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 1.1 by root, Fri Oct 25 23:09:45 2013 UTC vs.
Revision 1.67 by root, Sun Nov 29 21:35:06 2020 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
13#define ECB_NO_THREADS 1
12#include "ecb.h" 14#include "ecb.h"
13 15
14#if defined(__BORLANDC__) || defined(_MSC_VER) 16// compatibility with perl <5.18
15# define snprintf _snprintf // C compilers have this in stdio.h 17#ifndef HvNAMELEN_get
18# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
16#endif 19#endif
20#ifndef HvNAMELEN
21# define HvNAMELEN(hv) HvNAMELEN_get (hv)
22#endif
23#ifndef HvNAMEUTF8
24# define HvNAMEUTF8(hv) 0
25#endif
26#ifndef SvREFCNT_inc_NN
27# define SvREFCNT_inc_NN(sv) SvREFCNT_inc (sv)
28#endif
29#ifndef SvREFCNT_dec_NN
30# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
31#endif
17 32
33// known major and minor types
34enum cbor_type
35{
36 MAJOR_SHIFT = 5,
37 MINOR_MASK = 0x1f,
38
39 MAJOR_POS_INT = 0 << MAJOR_SHIFT,
40 MAJOR_NEG_INT = 1 << MAJOR_SHIFT,
41 MAJOR_BYTES = 2 << MAJOR_SHIFT,
42 MAJOR_TEXT = 3 << MAJOR_SHIFT,
43 MAJOR_ARRAY = 4 << MAJOR_SHIFT,
44 MAJOR_MAP = 5 << MAJOR_SHIFT,
45 MAJOR_TAG = 6 << MAJOR_SHIFT,
46 MAJOR_MISC = 7 << MAJOR_SHIFT,
47
48 // INT/STRING/ARRAY/MAP subtypes
49 LENGTH_EXT1 = 24,
50 LENGTH_EXT2 = 25,
51 LENGTH_EXT4 = 26,
52 LENGTH_EXT8 = 27,
53
54 // SIMPLE types (effectively MISC subtypes)
55 SIMPLE_FALSE = 20,
56 SIMPLE_TRUE = 21,
57 SIMPLE_NULL = 22,
58 SIMPLE_UNDEF = 23,
59
60 // MISC subtype (unused)
61 MISC_EXT1 = 24,
62 MISC_FLOAT16 = 25,
63 MISC_FLOAT32 = 26,
64 MISC_FLOAT64 = 27,
65
66 // BYTES/TEXT/ARRAY/MAP
67 MINOR_INDEF = 31,
68};
69
70// known tags
71enum cbor_tag
72{
73 // extensions
74 CBOR_TAG_STRINGREF = 25, // http://cbor.schmorp.de/stringref
75 CBOR_TAG_PERL_OBJECT = 26, // http://cbor.schmorp.de/perl-object
76 CBOR_TAG_GENERIC_OBJECT = 27, // http://cbor.schmorp.de/generic-object
77 CBOR_TAG_VALUE_SHAREABLE = 28, // http://cbor.schmorp.de/value-sharing
78 CBOR_TAG_VALUE_SHAREDREF = 29, // http://cbor.schmorp.de/value-sharing
79 CBOR_TAG_STRINGREF_NAMESPACE = 256, // http://cbor.schmorp.de/stringref
80 CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
81
82 // rfc7049
83 CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
84 CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
85 CBOR_TAG_POS_BIGNUM = 2, // byte string
86 CBOR_TAG_NEG_BIGNUM = 3, // byte string
87 CBOR_TAG_DECIMAL = 4, // decimal fraction, array
88 CBOR_TAG_BIGFLOAT = 5, // array
89
90 CBOR_TAG_CONV_B64U = 21, // base64url, any
91 CBOR_TAG_CONV_B64 = 22, // base64, any
92 CBOR_TAG_CONV_HEX = 23, // base16, any
93 CBOR_TAG_CBOR = 24, // embedded cbor, byte string
94
95 CBOR_TAG_URI = 32, // URI rfc3986, utf-8
96 CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
97 CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
98 CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
99 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
100
101 CBOR_TAG_MAGIC = 55799, // self-describe cbor
102};
103
104// known forced types, also hardcoded in CBOR.pm
105enum
106{
107 AS_CBOR = 0,
108 AS_BYTES = 1,
109 AS_TEXT = 2,
110 AS_INT = 3,
111 AS_FLOAT16 = 4,
112 AS_FLOAT32 = 5,
113 AS_FLOAT64 = 6,
114 // possibly future enhancements: (generic) float, (generic) string
115};
116
18#define F_SHRINK 0x00000200UL 117#define F_SHRINK 0x00000001UL
19#define F_ALLOW_UNKNOWN 0x00002000UL 118#define F_ALLOW_UNKNOWN 0x00000002UL
119#define F_ALLOW_SHARING 0x00000004UL
120#define F_ALLOW_CYCLES 0x00000008UL
121#define F_FORBID_OBJECTS 0x00000010UL
122#define F_PACK_STRINGS 0x00000020UL
123#define F_TEXT_KEYS 0x00000040UL
124#define F_TEXT_STRINGS 0x00000080UL
125#define F_VALIDATE_UTF8 0x00000100UL
20 126
21#define INIT_SIZE 32 // initial scalar size to be allocated 127#define INIT_SIZE 32 // initial scalar size to be allocated
22 128
23#define SB do { 129#define SB do {
24#define SE } while (0) 130#define SE } while (0)
25
26#if __GNUC__ >= 3
27# define expect(expr,value) __builtin_expect ((expr), (value))
28# define INLINE static inline
29#else
30# define expect(expr,value) (expr)
31# define INLINE static
32#endif
33
34#define expect_false(expr) expect ((expr) != 0, 0)
35#define expect_true(expr) expect ((expr) != 0, 1)
36 131
37#define IN_RANGE_INC(type,val,beg,end) \ 132#define IN_RANGE_INC(type,val,beg,end) \
38 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 133 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
39 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 134 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
40 135
46#else 141#else
47# define CBOR_SLOW 0 142# define CBOR_SLOW 0
48# define CBOR_STASH cbor_stash 143# define CBOR_STASH cbor_stash
49#endif 144#endif
50 145
51static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: 146static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
52static SV *cbor_true, *cbor_false; 147static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
53 148
54typedef struct { 149typedef struct {
55 U32 flags; 150 U32 flags;
56 U32 max_depth; 151 U32 max_depth;
57 STRLEN max_size; 152 STRLEN max_size;
153 SV *filter;
58 154
59 SV *cb_object; 155 // for the incremental parser
60 HV *cb_sk_object; 156 STRLEN incr_pos; // the current offset into the text
157 STRLEN incr_need; // minimum bytes needed to decode
158 AV *incr_count; // for every nesting level, the number of outstanding values, or -1 for indef.
61} CBOR; 159} CBOR;
62 160
63INLINE void 161ecb_inline void
64cbor_init (CBOR *cbor) 162cbor_init (CBOR *cbor)
65{ 163{
66 Zero (cbor, 1, CBOR); 164 Zero (cbor, 1, CBOR);
67 cbor->max_depth = 512; 165 cbor->max_depth = 512;
68} 166}
69 167
168ecb_inline void
169cbor_free (CBOR *cbor)
170{
171 SvREFCNT_dec (cbor->filter);
172 SvREFCNT_dec (cbor->incr_count);
173}
174
70///////////////////////////////////////////////////////////////////////////// 175/////////////////////////////////////////////////////////////////////////////
71// utility functions 176// utility functions
72 177
73INLINE SV * 178ecb_inline SV *
74get_bool (const char *name) 179get_bool (const char *name)
75{ 180{
76 SV *sv = get_sv (name, 1); 181 SV *sv = get_sv (name, 1);
77 182
78 SvREADONLY_on (sv); 183 SvREADONLY_on (sv);
79 SvREADONLY_on (SvRV (sv)); 184 SvREADONLY_on (SvRV (sv));
80 185
81 return sv; 186 return sv;
82} 187}
83 188
84INLINE void 189ecb_inline void
85shrink (SV *sv) 190shrink (SV *sv)
86{ 191{
87 sv_utf8_downgrade (sv, 1); 192 sv_utf8_downgrade (sv, 1);
88 193
89 if (SvLEN (sv) > SvCUR (sv) + 1) 194 if (SvLEN (sv) > SvCUR (sv) + 1)
94 SvPV_renew (sv, SvCUR (sv) + 1); 199 SvPV_renew (sv, SvCUR (sv) + 1);
95#endif 200#endif
96 } 201 }
97} 202}
98 203
99///////////////////////////////////////////////////////////////////////////// 204// minimum length of a string to be registered for stringref
100// fp hell 205ecb_inline int
101 206minimum_string_length (UV idx)
102//TODO 207{
208 return idx <= 23 ? 3
209 : idx <= 0xffU ? 4
210 : idx <= 0xffffU ? 5
211 : idx <= 0xffffffffU ? 7
212 : 11;
213}
103 214
104///////////////////////////////////////////////////////////////////////////// 215/////////////////////////////////////////////////////////////////////////////
105// encoder 216// encoder
106 217
107// structure used for encoding CBOR 218// structure used for encoding CBOR
110 char *cur; // SvPVX (sv) + current output position 221 char *cur; // SvPVX (sv) + current output position
111 char *end; // SvEND (sv) 222 char *end; // SvEND (sv)
112 SV *sv; // result scalar 223 SV *sv; // result scalar
113 CBOR cbor; 224 CBOR cbor;
114 U32 depth; // recursion level 225 U32 depth; // recursion level
226 HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
227 UV stringref_idx;
228 HV *shareable; // ptr => index, or 0
229 UV shareable_idx;
115} enc_t; 230} enc_t;
116 231
117INLINE void 232ecb_inline void
118need (enc_t *enc, STRLEN len) 233need (enc_t *enc, STRLEN len)
119{ 234{
120 if (expect_false (enc->cur + len >= enc->end)) 235 if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
121 { 236 {
122 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 237 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
123 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 238 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
124 enc->cur = SvPVX (enc->sv) + cur; 239 enc->cur = SvPVX (enc->sv) + cur;
125 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 240 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126 } 241 }
127} 242}
128 243
129INLINE void 244ecb_inline void
130encode_ch (enc_t *enc, char ch) 245encode_ch (enc_t *enc, char ch)
131{ 246{
132 need (enc, 1); 247 need (enc, 1);
133 *enc->cur++ = ch; 248 *enc->cur++ = ch;
134} 249}
135 250
251// used for tags, intregers, element counts and so on
136static void 252static void
137encode_uint (enc_t *enc, int major, UV len) 253encode_uint (enc_t *enc, int major, UV len)
138{ 254{
139 need (enc, 9); 255 need (enc, 9);
140 256
141 if (len < 24) 257 if (ecb_expect_true (len < LENGTH_EXT1))
142 *enc->cur++ = major | len; 258 *enc->cur++ = major | len;
143 else if (len < 0x100) 259 else if (ecb_expect_true (len <= 0xffU))
144 { 260 {
145 *enc->cur++ = major | 24; 261 *enc->cur++ = major | LENGTH_EXT1;
146 *enc->cur++ = len; 262 *enc->cur++ = len;
147 } 263 }
148 else if (len < 0x10000) 264 else if (len <= 0xffffU)
149 { 265 {
150 *enc->cur++ = major | 25; 266 *enc->cur++ = major | LENGTH_EXT2;
151 *enc->cur++ = len >> 8; 267 *enc->cur++ = len >> 8;
152 *enc->cur++ = len; 268 *enc->cur++ = len;
153 } 269 }
154 else if (len < 0x100000000) 270 else if (len <= 0xffffffffU)
155 { 271 {
156 *enc->cur++ = major | 26; 272 *enc->cur++ = major | LENGTH_EXT4;
157 *enc->cur++ = len >> 24; 273 *enc->cur++ = len >> 24;
158 *enc->cur++ = len >> 16; 274 *enc->cur++ = len >> 16;
159 *enc->cur++ = len >> 8; 275 *enc->cur++ = len >> 8;
160 *enc->cur++ = len; 276 *enc->cur++ = len;
161 } 277 }
162 else if (len) 278 else
163 { 279 {
164 *enc->cur++ = major | 27; 280 *enc->cur++ = major | LENGTH_EXT8;
165 *enc->cur++ = len >> 56; 281 *enc->cur++ = len >> 56;
166 *enc->cur++ = len >> 48; 282 *enc->cur++ = len >> 48;
167 *enc->cur++ = len >> 40; 283 *enc->cur++ = len >> 40;
168 *enc->cur++ = len >> 32; 284 *enc->cur++ = len >> 32;
169 *enc->cur++ = len >> 24; 285 *enc->cur++ = len >> 24;
171 *enc->cur++ = len >> 8; 287 *enc->cur++ = len >> 8;
172 *enc->cur++ = len; 288 *enc->cur++ = len;
173 } 289 }
174} 290}
175 291
176static void 292// encodes a perl value into a CBOR integer
293ecb_inline void
294encode_int (enc_t *enc, SV *sv)
295{
296 if (SvIsUV (sv))
297 encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
298 else if (SvIVX (sv) >= 0)
299 encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
300 else
301 encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
302}
303
304ecb_inline void
305encode_tag (enc_t *enc, UV tag)
306{
307 encode_uint (enc, MAJOR_TAG, tag);
308}
309
310// exceptional (hopefully) slow path for byte strings that need to be utf8-encoded
311ecb_noinline static void
177encode_str (enc_t *enc, int utf8, char *str, STRLEN len) 312encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
178{ 313{
179 encode_uint (enc, utf8 ? 0x60 : 0x40, len); 314 STRLEN ulen = len;
315 U8 *p, *pend = (U8 *)str + len;
316
317 for (p = (U8 *)str; p < pend; ++p)
318 ulen += *p >> 7; // count set high bits
319
320 encode_uint (enc, MAJOR_TEXT, ulen);
321
322 need (enc, ulen);
323 for (p = (U8 *)str; p < pend; ++p)
324 if (*p < 0x80)
325 *enc->cur++ = *p;
326 else
327 {
328 *enc->cur++ = 0xc0 + (*p >> 6);
329 *enc->cur++ = 0x80 + (*p & 63);
330 }
331}
332
333ecb_inline void
334encode_str (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
335{
336 if (ecb_expect_false (upgrade_utf8))
337 if (!utf8)
338 {
339 encode_str_utf8 (enc, utf8, str, len);
340 return;
341 }
342
343 encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len);
180 need (enc, len); 344 need (enc, len);
181 memcpy (enc->cur, str, len); 345 memcpy (enc->cur, str, len);
182 enc->cur += len; 346 enc->cur += len;
183} 347}
184 348
349ecb_inline void
350encode_strref (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
351{
352 if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS))
353 {
354 SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
355
356 if (SvOK (*svp))
357 {
358 // already registered, use stringref
359 encode_tag (enc, CBOR_TAG_STRINGREF);
360 encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
361 return;
362 }
363 else if (len >= minimum_string_length (enc->stringref_idx))
364 {
365 // register only
366 sv_setuv (*svp, enc->stringref_idx);
367 ++enc->stringref_idx;
368 }
369 }
370
371 encode_str (enc, upgrade_utf8, utf8, str, len);
372}
373
374ecb_inline void
375encode_float16 (enc_t *enc, NV nv)
376{
377 need (enc, 1+2);
378
379 *enc->cur++ = MAJOR_MISC | MISC_FLOAT16;
380
381 uint16_t fp = ecb_float_to_binary16 (nv);
382
383 if (!ecb_big_endian ())
384 fp = ecb_bswap16 (fp);
385
386 memcpy (enc->cur, &fp, 2);
387 enc->cur += 2;
388}
389
390ecb_inline void
391encode_float32 (enc_t *enc, NV nv)
392{
393 need (enc, 1+4);
394
395 *enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
396
397 uint32_t fp = ecb_float_to_binary32 (nv);
398
399 if (!ecb_big_endian ())
400 fp = ecb_bswap32 (fp);
401
402 memcpy (enc->cur, &fp, 4);
403 enc->cur += 4;
404}
405
406ecb_inline void
407encode_float64 (enc_t *enc, NV nv)
408{
409 need (enc, 1+8);
410
411 *enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
412
413 uint64_t fp = ecb_double_to_binary64 (nv);
414
415 if (!ecb_big_endian ())
416 fp = ecb_bswap64 (fp);
417
418 memcpy (enc->cur, &fp, 8);
419 enc->cur += 8;
420}
421
422ecb_inline void
423encode_forced (enc_t *enc, UV type, SV *sv)
424{
425 switch (type)
426 {
427 case AS_CBOR:
428 {
429 STRLEN len;
430 char *str = SvPVbyte (sv, len);
431
432 need (enc, len);
433 memcpy (enc->cur, str, len);
434 enc->cur += len;
435 }
436 break;
437
438 case AS_BYTES:
439 {
440 STRLEN len;
441 char *str = SvPVbyte (sv, len);
442 encode_strref (enc, 0, 0, str, len);
443 }
444 break;
445
446 case AS_TEXT:
447 {
448 STRLEN len;
449 char *str = SvPVutf8 (sv, len);
450 encode_strref (enc, 1, 1, str, len);
451 }
452 break;
453
454 case AS_INT: encode_int (enc, sv); break;
455 case AS_FLOAT16: encode_float16 (enc, SvNV (sv)); break;
456 case AS_FLOAT32: encode_float32 (enc, SvNV (sv)); break;
457 case AS_FLOAT64: encode_float64 (enc, SvNV (sv)); break;
458
459 default:
460 croak ("encountered malformed CBOR::XS::Tagged object");
461 }
462}
463
185static void encode_sv (enc_t *enc, SV *sv); 464static void encode_sv (enc_t *enc, SV *sv);
186 465
187static void 466static void
188encode_av (enc_t *enc, AV *av) 467encode_av (enc_t *enc, AV *av)
189{ 468{
192 if (enc->depth >= enc->cbor.max_depth) 471 if (enc->depth >= enc->cbor.max_depth)
193 croak (ERR_NESTING_EXCEEDED); 472 croak (ERR_NESTING_EXCEEDED);
194 473
195 ++enc->depth; 474 ++enc->depth;
196 475
197 encode_uint (enc, 0x80, len + 1); 476 encode_uint (enc, MAJOR_ARRAY, len + 1);
198 477
478 if (ecb_expect_false (SvMAGICAL (av)))
199 for (i = 0; i <= len; ++i) 479 for (i = 0; i <= len; ++i)
200 { 480 {
201 SV **svp = av_fetch (av, i, 0); 481 SV **svp = av_fetch (av, i, 0);
202 encode_sv (enc, svp ? *svp : &PL_sv_undef); 482 encode_sv (enc, svp ? *svp : &PL_sv_undef);
203 } 483 }
484 else
485 for (i = 0; i <= len; ++i)
486 {
487 SV *sv = AvARRAY (av)[i];
488 encode_sv (enc, sv ? sv : &PL_sv_undef);
489 }
204 490
205 --enc->depth; 491 --enc->depth;
206} 492}
207 493
208static void 494static void
216 ++enc->depth; 502 ++enc->depth;
217 503
218 int pairs = hv_iterinit (hv); 504 int pairs = hv_iterinit (hv);
219 int mg = SvMAGICAL (hv); 505 int mg = SvMAGICAL (hv);
220 506
221 if (mg) 507 if (ecb_expect_false (mg))
222 encode_ch (enc, 0xa0 | 31); 508 encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
223 else 509 else
224 encode_uint (enc, 0xa0, pairs); 510 encode_uint (enc, MAJOR_MAP, pairs);
225 511
226 while ((he = hv_iternext (hv))) 512 while ((he = hv_iternext (hv)))
227 { 513 {
228 if (HeKLEN (he) == HEf_SVKEY) 514 if (HeKLEN (he) == HEf_SVKEY)
229 encode_sv (enc, HeSVKEY (he)); 515 encode_sv (enc, HeSVKEY (he));
230 else 516 else
231 encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he)); 517 encode_strref (enc, enc->cbor.flags & (F_TEXT_KEYS | F_TEXT_STRINGS), HeKUTF8 (he), HeKEY (he), HeKLEN (he));
232 518
233 encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he)); 519 encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
234 } 520 }
235 521
236 if (mg) 522 if (ecb_expect_false (mg))
237 encode_ch (enc, 0xe0 | 31); 523 encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
238 524
239 --enc->depth; 525 --enc->depth;
240} 526}
241 527
242// encode objects, arrays and special \0=false and \1=true values. 528// encode objects, arrays and special \0=false and \1=true values.
243static void 529static void
244encode_rv (enc_t *enc, SV *sv) 530encode_rv (enc_t *enc, SV *sv)
245{ 531{
246 svtype svt;
247
248 SvGETMAGIC (sv); 532 SvGETMAGIC (sv);
533
249 svt = SvTYPE (sv); 534 svtype svt = SvTYPE (sv);
250 535
251 if (expect_false (SvOBJECT (sv))) 536 if (ecb_expect_false (SvOBJECT (sv)))
252 { 537 {
253 HV *stash = !CBOR_SLOW || cbor_boolean_stash 538 HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
254 ? cbor_boolean_stash 539 ? types_boolean_stash
540 : gv_stashpv ("Types::Serialiser::Boolean", 1);
541 HV *error_stash = !CBOR_SLOW || types_error_stash
542 ? types_error_stash
543 : gv_stashpv ("Types::Serialiser::Error", 1);
544 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
545 ? cbor_tagged_stash
255 : gv_stashpv ("CBOR::XS::Boolean", 1); 546 : gv_stashpv ("CBOR::XS::Tagged" , 1);
256 547
257 if (SvSTASH (sv) == stash) 548 HV *stash = SvSTASH (sv);
258 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); 549
550 if (stash == boolean_stash)
551 {
552 encode_ch (enc, SvIV (sv) ? MAJOR_MISC | SIMPLE_TRUE : MAJOR_MISC | SIMPLE_FALSE);
553 return;
554 }
555 else if (stash == error_stash)
556 {
557 encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
558 return;
559 }
560 else if (stash == tagged_stash)
561 {
562 if (svt != SVt_PVAV)
563 croak ("encountered CBOR::XS::Tagged object that isn't an array");
564
565 switch (av_len ((AV *)sv))
566 {
567 case 2-1:
568 // actually a tagged value
569 encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1)));
570 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
571 break;
572
573 case 3-1:
574 // a forced type [value, type, undef]
575 encode_forced (enc, SvUV (*av_fetch ((AV *)sv, 1, 1)), *av_fetch ((AV *)sv, 0, 1));
576 break;
577
578 default:
579 croak ("encountered malformed CBOR::XS::Tagged object");
580 }
581
582 return;
583 }
584 }
585
586 if (ecb_expect_false (SvREFCNT (sv) > 1)
587 && ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING))
588 {
589 if (ecb_expect_false (!enc->shareable))
590 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
591
592 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
593
594 if (SvOK (*svp))
595 {
596 encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
597 encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
598 return;
599 }
259 else 600 else
260 { 601 {
261#if 0 //TODO 602 sv_setuv (*svp, enc->shareable_idx);
262 if (enc->cbor.flags & F_CONV_BLESSED) 603 ++enc->shareable_idx;
604 encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
605 }
606 }
607
608 if (ecb_expect_false (SvOBJECT (sv)))
609 {
610 HV *stash = SvSTASH (sv);
611 GV *method;
612
613 if (enc->cbor.flags & F_FORBID_OBJECTS)
614 croak ("encountered object '%s', but forbid_objects is enabled",
615 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
616 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
263 { 617 {
618 dSP;
619
620 ENTER; SAVETMPS;
621 PUSHMARK (SP);
264 // we re-bless the reference to get overload and other niceties right 622 // we re-bless the reference to get overload and other niceties right
265 GV *to_cbor = gv_fetchmethod_autoload (SvSTASH (sv), "TO_CBOR", 0);
266
267 if (to_cbor)
268 {
269 dSP;
270
271 ENTER; SAVETMPS; PUSHMARK (SP);
272 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); 623 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
273 624
274 // calling with G_SCALAR ensures that we always get a 1 return value
275 PUTBACK; 625 PUTBACK;
626 // G_SCALAR ensures that return value is 1
276 call_sv ((SV *)GvCV (to_cbor), G_SCALAR); 627 call_sv ((SV *)GvCV (method), G_SCALAR);
277 SPAGAIN; 628 SPAGAIN;
278 629
279 // catch this surprisingly common error 630 // catch this surprisingly common error
280 if (SvROK (TOPs) && SvRV (TOPs) == sv) 631 if (SvROK (TOPs) && SvRV (TOPs) == sv)
281 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); 632 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
282 633
283 sv = POPs;
284 PUTBACK;
285
286 encode_sv (enc, sv); 634 encode_sv (enc, POPs);
287 635
636 PUTBACK;
637
288 FREETMPS; LEAVE; 638 FREETMPS; LEAVE;
289 } 639 }
290 else if (enc->cbor.flags & F_ALLOW_BLESSED) 640 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
291 encode_str (enc, "null", 4, 0); 641 {
292 else 642 dSP;
293 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it", 643
294 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 644 ENTER; SAVETMPS;
645 PUSHMARK (SP);
646 EXTEND (SP, 2);
647 // we re-bless the reference to get overload and other niceties right
648 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
649 PUSHs (sv_cbor);
650
651 PUTBACK;
652 int count = call_sv ((SV *)GvCV (method), G_ARRAY);
653 SPAGAIN;
654
655 // catch this surprisingly common error
656 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
657 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
658
659 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
660 encode_uint (enc, MAJOR_ARRAY, count + 1);
661 encode_strref (enc, 0, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
662
663 {
664 int i;
665
666 for (i = 0; i < count; ++i)
667 encode_sv (enc, SP[i + 1 - count]);
668
669 SP -= count;
295 } 670 }
296 else if (enc->cbor.flags & F_ALLOW_BLESSED) 671
297 encode_str (enc, "null", 4, 0); 672 PUTBACK;
673
674 FREETMPS; LEAVE;
675 }
298 else 676 else
299 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 677 croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
300 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 678 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
301#endif
302 }
303 } 679 }
304 else if (svt == SVt_PVHV) 680 else if (svt == SVt_PVHV)
305 encode_hv (enc, (HV *)sv); 681 encode_hv (enc, (HV *)sv);
306 else if (svt == SVt_PVAV) 682 else if (svt == SVt_PVAV)
307 encode_av (enc, (AV *)sv); 683 encode_av (enc, (AV *)sv);
308 else if (svt < SVt_PVAV)
309 {
310 STRLEN len = 0;
311 char *pv = svt ? SvPV (sv, len) : 0;
312
313 if (len == 1 && *pv == '1')
314 encode_ch (enc, 0xe0 | 21);
315 else if (len == 1 && *pv == '0')
316 encode_ch (enc, 0xe0 | 20);
317 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
318 encode_ch (enc, 0xe0 | 23);
319 else
320 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
321 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
322 }
323 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
324 encode_ch (enc, 0xe0 | 23);
325 else 684 else
326 croak ("encountered %s, but CBOR can only represent references to arrays or hashes", 685 {
327 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 686 encode_tag (enc, CBOR_TAG_INDIRECTION);
687 encode_sv (enc, sv);
688 }
328} 689}
329 690
330static void 691static void
331encode_nv (enc_t *enc, SV *sv) 692encode_nv (enc_t *enc, SV *sv)
332{ 693{
333 double nv = SvNVX (sv); 694 double nv = SvNVX (sv);
334 695
335 need (enc, 9); 696 need (enc, 9);
336 697
337 if (expect_false (nv == (U32)nv)) 698 if (ecb_expect_false (nv == (NV)(U32)nv))
338 encode_uint (enc, 0x00, (U32)nv); 699 encode_uint (enc, MAJOR_POS_INT, (U32)nv);
339 //TODO: maybe I32? 700 //TODO: maybe I32?
340 else if (expect_false (nv == (float)nv)) 701 else if (ecb_expect_false (nv == (float)nv))
341 { 702 encode_float32 (enc, nv);
342 uint32_t fp = ecb_float_to_binary32 (nv);
343
344 *enc->cur++ = 0xe0 | 26;
345
346 if (!ecb_big_endian ())
347 fp = ecb_bswap32 (fp);
348
349 memcpy (enc->cur, &fp, 4);
350 enc->cur += 4;
351 }
352 else 703 else
353 { 704 encode_float64 (enc, nv);
354 uint64_t fp = ecb_double_to_binary64 (nv);
355
356 *enc->cur++ = 0xe0 | 27;
357
358 if (!ecb_big_endian ())
359 fp = ecb_bswap64 (fp);
360
361 memcpy (enc->cur, &fp, 8);
362 enc->cur += 8;
363 }
364} 705}
365 706
366static void 707static void
367encode_sv (enc_t *enc, SV *sv) 708encode_sv (enc_t *enc, SV *sv)
368{ 709{
370 711
371 if (SvPOKp (sv)) 712 if (SvPOKp (sv))
372 { 713 {
373 STRLEN len; 714 STRLEN len;
374 char *str = SvPV (sv, len); 715 char *str = SvPV (sv, len);
375 encode_str (enc, SvUTF8 (sv), str, len); 716 encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
376 } 717 }
377 else if (SvNOKp (sv)) 718 else if (SvNOKp (sv))
378 encode_nv (enc, sv); 719 encode_nv (enc, sv);
379 else if (SvIOKp (sv)) 720 else if (SvIOKp (sv))
380 { 721 encode_int (enc, sv);
381 if (SvIsUV (sv))
382 encode_uint (enc, 0x00, SvUVX (sv));
383 else if (SvIVX (sv) >= 0)
384 encode_uint (enc, 0x00, SvIVX (sv));
385 else
386 encode_uint (enc, 0x20, -(SvIVX (sv) + 1));
387 }
388 else if (SvROK (sv)) 722 else if (SvROK (sv))
389 encode_rv (enc, SvRV (sv)); 723 encode_rv (enc, SvRV (sv));
390 else if (!SvOK (sv)) 724 else if (!SvOK (sv))
391 encode_ch (enc, 0xe0 | 22); 725 encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
392 else if (enc->cbor.flags & F_ALLOW_UNKNOWN) 726 else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
393 encode_ch (enc, 0xe0 | 23); 727 encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
394 else 728 else
395 croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data", 729 croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",
396 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); 730 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
397} 731}
398 732
399static SV * 733static SV *
400encode_cbor (SV *scalar, CBOR *cbor) 734encode_cbor (SV *scalar, CBOR *cbor)
401{ 735{
402 enc_t enc; 736 enc_t enc = { 0 };
403 737
404 enc.cbor = *cbor; 738 enc.cbor = *cbor;
405 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 739 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
406 enc.cur = SvPVX (enc.sv); 740 enc.cur = SvPVX (enc.sv);
407 enc.end = SvEND (enc.sv); 741 enc.end = SvEND (enc.sv);
408 enc.depth = 0;
409 742
410 SvPOK_only (enc.sv); 743 SvPOK_only (enc.sv);
744
745 if (cbor->flags & F_PACK_STRINGS)
746 {
747 encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
748 enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
749 enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
750 }
751
411 encode_sv (&enc, scalar); 752 encode_sv (&enc, scalar);
412 753
413 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 754 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
414 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 755 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
415 756
429 U8 *end; // end of input string 770 U8 *end; // end of input string
430 const char *err; // parse error, if != 0 771 const char *err; // parse error, if != 0
431 CBOR cbor; 772 CBOR cbor;
432 U32 depth; // recursion depth 773 U32 depth; // recursion depth
433 U32 maxdepth; // recursion depth limit 774 U32 maxdepth; // recursion depth limit
775 AV *shareable;
776 AV *stringref;
777 SV *decode_tagged;
778 SV *err_sv; // optional sv for error, needs to be freed
434} dec_t; 779} dec_t;
435 780
436#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 781// set dec->err to ERRSV
782ecb_cold static void
783err_errsv (dec_t *dec)
784{
785 if (!dec->err)
786 {
787 dec->err_sv = newSVsv (ERRSV);
437 788
438#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data"); 789 // chop off the trailing \n
790 SvCUR_set (dec->err_sv, SvCUR (dec->err_sv) - 1);
791 *SvEND (dec->err_sv) = 0;
439 792
793 dec->err = SvPVutf8_nolen (dec->err_sv);
794 }
795}
796
797// the following functions are used to reduce code size and help the compiler to optimise
798ecb_cold static void
799err_set (dec_t *dec, const char *reason)
800{
801 if (!dec->err)
802 dec->err = reason;
803}
804
805ecb_cold static void
806err_unexpected_end (dec_t *dec)
807{
808 err_set (dec, "unexpected end of CBOR data");
809}
810
811#define ERR_DO(do) SB do; goto fail; SE
812#define ERR(reason) ERR_DO (err_set (dec, reason))
813#define ERR_ERRSV ERR_DO (err_errsv (dec))
814
815#define WANT(len) if (ecb_expect_false ((uintptr_t)(dec->end - dec->cur) < (STRLEN)len)) ERR_DO (err_unexpected_end (dec))
816
440#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED) 817#define DEC_INC_DEPTH if (ecb_expect_false (++dec->depth > dec->cbor.max_depth)) ERR (ERR_NESTING_EXCEEDED)
441#define DEC_DEC_DEPTH --dec->depth 818#define DEC_DEC_DEPTH --dec->depth
442 819
443static UV 820static UV
444decode_uint (dec_t *dec) 821decode_uint (dec_t *dec)
445{ 822{
446 switch (*dec->cur & 31) 823 U8 m = *dec->cur & MINOR_MASK;
447 { 824 ++dec->cur;
448 case 0: case 1: case 2: case 3: case 4: case 5: case 6: case 7:
449 case 8: case 9: case 10: case 11: case 12: case 13: case 14: case 15:
450 case 16: case 17: case 18: case 19: case 20: case 21: case 22: case 23:
451 return *dec->cur++ & 31;
452 825
453 case 24: 826 if (ecb_expect_true (m < LENGTH_EXT1))
827 return m;
828 else if (ecb_expect_true (m == LENGTH_EXT1))
829 {
454 WANT (2); 830 WANT (1);
455 dec->cur += 2; 831 dec->cur += 1;
456 return dec->cur[-1]; 832 return dec->cur[-1];
457 833 }
458 case 25: 834 else if (ecb_expect_true (m == LENGTH_EXT2))
835 {
459 WANT (3); 836 WANT (2);
460 dec->cur += 3; 837 dec->cur += 2;
461 return (((UV)dec->cur[-2]) << 8) 838 return (((UV)dec->cur[-2]) << 8)
462 | ((UV)dec->cur[-1]); 839 | ((UV)dec->cur[-1]);
463 840 }
464 case 26: 841 else if (ecb_expect_true (m == LENGTH_EXT4))
842 {
465 WANT (5); 843 WANT (4);
466 dec->cur += 5; 844 dec->cur += 4;
467 return (((UV)dec->cur[-4]) << 24) 845 return (((UV)dec->cur[-4]) << 24)
468 | (((UV)dec->cur[-3]) << 16) 846 | (((UV)dec->cur[-3]) << 16)
469 | (((UV)dec->cur[-2]) << 8) 847 | (((UV)dec->cur[-2]) << 8)
470 | ((UV)dec->cur[-1]); 848 | ((UV)dec->cur[-1]);
471 849 }
472 case 27: 850 else if (ecb_expect_true (m == LENGTH_EXT8))
851 {
473 WANT (9); 852 WANT (8);
474 dec->cur += 9; 853 dec->cur += 8;
854
855 return
856#if UVSIZE < 8
857 0
858#else
475 return (((UV)dec->cur[-8]) << 56) 859 (((UV)dec->cur[-8]) << 56)
476 | (((UV)dec->cur[-7]) << 48) 860 | (((UV)dec->cur[-7]) << 48)
477 | (((UV)dec->cur[-6]) << 40) 861 | (((UV)dec->cur[-6]) << 40)
478 | (((UV)dec->cur[-5]) << 32) 862 | (((UV)dec->cur[-5]) << 32)
863#endif
479 | (((UV)dec->cur[-4]) << 24) 864 | (((UV)dec->cur[-4]) << 24)
480 | (((UV)dec->cur[-3]) << 16) 865 | (((UV)dec->cur[-3]) << 16)
481 | (((UV)dec->cur[-2]) << 8) 866 | (((UV)dec->cur[-2]) << 8)
482 | ((UV)dec->cur[-1]); 867 | ((UV)dec->cur[-1]);
483 868 }
484 default: 869 else
485 ERR ("corrupted CBOR data (unsupported integer minor encoding)"); 870 ERR ("corrupted CBOR data (unsupported integer minor encoding)");
486 }
487 871
488fail: 872fail:
489 return 0; 873 return 0;
490} 874}
491 875
496{ 880{
497 AV *av = newAV (); 881 AV *av = newAV ();
498 882
499 DEC_INC_DEPTH; 883 DEC_INC_DEPTH;
500 884
501 if ((*dec->cur & 31) == 31) 885 if (*dec->cur == (MAJOR_ARRAY | MINOR_INDEF))
502 { 886 {
503 ++dec->cur; 887 ++dec->cur;
504 888
505 for (;;) 889 for (;;)
506 { 890 {
507 WANT (1); 891 WANT (1);
508 892
509 if (*dec->cur == 0xe0 | 31) 893 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
510 { 894 {
511 ++dec->cur; 895 ++dec->cur;
512 break; 896 break;
513 } 897 }
514 898
515 av_push (av, decode_sv (dec)); 899 av_push (av, decode_sv (dec));
516 } 900 }
517 } 901 }
518 else 902 else
519 { 903 {
520 int i, len = decode_uint (dec); 904 UV i, len = decode_uint (dec);
521 905
906 WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays
522 av_fill (av, len - 1); 907 av_fill (av, len - 1);
523 908
524 for (i = 0; i < len; ++i) 909 for (i = 0; i < len; ++i)
525 AvARRAY (av)[i] = decode_sv (dec); 910 AvARRAY (av)[i] = decode_sv (dec);
526 } 911 }
527 912
528 DEC_DEC_DEPTH; 913 DEC_DEC_DEPTH;
529 return newRV_noinc ((SV *)av); 914 return newRV_noinc ((SV *)av);
530 915
531fail: 916fail:
532 SvREFCNT_dec (av); 917 SvREFCNT_dec_NN (av);
533 DEC_DEC_DEPTH; 918 DEC_DEC_DEPTH;
534 return &PL_sv_undef; 919 return &PL_sv_undef;
535} 920}
536 921
922static void
923decode_he (dec_t *dec, HV *hv)
924{
925 // for speed reasons, we specialcase single-string
926 // byte or utf-8 strings as keys, but only when !stringref
927
928 if (ecb_expect_true (!dec->stringref))
929 if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8))
930 {
931 STRLEN len = decode_uint (dec);
932 char *key = (char *)dec->cur;
933
934 WANT (len);
935 dec->cur += len;
936
937 hv_store (hv, key, len, decode_sv (dec), 0);
938
939 return;
940 }
941 else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8))
942 {
943 STRLEN len = decode_uint (dec);
944 char *key = (char *)dec->cur;
945
946 WANT (len);
947 dec->cur += len;
948
949 if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
950 if (!is_utf8_string (key, len))
951 ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
952
953 hv_store (hv, key, -len, decode_sv (dec), 0);
954
955 return;
956 }
957
958 SV *k = decode_sv (dec);
959 SV *v = decode_sv (dec);
960
961 // we leak memory if uncaught exceptions are thrown by random magical
962 // methods, and this is hopefully the only place where it can happen,
963 // so if there is a chance of an exception, take the very slow path.
964 // since catching exceptions is "undocumented/internal/forbidden" by
965 // the new p5p powers, we need to call out to a perl function :/
966 if (ecb_expect_false (SvAMAGIC (k)))
967 {
968 dSP;
969
970 ENTER; SAVETMPS;
971 PUSHMARK (SP);
972 EXTEND (SP, 3);
973 PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
974 PUSHs (sv_2mortal (k));
975 PUSHs (sv_2mortal (v));
976
977 PUTBACK;
978 call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
979 SPAGAIN;
980
981 FREETMPS; LEAVE;
982
983 if (SvTRUE (ERRSV))
984 ERR_ERRSV;
985
986 return;
987 }
988
989 hv_store_ent (hv, k, v, 0);
990 SvREFCNT_dec_NN (k);
991
992fail:
993 ;
994}
995
537static SV * 996static SV *
538decode_hv (dec_t *dec) 997decode_hv (dec_t *dec)
539{ 998{
540 HV *hv = newHV (); 999 HV *hv = newHV ();
541 1000
542 DEC_INC_DEPTH; 1001 DEC_INC_DEPTH;
543 1002
544 if ((*dec->cur & 31) == 31) 1003 if (*dec->cur == (MAJOR_MAP | MINOR_INDEF))
545 { 1004 {
546 ++dec->cur; 1005 ++dec->cur;
547 1006
548 for (;;) 1007 for (;;)
549 { 1008 {
550 WANT (1); 1009 WANT (1);
551 1010
552 if (*dec->cur == 0xe0 | 31) 1011 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
553 { 1012 {
554 ++dec->cur; 1013 ++dec->cur;
555 break; 1014 break;
556 } 1015 }
557 1016
558 SV *k = decode_sv (dec); 1017 decode_he (dec, hv);
559 SV *v = decode_sv (dec);
560
561 hv_store_ent (hv, k, v, 0);
562 } 1018 }
563 } 1019 }
564 else 1020 else
565 { 1021 {
566 int len = decode_uint (dec); 1022 UV pairs = decode_uint (dec);
567 1023
1024 WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
1025
568 while (len--) 1026 while (pairs--)
569 { 1027 decode_he (dec, hv);
570 SV *k = decode_sv (dec);
571 SV *v = decode_sv (dec);
572
573 hv_store_ent (hv, k, v, 0);
574 }
575 } 1028 }
576 1029
577 DEC_DEC_DEPTH; 1030 DEC_DEC_DEPTH;
578 return newRV_noinc ((SV *)hv); 1031 return newRV_noinc ((SV *)hv);
579 1032
580#if 0 1033fail:
581 SV *sv; 1034 SvREFCNT_dec_NN (hv);
582 HV *hv = newHV ();
583
584 DEC_INC_DEPTH; 1035 DEC_DEC_DEPTH;
585 decode_ws (dec); 1036 return &PL_sv_undef;
1037}
586 1038
587 for (;;) 1039static SV *
588 { 1040decode_str (dec_t *dec, int utf8)
589 // heuristic: assume that 1041{
590 // a) decode_str + hv_store_ent are abysmally slow. 1042 SV *sv = 0;
591 // b) most hash keys are short, simple ascii text. 1043
592 // => try to "fast-match" such strings to avoid 1044 if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
593 // the overhead of decode_str + hv_store_ent.
594 { 1045 {
595 SV *value; 1046 // indefinite length strings
596 char *p = dec->cur; 1047 ++dec->cur;
597 char *e = p + 24; // only try up to 24 bytes
598 1048
1049 U8 major = *dec->cur & MAJOR_MISC;
1050
1051 sv = newSVpvn ("", 0);
1052
599 for (;;) 1053 for (;;)
600 { 1054 {
601 // the >= 0x80 is false on most architectures 1055 WANT (1);
602 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') 1056
1057 if ((*dec->cur - major) > LENGTH_EXT8)
1058 if (*dec->cur == (MAJOR_MISC | MINOR_INDEF))
603 { 1059 {
604 // slow path, back up and use decode_str 1060 ++dec->cur;
605 SV *key = decode_str (dec);
606 if (!key)
607 goto fail;
608
609 decode_ws (dec); EXPECT_CH (':');
610
611 decode_ws (dec);
612 value = decode_sv (dec);
613 if (!value)
614 {
615 SvREFCNT_dec (key);
616 goto fail;
617 }
618
619 hv_store_ent (hv, key, value, 0);
620 SvREFCNT_dec (key);
621
622 break; 1061 break;
623 } 1062 }
624 else if (*p == '"')
625 {
626 // fast path, got a simple key
627 char *key = dec->cur;
628 int len = p - key;
629 dec->cur = p + 1;
630
631 decode_ws (dec); EXPECT_CH (':');
632
633 decode_ws (dec);
634 value = decode_sv (dec);
635 if (!value)
636 goto fail;
637
638 hv_store (hv, key, len, value, 0);
639
640 break;
641 }
642
643 ++p;
644 }
645 }
646
647 decode_ws (dec);
648
649 if (*dec->cur == '}')
650 {
651 ++dec->cur;
652 break;
653 }
654
655 if (*dec->cur != ',')
656 ERR (", or } expected while parsing object/hash");
657
658 ++dec->cur;
659
660 decode_ws (dec);
661
662 if (*dec->cur == '}' && dec->cbor.flags & F_RELAXED)
663 {
664 ++dec->cur;
665 break;
666 }
667 }
668
669 DEC_DEC_DEPTH;
670 sv = newRV_noinc ((SV *)hv);
671
672 // check filter callbacks
673 if (dec->cbor.flags & F_HOOK)
674 {
675 if (dec->cbor.cb_sk_object && HvKEYS (hv) == 1)
676 {
677 HE *cb, *he;
678
679 hv_iterinit (hv);
680 he = hv_iternext (hv);
681 hv_iterinit (hv);
682
683 // the next line creates a mortal sv each time its called.
684 // might want to optimise this for common cases.
685 cb = hv_fetch_ent (dec->cbor.cb_sk_object, hv_iterkeysv (he), 0, 0);
686
687 if (cb)
688 { 1063 else
689 dSP; 1064 ERR ("corrupted CBOR data (invalid chunks in indefinite length string)");
690 int count;
691 1065
692 ENTER; SAVETMPS; PUSHMARK (SP); 1066 STRLEN len = decode_uint (dec);
693 XPUSHs (HeVAL (he));
694 sv_2mortal (sv);
695 1067
696 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
697
698 if (count == 1)
699 {
700 sv = newSVsv (POPs);
701 FREETMPS; LEAVE;
702 return sv;
703 }
704
705 SvREFCNT_inc (sv);
706 FREETMPS; LEAVE;
707 }
708 }
709
710 if (dec->cbor.cb_object)
711 {
712 dSP;
713 int count;
714
715 ENTER; SAVETMPS; PUSHMARK (SP);
716 XPUSHs (sv_2mortal (sv));
717
718 PUTBACK; count = call_sv (dec->cbor.cb_object, G_ARRAY); SPAGAIN;
719
720 if (count == 1)
721 {
722 sv = newSVsv (POPs);
723 FREETMPS; LEAVE;
724 return sv;
725 }
726
727 SvREFCNT_inc (sv);
728 FREETMPS; LEAVE;
729 }
730 }
731
732 return sv;
733#endif
734
735fail:
736 SvREFCNT_dec (hv);
737 DEC_DEC_DEPTH;
738 return &PL_sv_undef;
739}
740
741static SV *
742decode_str (dec_t *dec, int utf8)
743{
744 SV *sv;
745
746 if ((*dec->cur & 31) == 31)
747 {
748 ++dec->cur;
749
750 sv = newSVpvn ("", 0);
751
752 // not very fast, and certainly not robust against illegal input
753 for (;;)
754 {
755 WANT (1); 1068 WANT (len);
756 1069 sv_catpvn (sv, dec->cur, len);
757 if (*dec->cur == 0xe0 | 31) 1070 dec->cur += len;
758 {
759 ++dec->cur;
760 break;
761 }
762
763 SV *sv2 = decode_sv (dec);
764 sv_catsv (sv, sv2);
765 } 1071 }
766 } 1072 }
767 else 1073 else
768 { 1074 {
769 STRLEN len = decode_uint (dec); 1075 STRLEN len = decode_uint (dec);
770 1076
771 WANT (len); 1077 WANT (len);
772 sv = newSVpvn (dec->cur, len); 1078 sv = newSVpvn (dec->cur, len);
773 dec->cur += len; 1079 dec->cur += len;
1080
1081 if (ecb_expect_false (dec->stringref)
1082 && SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
1083 av_push (dec->stringref, SvREFCNT_inc_NN (sv));
774 } 1084 }
775 1085
776 if (utf8) 1086 if (utf8)
1087 {
1088 if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
1089 if (!is_utf8_string (SvPVX (sv), SvCUR (sv)))
1090 ERR ("corrupted CBOR data (invalid UTF-8 in text string)");
1091
777 SvUTF8_on (sv); 1092 SvUTF8_on (sv);
1093 }
778 1094
779 return sv; 1095 return sv;
780 1096
781fail: 1097fail:
1098 SvREFCNT_dec (sv);
1099 return &PL_sv_undef;
1100}
1101
1102static SV *
1103decode_tagged (dec_t *dec)
1104{
1105 SV *sv = 0;
1106 UV tag = decode_uint (dec);
1107
1108 WANT (1);
1109
1110 switch (tag)
1111 {
1112 case CBOR_TAG_MAGIC:
1113 sv = decode_sv (dec);
1114 break;
1115
1116 case CBOR_TAG_INDIRECTION:
1117 sv = newRV_noinc (decode_sv (dec));
1118 break;
1119
1120 case CBOR_TAG_STRINGREF_NAMESPACE:
1121 {
1122 // do not use SAVETMPS/FREETMPS, as these will
1123 // erase mortalised caches, e.g. "shareable"
1124 ENTER;
1125
1126 SAVESPTR (dec->stringref);
1127 dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
1128
1129 sv = decode_sv (dec);
1130
1131 LEAVE;
1132 }
1133 break;
1134
1135 case CBOR_TAG_STRINGREF:
1136 {
1137 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
1138 ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
1139
1140 UV idx = decode_uint (dec);
1141
1142 if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
1143 ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
1144
1145 sv = newSVsv (AvARRAY (dec->stringref)[idx]);
1146 }
1147 break;
1148
1149 case CBOR_TAG_VALUE_SHAREABLE:
1150 {
1151 if (ecb_expect_false (!dec->shareable))
1152 dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
1153
1154 if (dec->cbor.flags & F_ALLOW_CYCLES)
1155 {
1156 sv = newSV (0);
1157 av_push (dec->shareable, SvREFCNT_inc_NN (sv));
1158
1159 SV *osv = decode_sv (dec);
1160 sv_setsv (sv, osv);
1161 SvREFCNT_dec_NN (osv);
1162 }
1163 else
1164 {
1165 av_push (dec->shareable, &PL_sv_undef);
1166 int idx = AvFILLp (dec->shareable);
1167 sv = decode_sv (dec);
1168 av_store (dec->shareable, idx, SvREFCNT_inc_NN (sv));
1169 }
1170 }
1171 break;
1172
1173 case CBOR_TAG_VALUE_SHAREDREF:
1174 {
1175 if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
1176 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
1177
1178 UV idx = decode_uint (dec);
1179
1180 if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
1181 ERR ("corrupted CBOR data (sharedref index out of bounds)");
1182
1183 sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
1184
1185 if (sv == &PL_sv_undef)
1186 ERR ("cyclic CBOR data structure found, but allow_cycles is not enabled");
1187 }
1188 break;
1189
1190 case CBOR_TAG_PERL_OBJECT:
1191 {
1192 if (dec->cbor.flags & F_FORBID_OBJECTS)
1193 goto filter;
1194
1195 sv = decode_sv (dec);
1196
1197 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
1198 ERR ("corrupted CBOR data (non-array perl object)");
1199
1200 AV *av = (AV *)SvRV (sv);
1201 int len = av_len (av) + 1;
1202 HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
1203
1204 if (!stash)
1205 ERR ("cannot decode perl-object (package does not exist)");
1206
1207 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1208
1209 if (!method)
1210 ERR ("cannot decode perl-object (package does not have a THAW method)");
1211
1212 dSP;
1213
1214 ENTER; SAVETMPS;
1215 PUSHMARK (SP);
1216 EXTEND (SP, len + 1);
1217 // we re-bless the reference to get overload and other niceties right
1218 PUSHs (*av_fetch (av, 0, 1));
1219 PUSHs (sv_cbor);
1220
1221 int i;
1222
1223 for (i = 1; i < len; ++i)
1224 PUSHs (*av_fetch (av, i, 1));
1225
1226 PUTBACK;
1227 call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
1228 SPAGAIN;
1229
1230 if (SvTRUE (ERRSV))
1231 {
1232 FREETMPS; LEAVE;
1233 ERR_ERRSV;
1234 }
1235
1236 SvREFCNT_dec_NN (sv);
1237 sv = SvREFCNT_inc (POPs);
1238
1239 PUTBACK;
1240
1241 FREETMPS; LEAVE;
1242 }
1243 break;
1244
1245 default:
1246 filter:
1247 {
1248 SV *tag_sv = newSVuv (tag);
1249
1250 sv = decode_sv (dec);
1251
1252 dSP;
1253 ENTER; SAVETMPS;
1254 PUSHMARK (SP);
1255 EXTEND (SP, 2);
1256 PUSHs (tag_sv);
1257 PUSHs (sv);
1258
1259 PUTBACK;
1260 int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
1261 SPAGAIN;
1262
1263 if (SvTRUE (ERRSV))
1264 {
1265 SvREFCNT_dec_NN (tag_sv);
1266 FREETMPS; LEAVE;
1267 ERR_ERRSV;
1268 }
1269
1270 if (count)
1271 {
1272 SvREFCNT_dec_NN (tag_sv);
1273 SvREFCNT_dec_NN (sv);
1274 sv = SvREFCNT_inc_NN (TOPs);
1275 SP -= count;
1276 }
1277 else
1278 {
1279 AV *av = newAV ();
1280 av_push (av, tag_sv);
1281 av_push (av, sv);
1282
1283 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
1284 ? cbor_tagged_stash
1285 : gv_stashpv ("CBOR::XS::Tagged" , 1);
1286 sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
1287 }
1288
1289 PUTBACK;
1290
1291 FREETMPS; LEAVE;
1292 }
1293 break;
1294 }
1295
1296 return sv;
1297
1298fail:
1299 SvREFCNT_dec (sv);
782 return &PL_sv_undef; 1300 return &PL_sv_undef;
783} 1301}
784 1302
785static SV * 1303static SV *
786decode_sv (dec_t *dec) 1304decode_sv (dec_t *dec)
787{ 1305{
788 WANT (1); 1306 WANT (1);
789 1307
790 switch (*dec->cur >> 5) 1308 switch (*dec->cur >> MAJOR_SHIFT)
791 { 1309 {
792 case 0: // unsigned int 1310 case MAJOR_POS_INT >> MAJOR_SHIFT: return newSVuv (decode_uint (dec));
793 //TODO: 64 bit values on 3 2bit perls 1311 case MAJOR_NEG_INT >> MAJOR_SHIFT: return newSViv (-1 - (IV)decode_uint (dec));
794 return newSVuv (decode_uint (dec)); 1312 case MAJOR_BYTES >> MAJOR_SHIFT: return decode_str (dec, 0);
795 case 1: // negative int 1313 case MAJOR_TEXT >> MAJOR_SHIFT: return decode_str (dec, 1);
796 return newSViv (-1 - (IV)decode_uint (dec)); 1314 case MAJOR_ARRAY >> MAJOR_SHIFT: return decode_av (dec);
797 case 2: // octet string 1315 case MAJOR_MAP >> MAJOR_SHIFT: return decode_hv (dec);
798 return decode_str (dec, 0); 1316 case MAJOR_TAG >> MAJOR_SHIFT: return decode_tagged (dec);
799 case 3: // utf-8 string 1317
800 return decode_str (dec, 1); 1318 case MAJOR_MISC >> MAJOR_SHIFT:
801 case 4: // array
802 return decode_av (dec);
803 case 5: // map
804 return decode_hv (dec);
805 case 6: // tag
806 abort ();
807 break;
808 case 7: // misc
809 switch (*dec->cur++ & 31) 1319 switch (*dec->cur++ & MINOR_MASK)
810 { 1320 {
811 case 20: 1321 case SIMPLE_FALSE:
812#if CBOR_SLOW 1322#if CBOR_SLOW
813 cbor_false = get_bool ("CBOR::XS::false"); 1323 types_false = get_bool ("Types::Serialiser::false");
814#endif 1324#endif
815 return newSVsv (cbor_false); 1325 return newSVsv (types_false);
816 case 21: 1326 case SIMPLE_TRUE:
817#if CBOR_SLOW 1327#if CBOR_SLOW
818 cbor_true = get_bool ("CBOR::XS::true"); 1328 types_true = get_bool ("Types::Serialiser::true");
819#endif 1329#endif
820 return newSVsv (cbor_true); 1330 return newSVsv (types_true);
821 case 22: 1331 case SIMPLE_NULL:
822 return newSVsv (&PL_sv_undef); 1332 return newSVsv (&PL_sv_undef);
1333 case SIMPLE_UNDEF:
1334#if CBOR_SLOW
1335 types_error = get_bool ("Types::Serialiser::error");
1336#endif
1337 return newSVsv (types_error);
823 1338
824 case 25: 1339 case MISC_FLOAT16:
825 // half float
826 abort ();
827 break; 1340 {
1341 WANT (2);
828 1342
1343 uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
1344 dec->cur += 2;
1345
1346 return newSVnv (ecb_binary16_to_float (fp));
1347 }
1348
829 case 26: 1349 case MISC_FLOAT32:
830 { 1350 {
831 uint32_t fp; 1351 uint32_t fp;
832 WANT (4); 1352 WANT (4);
833 memcpy (&fp, dec->cur, 4); 1353 memcpy (&fp, dec->cur, 4);
834 dec->cur += 4; 1354 dec->cur += 4;
837 fp = ecb_bswap32 (fp); 1357 fp = ecb_bswap32 (fp);
838 1358
839 return newSVnv (ecb_binary32_to_float (fp)); 1359 return newSVnv (ecb_binary32_to_float (fp));
840 } 1360 }
841 1361
842 case 27: 1362 case MISC_FLOAT64:
843 { 1363 {
844 uint64_t fp; 1364 uint64_t fp;
845 WANT (8); 1365 WANT (8);
846 memcpy (&fp, dec->cur, 8); 1366 memcpy (&fp, dec->cur, 8);
847 dec->cur += 8; 1367 dec->cur += 8;
850 fp = ecb_bswap64 (fp); 1370 fp = ecb_bswap64 (fp);
851 1371
852 return newSVnv (ecb_binary64_to_double (fp)); 1372 return newSVnv (ecb_binary64_to_double (fp));
853 } 1373 }
854 1374
855 // 0..19 unassigned 1375 // 0..19 unassigned simple
856 // 24 reserved + unassigned (reserved values are not encodable) 1376 // 24 reserved + unassigned simple (reserved values are not encodable)
1377 // 28-30 unassigned misc
1378 // 31 break code
857 default: 1379 default:
858 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); 1380 ERR ("corrupted CBOR data (reserved/unassigned/unexpected major 7 value)");
859 } 1381 }
860 1382
861 break; 1383 break;
862 } 1384 }
863#if 0
864 switch (*dec->cur)
865 {
866 //case '"': ++dec->cur; return decode_str (dec);
867 case '[': ++dec->cur; return decode_av (dec);
868 case '{': ++dec->cur; return decode_hv (dec);
869
870 case '-':
871 case '0': case '1': case '2': case '3': case '4':
872 case '5': case '6': case '7': case '8': case '9':
873 //TODO return decode_num (dec);
874
875 case 't':
876 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
877 {
878 dec->cur += 4;
879#if CBOR_SLOW
880 cbor_true = get_bool ("CBOR::XS::true");
881#endif
882 return newSVsv (cbor_true);
883 }
884 else
885 ERR ("'true' expected");
886
887 break;
888
889 case 'f':
890 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
891 {
892 dec->cur += 5;
893#if CBOR_SLOW
894 cbor_false = get_bool ("CBOR::XS::false");
895#endif
896 return newSVsv (cbor_false);
897 }
898 else
899 ERR ("'false' expected");
900
901 break;
902
903 case 'n':
904 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
905 {
906 dec->cur += 4;
907 return newSVsv (&PL_sv_undef);
908 }
909 else
910 ERR ("'null' expected");
911
912 break;
913
914 default:
915 ERR ("malformed CBOR string, neither array, object, number, string or atom");
916 break;
917 }
918#endif
919 1385
920fail: 1386fail:
921 return &PL_sv_undef; 1387 return &PL_sv_undef;
922} 1388}
923 1389
924static SV * 1390static SV *
925decode_cbor (SV *string, CBOR *cbor, char **offset_return) 1391decode_cbor (SV *string, CBOR *cbor, char **offset_return)
926{ 1392{
927 dec_t dec; 1393 dec_t dec = { 0 };
928 SV *sv; 1394 SV *sv;
1395 STRLEN len;
1396 char *data = SvPVbyte (string, len);
929 1397
930 /* work around bugs in 5.10 where manipulating magic values
931 * makes perl ignore the magic in subsequent accesses.
932 * also make a copy of non-PV values, to get them into a clean
933 * state (SvPV should do that, but it's buggy, see below).
934 */
935 /*SvGETMAGIC (string);*/
936 if (SvMAGICAL (string) || !SvPOK (string))
937 string = sv_2mortal (newSVsv (string));
938
939 SvUPGRADE (string, SVt_PV);
940
941 /* work around a bug in perl 5.10, which causes SvCUR to fail an
942 * assertion with -DDEBUGGING, although SvCUR is documented to
943 * return the xpv_cur field which certainly exists after upgrading.
944 * according to nicholas clark, calling SvPOK fixes this.
945 * But it doesn't fix it, so try another workaround, call SvPV_nolen
946 * and hope for the best.
947 * Damnit, SvPV_nolen still trips over yet another assertion. This
948 * assertion business is seriously broken, try yet another workaround
949 * for the broken -DDEBUGGING.
950 */
951 {
952#ifdef DEBUGGING
953 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
954#else
955 STRLEN offset = SvCUR (string);
956#endif
957
958 if (offset > cbor->max_size && cbor->max_size) 1398 if (len > cbor->max_size && cbor->max_size)
959 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu", 1399 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
960 (unsigned long)SvCUR (string), (unsigned long)cbor->max_size); 1400 (unsigned long)len, (unsigned long)cbor->max_size);
961 }
962
963 sv_utf8_downgrade (string, 0);
964 1401
965 dec.cbor = *cbor; 1402 dec.cbor = *cbor;
966 dec.cur = (U8 *)SvPVX (string); 1403 dec.cur = (U8 *)data;
967 dec.end = (U8 *)SvEND (string); 1404 dec.end = (U8 *)data + len;
968 dec.err = 0;
969 dec.depth = 0;
970
971 if (dec.cbor.cb_object || dec.cbor.cb_sk_object)
972 ;//TODO dec.cbor.flags |= F_HOOK;
973 1405
974 sv = decode_sv (&dec); 1406 sv = decode_sv (&dec);
975 1407
976 if (offset_return) 1408 if (offset_return)
977 *offset_return = dec.cur; 1409 *offset_return = dec.cur;
978 1410
979 if (!(offset_return || !sv)) 1411 if (!(offset_return || !sv))
980 {
981 if (*dec.cur && !dec.err) 1412 if (dec.cur != dec.end && !dec.err)
982 {
983 dec.err = "garbage after CBOR object"; 1413 dec.err = "garbage after CBOR object";
984 SvREFCNT_dec (sv); 1414
985 sv = 0; 1415 if (dec.err)
1416 {
1417 if (dec.shareable)
986 } 1418 {
987 } 1419 // need to break cyclic links, which would all be in shareable
1420 int i;
1421 SV **svp;
988 1422
989 if (!sv) 1423 for (i = av_len (dec.shareable) + 1; i--; )
1424 if ((svp = av_fetch (dec.shareable, i, 0)))
1425 sv_setsv (*svp, &PL_sv_undef);
1426 }
1427
1428 SvREFCNT_dec_NN (sv);
1429
1430 if (dec.err_sv)
1431 sv_2mortal (dec.err_sv);
1432
990 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); 1433 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
1434 }
991 1435
992 sv = sv_2mortal (sv); 1436 sv = sv_2mortal (sv);
993 1437
994 return sv; 1438 return sv;
995} 1439}
996 1440
1441/////////////////////////////////////////////////////////////////////////////
1442// incremental parser
1443
1444#define INCR_DONE(cbor) (AvFILLp (cbor->incr_count) < 0)
1445
1446// returns 0 for notyet, 1 for success or error
1447static int
1448incr_parse (CBOR *self, SV *cborstr)
1449{
1450 STRLEN cur;
1451 SvPV (cborstr, cur);
1452
1453 while (ecb_expect_true (self->incr_need <= cur))
1454 {
1455 // table of integer count bytes
1456 static I8 incr_len[MINOR_MASK + 1] = {
1457 0, 0, 0, 0, 0, 0, 0, 0,
1458 0, 0, 0, 0, 0, 0, 0, 0,
1459 0, 0, 0, 0, 0, 0, 0, 0,
1460 1, 2, 4, 8,-1,-1,-1,-2
1461 };
1462
1463 const U8 *p = SvPVX (cborstr) + self->incr_pos;
1464 U8 m = *p & MINOR_MASK;
1465 IV count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
1466 I8 ilen = incr_len[m];
1467
1468 self->incr_need = self->incr_pos + 1;
1469
1470 if (ecb_expect_false (ilen < 0))
1471 {
1472 if (m != MINOR_INDEF)
1473 return 1; // error
1474
1475 if (*p == (MAJOR_MISC | MINOR_INDEF))
1476 {
1477 if (count >= 0)
1478 return 1; // error
1479
1480 count = 1;
1481 }
1482 else
1483 {
1484 av_push (self->incr_count, newSViv (-1)); //TODO: nest
1485 count = -1;
1486 }
1487 }
1488 else
1489 {
1490 self->incr_need += ilen;
1491 if (ecb_expect_false (self->incr_need > cur))
1492 return 0;
1493
1494 int major = *p >> MAJOR_SHIFT;
1495
1496 switch (major)
1497 {
1498 case MAJOR_TAG >> MAJOR_SHIFT:
1499 ++count; // tags merely prefix another value
1500 break;
1501
1502 case MAJOR_BYTES >> MAJOR_SHIFT:
1503 case MAJOR_TEXT >> MAJOR_SHIFT:
1504 case MAJOR_ARRAY >> MAJOR_SHIFT:
1505 case MAJOR_MAP >> MAJOR_SHIFT:
1506 {
1507 UV len;
1508
1509 if (ecb_expect_false (ilen))
1510 {
1511 len = 0;
1512
1513 do {
1514 len = (len << 8) | *++p;
1515 } while (--ilen);
1516 }
1517 else
1518 len = m;
1519
1520 switch (major)
1521 {
1522 case MAJOR_BYTES >> MAJOR_SHIFT:
1523 case MAJOR_TEXT >> MAJOR_SHIFT:
1524 self->incr_need += len;
1525 if (ecb_expect_false (self->incr_need > cur))
1526 return 0;
1527
1528 break;
1529
1530 case MAJOR_MAP >> MAJOR_SHIFT:
1531 len <<= 1;
1532 case MAJOR_ARRAY >> MAJOR_SHIFT:
1533 if (len)
1534 {
1535 av_push (self->incr_count, newSViv (len + 1)); //TODO: nest
1536 count = len + 1;
1537 }
1538 break;
1539 }
1540 }
1541 }
1542 }
1543
1544 self->incr_pos = self->incr_need;
1545
1546 if (count > 0)
1547 {
1548 while (!--count)
1549 {
1550 if (!AvFILLp (self->incr_count))
1551 return 1; // done
1552
1553 SvREFCNT_dec_NN (av_pop (self->incr_count));
1554 count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
1555 }
1556
1557 SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]) = count;
1558 }
1559 }
1560
1561 return 0;
1562}
1563
1564
997///////////////////////////////////////////////////////////////////////////// 1565/////////////////////////////////////////////////////////////////////////////
998// XS interface functions 1566// XS interface functions
999 1567
1000MODULE = CBOR::XS PACKAGE = CBOR::XS 1568MODULE = CBOR::XS PACKAGE = CBOR::XS
1001 1569
1002BOOT: 1570BOOT:
1003{ 1571{
1004 cbor_stash = gv_stashpv ("CBOR::XS" , 1); 1572 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1005 cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); 1573 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1006 1574
1007 cbor_true = get_bool ("CBOR::XS::true"); 1575 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1008 cbor_false = get_bool ("CBOR::XS::false"); 1576 types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
1577
1578 types_true = get_bool ("Types::Serialiser::true" );
1579 types_false = get_bool ("Types::Serialiser::false");
1580 types_error = get_bool ("Types::Serialiser::error");
1581
1582 default_filter = newSVpv ("CBOR::XS::default_filter", 0);
1583
1584 sv_cbor = newSVpv ("CBOR", 0);
1585 SvREADONLY_on (sv_cbor);
1586
1587 assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1009} 1588}
1010 1589
1011PROTOTYPES: DISABLE 1590PROTOTYPES: DISABLE
1012 1591
1013void CLONE (...) 1592void CLONE (...)
1014 CODE: 1593 CODE:
1015 cbor_stash = 0; 1594 cbor_stash = 0;
1595 cbor_tagged_stash = 0;
1596 types_error_stash = 0;
1016 cbor_boolean_stash = 0; 1597 types_boolean_stash = 0;
1017 1598
1018void new (char *klass) 1599void new (char *klass)
1019 PPCODE: 1600 PPCODE:
1020{ 1601{
1021 SV *pv = NEWSV (0, sizeof (CBOR)); 1602 SV *pv = NEWSV (0, sizeof (CBOR));
1029 1610
1030void shrink (CBOR *self, int enable = 1) 1611void shrink (CBOR *self, int enable = 1)
1031 ALIAS: 1612 ALIAS:
1032 shrink = F_SHRINK 1613 shrink = F_SHRINK
1033 allow_unknown = F_ALLOW_UNKNOWN 1614 allow_unknown = F_ALLOW_UNKNOWN
1615 allow_sharing = F_ALLOW_SHARING
1616 allow_cycles = F_ALLOW_CYCLES
1617 forbid_objects = F_FORBID_OBJECTS
1618 pack_strings = F_PACK_STRINGS
1619 text_keys = F_TEXT_KEYS
1620 text_strings = F_TEXT_STRINGS
1621 validate_utf8 = F_VALIDATE_UTF8
1034 PPCODE: 1622 PPCODE:
1035{ 1623{
1036 if (enable) 1624 if (enable)
1037 self->flags |= ix; 1625 self->flags |= ix;
1038 else 1626 else
1043 1631
1044void get_shrink (CBOR *self) 1632void get_shrink (CBOR *self)
1045 ALIAS: 1633 ALIAS:
1046 get_shrink = F_SHRINK 1634 get_shrink = F_SHRINK
1047 get_allow_unknown = F_ALLOW_UNKNOWN 1635 get_allow_unknown = F_ALLOW_UNKNOWN
1636 get_allow_sharing = F_ALLOW_SHARING
1637 get_allow_cycles = F_ALLOW_CYCLES
1638 get_forbid_objects = F_FORBID_OBJECTS
1639 get_pack_strings = F_PACK_STRINGS
1640 get_text_keys = F_TEXT_KEYS
1641 get_text_strings = F_TEXT_STRINGS
1642 get_validate_utf8 = F_VALIDATE_UTF8
1048 PPCODE: 1643 PPCODE:
1049 XPUSHs (boolSV (self->flags & ix)); 1644 XPUSHs (boolSV (self->flags & ix));
1050 1645
1051void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1646void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1052 PPCODE: 1647 PPCODE:
1068 CODE: 1663 CODE:
1069 RETVAL = self->max_size; 1664 RETVAL = self->max_size;
1070 OUTPUT: 1665 OUTPUT:
1071 RETVAL 1666 RETVAL
1072 1667
1073#if 0 //TODO 1668void filter (CBOR *self, SV *filter = 0)
1074
1075void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1076 PPCODE: 1669 PPCODE:
1077{
1078 SvREFCNT_dec (self->cb_object); 1670 SvREFCNT_dec (self->filter);
1079 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; 1671 self->filter = filter ? newSVsv (filter) : filter;
1080
1081 XPUSHs (ST (0)); 1672 XPUSHs (ST (0));
1082}
1083 1673
1084void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef) 1674SV *get_filter (CBOR *self)
1085 PPCODE: 1675 CODE:
1086{ 1676 RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1087 if (!self->cb_sk_object) 1677 OUTPUT:
1088 self->cb_sk_object = newHV (); 1678 RETVAL
1089
1090 if (SvOK (cb))
1091 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1092 else
1093 {
1094 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1095
1096 if (!HvKEYS (self->cb_sk_object))
1097 {
1098 SvREFCNT_dec (self->cb_sk_object);
1099 self->cb_sk_object = 0;
1100 }
1101 }
1102
1103 XPUSHs (ST (0));
1104}
1105
1106#endif
1107 1679
1108void encode (CBOR *self, SV *scalar) 1680void encode (CBOR *self, SV *scalar)
1109 PPCODE: 1681 PPCODE:
1110 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; 1682 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1111 XPUSHs (scalar); 1683 XPUSHs (scalar);
1124 EXTEND (SP, 2); 1696 EXTEND (SP, 2);
1125 PUSHs (sv); 1697 PUSHs (sv);
1126 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); 1698 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1127} 1699}
1128 1700
1701void incr_parse (CBOR *self, SV *cborstr)
1702 ALIAS:
1703 incr_parse_multiple = 1
1704 PPCODE:
1705{
1706 if (SvUTF8 (cborstr))
1707 sv_utf8_downgrade (cborstr, 0);
1708
1709 if (!self->incr_count)
1710 {
1711 self->incr_count = newAV ();
1712 self->incr_pos = 0;
1713 self->incr_need = 1;
1714
1715 av_push (self->incr_count, newSViv (1));
1716 }
1717
1718 do
1719 {
1720 if (!incr_parse (self, cborstr))
1721 {
1722 if (self->incr_need > self->max_size && self->max_size)
1723 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
1724 (unsigned long)self->incr_need, (unsigned long)self->max_size);
1725
1726 break;
1727 }
1728
1729 SV *sv;
1730 char *offset;
1731
1732 PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
1733 XPUSHs (sv);
1734
1735 sv_chop (cborstr, offset);
1736
1737 av_clear (self->incr_count);
1738 av_push (self->incr_count, newSViv (1));
1739
1740 self->incr_pos = 0;
1741 self->incr_need = self->incr_pos + 1;
1742 }
1743 while (ix);
1744}
1745
1746void incr_reset (CBOR *self)
1747 CODE:
1748{
1749 SvREFCNT_dec (self->incr_count);
1750 self->incr_count = 0;
1751}
1752
1129void DESTROY (CBOR *self) 1753void DESTROY (CBOR *self)
1130 CODE: 1754 PPCODE:
1131 SvREFCNT_dec (self->cb_sk_object); 1755 cbor_free (self);
1132 SvREFCNT_dec (self->cb_object);
1133 1756
1134PROTOTYPES: ENABLE 1757PROTOTYPES: ENABLE
1135 1758
1136void encode_cbor (SV *scalar) 1759void encode_cbor (SV *scalar)
1760 ALIAS:
1761 encode_cbor = 0
1762 encode_cbor_sharing = F_ALLOW_SHARING
1137 PPCODE: 1763 PPCODE:
1138{ 1764{
1139 CBOR cbor; 1765 CBOR cbor;
1140 cbor_init (&cbor); 1766 cbor_init (&cbor);
1767 cbor.flags |= ix;
1141 PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN; 1768 PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1142 XPUSHs (scalar); 1769 XPUSHs (scalar);
1143} 1770}
1144 1771
1145void decode_cbor (SV *cborstr) 1772void decode_cbor (SV *cborstr)
1149 cbor_init (&cbor); 1776 cbor_init (&cbor);
1150 PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN; 1777 PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1151 XPUSHs (cborstr); 1778 XPUSHs (cborstr);
1152} 1779}
1153 1780
1781#ifdef __AFL_COMPILER
1782
1783void
1784afl_init ()
1785 CODE:
1786 __AFL_INIT ();
1787
1788int
1789afl_loop (unsigned int count = 10000)
1790 CODE:
1791 RETVAL = __AFL_LOOP (count);
1792 OUTPUT:
1793 RETVAL
1794
1795#endif
1796

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines