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 1.2 by root, Sat Oct 26 10:41:12 2013 UTC vs.
Revision 1.19 by root, Wed Nov 20 02:03:09 2013 UTC

9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11 11
12#include "ecb.h" 12#include "ecb.h"
13 13
14#if defined(__BORLANDC__) || defined(_MSC_VER) 14// compatibility with perl <5.18
15# define snprintf _snprintf // C compilers have this in stdio.h 15#ifndef HvNAMELEN_get
16# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
16#endif 17#endif
18#ifndef HvNAMELEN
19# define HvNAMELEN(hv) HvNAMELEN_get (hv)
20#endif
21#ifndef HvNAMEUTF8
22# define HvNAMEUTF8(hv) 0
23#endif
17 24
25// known tags
26enum cbor_tag
27{
28 // inofficial extensions (pending iana registration)
29 CBOR_TAG_PERL_OBJECT = 24, // http://cbor.schmorp.de/perl-object
30 CBOR_TAG_GENERIC_OBJECT = 25, // http://cbor.schmorp.de/generic-object
31 CBOR_TAG_VALUE_SHAREABLE = 26, // http://cbor.schmorp.de/value-sharing
32 CBOR_TAG_VALUE_SHAREDREF = 27, // http://cbor.schmorp.de/value-sharing
33 CBOR_TAG_STRINGREF_NAMESPACE = 65537, // http://cbor.schmorp.de/stringref
34 CBOR_TAG_STRINGREF = 28, // http://cbor.schmorp.de/stringref
35 CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
36
37 // rfc7049
38 CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
39 CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
40 CBOR_TAG_POS_BIGNUM = 2, // byte string
41 CBOR_TAG_NEG_BIGNUM = 3, // byte string
42 CBOR_TAG_DECIMAL = 4, // decimal fraction, array
43 CBOR_TAG_BIGFLOAT = 5, // array
44
45 CBOR_TAG_CONV_B64U = 21, // base64url, any
46 CBOR_TAG_CONV_B64 = 22, // base64, any
47 CBOR_TAG_CONV_HEX = 23, // base16, any
48 CBOR_TAG_CBOR = 24, // embedded cbor, byte string
49
50 CBOR_TAG_URI = 32, // URI rfc3986, utf-8
51 CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
52 CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
53 CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
54 CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
55
56 CBOR_TAG_MAGIC = 55799 // self-describe cbor
57};
58
18#define F_SHRINK 0x00000200UL 59#define F_SHRINK 0x00000001UL
19#define F_ALLOW_UNKNOWN 0x00002000UL 60#define F_ALLOW_UNKNOWN 0x00000002UL
61#define F_ALLOW_SHARING 0x00000004UL //TODO
62#define F_DEDUP_STRINGS 0x00000008UL //TODO
63#define F_DEDUP_KEYS 0x00000010UL //TODO
20 64
21#define INIT_SIZE 32 // initial scalar size to be allocated 65#define INIT_SIZE 32 // initial scalar size to be allocated
22 66
23#define SB do { 67#define SB do {
24#define SE } while (0) 68#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 69
37#define IN_RANGE_INC(type,val,beg,end) \ 70#define IN_RANGE_INC(type,val,beg,end) \
38 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 71 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
39 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 72 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
40 73
46#else 79#else
47# define CBOR_SLOW 0 80# define CBOR_SLOW 0
48# define CBOR_STASH cbor_stash 81# define CBOR_STASH cbor_stash
49#endif 82#endif
50 83
51static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: 84static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
52static SV *cbor_true, *cbor_false; 85static SV *types_true, *types_false, *types_error, *sv_cbor;
53 86
54typedef struct { 87typedef struct {
55 U32 flags; 88 U32 flags;
56 U32 max_depth; 89 U32 max_depth;
57 STRLEN max_size; 90 STRLEN max_size;
58
59 SV *cb_object;
60 HV *cb_sk_object;
61} CBOR; 91} CBOR;
62 92
63INLINE void 93ecb_inline void
64cbor_init (CBOR *cbor) 94cbor_init (CBOR *cbor)
65{ 95{
66 Zero (cbor, 1, CBOR); 96 Zero (cbor, 1, CBOR);
67 cbor->max_depth = 512; 97 cbor->max_depth = 512;
68} 98}
69 99
70///////////////////////////////////////////////////////////////////////////// 100/////////////////////////////////////////////////////////////////////////////
71// utility functions 101// utility functions
72 102
73INLINE SV * 103ecb_inline SV *
74get_bool (const char *name) 104get_bool (const char *name)
75{ 105{
76 SV *sv = get_sv (name, 1); 106 SV *sv = get_sv (name, 1);
77 107
78 SvREADONLY_on (sv); 108 SvREADONLY_on (sv);
79 SvREADONLY_on (SvRV (sv)); 109 SvREADONLY_on (SvRV (sv));
80 110
81 return sv; 111 return sv;
82} 112}
83 113
84INLINE void 114ecb_inline void
85shrink (SV *sv) 115shrink (SV *sv)
86{ 116{
87 sv_utf8_downgrade (sv, 1); 117 sv_utf8_downgrade (sv, 1);
88 118
89 if (SvLEN (sv) > SvCUR (sv) + 1) 119 if (SvLEN (sv) > SvCUR (sv) + 1)
95#endif 125#endif
96 } 126 }
97} 127}
98 128
99///////////////////////////////////////////////////////////////////////////// 129/////////////////////////////////////////////////////////////////////////////
100// fp hell
101
102//TODO
103
104/////////////////////////////////////////////////////////////////////////////
105// encoder 130// encoder
106 131
107// structure used for encoding CBOR 132// structure used for encoding CBOR
108typedef struct 133typedef struct
109{ 134{
110 char *cur; // SvPVX (sv) + current output position 135 char *cur; // SvPVX (sv) + current output position
111 char *end; // SvEND (sv) 136 char *end; // SvEND (sv)
112 SV *sv; // result scalar 137 SV *sv; // result scalar
113 CBOR cbor; 138 CBOR cbor;
114 U32 depth; // recursion level 139 U32 depth; // recursion level
140 HV *stringref; // string => index, or 0
141 HV *shareable; // ptr => index, or 0
142 UV shareable_idx;
115} enc_t; 143} enc_t;
116 144
117INLINE void 145ecb_inline void
118need (enc_t *enc, STRLEN len) 146need (enc_t *enc, STRLEN len)
119{ 147{
120 if (expect_false (enc->cur + len >= enc->end)) 148 if (ecb_expect_false (enc->cur + len >= enc->end))
121 { 149 {
122 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); 150 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
123 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); 151 SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
124 enc->cur = SvPVX (enc->sv) + cur; 152 enc->cur = SvPVX (enc->sv) + cur;
125 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 153 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126 } 154 }
127} 155}
128 156
129INLINE void 157ecb_inline void
130encode_ch (enc_t *enc, char ch) 158encode_ch (enc_t *enc, char ch)
131{ 159{
132 need (enc, 1); 160 need (enc, 1);
133 *enc->cur++ = ch; 161 *enc->cur++ = ch;
134} 162}
138{ 166{
139 need (enc, 9); 167 need (enc, 9);
140 168
141 if (len < 24) 169 if (len < 24)
142 *enc->cur++ = major | len; 170 *enc->cur++ = major | len;
143 else if (len < 0x100) 171 else if (len <= 0xff)
144 { 172 {
145 *enc->cur++ = major | 24; 173 *enc->cur++ = major | 24;
146 *enc->cur++ = len; 174 *enc->cur++ = len;
147 } 175 }
148 else if (len < 0x10000) 176 else if (len <= 0xffff)
149 { 177 {
150 *enc->cur++ = major | 25; 178 *enc->cur++ = major | 25;
151 *enc->cur++ = len >> 8; 179 *enc->cur++ = len >> 8;
152 *enc->cur++ = len; 180 *enc->cur++ = len;
153 } 181 }
154 else if (len < 0x100000000) 182 else if (len <= 0xffffffff)
155 { 183 {
156 *enc->cur++ = major | 26; 184 *enc->cur++ = major | 26;
157 *enc->cur++ = len >> 24; 185 *enc->cur++ = len >> 24;
158 *enc->cur++ = len >> 16; 186 *enc->cur++ = len >> 16;
159 *enc->cur++ = len >> 8; 187 *enc->cur++ = len >> 8;
160 *enc->cur++ = len; 188 *enc->cur++ = len;
161 } 189 }
162 else if (len) 190 else
163 { 191 {
164 *enc->cur++ = major | 27; 192 *enc->cur++ = major | 27;
165 *enc->cur++ = len >> 56; 193 *enc->cur++ = len >> 56;
166 *enc->cur++ = len >> 48; 194 *enc->cur++ = len >> 48;
167 *enc->cur++ = len >> 40; 195 *enc->cur++ = len >> 40;
180 need (enc, len); 208 need (enc, len);
181 memcpy (enc->cur, str, len); 209 memcpy (enc->cur, str, len);
182 enc->cur += len; 210 enc->cur += len;
183} 211}
184 212
213ecb_inline void
214encode_tag (enc_t *enc, UV tag)
215{
216 encode_uint (enc, 0xc0, tag);
217}
218
185static void encode_sv (enc_t *enc, SV *sv); 219static void encode_sv (enc_t *enc, SV *sv);
186 220
187static void 221static void
188encode_av (enc_t *enc, AV *av) 222encode_av (enc_t *enc, AV *av)
189{ 223{
228 if (HeKLEN (he) == HEf_SVKEY) 262 if (HeKLEN (he) == HEf_SVKEY)
229 encode_sv (enc, HeSVKEY (he)); 263 encode_sv (enc, HeSVKEY (he));
230 else 264 else
231 encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he)); 265 encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));
232 266
233 encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he)); 267 encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
234 } 268 }
235 269
236 if (mg) 270 if (mg)
237 encode_ch (enc, 0xe0 | 31); 271 encode_ch (enc, 0xe0 | 31);
238 272
241 275
242// encode objects, arrays and special \0=false and \1=true values. 276// encode objects, arrays and special \0=false and \1=true values.
243static void 277static void
244encode_rv (enc_t *enc, SV *sv) 278encode_rv (enc_t *enc, SV *sv)
245{ 279{
246 svtype svt;
247
248 SvGETMAGIC (sv); 280 SvGETMAGIC (sv);
249 svt = SvTYPE (sv);
250 281
251 if (expect_false (SvOBJECT (sv))) 282 if (ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)
283 && ecb_expect_false (SvREFCNT (sv) > 1))
252 { 284 {
253 HV *stash = !CBOR_SLOW || cbor_boolean_stash 285 if (!enc->shareable)
254 ? cbor_boolean_stash 286 enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
255 : gv_stashpv ("CBOR::XS::Boolean", 1);
256 287
257 if (SvSTASH (sv) == stash) 288 SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
258 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); 289
290 if (SvOK (*svp))
291 {
292 encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
293 encode_uint (enc, 0x00, SvUV (*svp));
294 return;
295 }
259 else 296 else
260 { 297 {
261#if 0 //TODO 298 sv_setuv (*svp, enc->shareable_idx);
262 if (enc->cbor.flags & F_CONV_BLESSED) 299 ++enc->shareable_idx;
263 { 300 encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
264 // 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)));
273
274 // calling with G_SCALAR ensures that we always get a 1 return value
275 PUTBACK;
276 call_sv ((SV *)GvCV (to_cbor), G_SCALAR);
277 SPAGAIN;
278
279 // catch this surprisingly common error
280 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)));
282
283 sv = POPs;
284 PUTBACK;
285
286 encode_sv (enc, sv);
287
288 FREETMPS; LEAVE;
289 }
290 else if (enc->cbor.flags & F_ALLOW_BLESSED)
291 encode_str (enc, "null", 4, 0);
292 else
293 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it",
294 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
295 }
296 else if (enc->cbor.flags & F_ALLOW_BLESSED)
297 encode_str (enc, "null", 4, 0);
298 else
299 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
300 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
301#endif
302 } 301 }
302 }
303
304 svtype svt = SvTYPE (sv);
305
306 if (ecb_expect_false (SvOBJECT (sv)))
307 {
308 HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
309 ? types_boolean_stash
310 : gv_stashpv ("Types::Serialiser::Boolean", 1);
311 HV *error_stash = !CBOR_SLOW || types_error_stash
312 ? types_error_stash
313 : gv_stashpv ("Types::Serialiser::Error", 1);
314 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
315 ? cbor_tagged_stash
316 : gv_stashpv ("CBOR::XS::Tagged" , 1);
317
318 HV *stash = SvSTASH (sv);
319 GV *method;
320
321 if (stash == boolean_stash)
322 encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);
323 else if (stash == error_stash)
324 encode_ch (enc, 0xe0 | 23);
325 else if (stash == tagged_stash)
326 {
327 if (svt != SVt_PVAV)
328 croak ("encountered CBOR::XS::Tagged object that isn't an array");
329
330 encode_uint (enc, 0xc0, SvUV (*av_fetch ((AV *)sv, 0, 1)));
331 encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
332 }
333 else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
334 {
335 dSP;
336
337 ENTER; SAVETMPS; PUSHMARK (SP);
338 // we re-bless the reference to get overload and other niceties right
339 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
340
341 PUTBACK;
342 // G_SCALAR ensures that return value is 1
343 call_sv ((SV *)GvCV (method), G_SCALAR);
344 SPAGAIN;
345
346 // catch this surprisingly common error
347 if (SvROK (TOPs) && SvRV (TOPs) == sv)
348 croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
349
350 encode_sv (enc, POPs);
351
352 PUTBACK;
353
354 FREETMPS; LEAVE;
355 }
356 else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
357 {
358 dSP;
359
360 ENTER; SAVETMPS; PUSHMARK (SP);
361 EXTEND (SP, 2);
362 // we re-bless the reference to get overload and other niceties right
363 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
364 PUSHs (sv_cbor);
365
366 PUTBACK;
367 int count = call_sv ((SV *)GvCV (method), G_ARRAY);
368 SPAGAIN;
369
370 // catch this surprisingly common error
371 if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
372 croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
373
374 encode_tag (enc, CBOR_TAG_PERL_OBJECT);
375 encode_uint (enc, 0x80, count + 1);
376 encode_str (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
377
378 while (count)
379 encode_sv (enc, SP[1 - count--]);
380
381 PUTBACK;
382
383 FREETMPS; LEAVE;
384 }
385 else
386 croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
387 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
303 } 388 }
304 else if (svt == SVt_PVHV) 389 else if (svt == SVt_PVHV)
305 encode_hv (enc, (HV *)sv); 390 encode_hv (enc, (HV *)sv);
306 else if (svt == SVt_PVAV) 391 else if (svt == SVt_PVAV)
307 encode_av (enc, (AV *)sv); 392 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 393 else
326 croak ("encountered %s, but CBOR can only represent references to arrays or hashes", 394 {
327 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 395 encode_tag (enc, CBOR_TAG_INDIRECTION);
396 encode_sv (enc, sv);
397 }
328} 398}
329 399
330static void 400static void
331encode_nv (enc_t *enc, SV *sv) 401encode_nv (enc_t *enc, SV *sv)
332{ 402{
333 double nv = SvNVX (sv); 403 double nv = SvNVX (sv);
334 404
335 need (enc, 9); 405 need (enc, 9);
336 406
337 if (expect_false (nv == (U32)nv)) 407 if (ecb_expect_false (nv == (U32)nv))
338 encode_uint (enc, 0x00, (U32)nv); 408 encode_uint (enc, 0x00, (U32)nv);
339 //TODO: maybe I32? 409 //TODO: maybe I32?
340 else if (expect_false (nv == (float)nv)) 410 else if (ecb_expect_false (nv == (float)nv))
341 { 411 {
342 uint32_t fp = ecb_float_to_binary32 (nv); 412 uint32_t fp = ecb_float_to_binary32 (nv);
343 413
344 *enc->cur++ = 0xe0 | 26; 414 *enc->cur++ = 0xe0 | 26;
345 415
397} 467}
398 468
399static SV * 469static SV *
400encode_cbor (SV *scalar, CBOR *cbor) 470encode_cbor (SV *scalar, CBOR *cbor)
401{ 471{
402 enc_t enc; 472 enc_t enc = { };
403 473
404 enc.cbor = *cbor; 474 enc.cbor = *cbor;
405 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 475 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
406 enc.cur = SvPVX (enc.sv); 476 enc.cur = SvPVX (enc.sv);
407 enc.end = SvEND (enc.sv); 477 enc.end = SvEND (enc.sv);
408 enc.depth = 0;
409 478
410 SvPOK_only (enc.sv); 479 SvPOK_only (enc.sv);
411 encode_sv (&enc, scalar); 480 encode_sv (&enc, scalar);
412 481
413 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 482 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
429 U8 *end; // end of input string 498 U8 *end; // end of input string
430 const char *err; // parse error, if != 0 499 const char *err; // parse error, if != 0
431 CBOR cbor; 500 CBOR cbor;
432 U32 depth; // recursion depth 501 U32 depth; // recursion depth
433 U32 maxdepth; // recursion depth limit 502 U32 maxdepth; // recursion depth limit
503 AV *shareable;
434} dec_t; 504} dec_t;
435 505
436#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE 506#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
437 507
438#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data"); 508#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
439 509
440#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED) 510#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)
441#define DEC_DEC_DEPTH --dec->depth 511#define DEC_DEC_DEPTH --dec->depth
442 512
443static UV 513static UV
532 SvREFCNT_dec (av); 602 SvREFCNT_dec (av);
533 DEC_DEC_DEPTH; 603 DEC_DEC_DEPTH;
534 return &PL_sv_undef; 604 return &PL_sv_undef;
535} 605}
536 606
607static void
608decode_he (dec_t *dec, HV *hv)
609{
610 // for speed reasons, we specialcase single-string
611 // byte or utf-8 strings as keys.
612
613 if (*dec->cur >= 0x40 && *dec->cur <= 0x40 + 27)
614 {
615 I32 len = decode_uint (dec);
616 char *key = (char *)dec->cur;
617
618 dec->cur += len;
619
620 hv_store (hv, key, len, decode_sv (dec), 0);
621 }
622 else if (*dec->cur >= 0x60 && *dec->cur <= 0x60 + 27)
623 {
624 I32 len = decode_uint (dec);
625 char *key = (char *)dec->cur;
626
627 dec->cur += len;
628
629 hv_store (hv, key, -len, decode_sv (dec), 0);
630 }
631 else
632 {
633 SV *k = decode_sv (dec);
634 SV *v = decode_sv (dec);
635
636 hv_store_ent (hv, k, v, 0);
637 SvREFCNT_dec (k);
638 }
639}
640
537static SV * 641static SV *
538decode_hv (dec_t *dec) 642decode_hv (dec_t *dec)
539{ 643{
540 HV *hv = newHV (); 644 HV *hv = newHV ();
541 645
553 { 657 {
554 ++dec->cur; 658 ++dec->cur;
555 break; 659 break;
556 } 660 }
557 661
558 SV *k = decode_sv (dec); 662 decode_he (dec, hv);
559 SV *v = decode_sv (dec);
560
561 hv_store_ent (hv, k, v, 0);
562 } 663 }
563 } 664 }
564 else 665 else
565 { 666 {
566 int len = decode_uint (dec); 667 int pairs = decode_uint (dec);
567 668
568 while (len--) 669 while (pairs--)
569 { 670 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 } 671 }
576 672
577 DEC_DEC_DEPTH; 673 DEC_DEC_DEPTH;
578 return newRV_noinc ((SV *)hv); 674 return newRV_noinc ((SV *)hv);
579
580#if 0
581 SV *sv;
582 HV *hv = newHV ();
583
584 DEC_INC_DEPTH;
585 decode_ws (dec);
586
587 for (;;)
588 {
589 // heuristic: assume that
590 // a) decode_str + hv_store_ent are abysmally slow.
591 // b) most hash keys are short, simple ascii text.
592 // => try to "fast-match" such strings to avoid
593 // the overhead of decode_str + hv_store_ent.
594 {
595 SV *value;
596 char *p = dec->cur;
597 char *e = p + 24; // only try up to 24 bytes
598
599 for (;;)
600 {
601 // the >= 0x80 is false on most architectures
602 if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
603 {
604 // slow path, back up and use decode_str
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;
623 }
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 {
689 dSP;
690 int count;
691
692 ENTER; SAVETMPS; PUSHMARK (SP);
693 XPUSHs (HeVAL (he));
694 sv_2mortal (sv);
695
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
734 675
735fail: 676fail:
736 SvREFCNT_dec (hv); 677 SvREFCNT_dec (hv);
737 DEC_DEC_DEPTH; 678 DEC_DEC_DEPTH;
738 return &PL_sv_undef; 679 return &PL_sv_undef;
739} 680}
740 681
741static SV * 682static SV *
742decode_str (dec_t *dec, int utf8) 683decode_str (dec_t *dec, int utf8)
743{ 684{
744 SV *sv; 685 SV *sv = 0;
745 686
746 if ((*dec->cur & 31) == 31) 687 if ((*dec->cur & 31) == 31)
747 { 688 {
748 ++dec->cur; 689 ++dec->cur;
749 690
758 { 699 {
759 ++dec->cur; 700 ++dec->cur;
760 break; 701 break;
761 } 702 }
762 703
763 SV *sv2 = decode_sv (dec);
764 sv_catsv (sv, sv2); 704 sv_catsv (sv, decode_sv (dec));
765 } 705 }
766 } 706 }
767 else 707 else
768 { 708 {
769 STRLEN len = decode_uint (dec); 709 STRLEN len = decode_uint (dec);
777 SvUTF8_on (sv); 717 SvUTF8_on (sv);
778 718
779 return sv; 719 return sv;
780 720
781fail: 721fail:
722 SvREFCNT_dec (sv);
723 return &PL_sv_undef;
724}
725
726static SV *
727decode_tagged (dec_t *dec)
728{
729 SV *sv = 0;
730 UV tag = decode_uint (dec);
731
732 WANT (1);
733
734 switch (tag)
735 {
736 case CBOR_TAG_MAGIC:
737 return decode_sv (dec);
738
739 case CBOR_TAG_INDIRECTION:
740 return newRV_noinc (decode_sv (dec));
741
742 case CBOR_TAG_VALUE_SHAREABLE:
743 {
744 if (ecb_expect_false (!dec->shareable))
745 dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
746
747 sv = newSV (0);
748 av_push (dec->shareable, SvREFCNT_inc_NN (sv));
749
750 SV *osv = decode_sv (dec);
751 sv_setsv (sv, osv);
752 SvREFCNT_dec_NN (osv);
753 }
754
755 return sv;
756
757 case CBOR_TAG_VALUE_SHAREDREF:
758 {
759 if ((*dec->cur >> 5) != 0)
760 ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
761
762 UV idx = decode_uint (dec);
763
764 if (!dec->shareable || idx > AvFILLp (dec->shareable))
765 ERR ("corrupted CBOR data (sharedref index out of bounds)");
766
767 return SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
768 }
769
770 case CBOR_TAG_PERL_OBJECT:
771 {
772 sv = decode_sv (dec);
773
774 if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
775 ERR ("corrupted CBOR data (non-array perl object)");
776
777 AV *av = (AV *)SvRV (sv);
778 int len = av_len (av) + 1;
779 HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
780
781 if (!stash)
782 ERR ("cannot decode perl-object (package does not exist)");
783
784 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
785
786 if (!method)
787 ERR ("cannot decode perl-object (package does not have a THAW method)");
788
789 dSP;
790
791 ENTER; SAVETMPS; PUSHMARK (SP);
792 EXTEND (SP, len + 1);
793 // we re-bless the reference to get overload and other niceties right
794 PUSHs (*av_fetch (av, 0, 1));
795 PUSHs (sv_cbor);
796
797 int i;
798
799 for (i = 1; i < len; ++i)
800 PUSHs (*av_fetch (av, i, 1));
801
802 PUTBACK;
803 call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
804 SPAGAIN;
805
806 if (SvTRUE (ERRSV))
807 {
808 FREETMPS; LEAVE;
809 ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
810 }
811
812 SvREFCNT_dec (sv);
813 sv = SvREFCNT_inc (POPs);
814
815 PUTBACK;
816
817 FREETMPS; LEAVE;
818
819 return sv;
820 }
821
822 default:
823 {
824 sv = decode_sv (dec);
825
826 AV *av = newAV ();
827 av_push (av, newSVuv (tag));
828 av_push (av, sv);
829
830 HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
831 ? cbor_tagged_stash
832 : gv_stashpv ("CBOR::XS::Tagged" , 1);
833
834 return sv_bless (newRV_noinc ((SV *)av), tagged_stash);
835 }
836 }
837
838fail:
839 SvREFCNT_dec (sv);
782 return &PL_sv_undef; 840 return &PL_sv_undef;
783} 841}
784 842
785static SV * 843static SV *
786decode_sv (dec_t *dec) 844decode_sv (dec_t *dec)
788 WANT (1); 846 WANT (1);
789 847
790 switch (*dec->cur >> 5) 848 switch (*dec->cur >> 5)
791 { 849 {
792 case 0: // unsigned int 850 case 0: // unsigned int
793 //TODO: 64 bit values on 3 2bit perls
794 return newSVuv (decode_uint (dec)); 851 return newSVuv (decode_uint (dec));
795 case 1: // negative int 852 case 1: // negative int
796 return newSViv (-1 - (IV)decode_uint (dec)); 853 return newSViv (-1 - (IV)decode_uint (dec));
797 case 2: // octet string 854 case 2: // octet string
798 return decode_str (dec, 0); 855 return decode_str (dec, 0);
801 case 4: // array 858 case 4: // array
802 return decode_av (dec); 859 return decode_av (dec);
803 case 5: // map 860 case 5: // map
804 return decode_hv (dec); 861 return decode_hv (dec);
805 case 6: // tag 862 case 6: // tag
806 abort (); 863 return decode_tagged (dec);
807 break;
808 case 7: // misc 864 case 7: // misc
809 switch (*dec->cur++ & 31) 865 switch (*dec->cur++ & 31)
810 { 866 {
811 case 20: 867 case 20:
812#if CBOR_SLOW 868#if CBOR_SLOW
813 cbor_false = get_bool ("CBOR::XS::false"); 869 types_false = get_bool ("Types::Serialiser::false");
814#endif 870#endif
815 return newSVsv (cbor_false); 871 return newSVsv (types_false);
816 case 21: 872 case 21:
817#if CBOR_SLOW 873#if CBOR_SLOW
818 cbor_true = get_bool ("CBOR::XS::true"); 874 types_true = get_bool ("Types::Serialiser::true");
819#endif 875#endif
820 return newSVsv (cbor_true); 876 return newSVsv (types_true);
821 case 22: 877 case 22:
822 return newSVsv (&PL_sv_undef); 878 return newSVsv (&PL_sv_undef);
879 case 23:
880#if CBOR_SLOW
881 types_error = get_bool ("Types::Serialiser::error");
882#endif
883 return newSVsv (types_error);
823 884
824 case 25: 885 case 25:
825 { 886 {
826 WANT (2); 887 WANT (2);
827 888
863 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); 924 ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
864 } 925 }
865 926
866 break; 927 break;
867 } 928 }
868#if 0
869 switch (*dec->cur)
870 {
871 //case '"': ++dec->cur; return decode_str (dec);
872 case '[': ++dec->cur; return decode_av (dec);
873 case '{': ++dec->cur; return decode_hv (dec);
874
875 case '-':
876 case '0': case '1': case '2': case '3': case '4':
877 case '5': case '6': case '7': case '8': case '9':
878 //TODO return decode_num (dec);
879
880 case 't':
881 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
882 {
883 dec->cur += 4;
884#if CBOR_SLOW
885 cbor_true = get_bool ("CBOR::XS::true");
886#endif
887 return newSVsv (cbor_true);
888 }
889 else
890 ERR ("'true' expected");
891
892 break;
893
894 case 'f':
895 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
896 {
897 dec->cur += 5;
898#if CBOR_SLOW
899 cbor_false = get_bool ("CBOR::XS::false");
900#endif
901 return newSVsv (cbor_false);
902 }
903 else
904 ERR ("'false' expected");
905
906 break;
907
908 case 'n':
909 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
910 {
911 dec->cur += 4;
912 return newSVsv (&PL_sv_undef);
913 }
914 else
915 ERR ("'null' expected");
916
917 break;
918
919 default:
920 ERR ("malformed CBOR string, neither array, object, number, string or atom");
921 break;
922 }
923#endif
924 929
925fail: 930fail:
926 return &PL_sv_undef; 931 return &PL_sv_undef;
927} 932}
928 933
929static SV * 934static SV *
930decode_cbor (SV *string, CBOR *cbor, char **offset_return) 935decode_cbor (SV *string, CBOR *cbor, char **offset_return)
931{ 936{
932 dec_t dec; 937 dec_t dec = { };
933 SV *sv; 938 SV *sv;
939 STRLEN len;
940 char *data = SvPVbyte (string, len);
934 941
935 /* work around bugs in 5.10 where manipulating magic values
936 * makes perl ignore the magic in subsequent accesses.
937 * also make a copy of non-PV values, to get them into a clean
938 * state (SvPV should do that, but it's buggy, see below).
939 */
940 /*SvGETMAGIC (string);*/
941 if (SvMAGICAL (string) || !SvPOK (string))
942 string = sv_2mortal (newSVsv (string));
943
944 SvUPGRADE (string, SVt_PV);
945
946 /* work around a bug in perl 5.10, which causes SvCUR to fail an
947 * assertion with -DDEBUGGING, although SvCUR is documented to
948 * return the xpv_cur field which certainly exists after upgrading.
949 * according to nicholas clark, calling SvPOK fixes this.
950 * But it doesn't fix it, so try another workaround, call SvPV_nolen
951 * and hope for the best.
952 * Damnit, SvPV_nolen still trips over yet another assertion. This
953 * assertion business is seriously broken, try yet another workaround
954 * for the broken -DDEBUGGING.
955 */
956 {
957#ifdef DEBUGGING
958 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
959#else
960 STRLEN offset = SvCUR (string);
961#endif
962
963 if (offset > cbor->max_size && cbor->max_size) 942 if (len > cbor->max_size && cbor->max_size)
964 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu", 943 croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
965 (unsigned long)SvCUR (string), (unsigned long)cbor->max_size); 944 (unsigned long)len, (unsigned long)cbor->max_size);
966 }
967
968 sv_utf8_downgrade (string, 0);
969 945
970 dec.cbor = *cbor; 946 dec.cbor = *cbor;
971 dec.cur = (U8 *)SvPVX (string); 947 dec.cur = (U8 *)data;
972 dec.end = (U8 *)SvEND (string); 948 dec.end = (U8 *)data + len;
973 dec.err = 0;
974 dec.depth = 0;
975
976 if (dec.cbor.cb_object || dec.cbor.cb_sk_object)
977 ;//TODO dec.cbor.flags |= F_HOOK;
978 949
979 sv = decode_sv (&dec); 950 sv = decode_sv (&dec);
980 951
981 if (offset_return) 952 if (offset_return)
982 *offset_return = dec.cur; 953 *offset_return = dec.cur;
986 dec.err = "garbage after CBOR object"; 957 dec.err = "garbage after CBOR object";
987 958
988 if (dec.err) 959 if (dec.err)
989 { 960 {
990 SvREFCNT_dec (sv); 961 SvREFCNT_dec (sv);
991 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); 962 croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
992 } 963 }
993 964
994 sv = sv_2mortal (sv); 965 sv = sv_2mortal (sv);
995 966
996 return sv; 967 return sv;
1002MODULE = CBOR::XS PACKAGE = CBOR::XS 973MODULE = CBOR::XS PACKAGE = CBOR::XS
1003 974
1004BOOT: 975BOOT:
1005{ 976{
1006 cbor_stash = gv_stashpv ("CBOR::XS" , 1); 977 cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1007 cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); 978 cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1008 979
1009 cbor_true = get_bool ("CBOR::XS::true"); 980 types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1010 cbor_false = get_bool ("CBOR::XS::false"); 981 types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
982
983 types_true = get_bool ("Types::Serialiser::true" );
984 types_false = get_bool ("Types::Serialiser::false");
985 types_error = get_bool ("Types::Serialiser::error");
986
987 sv_cbor = newSVpv ("CBOR", 0);
988 SvREADONLY_on (sv_cbor);
1011} 989}
1012 990
1013PROTOTYPES: DISABLE 991PROTOTYPES: DISABLE
1014 992
1015void CLONE (...) 993void CLONE (...)
1016 CODE: 994 CODE:
1017 cbor_stash = 0; 995 cbor_stash = 0;
996 cbor_tagged_stash = 0;
997 types_error_stash = 0;
1018 cbor_boolean_stash = 0; 998 types_boolean_stash = 0;
1019 999
1020void new (char *klass) 1000void new (char *klass)
1021 PPCODE: 1001 PPCODE:
1022{ 1002{
1023 SV *pv = NEWSV (0, sizeof (CBOR)); 1003 SV *pv = NEWSV (0, sizeof (CBOR));
1031 1011
1032void shrink (CBOR *self, int enable = 1) 1012void shrink (CBOR *self, int enable = 1)
1033 ALIAS: 1013 ALIAS:
1034 shrink = F_SHRINK 1014 shrink = F_SHRINK
1035 allow_unknown = F_ALLOW_UNKNOWN 1015 allow_unknown = F_ALLOW_UNKNOWN
1016 allow_sharing = F_ALLOW_SHARING
1017 dedup_keys = F_DEDUP_KEYS
1018 dedup_strings = F_DEDUP_STRINGS
1036 PPCODE: 1019 PPCODE:
1037{ 1020{
1038 if (enable) 1021 if (enable)
1039 self->flags |= ix; 1022 self->flags |= ix;
1040 else 1023 else
1045 1028
1046void get_shrink (CBOR *self) 1029void get_shrink (CBOR *self)
1047 ALIAS: 1030 ALIAS:
1048 get_shrink = F_SHRINK 1031 get_shrink = F_SHRINK
1049 get_allow_unknown = F_ALLOW_UNKNOWN 1032 get_allow_unknown = F_ALLOW_UNKNOWN
1033 get_allow_sharing = F_ALLOW_SHARING
1034 get_dedup_keys = F_DEDUP_KEYS
1035 get_dedup_strings = F_DEDUP_STRINGS
1050 PPCODE: 1036 PPCODE:
1051 XPUSHs (boolSV (self->flags & ix)); 1037 XPUSHs (boolSV (self->flags & ix));
1052 1038
1053void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) 1039void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1054 PPCODE: 1040 PPCODE:
1069int get_max_size (CBOR *self) 1055int get_max_size (CBOR *self)
1070 CODE: 1056 CODE:
1071 RETVAL = self->max_size; 1057 RETVAL = self->max_size;
1072 OUTPUT: 1058 OUTPUT:
1073 RETVAL 1059 RETVAL
1074
1075#if 0 //TODO
1076
1077void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1078 PPCODE:
1079{
1080 SvREFCNT_dec (self->cb_object);
1081 self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1082
1083 XPUSHs (ST (0));
1084}
1085
1086void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)
1087 PPCODE:
1088{
1089 if (!self->cb_sk_object)
1090 self->cb_sk_object = newHV ();
1091
1092 if (SvOK (cb))
1093 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1094 else
1095 {
1096 hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1097
1098 if (!HvKEYS (self->cb_sk_object))
1099 {
1100 SvREFCNT_dec (self->cb_sk_object);
1101 self->cb_sk_object = 0;
1102 }
1103 }
1104
1105 XPUSHs (ST (0));
1106}
1107
1108#endif
1109 1060
1110void encode (CBOR *self, SV *scalar) 1061void encode (CBOR *self, SV *scalar)
1111 PPCODE: 1062 PPCODE:
1112 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; 1063 PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1113 XPUSHs (scalar); 1064 XPUSHs (scalar);
1126 EXTEND (SP, 2); 1077 EXTEND (SP, 2);
1127 PUSHs (sv); 1078 PUSHs (sv);
1128 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); 1079 PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1129} 1080}
1130 1081
1131void DESTROY (CBOR *self)
1132 CODE:
1133 SvREFCNT_dec (self->cb_sk_object);
1134 SvREFCNT_dec (self->cb_object);
1135
1136PROTOTYPES: ENABLE 1082PROTOTYPES: ENABLE
1137 1083
1138void encode_cbor (SV *scalar) 1084void encode_cbor (SV *scalar)
1139 PPCODE: 1085 PPCODE:
1140{ 1086{

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines