1 |
#include "EXTERN.h" |
2 |
#include "perl.h" |
3 |
#include "XSUB.h" |
4 |
|
5 |
#include <assert.h> |
6 |
#include <string.h> |
7 |
#include <stdlib.h> |
8 |
#include <stdio.h> |
9 |
#include <limits.h> |
10 |
#include <float.h> |
11 |
|
12 |
#include "ecb.h" |
13 |
|
14 |
#define F_SHRINK 0x00000200UL |
15 |
#define F_ALLOW_UNKNOWN 0x00002000UL |
16 |
|
17 |
#define INIT_SIZE 32 // initial scalar size to be allocated |
18 |
|
19 |
#define SB do { |
20 |
#define SE } while (0) |
21 |
|
22 |
#if __GNUC__ >= 3 |
23 |
# define expect(expr,value) __builtin_expect ((expr), (value)) |
24 |
# define INLINE static inline |
25 |
#else |
26 |
# define expect(expr,value) (expr) |
27 |
# define INLINE static |
28 |
#endif |
29 |
|
30 |
#define expect_false(expr) expect ((expr) != 0, 0) |
31 |
#define expect_true(expr) expect ((expr) != 0, 1) |
32 |
|
33 |
#define IN_RANGE_INC(type,val,beg,end) \ |
34 |
((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ |
35 |
<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) |
36 |
|
37 |
#define ERR_NESTING_EXCEEDED "cbor text or perl structure exceeds maximum nesting level (max_depth set too low?)" |
38 |
|
39 |
#ifdef USE_ITHREADS |
40 |
# define CBOR_SLOW 1 |
41 |
# define CBOR_STASH (cbor_stash ? cbor_stash : gv_stashpv ("CBOR::XS", 1)) |
42 |
#else |
43 |
# define CBOR_SLOW 0 |
44 |
# define CBOR_STASH cbor_stash |
45 |
#endif |
46 |
|
47 |
static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS:: |
48 |
static SV *cbor_true, *cbor_false; |
49 |
|
50 |
typedef struct { |
51 |
U32 flags; |
52 |
U32 max_depth; |
53 |
STRLEN max_size; |
54 |
|
55 |
SV *cb_object; |
56 |
HV *cb_sk_object; |
57 |
} CBOR; |
58 |
|
59 |
INLINE void |
60 |
cbor_init (CBOR *cbor) |
61 |
{ |
62 |
Zero (cbor, 1, CBOR); |
63 |
cbor->max_depth = 512; |
64 |
} |
65 |
|
66 |
///////////////////////////////////////////////////////////////////////////// |
67 |
// utility functions |
68 |
|
69 |
INLINE SV * |
70 |
get_bool (const char *name) |
71 |
{ |
72 |
SV *sv = get_sv (name, 1); |
73 |
|
74 |
SvREADONLY_on (sv); |
75 |
SvREADONLY_on (SvRV (sv)); |
76 |
|
77 |
return sv; |
78 |
} |
79 |
|
80 |
INLINE void |
81 |
shrink (SV *sv) |
82 |
{ |
83 |
sv_utf8_downgrade (sv, 1); |
84 |
|
85 |
if (SvLEN (sv) > SvCUR (sv) + 1) |
86 |
{ |
87 |
#ifdef SvPV_shrink_to_cur |
88 |
SvPV_shrink_to_cur (sv); |
89 |
#elif defined (SvPV_renew) |
90 |
SvPV_renew (sv, SvCUR (sv) + 1); |
91 |
#endif |
92 |
} |
93 |
} |
94 |
|
95 |
///////////////////////////////////////////////////////////////////////////// |
96 |
// fp hell |
97 |
|
98 |
//TODO |
99 |
|
100 |
///////////////////////////////////////////////////////////////////////////// |
101 |
// encoder |
102 |
|
103 |
// structure used for encoding CBOR |
104 |
typedef struct |
105 |
{ |
106 |
char *cur; // SvPVX (sv) + current output position |
107 |
char *end; // SvEND (sv) |
108 |
SV *sv; // result scalar |
109 |
CBOR cbor; |
110 |
U32 depth; // recursion level |
111 |
} enc_t; |
112 |
|
113 |
INLINE void |
114 |
need (enc_t *enc, STRLEN len) |
115 |
{ |
116 |
if (expect_false (enc->cur + len >= enc->end)) |
117 |
{ |
118 |
STRLEN cur = enc->cur - (char *)SvPVX (enc->sv); |
119 |
SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1); |
120 |
enc->cur = SvPVX (enc->sv) + cur; |
121 |
enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
122 |
} |
123 |
} |
124 |
|
125 |
INLINE void |
126 |
encode_ch (enc_t *enc, char ch) |
127 |
{ |
128 |
need (enc, 1); |
129 |
*enc->cur++ = ch; |
130 |
} |
131 |
|
132 |
static void |
133 |
encode_uint (enc_t *enc, int major, UV len) |
134 |
{ |
135 |
need (enc, 9); |
136 |
|
137 |
if (len < 24) |
138 |
*enc->cur++ = major | len; |
139 |
else if (len <= 0xff) |
140 |
{ |
141 |
*enc->cur++ = major | 24; |
142 |
*enc->cur++ = len; |
143 |
} |
144 |
else if (len <= 0xffff) |
145 |
{ |
146 |
*enc->cur++ = major | 25; |
147 |
*enc->cur++ = len >> 8; |
148 |
*enc->cur++ = len; |
149 |
} |
150 |
else if (len <= 0xffffffff) |
151 |
{ |
152 |
*enc->cur++ = major | 26; |
153 |
*enc->cur++ = len >> 24; |
154 |
*enc->cur++ = len >> 16; |
155 |
*enc->cur++ = len >> 8; |
156 |
*enc->cur++ = len; |
157 |
} |
158 |
else |
159 |
{ |
160 |
*enc->cur++ = major | 27; |
161 |
*enc->cur++ = len >> 56; |
162 |
*enc->cur++ = len >> 48; |
163 |
*enc->cur++ = len >> 40; |
164 |
*enc->cur++ = len >> 32; |
165 |
*enc->cur++ = len >> 24; |
166 |
*enc->cur++ = len >> 16; |
167 |
*enc->cur++ = len >> 8; |
168 |
*enc->cur++ = len; |
169 |
} |
170 |
} |
171 |
|
172 |
static void |
173 |
encode_str (enc_t *enc, int utf8, char *str, STRLEN len) |
174 |
{ |
175 |
encode_uint (enc, utf8 ? 0x60 : 0x40, len); |
176 |
need (enc, len); |
177 |
memcpy (enc->cur, str, len); |
178 |
enc->cur += len; |
179 |
} |
180 |
|
181 |
static void encode_sv (enc_t *enc, SV *sv); |
182 |
|
183 |
static void |
184 |
encode_av (enc_t *enc, AV *av) |
185 |
{ |
186 |
int i, len = av_len (av); |
187 |
|
188 |
if (enc->depth >= enc->cbor.max_depth) |
189 |
croak (ERR_NESTING_EXCEEDED); |
190 |
|
191 |
++enc->depth; |
192 |
|
193 |
encode_uint (enc, 0x80, len + 1); |
194 |
|
195 |
for (i = 0; i <= len; ++i) |
196 |
{ |
197 |
SV **svp = av_fetch (av, i, 0); |
198 |
encode_sv (enc, svp ? *svp : &PL_sv_undef); |
199 |
} |
200 |
|
201 |
--enc->depth; |
202 |
} |
203 |
|
204 |
static void |
205 |
encode_hv (enc_t *enc, HV *hv) |
206 |
{ |
207 |
HE *he; |
208 |
|
209 |
if (enc->depth >= enc->cbor.max_depth) |
210 |
croak (ERR_NESTING_EXCEEDED); |
211 |
|
212 |
++enc->depth; |
213 |
|
214 |
int pairs = hv_iterinit (hv); |
215 |
int mg = SvMAGICAL (hv); |
216 |
|
217 |
if (mg) |
218 |
encode_ch (enc, 0xa0 | 31); |
219 |
else |
220 |
encode_uint (enc, 0xa0, pairs); |
221 |
|
222 |
while ((he = hv_iternext (hv))) |
223 |
{ |
224 |
if (HeKLEN (he) == HEf_SVKEY) |
225 |
encode_sv (enc, HeSVKEY (he)); |
226 |
else |
227 |
encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he)); |
228 |
|
229 |
encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he)); |
230 |
} |
231 |
|
232 |
if (mg) |
233 |
encode_ch (enc, 0xe0 | 31); |
234 |
|
235 |
--enc->depth; |
236 |
} |
237 |
|
238 |
// encode objects, arrays and special \0=false and \1=true values. |
239 |
static void |
240 |
encode_rv (enc_t *enc, SV *sv) |
241 |
{ |
242 |
svtype svt; |
243 |
|
244 |
SvGETMAGIC (sv); |
245 |
svt = SvTYPE (sv); |
246 |
|
247 |
if (expect_false (SvOBJECT (sv))) |
248 |
{ |
249 |
HV *stash = !CBOR_SLOW || cbor_boolean_stash |
250 |
? cbor_boolean_stash |
251 |
: gv_stashpv ("CBOR::XS::Boolean", 1); |
252 |
|
253 |
if (SvSTASH (sv) == stash) |
254 |
encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20); |
255 |
else |
256 |
{ |
257 |
#if 0 //TODO |
258 |
if (enc->cbor.flags & F_CONV_BLESSED) |
259 |
{ |
260 |
// we re-bless the reference to get overload and other niceties right |
261 |
GV *to_cbor = gv_fetchmethod_autoload (SvSTASH (sv), "TO_CBOR", 0); |
262 |
|
263 |
if (to_cbor) |
264 |
{ |
265 |
dSP; |
266 |
|
267 |
ENTER; SAVETMPS; PUSHMARK (SP); |
268 |
XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
269 |
|
270 |
// calling with G_SCALAR ensures that we always get a 1 return value |
271 |
PUTBACK; |
272 |
call_sv ((SV *)GvCV (to_cbor), G_SCALAR); |
273 |
SPAGAIN; |
274 |
|
275 |
// catch this surprisingly common error |
276 |
if (SvROK (TOPs) && SvRV (TOPs) == sv) |
277 |
croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
278 |
|
279 |
sv = POPs; |
280 |
PUTBACK; |
281 |
|
282 |
encode_sv (enc, sv); |
283 |
|
284 |
FREETMPS; LEAVE; |
285 |
} |
286 |
else if (enc->cbor.flags & F_ALLOW_BLESSED) |
287 |
encode_str (enc, "null", 4, 0); |
288 |
else |
289 |
croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it", |
290 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
291 |
} |
292 |
else if (enc->cbor.flags & F_ALLOW_BLESSED) |
293 |
encode_str (enc, "null", 4, 0); |
294 |
else |
295 |
croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
296 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
297 |
#endif |
298 |
} |
299 |
} |
300 |
else if (svt == SVt_PVHV) |
301 |
encode_hv (enc, (HV *)sv); |
302 |
else if (svt == SVt_PVAV) |
303 |
encode_av (enc, (AV *)sv); |
304 |
else if (svt < SVt_PVAV) |
305 |
{ |
306 |
STRLEN len = 0; |
307 |
char *pv = svt ? SvPV (sv, len) : 0; |
308 |
|
309 |
if (len == 1 && *pv == '1') |
310 |
encode_ch (enc, 0xe0 | 21); |
311 |
else if (len == 1 && *pv == '0') |
312 |
encode_ch (enc, 0xe0 | 20); |
313 |
else if (enc->cbor.flags & F_ALLOW_UNKNOWN) |
314 |
encode_ch (enc, 0xe0 | 23); |
315 |
else |
316 |
croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
317 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
318 |
} |
319 |
else if (enc->cbor.flags & F_ALLOW_UNKNOWN) |
320 |
encode_ch (enc, 0xe0 | 23); |
321 |
else |
322 |
croak ("encountered %s, but CBOR can only represent references to arrays or hashes", |
323 |
SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
324 |
} |
325 |
|
326 |
static void |
327 |
encode_nv (enc_t *enc, SV *sv) |
328 |
{ |
329 |
double nv = SvNVX (sv); |
330 |
|
331 |
need (enc, 9); |
332 |
|
333 |
if (expect_false (nv == (U32)nv)) |
334 |
encode_uint (enc, 0x00, (U32)nv); |
335 |
//TODO: maybe I32? |
336 |
else if (expect_false (nv == (float)nv)) |
337 |
{ |
338 |
uint32_t fp = ecb_float_to_binary32 (nv); |
339 |
|
340 |
*enc->cur++ = 0xe0 | 26; |
341 |
|
342 |
if (!ecb_big_endian ()) |
343 |
fp = ecb_bswap32 (fp); |
344 |
|
345 |
memcpy (enc->cur, &fp, 4); |
346 |
enc->cur += 4; |
347 |
} |
348 |
else |
349 |
{ |
350 |
uint64_t fp = ecb_double_to_binary64 (nv); |
351 |
|
352 |
*enc->cur++ = 0xe0 | 27; |
353 |
|
354 |
if (!ecb_big_endian ()) |
355 |
fp = ecb_bswap64 (fp); |
356 |
|
357 |
memcpy (enc->cur, &fp, 8); |
358 |
enc->cur += 8; |
359 |
} |
360 |
} |
361 |
|
362 |
static void |
363 |
encode_sv (enc_t *enc, SV *sv) |
364 |
{ |
365 |
SvGETMAGIC (sv); |
366 |
|
367 |
if (SvPOKp (sv)) |
368 |
{ |
369 |
STRLEN len; |
370 |
char *str = SvPV (sv, len); |
371 |
encode_str (enc, SvUTF8 (sv), str, len); |
372 |
} |
373 |
else if (SvNOKp (sv)) |
374 |
encode_nv (enc, sv); |
375 |
else if (SvIOKp (sv)) |
376 |
{ |
377 |
if (SvIsUV (sv)) |
378 |
encode_uint (enc, 0x00, SvUVX (sv)); |
379 |
else if (SvIVX (sv) >= 0) |
380 |
encode_uint (enc, 0x00, SvIVX (sv)); |
381 |
else |
382 |
encode_uint (enc, 0x20, -(SvIVX (sv) + 1)); |
383 |
} |
384 |
else if (SvROK (sv)) |
385 |
encode_rv (enc, SvRV (sv)); |
386 |
else if (!SvOK (sv)) |
387 |
encode_ch (enc, 0xe0 | 22); |
388 |
else if (enc->cbor.flags & F_ALLOW_UNKNOWN) |
389 |
encode_ch (enc, 0xe0 | 23); |
390 |
else |
391 |
croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data", |
392 |
SvPV_nolen (sv), (unsigned int)SvFLAGS (sv)); |
393 |
} |
394 |
|
395 |
static SV * |
396 |
encode_cbor (SV *scalar, CBOR *cbor) |
397 |
{ |
398 |
enc_t enc; |
399 |
|
400 |
enc.cbor = *cbor; |
401 |
enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
402 |
enc.cur = SvPVX (enc.sv); |
403 |
enc.end = SvEND (enc.sv); |
404 |
enc.depth = 0; |
405 |
|
406 |
SvPOK_only (enc.sv); |
407 |
encode_sv (&enc, scalar); |
408 |
|
409 |
SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
410 |
*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
411 |
|
412 |
if (enc.cbor.flags & F_SHRINK) |
413 |
shrink (enc.sv); |
414 |
|
415 |
return enc.sv; |
416 |
} |
417 |
|
418 |
///////////////////////////////////////////////////////////////////////////// |
419 |
// decoder |
420 |
|
421 |
// structure used for decoding CBOR |
422 |
typedef struct |
423 |
{ |
424 |
U8 *cur; // current parser pointer |
425 |
U8 *end; // end of input string |
426 |
const char *err; // parse error, if != 0 |
427 |
CBOR cbor; |
428 |
U32 depth; // recursion depth |
429 |
U32 maxdepth; // recursion depth limit |
430 |
} dec_t; |
431 |
|
432 |
#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE |
433 |
|
434 |
#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data"); |
435 |
|
436 |
#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED) |
437 |
#define DEC_DEC_DEPTH --dec->depth |
438 |
|
439 |
static UV |
440 |
decode_uint (dec_t *dec) |
441 |
{ |
442 |
switch (*dec->cur & 31) |
443 |
{ |
444 |
case 0: case 1: case 2: case 3: case 4: case 5: case 6: case 7: |
445 |
case 8: case 9: case 10: case 11: case 12: case 13: case 14: case 15: |
446 |
case 16: case 17: case 18: case 19: case 20: case 21: case 22: case 23: |
447 |
return *dec->cur++ & 31; |
448 |
|
449 |
case 24: |
450 |
WANT (2); |
451 |
dec->cur += 2; |
452 |
return dec->cur[-1]; |
453 |
|
454 |
case 25: |
455 |
WANT (3); |
456 |
dec->cur += 3; |
457 |
return (((UV)dec->cur[-2]) << 8) |
458 |
| ((UV)dec->cur[-1]); |
459 |
|
460 |
case 26: |
461 |
WANT (5); |
462 |
dec->cur += 5; |
463 |
return (((UV)dec->cur[-4]) << 24) |
464 |
| (((UV)dec->cur[-3]) << 16) |
465 |
| (((UV)dec->cur[-2]) << 8) |
466 |
| ((UV)dec->cur[-1]); |
467 |
|
468 |
case 27: |
469 |
WANT (9); |
470 |
dec->cur += 9; |
471 |
return (((UV)dec->cur[-8]) << 56) |
472 |
| (((UV)dec->cur[-7]) << 48) |
473 |
| (((UV)dec->cur[-6]) << 40) |
474 |
| (((UV)dec->cur[-5]) << 32) |
475 |
| (((UV)dec->cur[-4]) << 24) |
476 |
| (((UV)dec->cur[-3]) << 16) |
477 |
| (((UV)dec->cur[-2]) << 8) |
478 |
| ((UV)dec->cur[-1]); |
479 |
|
480 |
default: |
481 |
ERR ("corrupted CBOR data (unsupported integer minor encoding)"); |
482 |
} |
483 |
|
484 |
fail: |
485 |
return 0; |
486 |
} |
487 |
|
488 |
static SV *decode_sv (dec_t *dec); |
489 |
|
490 |
static SV * |
491 |
decode_av (dec_t *dec) |
492 |
{ |
493 |
AV *av = newAV (); |
494 |
|
495 |
DEC_INC_DEPTH; |
496 |
|
497 |
if ((*dec->cur & 31) == 31) |
498 |
{ |
499 |
++dec->cur; |
500 |
|
501 |
for (;;) |
502 |
{ |
503 |
WANT (1); |
504 |
|
505 |
if (*dec->cur == (0xe0 | 31)) |
506 |
{ |
507 |
++dec->cur; |
508 |
break; |
509 |
} |
510 |
|
511 |
av_push (av, decode_sv (dec)); |
512 |
} |
513 |
} |
514 |
else |
515 |
{ |
516 |
int i, len = decode_uint (dec); |
517 |
|
518 |
av_fill (av, len - 1); |
519 |
|
520 |
for (i = 0; i < len; ++i) |
521 |
AvARRAY (av)[i] = decode_sv (dec); |
522 |
} |
523 |
|
524 |
DEC_DEC_DEPTH; |
525 |
return newRV_noinc ((SV *)av); |
526 |
|
527 |
fail: |
528 |
SvREFCNT_dec (av); |
529 |
DEC_DEC_DEPTH; |
530 |
return &PL_sv_undef; |
531 |
} |
532 |
|
533 |
static SV * |
534 |
decode_hv (dec_t *dec) |
535 |
{ |
536 |
HV *hv = newHV (); |
537 |
|
538 |
DEC_INC_DEPTH; |
539 |
|
540 |
if ((*dec->cur & 31) == 31) |
541 |
{ |
542 |
++dec->cur; |
543 |
|
544 |
for (;;) |
545 |
{ |
546 |
WANT (1); |
547 |
|
548 |
if (*dec->cur == (0xe0 | 31)) |
549 |
{ |
550 |
++dec->cur; |
551 |
break; |
552 |
} |
553 |
|
554 |
SV *k = decode_sv (dec); |
555 |
SV *v = decode_sv (dec); |
556 |
|
557 |
hv_store_ent (hv, k, v, 0); |
558 |
} |
559 |
} |
560 |
else |
561 |
{ |
562 |
int len = decode_uint (dec); |
563 |
|
564 |
while (len--) |
565 |
{ |
566 |
SV *k = decode_sv (dec); |
567 |
SV *v = decode_sv (dec); |
568 |
|
569 |
hv_store_ent (hv, k, v, 0); |
570 |
} |
571 |
} |
572 |
|
573 |
DEC_DEC_DEPTH; |
574 |
return newRV_noinc ((SV *)hv); |
575 |
|
576 |
#if 0 |
577 |
SV *sv; |
578 |
HV *hv = newHV (); |
579 |
|
580 |
DEC_INC_DEPTH; |
581 |
decode_ws (dec); |
582 |
|
583 |
for (;;) |
584 |
{ |
585 |
// heuristic: assume that |
586 |
// a) decode_str + hv_store_ent are abysmally slow. |
587 |
// b) most hash keys are short, simple ascii text. |
588 |
// => try to "fast-match" such strings to avoid |
589 |
// the overhead of decode_str + hv_store_ent. |
590 |
{ |
591 |
SV *value; |
592 |
char *p = dec->cur; |
593 |
char *e = p + 24; // only try up to 24 bytes |
594 |
|
595 |
for (;;) |
596 |
{ |
597 |
// the >= 0x80 is false on most architectures |
598 |
if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
599 |
{ |
600 |
// slow path, back up and use decode_str |
601 |
SV *key = decode_str (dec); |
602 |
if (!key) |
603 |
goto fail; |
604 |
|
605 |
decode_ws (dec); EXPECT_CH (':'); |
606 |
|
607 |
decode_ws (dec); |
608 |
value = decode_sv (dec); |
609 |
if (!value) |
610 |
{ |
611 |
SvREFCNT_dec (key); |
612 |
goto fail; |
613 |
} |
614 |
|
615 |
hv_store_ent (hv, key, value, 0); |
616 |
SvREFCNT_dec (key); |
617 |
|
618 |
break; |
619 |
} |
620 |
else if (*p == '"') |
621 |
{ |
622 |
// fast path, got a simple key |
623 |
char *key = dec->cur; |
624 |
int len = p - key; |
625 |
dec->cur = p + 1; |
626 |
|
627 |
decode_ws (dec); EXPECT_CH (':'); |
628 |
|
629 |
decode_ws (dec); |
630 |
value = decode_sv (dec); |
631 |
if (!value) |
632 |
goto fail; |
633 |
|
634 |
hv_store (hv, key, len, value, 0); |
635 |
|
636 |
break; |
637 |
} |
638 |
|
639 |
++p; |
640 |
} |
641 |
} |
642 |
|
643 |
decode_ws (dec); |
644 |
|
645 |
if (*dec->cur == '}') |
646 |
{ |
647 |
++dec->cur; |
648 |
break; |
649 |
} |
650 |
|
651 |
if (*dec->cur != ',') |
652 |
ERR (", or } expected while parsing object/hash"); |
653 |
|
654 |
++dec->cur; |
655 |
|
656 |
decode_ws (dec); |
657 |
|
658 |
if (*dec->cur == '}' && dec->cbor.flags & F_RELAXED) |
659 |
{ |
660 |
++dec->cur; |
661 |
break; |
662 |
} |
663 |
} |
664 |
|
665 |
DEC_DEC_DEPTH; |
666 |
sv = newRV_noinc ((SV *)hv); |
667 |
|
668 |
// check filter callbacks |
669 |
if (dec->cbor.flags & F_HOOK) |
670 |
{ |
671 |
if (dec->cbor.cb_sk_object && HvKEYS (hv) == 1) |
672 |
{ |
673 |
HE *cb, *he; |
674 |
|
675 |
hv_iterinit (hv); |
676 |
he = hv_iternext (hv); |
677 |
hv_iterinit (hv); |
678 |
|
679 |
// the next line creates a mortal sv each time its called. |
680 |
// might want to optimise this for common cases. |
681 |
cb = hv_fetch_ent (dec->cbor.cb_sk_object, hv_iterkeysv (he), 0, 0); |
682 |
|
683 |
if (cb) |
684 |
{ |
685 |
dSP; |
686 |
int count; |
687 |
|
688 |
ENTER; SAVETMPS; PUSHMARK (SP); |
689 |
XPUSHs (HeVAL (he)); |
690 |
sv_2mortal (sv); |
691 |
|
692 |
PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
693 |
|
694 |
if (count == 1) |
695 |
{ |
696 |
sv = newSVsv (POPs); |
697 |
FREETMPS; LEAVE; |
698 |
return sv; |
699 |
} |
700 |
|
701 |
SvREFCNT_inc (sv); |
702 |
FREETMPS; LEAVE; |
703 |
} |
704 |
} |
705 |
|
706 |
if (dec->cbor.cb_object) |
707 |
{ |
708 |
dSP; |
709 |
int count; |
710 |
|
711 |
ENTER; SAVETMPS; PUSHMARK (SP); |
712 |
XPUSHs (sv_2mortal (sv)); |
713 |
|
714 |
PUTBACK; count = call_sv (dec->cbor.cb_object, G_ARRAY); SPAGAIN; |
715 |
|
716 |
if (count == 1) |
717 |
{ |
718 |
sv = newSVsv (POPs); |
719 |
FREETMPS; LEAVE; |
720 |
return sv; |
721 |
} |
722 |
|
723 |
SvREFCNT_inc (sv); |
724 |
FREETMPS; LEAVE; |
725 |
} |
726 |
} |
727 |
|
728 |
return sv; |
729 |
#endif |
730 |
|
731 |
fail: |
732 |
SvREFCNT_dec (hv); |
733 |
DEC_DEC_DEPTH; |
734 |
return &PL_sv_undef; |
735 |
} |
736 |
|
737 |
static SV * |
738 |
decode_str (dec_t *dec, int utf8) |
739 |
{ |
740 |
SV *sv; |
741 |
|
742 |
if ((*dec->cur & 31) == 31) |
743 |
{ |
744 |
++dec->cur; |
745 |
|
746 |
sv = newSVpvn ("", 0); |
747 |
|
748 |
// not very fast, and certainly not robust against illegal input |
749 |
for (;;) |
750 |
{ |
751 |
WANT (1); |
752 |
|
753 |
if (*dec->cur == (0xe0 | 31)) |
754 |
{ |
755 |
++dec->cur; |
756 |
break; |
757 |
} |
758 |
|
759 |
SV *sv2 = decode_sv (dec); |
760 |
sv_catsv (sv, sv2); |
761 |
} |
762 |
} |
763 |
else |
764 |
{ |
765 |
STRLEN len = decode_uint (dec); |
766 |
|
767 |
WANT (len); |
768 |
sv = newSVpvn (dec->cur, len); |
769 |
dec->cur += len; |
770 |
} |
771 |
|
772 |
if (utf8) |
773 |
SvUTF8_on (sv); |
774 |
|
775 |
return sv; |
776 |
|
777 |
fail: |
778 |
return &PL_sv_undef; |
779 |
} |
780 |
|
781 |
static SV * |
782 |
decode_tagged (dec_t *dec) |
783 |
{ |
784 |
UV tag = decode_uint (dec); |
785 |
SV *sv = decode_sv (dec); |
786 |
|
787 |
if (tag == 55799) // 2.4.5 Self-Describe CBOR |
788 |
return sv; |
789 |
|
790 |
AV *av = newAV (); |
791 |
av_push (av, newSVuv (tag)); |
792 |
av_push (av, sv); |
793 |
return newRV_noinc ((SV *)av); |
794 |
} |
795 |
|
796 |
static SV * |
797 |
decode_sv (dec_t *dec) |
798 |
{ |
799 |
WANT (1); |
800 |
|
801 |
switch (*dec->cur >> 5) |
802 |
{ |
803 |
case 0: // unsigned int |
804 |
//TODO: 64 bit values on 3 2bit perls |
805 |
return newSVuv (decode_uint (dec)); |
806 |
case 1: // negative int |
807 |
return newSViv (-1 - (IV)decode_uint (dec)); |
808 |
case 2: // octet string |
809 |
return decode_str (dec, 0); |
810 |
case 3: // utf-8 string |
811 |
return decode_str (dec, 1); |
812 |
case 4: // array |
813 |
return decode_av (dec); |
814 |
case 5: // map |
815 |
return decode_hv (dec); |
816 |
case 6: // tag |
817 |
return decode_tagged (dec); |
818 |
case 7: // misc |
819 |
switch (*dec->cur++ & 31) |
820 |
{ |
821 |
case 20: |
822 |
#if CBOR_SLOW |
823 |
cbor_false = get_bool ("CBOR::XS::false"); |
824 |
#endif |
825 |
return newSVsv (cbor_false); |
826 |
case 21: |
827 |
#if CBOR_SLOW |
828 |
cbor_true = get_bool ("CBOR::XS::true"); |
829 |
#endif |
830 |
return newSVsv (cbor_true); |
831 |
case 22: |
832 |
return newSVsv (&PL_sv_undef); |
833 |
|
834 |
case 25: |
835 |
{ |
836 |
WANT (2); |
837 |
|
838 |
uint16_t fp = (dec->cur[0] << 8) | dec->cur[1]; |
839 |
dec->cur += 2; |
840 |
|
841 |
return newSVnv (ecb_binary16_to_float (fp)); |
842 |
} |
843 |
|
844 |
case 26: |
845 |
{ |
846 |
uint32_t fp; |
847 |
WANT (4); |
848 |
memcpy (&fp, dec->cur, 4); |
849 |
dec->cur += 4; |
850 |
|
851 |
if (!ecb_big_endian ()) |
852 |
fp = ecb_bswap32 (fp); |
853 |
|
854 |
return newSVnv (ecb_binary32_to_float (fp)); |
855 |
} |
856 |
|
857 |
case 27: |
858 |
{ |
859 |
uint64_t fp; |
860 |
WANT (8); |
861 |
memcpy (&fp, dec->cur, 8); |
862 |
dec->cur += 8; |
863 |
|
864 |
if (!ecb_big_endian ()) |
865 |
fp = ecb_bswap64 (fp); |
866 |
|
867 |
return newSVnv (ecb_binary64_to_double (fp)); |
868 |
} |
869 |
|
870 |
// 0..19 unassigned |
871 |
// 24 reserved + unassigned (reserved values are not encodable) |
872 |
default: |
873 |
ERR ("corrupted CBOR data (reserved/unassigned major 7 value)"); |
874 |
} |
875 |
|
876 |
break; |
877 |
} |
878 |
#if 0 |
879 |
switch (*dec->cur) |
880 |
{ |
881 |
//case '"': ++dec->cur; return decode_str (dec); |
882 |
case '[': ++dec->cur; return decode_av (dec); |
883 |
case '{': ++dec->cur; return decode_hv (dec); |
884 |
|
885 |
case '-': |
886 |
case '0': case '1': case '2': case '3': case '4': |
887 |
case '5': case '6': case '7': case '8': case '9': |
888 |
//TODO return decode_num (dec); |
889 |
|
890 |
case 't': |
891 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
892 |
{ |
893 |
dec->cur += 4; |
894 |
#if CBOR_SLOW |
895 |
cbor_true = get_bool ("CBOR::XS::true"); |
896 |
#endif |
897 |
return newSVsv (cbor_true); |
898 |
} |
899 |
else |
900 |
ERR ("'true' expected"); |
901 |
|
902 |
break; |
903 |
|
904 |
case 'f': |
905 |
if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
906 |
{ |
907 |
dec->cur += 5; |
908 |
#if CBOR_SLOW |
909 |
cbor_false = get_bool ("CBOR::XS::false"); |
910 |
#endif |
911 |
return newSVsv (cbor_false); |
912 |
} |
913 |
else |
914 |
ERR ("'false' expected"); |
915 |
|
916 |
break; |
917 |
|
918 |
case 'n': |
919 |
if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) |
920 |
{ |
921 |
dec->cur += 4; |
922 |
return newSVsv (&PL_sv_undef); |
923 |
} |
924 |
else |
925 |
ERR ("'null' expected"); |
926 |
|
927 |
break; |
928 |
|
929 |
default: |
930 |
ERR ("malformed CBOR string, neither array, object, number, string or atom"); |
931 |
break; |
932 |
} |
933 |
#endif |
934 |
|
935 |
fail: |
936 |
return &PL_sv_undef; |
937 |
} |
938 |
|
939 |
static SV * |
940 |
decode_cbor (SV *string, CBOR *cbor, char **offset_return) |
941 |
{ |
942 |
dec_t dec; |
943 |
SV *sv; |
944 |
|
945 |
/* work around bugs in 5.10 where manipulating magic values |
946 |
* makes perl ignore the magic in subsequent accesses. |
947 |
* also make a copy of non-PV values, to get them into a clean |
948 |
* state (SvPV should do that, but it's buggy, see below). |
949 |
*/ |
950 |
/*SvGETMAGIC (string);*/ |
951 |
if (SvMAGICAL (string) || !SvPOK (string)) |
952 |
string = sv_2mortal (newSVsv (string)); |
953 |
|
954 |
SvUPGRADE (string, SVt_PV); |
955 |
|
956 |
/* work around a bug in perl 5.10, which causes SvCUR to fail an |
957 |
* assertion with -DDEBUGGING, although SvCUR is documented to |
958 |
* return the xpv_cur field which certainly exists after upgrading. |
959 |
* according to nicholas clark, calling SvPOK fixes this. |
960 |
* But it doesn't fix it, so try another workaround, call SvPV_nolen |
961 |
* and hope for the best. |
962 |
* Damnit, SvPV_nolen still trips over yet another assertion. This |
963 |
* assertion business is seriously broken, try yet another workaround |
964 |
* for the broken -DDEBUGGING. |
965 |
*/ |
966 |
{ |
967 |
#ifdef DEBUGGING |
968 |
STRLEN offset = SvOK (string) ? sv_len (string) : 0; |
969 |
#else |
970 |
STRLEN offset = SvCUR (string); |
971 |
#endif |
972 |
|
973 |
if (offset > cbor->max_size && cbor->max_size) |
974 |
croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu", |
975 |
(unsigned long)SvCUR (string), (unsigned long)cbor->max_size); |
976 |
} |
977 |
|
978 |
sv_utf8_downgrade (string, 0); |
979 |
|
980 |
dec.cbor = *cbor; |
981 |
dec.cur = (U8 *)SvPVX (string); |
982 |
dec.end = (U8 *)SvEND (string); |
983 |
dec.err = 0; |
984 |
dec.depth = 0; |
985 |
|
986 |
if (dec.cbor.cb_object || dec.cbor.cb_sk_object) |
987 |
;//TODO dec.cbor.flags |= F_HOOK; |
988 |
|
989 |
sv = decode_sv (&dec); |
990 |
|
991 |
if (offset_return) |
992 |
*offset_return = dec.cur; |
993 |
|
994 |
if (!(offset_return || !sv)) |
995 |
if (dec.cur != dec.end && !dec.err) |
996 |
dec.err = "garbage after CBOR object"; |
997 |
|
998 |
if (dec.err) |
999 |
{ |
1000 |
SvREFCNT_dec (sv); |
1001 |
croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur); |
1002 |
} |
1003 |
|
1004 |
sv = sv_2mortal (sv); |
1005 |
|
1006 |
return sv; |
1007 |
} |
1008 |
|
1009 |
///////////////////////////////////////////////////////////////////////////// |
1010 |
// XS interface functions |
1011 |
|
1012 |
MODULE = CBOR::XS PACKAGE = CBOR::XS |
1013 |
|
1014 |
BOOT: |
1015 |
{ |
1016 |
cbor_stash = gv_stashpv ("CBOR::XS" , 1); |
1017 |
cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1); |
1018 |
|
1019 |
cbor_true = get_bool ("CBOR::XS::true"); |
1020 |
cbor_false = get_bool ("CBOR::XS::false"); |
1021 |
} |
1022 |
|
1023 |
PROTOTYPES: DISABLE |
1024 |
|
1025 |
void CLONE (...) |
1026 |
CODE: |
1027 |
cbor_stash = 0; |
1028 |
cbor_boolean_stash = 0; |
1029 |
|
1030 |
void new (char *klass) |
1031 |
PPCODE: |
1032 |
{ |
1033 |
SV *pv = NEWSV (0, sizeof (CBOR)); |
1034 |
SvPOK_only (pv); |
1035 |
cbor_init ((CBOR *)SvPVX (pv)); |
1036 |
XPUSHs (sv_2mortal (sv_bless ( |
1037 |
newRV_noinc (pv), |
1038 |
strEQ (klass, "CBOR::XS") ? CBOR_STASH : gv_stashpv (klass, 1) |
1039 |
))); |
1040 |
} |
1041 |
|
1042 |
void shrink (CBOR *self, int enable = 1) |
1043 |
ALIAS: |
1044 |
shrink = F_SHRINK |
1045 |
allow_unknown = F_ALLOW_UNKNOWN |
1046 |
PPCODE: |
1047 |
{ |
1048 |
if (enable) |
1049 |
self->flags |= ix; |
1050 |
else |
1051 |
self->flags &= ~ix; |
1052 |
|
1053 |
XPUSHs (ST (0)); |
1054 |
} |
1055 |
|
1056 |
void get_shrink (CBOR *self) |
1057 |
ALIAS: |
1058 |
get_shrink = F_SHRINK |
1059 |
get_allow_unknown = F_ALLOW_UNKNOWN |
1060 |
PPCODE: |
1061 |
XPUSHs (boolSV (self->flags & ix)); |
1062 |
|
1063 |
void max_depth (CBOR *self, U32 max_depth = 0x80000000UL) |
1064 |
PPCODE: |
1065 |
self->max_depth = max_depth; |
1066 |
XPUSHs (ST (0)); |
1067 |
|
1068 |
U32 get_max_depth (CBOR *self) |
1069 |
CODE: |
1070 |
RETVAL = self->max_depth; |
1071 |
OUTPUT: |
1072 |
RETVAL |
1073 |
|
1074 |
void max_size (CBOR *self, U32 max_size = 0) |
1075 |
PPCODE: |
1076 |
self->max_size = max_size; |
1077 |
XPUSHs (ST (0)); |
1078 |
|
1079 |
int get_max_size (CBOR *self) |
1080 |
CODE: |
1081 |
RETVAL = self->max_size; |
1082 |
OUTPUT: |
1083 |
RETVAL |
1084 |
|
1085 |
#if 0 //TODO |
1086 |
|
1087 |
void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef) |
1088 |
PPCODE: |
1089 |
{ |
1090 |
SvREFCNT_dec (self->cb_object); |
1091 |
self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
1092 |
|
1093 |
XPUSHs (ST (0)); |
1094 |
} |
1095 |
|
1096 |
void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef) |
1097 |
PPCODE: |
1098 |
{ |
1099 |
if (!self->cb_sk_object) |
1100 |
self->cb_sk_object = newHV (); |
1101 |
|
1102 |
if (SvOK (cb)) |
1103 |
hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
1104 |
else |
1105 |
{ |
1106 |
hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
1107 |
|
1108 |
if (!HvKEYS (self->cb_sk_object)) |
1109 |
{ |
1110 |
SvREFCNT_dec (self->cb_sk_object); |
1111 |
self->cb_sk_object = 0; |
1112 |
} |
1113 |
} |
1114 |
|
1115 |
XPUSHs (ST (0)); |
1116 |
} |
1117 |
|
1118 |
#endif |
1119 |
|
1120 |
void encode (CBOR *self, SV *scalar) |
1121 |
PPCODE: |
1122 |
PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN; |
1123 |
XPUSHs (scalar); |
1124 |
|
1125 |
void decode (CBOR *self, SV *cborstr) |
1126 |
PPCODE: |
1127 |
PUTBACK; cborstr = decode_cbor (cborstr, self, 0); SPAGAIN; |
1128 |
XPUSHs (cborstr); |
1129 |
|
1130 |
void decode_prefix (CBOR *self, SV *cborstr) |
1131 |
PPCODE: |
1132 |
{ |
1133 |
SV *sv; |
1134 |
char *offset; |
1135 |
PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN; |
1136 |
EXTEND (SP, 2); |
1137 |
PUSHs (sv); |
1138 |
PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr)))); |
1139 |
} |
1140 |
|
1141 |
void DESTROY (CBOR *self) |
1142 |
CODE: |
1143 |
SvREFCNT_dec (self->cb_sk_object); |
1144 |
SvREFCNT_dec (self->cb_object); |
1145 |
|
1146 |
PROTOTYPES: ENABLE |
1147 |
|
1148 |
void encode_cbor (SV *scalar) |
1149 |
PPCODE: |
1150 |
{ |
1151 |
CBOR cbor; |
1152 |
cbor_init (&cbor); |
1153 |
PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN; |
1154 |
XPUSHs (scalar); |
1155 |
} |
1156 |
|
1157 |
void decode_cbor (SV *cborstr) |
1158 |
PPCODE: |
1159 |
{ |
1160 |
CBOR cbor; |
1161 |
cbor_init (&cbor); |
1162 |
PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN; |
1163 |
XPUSHs (cborstr); |
1164 |
} |
1165 |
|