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 1.28 by root, Sat Nov 23 18:30:59 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#ifndef SvREFCNT_dec_NN
25# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
26#endif
27
28// known tags
29enum cbor_tag
30{
31 // inofficial extensions (pending iana registration)
32 CBOR_TAG_PERL_OBJECT = 24, // http://cbor.schmorp.de/perl-object
33 CBOR_TAG_GENERIC_OBJECT = 25, // http://cbor.schmorp.de/generic-object
34 CBOR_TAG_VALUE_SHAREABLE = 26, // http://cbor.schmorp.de/value-sharing
35 CBOR_TAG_VALUE_SHAREDREF = 27, // http://cbor.schmorp.de/value-sharing
36 CBOR_TAG_STRINGREF_NAMESPACE = 65537, // http://cbor.schmorp.de/stringref
37 CBOR_TAG_STRINGREF = 28, // http://cbor.schmorp.de/stringref
38 CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
39
40 // rfc7049
41 CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
42 CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
43 CBOR_TAG_POS_BIGNUM = 2, // byte string
44 CBOR_TAG_NEG_BIGNUM = 3, // byte string
45 CBOR_TAG_DECIMAL = 4, // decimal fraction, array
46 CBOR_TAG_BIGFLOAT = 5, // array
47
48 CBOR_TAG_CONV_B64U = 21, // base64url, any
49 CBOR_TAG_CONV_B64 = 22, // base64, any
50 CBOR_TAG_CONV_HEX = 23, // base16, any
51 CBOR_TAG_CBOR = 24, // embedded cbor, byte string
52
53 CBOR_TAG_URI = 32, // URI rfc3986, utf-8
54 CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
55 CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
56 CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
57 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
58
59 CBOR_TAG_MAGIC = 55799 // self-describe cbor
60};
61
14#define F_SHRINK 0x00000200UL 62#define F_SHRINK 0x00000001UL
15#define F_ALLOW_UNKNOWN 0x00002000UL 63#define F_ALLOW_UNKNOWN 0x00000002UL
64#define F_ALLOW_SHARING 0x00000004UL //TODO
65#define F_ALLOW_STRINGREF 0x00000008UL //TODO
16 66
17#define INIT_SIZE 32 // initial scalar size to be allocated 67#define INIT_SIZE 32 // initial scalar size to be allocated
18 68
19#define SB do { 69#define SB do {
20#define SE } while (0) 70#define SE } while (0)
31#else 81#else
32# define CBOR_SLOW 0 82# define CBOR_SLOW 0
33# define CBOR_STASH cbor_stash 83# define CBOR_STASH cbor_stash
34#endif 84#endif
35 85
36static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: 86static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
37static SV *cbor_true, *cbor_false; 87static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
38 88
39typedef struct { 89typedef struct {
40 U32 flags; 90 U32 flags;
41 U32 max_depth; 91 U32 max_depth;
42 STRLEN max_size; 92 STRLEN max_size;
43 93 SV *filter;
44 SV *cb_object;
45 HV *cb_sk_object;
46} CBOR; 94} CBOR;
47 95
48ecb_inline void 96ecb_inline void
49cbor_init (CBOR *cbor) 97cbor_init (CBOR *cbor)
50{ 98{
51 Zero (cbor, 1, CBOR); 99 Zero (cbor, 1, CBOR);
52 cbor->max_depth = 512; 100 cbor->max_depth = 512;
101}
102
103ecb_inline void
104cbor_free (CBOR *cbor)
105{
106 SvREFCNT_dec (cbor->filter);
53} 107}
54 108
55///////////////////////////////////////////////////////////////////////////// 109/////////////////////////////////////////////////////////////////////////////
56// utility functions 110// utility functions
57 111
79 SvPV_renew (sv, SvCUR (sv) + 1); 133 SvPV_renew (sv, SvCUR (sv) + 1);
80#endif 134#endif
81 } 135 }
82} 136}
83 137
84///////////////////////////////////////////////////////////////////////////// 138// minimum length of a string to be registered for stringref
85// fp hell 139ecb_inline int
86 140minimum_string_length (UV idx)
87//TODO 141{
142 return idx > 23
143 ? idx > 0xffU
144 ? idx > 0xffffU
145 ? idx > 0xffffffffU
146 ? 7
147 : 6
148 : 5
149 : 4
150 : 3;
151}
88 152
89///////////////////////////////////////////////////////////////////////////// 153/////////////////////////////////////////////////////////////////////////////
90// encoder 154// encoder
91 155
92// structure used for encoding CBOR 156// structure used for encoding CBOR
95 char *cur; // SvPVX (sv) + current output position 159 char *cur; // SvPVX (sv) + current output position
96 char *end; // SvEND (sv) 160 char *end; // SvEND (sv)
97 SV *sv; // result scalar 161 SV *sv; // result scalar
98 CBOR cbor; 162 CBOR cbor;
99 U32 depth; // recursion level 163 U32 depth; // recursion level
164 HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
165 UV stringref_idx;
166 HV *shareable; // ptr => index, or 0
167 UV shareable_idx;
100} enc_t; 168} enc_t;
101 169
102ecb_inline void 170ecb_inline void
103need (enc_t *enc, STRLEN len) 171need (enc_t *enc, STRLEN len)
104{ 172{
156 *enc->cur++ = len >> 8; 224 *enc->cur++ = len >> 8;
157 *enc->cur++ = len; 225 *enc->cur++ = len;
158 } 226 }
159} 227}
160 228
229ecb_inline void
230encode_tag (enc_t *enc, UV tag)
231{
232 encode_uint (enc, 0xc0, tag);
233}
234
161static void 235static void
162encode_str (enc_t *enc, int utf8, char *str, STRLEN len) 236encode_str (enc_t *enc, int utf8, char *str, STRLEN len)
163{ 237{
238 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_STRINGREF))
239 {
240 SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
241
242 if (SvOK (*svp))
243 {
244 // already registered, use stringref
245 encode_tag (enc, CBOR_TAG_STRINGREF);
246 encode_uint (enc, 0x00, SvUV (*svp));
247 return;
248 }
249 else if (len >= minimum_string_length (enc->stringref_idx))
250 {
251 // register only
252 sv_setuv (*svp, enc->stringref_idx);
253 ++enc->stringref_idx;
254 }
255 }
256
164 encode_uint (enc, utf8 ? 0x60 : 0x40, len); 257 encode_uint (enc, utf8 ? 0x60 : 0x40, len);
165 need (enc, len); 258 need (enc, len);
166 memcpy (enc->cur, str, len); 259 memcpy (enc->cur, str, len);
167 enc->cur += len; 260 enc->cur += len;
168} 261}
226 319
227// encode objects, arrays and special \0=false and \1=true values. 320// encode objects, arrays and special \0=false and \1=true values.
228static void 321static void
229encode_rv (enc_t *enc, SV *sv) 322encode_rv (enc_t *enc, SV *sv)
230{ 323{
231 svtype svt;
232
233 SvGETMAGIC (sv); 324 SvGETMAGIC (sv);
325
326 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)
327 && ecb_expect_false (SvREFCNT (sv) > 1))
328 {
329 if (!enc->shareable)
330 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
331
332 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
333
334 if (SvOK (*svp))
335 {
336 encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
337 encode_uint (enc, 0x00, SvUV (*svp));
338 return;
339 }
340 else
341 {
342 sv_setuv (*svp, enc->shareable_idx);
343 ++enc->shareable_idx;
344 encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
345 }
346 }
347
234 svt = SvTYPE (sv); 348 svtype svt = SvTYPE (sv);
235 349
236 if (ecb_expect_false (SvOBJECT (sv))) 350 if (ecb_expect_false (SvOBJECT (sv)))
237 { 351 {
238 HV *stash = !CBOR_SLOW || cbor_boolean_stash 352 HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
239 ? cbor_boolean_stash 353 ? types_boolean_stash
354 : gv_stashpv ("Types::Serialiser::Boolean", 1);
355 HV *error_stash = !CBOR_SLOW || types_error_stash
356 ? types_error_stash
357 : gv_stashpv ("Types::Serialiser::Error", 1);
358 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
359 ? cbor_tagged_stash
240 : gv_stashpv ("CBOR::XS::Boolean", 1); 360 : gv_stashpv ("CBOR::XS::Tagged" , 1);
241 361
242 if (SvSTASH (sv) == stash) 362 HV *stash = SvSTASH (sv);
363 GV *method;
364
365 if (stash == boolean_stash)
243 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); 366 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);
367 else if (stash == error_stash)
368 encode_ch (enc, 0xe0 | 23);
369 else if (stash == tagged_stash)
370 {
371 if (svt != SVt_PVAV)
372 croak ("encountered CBOR::XS::Tagged object that isn't an array");
373
374 encode_uint (enc, 0xc0, SvUV (*av_fetch ((AV *)sv, 0, 1)));
375 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
376 }
377 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
378 {
379 dSP;
380
381 ENTER; SAVETMPS; PUSHMARK (SP);
382 // we re-bless the reference to get overload and other niceties right
383 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
384
385 PUTBACK;
386 // G_SCALAR ensures that return value is 1
387 call_sv ((SV *)GvCV (method), G_SCALAR);
388 SPAGAIN;
389
390 // catch this surprisingly common error
391 if (SvROK (TOPs) && SvRV (TOPs) == sv)
392 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
393
394 encode_sv (enc, POPs);
395
396 PUTBACK;
397
398 FREETMPS; LEAVE;
399 }
400 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
401 {
402 dSP;
403
404 ENTER; SAVETMPS; PUSHMARK (SP);
405 EXTEND (SP, 2);
406 // we re-bless the reference to get overload and other niceties right
407 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
408 PUSHs (sv_cbor);
409
410 PUTBACK;
411 int count = call_sv ((SV *)GvCV (method), G_ARRAY);
412 SPAGAIN;
413
414 // catch this surprisingly common error
415 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
416 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
417
418 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
419 encode_uint (enc, 0x80, count + 1);
420 encode_str (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
421
422 while (count)
423 encode_sv (enc, SP[1 - count--]);
424
425 PUTBACK;
426
427 FREETMPS; LEAVE;
428 }
244 else 429 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", 430 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)))); 431 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
286#endif
287 }
288 } 432 }
289 else if (svt == SVt_PVHV) 433 else if (svt == SVt_PVHV)
290 encode_hv (enc, (HV *)sv); 434 encode_hv (enc, (HV *)sv);
291 else if (svt == SVt_PVAV) 435 else if (svt == SVt_PVAV)
292 encode_av (enc, (AV *)sv); 436 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 437 else
311 croak ("encountered %s, but CBOR can only represent references to arrays or hashes", 438 {
312 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 439 encode_tag (enc, CBOR_TAG_INDIRECTION);
440 encode_sv (enc, sv);
441 }
313} 442}
314 443
315static void 444static void
316encode_nv (enc_t *enc, SV *sv) 445encode_nv (enc_t *enc, SV *sv)
317{ 446{
382} 511}
383 512
384static SV * 513static SV *
385encode_cbor (SV *scalar, CBOR *cbor) 514encode_cbor (SV *scalar, CBOR *cbor)
386{ 515{
387 enc_t enc; 516 enc_t enc = { };
388 517
389 enc.cbor = *cbor; 518 enc.cbor = *cbor;
390 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 519 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
391 enc.cur = SvPVX (enc.sv); 520 enc.cur = SvPVX (enc.sv);
392 enc.end = SvEND (enc.sv); 521 enc.end = SvEND (enc.sv);
393 enc.depth = 0;
394 522
395 SvPOK_only (enc.sv); 523 SvPOK_only (enc.sv);
524
525 if (cbor->flags & F_ALLOW_STRINGREF)
526 {
527 encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
528 enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
529 enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
530 }
531
396 encode_sv (&enc, scalar); 532 encode_sv (&enc, scalar);
397 533
398 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 534 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
399 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 535 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
400 536
414 U8 *end; // end of input string 550 U8 *end; // end of input string
415 const char *err; // parse error, if != 0 551 const char *err; // parse error, if != 0
416 CBOR cbor; 552 CBOR cbor;
417 U32 depth; // recursion depth 553 U32 depth; // recursion depth
418 U32 maxdepth; // recursion depth limit 554 U32 maxdepth; // recursion depth limit
555 AV *shareable;
556 AV *stringref;
557 SV *decode_tagged;
419} dec_t; 558} dec_t;
420 559
421#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 560#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
422 561
423#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data") 562#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
517 SvREFCNT_dec (av); 656 SvREFCNT_dec (av);
518 DEC_DEC_DEPTH; 657 DEC_DEC_DEPTH;
519 return &PL_sv_undef; 658 return &PL_sv_undef;
520} 659}
521 660
661static void
662decode_he (dec_t *dec, HV *hv)
663{
664 // for speed reasons, we specialcase single-string
665 // byte or utf-8 strings as keys, but only when !stringref
666
667 if (ecb_expect_true (!dec->stringref))
668 if (*dec->cur >= 0x40 && *dec->cur <= 0x40 + 27)
669 {
670 I32 len = decode_uint (dec);
671 char *key = (char *)dec->cur;
672
673 dec->cur += len;
674
675 if (ecb_expect_false (dec->stringref))
676 av_push (dec->stringref, newSVpvn (key, len));
677
678 hv_store (hv, key, len, decode_sv (dec), 0);
679
680 return;
681 }
682 else if (*dec->cur >= 0x60 && *dec->cur <= 0x60 + 27)
683 {
684 I32 len = decode_uint (dec);
685 char *key = (char *)dec->cur;
686
687 dec->cur += len;
688
689 if (ecb_expect_false (dec->stringref))
690 av_push (dec->stringref, newSVpvn_utf8 (key, len, 1));
691
692 hv_store (hv, key, -len, decode_sv (dec), 0);
693
694 return;
695 }
696
697 SV *k = decode_sv (dec);
698 SV *v = decode_sv (dec);
699
700 hv_store_ent (hv, k, v, 0);
701 SvREFCNT_dec (k);
702}
703
522static SV * 704static SV *
523decode_hv (dec_t *dec) 705decode_hv (dec_t *dec)
524{ 706{
525 HV *hv = newHV (); 707 HV *hv = newHV ();
526 708
538 { 720 {
539 ++dec->cur; 721 ++dec->cur;
540 break; 722 break;
541 } 723 }
542 724
543 SV *k = decode_sv (dec); 725 decode_he (dec, hv);
544 SV *v = decode_sv (dec);
545
546 hv_store_ent (hv, k, v, 0);
547 } 726 }
548 } 727 }
549 else 728 else
550 { 729 {
551 int len = decode_uint (dec); 730 int pairs = decode_uint (dec);
552 731
553 while (len--) 732 while (pairs--)
554 { 733 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 } 734 }
561 735
562 DEC_DEC_DEPTH; 736 DEC_DEC_DEPTH;
563 return newRV_noinc ((SV *)hv); 737 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
719 738
720fail: 739fail:
721 SvREFCNT_dec (hv); 740 SvREFCNT_dec (hv);
722 DEC_DEC_DEPTH; 741 DEC_DEC_DEPTH;
723 return &PL_sv_undef; 742 return &PL_sv_undef;
724} 743}
725 744
726static SV * 745static SV *
727decode_str (dec_t *dec, int utf8) 746decode_str (dec_t *dec, int utf8)
728{ 747{
729 SV *sv; 748 SV *sv = 0;
730 749
731 if ((*dec->cur & 31) == 31) 750 if ((*dec->cur & 31) == 31)
732 { 751 {
733 ++dec->cur; 752 ++dec->cur;
734 753
743 { 762 {
744 ++dec->cur; 763 ++dec->cur;
745 break; 764 break;
746 } 765 }
747 766
748 SV *sv2 = decode_sv (dec);
749 sv_catsv (sv, sv2); 767 sv_catsv (sv, decode_sv (dec));
750 } 768 }
751 } 769 }
752 else 770 else
753 { 771 {
754 STRLEN len = decode_uint (dec); 772 STRLEN len = decode_uint (dec);
755 773
756 WANT (len); 774 WANT (len);
757 sv = newSVpvn (dec->cur, len); 775 sv = newSVpvn (dec->cur, len);
758 dec->cur += len; 776 dec->cur += len;
777
778 if (ecb_expect_false (dec->stringref)
779 && SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
780 av_push (dec->stringref, SvREFCNT_inc_NN (sv));
759 } 781 }
760 782
761 if (utf8) 783 if (utf8)
762 SvUTF8_on (sv); 784 SvUTF8_on (sv);
763 785
764 return sv; 786 return sv;
765 787
766fail: 788fail:
789 SvREFCNT_dec (sv);
767 return &PL_sv_undef; 790 return &PL_sv_undef;
768} 791}
769 792
770static SV * 793static SV *
771decode_tagged (dec_t *dec) 794decode_tagged (dec_t *dec)
772{ 795{
796 SV *sv = 0;
773 UV tag = decode_uint (dec); 797 UV tag = decode_uint (dec);
798
799 WANT (1);
800
801 switch (tag)
802 {
803 case CBOR_TAG_MAGIC:
804 sv = decode_sv (dec);
805 break;
806
807 case CBOR_TAG_INDIRECTION:
808 sv = newRV_noinc (decode_sv (dec));
809 break;
810
811 case CBOR_TAG_STRINGREF_NAMESPACE:
812 {
813 ENTER; SAVETMPS;
814
815 SAVESPTR (dec->stringref);
816 dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
817
818 sv = decode_sv (dec);
819
820 FREETMPS; LEAVE;
821 }
822 break;
823
824 case CBOR_TAG_STRINGREF:
825 {
826 if ((*dec->cur >> 5) != 0)
827 ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
828
829 UV idx = decode_uint (dec);
830
831 if (!dec->stringref || (int)idx > AvFILLp (dec->stringref))
832 ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
833
834 sv = newSVsv (AvARRAY (dec->stringref)[idx]);
835 }
836 break;
837
838 case CBOR_TAG_VALUE_SHAREABLE:
839 {
840 if (ecb_expect_false (!dec->shareable))
841 dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
842
843 sv = newSV (0);
844 av_push (dec->shareable, SvREFCNT_inc_NN (sv));
845
774 SV *sv = decode_sv (dec); 846 SV *osv = decode_sv (dec);
847 sv_setsv (sv, osv);
848 SvREFCNT_dec_NN (osv);
849 }
850 break;
775 851
776 if (tag == 55799) // 2.4.5 Self-Describe CBOR 852 case CBOR_TAG_VALUE_SHAREDREF:
853 {
854 if ((*dec->cur >> 5) != 0)
855 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
856
857 UV idx = decode_uint (dec);
858
859 if (!dec->shareable || (int)idx > AvFILLp (dec->shareable))
860 ERR ("corrupted CBOR data (sharedref index out of bounds)");
861
862 sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
863 }
864 break;
865
866 case CBOR_TAG_PERL_OBJECT:
867 {
868 sv = decode_sv (dec);
869
870 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
871 ERR ("corrupted CBOR data (non-array perl object)");
872
873 AV *av = (AV *)SvRV (sv);
874 int len = av_len (av) + 1;
875 HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
876
877 if (!stash)
878 ERR ("cannot decode perl-object (package does not exist)");
879
880 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
881
882 if (!method)
883 ERR ("cannot decode perl-object (package does not have a THAW method)");
884
885 dSP;
886
887 ENTER; SAVETMPS; PUSHMARK (SP);
888 EXTEND (SP, len + 1);
889 // we re-bless the reference to get overload and other niceties right
890 PUSHs (*av_fetch (av, 0, 1));
891 PUSHs (sv_cbor);
892
893 int i;
894
895 for (i = 1; i < len; ++i)
896 PUSHs (*av_fetch (av, i, 1));
897
898 PUTBACK;
899 call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
900 SPAGAIN;
901
902 if (SvTRUE (ERRSV))
903 {
904 FREETMPS; LEAVE;
905 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
906 }
907
908 SvREFCNT_dec (sv);
909 sv = SvREFCNT_inc (POPs);
910
911 PUTBACK;
912
913 FREETMPS; LEAVE;
914 }
915 break;
916
917 default:
918 {
919 sv = decode_sv (dec);
920
921 dSP;
922 ENTER; SAVETMPS; PUSHMARK (SP);
923 EXTEND (SP, 2);
924 PUSHs (newSVuv (tag));
925 PUSHs (sv);
926
927 PUTBACK;
928 int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
929 SPAGAIN;
930
931 if (SvTRUE (ERRSV))
932 {
933 FREETMPS; LEAVE;
934 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
935 }
936
937 if (count)
938 {
939 SvREFCNT_dec (sv);
940 sv = SvREFCNT_inc (POPs);
941 }
942 else
943 {
944 AV *av = newAV ();
945 av_push (av, newSVuv (tag));
946 av_push (av, sv);
947
948 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
949 ? cbor_tagged_stash
950 : gv_stashpv ("CBOR::XS::Tagged" , 1);
951 sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
952 }
953
954 PUTBACK;
955
956 FREETMPS; LEAVE;
957 }
958 break;
959 }
960
777 return sv; 961 return sv;
778 962
779 AV *av = newAV (); 963fail:
780 av_push (av, newSVuv (tag)); 964 SvREFCNT_dec (sv);
781 av_push (av, sv); 965 return &PL_sv_undef;
782 return newRV_noinc ((SV *)av);
783} 966}
784 967
785static SV * 968static SV *
786decode_sv (dec_t *dec) 969decode_sv (dec_t *dec)
787{ 970{
788 WANT (1); 971 WANT (1);
789 972
790 switch (*dec->cur >> 5) 973 switch (*dec->cur >> 5)
791 { 974 {
792 case 0: // unsigned int 975 case 0: // unsigned int
793 //TODO: 64 bit values on 3 2bit perls
794 return newSVuv (decode_uint (dec)); 976 return newSVuv (decode_uint (dec));
795 case 1: // negative int 977 case 1: // negative int
796 return newSViv (-1 - (IV)decode_uint (dec)); 978 return newSViv (-1 - (IV)decode_uint (dec));
797 case 2: // octet string 979 case 2: // octet string
798 return decode_str (dec, 0); 980 return decode_str (dec, 0);
807 case 7: // misc 989 case 7: // misc
808 switch (*dec->cur++ & 31) 990 switch (*dec->cur++ & 31)
809 { 991 {
810 case 20: 992 case 20:
811#if CBOR_SLOW 993#if CBOR_SLOW
812 cbor_false = get_bool ("CBOR::XS::false"); 994 types_false = get_bool ("Types::Serialiser::false");
813#endif 995#endif
814 return newSVsv (cbor_false); 996 return newSVsv (types_false);
815 case 21: 997 case 21:
816#if CBOR_SLOW 998#if CBOR_SLOW
817 cbor_true = get_bool ("CBOR::XS::true"); 999 types_true = get_bool ("Types::Serialiser::true");
818#endif 1000#endif
819 return newSVsv (cbor_true); 1001 return newSVsv (types_true);
820 case 22: 1002 case 22:
821 return newSVsv (&PL_sv_undef); 1003 return newSVsv (&PL_sv_undef);
1004 case 23:
1005#if CBOR_SLOW
1006 types_error = get_bool ("Types::Serialiser::error");
1007#endif
1008 return newSVsv (types_error);
822 1009
823 case 25: 1010 case 25:
824 { 1011 {
825 WANT (2); 1012 WANT (2);
826 1013
862 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); 1049 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
863 } 1050 }
864 1051
865 break; 1052 break;
866 } 1053 }
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 1054
924fail: 1055fail:
925 return &PL_sv_undef; 1056 return &PL_sv_undef;
926} 1057}
927 1058
928static SV * 1059static SV *
929decode_cbor (SV *string, CBOR *cbor, char **offset_return) 1060decode_cbor (SV *string, CBOR *cbor, char **offset_return)
930{ 1061{
931 dec_t dec; 1062 dec_t dec = { };
932 SV *sv; 1063 SV *sv;
1064 STRLEN len;
1065 char *data = SvPVbyte (string, len);
933 1066
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) 1067 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", 1068 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); 1069 (unsigned long)len, (unsigned long)cbor->max_size);
965 }
966
967 sv_utf8_downgrade (string, 0);
968 1070
969 dec.cbor = *cbor; 1071 dec.cbor = *cbor;
970 dec.cur = (U8 *)SvPVX (string); 1072 dec.cur = (U8 *)data;
971 dec.end = (U8 *)SvEND (string); 1073 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 1074
978 sv = decode_sv (&dec); 1075 sv = decode_sv (&dec);
979 1076
980 if (offset_return) 1077 if (offset_return)
981 *offset_return = dec.cur; 1078 *offset_return = dec.cur;
985 dec.err = "garbage after CBOR object"; 1082 dec.err = "garbage after CBOR object";
986 1083
987 if (dec.err) 1084 if (dec.err)
988 { 1085 {
989 SvREFCNT_dec (sv); 1086 SvREFCNT_dec (sv);
990 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); 1087 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
991 } 1088 }
992 1089
993 sv = sv_2mortal (sv); 1090 sv = sv_2mortal (sv);
994 1091
995 return sv; 1092 return sv;
1001MODULE = CBOR::XS PACKAGE = CBOR::XS 1098MODULE = CBOR::XS PACKAGE = CBOR::XS
1002 1099
1003BOOT: 1100BOOT:
1004{ 1101{
1005 cbor_stash = gv_stashpv ("CBOR::XS" , 1); 1102 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1006 cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); 1103 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1007 1104
1008 cbor_true = get_bool ("CBOR::XS::true"); 1105 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1009 cbor_false = get_bool ("CBOR::XS::false"); 1106 types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
1107
1108 types_true = get_bool ("Types::Serialiser::true" );
1109 types_false = get_bool ("Types::Serialiser::false");
1110 types_error = get_bool ("Types::Serialiser::error");
1111
1112 default_filter = newSVpv ("CBOR::XS::default_filter", 0);
1113
1114 sv_cbor = newSVpv ("CBOR", 0);
1115 SvREADONLY_on (sv_cbor);
1010} 1116}
1011 1117
1012PROTOTYPES: DISABLE 1118PROTOTYPES: DISABLE
1013 1119
1014void CLONE (...) 1120void CLONE (...)
1015 CODE: 1121 CODE:
1016 cbor_stash = 0; 1122 cbor_stash = 0;
1123 cbor_tagged_stash = 0;
1124 types_error_stash = 0;
1017 cbor_boolean_stash = 0; 1125 types_boolean_stash = 0;
1018 1126
1019void new (char *klass) 1127void new (char *klass)
1020 PPCODE: 1128 PPCODE:
1021{ 1129{
1022 SV *pv = NEWSV (0, sizeof (CBOR)); 1130 SV *pv = NEWSV (0, sizeof (CBOR));
1030 1138
1031void shrink (CBOR *self, int enable = 1) 1139void shrink (CBOR *self, int enable = 1)
1032 ALIAS: 1140 ALIAS:
1033 shrink = F_SHRINK 1141 shrink = F_SHRINK
1034 allow_unknown = F_ALLOW_UNKNOWN 1142 allow_unknown = F_ALLOW_UNKNOWN
1143 allow_sharing = F_ALLOW_SHARING
1144 allow_stringref = F_ALLOW_STRINGREF
1035 PPCODE: 1145 PPCODE:
1036{ 1146{
1037 if (enable) 1147 if (enable)
1038 self->flags |= ix; 1148 self->flags |= ix;
1039 else 1149 else
1044 1154
1045void get_shrink (CBOR *self) 1155void get_shrink (CBOR *self)
1046 ALIAS: 1156 ALIAS:
1047 get_shrink = F_SHRINK 1157 get_shrink = F_SHRINK
1048 get_allow_unknown = F_ALLOW_UNKNOWN 1158 get_allow_unknown = F_ALLOW_UNKNOWN
1159 get_allow_sharing = F_ALLOW_SHARING
1160 get_allow_stringref = F_ALLOW_STRINGREF
1049 PPCODE: 1161 PPCODE:
1050 XPUSHs (boolSV (self->flags & ix)); 1162 XPUSHs (boolSV (self->flags & ix));
1051 1163
1052void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1164void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1053 PPCODE: 1165 PPCODE:
1069 CODE: 1181 CODE:
1070 RETVAL = self->max_size; 1182 RETVAL = self->max_size;
1071 OUTPUT: 1183 OUTPUT:
1072 RETVAL 1184 RETVAL
1073 1185
1074#if 0 //TODO 1186void filter (CBOR *self, SV *filter = 0)
1075
1076void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1077 PPCODE: 1187 PPCODE:
1078{
1079 SvREFCNT_dec (self->cb_object); 1188 SvREFCNT_dec (self->filter);
1080 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; 1189 self->filter = filter ? newSVsv (filter) : filter;
1081
1082 XPUSHs (ST (0)); 1190 XPUSHs (ST (0));
1083}
1084 1191
1085void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef) 1192SV *get_filter (CBOR *self)
1086 PPCODE: 1193 CODE:
1087{ 1194 RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1088 if (!self->cb_sk_object) 1195 OUTPUT:
1089 self->cb_sk_object = newHV (); 1196 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 1197
1109void encode (CBOR *self, SV *scalar) 1198void encode (CBOR *self, SV *scalar)
1110 PPCODE: 1199 PPCODE:
1111 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; 1200 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1112 XPUSHs (scalar); 1201 XPUSHs (scalar);
1126 PUSHs (sv); 1215 PUSHs (sv);
1127 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); 1216 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1128} 1217}
1129 1218
1130void DESTROY (CBOR *self) 1219void DESTROY (CBOR *self)
1131 CODE: 1220 PPCODE:
1132 SvREFCNT_dec (self->cb_sk_object); 1221 cbor_free (self);
1133 SvREFCNT_dec (self->cb_object);
1134 1222
1135PROTOTYPES: ENABLE 1223PROTOTYPES: ENABLE
1136 1224
1137void encode_cbor (SV *scalar) 1225void encode_cbor (SV *scalar)
1138 PPCODE: 1226 PPCODE:

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines