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 1.70 by root, Fri Dec 4 02:57:14 2020 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	// known major and minor types
		34	enum cbor_type
		35	{
		36	MAJOR_SHIFT = 5,
		37	MINOR_MASK = 0x1f,
		38
		39	MAJOR_POS_INT = 0 << MAJOR_SHIFT,
		40	MAJOR_NEG_INT = 1 << MAJOR_SHIFT,
		41	MAJOR_BYTES = 2 << MAJOR_SHIFT,
		42	MAJOR_TEXT = 3 << MAJOR_SHIFT,
		43	MAJOR_ARRAY = 4 << MAJOR_SHIFT,
		44	MAJOR_MAP = 5 << MAJOR_SHIFT,
		45	MAJOR_TAG = 6 << MAJOR_SHIFT,
		46	MAJOR_MISC = 7 << MAJOR_SHIFT,
		47
		48	// INT/STRING/ARRAY/MAP subtypes
		49	LENGTH_EXT1 = 24,
		50	LENGTH_EXT2 = 25,
		51	LENGTH_EXT4 = 26,
		52	LENGTH_EXT8 = 27,
		53
		54	// SIMPLE types (effectively MISC subtypes)
		55	SIMPLE_FALSE = 20,
		56	SIMPLE_TRUE = 21,
		57	SIMPLE_NULL = 22,
		58	SIMPLE_UNDEF = 23,
		59
		60	// MISC subtype (unused)
		61	MISC_EXT1 = 24,
		62	MISC_FLOAT16 = 25,
		63	MISC_FLOAT32 = 26,
		64	MISC_FLOAT64 = 27,
		65
		66	// BYTES/TEXT/ARRAY/MAP
		67	MINOR_INDEF = 31,
		68	};
		69
		70	// known tags
		71	enum cbor_tag
		72	{
		73	// extensions
		74	CBOR_TAG_STRINGREF = 25, // http://cbor.schmorp.de/stringref
		75	CBOR_TAG_PERL_OBJECT = 26, // http://cbor.schmorp.de/perl-object
		76	CBOR_TAG_GENERIC_OBJECT = 27, // http://cbor.schmorp.de/generic-object
		77	CBOR_TAG_VALUE_SHAREABLE = 28, // http://cbor.schmorp.de/value-sharing
		78	CBOR_TAG_VALUE_SHAREDREF = 29, // http://cbor.schmorp.de/value-sharing
		79	CBOR_TAG_STRINGREF_NAMESPACE = 256, // http://cbor.schmorp.de/stringref
		80	CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
		81
		82	// rfc7049
		83	CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
		84	CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
		85	CBOR_TAG_POS_BIGNUM = 2, // byte string
		86	CBOR_TAG_NEG_BIGNUM = 3, // byte string
		87	CBOR_TAG_DECIMAL = 4, // decimal fraction, array
		88	CBOR_TAG_BIGFLOAT = 5, // array
		89
		90	CBOR_TAG_CONV_B64U = 21, // base64url, any
		91	CBOR_TAG_CONV_B64 = 22, // base64, any
		92	CBOR_TAG_CONV_HEX = 23, // base16, any
		93	CBOR_TAG_CBOR = 24, // embedded cbor, byte string
		94
		95	CBOR_TAG_URI = 32, // URI rfc3986, utf-8
		96	CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
		97	CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
		98	CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
		99	CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
		100
		101	CBOR_TAG_MAGIC = 55799, // self-describe cbor
		102	};
		103
		104	// known forced types, also hardcoded in CBOR.pm
		105	enum
		106	{
		107	AS_CBOR = 0,
		108	AS_INT = 1,
		109	AS_BYTES = 2,
		110	AS_TEXT = 3,
		111	AS_FLOAT16 = 4,
		112	AS_FLOAT32 = 5,
		113	AS_FLOAT64 = 6,
		114	AS_MAP = 7,
		115	// possibly future enhancements: (generic) float, (generic) string
		116	};
		117
14	#define F_SHRINK 0x00000200UL	118	#define F_SHRINK 0x00000001UL
15	#define F_ALLOW_UNKNOWN 0x00002000UL	119	#define F_ALLOW_UNKNOWN 0x00000002UL
		120	#define F_ALLOW_SHARING 0x00000004UL
		121	#define F_ALLOW_CYCLES 0x00000008UL
		122	#define F_FORBID_OBJECTS 0x00000010UL
		123	#define F_PACK_STRINGS 0x00000020UL
		124	#define F_TEXT_KEYS 0x00000040UL
		125	#define F_TEXT_STRINGS 0x00000080UL
		126	#define F_VALIDATE_UTF8 0x00000100UL
16		127
17	#define INIT_SIZE 32 // initial scalar size to be allocated	128	#define INIT_SIZE 32 // initial scalar size to be allocated
18		129
19	#define SB do {	130	#define SB do {
20	#define SE } while (0)	131	#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		132
33	#define IN_RANGE_INC(type,val,beg,end) \	133	#define IN_RANGE_INC(type,val,beg,end) \
34	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \	134	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
35	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))	135	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
36		136
…		…
42	#else	142	#else
43	# define CBOR_SLOW 0	143	# define CBOR_SLOW 0
44	# define CBOR_STASH cbor_stash	144	# define CBOR_STASH cbor_stash
45	#endif	145	#endif
46		146
47	static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS::	147	static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
48	static SV *cbor_true, *cbor_false;	148	static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
49		149
50	typedef struct {	150	typedef struct {
51	U32 flags;	151	U32 flags;
52	U32 max_depth;	152	U32 max_depth;
53	STRLEN max_size;	153	STRLEN max_size;
		154	SV *filter;
54		155
55	SV *cb_object;	156	// for the incremental parser
56	HV *cb_sk_object;	157	STRLEN incr_pos; // the current offset into the text
		158	STRLEN incr_need; // minimum bytes needed to decode
		159	AV *incr_count; // for every nesting level, the number of outstanding values, or -1 for indef.
57	} CBOR;	160	} CBOR;
58		161
59	INLINE void	162	ecb_inline void
60	cbor_init (CBOR *cbor)	163	cbor_init (CBOR *cbor)
61	{	164	{
62	Zero (cbor, 1, CBOR);	165	Zero (cbor, 1, CBOR);
63	cbor->max_depth = 512;	166	cbor->max_depth = 512;
64	}	167	}
65		168
		169	ecb_inline void
		170	cbor_free (CBOR *cbor)
		171	{
		172	SvREFCNT_dec (cbor->filter);
		173	SvREFCNT_dec (cbor->incr_count);
		174	}
		175
66	/////////////////////////////////////////////////////////////////////////////	176	/////////////////////////////////////////////////////////////////////////////
67	// utility functions	177	// utility functions
68		178
69	INLINE SV *	179	ecb_inline SV *
70	get_bool (const char *name)	180	get_bool (const char *name)
71	{	181	{
72	SV *sv = get_sv (name, 1);	182	SV *sv = get_sv (name, 1);
73		183
74	SvREADONLY_on (sv);	184	SvREADONLY_on (sv);
75	SvREADONLY_on (SvRV (sv));	185	SvREADONLY_on (SvRV (sv));
76		186
77	return sv;	187	return sv;
78	}	188	}
79		189
80	INLINE void	190	ecb_inline void
81	shrink (SV *sv)	191	shrink (SV *sv)
82	{	192	{
83	sv_utf8_downgrade (sv, 1);	193	sv_utf8_downgrade (sv, 1);
84		194
85	if (SvLEN (sv) > SvCUR (sv) + 1)	195	if (SvLEN (sv) > SvCUR (sv) + 1)
…		…
90	SvPV_renew (sv, SvCUR (sv) + 1);	200	SvPV_renew (sv, SvCUR (sv) + 1);
91	#endif	201	#endif
92	}	202	}
93	}	203	}
94		204
95	/////////////////////////////////////////////////////////////////////////////	205	// minimum length of a string to be registered for stringref
96	// fp hell	206	ecb_inline int
97		207	minimum_string_length (UV idx)
98	//TODO	208	{
		209	return idx <= 23 ? 3
		210	: idx <= 0xffU ? 4
		211	: idx <= 0xffffU ? 5
		212	: idx <= 0xffffffffU ? 7
		213	: 11;
		214	}
99		215
100	/////////////////////////////////////////////////////////////////////////////	216	/////////////////////////////////////////////////////////////////////////////
101	// encoder	217	// encoder
102		218
103	// structure used for encoding CBOR	219	// structure used for encoding CBOR
…		…
106	char *cur; // SvPVX (sv) + current output position	222	char *cur; // SvPVX (sv) + current output position
107	char *end; // SvEND (sv)	223	char *end; // SvEND (sv)
108	SV *sv; // result scalar	224	SV *sv; // result scalar
109	CBOR cbor;	225	CBOR cbor;
110	U32 depth; // recursion level	226	U32 depth; // recursion level
		227	HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
		228	UV stringref_idx;
		229	HV *shareable; // ptr => index, or 0
		230	UV shareable_idx;
111	} enc_t;	231	} enc_t;
112		232
113	INLINE void	233	ecb_inline void
114	need (enc_t *enc, STRLEN len)	234	need (enc_t *enc, STRLEN len)
115	{	235	{
116	if (expect_false (enc->cur + len >= enc->end))	236	if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
117	{	237	{
118	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);	238	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
119	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);	239	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
120	enc->cur = SvPVX (enc->sv) + cur;	240	enc->cur = SvPVX (enc->sv) + cur;
121	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	241	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
122	}	242	}
123	}	243	}
124		244
125	INLINE void	245	static void encode_sv (enc_t *enc, SV *sv);
		246
		247	ecb_inline void
126	encode_ch (enc_t *enc, char ch)	248	encode_ch (enc_t *enc, char ch)
127	{	249	{
128	need (enc, 1);	250	need (enc, 1);
129	*enc->cur++ = ch;	251	*enc->cur++ = ch;
130	}	252	}
131		253
		254	// used for tags, intregers, element counts and so on
132	static void	255	static void
133	encode_uint (enc_t *enc, int major, UV len)	256	encode_uint (enc_t *enc, int major, UV len)
134	{	257	{
135	need (enc, 9);	258	need (enc, 9);
136		259
137	if (len < 24)	260	if (ecb_expect_true (len < LENGTH_EXT1))
138	*enc->cur++ = major | len;	261	*enc->cur++ = major | len;
139	else if (len < 0x100)	262	else if (ecb_expect_true (len <= 0xffU))
140	{	263	{
141	*enc->cur++ = major | 24;	264	*enc->cur++ = major | LENGTH_EXT1;
142	*enc->cur++ = len;	265	*enc->cur++ = len;
143	}	266	}
144	else if (len < 0x10000)	267	else if (len <= 0xffffU)
145	{	268	{
146	*enc->cur++ = major | 25;	269	*enc->cur++ = major | LENGTH_EXT2;
147	*enc->cur++ = len >> 8;	270	*enc->cur++ = len >> 8;
148	*enc->cur++ = len;	271	*enc->cur++ = len;
149	}	272	}
150	else if (len < 0x100000000)	273	else if (len <= 0xffffffffU)
151	{	274	{
152	*enc->cur++ = major | 26;	275	*enc->cur++ = major | LENGTH_EXT4;
153	*enc->cur++ = len >> 24;	276	*enc->cur++ = len >> 24;
154	*enc->cur++ = len >> 16;	277	*enc->cur++ = len >> 16;
155	*enc->cur++ = len >> 8;	278	*enc->cur++ = len >> 8;
156	*enc->cur++ = len;	279	*enc->cur++ = len;
157	}	280	}
158	else if (len)	281	else
159	{	282	{
160	*enc->cur++ = major | 27;	283	*enc->cur++ = major | LENGTH_EXT8;
161	*enc->cur++ = len >> 56;	284	*enc->cur++ = len >> 56;
162	*enc->cur++ = len >> 48;	285	*enc->cur++ = len >> 48;
163	*enc->cur++ = len >> 40;	286	*enc->cur++ = len >> 40;
164	*enc->cur++ = len >> 32;	287	*enc->cur++ = len >> 32;
165	*enc->cur++ = len >> 24;	288	*enc->cur++ = len >> 24;
…		…
167	*enc->cur++ = len >> 8;	290	*enc->cur++ = len >> 8;
168	*enc->cur++ = len;	291	*enc->cur++ = len;
169	}	292	}
170	}	293	}
171		294
172	static void	295	// encodes a perl value into a CBOR integer
		296	ecb_inline void
		297	encode_int (enc_t *enc, SV *sv)
		298	{
		299	if (SvIsUV (sv))
		300	encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
		301	else if (SvIVX (sv) >= 0)
		302	encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
		303	else
		304	encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
		305	}
		306
		307	ecb_inline void
		308	encode_tag (enc_t *enc, UV tag)
		309	{
		310	encode_uint (enc, MAJOR_TAG, tag);
		311	}
		312
		313	// exceptional (hopefully) slow path for byte strings that need to be utf8-encoded
		314	ecb_noinline static void
173	encode_str (enc_t *enc, int utf8, char *str, STRLEN len)	315	encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
174	{	316	{
175	encode_uint (enc, utf8 ? 0x60 : 0x40, len);	317	STRLEN ulen = len;
		318	U8 *p, *pend = (U8 *)str + len;
		319
		320	for (p = (U8 *)str; p < pend; ++p)
		321	ulen += *p >> 7; // count set high bits
		322
		323	encode_uint (enc, MAJOR_TEXT, ulen);
		324
		325	need (enc, ulen);
		326	for (p = (U8 *)str; p < pend; ++p)
		327	if (*p < 0x80)
		328	*enc->cur++ = *p;
		329	else
		330	{
		331	*enc->cur++ = 0xc0 + (*p >> 6);
		332	*enc->cur++ = 0x80 + (*p & 63);
		333	}
		334	}
		335
		336	ecb_inline void
		337	encode_str (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		338	{
		339	if (ecb_expect_false (upgrade_utf8))
		340	if (!utf8)
		341	{
		342	encode_str_utf8 (enc, utf8, str, len);
		343	return;
		344	}
		345
		346	encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len);
176	need (enc, len);	347	need (enc, len);
177	memcpy (enc->cur, str, len);	348	memcpy (enc->cur, str, len);
178	enc->cur += len;	349	enc->cur += len;
179	}	350	}
180		351
181	static void encode_sv (enc_t *enc, SV *sv);	352	ecb_inline void
		353	encode_strref (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		354	{
		355	if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS))
		356	{
		357	SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
		358
		359	if (SvOK (*svp))
		360	{
		361	// already registered, use stringref
		362	encode_tag (enc, CBOR_TAG_STRINGREF);
		363	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		364	return;
		365	}
		366	else if (len >= minimum_string_length (enc->stringref_idx))
		367	{
		368	// register only
		369	sv_setuv (*svp, enc->stringref_idx);
		370	++enc->stringref_idx;
		371	}
		372	}
		373
		374	encode_str (enc, upgrade_utf8, utf8, str, len);
		375	}
		376
		377	ecb_inline void
		378	encode_float16 (enc_t *enc, NV nv)
		379	{
		380	need (enc, 1+2);
		381
		382	*enc->cur++ = MAJOR_MISC | MISC_FLOAT16;
		383
		384	uint16_t fp = ecb_float_to_binary16 (nv);
		385
		386	if (!ecb_big_endian ())
		387	fp = ecb_bswap16 (fp);
		388
		389	memcpy (enc->cur, &fp, 2);
		390	enc->cur += 2;
		391	}
		392
		393	ecb_inline void
		394	encode_float32 (enc_t *enc, NV nv)
		395	{
		396	need (enc, 1+4);
		397
		398	*enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
		399
		400	uint32_t fp = ecb_float_to_binary32 (nv);
		401
		402	if (!ecb_big_endian ())
		403	fp = ecb_bswap32 (fp);
		404
		405	memcpy (enc->cur, &fp, 4);
		406	enc->cur += 4;
		407	}
		408
		409	ecb_inline void
		410	encode_float64 (enc_t *enc, NV nv)
		411	{
		412	need (enc, 1+8);
		413
		414	*enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
		415
		416	uint64_t fp = ecb_double_to_binary64 (nv);
		417
		418	if (!ecb_big_endian ())
		419	fp = ecb_bswap64 (fp);
		420
		421	memcpy (enc->cur, &fp, 8);
		422	enc->cur += 8;
		423	}
		424
		425	ecb_inline void
		426	encode_bool (enc_t *enc, int istrue)
		427	{
		428	encode_ch (enc, istrue ? MAJOR_MISC | SIMPLE_TRUE : MAJOR_MISC | SIMPLE_FALSE);
		429	}
		430
		431	// encodes an arrayref containing key-value pairs as CBOR map
		432	ecb_inline void
		433	encode_array_as_map (enc_t *enc, SV *sv)
		434	{
		435	if (enc->depth >= enc->cbor.max_depth)
		436	croak (ERR_NESTING_EXCEEDED);
		437
		438	++enc->depth;
		439
		440	// as_map does error checking for us, but we re-check in case
		441	// things have changed.
		442
		443	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		444	croak ("CBOR::XS::as_map requires an array reference (did you change the array after calling as_map?)");
		445
		446	AV *av = (AV *)SvRV (sv);
		447	int i, len = av_len (av);
		448
		449	if (!(len & 1))
		450	croak ("CBOR::XS::as_map requires an even number of elements (did you change the array after calling as_map?)");
		451
		452	encode_uint (enc, MAJOR_MAP, (len + 1) >> 1);
		453
		454	for (i = 0; i <= len; ++i)
		455	{
		456	SV **svp = av_fetch (av, i, 0);
		457	encode_sv (enc, svp ? *svp : &PL_sv_undef);
		458	}
		459
		460	--enc->depth;
		461	}
		462
		463	ecb_inline void
		464	encode_forced (enc_t *enc, UV type, SV *sv)
		465	{
		466	switch (type)
		467	{
		468	case AS_CBOR:
		469	{
		470	STRLEN len;
		471	char *str = SvPVbyte (sv, len);
		472
		473	need (enc, len);
		474	memcpy (enc->cur, str, len);
		475	enc->cur += len;
		476	}
		477	break;
		478
		479	case AS_BYTES:
		480	{
		481	STRLEN len;
		482	char *str = SvPVbyte (sv, len);
		483	encode_strref (enc, 0, 0, str, len);
		484	}
		485	break;
		486
		487	case AS_TEXT:
		488	{
		489	STRLEN len;
		490	char *str = SvPVutf8 (sv, len);
		491	encode_strref (enc, 1, 1, str, len);
		492	}
		493	break;
		494
		495	case AS_INT: encode_int (enc, sv); break;
		496
		497	case AS_FLOAT16: encode_float16 (enc, SvNV (sv)); break;
		498	case AS_FLOAT32: encode_float32 (enc, SvNV (sv)); break;
		499	case AS_FLOAT64: encode_float64 (enc, SvNV (sv)); break;
		500
		501	case AS_MAP: encode_array_as_map (enc, sv); break;
		502
		503	default:
		504	croak ("encountered malformed CBOR::XS::Tagged object");
		505	}
		506	}
182		507
183	static void	508	static void
184	encode_av (enc_t *enc, AV *av)	509	encode_av (enc_t *enc, AV *av)
185	{	510	{
186	int i, len = av_len (av);	511	int i, len = av_len (av);
…		…
188	if (enc->depth >= enc->cbor.max_depth)	513	if (enc->depth >= enc->cbor.max_depth)
189	croak (ERR_NESTING_EXCEEDED);	514	croak (ERR_NESTING_EXCEEDED);
190		515
191	++enc->depth;	516	++enc->depth;
192		517
193	encode_uint (enc, 0x80, len + 1);	518	encode_uint (enc, MAJOR_ARRAY, len + 1);
194		519
		520	if (ecb_expect_false (SvMAGICAL (av)))
195	for (i = 0; i <= len; ++i)	521	for (i = 0; i <= len; ++i)
196	{	522	{
197	SV **svp = av_fetch (av, i, 0);	523	SV **svp = av_fetch (av, i, 0);
198	encode_sv (enc, svp ? *svp : &PL_sv_undef);	524	encode_sv (enc, svp ? *svp : &PL_sv_undef);
199	}	525	}
		526	else
		527	for (i = 0; i <= len; ++i)
		528	{
		529	SV *sv = AvARRAY (av)[i];
		530	encode_sv (enc, sv ? sv : &PL_sv_undef);
		531	}
200		532
201	--enc->depth;	533	--enc->depth;
202	}	534	}
203		535
204	static void	536	static void
…		…
212	++enc->depth;	544	++enc->depth;
213		545
214	int pairs = hv_iterinit (hv);	546	int pairs = hv_iterinit (hv);
215	int mg = SvMAGICAL (hv);	547	int mg = SvMAGICAL (hv);
216		548
217	if (mg)	549	if (ecb_expect_false (mg))
218	encode_ch (enc, 0xa0 | 31);	550	encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
219	else	551	else
220	encode_uint (enc, 0xa0, pairs);	552	encode_uint (enc, MAJOR_MAP, pairs);
221		553
222	while ((he = hv_iternext (hv)))	554	while ((he = hv_iternext (hv)))
223	{	555	{
224	if (HeKLEN (he) == HEf_SVKEY)	556	if (HeKLEN (he) == HEf_SVKEY)
225	encode_sv (enc, HeSVKEY (he));	557	encode_sv (enc, HeSVKEY (he));
226	else	558	else
227	encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));	559	encode_strref (enc, enc->cbor.flags & (F_TEXT_KEYS | F_TEXT_STRINGS), HeKUTF8 (he), HeKEY (he), HeKLEN (he));
228		560
229	encode_sv (enc, expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));	561	encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
230	}	562	}
231		563
232	if (mg)	564	if (ecb_expect_false (mg))
233	encode_ch (enc, 0xe0 | 31);	565	encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
234		566
235	--enc->depth;	567	--enc->depth;
236	}	568	}
237		569
238	// encode objects, arrays and special \0=false and \1=true values.	570	// encode objects, arrays and special \0=false and \1=true values.
239	static void	571	static void
240	encode_rv (enc_t *enc, SV *sv)	572	encode_rv (enc_t *enc, SV *sv)
241	{	573	{
242	svtype svt;
243
244	SvGETMAGIC (sv);	574	SvGETMAGIC (sv);
		575
245	svt = SvTYPE (sv);	576	svtype svt = SvTYPE (sv);
246		577
247	if (expect_false (SvOBJECT (sv)))	578	if (ecb_expect_false (SvOBJECT (sv)))
248	{	579	{
249	HV *stash = !CBOR_SLOW || cbor_boolean_stash	580	HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
250	? cbor_boolean_stash	581	? types_boolean_stash
		582	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		583	HV *error_stash = !CBOR_SLOW || types_error_stash
		584	? types_error_stash
		585	: gv_stashpv ("Types::Serialiser::Error", 1);
		586	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		587	? cbor_tagged_stash
251	: gv_stashpv ("CBOR::XS::Boolean", 1);	588	: gv_stashpv ("CBOR::XS::Tagged" , 1);
252		589
253	if (SvSTASH (sv) == stash)	590	HV *stash = SvSTASH (sv);
254	encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);	591
		592	if (stash == boolean_stash)
		593	{
		594	encode_bool (enc, SvIV (sv));
		595	return;
		596	}
		597	else if (stash == error_stash)
		598	{
		599	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
		600	return;
		601	}
		602	else if (stash == tagged_stash)
		603	{
		604	if (svt != SVt_PVAV)
		605	croak ("encountered CBOR::XS::Tagged object that isn't an array");
		606
		607	switch (av_len ((AV *)sv))
		608	{
		609	case 2-1:
		610	// actually a tagged value
		611	encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1)));
		612	encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
		613	break;
		614
		615	case 3-1:
		616	// a forced type [value, type, undef]
		617	encode_forced (enc, SvUV (*av_fetch ((AV *)sv, 1, 1)), *av_fetch ((AV *)sv, 0, 1));
		618	break;
		619
		620	default:
		621	croak ("encountered malformed CBOR::XS::Tagged object");
		622	}
		623
		624	return;
		625	}
		626	}
		627
		628	if (ecb_expect_false (SvREFCNT (sv) > 1)
		629	&& ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING))
		630	{
		631	if (ecb_expect_false (!enc->shareable))
		632	enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
		633
		634	SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
		635
		636	if (SvOK (*svp))
		637	{
		638	encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
		639	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		640	return;
		641	}
255	else	642	else
256	{	643	{
257	#if 0 //TODO	644	sv_setuv (*svp, enc->shareable_idx);
258	if (enc->cbor.flags & F_CONV_BLESSED)	645	++enc->shareable_idx;
		646	encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
		647	}
		648	}
		649
		650	if (ecb_expect_false (SvOBJECT (sv)))
		651	{
		652	HV *stash = SvSTASH (sv);
		653	GV *method;
		654
		655	if (enc->cbor.flags & F_FORBID_OBJECTS)
		656	croak ("encountered object '%s', but forbid_objects is enabled",
		657	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		658	else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
259	{	659	{
		660	dSP;
		661
		662	ENTER; SAVETMPS;
		663	PUSHMARK (SP);
260	// we re-bless the reference to get overload and other niceties right	664	// 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)));	665	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
269		666
270	// calling with G_SCALAR ensures that we always get a 1 return value
271	PUTBACK;	667	PUTBACK;
		668	// G_SCALAR ensures that return value is 1
272	call_sv ((SV *)GvCV (to_cbor), G_SCALAR);	669	call_sv ((SV *)GvCV (method), G_SCALAR);
273	SPAGAIN;	670	SPAGAIN;
274		671
275	// catch this surprisingly common error	672	// catch this surprisingly common error
276	if (SvROK (TOPs) && SvRV (TOPs) == sv)	673	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)));	674	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
278		675
279	sv = POPs;
280	PUTBACK;
281
282	encode_sv (enc, sv);	676	encode_sv (enc, POPs);
283		677
		678	PUTBACK;
		679
284	FREETMPS; LEAVE;	680	FREETMPS; LEAVE;
285	}	681	}
286	else if (enc->cbor.flags & F_ALLOW_BLESSED)	682	else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
287	encode_str (enc, "null", 4, 0);	683	{
288	else	684	dSP;
289	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it",	685
290	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	686	ENTER; SAVETMPS;
		687	PUSHMARK (SP);
		688	EXTEND (SP, 2);
		689	// we re-bless the reference to get overload and other niceties right
		690	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		691	PUSHs (sv_cbor);
		692
		693	PUTBACK;
		694	int count = call_sv ((SV *)GvCV (method), G_ARRAY);
		695	SPAGAIN;
		696
		697	// catch this surprisingly common error
		698	if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
		699	croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
		700
		701	encode_tag (enc, CBOR_TAG_PERL_OBJECT);
		702	encode_uint (enc, MAJOR_ARRAY, count + 1);
		703	encode_strref (enc, 0, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
		704
		705	{
		706	int i;
		707
		708	for (i = 0; i < count; ++i)
		709	encode_sv (enc, SP[i + 1 - count]);
		710
		711	SP -= count;
291	}	712	}
292	else if (enc->cbor.flags & F_ALLOW_BLESSED)	713
293	encode_str (enc, "null", 4, 0);	714	PUTBACK;
		715
		716	FREETMPS; LEAVE;
		717	}
294	else	718	else
295	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",	719	croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
296	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	720	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
297	#endif
298	}
299	}	721	}
300	else if (svt == SVt_PVHV)	722	else if (svt == SVt_PVHV)
301	encode_hv (enc, (HV *)sv);	723	encode_hv (enc, (HV *)sv);
302	else if (svt == SVt_PVAV)	724	else if (svt == SVt_PVAV)
303	encode_av (enc, (AV *)sv);	725	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	726	else
322	croak ("encountered %s, but CBOR can only represent references to arrays or hashes",	727	{
323	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	728	encode_tag (enc, CBOR_TAG_INDIRECTION);
		729	encode_sv (enc, sv);
		730	}
324	}	731	}
325		732
326	static void	733	static void
327	encode_nv (enc_t *enc, SV *sv)	734	encode_nv (enc_t *enc, SV *sv)
328	{	735	{
329	double nv = SvNVX (sv);	736	double nv = SvNVX (sv);
330		737
331	need (enc, 9);	738	need (enc, 9);
332		739
333	if (expect_false (nv == (U32)nv))	740	if (ecb_expect_false (nv == (NV)(U32)nv))
334	encode_uint (enc, 0x00, (U32)nv);	741	encode_uint (enc, MAJOR_POS_INT, (U32)nv);
335	//TODO: maybe I32?	742	//TODO: maybe I32?
336	else if (expect_false (nv == (float)nv))	743	else if (ecb_expect_false (nv == (float)nv))
337	{	744	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	745	else
349	{	746	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	}	747	}
361		748
362	static void	749	static void
363	encode_sv (enc_t *enc, SV *sv)	750	encode_sv (enc_t *enc, SV *sv)
364	{	751	{
…		…
366		753
367	if (SvPOKp (sv))	754	if (SvPOKp (sv))
368	{	755	{
369	STRLEN len;	756	STRLEN len;
370	char *str = SvPV (sv, len);	757	char *str = SvPV (sv, len);
371	encode_str (enc, SvUTF8 (sv), str, len);	758	encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
372	}	759	}
373	else if (SvNOKp (sv))	760	else if (SvNOKp (sv))
374	encode_nv (enc, sv);	761	encode_nv (enc, sv);
375	else if (SvIOKp (sv))	762	else if (SvIOKp (sv))
376	{	763	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))	764	else if (SvROK (sv))
385	encode_rv (enc, SvRV (sv));	765	encode_rv (enc, SvRV (sv));
386	else if (!SvOK (sv))	766	else if (!SvOK (sv))
387	encode_ch (enc, 0xe0 | 22);	767	encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
388	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)	768	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
389	encode_ch (enc, 0xe0 | 23);	769	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
390	else	770	else
391	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",	771	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",
392	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));	772	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
393	}	773	}
394		774
395	static SV *	775	static SV *
396	encode_cbor (SV *scalar, CBOR *cbor)	776	encode_cbor (SV *scalar, CBOR *cbor)
397	{	777	{
398	enc_t enc;	778	enc_t enc = { 0 };
399		779
400	enc.cbor = *cbor;	780	enc.cbor = *cbor;
401	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	781	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
402	enc.cur = SvPVX (enc.sv);	782	enc.cur = SvPVX (enc.sv);
403	enc.end = SvEND (enc.sv);	783	enc.end = SvEND (enc.sv);
404	enc.depth = 0;
405		784
406	SvPOK_only (enc.sv);	785	SvPOK_only (enc.sv);
		786
		787	if (cbor->flags & F_PACK_STRINGS)
		788	{
		789	encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
		790	enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
		791	enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
		792	}
		793
407	encode_sv (&enc, scalar);	794	encode_sv (&enc, scalar);
408		795
409	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	796	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
410	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	797	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
411		798
…		…
425	U8 *end; // end of input string	812	U8 *end; // end of input string
426	const char *err; // parse error, if != 0	813	const char *err; // parse error, if != 0
427	CBOR cbor;	814	CBOR cbor;
428	U32 depth; // recursion depth	815	U32 depth; // recursion depth
429	U32 maxdepth; // recursion depth limit	816	U32 maxdepth; // recursion depth limit
		817	AV *shareable;
		818	AV *stringref;
		819	SV *decode_tagged;
		820	SV *err_sv; // optional sv for error, needs to be freed
430	} dec_t;	821	} dec_t;
431		822
432	#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE	823	// set dec->err to ERRSV
		824	ecb_cold static void
		825	err_errsv (dec_t *dec)
		826	{
		827	if (!dec->err)
		828	{
		829	dec->err_sv = newSVsv (ERRSV);
433		830
434	#define WANT(len) if (expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data");	831	// chop off the trailing \n
		832	SvCUR_set (dec->err_sv, SvCUR (dec->err_sv) - 1);
		833	*SvEND (dec->err_sv) = 0;
435		834
		835	dec->err = SvPVutf8_nolen (dec->err_sv);
		836	}
		837	}
		838
		839	// the following functions are used to reduce code size and help the compiler to optimise
		840	ecb_cold static void
		841	err_set (dec_t *dec, const char *reason)
		842	{
		843	if (!dec->err)
		844	dec->err = reason;
		845	}
		846
		847	ecb_cold static void
		848	err_unexpected_end (dec_t *dec)
		849	{
		850	err_set (dec, "unexpected end of CBOR data");
		851	}
		852
		853	#define ERR_DO(do) SB do; goto fail; SE
		854	#define ERR(reason) ERR_DO (err_set (dec, reason))
		855	#define ERR_ERRSV ERR_DO (err_errsv (dec))
		856
		857	#define WANT(len) if (ecb_expect_false ((uintptr_t)(dec->end - dec->cur) < (STRLEN)len)) ERR_DO (err_unexpected_end (dec))
		858
436	#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)	859	#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	860	#define DEC_DEC_DEPTH --dec->depth
438		861
439	static UV	862	static UV
440	decode_uint (dec_t *dec)	863	decode_uint (dec_t *dec)
441	{	864	{
442	switch (*dec->cur & 31)	865	U8 m = *dec->cur & MINOR_MASK;
443	{	866	++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		867
449	case 24:	868	if (ecb_expect_true (m < LENGTH_EXT1))
		869	return m;
		870	else if (ecb_expect_true (m == LENGTH_EXT1))
		871	{
450	WANT (2);	872	WANT (1);
451	dec->cur += 2;	873	dec->cur += 1;
452	return dec->cur[-1];	874	return dec->cur[-1];
453		875	}
454	case 25:	876	else if (ecb_expect_true (m == LENGTH_EXT2))
		877	{
455	WANT (3);	878	WANT (2);
456	dec->cur += 3;	879	dec->cur += 2;
457	return (((UV)dec->cur[-2]) << 8)	880	return (((UV)dec->cur[-2]) << 8)
458	| ((UV)dec->cur[-1]);	881	| ((UV)dec->cur[-1]);
459		882	}
460	case 26:	883	else if (ecb_expect_true (m == LENGTH_EXT4))
		884	{
461	WANT (5);	885	WANT (4);
462	dec->cur += 5;	886	dec->cur += 4;
463	return (((UV)dec->cur[-4]) << 24)	887	return (((UV)dec->cur[-4]) << 24)
464	| (((UV)dec->cur[-3]) << 16)	888	| (((UV)dec->cur[-3]) << 16)
465	| (((UV)dec->cur[-2]) << 8)	889	| (((UV)dec->cur[-2]) << 8)
466	| ((UV)dec->cur[-1]);	890	| ((UV)dec->cur[-1]);
467		891	}
468	case 27:	892	else if (ecb_expect_true (m == LENGTH_EXT8))
		893	{
469	WANT (9);	894	WANT (8);
470	dec->cur += 9;	895	dec->cur += 8;
		896
		897	return
		898	#if UVSIZE < 8
		899	0
		900	#else
471	return (((UV)dec->cur[-8]) << 56)	901	(((UV)dec->cur[-8]) << 56)
472	| (((UV)dec->cur[-7]) << 48)	902	| (((UV)dec->cur[-7]) << 48)
473	| (((UV)dec->cur[-6]) << 40)	903	| (((UV)dec->cur[-6]) << 40)
474	| (((UV)dec->cur[-5]) << 32)	904	| (((UV)dec->cur[-5]) << 32)
		905	#endif
475	| (((UV)dec->cur[-4]) << 24)	906	| (((UV)dec->cur[-4]) << 24)
476	| (((UV)dec->cur[-3]) << 16)	907	| (((UV)dec->cur[-3]) << 16)
477	| (((UV)dec->cur[-2]) << 8)	908	| (((UV)dec->cur[-2]) << 8)
478	| ((UV)dec->cur[-1]);	909	| ((UV)dec->cur[-1]);
479		910	}
480	default:	911	else
481	ERR ("corrupted CBOR data (unsupported integer minor encoding)");	912	ERR ("corrupted CBOR data (unsupported integer minor encoding)");
482	}
483		913
484	fail:	914	fail:
485	return 0;	915	return 0;
486	}	916	}
487		917
…		…
492	{	922	{
493	AV *av = newAV ();	923	AV *av = newAV ();
494		924
495	DEC_INC_DEPTH;	925	DEC_INC_DEPTH;
496		926
497	if ((*dec->cur & 31) == 31)	927	if (*dec->cur == (MAJOR_ARRAY | MINOR_INDEF))
498	{	928	{
499	++dec->cur;	929	++dec->cur;
500		930
501	for (;;)	931	for (;;)
502	{	932	{
503	WANT (1);	933	WANT (1);
504		934
505	if (*dec->cur == (0xe0 | 31))	935	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
506	{	936	{
507	++dec->cur;	937	++dec->cur;
508	break;	938	break;
509	}	939	}
510		940
511	av_push (av, decode_sv (dec));	941	av_push (av, decode_sv (dec));
512	}	942	}
513	}	943	}
514	else	944	else
515	{	945	{
516	int i, len = decode_uint (dec);	946	UV i, len = decode_uint (dec);
517		947
		948	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);	949	av_fill (av, len - 1);
519		950
520	for (i = 0; i < len; ++i)	951	for (i = 0; i < len; ++i)
521	AvARRAY (av)[i] = decode_sv (dec);	952	AvARRAY (av)[i] = decode_sv (dec);
522	}	953	}
523		954
524	DEC_DEC_DEPTH;	955	DEC_DEC_DEPTH;
525	return newRV_noinc ((SV *)av);	956	return newRV_noinc ((SV *)av);
526		957
527	fail:	958	fail:
528	SvREFCNT_dec (av);	959	SvREFCNT_dec_NN (av);
529	DEC_DEC_DEPTH;	960	DEC_DEC_DEPTH;
530	return &PL_sv_undef;	961	return &PL_sv_undef;
531	}	962	}
532		963
		964	static void
		965	decode_he (dec_t *dec, HV *hv)
		966	{
		967	// for speed reasons, we specialcase single-string
		968	// byte or utf-8 strings as keys, but only when !stringref
		969
		970	if (ecb_expect_true (!dec->stringref))
		971	if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8))
		972	{
		973	STRLEN len = decode_uint (dec);
		974	char *key = (char *)dec->cur;
		975
		976	WANT (len);
		977	dec->cur += len;
		978
		979	hv_store (hv, key, len, decode_sv (dec), 0);
		980
		981	return;
		982	}
		983	else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8))
		984	{
		985	STRLEN len = decode_uint (dec);
		986	char *key = (char *)dec->cur;
		987
		988	WANT (len);
		989	dec->cur += len;
		990
		991	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		992	if (!is_utf8_string (key, len))
		993	ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
		994
		995	hv_store (hv, key, -len, decode_sv (dec), 0);
		996
		997	return;
		998	}
		999
		1000	SV *k = decode_sv (dec);
		1001	SV *v = decode_sv (dec);
		1002
		1003	// we leak memory if uncaught exceptions are thrown by random magical
		1004	// methods, and this is hopefully the only place where it can happen,
		1005	// so if there is a chance of an exception, take the very slow path.
		1006	// since catching exceptions is "undocumented/internal/forbidden" by
		1007	// the new p5p powers, we need to call out to a perl function :/
		1008	if (ecb_expect_false (SvAMAGIC (k)))
		1009	{
		1010	dSP;
		1011
		1012	ENTER; SAVETMPS;
		1013	PUSHMARK (SP);
		1014	EXTEND (SP, 3);
		1015	PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
		1016	PUSHs (sv_2mortal (k));
		1017	PUSHs (sv_2mortal (v));
		1018
		1019	PUTBACK;
		1020	call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
		1021	SPAGAIN;
		1022
		1023	FREETMPS; LEAVE;
		1024
		1025	if (SvTRUE (ERRSV))
		1026	ERR_ERRSV;
		1027
		1028	return;
		1029	}
		1030
		1031	hv_store_ent (hv, k, v, 0);
		1032	SvREFCNT_dec_NN (k);
		1033
		1034	fail:
		1035	;
		1036	}
		1037
533	static SV *	1038	static SV *
534	decode_hv (dec_t *dec)	1039	decode_hv (dec_t *dec)
535	{	1040	{
536	HV *hv = newHV ();	1041	HV *hv = newHV ();
537		1042
538	DEC_INC_DEPTH;	1043	DEC_INC_DEPTH;
539		1044
540	if ((*dec->cur & 31) == 31)	1045	if (*dec->cur == (MAJOR_MAP | MINOR_INDEF))
541	{	1046	{
542	++dec->cur;	1047	++dec->cur;
543		1048
544	for (;;)	1049	for (;;)
545	{	1050	{
546	WANT (1);	1051	WANT (1);
547		1052
548	if (*dec->cur == (0xe0 | 31))	1053	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
549	{	1054	{
550	++dec->cur;	1055	++dec->cur;
551	break;	1056	break;
552	}	1057	}
553		1058
554	SV *k = decode_sv (dec);	1059	decode_he (dec, hv);
555	SV *v = decode_sv (dec);
556
557	hv_store_ent (hv, k, v, 0);
558	}	1060	}
559	}	1061	}
560	else	1062	else
561	{	1063	{
562	int len = decode_uint (dec);	1064	UV pairs = decode_uint (dec);
563		1065
		1066	WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
		1067
564	while (len--)	1068	while (pairs--)
565	{	1069	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	}	1070	}
572		1071
573	DEC_DEC_DEPTH;	1072	DEC_DEC_DEPTH;
574	return newRV_noinc ((SV *)hv);	1073	return newRV_noinc ((SV *)hv);
575		1074
576	#if 0	1075	fail:
577	SV *sv;	1076	SvREFCNT_dec_NN (hv);
578	HV *hv = newHV ();
579
580	DEC_INC_DEPTH;	1077	DEC_DEC_DEPTH;
581	decode_ws (dec);	1078	return &PL_sv_undef;
		1079	}
582		1080
583	for (;;)	1081	static SV *
584	{	1082	decode_str (dec_t *dec, int utf8)
585	// heuristic: assume that	1083	{
586	// a) decode_str + hv_store_ent are abysmally slow.	1084	SV *sv = 0;
587	// b) most hash keys are short, simple ascii text.	1085
588	// => try to "fast-match" such strings to avoid	1086	if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
589	// the overhead of decode_str + hv_store_ent.
590	{	1087	{
591	SV *value;	1088	// indefinite length strings
592	char *p = dec->cur;	1089	++dec->cur;
593	char *e = p + 24; // only try up to 24 bytes
594		1090
		1091	U8 major = *dec->cur & MAJOR_MISC;
		1092
		1093	sv = newSVpvn ("", 0);
		1094
595	for (;;)	1095	for (;;)
596	{	1096	{
597	// the >= 0x80 is false on most architectures	1097	WANT (1);
598	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1098
		1099	if ((*dec->cur - major) > LENGTH_EXT8)
		1100	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF))
599	{	1101	{
600	// slow path, back up and use decode_str	1102	++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;	1103	break;
619	}	1104	}
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	{	1105	else
685	dSP;	1106	ERR ("corrupted CBOR data (invalid chunks in indefinite length string)");
686	int count;
687		1107
688	ENTER; SAVETMPS; PUSHMARK (SP);	1108	STRLEN len = decode_uint (dec);
689	XPUSHs (HeVAL (he));
690	sv_2mortal (sv);
691		1109
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);	1110	WANT (len);
752		1111	sv_catpvn (sv, dec->cur, len);
753	if (*dec->cur == (0xe0 | 31))	1112	dec->cur += len;
754	{
755	++dec->cur;
756	break;
757	}
758
759	SV *sv2 = decode_sv (dec);
760	sv_catsv (sv, sv2);
761	}	1113	}
762	}	1114	}
763	else	1115	else
764	{	1116	{
765	STRLEN len = decode_uint (dec);	1117	STRLEN len = decode_uint (dec);
766		1118
767	WANT (len);	1119	WANT (len);
768	sv = newSVpvn (dec->cur, len);	1120	sv = newSVpvn (dec->cur, len);
769	dec->cur += len;	1121	dec->cur += len;
		1122
		1123	if (ecb_expect_false (dec->stringref)
		1124	&& SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
		1125	av_push (dec->stringref, SvREFCNT_inc_NN (sv));
770	}	1126	}
771		1127
772	if (utf8)	1128	if (utf8)
		1129	{
		1130	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		1131	if (!is_utf8_string (SvPVX (sv), SvCUR (sv)))
		1132	ERR ("corrupted CBOR data (invalid UTF-8 in text string)");
		1133
773	SvUTF8_on (sv);	1134	SvUTF8_on (sv);
		1135	}
774		1136
775	return sv;	1137	return sv;
776		1138
777	fail:	1139	fail:
		1140	SvREFCNT_dec (sv);
778	return &PL_sv_undef;	1141	return &PL_sv_undef;
779	}	1142	}
780		1143
781	static SV *	1144	static SV *
782	decode_tagged (dec_t *dec)	1145	decode_tagged (dec_t *dec)
783	{	1146	{
		1147	SV *sv = 0;
784	UV tag = decode_uint (dec);	1148	UV tag = decode_uint (dec);
		1149
		1150	WANT (1);
		1151
		1152	switch (tag)
		1153	{
		1154	case CBOR_TAG_MAGIC:
785	SV *sv = decode_sv (dec);	1155	sv = decode_sv (dec);
		1156	break;
786		1157
787	if (tag == 55799) // 2.4.5 Self-Describe CBOR	1158	case CBOR_TAG_INDIRECTION:
		1159	sv = newRV_noinc (decode_sv (dec));
		1160	break;
		1161
		1162	case CBOR_TAG_STRINGREF_NAMESPACE:
		1163	{
		1164	// do not use SAVETMPS/FREETMPS, as these will
		1165	// erase mortalised caches, e.g. "shareable"
		1166	ENTER;
		1167
		1168	SAVESPTR (dec->stringref);
		1169	dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
		1170
		1171	sv = decode_sv (dec);
		1172
		1173	LEAVE;
		1174	}
		1175	break;
		1176
		1177	case CBOR_TAG_STRINGREF:
		1178	{
		1179	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1180	ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
		1181
		1182	UV idx = decode_uint (dec);
		1183
		1184	if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
		1185	ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
		1186
		1187	sv = newSVsv (AvARRAY (dec->stringref)[idx]);
		1188	}
		1189	break;
		1190
		1191	case CBOR_TAG_VALUE_SHAREABLE:
		1192	{
		1193	if (ecb_expect_false (!dec->shareable))
		1194	dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
		1195
		1196	if (dec->cbor.flags & F_ALLOW_CYCLES)
		1197	{
		1198	sv = newSV (0);
		1199	av_push (dec->shareable, SvREFCNT_inc_NN (sv));
		1200
		1201	SV *osv = decode_sv (dec);
		1202	sv_setsv (sv, osv);
		1203	SvREFCNT_dec_NN (osv);
		1204	}
		1205	else
		1206	{
		1207	av_push (dec->shareable, &PL_sv_undef);
		1208	int idx = AvFILLp (dec->shareable);
		1209	sv = decode_sv (dec);
		1210	av_store (dec->shareable, idx, SvREFCNT_inc_NN (sv));
		1211	}
		1212	}
		1213	break;
		1214
		1215	case CBOR_TAG_VALUE_SHAREDREF:
		1216	{
		1217	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1218	ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
		1219
		1220	UV idx = decode_uint (dec);
		1221
		1222	if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
		1223	ERR ("corrupted CBOR data (sharedref index out of bounds)");
		1224
		1225	sv = SvREFCNT_inc_NN (AvARRAY (dec->shareable)[idx]);
		1226
		1227	if (sv == &PL_sv_undef)
		1228	ERR ("cyclic CBOR data structure found, but allow_cycles is not enabled");
		1229	}
		1230	break;
		1231
		1232	case CBOR_TAG_PERL_OBJECT:
		1233	{
		1234	if (dec->cbor.flags & F_FORBID_OBJECTS)
		1235	goto filter;
		1236
		1237	sv = decode_sv (dec);
		1238
		1239	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		1240	ERR ("corrupted CBOR data (non-array perl object)");
		1241
		1242	AV *av = (AV *)SvRV (sv);
		1243	int len = av_len (av) + 1;
		1244	HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
		1245
		1246	if (!stash)
		1247	ERR ("cannot decode perl-object (package does not exist)");
		1248
		1249	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1250
		1251	if (!method)
		1252	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1253
		1254	dSP;
		1255
		1256	ENTER; SAVETMPS;
		1257	PUSHMARK (SP);
		1258	EXTEND (SP, len + 1);
		1259	// we re-bless the reference to get overload and other niceties right
		1260	PUSHs (*av_fetch (av, 0, 1));
		1261	PUSHs (sv_cbor);
		1262
		1263	int i;
		1264
		1265	for (i = 1; i < len; ++i)
		1266	PUSHs (*av_fetch (av, i, 1));
		1267
		1268	PUTBACK;
		1269	call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
		1270	SPAGAIN;
		1271
		1272	if (SvTRUE (ERRSV))
		1273	{
		1274	FREETMPS; LEAVE;
		1275	ERR_ERRSV;
		1276	}
		1277
		1278	SvREFCNT_dec_NN (sv);
		1279	sv = SvREFCNT_inc (POPs);
		1280
		1281	PUTBACK;
		1282
		1283	FREETMPS; LEAVE;
		1284	}
		1285	break;
		1286
		1287	default:
		1288	filter:
		1289	{
		1290	SV *tag_sv = newSVuv (tag);
		1291
		1292	sv = decode_sv (dec);
		1293
		1294	dSP;
		1295	ENTER; SAVETMPS;
		1296	PUSHMARK (SP);
		1297	EXTEND (SP, 2);
		1298	PUSHs (tag_sv);
		1299	PUSHs (sv);
		1300
		1301	PUTBACK;
		1302	int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
		1303	SPAGAIN;
		1304
		1305	if (SvTRUE (ERRSV))
		1306	{
		1307	SvREFCNT_dec_NN (tag_sv);
		1308	FREETMPS; LEAVE;
		1309	ERR_ERRSV;
		1310	}
		1311
		1312	if (count)
		1313	{
		1314	SvREFCNT_dec_NN (tag_sv);
		1315	SvREFCNT_dec_NN (sv);
		1316	sv = SvREFCNT_inc_NN (TOPs);
		1317	SP -= count;
		1318	}
		1319	else
		1320	{
		1321	AV *av = newAV ();
		1322	av_push (av, tag_sv);
		1323	av_push (av, sv);
		1324
		1325	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		1326	? cbor_tagged_stash
		1327	: gv_stashpv ("CBOR::XS::Tagged" , 1);
		1328	sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
		1329	}
		1330
		1331	PUTBACK;
		1332
		1333	FREETMPS; LEAVE;
		1334	}
		1335	break;
		1336	}
		1337
788	return sv;	1338	return sv;
789		1339
790	AV *av = newAV ();	1340	fail:
791	av_push (av, newSVuv (tag));	1341	SvREFCNT_dec (sv);
792	av_push (av, sv);	1342	return &PL_sv_undef;
793	return newRV_noinc ((SV *)av);
794	}	1343	}
795		1344
796	static SV *	1345	static SV *
797	decode_sv (dec_t *dec)	1346	decode_sv (dec_t *dec)
798	{	1347	{
799	WANT (1);	1348	WANT (1);
800		1349
801	switch (*dec->cur >> 5)	1350	switch (*dec->cur >> MAJOR_SHIFT)
802	{	1351	{
803	case 0: // unsigned int	1352	case MAJOR_POS_INT >> MAJOR_SHIFT: return newSVuv (decode_uint (dec));
804	//TODO: 64 bit values on 3 2bit perls	1353	case MAJOR_NEG_INT >> MAJOR_SHIFT: return newSViv (-1 - (IV)decode_uint (dec));
805	return newSVuv (decode_uint (dec));	1354	case MAJOR_BYTES >> MAJOR_SHIFT: return decode_str (dec, 0);
806	case 1: // negative int	1355	case MAJOR_TEXT >> MAJOR_SHIFT: return decode_str (dec, 1);
807	return newSViv (-1 - (IV)decode_uint (dec));	1356	case MAJOR_ARRAY >> MAJOR_SHIFT: return decode_av (dec);
808	case 2: // octet string	1357	case MAJOR_MAP >> MAJOR_SHIFT: return decode_hv (dec);
809	return decode_str (dec, 0);	1358	case MAJOR_TAG >> MAJOR_SHIFT: return decode_tagged (dec);
810	case 3: // utf-8 string	1359
811	return decode_str (dec, 1);	1360	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)	1361	switch (*dec->cur++ & MINOR_MASK)
820	{	1362	{
821	case 20:	1363	case SIMPLE_FALSE:
822	#if CBOR_SLOW	1364	#if CBOR_SLOW
823	cbor_false = get_bool ("CBOR::XS::false");	1365	types_false = get_bool ("Types::Serialiser::false");
824	#endif	1366	#endif
825	return newSVsv (cbor_false);	1367	return newSVsv (types_false);
826	case 21:	1368	case SIMPLE_TRUE:
827	#if CBOR_SLOW	1369	#if CBOR_SLOW
828	cbor_true = get_bool ("CBOR::XS::true");	1370	types_true = get_bool ("Types::Serialiser::true");
829	#endif	1371	#endif
830	return newSVsv (cbor_true);	1372	return newSVsv (types_true);
831	case 22:	1373	case SIMPLE_NULL:
832	return newSVsv (&PL_sv_undef);	1374	return newSVsv (&PL_sv_undef);
		1375	case SIMPLE_UNDEF:
		1376	#if CBOR_SLOW
		1377	types_error = get_bool ("Types::Serialiser::error");
		1378	#endif
		1379	return newSVsv (types_error);
833		1380
834	case 25:	1381	case MISC_FLOAT16:
835	{	1382	{
836	WANT (2);	1383	WANT (2);
837		1384
838	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];	1385	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
839	dec->cur += 2;	1386	dec->cur += 2;
840		1387
841	return newSVnv (ecb_binary16_to_float (fp));	1388	return newSVnv (ecb_binary16_to_float (fp));
842	}	1389	}
843		1390
844	case 26:	1391	case MISC_FLOAT32:
845	{	1392	{
846	uint32_t fp;	1393	uint32_t fp;
847	WANT (4);	1394	WANT (4);
848	memcpy (&fp, dec->cur, 4);	1395	memcpy (&fp, dec->cur, 4);
849	dec->cur += 4;	1396	dec->cur += 4;
…		…
852	fp = ecb_bswap32 (fp);	1399	fp = ecb_bswap32 (fp);
853		1400
854	return newSVnv (ecb_binary32_to_float (fp));	1401	return newSVnv (ecb_binary32_to_float (fp));
855	}	1402	}
856		1403
857	case 27:	1404	case MISC_FLOAT64:
858	{	1405	{
859	uint64_t fp;	1406	uint64_t fp;
860	WANT (8);	1407	WANT (8);
861	memcpy (&fp, dec->cur, 8);	1408	memcpy (&fp, dec->cur, 8);
862	dec->cur += 8;	1409	dec->cur += 8;
…		…
865	fp = ecb_bswap64 (fp);	1412	fp = ecb_bswap64 (fp);
866		1413
867	return newSVnv (ecb_binary64_to_double (fp));	1414	return newSVnv (ecb_binary64_to_double (fp));
868	}	1415	}
869		1416
870	// 0..19 unassigned	1417	// 0..19 unassigned simple
871	// 24 reserved + unassigned (reserved values are not encodable)	1418	// 24 reserved + unassigned simple (reserved values are not encodable)
		1419	// 28-30 unassigned misc
		1420	// 31 break code
872	default:	1421	default:
873	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");	1422	ERR ("corrupted CBOR data (reserved/unassigned/unexpected major 7 value)");
874	}	1423	}
875		1424
876	break;	1425	break;
877	}	1426	}
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		1427
935	fail:	1428	fail:
936	return &PL_sv_undef;	1429	return &PL_sv_undef;
937	}	1430	}
938		1431
939	static SV *	1432	static SV *
940	decode_cbor (SV *string, CBOR *cbor, char **offset_return)	1433	decode_cbor (SV *string, CBOR *cbor, char **offset_return)
941	{	1434	{
942	dec_t dec;	1435	dec_t dec = { 0 };
943	SV *sv;	1436	SV *sv;
		1437	STRLEN len;
		1438	char *data = SvPVbyte (string, len);
944		1439
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)	1440	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",	1441	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);	1442	(unsigned long)len, (unsigned long)cbor->max_size);
976	}
977
978	sv_utf8_downgrade (string, 0);
979		1443
980	dec.cbor = *cbor;	1444	dec.cbor = *cbor;
981	dec.cur = (U8 *)SvPVX (string);	1445	dec.cur = (U8 *)data;
982	dec.end = (U8 *)SvEND (string);	1446	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		1447
989	sv = decode_sv (&dec);	1448	sv = decode_sv (&dec);
990		1449
991	if (offset_return)	1450	if (offset_return)
992	*offset_return = dec.cur;	1451	*offset_return = dec.cur;
…		…
995	if (dec.cur != dec.end && !dec.err)	1454	if (dec.cur != dec.end && !dec.err)
996	dec.err = "garbage after CBOR object";	1455	dec.err = "garbage after CBOR object";
997		1456
998	if (dec.err)	1457	if (dec.err)
999	{	1458	{
		1459	if (dec.shareable)
		1460	{
		1461	// need to break cyclic links, which would all be in shareable
		1462	int i;
		1463	SV **svp;
		1464
		1465	for (i = av_len (dec.shareable) + 1; i--; )
		1466	if ((svp = av_fetch (dec.shareable, i, 0)))
		1467	sv_setsv (*svp, &PL_sv_undef);
		1468	}
		1469
1000	SvREFCNT_dec (sv);	1470	SvREFCNT_dec_NN (sv);
		1471
		1472	if (dec.err_sv)
		1473	sv_2mortal (dec.err_sv);
		1474
1001	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur);	1475	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)data, (int)(uint8_t)*dec.cur);
1002	}	1476	}
1003		1477
1004	sv = sv_2mortal (sv);	1478	sv = sv_2mortal (sv);
1005		1479
1006	return sv;	1480	return sv;
1007	}	1481	}
1008		1482
		1483	/////////////////////////////////////////////////////////////////////////////
		1484	// incremental parser
		1485
		1486	#define INCR_DONE(cbor) (AvFILLp (cbor->incr_count) < 0)
		1487
		1488	// returns 0 for notyet, 1 for success or error
		1489	static int
		1490	incr_parse (CBOR *self, SV *cborstr)
		1491	{
		1492	STRLEN cur;
		1493	SvPV (cborstr, cur);
		1494
		1495	while (ecb_expect_true (self->incr_need <= cur))
		1496	{
		1497	// table of integer count bytes
		1498	static I8 incr_len[MINOR_MASK + 1] = {
		1499	0, 0, 0, 0, 0, 0, 0, 0,
		1500	0, 0, 0, 0, 0, 0, 0, 0,
		1501	0, 0, 0, 0, 0, 0, 0, 0,
		1502	1, 2, 4, 8,-1,-1,-1,-2
		1503	};
		1504
		1505	const U8 *p = SvPVX (cborstr) + self->incr_pos;
		1506	U8 m = *p & MINOR_MASK;
		1507	IV count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1508	I8 ilen = incr_len[m];
		1509
		1510	self->incr_need = self->incr_pos + 1;
		1511
		1512	if (ecb_expect_false (ilen < 0))
		1513	{
		1514	if (m != MINOR_INDEF)
		1515	return 1; // error
		1516
		1517	if (*p == (MAJOR_MISC | MINOR_INDEF))
		1518	{
		1519	if (count >= 0)
		1520	return 1; // error
		1521
		1522	count = 1;
		1523	}
		1524	else
		1525	{
		1526	av_push (self->incr_count, newSViv (-1)); //TODO: nest
		1527	count = -1;
		1528	}
		1529	}
		1530	else
		1531	{
		1532	self->incr_need += ilen;
		1533	if (ecb_expect_false (self->incr_need > cur))
		1534	return 0;
		1535
		1536	int major = *p >> MAJOR_SHIFT;
		1537
		1538	switch (major)
		1539	{
		1540	case MAJOR_TAG >> MAJOR_SHIFT:
		1541	++count; // tags merely prefix another value
		1542	break;
		1543
		1544	case MAJOR_BYTES >> MAJOR_SHIFT:
		1545	case MAJOR_TEXT >> MAJOR_SHIFT:
		1546	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1547	case MAJOR_MAP >> MAJOR_SHIFT:
		1548	{
		1549	UV len;
		1550
		1551	if (ecb_expect_false (ilen))
		1552	{
		1553	len = 0;
		1554
		1555	do {
		1556	len = (len << 8) | *++p;
		1557	} while (--ilen);
		1558	}
		1559	else
		1560	len = m;
		1561
		1562	switch (major)
		1563	{
		1564	case MAJOR_BYTES >> MAJOR_SHIFT:
		1565	case MAJOR_TEXT >> MAJOR_SHIFT:
		1566	self->incr_need += len;
		1567	if (ecb_expect_false (self->incr_need > cur))
		1568	return 0;
		1569
		1570	break;
		1571
		1572	case MAJOR_MAP >> MAJOR_SHIFT:
		1573	len <<= 1;
		1574	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1575	if (len)
		1576	{
		1577	av_push (self->incr_count, newSViv (len + 1)); //TODO: nest
		1578	count = len + 1;
		1579	}
		1580	break;
		1581	}
		1582	}
		1583	}
		1584	}
		1585
		1586	self->incr_pos = self->incr_need;
		1587
		1588	if (count > 0)
		1589	{
		1590	while (!--count)
		1591	{
		1592	if (!AvFILLp (self->incr_count))
		1593	return 1; // done
		1594
		1595	SvREFCNT_dec_NN (av_pop (self->incr_count));
		1596	count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1597	}
		1598
		1599	SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]) = count;
		1600	}
		1601	}
		1602
		1603	return 0;
		1604	}
		1605
		1606
1009	/////////////////////////////////////////////////////////////////////////////	1607	/////////////////////////////////////////////////////////////////////////////
1010	// XS interface functions	1608	// XS interface functions
1011		1609
1012	MODULE = CBOR::XS PACKAGE = CBOR::XS	1610	MODULE = CBOR::XS PACKAGE = CBOR::XS
1013		1611
1014	BOOT:	1612	BOOT:
1015	{	1613	{
1016	cbor_stash = gv_stashpv ("CBOR::XS" , 1);	1614	cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1017	cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1);	1615	cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1018		1616
1019	cbor_true = get_bool ("CBOR::XS::true");	1617	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1020	cbor_false = get_bool ("CBOR::XS::false");	1618	types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
		1619
		1620	types_true = get_bool ("Types::Serialiser::true" );
		1621	types_false = get_bool ("Types::Serialiser::false");
		1622	types_error = get_bool ("Types::Serialiser::error");
		1623
		1624	default_filter = newSVpv ("CBOR::XS::default_filter", 0);
		1625
		1626	sv_cbor = newSVpv ("CBOR", 0);
		1627	SvREADONLY_on (sv_cbor);
		1628
		1629	assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1021	}	1630	}
1022		1631
1023	PROTOTYPES: DISABLE	1632	PROTOTYPES: DISABLE
1024		1633
1025	void CLONE (...)	1634	void CLONE (...)
1026	CODE:	1635	CODE:
1027	cbor_stash = 0;	1636	cbor_stash = 0;
		1637	cbor_tagged_stash = 0;
		1638	types_error_stash = 0;
1028	cbor_boolean_stash = 0;	1639	types_boolean_stash = 0;
1029		1640
1030	void new (char *klass)	1641	void new (char *klass)
1031	PPCODE:	1642	PPCODE:
1032	{	1643	{
1033	SV *pv = NEWSV (0, sizeof (CBOR));	1644	SV *pv = NEWSV (0, sizeof (CBOR));
…		…
1041		1652
1042	void shrink (CBOR *self, int enable = 1)	1653	void shrink (CBOR *self, int enable = 1)
1043	ALIAS:	1654	ALIAS:
1044	shrink = F_SHRINK	1655	shrink = F_SHRINK
1045	allow_unknown = F_ALLOW_UNKNOWN	1656	allow_unknown = F_ALLOW_UNKNOWN
		1657	allow_sharing = F_ALLOW_SHARING
		1658	allow_cycles = F_ALLOW_CYCLES
		1659	forbid_objects = F_FORBID_OBJECTS
		1660	pack_strings = F_PACK_STRINGS
		1661	text_keys = F_TEXT_KEYS
		1662	text_strings = F_TEXT_STRINGS
		1663	validate_utf8 = F_VALIDATE_UTF8
1046	PPCODE:	1664	PPCODE:
1047	{	1665	{
1048	if (enable)	1666	if (enable)
1049	self->flags |= ix;	1667	self->flags |= ix;
1050	else	1668	else
…		…
1055		1673
1056	void get_shrink (CBOR *self)	1674	void get_shrink (CBOR *self)
1057	ALIAS:	1675	ALIAS:
1058	get_shrink = F_SHRINK	1676	get_shrink = F_SHRINK
1059	get_allow_unknown = F_ALLOW_UNKNOWN	1677	get_allow_unknown = F_ALLOW_UNKNOWN
		1678	get_allow_sharing = F_ALLOW_SHARING
		1679	get_allow_cycles = F_ALLOW_CYCLES
		1680	get_forbid_objects = F_FORBID_OBJECTS
		1681	get_pack_strings = F_PACK_STRINGS
		1682	get_text_keys = F_TEXT_KEYS
		1683	get_text_strings = F_TEXT_STRINGS
		1684	get_validate_utf8 = F_VALIDATE_UTF8
1060	PPCODE:	1685	PPCODE:
1061	XPUSHs (boolSV (self->flags & ix));	1686	XPUSHs (boolSV (self->flags & ix));
1062		1687
1063	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)	1688	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1064	PPCODE:	1689	PPCODE:
…		…
1080	CODE:	1705	CODE:
1081	RETVAL = self->max_size;	1706	RETVAL = self->max_size;
1082	OUTPUT:	1707	OUTPUT:
1083	RETVAL	1708	RETVAL
1084		1709
1085	#if 0 //TODO	1710	void filter (CBOR *self, SV *filter = 0)
1086
1087	void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1088	PPCODE:	1711	PPCODE:
1089	{
1090	SvREFCNT_dec (self->cb_object);	1712	SvREFCNT_dec (self->filter);
1091	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;	1713	self->filter = filter ? newSVsv (filter) : filter;
1092
1093	XPUSHs (ST (0));	1714	XPUSHs (ST (0));
1094	}
1095		1715
1096	void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)	1716	SV *get_filter (CBOR *self)
1097	PPCODE:	1717	CODE:
1098	{	1718	RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1099	if (!self->cb_sk_object)	1719	OUTPUT:
1100	self->cb_sk_object = newHV ();	1720	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		1721
1120	void encode (CBOR *self, SV *scalar)	1722	void encode (CBOR *self, SV *scalar)
1121	PPCODE:	1723	PPCODE:
1122	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;	1724	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1123	XPUSHs (scalar);	1725	XPUSHs (scalar);
…		…
1136	EXTEND (SP, 2);	1738	EXTEND (SP, 2);
1137	PUSHs (sv);	1739	PUSHs (sv);
1138	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));	1740	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1139	}	1741	}
1140		1742
		1743	void incr_parse (CBOR *self, SV *cborstr)
		1744	ALIAS:
		1745	incr_parse_multiple = 1
		1746	PPCODE:
		1747	{
		1748	if (SvUTF8 (cborstr))
		1749	sv_utf8_downgrade (cborstr, 0);
		1750
		1751	if (!self->incr_count)
		1752	{
		1753	self->incr_count = newAV ();
		1754	self->incr_pos = 0;
		1755	self->incr_need = 1;
		1756
		1757	av_push (self->incr_count, newSViv (1));
		1758	}
		1759
		1760	do
		1761	{
		1762	if (!incr_parse (self, cborstr))
		1763	{
		1764	if (self->incr_need > self->max_size && self->max_size)
		1765	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
		1766	(unsigned long)self->incr_need, (unsigned long)self->max_size);
		1767
		1768	break;
		1769	}
		1770
		1771	SV *sv;
		1772	char *offset;
		1773
		1774	PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
		1775	XPUSHs (sv);
		1776
		1777	sv_chop (cborstr, offset);
		1778
		1779	av_clear (self->incr_count);
		1780	av_push (self->incr_count, newSViv (1));
		1781
		1782	self->incr_pos = 0;
		1783	self->incr_need = self->incr_pos + 1;
		1784	}
		1785	while (ix);
		1786	}
		1787
		1788	void incr_reset (CBOR *self)
		1789	CODE:
		1790	{
		1791	SvREFCNT_dec (self->incr_count);
		1792	self->incr_count = 0;
		1793	}
		1794
1141	void DESTROY (CBOR *self)	1795	void DESTROY (CBOR *self)
1142	CODE:	1796	PPCODE:
1143	SvREFCNT_dec (self->cb_sk_object);	1797	cbor_free (self);
1144	SvREFCNT_dec (self->cb_object);
1145		1798
1146	PROTOTYPES: ENABLE	1799	PROTOTYPES: ENABLE
1147		1800
1148	void encode_cbor (SV *scalar)	1801	void encode_cbor (SV *scalar)
		1802	ALIAS:
		1803	encode_cbor = 0
		1804	encode_cbor_sharing = F_ALLOW_SHARING
1149	PPCODE:	1805	PPCODE:
1150	{	1806	{
1151	CBOR cbor;	1807	CBOR cbor;
1152	cbor_init (&cbor);	1808	cbor_init (&cbor);
		1809	cbor.flags |= ix;
1153	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;	1810	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1154	XPUSHs (scalar);	1811	XPUSHs (scalar);
1155	}	1812	}
1156		1813
1157	void decode_cbor (SV *cborstr)	1814	void decode_cbor (SV *cborstr)
…		…
1161	cbor_init (&cbor);	1818	cbor_init (&cbor);
1162	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;	1819	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1163	XPUSHs (cborstr);	1820	XPUSHs (cborstr);
1164	}	1821	}
1165		1822
		1823	#ifdef __AFL_COMPILER
		1824
		1825	void
		1826	afl_init ()
		1827	CODE:
		1828	__AFL_INIT ();
		1829
		1830	int
		1831	afl_loop (unsigned int count = 10000)
		1832	CODE:
		1833	RETVAL = __AFL_LOOP (count);
		1834	OUTPUT:
		1835	RETVAL
		1836
		1837	#endif
		1838

Diff Legend

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