ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/CBOR-XS/XS.xs

Comparing CBOR-XS/XS.xs (file contents):
Revision 1.5 by root, Sat Oct 26 21:14:20 2013 UTC vs.
Revision 1.80 by root, Fri Sep 8 20:03:06 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
		16	// compatibility with perl <5.18
		17	#ifndef HvNAMELEN_get
		18	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		19	#endif
		20	#ifndef HvNAMELEN
		21	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		22	#endif
		23	#ifndef HvNAMEUTF8
		24	# define HvNAMEUTF8(hv) 0
		25	#endif
		26	#ifndef SvREFCNT_inc_NN
		27	# define SvREFCNT_inc_NN(sv) SvREFCNT_inc (sv)
		28	#endif
		29	#ifndef SvREFCNT_dec_NN
		30	# define SvREFCNT_dec_NN(sv) SvREFCNT_dec (sv)
		31	#endif
		32
		33	// perl's is_utf8_string interprets len=0 as "calculate len", but we want it to mean 0
		34	#define cbor_is_utf8_string(str,len) (!(len) || is_utf8_string ((str), (len)))
		35
		36	// known major and minor types
		37	enum cbor_type
		38	{
		39	MAJOR_SHIFT = 5,
		40	MINOR_MASK = 0x1f,
		41
		42	MAJOR_POS_INT = 0 << MAJOR_SHIFT,
		43	MAJOR_NEG_INT = 1 << MAJOR_SHIFT,
		44	MAJOR_BYTES = 2 << MAJOR_SHIFT,
		45	MAJOR_TEXT = 3 << MAJOR_SHIFT,
		46	MAJOR_ARRAY = 4 << MAJOR_SHIFT,
		47	MAJOR_MAP = 5 << MAJOR_SHIFT,
		48	MAJOR_TAG = 6 << MAJOR_SHIFT,
		49	MAJOR_MISC = 7 << MAJOR_SHIFT,
		50
		51	// INT/STRING/ARRAY/MAP subtypes
		52	LENGTH_EXT1 = 24,
		53	LENGTH_EXT2 = 25,
		54	LENGTH_EXT4 = 26,
		55	LENGTH_EXT8 = 27,
		56
		57	// SIMPLE types (effectively MISC subtypes)
		58	SIMPLE_FALSE = 20,
		59	SIMPLE_TRUE = 21,
		60	SIMPLE_NULL = 22,
		61	SIMPLE_UNDEF = 23,
		62
		63	// MISC subtype (unused)
		64	MISC_EXT1 = 24,
		65	MISC_FLOAT16 = 25,
		66	MISC_FLOAT32 = 26,
		67	MISC_FLOAT64 = 27,
		68
		69	// BYTES/TEXT/ARRAY/MAP
		70	MINOR_INDEF = 31,
		71	};
		72
		73	// known tags
		74	enum cbor_tag
		75	{
		76	// extensions
		77	CBOR_TAG_STRINGREF = 25, // http://cbor.schmorp.de/stringref
		78	CBOR_TAG_PERL_OBJECT = 26, // http://cbor.schmorp.de/perl-object
		79	CBOR_TAG_GENERIC_OBJECT = 27, // http://cbor.schmorp.de/generic-object
		80	CBOR_TAG_VALUE_SHAREABLE = 28, // http://cbor.schmorp.de/value-sharing
		81	CBOR_TAG_VALUE_SHAREDREF = 29, // http://cbor.schmorp.de/value-sharing
		82	CBOR_TAG_STRINGREF_NAMESPACE = 256, // http://cbor.schmorp.de/stringref
		83	CBOR_TAG_INDIRECTION = 22098, // http://cbor.schmorp.de/indirection
		84
		85	// rfc7049
		86	CBOR_TAG_DATETIME = 0, // rfc4287, utf-8
		87	CBOR_TAG_TIMESTAMP = 1, // unix timestamp, any
		88	CBOR_TAG_POS_BIGNUM = 2, // byte string
		89	CBOR_TAG_NEG_BIGNUM = 3, // byte string
		90	CBOR_TAG_DECIMAL = 4, // decimal fraction, array
		91	CBOR_TAG_BIGFLOAT = 5, // array
		92
		93	CBOR_TAG_CONV_B64U = 21, // base64url, any
		94	CBOR_TAG_CONV_B64 = 22, // base64, any
		95	CBOR_TAG_CONV_HEX = 23, // base16, any
		96	CBOR_TAG_CBOR = 24, // embedded cbor, byte string
		97
		98	CBOR_TAG_URI = 32, // URI rfc3986, utf-8
		99	CBOR_TAG_B64U = 33, // base64url rfc4648, utf-8
		100	CBOR_TAG_B64 = 34, // base6 rfc46484, utf-8
		101	CBOR_TAG_REGEX = 35, // regex pcre/ecma262, utf-8
		102	CBOR_TAG_MIME = 36, // mime message rfc2045, utf-8
		103
		104	CBOR_TAG_MAGIC = 55799, // self-describe cbor
		105	};
		106
		107	// known forced types, also hardcoded in CBOR.pm
		108	enum
		109	{
		110	AS_CBOR = 0,
		111	AS_INT = 1,
		112	AS_BYTES = 2,
		113	AS_TEXT = 3,
		114	AS_FLOAT16 = 4,
		115	AS_FLOAT32 = 5,
		116	AS_FLOAT64 = 6,
		117	AS_MAP = 7,
		118	// possibly future enhancements: (generic) float, (generic) string
		119	};
		120
14	#define F_SHRINK 0x00000200UL	121	#define F_SHRINK 0x00000001UL
15	#define F_ALLOW_UNKNOWN 0x00002000UL	122	#define F_ALLOW_UNKNOWN 0x00000002UL
		123	#define F_ALLOW_SHARING 0x00000004UL
		124	#define F_ALLOW_CYCLES 0x00000008UL
		125	#define F_ALLOW_WEAK_CYCLES 0x00000010UL
		126	#define F_FORBID_OBJECTS 0x00000020UL
		127	#define F_PACK_STRINGS 0x00000040UL
		128	#define F_TEXT_KEYS 0x00000080UL
		129	#define F_TEXT_STRINGS 0x00000100UL
		130	#define F_VALIDATE_UTF8 0x00000200UL
16		131
17	#define INIT_SIZE 32 // initial scalar size to be allocated	132	#define INIT_SIZE 32 // initial scalar size to be allocated
18		133
19	#define SB do {	134	#define SB do {
20	#define SE } while (0)	135	#define SE } while (0)
…		…
31	#else	146	#else
32	# define CBOR_SLOW 0	147	# define CBOR_SLOW 0
33	# define CBOR_STASH cbor_stash	148	# define CBOR_STASH cbor_stash
34	#endif	149	#endif
35		150
36	static HV *cbor_stash, *cbor_boolean_stash; // CBOR::XS::	151	static HV *cbor_stash, *types_boolean_stash, *types_error_stash, *cbor_tagged_stash; // CBOR::XS::
37	static SV *cbor_true, *cbor_false;	152	static SV *types_true, *types_false, *types_error, *sv_cbor, *default_filter;
38		153
39	typedef struct {	154	typedef struct {
40	U32 flags;	155	U32 flags;
41	U32 max_depth;	156	U32 max_depth;
42	STRLEN max_size;	157	STRLEN max_size;
		158	SV *filter;
43		159
44	SV *cb_object;	160	// for the incremental parser
45	HV *cb_sk_object;	161	STRLEN incr_pos; // the current offset into the text
		162	STRLEN incr_need; // minimum bytes needed to decode
		163	AV *incr_count; // for every nesting level, the number of outstanding values, or -1 for indef.
46	} CBOR;	164	} CBOR;
47		165
48	ecb_inline void	166	ecb_inline void
49	cbor_init (CBOR *cbor)	167	cbor_init (CBOR *cbor)
50	{	168	{
51	Zero (cbor, 1, CBOR);	169	Zero (cbor, 1, CBOR);
52	cbor->max_depth = 512;	170	cbor->max_depth = 512;
		171	}
		172
		173	ecb_inline void
		174	cbor_free (CBOR *cbor)
		175	{
		176	SvREFCNT_dec (cbor->filter);
		177	SvREFCNT_dec (cbor->incr_count);
53	}	178	}
54		179
55	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	181	// utility functions
57		182
…		…
79	SvPV_renew (sv, SvCUR (sv) + 1);	204	SvPV_renew (sv, SvCUR (sv) + 1);
80	#endif	205	#endif
81	}	206	}
82	}	207	}
83		208
84	/////////////////////////////////////////////////////////////////////////////	209	// minimum length of a string to be registered for stringref
85	// fp hell	210	ecb_inline STRLEN
86		211	minimum_string_length (UV idx)
87	//TODO	212	{
		213	return idx <= 23 ? 3
		214	: idx <= 0xffU ? 4
		215	: idx <= 0xffffU ? 5
		216	: idx <= 0xffffffffU ? 7
		217	: 11;
		218	}
88		219
89	/////////////////////////////////////////////////////////////////////////////	220	/////////////////////////////////////////////////////////////////////////////
90	// encoder	221	// encoder
91		222
92	// structure used for encoding CBOR	223	// structure used for encoding CBOR
…		…
95	char *cur; // SvPVX (sv) + current output position	226	char *cur; // SvPVX (sv) + current output position
96	char *end; // SvEND (sv)	227	char *end; // SvEND (sv)
97	SV *sv; // result scalar	228	SV *sv; // result scalar
98	CBOR cbor;	229	CBOR cbor;
99	U32 depth; // recursion level	230	U32 depth; // recursion level
		231	HV *stringref[2]; // string => index, or 0 ([0] = bytes, [1] = utf-8)
		232	UV stringref_idx;
		233	HV *shareable; // ptr => index, or 0
		234	UV shareable_idx;
100	} enc_t;	235	} enc_t;
101		236
102	ecb_inline void	237	ecb_inline void
103	need (enc_t *enc, STRLEN len)	238	need (enc_t *enc, STRLEN len)
104	{	239	{
105	if (ecb_expect_false (enc->cur + len >= enc->end))	240	if (ecb_expect_false ((uintptr_t)(enc->end - enc->cur) < len))
106	{	241	{
107	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);	242	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
108	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);	243	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
109	enc->cur = SvPVX (enc->sv) + cur;	244	enc->cur = SvPVX (enc->sv) + cur;
110	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	245	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
111	}	246	}
112	}	247	}
113		248
		249	static void encode_sv (enc_t *enc, SV *sv);
		250
114	ecb_inline void	251	ecb_inline void
115	encode_ch (enc_t *enc, char ch)	252	encode_ch (enc_t *enc, char ch)
116	{	253	{
117	need (enc, 1);	254	need (enc, 1);
118	*enc->cur++ = ch;	255	*enc->cur++ = ch;
119	}	256	}
120		257
		258	// used for tags, intregers, element counts and so on
121	static void	259	static void
122	encode_uint (enc_t *enc, int major, UV len)	260	encode_uint (enc_t *enc, int major, UV len)
123	{	261	{
124	need (enc, 9);	262	need (enc, 9);
125		263
126	if (len < 24)	264	if (ecb_expect_true (len < LENGTH_EXT1))
127	*enc->cur++ = major | len;	265	*enc->cur++ = major | len;
128	else if (len <= 0xff)	266	else if (ecb_expect_true (len <= 0xffU))
129	{	267	{
130	*enc->cur++ = major | 24;	268	*enc->cur++ = major | LENGTH_EXT1;
131	*enc->cur++ = len;	269	*enc->cur++ = len;
132	}	270	}
133	else if (len <= 0xffff)	271	else if (len <= 0xffffU)
134	{	272	{
135	*enc->cur++ = major | 25;	273	*enc->cur++ = major | LENGTH_EXT2;
136	*enc->cur++ = len >> 8;	274	*enc->cur++ = len >> 8;
137	*enc->cur++ = len;	275	*enc->cur++ = len;
138	}	276	}
139	else if (len <= 0xffffffff)	277	else if (len <= 0xffffffffU)
140	{	278	{
141	*enc->cur++ = major | 26;	279	*enc->cur++ = major | LENGTH_EXT4;
142	*enc->cur++ = len >> 24;	280	*enc->cur++ = len >> 24;
143	*enc->cur++ = len >> 16;	281	*enc->cur++ = len >> 16;
144	*enc->cur++ = len >> 8;	282	*enc->cur++ = len >> 8;
145	*enc->cur++ = len;	283	*enc->cur++ = len;
146	}	284	}
147	else	285	else
148	{	286	{
149	*enc->cur++ = major | 27;	287	*enc->cur++ = major | LENGTH_EXT8;
150	*enc->cur++ = len >> 56;	288	*enc->cur++ = len >> 56;
151	*enc->cur++ = len >> 48;	289	*enc->cur++ = len >> 48;
152	*enc->cur++ = len >> 40;	290	*enc->cur++ = len >> 40;
153	*enc->cur++ = len >> 32;	291	*enc->cur++ = len >> 32;
154	*enc->cur++ = len >> 24;	292	*enc->cur++ = len >> 24;
…		…
156	*enc->cur++ = len >> 8;	294	*enc->cur++ = len >> 8;
157	*enc->cur++ = len;	295	*enc->cur++ = len;
158	}	296	}
159	}	297	}
160		298
161	static void	299	// encodes a perl value into a CBOR integer
		300	ecb_inline void
		301	encode_int (enc_t *enc, SV *sv)
		302	{
		303	if (SvIsUV (sv))
		304	encode_uint (enc, MAJOR_POS_INT, SvUVX (sv));
		305	else if (SvIVX (sv) >= 0)
		306	encode_uint (enc, MAJOR_POS_INT, SvIVX (sv));
		307	else
		308	encode_uint (enc, MAJOR_NEG_INT, -(SvIVX (sv) + 1));
		309	}
		310
		311	ecb_inline void
		312	encode_tag (enc_t *enc, UV tag)
		313	{
		314	encode_uint (enc, MAJOR_TAG, tag);
		315	}
		316
		317	// exceptional (hopefully) slow path for byte strings that need to be utf8-encoded
		318	ecb_noinline static void
162	encode_str (enc_t *enc, int utf8, char *str, STRLEN len)	319	encode_str_utf8 (enc_t *enc, int utf8, char *str, STRLEN len)
163	{	320	{
164	encode_uint (enc, utf8 ? 0x60 : 0x40, len);	321	STRLEN ulen = len;
		322	U8 *p, *pend = (U8 *)str + len;
		323
		324	for (p = (U8 *)str; p < pend; ++p)
		325	ulen += *p >> 7; // count set high bits
		326
		327	encode_uint (enc, MAJOR_TEXT, ulen);
		328
		329	need (enc, ulen);
		330	for (p = (U8 *)str; p < pend; ++p)
		331	if (*p < 0x80)
		332	*enc->cur++ = *p;
		333	else
		334	{
		335	*enc->cur++ = 0xc0 + (*p >> 6);
		336	*enc->cur++ = 0x80 + (*p & 63);
		337	}
		338	}
		339
		340	ecb_inline void
		341	encode_str (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		342	{
		343	if (ecb_expect_false (upgrade_utf8))
		344	if (!utf8)
		345	{
		346	encode_str_utf8 (enc, utf8, str, len);
		347	return;
		348	}
		349
		350	encode_uint (enc, utf8 ? MAJOR_TEXT : MAJOR_BYTES, len);
165	need (enc, len);	351	need (enc, len);
166	memcpy (enc->cur, str, len);	352	memcpy (enc->cur, str, len);
167	enc->cur += len;	353	enc->cur += len;
168	}	354	}
169		355
170	static void encode_sv (enc_t *enc, SV *sv);	356	ecb_inline void
		357	encode_strref (enc_t *enc, int upgrade_utf8, int utf8, char *str, STRLEN len)
		358	{
		359	if (ecb_expect_false (enc->cbor.flags & F_PACK_STRINGS))
		360	{
		361	SV **svp = hv_fetch (enc->stringref[!!utf8], str, len, 1);
		362
		363	if (SvOK (*svp))
		364	{
		365	// already registered, use stringref
		366	encode_tag (enc, CBOR_TAG_STRINGREF);
		367	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		368	return;
		369	}
		370	else if (len >= minimum_string_length (enc->stringref_idx))
		371	{
		372	// register only
		373	sv_setuv (*svp, enc->stringref_idx);
		374	++enc->stringref_idx;
		375	}
		376	}
		377
		378	encode_str (enc, upgrade_utf8, utf8, str, len);
		379	}
		380
		381	ecb_inline void
		382	encode_float16 (enc_t *enc, NV nv)
		383	{
		384	need (enc, 1+2);
		385
		386	*enc->cur++ = MAJOR_MISC | MISC_FLOAT16;
		387
		388	uint16_t fp = ecb_float_to_binary16 (nv);
		389
		390	if (!ecb_big_endian ())
		391	fp = ecb_bswap16 (fp);
		392
		393	memcpy (enc->cur, &fp, 2);
		394	enc->cur += 2;
		395	}
		396
		397	ecb_inline void
		398	encode_float32 (enc_t *enc, NV nv)
		399	{
		400	need (enc, 1+4);
		401
		402	*enc->cur++ = MAJOR_MISC | MISC_FLOAT32;
		403
		404	uint32_t fp = ecb_float_to_binary32 (nv);
		405
		406	if (!ecb_big_endian ())
		407	fp = ecb_bswap32 (fp);
		408
		409	memcpy (enc->cur, &fp, 4);
		410	enc->cur += 4;
		411	}
		412
		413	ecb_inline void
		414	encode_float64 (enc_t *enc, NV nv)
		415	{
		416	need (enc, 1+8);
		417
		418	*enc->cur++ = MAJOR_MISC | MISC_FLOAT64;
		419
		420	uint64_t fp = ecb_double_to_binary64 (nv);
		421
		422	if (!ecb_big_endian ())
		423	fp = ecb_bswap64 (fp);
		424
		425	memcpy (enc->cur, &fp, 8);
		426	enc->cur += 8;
		427	}
		428
		429	ecb_inline void
		430	encode_bool (enc_t *enc, int istrue)
		431	{
		432	encode_ch (enc, istrue ? MAJOR_MISC | SIMPLE_TRUE : MAJOR_MISC | SIMPLE_FALSE);
		433	}
		434
		435	// encodes an arrayref containing key-value pairs as CBOR map
		436	ecb_inline void
		437	encode_array_as_map (enc_t *enc, SV *sv)
		438	{
		439	if (enc->depth >= enc->cbor.max_depth)
		440	croak (ERR_NESTING_EXCEEDED);
		441
		442	++enc->depth;
		443
		444	// as_map does error checking for us, but we re-check in case
		445	// things have changed.
		446
		447	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		448	croak ("CBOR::XS::as_map requires an array reference (did you change the array after calling as_map?)");
		449
		450	AV *av = (AV *)SvRV (sv);
		451	int i, len = av_len (av);
		452
		453	if (!(len & 1))
		454	croak ("CBOR::XS::as_map requires an even number of elements (did you change the array after calling as_map?)");
		455
		456	encode_uint (enc, MAJOR_MAP, (len + 1) >> 1);
		457
		458	for (i = 0; i <= len; ++i)
		459	{
		460	SV **svp = av_fetch (av, i, 0);
		461	encode_sv (enc, svp ? *svp : &PL_sv_undef);
		462	}
		463
		464	--enc->depth;
		465	}
		466
		467	ecb_inline void
		468	encode_forced (enc_t *enc, UV type, SV *sv)
		469	{
		470	switch (type)
		471	{
		472	case AS_CBOR:
		473	{
		474	STRLEN len;
		475	char *str = SvPVbyte (sv, len);
		476
		477	need (enc, len);
		478	memcpy (enc->cur, str, len);
		479	enc->cur += len;
		480	}
		481	break;
		482
		483	case AS_BYTES:
		484	{
		485	STRLEN len;
		486	char *str = SvPVbyte (sv, len);
		487	encode_strref (enc, 0, 0, str, len);
		488	}
		489	break;
		490
		491	case AS_TEXT:
		492	{
		493	STRLEN len;
		494	char *str = SvPVutf8 (sv, len);
		495	encode_strref (enc, 1, 1, str, len);
		496	}
		497	break;
		498
		499	case AS_INT: encode_int (enc, sv); break;
		500
		501	case AS_FLOAT16: encode_float16 (enc, SvNV (sv)); break;
		502	case AS_FLOAT32: encode_float32 (enc, SvNV (sv)); break;
		503	case AS_FLOAT64: encode_float64 (enc, SvNV (sv)); break;
		504
		505	case AS_MAP: encode_array_as_map (enc, sv); break;
		506
		507	default:
		508	croak ("encountered malformed CBOR::XS::Tagged object");
		509	}
		510	}
171		511
172	static void	512	static void
173	encode_av (enc_t *enc, AV *av)	513	encode_av (enc_t *enc, AV *av)
174	{	514	{
175	int i, len = av_len (av);	515	int i, len = av_len (av);
…		…
177	if (enc->depth >= enc->cbor.max_depth)	517	if (enc->depth >= enc->cbor.max_depth)
178	croak (ERR_NESTING_EXCEEDED);	518	croak (ERR_NESTING_EXCEEDED);
179		519
180	++enc->depth;	520	++enc->depth;
181		521
182	encode_uint (enc, 0x80, len + 1);	522	encode_uint (enc, MAJOR_ARRAY, len + 1);
183		523
		524	if (ecb_expect_false (SvMAGICAL (av)))
184	for (i = 0; i <= len; ++i)	525	for (i = 0; i <= len; ++i)
185	{	526	{
186	SV **svp = av_fetch (av, i, 0);	527	SV **svp = av_fetch (av, i, 0);
187	encode_sv (enc, svp ? *svp : &PL_sv_undef);	528	encode_sv (enc, svp ? *svp : &PL_sv_undef);
188	}	529	}
		530	else
		531	for (i = 0; i <= len; ++i)
		532	{
		533	SV *sv = AvARRAY (av)[i];
		534	encode_sv (enc, sv ? sv : &PL_sv_undef);
		535	}
189		536
190	--enc->depth;	537	--enc->depth;
191	}	538	}
192		539
193	static void	540	static void
…		…
201	++enc->depth;	548	++enc->depth;
202		549
203	int pairs = hv_iterinit (hv);	550	int pairs = hv_iterinit (hv);
204	int mg = SvMAGICAL (hv);	551	int mg = SvMAGICAL (hv);
205		552
206	if (mg)	553	if (ecb_expect_false (mg))
207	encode_ch (enc, 0xa0 | 31);	554	encode_ch (enc, MAJOR_MAP | MINOR_INDEF);
208	else	555	else
209	encode_uint (enc, 0xa0, pairs);	556	encode_uint (enc, MAJOR_MAP, pairs);
210		557
211	while ((he = hv_iternext (hv)))	558	while ((he = hv_iternext (hv)))
212	{	559	{
213	if (HeKLEN (he) == HEf_SVKEY)	560	if (HeKLEN (he) == HEf_SVKEY)
214	encode_sv (enc, HeSVKEY (he));	561	encode_sv (enc, HeSVKEY (he));
215	else	562	else
216	encode_str (enc, HeKUTF8 (he), HeKEY (he), HeKLEN (he));	563	encode_strref (enc, enc->cbor.flags & (F_TEXT_KEYS | F_TEXT_STRINGS), HeKUTF8 (he), HeKEY (he), HeKLEN (he));
217		564
218	encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));	565	encode_sv (enc, ecb_expect_false (mg) ? hv_iterval (hv, he) : HeVAL (he));
219	}	566	}
220		567
221	if (mg)	568	if (ecb_expect_false (mg))
222	encode_ch (enc, 0xe0 | 31);	569	encode_ch (enc, MAJOR_MISC | MINOR_INDEF);
223		570
224	--enc->depth;	571	--enc->depth;
225	}	572	}
226		573
227	// encode objects, arrays and special \0=false and \1=true values.	574	// encode objects, arrays and special \0=false and \1=true values.
228	static void	575	static void
229	encode_rv (enc_t *enc, SV *sv)	576	encode_rv (enc_t *enc, SV *sv)
230	{	577	{
231	svtype svt;
232
233	SvGETMAGIC (sv);	578	SvGETMAGIC (sv);
		579
234	svt = SvTYPE (sv);	580	svtype svt = SvTYPE (sv);
235		581
236	if (ecb_expect_false (SvOBJECT (sv)))	582	if (ecb_expect_false (SvOBJECT (sv)))
237	{	583	{
238	HV *stash = !CBOR_SLOW || cbor_boolean_stash	584	HV *boolean_stash = !CBOR_SLOW || types_boolean_stash
239	? cbor_boolean_stash	585	? types_boolean_stash
		586	: gv_stashpv ("Types::Serialiser::Boolean", 1);
		587	HV *error_stash = !CBOR_SLOW || types_error_stash
		588	? types_error_stash
		589	: gv_stashpv ("Types::Serialiser::Error", 1);
		590	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		591	? cbor_tagged_stash
240	: gv_stashpv ("CBOR::XS::Boolean", 1);	592	: gv_stashpv ("CBOR::XS::Tagged" , 1);
241		593
242	if (SvSTASH (sv) == stash)	594	HV *stash = SvSTASH (sv);
243	encode_ch (enc, SvIV (sv) ? 0xe0 | 21 : 0xe0 | 20);	595
		596	if (stash == boolean_stash)
		597	{
		598	encode_bool (enc, SvIV (sv));
		599	return;
		600	}
		601	else if (stash == error_stash)
		602	{
		603	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
		604	return;
		605	}
		606	else if (stash == tagged_stash)
		607	{
		608	if (svt != SVt_PVAV)
		609	croak ("encountered CBOR::XS::Tagged object that isn't an array");
		610
		611	switch (av_len ((AV *)sv))
		612	{
		613	case 2-1:
		614	// actually a tagged value
		615	encode_uint (enc, MAJOR_TAG, SvUV (*av_fetch ((AV *)sv, 0, 1)));
		616	encode_sv (enc, *av_fetch ((AV *)sv, 1, 1));
		617	break;
		618
		619	case 3-1:
		620	// a forced type [value, type, undef]
		621	encode_forced (enc, SvUV (*av_fetch ((AV *)sv, 1, 1)), *av_fetch ((AV *)sv, 0, 1));
		622	break;
		623
		624	default:
		625	croak ("encountered malformed CBOR::XS::Tagged object");
		626	}
		627
		628	return;
		629	}
		630	}
		631
		632	if (ecb_expect_false (SvREFCNT (sv) > 1)
		633	&& ecb_expect_false (enc->cbor.flags & F_ALLOW_SHARING))
		634	{
		635	if (ecb_expect_false (!enc->shareable))
		636	enc->shareable = (HV *)sv_2mortal ((SV *)newHV ());
		637
		638	SV **svp = hv_fetch (enc->shareable, (char *)&sv, sizeof (sv), 1);
		639
		640	if (SvOK (*svp))
		641	{
		642	encode_tag (enc, CBOR_TAG_VALUE_SHAREDREF);
		643	encode_uint (enc, MAJOR_POS_INT, SvUV (*svp));
		644	return;
		645	}
244	else	646	else
245	{	647	{
246	#if 0 //TODO	648	sv_setuv (*svp, enc->shareable_idx);
247	if (enc->cbor.flags & F_CONV_BLESSED)	649	++enc->shareable_idx;
		650	encode_tag (enc, CBOR_TAG_VALUE_SHAREABLE);
		651	}
		652	}
		653
		654	if (ecb_expect_false (SvOBJECT (sv)))
		655	{
		656	HV *stash = SvSTASH (sv);
		657	GV *method;
		658
		659	if (enc->cbor.flags & F_FORBID_OBJECTS)
		660	croak ("encountered object '%s', but forbid_objects is enabled",
		661	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		662	else if ((method = gv_fetchmethod_autoload (stash, "TO_CBOR", 0)))
248	{	663	{
		664	dSP;
		665
		666	ENTER; SAVETMPS;
		667	PUSHMARK (SP);
249	// we re-bless the reference to get overload and other niceties right	668	// we re-bless the reference to get overload and other niceties right
250	GV *to_cbor = gv_fetchmethod_autoload (SvSTASH (sv), "TO_CBOR", 0);
251
252	if (to_cbor)
253	{
254	dSP;
255
256	ENTER; SAVETMPS; PUSHMARK (SP);
257	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	669	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
258		670
259	// calling with G_SCALAR ensures that we always get a 1 return value
260	PUTBACK;	671	PUTBACK;
		672	// G_SCALAR ensures that return value is 1
261	call_sv ((SV *)GvCV (to_cbor), G_SCALAR);	673	call_sv ((SV *)GvCV (method), G_SCALAR);
262	SPAGAIN;	674	SPAGAIN;
263		675
264	// catch this surprisingly common error	676	// catch this surprisingly common error
265	if (SvROK (TOPs) && SvRV (TOPs) == sv)	677	if (SvROK (TOPs) && SvRV (TOPs) == sv)
266	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));	678	croak ("%s::TO_CBOR method returned same object as was passed instead of a new one", HvNAME (stash));
267		679
268	sv = POPs;
269	PUTBACK;
270
271	encode_sv (enc, sv);	680	encode_sv (enc, POPs);
272		681
		682	PUTBACK;
		683
273	FREETMPS; LEAVE;	684	FREETMPS; LEAVE;
274	}	685	}
275	else if (enc->cbor.flags & F_ALLOW_BLESSED)	686	else if ((method = gv_fetchmethod_autoload (stash, "FREEZE", 0)) != 0)
276	encode_str (enc, "null", 4, 0);	687	{
277	else	688	dSP;
278	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_CBOR method available on it",	689
279	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	690	ENTER; SAVETMPS;
		691	PUSHMARK (SP);
		692	EXTEND (SP, 2);
		693	// we re-bless the reference to get overload and other niceties right
		694	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		695	PUSHs (sv_cbor);
		696
		697	PUTBACK;
		698	int count = call_sv ((SV *)GvCV (method), G_ARRAY);
		699	SPAGAIN;
		700
		701	// catch this surprisingly common error
		702	if (count == 1 && SvROK (TOPs) && SvRV (TOPs) == sv)
		703	croak ("%s::FREEZE(CBOR) method returned same object as was passed instead of a new one", HvNAME (stash));
		704
		705	encode_tag (enc, CBOR_TAG_PERL_OBJECT);
		706	encode_uint (enc, MAJOR_ARRAY, count + 1);
		707	encode_strref (enc, 0, HvNAMEUTF8 (stash), HvNAME (stash), HvNAMELEN (stash));
		708
		709	{
		710	int i;
		711
		712	for (i = 0; i < count; ++i)
		713	encode_sv (enc, SP[i + 1 - count]);
		714
		715	SP -= count;
280	}	716	}
281	else if (enc->cbor.flags & F_ALLOW_BLESSED)	717
282	encode_str (enc, "null", 4, 0);	718	PUTBACK;
		719
		720	FREETMPS; LEAVE;
		721	}
283	else	722	else
284	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",	723	croak ("encountered object '%s', but no TO_CBOR or FREEZE methods available on it",
285	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	724	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
286	#endif
287	}
288	}	725	}
289	else if (svt == SVt_PVHV)	726	else if (svt == SVt_PVHV)
290	encode_hv (enc, (HV *)sv);	727	encode_hv (enc, (HV *)sv);
291	else if (svt == SVt_PVAV)	728	else if (svt == SVt_PVAV)
292	encode_av (enc, (AV *)sv);	729	encode_av (enc, (AV *)sv);
293	else if (svt < SVt_PVAV)
294	{
295	STRLEN len = 0;
296	char *pv = svt ? SvPV (sv, len) : 0;
297
298	if (len == 1 && *pv == '1')
299	encode_ch (enc, 0xe0 | 21);
300	else if (len == 1 && *pv == '0')
301	encode_ch (enc, 0xe0 | 20);
302	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
303	encode_ch (enc, 0xe0 | 23);
304	else
305	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
306	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
307	}
308	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
309	encode_ch (enc, 0xe0 | 23);
310	else	730	else
311	croak ("encountered %s, but CBOR can only represent references to arrays or hashes",	731	{
312	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	732	encode_tag (enc, CBOR_TAG_INDIRECTION);
		733	encode_sv (enc, sv);
		734	}
313	}	735	}
314		736
315	static void	737	static void
316	encode_nv (enc_t *enc, SV *sv)	738	encode_nv (enc_t *enc, SV *sv)
317	{	739	{
318	double nv = SvNVX (sv);	740	double nv = SvNVX (sv);
319		741
320	need (enc, 9);	742	need (enc, 9);
321		743
322	if (ecb_expect_false (nv == (U32)nv))	744	if (ecb_expect_false (nv == (NV)(U32)nv))
323	encode_uint (enc, 0x00, (U32)nv);	745	encode_uint (enc, MAJOR_POS_INT, (U32)nv);
324	//TODO: maybe I32?	746	//TODO: maybe I32?
325	else if (ecb_expect_false (nv == (float)nv))	747	else if (ecb_expect_false (nv == (float)nv))
326	{	748	encode_float32 (enc, nv);
327	uint32_t fp = ecb_float_to_binary32 (nv);
328
329	*enc->cur++ = 0xe0 | 26;
330
331	if (!ecb_big_endian ())
332	fp = ecb_bswap32 (fp);
333
334	memcpy (enc->cur, &fp, 4);
335	enc->cur += 4;
336	}
337	else	749	else
338	{	750	encode_float64 (enc, nv);
339	uint64_t fp = ecb_double_to_binary64 (nv);
340
341	*enc->cur++ = 0xe0 | 27;
342
343	if (!ecb_big_endian ())
344	fp = ecb_bswap64 (fp);
345
346	memcpy (enc->cur, &fp, 8);
347	enc->cur += 8;
348	}
349	}	751	}
350		752
351	static void	753	static void
352	encode_sv (enc_t *enc, SV *sv)	754	encode_sv (enc_t *enc, SV *sv)
353	{	755	{
…		…
355		757
356	if (SvPOKp (sv))	758	if (SvPOKp (sv))
357	{	759	{
358	STRLEN len;	760	STRLEN len;
359	char *str = SvPV (sv, len);	761	char *str = SvPV (sv, len);
360	encode_str (enc, SvUTF8 (sv), str, len);	762	encode_strref (enc, enc->cbor.flags & F_TEXT_STRINGS, SvUTF8 (sv), str, len);
361	}	763	}
362	else if (SvNOKp (sv))	764	else if (SvNOKp (sv))
363	encode_nv (enc, sv);	765	encode_nv (enc, sv);
364	else if (SvIOKp (sv))	766	else if (SvIOKp (sv))
365	{	767	encode_int (enc, sv);
366	if (SvIsUV (sv))
367	encode_uint (enc, 0x00, SvUVX (sv));
368	else if (SvIVX (sv) >= 0)
369	encode_uint (enc, 0x00, SvIVX (sv));
370	else
371	encode_uint (enc, 0x20, -(SvIVX (sv) + 1));
372	}
373	else if (SvROK (sv))	768	else if (SvROK (sv))
374	encode_rv (enc, SvRV (sv));	769	encode_rv (enc, SvRV (sv));
375	else if (!SvOK (sv))	770	else if (!SvOK (sv))
376	encode_ch (enc, 0xe0 | 22);	771	encode_ch (enc, MAJOR_MISC | SIMPLE_NULL);
377	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)	772	else if (enc->cbor.flags & F_ALLOW_UNKNOWN)
378	encode_ch (enc, 0xe0 | 23);	773	encode_ch (enc, MAJOR_MISC | SIMPLE_UNDEF);
379	else	774	else
380	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",	775	croak ("encountered perl type (%s,0x%x) that CBOR cannot handle, check your input data",
381	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));	776	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
382	}	777	}
383		778
384	static SV *	779	static SV *
385	encode_cbor (SV *scalar, CBOR *cbor)	780	encode_cbor (SV *scalar, CBOR *cbor)
386	{	781	{
387	enc_t enc;	782	enc_t enc = { 0 };
388		783
389	enc.cbor = *cbor;	784	enc.cbor = *cbor;
390	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	785	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
391	enc.cur = SvPVX (enc.sv);	786	enc.cur = SvPVX (enc.sv);
392	enc.end = SvEND (enc.sv);	787	enc.end = SvEND (enc.sv);
393	enc.depth = 0;
394		788
395	SvPOK_only (enc.sv);	789	SvPOK_only (enc.sv);
		790
		791	if (cbor->flags & F_PACK_STRINGS)
		792	{
		793	encode_tag (&enc, CBOR_TAG_STRINGREF_NAMESPACE);
		794	enc.stringref[0]= (HV *)sv_2mortal ((SV *)newHV ());
		795	enc.stringref[1]= (HV *)sv_2mortal ((SV *)newHV ());
		796	}
		797
396	encode_sv (&enc, scalar);	798	encode_sv (&enc, scalar);
397		799
398	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	800	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
399	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	801	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
400		802
…		…
414	U8 *end; // end of input string	816	U8 *end; // end of input string
415	const char *err; // parse error, if != 0	817	const char *err; // parse error, if != 0
416	CBOR cbor;	818	CBOR cbor;
417	U32 depth; // recursion depth	819	U32 depth; // recursion depth
418	U32 maxdepth; // recursion depth limit	820	U32 maxdepth; // recursion depth limit
		821	AV *shareable;
		822	AV *stringref;
		823	SV *decode_tagged;
		824	SV *err_sv; // optional sv for error, needs to be freed
419	} dec_t;	825	} dec_t;
420		826
421	#define ERR(reason) SB if (!dec->err) dec->err = reason; goto fail; SE	827	// set dec->err to ERRSV
		828	ecb_cold static void
		829	err_errsv (dec_t *dec)
		830	{
		831	if (!dec->err)
		832	{
		833	dec->err_sv = newSVsv (ERRSV);
422		834
423	#define WANT(len) if (ecb_expect_false (dec->cur + len > dec->end)) ERR ("unexpected end of CBOR data")	835	// chop off the trailing \n
		836	SvCUR_set (dec->err_sv, SvCUR (dec->err_sv) - 1);
		837	*SvEND (dec->err_sv) = 0;
424		838
		839	dec->err = SvPVutf8_nolen (dec->err_sv);
		840	}
		841	}
		842
		843	// the following functions are used to reduce code size and help the compiler to optimise
		844	ecb_cold static void
		845	err_set (dec_t *dec, const char *reason)
		846	{
		847	if (!dec->err)
		848	dec->err = reason;
		849	}
		850
		851	ecb_cold static void
		852	err_unexpected_end (dec_t *dec)
		853	{
		854	err_set (dec, "unexpected end of CBOR data");
		855	}
		856
		857	#define ERR_DO(do) SB do; goto fail; SE
		858	#define ERR(reason) ERR_DO (err_set (dec, reason))
		859	#define ERR_ERRSV ERR_DO (err_errsv (dec))
		860
		861	#define WANT(len) if (ecb_expect_false ((uintptr_t)(dec->end - dec->cur) < (STRLEN)len)) ERR_DO (err_unexpected_end (dec))
		862
425	#define DEC_INC_DEPTH if (++dec->depth > dec->cbor.max_depth) ERR (ERR_NESTING_EXCEEDED)	863	#define DEC_INC_DEPTH if (ecb_expect_false (++dec->depth > dec->cbor.max_depth)) ERR (ERR_NESTING_EXCEEDED)
426	#define DEC_DEC_DEPTH --dec->depth	864	#define DEC_DEC_DEPTH --dec->depth
427		865
428	static UV	866	static UV
429	decode_uint (dec_t *dec)	867	decode_uint (dec_t *dec)
430	{	868	{
431	switch (*dec->cur & 31)	869	U8 m = *dec->cur & MINOR_MASK;
432	{	870	++dec->cur;
433	case 0: case 1: case 2: case 3: case 4: case 5: case 6: case 7:
434	case 8: case 9: case 10: case 11: case 12: case 13: case 14: case 15:
435	case 16: case 17: case 18: case 19: case 20: case 21: case 22: case 23:
436	return *dec->cur++ & 31;
437		871
438	case 24:	872	if (ecb_expect_true (m < LENGTH_EXT1))
		873	return m;
		874	else if (ecb_expect_true (m == LENGTH_EXT1))
		875	{
439	WANT (2);	876	WANT (1);
440	dec->cur += 2;	877	dec->cur += 1;
441	return dec->cur[-1];	878	return dec->cur[-1];
442		879	}
443	case 25:	880	else if (ecb_expect_true (m == LENGTH_EXT2))
		881	{
444	WANT (3);	882	WANT (2);
445	dec->cur += 3;	883	dec->cur += 2;
446	return (((UV)dec->cur[-2]) << 8)	884	return (((UV)dec->cur[-2]) << 8)
447	| ((UV)dec->cur[-1]);	885	| ((UV)dec->cur[-1]);
448		886	}
449	case 26:	887	else if (ecb_expect_true (m == LENGTH_EXT4))
		888	{
450	WANT (5);	889	WANT (4);
451	dec->cur += 5;	890	dec->cur += 4;
452	return (((UV)dec->cur[-4]) << 24)	891	return (((UV)dec->cur[-4]) << 24)
453	| (((UV)dec->cur[-3]) << 16)	892	| (((UV)dec->cur[-3]) << 16)
454	| (((UV)dec->cur[-2]) << 8)	893	| (((UV)dec->cur[-2]) << 8)
455	| ((UV)dec->cur[-1]);	894	| ((UV)dec->cur[-1]);
456		895	}
457	case 27:	896	else if (ecb_expect_true (m == LENGTH_EXT8))
		897	{
458	WANT (9);	898	WANT (8);
459	dec->cur += 9;	899	dec->cur += 8;
		900
		901	return
		902	#if UVSIZE < 8
		903	0
		904	#else
460	return (((UV)dec->cur[-8]) << 56)	905	(((UV)dec->cur[-8]) << 56)
461	| (((UV)dec->cur[-7]) << 48)	906	| (((UV)dec->cur[-7]) << 48)
462	| (((UV)dec->cur[-6]) << 40)	907	| (((UV)dec->cur[-6]) << 40)
463	| (((UV)dec->cur[-5]) << 32)	908	| (((UV)dec->cur[-5]) << 32)
		909	#endif
464	| (((UV)dec->cur[-4]) << 24)	910	| (((UV)dec->cur[-4]) << 24)
465	| (((UV)dec->cur[-3]) << 16)	911	| (((UV)dec->cur[-3]) << 16)
466	| (((UV)dec->cur[-2]) << 8)	912	| (((UV)dec->cur[-2]) << 8)
467	| ((UV)dec->cur[-1]);	913	| ((UV)dec->cur[-1]);
468		914	}
469	default:	915	else
470	ERR ("corrupted CBOR data (unsupported integer minor encoding)");	916	ERR ("corrupted CBOR data (unsupported integer minor encoding)");
471	}
472		917
473	fail:	918	fail:
474	return 0;	919	return 0;
475	}	920	}
476		921
…		…
481	{	926	{
482	AV *av = newAV ();	927	AV *av = newAV ();
483		928
484	DEC_INC_DEPTH;	929	DEC_INC_DEPTH;
485		930
486	if ((*dec->cur & 31) == 31)	931	if (*dec->cur == (MAJOR_ARRAY | MINOR_INDEF))
487	{	932	{
488	++dec->cur;	933	++dec->cur;
489		934
490	for (;;)	935	for (;;)
491	{	936	{
492	WANT (1);	937	WANT (1);
493		938
494	if (*dec->cur == (0xe0 | 31))	939	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
495	{	940	{
496	++dec->cur;	941	++dec->cur;
497	break;	942	break;
498	}	943	}
499		944
500	av_push (av, decode_sv (dec));	945	av_push (av, decode_sv (dec));
501	}	946	}
502	}	947	}
503	else	948	else
504	{	949	{
505	int i, len = decode_uint (dec);	950	UV i, len = decode_uint (dec);
506		951
		952	WANT (len); // complexity check for av_fill - need at least one byte per value, do not allow supersize arrays
507	av_fill (av, len - 1);	953	av_fill (av, len - 1);
508		954
509	for (i = 0; i < len; ++i)	955	for (i = 0; i < len; ++i)
510	AvARRAY (av)[i] = decode_sv (dec);	956	AvARRAY (av)[i] = decode_sv (dec);
511	}	957	}
512		958
513	DEC_DEC_DEPTH;	959	DEC_DEC_DEPTH;
514	return newRV_noinc ((SV *)av);	960	return newRV_noinc ((SV *)av);
515		961
516	fail:	962	fail:
517	SvREFCNT_dec (av);	963	SvREFCNT_dec_NN (av);
518	DEC_DEC_DEPTH;	964	DEC_DEC_DEPTH;
519	return &PL_sv_undef;	965	return &PL_sv_undef;
520	}	966	}
521		967
		968	static void
		969	decode_he (dec_t *dec, HV *hv)
		970	{
		971	// for speed reasons, we specialcase single-string
		972	// byte or utf-8 strings as keys, but only when !stringref
		973
		974	if (ecb_expect_true (!dec->stringref))
		975	if (ecb_expect_true ((U8)(*dec->cur - MAJOR_BYTES) <= LENGTH_EXT8))
		976	{
		977	STRLEN len = decode_uint (dec);
		978	char *key = (char *)dec->cur;
		979
		980	WANT (len);
		981	dec->cur += len;
		982
		983	hv_store (hv, key, len, decode_sv (dec), 0);
		984
		985	return;
		986	}
		987	else if (ecb_expect_true ((U8)(*dec->cur - MAJOR_TEXT) <= LENGTH_EXT8))
		988	{
		989	STRLEN len = decode_uint (dec);
		990	char *key = (char *)dec->cur;
		991
		992	WANT (len);
		993	dec->cur += len;
		994
		995	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		996	if (!cbor_is_utf8_string ((U8 *)key, len))
		997	ERR ("corrupted CBOR data (invalid UTF-8 in map key)");
		998
		999	hv_store (hv, key, -len, decode_sv (dec), 0);
		1000
		1001	return;
		1002	}
		1003
		1004	SV *k = decode_sv (dec);
		1005	SV *v = decode_sv (dec);
		1006
		1007	// we leak memory if uncaught exceptions are thrown by random magical
		1008	// methods, and this is hopefully the only place where it can happen,
		1009	// so if there is a chance of an exception, take the very slow path.
		1010	// since catching exceptions is "undocumented/internal/forbidden" by
		1011	// the new p5p powers, we need to call out to a perl function :/
		1012	if (ecb_expect_false (SvAMAGIC (k)))
		1013	{
		1014	dSP;
		1015
		1016	ENTER; SAVETMPS;
		1017	PUSHMARK (SP);
		1018	EXTEND (SP, 3);
		1019	PUSHs (sv_2mortal (newRV_inc ((SV *)hv)));
		1020	PUSHs (sv_2mortal (k));
		1021	PUSHs (sv_2mortal (v));
		1022
		1023	PUTBACK;
		1024	call_pv ("CBOR::XS::_hv_store", G_VOID | G_DISCARD | G_EVAL);
		1025	SPAGAIN;
		1026
		1027	FREETMPS; LEAVE;
		1028
		1029	if (SvTRUE (ERRSV))
		1030	ERR_ERRSV;
		1031
		1032	return;
		1033	}
		1034
		1035	hv_store_ent (hv, k, v, 0);
		1036	SvREFCNT_dec_NN (k);
		1037
		1038	fail:
		1039	;
		1040	}
		1041
522	static SV *	1042	static SV *
523	decode_hv (dec_t *dec)	1043	decode_hv (dec_t *dec)
524	{	1044	{
525	HV *hv = newHV ();	1045	HV *hv = newHV ();
526		1046
527	DEC_INC_DEPTH;	1047	DEC_INC_DEPTH;
528		1048
529	if ((*dec->cur & 31) == 31)	1049	if (*dec->cur == (MAJOR_MAP | MINOR_INDEF))
530	{	1050	{
531	++dec->cur;	1051	++dec->cur;
532		1052
533	for (;;)	1053	for (;;)
534	{	1054	{
535	WANT (1);	1055	WANT (1);
536		1056
537	if (*dec->cur == (0xe0 | 31))	1057	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF) || dec->err)
538	{	1058	{
539	++dec->cur;	1059	++dec->cur;
540	break;	1060	break;
541	}	1061	}
542		1062
543	SV *k = decode_sv (dec);	1063	decode_he (dec, hv);
544	SV *v = decode_sv (dec);
545
546	hv_store_ent (hv, k, v, 0);
547	}	1064	}
548	}	1065	}
549	else	1066	else
550	{	1067	{
551	int len = decode_uint (dec);	1068	UV pairs = decode_uint (dec);
552		1069
		1070	WANT (pairs); // complexity check - need at least one byte per value, do not allow supersize hashes
		1071
553	while (len--)	1072	while (pairs--)
554	{	1073	decode_he (dec, hv);
555	SV *k = decode_sv (dec);
556	SV *v = decode_sv (dec);
557
558	hv_store_ent (hv, k, v, 0);
559	}
560	}	1074	}
561		1075
562	DEC_DEC_DEPTH;	1076	DEC_DEC_DEPTH;
563	return newRV_noinc ((SV *)hv);	1077	return newRV_noinc ((SV *)hv);
564		1078
565	#if 0	1079	fail:
566	SV *sv;	1080	SvREFCNT_dec_NN (hv);
567	HV *hv = newHV ();
568
569	DEC_INC_DEPTH;	1081	DEC_DEC_DEPTH;
570	decode_ws (dec);	1082	return &PL_sv_undef;
		1083	}
571		1084
572	for (;;)	1085	static SV *
573	{	1086	decode_str (dec_t *dec, int utf8)
574	// heuristic: assume that	1087	{
575	// a) decode_str + hv_store_ent are abysmally slow.	1088	SV *sv = 0;
576	// b) most hash keys are short, simple ascii text.	1089
577	// => try to "fast-match" such strings to avoid	1090	if (ecb_expect_false ((*dec->cur & MINOR_MASK) == MINOR_INDEF))
578	// the overhead of decode_str + hv_store_ent.
579	{	1091	{
580	SV *value;	1092	// indefinite length strings
581	char *p = dec->cur;	1093	++dec->cur;
582	char *e = p + 24; // only try up to 24 bytes
583		1094
		1095	U8 major = *dec->cur & MAJOR_MISC;
		1096
		1097	sv = newSVpvn ("", 0);
		1098
584	for (;;)	1099	for (;;)
585	{	1100	{
586	// the >= 0x80 is false on most architectures	1101	WANT (1);
587	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1102
		1103	if ((*dec->cur - major) > LENGTH_EXT8)
		1104	if (*dec->cur == (MAJOR_MISC | MINOR_INDEF))
588	{	1105	{
589	// slow path, back up and use decode_str	1106	++dec->cur;
590	SV *key = decode_str (dec);
591	if (!key)
592	goto fail;
593
594	decode_ws (dec); EXPECT_CH (':');
595
596	decode_ws (dec);
597	value = decode_sv (dec);
598	if (!value)
599	{
600	SvREFCNT_dec (key);
601	goto fail;
602	}
603
604	hv_store_ent (hv, key, value, 0);
605	SvREFCNT_dec (key);
606
607	break;	1107	break;
608	}	1108	}
609	else if (*p == '"')
610	{
611	// fast path, got a simple key
612	char *key = dec->cur;
613	int len = p - key;
614	dec->cur = p + 1;
615
616	decode_ws (dec); EXPECT_CH (':');
617
618	decode_ws (dec);
619	value = decode_sv (dec);
620	if (!value)
621	goto fail;
622
623	hv_store (hv, key, len, value, 0);
624
625	break;
626	}
627
628	++p;
629	}
630	}
631
632	decode_ws (dec);
633
634	if (*dec->cur == '}')
635	{
636	++dec->cur;
637	break;
638	}
639
640	if (*dec->cur != ',')
641	ERR (", or } expected while parsing object/hash");
642
643	++dec->cur;
644
645	decode_ws (dec);
646
647	if (*dec->cur == '}' && dec->cbor.flags & F_RELAXED)
648	{
649	++dec->cur;
650	break;
651	}
652	}
653
654	DEC_DEC_DEPTH;
655	sv = newRV_noinc ((SV *)hv);
656
657	// check filter callbacks
658	if (dec->cbor.flags & F_HOOK)
659	{
660	if (dec->cbor.cb_sk_object && HvKEYS (hv) == 1)
661	{
662	HE *cb, *he;
663
664	hv_iterinit (hv);
665	he = hv_iternext (hv);
666	hv_iterinit (hv);
667
668	// the next line creates a mortal sv each time its called.
669	// might want to optimise this for common cases.
670	cb = hv_fetch_ent (dec->cbor.cb_sk_object, hv_iterkeysv (he), 0, 0);
671
672	if (cb)
673	{	1109	else
674	dSP;	1110	ERR ("corrupted CBOR data (invalid chunks in indefinite length string)");
675	int count;
676		1111
677	ENTER; SAVETMPS; PUSHMARK (SP);	1112	STRLEN len = decode_uint (dec);
678	XPUSHs (HeVAL (he));
679	sv_2mortal (sv);
680		1113
681	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
682
683	if (count == 1)
684	{
685	sv = newSVsv (POPs);
686	FREETMPS; LEAVE;
687	return sv;
688	}
689
690	SvREFCNT_inc (sv);
691	FREETMPS; LEAVE;
692	}
693	}
694
695	if (dec->cbor.cb_object)
696	{
697	dSP;
698	int count;
699
700	ENTER; SAVETMPS; PUSHMARK (SP);
701	XPUSHs (sv_2mortal (sv));
702
703	PUTBACK; count = call_sv (dec->cbor.cb_object, G_ARRAY); SPAGAIN;
704
705	if (count == 1)
706	{
707	sv = newSVsv (POPs);
708	FREETMPS; LEAVE;
709	return sv;
710	}
711
712	SvREFCNT_inc (sv);
713	FREETMPS; LEAVE;
714	}
715	}
716
717	return sv;
718	#endif
719
720	fail:
721	SvREFCNT_dec (hv);
722	DEC_DEC_DEPTH;
723	return &PL_sv_undef;
724	}
725
726	static SV *
727	decode_str (dec_t *dec, int utf8)
728	{
729	SV *sv;
730
731	if ((*dec->cur & 31) == 31)
732	{
733	++dec->cur;
734
735	sv = newSVpvn ("", 0);
736
737	// not very fast, and certainly not robust against illegal input
738	for (;;)
739	{
740	WANT (1);	1114	WANT (len);
741		1115	sv_catpvn (sv, dec->cur, len);
742	if (*dec->cur == (0xe0 | 31))	1116	dec->cur += len;
743	{
744	++dec->cur;
745	break;
746	}
747
748	SV *sv2 = decode_sv (dec);
749	sv_catsv (sv, sv2);
750	}	1117	}
751	}	1118	}
752	else	1119	else
753	{	1120	{
754	STRLEN len = decode_uint (dec);	1121	STRLEN len = decode_uint (dec);
755		1122
756	WANT (len);	1123	WANT (len);
757	sv = newSVpvn (dec->cur, len);	1124	sv = newSVpvn (dec->cur, len);
758	dec->cur += len;	1125	dec->cur += len;
		1126
		1127	if (ecb_expect_false (dec->stringref)
		1128	&& SvCUR (sv) >= minimum_string_length (AvFILLp (dec->stringref) + 1))
		1129	av_push (dec->stringref, SvREFCNT_inc_NN (sv));
759	}	1130	}
760		1131
761	if (utf8)	1132	if (utf8)
		1133	{
		1134	if (ecb_expect_false (dec->cbor.flags & F_VALIDATE_UTF8))
		1135	if (!cbor_is_utf8_string (SvPVX (sv), SvCUR (sv)))
		1136	ERR ("corrupted CBOR data (invalid UTF-8 in text string)");
		1137
762	SvUTF8_on (sv);	1138	SvUTF8_on (sv);
		1139	}
763		1140
764	return sv;	1141	return sv;
765		1142
766	fail:	1143	fail:
		1144	SvREFCNT_dec (sv);
767	return &PL_sv_undef;	1145	return &PL_sv_undef;
768	}	1146	}
769		1147
770	static SV *	1148	static SV *
771	decode_tagged (dec_t *dec)	1149	decode_tagged (dec_t *dec)
772	{	1150	{
		1151	SV *sv = 0;
773	UV tag = decode_uint (dec);	1152	UV tag = decode_uint (dec);
		1153
		1154	WANT (1);
		1155
		1156	switch (tag)
		1157	{
		1158	case CBOR_TAG_MAGIC:
774	SV *sv = decode_sv (dec);	1159	sv = decode_sv (dec);
		1160	break;
775		1161
776	if (tag == 55799) // 2.4.5 Self-Describe CBOR	1162	case CBOR_TAG_INDIRECTION:
		1163	sv = newRV_noinc (decode_sv (dec));
		1164	break;
		1165
		1166	case CBOR_TAG_STRINGREF_NAMESPACE:
		1167	{
		1168	// do not use SAVETMPS/FREETMPS, as these will
		1169	// erase mortalised caches, e.g. "shareable"
		1170	ENTER;
		1171
		1172	SAVESPTR (dec->stringref);
		1173	dec->stringref = (AV *)sv_2mortal ((SV *)newAV ());
		1174
		1175	sv = decode_sv (dec);
		1176
		1177	LEAVE;
		1178	}
		1179	break;
		1180
		1181	case CBOR_TAG_STRINGREF:
		1182	{
		1183	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1184	ERR ("corrupted CBOR data (stringref index not an unsigned integer)");
		1185
		1186	UV idx = decode_uint (dec);
		1187
		1188	if (!dec->stringref || idx >= (UV)(1 + AvFILLp (dec->stringref)))
		1189	ERR ("corrupted CBOR data (stringref index out of bounds or outside namespace)");
		1190
		1191	sv = newSVsv (AvARRAY (dec->stringref)[idx]);
		1192	}
		1193	break;
		1194
		1195	case CBOR_TAG_VALUE_SHAREABLE:
		1196	{
		1197	if (ecb_expect_false (!dec->shareable))
		1198	dec->shareable = (AV *)sv_2mortal ((SV *)newAV ());
		1199
		1200	if (ecb_expect_false (dec->cbor.flags & (F_ALLOW_CYCLES | F_ALLOW_WEAK_CYCLES)))
		1201	{
		1202	// if cycles are allowed, then we store an AV as value
		1203	// while it is being decoded, and gather unresolved
		1204	// references in it, to be re4solved after decoding.
		1205	int idx, i;
		1206	AV *av = newAV ();
		1207	av_push (dec->shareable, (SV *)av);
		1208	idx = AvFILLp (dec->shareable);
		1209
		1210	sv = decode_sv (dec);
		1211
		1212	// the AV now contains \undef for all unresolved references,
		1213	// so we fix them up here.
		1214	for (i = 0; i <= AvFILLp (av); ++i)
		1215	SvRV_set (AvARRAY (av)[i], SvREFCNT_inc_NN (SvRV (sv)));
		1216
		1217	// weaken all recursive references
		1218	if (dec->cbor.flags & F_ALLOW_WEAK_CYCLES)
		1219	for (i = 0; i <= AvFILLp (av); ++i)
		1220	sv_rvweaken (AvARRAY (av)[i]);
		1221
		1222	// now replace the AV by a reference to the completed value
		1223	SvREFCNT_dec_NN ((SV *)av);
		1224	AvARRAY (dec->shareable)[idx] = SvREFCNT_inc_NN (sv);
		1225	}
		1226	else
		1227	{
		1228	av_push (dec->shareable, &PL_sv_undef);
		1229	int idx = AvFILLp (dec->shareable);
		1230	sv = decode_sv (dec);
		1231	AvARRAY (dec->shareable)[idx] = SvREFCNT_inc_NN (sv);
		1232	}
		1233	}
		1234	break;
		1235
		1236	case CBOR_TAG_VALUE_SHAREDREF:
		1237	{
		1238	if ((*dec->cur >> MAJOR_SHIFT) != (MAJOR_POS_INT >> MAJOR_SHIFT))
		1239	ERR ("corrupted CBOR data (sharedref index not an unsigned integer)");
		1240
		1241	UV idx = decode_uint (dec);
		1242
		1243	if (!dec->shareable || idx >= (UV)(1 + AvFILLp (dec->shareable)))
		1244	ERR ("corrupted CBOR data (sharedref index out of bounds)");
		1245
		1246	sv = AvARRAY (dec->shareable)[idx];
		1247
		1248	// reference to cycle, we create a new \undef and use that, and also
		1249	// registerr it in the AV for later fixing
		1250	if (ecb_expect_false (SvTYPE (sv) == SVt_PVAV))
		1251	{
		1252	AV *av = (AV *)sv;
		1253	sv = newRV_noinc (&PL_sv_undef);
		1254	av_push (av, SvREFCNT_inc_NN (sv));
		1255	}
		1256	else if (ecb_expect_false (sv == &PL_sv_undef)) // not yet decoded, but cycles not allowed
		1257	ERR ("cyclic CBOR data structure found, but allow_cycles is not enabled");
		1258	else // we decoded the object earlier, no cycle
		1259	sv = newSVsv (sv);
		1260	}
		1261	break;
		1262
		1263	case CBOR_TAG_PERL_OBJECT:
		1264	{
		1265	if (dec->cbor.flags & F_FORBID_OBJECTS)
		1266	goto filter;
		1267
		1268	sv = decode_sv (dec);
		1269
		1270	if (!SvROK (sv) || SvTYPE (SvRV (sv)) != SVt_PVAV)
		1271	ERR ("corrupted CBOR data (non-array perl object)");
		1272
		1273	AV *av = (AV *)SvRV (sv);
		1274	int len = av_len (av) + 1;
		1275	HV *stash = gv_stashsv (*av_fetch (av, 0, 1), 0);
		1276
		1277	if (!stash)
		1278	ERR ("cannot decode perl-object (package does not exist)");
		1279
		1280	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1281
		1282	if (!method)
		1283	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1284
		1285	dSP;
		1286
		1287	ENTER; SAVETMPS;
		1288	PUSHMARK (SP);
		1289	EXTEND (SP, len + 1);
		1290	// we re-bless the reference to get overload and other niceties right
		1291	PUSHs (*av_fetch (av, 0, 1));
		1292	PUSHs (sv_cbor);
		1293
		1294	int i;
		1295
		1296	for (i = 1; i < len; ++i)
		1297	PUSHs (*av_fetch (av, i, 1));
		1298
		1299	PUTBACK;
		1300	call_sv ((SV *)GvCV (method), G_SCALAR | G_EVAL);
		1301	SPAGAIN;
		1302
		1303	if (SvTRUE (ERRSV))
		1304	{
		1305	FREETMPS; LEAVE;
		1306	ERR_ERRSV;
		1307	}
		1308
		1309	SvREFCNT_dec_NN (sv);
		1310	sv = SvREFCNT_inc (POPs);
		1311
		1312	PUTBACK;
		1313
		1314	FREETMPS; LEAVE;
		1315	}
		1316	break;
		1317
		1318	default:
		1319	filter:
		1320	{
		1321	SV *tag_sv = newSVuv (tag);
		1322
		1323	sv = decode_sv (dec);
		1324
		1325	dSP;
		1326	ENTER; SAVETMPS;
		1327	PUSHMARK (SP);
		1328	EXTEND (SP, 2);
		1329	PUSHs (tag_sv);
		1330	PUSHs (sv);
		1331
		1332	PUTBACK;
		1333	int count = call_sv (dec->cbor.filter ? dec->cbor.filter : default_filter, G_ARRAY | G_EVAL);
		1334	SPAGAIN;
		1335
		1336	if (SvTRUE (ERRSV))
		1337	{
		1338	SvREFCNT_dec_NN (tag_sv);
		1339	FREETMPS; LEAVE;
		1340	ERR_ERRSV;
		1341	}
		1342
		1343	if (count)
		1344	{
		1345	SvREFCNT_dec_NN (tag_sv);
		1346	SvREFCNT_dec_NN (sv);
		1347	sv = SvREFCNT_inc_NN (TOPs);
		1348	SP -= count;
		1349	}
		1350	else
		1351	{
		1352	AV *av = newAV ();
		1353	av_push (av, tag_sv);
		1354	av_push (av, sv);
		1355
		1356	HV *tagged_stash = !CBOR_SLOW || cbor_tagged_stash
		1357	? cbor_tagged_stash
		1358	: gv_stashpv ("CBOR::XS::Tagged" , 1);
		1359	sv = sv_bless (newRV_noinc ((SV *)av), tagged_stash);
		1360	}
		1361
		1362	PUTBACK;
		1363
		1364	FREETMPS; LEAVE;
		1365	}
		1366	break;
		1367	}
		1368
777	return sv;	1369	return sv;
778		1370
779	AV *av = newAV ();	1371	fail:
780	av_push (av, newSVuv (tag));	1372	SvREFCNT_dec (sv);
781	av_push (av, sv);	1373	return &PL_sv_undef;
782	return newRV_noinc ((SV *)av);
783	}	1374	}
784		1375
785	static SV *	1376	static SV *
786	decode_sv (dec_t *dec)	1377	decode_sv (dec_t *dec)
787	{	1378	{
788	WANT (1);	1379	WANT (1);
789		1380
790	switch (*dec->cur >> 5)	1381	switch (*dec->cur >> MAJOR_SHIFT)
791	{	1382	{
792	case 0: // unsigned int	1383	case MAJOR_POS_INT >> MAJOR_SHIFT: return newSVuv (decode_uint (dec));
793	//TODO: 64 bit values on 3 2bit perls	1384	case MAJOR_NEG_INT >> MAJOR_SHIFT: return newSViv (-1 - (IV)decode_uint (dec));
794	return newSVuv (decode_uint (dec));	1385	case MAJOR_BYTES >> MAJOR_SHIFT: return decode_str (dec, 0);
795	case 1: // negative int	1386	case MAJOR_TEXT >> MAJOR_SHIFT: return decode_str (dec, 1);
796	return newSViv (-1 - (IV)decode_uint (dec));	1387	case MAJOR_ARRAY >> MAJOR_SHIFT: return decode_av (dec);
797	case 2: // octet string	1388	case MAJOR_MAP >> MAJOR_SHIFT: return decode_hv (dec);
798	return decode_str (dec, 0);	1389	case MAJOR_TAG >> MAJOR_SHIFT: return decode_tagged (dec);
799	case 3: // utf-8 string	1390
800	return decode_str (dec, 1);	1391	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	return decode_tagged (dec);
807	case 7: // misc
808	switch (*dec->cur++ & 31)	1392	switch (*dec->cur++ & MINOR_MASK)
809	{	1393	{
810	case 20:	1394	case SIMPLE_FALSE:
811	#if CBOR_SLOW	1395	#if CBOR_SLOW
812	cbor_false = get_bool ("CBOR::XS::false");	1396	types_false = get_bool ("Types::Serialiser::false");
813	#endif	1397	#endif
814	return newSVsv (cbor_false);	1398	return newSVsv (types_false);
815	case 21:	1399	case SIMPLE_TRUE:
816	#if CBOR_SLOW	1400	#if CBOR_SLOW
817	cbor_true = get_bool ("CBOR::XS::true");	1401	types_true = get_bool ("Types::Serialiser::true");
818	#endif	1402	#endif
819	return newSVsv (cbor_true);	1403	return newSVsv (types_true);
820	case 22:	1404	case SIMPLE_NULL:
821	return newSVsv (&PL_sv_undef);	1405	return newSVsv (&PL_sv_undef);
		1406	case SIMPLE_UNDEF:
		1407	#if CBOR_SLOW
		1408	types_error = get_bool ("Types::Serialiser::error");
		1409	#endif
		1410	return newSVsv (types_error);
822		1411
823	case 25:	1412	case MISC_FLOAT16:
824	{	1413	{
825	WANT (2);	1414	WANT (2);
826		1415
827	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];	1416	uint16_t fp = (dec->cur[0] << 8) | dec->cur[1];
828	dec->cur += 2;	1417	dec->cur += 2;
829		1418
830	return newSVnv (ecb_binary16_to_float (fp));	1419	return newSVnv (ecb_binary16_to_float (fp));
831	}	1420	}
832		1421
833	case 26:	1422	case MISC_FLOAT32:
834	{	1423	{
835	uint32_t fp;	1424	uint32_t fp;
836	WANT (4);	1425	WANT (4);
837	memcpy (&fp, dec->cur, 4);	1426	memcpy (&fp, dec->cur, 4);
838	dec->cur += 4;	1427	dec->cur += 4;
…		…
841	fp = ecb_bswap32 (fp);	1430	fp = ecb_bswap32 (fp);
842		1431
843	return newSVnv (ecb_binary32_to_float (fp));	1432	return newSVnv (ecb_binary32_to_float (fp));
844	}	1433	}
845		1434
846	case 27:	1435	case MISC_FLOAT64:
847	{	1436	{
848	uint64_t fp;	1437	uint64_t fp;
849	WANT (8);	1438	WANT (8);
850	memcpy (&fp, dec->cur, 8);	1439	memcpy (&fp, dec->cur, 8);
851	dec->cur += 8;	1440	dec->cur += 8;
…		…
854	fp = ecb_bswap64 (fp);	1443	fp = ecb_bswap64 (fp);
855		1444
856	return newSVnv (ecb_binary64_to_double (fp));	1445	return newSVnv (ecb_binary64_to_double (fp));
857	}	1446	}
858		1447
859	// 0..19 unassigned	1448	// 0..19 unassigned simple
860	// 24 reserved + unassigned (reserved values are not encodable)	1449	// 24 reserved + unassigned simple (reserved values are not encodable)
		1450	// 28-30 unassigned misc
		1451	// 31 break code
861	default:	1452	default:
862	ERR ("corrupted CBOR data (reserved/unassigned major 7 value)");	1453	ERR ("corrupted CBOR data (reserved/unassigned/unexpected major 7 value)");
863	}	1454	}
864		1455
865	break;	1456	break;
866	}	1457	}
867	#if 0
868	switch (*dec->cur)
869	{
870	//case '"': ++dec->cur; return decode_str (dec);
871	case '[': ++dec->cur; return decode_av (dec);
872	case '{': ++dec->cur; return decode_hv (dec);
873
874	case '-':
875	case '0': case '1': case '2': case '3': case '4':
876	case '5': case '6': case '7': case '8': case '9':
877	//TODO return decode_num (dec);
878
879	case 't':
880	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
881	{
882	dec->cur += 4;
883	#if CBOR_SLOW
884	cbor_true = get_bool ("CBOR::XS::true");
885	#endif
886	return newSVsv (cbor_true);
887	}
888	else
889	ERR ("'true' expected");
890
891	break;
892
893	case 'f':
894	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
895	{
896	dec->cur += 5;
897	#if CBOR_SLOW
898	cbor_false = get_bool ("CBOR::XS::false");
899	#endif
900	return newSVsv (cbor_false);
901	}
902	else
903	ERR ("'false' expected");
904
905	break;
906
907	case 'n':
908	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
909	{
910	dec->cur += 4;
911	return newSVsv (&PL_sv_undef);
912	}
913	else
914	ERR ("'null' expected");
915
916	break;
917
918	default:
919	ERR ("malformed CBOR string, neither array, object, number, string or atom");
920	break;
921	}
922	#endif
923		1458
924	fail:	1459	fail:
925	return &PL_sv_undef;	1460	return &PL_sv_undef;
926	}	1461	}
927		1462
928	static SV *	1463	static SV *
929	decode_cbor (SV *string, CBOR *cbor, char **offset_return)	1464	decode_cbor (SV *string, CBOR *cbor, char **offset_return)
930	{	1465	{
931	dec_t dec;	1466	dec_t dec = { 0 };
932	SV *sv;	1467	SV *sv;
		1468	STRLEN len;
		1469	char *data = SvPVbyte (string, len);
933		1470
934	/* work around bugs in 5.10 where manipulating magic values
935	* makes perl ignore the magic in subsequent accesses.
936	* also make a copy of non-PV values, to get them into a clean
937	* state (SvPV should do that, but it's buggy, see below).
938	*/
939	/*SvGETMAGIC (string);*/
940	if (SvMAGICAL (string) || !SvPOK (string))
941	string = sv_2mortal (newSVsv (string));
942
943	SvUPGRADE (string, SVt_PV);
944
945	/* work around a bug in perl 5.10, which causes SvCUR to fail an
946	* assertion with -DDEBUGGING, although SvCUR is documented to
947	* return the xpv_cur field which certainly exists after upgrading.
948	* according to nicholas clark, calling SvPOK fixes this.
949	* But it doesn't fix it, so try another workaround, call SvPV_nolen
950	* and hope for the best.
951	* Damnit, SvPV_nolen still trips over yet another assertion. This
952	* assertion business is seriously broken, try yet another workaround
953	* for the broken -DDEBUGGING.
954	*/
955	{
956	#ifdef DEBUGGING
957	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
958	#else
959	STRLEN offset = SvCUR (string);
960	#endif
961
962	if (offset > cbor->max_size && cbor->max_size)	1471	if (len > cbor->max_size && cbor->max_size)
963	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",	1472	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
964	(unsigned long)SvCUR (string), (unsigned long)cbor->max_size);	1473	(unsigned long)len, (unsigned long)cbor->max_size);
965	}
966
967	sv_utf8_downgrade (string, 0);
968		1474
969	dec.cbor = *cbor;	1475	dec.cbor = *cbor;
970	dec.cur = (U8 *)SvPVX (string);	1476	dec.cur = (U8 *)data;
971	dec.end = (U8 *)SvEND (string);	1477	dec.end = (U8 *)data + len;
972	dec.err = 0;
973	dec.depth = 0;
974
975	if (dec.cbor.cb_object || dec.cbor.cb_sk_object)
976	;//TODO dec.cbor.flags |= F_HOOK;
977		1478
978	sv = decode_sv (&dec);	1479	sv = decode_sv (&dec);
979		1480
980	if (offset_return)	1481	if (offset_return)
981	*offset_return = dec.cur;	1482	*offset_return = dec.cur;
…		…
984	if (dec.cur != dec.end && !dec.err)	1485	if (dec.cur != dec.end && !dec.err)
985	dec.err = "garbage after CBOR object";	1486	dec.err = "garbage after CBOR object";
986		1487
987	if (dec.err)	1488	if (dec.err)
988	{	1489	{
		1490	if (dec.shareable)
		1491	{
		1492	// need to break cyclic links, which would all be in shareable
		1493	int i;
		1494	SV **svp;
		1495
		1496	for (i = av_len (dec.shareable) + 1; i--; )
		1497	if ((svp = av_fetch (dec.shareable, i, 0)))
		1498	sv_setsv (*svp, &PL_sv_undef);
		1499	}
		1500
989	SvREFCNT_dec (sv);	1501	SvREFCNT_dec_NN (sv);
		1502
		1503	if (dec.err_sv)
		1504	sv_2mortal (dec.err_sv);
		1505
990	croak ("%s, at offset %d (octet 0x%02x)", dec.err, dec.cur - (U8 *)SvPVX (string), (int)(uint8_t)*dec.cur);	1506	croak ("%s, at offset %ld (octet 0x%02x)", dec.err, (long)(dec.cur - (U8 *)data), (int)(uint8_t)*dec.cur);
991	}	1507	}
992		1508
993	sv = sv_2mortal (sv);	1509	sv = sv_2mortal (sv);
994		1510
995	return sv;	1511	return sv;
996	}	1512	}
997		1513
		1514	/////////////////////////////////////////////////////////////////////////////
		1515	// incremental parser
		1516
		1517	#define INCR_DONE(cbor) (AvFILLp (cbor->incr_count) < 0)
		1518
		1519	// returns 0 for notyet, 1 for success or error
		1520	static int
		1521	incr_parse (CBOR *self, SV *cborstr)
		1522	{
		1523	STRLEN cur;
		1524	SvPV (cborstr, cur);
		1525
		1526	while (ecb_expect_true (self->incr_need <= cur))
		1527	{
		1528	// table of integer count bytes
		1529	static I8 incr_len[MINOR_MASK + 1] = {
		1530	0, 0, 0, 0, 0, 0, 0, 0,
		1531	0, 0, 0, 0, 0, 0, 0, 0,
		1532	0, 0, 0, 0, 0, 0, 0, 0,
		1533	1, 2, 4, 8,-1,-1,-1,-2
		1534	};
		1535
		1536	const U8 *p = SvPVX (cborstr) + self->incr_pos;
		1537	U8 m = *p & MINOR_MASK;
		1538	IV count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1539	I8 ilen = incr_len[m];
		1540
		1541	self->incr_need = self->incr_pos + 1;
		1542
		1543	if (ecb_expect_false (ilen < 0))
		1544	{
		1545	if (m != MINOR_INDEF)
		1546	return 1; // error
		1547
		1548	if (*p == (MAJOR_MISC | MINOR_INDEF))
		1549	{
		1550	if (count >= 0)
		1551	return 1; // error
		1552
		1553	count = 1;
		1554	}
		1555	else
		1556	{
		1557	av_push (self->incr_count, newSViv (-1)); //TODO: nest
		1558	count = -1;
		1559	}
		1560	}
		1561	else
		1562	{
		1563	self->incr_need += ilen;
		1564	if (ecb_expect_false (self->incr_need > cur))
		1565	return 0;
		1566
		1567	int major = *p >> MAJOR_SHIFT;
		1568
		1569	switch (major)
		1570	{
		1571	case MAJOR_TAG >> MAJOR_SHIFT:
		1572	++count; // tags merely prefix another value
		1573	break;
		1574
		1575	case MAJOR_BYTES >> MAJOR_SHIFT:
		1576	case MAJOR_TEXT >> MAJOR_SHIFT:
		1577	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1578	case MAJOR_MAP >> MAJOR_SHIFT:
		1579	{
		1580	UV len;
		1581
		1582	if (ecb_expect_false (ilen))
		1583	{
		1584	len = 0;
		1585
		1586	do {
		1587	len = (len << 8) | *++p;
		1588	} while (--ilen);
		1589	}
		1590	else
		1591	len = m;
		1592
		1593	switch (major)
		1594	{
		1595	case MAJOR_BYTES >> MAJOR_SHIFT:
		1596	case MAJOR_TEXT >> MAJOR_SHIFT:
		1597	self->incr_need += len;
		1598	if (ecb_expect_false (self->incr_need > cur))
		1599	return 0;
		1600
		1601	break;
		1602
		1603	case MAJOR_MAP >> MAJOR_SHIFT:
		1604	len <<= 1;
		1605	/* FALLTHROUGH */
		1606	case MAJOR_ARRAY >> MAJOR_SHIFT:
		1607	if (len)
		1608	{
		1609	av_push (self->incr_count, newSViv (len + 1)); //TODO: nest
		1610	count = len + 1;
		1611	}
		1612	break;
		1613	}
		1614	}
		1615	}
		1616	}
		1617
		1618	self->incr_pos = self->incr_need;
		1619
		1620	if (count > 0)
		1621	{
		1622	while (!--count)
		1623	{
		1624	if (!AvFILLp (self->incr_count))
		1625	return 1; // done
		1626
		1627	SvREFCNT_dec_NN (av_pop (self->incr_count));
		1628	count = SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]);
		1629	}
		1630
		1631	SvIVX (AvARRAY (self->incr_count)[AvFILLp (self->incr_count)]) = count;
		1632	}
		1633	}
		1634
		1635	return 0;
		1636	}
		1637
		1638
998	/////////////////////////////////////////////////////////////////////////////	1639	/////////////////////////////////////////////////////////////////////////////
999	// XS interface functions	1640	// XS interface functions
1000		1641
1001	MODULE = CBOR::XS PACKAGE = CBOR::XS	1642	MODULE = CBOR::XS PACKAGE = CBOR::XS
1002		1643
1003	BOOT:	1644	BOOT:
1004	{	1645	{
1005	cbor_stash = gv_stashpv ("CBOR::XS" , 1);	1646	cbor_stash = gv_stashpv ("CBOR::XS" , 1);
1006	cbor_boolean_stash = gv_stashpv ("CBOR::XS::Boolean", 1);	1647	cbor_tagged_stash = gv_stashpv ("CBOR::XS::Tagged" , 1);
1007		1648
1008	cbor_true = get_bool ("CBOR::XS::true");	1649	types_boolean_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1009	cbor_false = get_bool ("CBOR::XS::false");	1650	types_error_stash = gv_stashpv ("Types::Serialiser::Error" , 1);
		1651
		1652	types_true = get_bool ("Types::Serialiser::true" );
		1653	types_false = get_bool ("Types::Serialiser::false");
		1654	types_error = get_bool ("Types::Serialiser::error");
		1655
		1656	default_filter = newSVpv ("CBOR::XS::default_filter", 0);
		1657
		1658	sv_cbor = newSVpv ("CBOR", 0);
		1659	SvREADONLY_on (sv_cbor);
		1660
		1661	assert (("STRLEN must be an unsigned type", 0 <= (STRLEN)-1));
1010	}	1662	}
1011		1663
1012	PROTOTYPES: DISABLE	1664	PROTOTYPES: DISABLE
1013		1665
1014	void CLONE (...)	1666	void CLONE (...)
1015	CODE:	1667	CODE:
1016	cbor_stash = 0;	1668	cbor_stash = 0;
		1669	cbor_tagged_stash = 0;
		1670	types_error_stash = 0;
1017	cbor_boolean_stash = 0;	1671	types_boolean_stash = 0;
1018		1672
1019	void new (char *klass)	1673	void new (char *klass)
1020	PPCODE:	1674	PPCODE:
1021	{	1675	{
1022	SV *pv = NEWSV (0, sizeof (CBOR));	1676	SV *pv = NEWSV (0, sizeof (CBOR));
…		…
1028	)));	1682	)));
1029	}	1683	}
1030		1684
1031	void shrink (CBOR *self, int enable = 1)	1685	void shrink (CBOR *self, int enable = 1)
1032	ALIAS:	1686	ALIAS:
1033	shrink = F_SHRINK	1687	shrink = F_SHRINK
1034	allow_unknown = F_ALLOW_UNKNOWN	1688	allow_unknown = F_ALLOW_UNKNOWN
		1689	allow_sharing = F_ALLOW_SHARING
		1690	allow_cycles = F_ALLOW_CYCLES
		1691	allow_weak_cycles = F_ALLOW_WEAK_CYCLES
		1692	forbid_objects = F_FORBID_OBJECTS
		1693	pack_strings = F_PACK_STRINGS
		1694	text_keys = F_TEXT_KEYS
		1695	text_strings = F_TEXT_STRINGS
		1696	validate_utf8 = F_VALIDATE_UTF8
1035	PPCODE:	1697	PPCODE:
1036	{	1698	{
1037	if (enable)	1699	if (enable)
1038	self->flags |= ix;	1700	self->flags |= ix;
1039	else	1701	else
…		…
1042	XPUSHs (ST (0));	1704	XPUSHs (ST (0));
1043	}	1705	}
1044		1706
1045	void get_shrink (CBOR *self)	1707	void get_shrink (CBOR *self)
1046	ALIAS:	1708	ALIAS:
1047	get_shrink = F_SHRINK	1709	get_shrink = F_SHRINK
1048	get_allow_unknown = F_ALLOW_UNKNOWN	1710	get_allow_unknown = F_ALLOW_UNKNOWN
		1711	get_allow_sharing = F_ALLOW_SHARING
		1712	get_allow_cycles = F_ALLOW_CYCLES
		1713	get_allow_weak_cycles = F_ALLOW_WEAK_CYCLES
		1714	get_forbid_objects = F_FORBID_OBJECTS
		1715	get_pack_strings = F_PACK_STRINGS
		1716	get_text_keys = F_TEXT_KEYS
		1717	get_text_strings = F_TEXT_STRINGS
		1718	get_validate_utf8 = F_VALIDATE_UTF8
1049	PPCODE:	1719	PPCODE:
1050	XPUSHs (boolSV (self->flags & ix));	1720	XPUSHs (boolSV (self->flags & ix));
1051		1721
1052	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)	1722	void max_depth (CBOR *self, U32 max_depth = 0x80000000UL)
1053	PPCODE:	1723	PPCODE:
…		…
1069	CODE:	1739	CODE:
1070	RETVAL = self->max_size;	1740	RETVAL = self->max_size;
1071	OUTPUT:	1741	OUTPUT:
1072	RETVAL	1742	RETVAL
1073		1743
1074	#if 0 //TODO	1744	void filter (CBOR *self, SV *filter = 0)
1075
1076	void filter_cbor_object (CBOR *self, SV *cb = &PL_sv_undef)
1077	PPCODE:	1745	PPCODE:
1078	{
1079	SvREFCNT_dec (self->cb_object);	1746	SvREFCNT_dec (self->filter);
1080	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;	1747	self->filter = filter ? newSVsv (filter) : filter;
1081
1082	XPUSHs (ST (0));	1748	XPUSHs (ST (0));
1083	}
1084		1749
1085	void filter_cbor_single_key_object (CBOR *self, SV *key, SV *cb = &PL_sv_undef)	1750	SV *get_filter (CBOR *self)
1086	PPCODE:	1751	CODE:
1087	{	1752	RETVAL = self->filter ? self->filter : NEWSV (0, 0);
1088	if (!self->cb_sk_object)	1753	OUTPUT:
1089	self->cb_sk_object = newHV ();	1754	RETVAL
1090
1091	if (SvOK (cb))
1092	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1093	else
1094	{
1095	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
1096
1097	if (!HvKEYS (self->cb_sk_object))
1098	{
1099	SvREFCNT_dec (self->cb_sk_object);
1100	self->cb_sk_object = 0;
1101	}
1102	}
1103
1104	XPUSHs (ST (0));
1105	}
1106
1107	#endif
1108		1755
1109	void encode (CBOR *self, SV *scalar)	1756	void encode (CBOR *self, SV *scalar)
1110	PPCODE:	1757	PPCODE:
1111	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;	1758	PUTBACK; scalar = encode_cbor (scalar, self); SPAGAIN;
1112	XPUSHs (scalar);	1759	XPUSHs (scalar);
…		…
1125	EXTEND (SP, 2);	1772	EXTEND (SP, 2);
1126	PUSHs (sv);	1773	PUSHs (sv);
1127	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));	1774	PUSHs (sv_2mortal (newSVuv (offset - SvPVX (cborstr))));
1128	}	1775	}
1129		1776
		1777	void incr_parse (CBOR *self, SV *cborstr)
		1778	ALIAS:
		1779	incr_parse_multiple = 1
		1780	PPCODE:
		1781	{
		1782	if (SvUTF8 (cborstr))
		1783	sv_utf8_downgrade (cborstr, 0);
		1784
		1785	if (!self->incr_count)
		1786	{
		1787	self->incr_count = newAV ();
		1788	self->incr_pos = 0;
		1789	self->incr_need = 1;
		1790
		1791	av_push (self->incr_count, newSViv (1));
		1792	}
		1793
		1794	do
		1795	{
		1796	if (!incr_parse (self, cborstr))
		1797	{
		1798	if (self->incr_need > self->max_size && self->max_size)
		1799	croak ("attempted decode of CBOR text of %lu bytes size, but max_size is set to %lu",
		1800	(unsigned long)self->incr_need, (unsigned long)self->max_size);
		1801
		1802	break;
		1803	}
		1804
		1805	SV *sv;
		1806	char *offset;
		1807
		1808	PUTBACK; sv = decode_cbor (cborstr, self, &offset); SPAGAIN;
		1809	XPUSHs (sv);
		1810
		1811	sv_chop (cborstr, offset);
		1812
		1813	av_clear (self->incr_count);
		1814	av_push (self->incr_count, newSViv (1));
		1815
		1816	self->incr_pos = 0;
		1817	self->incr_need = self->incr_pos + 1;
		1818	}
		1819	while (ix);
		1820	}
		1821
		1822	void incr_reset (CBOR *self)
		1823	CODE:
		1824	{
		1825	SvREFCNT_dec (self->incr_count);
		1826	self->incr_count = 0;
		1827	}
		1828
1130	void DESTROY (CBOR *self)	1829	void DESTROY (CBOR *self)
1131	CODE:	1830	PPCODE:
1132	SvREFCNT_dec (self->cb_sk_object);	1831	cbor_free (self);
1133	SvREFCNT_dec (self->cb_object);
1134		1832
1135	PROTOTYPES: ENABLE	1833	PROTOTYPES: ENABLE
1136		1834
1137	void encode_cbor (SV *scalar)	1835	void encode_cbor (SV *scalar)
		1836	ALIAS:
		1837	encode_cbor = 0
		1838	encode_cbor_sharing = F_ALLOW_SHARING
1138	PPCODE:	1839	PPCODE:
1139	{	1840	{
1140	CBOR cbor;	1841	CBOR cbor;
1141	cbor_init (&cbor);	1842	cbor_init (&cbor);
		1843	cbor.flags |= ix;
1142	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;	1844	PUTBACK; scalar = encode_cbor (scalar, &cbor); SPAGAIN;
1143	XPUSHs (scalar);	1845	XPUSHs (scalar);
1144	}	1846	}
1145		1847
1146	void decode_cbor (SV *cborstr)	1848	void decode_cbor (SV *cborstr)
…		…
1150	cbor_init (&cbor);	1852	cbor_init (&cbor);
1151	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;	1853	PUTBACK; cborstr = decode_cbor (cborstr, &cbor, 0); SPAGAIN;
1152	XPUSHs (cborstr);	1854	XPUSHs (cborstr);
1153	}	1855	}
1154		1856
		1857	#ifdef __AFL_COMPILER
		1858
		1859	void
		1860	afl_init ()
		1861	CODE:
		1862	__AFL_INIT ();
		1863
		1864	int
		1865	afl_loop (unsigned int count = 10000)
		1866	CODE:
		1867	RETVAL = __AFL_LOOP (count);
		1868	OUTPUT:
		1869	RETVAL
		1870
		1871	#endif
		1872

Diff Legend

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