ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/CBOR-XS/XS.xs

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.3 by root, Sat Oct 26 11:08:34 2013 UTC vs.
Revision 1.72 by root, Thu Oct 21 01:14:58 2021 UTC

…		…
6	#include <string.h>	6	#include <string.h>
7	#include <stdlib.h>	7	#include <stdlib.h>
8	#include <stdio.h>	8	#include <stdio.h>
9	#include <limits.h>	9	#include <limits.h>
10	#include <float.h>	10	#include <float.h>
		11	#include <inttypes.h>
11		12
		13	#define ECB_NO_THREADS 1
12	#include "ecb.h"	14	#include "ecb.h"
13		15
		16	// compatibility with perl <5.18
		17	#ifndef HvNAMELEN_get
		18	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		19	#endif
		20	#ifndef HvNAMELEN
		21	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		22	#endif
		23	#ifndef HvNAMEUTF8
		24	# define HvNAMEUTF8(hv) 0
		25	#endif
		26	#ifndef SvREFCNT_inc_NN
		27	# define SvREFCNT_inc_NN(sv) SvREFCNT_inc (sv)
		28	#endif
		29	#ifndef SvREFCNT_dec_NN
		30	# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
		31	#endif
		32
		33	// perl's is_utf8_string interprets len=0 as "calculate len", but we want it to mean 0
		34	#define cbor_is_utf8_string(str,len) (!(len) || is_utf8_string ((str), (len)))
		35
		36	// known major and minor types
		37	enum cbor_type
		38	{
		39	MAJOR_SHIFT = 5,
		40	MINOR_MASK = 0x1f,
		41
		42	MAJOR_POS_INT = 0 << MAJOR_SHIFT,
		43	MAJOR_NEG_INT = 1 << MAJOR_SHIFT,
		44	MAJOR_BYTES = 2 << MAJOR_SHIFT,
		45	MAJOR_TEXT = 3 << MAJOR_SHIFT,
		46	MAJOR_ARRAY = 4 << MAJOR_SHIFT,
		47	MAJOR_MAP = 5 << MAJOR_SHIFT,
		48	MAJOR_TAG = 6 << MAJOR_SHIFT,
		49	MAJOR_MISC = 7 << MAJOR_SHIFT,
		50
		51	// INT/STRING/ARRAY/MAP subtypes
		52	LENGTH_EXT1 = 24,
		53	LENGTH_EXT2 = 25,
		54	LENGTH_EXT4 = 26,
		55	LENGTH_EXT8 = 27,
		56
		57	// SIMPLE types (effectively MISC subtypes)
		58	SIMPLE_FALSE = 20,
		59	SIMPLE_TRUE = 21,
		60	SIMPLE_NULL = 22,
		61	SIMPLE_UNDEF = 23,
		62
		63	// MISC subtype (unused)
		64	MISC_EXT1 = 24,
		65	MISC_FLOAT16 = 25,
		66	MISC_FLOAT32 = 26,
		67	MISC_FLOAT64 = 27,
		68
		69	// BYTES/TEXT/ARRAY/MAP
		70	MINOR_INDEF = 31,
		71	};
		72
		73	// known tags
		74	enum cbor_tag
		75	{
		76	// extensions
		77	CBOR_TAG_STRINGREF = 25, // http://cbor.schmorp.de/stringref
		78	CBOR_TAG_PERL_OBJECT = 26, // http://cbor.schmorp.de/perl-object
		79	CBOR_TAG_GENERIC_OBJECT = 27, // http://cbor.schmorp.de/generic-object
		80	CBOR_TAG_VALUE_SHAREABLE = 28, // http://cbor.schmorp.de/value-sharing
		81	CBOR_TAG_VALUE_SHAREDREF = 29, // http://cbor.schmorp.de/value-sharing
		82	CBOR_TAG_STRINGREF_NAMESPACE = 256, // http://cbor.schmorp.de/stringref
		83	CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
		84
		85	// rfc7049
		86	CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
		87	CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
		88	CBOR_TAG_POS_BIGNUM = 2, // byte string
		89	CBOR_TAG_NEG_BIGNUM = 3, // byte string
		90	CBOR_TAG_DECIMAL = 4, // decimal fraction, array
		91	CBOR_TAG_BIGFLOAT = 5, // array
		92
		93	CBOR_TAG_CONV_B64U = 21, // base64url, any
		94	CBOR_TAG_CONV_B64 = 22, // base64, any
		95	CBOR_TAG_CONV_HEX = 23, // base16, any
		96	CBOR_TAG_CBOR = 24, // embedded cbor, byte string
		97
		98	CBOR_TAG_URI = 32, // URI rfc3986, utf-8
		99	CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
		100	CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
		101	CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
		102	CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
		103
		104	CBOR_TAG_MAGIC = 55799, // self-describe cbor
		105	};
		106
		107	// known forced types, also hardcoded in CBOR.pm
		108	enum
		109	{
		110	AS_CBOR = 0,
		111	AS_INT = 1,
		112	AS_BYTES = 2,
		113	AS_TEXT = 3,
		114	AS_FLOAT16 = 4,
		115	AS_FLOAT32 = 5,
		116	AS_FLOAT64 = 6,
		117	AS_MAP = 7,
		118	// possibly future enhancements: (generic) float, (generic) string
		119	};
		120
14	#define F_SHRINK 0x00000200UL	121	#define F_SHRINK 0x00000001UL
15	#define F_ALLOW_UNKNOWN 0x00002000UL	122	#define F_ALLOW_UNKNOWN 0x00000002UL
		123	#define F_ALLOW_SHARING 0x00000004UL
		124	#define F_ALLOW_CYCLES 0x00000008UL
		125	#define F_FORBID_OBJECTS 0x00000010UL
		126	#define F_PACK_STRINGS 0x00000020UL
		127	#define F_TEXT_KEYS 0x00000040UL
		128	#define F_TEXT_STRINGS 0x00000080UL
		129	#define F_VALIDATE_UTF8 0x00000100UL
16		130
17	#define INIT_SIZE 32 // initial scalar size to be allocated	131	#define INIT_SIZE 32 // initial scalar size to be allocated
18		132
19	#define SB do {	133	#define SB do {
20	#define SE } while (0)	134	#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		135
33	#define IN_RANGE_INC(type,val,beg,end) \	136	#define IN_RANGE_INC(type,val,beg,end) \
34	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	137	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
35	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	138	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
36		139
…		…
42	#else	145	#else
43	# define CBOR_SLOW 0	146	# define CBOR_SLOW 0
44	# define CBOR_STASH cbor_stash	147	# define CBOR_STASH cbor_stash
45	#endif	148	#endif
46		149
47	static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS::	150	static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
48	static SV *cbor_true, *cbor_false;	151	static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
49		152
50	typedef struct {	153	typedef struct {
51	U32 flags;	154	U32 flags;
52	U32 max_depth;	155	U32 max_depth;
53	STRLEN max_size;	156	STRLEN max_size;
		157	SV *filter;
54		158
55	SV *cb_object;	159	// for the incremental parser
56	HV *cb_sk_object;	160	STRLEN incr_pos; // the current offset into the text
		161	STRLEN incr_need; // minimum bytes needed to decode
		162	AV *incr_count; // for every nesting level, the number of outstanding values, or -1 for indef.
57	} CBOR;	163	} CBOR;
58		164
59	INLINE void	165	ecb_inline void
60	cbor_init (CBOR *cbor)	166	cbor_init (CBOR *cbor)
61	{	167	{
62	Zero (cbor, 1, CBOR);	168	Zero (cbor, 1, CBOR);
63	cbor->max_depth = 512;	169	cbor->max_depth = 512;
64	}	170	}
65		171
		172	ecb_inline void
		173	cbor_free (CBOR *cbor)
		174	{
		175	SvREFCNT_dec (cbor->filter);
		176	SvREFCNT_dec (cbor->incr_count);
		177	}
		178
66	/////////////////////////////////////////////////////////////////////////////	179	/////////////////////////////////////////////////////////////////////////////
67	// utility functions	180	// utility functions
68		181
69	INLINE SV *	182	ecb_inline SV *
70	get_bool (const char *name)	183	get_bool (const char *name)
71	{	184	{
72	SV *sv = get_sv (name, 1);	185	SV *sv = get_sv (name, 1);
73		186
74	SvREADONLY_on (sv);	187	SvREADONLY_on (sv);
75	SvREADONLY_on (SvRV (sv));	188	SvREADONLY_on (SvRV (sv));
76		189
77	return sv;	190	return sv;
78	}	191	}
79		192
80	INLINE void	193	ecb_inline void
81	shrink (SV *sv)	194	shrink (SV *sv)
82	{	195	{
83	sv_utf8_downgrade (sv, 1);	196	sv_utf8_downgrade (sv, 1);
84		197
85	if (SvLEN (sv) > SvCUR (sv) + 1)	198	if (SvLEN (sv) > SvCUR (sv) + 1)
…		…
90	SvPV_renew (sv, SvCUR (sv) + 1);	203	SvPV_renew (sv, SvCUR (sv) + 1);
91	#endif	204	#endif
92	}	205	}
93	}	206	}
94		207
95	/////////////////////////////////////////////////////////////////////////////	208	// minimum length of a string to be registered for stringref
96	// fp hell	209	ecb_inline STRLEN
97		210	minimum_string_length (UV idx)
98	//TODO	211	{
		212	return idx <= 23 ? 3
		213	: idx <= 0xffU ? 4
		214	: idx <= 0xffffU ? 5
		215	: idx <= 0xffffffffU ? 7
		216	: 11;
		217	}
99		218
100	/////////////////////////////////////////////////////////////////////////////	219	/////////////////////////////////////////////////////////////////////////////
101	// encoder	220	// encoder
102		221
103	// structure used for encoding CBOR	222	// structure used for encoding CBOR
…		…
106	char *cur; // SvPVX (sv) + current output position	225	char *cur; // SvPVX (sv) + current output position
107	char *end; // SvEND (sv)	226	char *end; // SvEND (sv)
108	SV *sv; // result scalar	227	SV *sv; // result scalar
109	CBOR cbor;	228	CBOR cbor;
110	U32 depth; // recursion level	229	U32 depth; // recursion level
		230	HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
		231	UV stringref_idx;
		232	HV *shareable; // ptr => index, or 0
		233	UV shareable_idx;
111	} enc_t;	234	} enc_t;
112		235
113	INLINE void	236	ecb_inline void
114	need (enc_t *enc, STRLEN len)	237	need (enc_t *enc, STRLEN len)
115	{	238	{
116	if (expect_false (enc->cur + len >= enc->end))	239	if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
117	{	240	{
118	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);	241	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
119	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);	242	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
120	enc->cur = SvPVX (enc->sv) + cur;	243	enc->cur = SvPVX (enc->sv) + cur;
121	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	244	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
122	}	245	}
123	}	246	}
124		247
125	INLINE void	248	static void encode_sv (enc_t *enc, SV *sv);
		249
		250	ecb_inline void
126	encode_ch (enc_t *enc, char ch)	251	encode_ch (enc_t *enc, char ch)
127	{	252	{
128	need (enc, 1);	253	need (enc, 1);
129	*enc->cur++ = ch;	254	*enc->cur++ = ch;
130	}	255	}
131		256
		257	// used for tags, intregers, element counts and so on
132	static void	258	static void
133	encode_uint (enc_t *enc, int major, UV len)	259	encode_uint (enc_t *enc, int major, UV len)
134	{	260	{
135	need (enc, 9);	261	need (enc, 9);
136		262
137	if (len < 24)	263	if (ecb_expect_true (len < LENGTH_EXT1))
138	*enc->cur++ = major | len;	264	*enc->cur++ = major | len;
139	else if (len < 0x100)	265	else if (ecb_expect_true (len <= 0xffU))
140	{	266	{
141	*enc->cur++ = major | 24;	267	*enc->cur++ = major | LENGTH_EXT1;
142	*enc->cur++ = len;	268	*enc->cur++ = len;
143	}	269	}
144	else if (len < 0x10000)	270	else if (len <= 0xffffU)
145	{	271	{
146	*enc->cur++ = major | 25;	272	*enc->cur++ = major | LENGTH_EXT2;
147	*enc->cur++ = len >> 8;	273	*enc->cur++ = len >> 8;
148	*enc->cur++ = len;	274	*enc->cur++ = len;
149	}	275	}
150	else if (len < 0x100000000)	276	else if (len <= 0xffffffffU)
151	{	277	{
152	*enc->cur++ = major | 26;	278	*enc->cur++ = major | LENGTH_EXT4;
153	*enc->cur++ = len >> 24;	279	*enc->cur++ = len >> 24;
154	*enc->cur++ = len >> 16;	280	*enc->cur++ = len >> 16;
155	*enc->cur++ = len >> 8;	281	*enc->cur++ = len >> 8;
156	*enc->cur++ = len;	282	*enc->cur++ = len;
157	}	283	}
158	else if (len)	284	else
159	{	285	{
160	*enc->cur++ = major | 27;	286	*enc->cur++ = major | LENGTH_EXT8;
161	*enc->cur++ = len >> 56;	287	*enc->cur++ = len >> 56;
162	*enc->cur++ = len >> 48;	288	*enc->cur++ = len >> 48;
163	*enc->cur++ = len >> 40;	289	*enc->cur++ = len >> 40;
164	*enc->cur++ = len >> 32;	290	*enc->cur++ = len >> 32;
165	*enc->cur++ = len >> 24;	291	*enc->cur++ = len >> 24;
…		…
167	*enc->cur++ = len >> 8;	293	*enc->cur++ = len >> 8;
168	*enc->cur++ = len;	294	*enc->cur++ = len;
169	}	295	}
170	}	296	}
171		297
172	static void	298	// encodes a perl value into a CBOR integer
		299	ecb_inline void
		300	encode_int (enc_t *enc, SV *sv)
		301	{
		302	if (SvIsUV (sv))
		303	encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
		304	else if (SvIVX (sv) >= 0)
		305	encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
		306	else
		307	encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
		308	}
		309
		310	ecb_inline void
		311	encode_tag (enc_t *enc, UV tag)
		312	{
		313	encode_uint (enc, MAJOR_TAG, tag);
		314	}
		315
		316	// exceptional (hopefully) slow path for byte strings that need to be utf8-encoded
		317	ecb_noinline static void
173	encode_str (enc_t *enc, int utf8, char *str, STRLEN len)	318	encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
174	{	319	{
175	encode_uint (enc, utf8 ? 0x60 : 0x40, len);	320	STRLEN ulen = len;
		321	U8 *p, *pend = (U8 *)str + len;
		322
		323	for (p = (U8 *)str; p < pend; ++p)
		324	ulen += *p >> 7; // count set high bits
		325
		326	encode_uint (enc, MAJOR_TEXT, ulen);
		327
		328	need (enc, ulen);
		329	for (p = (U8 *)str; p < pend; ++p)
		330	if (*p < 0x80)
		331	*enc->cur++ = *p;
		332	else
		333	{
		334	*enc->cur++ = 0xc0 + (*p >> 6);
		335	*enc->cur++ = 0x80 + (*p & 63);
		336	}
		337	}
		338
		339	ecb_inline void
		340	encode_str (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		341	{
		342	if (ecb_expect_false (upgrade_utf8))
		343	if (!utf8)
		344	{
		345	encode_str_utf8 (enc, utf8, str, len);
		346	return;
		347	}
		348
		349	encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len);
176	need (enc, len);	350	need (enc, len);
177	memcpy (enc->cur, str, len);	351	memcpy (enc->cur, str, len);
178	enc->cur += len;	352	enc->cur += len;
179	}	353	}
180		354
181	static void encode_sv (enc_t *enc, SV *sv);	355	ecb_inline void
		356	encode_strref (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		357	{
		358	if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS))
		359	{
		360	SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
		361
		362	if (SvOK (*svp))
		363	{
		364	// already registered, use stringref
		365	encode_tag (enc, CBOR_TAG_STRINGREF);
		366	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		367	return;
		368	}
		369	else if (len >= minimum_string_length (enc->stringref_idx))
		370	{
		371	// register only
		372	sv_setuv (*svp, enc->stringref_idx);
		373	++enc->stringref_idx;
		374	}
		375	}
		376
		377	encode_str (enc, upgrade_utf8, utf8, str, len);
		378	}
		379
		380	ecb_inline void
		381	encode_float16 (enc_t *enc, NV nv)
		382	{
		383	need (enc, 1+2);
		384
		385	*enc->cur++ = MAJOR_MISC | MISC_FLOAT16;
		386
		387	uint16_t fp = ecb_float_to_binary16 (nv);
		388
		389	if (!ecb_big_endian ())
		390	fp = ecb_bswap16 (fp);
		391
		392	memcpy (enc->cur, &fp, 2);
		393	enc->cur += 2;
		394	}
		395
		396	ecb_inline void
		397	encode_float32 (enc_t *enc, NV nv)
		398	{
		399	need (enc, 1+4);
		400
		401	*enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
		402
		403	uint32_t fp = ecb_float_to_binary32 (nv);
		404
		405	if (!ecb_big_endian ())
		406	fp = ecb_bswap32 (fp);
		407
		408	memcpy (enc->cur, &fp, 4);
		409	enc->cur += 4;
		410	}
		411
		412	ecb_inline void
		413	encode_float64 (enc_t *enc, NV nv)
		414	{
		415	need (enc, 1+8);
		416
		417	*enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
		418
		419	uint64_t fp = ecb_double_to_binary64 (nv);
		420
		421	if (!ecb_big_endian ())
		422	fp = ecb_bswap64 (fp);
		423
		424	memcpy (enc->cur, &fp, 8);
		425	enc->cur += 8;
		426	}
		427
		428	ecb_inline void
		429	encode_bool (enc_t *enc, int istrue)
		430	{
		431	encode_ch (enc, istrue ? MAJOR_MISC | SIMPLE_TRUE : MAJOR_MISC | SIMPLE_FALSE);
		432	}
		433
		434	// encodes an arrayref containing key-value pairs as CBOR map
		435	ecb_inline void
		436	encode_array_as_map (enc_t *enc, SV *sv)
		437	{
		438	if (enc->depth >= enc->cbor.max_depth)
		439	croak (ERR_NESTING_EXCEEDED);
		440
		441	++enc->depth;
		442
		443	// as_map does error checking for us, but we re-check in case
		444	// things have changed.
		445
		446	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		447	croak ("CBOR::XS::as_map requires an array reference (did you change the array after calling as_map?)");
		448
		449	AV *av = (AV *)SvRV (sv);
		450	int i, len = av_len (av);
		451
		452	if (!(len & 1))
		453	croak ("CBOR::XS::as_map requires an even number of elements (did you change the array after calling as_map?)");
		454
		455	encode_uint (enc, MAJOR_MAP, (len + 1) >> 1);
		456
		457	for (i = 0; i <= len; ++i)
		458	{
		459	SV **svp = av_fetch (av, i, 0);
		460	encode_sv (enc, svp ? *svp : &PL_sv_undef);
		461	}
		462
		463	--enc->depth;
		464	}
		465
		466	ecb_inline void
		467	encode_forced (enc_t *enc, UV type, SV *sv)
		468	{
		469	switch (type)
		470	{
		471	case AS_CBOR:
		472	{
		473	STRLEN len;
		474	char *str = SvPVbyte (sv, len);
		475
		476	need (enc, len);
		477	memcpy (enc->cur, str, len);
		478	enc->cur += len;
		479	}
		480	break;
		481
		482	case AS_BYTES:
		483	{
		484	STRLEN len;
		485	char *str = SvPVbyte (sv, len);
		486	encode_strref (enc, 0, 0, str, len);
		487	}
		488	break;
		489
		490	case AS_TEXT:
		491	{
		492	STRLEN len;
		493	char *str = SvPVutf8 (sv, len);
		494	encode_strref (enc, 1, 1, str, len);
		495	}
		496	break;
		497
		498	case AS_INT: encode_int (enc, sv); break;
		499
		500	case AS_FLOAT16: encode_float16 (enc, SvNV (sv)); break;
		501	case AS_FLOAT32: encode_float32 (enc, SvNV (sv)); break;
		502	case AS_FLOAT64: encode_float64 (enc, SvNV (sv)); break;
		503
		504	case AS_MAP: encode_array_as_map (enc, sv); break;
		505
		506	default:
		507	croak ("encountered malformed CBOR::XS::Tagged object");
		508	}
		509	}
182		510
183	static void	511	static void
184	encode_av (enc_t *enc, AV *av)	512	encode_av (enc_t *enc, AV *av)
185	{	513	{
186	int i, len = av_len (av);	514	int i, len = av_len (av);
…		…
188	if (enc->depth >= enc->cbor.max_depth)	516	if (enc->depth >= enc->cbor.max_depth)
189	croak (ERR_NESTING_EXCEEDED);	517	croak (ERR_NESTING_EXCEEDED);
190		518
191	++enc->depth;	519	++enc->depth;
192		520
193	encode_uint (enc, 0x80, len + 1);	521	encode_uint (enc, MAJOR_ARRAY, len + 1);
194		522
		523	if (ecb_expect_false (SvMAGICAL (av)))
195	for (i = 0; i <= len; ++i)	524	for (i = 0; i <= len; ++i)
196	{	525	{
197	SV **svp = av_fetch (av, i, 0);	526	SV **svp = av_fetch (av, i, 0);
198	encode_sv (enc, svp ? *svp : &PL_sv_undef);	527	encode_sv (enc, svp ? *svp : &PL_sv_undef);
199	}	528	}
		529	else
		530	for (i = 0; i <= len; ++i)
		531	{
		532	SV *sv = AvARRAY (av)[i];
		533	encode_sv (enc, sv ? sv : &PL_sv_undef);
		534	}
200		535
201	--enc->depth;	536	--enc->depth;
202	}	537	}
203		538
204	static void	539	static void
…		…
212	++enc->depth;	547	++enc->depth;
213		548
214	int pairs = hv_iterinit (hv);	549	int pairs = hv_iterinit (hv);
215	int mg = SvMAGICAL (hv);	550	int mg = SvMAGICAL (hv);
216		551
217	if (mg)	552	if (ecb_expect_false (mg))
218	encode_ch (enc, 0xa0 | 31);	553	encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
219	else	554	else
220	encode_uint (enc, 0xa0, pairs);	555	encode_uint (enc, MAJOR_MAP, pairs);
221		556
222	while ((he = hv_iternext (hv)))	557	while ((he = hv_iternext (hv)))
223	{	558	{
224	if (HeKLEN (he) == HEf_SVKEY)	559	if (HeKLEN (he) == HEf_SVKEY)
225	encode_sv (enc, HeSVKEY (he));	560	encode_sv (enc, HeSVKEY (he));
226	else	561	else
227	encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));	562	encode_strref (enc, enc->cbor.flags & (F_TEXT_KEYS | F_TEXT_STRINGS), HeKUTF8 (he), HeKEY (he), HeKLEN (he));
228		563
229	encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));	564	encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
230	}	565	}
231		566
232	if (mg)	567	if (ecb_expect_false (mg))
233	encode_ch (enc, 0xe0 | 31);	568	encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
234		569
235	--enc->depth;	570	--enc->depth;
236	}	571	}
237		572
238	// encode objects, arrays and special \0=false and \1=true values.	573	// encode objects, arrays and special \0=false and \1=true values.
239	static void	574	static void
240	encode_rv (enc_t *enc, SV *sv)	575	encode_rv (enc_t *enc, SV *sv)
241	{	576	{
242	svtype svt;
243
244	SvGETMAGIC (sv);	577	SvGETMAGIC (sv);
		578
245	svt = SvTYPE (sv);	579	svtype svt = SvTYPE (sv);
246		580
247	if (expect_false (SvOBJECT (sv)))	581	if (ecb_expect_false (SvOBJECT (sv)))
248	{	582	{
249	HV *stash = !CBOR_SLOW || cbor_boolean_stash	583	HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
250	? cbor_boolean_stash	584	? types_boolean_stash
		585	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		586	HV *error_stash = !CBOR_SLOW || types_error_stash
		587	? types_error_stash
		588	: gv_stashpv ("Types::Serialiser::Error", 1);
		589	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		590	? cbor_tagged_stash
251	: gv_stashpv ("CBOR::XS::Boolean", 1);	591	: gv_stashpv ("CBOR::XS::Tagged" , 1);
252		592
253	if (SvSTASH (sv) == stash)	593	HV *stash = SvSTASH (sv);
254	encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);	594
		595	if (stash == boolean_stash)
		596	{
		597	encode_bool (enc, SvIV (sv));
		598	return;
		599	}
		600	else if (stash == error_stash)
		601	{
		602	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
		603	return;
		604	}
		605	else if (stash == tagged_stash)
		606	{
		607	if (svt != SVt_PVAV)
		608	croak ("encountered CBOR::XS::Tagged object that isn't an array");
		609
		610	switch (av_len ((AV *)sv))
		611	{
		612	case 2-1:
		613	// actually a tagged value
		614	encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1)));
		615	encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
		616	break;
		617
		618	case 3-1:
		619	// a forced type [value, type, undef]
		620	encode_forced (enc, SvUV (*av_fetch ((AV *)sv, 1, 1)), *av_fetch ((AV *)sv, 0, 1));
		621	break;
		622
		623	default:
		624	croak ("encountered malformed CBOR::XS::Tagged object");
		625	}
		626
		627	return;
		628	}
		629	}
		630
		631	if (ecb_expect_false (SvREFCNT (sv) > 1)
		632	&& ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING))
		633	{
		634	if (ecb_expect_false (!enc->shareable))
		635	enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
		636
		637	SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
		638
		639	if (SvOK (*svp))
		640	{
		641	encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
		642	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		643	return;
		644	}
255	else	645	else
256	{	646	{
257	#if 0 //TODO	647	sv_setuv (*svp, enc->shareable_idx);
258	if (enc->cbor.flags & F_CONV_BLESSED)	648	++enc->shareable_idx;
		649	encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
		650	}
		651	}
		652
		653	if (ecb_expect_false (SvOBJECT (sv)))
		654	{
		655	HV *stash = SvSTASH (sv);
		656	GV *method;
		657
		658	if (enc->cbor.flags & F_FORBID_OBJECTS)
		659	croak ("encountered object '%s', but forbid_objects is enabled",
		660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		661	else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
259	{	662	{
		663	dSP;
		664
		665	ENTER; SAVETMPS;
		666	PUSHMARK (SP);
260	// we re-bless the reference to get overload and other niceties right	667	// 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)));	668	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
269		669
270	// calling with G_SCALAR ensures that we always get a 1 return value
271	PUTBACK;	670	PUTBACK;
		671	// G_SCALAR ensures that return value is 1
272	call_sv ((SV *)GvCV (to_cbor), G_SCALAR);	672	call_sv ((SV *)GvCV (method), G_SCALAR);
273	SPAGAIN;	673	SPAGAIN;
274		674
275	// catch this surprisingly common error	675	// catch this surprisingly common error
276	if (SvROK (TOPs) && SvRV (TOPs) == sv)	676	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)));	677	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
278		678
279	sv = POPs;
280	PUTBACK;
281
282	encode_sv (enc, sv);	679	encode_sv (enc, POPs);
283		680
		681	PUTBACK;
		682
284	FREETMPS; LEAVE;	683	FREETMPS; LEAVE;
285	}	684	}
286	else if (enc->cbor.flags & F_ALLOW_BLESSED)	685	else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
287	encode_str (enc, "null", 4, 0);	686	{
288	else	687	dSP;
289	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it",	688
290	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	689	ENTER; SAVETMPS;
		690	PUSHMARK (SP);
		691	EXTEND (SP, 2);
		692	// we re-bless the reference to get overload and other niceties right
		693	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		694	PUSHs (sv_cbor);
		695
		696	PUTBACK;
		697	int count = call_sv ((SV *)GvCV (method), G_ARRAY);
		698	SPAGAIN;
		699
		700	// catch this surprisingly common error
		701	if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
		702	croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
		703
		704	encode_tag (enc, CBOR_TAG_PERL_OBJECT);
		705	encode_uint (enc, MAJOR_ARRAY, count + 1);
		706	encode_strref (enc, 0, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
		707
		708	{
		709	int i;
		710
		711	for (i = 0; i < count; ++i)
		712	encode_sv (enc, SP[i + 1 - count]);
		713
		714	SP -= count;
291	}	715	}
292	else if (enc->cbor.flags & F_ALLOW_BLESSED)	716
293	encode_str (enc, "null", 4, 0);	717	PUTBACK;
		718
		719	FREETMPS; LEAVE;
		720	}
294	else	721	else
295	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",	722	croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
296	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	723	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
297	#endif
298	}
299	}	724	}
300	else if (svt == SVt_PVHV)	725	else if (svt == SVt_PVHV)
301	encode_hv (enc, (HV *)sv);	726	encode_hv (enc, (HV *)sv);
302	else if (svt == SVt_PVAV)	727	else if (svt == SVt_PVAV)
303	encode_av (enc, (AV *)sv);	728	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	729	else
322	croak ("encountered %s, but CBOR can only represent references to arrays or hashes",	730	{
323	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	731	encode_tag (enc, CBOR_TAG_INDIRECTION);
		732	encode_sv (enc, sv);
		733	}
324	}	734	}
325		735
326	static void	736	static void
327	encode_nv (enc_t *enc, SV *sv)	737	encode_nv (enc_t *enc, SV *sv)
328	{	738	{
329	double nv = SvNVX (sv);	739	double nv = SvNVX (sv);
330		740
331	need (enc, 9);	741	need (enc, 9);
332		742
333	if (expect_false (nv == (U32)nv))	743	if (ecb_expect_false (nv == (NV)(U32)nv))
334	encode_uint (enc, 0x00, (U32)nv);	744	encode_uint (enc, MAJOR_POS_INT, (U32)nv);
335	//TODO: maybe I32?	745	//TODO: maybe I32?
336	else if (expect_false (nv == (float)nv))	746	else if (ecb_expect_false (nv == (float)nv))
337	{	747	encode_float32 (enc, nv);
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	748	else
349	{	749	encode_float64 (enc, nv);
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	}	750	}
361		751
362	static void	752	static void
363	encode_sv (enc_t *enc, SV *sv)	753	encode_sv (enc_t *enc, SV *sv)
364	{	754	{
…		…
366		756
367	if (SvPOKp (sv))	757	if (SvPOKp (sv))
368	{	758	{
369	STRLEN len;	759	STRLEN len;
370	char *str = SvPV (sv, len);	760	char *str = SvPV (sv, len);
371	encode_str (enc, SvUTF8 (sv), str, len);	761	encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
372	}	762	}
373	else if (SvNOKp (sv))	763	else if (SvNOKp (sv))
374	encode_nv (enc, sv);	764	encode_nv (enc, sv);
375	else if (SvIOKp (sv))	765	else if (SvIOKp (sv))
376	{	766	encode_int (enc, sv);
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))	767	else if (SvROK (sv))
385	encode_rv (enc, SvRV (sv));	768	encode_rv (enc, SvRV (sv));
386	else if (!SvOK (sv))	769	else if (!SvOK (sv))
387	encode_ch (enc, 0xe0 | 22);	770	encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
388	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)	771	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
389	encode_ch (enc, 0xe0 | 23);	772	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
390	else	773	else
391	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",	774	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",
392	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));	775	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
393	}	776	}
394		777
395	static SV *	778	static SV *
396	encode_cbor (SV *scalar, CBOR *cbor)	779	encode_cbor (SV *scalar, CBOR *cbor)
397	{	780	{
398	enc_t enc;	781	enc_t enc = { 0 };
399		782
400	enc.cbor = *cbor;	783	enc.cbor = *cbor;
401	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	784	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
402	enc.cur = SvPVX (enc.sv);	785	enc.cur = SvPVX (enc.sv);
403	enc.end = SvEND (enc.sv);	786	enc.end = SvEND (enc.sv);
404	enc.depth = 0;
405		787
406	SvPOK_only (enc.sv);	788	SvPOK_only (enc.sv);
		789
		790	if (cbor->flags & F_PACK_STRINGS)
		791	{
		792	encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
		793	enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
		794	enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
		795	}
		796
407	encode_sv (&enc, scalar);	797	encode_sv (&enc, scalar);
408		798
409	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	799	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
410	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	800	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
411		801
…		…
425	U8 *end; // end of input string	815	U8 *end; // end of input string
426	const char *err; // parse error, if != 0	816	const char *err; // parse error, if != 0
427	CBOR cbor;	817	CBOR cbor;
428	U32 depth; // recursion depth	818	U32 depth; // recursion depth
429	U32 maxdepth; // recursion depth limit	819	U32 maxdepth; // recursion depth limit
		820	AV *shareable;
		821	AV *stringref;
		822	SV *decode_tagged;
		823	SV *err_sv; // optional sv for error, needs to be freed
430	} dec_t;	824	} dec_t;
431		825
432	#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE	826	// set dec->err to ERRSV
		827	ecb_cold static void
		828	err_errsv (dec_t *dec)
		829	{
		830	if (!dec->err)
		831	{
		832	dec->err_sv = newSVsv (ERRSV);
433		833
434	#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data");	834	// chop off the trailing \n
		835	SvCUR_set (dec->err_sv, SvCUR (dec->err_sv) - 1);
		836	*SvEND (dec->err_sv) = 0;
435		837
		838	dec->err = SvPVutf8_nolen (dec->err_sv);
		839	}
		840	}
		841
		842	// the following functions are used to reduce code size and help the compiler to optimise
		843	ecb_cold static void
		844	err_set (dec_t *dec, const char *reason)
		845	{
		846	if (!dec->err)
		847	dec->err = reason;
		848	}
		849
		850	ecb_cold static void
		851	err_unexpected_end (dec_t *dec)
		852	{
		853	err_set (dec, "unexpected end of CBOR data");
		854	}
		855
		856	#define ERR_DO(do) SB do; goto fail; SE
		857	#define ERR(reason) ERR_DO (err_set (dec, reason))
		858	#define ERR_ERRSV ERR_DO (err_errsv (dec))
		859
		860	#define WANT(len) if (ecb_expect_false ((uintptr_t)(dec->end - dec->cur) < (STRLEN)len)) ERR_DO (err_unexpected_end (dec))
		861
436	#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)	862	#define DEC_INC_DEPTH if (ecb_expect_false (++dec->depth > dec->cbor.max_depth)) ERR (ERR_NESTING_EXCEEDED)
437	#define DEC_DEC_DEPTH --dec->depth	863	#define DEC_DEC_DEPTH --dec->depth
438		864
439	static UV	865	static UV
440	decode_uint (dec_t *dec)	866	decode_uint (dec_t *dec)
441	{	867	{
442	switch (*dec->cur & 31)	868	U8 m = *dec->cur & MINOR_MASK;
443	{	869	++dec->cur;
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		870
449	case 24:	871	if (ecb_expect_true (m < LENGTH_EXT1))
		872	return m;
		873	else if (ecb_expect_true (m == LENGTH_EXT1))
		874	{
450	WANT (2);	875	WANT (1);
451	dec->cur += 2;	876	dec->cur += 1;
452	return dec->cur[-1];	877	return dec->cur[-1];
453		878	}
454	case 25:	879	else if (ecb_expect_true (m == LENGTH_EXT2))
		880	{
455	WANT (3);	881	WANT (2);
456	dec->cur += 3;	882	dec->cur += 2;
457	return (((UV)dec->cur[-2]) << 8)	883	return (((UV)dec->cur[-2]) << 8)
458	| ((UV)dec->cur[-1]);	884	| ((UV)dec->cur[-1]);
459		885	}
460	case 26:	886	else if (ecb_expect_true (m == LENGTH_EXT4))
		887	{
461	WANT (5);	888	WANT (4);
462	dec->cur += 5;	889	dec->cur += 4;
463	return (((UV)dec->cur[-4]) << 24)	890	return (((UV)dec->cur[-4]) << 24)
464	| (((UV)dec->cur[-3]) << 16)	891	| (((UV)dec->cur[-3]) << 16)
465	| (((UV)dec->cur[-2]) << 8)	892	| (((UV)dec->cur[-2]) << 8)
466	| ((UV)dec->cur[-1]);	893	| ((UV)dec->cur[-1]);
467		894	}
468	case 27:	895	else if (ecb_expect_true (m == LENGTH_EXT8))
		896	{
469	WANT (9);	897	WANT (8);
470	dec->cur += 9;	898	dec->cur += 8;
		899
		900	return
		901	#if UVSIZE < 8
		902	0
		903	#else
471	return (((UV)dec->cur[-8]) << 56)	904	(((UV)dec->cur[-8]) << 56)
472	| (((UV)dec->cur[-7]) << 48)	905	| (((UV)dec->cur[-7]) << 48)
473	| (((UV)dec->cur[-6]) << 40)	906	| (((UV)dec->cur[-6]) << 40)
474	| (((UV)dec->cur[-5]) << 32)	907	| (((UV)dec->cur[-5]) << 32)
		908	#endif
475	| (((UV)dec->cur[-4]) << 24)	909	| (((UV)dec->cur[-4]) << 24)
476	| (((UV)dec->cur[-3]) << 16)	910	| (((UV)dec->cur[-3]) << 16)
477	| (((UV)dec->cur[-2]) << 8)	911	| (((UV)dec->cur[-2]) << 8)
478	| ((UV)dec->cur[-1]);	912	| ((UV)dec->cur[-1]);
479		913	}
480	default:	914	else
481	ERR ("corrupted CBOR data (unsupported integer minor encoding)");	915	ERR ("corrupted CBOR data (unsupported integer minor encoding)");
482	}
483		916
484	fail:	917	fail:
485	return 0;	918	return 0;
486	}	919	}
487		920
…		…
492	{	925	{
493	AV *av = newAV ();	926	AV *av = newAV ();
494		927
495	DEC_INC_DEPTH;	928	DEC_INC_DEPTH;
496		929
497	if ((*dec->cur & 31) == 31)	930	if (*dec->cur == (MAJOR_ARRAY | MINOR_INDEF))
498	{	931	{
499	++dec->cur;	932	++dec->cur;
500		933
501	for (;;)	934	for (;;)
502	{	935	{
503	WANT (1);	936	WANT (1);
504		937
505	if (*dec->cur == (0xe0 | 31))	938	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
506	{	939	{
507	++dec->cur;	940	++dec->cur;
508	break;	941	break;
509	}	942	}
510		943
511	av_push (av, decode_sv (dec));	944	av_push (av, decode_sv (dec));
512	}	945	}
513	}	946	}
514	else	947	else
515	{	948	{
516	int i, len = decode_uint (dec);	949	UV i, len = decode_uint (dec);
517		950
		951	WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays
518	av_fill (av, len - 1);	952	av_fill (av, len - 1);
519		953
520	for (i = 0; i < len; ++i)	954	for (i = 0; i < len; ++i)
521	AvARRAY (av)[i] = decode_sv (dec);	955	AvARRAY (av)[i] = decode_sv (dec);
522	}	956	}
523		957
524	DEC_DEC_DEPTH;	958	DEC_DEC_DEPTH;
525	return newRV_noinc ((SV *)av);	959	return newRV_noinc ((SV *)av);
526		960
527	fail:	961	fail:
528	SvREFCNT_dec (av);	962	SvREFCNT_dec_NN (av);
529	DEC_DEC_DEPTH;	963	DEC_DEC_DEPTH;
530	return &PL_sv_undef;	964	return &PL_sv_undef;
531	}	965	}
532		966
		967	static void
		968	decode_he (dec_t *dec, HV *hv)
		969	{
		970	// for speed reasons, we specialcase single-string
		971	// byte or utf-8 strings as keys, but only when !stringref
		972
		973	if (ecb_expect_true (!dec->stringref))
		974	if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8))
		975	{
		976	STRLEN len = decode_uint (dec);
		977	char *key = (char *)dec->cur;
		978
		979	WANT (len);
		980	dec->cur += len;
		981
		982	hv_store (hv, key, len, decode_sv (dec), 0);
		983
		984	return;
		985	}
		986	else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8))
		987	{
		988	STRLEN len = decode_uint (dec);
		989	char *key = (char *)dec->cur;
		990	printf ("len %d\n", len);//D
		991
		992	WANT (len);
		993	dec->cur += len;
		994
		995	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		996	if (!cbor_is_utf8_string ((U8 *)key, len))
		997	ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
		998
		999	hv_store (hv, key, -len, decode_sv (dec), 0);
		1000
		1001	return;
		1002	}
		1003
		1004	SV *k = decode_sv (dec);
		1005	SV *v = decode_sv (dec);
		1006
		1007	// we leak memory if uncaught exceptions are thrown by random magical
		1008	// methods, and this is hopefully the only place where it can happen,
		1009	// so if there is a chance of an exception, take the very slow path.
		1010	// since catching exceptions is "undocumented/internal/forbidden" by
		1011	// the new p5p powers, we need to call out to a perl function :/
		1012	if (ecb_expect_false (SvAMAGIC (k)))
		1013	{
		1014	dSP;
		1015
		1016	ENTER; SAVETMPS;
		1017	PUSHMARK (SP);
		1018	EXTEND (SP, 3);
		1019	PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
		1020	PUSHs (sv_2mortal (k));
		1021	PUSHs (sv_2mortal (v));
		1022
		1023	PUTBACK;
		1024	call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
		1025	SPAGAIN;
		1026
		1027	FREETMPS; LEAVE;
		1028
		1029	if (SvTRUE (ERRSV))
		1030	ERR_ERRSV;
		1031
		1032	return;
		1033	}
		1034
		1035	hv_store_ent (hv, k, v, 0);
		1036	SvREFCNT_dec_NN (k);
		1037
		1038	fail:
		1039	;
		1040	}
		1041
533	static SV *	1042	static SV *
534	decode_hv (dec_t *dec)	1043	decode_hv (dec_t *dec)
535	{	1044	{
536	HV *hv = newHV ();	1045	HV *hv = newHV ();
537		1046
538	DEC_INC_DEPTH;	1047	DEC_INC_DEPTH;
539		1048
540	if ((*dec->cur & 31) == 31)	1049	if (*dec->cur == (MAJOR_MAP | MINOR_INDEF))
541	{	1050	{
542	++dec->cur;	1051	++dec->cur;
543		1052
544	for (;;)	1053	for (;;)
545	{	1054	{
546	WANT (1);	1055	WANT (1);
547		1056
548	if (*dec->cur == (0xe0 | 31))	1057	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
549	{	1058	{
550	++dec->cur;	1059	++dec->cur;
551	break;	1060	break;
552	}	1061	}
553		1062
554	SV *k = decode_sv (dec);	1063	decode_he (dec, hv);
555	SV *v = decode_sv (dec);
556
557	hv_store_ent (hv, k, v, 0);
558	}	1064	}
559	}	1065	}
560	else	1066	else
561	{	1067	{
562	int len = decode_uint (dec);	1068	UV pairs = decode_uint (dec);
563		1069
		1070	WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
		1071
564	while (len--)	1072	while (pairs--)
565	{	1073	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	}	1074	}
572		1075
573	DEC_DEC_DEPTH;	1076	DEC_DEC_DEPTH;
574	return newRV_noinc ((SV *)hv);	1077	return newRV_noinc ((SV *)hv);
575		1078
576	#if 0	1079	fail:
577	SV *sv;	1080	SvREFCNT_dec_NN (hv);
578	HV *hv = newHV ();
579
580	DEC_INC_DEPTH;	1081	DEC_DEC_DEPTH;
581	decode_ws (dec);	1082	return &PL_sv_undef;
		1083	}
582		1084
583	for (;;)	1085	static SV *
584	{	1086	decode_str (dec_t *dec, int utf8)
585	// heuristic: assume that	1087	{
586	// a) decode_str + hv_store_ent are abysmally slow.	1088	SV *sv = 0;
587	// b) most hash keys are short, simple ascii text.	1089
588	// => try to "fast-match" such strings to avoid	1090	if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
589	// the overhead of decode_str + hv_store_ent.
590	{	1091	{
591	SV *value;	1092	// indefinite length strings
592	char *p = dec->cur;	1093	++dec->cur;
593	char *e = p + 24; // only try up to 24 bytes
594		1094
		1095	U8 major = *dec->cur & MAJOR_MISC;
		1096
		1097	sv = newSVpvn ("", 0);
		1098
595	for (;;)	1099	for (;;)
596	{	1100	{
597	// the >= 0x80 is false on most architectures	1101	WANT (1);
598	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1102
		1103	if ((*dec->cur - major) > LENGTH_EXT8)
		1104	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF))
599	{	1105	{
600	// slow path, back up and use decode_str	1106	++dec->cur;
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;	1107	break;
619	}	1108	}
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	{	1109	else
685	dSP;	1110	ERR ("corrupted CBOR data (invalid chunks in indefinite length string)");
686	int count;
687		1111
688	ENTER; SAVETMPS; PUSHMARK (SP);	1112	STRLEN len = decode_uint (dec);
689	XPUSHs (HeVAL (he));
690	sv_2mortal (sv);
691		1113
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);	1114	WANT (len);
752		1115	sv_catpvn (sv, dec->cur, len);
753	if (*dec->cur == (0xe0 | 31))	1116	dec->cur += len;
754	{
755	++dec->cur;
756	break;
757	}
758
759	SV *sv2 = decode_sv (dec);
760	sv_catsv (sv, sv2);
761	}	1117	}
762	}	1118	}
763	else	1119	else
764	{	1120	{
765	STRLEN len = decode_uint (dec);	1121	STRLEN len = decode_uint (dec);
766		1122
767	WANT (len);	1123	WANT (len);
768	sv = newSVpvn (dec->cur, len);	1124	sv = newSVpvn (dec->cur, len);
769	dec->cur += len;	1125	dec->cur += len;
		1126
		1127	if (ecb_expect_false (dec->stringref)
		1128	&& SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
		1129	av_push (dec->stringref, SvREFCNT_inc_NN (sv));
770	}	1130	}
771		1131
772	if (utf8)	1132	if (utf8)
		1133	{
		1134	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		1135	if (!cbor_is_utf8_string (SvPVX (sv), SvCUR (sv)))
		1136	ERR ("corrupted CBOR data (invalid UTF-8 in text string)");
		1137
773	SvUTF8_on (sv);	1138	SvUTF8_on (sv);
		1139	}
774		1140
775	return sv;	1141	return sv;
776		1142
777	fail:	1143	fail:
		1144	SvREFCNT_dec (sv);
778	return &PL_sv_undef;	1145	return &PL_sv_undef;
779	}	1146	}
780		1147
781	static SV *	1148	static SV *
782	decode_tagged (dec_t *dec)	1149	decode_tagged (dec_t *dec)
783	{	1150	{
		1151	SV *sv = 0;
784	UV tag = decode_uint (dec);	1152	UV tag = decode_uint (dec);
		1153
		1154	WANT (1);
		1155
		1156	switch (tag)
		1157	{
		1158	case CBOR_TAG_MAGIC:
785	SV *sv = decode_sv (dec);	1159	sv = decode_sv (dec);
		1160	break;
786		1161
787	if (tag == 55799) // 2.4.5 Self-Describe CBOR	1162	case CBOR_TAG_INDIRECTION:
		1163	sv = newRV_noinc (decode_sv (dec));
		1164	break;
		1165
		1166	case CBOR_TAG_STRINGREF_NAMESPACE:
		1167	{
		1168	// do not use SAVETMPS/FREETMPS, as these will
		1169	// erase mortalised caches, e.g. "shareable"
		1170	ENTER;
		1171
		1172	SAVESPTR (dec->stringref);
		1173	dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
		1174
		1175	sv = decode_sv (dec);
		1176
		1177	LEAVE;
		1178	}
		1179	break;
		1180
		1181	case CBOR_TAG_STRINGREF:
		1182	{
		1183	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1184	ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
		1185
		1186	UV idx = decode_uint (dec);
		1187
		1188	if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
		1189	ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
		1190
		1191	sv = newSVsv (AvARRAY (dec->stringref)[idx]);
		1192	}
		1193	break;
		1194
		1195	case CBOR_TAG_VALUE_SHAREABLE:
		1196	{
		1197	if (ecb_expect_false (!dec->shareable))
		1198	dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
		1199
		1200	if (dec->cbor.flags & F_ALLOW_CYCLES)
		1201	{
		1202	sv = newSV (0);
		1203	av_push (dec->shareable, SvREFCNT_inc_NN (sv));
		1204
		1205	SV *osv = decode_sv (dec);
		1206	sv_setsv (sv, osv);
		1207	SvREFCNT_dec_NN (osv);
		1208	}
		1209	else
		1210	{
		1211	av_push (dec->shareable, &PL_sv_undef);
		1212	int idx = AvFILLp (dec->shareable);
		1213	sv = decode_sv (dec);
		1214	av_store (dec->shareable, idx, SvREFCNT_inc_NN (sv));
		1215	}
		1216	}
		1217	break;
		1218
		1219	case CBOR_TAG_VALUE_SHAREDREF:
		1220	{
		1221	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1222	ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
		1223
		1224	UV idx = decode_uint (dec);
		1225
		1226	if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
		1227	ERR ("corrupted CBOR data (sharedref index out of bounds)");
		1228
		1229	sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
		1230
		1231	if (sv == &PL_sv_undef)
		1232	ERR ("cyclic CBOR data structure found, but allow_cycles is not enabled");
		1233	}
		1234	break;
		1235
		1236	case CBOR_TAG_PERL_OBJECT:
		1237	{
		1238	if (dec->cbor.flags & F_FORBID_OBJECTS)
		1239	goto filter;
		1240
		1241	sv = decode_sv (dec);
		1242
		1243	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		1244	ERR ("corrupted CBOR data (non-array perl object)");
		1245
		1246	AV *av = (AV *)SvRV (sv);
		1247	int len = av_len (av) + 1;
		1248	HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
		1249
		1250	if (!stash)
		1251	ERR ("cannot decode perl-object (package does not exist)");
		1252
		1253	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1254
		1255	if (!method)
		1256	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1257
		1258	dSP;
		1259
		1260	ENTER; SAVETMPS;
		1261	PUSHMARK (SP);
		1262	EXTEND (SP, len + 1);
		1263	// we re-bless the reference to get overload and other niceties right
		1264	PUSHs (*av_fetch (av, 0, 1));
		1265	PUSHs (sv_cbor);
		1266
		1267	int i;
		1268
		1269	for (i = 1; i < len; ++i)
		1270	PUSHs (*av_fetch (av, i, 1));
		1271
		1272	PUTBACK;
		1273	call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
		1274	SPAGAIN;
		1275
		1276	if (SvTRUE (ERRSV))
		1277	{
		1278	FREETMPS; LEAVE;
		1279	ERR_ERRSV;
		1280	}
		1281
		1282	SvREFCNT_dec_NN (sv);
		1283	sv = SvREFCNT_inc (POPs);
		1284
		1285	PUTBACK;
		1286
		1287	FREETMPS; LEAVE;
		1288	}
		1289	break;
		1290
		1291	default:
		1292	filter:
		1293	{
		1294	SV *tag_sv = newSVuv (tag);
		1295
		1296	sv = decode_sv (dec);
		1297
		1298	dSP;
		1299	ENTER; SAVETMPS;
		1300	PUSHMARK (SP);
		1301	EXTEND (SP, 2);
		1302	PUSHs (tag_sv);
		1303	PUSHs (sv);
		1304
		1305	PUTBACK;
		1306	int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
		1307	SPAGAIN;
		1308
		1309	if (SvTRUE (ERRSV))
		1310	{
		1311	SvREFCNT_dec_NN (tag_sv);
		1312	FREETMPS; LEAVE;
		1313	ERR_ERRSV;
		1314	}
		1315
		1316	if (count)
		1317	{
		1318	SvREFCNT_dec_NN (tag_sv);
		1319	SvREFCNT_dec_NN (sv);
		1320	sv = SvREFCNT_inc_NN (TOPs);
		1321	SP -= count;
		1322	}
		1323	else
		1324	{
		1325	AV *av = newAV ();
		1326	av_push (av, tag_sv);
		1327	av_push (av, sv);
		1328
		1329	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		1330	? cbor_tagged_stash
		1331	: gv_stashpv ("CBOR::XS::Tagged" , 1);
		1332	sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
		1333	}
		1334
		1335	PUTBACK;
		1336
		1337	FREETMPS; LEAVE;
		1338	}
		1339	break;
		1340	}
		1341
788	return sv;	1342	return sv;
789		1343
790	AV *av = newAV ();	1344	fail:
791	av_push (av, newSVuv (tag));	1345	SvREFCNT_dec (sv);
792	av_push (av, sv);	1346	return &PL_sv_undef;
793	return newRV_noinc ((SV *)av);
794	}	1347	}
795		1348
796	static SV *	1349	static SV *
797	decode_sv (dec_t *dec)	1350	decode_sv (dec_t *dec)
798	{	1351	{
799	WANT (1);	1352	WANT (1);
800		1353
801	switch (*dec->cur >> 5)	1354	switch (*dec->cur >> MAJOR_SHIFT)
802	{	1355	{
803	case 0: // unsigned int	1356	case MAJOR_POS_INT >> MAJOR_SHIFT: return newSVuv (decode_uint (dec));
804	//TODO: 64 bit values on 3 2bit perls	1357	case MAJOR_NEG_INT >> MAJOR_SHIFT: return newSViv (-1 - (IV)decode_uint (dec));
805	return newSVuv (decode_uint (dec));	1358	case MAJOR_BYTES >> MAJOR_SHIFT: return decode_str (dec, 0);
806	case 1: // negative int	1359	case MAJOR_TEXT >> MAJOR_SHIFT: return decode_str (dec, 1);
807	return newSViv (-1 - (IV)decode_uint (dec));	1360	case MAJOR_ARRAY >> MAJOR_SHIFT: return decode_av (dec);
808	case 2: // octet string	1361	case MAJOR_MAP >> MAJOR_SHIFT: return decode_hv (dec);
809	return decode_str (dec, 0);	1362	case MAJOR_TAG >> MAJOR_SHIFT: return decode_tagged (dec);
810	case 3: // utf-8 string	1363
811	return decode_str (dec, 1);	1364	case MAJOR_MISC >> MAJOR_SHIFT:
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)	1365	switch (*dec->cur++ & MINOR_MASK)
820	{	1366	{
821	case 20:	1367	case SIMPLE_FALSE:
822	#if CBOR_SLOW	1368	#if CBOR_SLOW
823	cbor_false = get_bool ("CBOR::XS::false");	1369	types_false = get_bool ("Types::Serialiser::false");
824	#endif	1370	#endif
825	return newSVsv (cbor_false);	1371	return newSVsv (types_false);
826	case 21:	1372	case SIMPLE_TRUE:
827	#if CBOR_SLOW	1373	#if CBOR_SLOW
828	cbor_true = get_bool ("CBOR::XS::true");	1374	types_true = get_bool ("Types::Serialiser::true");
829	#endif	1375	#endif
830	return newSVsv (cbor_true);	1376	return newSVsv (types_true);
831	case 22:	1377	case SIMPLE_NULL:
832	return newSVsv (&PL_sv_undef);	1378	return newSVsv (&PL_sv_undef);
		1379	case SIMPLE_UNDEF:
		1380	#if CBOR_SLOW
		1381	types_error = get_bool ("Types::Serialiser::error");
		1382	#endif
		1383	return newSVsv (types_error);
833		1384
834	case 25:	1385	case MISC_FLOAT16:
835	{	1386	{
836	WANT (2);	1387	WANT (2);
837		1388
838	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];	1389	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
839	dec->cur += 2;	1390	dec->cur += 2;
840		1391
841	return newSVnv (ecb_binary16_to_float (fp));	1392	return newSVnv (ecb_binary16_to_float (fp));
842	}	1393	}
843		1394
844	case 26:	1395	case MISC_FLOAT32:
845	{	1396	{
846	uint32_t fp;	1397	uint32_t fp;
847	WANT (4);	1398	WANT (4);
848	memcpy (&fp, dec->cur, 4);	1399	memcpy (&fp, dec->cur, 4);
849	dec->cur += 4;	1400	dec->cur += 4;
…		…
852	fp = ecb_bswap32 (fp);	1403	fp = ecb_bswap32 (fp);
853		1404
854	return newSVnv (ecb_binary32_to_float (fp));	1405	return newSVnv (ecb_binary32_to_float (fp));
855	}	1406	}
856		1407
857	case 27:	1408	case MISC_FLOAT64:
858	{	1409	{
859	uint64_t fp;	1410	uint64_t fp;
860	WANT (8);	1411	WANT (8);
861	memcpy (&fp, dec->cur, 8);	1412	memcpy (&fp, dec->cur, 8);
862	dec->cur += 8;	1413	dec->cur += 8;
…		…
865	fp = ecb_bswap64 (fp);	1416	fp = ecb_bswap64 (fp);
866		1417
867	return newSVnv (ecb_binary64_to_double (fp));	1418	return newSVnv (ecb_binary64_to_double (fp));
868	}	1419	}
869		1420
870	// 0..19 unassigned	1421	// 0..19 unassigned simple
871	// 24 reserved + unassigned (reserved values are not encodable)	1422	// 24 reserved + unassigned simple (reserved values are not encodable)
		1423	// 28-30 unassigned misc
		1424	// 31 break code
872	default:	1425	default:
873	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");	1426	ERR ("corrupted CBOR data (reserved/unassigned/unexpected major 7 value)");
874	}	1427	}
875		1428
876	break;	1429	break;
877	}	1430	}
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		1431
935	fail:	1432	fail:
936	return &PL_sv_undef;	1433	return &PL_sv_undef;
937	}	1434	}
938		1435
939	static SV *	1436	static SV *
940	decode_cbor (SV *string, CBOR *cbor, char **offset_return)	1437	decode_cbor (SV *string, CBOR *cbor, char **offset_return)
941	{	1438	{
942	dec_t dec;	1439	dec_t dec = { 0 };
943	SV *sv;	1440	SV *sv;
		1441	STRLEN len;
		1442	char *data = SvPVbyte (string, len);
944		1443
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)	1444	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",	1445	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);	1446	(unsigned long)len, (unsigned long)cbor->max_size);
976	}
977
978	sv_utf8_downgrade (string, 0);
979		1447
980	dec.cbor = *cbor;	1448	dec.cbor = *cbor;
981	dec.cur = (U8 *)SvPVX (string);	1449	dec.cur = (U8 *)data;
982	dec.end = (U8 *)SvEND (string);	1450	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		1451
989	sv = decode_sv (&dec);	1452	sv = decode_sv (&dec);
990		1453
991	if (offset_return)	1454	if (offset_return)
992	*offset_return = dec.cur;	1455	*offset_return = dec.cur;
…		…
995	if (dec.cur != dec.end && !dec.err)	1458	if (dec.cur != dec.end && !dec.err)
996	dec.err = "garbage after CBOR object";	1459	dec.err = "garbage after CBOR object";
997		1460
998	if (dec.err)	1461	if (dec.err)
999	{	1462	{
		1463	if (dec.shareable)
		1464	{
		1465	// need to break cyclic links, which would all be in shareable
		1466	int i;
		1467	SV **svp;
		1468
		1469	for (i = av_len (dec.shareable) + 1; i--; )
		1470	if ((svp = av_fetch (dec.shareable, i, 0)))
		1471	sv_setsv (*svp, &PL_sv_undef);
		1472	}
		1473
1000	SvREFCNT_dec (sv);	1474	SvREFCNT_dec_NN (sv);
		1475
		1476	if (dec.err_sv)
		1477	sv_2mortal (dec.err_sv);
		1478
1001	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur);	1479	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
1002	}	1480	}
1003		1481
1004	sv = sv_2mortal (sv);	1482	sv = sv_2mortal (sv);
1005		1483
1006	return sv;	1484	return sv;
1007	}	1485	}
1008		1486
		1487	/////////////////////////////////////////////////////////////////////////////
		1488	// incremental parser
		1489
		1490	#define INCR_DONE(cbor) (AvFILLp (cbor->incr_count) < 0)
		1491
		1492	// returns 0 for notyet, 1 for success or error
		1493	static int
		1494	incr_parse (CBOR *self, SV *cborstr)
		1495	{
		1496	STRLEN cur;
		1497	SvPV (cborstr, cur);
		1498
		1499	while (ecb_expect_true (self->incr_need <= cur))
		1500	{
		1501	// table of integer count bytes
		1502	static I8 incr_len[MINOR_MASK + 1] = {
		1503	0, 0, 0, 0, 0, 0, 0, 0,
		1504	0, 0, 0, 0, 0, 0, 0, 0,
		1505	0, 0, 0, 0, 0, 0, 0, 0,
		1506	1, 2, 4, 8,-1,-1,-1,-2
		1507	};
		1508
		1509	const U8 *p = SvPVX (cborstr) + self->incr_pos;
		1510	U8 m = *p & MINOR_MASK;
		1511	IV count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1512	I8 ilen = incr_len[m];
		1513
		1514	self->incr_need = self->incr_pos + 1;
		1515
		1516	if (ecb_expect_false (ilen < 0))
		1517	{
		1518	if (m != MINOR_INDEF)
		1519	return 1; // error
		1520
		1521	if (*p == (MAJOR_MISC | MINOR_INDEF))
		1522	{
		1523	if (count >= 0)
		1524	return 1; // error
		1525
		1526	count = 1;
		1527	}
		1528	else
		1529	{
		1530	av_push (self->incr_count, newSViv (-1)); //TODO: nest
		1531	count = -1;
		1532	}
		1533	}
		1534	else
		1535	{
		1536	self->incr_need += ilen;
		1537	if (ecb_expect_false (self->incr_need > cur))
		1538	return 0;
		1539
		1540	int major = *p >> MAJOR_SHIFT;
		1541
		1542	switch (major)
		1543	{
		1544	case MAJOR_TAG >> MAJOR_SHIFT:
		1545	++count; // tags merely prefix another value
		1546	break;
		1547
		1548	case MAJOR_BYTES >> MAJOR_SHIFT:
		1549	case MAJOR_TEXT >> MAJOR_SHIFT:
		1550	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1551	case MAJOR_MAP >> MAJOR_SHIFT:
		1552	{
		1553	UV len;
		1554
		1555	if (ecb_expect_false (ilen))
		1556	{
		1557	len = 0;
		1558
		1559	do {
		1560	len = (len << 8) | *++p;
		1561	} while (--ilen);
		1562	}
		1563	else
		1564	len = m;
		1565
		1566	switch (major)
		1567	{
		1568	case MAJOR_BYTES >> MAJOR_SHIFT:
		1569	case MAJOR_TEXT >> MAJOR_SHIFT:
		1570	self->incr_need += len;
		1571	if (ecb_expect_false (self->incr_need > cur))
		1572	return 0;
		1573
		1574	break;
		1575
		1576	case MAJOR_MAP >> MAJOR_SHIFT:
		1577	len <<= 1;
		1578	/* FALLTHROUGH */
		1579	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1580	if (len)
		1581	{
		1582	av_push (self->incr_count, newSViv (len + 1)); //TODO: nest
		1583	count = len + 1;
		1584	}
		1585	break;
		1586	}
		1587	}
		1588	}
		1589	}
		1590
		1591	self->incr_pos = self->incr_need;
		1592
		1593	if (count > 0)
		1594	{
		1595	while (!--count)
		1596	{
		1597	if (!AvFILLp (self->incr_count))
		1598	return 1; // done
		1599
		1600	SvREFCNT_dec_NN (av_pop (self->incr_count));
		1601	count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1602	}
		1603
		1604	SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]) = count;
		1605	}
		1606	}
		1607
		1608	return 0;
		1609	}
		1610
		1611
1009	/////////////////////////////////////////////////////////////////////////////	1612	/////////////////////////////////////////////////////////////////////////////
1010	// XS interface functions	1613	// XS interface functions
1011		1614
1012	MODULE = CBOR::XS PACKAGE = CBOR::XS	1615	MODULE = CBOR::XS PACKAGE = CBOR::XS
1013		1616
1014	BOOT:	1617	BOOT:
1015	{	1618	{
1016	cbor_stash = gv_stashpv ("CBOR::XS" , 1);	1619	cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1017	cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1);	1620	cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1018		1621
1019	cbor_true = get_bool ("CBOR::XS::true");	1622	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1020	cbor_false = get_bool ("CBOR::XS::false");	1623	types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
		1624
		1625	types_true = get_bool ("Types::Serialiser::true" );
		1626	types_false = get_bool ("Types::Serialiser::false");
		1627	types_error = get_bool ("Types::Serialiser::error");
		1628
		1629	default_filter = newSVpv ("CBOR::XS::default_filter", 0);
		1630
		1631	sv_cbor = newSVpv ("CBOR", 0);
		1632	SvREADONLY_on (sv_cbor);
		1633
		1634	assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1021	}	1635	}
1022		1636
1023	PROTOTYPES: DISABLE	1637	PROTOTYPES: DISABLE
1024		1638
1025	void CLONE (...)	1639	void CLONE (...)
1026	CODE:	1640	CODE:
1027	cbor_stash = 0;	1641	cbor_stash = 0;
		1642	cbor_tagged_stash = 0;
		1643	types_error_stash = 0;
1028	cbor_boolean_stash = 0;	1644	types_boolean_stash = 0;
1029		1645
1030	void new (char *klass)	1646	void new (char *klass)
1031	PPCODE:	1647	PPCODE:
1032	{	1648	{
1033	SV *pv = NEWSV (0, sizeof (CBOR));	1649	SV *pv = NEWSV (0, sizeof (CBOR));
…		…
1041		1657
1042	void shrink (CBOR *self, int enable = 1)	1658	void shrink (CBOR *self, int enable = 1)
1043	ALIAS:	1659	ALIAS:
1044	shrink = F_SHRINK	1660	shrink = F_SHRINK
1045	allow_unknown = F_ALLOW_UNKNOWN	1661	allow_unknown = F_ALLOW_UNKNOWN
		1662	allow_sharing = F_ALLOW_SHARING
		1663	allow_cycles = F_ALLOW_CYCLES
		1664	forbid_objects = F_FORBID_OBJECTS
		1665	pack_strings = F_PACK_STRINGS
		1666	text_keys = F_TEXT_KEYS
		1667	text_strings = F_TEXT_STRINGS
		1668	validate_utf8 = F_VALIDATE_UTF8
1046	PPCODE:	1669	PPCODE:
1047	{	1670	{
1048	if (enable)	1671	if (enable)
1049	self->flags |= ix;	1672	self->flags |= ix;
1050	else	1673	else
…		…
1055		1678
1056	void get_shrink (CBOR *self)	1679	void get_shrink (CBOR *self)
1057	ALIAS:	1680	ALIAS:
1058	get_shrink = F_SHRINK	1681	get_shrink = F_SHRINK
1059	get_allow_unknown = F_ALLOW_UNKNOWN	1682	get_allow_unknown = F_ALLOW_UNKNOWN
		1683	get_allow_sharing = F_ALLOW_SHARING
		1684	get_allow_cycles = F_ALLOW_CYCLES
		1685	get_forbid_objects = F_FORBID_OBJECTS
		1686	get_pack_strings = F_PACK_STRINGS
		1687	get_text_keys = F_TEXT_KEYS
		1688	get_text_strings = F_TEXT_STRINGS
		1689	get_validate_utf8 = F_VALIDATE_UTF8
1060	PPCODE:	1690	PPCODE:
1061	XPUSHs (boolSV (self->flags & ix));	1691	XPUSHs (boolSV (self->flags & ix));
1062		1692
1063	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)	1693	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1064	PPCODE:	1694	PPCODE:
…		…
1080	CODE:	1710	CODE:
1081	RETVAL = self->max_size;	1711	RETVAL = self->max_size;
1082	OUTPUT:	1712	OUTPUT:
1083	RETVAL	1713	RETVAL
1084		1714
1085	#if 0 //TODO	1715	void filter (CBOR *self, SV *filter = 0)
1086
1087	void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1088	PPCODE:	1716	PPCODE:
1089	{
1090	SvREFCNT_dec (self->cb_object);	1717	SvREFCNT_dec (self->filter);
1091	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;	1718	self->filter = filter ? newSVsv (filter) : filter;
1092
1093	XPUSHs (ST (0));	1719	XPUSHs (ST (0));
1094	}
1095		1720
1096	void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)	1721	SV *get_filter (CBOR *self)
1097	PPCODE:	1722	CODE:
1098	{	1723	RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1099	if (!self->cb_sk_object)	1724	OUTPUT:
1100	self->cb_sk_object = newHV ();	1725	RETVAL
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		1726
1120	void encode (CBOR *self, SV *scalar)	1727	void encode (CBOR *self, SV *scalar)
1121	PPCODE:	1728	PPCODE:
1122	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;	1729	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1123	XPUSHs (scalar);	1730	XPUSHs (scalar);
…		…
1136	EXTEND (SP, 2);	1743	EXTEND (SP, 2);
1137	PUSHs (sv);	1744	PUSHs (sv);
1138	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));	1745	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1139	}	1746	}
1140		1747
		1748	void incr_parse (CBOR *self, SV *cborstr)
		1749	ALIAS:
		1750	incr_parse_multiple = 1
		1751	PPCODE:
		1752	{
		1753	if (SvUTF8 (cborstr))
		1754	sv_utf8_downgrade (cborstr, 0);
		1755
		1756	if (!self->incr_count)
		1757	{
		1758	self->incr_count = newAV ();
		1759	self->incr_pos = 0;
		1760	self->incr_need = 1;
		1761
		1762	av_push (self->incr_count, newSViv (1));
		1763	}
		1764
		1765	do
		1766	{
		1767	if (!incr_parse (self, cborstr))
		1768	{
		1769	if (self->incr_need > self->max_size && self->max_size)
		1770	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
		1771	(unsigned long)self->incr_need, (unsigned long)self->max_size);
		1772
		1773	break;
		1774	}
		1775
		1776	SV *sv;
		1777	char *offset;
		1778
		1779	PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
		1780	XPUSHs (sv);
		1781
		1782	sv_chop (cborstr, offset);
		1783
		1784	av_clear (self->incr_count);
		1785	av_push (self->incr_count, newSViv (1));
		1786
		1787	self->incr_pos = 0;
		1788	self->incr_need = self->incr_pos + 1;
		1789	}
		1790	while (ix);
		1791	}
		1792
		1793	void incr_reset (CBOR *self)
		1794	CODE:
		1795	{
		1796	SvREFCNT_dec (self->incr_count);
		1797	self->incr_count = 0;
		1798	}
		1799
1141	void DESTROY (CBOR *self)	1800	void DESTROY (CBOR *self)
1142	CODE:	1801	PPCODE:
1143	SvREFCNT_dec (self->cb_sk_object);	1802	cbor_free (self);
1144	SvREFCNT_dec (self->cb_object);
1145		1803
1146	PROTOTYPES: ENABLE	1804	PROTOTYPES: ENABLE
1147		1805
1148	void encode_cbor (SV *scalar)	1806	void encode_cbor (SV *scalar)
		1807	ALIAS:
		1808	encode_cbor = 0
		1809	encode_cbor_sharing = F_ALLOW_SHARING
1149	PPCODE:	1810	PPCODE:
1150	{	1811	{
1151	CBOR cbor;	1812	CBOR cbor;
1152	cbor_init (&cbor);	1813	cbor_init (&cbor);
		1814	cbor.flags |= ix;
1153	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;	1815	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1154	XPUSHs (scalar);	1816	XPUSHs (scalar);
1155	}	1817	}
1156		1818
1157	void decode_cbor (SV *cborstr)	1819	void decode_cbor (SV *cborstr)
…		…
1161	cbor_init (&cbor);	1823	cbor_init (&cbor);
1162	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;	1824	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1163	XPUSHs (cborstr);	1825	XPUSHs (cborstr);
1164	}	1826	}
1165		1827
		1828	#ifdef __AFL_COMPILER
		1829
		1830	void
		1831	afl_init ()
		1832	CODE:
		1833	__AFL_INIT ();
		1834
		1835	int
		1836	afl_loop (unsigned int count = 10000)
		1837	CODE:
		1838	RETVAL = __AFL_LOOP (count);
		1839	OUTPUT:
		1840	RETVAL
		1841
		1842	#endif
		1843

Diff Legend

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