
3401 else {
9cbac4c7
DM
3402 (void)SvOK_off(dstr);
3403 (void)SvIOKp_on(dstr);
3332b3c1
JH
3404 }
3405 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
2b1c7e3e 3406 if (sflags & SVf_IVisUV)
25da4f38 3407 SvIsUV_on(dstr);
45977657 3408 SvIV_set(dstr, SvIVX(sstr));
3332b3c1
JH
3409 if (sflags & SVp_NOK) {
3410 if (sflags & SVf_NOK)
3411 (void)SvNOK_on(dstr);
3412 else
3413 (void)SvNOKp_on(dstr);
9d6ce603 3414 SvNV_set(dstr, SvNVX(sstr));
3332b3c1
JH
3415 }
3416 }
3417 else if (sflags & SVp_NOK) {
3418 if (sflags & SVf_NOK)
3419 (void)SvNOK_only(dstr);
3420 else {
9cbac4c7 3421 (void)SvOK_off(dstr);
3332b3c1
JH
3422 SvNOKp_on(dstr);
3423 }
9d6ce603 3424 SvNV_set(dstr, SvNVX(sstr));
79072805
LW
3425 }
3426 else {
20408e3c 3427 if (dtype == SVt_PVGV) {
e476b1b5 3428 if (ckWARN(WARN_MISC))
9014280d 3429 Perl_warner(aTHX_ packWARN(WARN_MISC), "Undefined value assigned to typeglob");
20408e3c
GS
3430 }
3431 else
3432 (void)SvOK_off(dstr);
a0d0e21e 3433 }
27c9684d
AP
3434 if (SvTAINTED(sstr))
3435 SvTAINT(dstr);
79072805
LW
3436}
3437
954c1994
GS
3438/*
3439=for apidoc sv_setsv_mg
3440
3441Like C<sv_setsv>, but also handles 'set' magic.
3442
3443=cut
3444*/
3445
79072805 3446void
864dbfa3 3447Perl_sv_setsv_mg(pTHX_ SV *dstr, register SV *sstr)
ef50df4b
GS
3448{
3449 sv_setsv(dstr,sstr);
3450 SvSETMAGIC(dstr);
3451}
3452
f8c7b90f 3453#ifdef PERL_OLD_COPY_ON_WRITE
ed252734
NC
3454SV *
3455Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
3456{
3457 STRLEN cur = SvCUR(sstr);
3458 STRLEN len = SvLEN(sstr);
3459 register char *new_pv;
3460
3461 if (DEBUG_C_TEST) {
3462 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
3463 sstr, dstr);
3464 sv_dump(sstr);
3465 if (dstr)
3466 sv_dump(dstr);
3467 }
3468
3469 if (dstr) {
3470 if (SvTHINKFIRST(dstr))
3471 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
3f7c398e
SP
3472 else if (SvPVX_const(dstr))
3473 Safefree(SvPVX_const(dstr));
ed252734
NC
3474 }
3475 else
3476 new_SV(dstr);
862a34c6 3477 SvUPGRADE(dstr, SVt_PVIV);
ed252734
NC
3478
3479 assert (SvPOK(sstr));
3480 assert (SvPOKp(sstr));
3481 assert (!SvIOK(sstr));
3482 assert (!SvIOKp(sstr));
3483 assert (!SvNOK(sstr));
3484 assert (!SvNOKp(sstr));
3485
3486 if (SvIsCOW(sstr)) {
3487
3488 if (SvLEN(sstr) == 0) {
3489 /* source is a COW shared hash key. */
ed252734
NC
3490 DEBUG_C(PerlIO_printf(Perl_debug_log,
3491 "Fast copy on write: Sharing hash\n"));
d1db91c6 3492 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
ed252734
NC
3493 goto common_exit;
3494 }
3495 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
3496 } else {
3497 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
862a34c6 3498 SvUPGRADE(sstr, SVt_PVIV);
ed252734
NC
3499 SvREADONLY_on(sstr);
3500 SvFAKE_on(sstr);
3501 DEBUG_C(PerlIO_printf(Perl_debug_log,
3502 "Fast copy on write: Converting sstr to COW\n"));
3503 SV_COW_NEXT_SV_SET(dstr, sstr);
3504 }
3505 SV_COW_NEXT_SV_SET(sstr, dstr);
940132f3 3506 new_pv = SvPVX_mutable(sstr);
ed252734
NC
3507
3508 common_exit:
3509 SvPV_set(dstr, new_pv);
3510 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
3511 if (SvUTF8(sstr))
3512 SvUTF8_on(dstr);
87a1ef3d
SP
3513 SvLEN_set(dstr, len);
3514 SvCUR_set(dstr, cur);
ed252734
NC
3515 if (DEBUG_C_TEST) {
3516 sv_dump(dstr);
3517 }
3518 return dstr;
3519}
3520#endif
3521
954c1994
GS
3522/*
3523=for apidoc sv_setpvn
3524
3525Copies a string into an SV. The C<len> parameter indicates the number of
9e09f5f2
MHM
3526bytes to be copied. If the C<ptr> argument is NULL the SV will become
3527undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
954c1994
GS
3528
3529=cut
3530*/
3531
ef50df4b 3532void
864dbfa3 3533Perl_sv_setpvn(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
79072805 3534{
c6f8c383 3535 register char *dptr;
22c522df 3536
765f542d 3537 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3538 if (!ptr) {
a0d0e21e 3539 (void)SvOK_off(sv);
463ee0b2
LW
3540 return;
3541 }
22c522df
JH
3542 else {
3543 /* len is STRLEN which is unsigned, need to copy to signed */
a3b680e6 3544 const IV iv = len;
9c5ffd7c
JH
3545 if (iv < 0)
3546 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
22c522df 3547 }
862a34c6 3548 SvUPGRADE(sv, SVt_PV);
c6f8c383 3549
5902b6a9 3550 dptr = SvGROW(sv, len + 1);
c6f8c383
GA
3551 Move(ptr,dptr,len,char);
3552 dptr[len] = '\0';
79072805 3553 SvCUR_set(sv, len);
1aa99e6b 3554 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3555 SvTAINT(sv);
79072805
LW
3556}
3557
954c1994
GS
3558/*
3559=for apidoc sv_setpvn_mg
3560
3561Like C<sv_setpvn>, but also handles 'set' magic.
3562
3563=cut
3564*/
3565
79072805 3566void
864dbfa3 3567Perl_sv_setpvn_mg(pTHX_ register SV *sv, register const char *ptr, register STRLEN len)
ef50df4b
GS
3568{
3569 sv_setpvn(sv,ptr,len);
3570 SvSETMAGIC(sv);
3571}
3572
954c1994
GS
3573/*
3574=for apidoc sv_setpv
3575
3576Copies a string into an SV. The string must be null-terminated. Does not
3577handle 'set' magic. See C<sv_setpv_mg>.
3578
3579=cut
3580*/
3581
ef50df4b 3582void
864dbfa3 3583Perl_sv_setpv(pTHX_ register SV *sv, register const char *ptr)
79072805
LW
3584{
3585 register STRLEN len;
3586
765f542d 3587 SV_CHECK_THINKFIRST_COW_DROP(sv);
463ee0b2 3588 if (!ptr) {
a0d0e21e 3589 (void)SvOK_off(sv);
463ee0b2
LW
3590 return;
3591 }
79072805 3592 len = strlen(ptr);
862a34c6 3593 SvUPGRADE(sv, SVt_PV);
c6f8c383 3594
79072805 3595 SvGROW(sv, len + 1);
463ee0b2 3596 Move(ptr,SvPVX(sv),len+1,char);
79072805 3597 SvCUR_set(sv, len);
1aa99e6b 3598 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2
LW
3599 SvTAINT(sv);
3600}
3601
954c1994
GS
3602/*
3603=for apidoc sv_setpv_mg
3604
3605Like C<sv_setpv>, but also handles 'set' magic.
3606
3607=cut
3608*/
3609
463ee0b2 3610void
864dbfa3 3611Perl_sv_setpv_mg(pTHX_ register SV *sv, register const char *ptr)
ef50df4b
GS
3612{
3613 sv_setpv(sv,ptr);
3614 SvSETMAGIC(sv);
3615}
3616
954c1994
GS
3617/*
3618=for apidoc sv_usepvn
3619
3620Tells an SV to use C<ptr> to find its string value. Normally the string is
1c846c1f 3621stored inside the SV but sv_usepvn allows the SV to use an outside string.
954c1994
GS
3622The C<ptr> should point to memory that was allocated by C<malloc>. The
3623string length, C<len>, must be supplied. This function will realloc the
3624memory pointed to by C<ptr>, so that pointer should not be freed or used by
3625the programmer after giving it to sv_usepvn. Does not handle 'set' magic.
3626See C<sv_usepvn_mg>.
3627
3628=cut
3629*/
3630
ef50df4b 3631void
864dbfa3 3632Perl_sv_usepvn(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
463ee0b2 3633{
1936d2a7 3634 STRLEN allocate;
765f542d 3635 SV_CHECK_THINKFIRST_COW_DROP(sv);
862a34c6 3636 SvUPGRADE(sv, SVt_PV);
463ee0b2 3637 if (!ptr) {
a0d0e21e 3638 (void)SvOK_off(sv);
463ee0b2
LW
3639 return;
3640 }
3f7c398e 3641 if (SvPVX_const(sv))
8bd4d4c5 3642 SvPV_free(sv);
1936d2a7
NC
3643
3644 allocate = PERL_STRLEN_ROUNDUP(len + 1);
7a9b70e9 3645 ptr = saferealloc (ptr, allocate);
f880fe2f 3646 SvPV_set(sv, ptr);
463ee0b2 3647 SvCUR_set(sv, len);
1936d2a7 3648 SvLEN_set(sv, allocate);
463ee0b2 3649 *SvEND(sv) = '\0';
1aa99e6b 3650 (void)SvPOK_only_UTF8(sv); /* validate pointer */
463ee0b2 3651 SvTAINT(sv);
79072805
LW
3652}
3653
954c1994
GS
3654/*
3655=for apidoc sv_usepvn_mg
3656
3657Like C<sv_usepvn>, but also handles 'set' magic.
3658
3659=cut
3660*/
3661
ef50df4b 3662void
864dbfa3 3663Perl_sv_usepvn_mg(pTHX_ register SV *sv, register char *ptr, register STRLEN len)
ef50df4b 3664{
51c1089b 3665 sv_usepvn(sv,ptr,len);
ef50df4b
GS
3666 SvSETMAGIC(sv);
3667}
3668
f8c7b90f 3669#ifdef PERL_OLD_COPY_ON_WRITE
765f542d
NC
3670/* Need to do this *after* making the SV normal, as we need the buffer
3671 pointer to remain valid until after we've copied it. If we let go too early,
3672 another thread could invalidate it by unsharing last of the same hash key
3673 (which it can do by means other than releasing copy-on-write Svs)
3674 or by changing the other copy-on-write SVs in the loop. */
3675STATIC void
bdd68bc3 3676S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, STRLEN len, SV *after)
765f542d
NC
3677{
3678 if (len) { /* this SV was SvIsCOW_normal(sv) */
3679 /* we need to find the SV pointing to us. */
aec46f14 3680 SV * const current = SV_COW_NEXT_SV(after);
7a5fa8a2 3681
765f542d
NC
3682 if (current == sv) {
3683 /* The SV we point to points back to us (there were only two of us
3684 in the loop.)
3685 Hence other SV is no longer copy on write either. */
3686 SvFAKE_off(after);
3687 SvREADONLY_off(after);
3688 } else {
3689 /* We need to follow the pointers around the loop. */
3690 SV *next;
3691 while ((next = SV_COW_NEXT_SV(current)) != sv) {
3692 assert (next);
3693 current = next;
3694 /* don't loop forever if the structure is bust, and we have
3695 a pointer into a closed loop. */
3696 assert (current != after);
3f7c398e 3697 assert (SvPVX_const(current) == pvx);
765f542d
NC
3698 }
3699 /* Make the SV before us point to the SV after us. */
a29f6d03 3700 SV_COW_NEXT_SV_SET(current, after);
765f542d
NC
3701 }
3702 } else {
bdd68bc3 3703 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
765f542d
NC
3704 }
3705}
3706
3707int
3708Perl_sv_release_IVX(pTHX_ register SV *sv)
3709{
3710 if (SvIsCOW(sv))
3711 sv_force_normal_flags(sv, 0);
0c34ef67
MHM
3712 SvOOK_off(sv);
3713 return 0;
765f542d
NC
3714}
3715#endif
645c22ef
DM
3716/*
3717=for apidoc sv_force_normal_flags
3718
3719Undo various types of fakery on an SV: if the PV is a shared string, make
3720a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
765f542d
NC
3721an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
3722we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
3723then a copy-on-write scalar drops its PV buffer (if any) and becomes
3724SvPOK_off rather than making a copy. (Used where this scalar is about to be
d3050d9d 3725set to some other value.) In addition, the C<flags> parameter gets passed to
765f542d
NC