ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/CBOR-XS/XS.xs

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.1 by root, Fri Oct 25 23:09:45 2013 UTC vs.
Revision 1.78 by root, Fri Sep 8 06:01:12 2023 UTC

…		…
6	#include <string.h>	6	#include <string.h>
7	#include <stdlib.h>	7	#include <stdlib.h>
8	#include <stdio.h>	8	#include <stdio.h>
9	#include <limits.h>	9	#include <limits.h>
10	#include <float.h>	10	#include <float.h>
		11	#include <inttypes.h>
11		12
		13	#define ECB_NO_THREADS 1
12	#include "ecb.h"	14	#include "ecb.h"
13		15
14	#if defined(__BORLANDC__) || defined(_MSC_VER)	16	// compatibility with perl <5.18
15	# define snprintf _snprintf // C compilers have this in stdio.h	17	#ifndef HvNAMELEN_get
		18	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
16	#endif	19	#endif
		20	#ifndef HvNAMELEN
		21	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		22	#endif
		23	#ifndef HvNAMEUTF8
		24	# define HvNAMEUTF8(hv) 0
		25	#endif
		26	#ifndef SvREFCNT_inc_NN
		27	# define SvREFCNT_inc_NN(sv) SvREFCNT_inc (sv)
		28	#endif
		29	#ifndef SvREFCNT_dec_NN
		30	# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
		31	#endif
17		32
		33	// 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
18	#define F_SHRINK 0x00000200UL	121	#define F_SHRINK 0x00000001UL
19	#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
20		130
21	#define INIT_SIZE 32 // initial scalar size to be allocated	131	#define INIT_SIZE 32 // initial scalar size to be allocated
22		132
23	#define SB do {	133	#define SB do {
24	#define SE } while (0)	134	#define SE } while (0)
25
26	#if __GNUC__ >= 3
27	# define expect(expr,value) __builtin_expect ((expr), (value))
28	# define INLINE static inline
29	#else
30	# define expect(expr,value) (expr)
31	# define INLINE static
32	#endif
33
34	#define expect_false(expr) expect ((expr) != 0, 0)
35	#define expect_true(expr) expect ((expr) != 0, 1)
36		135
37	#define IN_RANGE_INC(type,val,beg,end) \	136	#define IN_RANGE_INC(type,val,beg,end) \
38	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	137	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
39	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	138	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
40		139
…		…
46	#else	145	#else
47	# define CBOR_SLOW 0	146	# define CBOR_SLOW 0
48	# define CBOR_STASH cbor_stash	147	# define CBOR_STASH cbor_stash
49	#endif	148	#endif
50		149
51	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::
52	static SV *cbor_true, *cbor_false;	151	static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
53		152
54	typedef struct {	153	typedef struct {
55	U32 flags;	154	U32 flags;
56	U32 max_depth;	155	U32 max_depth;
57	STRLEN max_size;	156	STRLEN max_size;
		157	SV *filter;
58		158
59	SV *cb_object;	159	// for the incremental parser
60	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.
61	} CBOR;	163	} CBOR;
62		164
63	INLINE void	165	ecb_inline void
64	cbor_init (CBOR *cbor)	166	cbor_init (CBOR *cbor)
65	{	167	{
66	Zero (cbor, 1, CBOR);	168	Zero (cbor, 1, CBOR);
67	cbor->max_depth = 512;	169	cbor->max_depth = 512;
68	}	170	}
69		171
		172	ecb_inline void
		173	cbor_free (CBOR *cbor)
		174	{
		175	SvREFCNT_dec (cbor->filter);
		176	SvREFCNT_dec (cbor->incr_count);
		177	}
		178
70	/////////////////////////////////////////////////////////////////////////////	179	/////////////////////////////////////////////////////////////////////////////
71	// utility functions	180	// utility functions
72		181
73	INLINE SV *	182	ecb_inline SV *
74	get_bool (const char *name)	183	get_bool (const char *name)
75	{	184	{
76	SV *sv = get_sv (name, 1);	185	SV *sv = get_sv (name, 1);
77		186
78	SvREADONLY_on (sv);	187	SvREADONLY_on (sv);
79	SvREADONLY_on (SvRV (sv));	188	SvREADONLY_on (SvRV (sv));
80		189
81	return sv;	190	return sv;
82	}	191	}
83		192
84	INLINE void	193	ecb_inline void
85	shrink (SV *sv)	194	shrink (SV *sv)
86	{	195	{
87	sv_utf8_downgrade (sv, 1);	196	sv_utf8_downgrade (sv, 1);
88		197
89	if (SvLEN (sv) > SvCUR (sv) + 1)	198	if (SvLEN (sv) > SvCUR (sv) + 1)
…		…
94	SvPV_renew (sv, SvCUR (sv) + 1);	203	SvPV_renew (sv, SvCUR (sv) + 1);
95	#endif	204	#endif
96	}	205	}
97	}	206	}
98		207
99	/////////////////////////////////////////////////////////////////////////////	208	// minimum length of a string to be registered for stringref
100	// fp hell	209	ecb_inline STRLEN
101		210	minimum_string_length (UV idx)
102	//TODO	211	{
		212	return idx <= 23 ? 3
		213	: idx <= 0xffU ? 4
		214	: idx <= 0xffffU ? 5
		215	: idx <= 0xffffffffU ? 7
		216	: 11;
		217	}
103		218
104	/////////////////////////////////////////////////////////////////////////////	219	/////////////////////////////////////////////////////////////////////////////
105	// encoder	220	// encoder
106		221
107	// structure used for encoding CBOR	222	// structure used for encoding CBOR
…		…
110	char *cur; // SvPVX (sv) + current output position	225	char *cur; // SvPVX (sv) + current output position
111	char *end; // SvEND (sv)	226	char *end; // SvEND (sv)
112	SV *sv; // result scalar	227	SV *sv; // result scalar
113	CBOR cbor;	228	CBOR cbor;
114	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;
115	} enc_t;	234	} enc_t;
116		235
117	INLINE void	236	ecb_inline void
118	need (enc_t *enc, STRLEN len)	237	need (enc_t *enc, STRLEN len)
119	{	238	{
120	if (expect_false (enc->cur + len >= enc->end))	239	if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
121	{	240	{
122	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);	241	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
123	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);	242	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
124	enc->cur = SvPVX (enc->sv) + cur;	243	enc->cur = SvPVX (enc->sv) + cur;
125	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	244	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126	}	245	}
127	}	246	}
128		247
129	INLINE void	248	static void encode_sv (enc_t *enc, SV *sv);
		249
		250	ecb_inline void
130	encode_ch (enc_t *enc, char ch)	251	encode_ch (enc_t *enc, char ch)
131	{	252	{
132	need (enc, 1);	253	need (enc, 1);
133	*enc->cur++ = ch;	254	*enc->cur++ = ch;
134	}	255	}
135		256
		257	// used for tags, intregers, element counts and so on
136	static void	258	static void
137	encode_uint (enc_t *enc, int major, UV len)	259	encode_uint (enc_t *enc, int major, UV len)
138	{	260	{
139	need (enc, 9);	261	need (enc, 9);
140		262
141	if (len < 24)	263	if (ecb_expect_true (len < LENGTH_EXT1))
142	*enc->cur++ = major | len;	264	*enc->cur++ = major | len;
143	else if (len < 0x100)	265	else if (ecb_expect_true (len <= 0xffU))
144	{	266	{
145	*enc->cur++ = major | 24;	267	*enc->cur++ = major | LENGTH_EXT1;
146	*enc->cur++ = len;	268	*enc->cur++ = len;
147	}	269	}
148	else if (len < 0x10000)	270	else if (len <= 0xffffU)
149	{	271	{
150	*enc->cur++ = major | 25;	272	*enc->cur++ = major | LENGTH_EXT2;
151	*enc->cur++ = len >> 8;	273	*enc->cur++ = len >> 8;
152	*enc->cur++ = len;	274	*enc->cur++ = len;
153	}	275	}
154	else if (len < 0x100000000)	276	else if (len <= 0xffffffffU)
155	{	277	{
156	*enc->cur++ = major | 26;	278	*enc->cur++ = major | LENGTH_EXT4;
157	*enc->cur++ = len >> 24;	279	*enc->cur++ = len >> 24;
158	*enc->cur++ = len >> 16;	280	*enc->cur++ = len >> 16;
159	*enc->cur++ = len >> 8;	281	*enc->cur++ = len >> 8;
160	*enc->cur++ = len;	282	*enc->cur++ = len;
161	}	283	}
162	else if (len)	284	else
163	{	285	{
164	*enc->cur++ = major | 27;	286	*enc->cur++ = major | LENGTH_EXT8;
165	*enc->cur++ = len >> 56;	287	*enc->cur++ = len >> 56;
166	*enc->cur++ = len >> 48;	288	*enc->cur++ = len >> 48;
167	*enc->cur++ = len >> 40;	289	*enc->cur++ = len >> 40;
168	*enc->cur++ = len >> 32;	290	*enc->cur++ = len >> 32;
169	*enc->cur++ = len >> 24;	291	*enc->cur++ = len >> 24;
…		…
171	*enc->cur++ = len >> 8;	293	*enc->cur++ = len >> 8;
172	*enc->cur++ = len;	294	*enc->cur++ = len;
173	}	295	}
174	}	296	}
175		297
176	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
177	encode_str (enc_t *enc, int utf8, char *str, STRLEN len)	318	encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
178	{	319	{
179	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);
180	need (enc, len);	350	need (enc, len);
181	memcpy (enc->cur, str, len);	351	memcpy (enc->cur, str, len);
182	enc->cur += len;	352	enc->cur += len;
183	}	353	}
184		354
185	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	}
186		510
187	static void	511	static void
188	encode_av (enc_t *enc, AV *av)	512	encode_av (enc_t *enc, AV *av)
189	{	513	{
190	int i, len = av_len (av);	514	int i, len = av_len (av);
…		…
192	if (enc->depth >= enc->cbor.max_depth)	516	if (enc->depth >= enc->cbor.max_depth)
193	croak (ERR_NESTING_EXCEEDED);	517	croak (ERR_NESTING_EXCEEDED);
194		518
195	++enc->depth;	519	++enc->depth;
196		520
197	encode_uint (enc, 0x80, len + 1);	521	encode_uint (enc, MAJOR_ARRAY, len + 1);
198		522
		523	if (ecb_expect_false (SvMAGICAL (av)))
199	for (i = 0; i <= len; ++i)	524	for (i = 0; i <= len; ++i)
200	{	525	{
201	SV **svp = av_fetch (av, i, 0);	526	SV **svp = av_fetch (av, i, 0);
202	encode_sv (enc, svp ? *svp : &PL_sv_undef);	527	encode_sv (enc, svp ? *svp : &PL_sv_undef);
203	}	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	}
204		535
205	--enc->depth;	536	--enc->depth;
206	}	537	}
207		538
208	static void	539	static void
…		…
216	++enc->depth;	547	++enc->depth;
217		548
218	int pairs = hv_iterinit (hv);	549	int pairs = hv_iterinit (hv);
219	int mg = SvMAGICAL (hv);	550	int mg = SvMAGICAL (hv);
220		551
221	if (mg)	552	if (ecb_expect_false (mg))
222	encode_ch (enc, 0xa0 | 31);	553	encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
223	else	554	else
224	encode_uint (enc, 0xa0, pairs);	555	encode_uint (enc, MAJOR_MAP, pairs);
225		556
226	while ((he = hv_iternext (hv)))	557	while ((he = hv_iternext (hv)))
227	{	558	{
228	if (HeKLEN (he) == HEf_SVKEY)	559	if (HeKLEN (he) == HEf_SVKEY)
229	encode_sv (enc, HeSVKEY (he));	560	encode_sv (enc, HeSVKEY (he));
230	else	561	else
231	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));
232		563
233	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));
234	}	565	}
235		566
236	if (mg)	567	if (ecb_expect_false (mg))
237	encode_ch (enc, 0xe0 | 31);	568	encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
238		569
239	--enc->depth;	570	--enc->depth;
240	}	571	}
241		572
242	// encode objects, arrays and special \0=false and \1=true values.	573	// encode objects, arrays and special \0=false and \1=true values.
243	static void	574	static void
244	encode_rv (enc_t *enc, SV *sv)	575	encode_rv (enc_t *enc, SV *sv)
245	{	576	{
246	svtype svt;
247
248	SvGETMAGIC (sv);	577	SvGETMAGIC (sv);
		578
249	svt = SvTYPE (sv);	579	svtype svt = SvTYPE (sv);
250		580
251	if (expect_false (SvOBJECT (sv)))	581	if (ecb_expect_false (SvOBJECT (sv)))
252	{	582	{
253	HV *stash = !CBOR_SLOW || cbor_boolean_stash	583	HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
254	? 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
255	: gv_stashpv ("CBOR::XS::Boolean", 1);	591	: gv_stashpv ("CBOR::XS::Tagged" , 1);
256		592
257	if (SvSTASH (sv) == stash)	593	HV *stash = SvSTASH (sv);
258	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	}
259	else	645	else
260	{	646	{
261	#if 0 //TODO	647	sv_setuv (*svp, enc->shareable_idx);
262	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)))
263	{	662	{
		663	dSP;
		664
		665	ENTER; SAVETMPS;
		666	PUSHMARK (SP);
264	// 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
265	GV *to_cbor = gv_fetchmethod_autoload (SvSTASH (sv), "TO_CBOR", 0);
266
267	if (to_cbor)
268	{
269	dSP;
270
271	ENTER; SAVETMPS; PUSHMARK (SP);
272	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	668	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
273		669
274	// calling with G_SCALAR ensures that we always get a 1 return value
275	PUTBACK;	670	PUTBACK;
		671	// G_SCALAR ensures that return value is 1
276	call_sv ((SV *)GvCV (to_cbor), G_SCALAR);	672	call_sv ((SV *)GvCV (method), G_SCALAR);
277	SPAGAIN;	673	SPAGAIN;
278		674
279	// catch this surprisingly common error	675	// catch this surprisingly common error
280	if (SvROK (TOPs) && SvRV (TOPs) == sv)	676	if (SvROK (TOPs) && SvRV (TOPs) == sv)
281	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));	677	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
282		678
283	sv = POPs;
284	PUTBACK;
285
286	encode_sv (enc, sv);	679	encode_sv (enc, POPs);
287		680
		681	PUTBACK;
		682
288	FREETMPS; LEAVE;	683	FREETMPS; LEAVE;
289	}	684	}
290	else if (enc->cbor.flags & F_ALLOW_BLESSED)	685	else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
291	encode_str (enc, "null", 4, 0);	686	{
292	else	687	dSP;
293	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it",	688
294	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;
295	}	715	}
296	else if (enc->cbor.flags & F_ALLOW_BLESSED)	716
297	encode_str (enc, "null", 4, 0);	717	PUTBACK;
		718
		719	FREETMPS; LEAVE;
		720	}
298	else	721	else
299	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",
300	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	723	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
301	#endif
302	}
303	}	724	}
304	else if (svt == SVt_PVHV)	725	else if (svt == SVt_PVHV)
305	encode_hv (enc, (HV *)sv);	726	encode_hv (enc, (HV *)sv);
306	else if (svt == SVt_PVAV)	727	else if (svt == SVt_PVAV)
307	encode_av (enc, (AV *)sv);	728	encode_av (enc, (AV *)sv);
308	else if (svt < SVt_PVAV)
309	{
310	STRLEN len = 0;
311	char *pv = svt ? SvPV (sv, len) : 0;
312
313	if (len == 1 && *pv == '1')
314	encode_ch (enc, 0xe0 | 21);
315	else if (len == 1 && *pv == '0')
316	encode_ch (enc, 0xe0 | 20);
317	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
318	encode_ch (enc, 0xe0 | 23);
319	else
320	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
321	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
322	}
323	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
324	encode_ch (enc, 0xe0 | 23);
325	else	729	else
326	croak ("encountered %s, but CBOR can only represent references to arrays or hashes",	730	{
327	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	731	encode_tag (enc, CBOR_TAG_INDIRECTION);
		732	encode_sv (enc, sv);
		733	}
328	}	734	}
329		735
330	static void	736	static void
331	encode_nv (enc_t *enc, SV *sv)	737	encode_nv (enc_t *enc, SV *sv)
332	{	738	{
333	double nv = SvNVX (sv);	739	double nv = SvNVX (sv);
334		740
335	need (enc, 9);	741	need (enc, 9);
336		742
337	if (expect_false (nv == (U32)nv))	743	if (ecb_expect_false (nv == (NV)(U32)nv))
338	encode_uint (enc, 0x00, (U32)nv);	744	encode_uint (enc, MAJOR_POS_INT, (U32)nv);
339	//TODO: maybe I32?	745	//TODO: maybe I32?
340	else if (expect_false (nv == (float)nv))	746	else if (ecb_expect_false (nv == (float)nv))
341	{	747	encode_float32 (enc, nv);
342	uint32_t fp = ecb_float_to_binary32 (nv);
343
344	*enc->cur++ = 0xe0 | 26;
345
346	if (!ecb_big_endian ())
347	fp = ecb_bswap32 (fp);
348
349	memcpy (enc->cur, &fp, 4);
350	enc->cur += 4;
351	}
352	else	748	else
353	{	749	encode_float64 (enc, nv);
354	uint64_t fp = ecb_double_to_binary64 (nv);
355
356	*enc->cur++ = 0xe0 | 27;
357
358	if (!ecb_big_endian ())
359	fp = ecb_bswap64 (fp);
360
361	memcpy (enc->cur, &fp, 8);
362	enc->cur += 8;
363	}
364	}	750	}
365		751
366	static void	752	static void
367	encode_sv (enc_t *enc, SV *sv)	753	encode_sv (enc_t *enc, SV *sv)
368	{	754	{
…		…
370		756
371	if (SvPOKp (sv))	757	if (SvPOKp (sv))
372	{	758	{
373	STRLEN len;	759	STRLEN len;
374	char *str = SvPV (sv, len);	760	char *str = SvPV (sv, len);
375	encode_str (enc, SvUTF8 (sv), str, len);	761	encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
376	}	762	}
377	else if (SvNOKp (sv))	763	else if (SvNOKp (sv))
378	encode_nv (enc, sv);	764	encode_nv (enc, sv);
379	else if (SvIOKp (sv))	765	else if (SvIOKp (sv))
380	{	766	encode_int (enc, sv);
381	if (SvIsUV (sv))
382	encode_uint (enc, 0x00, SvUVX (sv));
383	else if (SvIVX (sv) >= 0)
384	encode_uint (enc, 0x00, SvIVX (sv));
385	else
386	encode_uint (enc, 0x20, -(SvIVX (sv) + 1));
387	}
388	else if (SvROK (sv))	767	else if (SvROK (sv))
389	encode_rv (enc, SvRV (sv));	768	encode_rv (enc, SvRV (sv));
390	else if (!SvOK (sv))	769	else if (!SvOK (sv))
391	encode_ch (enc, 0xe0 | 22);	770	encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
392	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)	771	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
393	encode_ch (enc, 0xe0 | 23);	772	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
394	else	773	else
395	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",
396	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));	775	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
397	}	776	}
398		777
399	static SV *	778	static SV *
400	encode_cbor (SV *scalar, CBOR *cbor)	779	encode_cbor (SV *scalar, CBOR *cbor)
401	{	780	{
402	enc_t enc;	781	enc_t enc = { 0 };
403		782
404	enc.cbor = *cbor;	783	enc.cbor = *cbor;
405	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	784	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
406	enc.cur = SvPVX (enc.sv);	785	enc.cur = SvPVX (enc.sv);
407	enc.end = SvEND (enc.sv);	786	enc.end = SvEND (enc.sv);
408	enc.depth = 0;
409		787
410	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
411	encode_sv (&enc, scalar);	797	encode_sv (&enc, scalar);
412		798
413	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	799	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
414	*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
415		801
…		…
429	U8 *end; // end of input string	815	U8 *end; // end of input string
430	const char *err; // parse error, if != 0	816	const char *err; // parse error, if != 0
431	CBOR cbor;	817	CBOR cbor;
432	U32 depth; // recursion depth	818	U32 depth; // recursion depth
433	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
434	} dec_t;	824	} dec_t;
435		825
436	#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);
437		833
438	#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;
439		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
440	#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)
441	#define DEC_DEC_DEPTH --dec->depth	863	#define DEC_DEC_DEPTH --dec->depth
442		864
443	static UV	865	static UV
444	decode_uint (dec_t *dec)	866	decode_uint (dec_t *dec)
445	{	867	{
446	switch (*dec->cur & 31)	868	U8 m = *dec->cur & MINOR_MASK;
447	{	869	++dec->cur;
448	case 0: case 1: case 2: case 3: case 4: case 5: case 6: case 7:
449	case 8: case 9: case 10: case 11: case 12: case 13: case 14: case 15:
450	case 16: case 17: case 18: case 19: case 20: case 21: case 22: case 23:
451	return *dec->cur++ & 31;
452		870
453	case 24:	871	if (ecb_expect_true (m < LENGTH_EXT1))
		872	return m;
		873	else if (ecb_expect_true (m == LENGTH_EXT1))
		874	{
454	WANT (2);	875	WANT (1);
455	dec->cur += 2;	876	dec->cur += 1;
456	return dec->cur[-1];	877	return dec->cur[-1];
457		878	}
458	case 25:	879	else if (ecb_expect_true (m == LENGTH_EXT2))
		880	{
459	WANT (3);	881	WANT (2);
460	dec->cur += 3;	882	dec->cur += 2;
461	return (((UV)dec->cur[-2]) << 8)	883	return (((UV)dec->cur[-2]) << 8)
462	| ((UV)dec->cur[-1]);	884	| ((UV)dec->cur[-1]);
463		885	}
464	case 26:	886	else if (ecb_expect_true (m == LENGTH_EXT4))
		887	{
465	WANT (5);	888	WANT (4);
466	dec->cur += 5;	889	dec->cur += 4;
467	return (((UV)dec->cur[-4]) << 24)	890	return (((UV)dec->cur[-4]) << 24)
468	| (((UV)dec->cur[-3]) << 16)	891	| (((UV)dec->cur[-3]) << 16)
469	| (((UV)dec->cur[-2]) << 8)	892	| (((UV)dec->cur[-2]) << 8)
470	| ((UV)dec->cur[-1]);	893	| ((UV)dec->cur[-1]);
471		894	}
472	case 27:	895	else if (ecb_expect_true (m == LENGTH_EXT8))
		896	{
473	WANT (9);	897	WANT (8);
474	dec->cur += 9;	898	dec->cur += 8;
		899
		900	return
		901	#if UVSIZE < 8
		902	0
		903	#else
475	return (((UV)dec->cur[-8]) << 56)	904	(((UV)dec->cur[-8]) << 56)
476	| (((UV)dec->cur[-7]) << 48)	905	| (((UV)dec->cur[-7]) << 48)
477	| (((UV)dec->cur[-6]) << 40)	906	| (((UV)dec->cur[-6]) << 40)
478	| (((UV)dec->cur[-5]) << 32)	907	| (((UV)dec->cur[-5]) << 32)
		908	#endif
479	| (((UV)dec->cur[-4]) << 24)	909	| (((UV)dec->cur[-4]) << 24)
480	| (((UV)dec->cur[-3]) << 16)	910	| (((UV)dec->cur[-3]) << 16)
481	| (((UV)dec->cur[-2]) << 8)	911	| (((UV)dec->cur[-2]) << 8)
482	| ((UV)dec->cur[-1]);	912	| ((UV)dec->cur[-1]);
483		913	}
484	default:	914	else
485	ERR ("corrupted CBOR data (unsupported integer minor encoding)");	915	ERR ("corrupted CBOR data (unsupported integer minor encoding)");
486	}
487		916
488	fail:	917	fail:
489	return 0;	918	return 0;
490	}	919	}
491		920
…		…
496	{	925	{
497	AV *av = newAV ();	926	AV *av = newAV ();
498		927
499	DEC_INC_DEPTH;	928	DEC_INC_DEPTH;
500		929
501	if ((*dec->cur & 31) == 31)	930	if (*dec->cur == (MAJOR_ARRAY | MINOR_INDEF))
502	{	931	{
503	++dec->cur;	932	++dec->cur;
504		933
505	for (;;)	934	for (;;)
506	{	935	{
507	WANT (1);	936	WANT (1);
508		937
509	if (*dec->cur == 0xe0 | 31)	938	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
510	{	939	{
511	++dec->cur;	940	++dec->cur;
512	break;	941	break;
513	}	942	}
514		943
515	av_push (av, decode_sv (dec));	944	av_push (av, decode_sv (dec));
516	}	945	}
517	}	946	}
518	else	947	else
519	{	948	{
520	int i, len = decode_uint (dec);	949	UV i, len = decode_uint (dec);
521		950
		951	WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays
522	av_fill (av, len - 1);	952	av_fill (av, len - 1);
523		953
524	for (i = 0; i < len; ++i)	954	for (i = 0; i < len; ++i)
525	AvARRAY (av)[i] = decode_sv (dec);	955	AvARRAY (av)[i] = decode_sv (dec);
526	}	956	}
527		957
528	DEC_DEC_DEPTH;	958	DEC_DEC_DEPTH;
529	return newRV_noinc ((SV *)av);	959	return newRV_noinc ((SV *)av);
530		960
531	fail:	961	fail:
532	SvREFCNT_dec (av);	962	SvREFCNT_dec_NN (av);
533	DEC_DEC_DEPTH;	963	DEC_DEC_DEPTH;
534	return &PL_sv_undef;	964	return &PL_sv_undef;
535	}	965	}
536		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
		991	WANT (len);
		992	dec->cur += len;
		993
		994	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		995	if (!cbor_is_utf8_string ((U8 *)key, len))
		996	ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
		997
		998	hv_store (hv, key, -len, decode_sv (dec), 0);
		999
		1000	return;
		1001	}
		1002
		1003	SV *k = decode_sv (dec);
		1004	SV *v = decode_sv (dec);
		1005
		1006	// we leak memory if uncaught exceptions are thrown by random magical
		1007	// methods, and this is hopefully the only place where it can happen,
		1008	// so if there is a chance of an exception, take the very slow path.
		1009	// since catching exceptions is "undocumented/internal/forbidden" by
		1010	// the new p5p powers, we need to call out to a perl function :/
		1011	if (ecb_expect_false (SvAMAGIC (k)))
		1012	{
		1013	dSP;
		1014
		1015	ENTER; SAVETMPS;
		1016	PUSHMARK (SP);
		1017	EXTEND (SP, 3);
		1018	PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
		1019	PUSHs (sv_2mortal (k));
		1020	PUSHs (sv_2mortal (v));
		1021
		1022	PUTBACK;
		1023	call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
		1024	SPAGAIN;
		1025
		1026	FREETMPS; LEAVE;
		1027
		1028	if (SvTRUE (ERRSV))
		1029	ERR_ERRSV;
		1030
		1031	return;
		1032	}
		1033
		1034	hv_store_ent (hv, k, v, 0);
		1035	SvREFCNT_dec_NN (k);
		1036
		1037	fail:
		1038	;
		1039	}
		1040
537	static SV *	1041	static SV *
538	decode_hv (dec_t *dec)	1042	decode_hv (dec_t *dec)
539	{	1043	{
540	HV *hv = newHV ();	1044	HV *hv = newHV ();
541		1045
542	DEC_INC_DEPTH;	1046	DEC_INC_DEPTH;
543		1047
544	if ((*dec->cur & 31) == 31)	1048	if (*dec->cur == (MAJOR_MAP | MINOR_INDEF))
545	{	1049	{
546	++dec->cur;	1050	++dec->cur;
547		1051
548	for (;;)	1052	for (;;)
549	{	1053	{
550	WANT (1);	1054	WANT (1);
551		1055
552	if (*dec->cur == 0xe0 | 31)	1056	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
553	{	1057	{
554	++dec->cur;	1058	++dec->cur;
555	break;	1059	break;
556	}	1060	}
557		1061
558	SV *k = decode_sv (dec);	1062	decode_he (dec, hv);
559	SV *v = decode_sv (dec);
560
561	hv_store_ent (hv, k, v, 0);
562	}	1063	}
563	}	1064	}
564	else	1065	else
565	{	1066	{
566	int len = decode_uint (dec);	1067	UV pairs = decode_uint (dec);
567		1068
		1069	WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
		1070
568	while (len--)	1071	while (pairs--)
569	{	1072	decode_he (dec, hv);
570	SV *k = decode_sv (dec);
571	SV *v = decode_sv (dec);
572
573	hv_store_ent (hv, k, v, 0);
574	}
575	}	1073	}
576		1074
577	DEC_DEC_DEPTH;	1075	DEC_DEC_DEPTH;
578	return newRV_noinc ((SV *)hv);	1076	return newRV_noinc ((SV *)hv);
579		1077
580	#if 0	1078	fail:
581	SV *sv;	1079	SvREFCNT_dec_NN (hv);
582	HV *hv = newHV ();
583
584	DEC_INC_DEPTH;	1080	DEC_DEC_DEPTH;
585	decode_ws (dec);	1081	return &PL_sv_undef;
		1082	}
586		1083
587	for (;;)	1084	static SV *
588	{	1085	decode_str (dec_t *dec, int utf8)
589	// heuristic: assume that	1086	{
590	// a) decode_str + hv_store_ent are abysmally slow.	1087	SV *sv = 0;
591	// b) most hash keys are short, simple ascii text.	1088
592	// => try to "fast-match" such strings to avoid	1089	if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
593	// the overhead of decode_str + hv_store_ent.
594	{	1090	{
595	SV *value;	1091	// indefinite length strings
596	char *p = dec->cur;	1092	++dec->cur;
597	char *e = p + 24; // only try up to 24 bytes
598		1093
		1094	U8 major = *dec->cur & MAJOR_MISC;
		1095
		1096	sv = newSVpvn ("", 0);
		1097
599	for (;;)	1098	for (;;)
600	{	1099	{
601	// the >= 0x80 is false on most architectures	1100	WANT (1);
602	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1101
		1102	if ((*dec->cur - major) > LENGTH_EXT8)
		1103	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF))
603	{	1104	{
604	// slow path, back up and use decode_str	1105	++dec->cur;
605	SV *key = decode_str (dec);
606	if (!key)
607	goto fail;
608
609	decode_ws (dec); EXPECT_CH (':');
610
611	decode_ws (dec);
612	value = decode_sv (dec);
613	if (!value)
614	{
615	SvREFCNT_dec (key);
616	goto fail;
617	}
618
619	hv_store_ent (hv, key, value, 0);
620	SvREFCNT_dec (key);
621
622	break;	1106	break;
623	}	1107	}
624	else if (*p == '"')
625	{
626	// fast path, got a simple key
627	char *key = dec->cur;
628	int len = p - key;
629	dec->cur = p + 1;
630
631	decode_ws (dec); EXPECT_CH (':');
632
633	decode_ws (dec);
634	value = decode_sv (dec);
635	if (!value)
636	goto fail;
637
638	hv_store (hv, key, len, value, 0);
639
640	break;
641	}
642
643	++p;
644	}
645	}
646
647	decode_ws (dec);
648
649	if (*dec->cur == '}')
650	{
651	++dec->cur;
652	break;
653	}
654
655	if (*dec->cur != ',')
656	ERR (", or } expected while parsing object/hash");
657
658	++dec->cur;
659
660	decode_ws (dec);
661
662	if (*dec->cur == '}' && dec->cbor.flags & F_RELAXED)
663	{
664	++dec->cur;
665	break;
666	}
667	}
668
669	DEC_DEC_DEPTH;
670	sv = newRV_noinc ((SV *)hv);
671
672	// check filter callbacks
673	if (dec->cbor.flags & F_HOOK)
674	{
675	if (dec->cbor.cb_sk_object && HvKEYS (hv) == 1)
676	{
677	HE *cb, *he;
678
679	hv_iterinit (hv);
680	he = hv_iternext (hv);
681	hv_iterinit (hv);
682
683	// the next line creates a mortal sv each time its called.
684	// might want to optimise this for common cases.
685	cb = hv_fetch_ent (dec->cbor.cb_sk_object, hv_iterkeysv (he), 0, 0);
686
687	if (cb)
688	{	1108	else
689	dSP;	1109	ERR ("corrupted CBOR data (invalid chunks in indefinite length string)");
690	int count;
691		1110
692	ENTER; SAVETMPS; PUSHMARK (SP);	1111	STRLEN len = decode_uint (dec);
693	XPUSHs (HeVAL (he));
694	sv_2mortal (sv);
695		1112
696	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
697
698	if (count == 1)
699	{
700	sv = newSVsv (POPs);
701	FREETMPS; LEAVE;
702	return sv;
703	}
704
705	SvREFCNT_inc (sv);
706	FREETMPS; LEAVE;
707	}
708	}
709
710	if (dec->cbor.cb_object)
711	{
712	dSP;
713	int count;
714
715	ENTER; SAVETMPS; PUSHMARK (SP);
716	XPUSHs (sv_2mortal (sv));
717
718	PUTBACK; count = call_sv (dec->cbor.cb_object, G_ARRAY); SPAGAIN;
719
720	if (count == 1)
721	{
722	sv = newSVsv (POPs);
723	FREETMPS; LEAVE;
724	return sv;
725	}
726
727	SvREFCNT_inc (sv);
728	FREETMPS; LEAVE;
729	}
730	}
731
732	return sv;
733	#endif
734
735	fail:
736	SvREFCNT_dec (hv);
737	DEC_DEC_DEPTH;
738	return &PL_sv_undef;
739	}
740
741	static SV *
742	decode_str (dec_t *dec, int utf8)
743	{
744	SV *sv;
745
746	if ((*dec->cur & 31) == 31)
747	{
748	++dec->cur;
749
750	sv = newSVpvn ("", 0);
751
752	// not very fast, and certainly not robust against illegal input
753	for (;;)
754	{
755	WANT (1);	1113	WANT (len);
756		1114	sv_catpvn (sv, dec->cur, len);
757	if (*dec->cur == 0xe0 | 31)	1115	dec->cur += len;
758	{
759	++dec->cur;
760	break;
761	}
762
763	SV *sv2 = decode_sv (dec);
764	sv_catsv (sv, sv2);
765	}	1116	}
766	}	1117	}
767	else	1118	else
768	{	1119	{
769	STRLEN len = decode_uint (dec);	1120	STRLEN len = decode_uint (dec);
770		1121
771	WANT (len);	1122	WANT (len);
772	sv = newSVpvn (dec->cur, len);	1123	sv = newSVpvn (dec->cur, len);
773	dec->cur += len;	1124	dec->cur += len;
		1125
		1126	if (ecb_expect_false (dec->stringref)
		1127	&& SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
		1128	av_push (dec->stringref, SvREFCNT_inc_NN (sv));
774	}	1129	}
775		1130
776	if (utf8)	1131	if (utf8)
		1132	{
		1133	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		1134	if (!cbor_is_utf8_string (SvPVX (sv), SvCUR (sv)))
		1135	ERR ("corrupted CBOR data (invalid UTF-8 in text string)");
		1136
777	SvUTF8_on (sv);	1137	SvUTF8_on (sv);
		1138	}
778		1139
779	return sv;	1140	return sv;
780		1141
781	fail:	1142	fail:
		1143	SvREFCNT_dec (sv);
		1144	return &PL_sv_undef;
		1145	}
		1146
		1147	static SV *
		1148	decode_tagged (dec_t *dec)
		1149	{
		1150	SV *sv = 0;
		1151	UV tag = decode_uint (dec);
		1152
		1153	WANT (1);
		1154
		1155	switch (tag)
		1156	{
		1157	case CBOR_TAG_MAGIC:
		1158	sv = decode_sv (dec);
		1159	break;
		1160
		1161	case CBOR_TAG_INDIRECTION:
		1162	sv = newRV_noinc (decode_sv (dec));
		1163	break;
		1164
		1165	case CBOR_TAG_STRINGREF_NAMESPACE:
		1166	{
		1167	// do not use SAVETMPS/FREETMPS, as these will
		1168	// erase mortalised caches, e.g. "shareable"
		1169	ENTER;
		1170
		1171	SAVESPTR (dec->stringref);
		1172	dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
		1173
		1174	sv = decode_sv (dec);
		1175
		1176	LEAVE;
		1177	}
		1178	break;
		1179
		1180	case CBOR_TAG_STRINGREF:
		1181	{
		1182	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1183	ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
		1184
		1185	UV idx = decode_uint (dec);
		1186
		1187	if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
		1188	ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
		1189
		1190	sv = newSVsv (AvARRAY (dec->stringref)[idx]);
		1191	}
		1192	break;
		1193
		1194	case CBOR_TAG_VALUE_SHAREABLE:
		1195	{
		1196	if (ecb_expect_false (!dec->shareable))
		1197	dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
		1198
		1199	if (dec->cbor.flags & F_ALLOW_CYCLES)
		1200	{
		1201	// if cycles are allowed, then we store an AV as value
		1202	// while it is being decoded, and gather unresolved
		1203	// references in it, to be re4solved after decoding.
		1204	int idx, i;
		1205	AV *av = newAV ();
		1206	av_push (dec->shareable, (SV *)av);
		1207	idx = AvFILLp (dec->shareable);
		1208
		1209	sv = decode_sv (dec);
		1210
		1211	// the AV now contains \undef for all unresolved references,
		1212	// so we fix them up here.
		1213	for (i = 0; i <= AvFILLp (av); ++i)
		1214	SvRV_set (AvARRAY (av)[i], SvREFCNT_inc_NN (SvRV (sv)));
		1215
		1216	// now replace the AV by a reference to the completed value
		1217	SvREFCNT_dec_NN ((SV *)av);
		1218	AvARRAY (dec->shareable)[idx] = SvREFCNT_inc_NN (sv);
		1219	}
		1220	else
		1221	{
		1222	av_push (dec->shareable, &PL_sv_undef);
		1223	int idx = AvFILLp (dec->shareable);
		1224	sv = decode_sv (dec);
		1225	AvARRAY (dec->shareable)[idx] = SvREFCNT_inc_NN (sv);
		1226	}
		1227	}
		1228	break;
		1229
		1230	case CBOR_TAG_VALUE_SHAREDREF:
		1231	{
		1232	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1233	ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
		1234
		1235	UV idx = decode_uint (dec);
		1236
		1237	if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
		1238	ERR ("corrupted CBOR data (sharedref index out of bounds)");
		1239
		1240	sv = AvARRAY (dec->shareable)[idx];
		1241
		1242	// reference to cycle, we create a new \undef and use that, and also
		1243	// registerr it in the AV for later fixing
		1244	if (SvTYPE (sv) == SVt_PVAV)
		1245	{
		1246	AV *av = (AV *)sv;
		1247	sv = newRV_noinc (&PL_sv_undef);
		1248	av_push (av, sv);
		1249	}
		1250	else if (sv == &PL_sv_undef) // not yet decoded, but cycles not allowed
		1251	ERR ("cyclic CBOR data structure found, but allow_cycles is not enabled");
		1252	else // we decoded the object earlier, no cycle
		1253	sv = newSVsv (sv);
		1254	}
		1255	break;
		1256
		1257	case CBOR_TAG_PERL_OBJECT:
		1258	{
		1259	if (dec->cbor.flags & F_FORBID_OBJECTS)
		1260	goto filter;
		1261
		1262	sv = decode_sv (dec);
		1263
		1264	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		1265	ERR ("corrupted CBOR data (non-array perl object)");
		1266
		1267	AV *av = (AV *)SvRV (sv);
		1268	int len = av_len (av) + 1;
		1269	HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
		1270
		1271	if (!stash)
		1272	ERR ("cannot decode perl-object (package does not exist)");
		1273
		1274	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1275
		1276	if (!method)
		1277	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1278
		1279	dSP;
		1280
		1281	ENTER; SAVETMPS;
		1282	PUSHMARK (SP);
		1283	EXTEND (SP, len + 1);
		1284	// we re-bless the reference to get overload and other niceties right
		1285	PUSHs (*av_fetch (av, 0, 1));
		1286	PUSHs (sv_cbor);
		1287
		1288	int i;
		1289
		1290	for (i = 1; i < len; ++i)
		1291	PUSHs (*av_fetch (av, i, 1));
		1292
		1293	PUTBACK;
		1294	call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
		1295	SPAGAIN;
		1296
		1297	if (SvTRUE (ERRSV))
		1298	{
		1299	FREETMPS; LEAVE;
		1300	ERR_ERRSV;
		1301	}
		1302
		1303	SvREFCNT_dec_NN (sv);
		1304	sv = SvREFCNT_inc (POPs);
		1305
		1306	PUTBACK;
		1307
		1308	FREETMPS; LEAVE;
		1309	}
		1310	break;
		1311
		1312	default:
		1313	filter:
		1314	{
		1315	SV *tag_sv = newSVuv (tag);
		1316
		1317	sv = decode_sv (dec);
		1318
		1319	dSP;
		1320	ENTER; SAVETMPS;
		1321	PUSHMARK (SP);
		1322	EXTEND (SP, 2);
		1323	PUSHs (tag_sv);
		1324	PUSHs (sv);
		1325
		1326	PUTBACK;
		1327	int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
		1328	SPAGAIN;
		1329
		1330	if (SvTRUE (ERRSV))
		1331	{
		1332	SvREFCNT_dec_NN (tag_sv);
		1333	FREETMPS; LEAVE;
		1334	ERR_ERRSV;
		1335	}
		1336
		1337	if (count)
		1338	{
		1339	SvREFCNT_dec_NN (tag_sv);
		1340	SvREFCNT_dec_NN (sv);
		1341	sv = SvREFCNT_inc_NN (TOPs);
		1342	SP -= count;
		1343	}
		1344	else
		1345	{
		1346	AV *av = newAV ();
		1347	av_push (av, tag_sv);
		1348	av_push (av, sv);
		1349
		1350	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		1351	? cbor_tagged_stash
		1352	: gv_stashpv ("CBOR::XS::Tagged" , 1);
		1353	sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
		1354	}
		1355
		1356	PUTBACK;
		1357
		1358	FREETMPS; LEAVE;
		1359	}
		1360	break;
		1361	}
		1362
		1363	return sv;
		1364
		1365	fail:
		1366	SvREFCNT_dec (sv);
782	return &PL_sv_undef;	1367	return &PL_sv_undef;
783	}	1368	}
784		1369
785	static SV *	1370	static SV *
786	decode_sv (dec_t *dec)	1371	decode_sv (dec_t *dec)
787	{	1372	{
788	WANT (1);	1373	WANT (1);
789		1374
790	switch (*dec->cur >> 5)	1375	switch (*dec->cur >> MAJOR_SHIFT)
791	{	1376	{
792	case 0: // unsigned int	1377	case MAJOR_POS_INT >> MAJOR_SHIFT: return newSVuv (decode_uint (dec));
793	//TODO: 64 bit values on 3 2bit perls	1378	case MAJOR_NEG_INT >> MAJOR_SHIFT: return newSViv (-1 - (IV)decode_uint (dec));
794	return newSVuv (decode_uint (dec));	1379	case MAJOR_BYTES >> MAJOR_SHIFT: return decode_str (dec, 0);
795	case 1: // negative int	1380	case MAJOR_TEXT >> MAJOR_SHIFT: return decode_str (dec, 1);
796	return newSViv (-1 - (IV)decode_uint (dec));	1381	case MAJOR_ARRAY >> MAJOR_SHIFT: return decode_av (dec);
797	case 2: // octet string	1382	case MAJOR_MAP >> MAJOR_SHIFT: return decode_hv (dec);
798	return decode_str (dec, 0);	1383	case MAJOR_TAG >> MAJOR_SHIFT: return decode_tagged (dec);
799	case 3: // utf-8 string	1384
800	return decode_str (dec, 1);	1385	case MAJOR_MISC >> MAJOR_SHIFT:
801	case 4: // array
802	return decode_av (dec);
803	case 5: // map
804	return decode_hv (dec);
805	case 6: // tag
806	abort ();
807	break;
808	case 7: // misc
809	switch (*dec->cur++ & 31)	1386	switch (*dec->cur++ & MINOR_MASK)
810	{	1387	{
811	case 20:	1388	case SIMPLE_FALSE:
812	#if CBOR_SLOW	1389	#if CBOR_SLOW
813	cbor_false = get_bool ("CBOR::XS::false");	1390	types_false = get_bool ("Types::Serialiser::false");
814	#endif	1391	#endif
815	return newSVsv (cbor_false);	1392	return newSVsv (types_false);
816	case 21:	1393	case SIMPLE_TRUE:
817	#if CBOR_SLOW	1394	#if CBOR_SLOW
818	cbor_true = get_bool ("CBOR::XS::true");	1395	types_true = get_bool ("Types::Serialiser::true");
819	#endif	1396	#endif
820	return newSVsv (cbor_true);	1397	return newSVsv (types_true);
821	case 22:	1398	case SIMPLE_NULL:
822	return newSVsv (&PL_sv_undef);	1399	return newSVsv (&PL_sv_undef);
		1400	case SIMPLE_UNDEF:
		1401	#if CBOR_SLOW
		1402	types_error = get_bool ("Types::Serialiser::error");
		1403	#endif
		1404	return newSVsv (types_error);
823		1405
824	case 25:	1406	case MISC_FLOAT16:
825	// half float
826	abort ();
827	break;	1407	{
		1408	WANT (2);
828		1409
		1410	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
		1411	dec->cur += 2;
		1412
		1413	return newSVnv (ecb_binary16_to_float (fp));
		1414	}
		1415
829	case 26:	1416	case MISC_FLOAT32:
830	{	1417	{
831	uint32_t fp;	1418	uint32_t fp;
832	WANT (4);	1419	WANT (4);
833	memcpy (&fp, dec->cur, 4);	1420	memcpy (&fp, dec->cur, 4);
834	dec->cur += 4;	1421	dec->cur += 4;
…		…
837	fp = ecb_bswap32 (fp);	1424	fp = ecb_bswap32 (fp);
838		1425
839	return newSVnv (ecb_binary32_to_float (fp));	1426	return newSVnv (ecb_binary32_to_float (fp));
840	}	1427	}
841		1428
842	case 27:	1429	case MISC_FLOAT64:
843	{	1430	{
844	uint64_t fp;	1431	uint64_t fp;
845	WANT (8);	1432	WANT (8);
846	memcpy (&fp, dec->cur, 8);	1433	memcpy (&fp, dec->cur, 8);
847	dec->cur += 8;	1434	dec->cur += 8;
…		…
850	fp = ecb_bswap64 (fp);	1437	fp = ecb_bswap64 (fp);
851		1438
852	return newSVnv (ecb_binary64_to_double (fp));	1439	return newSVnv (ecb_binary64_to_double (fp));
853	}	1440	}
854		1441
855	// 0..19 unassigned	1442	// 0..19 unassigned simple
856	// 24 reserved + unassigned (reserved values are not encodable)	1443	// 24 reserved + unassigned simple (reserved values are not encodable)
		1444	// 28-30 unassigned misc
		1445	// 31 break code
857	default:	1446	default:
858	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");	1447	ERR ("corrupted CBOR data (reserved/unassigned/unexpected major 7 value)");
859	}	1448	}
860		1449
861	break;	1450	break;
862	}	1451	}
863	#if 0
864	switch (*dec->cur)
865	{
866	//case '"': ++dec->cur; return decode_str (dec);
867	case '[': ++dec->cur; return decode_av (dec);
868	case '{': ++dec->cur; return decode_hv (dec);
869
870	case '-':
871	case '0': case '1': case '2': case '3': case '4':
872	case '5': case '6': case '7': case '8': case '9':
873	//TODO return decode_num (dec);
874
875	case 't':
876	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
877	{
878	dec->cur += 4;
879	#if CBOR_SLOW
880	cbor_true = get_bool ("CBOR::XS::true");
881	#endif
882	return newSVsv (cbor_true);
883	}
884	else
885	ERR ("'true' expected");
886
887	break;
888
889	case 'f':
890	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
891	{
892	dec->cur += 5;
893	#if CBOR_SLOW
894	cbor_false = get_bool ("CBOR::XS::false");
895	#endif
896	return newSVsv (cbor_false);
897	}
898	else
899	ERR ("'false' expected");
900
901	break;
902
903	case 'n':
904	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
905	{
906	dec->cur += 4;
907	return newSVsv (&PL_sv_undef);
908	}
909	else
910	ERR ("'null' expected");
911
912	break;
913
914	default:
915	ERR ("malformed CBOR string, neither array, object, number, string or atom");
916	break;
917	}
918	#endif
919		1452
920	fail:	1453	fail:
921	return &PL_sv_undef;	1454	return &PL_sv_undef;
922	}	1455	}
923		1456
924	static SV *	1457	static SV *
925	decode_cbor (SV *string, CBOR *cbor, char **offset_return)	1458	decode_cbor (SV *string, CBOR *cbor, char **offset_return)
926	{	1459	{
927	dec_t dec;	1460	dec_t dec = { 0 };
928	SV *sv;	1461	SV *sv;
		1462	STRLEN len;
		1463	char *data = SvPVbyte (string, len);
929		1464
930	/* work around bugs in 5.10 where manipulating magic values
931	* makes perl ignore the magic in subsequent accesses.
932	* also make a copy of non-PV values, to get them into a clean
933	* state (SvPV should do that, but it's buggy, see below).
934	*/
935	/*SvGETMAGIC (string);*/
936	if (SvMAGICAL (string) || !SvPOK (string))
937	string = sv_2mortal (newSVsv (string));
938
939	SvUPGRADE (string, SVt_PV);
940
941	/* work around a bug in perl 5.10, which causes SvCUR to fail an
942	* assertion with -DDEBUGGING, although SvCUR is documented to
943	* return the xpv_cur field which certainly exists after upgrading.
944	* according to nicholas clark, calling SvPOK fixes this.
945	* But it doesn't fix it, so try another workaround, call SvPV_nolen
946	* and hope for the best.
947	* Damnit, SvPV_nolen still trips over yet another assertion. This
948	* assertion business is seriously broken, try yet another workaround
949	* for the broken -DDEBUGGING.
950	*/
951	{
952	#ifdef DEBUGGING
953	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
954	#else
955	STRLEN offset = SvCUR (string);
956	#endif
957
958	if (offset > cbor->max_size && cbor->max_size)	1465	if (len > cbor->max_size && cbor->max_size)
959	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",	1466	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
960	(unsigned long)SvCUR (string), (unsigned long)cbor->max_size);	1467	(unsigned long)len, (unsigned long)cbor->max_size);
961	}
962
963	sv_utf8_downgrade (string, 0);
964		1468
965	dec.cbor = *cbor;	1469	dec.cbor = *cbor;
966	dec.cur = (U8 *)SvPVX (string);	1470	dec.cur = (U8 *)data;
967	dec.end = (U8 *)SvEND (string);	1471	dec.end = (U8 *)data + len;
968	dec.err = 0;
969	dec.depth = 0;
970
971	if (dec.cbor.cb_object || dec.cbor.cb_sk_object)
972	;//TODO dec.cbor.flags |= F_HOOK;
973		1472
974	sv = decode_sv (&dec);	1473	sv = decode_sv (&dec);
975		1474
976	if (offset_return)	1475	if (offset_return)
977	*offset_return = dec.cur;	1476	*offset_return = dec.cur;
978		1477
979	if (!(offset_return || !sv))	1478	if (!(offset_return || !sv))
980	{
981	if (*dec.cur && !dec.err)	1479	if (dec.cur != dec.end && !dec.err)
982	{
983	dec.err = "garbage after CBOR object";	1480	dec.err = "garbage after CBOR object";
984	SvREFCNT_dec (sv);	1481
985	sv = 0;	1482	if (dec.err)
		1483	{
		1484	if (dec.shareable)
986	}	1485	{
987	}	1486	// need to break cyclic links, which would all be in shareable
		1487	int i;
		1488	SV **svp;
988		1489
989	if (!sv)	1490	for (i = av_len (dec.shareable) + 1; i--; )
		1491	if ((svp = av_fetch (dec.shareable, i, 0)))
		1492	sv_setsv (*svp, &PL_sv_undef);
		1493	}
		1494
		1495	SvREFCNT_dec_NN (sv);
		1496
		1497	if (dec.err_sv)
		1498	sv_2mortal (dec.err_sv);
		1499
990	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur);	1500	croak ("%s, at offset %ld (octet 0x%02x)", dec.err, (long)(dec.cur - (U8 *)data), (int)(uint8_t)*dec.cur);
		1501	}
991		1502
992	sv = sv_2mortal (sv);	1503	sv = sv_2mortal (sv);
993		1504
994	return sv;	1505	return sv;
995	}	1506	}
996		1507
		1508	/////////////////////////////////////////////////////////////////////////////
		1509	// incremental parser
		1510
		1511	#define INCR_DONE(cbor) (AvFILLp (cbor->incr_count) < 0)
		1512
		1513	// returns 0 for notyet, 1 for success or error
		1514	static int
		1515	incr_parse (CBOR *self, SV *cborstr)
		1516	{
		1517	STRLEN cur;
		1518	SvPV (cborstr, cur);
		1519
		1520	while (ecb_expect_true (self->incr_need <= cur))
		1521	{
		1522	// table of integer count bytes
		1523	static I8 incr_len[MINOR_MASK + 1] = {
		1524	0, 0, 0, 0, 0, 0, 0, 0,
		1525	0, 0, 0, 0, 0, 0, 0, 0,
		1526	0, 0, 0, 0, 0, 0, 0, 0,
		1527	1, 2, 4, 8,-1,-1,-1,-2
		1528	};
		1529
		1530	const U8 *p = SvPVX (cborstr) + self->incr_pos;
		1531	U8 m = *p & MINOR_MASK;
		1532	IV count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1533	I8 ilen = incr_len[m];
		1534
		1535	self->incr_need = self->incr_pos + 1;
		1536
		1537	if (ecb_expect_false (ilen < 0))
		1538	{
		1539	if (m != MINOR_INDEF)
		1540	return 1; // error
		1541
		1542	if (*p == (MAJOR_MISC | MINOR_INDEF))
		1543	{
		1544	if (count >= 0)
		1545	return 1; // error
		1546
		1547	count = 1;
		1548	}
		1549	else
		1550	{
		1551	av_push (self->incr_count, newSViv (-1)); //TODO: nest
		1552	count = -1;
		1553	}
		1554	}
		1555	else
		1556	{
		1557	self->incr_need += ilen;
		1558	if (ecb_expect_false (self->incr_need > cur))
		1559	return 0;
		1560
		1561	int major = *p >> MAJOR_SHIFT;
		1562
		1563	switch (major)
		1564	{
		1565	case MAJOR_TAG >> MAJOR_SHIFT:
		1566	++count; // tags merely prefix another value
		1567	break;
		1568
		1569	case MAJOR_BYTES >> MAJOR_SHIFT:
		1570	case MAJOR_TEXT >> MAJOR_SHIFT:
		1571	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1572	case MAJOR_MAP >> MAJOR_SHIFT:
		1573	{
		1574	UV len;
		1575
		1576	if (ecb_expect_false (ilen))
		1577	{
		1578	len = 0;
		1579
		1580	do {
		1581	len = (len << 8) | *++p;
		1582	} while (--ilen);
		1583	}
		1584	else
		1585	len = m;
		1586
		1587	switch (major)
		1588	{
		1589	case MAJOR_BYTES >> MAJOR_SHIFT:
		1590	case MAJOR_TEXT >> MAJOR_SHIFT:
		1591	self->incr_need += len;
		1592	if (ecb_expect_false (self->incr_need > cur))
		1593	return 0;
		1594
		1595	break;
		1596
		1597	case MAJOR_MAP >> MAJOR_SHIFT:
		1598	len <<= 1;
		1599	/* FALLTHROUGH */
		1600	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1601	if (len)
		1602	{
		1603	av_push (self->incr_count, newSViv (len + 1)); //TODO: nest
		1604	count = len + 1;
		1605	}
		1606	break;
		1607	}
		1608	}
		1609	}
		1610	}
		1611
		1612	self->incr_pos = self->incr_need;
		1613
		1614	if (count > 0)
		1615	{
		1616	while (!--count)
		1617	{
		1618	if (!AvFILLp (self->incr_count))
		1619	return 1; // done
		1620
		1621	SvREFCNT_dec_NN (av_pop (self->incr_count));
		1622	count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1623	}
		1624
		1625	SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]) = count;
		1626	}
		1627	}
		1628
		1629	return 0;
		1630	}
		1631
		1632
997	/////////////////////////////////////////////////////////////////////////////	1633	/////////////////////////////////////////////////////////////////////////////
998	// XS interface functions	1634	// XS interface functions
999		1635
1000	MODULE = CBOR::XS PACKAGE = CBOR::XS	1636	MODULE = CBOR::XS PACKAGE = CBOR::XS
1001		1637
1002	BOOT:	1638	BOOT:
1003	{	1639	{
1004	cbor_stash = gv_stashpv ("CBOR::XS" , 1);	1640	cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1005	cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1);	1641	cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1006		1642
1007	cbor_true = get_bool ("CBOR::XS::true");	1643	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1008	cbor_false = get_bool ("CBOR::XS::false");	1644	types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
		1645
		1646	types_true = get_bool ("Types::Serialiser::true" );
		1647	types_false = get_bool ("Types::Serialiser::false");
		1648	types_error = get_bool ("Types::Serialiser::error");
		1649
		1650	default_filter = newSVpv ("CBOR::XS::default_filter", 0);
		1651
		1652	sv_cbor = newSVpv ("CBOR", 0);
		1653	SvREADONLY_on (sv_cbor);
		1654
		1655	assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1009	}	1656	}
1010		1657
1011	PROTOTYPES: DISABLE	1658	PROTOTYPES: DISABLE
1012		1659
1013	void CLONE (...)	1660	void CLONE (...)
1014	CODE:	1661	CODE:
1015	cbor_stash = 0;	1662	cbor_stash = 0;
		1663	cbor_tagged_stash = 0;
		1664	types_error_stash = 0;
1016	cbor_boolean_stash = 0;	1665	types_boolean_stash = 0;
1017		1666
1018	void new (char *klass)	1667	void new (char *klass)
1019	PPCODE:	1668	PPCODE:
1020	{	1669	{
1021	SV *pv = NEWSV (0, sizeof (CBOR));	1670	SV *pv = NEWSV (0, sizeof (CBOR));
…		…
1029		1678
1030	void shrink (CBOR *self, int enable = 1)	1679	void shrink (CBOR *self, int enable = 1)
1031	ALIAS:	1680	ALIAS:
1032	shrink = F_SHRINK	1681	shrink = F_SHRINK
1033	allow_unknown = F_ALLOW_UNKNOWN	1682	allow_unknown = F_ALLOW_UNKNOWN
		1683	allow_sharing = F_ALLOW_SHARING
		1684	allow_cycles = F_ALLOW_CYCLES
		1685	forbid_objects = F_FORBID_OBJECTS
		1686	pack_strings = F_PACK_STRINGS
		1687	text_keys = F_TEXT_KEYS
		1688	text_strings = F_TEXT_STRINGS
		1689	validate_utf8 = F_VALIDATE_UTF8
1034	PPCODE:	1690	PPCODE:
1035	{	1691	{
1036	if (enable)	1692	if (enable)
1037	self->flags |= ix;	1693	self->flags |= ix;
1038	else	1694	else
…		…
1043		1699
1044	void get_shrink (CBOR *self)	1700	void get_shrink (CBOR *self)
1045	ALIAS:	1701	ALIAS:
1046	get_shrink = F_SHRINK	1702	get_shrink = F_SHRINK
1047	get_allow_unknown = F_ALLOW_UNKNOWN	1703	get_allow_unknown = F_ALLOW_UNKNOWN
		1704	get_allow_sharing = F_ALLOW_SHARING
		1705	get_allow_cycles = F_ALLOW_CYCLES
		1706	get_forbid_objects = F_FORBID_OBJECTS
		1707	get_pack_strings = F_PACK_STRINGS
		1708	get_text_keys = F_TEXT_KEYS
		1709	get_text_strings = F_TEXT_STRINGS
		1710	get_validate_utf8 = F_VALIDATE_UTF8
1048	PPCODE:	1711	PPCODE:
1049	XPUSHs (boolSV (self->flags & ix));	1712	XPUSHs (boolSV (self->flags & ix));
1050		1713
1051	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)	1714	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1052	PPCODE:	1715	PPCODE:
…		…
1068	CODE:	1731	CODE:
1069	RETVAL = self->max_size;	1732	RETVAL = self->max_size;
1070	OUTPUT:	1733	OUTPUT:
1071	RETVAL	1734	RETVAL
1072		1735
1073	#if 0 //TODO	1736	void filter (CBOR *self, SV *filter = 0)
1074
1075	void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1076	PPCODE:	1737	PPCODE:
1077	{
1078	SvREFCNT_dec (self->cb_object);	1738	SvREFCNT_dec (self->filter);
1079	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;	1739	self->filter = filter ? newSVsv (filter) : filter;
1080
1081	XPUSHs (ST (0));	1740	XPUSHs (ST (0));
1082	}
1083		1741
1084	void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)	1742	SV *get_filter (CBOR *self)
1085	PPCODE:	1743	CODE:
1086	{	1744	RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1087	if (!self->cb_sk_object)	1745	OUTPUT:
1088	self->cb_sk_object = newHV ();	1746	RETVAL
1089
1090	if (SvOK (cb))
1091	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1092	else
1093	{
1094	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1095
1096	if (!HvKEYS (self->cb_sk_object))
1097	{
1098	SvREFCNT_dec (self->cb_sk_object);
1099	self->cb_sk_object = 0;
1100	}
1101	}
1102
1103	XPUSHs (ST (0));
1104	}
1105
1106	#endif
1107		1747
1108	void encode (CBOR *self, SV *scalar)	1748	void encode (CBOR *self, SV *scalar)
1109	PPCODE:	1749	PPCODE:
1110	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;	1750	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1111	XPUSHs (scalar);	1751	XPUSHs (scalar);
…		…
1124	EXTEND (SP, 2);	1764	EXTEND (SP, 2);
1125	PUSHs (sv);	1765	PUSHs (sv);
1126	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));	1766	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1127	}	1767	}
1128		1768
		1769	void incr_parse (CBOR *self, SV *cborstr)
		1770	ALIAS:
		1771	incr_parse_multiple = 1
		1772	PPCODE:
		1773	{
		1774	if (SvUTF8 (cborstr))
		1775	sv_utf8_downgrade (cborstr, 0);
		1776
		1777	if (!self->incr_count)
		1778	{
		1779	self->incr_count = newAV ();
		1780	self->incr_pos = 0;
		1781	self->incr_need = 1;
		1782
		1783	av_push (self->incr_count, newSViv (1));
		1784	}
		1785
		1786	do
		1787	{
		1788	if (!incr_parse (self, cborstr))
		1789	{
		1790	if (self->incr_need > self->max_size && self->max_size)
		1791	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
		1792	(unsigned long)self->incr_need, (unsigned long)self->max_size);
		1793
		1794	break;
		1795	}
		1796
		1797	SV *sv;
		1798	char *offset;
		1799
		1800	PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
		1801	XPUSHs (sv);
		1802
		1803	sv_chop (cborstr, offset);
		1804
		1805	av_clear (self->incr_count);
		1806	av_push (self->incr_count, newSViv (1));
		1807
		1808	self->incr_pos = 0;
		1809	self->incr_need = self->incr_pos + 1;
		1810	}
		1811	while (ix);
		1812	}
		1813
		1814	void incr_reset (CBOR *self)
		1815	CODE:
		1816	{
		1817	SvREFCNT_dec (self->incr_count);
		1818	self->incr_count = 0;
		1819	}
		1820
1129	void DESTROY (CBOR *self)	1821	void DESTROY (CBOR *self)
1130	CODE:	1822	PPCODE:
1131	SvREFCNT_dec (self->cb_sk_object);	1823	cbor_free (self);
1132	SvREFCNT_dec (self->cb_object);
1133		1824
1134	PROTOTYPES: ENABLE	1825	PROTOTYPES: ENABLE
1135		1826
1136	void encode_cbor (SV *scalar)	1827	void encode_cbor (SV *scalar)
		1828	ALIAS:
		1829	encode_cbor = 0
		1830	encode_cbor_sharing = F_ALLOW_SHARING
1137	PPCODE:	1831	PPCODE:
1138	{	1832	{
1139	CBOR cbor;	1833	CBOR cbor;
1140	cbor_init (&cbor);	1834	cbor_init (&cbor);
		1835	cbor.flags |= ix;
1141	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;	1836	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1142	XPUSHs (scalar);	1837	XPUSHs (scalar);
1143	}	1838	}
1144		1839
1145	void decode_cbor (SV *cborstr)	1840	void decode_cbor (SV *cborstr)
…		…
1149	cbor_init (&cbor);	1844	cbor_init (&cbor);
1150	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;	1845	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1151	XPUSHs (cborstr);	1846	XPUSHs (cborstr);
1152	}	1847	}
1153		1848
		1849	#ifdef __AFL_COMPILER
		1850
		1851	void
		1852	afl_init ()
		1853	CODE:
		1854	__AFL_INIT ();
		1855
		1856	int
		1857	afl_loop (unsigned int count = 10000)
		1858	CODE:
		1859	RETVAL = __AFL_LOOP (count);
		1860	OUTPUT:
		1861	RETVAL
		1862
		1863	#endif
		1864

Diff Legend

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