[ViewVC] Diff of: cvs/CBOR-XS/XS.xs

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.4 by root, Sat Oct 26 21:06:37 2013 UTC vs.
Revision 1.18 by root, Wed Nov 20 01:09:46 2013 UTC

…		…
9	#include <limits.h>	9	#include <limits.h>
10	#include <float.h>	10	#include <float.h>
11		11
12	#include "ecb.h"	12	#include "ecb.h"
13		13
		14	// compatibility with perl <5.18
		15	#ifndef HvNAMELEN_get
		16	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		17	#endif
		18	#ifndef HvNAMELEN
		19	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		20	#endif
		21	#ifndef HvNAMEUTF8
		22	# define HvNAMEUTF8(hv) 0
		23	#endif
		24
		25	// known tags
		26	enum cbor_tag
		27	{
		28	// inofficial extensions (pending iana registration)
		29	CBOR_TAG_PERL_OBJECT = 24, // http://cbor.schmorp.de/perl-object
		30	CBOR_TAG_GENERIC_OBJECT = 25, // http://cbor.schmorp.de/generic-object
		31	CBOR_TAG_VALUE_SHARABLE = 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
14	#define F_SHRINK 0x00000200UL	59	#define F_SHRINK 0x00000001UL
15	#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
16		64
17	#define INIT_SIZE 32 // initial scalar size to be allocated	65	#define INIT_SIZE 32 // initial scalar size to be allocated
18		66
19	#define SB do {	67	#define SB do {
20	#define SE } while (0)	68	#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		69
33	#define IN_RANGE_INC(type,val,beg,end) \	70	#define IN_RANGE_INC(type,val,beg,end) \
34	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	71	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
35	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	72	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
36		73
…		…
42	#else	79	#else
43	# define CBOR_SLOW 0	80	# define CBOR_SLOW 0
44	# define CBOR_STASH cbor_stash	81	# define CBOR_STASH cbor_stash
45	#endif	82	#endif
46		83
47	static HV cbor_stash, cbor_boolean_stash; // CBOR::XS::	84	static HV cbor_stash, types_boolean_stash, types_error_stash, cbor_tagged_stash; // CBOR::XS::
48	static SV cbor_true, cbor_false;	85	static SV types_true, types_false, types_error, sv_cbor;
49		86
50	typedef struct {	87	typedef struct {
51	U32 flags;	88	U32 flags;
52	U32 max_depth;	89	U32 max_depth;
53	STRLEN max_size;	90	STRLEN max_size;
54
55	SV *cb_object;
56	HV *cb_sk_object;
57	} CBOR;	91	} CBOR;
58		92
59	INLINE void	93	ecb_inline void
60	cbor_init (CBOR *cbor)	94	cbor_init (CBOR *cbor)
61	{	95	{
62	Zero (cbor, 1, CBOR);	96	Zero (cbor, 1, CBOR);
63	cbor->max_depth = 512;	97	cbor->max_depth = 512;
64	}	98	}
65		99
66	/////////////////////////////////////////////////////////////////////////////	100	/////////////////////////////////////////////////////////////////////////////
67	// utility functions	101	// utility functions
68		102
69	INLINE SV *	103	ecb_inline SV *
70	get_bool (const char *name)	104	get_bool (const char *name)
71	{	105	{
72	SV *sv = get_sv (name, 1);	106	SV *sv = get_sv (name, 1);
73		107
74	SvREADONLY_on (sv);	108	SvREADONLY_on (sv);
75	SvREADONLY_on (SvRV (sv));	109	SvREADONLY_on (SvRV (sv));
76		110
77	return sv;	111	return sv;
78	}	112	}
79		113
80	INLINE void	114	ecb_inline void
81	shrink (SV *sv)	115	shrink (SV *sv)
82	{	116	{
83	sv_utf8_downgrade (sv, 1);	117	sv_utf8_downgrade (sv, 1);
84		118
85	if (SvLEN (sv) > SvCUR (sv) + 1)	119	if (SvLEN (sv) > SvCUR (sv) + 1)
…		…
91	#endif	125	#endif
92	}	126	}
93	}	127	}
94		128
95	/////////////////////////////////////////////////////////////////////////////	129	/////////////////////////////////////////////////////////////////////////////
96	// fp hell
97
98	//TODO
99
100	/////////////////////////////////////////////////////////////////////////////
101	// encoder	130	// encoder
102		131
103	// structure used for encoding CBOR	132	// structure used for encoding CBOR
104	typedef struct	133	typedef struct
105	{	134	{
106	char *cur; // SvPVX (sv) + current output position	135	char *cur; // SvPVX (sv) + current output position
107	char *end; // SvEND (sv)	136	char *end; // SvEND (sv)
108	SV *sv; // result scalar	137	SV *sv; // result scalar
109	CBOR cbor;	138	CBOR cbor;
110	U32 depth; // recursion level	139	U32 depth; // recursion level
		140	HV *stringref; // string => index, or 0
		141	HV *sharable; // ptr => index, or 0
		142	HV *sharable_idx;
111	} enc_t;	143	} enc_t;
112		144
113	INLINE void	145	ecb_inline void
114	need (enc_t *enc, STRLEN len)	146	need (enc_t *enc, STRLEN len)
115	{	147	{
116	if (expect_false (enc->cur + len >= enc->end))	148	if (ecb_expect_false (enc->cur + len >= enc->end))
117	{	149	{
118	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);	150	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
119	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);	151	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
120	enc->cur = SvPVX (enc->sv) + cur;	152	enc->cur = SvPVX (enc->sv) + cur;
121	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	153	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
122	}	154	}
123	}	155	}
124		156
125	INLINE void	157	ecb_inline void
126	encode_ch (enc_t *enc, char ch)	158	encode_ch (enc_t *enc, char ch)
127	{	159	{
128	need (enc, 1);	160	need (enc, 1);
129	*enc->cur++ = ch;	161	*enc->cur++ = ch;
130	}	162	}
…		…
176	need (enc, len);	208	need (enc, len);
177	memcpy (enc->cur, str, len);	209	memcpy (enc->cur, str, len);
178	enc->cur += len;	210	enc->cur += len;
179	}	211	}
180		212
		213	ecb_inline void
		214	encode_tag (enc_t *enc, UV tag)
		215	{
		216	encode_uint (enc, 0xc0, tag);
		217	}
		218
		219	static int
		220	encode_sharable2 (enc_t enc, SV sv)
		221	{
		222	if (!enc->sharable)
		223	enc->sharable = (HV )sv_2mortal ((SV )newHV ());
		224
		225	SV **svp = hv_fetch (enc->sharable, &sv, sizeof (sv), 1);
		226
		227	if (SvOK (*svp))
		228	{
		229	encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
		230	encode_uint (enc, 0x00, SvUV (*svp));
		231
		232	return 1;
		233	}
		234	else
		235	{
		236	sv_setuv (*svp, enc->sharable_idx++);
		237	encode_tag (enc, CBOR_TAG_VALUE_SHARABLE);
		238
		239	return 0;
		240	}
		241	}
		242
		243	ecb_inline int
		244	encode_sharable (enc_t enc, SV sv)
		245	{
		246	if (ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING)
		247	&& ecb_expect_false (SvREFCNT (sv) > 1))
		248	return encode_sharable2 (enc, sv);
		249
		250	return 0;
		251	}
		252
181	static void encode_sv (enc_t enc, SV sv);	253	static void encode_sv (enc_t enc, SV sv);
182		254
183	static void	255	static void
184	encode_av (enc_t enc, AV av)	256	encode_av (enc_t enc, AV av)
185	{	257	{
…		…
224	if (HeKLEN (he) == HEf_SVKEY)	296	if (HeKLEN (he) == HEf_SVKEY)
225	encode_sv (enc, HeSVKEY (he));	297	encode_sv (enc, HeSVKEY (he));
226	else	298	else
227	encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));	299	encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));
228		300
229	encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));	301	encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
230	}	302	}
231		303
232	if (mg)	304	if (mg)
233	encode_ch (enc, 0xe0 \| 31);	305	encode_ch (enc, 0xe0 \| 31);
234		306
…		…
242	svtype svt;	314	svtype svt;
243		315
244	SvGETMAGIC (sv);	316	SvGETMAGIC (sv);
245	svt = SvTYPE (sv);	317	svt = SvTYPE (sv);
246		318
		319	if (encode_sharable (enc, sv))
		320	return;
		321
247	if (expect_false (SvOBJECT (sv)))	322	if (ecb_expect_false (SvOBJECT (sv)))
248	{	323	{
249	HV *stash = !CBOR_SLOW \|\| cbor_boolean_stash	324	HV *boolean_stash = !CBOR_SLOW \|\| types_boolean_stash
250	? cbor_boolean_stash	325	? types_boolean_stash
		326	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		327	HV *error_stash = !CBOR_SLOW \|\| types_error_stash
		328	? types_error_stash
		329	: gv_stashpv ("Types::Serialiser::Error", 1);
		330	HV *tagged_stash = !CBOR_SLOW \|\| cbor_tagged_stash
		331	? cbor_tagged_stash
251	: gv_stashpv ("CBOR::XS::Boolean", 1);	332	: gv_stashpv ("CBOR::XS::Tagged" , 1);
252		333
253	if (SvSTASH (sv) == stash)	334	HV *stash = SvSTASH (sv);
		335	GV *method;
		336
		337	if (stash == boolean_stash)
254	encode_ch (enc, SvIV (sv) ? 0xe0 \| 21 : 0xe0 \| 20);	338	encode_ch (enc, SvIV (sv) ? 0xe0 \| 21 : 0xe0 \| 20);
		339	else if (stash == error_stash)
		340	encode_ch (enc, 0xe0 \| 23);
		341	else if (stash == tagged_stash)
		342	{
		343	if (svt != SVt_PVAV)
		344	croak ("encountered CBOR::XS::Tagged object that isn't an array");
		345
		346	encode_uint (enc, 0xc0, SvUV (av_fetch ((AV )sv, 0, 1)));
		347	encode_sv (enc, av_fetch ((AV )sv, 1, 1));
		348	}
		349	else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
		350	{
		351	dSP;
		352
		353	ENTER; SAVETMPS; PUSHMARK (SP);
		354	// we re-bless the reference to get overload and other niceties right
		355	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		356
		357	PUTBACK;
		358	// G_SCALAR ensures that return value is 1
		359	call_sv ((SV *)GvCV (method), G_SCALAR);
		360	SPAGAIN;
		361
		362	// catch this surprisingly common error
		363	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		364	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
		365
		366	encode_sv (enc, POPs);
		367
		368	PUTBACK;
		369
		370	FREETMPS; LEAVE;
		371	}
		372	else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
		373	{
		374	dSP;
		375
		376	ENTER; SAVETMPS; PUSHMARK (SP);
		377	EXTEND (SP, 2);
		378	// we re-bless the reference to get overload and other niceties right
		379	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		380	PUSHs (sv_cbor);
		381
		382	PUTBACK;
		383	int count = call_sv ((SV *)GvCV (method), G_ARRAY);
		384	SPAGAIN;
		385
		386	// catch this surprisingly common error
		387	if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
		388	croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
		389
		390	encode_tag (enc, CBOR_TAG_PERL_OBJECT);
		391	encode_uint (enc, 0x80, count + 1);
		392	encode_str (enc, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
		393
		394	while (count)
		395	encode_sv (enc, SP[1 - count--]);
		396
		397	PUTBACK;
		398
		399	FREETMPS; LEAVE;
		400	}
255	else	401	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",	402	croak ("encountered object '%s', but no TO_CBOR or FREEZE methods 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))));	403	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
297	#endif
298	}
299	}	404	}
300	else if (svt == SVt_PVHV)	405	else if (svt == SVt_PVHV)
301	encode_hv (enc, (HV *)sv);	406	encode_hv (enc, (HV *)sv);
302	else if (svt == SVt_PVAV)	407	else if (svt == SVt_PVAV)
303	encode_av (enc, (AV *)sv);	408	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	409	else
322	croak ("encountered %s, but CBOR can only represent references to arrays or hashes",	410	{
323	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	411	encode_tag (enc, CBOR_TAG_INDIRECTION);
		412	encode_sv (enc, sv);
		413	}
324	}	414	}
325		415
326	static void	416	static void
327	encode_nv (enc_t enc, SV sv)	417	encode_nv (enc_t enc, SV sv)
328	{	418	{
329	double nv = SvNVX (sv);	419	double nv = SvNVX (sv);
330		420
331	need (enc, 9);	421	need (enc, 9);
332		422
333	if (expect_false (nv == (U32)nv))	423	if (ecb_expect_false (nv == (U32)nv))
334	encode_uint (enc, 0x00, (U32)nv);	424	encode_uint (enc, 0x00, (U32)nv);
335	//TODO: maybe I32?	425	//TODO: maybe I32?
336	else if (expect_false (nv == (float)nv))	426	else if (ecb_expect_false (nv == (float)nv))
337	{	427	{
338	uint32_t fp = ecb_float_to_binary32 (nv);	428	uint32_t fp = ecb_float_to_binary32 (nv);
339		429
340	*enc->cur++ = 0xe0 \| 26;	430	*enc->cur++ = 0xe0 \| 26;
341		431
…		…
361		451
362	static void	452	static void
363	encode_sv (enc_t enc, SV sv)	453	encode_sv (enc_t enc, SV sv)
364	{	454	{
365	SvGETMAGIC (sv);	455	SvGETMAGIC (sv);
		456
		457	if (encode_sharable (enc, sv))
		458	return;
366		459
367	if (SvPOKp (sv))	460	if (SvPOKp (sv))
368	{	461	{
369	STRLEN len;	462	STRLEN len;
370	char *str = SvPV (sv, len);	463	char *str = SvPV (sv, len);
…		…
393	}	486	}
394		487
395	static SV *	488	static SV *
396	encode_cbor (SV scalar, CBOR cbor)	489	encode_cbor (SV scalar, CBOR cbor)
397	{	490	{
398	enc_t enc;	491	enc_t enc = { };
399		492
400	enc.cbor = *cbor;	493	enc.cbor = *cbor;
401	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	494	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
402	enc.cur = SvPVX (enc.sv);	495	enc.cur = SvPVX (enc.sv);
403	enc.end = SvEND (enc.sv);	496	enc.end = SvEND (enc.sv);
404	enc.depth = 0;
405		497
406	SvPOK_only (enc.sv);	498	SvPOK_only (enc.sv);
407	encode_sv (&enc, scalar);	499	encode_sv (&enc, scalar);
408		500
409	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	501	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
425	U8 *end; // end of input string	517	U8 *end; // end of input string
426	const char *err; // parse error, if != 0	518	const char *err; // parse error, if != 0
427	CBOR cbor;	519	CBOR cbor;
428	U32 depth; // recursion depth	520	U32 depth; // recursion depth
429	U32 maxdepth; // recursion depth limit	521	U32 maxdepth; // recursion depth limit
		522	AV *sharable;
430	} dec_t;	523	} dec_t;
431		524
432	#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE	525	#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE
433		526
434	#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data");	527	#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")
435		528
436	#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)	529	#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)
437	#define DEC_DEC_DEPTH --dec->depth	530	#define DEC_DEC_DEPTH --dec->depth
438		531
439	static UV	532	static UV
…		…
528	SvREFCNT_dec (av);	621	SvREFCNT_dec (av);
529	DEC_DEC_DEPTH;	622	DEC_DEC_DEPTH;
530	return &PL_sv_undef;	623	return &PL_sv_undef;
531	}	624	}
532		625
		626	static void
		627	decode_he (dec_t dec, HV hv)
		628	{
		629	// for speed reasons, we specialcase single-string
		630	// byte or utf-8 strings as keys.
		631
		632	if (dec->cur >= 0x40 && dec->cur <= 0x40 + 27)
		633	{
		634	I32 len = decode_uint (dec);
		635	char key = (char )dec->cur;
		636
		637	dec->cur += len;
		638
		639	hv_store (hv, key, len, decode_sv (dec), 0);
		640	}
		641	else if (dec->cur >= 0x60 && dec->cur <= 0x60 + 27)
		642	{
		643	I32 len = decode_uint (dec);
		644	char key = (char )dec->cur;
		645
		646	dec->cur += len;
		647
		648	hv_store (hv, key, -len, decode_sv (dec), 0);
		649	}
		650	else
		651	{
		652	SV *k = decode_sv (dec);
		653	SV *v = decode_sv (dec);
		654
		655	hv_store_ent (hv, k, v, 0);
		656	SvREFCNT_dec (k);
		657	}
		658	}
		659
533	static SV *	660	static SV *
534	decode_hv (dec_t *dec)	661	decode_hv (dec_t *dec)
535	{	662	{
536	HV *hv = newHV ();	663	HV *hv = newHV ();
537		664
…		…
549	{	676	{
550	++dec->cur;	677	++dec->cur;
551	break;	678	break;
552	}	679	}
553		680
554	SV *k = decode_sv (dec);	681	decode_he (dec, hv);
555	SV *v = decode_sv (dec);
556
557	hv_store_ent (hv, k, v, 0);
558	}	682	}
559	}	683	}
560	else	684	else
561	{	685	{
562	int len = decode_uint (dec);	686	int pairs = decode_uint (dec);
563		687
564	while (len--)	688	while (pairs--)
565	{	689	decode_he (dec, hv);
566	SV *k = decode_sv (dec);
567	SV *v = decode_sv (dec);
568
569	hv_store_ent (hv, k, v, 0);
570	}
571	}	690	}
572		691
573	DEC_DEC_DEPTH;	692	DEC_DEC_DEPTH;
574	return newRV_noinc ((SV *)hv);	693	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		694
731	fail:	695	fail:
732	SvREFCNT_dec (hv);	696	SvREFCNT_dec (hv);
733	DEC_DEC_DEPTH;	697	DEC_DEC_DEPTH;
734	return &PL_sv_undef;	698	return &PL_sv_undef;
735	}	699	}
736		700
737	static SV *	701	static SV *
738	decode_str (dec_t *dec, int utf8)	702	decode_str (dec_t *dec, int utf8)
739	{	703	{
740	SV *sv;	704	SV *sv = 0;
741		705
742	if ((*dec->cur & 31) == 31)	706	if ((*dec->cur & 31) == 31)
743	{	707	{
744	++dec->cur;	708	++dec->cur;
745		709
…		…
754	{	718	{
755	++dec->cur;	719	++dec->cur;
756	break;	720	break;
757	}	721	}
758		722
759	SV *sv2 = decode_sv (dec);
760	sv_catsv (sv, sv2);	723	sv_catsv (sv, decode_sv (dec));
761	}	724	}
762	}	725	}
763	else	726	else
764	{	727	{
765	STRLEN len = decode_uint (dec);	728	STRLEN len = decode_uint (dec);
…		…
773	SvUTF8_on (sv);	736	SvUTF8_on (sv);
774		737
775	return sv;	738	return sv;
776		739
777	fail:	740	fail:
		741	SvREFCNT_dec (sv);
778	return &PL_sv_undef;	742	return &PL_sv_undef;
779	}	743	}
780		744
781	static SV *	745	static SV *
782	decode_tagged (dec_t *dec)	746	decode_tagged (dec_t *dec)
783	{	747	{
784	UV tag = decode_uint (dec);	748	UV tag = decode_uint (dec);
785	SV *sv = decode_sv (dec);	749	SV *sv = decode_sv (dec);
786		750
787	if (tag == 55799) // 2.4.5 Self-Describe CBOR	751	switch (tag)
		752	{
		753	case CBOR_TAG_MAGIC:
788	return sv;	754	return sv;
789		755
		756	case CBOR_TAG_INDIRECTION:
		757	return newRV_noinc (sv);
		758
		759	case CBOR_TAG_VALUE_SHARABLE:
		760	if (ecb_expect_false (!dec->sharable))
		761	dec->sharable = (AV )sv_2mortal ((SV )newAV ());
		762
		763	av_push (dec->sharable, SvREFCNT_inc_NN (sv));
		764
		765	return sv;
		766
		767	case CBOR_TAG_VALUE_SHAREDREF:
		768	{
		769	// TODO: should verify that the sv atcually was a CBOR unsigned integer
		770	UV idx = SvUV (sv);
		771
		772	if (!dec->sharable \|\| idx > AvFILLp (dec->sharable))
		773	ERR ("corrupted CBOR data (sharedref index out of bounds)");
		774
		775	SvREFCNT_dec (sv);
		776
		777	return SvREFCNT_inc_NN (AvARRAY (dec->sharable)[idx]);
		778	}
		779
		780	case CBOR_TAG_PERL_OBJECT:
		781	{
		782	if (!SvROK (sv) \|\| SvTYPE (SvRV (sv)) != SVt_PVAV)
		783	ERR ("corrupted CBOR data (non-array perl object)");
		784
		785	AV av = (AV )SvRV (sv);
		786	int len = av_len (av) + 1;
		787	HV stash = gv_stashsv (av_fetch (av, 0, 1), 0);
		788
		789	if (!stash)
		790	ERR ("cannot decode perl-object (package does not exist)");
		791
		792	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		793
		794	if (!method)
		795	ERR ("cannot decode perl-object (package does not have a THAW method)");
		796
		797	dSP;
		798
		799	ENTER; SAVETMPS; PUSHMARK (SP);
		800	EXTEND (SP, len + 1);
		801	// we re-bless the reference to get overload and other niceties right
		802	PUSHs (*av_fetch (av, 0, 1));
		803	PUSHs (sv_cbor);
		804
		805	int i;
		806
		807	for (i = 1; i < len; ++i)
		808	PUSHs (*av_fetch (av, i, 1));
		809
		810	PUTBACK;
		811	call_sv ((SV *)GvCV (method), G_SCALAR \| G_EVAL);
		812	SPAGAIN;
		813
		814	if (SvTRUE (ERRSV))
		815	{
		816	FREETMPS; LEAVE;
		817	ERR (SvPVutf8_nolen (sv_2mortal (SvREFCNT_inc (ERRSV))));
		818	}
		819
		820	SvREFCNT_dec (sv);
		821	sv = SvREFCNT_inc (POPs);
		822
		823	PUTBACK;
		824
		825	FREETMPS; LEAVE;
		826
		827	return sv;
		828	}
		829
		830	default:
		831	{
790	AV *av = newAV ();	832	AV *av = newAV ();
791	av_push (av, newSVuv (tag));	833	av_push (av, newSVuv (tag));
792	av_push (av, sv);	834	av_push (av, sv);
793	return newRV_noinc ((SV *)av);	835
		836	HV *tagged_stash = !CBOR_SLOW \|\| cbor_tagged_stash
		837	? cbor_tagged_stash
		838	: gv_stashpv ("CBOR::XS::Tagged" , 1);
		839
		840	return sv_bless (newRV_noinc ((SV *)av), tagged_stash);
		841	}
		842	}
		843
		844	fail:
		845	SvREFCNT_dec (sv);
		846	return &PL_sv_undef;
794	}	847	}
795		848
796	static SV *	849	static SV *
797	decode_sv (dec_t *dec)	850	decode_sv (dec_t *dec)
798	{	851	{
799	WANT (1);	852	WANT (1);
800		853
801	switch (*dec->cur >> 5)	854	switch (*dec->cur >> 5)
802	{	855	{
803	case 0: // unsigned int	856	case 0: // unsigned int
804	//TODO: 64 bit values on 3 2bit perls
805	return newSVuv (decode_uint (dec));	857	return newSVuv (decode_uint (dec));
806	case 1: // negative int	858	case 1: // negative int
807	return newSViv (-1 - (IV)decode_uint (dec));	859	return newSViv (-1 - (IV)decode_uint (dec));
808	case 2: // octet string	860	case 2: // octet string
809	return decode_str (dec, 0);	861	return decode_str (dec, 0);
…		…
818	case 7: // misc	870	case 7: // misc
819	switch (*dec->cur++ & 31)	871	switch (*dec->cur++ & 31)
820	{	872	{
821	case 20:	873	case 20:
822	#if CBOR_SLOW	874	#if CBOR_SLOW
823	cbor_false = get_bool ("CBOR::XS::false");	875	types_false = get_bool ("Types::Serialiser::false");
824	#endif	876	#endif
825	return newSVsv (cbor_false);	877	return newSVsv (types_false);
826	case 21:	878	case 21:
827	#if CBOR_SLOW	879	#if CBOR_SLOW
828	cbor_true = get_bool ("CBOR::XS::true");	880	types_true = get_bool ("Types::Serialiser::true");
829	#endif	881	#endif
830	return newSVsv (cbor_true);	882	return newSVsv (types_true);
831	case 22:	883	case 22:
832	return newSVsv (&PL_sv_undef);	884	return newSVsv (&PL_sv_undef);
		885	case 23:
		886	#if CBOR_SLOW
		887	types_error = get_bool ("Types::Serialiser::error");
		888	#endif
		889	return newSVsv (types_error);
833		890
834	case 25:	891	case 25:
835	{	892	{
836	WANT (2);	893	WANT (2);
837		894
…		…
873	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");	930	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");
874	}	931	}
875		932
876	break;	933	break;
877	}	934	}
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
935	fail:	936	fail:
936	return &PL_sv_undef;	937	return &PL_sv_undef;
937	}	938	}
938		939
939	static SV *	940	static SV *
940	decode_cbor (SV string, CBOR cbor, char **offset_return)	941	decode_cbor (SV string, CBOR cbor, char **offset_return)
941	{	942	{
942	dec_t dec;	943	dec_t dec = { };
943	SV *sv;	944	SV *sv;
		945	STRLEN len;
		946	char *data = SvPVbyte (string, len);
944		947
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)	948	if (len > cbor->max_size && cbor->max_size)
974	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",	949	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);	950	(unsigned long)len, (unsigned long)cbor->max_size);
976	}
977
978	sv_utf8_downgrade (string, 0);
979		951
980	dec.cbor = *cbor;	952	dec.cbor = *cbor;
981	dec.cur = (U8 *)SvPVX (string);	953	dec.cur = (U8 *)data;
982	dec.end = (U8 *)SvEND (string);	954	dec.end = (U8 *)data + len;
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		955
989	sv = decode_sv (&dec);	956	sv = decode_sv (&dec);
990		957
991	if (offset_return)	958	if (offset_return)
992	*offset_return = dec.cur;	959	*offset_return = dec.cur;
…		…
996	dec.err = "garbage after CBOR object";	963	dec.err = "garbage after CBOR object";
997		964
998	if (dec.err)	965	if (dec.err)
999	{	966	{
1000	SvREFCNT_dec (sv);	967	SvREFCNT_dec (sv);
1001	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 )SvPVX (string), (int)(uint8_t)dec.cur);	968	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 )data, (int)(uint8_t)dec.cur);
1002	}	969	}
1003		970
1004	sv = sv_2mortal (sv);	971	sv = sv_2mortal (sv);
1005		972
1006	return sv;	973	return sv;
…		…
1012	MODULE = CBOR::XS PACKAGE = CBOR::XS	979	MODULE = CBOR::XS PACKAGE = CBOR::XS
1013		980
1014	BOOT:	981	BOOT:
1015	{	982	{
1016	cbor_stash = gv_stashpv ("CBOR::XS" , 1);	983	cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1017	cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1);	984	cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1018		985
1019	cbor_true = get_bool ("CBOR::XS::true");	986	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1020	cbor_false = get_bool ("CBOR::XS::false");	987	types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
		988
		989	types_true = get_bool ("Types::Serialiser::true" );
		990	types_false = get_bool ("Types::Serialiser::false");
		991	types_error = get_bool ("Types::Serialiser::error");
		992
		993	sv_cbor = newSVpv ("CBOR", 0);
		994	SvREADONLY_on (sv_cbor);
1021	}	995	}
1022		996
1023	PROTOTYPES: DISABLE	997	PROTOTYPES: DISABLE
1024		998
1025	void CLONE (...)	999	void CLONE (...)
1026	CODE:	1000	CODE:
1027	cbor_stash = 0;	1001	cbor_stash = 0;
		1002	cbor_tagged_stash = 0;
		1003	types_error_stash = 0;
1028	cbor_boolean_stash = 0;	1004	types_boolean_stash = 0;
1029		1005
1030	void new (char *klass)	1006	void new (char *klass)
1031	PPCODE:	1007	PPCODE:
1032	{	1008	{
1033	SV *pv = NEWSV (0, sizeof (CBOR));	1009	SV *pv = NEWSV (0, sizeof (CBOR));
…		…
1041		1017
1042	void shrink (CBOR *self, int enable = 1)	1018	void shrink (CBOR *self, int enable = 1)
1043	ALIAS:	1019	ALIAS:
1044	shrink = F_SHRINK	1020	shrink = F_SHRINK
1045	allow_unknown = F_ALLOW_UNKNOWN	1021	allow_unknown = F_ALLOW_UNKNOWN
		1022	allow_sharing = F_ALLOW_SHARING
		1023	dedup_keys = F_DEDUP_KEYS
		1024	dedup_strings = F_DEDUP_STRINGS
1046	PPCODE:	1025	PPCODE:
1047	{	1026	{
1048	if (enable)	1027	if (enable)
1049	self->flags \|= ix;	1028	self->flags \|= ix;
1050	else	1029	else
…		…
1055		1034
1056	void get_shrink (CBOR *self)	1035	void get_shrink (CBOR *self)
1057	ALIAS:	1036	ALIAS:
1058	get_shrink = F_SHRINK	1037	get_shrink = F_SHRINK
1059	get_allow_unknown = F_ALLOW_UNKNOWN	1038	get_allow_unknown = F_ALLOW_UNKNOWN
		1039	get_allow_sharing = F_ALLOW_SHARING
		1040	get_dedup_keys = F_DEDUP_KEYS
		1041	get_dedup_strings = F_DEDUP_STRINGS
1060	PPCODE:	1042	PPCODE:
1061	XPUSHs (boolSV (self->flags & ix));	1043	XPUSHs (boolSV (self->flags & ix));
1062		1044
1063	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)	1045	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1064	PPCODE:	1046	PPCODE:
…		…
1079	int get_max_size (CBOR *self)	1061	int get_max_size (CBOR *self)
1080	CODE:	1062	CODE:
1081	RETVAL = self->max_size;	1063	RETVAL = self->max_size;
1082	OUTPUT:	1064	OUTPUT:
1083	RETVAL	1065	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		1066
1120	void encode (CBOR self, SV scalar)	1067	void encode (CBOR self, SV scalar)
1121	PPCODE:	1068	PPCODE:
1122	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;	1069	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1123	XPUSHs (scalar);	1070	XPUSHs (scalar);
…		…
1136	EXTEND (SP, 2);	1083	EXTEND (SP, 2);
1137	PUSHs (sv);	1084	PUSHs (sv);
1138	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));	1085	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1139	}	1086	}
1140		1087
1141	void DESTROY (CBOR *self)
1142	CODE:
1143	SvREFCNT_dec (self->cb_sk_object);
1144	SvREFCNT_dec (self->cb_object);
1145
1146	PROTOTYPES: ENABLE	1088	PROTOTYPES: ENABLE
1147		1089
1148	void encode_cbor (SV *scalar)	1090	void encode_cbor (SV *scalar)
1149	PPCODE:	1091	PPCODE:
1150	{	1092	{

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing CBOR-XS/XS.xs (file contents): Revision 1.4 by root, Sat Oct 26 21:06:37 2013 UTC vs. Revision 1.18 by root, Wed Nov 20 01:09:46 2013 UTC

Diff Legend

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.4 by root, Sat Oct 26 21:06:37 2013 UTC vs.
Revision 1.18 by root, Wed Nov 20 01:09:46 2013 UTC