ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC vs.
Revision 1.134 by root, Thu Nov 15 20:13:03 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	Zero (json, 1, JSON);
		131	json->max_depth = 512;
		132	}
81		133
82	/////////////////////////////////////////////////////////////////////////////	134	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	135	// utility functions
84		136
85	inline void	137	INLINE SV *
		138	get_bool (const char *name)
		139	{
		140	SV *sv = get_sv (name, 1);
		141
		142	SvREADONLY_on (sv);
		143	SvREADONLY_on (SvRV (sv));
		144
		145	return sv;
		146	}
		147
		148	INLINE void
86	shrink (SV *sv)	149	shrink (SV *sv)
87	{	150	{
88	sv_utf8_downgrade (sv, 1);	151	sv_utf8_downgrade (sv, 1);
		152
89	if (SvLEN (sv) > SvCUR (sv) + 1)	153	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	154	{
91	#ifdef SvPV_shrink_to_cur	155	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	156	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	157	#elif defined (SvPV_renew)
94	SvPV_renew (sv, SvCUR (sv) + 1);	158	SvPV_renew (sv, SvCUR (sv) + 1);
95	#endif	159	#endif
96	}	160	}
97	}	161	}
98		162
		163	/* adds two STRLENs together, slow, and with paranoia */
		164	STRLEN
		165	strlen_sum (STRLEN l1, STRLEN l2)
		166	{
		167	size_t sum = l1 + l2;
		168
		169	if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
		170	croak ("JSON::XS: string size overflow");
		171
		172	return sum;
		173	}
		174
		175	/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
		176	static char *
		177	json_sv_grow (SV *sv, size_t len1, size_t len2)
		178	{
		179	len1 = strlen_sum (len1, len2);
		180	len1 = strlen_sum (len1, len1 >> 1);
		181
		182	if (len1 > 4096 - 24)
		183	len1 = (len1 | 4095) - 24;
		184
		185	return SvGROW (sv, len1);
		186	}
		187
99	// decode an utf-8 character and return it, or (UV)-1 in	188	// decode a utf-8 character and return it, or (UV)-1 in
100	// case of an error.	189	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	190	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	191	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	192	// note that we never call this function for a ascii codepoints
104	inline UV	193	INLINE UV
105	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	194	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
106	{	195	{
107	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))	196	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	197	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	198	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	199	{
111	*clen = 2;	200	*clen = 2;
112	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	201	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
113	}	202	}
114	else	203	else
		204	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		205	}
		206
		207	// likewise for encoding, also never called for ascii codepoints
		208	// this function takes advantage of this fact, although current gccs
		209	// seem to optimise the check for >= 0x80 away anyways
		210	INLINE unsigned char *
		211	encode_utf8 (unsigned char *s, UV ch)
		212	{
		213	if (expect_false (ch < 0x000080))
		214	*s++ = ch;
		215	else if (expect_true (ch < 0x000800))
		216	*s++ = 0xc0 | ( ch >> 6),
		217	*s++ = 0x80 | ( ch & 0x3f);
		218	else if ( ch < 0x010000)
		219	*s++ = 0xe0 | ( ch >> 12),
		220	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		221	*s++ = 0x80 | ( ch & 0x3f);
		222	else if ( ch < 0x110000)
		223	*s++ = 0xf0 | ( ch >> 18),
		224	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		225	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		226	*s++ = 0x80 | ( ch & 0x3f);
		227
		228	return s;
		229	}
		230
		231	// convert offset pointer to character index, sv must be string
		232	static STRLEN
		233	ptr_to_index (SV *sv, char *offset)
		234	{
		235	return SvUTF8 (sv)
		236	? utf8_distance (offset, SvPVX (sv))
		237	: offset - SvPVX (sv);
		238	}
		239
		240	/////////////////////////////////////////////////////////////////////////////
		241	// fp hell
		242
		243	// scan a group of digits, and a trailing exponent
		244	static void
		245	json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
		246	{
		247	UV uaccum = 0;
		248	int eaccum = 0;
		249
		250	// if we recurse too deep, skip all remaining digits
		251	// to avoid a stack overflow attack
		252	if (expect_false (--maxdepth <= 0))
		253	while (((U8)*s - '0') < 10)
		254	++s;
		255
		256	for (;;)
		257	{
		258	U8 dig = (U8)*s - '0';
		259
		260	if (expect_false (dig >= 10))
		261	{
		262	if (dig == (U8)((U8)'.' - (U8)'0'))
		263	{
		264	++s;
		265	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		266	}
		267	else if ((dig | ' ') == 'e' - '0')
		268	{
		269	int exp2 = 0;
		270	int neg = 0;
		271
		272	++s;
		273
		274	if (*s == '-')
		275	{
		276	++s;
		277	neg = 1;
		278	}
		279	else if (*s == '+')
		280	++s;
		281
		282	while ((dig = (U8)*s - '0') < 10)
		283	exp2 = exp2 * 10 + *s++ - '0';
		284
		285	*expo += neg ? -exp2 : exp2;
		286	}
		287
		288	break;
		289	}
		290
		291	++s;
		292
		293	uaccum = uaccum * 10 + dig;
		294	++eaccum;
		295
		296	// if we have too many digits, then recurse for more
		297	// we actually do this for rather few digits
		298	if (uaccum >= (UV_MAX - 9) / 10)
		299	{
		300	if (postdp) *expo -= eaccum;
		301	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		302	if (postdp) *expo += eaccum;
		303
		304	break;
		305	}
115	{	306	}
116	*clen = (STRLEN)-1;	307
		308	// this relies greatly on the quality of the pow ()
		309	// implementation of the platform, but a good
		310	// implementation is hard to beat.
		311	// (IEEE 754 conformant ones are required to be exact)
		312	if (postdp) *expo -= eaccum;
		313	*accum += uaccum * Perl_pow (10., *expo);
		314	*expo += eaccum;
		315	}
		316
		317	static NV
		318	json_atof (const char *s)
		319	{
		320	NV accum = 0.;
		321	int expo = 0;
		322	int neg = 0;
		323
		324	if (*s == '-')
		325	{
		326	++s;
		327	neg = 1;
		328	}
		329
		330	// a recursion depth of ten gives us >>500 bits
		331	json_atof_scan1 (s, &accum, &expo, 0, 10);
		332
		333	return neg ? -accum : accum;
		334	}
		335
		336	// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
		337	static int
		338	ref_bool_type (SV *sv)
		339	{
		340	svtype svt = SvTYPE (sv);
		341
		342	if (svt < SVt_PVAV)
		343	{
		344	STRLEN len = 0;
		345	char *pv = svt ? SvPV (sv, len) : 0;
		346
		347	if (len == 1)
		348	if (*pv == '1')
		349	return 1;
		350	else if (*pv == '0')
		351	return 0;
		352	}
		353
		354	return -1;
		355	}
		356
		357	// returns whether scalar is not a reference in the sense of allow_nonref
		358	static int
		359	json_nonref (SV *scalar)
		360	{
		361	if (!SvROK (scalar))
		362	return 1;
		363
		364	scalar = SvRV (scalar);
		365
		366	if (SvTYPE (scalar) >= SVt_PVMG)
		367	{
		368	if (SvSTASH (scalar) == bool_stash)
117	return (UV)-1;	369	return 1;
		370
		371	if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
		372	return 1;
118	}	373	}
		374
		375	return 0;
119	}	376	}
120		377
121	/////////////////////////////////////////////////////////////////////////////	378	/////////////////////////////////////////////////////////////////////////////
122	// encoder	379	// encoder
123		380
…		…
127	char *cur; // SvPVX (sv) + current output position	384	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	385	char *end; // SvEND (sv)
129	SV *sv; // result scalar	386	SV *sv; // result scalar
130	JSON json;	387	JSON json;
131	U32 indent; // indentation level	388	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	389	UV limit; // escape character values >= this value when encoding
133	} enc_t;	390	} enc_t;
134		391
135	inline void	392	INLINE void
136	need (enc_t *enc, STRLEN len)	393	need (enc_t *enc, STRLEN len)
137	{	394	{
138	if (expect_false (enc->cur + len >= enc->end))	395	if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
139	{	396	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	397	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	398	char *buf = json_sv_grow (enc->sv, cur, len);
142	enc->cur = SvPVX (enc->sv) + cur;	399	enc->cur = buf + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	400	enc->end = buf + SvLEN (enc->sv) - 1;
144	}	401	}
145	}	402	}
146		403
147	inline void	404	INLINE void
148	encode_ch (enc_t *enc, char ch)	405	encode_ch (enc_t *enc, char ch)
149	{	406	{
150	need (enc, 1);	407	need (enc, 1);
151	*enc->cur++ = ch;	408	*enc->cur++ = ch;
152	}	409	}
…		…
164		421
165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case	422	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
166	{	423	{
167	if (expect_false (ch == '"')) // but with slow exceptions	424	if (expect_false (ch == '"')) // but with slow exceptions
168	{	425	{
169	need (enc, len += 1);	426	need (enc, len + 1);
170	*enc->cur++ = '\\';	427	*enc->cur++ = '\\';
171	*enc->cur++ = '"';	428	*enc->cur++ = '"';
172	}	429	}
173	else if (expect_false (ch == '\\'))	430	else if (expect_false (ch == '\\'))
174	{	431	{
175	need (enc, len += 1);	432	need (enc, len + 1);
176	*enc->cur++ = '\\';	433	*enc->cur++ = '\\';
177	*enc->cur++ = '\\';	434	*enc->cur++ = '\\';
178	}	435	}
179	else	436	else
180	*enc->cur++ = ch;	437	*enc->cur++ = ch;
…		…
183	}	440	}
184	else	441	else
185	{	442	{
186	switch (ch)	443	switch (ch)
187	{	444	{
188	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;	445	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;	446	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;	447	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;	448	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;	449	case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
193		450
194	default:	451	default:
195	{	452	{
196	STRLEN clen;	453	STRLEN clen;
197	UV uch;	454	UV uch;
…		…
206	{	463	{
207	uch = ch;	464	uch = ch;
208	clen = 1;	465	clen = 1;
209	}	466	}
210		467
211	if (uch > 0x10FFFFUL)	468	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	{	469	{
216	if (uch > 0xFFFFUL)	470	if (uch >= 0x10000UL)
217	{	471	{
		472	if (uch >= 0x110000UL)
		473	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		474
218	need (enc, len += 11);	475	need (enc, len + 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	476	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	477	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	478	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	479	enc->cur += 12;
223	}	480	}
224	else	481	else
225	{	482	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	483	need (enc, len + 5);
228	*enc->cur++ = '\\';	484	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	485	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	486	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	487	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	488	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	489	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	490	}
235		491
236	str += clen;	492	str += clen;
237	}	493	}
238	else if (enc->json.flags & F_LATIN1)	494	else if (enc->json.flags & F_LATIN1)
…		…
240	*enc->cur++ = uch;	496	*enc->cur++ = uch;
241	str += clen;	497	str += clen;
242	}	498	}
243	else if (is_utf8)	499	else if (is_utf8)
244	{	500	{
245	need (enc, len += clen);	501	need (enc, len + clen);
246	do	502	do
247	{	503	{
248	*enc->cur++ = *str++;	504	*enc->cur++ = *str++;
249	}	505	}
250	while (--clen);	506	while (--clen);
251	}	507	}
252	else	508	else
253	{	509	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	510	need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	511	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	512	++str;
257	}	513	}
258	}	514	}
259	}	515	}
260	}	516	}
261		517
262	--len;	518	--len;
263	}	519	}
264	}	520	}
265		521
266	inline void	522	INLINE void
267	encode_indent (enc_t *enc)	523	encode_indent (enc_t *enc)
268	{	524	{
269	if (enc->json.flags & F_INDENT)	525	if (enc->json.flags & F_INDENT)
270	{	526	{
271	int spaces = enc->indent * INDENT_STEP;	527	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	530	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	531	enc->cur += spaces;
276	}	532	}
277	}	533	}
278		534
279	inline void	535	INLINE void
280	encode_space (enc_t *enc)	536	encode_space (enc_t *enc)
281	{	537	{
282	need (enc, 1);	538	need (enc, 1);
283	encode_ch (enc, ' ');	539	encode_ch (enc, ' ');
284	}	540	}
285		541
286	inline void	542	INLINE void
287	encode_nl (enc_t *enc)	543	encode_nl (enc_t *enc)
288	{	544	{
289	if (enc->json.flags & F_INDENT)	545	if (enc->json.flags & F_INDENT)
290	{	546	{
291	need (enc, 1);	547	need (enc, 1);
292	encode_ch (enc, '\n');	548	encode_ch (enc, '\n');
293	}	549	}
294	}	550	}
295		551
296	inline void	552	INLINE void
297	encode_comma (enc_t *enc)	553	encode_comma (enc_t *enc)
298	{	554	{
299	encode_ch (enc, ',');	555	encode_ch (enc, ',');
300		556
301	if (enc->json.flags & F_INDENT)	557	if (enc->json.flags & F_INDENT)
…		…
309	static void	565	static void
310	encode_av (enc_t *enc, AV *av)	566	encode_av (enc_t *enc, AV *av)
311	{	567	{
312	int i, len = av_len (av);	568	int i, len = av_len (av);
313		569
314	if (enc->indent >= enc->maxdepth)	570	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	571	croak (ERR_NESTING_EXCEEDED);
316		572
317	encode_ch (enc, '['); encode_nl (enc);	573	encode_ch (enc, '[');
318	++enc->indent;
319		574
		575	if (len >= 0)
		576	{
		577	encode_nl (enc); ++enc->indent;
		578
320	for (i = 0; i <= len; ++i)	579	for (i = 0; i <= len; ++i)
321	{	580	{
322	SV **svp = av_fetch (av, i, 0);	581	SV **svp = av_fetch (av, i, 0);
323		582
324	encode_indent (enc);	583	encode_indent (enc);
325		584
326	if (svp)	585	if (svp)
327	encode_sv (enc, *svp);	586	encode_sv (enc, *svp);
328	else	587	else
329	encode_str (enc, "null", 4, 0);	588	encode_str (enc, "null", 4, 0);
330		589
331	if (i < len)	590	if (i < len)
332	encode_comma (enc);	591	encode_comma (enc);
333	}	592	}
334		593
		594	encode_nl (enc); --enc->indent; encode_indent (enc);
		595	}
		596
335	encode_nl (enc);	597	encode_ch (enc, ']');
336
337	--enc->indent;
338	encode_indent (enc); encode_ch (enc, ']');
339	}	598	}
340		599
341	static void	600	static void
342	encode_hk (enc_t *enc, HE *he)	601	encode_hk (enc_t *enc, HE *he)
343	{	602	{
…		…
346	if (HeKLEN (he) == HEf_SVKEY)	605	if (HeKLEN (he) == HEf_SVKEY)
347	{	606	{
348	SV *sv = HeSVKEY (he);	607	SV *sv = HeSVKEY (he);
349	STRLEN len;	608	STRLEN len;
350	char *str;	609	char *str;
351		610
352	SvGETMAGIC (sv);	611	SvGETMAGIC (sv);
353	str = SvPV (sv, len);	612	str = SvPV (sv, len);
354		613
355	encode_str (enc, str, len, SvUTF8 (sv));	614	encode_str (enc, str, len, SvUTF8 (sv));
356	}	615	}
…		…
391		650
392	static void	651	static void
393	encode_hv (enc_t *enc, HV *hv)	652	encode_hv (enc_t *enc, HV *hv)
394	{	653	{
395	HE *he;	654	HE *he;
396	int count;
397		655
398	if (enc->indent >= enc->maxdepth)	656	if (enc->indent >= enc->json.max_depth)
399	croak ("data structure too deep (hit recursion limit)");	657	croak (ERR_NESTING_EXCEEDED);
400		658
401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	659	encode_ch (enc, '{');
402		660
403	// for canonical output we have to sort by keys first	661	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	662	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	663	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	664	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	665	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
408	{	666	{
409	int count = hv_iterinit (hv);	667	int count = hv_iterinit (hv);
410		668
411	if (SvMAGICAL (hv))	669	if (SvMAGICAL (hv))
412	{	670	{
…		…
423	}	681	}
424		682
425	if (count)	683	if (count)
426	{	684	{
427	int i, fast = 1;	685	int i, fast = 1;
428	#if defined(__BORLANDC__) || defined(_MSC_VER)	686	HE *hes_stack [STACK_HES];
429	HE **hes = _alloca (count * sizeof (HE));	687	HE **hes = hes_stack;
430	#else	688
431	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	689	// allocate larger arrays on the heap
432	#endif	690	if (count > STACK_HES)
		691	{
		692	SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
		693	hes = (HE **)SvPVX (sv);
		694	}
433		695
434	i = 0;	696	i = 0;
435	while ((he = hv_iternext (hv)))	697	while ((he = hv_iternext (hv)))
436	{	698	{
437	hes [i++] = he;	699	hes [i++] = he;
…		…
459		721
460	FREETMPS;	722	FREETMPS;
461	LEAVE;	723	LEAVE;
462	}	724	}
463		725
		726	encode_nl (enc); ++enc->indent;
		727
464	while (count--)	728	while (count--)
465	{	729	{
466	encode_indent (enc);	730	encode_indent (enc);
467	he = hes [count];	731	he = hes [count];
468	encode_hk (enc, he);	732	encode_hk (enc, he);
469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	733	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
470		734
471	if (count)	735	if (count)
472	encode_comma (enc);	736	encode_comma (enc);
473	}	737	}
		738
		739	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	740	}
475	}	741	}
476	else	742	else
477	{	743	{
478	if (hv_iterinit (hv) || SvMAGICAL (hv))	744	if (hv_iterinit (hv) || SvMAGICAL (hv))
479	if ((he = hv_iternext (hv)))	745	if ((he = hv_iternext (hv)))
		746	{
		747	encode_nl (enc); ++enc->indent;
		748
480	for (;;)	749	for (;;)
481	{	750	{
482	encode_indent (enc);	751	encode_indent (enc);
483	encode_hk (enc, he);	752	encode_hk (enc, he);
484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));	753	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
485		754
486	if (!(he = hv_iternext (hv)))	755	if (!(he = hv_iternext (hv)))
487	break;	756	break;
488		757
489	encode_comma (enc);	758	encode_comma (enc);
490	}	759	}
491	}
492		760
		761	encode_nl (enc); --enc->indent; encode_indent (enc);
		762	}
		763	}
		764
493	encode_nl (enc);	765	encode_ch (enc, '}');
494
495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
496	}	766	}
497		767
498	// encode objects, arrays and special \0=false and \1=true values.	768	// encode objects, arrays and special \0=false and \1=true values.
499	static void	769	static void
500	encode_rv (enc_t *enc, SV *sv)	770	encode_rv (enc_t *enc, SV *sv)
501	{	771	{
502	svtype svt;	772	svtype svt;
		773	GV *method;
503		774
504	SvGETMAGIC (sv);	775	SvGETMAGIC (sv);
505	svt = SvTYPE (sv);	776	svt = SvTYPE (sv);
506		777
507	if (expect_false (SvOBJECT (sv)))	778	if (expect_false (SvOBJECT (sv)))
508	{	779	{
509	HV *stash = !JSON_SLOW || json_boolean_stash	780	HV *stash = SvSTASH (sv);
510	? json_boolean_stash
511	: gv_stashpv ("JSON::XS::Boolean", 1);
512		781
513	if (SvSTASH (sv) == stash)	782	if (stash == bool_stash)
514	{	783	{
515	if (SvIV (sv))
516	encode_str (enc, "true", 4, 0);	784	if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
517	else
518	encode_str (enc, "false", 5, 0);	785	else encode_str (enc, "false", 5, 0);
519	}	786	}
		787	else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
		788	{
		789	int count;
		790	dSP;
		791
		792	ENTER; SAVETMPS;
		793	PUSHMARK (SP);
		794	EXTEND (SP, 2);
		795	// we re-bless the reference to get overload and other niceties right
		796	PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		797	PUSHs (sv_json);
		798
		799	PUTBACK;
		800	count = call_sv ((SV *)GvCV (method), G_ARRAY);
		801	SPAGAIN;
		802
		803	// catch this surprisingly common error
		804	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		805	croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		806
		807	encode_ch (enc, '(');
		808	encode_ch (enc, '"');
		809	encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
		810	encode_ch (enc, '"');
		811	encode_ch (enc, ')');
		812	encode_ch (enc, '[');
		813
		814	if (count)
		815	{
		816	int i;
		817
		818	for (i = 0; i < count - 1; ++i)
		819	{
		820	encode_sv (enc, SP[i + 1 - count]);
		821	encode_ch (enc, ',');
		822	}
		823
		824	encode_sv (enc, TOPs);
		825	SP -= count;
		826	}
		827
		828	encode_ch (enc, ']');
		829
		830	FREETMPS; LEAVE;
		831	}
		832	else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
		833	{
		834	dSP;
		835
		836	ENTER; SAVETMPS;
		837	PUSHMARK (SP);
		838	// we re-bless the reference to get overload and other niceties right
		839	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
		840
		841	// calling with G_SCALAR ensures that we always get a 1 return value
		842	PUTBACK;
		843	call_sv ((SV *)GvCV (method), G_SCALAR);
		844	SPAGAIN;
		845
		846	// catch this surprisingly common error
		847	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		848	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		849
		850	sv = POPs;
		851	PUTBACK;
		852
		853	encode_sv (enc, sv);
		854
		855	FREETMPS; LEAVE;
		856	}
		857	else if (enc->json.flags & F_ALLOW_BLESSED)
		858	encode_str (enc, "null", 4, 0);
520	else	859	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",	860	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))));	861	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
567	}
568	}	862	}
569	else if (svt == SVt_PVHV)	863	else if (svt == SVt_PVHV)
570	encode_hv (enc, (HV *)sv);	864	encode_hv (enc, (HV *)sv);
571	else if (svt == SVt_PVAV)	865	else if (svt == SVt_PVAV)
572	encode_av (enc, (AV *)sv);	866	encode_av (enc, (AV *)sv);
573	else if (svt < SVt_PVAV)	867	else if (svt < SVt_PVAV)
574	{	868	{
575	STRLEN len = 0;	869	int bool_type = ref_bool_type (sv);
576	char *pv = svt ? SvPV (sv, len) : 0;
577		870
578	if (len == 1 && *pv == '1')	871	if (bool_type == 1)
579	encode_str (enc, "true", 4, 0);	872	encode_str (enc, "true", 4, 0);
580	else if (len == 1 && *pv == '0')	873	else if (bool_type == 0)
581	encode_str (enc, "false", 5, 0);	874	encode_str (enc, "false", 5, 0);
		875	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		876	encode_str (enc, "null", 4, 0);
582	else	877	else
583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	878	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	879	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
585	}	880	}
		881	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		882	encode_str (enc, "null", 4, 0);
586	else	883	else
587	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	884	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
588	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	885	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
589	}	886	}
590		887
…		…
608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	905	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
609	enc->cur += strlen (enc->cur);	906	enc->cur += strlen (enc->cur);
610	}	907	}
611	else if (SvIOKp (sv))	908	else if (SvIOKp (sv))
612	{	909	{
613	// we assume we can always read an IV as a UV	910	// we assume we can always read an IV as a UV and vice versa
614	if (SvUV (sv) & ~(UV)0x7fff)	911	// we assume two's complement
615	{	912	// we assume no aliasing issues in the union
616	// large integer, use the (rather slow) snprintf way.	913	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
617	need (enc, sizeof (UV) * 3);	914	: 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	{	915	{
625	// optimise the "small number case"	916	// optimise the "small number case"
626	// code will likely be branchless and use only a single multiplication	917	// code will likely be branchless and use only a single multiplication
		918	// works for numbers up to 59074
627	I32 i = SvIV (sv);	919	I32 i = SvIVX (sv);
628	U32 u;	920	U32 u;
629	char digit, nz = 0;	921	char digit, nz = 0;
630		922
631	need (enc, 6);	923	need (enc, 6);
632		924
…		…
638		930
639	// now output digit by digit, each time masking out the integer part	931	// 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,	932	// and multiplying by 5 while moving the decimal point one to the right,
641	// resulting in a net multiplication by 10.	933	// resulting in a net multiplication by 10.
642	// we always write the digit to memory but conditionally increment	934	// we always write the digit to memory but conditionally increment
643	// the pointer, to ease the usage of conditional move instructions.	935	// 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;	936	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;	937	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;	938	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;	939	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'	940	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
649	}	941	}
		942	else
		943	{
		944	// large integer, use the (rather slow) snprintf way.
		945	need (enc, IVUV_MAXCHARS);
		946	enc->cur +=
		947	SvIsUV(sv)
		948	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		949	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		950	}
650	}	951	}
651	else if (SvROK (sv))	952	else if (SvROK (sv))
652	encode_rv (enc, SvRV (sv));	953	encode_rv (enc, SvRV (sv));
653	else if (!SvOK (sv))	954	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
654	encode_str (enc, "null", 4, 0);	955	encode_str (enc, "null", 4, 0);
655	else	956	else
656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	957	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
657	SvPV_nolen (sv), SvFLAGS (sv));	958	SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
658	}	959	}
659		960
660	static SV *	961	static SV *
661	encode_json (SV *scalar, JSON *json)	962	encode_json (SV *scalar, JSON *json)
662	{	963	{
663	enc_t enc;	964	enc_t enc;
664		965
665	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))	966	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)");	967	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
667		968
668	enc.json = *json;	969	enc.json = *json;
669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	970	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
670	enc.cur = SvPVX (enc.sv);	971	enc.cur = SvPVX (enc.sv);
671	enc.end = SvEND (enc.sv);	972	enc.end = SvEND (enc.sv);
672	enc.indent = 0;	973	enc.indent = 0;
673	enc.maxdepth = DEC_DEPTH (enc.json.flags);	974	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		975	: enc.json.flags & F_LATIN1 ? 0x000100UL
		976	: 0x110000UL;
674		977
675	SvPOK_only (enc.sv);	978	SvPOK_only (enc.sv);
676	encode_sv (&enc, scalar);	979	encode_sv (&enc, scalar);
		980	encode_nl (&enc);
677		981
678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	982	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	983	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
680		984
681	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))	985	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
…		…
699	JSON json;	1003	JSON json;
700	U32 depth; // recursion depth	1004	U32 depth; // recursion depth
701	U32 maxdepth; // recursion depth limit	1005	U32 maxdepth; // recursion depth limit
702	} dec_t;	1006	} dec_t;
703		1007
704	inline void	1008	INLINE void
		1009	decode_comment (dec_t *dec)
		1010	{
		1011	// only '#'-style comments allowed a.t.m.
		1012
		1013	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		1014	++dec->cur;
		1015	}
		1016
		1017	INLINE void
705	decode_ws (dec_t *dec)	1018	decode_ws (dec_t *dec)
706	{	1019	{
707	for (;;)	1020	for (;;)
708	{	1021	{
709	char ch = *dec->cur;	1022	char ch = *dec->cur;
710		1023
711	if (ch > 0x20	1024	if (ch > 0x20)
		1025	{
		1026	if (expect_false (ch == '#'))
		1027	{
		1028	if (dec->json.flags & F_RELAXED)
		1029	decode_comment (dec);
		1030	else
		1031	break;
		1032	}
		1033	else
		1034	break;
		1035	}
712	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	1036	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
713	break;	1037	break; // parse error, but let higher level handle it, gives better error messages
714
715	if (ch == '#' && dec->json.flags & F_RELAXED)
716	++dec->cur;
717		1038
718	++dec->cur;	1039	++dec->cur;
719	}	1040	}
720	}	1041	}
721		1042
…		…
725	if (*dec->cur != ch) \	1046	if (*dec->cur != ch) \
726	ERR (# ch " expected"); \	1047	ERR (# ch " expected"); \
727	++dec->cur; \	1048	++dec->cur; \
728	SE	1049	SE
729		1050
730	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	1051	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
731	#define DEC_DEC_DEPTH --dec->depth	1052	#define DEC_DEC_DEPTH --dec->depth
732		1053
733	static SV *decode_sv (dec_t *dec);	1054	static SV *decode_sv (dec_t *dec);
734		1055
735	static signed char decode_hexdigit[256];	1056	static signed char decode_hexdigit[256];
…		…
827		1148
828	if (hi >= 0x80)	1149	if (hi >= 0x80)
829	{	1150	{
830	utf8 = 1;	1151	utf8 = 1;
831		1152
832	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1153	cur = encode_utf8 (cur, hi);
833	}	1154	}
834	else	1155	else
835	*cur++ = hi;	1156	*cur++ = hi;
836	}	1157	}
837	break;	1158	break;
…		…
839	default:	1160	default:
840	--dec_cur;	1161	--dec_cur;
841	ERR ("illegal backslash escape sequence in string");	1162	ERR ("illegal backslash escape sequence in string");
842	}	1163	}
843	}	1164	}
844	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1165	else if (expect_true (ch >= 0x20 && ch < 0x80))
845	*cur++ = ch;	1166	*cur++ = ch;
846	else if (ch >= 0x80)	1167	else if (ch >= 0x80)
847	{	1168	{
848	STRLEN clen;	1169	STRLEN clen;
849	UV uch;
850		1170
851	--dec_cur;	1171	--dec_cur;
852		1172
853	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1173	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
854	if (clen == (STRLEN)-1)	1174	if (clen == (STRLEN)-1)
855	ERR ("malformed UTF-8 character in JSON string");	1175	ERR ("malformed UTF-8 character in JSON string");
856		1176
857	do	1177	do
858	*cur++ = *dec_cur++;	1178	*cur++ = *dec_cur++;
859	while (--clen);	1179	while (--clen);
860		1180
861	utf8 = 1;	1181	utf8 = 1;
862	}	1182	}
		1183	else if (ch == '\t' && dec->json.flags & F_RELAXED)
		1184	*cur++ = ch;
863	else	1185	else
864	{	1186	{
865	--dec_cur;	1187	--dec_cur;
866		1188
867	if (!ch)	1189	if (!ch)
…		…
875	{	1197	{
876	STRLEN len = cur - buf;	1198	STRLEN len = cur - buf;
877		1199
878	if (sv)	1200	if (sv)
879	{	1201	{
880	SvGROW (sv, SvCUR (sv) + len + 1);	1202	STRLEN cur = SvCUR (sv);
		1203
		1204	if (SvLEN (sv) - cur <= len)
		1205	json_sv_grow (sv, cur, len);
		1206
881	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1207	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
882	SvCUR_set (sv, SvCUR (sv) + len);	1208	SvCUR_set (sv, SvCUR (sv) + len);
883	}	1209	}
884	else	1210	else
885	sv = newSVpvn (buf, len);	1211	sv = newSVpvn (buf, len);
…		…
972		1298
973	if (!is_nv)	1299	if (!is_nv)
974	{	1300	{
975	int len = dec->cur - start;	1301	int len = dec->cur - start;
976		1302
977	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1303	// special case the rather common 1..5-digit-int case
978	if (*start == '-')	1304	if (*start == '-')
979	switch (len)	1305	switch (len)
980	{	1306	{
981	case 2: return newSViv (-( start [1] - '0' * 1));	1307	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
982	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1308	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
983	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1309	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
984	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1310	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1311	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
985	}	1312	}
986	else	1313	else
987	switch (len)	1314	switch (len)
988	{	1315	{
989	case 1: return newSViv ( start [0] - '0' * 1);	1316	case 1: return newSViv ( start [0] - '0' * 1);
990	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1317	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
991	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1318	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
992	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1319	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1320	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
993	}	1321	}
994		1322
995	{	1323	{
996	UV uv;	1324	UV uv;
997	int numtype = grok_number (start, len, &uv);	1325	int numtype = grok_number (start, len, &uv);
…		…
1006	}	1334	}
1007		1335
1008	len -= *start == '-' ? 1 : 0;	1336	len -= *start == '-' ? 1 : 0;
1009		1337
1010	// does not fit into IV or UV, try NV	1338	// does not fit into IV or UV, try NV
1011	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1339	if (len <= NV_DIG)
1012	#if defined (LDBL_DIG)
1013	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1014	#endif
1015	)
1016	// fits into NV without loss of precision	1340	// fits into NV without loss of precision
1017	return newSVnv (Atof (start));	1341	return newSVnv (json_atof (start));
1018		1342
1019	// everything else fails, convert it to a string	1343	// everything else fails, convert it to a string
1020	return newSVpvn (start, dec->cur - start);	1344	return newSVpvn (start, dec->cur - start);
1021	}	1345	}
1022		1346
1023	// loss of precision here	1347	// loss of precision here
1024	return newSVnv (Atof (start));	1348	return newSVnv (json_atof (start));
1025		1349
1026	fail:	1350	fail:
1027	return 0;	1351	return 0;
1028	}	1352	}
1029		1353
…		…
1053	if (*dec->cur == ']')	1377	if (*dec->cur == ']')
1054	{	1378	{
1055	++dec->cur;	1379	++dec->cur;
1056	break;	1380	break;
1057	}	1381	}
1058		1382
1059	if (*dec->cur != ',')	1383	if (*dec->cur != ',')
1060	ERR (", or ] expected while parsing array");	1384	ERR (", or ] expected while parsing array");
1061		1385
1062	++dec->cur;	1386	++dec->cur;
1063		1387
…		…
1105	char *p = dec->cur;	1429	char *p = dec->cur;
1106	char *e = p + 24; // only try up to 24 bytes	1430	char *e = p + 24; // only try up to 24 bytes
1107		1431
1108	for (;;)	1432	for (;;)
1109	{	1433	{
1110	// the >= 0x80 is true on most architectures	1434	// the >= 0x80 is false on most architectures
1111	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')	1435	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
1112	{	1436	{
1113	// slow path, back up and use decode_str	1437	// slow path, back up and use decode_str
1114	SV *key = decode_str (dec);	1438	SV *key = decode_str (dec);
1115	if (!key)	1439	if (!key)
…		…
1187		1511
1188	hv_iterinit (hv);	1512	hv_iterinit (hv);
1189	he = hv_iternext (hv);	1513	he = hv_iternext (hv);
1190	hv_iterinit (hv);	1514	hv_iterinit (hv);
1191		1515
1192	// the next line creates a mortal sv each time its called.	1516	// the next line creates a mortal sv each time it's called.
1193	// might want to optimise this for common cases.	1517	// might want to optimise this for common cases.
1194	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);	1518	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1195		1519
1196	if (cb)	1520	if (cb)
1197	{	1521	{
1198	dSP;	1522	dSP;
1199	int count;	1523	int count;
1200		1524
1201	ENTER; SAVETMPS; PUSHMARK (SP);	1525	ENTER; SAVETMPS;
		1526	PUSHMARK (SP);
1202	XPUSHs (HeVAL (he));	1527	XPUSHs (HeVAL (he));
		1528	sv_2mortal (sv);
1203		1529
1204	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1530	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1205		1531
1206	if (count == 1)	1532	if (count == 1)
1207	{	1533	{
1208	sv = newSVsv (POPs);	1534	sv = newSVsv (POPs);
1209	FREETMPS; LEAVE;	1535	FREETMPS; LEAVE;
1210	return sv;	1536	return sv;
1211	}	1537	}
		1538	else if (count)
		1539	croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1212		1540
		1541	SvREFCNT_inc (sv);
1213	FREETMPS; LEAVE;	1542	FREETMPS; LEAVE;
1214	}	1543	}
1215	}	1544	}
1216		1545
1217	if (dec->json.cb_object)	1546	if (dec->json.cb_object)
1218	{	1547	{
1219	dSP;	1548	dSP;
1220	int count;	1549	int count;
1221		1550
1222	ENTER; SAVETMPS; PUSHMARK (SP);	1551	ENTER; SAVETMPS;
		1552	PUSHMARK (SP);
1223	XPUSHs (sv_2mortal (sv));	1553	XPUSHs (sv_2mortal (sv));
1224		1554
1225	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1555	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1226		1556
1227	if (count == 1)	1557	if (count == 1)
1228	{	1558	{
1229	sv = newSVsv (POPs);	1559	sv = newSVsv (POPs);
1230	FREETMPS; LEAVE;	1560	FREETMPS; LEAVE;
1231	return sv;	1561	return sv;
1232	}	1562	}
		1563	else if (count)
		1564	croak ("filter_json_object callbacks must not return more than one scalar");
1233		1565
1234	SvREFCNT_inc (sv);	1566	SvREFCNT_inc (sv);
1235	FREETMPS; LEAVE;	1567	FREETMPS; LEAVE;
1236	}	1568	}
1237	}	1569	}
…		…
1243	DEC_DEC_DEPTH;	1575	DEC_DEC_DEPTH;
1244	return 0;	1576	return 0;
1245	}	1577	}
1246		1578
1247	static SV *	1579	static SV *
		1580	decode_tag (dec_t *dec)
		1581	{
		1582	SV *tag = 0;
		1583	SV *val = 0;
		1584
		1585	if (!(dec->json.flags & F_ALLOW_TAGS))
		1586	ERR ("malformed JSON string, neither array, object, number, string or atom");
		1587
		1588	++dec->cur;
		1589
		1590	decode_ws (dec);
		1591
		1592	tag = decode_sv (dec);
		1593	if (!tag)
		1594	goto fail;
		1595
		1596	if (!SvPOK (tag))
		1597	ERR ("malformed JSON string, (tag) must be a string");
		1598
		1599	decode_ws (dec);
		1600
		1601	if (*dec->cur != ')')
		1602	ERR (") expected after tag");
		1603
		1604	++dec->cur;
		1605
		1606	decode_ws (dec);
		1607
		1608	val = decode_sv (dec);
		1609	if (!val)
		1610	goto fail;
		1611
		1612	if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
		1613	ERR ("malformed JSON string, tag value must be an array");
		1614
		1615	{
		1616	AV *av = (AV *)SvRV (val);
		1617	int i, len = av_len (av) + 1;
		1618	HV *stash = gv_stashsv (tag, 0);
		1619	SV *sv;
		1620
		1621	if (!stash)
		1622	ERR ("cannot decode perl-object (package does not exist)");
		1623
		1624	GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
		1625
		1626	if (!method)
		1627	ERR ("cannot decode perl-object (package does not have a THAW method)");
		1628
		1629	dSP;
		1630
		1631	ENTER; SAVETMPS;
		1632	PUSHMARK (SP);
		1633	EXTEND (SP, len + 2);
		1634	// we re-bless the reference to get overload and other niceties right
		1635	PUSHs (tag);
		1636	PUSHs (sv_json);
		1637
		1638	for (i = 0; i < len; ++i)
		1639	PUSHs (*av_fetch (av, i, 1));
		1640
		1641	PUTBACK;
		1642	call_sv ((SV *)GvCV (method), G_SCALAR);
		1643	SPAGAIN;
		1644
		1645	SvREFCNT_dec (tag);
		1646	SvREFCNT_dec (val);
		1647	sv = SvREFCNT_inc (POPs);
		1648
		1649	PUTBACK;
		1650
		1651	FREETMPS; LEAVE;
		1652
		1653	return sv;
		1654	}
		1655
		1656	fail:
		1657	SvREFCNT_dec (tag);
		1658	SvREFCNT_dec (val);
		1659	return 0;
		1660	}
		1661
		1662	static SV *
1248	decode_sv (dec_t *dec)	1663	decode_sv (dec_t *dec)
1249	{	1664	{
1250	// the beauty of JSON: you need exactly one character lookahead	1665	// the beauty of JSON: you need exactly one character lookahead
1251	// to parse anything.	1666	// to parse everything.
1252	switch (*dec->cur)	1667	switch (*dec->cur)
1253	{	1668	{
1254	case '"': ++dec->cur; return decode_str (dec);	1669	case '"': ++dec->cur; return decode_str (dec);
1255	case '[': ++dec->cur; return decode_av (dec);	1670	case '[': ++dec->cur; return decode_av (dec);
1256	case '{': ++dec->cur; return decode_hv (dec);	1671	case '{': ++dec->cur; return decode_hv (dec);
		1672	case '(': return decode_tag (dec);
1257		1673
1258	case '-':	1674	case '-':
1259	case '0': case '1': case '2': case '3': case '4':	1675	case '0': case '1': case '2': case '3': case '4':
1260	case '5': case '6': case '7': case '8': case '9':	1676	case '5': case '6': case '7': case '8': case '9':
1261	return decode_num (dec);	1677	return decode_num (dec);
1262		1678
		1679	case 'f':
		1680	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
		1681	{
		1682	dec->cur += 5;
		1683
		1684	if (expect_false (!dec->json.v_false))
		1685	dec->json.v_false = GET_BOOL (false);
		1686
		1687	return newSVsv (dec->json.v_false);
		1688	}
		1689	else
		1690	ERR ("'false' expected");
		1691
		1692	break;
		1693
1263	case 't':	1694	case 't':
1264	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1695	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1265	{	1696	{
1266	dec->cur += 4;	1697	dec->cur += 4;
1267	#if JSON_SLOW	1698
1268	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1699	if (expect_false (!dec->json.v_true))
1269	#endif	1700	dec->json.v_true = GET_BOOL (true);
1270	return SvREFCNT_inc (json_true);	1701
		1702	return newSVsv (dec->json.v_true);
1271	}	1703	}
1272	else	1704	else
1273	ERR ("'true' expected");	1705	ERR ("'true' expected");
1274
1275	break;
1276
1277	case 'f':
1278	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1279	{
1280	dec->cur += 5;
1281	#if JSON_SLOW
1282	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1283	#endif
1284	return SvREFCNT_inc (json_false);
1285	}
1286	else
1287	ERR ("'false' expected");
1288		1706
1289	break;	1707	break;
1290		1708
1291	case 'n':	1709	case 'n':
1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1710	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
…		…
1298	ERR ("'null' expected");	1716	ERR ("'null' expected");
1299		1717
1300	break;	1718	break;
1301		1719
1302	default:	1720	default:
1303	ERR ("malformed JSON string, neither array, object, number, string or atom");	1721	ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1304	break;	1722	break;
1305	}	1723	}
1306		1724
1307	fail:	1725	fail:
1308	return 0;	1726	return 0;
1309	}	1727	}
1310		1728
1311	static SV *	1729	static SV *
1312	decode_json (SV *string, JSON *json, UV *offset_return)	1730	decode_json (SV *string, JSON *json, STRLEN *offset_return)
1313	{	1731	{
1314	dec_t dec;	1732	dec_t dec;
1315	UV offset;
1316	SV *sv;	1733	SV *sv;
1317		1734
		1735	/* work around bugs in 5.10 where manipulating magic values
		1736	* makes perl ignore the magic in subsequent accesses.
		1737	* also make a copy of non-PV values, to get them into a clean
		1738	* state (SvPV should do that, but it's buggy, see below).
		1739	*
		1740	* SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
		1741	* as reported by Reini Urban.
		1742	*/
1318	SvGETMAGIC (string);	1743	/*SvGETMAGIC (string);*/
		1744	if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
		1745	string = sv_2mortal (newSVsv (string));
		1746
1319	SvUPGRADE (string, SVt_PV);	1747	SvUPGRADE (string, SVt_PV);
1320		1748
1321	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1749	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1750	* assertion with -DDEBUGGING, although SvCUR is documented to
		1751	* return the xpv_cur field which certainly exists after upgrading.
		1752	* according to nicholas clark, calling SvPOK fixes this.
		1753	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1754	* and hope for the best.
		1755	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1756	* assertion business is seriously broken, try yet another workaround
		1757	* for the broken -DDEBUGGING.
		1758	*/
		1759	{
		1760	#ifdef DEBUGGING
		1761	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1762	#else
		1763	STRLEN offset = SvCUR (string);
		1764	#endif
		1765
		1766	if (offset > json->max_size && json->max_size)
1322	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1767	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1323	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1768	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1769	}
1324		1770
1325	if (json->flags & F_UTF8)	1771	if (DECODE_WANTS_OCTETS (json))
1326	sv_utf8_downgrade (string, 0);	1772	sv_utf8_downgrade (string, 0);
1327	else	1773	else
1328	sv_utf8_upgrade (string);	1774	sv_utf8_upgrade (string);
1329		1775
1330	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1776	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1331		1777
1332	dec.json = *json;	1778	dec.json = *json;
1333	dec.cur = SvPVX (string);	1779	dec.cur = SvPVX (string);
1334	dec.end = SvEND (string);	1780	dec.end = SvEND (string);
1335	dec.err = 0;	1781	dec.err = 0;
1336	dec.depth = 0;	1782	dec.depth = 0;
1337	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1338		1783
1339	if (dec.json.cb_object || dec.json.cb_sk_object)	1784	if (dec.json.cb_object || dec.json.cb_sk_object)
1340	dec.json.flags |= F_HOOK;	1785	dec.json.flags |= F_HOOK;
1341		1786
1342	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1787	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1343		1788
1344	decode_ws (&dec);	1789	decode_ws (&dec);
1345	sv = decode_sv (&dec);	1790	sv = decode_sv (&dec);
1346		1791
1347	if (!(offset_return || !sv))	1792	if (offset_return)
		1793	*offset_return = dec.cur - SvPVX (string);
		1794	else if (sv)
1348	{	1795	{
1349	// check for trailing garbage	1796	// check for trailing garbage
1350	decode_ws (&dec);	1797	decode_ws (&dec);
1351		1798
1352	if (*dec.cur)	1799	if (*dec.cur)
1353	{	1800	{
1354	dec.err = "garbage after JSON object";	1801	dec.err = "garbage after JSON object";
1355	SvREFCNT_dec (sv);	1802	SvREFCNT_dec (sv);
1356	sv = 0;	1803	sv = 0;
1357	}	1804	}
1358	}
1359
1360	if (offset_return || !sv)
1361	{
1362	offset = dec.json.flags & F_UTF8
1363	? dec.cur - SvPVX (string)
1364	: utf8_distance (dec.cur, SvPVX (string));
1365
1366	if (offset_return)
1367	*offset_return = offset;
1368	}	1805	}
1369		1806
1370	if (!sv)	1807	if (!sv)
1371	{	1808	{
1372	SV *uni = sv_newmortal ();	1809	SV *uni = sv_newmortal ();
…		…
1378	SAVEVPTR (PL_curcop);	1815	SAVEVPTR (PL_curcop);
1379	PL_curcop = &cop;	1816	PL_curcop = &cop;
1380	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1817	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1381	LEAVE;	1818	LEAVE;
1382		1819
1383	croak ("%s, at character offset %d [\"%s\"]",	1820	croak ("%s, at character offset %d (before \"%s\")",
1384	dec.err,	1821	dec.err,
1385	(int)offset,	1822	(int)ptr_to_index (string, dec.cur),
1386	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1823	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1387	}	1824	}
1388		1825
1389	sv = sv_2mortal (sv);	1826	sv = sv_2mortal (sv);
1390		1827
1391	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1828	if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1392	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1829	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1393		1830
1394	return sv;	1831	return sv;
		1832	}
		1833
		1834	/////////////////////////////////////////////////////////////////////////////
		1835	// incremental parser
		1836
		1837	static void
		1838	incr_parse (JSON *self)
		1839	{
		1840	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1841
		1842	// the state machine here is a bit convoluted and could be simplified a lot
		1843	// but this would make it slower, so...
		1844
		1845	for (;;)
		1846	{
		1847	//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
		1848	switch (self->incr_mode)
		1849	{
		1850	// only used for initial whitespace skipping
		1851	case INCR_M_WS:
		1852	for (;;)
		1853	{
		1854	if (*p > 0x20)
		1855	{
		1856	if (*p == '#')
		1857	{
		1858	self->incr_mode = INCR_M_C0;
		1859	goto incr_m_c;
		1860	}
		1861	else
		1862	{
		1863	self->incr_mode = INCR_M_JSON;
		1864	goto incr_m_json;
		1865	}
		1866	}
		1867	else if (!*p)
		1868	goto interrupt;
		1869
		1870	++p;
		1871	}
		1872
		1873	// skip a single char inside a string (for \\-processing)
		1874	case INCR_M_BS:
		1875	if (!*p)
		1876	goto interrupt;
		1877
		1878	++p;
		1879	self->incr_mode = INCR_M_STR;
		1880	goto incr_m_str;
		1881
		1882	// inside #-style comments
		1883	case INCR_M_C0:
		1884	case INCR_M_C1:
		1885	incr_m_c:
		1886	for (;;)
		1887	{
		1888	if (*p == '\n')
		1889	{
		1890	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1891	break;
		1892	}
		1893	else if (!*p)
		1894	goto interrupt;
		1895
		1896	++p;
		1897	}
		1898
		1899	break;
		1900
		1901	// inside a string
		1902	case INCR_M_STR:
		1903	incr_m_str:
		1904	for (;;)
		1905	{
		1906	if (*p == '"')
		1907	{
		1908	++p;
		1909	self->incr_mode = INCR_M_JSON;
		1910
		1911	if (!self->incr_nest)
		1912	goto interrupt;
		1913
		1914	goto incr_m_json;
		1915	}
		1916	else if (*p == '\\')
		1917	{
		1918	++p; // "virtually" consumes character after \
		1919
		1920	if (!*p) // if at end of string we have to switch modes
		1921	{
		1922	self->incr_mode = INCR_M_BS;
		1923	goto interrupt;
		1924	}
		1925	}
		1926	else if (!*p)
		1927	goto interrupt;
		1928
		1929	++p;
		1930	}
		1931
		1932	// after initial ws, outside string
		1933	case INCR_M_JSON:
		1934	incr_m_json:
		1935	for (;;)
		1936	{
		1937	switch (*p++)
		1938	{
		1939	case 0:
		1940	--p;
		1941	goto interrupt;
		1942
		1943	case 0x09:
		1944	case 0x0a:
		1945	case 0x0d:
		1946	case 0x20:
		1947	if (!self->incr_nest)
		1948	{
		1949	--p; // do not eat the whitespace, let the next round do it
		1950	goto interrupt;
		1951	}
		1952	break;
		1953
		1954	case '"':
		1955	self->incr_mode = INCR_M_STR;
		1956	goto incr_m_str;
		1957
		1958	case '[':
		1959	case '{':
		1960	case '(':
		1961	if (++self->incr_nest > self->max_depth)
		1962	croak (ERR_NESTING_EXCEEDED);
		1963	break;
		1964
		1965	case ']':
		1966	case '}':
		1967	if (--self->incr_nest <= 0)
		1968	goto interrupt;
		1969	break;
		1970
		1971	case ')':
		1972	--self->incr_nest;
		1973	break;
		1974
		1975	case '#':
		1976	self->incr_mode = INCR_M_C1;
		1977	goto incr_m_c;
		1978	}
		1979	}
		1980	}
		1981
		1982	modechange:
		1983	;
		1984	}
		1985
		1986	interrupt:
		1987	self->incr_pos = p - SvPVX (self->incr_text);
		1988	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1989	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1395	}	1990	}
1396		1991
1397	/////////////////////////////////////////////////////////////////////////////	1992	/////////////////////////////////////////////////////////////////////////////
1398	// XS interface functions	1993	// XS interface functions
1399		1994
…		…
1408	i >= '0' && i <= '9' ? i - '0'	2003	i >= '0' && i <= '9' ? i - '0'
1409	: i >= 'a' && i <= 'f' ? i - 'a' + 10	2004	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1410	: i >= 'A' && i <= 'F' ? i - 'A' + 10	2005	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1411	: -1;	2006	: -1;
1412		2007
1413	json_stash = gv_stashpv ("JSON::XS" , 1);	2008	json_stash = gv_stashpv ("JSON::XS" , 1);
1414	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	2009	bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
		2010	bool_false = get_bool ("Types::Serialiser::false");
		2011	bool_true = get_bool ("Types::Serialiser::true");
1415		2012
1416	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	2013	sv_json = newSVpv ("JSON", 0);
1417	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	2014	SvREADONLY_on (sv_json);
		2015
		2016	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1418	}	2017	}
1419		2018
1420	PROTOTYPES: DISABLE	2019	PROTOTYPES: DISABLE
1421		2020
1422	void CLONE (...)	2021	void CLONE (...)
1423	CODE:	2022	CODE:
		2023	// as long as these writes are atomic, the race should not matter
		2024	// as existing threads either already use 0, or use the old value,
		2025	// which is sitll correct for the initial thread.
1424	json_stash = 0;	2026	json_stash = 0;
1425	json_boolean_stash = 0;	2027	bool_stash = 0;
		2028	bool_false = 0;
		2029	bool_true = 0;
1426		2030
1427	void new (char *klass)	2031	void new (char *klass)
1428	PPCODE:	2032	PPCODE:
1429	{	2033	{
1430	SV *pv = NEWSV (0, sizeof (JSON));	2034	SV *pv = NEWSV (0, sizeof (JSON));
1431	SvPOK_only (pv);	2035	SvPOK_only (pv);
1432	Zero (SvPVX (pv), 1, JSON);	2036	json_init ((JSON *)SvPVX (pv));
1433	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	2037	XPUSHs (sv_2mortal (sv_bless (
1434	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	2038	newRV_noinc (pv),
		2039	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		2040	)));
1435	}	2041	}
		2042
		2043	void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
		2044	PPCODE:
		2045	self->v_false = newSVsv (v_false);
		2046	self->v_true = newSVsv (v_true);
		2047	XPUSHs (ST (0));
		2048
		2049	void get_boolean_values (JSON *self)
		2050	PPCODE:
		2051	if (self->v_false && self->v_true)
		2052	{
		2053	EXTEND (SP, 2);
		2054	PUSHs (self->v_false);
		2055	PUSHs (self->v_true);
		2056	}
1436		2057
1437	void ascii (JSON *self, int enable = 1)	2058	void ascii (JSON *self, int enable = 1)
1438	ALIAS:	2059	ALIAS:
1439	ascii = F_ASCII	2060	ascii = F_ASCII
1440	latin1 = F_LATIN1	2061	latin1 = F_LATIN1
…		…
1447	allow_nonref = F_ALLOW_NONREF	2068	allow_nonref = F_ALLOW_NONREF
1448	shrink = F_SHRINK	2069	shrink = F_SHRINK
1449	allow_blessed = F_ALLOW_BLESSED	2070	allow_blessed = F_ALLOW_BLESSED
1450	convert_blessed = F_CONV_BLESSED	2071	convert_blessed = F_CONV_BLESSED
1451	relaxed = F_RELAXED	2072	relaxed = F_RELAXED
		2073	allow_unknown = F_ALLOW_UNKNOWN
		2074	allow_tags = F_ALLOW_TAGS
1452	PPCODE:	2075	PPCODE:
1453	{	2076	{
1454	if (enable)	2077	if (enable)
1455	self->flags |= ix;	2078	self->flags |= ix;
1456	else	2079	else
1457	self->flags &= ~ix;	2080	self->flags &= ~ix;
1458		2081
1459	XPUSHs (ST (0));	2082	XPUSHs (ST (0));
1460	}	2083	}
1461		2084
1462	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	2085	void get_ascii (JSON *self)
		2086	ALIAS:
		2087	get_ascii = F_ASCII
		2088	get_latin1 = F_LATIN1
		2089	get_utf8 = F_UTF8
		2090	get_indent = F_INDENT
		2091	get_canonical = F_CANONICAL
		2092	get_space_before = F_SPACE_BEFORE
		2093	get_space_after = F_SPACE_AFTER
		2094	get_allow_nonref = F_ALLOW_NONREF
		2095	get_shrink = F_SHRINK
		2096	get_allow_blessed = F_ALLOW_BLESSED
		2097	get_convert_blessed = F_CONV_BLESSED
		2098	get_relaxed = F_RELAXED
		2099	get_allow_unknown = F_ALLOW_UNKNOWN
		2100	get_allow_tags = F_ALLOW_TAGS
1463	PPCODE:	2101	PPCODE:
1464	{	2102	XPUSHs (boolSV (self->flags & ix));
1465	UV log2 = 0;
1466		2103
1467	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	2104	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1468		2105	PPCODE:
1469	while ((1UL << log2) < max_depth)	2106	self->max_depth = max_depth;
1470	++log2;
1471
1472	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1473
1474	XPUSHs (ST (0));	2107	XPUSHs (ST (0));
1475	}
1476		2108
		2109	U32 get_max_depth (JSON *self)
		2110	CODE:
		2111	RETVAL = self->max_depth;
		2112	OUTPUT:
		2113	RETVAL
		2114
1477	void max_size (JSON *self, UV max_size = 0)	2115	void max_size (JSON *self, U32 max_size = 0)
1478	PPCODE:	2116	PPCODE:
1479	{	2117	self->max_size = max_size;
1480	UV log2 = 0;
1481
1482	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1483	if (max_size == 1) max_size = 2;
1484
1485	while ((1UL << log2) < max_size)
1486	++log2;
1487
1488	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1489
1490	XPUSHs (ST (0));	2118	XPUSHs (ST (0));
1491	}	2119
		2120	int get_max_size (JSON *self)
		2121	CODE:
		2122	RETVAL = self->max_size;
		2123	OUTPUT:
		2124	RETVAL
1492		2125
1493	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)	2126	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1494	PPCODE:	2127	PPCODE:
1495	{	2128	{
1496	SvREFCNT_dec (self->cb_object);	2129	SvREFCNT_dec (self->cb_object);
…		…
1500	}	2133	}
1501		2134
1502	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)	2135	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1503	PPCODE:	2136	PPCODE:
1504	{	2137	{
1505	if (!self->cb_sk_object)	2138	if (!self->cb_sk_object)
1506	self->cb_sk_object = newHV ();	2139	self->cb_sk_object = newHV ();
1507		2140
1508	if (SvOK (cb))	2141	if (SvOK (cb))
1509	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);	2142	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1510	else	2143	else
…		…
1521	XPUSHs (ST (0));	2154	XPUSHs (ST (0));
1522	}	2155	}
1523		2156
1524	void encode (JSON *self, SV *scalar)	2157	void encode (JSON *self, SV *scalar)
1525	PPCODE:	2158	PPCODE:
1526	XPUSHs (encode_json (scalar, self));	2159	PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
		2160	XPUSHs (scalar);
1527		2161
1528	void decode (JSON *self, SV *jsonstr)	2162	void decode (JSON *self, SV *jsonstr)
1529	PPCODE:	2163	PPCODE:
1530	XPUSHs (decode_json (jsonstr, self, 0));	2164	PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
		2165	XPUSHs (jsonstr);
1531		2166
1532	void decode_prefix (JSON *self, SV *jsonstr)	2167	void decode_prefix (JSON *self, SV *jsonstr)
1533	PPCODE:	2168	PPCODE:
1534	{	2169	{
1535	UV offset;	2170	SV *sv;
		2171	STRLEN offset;
		2172	PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1536	EXTEND (SP, 2);	2173	EXTEND (SP, 2);
1537	PUSHs (decode_json (jsonstr, self, &offset));	2174	PUSHs (sv);
1538	PUSHs (sv_2mortal (newSVuv (offset)));	2175	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
		2176	}
		2177
		2178	void incr_parse (JSON *self, SV *jsonstr = 0)
		2179	PPCODE:
		2180	{
		2181	if (!self->incr_text)
		2182	self->incr_text = newSVpvn ("", 0);
		2183
		2184	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		2185	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		2186	if (DECODE_WANTS_OCTETS (self))
		2187	{
		2188	if (self->incr_pos)
		2189	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		2190	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		2191
		2192	sv_utf8_downgrade (self->incr_text, 0);
		2193	}
		2194	else
		2195	{
		2196	sv_utf8_upgrade (self->incr_text);
		2197
		2198	if (self->incr_pos)
		2199	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		2200	- (U8 *)SvPVX (self->incr_text);
		2201	}
		2202
		2203	// append data, if any
		2204	if (jsonstr)
		2205	{
		2206	/* make sure both strings have same encoding */
		2207	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		2208	if (SvUTF8 (jsonstr))
		2209	sv_utf8_downgrade (jsonstr, 0);
		2210	else
		2211	sv_utf8_upgrade (jsonstr);
		2212
		2213	/* and then just blindly append */
		2214	{
		2215	STRLEN len;
		2216	const char *str = SvPV (jsonstr, len);
		2217	STRLEN cur = SvCUR (self->incr_text);
		2218
		2219	if (SvLEN (self->incr_text) - cur <= len)
		2220	json_sv_grow (self->incr_text, cur, len);
		2221
		2222	Move (str, SvEND (self->incr_text), len, char);
		2223	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		2224	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		2225	}
		2226	}
		2227
		2228	if (GIMME_V != G_VOID)
		2229	do
		2230	{
		2231	SV *sv;
		2232	STRLEN offset;
		2233
		2234	if (!INCR_DONE (self))
		2235	{
		2236	incr_parse (self);
		2237
		2238	if (self->incr_pos > self->max_size && self->max_size)
		2239	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		2240	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		2241
		2242	if (!INCR_DONE (self))
		2243	{
		2244	// as an optimisation, do not accumulate white space in the incr buffer
		2245	if (self->incr_mode == INCR_M_WS && self->incr_pos)
		2246	{
		2247	self->incr_pos = 0;
		2248	SvCUR_set (self->incr_text, 0);
		2249	}
		2250
		2251	break;
		2252	}
		2253	}
		2254
		2255	PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
		2256	XPUSHs (sv);
		2257
		2258	self->incr_pos -= offset;
		2259	self->incr_nest = 0;
		2260	self->incr_mode = 0;
		2261
		2262	sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
		2263	}
		2264	while (GIMME_V == G_ARRAY);
		2265	}
		2266
		2267	SV *incr_text (JSON *self)
		2268	ATTRS: lvalue
		2269	CODE:
		2270	{
		2271	if (self->incr_pos)
		2272	croak ("incr_text can not be called when the incremental parser already started parsing");
		2273
		2274	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2275	}
		2276	OUTPUT:
		2277	RETVAL
		2278
		2279	void incr_skip (JSON *self)
		2280	CODE:
		2281	{
		2282	if (self->incr_pos)
		2283	{
		2284	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2285	self->incr_pos = 0;
		2286	self->incr_nest = 0;
		2287	self->incr_mode = 0;
		2288	}
		2289	}
		2290
		2291	void incr_reset (JSON *self)
		2292	CODE:
		2293	{
		2294	SvREFCNT_dec (self->incr_text);
		2295	self->incr_text = 0;
		2296	self->incr_pos = 0;
		2297	self->incr_nest = 0;
		2298	self->incr_mode = 0;
1539	}	2299	}
1540		2300
1541	void DESTROY (JSON *self)	2301	void DESTROY (JSON *self)
1542	CODE:	2302	CODE:
		2303	SvREFCNT_dec (self->v_false);
		2304	SvREFCNT_dec (self->v_true);
1543	SvREFCNT_dec (self->cb_sk_object);	2305	SvREFCNT_dec (self->cb_sk_object);
1544	SvREFCNT_dec (self->cb_object);	2306	SvREFCNT_dec (self->cb_object);
		2307	SvREFCNT_dec (self->incr_text);
1545		2308
1546	PROTOTYPES: ENABLE	2309	PROTOTYPES: ENABLE
1547		2310
1548	void to_json (SV *scalar)	2311	void encode_json (SV *scalar)
1549	PPCODE:	2312	PPCODE:
1550	{	2313	{
1551	JSON json = { F_DEFAULT | F_UTF8 };	2314	JSON json;
1552	XPUSHs (encode_json (scalar, &json));	2315	json_init (&json);
		2316	json.flags |= F_UTF8;
		2317	PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
		2318	XPUSHs (scalar);
1553	}	2319	}
1554		2320
1555	void from_json (SV *jsonstr)	2321	void decode_json (SV *jsonstr)
1556	PPCODE:	2322	PPCODE:
1557	{	2323	{
1558	JSON json = { F_DEFAULT | F_UTF8 };	2324	JSON json;
1559	XPUSHs (decode_json (jsonstr, &json, 0));	2325	json_init (&json);
		2326	json.flags |= F_UTF8;
		2327	PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
		2328	XPUSHs (jsonstr);
1560	}	2329	}
1561		2330

Diff Legend

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