ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/CBOR-XS/XS.xs
Revision: 1.27
Committed: Fri Nov 22 15:28:38 2013 UTC (10 years, 5 months ago) by root
Branch: MAIN
CVS Tags: rel-0_09
Changes since 1.26: +62 -9 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 #include "EXTERN.h"
2 #include "perl.h"
3 #include "XSUB.h"
4
5 #include <assert.h>
6 #include <string.h>
7 #include <stdlib.h>
8 #include <stdio.h>
9 #include <limits.h>
10 #include <float.h>
11
12 #include "ecb.h"
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
26 enum 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
59 #define F_SHRINK 0x00000001UL
60 #define F_ALLOW_UNKNOWN 0x00000002UL
61 #define F_ALLOW_SHARING 0x00000004UL //TODO
62 #define F_ALLOW_STRINGREF 0x00000008UL //TODO
63
64 #define INIT_SIZE 32 // initial scalar size to be allocated
65
66 #define SB do {
67 #define SE } while (0)
68
69 #define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72
73 #define ERR_NESTING_EXCEEDED "cbor text or perl structure exceeds maximum nesting level (max_depth set too low?)"
74
75 #ifdef USE_ITHREADS
76 # define CBOR_SLOW 1
77 # define CBOR_STASH (cbor_stash ? cbor_stash : gv_stashpv ("CBOR::XS", 1))
78 #else
79 # define CBOR_SLOW 0
80 # define CBOR_STASH cbor_stash
81 #endif
82
83 static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
84 static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
85
86 typedef struct {
87 U32 flags;
88 U32 max_depth;
89 STRLEN max_size;
90 SV *filter;
91 } CBOR;
92
93 ecb_inline void
94 cbor_init (CBOR *cbor)
95 {
96 Zero (cbor, 1, CBOR);
97 cbor->max_depth = 512;
98 }
99
100 ecb_inline void
101 cbor_free (CBOR *cbor)
102 {
103 SvREFCNT_dec (cbor->filter);
104 }
105
106 /////////////////////////////////////////////////////////////////////////////
107 // utility functions
108
109 ecb_inline SV *
110 get_bool (const char *name)
111 {
112 SV *sv = get_sv (name, 1);
113
114 SvREADONLY_on (sv);
115 SvREADONLY_on (SvRV (sv));
116
117 return sv;
118 }
119
120 ecb_inline void
121 shrink (SV *sv)
122 {
123 sv_utf8_downgrade (sv, 1);
124
125 if (SvLEN (sv) > SvCUR (sv) + 1)
126 {
127 #ifdef SvPV_shrink_to_cur
128 SvPV_shrink_to_cur (sv);
129 #elif defined (SvPV_renew)
130 SvPV_renew (sv, SvCUR (sv) + 1);
131 #endif
132 }
133 }
134
135 // minimum length of a string to be registered for stringref
136 ecb_inline int
137 minimum_string_length (UV idx)
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 }
149
150 /////////////////////////////////////////////////////////////////////////////
151 // encoder
152
153 // structure used for encoding CBOR
154 typedef struct
155 {
156 char *cur; // SvPVX (sv) + current output position
157 char *end; // SvEND (sv)
158 SV *sv; // result scalar
159 CBOR cbor;
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;
165 } enc_t;
166
167 ecb_inline void
168 need (enc_t *enc, STRLEN len)
169 {
170 if (ecb_expect_false (enc->cur + len >= enc->end))
171 {
172 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
173 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
174 enc->cur = SvPVX (enc->sv) + cur;
175 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
176 }
177 }
178
179 ecb_inline void
180 encode_ch (enc_t *enc, char ch)
181 {
182 need (enc, 1);
183 *enc->cur++ = ch;
184 }
185
186 static void
187 encode_uint (enc_t *enc, int major, UV len)
188 {
189 need (enc, 9);
190
191 if (len < 24)
192 *enc->cur++ = major | len;
193 else if (len <= 0xff)
194 {
195 *enc->cur++ = major | 24;
196 *enc->cur++ = len;
197 }
198 else if (len <= 0xffff)
199 {
200 *enc->cur++ = major | 25;
201 *enc->cur++ = len >> 8;
202 *enc->cur++ = len;
203 }
204 else if (len <= 0xffffffff)
205 {
206 *enc->cur++ = major | 26;
207 *enc->cur++ = len >> 24;
208 *enc->cur++ = len >> 16;
209 *enc->cur++ = len >> 8;
210 *enc->cur++ = len;
211 }
212 else
213 {
214 *enc->cur++ = major | 27;
215 *enc->cur++ = len >> 56;
216 *enc->cur++ = len >> 48;
217 *enc->cur++ = len >> 40;
218 *enc->cur++ = len >> 32;
219 *enc->cur++ = len >> 24;
220 *enc->cur++ = len >> 16;
221 *enc->cur++ = len >> 8;
222 *enc->cur++ = len;
223 }
224 }
225
226 ecb_inline void
227 encode_tag (enc_t *enc, UV tag)
228 {
229 encode_uint (enc, 0xc0, tag);
230 }
231
232 static void
233 encode_str (enc_t *enc, int utf8, char *str, STRLEN len)
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
254 encode_uint (enc, utf8 ? 0x60 : 0x40, len);
255 need (enc, len);
256 memcpy (enc->cur, str, len);
257 enc->cur += len;
258 }
259
260 static void encode_sv (enc_t *enc, SV *sv);
261
262 static void
263 encode_av (enc_t *enc, AV *av)
264 {
265 int i, len = av_len (av);
266
267 if (enc->depth >= enc->cbor.max_depth)
268 croak (ERR_NESTING_EXCEEDED);
269
270 ++enc->depth;
271
272 encode_uint (enc, 0x80, len + 1);
273
274 for (i = 0; i <= len; ++i)
275 {
276 SV **svp = av_fetch (av, i, 0);
277 encode_sv (enc, svp ? *svp : &PL_sv_undef);
278 }
279
280 --enc->depth;
281 }
282
283 static void
284 encode_hv (enc_t *enc, HV *hv)
285 {
286 HE *he;
287
288 if (enc->depth >= enc->cbor.max_depth)
289 croak (ERR_NESTING_EXCEEDED);
290
291 ++enc->depth;
292
293 int pairs = hv_iterinit (hv);
294 int mg = SvMAGICAL (hv);
295
296 if (mg)
297 encode_ch (enc, 0xa0 | 31);
298 else
299 encode_uint (enc, 0xa0, pairs);
300
301 while ((he = hv_iternext (hv)))
302 {
303 if (HeKLEN (he) == HEf_SVKEY)
304 encode_sv (enc, HeSVKEY (he));
305 else
306 encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));
307
308 encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
309 }
310
311 if (mg)
312 encode_ch (enc, 0xe0 | 31);
313
314 --enc->depth;
315 }
316
317 // encode objects, arrays and special \0=false and \1=true values.
318 static void
319 encode_rv (enc_t *enc, SV *sv)
320 {
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
345 svtype svt = SvTYPE (sv);
346
347 if (ecb_expect_false (SvOBJECT (sv)))
348 {
349 HV *boolean_stash = !CBOR_SLOW || types_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
357 : gv_stashpv ("CBOR::XS::Tagged" , 1);
358
359 HV *stash = SvSTASH (sv);
360 GV *method;
361
362 if (stash == boolean_stash)
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 }
426 else
427 croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
428 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
429 }
430 else if (svt == SVt_PVHV)
431 encode_hv (enc, (HV *)sv);
432 else if (svt == SVt_PVAV)
433 encode_av (enc, (AV *)sv);
434 else
435 {
436 encode_tag (enc, CBOR_TAG_INDIRECTION);
437 encode_sv (enc, sv);
438 }
439 }
440
441 static void
442 encode_nv (enc_t *enc, SV *sv)
443 {
444 double nv = SvNVX (sv);
445
446 need (enc, 9);
447
448 if (ecb_expect_false (nv == (U32)nv))
449 encode_uint (enc, 0x00, (U32)nv);
450 //TODO: maybe I32?
451 else if (ecb_expect_false (nv == (float)nv))
452 {
453 uint32_t fp = ecb_float_to_binary32 (nv);
454
455 *enc->cur++ = 0xe0 | 26;
456
457 if (!ecb_big_endian ())
458 fp = ecb_bswap32 (fp);
459
460 memcpy (enc->cur, &fp, 4);
461 enc->cur += 4;
462 }
463 else
464 {
465 uint64_t fp = ecb_double_to_binary64 (nv);
466
467 *enc->cur++ = 0xe0 | 27;
468
469 if (!ecb_big_endian ())
470 fp = ecb_bswap64 (fp);
471
472 memcpy (enc->cur, &fp, 8);
473 enc->cur += 8;
474 }
475 }
476
477 static void
478 encode_sv (enc_t *enc, SV *sv)
479 {
480 SvGETMAGIC (sv);
481
482 if (SvPOKp (sv))
483 {
484 STRLEN len;
485 char *str = SvPV (sv, len);
486 encode_str (enc, SvUTF8 (sv), str, len);
487 }
488 else if (SvNOKp (sv))
489 encode_nv (enc, sv);
490 else if (SvIOKp (sv))
491 {
492 if (SvIsUV (sv))
493 encode_uint (enc, 0x00, SvUVX (sv));
494 else if (SvIVX (sv) >= 0)
495 encode_uint (enc, 0x00, SvIVX (sv));
496 else
497 encode_uint (enc, 0x20, -(SvIVX (sv) + 1));
498 }
499 else if (SvROK (sv))
500 encode_rv (enc, SvRV (sv));
501 else if (!SvOK (sv))
502 encode_ch (enc, 0xe0 | 22);
503 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
504 encode_ch (enc, 0xe0 | 23);
505 else
506 croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",
507 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
508 }
509
510 static SV *
511 encode_cbor (SV *scalar, CBOR *cbor)
512 {
513 enc_t enc = { };
514
515 enc.cbor = *cbor;
516 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
517 enc.cur = SvPVX (enc.sv);
518 enc.end = SvEND (enc.sv);
519
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
529 encode_sv (&enc, scalar);
530
531 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
532 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
533
534 if (enc.cbor.flags & F_SHRINK)
535 shrink (enc.sv);
536
537 return enc.sv;
538 }
539
540 /////////////////////////////////////////////////////////////////////////////
541 // decoder
542
543 // structure used for decoding CBOR
544 typedef struct
545 {
546 U8 *cur; // current parser pointer
547 U8 *end; // end of input string
548 const char *err; // parse error, if != 0
549 CBOR cbor;
550 U32 depth; // recursion depth
551 U32 maxdepth; // recursion depth limit
552 AV *shareable;
553 AV *stringref;
554 SV *decode_tagged;
555 } dec_t;
556
557 #define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
558
559 #define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
560
561 #define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)
562 #define DEC_DEC_DEPTH --dec->depth
563
564 static UV
565 decode_uint (dec_t *dec)
566 {
567 switch (*dec->cur & 31)
568 {
569 case 0: case 1: case 2: case 3: case 4: case 5: case 6: case 7:
570 case 8: case 9: case 10: case 11: case 12: case 13: case 14: case 15:
571 case 16: case 17: case 18: case 19: case 20: case 21: case 22: case 23:
572 return *dec->cur++ & 31;
573
574 case 24:
575 WANT (2);
576 dec->cur += 2;
577 return dec->cur[-1];
578
579 case 25:
580 WANT (3);
581 dec->cur += 3;
582 return (((UV)dec->cur[-2]) << 8)
583 | ((UV)dec->cur[-1]);
584
585 case 26:
586 WANT (5);
587 dec->cur += 5;
588 return (((UV)dec->cur[-4]) << 24)
589 | (((UV)dec->cur[-3]) << 16)
590 | (((UV)dec->cur[-2]) << 8)
591 | ((UV)dec->cur[-1]);
592
593 case 27:
594 WANT (9);
595 dec->cur += 9;
596 return (((UV)dec->cur[-8]) << 56)
597 | (((UV)dec->cur[-7]) << 48)
598 | (((UV)dec->cur[-6]) << 40)
599 | (((UV)dec->cur[-5]) << 32)
600 | (((UV)dec->cur[-4]) << 24)
601 | (((UV)dec->cur[-3]) << 16)
602 | (((UV)dec->cur[-2]) << 8)
603 | ((UV)dec->cur[-1]);
604
605 default:
606 ERR ("corrupted CBOR data (unsupported integer minor encoding)");
607 }
608
609 fail:
610 return 0;
611 }
612
613 static SV *decode_sv (dec_t *dec);
614
615 static SV *
616 decode_av (dec_t *dec)
617 {
618 AV *av = newAV ();
619
620 DEC_INC_DEPTH;
621
622 if ((*dec->cur & 31) == 31)
623 {
624 ++dec->cur;
625
626 for (;;)
627 {
628 WANT (1);
629
630 if (*dec->cur == (0xe0 | 31))
631 {
632 ++dec->cur;
633 break;
634 }
635
636 av_push (av, decode_sv (dec));
637 }
638 }
639 else
640 {
641 int i, len = decode_uint (dec);
642
643 av_fill (av, len - 1);
644
645 for (i = 0; i < len; ++i)
646 AvARRAY (av)[i] = decode_sv (dec);
647 }
648
649 DEC_DEC_DEPTH;
650 return newRV_noinc ((SV *)av);
651
652 fail:
653 SvREFCNT_dec (av);
654 DEC_DEC_DEPTH;
655 return &PL_sv_undef;
656 }
657
658 static void
659 decode_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
701 static SV *
702 decode_hv (dec_t *dec)
703 {
704 HV *hv = newHV ();
705
706 DEC_INC_DEPTH;
707
708 if ((*dec->cur & 31) == 31)
709 {
710 ++dec->cur;
711
712 for (;;)
713 {
714 WANT (1);
715
716 if (*dec->cur == (0xe0 | 31))
717 {
718 ++dec->cur;
719 break;
720 }
721
722 decode_he (dec, hv);
723 }
724 }
725 else
726 {
727 int pairs = decode_uint (dec);
728
729 while (pairs--)
730 decode_he (dec, hv);
731 }
732
733 DEC_DEC_DEPTH;
734 return newRV_noinc ((SV *)hv);
735
736 fail:
737 SvREFCNT_dec (hv);
738 DEC_DEC_DEPTH;
739 return &PL_sv_undef;
740 }
741
742 static SV *
743 decode_str (dec_t *dec, int utf8)
744 {
745 SV *sv = 0;
746
747 if ((*dec->cur & 31) == 31)
748 {
749 ++dec->cur;
750
751 sv = newSVpvn ("", 0);
752
753 // not very fast, and certainly not robust against illegal input
754 for (;;)
755 {
756 WANT (1);
757
758 if (*dec->cur == (0xe0 | 31))
759 {
760 ++dec->cur;
761 break;
762 }
763
764 sv_catsv (sv, decode_sv (dec));
765 }
766 }
767 else
768 {
769 STRLEN len = decode_uint (dec);
770
771 WANT (len);
772 sv = newSVpvn (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));
778 }
779
780 if (utf8)
781 SvUTF8_on (sv);
782
783 return sv;
784
785 fail:
786 SvREFCNT_dec (sv);
787 return &PL_sv_undef;
788 }
789
790 static SV *
791 decode_tagged (dec_t *dec)
792 {
793 SV *sv = 0;
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
843 SV *osv = decode_sv (dec);
844 sv_setsv (sv, osv);
845 SvREFCNT_dec_NN (osv);
846 }
847 break;
848
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
958 return sv;
959
960 fail:
961 SvREFCNT_dec (sv);
962 return &PL_sv_undef;
963 }
964
965 static SV *
966 decode_sv (dec_t *dec)
967 {
968 WANT (1);
969
970 switch (*dec->cur >> 5)
971 {
972 case 0: // unsigned int
973 return newSVuv (decode_uint (dec));
974 case 1: // negative int
975 return newSViv (-1 - (IV)decode_uint (dec));
976 case 2: // octet string
977 return decode_str (dec, 0);
978 case 3: // utf-8 string
979 return decode_str (dec, 1);
980 case 4: // array
981 return decode_av (dec);
982 case 5: // map
983 return decode_hv (dec);
984 case 6: // tag
985 return decode_tagged (dec);
986 case 7: // misc
987 switch (*dec->cur++ & 31)
988 {
989 case 20:
990 #if CBOR_SLOW
991 types_false = get_bool ("Types::Serialiser::false");
992 #endif
993 return newSVsv (types_false);
994 case 21:
995 #if CBOR_SLOW
996 types_true = get_bool ("Types::Serialiser::true");
997 #endif
998 return newSVsv (types_true);
999 case 22:
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);
1006
1007 case 25:
1008 {
1009 WANT (2);
1010
1011 uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
1012 dec->cur += 2;
1013
1014 return newSVnv (ecb_binary16_to_float (fp));
1015 }
1016
1017 case 26:
1018 {
1019 uint32_t fp;
1020 WANT (4);
1021 memcpy (&fp, dec->cur, 4);
1022 dec->cur += 4;
1023
1024 if (!ecb_big_endian ())
1025 fp = ecb_bswap32 (fp);
1026
1027 return newSVnv (ecb_binary32_to_float (fp));
1028 }
1029
1030 case 27:
1031 {
1032 uint64_t fp;
1033 WANT (8);
1034 memcpy (&fp, dec->cur, 8);
1035 dec->cur += 8;
1036
1037 if (!ecb_big_endian ())
1038 fp = ecb_bswap64 (fp);
1039
1040 return newSVnv (ecb_binary64_to_double (fp));
1041 }
1042
1043 // 0..19 unassigned
1044 // 24 reserved + unassigned (reserved values are not encodable)
1045 default:
1046 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
1047 }
1048
1049 break;
1050 }
1051
1052 fail:
1053 return &PL_sv_undef;
1054 }
1055
1056 static SV *
1057 decode_cbor (SV *string, CBOR *cbor, char **offset_return)
1058 {
1059 dec_t dec = { };
1060 SV *sv;
1061 STRLEN len;
1062 char *data = SvPVbyte (string, len);
1063
1064 if (len > cbor->max_size && cbor->max_size)
1065 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
1066 (unsigned long)len, (unsigned long)cbor->max_size);
1067
1068 dec.cbor = *cbor;
1069 dec.cur = (U8 *)data;
1070 dec.end = (U8 *)data + len;
1071
1072 sv = decode_sv (&dec);
1073
1074 if (offset_return)
1075 *offset_return = dec.cur;
1076
1077 if (!(offset_return || !sv))
1078 if (dec.cur != dec.end && !dec.err)
1079 dec.err = "garbage after CBOR object";
1080
1081 if (dec.err)
1082 {
1083 SvREFCNT_dec (sv);
1084 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
1085 }
1086
1087 sv = sv_2mortal (sv);
1088
1089 return sv;
1090 }
1091
1092 /////////////////////////////////////////////////////////////////////////////
1093 // XS interface functions
1094
1095 MODULE = CBOR::XS PACKAGE = CBOR::XS
1096
1097 BOOT:
1098 {
1099 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1100 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1101
1102 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
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);
1113 }
1114
1115 PROTOTYPES: DISABLE
1116
1117 void CLONE (...)
1118 CODE:
1119 cbor_stash = 0;
1120 cbor_tagged_stash = 0;
1121 types_error_stash = 0;
1122 types_boolean_stash = 0;
1123
1124 void new (char *klass)
1125 PPCODE:
1126 {
1127 SV *pv = NEWSV (0, sizeof (CBOR));
1128 SvPOK_only (pv);
1129 cbor_init ((CBOR *)SvPVX (pv));
1130 XPUSHs (sv_2mortal (sv_bless (
1131 newRV_noinc (pv),
1132 strEQ (klass, "CBOR::XS") ? CBOR_STASH : gv_stashpv (klass, 1)
1133 )));
1134 }
1135
1136 void shrink (CBOR *self, int enable = 1)
1137 ALIAS:
1138 shrink = F_SHRINK
1139 allow_unknown = F_ALLOW_UNKNOWN
1140 allow_sharing = F_ALLOW_SHARING
1141 allow_stringref = F_ALLOW_STRINGREF
1142 PPCODE:
1143 {
1144 if (enable)
1145 self->flags |= ix;
1146 else
1147 self->flags &= ~ix;
1148
1149 XPUSHs (ST (0));
1150 }
1151
1152 void get_shrink (CBOR *self)
1153 ALIAS:
1154 get_shrink = F_SHRINK
1155 get_allow_unknown = F_ALLOW_UNKNOWN
1156 get_allow_sharing = F_ALLOW_SHARING
1157 get_allow_stringref = F_ALLOW_STRINGREF
1158 PPCODE:
1159 XPUSHs (boolSV (self->flags & ix));
1160
1161 void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1162 PPCODE:
1163 self->max_depth = max_depth;
1164 XPUSHs (ST (0));
1165
1166 U32 get_max_depth (CBOR *self)
1167 CODE:
1168 RETVAL = self->max_depth;
1169 OUTPUT:
1170 RETVAL
1171
1172 void max_size (CBOR *self, U32 max_size = 0)
1173 PPCODE:
1174 self->max_size = max_size;
1175 XPUSHs (ST (0));
1176
1177 int get_max_size (CBOR *self)
1178 CODE:
1179 RETVAL = self->max_size;
1180 OUTPUT:
1181 RETVAL
1182
1183 void filter (CBOR *self, SV *filter = 0)
1184 PPCODE:
1185 SvREFCNT_dec (self->filter);
1186 self->filter = filter ? newSVsv (filter) : filter;
1187 XPUSHs (ST (0));
1188
1189 SV *get_filter (CBOR *self)
1190 CODE:
1191 RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1192 OUTPUT:
1193 RETVAL
1194
1195 void encode (CBOR *self, SV *scalar)
1196 PPCODE:
1197 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1198 XPUSHs (scalar);
1199
1200 void decode (CBOR *self, SV *cborstr)
1201 PPCODE:
1202 PUTBACK; cborstr = decode_cbor (cborstr, self, 0); SPAGAIN;
1203 XPUSHs (cborstr);
1204
1205 void decode_prefix (CBOR *self, SV *cborstr)
1206 PPCODE:
1207 {
1208 SV *sv;
1209 char *offset;
1210 PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
1211 EXTEND (SP, 2);
1212 PUSHs (sv);
1213 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1214 }
1215
1216 void DESTROY (CBOR *self)
1217 PPCODE:
1218 cbor_free (self);
1219
1220 PROTOTYPES: ENABLE
1221
1222 void encode_cbor (SV *scalar)
1223 PPCODE:
1224 {
1225 CBOR cbor;
1226 cbor_init (&cbor);
1227 PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1228 XPUSHs (scalar);
1229 }
1230
1231 void decode_cbor (SV *cborstr)
1232 PPCODE:
1233 {
1234 CBOR cbor;
1235 cbor_init (&cbor);
1236 PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1237 XPUSHs (cborstr);
1238 }
1239