ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/CBOR-XS/XS.xs
(Generate patch)

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.5 by root, Sat Oct 26 21:14:20 2013 UTC vs.
Revision 1.18 by root, Wed Nov 20 01:09:46 2013 UTC

9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11 11
12#include "ecb.h" 12#include "ecb.h"
13 13
14// compatibility with perl <5.18
15#ifndef HvNAMELEN_get
16# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
17#endif
18#ifndef HvNAMELEN
19# define HvNAMELEN(hv) HvNAMELEN_get (hv)
20#endif
21#ifndef HvNAMEUTF8
22# define HvNAMEUTF8(hv) 0
23#endif
24
25// known tags
26enum cbor_tag
27{
28 // inofficial extensions (pending iana registration)
29 CBOR_TAG_PERL_OBJECT = 24, // http://cbor.schmorp.de/perl-object
30 CBOR_TAG_GENERIC_OBJECT = 25, // http://cbor.schmorp.de/generic-object
31 CBOR_TAG_VALUE_SHARABLE = 26, // http://cbor.schmorp.de/value-sharing
32 CBOR_TAG_VALUE_SHAREDREF = 27, // http://cbor.schmorp.de/value-sharing
33 CBOR_TAG_STRINGREF_NAMESPACE = 65537, // http://cbor.schmorp.de/stringref
34 CBOR_TAG_STRINGREF = 28, // http://cbor.schmorp.de/stringref
35 CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
36
37 // rfc7049
38 CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
39 CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
40 CBOR_TAG_POS_BIGNUM = 2, // byte string
41 CBOR_TAG_NEG_BIGNUM = 3, // byte string
42 CBOR_TAG_DECIMAL = 4, // decimal fraction, array
43 CBOR_TAG_BIGFLOAT = 5, // array
44
45 CBOR_TAG_CONV_B64U = 21, // base64url, any
46 CBOR_TAG_CONV_B64 = 22, // base64, any
47 CBOR_TAG_CONV_HEX = 23, // base16, any
48 CBOR_TAG_CBOR = 24, // embedded cbor, byte string
49
50 CBOR_TAG_URI = 32, // URI rfc3986, utf-8
51 CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
52 CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
53 CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
54 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
55
56 CBOR_TAG_MAGIC = 55799 // self-describe cbor
57};
58
14#define F_SHRINK 0x00000200UL 59#define F_SHRINK 0x00000001UL
15#define F_ALLOW_UNKNOWN 0x00002000UL 60#define F_ALLOW_UNKNOWN 0x00000002UL
61#define F_ALLOW_SHARING 0x00000004UL //TODO
62#define F_DEDUP_STRINGS 0x00000008UL //TODO
63#define F_DEDUP_KEYS 0x00000010UL //TODO
16 64
17#define INIT_SIZE 32 // initial scalar size to be allocated 65#define INIT_SIZE 32 // initial scalar size to be allocated
18 66
19#define SB do { 67#define SB do {
20#define SE } while (0) 68#define SE } while (0)
31#else 79#else
32# define CBOR_SLOW 0 80# define CBOR_SLOW 0
33# define CBOR_STASH cbor_stash 81# define CBOR_STASH cbor_stash
34#endif 82#endif
35 83
36static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: 84static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
37static SV *cbor_true, *cbor_false; 85static SV *types_true, *types_false, *types_error, *sv_cbor;
38 86
39typedef struct { 87typedef struct {
40 U32 flags; 88 U32 flags;
41 U32 max_depth; 89 U32 max_depth;
42 STRLEN max_size; 90 STRLEN max_size;
43
44 SV *cb_object;
45 HV *cb_sk_object;
46} CBOR; 91} CBOR;
47 92
48ecb_inline void 93ecb_inline void
49cbor_init (CBOR *cbor) 94cbor_init (CBOR *cbor)
50{ 95{
80#endif 125#endif
81 } 126 }
82} 127}
83 128
84///////////////////////////////////////////////////////////////////////////// 129/////////////////////////////////////////////////////////////////////////////
85// fp hell
86
87//TODO
88
89/////////////////////////////////////////////////////////////////////////////
90// encoder 130// encoder
91 131
92// structure used for encoding CBOR 132// structure used for encoding CBOR
93typedef struct 133typedef struct
94{ 134{
95 char *cur; // SvPVX (sv) + current output position 135 char *cur; // SvPVX (sv) + current output position
96 char *end; // SvEND (sv) 136 char *end; // SvEND (sv)
97 SV *sv; // result scalar 137 SV *sv; // result scalar
98 CBOR cbor; 138 CBOR cbor;
99 U32 depth; // recursion level 139 U32 depth; // recursion level
140 HV *stringref; // string => index, or 0
141 HV *sharable; // ptr => index, or 0
142 HV *sharable_idx;
100} enc_t; 143} enc_t;
101 144
102ecb_inline void 145ecb_inline void
103need (enc_t *enc, STRLEN len) 146need (enc_t *enc, STRLEN len)
104{ 147{
165 need (enc, len); 208 need (enc, len);
166 memcpy (enc->cur, str, len); 209 memcpy (enc->cur, str, len);
167 enc->cur += len; 210 enc->cur += len;
168} 211}
169 212
213ecb_inline void
214encode_tag (enc_t *enc, UV tag)
215{
216 encode_uint (enc, 0xc0, tag);
217}
218
219static int
220encode_sharable2 (enc_t *enc, SV *sv)
221{
222 if (!enc->sharable)
223 enc->sharable = (HV *)sv_2mortal ((SV *)newHV ());
224
225 SV **svp = hv_fetch (enc->sharable, &sv, sizeof (sv), 1);
226
227 if (SvOK (*svp))
228 {
229 encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
230 encode_uint (enc, 0x00, SvUV (*svp));
231
232 return 1;
233 }
234 else
235 {
236 sv_setuv (*svp, enc->sharable_idx++);
237 encode_tag (enc, CBOR_TAG_VALUE_SHARABLE);
238
239 return 0;
240 }
241}
242
243ecb_inline int
244encode_sharable (enc_t *enc, SV *sv)
245{
246 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)
247 && ecb_expect_false (SvREFCNT (sv) > 1))
248 return encode_sharable2 (enc, sv);
249
250 return 0;
251}
252
170static void encode_sv (enc_t *enc, SV *sv); 253static void encode_sv (enc_t *enc, SV *sv);
171 254
172static void 255static void
173encode_av (enc_t *enc, AV *av) 256encode_av (enc_t *enc, AV *av)
174{ 257{
231 svtype svt; 314 svtype svt;
232 315
233 SvGETMAGIC (sv); 316 SvGETMAGIC (sv);
234 svt = SvTYPE (sv); 317 svt = SvTYPE (sv);
235 318
319 if (encode_sharable (enc, sv))
320 return;
321
236 if (ecb_expect_false (SvOBJECT (sv))) 322 if (ecb_expect_false (SvOBJECT (sv)))
237 { 323 {
238 HV *stash = !CBOR_SLOW || cbor_boolean_stash 324 HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
239 ? cbor_boolean_stash 325 ? types_boolean_stash
326 : gv_stashpv ("Types::Serialiser::Boolean", 1);
327 HV *error_stash = !CBOR_SLOW || types_error_stash
328 ? types_error_stash
329 : gv_stashpv ("Types::Serialiser::Error", 1);
330 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
331 ? cbor_tagged_stash
240 : gv_stashpv ("CBOR::XS::Boolean", 1); 332 : gv_stashpv ("CBOR::XS::Tagged" , 1);
241 333
242 if (SvSTASH (sv) == stash) 334 HV *stash = SvSTASH (sv);
335 GV *method;
336
337 if (stash == boolean_stash)
243 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); 338 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);
339 else if (stash == error_stash)
340 encode_ch (enc, 0xe0 | 23);
341 else if (stash == tagged_stash)
342 {
343 if (svt != SVt_PVAV)
344 croak ("encountered CBOR::XS::Tagged object that isn't an array");
345
346 encode_uint (enc, 0xc0, SvUV (*av_fetch ((AV *)sv, 0, 1)));
347 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
348 }
349 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
350 {
351 dSP;
352
353 ENTER; SAVETMPS; PUSHMARK (SP);
354 // we re-bless the reference to get overload and other niceties right
355 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
356
357 PUTBACK;
358 // G_SCALAR ensures that return value is 1
359 call_sv ((SV *)GvCV (method), G_SCALAR);
360 SPAGAIN;
361
362 // catch this surprisingly common error
363 if (SvROK (TOPs) && SvRV (TOPs) == sv)
364 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
365
366 encode_sv (enc, POPs);
367
368 PUTBACK;
369
370 FREETMPS; LEAVE;
371 }
372 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
373 {
374 dSP;
375
376 ENTER; SAVETMPS; PUSHMARK (SP);
377 EXTEND (SP, 2);
378 // we re-bless the reference to get overload and other niceties right
379 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
380 PUSHs (sv_cbor);
381
382 PUTBACK;
383 int count = call_sv ((SV *)GvCV (method), G_ARRAY);
384 SPAGAIN;
385
386 // catch this surprisingly common error
387 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
388 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
389
390 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
391 encode_uint (enc, 0x80, count + 1);
392 encode_str (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
393
394 while (count)
395 encode_sv (enc, SP[1 - count--]);
396
397 PUTBACK;
398
399 FREETMPS; LEAVE;
400 }
244 else 401 else
245 {
246#if 0 //TODO
247 if (enc->cbor.flags & F_CONV_BLESSED)
248 {
249 // we re-bless the reference to get overload and other niceties right
250 GV *to_cbor = gv_fetchmethod_autoload (SvSTASH (sv), "TO_CBOR", 0);
251
252 if (to_cbor)
253 {
254 dSP;
255
256 ENTER; SAVETMPS; PUSHMARK (SP);
257 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
258
259 // calling with G_SCALAR ensures that we always get a 1 return value
260 PUTBACK;
261 call_sv ((SV *)GvCV (to_cbor), G_SCALAR);
262 SPAGAIN;
263
264 // catch this surprisingly common error
265 if (SvROK (TOPs) && SvRV (TOPs) == sv)
266 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
267
268 sv = POPs;
269 PUTBACK;
270
271 encode_sv (enc, sv);
272
273 FREETMPS; LEAVE;
274 }
275 else if (enc->cbor.flags & F_ALLOW_BLESSED)
276 encode_str (enc, "null", 4, 0);
277 else
278 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it", 402 croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
279 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
280 }
281 else if (enc->cbor.flags & F_ALLOW_BLESSED)
282 encode_str (enc, "null", 4, 0);
283 else
284 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
285 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 403 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
286#endif
287 }
288 } 404 }
289 else if (svt == SVt_PVHV) 405 else if (svt == SVt_PVHV)
290 encode_hv (enc, (HV *)sv); 406 encode_hv (enc, (HV *)sv);
291 else if (svt == SVt_PVAV) 407 else if (svt == SVt_PVAV)
292 encode_av (enc, (AV *)sv); 408 encode_av (enc, (AV *)sv);
293 else if (svt < SVt_PVAV)
294 {
295 STRLEN len = 0;
296 char *pv = svt ? SvPV (sv, len) : 0;
297
298 if (len == 1 && *pv == '1')
299 encode_ch (enc, 0xe0 | 21);
300 else if (len == 1 && *pv == '0')
301 encode_ch (enc, 0xe0 | 20);
302 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
303 encode_ch (enc, 0xe0 | 23);
304 else
305 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
306 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
307 }
308 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
309 encode_ch (enc, 0xe0 | 23);
310 else 409 else
311 croak ("encountered %s, but CBOR can only represent references to arrays or hashes", 410 {
312 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 411 encode_tag (enc, CBOR_TAG_INDIRECTION);
412 encode_sv (enc, sv);
413 }
313} 414}
314 415
315static void 416static void
316encode_nv (enc_t *enc, SV *sv) 417encode_nv (enc_t *enc, SV *sv)
317{ 418{
350 451
351static void 452static void
352encode_sv (enc_t *enc, SV *sv) 453encode_sv (enc_t *enc, SV *sv)
353{ 454{
354 SvGETMAGIC (sv); 455 SvGETMAGIC (sv);
456
457 if (encode_sharable (enc, sv))
458 return;
355 459
356 if (SvPOKp (sv)) 460 if (SvPOKp (sv))
357 { 461 {
358 STRLEN len; 462 STRLEN len;
359 char *str = SvPV (sv, len); 463 char *str = SvPV (sv, len);
382} 486}
383 487
384static SV * 488static SV *
385encode_cbor (SV *scalar, CBOR *cbor) 489encode_cbor (SV *scalar, CBOR *cbor)
386{ 490{
387 enc_t enc; 491 enc_t enc = { };
388 492
389 enc.cbor = *cbor; 493 enc.cbor = *cbor;
390 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 494 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
391 enc.cur = SvPVX (enc.sv); 495 enc.cur = SvPVX (enc.sv);
392 enc.end = SvEND (enc.sv); 496 enc.end = SvEND (enc.sv);
393 enc.depth = 0;
394 497
395 SvPOK_only (enc.sv); 498 SvPOK_only (enc.sv);
396 encode_sv (&enc, scalar); 499 encode_sv (&enc, scalar);
397 500
398 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 501 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
414 U8 *end; // end of input string 517 U8 *end; // end of input string
415 const char *err; // parse error, if != 0 518 const char *err; // parse error, if != 0
416 CBOR cbor; 519 CBOR cbor;
417 U32 depth; // recursion depth 520 U32 depth; // recursion depth
418 U32 maxdepth; // recursion depth limit 521 U32 maxdepth; // recursion depth limit
522 AV *sharable;
419} dec_t; 523} dec_t;
420 524
421#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 525#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
422 526
423#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data") 527#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
517 SvREFCNT_dec (av); 621 SvREFCNT_dec (av);
518 DEC_DEC_DEPTH; 622 DEC_DEC_DEPTH;
519 return &PL_sv_undef; 623 return &PL_sv_undef;
520} 624}
521 625
626static void
627decode_he (dec_t *dec, HV *hv)
628{
629 // for speed reasons, we specialcase single-string
630 // byte or utf-8 strings as keys.
631
632 if (*dec->cur >= 0x40 && *dec->cur <= 0x40 + 27)
633 {
634 I32 len = decode_uint (dec);
635 char *key = (char *)dec->cur;
636
637 dec->cur += len;
638
639 hv_store (hv, key, len, decode_sv (dec), 0);
640 }
641 else if (*dec->cur >= 0x60 && *dec->cur <= 0x60 + 27)
642 {
643 I32 len = decode_uint (dec);
644 char *key = (char *)dec->cur;
645
646 dec->cur += len;
647
648 hv_store (hv, key, -len, decode_sv (dec), 0);
649 }
650 else
651 {
652 SV *k = decode_sv (dec);
653 SV *v = decode_sv (dec);
654
655 hv_store_ent (hv, k, v, 0);
656 SvREFCNT_dec (k);
657 }
658}
659
522static SV * 660static SV *
523decode_hv (dec_t *dec) 661decode_hv (dec_t *dec)
524{ 662{
525 HV *hv = newHV (); 663 HV *hv = newHV ();
526 664
538 { 676 {
539 ++dec->cur; 677 ++dec->cur;
540 break; 678 break;
541 } 679 }
542 680
543 SV *k = decode_sv (dec); 681 decode_he (dec, hv);
544 SV *v = decode_sv (dec);
545
546 hv_store_ent (hv, k, v, 0);
547 } 682 }
548 } 683 }
549 else 684 else
550 { 685 {
551 int len = decode_uint (dec); 686 int pairs = decode_uint (dec);
552 687
553 while (len--) 688 while (pairs--)
554 { 689 decode_he (dec, hv);
555 SV *k = decode_sv (dec);
556 SV *v = decode_sv (dec);
557
558 hv_store_ent (hv, k, v, 0);
559 }
560 } 690 }
561 691
562 DEC_DEC_DEPTH; 692 DEC_DEC_DEPTH;
563 return newRV_noinc ((SV *)hv); 693 return newRV_noinc ((SV *)hv);
564
565#if 0
566 SV *sv;
567 HV *hv = newHV ();
568
569 DEC_INC_DEPTH;
570 decode_ws (dec);
571
572 for (;;)
573 {
574 // heuristic: assume that
575 // a) decode_str + hv_store_ent are abysmally slow.
576 // b) most hash keys are short, simple ascii text.
577 // => try to "fast-match" such strings to avoid
578 // the overhead of decode_str + hv_store_ent.
579 {
580 SV *value;
581 char *p = dec->cur;
582 char *e = p + 24; // only try up to 24 bytes
583
584 for (;;)
585 {
586 // the >= 0x80 is false on most architectures
587 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
588 {
589 // slow path, back up and use decode_str
590 SV *key = decode_str (dec);
591 if (!key)
592 goto fail;
593
594 decode_ws (dec); EXPECT_CH (':');
595
596 decode_ws (dec);
597 value = decode_sv (dec);
598 if (!value)
599 {
600 SvREFCNT_dec (key);
601 goto fail;
602 }
603
604 hv_store_ent (hv, key, value, 0);
605 SvREFCNT_dec (key);
606
607 break;
608 }
609 else if (*p == '"')
610 {
611 // fast path, got a simple key
612 char *key = dec->cur;
613 int len = p - key;
614 dec->cur = p + 1;
615
616 decode_ws (dec); EXPECT_CH (':');
617
618 decode_ws (dec);
619 value = decode_sv (dec);
620 if (!value)
621 goto fail;
622
623 hv_store (hv, key, len, value, 0);
624
625 break;
626 }
627
628 ++p;
629 }
630 }
631
632 decode_ws (dec);
633
634 if (*dec->cur == '}')
635 {
636 ++dec->cur;
637 break;
638 }
639
640 if (*dec->cur != ',')
641 ERR (", or } expected while parsing object/hash");
642
643 ++dec->cur;
644
645 decode_ws (dec);
646
647 if (*dec->cur == '}' && dec->cbor.flags & F_RELAXED)
648 {
649 ++dec->cur;
650 break;
651 }
652 }
653
654 DEC_DEC_DEPTH;
655 sv = newRV_noinc ((SV *)hv);
656
657 // check filter callbacks
658 if (dec->cbor.flags & F_HOOK)
659 {
660 if (dec->cbor.cb_sk_object && HvKEYS (hv) == 1)
661 {
662 HE *cb, *he;
663
664 hv_iterinit (hv);
665 he = hv_iternext (hv);
666 hv_iterinit (hv);
667
668 // the next line creates a mortal sv each time its called.
669 // might want to optimise this for common cases.
670 cb = hv_fetch_ent (dec->cbor.cb_sk_object, hv_iterkeysv (he), 0, 0);
671
672 if (cb)
673 {
674 dSP;
675 int count;
676
677 ENTER; SAVETMPS; PUSHMARK (SP);
678 XPUSHs (HeVAL (he));
679 sv_2mortal (sv);
680
681 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
682
683 if (count == 1)
684 {
685 sv = newSVsv (POPs);
686 FREETMPS; LEAVE;
687 return sv;
688 }
689
690 SvREFCNT_inc (sv);
691 FREETMPS; LEAVE;
692 }
693 }
694
695 if (dec->cbor.cb_object)
696 {
697 dSP;
698 int count;
699
700 ENTER; SAVETMPS; PUSHMARK (SP);
701 XPUSHs (sv_2mortal (sv));
702
703 PUTBACK; count = call_sv (dec->cbor.cb_object, G_ARRAY); SPAGAIN;
704
705 if (count == 1)
706 {
707 sv = newSVsv (POPs);
708 FREETMPS; LEAVE;
709 return sv;
710 }
711
712 SvREFCNT_inc (sv);
713 FREETMPS; LEAVE;
714 }
715 }
716
717 return sv;
718#endif
719 694
720fail: 695fail:
721 SvREFCNT_dec (hv); 696 SvREFCNT_dec (hv);
722 DEC_DEC_DEPTH; 697 DEC_DEC_DEPTH;
723 return &PL_sv_undef; 698 return &PL_sv_undef;
724} 699}
725 700
726static SV * 701static SV *
727decode_str (dec_t *dec, int utf8) 702decode_str (dec_t *dec, int utf8)
728{ 703{
729 SV *sv; 704 SV *sv = 0;
730 705
731 if ((*dec->cur & 31) == 31) 706 if ((*dec->cur & 31) == 31)
732 { 707 {
733 ++dec->cur; 708 ++dec->cur;
734 709
743 { 718 {
744 ++dec->cur; 719 ++dec->cur;
745 break; 720 break;
746 } 721 }
747 722
748 SV *sv2 = decode_sv (dec);
749 sv_catsv (sv, sv2); 723 sv_catsv (sv, decode_sv (dec));
750 } 724 }
751 } 725 }
752 else 726 else
753 { 727 {
754 STRLEN len = decode_uint (dec); 728 STRLEN len = decode_uint (dec);
762 SvUTF8_on (sv); 736 SvUTF8_on (sv);
763 737
764 return sv; 738 return sv;
765 739
766fail: 740fail:
741 SvREFCNT_dec (sv);
767 return &PL_sv_undef; 742 return &PL_sv_undef;
768} 743}
769 744
770static SV * 745static SV *
771decode_tagged (dec_t *dec) 746decode_tagged (dec_t *dec)
772{ 747{
773 UV tag = decode_uint (dec); 748 UV tag = decode_uint (dec);
774 SV *sv = decode_sv (dec); 749 SV *sv = decode_sv (dec);
775 750
776 if (tag == 55799) // 2.4.5 Self-Describe CBOR 751 switch (tag)
752 {
753 case CBOR_TAG_MAGIC:
777 return sv; 754 return sv;
778 755
756 case CBOR_TAG_INDIRECTION:
757 return newRV_noinc (sv);
758
759 case CBOR_TAG_VALUE_SHARABLE:
760 if (ecb_expect_false (!dec->sharable))
761 dec->sharable = (AV *)sv_2mortal ((SV *)newAV ());
762
763 av_push (dec->sharable, SvREFCNT_inc_NN (sv));
764
765 return sv;
766
767 case CBOR_TAG_VALUE_SHAREDREF:
768 {
769 // TODO: should verify that the sv atcually was a CBOR unsigned integer
770 UV idx = SvUV (sv);
771
772 if (!dec->sharable || idx > AvFILLp (dec->sharable))
773 ERR ("corrupted CBOR data (sharedref index out of bounds)");
774
775 SvREFCNT_dec (sv);
776
777 return SvREFCNT_inc_NN (AvARRAY (dec->sharable)[idx]);
778 }
779
780 case CBOR_TAG_PERL_OBJECT:
781 {
782 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
783 ERR ("corrupted CBOR data (non-array perl object)");
784
785 AV *av = (AV *)SvRV (sv);
786 int len = av_len (av) + 1;
787 HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
788
789 if (!stash)
790 ERR ("cannot decode perl-object (package does not exist)");
791
792 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
793
794 if (!method)
795 ERR ("cannot decode perl-object (package does not have a THAW method)");
796
797 dSP;
798
799 ENTER; SAVETMPS; PUSHMARK (SP);
800 EXTEND (SP, len + 1);
801 // we re-bless the reference to get overload and other niceties right
802 PUSHs (*av_fetch (av, 0, 1));
803 PUSHs (sv_cbor);
804
805 int i;
806
807 for (i = 1; i < len; ++i)
808 PUSHs (*av_fetch (av, i, 1));
809
810 PUTBACK;
811 call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
812 SPAGAIN;
813
814 if (SvTRUE (ERRSV))
815 {
816 FREETMPS; LEAVE;
817 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
818 }
819
820 SvREFCNT_dec (sv);
821 sv = SvREFCNT_inc (POPs);
822
823 PUTBACK;
824
825 FREETMPS; LEAVE;
826
827 return sv;
828 }
829
830 default:
831 {
779 AV *av = newAV (); 832 AV *av = newAV ();
780 av_push (av, newSVuv (tag)); 833 av_push (av, newSVuv (tag));
781 av_push (av, sv); 834 av_push (av, sv);
782 return newRV_noinc ((SV *)av); 835
836 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
837 ? cbor_tagged_stash
838 : gv_stashpv ("CBOR::XS::Tagged" , 1);
839
840 return sv_bless (newRV_noinc ((SV *)av), tagged_stash);
841 }
842 }
843
844fail:
845 SvREFCNT_dec (sv);
846 return &PL_sv_undef;
783} 847}
784 848
785static SV * 849static SV *
786decode_sv (dec_t *dec) 850decode_sv (dec_t *dec)
787{ 851{
788 WANT (1); 852 WANT (1);
789 853
790 switch (*dec->cur >> 5) 854 switch (*dec->cur >> 5)
791 { 855 {
792 case 0: // unsigned int 856 case 0: // unsigned int
793 //TODO: 64 bit values on 3 2bit perls
794 return newSVuv (decode_uint (dec)); 857 return newSVuv (decode_uint (dec));
795 case 1: // negative int 858 case 1: // negative int
796 return newSViv (-1 - (IV)decode_uint (dec)); 859 return newSViv (-1 - (IV)decode_uint (dec));
797 case 2: // octet string 860 case 2: // octet string
798 return decode_str (dec, 0); 861 return decode_str (dec, 0);
807 case 7: // misc 870 case 7: // misc
808 switch (*dec->cur++ & 31) 871 switch (*dec->cur++ & 31)
809 { 872 {
810 case 20: 873 case 20:
811#if CBOR_SLOW 874#if CBOR_SLOW
812 cbor_false = get_bool ("CBOR::XS::false"); 875 types_false = get_bool ("Types::Serialiser::false");
813#endif 876#endif
814 return newSVsv (cbor_false); 877 return newSVsv (types_false);
815 case 21: 878 case 21:
816#if CBOR_SLOW 879#if CBOR_SLOW
817 cbor_true = get_bool ("CBOR::XS::true"); 880 types_true = get_bool ("Types::Serialiser::true");
818#endif 881#endif
819 return newSVsv (cbor_true); 882 return newSVsv (types_true);
820 case 22: 883 case 22:
821 return newSVsv (&PL_sv_undef); 884 return newSVsv (&PL_sv_undef);
885 case 23:
886#if CBOR_SLOW
887 types_error = get_bool ("Types::Serialiser::error");
888#endif
889 return newSVsv (types_error);
822 890
823 case 25: 891 case 25:
824 { 892 {
825 WANT (2); 893 WANT (2);
826 894
862 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); 930 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
863 } 931 }
864 932
865 break; 933 break;
866 } 934 }
867#if 0
868 switch (*dec->cur)
869 {
870 //case '"': ++dec->cur; return decode_str (dec);
871 case '[': ++dec->cur; return decode_av (dec);
872 case '{': ++dec->cur; return decode_hv (dec);
873
874 case '-':
875 case '0': case '1': case '2': case '3': case '4':
876 case '5': case '6': case '7': case '8': case '9':
877 //TODO return decode_num (dec);
878
879 case 't':
880 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
881 {
882 dec->cur += 4;
883#if CBOR_SLOW
884 cbor_true = get_bool ("CBOR::XS::true");
885#endif
886 return newSVsv (cbor_true);
887 }
888 else
889 ERR ("'true' expected");
890
891 break;
892
893 case 'f':
894 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
895 {
896 dec->cur += 5;
897#if CBOR_SLOW
898 cbor_false = get_bool ("CBOR::XS::false");
899#endif
900 return newSVsv (cbor_false);
901 }
902 else
903 ERR ("'false' expected");
904
905 break;
906
907 case 'n':
908 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
909 {
910 dec->cur += 4;
911 return newSVsv (&PL_sv_undef);
912 }
913 else
914 ERR ("'null' expected");
915
916 break;
917
918 default:
919 ERR ("malformed CBOR string, neither array, object, number, string or atom");
920 break;
921 }
922#endif
923 935
924fail: 936fail:
925 return &PL_sv_undef; 937 return &PL_sv_undef;
926} 938}
927 939
928static SV * 940static SV *
929decode_cbor (SV *string, CBOR *cbor, char **offset_return) 941decode_cbor (SV *string, CBOR *cbor, char **offset_return)
930{ 942{
931 dec_t dec; 943 dec_t dec = { };
932 SV *sv; 944 SV *sv;
945 STRLEN len;
946 char *data = SvPVbyte (string, len);
933 947
934 /* work around bugs in 5.10 where manipulating magic values
935 * makes perl ignore the magic in subsequent accesses.
936 * also make a copy of non-PV values, to get them into a clean
937 * state (SvPV should do that, but it's buggy, see below).
938 */
939 /*SvGETMAGIC (string);*/
940 if (SvMAGICAL (string) || !SvPOK (string))
941 string = sv_2mortal (newSVsv (string));
942
943 SvUPGRADE (string, SVt_PV);
944
945 /* work around a bug in perl 5.10, which causes SvCUR to fail an
946 * assertion with -DDEBUGGING, although SvCUR is documented to
947 * return the xpv_cur field which certainly exists after upgrading.
948 * according to nicholas clark, calling SvPOK fixes this.
949 * But it doesn't fix it, so try another workaround, call SvPV_nolen
950 * and hope for the best.
951 * Damnit, SvPV_nolen still trips over yet another assertion. This
952 * assertion business is seriously broken, try yet another workaround
953 * for the broken -DDEBUGGING.
954 */
955 {
956#ifdef DEBUGGING
957 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
958#else
959 STRLEN offset = SvCUR (string);
960#endif
961
962 if (offset > cbor->max_size && cbor->max_size) 948 if (len > cbor->max_size && cbor->max_size)
963 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu", 949 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
964 (unsigned long)SvCUR (string), (unsigned long)cbor->max_size); 950 (unsigned long)len, (unsigned long)cbor->max_size);
965 }
966
967 sv_utf8_downgrade (string, 0);
968 951
969 dec.cbor = *cbor; 952 dec.cbor = *cbor;
970 dec.cur = (U8 *)SvPVX (string); 953 dec.cur = (U8 *)data;
971 dec.end = (U8 *)SvEND (string); 954 dec.end = (U8 *)data + len;
972 dec.err = 0;
973 dec.depth = 0;
974
975 if (dec.cbor.cb_object || dec.cbor.cb_sk_object)
976 ;//TODO dec.cbor.flags |= F_HOOK;
977 955
978 sv = decode_sv (&dec); 956 sv = decode_sv (&dec);
979 957
980 if (offset_return) 958 if (offset_return)
981 *offset_return = dec.cur; 959 *offset_return = dec.cur;
985 dec.err = "garbage after CBOR object"; 963 dec.err = "garbage after CBOR object";
986 964
987 if (dec.err) 965 if (dec.err)
988 { 966 {
989 SvREFCNT_dec (sv); 967 SvREFCNT_dec (sv);
990 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); 968 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
991 } 969 }
992 970
993 sv = sv_2mortal (sv); 971 sv = sv_2mortal (sv);
994 972
995 return sv; 973 return sv;
1001MODULE = CBOR::XS PACKAGE = CBOR::XS 979MODULE = CBOR::XS PACKAGE = CBOR::XS
1002 980
1003BOOT: 981BOOT:
1004{ 982{
1005 cbor_stash = gv_stashpv ("CBOR::XS" , 1); 983 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1006 cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); 984 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1007 985
1008 cbor_true = get_bool ("CBOR::XS::true"); 986 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1009 cbor_false = get_bool ("CBOR::XS::false"); 987 types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
988
989 types_true = get_bool ("Types::Serialiser::true" );
990 types_false = get_bool ("Types::Serialiser::false");
991 types_error = get_bool ("Types::Serialiser::error");
992
993 sv_cbor = newSVpv ("CBOR", 0);
994 SvREADONLY_on (sv_cbor);
1010} 995}
1011 996
1012PROTOTYPES: DISABLE 997PROTOTYPES: DISABLE
1013 998
1014void CLONE (...) 999void CLONE (...)
1015 CODE: 1000 CODE:
1016 cbor_stash = 0; 1001 cbor_stash = 0;
1002 cbor_tagged_stash = 0;
1003 types_error_stash = 0;
1017 cbor_boolean_stash = 0; 1004 types_boolean_stash = 0;
1018 1005
1019void new (char *klass) 1006void new (char *klass)
1020 PPCODE: 1007 PPCODE:
1021{ 1008{
1022 SV *pv = NEWSV (0, sizeof (CBOR)); 1009 SV *pv = NEWSV (0, sizeof (CBOR));
1030 1017
1031void shrink (CBOR *self, int enable = 1) 1018void shrink (CBOR *self, int enable = 1)
1032 ALIAS: 1019 ALIAS:
1033 shrink = F_SHRINK 1020 shrink = F_SHRINK
1034 allow_unknown = F_ALLOW_UNKNOWN 1021 allow_unknown = F_ALLOW_UNKNOWN
1022 allow_sharing = F_ALLOW_SHARING
1023 dedup_keys = F_DEDUP_KEYS
1024 dedup_strings = F_DEDUP_STRINGS
1035 PPCODE: 1025 PPCODE:
1036{ 1026{
1037 if (enable) 1027 if (enable)
1038 self->flags |= ix; 1028 self->flags |= ix;
1039 else 1029 else
1044 1034
1045void get_shrink (CBOR *self) 1035void get_shrink (CBOR *self)
1046 ALIAS: 1036 ALIAS:
1047 get_shrink = F_SHRINK 1037 get_shrink = F_SHRINK
1048 get_allow_unknown = F_ALLOW_UNKNOWN 1038 get_allow_unknown = F_ALLOW_UNKNOWN
1039 get_allow_sharing = F_ALLOW_SHARING
1040 get_dedup_keys = F_DEDUP_KEYS
1041 get_dedup_strings = F_DEDUP_STRINGS
1049 PPCODE: 1042 PPCODE:
1050 XPUSHs (boolSV (self->flags & ix)); 1043 XPUSHs (boolSV (self->flags & ix));
1051 1044
1052void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1045void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1053 PPCODE: 1046 PPCODE:
1068int get_max_size (CBOR *self) 1061int get_max_size (CBOR *self)
1069 CODE: 1062 CODE:
1070 RETVAL = self->max_size; 1063 RETVAL = self->max_size;
1071 OUTPUT: 1064 OUTPUT:
1072 RETVAL 1065 RETVAL
1073
1074#if 0 //TODO
1075
1076void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1077 PPCODE:
1078{
1079 SvREFCNT_dec (self->cb_object);
1080 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1081
1082 XPUSHs (ST (0));
1083}
1084
1085void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)
1086 PPCODE:
1087{
1088 if (!self->cb_sk_object)
1089 self->cb_sk_object = newHV ();
1090
1091 if (SvOK (cb))
1092 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1093 else
1094 {
1095 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1096
1097 if (!HvKEYS (self->cb_sk_object))
1098 {
1099 SvREFCNT_dec (self->cb_sk_object);
1100 self->cb_sk_object = 0;
1101 }
1102 }
1103
1104 XPUSHs (ST (0));
1105}
1106
1107#endif
1108 1066
1109void encode (CBOR *self, SV *scalar) 1067void encode (CBOR *self, SV *scalar)
1110 PPCODE: 1068 PPCODE:
1111 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; 1069 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1112 XPUSHs (scalar); 1070 XPUSHs (scalar);
1125 EXTEND (SP, 2); 1083 EXTEND (SP, 2);
1126 PUSHs (sv); 1084 PUSHs (sv);
1127 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); 1085 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1128} 1086}
1129 1087
1130void DESTROY (CBOR *self)
1131 CODE:
1132 SvREFCNT_dec (self->cb_sk_object);
1133 SvREFCNT_dec (self->cb_object);
1134
1135PROTOTYPES: ENABLE 1088PROTOTYPES: ENABLE
1136 1089
1137void encode_cbor (SV *scalar) 1090void encode_cbor (SV *scalar)
1138 PPCODE: 1091 PPCODE:
1139{ 1092{

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines