ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC vs.
Revision 1.135 by root, Thu Nov 15 20:49:12 2018 UTC

…		…
4		4
5	#include <assert.h>	5	#include <assert.h>
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 <float.h>	10	#include <float.h>
		11	#include <inttypes.h>
10		12
11	#if defined(__BORLANDC__) || defined(_MSC_VER)	13	#if defined(__BORLANDC__) || defined(_MSC_VER)
12	# define snprintf _snprintf // C compilers have this in stdio.h	14	# define snprintf _snprintf // C compilers have this in stdio.h
13	#endif	15	#endif
14		16
15	// some old perls do not have this, try to make it work, no	17	// some old perls do not have this, try to make it work, no
16	// guarentees, though. if it breaks, you get to keep the pieces.	18	// guarantees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	19	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	20	# define UTF8_MAXBYTES 13
19	#endif	21	#endif
		22
		23	// compatibility with perl <5.18
		24	#ifndef HvNAMELEN_get
		25	# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
		26	#endif
		27	#ifndef HvNAMELEN
		28	# define HvNAMELEN(hv) HvNAMELEN_get (hv)
		29	#endif
		30	#ifndef HvNAMEUTF8
		31	# define HvNAMEUTF8(hv) 0
		32	#endif
		33
		34	// three extra for rounding, sign, and end of string
		35	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20		36
21	#define F_ASCII 0x00000001UL	37	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	38	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	39	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	40	#define F_INDENT 0x00000008UL
…		…
28	#define F_ALLOW_NONREF 0x00000100UL	44	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	45	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	46	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	47	#define F_CONV_BLESSED 0x00000800UL
32	#define F_RELAXED 0x00001000UL	48	#define F_RELAXED 0x00001000UL
33		49	#define F_ALLOW_UNKNOWN 0x00002000UL
34	#define F_MAXDEPTH 0xf8000000UL	50	#define F_ALLOW_TAGS 0x00004000UL
35	#define S_MAXDEPTH 27
36	#define F_MAXSIZE 0x01f00000UL
37	#define S_MAXSIZE 20
38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	51	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39		52
40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	53	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
44	#define F_DEFAULT (9UL << S_MAXDEPTH)
45		54
46	#define INIT_SIZE 32 // initial scalar size to be allocated	55	#define INIT_SIZE 64 // initial scalar size to be allocated
47	#define INDENT_STEP 3 // spaces per indentation level	56	#define INDENT_STEP 3 // spaces per indentation level
48		57
49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	58	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
		59
		60	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
50		61
51	#define SB do {	62	#define SB do {
52	#define SE } while (0)	63	#define SE } while (0)
53		64
54	#if __GNUC__ >= 3	65	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	66	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	67	# define INLINE static inline
57	#else	68	#else
58	# define expect(expr,value) (expr)	69	# define expect(expr,value) (expr)
59	# define inline static	70	# define INLINE static
60	#endif	71	#endif
61		72
62	#define expect_false(expr) expect ((expr) != 0, 0)	73	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	74	#define expect_true(expr) expect ((expr) != 0, 1)
64		75
		76	#define IN_RANGE_INC(type,val,beg,end) \
		77	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		78	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		79
		80	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		81
65	#ifdef USE_ITHREADS	82	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	83	# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
		84	# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
		85	# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
68	#else	86	#else
69	# define JSON_SLOW 0
70	# define JSON_STASH json_stash	87	# define JSON_STASH json_stash
		88	# define BOOL_STASH bool_stash
		89	# define GET_BOOL(value) bool_ ## value
71	#endif	90	#endif
72		91
73	static HV *json_stash, *json_boolean_stash; // JSON::XS::	92	// the amount of HEs to allocate on the stack, when sorting keys
74	static SV *json_true, *json_false;	93	#define STACK_HES 64
		94
		95	static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
		96	static SV *bool_false, *bool_true;
		97	static SV *sv_json;
		98
		99	enum {
		100	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		101	INCR_M_STR, // inside string
		102	INCR_M_BS, // inside backslash
		103	INCR_M_C0, // inside comment in initial whitespace sequence
		104	INCR_M_C1, // inside comment in other places
		105	INCR_M_JSON // outside anything, count nesting
		106	};
		107
		108	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
75		109
76	typedef struct {	110	typedef struct {
77	U32 flags;	111	U32 flags;
		112	U32 max_depth;
		113	STRLEN max_size;
		114
78	SV *cb_object;	115	SV *cb_object;
79	HV *cb_sk_object;	116	HV *cb_sk_object;
		117
		118	// for the incremental parser
		119	SV *incr_text; // the source text so far
		120	STRLEN incr_pos; // the current offset into the text
		121	int incr_nest; // {[]}-nesting level
		122	unsigned char incr_mode;
		123
		124	SV *v_false, *v_true;
80	} JSON;	125	} JSON;
		126
		127	INLINE void
		128	json_init (JSON *json)
		129	{
		130	static const JSON init = { F_ALLOW_NONREF, 512 };
		131
		132	*json = init;
		133	}
81		134
82	/////////////////////////////////////////////////////////////////////////////	135	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	136	// utility functions
84		137
85	inline void	138	INLINE SV *
		139	get_bool (const char *name)
		140	{
		141	SV *sv = get_sv (name, 1);
		142
		143	SvREADONLY_on (sv);
		144	SvREADONLY_on (SvRV (sv));
		145
		146	return sv;
		147	}
		148
		149	INLINE void
86	shrink (SV *sv)	150	shrink (SV *sv)
87	{	151	{
88	sv_utf8_downgrade (sv, 1);	152	sv_utf8_downgrade (sv, 1);
		153
89	if (SvLEN (sv) > SvCUR (sv) + 1)	154	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	155	{
91	#ifdef SvPV_shrink_to_cur	156	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	157	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	158	#elif defined (SvPV_renew)
94	SvPV_renew (sv, SvCUR (sv) + 1);	159	SvPV_renew (sv, SvCUR (sv) + 1);
95	#endif	160	#endif
96	}	161	}
97	}	162	}
98		163
		164	/* adds two STRLENs together, slow, and with paranoia */
		165	STRLEN
		166	strlen_sum (STRLEN l1, STRLEN l2)
		167	{
		168	size_t sum = l1 + l2;
		169
		170	if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
		171	croak ("JSON::XS: string size overflow");
		172
		173	return sum;
		174	}
		175
		176	/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
		177	static char *
		178	json_sv_grow (SV *sv, size_t len1, size_t len2)
		179	{
		180	len1 = strlen_sum (len1, len2);
		181	len1 = strlen_sum (len1, len1 >> 1);
		182
		183	if (len1 > 4096 - 24)
		184	len1 = (len1 | 4095) - 24;
		185
		186	return SvGROW (sv, len1);
		187	}
		188
99	// decode an utf-8 character and return it, or (UV)-1 in	189	// decode a utf-8 character and return it, or (UV)-1 in
100	// case of an error.	190	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	191	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	192	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	193	// note that we never call this function for a ascii codepoints
104	inline UV	194	INLINE UV
105	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	195	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
106	{	196	{
107	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))	197	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	198	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	199	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	200	{
111	*clen = 2;	201	*clen = 2;
112	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	202	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
113	}	203	}
114	else	204	else
		205	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		206	}
		207
		208	// likewise for encoding, also never called for ascii codepoints
		209	// this function takes advantage of this fact, although current gccs
		210	// seem to optimise the check for >= 0x80 away anyways
		211	INLINE unsigned char *
		212	encode_utf8 (unsigned char *s, UV ch)
		213	{
		214	if (expect_false (ch < 0x000080))
		215	*s++ = ch;
		216	else if (expect_true (ch < 0x000800))
		217	*s++ = 0xc0 | ( ch >> 6),
		218	*s++ = 0x80 | ( ch & 0x3f);
		219	else if ( ch < 0x010000)
		220	*s++ = 0xe0 | ( ch >> 12),
		221	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		222	*s++ = 0x80 | ( ch & 0x3f);
		223	else if ( ch < 0x110000)
		224	*s++ = 0xf0 | ( ch >> 18),
		225	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		226	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		227	*s++ = 0x80 | ( ch & 0x3f);
		228
		229	return s;
		230	}
		231
		232	// convert offset pointer to character index, sv must be string
		233	static STRLEN
		234	ptr_to_index (SV *sv, char *offset)
		235	{
		236	return SvUTF8 (sv)
		237	? utf8_distance (offset, SvPVX (sv))
		238	: offset - SvPVX (sv);
		239	}
		240
		241	/////////////////////////////////////////////////////////////////////////////
		242	// fp hell
		243
		244	// scan a group of digits, and a trailing exponent
		245	static void
		246	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		247	{
		248	UV uaccum = 0;
		249	int eaccum = 0;
		250
		251	// if we recurse too deep, skip all remaining digits
		252	// to avoid a stack overflow attack
		253	if (expect_false (--maxdepth <= 0))
		254	while (((U8)*s - '0') < 10)
		255	++s;
		256
		257	for (;;)
		258	{
		259	U8 dig = (U8)*s - '0';
		260
		261	if (expect_false (dig >= 10))
		262	{
		263	if (dig == (U8)((U8)'.' - (U8)'0'))
		264	{
		265	++s;
		266	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		267	}
		268	else if ((dig | ' ') == 'e' - '0')
		269	{
		270	int exp2 = 0;
		271	int neg = 0;
		272
		273	++s;
		274
		275	if (*s == '-')
		276	{
		277	++s;
		278	neg = 1;
		279	}
		280	else if (*s == '+')
		281	++s;
		282
		283	while ((dig = (U8)*s - '0') < 10)
		284	exp2 = exp2 * 10 + *s++ - '0';
		285
		286	*expo += neg ? -exp2 : exp2;
		287	}
		288
		289	break;
		290	}
		291
		292	++s;
		293
		294	uaccum = uaccum * 10 + dig;
		295	++eaccum;
		296
		297	// if we have too many digits, then recurse for more
		298	// we actually do this for rather few digits
		299	if (uaccum >= (UV_MAX - 9) / 10)
		300	{
		301	if (postdp) *expo -= eaccum;
		302	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		303	if (postdp) *expo += eaccum;
		304
		305	break;
		306	}
115	{	307	}
116	*clen = (STRLEN)-1;	308
		309	// this relies greatly on the quality of the pow ()
		310	// implementation of the platform, but a good
		311	// implementation is hard to beat.
		312	// (IEEE 754 conformant ones are required to be exact)
		313	if (postdp) *expo -= eaccum;
		314	*accum += uaccum * Perl_pow (10., *expo);
		315	*expo += eaccum;
		316	}
		317
		318	static NV
		319	json_atof (const char *s)
		320	{
		321	NV accum = 0.;
		322	int expo = 0;
		323	int neg = 0;
		324
		325	if (*s == '-')
		326	{
		327	++s;
		328	neg = 1;
		329	}
		330
		331	// a recursion depth of ten gives us >>500 bits
		332	json_atof_scan1 (s, &accum, &expo, 0, 10);
		333
		334	return neg ? -accum : accum;
		335	}
		336
		337	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		338	static int
		339	ref_bool_type (SV *sv)
		340	{
		341	svtype svt = SvTYPE (sv);
		342
		343	if (svt < SVt_PVAV)
		344	{
		345	STRLEN len = 0;
		346	char *pv = svt ? SvPV (sv, len) : 0;
		347
		348	if (len == 1)
		349	if (*pv == '1')
		350	return 1;
		351	else if (*pv == '0')
		352	return 0;
		353	}
		354
		355	return -1;
		356	}
		357
		358	// returns whether scalar is not a reference in the sense of allow_nonref
		359	static int
		360	json_nonref (SV *scalar)
		361	{
		362	if (!SvROK (scalar))
		363	return 1;
		364
		365	scalar = SvRV (scalar);
		366
		367	if (SvTYPE (scalar) >= SVt_PVMG)
		368	{
		369	if (SvSTASH (scalar) == bool_stash)
117	return (UV)-1;	370	return 1;
		371
		372	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		373	return 1;
118	}	374	}
		375
		376	return 0;
119	}	377	}
120		378
121	/////////////////////////////////////////////////////////////////////////////	379	/////////////////////////////////////////////////////////////////////////////
122	// encoder	380	// encoder
123		381
…		…
127	char *cur; // SvPVX (sv) + current output position	385	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	386	char *end; // SvEND (sv)
129	SV *sv; // result scalar	387	SV *sv; // result scalar
130	JSON json;	388	JSON json;
131	U32 indent; // indentation level	389	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	390	UV limit; // escape character values >= this value when encoding
133	} enc_t;	391	} enc_t;
134		392
135	inline void	393	INLINE void
136	need (enc_t *enc, STRLEN len)	394	need (enc_t *enc, STRLEN len)
137	{	395	{
138	if (expect_false (enc->cur + len >= enc->end))	396	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
139	{	397	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	398	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	399	char *buf = json_sv_grow (enc->sv, cur, len);
142	enc->cur = SvPVX (enc->sv) + cur;	400	enc->cur = buf + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	401	enc->end = buf + SvLEN (enc->sv) - 1;
144	}	402	}
145	}	403	}
146		404
147	inline void	405	INLINE void
148	encode_ch (enc_t *enc, char ch)	406	encode_ch (enc_t *enc, char ch)
149	{	407	{
150	need (enc, 1);	408	need (enc, 1);
151	*enc->cur++ = ch;	409	*enc->cur++ = ch;
152	}	410	}
…		…
164		422
165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case	423	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
166	{	424	{
167	if (expect_false (ch == '"')) // but with slow exceptions	425	if (expect_false (ch == '"')) // but with slow exceptions
168	{	426	{
169	need (enc, len += 1);	427	need (enc, len + 1);
170	*enc->cur++ = '\\';	428	*enc->cur++ = '\\';
171	*enc->cur++ = '"';	429	*enc->cur++ = '"';
172	}	430	}
173	else if (expect_false (ch == '\\'))	431	else if (expect_false (ch == '\\'))
174	{	432	{
175	need (enc, len += 1);	433	need (enc, len + 1);
176	*enc->cur++ = '\\';	434	*enc->cur++ = '\\';
177	*enc->cur++ = '\\';	435	*enc->cur++ = '\\';
178	}	436	}
179	else	437	else
180	*enc->cur++ = ch;	438	*enc->cur++ = ch;
…		…
183	}	441	}
184	else	442	else
185	{	443	{
186	switch (ch)	444	switch (ch)
187	{	445	{
188	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;	446	case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
189	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;	447	case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
190	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;	448	case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
191	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;	449	case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
192	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;	450	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
193		451
194	default:	452	default:
195	{	453	{
196	STRLEN clen;	454	STRLEN clen;
197	UV uch;	455	UV uch;
…		…
206	{	464	{
207	uch = ch;	465	uch = ch;
208	clen = 1;	466	clen = 1;
209	}	467	}
210		468
211	if (uch > 0x10FFFFUL)	469	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
213
214	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
215	{	470	{
216	if (uch > 0xFFFFUL)	471	if (uch >= 0x10000UL)
217	{	472	{
		473	if (uch >= 0x110000UL)
		474	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		475
218	need (enc, len += 11);	476	need (enc, len + 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	477	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	478	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	479	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	480	enc->cur += 12;
223	}	481	}
224	else	482	else
225	{	483	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	484	need (enc, len + 5);
228	*enc->cur++ = '\\';	485	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	486	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	487	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	488	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	489	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	490	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	491	}
235		492
236	str += clen;	493	str += clen;
237	}	494	}
238	else if (enc->json.flags & F_LATIN1)	495	else if (enc->json.flags & F_LATIN1)
…		…
240	*enc->cur++ = uch;	497	*enc->cur++ = uch;
241	str += clen;	498	str += clen;
242	}	499	}
243	else if (is_utf8)	500	else if (is_utf8)
244	{	501	{
245	need (enc, len += clen);	502	need (enc, len + clen);
246	do	503	do
247	{	504	{
248	*enc->cur++ = *str++;	505	*enc->cur++ = *str++;
249	}	506	}
250	while (--clen);	507	while (--clen);
251	}	508	}
252	else	509	else
253	{	510	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	511	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	512	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	513	++str;
257	}	514	}
258	}	515	}
259	}	516	}
260	}	517	}
261		518
262	--len;	519	--len;
263	}	520	}
264	}	521	}
265		522
266	inline void	523	INLINE void
267	encode_indent (enc_t *enc)	524	encode_indent (enc_t *enc)
268	{	525	{
269	if (enc->json.flags & F_INDENT)	526	if (enc->json.flags & F_INDENT)
270	{	527	{
271	int spaces = enc->indent * INDENT_STEP;	528	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	531	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	532	enc->cur += spaces;
276	}	533	}
277	}	534	}
278		535
279	inline void	536	INLINE void
280	encode_space (enc_t *enc)	537	encode_space (enc_t *enc)
281	{	538	{
282	need (enc, 1);	539	need (enc, 1);
283	encode_ch (enc, ' ');	540	encode_ch (enc, ' ');
284	}	541	}
285		542
286	inline void	543	INLINE void
287	encode_nl (enc_t *enc)	544	encode_nl (enc_t *enc)
288	{	545	{
289	if (enc->json.flags & F_INDENT)	546	if (enc->json.flags & F_INDENT)
290	{	547	{
291	need (enc, 1);	548	need (enc, 1);
292	encode_ch (enc, '\n');	549	encode_ch (enc, '\n');
293	}	550	}
294	}	551	}
295		552
296	inline void	553	INLINE void
297	encode_comma (enc_t *enc)	554	encode_comma (enc_t *enc)
298	{	555	{
299	encode_ch (enc, ',');	556	encode_ch (enc, ',');
300		557
301	if (enc->json.flags & F_INDENT)	558	if (enc->json.flags & F_INDENT)
…		…
309	static void	566	static void
310	encode_av (enc_t *enc, AV *av)	567	encode_av (enc_t *enc, AV *av)
311	{	568	{
312	int i, len = av_len (av);	569	int i, len = av_len (av);
313		570
314	if (enc->indent >= enc->maxdepth)	571	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	572	croak (ERR_NESTING_EXCEEDED);
316		573
317	encode_ch (enc, '['); encode_nl (enc);	574	encode_ch (enc, '[');
318	++enc->indent;
319		575
		576	if (len >= 0)
		577	{
		578	encode_nl (enc); ++enc->indent;
		579
320	for (i = 0; i <= len; ++i)	580	for (i = 0; i <= len; ++i)
321	{	581	{
322	SV **svp = av_fetch (av, i, 0);	582	SV **svp = av_fetch (av, i, 0);
323		583
324	encode_indent (enc);	584	encode_indent (enc);
325		585
326	if (svp)	586	if (svp)
327	encode_sv (enc, *svp);	587	encode_sv (enc, *svp);
328	else	588	else
329	encode_str (enc, "null", 4, 0);	589	encode_str (enc, "null", 4, 0);
330		590
331	if (i < len)	591	if (i < len)
332	encode_comma (enc);	592	encode_comma (enc);
333	}	593	}
334		594
		595	encode_nl (enc); --enc->indent; encode_indent (enc);
		596	}
		597
335	encode_nl (enc);	598	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	599	}
340		600
341	static void	601	static void
342	encode_hk (enc_t *enc, HE *he)	602	encode_hk (enc_t *enc, HE *he)
343	{	603	{
…		…
346	if (HeKLEN (he) == HEf_SVKEY)	606	if (HeKLEN (he) == HEf_SVKEY)
347	{	607	{
348	SV *sv = HeSVKEY (he);	608	SV *sv = HeSVKEY (he);
349	STRLEN len;	609	STRLEN len;
350	char *str;	610	char *str;
351		611
352	SvGETMAGIC (sv);	612	SvGETMAGIC (sv);
353	str = SvPV (sv, len);	613	str = SvPV (sv, len);
354		614
355	encode_str (enc, str, len, SvUTF8 (sv));	615	encode_str (enc, str, len, SvUTF8 (sv));
356	}	616	}
…		…
391		651
392	static void	652	static void
393	encode_hv (enc_t *enc, HV *hv)	653	encode_hv (enc_t *enc, HV *hv)
394	{	654	{
395	HE *he;	655	HE *he;
396	int count;
397		656
398	if (enc->indent >= enc->maxdepth)	657	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	658	croak (ERR_NESTING_EXCEEDED);
400		659
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	660	encode_ch (enc, '{');
402		661
403	// for canonical output we have to sort by keys first	662	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	663	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	664	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	665	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	666	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	667	{
409	int count = hv_iterinit (hv);	668	int count = hv_iterinit (hv);
410		669
411	if (SvMAGICAL (hv))	670	if (SvMAGICAL (hv))
412	{	671	{
…		…
423	}	682	}
424		683
425	if (count)	684	if (count)
426	{	685	{
427	int i, fast = 1;	686	int i, fast = 1;
428	#if defined(__BORLANDC__) || defined(_MSC_VER)	687	HE *hes_stack [STACK_HES];
429	HE **hes = _alloca (count * sizeof (HE));	688	HE **hes = hes_stack;
430	#else	689
431	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	690	// allocate larger arrays on the heap
432	#endif	691	if (count > STACK_HES)
		692	{
		693	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		694	hes = (HE **)SvPVX (sv);
		695	}
433		696
434	i = 0;	697	i = 0;
435	while ((he = hv_iternext (hv)))	698	while ((he = hv_iternext (hv)))
436	{	699	{
437	hes [i++] = he;	700	hes [i++] = he;
…		…
459		722
460	FREETMPS;	723	FREETMPS;
461	LEAVE;	724	LEAVE;
462	}	725	}
463		726
		727	encode_nl (enc); ++enc->indent;
		728
464	while (count--)	729	while (count--)
465	{	730	{
466	encode_indent (enc);	731	encode_indent (enc);
467	he = hes [count];	732	he = hes [count];
468	encode_hk (enc, he);	733	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	734	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		735
471	if (count)	736	if (count)
472	encode_comma (enc);	737	encode_comma (enc);
473	}	738	}
		739
		740	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	741	}
475	}	742	}
476	else	743	else
477	{	744	{
478	if (hv_iterinit (hv) || SvMAGICAL (hv))	745	if (hv_iterinit (hv) || SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	746	if ((he = hv_iternext (hv)))
		747	{
		748	encode_nl (enc); ++enc->indent;
		749
480	for (;;)	750	for (;;)
481	{	751	{
482	encode_indent (enc);	752	encode_indent (enc);
483	encode_hk (enc, he);	753	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	754	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		755
486	if (!(he = hv_iternext (hv)))	756	if (!(he = hv_iternext (hv)))
487	break;	757	break;
488		758
489	encode_comma (enc);	759	encode_comma (enc);
490	}	760	}
491	}
492		761
		762	encode_nl (enc); --enc->indent; encode_indent (enc);
		763	}
		764	}
		765
493	encode_nl (enc);	766	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	767	}
497		768
498	// encode objects, arrays and special \0=false and \1=true values.	769	// encode objects, arrays and special \0=false and \1=true values.
499	static void	770	static void
500	encode_rv (enc_t *enc, SV *sv)	771	encode_rv (enc_t *enc, SV *sv)
501	{	772	{
502	svtype svt;	773	svtype svt;
		774	GV *method;
503		775
504	SvGETMAGIC (sv);	776	SvGETMAGIC (sv);
505	svt = SvTYPE (sv);	777	svt = SvTYPE (sv);
506		778
507	if (expect_false (SvOBJECT (sv)))	779	if (expect_false (SvOBJECT (sv)))
508	{	780	{
509	HV *stash = !JSON_SLOW || json_boolean_stash	781	HV *stash = SvSTASH (sv);
510	? json_boolean_stash
511	: gv_stashpv ("JSON::XS::Boolean", 1);
512		782
513	if (SvSTASH (sv) == stash)	783	if (stash == bool_stash)
514	{	784	{
515	if (SvIV (sv))
516	encode_str (enc, "true", 4, 0);	785	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
517	else
518	encode_str (enc, "false", 5, 0);	786	else encode_str (enc, "false", 5, 0);
519	}	787	}
		788	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		789	{
		790	int count;
		791	dSP;
		792
		793	ENTER; SAVETMPS;
		794	PUSHMARK (SP);
		795	EXTEND (SP, 2);
		796	// we re-bless the reference to get overload and other niceties right
		797	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		798	PUSHs (sv_json);
		799
		800	PUTBACK;
		801	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		802	SPAGAIN;
		803
		804	// catch this surprisingly common error
		805	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		806	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		807
		808	encode_ch (enc, '(');
		809	encode_ch (enc, '"');
		810	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		811	encode_ch (enc, '"');
		812	encode_ch (enc, ')');
		813	encode_ch (enc, '[');
		814
		815	if (count)
		816	{
		817	int i;
		818
		819	for (i = 0; i < count - 1; ++i)
		820	{
		821	encode_sv (enc, SP[i + 1 - count]);
		822	encode_ch (enc, ',');
		823	}
		824
		825	encode_sv (enc, TOPs);
		826	SP -= count;
		827	}
		828
		829	encode_ch (enc, ']');
		830
		831	FREETMPS; LEAVE;
		832	}
		833	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		834	{
		835	dSP;
		836
		837	ENTER; SAVETMPS;
		838	PUSHMARK (SP);
		839	// we re-bless the reference to get overload and other niceties right
		840	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		841
		842	// calling with G_SCALAR ensures that we always get a 1 return value
		843	PUTBACK;
		844	call_sv ((SV *)GvCV (method), G_SCALAR);
		845	SPAGAIN;
		846
		847	// catch this surprisingly common error
		848	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		849	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		850
		851	sv = POPs;
		852	PUTBACK;
		853
		854	encode_sv (enc, sv);
		855
		856	FREETMPS; LEAVE;
		857	}
		858	else if (enc->json.flags & F_ALLOW_BLESSED)
		859	encode_str (enc, "null", 4, 0);
520	else	860	else
521	{
522	#if 0
523	if (0 && sv_derived_from (rv, "JSON::Literal"))
524	{
525	// not yet
526	}
527	#endif
528	if (enc->json.flags & F_CONV_BLESSED)
529	{
530	// we re-bless the reference to get overload and other niceties right
531	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
532
533	if (to_json)
534	{
535	dSP;
536
537	ENTER; SAVETMPS; PUSHMARK (SP);
538	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
539
540	// calling with G_SCALAR ensures that we always get a 1 return value
541	PUTBACK;
542	call_sv ((SV *)GvCV (to_json), G_SCALAR);
543	SPAGAIN;
544
545	// catch this surprisingly common error
546	if (SvROK (TOPs) && SvRV (TOPs) == sv)
547	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
548
549	sv = POPs;
550	PUTBACK;
551
552	encode_sv (enc, sv);
553
554	FREETMPS; LEAVE;
555	}
556	else if (enc->json.flags & F_ALLOW_BLESSED)
557	encode_str (enc, "null", 4, 0);
558	else
559	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
560	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
561	}
562	else if (enc->json.flags & F_ALLOW_BLESSED)
563	encode_str (enc, "null", 4, 0);
564	else
565	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",	861	croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	862	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
567	}
568	}	863	}
569	else if (svt == SVt_PVHV)	864	else if (svt == SVt_PVHV)
570	encode_hv (enc, (HV *)sv);	865	encode_hv (enc, (HV *)sv);
571	else if (svt == SVt_PVAV)	866	else if (svt == SVt_PVAV)
572	encode_av (enc, (AV *)sv);	867	encode_av (enc, (AV *)sv);
573	else if (svt < SVt_PVAV)	868	else if (svt < SVt_PVAV)
574	{	869	{
575	STRLEN len = 0;	870	int bool_type = ref_bool_type (sv);
576	char *pv = svt ? SvPV (sv, len) : 0;
577		871
578	if (len == 1 && *pv == '1')	872	if (bool_type == 1)
579	encode_str (enc, "true", 4, 0);	873	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	874	else if (bool_type == 0)
581	encode_str (enc, "false", 5, 0);	875	encode_str (enc, "false", 5, 0);
		876	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		877	encode_str (enc, "null", 4, 0);
582	else	878	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	879	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	880	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	881	}
		882	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		883	encode_str (enc, "null", 4, 0);
586	else	884	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	885	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	886	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	887	}
590		888
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	906	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	907	enc->cur += strlen (enc->cur);
610	}	908	}
611	else if (SvIOKp (sv))	909	else if (SvIOKp (sv))
612	{	910	{
613	// we assume we can always read an IV as a UV	911	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	912	// we assume two's complement
615	{	913	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	914	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	915	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
618	enc->cur +=
619	SvIsUV(sv)
620	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
621	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
622	}
623	else
624	{	916	{
625	// optimise the "small number case"	917	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	918	// code will likely be branchless and use only a single multiplication
		919	// works for numbers up to 59074
627	I32 i = SvIV (sv);	920	I32 i = SvIVX (sv);
628	U32 u;	921	U32 u;
629	char digit, nz = 0;	922	char digit, nz = 0;
630		923
631	need (enc, 6);	924	need (enc, 6);
632		925
…		…
638		931
639	// now output digit by digit, each time masking out the integer part	932	// now output digit by digit, each time masking out the integer part
640	// and multiplying by 5 while moving the decimal point one to the right,	933	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	934	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	935	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	936	// the pointer, to enable the use of conditional move instructions.
644	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;	937	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
645	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;	938	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
646	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;	939	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
647	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;	940	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
648	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	941	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	942	}
		943	else
		944	{
		945	// large integer, use the (rather slow) snprintf way.
		946	need (enc, IVUV_MAXCHARS);
		947	enc->cur +=
		948	SvIsUV(sv)
		949	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		950	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		951	}
650	}	952	}
651	else if (SvROK (sv))	953	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	954	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	955	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	956	encode_str (enc, "null", 4, 0);
655	else	957	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	958	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
657	SvPV_nolen (sv), SvFLAGS (sv));	959	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
658	}	960	}
659		961
660	static SV *	962	static SV *
661	encode_json (SV *scalar, JSON *json)	963	encode_json (SV *scalar, JSON *json)
662	{	964	{
663	enc_t enc;	965	enc_t enc;
664		966
665	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))	967	if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
666	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	968	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
667		969
668	enc.json = *json;	970	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	971	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	972	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	973	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	974	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	975	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		976	: enc.json.flags & F_LATIN1 ? 0x000100UL
		977	: 0x110000UL;
674		978
675	SvPOK_only (enc.sv);	979	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	980	encode_sv (&enc, scalar);
		981	encode_nl (&enc);
677		982
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	983	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	984	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		985
681	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))	986	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
…		…
699	JSON json;	1004	JSON json;
700	U32 depth; // recursion depth	1005	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	1006	U32 maxdepth; // recursion depth limit
702	} dec_t;	1007	} dec_t;
703		1008
704	inline void	1009	INLINE void
705	decode_comment (dec_t *dec)	1010	decode_comment (dec_t *dec)
706	{	1011	{
707	// only '#'-style comments allowed a.t.m.	1012	// only '#'-style comments allowed a.t.m.
708		1013
709	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)	1014	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
710	++dec->cur;	1015	++dec->cur;
711	}	1016	}
712		1017
713	inline void	1018	INLINE void
714	decode_ws (dec_t *dec)	1019	decode_ws (dec_t *dec)
715	{	1020	{
716	for (;;)	1021	for (;;)
717	{	1022	{
718	char ch = *dec->cur;	1023	char ch = *dec->cur;
…		…
742	if (*dec->cur != ch) \	1047	if (*dec->cur != ch) \
743	ERR (# ch " expected"); \	1048	ERR (# ch " expected"); \
744	++dec->cur; \	1049	++dec->cur; \
745	SE	1050	SE
746		1051
747	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	1052	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
748	#define DEC_DEC_DEPTH --dec->depth	1053	#define DEC_DEC_DEPTH --dec->depth
749		1054
750	static SV *decode_sv (dec_t *dec);	1055	static SV *decode_sv (dec_t *dec);
751		1056
752	static signed char decode_hexdigit[256];	1057	static signed char decode_hexdigit[256];
…		…
844		1149
845	if (hi >= 0x80)	1150	if (hi >= 0x80)
846	{	1151	{
847	utf8 = 1;	1152	utf8 = 1;
848		1153
849	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1154	cur = encode_utf8 (cur, hi);
850	}	1155	}
851	else	1156	else
852	*cur++ = hi;	1157	*cur++ = hi;
853	}	1158	}
854	break;	1159	break;
…		…
856	default:	1161	default:
857	--dec_cur;	1162	--dec_cur;
858	ERR ("illegal backslash escape sequence in string");	1163	ERR ("illegal backslash escape sequence in string");
859	}	1164	}
860	}	1165	}
861	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1166	else if (expect_true (ch >= 0x20 && ch < 0x80))
862	*cur++ = ch;	1167	*cur++ = ch;
863	else if (ch >= 0x80)	1168	else if (ch >= 0x80)
864	{	1169	{
865	STRLEN clen;	1170	STRLEN clen;
866	UV uch;
867		1171
868	--dec_cur;	1172	--dec_cur;
869		1173
870	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1174	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
871	if (clen == (STRLEN)-1)	1175	if (clen == (STRLEN)-1)
872	ERR ("malformed UTF-8 character in JSON string");	1176	ERR ("malformed UTF-8 character in JSON string");
873		1177
874	do	1178	do
875	*cur++ = *dec_cur++;	1179	*cur++ = *dec_cur++;
876	while (--clen);	1180	while (--clen);
877		1181
878	utf8 = 1;	1182	utf8 = 1;
879	}	1183	}
		1184	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1185	*cur++ = ch;
880	else	1186	else
881	{	1187	{
882	--dec_cur;	1188	--dec_cur;
883		1189
884	if (!ch)	1190	if (!ch)
…		…
892	{	1198	{
893	STRLEN len = cur - buf;	1199	STRLEN len = cur - buf;
894		1200
895	if (sv)	1201	if (sv)
896	{	1202	{
897	SvGROW (sv, SvCUR (sv) + len + 1);	1203	STRLEN cur = SvCUR (sv);
		1204
		1205	if (SvLEN (sv) - cur <= len)
		1206	json_sv_grow (sv, cur, len);
		1207
898	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1208	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
899	SvCUR_set (sv, SvCUR (sv) + len);	1209	SvCUR_set (sv, SvCUR (sv) + len);
900	}	1210	}
901	else	1211	else
902	sv = newSVpvn (buf, len);	1212	sv = newSVpvn (buf, len);
…		…
989		1299
990	if (!is_nv)	1300	if (!is_nv)
991	{	1301	{
992	int len = dec->cur - start;	1302	int len = dec->cur - start;
993		1303
994	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1304	// special case the rather common 1..5-digit-int case
995	if (*start == '-')	1305	if (*start == '-')
996	switch (len)	1306	switch (len)
997	{	1307	{
998	case 2: return newSViv (-( start [1] - '0' * 1));	1308	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
999	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1309	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1000	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1310	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1001	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1311	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1312	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1002	}	1313	}
1003	else	1314	else
1004	switch (len)	1315	switch (len)
1005	{	1316	{
1006	case 1: return newSViv ( start [0] - '0' * 1);	1317	case 1: return newSViv ( start [0] - '0' * 1);
1007	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1318	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1008	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1319	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1009	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1320	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1321	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1010	}	1322	}
1011		1323
1012	{	1324	{
1013	UV uv;	1325	UV uv;
1014	int numtype = grok_number (start, len, &uv);	1326	int numtype = grok_number (start, len, &uv);
…		…
1023	}	1335	}
1024		1336
1025	len -= *start == '-' ? 1 : 0;	1337	len -= *start == '-' ? 1 : 0;
1026		1338
1027	// does not fit into IV or UV, try NV	1339	// does not fit into IV or UV, try NV
1028	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1340	if (len <= NV_DIG)
1029	#if defined (LDBL_DIG)
1030	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1031	#endif
1032	)
1033	// fits into NV without loss of precision	1341	// fits into NV without loss of precision
1034	return newSVnv (Atof (start));	1342	return newSVnv (json_atof (start));
1035		1343
1036	// everything else fails, convert it to a string	1344	// everything else fails, convert it to a string
1037	return newSVpvn (start, dec->cur - start);	1345	return newSVpvn (start, dec->cur - start);
1038	}	1346	}
1039		1347
1040	// loss of precision here	1348	// loss of precision here
1041	return newSVnv (Atof (start));	1349	return newSVnv (json_atof (start));
1042		1350
1043	fail:	1351	fail:
1044	return 0;	1352	return 0;
1045	}	1353	}
1046		1354
…		…
1070	if (*dec->cur == ']')	1378	if (*dec->cur == ']')
1071	{	1379	{
1072	++dec->cur;	1380	++dec->cur;
1073	break;	1381	break;
1074	}	1382	}
1075		1383
1076	if (*dec->cur != ',')	1384	if (*dec->cur != ',')
1077	ERR (", or ] expected while parsing array");	1385	ERR (", or ] expected while parsing array");
1078		1386
1079	++dec->cur;	1387	++dec->cur;
1080		1388
…		…
1122	char *p = dec->cur;	1430	char *p = dec->cur;
1123	char *e = p + 24; // only try up to 24 bytes	1431	char *e = p + 24; // only try up to 24 bytes
1124		1432
1125	for (;;)	1433	for (;;)
1126	{	1434	{
1127	// the >= 0x80 is true on most architectures	1435	// the >= 0x80 is false on most architectures
1128	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1436	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1129	{	1437	{
1130	// slow path, back up and use decode_str	1438	// slow path, back up and use decode_str
1131	SV *key = decode_str (dec);	1439	SV *key = decode_str (dec);
1132	if (!key)	1440	if (!key)
…		…
1204		1512
1205	hv_iterinit (hv);	1513	hv_iterinit (hv);
1206	he = hv_iternext (hv);	1514	he = hv_iternext (hv);
1207	hv_iterinit (hv);	1515	hv_iterinit (hv);
1208		1516
1209	// the next line creates a mortal sv each time its called.	1517	// the next line creates a mortal sv each time it's called.
1210	// might want to optimise this for common cases.	1518	// might want to optimise this for common cases.
1211	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);	1519	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1212		1520
1213	if (cb)	1521	if (cb)
1214	{	1522	{
1215	dSP;	1523	dSP;
1216	int count;	1524	int count;
1217		1525
1218	ENTER; SAVETMPS; PUSHMARK (SP);	1526	ENTER; SAVETMPS;
		1527	PUSHMARK (SP);
1219	XPUSHs (HeVAL (he));	1528	XPUSHs (HeVAL (he));
		1529	sv_2mortal (sv);
1220		1530
1221	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1531	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1222		1532
1223	if (count == 1)	1533	if (count == 1)
1224	{	1534	{
1225	sv = newSVsv (POPs);	1535	sv = newSVsv (POPs);
1226	FREETMPS; LEAVE;	1536	FREETMPS; LEAVE;
1227	return sv;	1537	return sv;
1228	}	1538	}
		1539	else if (count)
		1540	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1229		1541
		1542	SvREFCNT_inc (sv);
1230	FREETMPS; LEAVE;	1543	FREETMPS; LEAVE;
1231	}	1544	}
1232	}	1545	}
1233		1546
1234	if (dec->json.cb_object)	1547	if (dec->json.cb_object)
1235	{	1548	{
1236	dSP;	1549	dSP;
1237	int count;	1550	int count;
1238		1551
1239	ENTER; SAVETMPS; PUSHMARK (SP);	1552	ENTER; SAVETMPS;
		1553	PUSHMARK (SP);
1240	XPUSHs (sv_2mortal (sv));	1554	XPUSHs (sv_2mortal (sv));
1241		1555
1242	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1556	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1243		1557
1244	if (count == 1)	1558	if (count == 1)
1245	{	1559	{
1246	sv = newSVsv (POPs);	1560	sv = newSVsv (POPs);
1247	FREETMPS; LEAVE;	1561	FREETMPS; LEAVE;
1248	return sv;	1562	return sv;
1249	}	1563	}
		1564	else if (count)
		1565	croak ("filter_json_object callbacks must not return more than one scalar");
1250		1566
1251	SvREFCNT_inc (sv);	1567	SvREFCNT_inc (sv);
1252	FREETMPS; LEAVE;	1568	FREETMPS; LEAVE;
1253	}	1569	}
1254	}	1570	}
…		…
1260	DEC_DEC_DEPTH;	1576	DEC_DEC_DEPTH;
1261	return 0;	1577	return 0;
1262	}	1578	}
1263		1579
1264	static SV *	1580	static SV *
		1581	decode_tag (dec_t *dec)
		1582	{
		1583	SV *tag = 0;
		1584	SV *val = 0;
		1585
		1586	if (!(dec->json.flags & F_ALLOW_TAGS))
		1587	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1588
		1589	++dec->cur;
		1590
		1591	decode_ws (dec);
		1592
		1593	tag = decode_sv (dec);
		1594	if (!tag)
		1595	goto fail;
		1596
		1597	if (!SvPOK (tag))
		1598	ERR ("malformed JSON string, (tag) must be a string");
		1599
		1600	decode_ws (dec);
		1601
		1602	if (*dec->cur != ')')
		1603	ERR (") expected after tag");
		1604
		1605	++dec->cur;
		1606
		1607	decode_ws (dec);
		1608
		1609	val = decode_sv (dec);
		1610	if (!val)
		1611	goto fail;
		1612
		1613	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1614	ERR ("malformed JSON string, tag value must be an array");
		1615
		1616	{
		1617	AV *av = (AV *)SvRV (val);
		1618	int i, len = av_len (av) + 1;
		1619	HV *stash = gv_stashsv (tag, 0);
		1620	SV *sv;
		1621
		1622	if (!stash)
		1623	ERR ("cannot decode perl-object (package does not exist)");
		1624
		1625	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1626
		1627	if (!method)
		1628	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1629
		1630	dSP;
		1631
		1632	ENTER; SAVETMPS;
		1633	PUSHMARK (SP);
		1634	EXTEND (SP, len + 2);
		1635	// we re-bless the reference to get overload and other niceties right
		1636	PUSHs (tag);
		1637	PUSHs (sv_json);
		1638
		1639	for (i = 0; i < len; ++i)
		1640	PUSHs (*av_fetch (av, i, 1));
		1641
		1642	PUTBACK;
		1643	call_sv ((SV *)GvCV (method), G_SCALAR);
		1644	SPAGAIN;
		1645
		1646	SvREFCNT_dec (tag);
		1647	SvREFCNT_dec (val);
		1648	sv = SvREFCNT_inc (POPs);
		1649
		1650	PUTBACK;
		1651
		1652	FREETMPS; LEAVE;
		1653
		1654	return sv;
		1655	}
		1656
		1657	fail:
		1658	SvREFCNT_dec (tag);
		1659	SvREFCNT_dec (val);
		1660	return 0;
		1661	}
		1662
		1663	static SV *
1265	decode_sv (dec_t *dec)	1664	decode_sv (dec_t *dec)
1266	{	1665	{
1267	// the beauty of JSON: you need exactly one character lookahead	1666	// the beauty of JSON: you need exactly one character lookahead
1268	// to parse anything.	1667	// to parse everything.
1269	switch (*dec->cur)	1668	switch (*dec->cur)
1270	{	1669	{
1271	case '"': ++dec->cur; return decode_str (dec);	1670	case '"': ++dec->cur; return decode_str (dec);
1272	case '[': ++dec->cur; return decode_av (dec);	1671	case '[': ++dec->cur; return decode_av (dec);
1273	case '{': ++dec->cur; return decode_hv (dec);	1672	case '{': ++dec->cur; return decode_hv (dec);
		1673	case '(': return decode_tag (dec);
1274		1674
1275	case '-':	1675	case '-':
1276	case '0': case '1': case '2': case '3': case '4':	1676	case '0': case '1': case '2': case '3': case '4':
1277	case '5': case '6': case '7': case '8': case '9':	1677	case '5': case '6': case '7': case '8': case '9':
1278	return decode_num (dec);	1678	return decode_num (dec);
1279		1679
		1680	case 'f':
		1681	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
		1682	{
		1683	dec->cur += 5;
		1684
		1685	if (expect_false (!dec->json.v_false))
		1686	dec->json.v_false = GET_BOOL (false);
		1687
		1688	return newSVsv (dec->json.v_false);
		1689	}
		1690	else
		1691	ERR ("'false' expected");
		1692
		1693	break;
		1694
1280	case 't':	1695	case 't':
1281	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1696	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1282	{	1697	{
1283	dec->cur += 4;	1698	dec->cur += 4;
1284	#if JSON_SLOW	1699
1285	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1700	if (expect_false (!dec->json.v_true))
1286	#endif	1701	dec->json.v_true = GET_BOOL (true);
1287	return SvREFCNT_inc (json_true);	1702
		1703	return newSVsv (dec->json.v_true);
1288	}	1704	}
1289	else	1705	else
1290	ERR ("'true' expected");	1706	ERR ("'true' expected");
1291
1292	break;
1293
1294	case 'f':
1295	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1296	{
1297	dec->cur += 5;
1298	#if JSON_SLOW
1299	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1300	#endif
1301	return SvREFCNT_inc (json_false);
1302	}
1303	else
1304	ERR ("'false' expected");
1305		1707
1306	break;	1708	break;
1307		1709
1308	case 'n':	1710	case 'n':
1309	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1711	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
…		…
1315	ERR ("'null' expected");	1717	ERR ("'null' expected");
1316		1718
1317	break;	1719	break;
1318		1720
1319	default:	1721	default:
1320	ERR ("malformed JSON string, neither array, object, number, string or atom");	1722	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1321	break;	1723	break;
1322	}	1724	}
1323		1725
1324	fail:	1726	fail:
1325	return 0;	1727	return 0;
1326	}	1728	}
1327		1729
1328	static SV *	1730	static SV *
1329	decode_json (SV *string, JSON *json, UV *offset_return)	1731	decode_json (SV *string, JSON *json, STRLEN *offset_return)
1330	{	1732	{
1331	dec_t dec;	1733	dec_t dec;
1332	UV offset;
1333	SV *sv;	1734	SV *sv;
1334		1735
		1736	/* work around bugs in 5.10 where manipulating magic values
		1737	* makes perl ignore the magic in subsequent accesses.
		1738	* also make a copy of non-PV values, to get them into a clean
		1739	* state (SvPV should do that, but it's buggy, see below).
		1740	*
		1741	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1742	* as reported by Reini Urban.
		1743	*/
1335	SvGETMAGIC (string);	1744	/*SvGETMAGIC (string);*/
		1745	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1746	string = sv_2mortal (newSVsv (string));
		1747
1336	SvUPGRADE (string, SVt_PV);	1748	SvUPGRADE (string, SVt_PV);
1337		1749
1338	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1750	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1751	* assertion with -DDEBUGGING, although SvCUR is documented to
		1752	* return the xpv_cur field which certainly exists after upgrading.
		1753	* according to nicholas clark, calling SvPOK fixes this.
		1754	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1755	* and hope for the best.
		1756	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1757	* assertion business is seriously broken, try yet another workaround
		1758	* for the broken -DDEBUGGING.
		1759	*/
		1760	{
		1761	#ifdef DEBUGGING
		1762	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1763	#else
		1764	STRLEN offset = SvCUR (string);
		1765	#endif
		1766
		1767	if (offset > json->max_size && json->max_size)
1339	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1768	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1340	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1769	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1770	}
1341		1771
1342	if (json->flags & F_UTF8)	1772	if (DECODE_WANTS_OCTETS (json))
1343	sv_utf8_downgrade (string, 0);	1773	sv_utf8_downgrade (string, 0);
1344	else	1774	else
1345	sv_utf8_upgrade (string);	1775	sv_utf8_upgrade (string);
1346		1776
1347	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1777	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1348		1778
1349	dec.json = *json;	1779	dec.json = *json;
1350	dec.cur = SvPVX (string);	1780	dec.cur = SvPVX (string);
1351	dec.end = SvEND (string);	1781	dec.end = SvEND (string);
1352	dec.err = 0;	1782	dec.err = 0;
1353	dec.depth = 0;	1783	dec.depth = 0;
1354	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1355		1784
1356	if (dec.json.cb_object || dec.json.cb_sk_object)	1785	if (dec.json.cb_object || dec.json.cb_sk_object)
1357	dec.json.flags |= F_HOOK;	1786	dec.json.flags |= F_HOOK;
1358		1787
1359	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1788	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1360		1789
1361	decode_ws (&dec);	1790	decode_ws (&dec);
1362	sv = decode_sv (&dec);	1791	sv = decode_sv (&dec);
1363		1792
1364	if (!(offset_return || !sv))	1793	if (offset_return)
		1794	*offset_return = dec.cur - SvPVX (string);
		1795	else if (sv)
1365	{	1796	{
1366	// check for trailing garbage	1797	// check for trailing garbage
1367	decode_ws (&dec);	1798	decode_ws (&dec);
1368		1799
1369	if (*dec.cur)	1800	if (dec.cur != dec.end)
1370	{	1801	{
1371	dec.err = "garbage after JSON object";	1802	dec.err = "garbage after JSON object";
1372	SvREFCNT_dec (sv);	1803	SvREFCNT_dec (sv);
1373	sv = 0;	1804	sv = 0;
1374	}	1805	}
1375	}
1376
1377	if (offset_return || !sv)
1378	{
1379	offset = dec.json.flags & F_UTF8
1380	? dec.cur - SvPVX (string)
1381	: utf8_distance (dec.cur, SvPVX (string));
1382
1383	if (offset_return)
1384	*offset_return = offset;
1385	}	1806	}
1386		1807
1387	if (!sv)	1808	if (!sv)
1388	{	1809	{
1389	SV *uni = sv_newmortal ();	1810	SV *uni = sv_newmortal ();
…		…
1395	SAVEVPTR (PL_curcop);	1816	SAVEVPTR (PL_curcop);
1396	PL_curcop = &cop;	1817	PL_curcop = &cop;
1397	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1818	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1398	LEAVE;	1819	LEAVE;
1399		1820
1400	croak ("%s, at character offset %d [\"%s\"]",	1821	croak ("%s, at character offset %d (before \"%s\")",
1401	dec.err,	1822	dec.err,
1402	(int)offset,	1823	(int)ptr_to_index (string, dec.cur),
1403	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1824	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1404	}	1825	}
1405		1826
1406	sv = sv_2mortal (sv);	1827	sv = sv_2mortal (sv);
1407		1828
1408	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1829	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1409	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1830	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1410		1831
1411	return sv;	1832	return sv;
		1833	}
		1834
		1835	/////////////////////////////////////////////////////////////////////////////
		1836	// incremental parser
		1837
		1838	static void
		1839	incr_parse (JSON *self)
		1840	{
		1841	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1842
		1843	// the state machine here is a bit convoluted and could be simplified a lot
		1844	// but this would make it slower, so...
		1845
		1846	for (;;)
		1847	{
		1848	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1849	switch (self->incr_mode)
		1850	{
		1851	// only used for initial whitespace skipping
		1852	case INCR_M_WS:
		1853	for (;;)
		1854	{
		1855	if (*p > 0x20)
		1856	{
		1857	if (*p == '#')
		1858	{
		1859	self->incr_mode = INCR_M_C0;
		1860	goto incr_m_c;
		1861	}
		1862	else
		1863	{
		1864	self->incr_mode = INCR_M_JSON;
		1865	goto incr_m_json;
		1866	}
		1867	}
		1868	else if (!*p)
		1869	goto interrupt;
		1870
		1871	++p;
		1872	}
		1873
		1874	// skip a single char inside a string (for \\-processing)
		1875	case INCR_M_BS:
		1876	if (!*p)
		1877	goto interrupt;
		1878
		1879	++p;
		1880	self->incr_mode = INCR_M_STR;
		1881	goto incr_m_str;
		1882
		1883	// inside #-style comments
		1884	case INCR_M_C0:
		1885	case INCR_M_C1:
		1886	incr_m_c:
		1887	for (;;)
		1888	{
		1889	if (*p == '\n')
		1890	{
		1891	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1892	break;
		1893	}
		1894	else if (!*p)
		1895	goto interrupt;
		1896
		1897	++p;
		1898	}
		1899
		1900	break;
		1901
		1902	// inside a string
		1903	case INCR_M_STR:
		1904	incr_m_str:
		1905	for (;;)
		1906	{
		1907	if (*p == '"')
		1908	{
		1909	++p;
		1910	self->incr_mode = INCR_M_JSON;
		1911
		1912	if (!self->incr_nest)
		1913	goto interrupt;
		1914
		1915	goto incr_m_json;
		1916	}
		1917	else if (*p == '\\')
		1918	{
		1919	++p; // "virtually" consumes character after \
		1920
		1921	if (!*p) // if at end of string we have to switch modes
		1922	{
		1923	self->incr_mode = INCR_M_BS;
		1924	goto interrupt;
		1925	}
		1926	}
		1927	else if (!*p)
		1928	goto interrupt;
		1929
		1930	++p;
		1931	}
		1932
		1933	// after initial ws, outside string
		1934	case INCR_M_JSON:
		1935	incr_m_json:
		1936	for (;;)
		1937	{
		1938	switch (*p++)
		1939	{
		1940	case 0:
		1941	--p;
		1942	goto interrupt;
		1943
		1944	case 0x09:
		1945	case 0x0a:
		1946	case 0x0d:
		1947	case 0x20:
		1948	if (!self->incr_nest)
		1949	{
		1950	--p; // do not eat the whitespace, let the next round do it
		1951	goto interrupt;
		1952	}
		1953	break;
		1954
		1955	case '"':
		1956	self->incr_mode = INCR_M_STR;
		1957	goto incr_m_str;
		1958
		1959	case '[':
		1960	case '{':
		1961	case '(':
		1962	if (++self->incr_nest > self->max_depth)
		1963	croak (ERR_NESTING_EXCEEDED);
		1964	break;
		1965
		1966	case ']':
		1967	case '}':
		1968	if (--self->incr_nest <= 0)
		1969	goto interrupt;
		1970	break;
		1971
		1972	case ')':
		1973	--self->incr_nest;
		1974	break;
		1975
		1976	case '#':
		1977	self->incr_mode = INCR_M_C1;
		1978	goto incr_m_c;
		1979	}
		1980	}
		1981	}
		1982
		1983	modechange:
		1984	;
		1985	}
		1986
		1987	interrupt:
		1988	self->incr_pos = p - SvPVX (self->incr_text);
		1989	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1990	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1412	}	1991	}
1413		1992
1414	/////////////////////////////////////////////////////////////////////////////	1993	/////////////////////////////////////////////////////////////////////////////
1415	// XS interface functions	1994	// XS interface functions
1416		1995
…		…
1425	i >= '0' && i <= '9' ? i - '0'	2004	i >= '0' && i <= '9' ? i - '0'
1426	: i >= 'a' && i <= 'f' ? i - 'a' + 10	2005	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1427	: i >= 'A' && i <= 'F' ? i - 'A' + 10	2006	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1428	: -1;	2007	: -1;
1429		2008
1430	json_stash = gv_stashpv ("JSON::XS" , 1);	2009	json_stash = gv_stashpv ("JSON::XS" , 1);
1431	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	2010	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		2011	bool_false = get_bool ("Types::Serialiser::false");
		2012	bool_true = get_bool ("Types::Serialiser::true");
1432		2013
1433	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	2014	sv_json = newSVpv ("JSON", 0);
1434	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	2015	SvREADONLY_on (sv_json);
		2016
		2017	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1435	}	2018	}
1436		2019
1437	PROTOTYPES: DISABLE	2020	PROTOTYPES: DISABLE
1438		2021
1439	void CLONE (...)	2022	void CLONE (...)
1440	CODE:	2023	CODE:
		2024	// as long as these writes are atomic, the race should not matter
		2025	// as existing threads either already use 0, or use the old value,
		2026	// which is sitll correct for the initial thread.
1441	json_stash = 0;	2027	json_stash = 0;
1442	json_boolean_stash = 0;	2028	bool_stash = 0;
		2029	bool_false = 0;
		2030	bool_true = 0;
1443		2031
1444	void new (char *klass)	2032	void new (char *klass)
1445	PPCODE:	2033	PPCODE:
1446	{	2034	{
1447	SV *pv = NEWSV (0, sizeof (JSON));	2035	SV *pv = NEWSV (0, sizeof (JSON));
1448	SvPOK_only (pv);	2036	SvPOK_only (pv);
1449	Zero (SvPVX (pv), 1, JSON);	2037	json_init ((JSON *)SvPVX (pv));
1450	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	2038	XPUSHs (sv_2mortal (sv_bless (
1451	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	2039	newRV_noinc (pv),
		2040	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2041	)));
1452	}	2042	}
		2043
		2044	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2045	PPCODE:
		2046	self->v_false = newSVsv (v_false);
		2047	self->v_true = newSVsv (v_true);
		2048	XPUSHs (ST (0));
		2049
		2050	void get_boolean_values (JSON *self)
		2051	PPCODE:
		2052	if (self->v_false && self->v_true)
		2053	{
		2054	EXTEND (SP, 2);
		2055	PUSHs (self->v_false);
		2056	PUSHs (self->v_true);
		2057	}
1453		2058
1454	void ascii (JSON *self, int enable = 1)	2059	void ascii (JSON *self, int enable = 1)
1455	ALIAS:	2060	ALIAS:
1456	ascii = F_ASCII	2061	ascii = F_ASCII
1457	latin1 = F_LATIN1	2062	latin1 = F_LATIN1
…		…
1464	allow_nonref = F_ALLOW_NONREF	2069	allow_nonref = F_ALLOW_NONREF
1465	shrink = F_SHRINK	2070	shrink = F_SHRINK
1466	allow_blessed = F_ALLOW_BLESSED	2071	allow_blessed = F_ALLOW_BLESSED
1467	convert_blessed = F_CONV_BLESSED	2072	convert_blessed = F_CONV_BLESSED
1468	relaxed = F_RELAXED	2073	relaxed = F_RELAXED
		2074	allow_unknown = F_ALLOW_UNKNOWN
		2075	allow_tags = F_ALLOW_TAGS
1469	PPCODE:	2076	PPCODE:
1470	{	2077	{
1471	if (enable)	2078	if (enable)
1472	self->flags |= ix;	2079	self->flags |= ix;
1473	else	2080	else
1474	self->flags &= ~ix;	2081	self->flags &= ~ix;
1475		2082
1476	XPUSHs (ST (0));	2083	XPUSHs (ST (0));
1477	}	2084	}
1478		2085
1479	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	2086	void get_ascii (JSON *self)
		2087	ALIAS:
		2088	get_ascii = F_ASCII
		2089	get_latin1 = F_LATIN1
		2090	get_utf8 = F_UTF8
		2091	get_indent = F_INDENT
		2092	get_canonical = F_CANONICAL
		2093	get_space_before = F_SPACE_BEFORE
		2094	get_space_after = F_SPACE_AFTER
		2095	get_allow_nonref = F_ALLOW_NONREF
		2096	get_shrink = F_SHRINK
		2097	get_allow_blessed = F_ALLOW_BLESSED
		2098	get_convert_blessed = F_CONV_BLESSED
		2099	get_relaxed = F_RELAXED
		2100	get_allow_unknown = F_ALLOW_UNKNOWN
		2101	get_allow_tags = F_ALLOW_TAGS
1480	PPCODE:	2102	PPCODE:
1481	{	2103	XPUSHs (boolSV (self->flags & ix));
1482	UV log2 = 0;
1483		2104
1484	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2105	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1485		2106	PPCODE:
1486	while ((1UL << log2) < max_depth)	2107	self->max_depth = max_depth;
1487	++log2;
1488
1489	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1490
1491	XPUSHs (ST (0));	2108	XPUSHs (ST (0));
1492	}
1493		2109
		2110	U32 get_max_depth (JSON *self)
		2111	CODE:
		2112	RETVAL = self->max_depth;
		2113	OUTPUT:
		2114	RETVAL
		2115
1494	void max_size (JSON *self, UV max_size = 0)	2116	void max_size (JSON *self, U32 max_size = 0)
1495	PPCODE:	2117	PPCODE:
1496	{	2118	self->max_size = max_size;
1497	UV log2 = 0;
1498
1499	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1500	if (max_size == 1) max_size = 2;
1501
1502	while ((1UL << log2) < max_size)
1503	++log2;
1504
1505	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1506
1507	XPUSHs (ST (0));	2119	XPUSHs (ST (0));
1508	}	2120
		2121	int get_max_size (JSON *self)
		2122	CODE:
		2123	RETVAL = self->max_size;
		2124	OUTPUT:
		2125	RETVAL
1509		2126
1510	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)	2127	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1511	PPCODE:	2128	PPCODE:
1512	{	2129	{
1513	SvREFCNT_dec (self->cb_object);	2130	SvREFCNT_dec (self->cb_object);
…		…
1517	}	2134	}
1518		2135
1519	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)	2136	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1520	PPCODE:	2137	PPCODE:
1521	{	2138	{
1522	if (!self->cb_sk_object)	2139	if (!self->cb_sk_object)
1523	self->cb_sk_object = newHV ();	2140	self->cb_sk_object = newHV ();
1524		2141
1525	if (SvOK (cb))	2142	if (SvOK (cb))
1526	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);	2143	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1527	else	2144	else
…		…
1538	XPUSHs (ST (0));	2155	XPUSHs (ST (0));
1539	}	2156	}
1540		2157
1541	void encode (JSON *self, SV *scalar)	2158	void encode (JSON *self, SV *scalar)
1542	PPCODE:	2159	PPCODE:
1543	XPUSHs (encode_json (scalar, self));	2160	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2161	XPUSHs (scalar);
1544		2162
1545	void decode (JSON *self, SV *jsonstr)	2163	void decode (JSON *self, SV *jsonstr)
1546	PPCODE:	2164	PPCODE:
1547	XPUSHs (decode_json (jsonstr, self, 0));	2165	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2166	XPUSHs (jsonstr);
1548		2167
1549	void decode_prefix (JSON *self, SV *jsonstr)	2168	void decode_prefix (JSON *self, SV *jsonstr)
1550	PPCODE:	2169	PPCODE:
1551	{	2170	{
1552	UV offset;	2171	SV *sv;
		2172	STRLEN offset;
		2173	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1553	EXTEND (SP, 2);	2174	EXTEND (SP, 2);
1554	PUSHs (decode_json (jsonstr, self, &offset));	2175	PUSHs (sv);
1555	PUSHs (sv_2mortal (newSVuv (offset)));	2176	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2177	}
		2178
		2179	void incr_parse (JSON *self, SV *jsonstr = 0)
		2180	PPCODE:
		2181	{
		2182	if (!self->incr_text)
		2183	self->incr_text = newSVpvn ("", 0);
		2184
		2185	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2186	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2187	if (DECODE_WANTS_OCTETS (self))
		2188	{
		2189	if (self->incr_pos)
		2190	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2191	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2192
		2193	sv_utf8_downgrade (self->incr_text, 0);
		2194	}
		2195	else
		2196	{
		2197	sv_utf8_upgrade (self->incr_text);
		2198
		2199	if (self->incr_pos)
		2200	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2201	- (U8 *)SvPVX (self->incr_text);
		2202	}
		2203
		2204	// append data, if any
		2205	if (jsonstr)
		2206	{
		2207	/* make sure both strings have same encoding */
		2208	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2209	if (SvUTF8 (jsonstr))
		2210	sv_utf8_downgrade (jsonstr, 0);
		2211	else
		2212	sv_utf8_upgrade (jsonstr);
		2213
		2214	/* and then just blindly append */
		2215	{
		2216	STRLEN len;
		2217	const char *str = SvPV (jsonstr, len);
		2218	STRLEN cur = SvCUR (self->incr_text);
		2219
		2220	if (SvLEN (self->incr_text) - cur <= len)
		2221	json_sv_grow (self->incr_text, cur, len);
		2222
		2223	Move (str, SvEND (self->incr_text), len, char);
		2224	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2225	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2226	}
		2227	}
		2228
		2229	if (GIMME_V != G_VOID)
		2230	do
		2231	{
		2232	SV *sv;
		2233	STRLEN offset;
		2234
		2235	if (!INCR_DONE (self))
		2236	{
		2237	incr_parse (self);
		2238
		2239	if (self->incr_pos > self->max_size && self->max_size)
		2240	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2241	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2242
		2243	if (!INCR_DONE (self))
		2244	{
		2245	// as an optimisation, do not accumulate white space in the incr buffer
		2246	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2247	{
		2248	self->incr_pos = 0;
		2249	SvCUR_set (self->incr_text, 0);
		2250	}
		2251
		2252	break;
		2253	}
		2254	}
		2255
		2256	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2257	XPUSHs (sv);
		2258
		2259	self->incr_pos -= offset;
		2260	self->incr_nest = 0;
		2261	self->incr_mode = 0;
		2262
		2263	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2264	}
		2265	while (GIMME_V == G_ARRAY);
		2266	}
		2267
		2268	SV *incr_text (JSON *self)
		2269	ATTRS: lvalue
		2270	CODE:
		2271	{
		2272	if (self->incr_pos)
		2273	croak ("incr_text can not be called when the incremental parser already started parsing");
		2274
		2275	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2276	}
		2277	OUTPUT:
		2278	RETVAL
		2279
		2280	void incr_skip (JSON *self)
		2281	CODE:
		2282	{
		2283	if (self->incr_pos)
		2284	{
		2285	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2286	self->incr_pos = 0;
		2287	self->incr_nest = 0;
		2288	self->incr_mode = 0;
		2289	}
		2290	}
		2291
		2292	void incr_reset (JSON *self)
		2293	CODE:
		2294	{
		2295	SvREFCNT_dec (self->incr_text);
		2296	self->incr_text = 0;
		2297	self->incr_pos = 0;
		2298	self->incr_nest = 0;
		2299	self->incr_mode = 0;
1556	}	2300	}
1557		2301
1558	void DESTROY (JSON *self)	2302	void DESTROY (JSON *self)
1559	CODE:	2303	CODE:
		2304	SvREFCNT_dec (self->v_false);
		2305	SvREFCNT_dec (self->v_true);
1560	SvREFCNT_dec (self->cb_sk_object);	2306	SvREFCNT_dec (self->cb_sk_object);
1561	SvREFCNT_dec (self->cb_object);	2307	SvREFCNT_dec (self->cb_object);
		2308	SvREFCNT_dec (self->incr_text);
1562		2309
1563	PROTOTYPES: ENABLE	2310	PROTOTYPES: ENABLE
1564		2311
1565	void to_json (SV *scalar)	2312	void encode_json (SV *scalar)
1566	PPCODE:	2313	PPCODE:
1567	{	2314	{
1568	JSON json = { F_DEFAULT | F_UTF8 };	2315	JSON json;
1569	XPUSHs (encode_json (scalar, &json));	2316	json_init (&json);
		2317	json.flags |= F_UTF8;
		2318	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2319	XPUSHs (scalar);
1570	}	2320	}
1571		2321
1572	void from_json (SV *jsonstr)	2322	void decode_json (SV *jsonstr)
1573	PPCODE:	2323	PPCODE:
1574	{	2324	{
1575	JSON json = { F_DEFAULT | F_UTF8 };	2325	JSON json;
1576	XPUSHs (decode_json (jsonstr, &json, 0));	2326	json_init (&json);
		2327	json.flags |= F_UTF8;
		2328	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2329	XPUSHs (jsonstr);
1577	}	2330	}
1578		2331

Diff Legend

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