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 1.27 by root, Fri Nov 22 15:28:38 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_SHAREABLE = 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_ALLOW_STRINGREF 0x00000008UL //TODO
16 63
17#define INIT_SIZE 32 // initial scalar size to be allocated 64#define INIT_SIZE 32 // initial scalar size to be allocated
18 65
19#define SB do { 66#define SB do {
20#define SE } while (0) 67#define SE } while (0)
31#else 78#else
32# define CBOR_SLOW 0 79# define CBOR_SLOW 0
33# define CBOR_STASH cbor_stash 80# define CBOR_STASH cbor_stash
34#endif 81#endif
35 82
36static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: 83static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
37static SV *cbor_true, *cbor_false; 84static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
38 85
39typedef struct { 86typedef struct {
40 U32 flags; 87 U32 flags;
41 U32 max_depth; 88 U32 max_depth;
42 STRLEN max_size; 89 STRLEN max_size;
43 90 SV *filter;
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{
51 Zero (cbor, 1, CBOR); 96 Zero (cbor, 1, CBOR);
52 cbor->max_depth = 512; 97 cbor->max_depth = 512;
98}
99
100ecb_inline void
101cbor_free (CBOR *cbor)
102{
103 SvREFCNT_dec (cbor->filter);
53} 104}
54 105
55///////////////////////////////////////////////////////////////////////////// 106/////////////////////////////////////////////////////////////////////////////
56// utility functions 107// utility functions
57 108
79 SvPV_renew (sv, SvCUR (sv) + 1); 130 SvPV_renew (sv, SvCUR (sv) + 1);
80#endif 131#endif
81 } 132 }
82} 133}
83 134
84///////////////////////////////////////////////////////////////////////////// 135// minimum length of a string to be registered for stringref
85// fp hell 136ecb_inline int
86 137minimum_string_length (UV idx)
87//TODO 138{
139 return idx > 23
140 ? idx > 0xffU
141 ? idx > 0xffffU
142 ? idx > 0xffffffffU
143 ? 7
144 : 6
145 : 5
146 : 4
147 : 3;
148}
88 149
89///////////////////////////////////////////////////////////////////////////// 150/////////////////////////////////////////////////////////////////////////////
90// encoder 151// encoder
91 152
92// structure used for encoding CBOR 153// structure used for encoding CBOR
95 char *cur; // SvPVX (sv) + current output position 156 char *cur; // SvPVX (sv) + current output position
96 char *end; // SvEND (sv) 157 char *end; // SvEND (sv)
97 SV *sv; // result scalar 158 SV *sv; // result scalar
98 CBOR cbor; 159 CBOR cbor;
99 U32 depth; // recursion level 160 U32 depth; // recursion level
161 HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
162 UV stringref_idx;
163 HV *shareable; // ptr => index, or 0
164 UV shareable_idx;
100} enc_t; 165} enc_t;
101 166
102ecb_inline void 167ecb_inline void
103need (enc_t *enc, STRLEN len) 168need (enc_t *enc, STRLEN len)
104{ 169{
156 *enc->cur++ = len >> 8; 221 *enc->cur++ = len >> 8;
157 *enc->cur++ = len; 222 *enc->cur++ = len;
158 } 223 }
159} 224}
160 225
226ecb_inline void
227encode_tag (enc_t *enc, UV tag)
228{
229 encode_uint (enc, 0xc0, tag);
230}
231
161static void 232static void
162encode_str (enc_t *enc, int utf8, char *str, STRLEN len) 233encode_str (enc_t *enc, int utf8, char *str, STRLEN len)
163{ 234{
235 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_STRINGREF))
236 {
237 SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
238
239 if (SvOK (*svp))
240 {
241 // already registered, use stringref
242 encode_tag (enc, CBOR_TAG_STRINGREF);
243 encode_uint (enc, 0x00, SvUV (*svp));
244 return;
245 }
246 else if (len >= minimum_string_length (enc->stringref_idx))
247 {
248 // register only
249 sv_setuv (*svp, enc->stringref_idx);
250 ++enc->stringref_idx;
251 }
252 }
253
164 encode_uint (enc, utf8 ? 0x60 : 0x40, len); 254 encode_uint (enc, utf8 ? 0x60 : 0x40, len);
165 need (enc, len); 255 need (enc, len);
166 memcpy (enc->cur, str, len); 256 memcpy (enc->cur, str, len);
167 enc->cur += len; 257 enc->cur += len;
168} 258}
226 316
227// encode objects, arrays and special \0=false and \1=true values. 317// encode objects, arrays and special \0=false and \1=true values.
228static void 318static void
229encode_rv (enc_t *enc, SV *sv) 319encode_rv (enc_t *enc, SV *sv)
230{ 320{
231 svtype svt;
232
233 SvGETMAGIC (sv); 321 SvGETMAGIC (sv);
322
323 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)
324 && ecb_expect_false (SvREFCNT (sv) > 1))
325 {
326 if (!enc->shareable)
327 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
328
329 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
330
331 if (SvOK (*svp))
332 {
333 encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
334 encode_uint (enc, 0x00, SvUV (*svp));
335 return;
336 }
337 else
338 {
339 sv_setuv (*svp, enc->shareable_idx);
340 ++enc->shareable_idx;
341 encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
342 }
343 }
344
234 svt = SvTYPE (sv); 345 svtype svt = SvTYPE (sv);
235 346
236 if (ecb_expect_false (SvOBJECT (sv))) 347 if (ecb_expect_false (SvOBJECT (sv)))
237 { 348 {
238 HV *stash = !CBOR_SLOW || cbor_boolean_stash 349 HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
239 ? cbor_boolean_stash 350 ? types_boolean_stash
351 : gv_stashpv ("Types::Serialiser::Boolean", 1);
352 HV *error_stash = !CBOR_SLOW || types_error_stash
353 ? types_error_stash
354 : gv_stashpv ("Types::Serialiser::Error", 1);
355 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
356 ? cbor_tagged_stash
240 : gv_stashpv ("CBOR::XS::Boolean", 1); 357 : gv_stashpv ("CBOR::XS::Tagged" , 1);
241 358
242 if (SvSTASH (sv) == stash) 359 HV *stash = SvSTASH (sv);
360 GV *method;
361
362 if (stash == boolean_stash)
243 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); 363 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);
364 else if (stash == error_stash)
365 encode_ch (enc, 0xe0 | 23);
366 else if (stash == tagged_stash)
367 {
368 if (svt != SVt_PVAV)
369 croak ("encountered CBOR::XS::Tagged object that isn't an array");
370
371 encode_uint (enc, 0xc0, SvUV (*av_fetch ((AV *)sv, 0, 1)));
372 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
373 }
374 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
375 {
376 dSP;
377
378 ENTER; SAVETMPS; PUSHMARK (SP);
379 // we re-bless the reference to get overload and other niceties right
380 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
381
382 PUTBACK;
383 // G_SCALAR ensures that return value is 1
384 call_sv ((SV *)GvCV (method), G_SCALAR);
385 SPAGAIN;
386
387 // catch this surprisingly common error
388 if (SvROK (TOPs) && SvRV (TOPs) == sv)
389 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
390
391 encode_sv (enc, POPs);
392
393 PUTBACK;
394
395 FREETMPS; LEAVE;
396 }
397 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
398 {
399 dSP;
400
401 ENTER; SAVETMPS; PUSHMARK (SP);
402 EXTEND (SP, 2);
403 // we re-bless the reference to get overload and other niceties right
404 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
405 PUSHs (sv_cbor);
406
407 PUTBACK;
408 int count = call_sv ((SV *)GvCV (method), G_ARRAY);
409 SPAGAIN;
410
411 // catch this surprisingly common error
412 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
413 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
414
415 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
416 encode_uint (enc, 0x80, count + 1);
417 encode_str (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
418
419 while (count)
420 encode_sv (enc, SP[1 - count--]);
421
422 PUTBACK;
423
424 FREETMPS; LEAVE;
425 }
244 else 426 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", 427 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)))); 428 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
286#endif
287 }
288 } 429 }
289 else if (svt == SVt_PVHV) 430 else if (svt == SVt_PVHV)
290 encode_hv (enc, (HV *)sv); 431 encode_hv (enc, (HV *)sv);
291 else if (svt == SVt_PVAV) 432 else if (svt == SVt_PVAV)
292 encode_av (enc, (AV *)sv); 433 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 434 else
311 croak ("encountered %s, but CBOR can only represent references to arrays or hashes", 435 {
312 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 436 encode_tag (enc, CBOR_TAG_INDIRECTION);
437 encode_sv (enc, sv);
438 }
313} 439}
314 440
315static void 441static void
316encode_nv (enc_t *enc, SV *sv) 442encode_nv (enc_t *enc, SV *sv)
317{ 443{
382} 508}
383 509
384static SV * 510static SV *
385encode_cbor (SV *scalar, CBOR *cbor) 511encode_cbor (SV *scalar, CBOR *cbor)
386{ 512{
387 enc_t enc; 513 enc_t enc = { };
388 514
389 enc.cbor = *cbor; 515 enc.cbor = *cbor;
390 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 516 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
391 enc.cur = SvPVX (enc.sv); 517 enc.cur = SvPVX (enc.sv);
392 enc.end = SvEND (enc.sv); 518 enc.end = SvEND (enc.sv);
393 enc.depth = 0;
394 519
395 SvPOK_only (enc.sv); 520 SvPOK_only (enc.sv);
521
522 if (cbor->flags & F_ALLOW_STRINGREF)
523 {
524 encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
525 enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
526 enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
527 }
528
396 encode_sv (&enc, scalar); 529 encode_sv (&enc, scalar);
397 530
398 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 531 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
399 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 532 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
400 533
414 U8 *end; // end of input string 547 U8 *end; // end of input string
415 const char *err; // parse error, if != 0 548 const char *err; // parse error, if != 0
416 CBOR cbor; 549 CBOR cbor;
417 U32 depth; // recursion depth 550 U32 depth; // recursion depth
418 U32 maxdepth; // recursion depth limit 551 U32 maxdepth; // recursion depth limit
552 AV *shareable;
553 AV *stringref;
554 SV *decode_tagged;
419} dec_t; 555} dec_t;
420 556
421#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 557#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
422 558
423#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data") 559#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
517 SvREFCNT_dec (av); 653 SvREFCNT_dec (av);
518 DEC_DEC_DEPTH; 654 DEC_DEC_DEPTH;
519 return &PL_sv_undef; 655 return &PL_sv_undef;
520} 656}
521 657
658static void
659decode_he (dec_t *dec, HV *hv)
660{
661 // for speed reasons, we specialcase single-string
662 // byte or utf-8 strings as keys, but only when !stringref
663
664 if (ecb_expect_true (!dec->stringref))
665 if (*dec->cur >= 0x40 && *dec->cur <= 0x40 + 27)
666 {
667 I32 len = decode_uint (dec);
668 char *key = (char *)dec->cur;
669
670 dec->cur += len;
671
672 if (ecb_expect_false (dec->stringref))
673 av_push (dec->stringref, newSVpvn (key, len));
674
675 hv_store (hv, key, len, decode_sv (dec), 0);
676
677 return;
678 }
679 else if (*dec->cur >= 0x60 && *dec->cur <= 0x60 + 27)
680 {
681 I32 len = decode_uint (dec);
682 char *key = (char *)dec->cur;
683
684 dec->cur += len;
685
686 if (ecb_expect_false (dec->stringref))
687 av_push (dec->stringref, newSVpvn_utf8 (key, len, 1));
688
689 hv_store (hv, key, -len, decode_sv (dec), 0);
690
691 return;
692 }
693
694 SV *k = decode_sv (dec);
695 SV *v = decode_sv (dec);
696
697 hv_store_ent (hv, k, v, 0);
698 SvREFCNT_dec (k);
699}
700
522static SV * 701static SV *
523decode_hv (dec_t *dec) 702decode_hv (dec_t *dec)
524{ 703{
525 HV *hv = newHV (); 704 HV *hv = newHV ();
526 705
538 { 717 {
539 ++dec->cur; 718 ++dec->cur;
540 break; 719 break;
541 } 720 }
542 721
543 SV *k = decode_sv (dec); 722 decode_he (dec, hv);
544 SV *v = decode_sv (dec);
545
546 hv_store_ent (hv, k, v, 0);
547 } 723 }
548 } 724 }
549 else 725 else
550 { 726 {
551 int len = decode_uint (dec); 727 int pairs = decode_uint (dec);
552 728
553 while (len--) 729 while (pairs--)
554 { 730 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 } 731 }
561 732
562 DEC_DEC_DEPTH; 733 DEC_DEC_DEPTH;
563 return newRV_noinc ((SV *)hv); 734 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
719 735
720fail: 736fail:
721 SvREFCNT_dec (hv); 737 SvREFCNT_dec (hv);
722 DEC_DEC_DEPTH; 738 DEC_DEC_DEPTH;
723 return &PL_sv_undef; 739 return &PL_sv_undef;
724} 740}
725 741
726static SV * 742static SV *
727decode_str (dec_t *dec, int utf8) 743decode_str (dec_t *dec, int utf8)
728{ 744{
729 SV *sv; 745 SV *sv = 0;
730 746
731 if ((*dec->cur & 31) == 31) 747 if ((*dec->cur & 31) == 31)
732 { 748 {
733 ++dec->cur; 749 ++dec->cur;
734 750
743 { 759 {
744 ++dec->cur; 760 ++dec->cur;
745 break; 761 break;
746 } 762 }
747 763
748 SV *sv2 = decode_sv (dec);
749 sv_catsv (sv, sv2); 764 sv_catsv (sv, decode_sv (dec));
750 } 765 }
751 } 766 }
752 else 767 else
753 { 768 {
754 STRLEN len = decode_uint (dec); 769 STRLEN len = decode_uint (dec);
755 770
756 WANT (len); 771 WANT (len);
757 sv = newSVpvn (dec->cur, len); 772 sv = newSVpvn (dec->cur, len);
758 dec->cur += len; 773 dec->cur += len;
774
775 if (ecb_expect_false (dec->stringref)
776 && SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
777 av_push (dec->stringref, SvREFCNT_inc_NN (sv));
759 } 778 }
760 779
761 if (utf8) 780 if (utf8)
762 SvUTF8_on (sv); 781 SvUTF8_on (sv);
763 782
764 return sv; 783 return sv;
765 784
766fail: 785fail:
786 SvREFCNT_dec (sv);
767 return &PL_sv_undef; 787 return &PL_sv_undef;
768} 788}
769 789
770static SV * 790static SV *
771decode_tagged (dec_t *dec) 791decode_tagged (dec_t *dec)
772{ 792{
793 SV *sv = 0;
773 UV tag = decode_uint (dec); 794 UV tag = decode_uint (dec);
795
796 WANT (1);
797
798 switch (tag)
799 {
800 case CBOR_TAG_MAGIC:
801 sv = decode_sv (dec);
802 break;
803
804 case CBOR_TAG_INDIRECTION:
805 sv = newRV_noinc (decode_sv (dec));
806 break;
807
808 case CBOR_TAG_STRINGREF_NAMESPACE:
809 {
810 ENTER; SAVETMPS;
811
812 SAVESPTR (dec->stringref);
813 dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
814
815 sv = decode_sv (dec);
816
817 FREETMPS; LEAVE;
818 }
819 break;
820
821 case CBOR_TAG_STRINGREF:
822 {
823 if ((*dec->cur >> 5) != 0)
824 ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
825
826 UV idx = decode_uint (dec);
827
828 if (!dec->stringref || (int)idx > AvFILLp (dec->stringref))
829 ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
830
831 sv = newSVsv (AvARRAY (dec->stringref)[idx]);
832 }
833 break;
834
835 case CBOR_TAG_VALUE_SHAREABLE:
836 {
837 if (ecb_expect_false (!dec->shareable))
838 dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
839
840 sv = newSV (0);
841 av_push (dec->shareable, SvREFCNT_inc_NN (sv));
842
774 SV *sv = decode_sv (dec); 843 SV *osv = decode_sv (dec);
844 sv_setsv (sv, osv);
845 SvREFCNT_dec_NN (osv);
846 }
847 break;
775 848
776 if (tag == 55799) // 2.4.5 Self-Describe CBOR 849 case CBOR_TAG_VALUE_SHAREDREF:
850 {
851 if ((*dec->cur >> 5) != 0)
852 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
853
854 UV idx = decode_uint (dec);
855
856 if (!dec->shareable || (int)idx > AvFILLp (dec->shareable))
857 ERR ("corrupted CBOR data (sharedref index out of bounds)");
858
859 sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
860 }
861 break;
862
863 case CBOR_TAG_PERL_OBJECT:
864 {
865 sv = decode_sv (dec);
866
867 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
868 ERR ("corrupted CBOR data (non-array perl object)");
869
870 AV *av = (AV *)SvRV (sv);
871 int len = av_len (av) + 1;
872 HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
873
874 if (!stash)
875 ERR ("cannot decode perl-object (package does not exist)");
876
877 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
878
879 if (!method)
880 ERR ("cannot decode perl-object (package does not have a THAW method)");
881
882 dSP;
883
884 ENTER; SAVETMPS; PUSHMARK (SP);
885 EXTEND (SP, len + 1);
886 // we re-bless the reference to get overload and other niceties right
887 PUSHs (*av_fetch (av, 0, 1));
888 PUSHs (sv_cbor);
889
890 int i;
891
892 for (i = 1; i < len; ++i)
893 PUSHs (*av_fetch (av, i, 1));
894
895 PUTBACK;
896 call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
897 SPAGAIN;
898
899 if (SvTRUE (ERRSV))
900 {
901 FREETMPS; LEAVE;
902 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
903 }
904
905 SvREFCNT_dec (sv);
906 sv = SvREFCNT_inc (POPs);
907
908 PUTBACK;
909
910 FREETMPS; LEAVE;
911 }
912 break;
913
914 default:
915 {
916 sv = decode_sv (dec);
917
918 dSP;
919 ENTER; SAVETMPS; PUSHMARK (SP);
920 EXTEND (SP, 2);
921 PUSHs (newSVuv (tag));
922 PUSHs (sv);
923
924 PUTBACK;
925 int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
926 SPAGAIN;
927
928 if (SvTRUE (ERRSV))
929 {
930 FREETMPS; LEAVE;
931 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
932 }
933
934 if (count)
935 {
936 SvREFCNT_dec (sv);
937 sv = SvREFCNT_inc (POPs);
938 }
939 else
940 {
941 AV *av = newAV ();
942 av_push (av, newSVuv (tag));
943 av_push (av, sv);
944
945 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
946 ? cbor_tagged_stash
947 : gv_stashpv ("CBOR::XS::Tagged" , 1);
948 sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
949 }
950
951 PUTBACK;
952
953 FREETMPS; LEAVE;
954 }
955 break;
956 }
957
777 return sv; 958 return sv;
778 959
779 AV *av = newAV (); 960fail:
780 av_push (av, newSVuv (tag)); 961 SvREFCNT_dec (sv);
781 av_push (av, sv); 962 return &PL_sv_undef;
782 return newRV_noinc ((SV *)av);
783} 963}
784 964
785static SV * 965static SV *
786decode_sv (dec_t *dec) 966decode_sv (dec_t *dec)
787{ 967{
788 WANT (1); 968 WANT (1);
789 969
790 switch (*dec->cur >> 5) 970 switch (*dec->cur >> 5)
791 { 971 {
792 case 0: // unsigned int 972 case 0: // unsigned int
793 //TODO: 64 bit values on 3 2bit perls
794 return newSVuv (decode_uint (dec)); 973 return newSVuv (decode_uint (dec));
795 case 1: // negative int 974 case 1: // negative int
796 return newSViv (-1 - (IV)decode_uint (dec)); 975 return newSViv (-1 - (IV)decode_uint (dec));
797 case 2: // octet string 976 case 2: // octet string
798 return decode_str (dec, 0); 977 return decode_str (dec, 0);
807 case 7: // misc 986 case 7: // misc
808 switch (*dec->cur++ & 31) 987 switch (*dec->cur++ & 31)
809 { 988 {
810 case 20: 989 case 20:
811#if CBOR_SLOW 990#if CBOR_SLOW
812 cbor_false = get_bool ("CBOR::XS::false"); 991 types_false = get_bool ("Types::Serialiser::false");
813#endif 992#endif
814 return newSVsv (cbor_false); 993 return newSVsv (types_false);
815 case 21: 994 case 21:
816#if CBOR_SLOW 995#if CBOR_SLOW
817 cbor_true = get_bool ("CBOR::XS::true"); 996 types_true = get_bool ("Types::Serialiser::true");
818#endif 997#endif
819 return newSVsv (cbor_true); 998 return newSVsv (types_true);
820 case 22: 999 case 22:
821 return newSVsv (&PL_sv_undef); 1000 return newSVsv (&PL_sv_undef);
1001 case 23:
1002#if CBOR_SLOW
1003 types_error = get_bool ("Types::Serialiser::error");
1004#endif
1005 return newSVsv (types_error);
822 1006
823 case 25: 1007 case 25:
824 { 1008 {
825 WANT (2); 1009 WANT (2);
826 1010
862 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); 1046 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
863 } 1047 }
864 1048
865 break; 1049 break;
866 } 1050 }
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 1051
924fail: 1052fail:
925 return &PL_sv_undef; 1053 return &PL_sv_undef;
926} 1054}
927 1055
928static SV * 1056static SV *
929decode_cbor (SV *string, CBOR *cbor, char **offset_return) 1057decode_cbor (SV *string, CBOR *cbor, char **offset_return)
930{ 1058{
931 dec_t dec; 1059 dec_t dec = { };
932 SV *sv; 1060 SV *sv;
1061 STRLEN len;
1062 char *data = SvPVbyte (string, len);
933 1063
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) 1064 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", 1065 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); 1066 (unsigned long)len, (unsigned long)cbor->max_size);
965 }
966
967 sv_utf8_downgrade (string, 0);
968 1067
969 dec.cbor = *cbor; 1068 dec.cbor = *cbor;
970 dec.cur = (U8 *)SvPVX (string); 1069 dec.cur = (U8 *)data;
971 dec.end = (U8 *)SvEND (string); 1070 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 1071
978 sv = decode_sv (&dec); 1072 sv = decode_sv (&dec);
979 1073
980 if (offset_return) 1074 if (offset_return)
981 *offset_return = dec.cur; 1075 *offset_return = dec.cur;
985 dec.err = "garbage after CBOR object"; 1079 dec.err = "garbage after CBOR object";
986 1080
987 if (dec.err) 1081 if (dec.err)
988 { 1082 {
989 SvREFCNT_dec (sv); 1083 SvREFCNT_dec (sv);
990 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); 1084 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
991 } 1085 }
992 1086
993 sv = sv_2mortal (sv); 1087 sv = sv_2mortal (sv);
994 1088
995 return sv; 1089 return sv;
1001MODULE = CBOR::XS PACKAGE = CBOR::XS 1095MODULE = CBOR::XS PACKAGE = CBOR::XS
1002 1096
1003BOOT: 1097BOOT:
1004{ 1098{
1005 cbor_stash = gv_stashpv ("CBOR::XS" , 1); 1099 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1006 cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); 1100 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1007 1101
1008 cbor_true = get_bool ("CBOR::XS::true"); 1102 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1009 cbor_false = get_bool ("CBOR::XS::false"); 1103 types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
1104
1105 types_true = get_bool ("Types::Serialiser::true" );
1106 types_false = get_bool ("Types::Serialiser::false");
1107 types_error = get_bool ("Types::Serialiser::error");
1108
1109 default_filter = newSVpv ("CBOR::XS::default_filter", 0);
1110
1111 sv_cbor = newSVpv ("CBOR", 0);
1112 SvREADONLY_on (sv_cbor);
1010} 1113}
1011 1114
1012PROTOTYPES: DISABLE 1115PROTOTYPES: DISABLE
1013 1116
1014void CLONE (...) 1117void CLONE (...)
1015 CODE: 1118 CODE:
1016 cbor_stash = 0; 1119 cbor_stash = 0;
1120 cbor_tagged_stash = 0;
1121 types_error_stash = 0;
1017 cbor_boolean_stash = 0; 1122 types_boolean_stash = 0;
1018 1123
1019void new (char *klass) 1124void new (char *klass)
1020 PPCODE: 1125 PPCODE:
1021{ 1126{
1022 SV *pv = NEWSV (0, sizeof (CBOR)); 1127 SV *pv = NEWSV (0, sizeof (CBOR));
1030 1135
1031void shrink (CBOR *self, int enable = 1) 1136void shrink (CBOR *self, int enable = 1)
1032 ALIAS: 1137 ALIAS:
1033 shrink = F_SHRINK 1138 shrink = F_SHRINK
1034 allow_unknown = F_ALLOW_UNKNOWN 1139 allow_unknown = F_ALLOW_UNKNOWN
1140 allow_sharing = F_ALLOW_SHARING
1141 allow_stringref = F_ALLOW_STRINGREF
1035 PPCODE: 1142 PPCODE:
1036{ 1143{
1037 if (enable) 1144 if (enable)
1038 self->flags |= ix; 1145 self->flags |= ix;
1039 else 1146 else
1044 1151
1045void get_shrink (CBOR *self) 1152void get_shrink (CBOR *self)
1046 ALIAS: 1153 ALIAS:
1047 get_shrink = F_SHRINK 1154 get_shrink = F_SHRINK
1048 get_allow_unknown = F_ALLOW_UNKNOWN 1155 get_allow_unknown = F_ALLOW_UNKNOWN
1156 get_allow_sharing = F_ALLOW_SHARING
1157 get_allow_stringref = F_ALLOW_STRINGREF
1049 PPCODE: 1158 PPCODE:
1050 XPUSHs (boolSV (self->flags & ix)); 1159 XPUSHs (boolSV (self->flags & ix));
1051 1160
1052void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1161void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1053 PPCODE: 1162 PPCODE:
1069 CODE: 1178 CODE:
1070 RETVAL = self->max_size; 1179 RETVAL = self->max_size;
1071 OUTPUT: 1180 OUTPUT:
1072 RETVAL 1181 RETVAL
1073 1182
1074#if 0 //TODO 1183void filter (CBOR *self, SV *filter = 0)
1075
1076void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1077 PPCODE: 1184 PPCODE:
1078{
1079 SvREFCNT_dec (self->cb_object); 1185 SvREFCNT_dec (self->filter);
1080 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; 1186 self->filter = filter ? newSVsv (filter) : filter;
1081
1082 XPUSHs (ST (0)); 1187 XPUSHs (ST (0));
1083}
1084 1188
1085void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef) 1189SV *get_filter (CBOR *self)
1086 PPCODE: 1190 CODE:
1087{ 1191 RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1088 if (!self->cb_sk_object) 1192 OUTPUT:
1089 self->cb_sk_object = newHV (); 1193 RETVAL
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 1194
1109void encode (CBOR *self, SV *scalar) 1195void encode (CBOR *self, SV *scalar)
1110 PPCODE: 1196 PPCODE:
1111 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; 1197 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1112 XPUSHs (scalar); 1198 XPUSHs (scalar);
1126 PUSHs (sv); 1212 PUSHs (sv);
1127 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); 1213 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1128} 1214}
1129 1215
1130void DESTROY (CBOR *self) 1216void DESTROY (CBOR *self)
1131 CODE: 1217 PPCODE:
1132 SvREFCNT_dec (self->cb_sk_object); 1218 cbor_free (self);
1133 SvREFCNT_dec (self->cb_object);
1134 1219
1135PROTOTYPES: ENABLE 1220PROTOTYPES: ENABLE
1136 1221
1137void encode_cbor (SV *scalar) 1222void encode_cbor (SV *scalar)
1138 PPCODE: 1223 PPCODE:

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines