ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.2 by root, Thu Mar 22 17:28:50 2007 UTC vs.
Revision 1.85 by root, Sat Apr 5 18:15:46 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
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>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040	31	#define F_ALLOW_NONREF 0x00000100UL
		32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36	#define F_ALLOW_UNKNOWN 0x00002000UL
		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
16		38
17	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
18	#define F_DEFAULT 0
19		40
20	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
		42	#define INDENT_STEP 3 // spaces per indentation level
		43
		44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
21		45
22	#define SB do {	46	#define SB do {
23	#define SE } while (0)	47	#define SE } while (0)
24		48
25	static HV *json_stash;	49	#if __GNUC__ >= 3
		50	# define expect(expr,value) __builtin_expect ((expr), (value))
		51	# define INLINE static inline
		52	#else
		53	# define expect(expr,value) (expr)
		54	# define INLINE static
		55	#endif
		56
		57	#define expect_false(expr) expect ((expr) != 0, 0)
		58	#define expect_true(expr) expect ((expr) != 0, 1)
		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
		74	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		75	static SV *json_true, *json_false;
		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		85
		86	typedef struct {
		87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
		91	SV *cb_object;
		92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	unsigned char incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
		99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
		107
		108	/////////////////////////////////////////////////////////////////////////////
		109	// utility functions
		110
		111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
		123	shrink (SV *sv)
		124	{
		125	sv_utf8_downgrade (sv, 1);
		126	if (SvLEN (sv) > SvCUR (sv) + 1)
		127	{
		128	#ifdef SvPV_shrink_to_cur
		129	SvPV_shrink_to_cur (sv);
		130	#elif defined (SvPV_renew)
		131	SvPV_renew (sv, SvCUR (sv) + 1);
		132	#endif
		133	}
		134	}
		135
		136	// decode an utf-8 character and return it, or (UV)-1 in
		137	// case of an error.
		138	// we special-case "safe" characters from U+80 .. U+7FF,
		139	// but use the very good perl function to parse anything else.
		140	// note that we never call this function for a ascii codepoints
		141	INLINE UV
		142	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		143	{
		144	if (expect_true (len >= 2
		145	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		146	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		147	{
		148	*clen = 2;
		149	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		150	}
		151	else
		152	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		153	}
		154
		155	// likewise for encoding, also never called for ascii codepoints
		156	// this function takes advantage of this fact, although current gccs
		157	// seem to optimise the check for >= 0x80 away anyways
		158	INLINE unsigned char *
		159	encode_utf8 (unsigned char *s, UV ch)
		160	{
		161	if (expect_false (ch < 0x000080))
		162	*s++ = ch;
		163	else if (expect_true (ch < 0x000800))
		164	*s++ = 0xc0 | ( ch >> 6),
		165	*s++ = 0x80 | ( ch & 0x3f);
		166	else if ( ch < 0x010000)
		167	*s++ = 0xe0 | ( ch >> 12),
		168	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		169	*s++ = 0x80 | ( ch & 0x3f);
		170	else if ( ch < 0x110000)
		171	*s++ = 0xf0 | ( ch >> 18),
		172	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		173	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		174	*s++ = 0x80 | ( ch & 0x3f);
		175
		176	return s;
		177	}
		178
		179	/////////////////////////////////////////////////////////////////////////////
		180	// encoder
26		181
27	// structure used for encoding JSON	182	// structure used for encoding JSON
28	typedef struct	183	typedef struct
29	{	184	{
30	char *cur;	185	char *cur; // SvPVX (sv) + current output position
31	STRLEN len; // SvLEN (sv)
32	char *end; // SvEND (sv)	186	char *end; // SvEND (sv)
33	SV *sv;	187	SV *sv; // result scalar
34	UV flags;	188	JSON json;
35	int max_recurse;	189	U32 indent; // indentation level
36	int indent;	190	UV limit; // escape character values >= this value when encoding
37	} enc_t;	191	} enc_t;
38		192
39	// structure used for decoding JSON	193	INLINE void
40	typedef struct
41	{
42	char *cur;
43	char *end;
44	char *err;
45	UV flags;
46	} dec_t;
47
48	static UV *
49	SvJSON (SV *sv)
50	{
51	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
52	croak ("object is not of type JSON::XS");
53
54	return &SvUVX (SvRV (sv));
55	}
56
57	/////////////////////////////////////////////////////////////////////////////
58
59	static void
60	need (enc_t *enc, STRLEN len)	194	need (enc_t *enc, STRLEN len)
61	{	195	{
62	if (enc->cur + len >= enc->end)	196	if (expect_false (enc->cur + len >= enc->end))
63	{	197	{
64	STRLEN cur = enc->cur - SvPVX (enc->sv);	198	STRLEN cur = enc->cur - SvPVX (enc->sv);
65	SvGROW (enc->sv, cur + len + 1);	199	SvGROW (enc->sv, cur + len + 1);
66	enc->cur = SvPVX (enc->sv) + cur;	200	enc->cur = SvPVX (enc->sv) + cur;
67	enc->end = SvEND (enc->sv);	201	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
68	}	202	}
69	}	203	}
70		204
71	static void	205	INLINE void
72	encode_ch (enc_t *enc, char ch)	206	encode_ch (enc_t *enc, char ch)
73	{	207	{
74	need (enc, 1);	208	need (enc, 1);
75	*enc->cur++ = ch;	209	*enc->cur++ = ch;
76	}	210	}
…		…
78	static void	212	static void
79	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	213	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
80	{	214	{
81	char *end = str + len;	215	char *end = str + len;
82		216
		217	need (enc, len);
		218
83	while (str < end)	219	while (str < end)
84	{	220	{
85	unsigned char ch = *(unsigned char *)str;	221	unsigned char ch = *(unsigned char *)str;
		222
86	if (ch >= 0x20 && ch < 0x80) // most common case	223	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
87	{	224	{
		225	if (expect_false (ch == '"')) // but with slow exceptions
		226	{
		227	need (enc, len += 1);
		228	*enc->cur++ = '\\';
		229	*enc->cur++ = '"';
		230	}
		231	else if (expect_false (ch == '\\'))
		232	{
		233	need (enc, len += 1);
		234	*enc->cur++ = '\\';
		235	*enc->cur++ = '\\';
		236	}
		237	else
88	*enc->cur++ = ch;	238	*enc->cur++ = ch;
		239
89	str++;	240	++str;
90	}	241	}
91	else	242	else
92	{	243	{
93	STRLEN clen;	244	switch (ch)
94	UV uch;
95
96	if (is_utf8)
97	{	245	{
98	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	246	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
99	if (clen < 0)	247	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
100	croak ("malformed UTF-8 character in string, cannot convert to JSON");	248	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
101	}	249	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
102	else	250	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
103	{
104	uch = ch;
105	clen = 1;
106	}
107		251
108	need (enc, len += 6);	252	default:
109
110	if (uch < 0xa0 || enc->flags & F_ASCII)
111	{
112	if (uch > 0xFFFFUL)
113	{	253	{
		254	STRLEN clen;
		255	UV uch;
		256
		257	if (is_utf8)
		258	{
		259	uch = decode_utf8 (str, end - str, &clen);
		260	if (clen == (STRLEN)-1)
		261	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		262	}
		263	else
		264	{
		265	uch = ch;
114	len += 6;	266	clen = 1;
		267	}
		268
		269	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		270	{
		271	if (uch >= 0x10000UL)
		272	{
		273	if (uch >= 0x110000UL)
		274	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		275
115	need (enc, len += 6);	276	need (enc, len += 11);
116	sprintf (enc->cur, "\\u%04x\\u%04x",	277	sprintf (enc->cur, "\\u%04x\\u%04x",
117	(uch - 0x10000) / 0x400 + 0xD800,	278	(int)((uch - 0x10000) / 0x400 + 0xD800),
118	(uch - 0x10000) % 0x400 + 0xDC00);	279	(int)((uch - 0x10000) % 0x400 + 0xDC00));
119	enc->cur += 12;	280	enc->cur += 12;
		281	}
		282	else
		283	{
		284	static char hexdigit [16] = "0123456789abcdef";
		285	need (enc, len += 5);
		286	*enc->cur++ = '\\';
		287	*enc->cur++ = 'u';
		288	*enc->cur++ = hexdigit [ uch >> 12 ];
		289	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		290	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		291	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		292	}
		293
		294	str += clen;
120	}	295	}
		296	else if (enc->json.flags & F_LATIN1)
		297	{
		298	*enc->cur++ = uch;
		299	str += clen;
		300	}
		301	else if (is_utf8)
		302	{
		303	need (enc, len += clen);
		304	do
		305	{
		306	*enc->cur++ = *str++;
		307	}
		308	while (--clen);
		309	}
121	else	310	else
122	{	311	{
123	sprintf (enc->cur, "\\u%04x", uch);	312	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
124	enc->cur += 6;	313	enc->cur = encode_utf8 (enc->cur, uch);
		314	++str;
		315	}
125	}	316	}
126	}	317	}
127	else if (is_utf8)
128	{
129	memcpy (enc->cur, str, clen);
130	enc->cur += clen;
131	}
132	else
133	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
134
135	str += clen;
136	}	318	}
137		319
138	--len;	320	--len;
139	}	321	}
140	}	322	}
141		323
142	#define INDENT SB \	324	INLINE void
		325	encode_indent (enc_t *enc)
		326	{
143	if (enc->flags & F_INDENT) \	327	if (enc->json.flags & F_INDENT)
144	{ \	328	{
145	int i_; \	329	int spaces = enc->indent * INDENT_STEP;
146	need (enc, enc->indent); \	330
147	for (i_ = enc->indent * 3; i_--; )\	331	need (enc, spaces);
		332	memset (enc->cur, ' ', spaces);
		333	enc->cur += spaces;
		334	}
		335	}
		336
		337	INLINE void
		338	encode_space (enc_t *enc)
		339	{
		340	need (enc, 1);
		341	encode_ch (enc, ' ');
		342	}
		343
		344	INLINE void
		345	encode_nl (enc_t *enc)
		346	{
		347	if (enc->json.flags & F_INDENT)
		348	{
		349	need (enc, 1);
148	encode_ch (enc, ' '); \	350	encode_ch (enc, '\n');
149	} \	351	}
150	SE	352	}
151		353
152	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	354	INLINE void
153	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	355	encode_comma (enc_t *enc)
154	#define COMMA SB \	356	{
155	encode_ch (enc, ','); \	357	encode_ch (enc, ',');
		358
156	if (enc->flags & F_INDENT) \	359	if (enc->json.flags & F_INDENT)
157	NL; \	360	encode_nl (enc);
158	else if (enc->flags & F_SPACE_AFTER) \	361	else if (enc->json.flags & F_SPACE_AFTER)
159	SPACE; \	362	encode_space (enc);
160	SE	363	}
161		364
162	static void encode_sv (enc_t *enc, SV *sv);	365	static void encode_sv (enc_t *enc, SV *sv);
163		366
164	static void	367	static void
165	encode_av (enc_t *enc, AV *av)	368	encode_av (enc_t *enc, AV *av)
166	{	369	{
167	int i, len = av_len (av);	370	int i, len = av_len (av);
168		371
		372	if (enc->indent >= enc->json.max_depth)
		373	croak (ERR_NESTING_EXCEEDED);
		374
169	encode_ch (enc, '['); NL;	375	encode_ch (enc, '[');
170	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
171		380
172	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
173	{	382	{
174	INDENT;	383	SV **svp = av_fetch (av, i, 0);
175	encode_sv (enc, *av_fetch (av, i, 0));
176		384
		385	encode_indent (enc);
		386
		387	if (svp)
		388	encode_sv (enc, *svp);
		389	else
		390	encode_str (enc, "null", 4, 0);
		391
177	if (i < len)	392	if (i < len)
178	COMMA;	393	encode_comma (enc);
179	}	394	}
180		395
181	NL;	396	encode_nl (enc); --enc->indent; encode_indent (enc);
182		397	}
183	--enc->indent;	398
184	INDENT; encode_ch (enc, ']');	399	encode_ch (enc, ']');
185	}	400	}
186		401
187	static void	402	static void
188	encode_he (enc_t *enc, HE *he)	403	encode_hk (enc_t *enc, HE *he)
189	{	404	{
190	encode_ch (enc, '"');	405	encode_ch (enc, '"');
191		406
192	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
193	{	408	{
194	SV *sv = HeSVKEY (he);	409	SV *sv = HeSVKEY (he);
195	STRLEN len;	410	STRLEN len;
		411	char *str;
		412
		413	SvGETMAGIC (sv);
196	char *str = SvPV (sv, len);	414	str = SvPV (sv, len);
197		415
198	encode_str (enc, str, len, SvUTF8 (sv));	416	encode_str (enc, str, len, SvUTF8 (sv));
199	}	417	}
200	else	418	else
201	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	419	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
202		420
203	encode_ch (enc, '"');	421	encode_ch (enc, '"');
204		422
205	if (enc->flags & F_SPACE_BEFORE) SPACE;	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
206	encode_ch (enc, ':');	424	encode_ch (enc, ':');
207	if (enc->flags & F_SPACE_AFTER ) SPACE;	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
208	encode_sv (enc, HeVAL (he));
209	}	426	}
210		427
211	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
212	static int	429	static int
213	he_cmp_fast (const void *a_, const void *b_)	430	he_cmp_fast (const void *a_, const void *b_)
…		…
218	HE *b = *(HE **)b_;	435	HE *b = *(HE **)b_;
219		436
220	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
221	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
222		439
223	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
224	cmp = la < lb ? -1 : la == lb ? 0 : 1;	441	cmp = lb - la;
225		442
226	return cmp;	443	return cmp;
227	}	444	}
228		445
229	// compare hash entries, used when some keys are sv's or utf-x	446	// compare hash entries, used when some keys are sv's or utf-x
230	static int	447	static int
231	he_cmp_slow (const void *a, const void *b)	448	he_cmp_slow (const void *a, const void *b)
232	{	449	{
233	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	450	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
234	}	451	}
235		452
236	static void	453	static void
237	encode_hv (enc_t *enc, HV *hv)	454	encode_hv (enc_t *enc, HV *hv)
238	{	455	{
239	int count, i;	456	HE *he;
240		457
241	encode_ch (enc, '{'); NL; ++enc->indent;	458	if (enc->indent >= enc->json.max_depth)
		459	croak (ERR_NESTING_EXCEEDED);
242		460
243	if ((count = hv_iterinit (hv)))	461	encode_ch (enc, '{');
244	{	462
245	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
246	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
247	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x.
248	// that randomises hash orderings	466	// that randomises hash orderings
249	if (enc->flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
		468	{
		469	int count = hv_iterinit (hv);
		470
		471	if (SvMAGICAL (hv))
250	{	472	{
251	HE *he, *hes [count];	473	// need to count by iterating. could improve by dynamically building the vector below
		474	// but I don't care for the speed of this special case.
		475	// note also that we will run into undefined behaviour when the two iterations
		476	// do not result in the same count, something I might care for in some later release.
		477
		478	count = 0;
		479	while (hv_iternext (hv))
		480	++count;
		481
		482	hv_iterinit (hv);
		483	}
		484
		485	if (count)
		486	{
252	int fast = 1;	487	int i, fast = 1;
		488	#if defined(__BORLANDC__) || defined(_MSC_VER)
		489	HE **hes = _alloca (count * sizeof (HE));
		490	#else
		491	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		492	#endif
253		493
254	i = 0;	494	i = 0;
255	while ((he = hv_iternext (hv)))	495	while ((he = hv_iternext (hv)))
256	{	496	{
257	hes [i++] = he;	497	hes [i++] = he;
…		…
263		503
264	if (fast)	504	if (fast)
265	qsort (hes, count, sizeof (HE *), he_cmp_fast);	505	qsort (hes, count, sizeof (HE *), he_cmp_fast);
266	else	506	else
267	{	507	{
268	// hack to disable "use bytes"	508	// hack to forcefully disable "use bytes"
269	COP *oldcop = PL_curcop, cop;	509	COP cop = *PL_curcop;
270	cop.op_private = 0;	510	cop.op_private = 0;
		511
		512	ENTER;
		513	SAVETMPS;
		514
		515	SAVEVPTR (PL_curcop);
271	PL_curcop = &cop;	516	PL_curcop = &cop;
272		517
273	SAVETMPS;
274	qsort (hes, count, sizeof (HE *), he_cmp_slow);	518	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		519
275	FREETMPS;	520	FREETMPS;
276		521	LEAVE;
277	PL_curcop = oldcop;
278	}	522	}
279		523
280	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
281	{	527	{
282	INDENT;	528	encode_indent (enc);
		529	he = hes [count];
283	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
284		532
285	if (i < count - 1)	533	if (count)
286	COMMA;	534	encode_comma (enc);
287	}	535	}
288		536
		537	encode_nl (enc); --enc->indent; encode_indent (enc);
		538	}
		539	}
		540	else
		541	{
		542	if (hv_iterinit (hv) || SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
289	NL;	544	{
		545	encode_nl (enc); ++enc->indent;
		546
		547	for (;;)
		548	{
		549	encode_indent (enc);
		550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		552
		553	if (!(he = hv_iternext (hv)))
		554	break;
		555
		556	encode_comma (enc);
		557	}
		558
		559	encode_nl (enc); --enc->indent; encode_indent (enc);
		560	}
		561	}
		562
		563	encode_ch (enc, '}');
		564	}
		565
		566	// encode objects, arrays and special \0=false and \1=true values.
		567	static void
		568	encode_rv (enc_t *enc, SV *sv)
		569	{
		570	svtype svt;
		571
		572	SvGETMAGIC (sv);
		573	svt = SvTYPE (sv);
		574
		575	if (expect_false (SvOBJECT (sv)))
		576	{
		577	HV *stash = !JSON_SLOW || json_boolean_stash
		578	? json_boolean_stash
		579	: gv_stashpv ("JSON::XS::Boolean", 1);
		580
		581	if (SvSTASH (sv) == stash)
		582	{
		583	if (SvIV (sv))
		584	encode_str (enc, "true", 4, 0);
		585	else
		586	encode_str (enc, "false", 5, 0);
290	}	587	}
291	else	588	else
292	{	589	{
293	SV *sv;	590	#if 0
294	HE *he = hv_iternext (hv);	591	if (0 && sv_derived_from (rv, "JSON::Literal"))
295
296	for (;;)
297	{	592	{
298	INDENT;	593	// not yet
299	encode_he (enc, he);
300
301	if (!(he = hv_iternext (hv)))
302	break;
303
304	COMMA;
305	}	594	}
306		595	#endif
		596	if (enc->json.flags & F_CONV_BLESSED)
307	NL;	597	{
		598	// we re-bless the reference to get overload and other niceties right
		599	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		600
		601	if (to_json)
		602	{
		603	dSP;
		604
		605	ENTER; SAVETMPS; PUSHMARK (SP);
		606	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		607
		608	// calling with G_SCALAR ensures that we always get a 1 return value
		609	PUTBACK;
		610	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		611	SPAGAIN;
		612
		613	// catch this surprisingly common error
		614	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		615	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		616
		617	sv = POPs;
		618	PUTBACK;
		619
		620	encode_sv (enc, sv);
		621
		622	FREETMPS; LEAVE;
		623	}
		624	else if (enc->json.flags & F_ALLOW_BLESSED)
		625	encode_str (enc, "null", 4, 0);
		626	else
		627	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		628	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		629	}
		630	else if (enc->json.flags & F_ALLOW_BLESSED)
		631	encode_str (enc, "null", 4, 0);
		632	else
		633	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		634	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
308	}	635	}
309	}	636	}
		637	else if (svt == SVt_PVHV)
		638	encode_hv (enc, (HV *)sv);
		639	else if (svt == SVt_PVAV)
		640	encode_av (enc, (AV *)sv);
		641	else if (svt < SVt_PVAV)
		642	{
		643	STRLEN len = 0;
		644	char *pv = svt ? SvPV (sv, len) : 0;
310		645
311	--enc->indent; INDENT; encode_ch (enc, '}');	646	if (len == 1 && *pv == '1')
		647	encode_str (enc, "true", 4, 0);
		648	else if (len == 1 && *pv == '0')
		649	encode_str (enc, "false", 5, 0);
		650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		651	encode_str (enc, "null", 4, 0);
		652	else
		653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
		658	else
		659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
312	}	661	}
313		662
314	static void	663	static void
315	encode_sv (enc_t *enc, SV *sv)	664	encode_sv (enc_t *enc, SV *sv)
316	{	665	{
		666	SvGETMAGIC (sv);
		667
317	if (SvPOKp (sv))	668	if (SvPOKp (sv))
318	{	669	{
319	STRLEN len;	670	STRLEN len;
320	char *str = SvPV (sv, len);	671	char *str = SvPV (sv, len);
321	encode_ch (enc, '"');	672	encode_ch (enc, '"');
322	encode_str (enc, str, len, SvUTF8 (sv));	673	encode_str (enc, str, len, SvUTF8 (sv));
323	encode_ch (enc, '"');	674	encode_ch (enc, '"');
324	}	675	}
325	else if (SvNOKp (sv))	676	else if (SvNOKp (sv))
326	{	677	{
		678	// trust that perl will do the right thing w.r.t. JSON syntax.
327	need (enc, NV_DIG + 32);	679	need (enc, NV_DIG + 32);
328	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
329	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
330	}	682	}
331	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
332	{	684	{
		685	// we assume we can always read an IV as a UV and vice versa
		686	// we assume two's complement
		687	// we assume no aliasing issues in the union
		688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		690	{
		691	// optimise the "small number case"
		692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
		694	I32 i = SvIVX (sv);
		695	U32 u;
		696	char digit, nz = 0;
		697
333	need (enc, 64);	698	need (enc, 6);
		699
		700	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		701	u = i < 0 ? -i : i;
		702
		703	// convert to 4.28 fixed-point representation
		704	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		705
		706	// now output digit by digit, each time masking out the integer part
		707	// and multiplying by 5 while moving the decimal point one to the right,
		708	// resulting in a net multiplication by 10.
		709	// we always write the digit to memory but conditionally increment
		710	// the pointer, to enable the use of conditional move instructions.
		711	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		712	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		713	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		714	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
334	enc->cur +=	721	enc->cur +=
335	SvIsUV(sv)	722	SvIsUV(sv)
336	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
337	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
338	}	726	}
339	else if (SvROK (sv))	727	else if (SvROK (sv))
340	{	728	encode_rv (enc, SvRV (sv));
341	if (!--enc->max_recurse)	729	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
342	croak ("data structure too deep (hit recursion limit)");
343
344	sv = SvRV (sv);
345
346	switch (SvTYPE (sv))
347	{
348	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
349	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
350
351	default:
352	croak ("JSON can only represent references to arrays or hashes");
353	}
354	}
355	else if (!SvOK (sv))
356	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
357	else	731	else
358	croak ("encountered perl type that JSON cannot handle");	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		733	SvPV_nolen (sv), SvFLAGS (sv));
359	}	734	}
360		735
361	static SV *	736	static SV *
362	encode_json (SV *scalar, UV flags)	737	encode_json (SV *scalar, JSON *json)
363	{	738	{
364	enc_t enc;	739	enc_t enc;
365	enc.flags = flags;	740
		741	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		742	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		743
		744	enc.json = *json;
366	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
367	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
368	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
369	enc.max_recurse = 0;
370	enc.indent = 0;	748	enc.indent = 0;
		749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		750	: enc.json.flags & F_LATIN1 ? 0x000100UL
		751	: 0x110000UL;
371		752
372	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
373	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
374		755
		756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		757	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		758
375	if (!(flags & (F_ASCII | F_UTF8)))	759	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
376	SvUTF8_on (enc.sv);	760	SvUTF8_on (enc.sv);
377		761
378	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	762	if (enc.json.flags & F_SHRINK)
		763	shrink (enc.sv);
		764
379	return enc.sv;	765	return enc.sv;
380	}	766	}
381		767
382	/////////////////////////////////////////////////////////////////////////////	768	/////////////////////////////////////////////////////////////////////////////
		769	// decoder
383		770
384	#define WS \	771	// structure used for decoding JSON
		772	typedef struct
		773	{
		774	char *cur; // current parser pointer
		775	char *end; // end of input string
		776	const char *err; // parse error, if != 0
		777	JSON json;
		778	U32 depth; // recursion depth
		779	U32 maxdepth; // recursion depth limit
		780	} dec_t;
		781
		782	INLINE void
		783	decode_comment (dec_t *dec)
		784	{
		785	// only '#'-style comments allowed a.t.m.
		786
		787	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		788	++dec->cur;
		789	}
		790
		791	INLINE void
		792	decode_ws (dec_t *dec)
		793	{
385	for (;;) \	794	for (;;)
386	{ \	795	{
387	char ch = *dec->cur; \	796	char ch = *dec->cur;
		797
388	if (ch > 0x20 \	798	if (ch > 0x20)
		799	{
		800	if (expect_false (ch == '#'))
		801	{
		802	if (dec->json.flags & F_RELAXED)
		803	decode_comment (dec);
		804	else
		805	break;
		806	}
		807	else
		808	break;
		809	}
389	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
390	break; \	811	break; // parse error, but let higher level handle it, gives better error messages
		812
391	++dec->cur; \	813	++dec->cur;
392	}	814	}
		815	}
393		816
394	#define ERR(reason) SB dec->err = reason; goto fail; SE	817	#define ERR(reason) SB dec->err = reason; goto fail; SE
		818
395	#define EXPECT_CH(ch) SB \	819	#define EXPECT_CH(ch) SB \
396	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
397	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
398	++dec->cur; \	822	++dec->cur; \
399	SE	823	SE
400		824
		825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		826	#define DEC_DEC_DEPTH --dec->depth
		827
401	static SV *decode_sv (dec_t *dec);	828	static SV *decode_sv (dec_t *dec);
402
403	#define APPEND_CH(ch) SB \
404	SvGROW (sv, cur + 1 + 1); \
405	SvPVX (sv)[cur++] = (ch); \
406	SE
407		829
408	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
409		831
410	static UV	832	static UV
411	decode_4hex (dec_t *dec)	833	decode_4hex (dec_t *dec)
412	{	834	{
413	signed char d1, d2, d3, d4;	835	signed char d1, d2, d3, d4;
		836	unsigned char *cur = (unsigned char *)dec->cur;
414		837
415	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	838	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
416	if (d1 < 0) ERR ("four hexadecimal digits expected");	839	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
417	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	840	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
418	if (d2 < 0) ERR ("four hexadecimal digits expected");	841	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
419	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
420	if (d3 < 0) ERR ("four hexadecimal digits expected");
421	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
422	if (d4 < 0) ERR ("four hexadecimal digits expected");
423		842
424	dec->cur += 4;	843	dec->cur += 4;
425		844
426	return ((UV)d1) << 12	845	return ((UV)d1) << 12
427	| ((UV)d2) << 8	846	| ((UV)d2) << 8
…		…
433	}	852	}
434		853
435	static SV *	854	static SV *
436	decode_str (dec_t *dec)	855	decode_str (dec_t *dec)
437	{	856	{
438	SV *sv = NEWSV (0,2);	857	SV *sv = 0;
439	STRLEN cur = 0;
440	int utf8 = 0;	858	int utf8 = 0;
		859	char *dec_cur = dec->cur;
441		860
442	for (;;)	861	do
443	{	862	{
444	unsigned char ch = *(unsigned char *)dec->cur;	863	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		864	char *cur = buf;
445		865
446	if (ch == '"')	866	do
447	break;
448	else if (ch == '\\')
449	{	867	{
450	switch (*++dec->cur)	868	unsigned char ch = *(unsigned char *)dec_cur++;
		869
		870	if (expect_false (ch == '"'))
451	{	871	{
452	case '\\':	872	--dec_cur;
453	case '/':	873	break;
454	case '"': APPEND_CH (*dec->cur++); break;	874	}
455		875	else if (expect_false (ch == '\\'))
456	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	876	{
457	case 't': APPEND_CH ('\011'); ++dec->cur; break;	877	switch (*dec_cur)
458	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
459	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
460	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
461
462	case 'u':
463	{	878	{
464	UV lo, hi;	879	case '\\':
465	++dec->cur;	880	case '/':
		881	case '"': *cur++ = *dec_cur++; break;
466		882
467	hi = decode_4hex (dec);	883	case 'b': ++dec_cur; *cur++ = '\010'; break;
468	if (hi == (UV)-1)	884	case 't': ++dec_cur; *cur++ = '\011'; break;
469	goto fail;	885	case 'n': ++dec_cur; *cur++ = '\012'; break;
		886	case 'f': ++dec_cur; *cur++ = '\014'; break;
		887	case 'r': ++dec_cur; *cur++ = '\015'; break;
470		888
471	// possibly a surrogate pair	889	case 'u':
472	if (hi >= 0xd800 && hi < 0xdc00)
473	{	890	{
474	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	891	UV lo, hi;
475	ERR ("illegal surrogate character");	892	++dec_cur;
476		893
477	dec->cur += 2;	894	dec->cur = dec_cur;
478
479	lo = decode_4hex (dec);	895	hi = decode_4hex (dec);
		896	dec_cur = dec->cur;
480	if (lo == (UV)-1)	897	if (hi == (UV)-1)
481	goto fail;	898	goto fail;
482		899
		900	// possibly a surrogate pair
		901	if (hi >= 0xd800)
		902	if (hi < 0xdc00)
		903	{
		904	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		905	ERR ("missing low surrogate character in surrogate pair");
		906
		907	dec_cur += 2;
		908
		909	dec->cur = dec_cur;
		910	lo = decode_4hex (dec);
		911	dec_cur = dec->cur;
		912	if (lo == (UV)-1)
		913	goto fail;
		914
483	if (lo < 0xdc00 || lo >= 0xe000)	915	if (lo < 0xdc00 || lo >= 0xe000)
484	ERR ("surrogate pair expected");	916	ERR ("surrogate pair expected");
485		917
486	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	918	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		919	}
		920	else if (hi < 0xe000)
		921	ERR ("missing high surrogate character in surrogate pair");
		922
		923	if (hi >= 0x80)
		924	{
		925	utf8 = 1;
		926
		927	cur = encode_utf8 (cur, hi);
		928	}
		929	else
		930	*cur++ = hi;
487	}	931	}
488	else if (lo >= 0xdc00 && lo < 0xe000)
489	ERR ("illegal surrogate character");
490
491	if (hi >= 0x80)
492	{	932	break;
493	utf8 = 1;
494		933
495	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
496	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
497	}
498	else	934	default:
499	APPEND_CH (hi);	935	--dec_cur;
		936	ERR ("illegal backslash escape sequence in string");
500	}	937	}
		938	}
		939	else if (expect_true (ch >= 0x20 && ch < 0x80))
		940	*cur++ = ch;
		941	else if (ch >= 0x80)
		942	{
		943	STRLEN clen;
		944	UV uch;
		945
		946	--dec_cur;
		947
		948	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		949	if (clen == (STRLEN)-1)
		950	ERR ("malformed UTF-8 character in JSON string");
		951
		952	do
		953	*cur++ = *dec_cur++;
		954	while (--clen);
		955
		956	utf8 = 1;
		957	}
		958	else
		959	{
		960	--dec_cur;
		961
		962	if (!ch)
		963	ERR ("unexpected end of string while parsing JSON string");
501	break;	964	else
		965	ERR ("invalid character encountered while parsing JSON string");
502	}	966	}
503	}	967	}
504	else if (ch >= 0x20 && ch <= 0x7f)	968	while (cur < buf + SHORT_STRING_LEN);
505	APPEND_CH (*dec->cur++);	969
506	else if (ch >= 0x80)	970	{
		971	STRLEN len = cur - buf;
		972
		973	if (sv)
507	{	974	{
508	STRLEN clen;	975	SvGROW (sv, SvCUR (sv) + len + 1);
509	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	976	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
510	if (clen < 0)	977	SvCUR_set (sv, SvCUR (sv) + len);
511	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
512
513	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
514	memcpy (SvPVX (sv) + cur, dec->cur, clen);
515	dec->cur += clen;
516	}	978	}
517	else	979	else
518	ERR ("invalid character encountered");	980	sv = newSVpvn (buf, len);
519	}	981	}
		982	}
		983	while (*dec_cur != '"');
520		984
521	++dec->cur;	985	++dec_cur;
522		986
		987	if (sv)
		988	{
523	SvPOK_only (sv);	989	SvPOK_only (sv);
524
525	SvCUR_set (sv, cur);
526	*SvEND (sv) = 0;	990	*SvEND (sv) = 0;
527		991
528	if (utf8)	992	if (utf8)
529	SvUTF8_on (sv);	993	SvUTF8_on (sv);
		994	}
		995	else
		996	sv = newSVpvn ("", 0);
530		997
		998	dec->cur = dec_cur;
531	return sv;	999	return sv;
532		1000
533	fail:	1001	fail:
534	SvREFCNT_dec (sv);	1002	dec->cur = dec_cur;
535	return 0;	1003	return 0;
536	}	1004	}
537		1005
538	static SV *	1006	static SV *
539	decode_num (dec_t *dec)	1007	decode_num (dec_t *dec)
…		…
549	{	1017	{
550	++dec->cur;	1018	++dec->cur;
551	if (*dec->cur >= '0' && *dec->cur <= '9')	1019	if (*dec->cur >= '0' && *dec->cur <= '9')
552	ERR ("malformed number (leading zero must not be followed by another digit)");	1020	ERR ("malformed number (leading zero must not be followed by another digit)");
553	}	1021	}
554		1022	else if (*dec->cur < '0' || *dec->cur > '9')
555	// int	1023	ERR ("malformed number (no digits after initial minus)");
		1024	else
		1025	do
		1026	{
		1027	++dec->cur;
		1028	}
556	while (*dec->cur >= '0' && *dec->cur <= '9')	1029	while (*dec->cur >= '0' && *dec->cur <= '9');
557	++dec->cur;
558		1030
559	// [frac]	1031	// [frac]
560	if (*dec->cur == '.')	1032	if (*dec->cur == '.')
561	{	1033	{
562	is_nv = 1;	1034	++dec->cur;
		1035
		1036	if (*dec->cur < '0' || *dec->cur > '9')
		1037	ERR ("malformed number (no digits after decimal point)");
563		1038
564	do	1039	do
565	{	1040	{
566	++dec->cur;	1041	++dec->cur;
567	}	1042	}
568	while (*dec->cur >= '0' && *dec->cur <= '9');	1043	while (*dec->cur >= '0' && *dec->cur <= '9');
		1044
		1045	is_nv = 1;
569	}	1046	}
570		1047
571	// [exp]	1048	// [exp]
572	if (*dec->cur == 'e' || *dec->cur == 'E')	1049	if (*dec->cur == 'e' || *dec->cur == 'E')
573	{	1050	{
574	is_nv = 1;
575
576	++dec->cur;	1051	++dec->cur;
		1052
577	if (*dec->cur == '-' || *dec->cur == '+')	1053	if (*dec->cur == '-' || *dec->cur == '+')
578	++dec->cur;	1054	++dec->cur;
579		1055
		1056	if (*dec->cur < '0' || *dec->cur > '9')
		1057	ERR ("malformed number (no digits after exp sign)");
		1058
		1059	do
		1060	{
		1061	++dec->cur;
		1062	}
580	while (*dec->cur >= '0' && *dec->cur <= '9')	1063	while (*dec->cur >= '0' && *dec->cur <= '9');
581	++dec->cur;	1064
		1065	is_nv = 1;
582	}	1066	}
583		1067
584	if (!is_nv)	1068	if (!is_nv)
585	{	1069	{
586	UV uv;	1070	int len = dec->cur - start;
587	int numtype = grok_number (start, dec->cur - start, &uv);	1071
588	if (numtype & IS_NUMBER_IN_UV)	1072	// special case the rather common 1..5-digit-int case
589	if (numtype & IS_NUMBER_NEG)	1073	if (*start == '-')
		1074	switch (len)
590	{	1075	{
591	if (uv < (UV)IV_MIN)	1076	case 2: return newSViv (-( start [1] - '0' * 1));
592	return newSViv (-(IV)uv);	1077	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1078	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1079	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1080	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
593	}	1081	}
		1082	else
		1083	switch (len)
		1084	{
		1085	case 1: return newSViv ( start [0] - '0' * 1);
		1086	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1087	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1088	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1089	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1090	}
		1091
		1092	{
		1093	UV uv;
		1094	int numtype = grok_number (start, len, &uv);
		1095	if (numtype & IS_NUMBER_IN_UV)
		1096	if (numtype & IS_NUMBER_NEG)
		1097	{
		1098	if (uv < (UV)IV_MIN)
		1099	return newSViv (-(IV)uv);
		1100	}
594	else	1101	else
595	return newSVuv (uv);	1102	return newSVuv (uv);
596	}	1103	}
597		1104
		1105	len -= *start == '-' ? 1 : 0;
		1106
		1107	// does not fit into IV or UV, try NV
		1108	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1109	#if defined (LDBL_DIG)
		1110	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1111	#endif
		1112	)
		1113	// fits into NV without loss of precision
		1114	return newSVnv (Atof (start));
		1115
		1116	// everything else fails, convert it to a string
		1117	return newSVpvn (start, dec->cur - start);
		1118	}
		1119
		1120	// loss of precision here
598	return newSVnv (Atof (start));	1121	return newSVnv (Atof (start));
599		1122
600	fail:	1123	fail:
601	return 0;	1124	return 0;
602	}	1125	}
…		…
604	static SV *	1127	static SV *
605	decode_av (dec_t *dec)	1128	decode_av (dec_t *dec)
606	{	1129	{
607	AV *av = newAV ();	1130	AV *av = newAV ();
608		1131
		1132	DEC_INC_DEPTH;
		1133	decode_ws (dec);
		1134
		1135	if (*dec->cur == ']')
		1136	++dec->cur;
		1137	else
609	for (;;)	1138	for (;;)
610	{	1139	{
611	SV *value;	1140	SV *value;
612		1141
613	value = decode_sv (dec);	1142	value = decode_sv (dec);
614	if (!value)	1143	if (!value)
615	goto fail;	1144	goto fail;
616		1145
617	av_push (av, value);	1146	av_push (av, value);
618		1147
619	WS;	1148	decode_ws (dec);
620		1149
621	if (*dec->cur == ']')	1150	if (*dec->cur == ']')
622	{	1151	{
623	++dec->cur;	1152	++dec->cur;
624	break;	1153	break;
		1154	}
625	}	1155
626
627	if (*dec->cur != ',')	1156	if (*dec->cur != ',')
628	ERR (", or ] expected while parsing array");	1157	ERR (", or ] expected while parsing array");
629		1158
630	++dec->cur;	1159	++dec->cur;
		1160
		1161	decode_ws (dec);
		1162
		1163	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1164	{
		1165	++dec->cur;
		1166	break;
		1167	}
631	}	1168	}
632		1169
		1170	DEC_DEC_DEPTH;
633	return newRV_noinc ((SV *)av);	1171	return newRV_noinc ((SV *)av);
634		1172
635	fail:	1173	fail:
636	SvREFCNT_dec (av);	1174	SvREFCNT_dec (av);
		1175	DEC_DEC_DEPTH;
637	return 0;	1176	return 0;
638	}	1177	}
639		1178
640	static SV *	1179	static SV *
641	decode_hv (dec_t *dec)	1180	decode_hv (dec_t *dec)
642	{	1181	{
		1182	SV *sv;
643	HV *hv = newHV ();	1183	HV *hv = newHV ();
644		1184
		1185	DEC_INC_DEPTH;
		1186	decode_ws (dec);
		1187
		1188	if (*dec->cur == '}')
		1189	++dec->cur;
		1190	else
645	for (;;)	1191	for (;;)
646	{	1192	{
647	SV *key, *value;
648
649	WS; EXPECT_CH ('"');	1193	EXPECT_CH ('"');
650		1194
651	key = decode_str (dec);	1195	// heuristic: assume that
652	if (!key)	1196	// a) decode_str + hv_store_ent are abysmally slow.
653	goto fail;	1197	// b) most hash keys are short, simple ascii text.
654		1198	// => try to "fast-match" such strings to avoid
655	WS; EXPECT_CH (':');	1199	// the overhead of decode_str + hv_store_ent.
656
657	value = decode_sv (dec);
658	if (!value)
659	{	1200	{
		1201	SV *value;
		1202	char *p = dec->cur;
		1203	char *e = p + 24; // only try up to 24 bytes
		1204
		1205	for (;;)
		1206	{
		1207	// the >= 0x80 is false on most architectures
		1208	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1209	{
		1210	// slow path, back up and use decode_str
		1211	SV *key = decode_str (dec);
		1212	if (!key)
		1213	goto fail;
		1214
		1215	decode_ws (dec); EXPECT_CH (':');
		1216
		1217	decode_ws (dec);
		1218	value = decode_sv (dec);
		1219	if (!value)
		1220	{
		1221	SvREFCNT_dec (key);
		1222	goto fail;
		1223	}
		1224
		1225	hv_store_ent (hv, key, value, 0);
660	SvREFCNT_dec (key);	1226	SvREFCNT_dec (key);
		1227
		1228	break;
		1229	}
		1230	else if (*p == '"')
		1231	{
		1232	// fast path, got a simple key
		1233	char *key = dec->cur;
		1234	int len = p - key;
		1235	dec->cur = p + 1;
		1236
		1237	decode_ws (dec); EXPECT_CH (':');
		1238
		1239	decode_ws (dec);
		1240	value = decode_sv (dec);
		1241	if (!value)
661	goto fail;	1242	goto fail;
		1243
		1244	hv_store (hv, key, len, value, 0);
		1245
		1246	break;
		1247	}
		1248
		1249	++p;
		1250	}
662	}	1251	}
663		1252
664	//TODO: optimise	1253	decode_ws (dec);
665	hv_store_ent (hv, key, value, 0);
666		1254
667	WS;
668
669	if (*dec->cur == '}')	1255	if (*dec->cur == '}')
		1256	{
		1257	++dec->cur;
		1258	break;
		1259	}
		1260
		1261	if (*dec->cur != ',')
		1262	ERR (", or } expected while parsing object/hash");
		1263
		1264	++dec->cur;
		1265
		1266	decode_ws (dec);
		1267
		1268	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1269	{
		1270	++dec->cur;
		1271	break;
		1272	}
		1273	}
		1274
		1275	DEC_DEC_DEPTH;
		1276	sv = newRV_noinc ((SV *)hv);
		1277
		1278	// check filter callbacks
		1279	if (dec->json.flags & F_HOOK)
		1280	{
		1281	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
670	{	1282	{
671	++dec->cur;	1283	HE *cb, *he;
672	break;	1284
		1285	hv_iterinit (hv);
		1286	he = hv_iternext (hv);
		1287	hv_iterinit (hv);
		1288
		1289	// the next line creates a mortal sv each time its called.
		1290	// might want to optimise this for common cases.
		1291	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1292
		1293	if (cb)
		1294	{
		1295	dSP;
		1296	int count;
		1297
		1298	ENTER; SAVETMPS; PUSHMARK (SP);
		1299	XPUSHs (HeVAL (he));
		1300
		1301	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1302
		1303	if (count == 1)
		1304	{
		1305	sv = newSVsv (POPs);
		1306	FREETMPS; LEAVE;
		1307	return sv;
		1308	}
		1309
		1310	FREETMPS; LEAVE;
		1311	}
673	}	1312	}
674		1313
675	if (*dec->cur != ',')	1314	if (dec->json.cb_object)
676	ERR (", or } expected while parsing object/hash");	1315	{
		1316	dSP;
		1317	int count;
677		1318
678	++dec->cur;	1319	ENTER; SAVETMPS; PUSHMARK (SP);
679	}	1320	XPUSHs (sv_2mortal (sv));
680		1321
681	return newRV_noinc ((SV *)hv);	1322	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1323
		1324	if (count == 1)
		1325	{
		1326	sv = newSVsv (POPs);
		1327	FREETMPS; LEAVE;
		1328	return sv;
		1329	}
		1330
		1331	SvREFCNT_inc (sv);
		1332	FREETMPS; LEAVE;
		1333	}
		1334	}
		1335
		1336	return sv;
682		1337
683	fail:	1338	fail:
684	SvREFCNT_dec (hv);	1339	SvREFCNT_dec (hv);
		1340	DEC_DEC_DEPTH;
685	return 0;	1341	return 0;
686	}	1342	}
687		1343
688	static SV *	1344	static SV *
689	decode_sv (dec_t *dec)	1345	decode_sv (dec_t *dec)
690	{	1346	{
691	WS;	1347	// the beauty of JSON: you need exactly one character lookahead
		1348	// to parse everything.
692	switch (*dec->cur)	1349	switch (*dec->cur)
693	{	1350	{
694	case '"': ++dec->cur; return decode_str (dec);	1351	case '"': ++dec->cur; return decode_str (dec);
695	case '[': ++dec->cur; return decode_av (dec);	1352	case '[': ++dec->cur; return decode_av (dec);
696	case '{': ++dec->cur; return decode_hv (dec);	1353	case '{': ++dec->cur; return decode_hv (dec);
697		1354
698	case '-':	1355	case '-':
699	case '0': case '1': case '2': case '3': case '4':	1356	case '0': case '1': case '2': case '3': case '4':
700	case '5': case '6': case '7': case '8': case '9':	1357	case '5': case '6': case '7': case '8': case '9':
701	return decode_num (dec);	1358	return decode_num (dec);
702		1359
703	case 't':	1360	case 't':
704	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1361	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
705	{	1362	{
706	dec->cur += 4;	1363	dec->cur += 4;
		1364	#if JSON_SLOW
		1365	json_true = get_bool ("JSON::XS::true");
		1366	#endif
707	return newSViv (1);	1367	return newSVsv (json_true);
708	}	1368	}
709	else	1369	else
710	ERR ("'true' expected");	1370	ERR ("'true' expected");
711		1371
712	break;	1372	break;
713		1373
714	case 'f':	1374	case 'f':
715	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1375	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
716	{	1376	{
717	dec->cur += 5;	1377	dec->cur += 5;
		1378	#if JSON_SLOW
		1379	json_false = get_bool ("JSON::XS::false");
		1380	#endif
718	return newSViv (0);	1381	return newSVsv (json_false);
719	}	1382	}
720	else	1383	else
721	ERR ("'false' expected");	1384	ERR ("'false' expected");
722		1385
723	break;	1386	break;
724		1387
725	case 'n':	1388	case 'n':
726	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1389	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
727	{	1390	{
728	dec->cur += 4;	1391	dec->cur += 4;
729	return newSViv (1);	1392	return newSVsv (&PL_sv_undef);
730	}	1393	}
731	else	1394	else
732	ERR ("'null' expected");	1395	ERR ("'null' expected");
733		1396
734	break;	1397	break;
735		1398
736	default:	1399	default:
737	ERR ("malformed json string");	1400	ERR ("malformed JSON string, neither array, object, number, string or atom");
738	break;	1401	break;
739	}	1402	}
740		1403
741	fail:	1404	fail:
742	return 0;	1405	return 0;
743	}	1406	}
744		1407
745	static SV *	1408	static SV *
746	decode_json (SV *string, UV flags)	1409	decode_json (SV *string, JSON *json, STRLEN *offset_return)
747	{	1410	{
		1411	dec_t dec;
		1412	STRLEN offset;
748	SV *sv;	1413	SV *sv;
749		1414
		1415	SvGETMAGIC (string);
		1416	SvUPGRADE (string, SVt_PV);
		1417
		1418	if (SvCUR (string) > json->max_size && json->max_size)
		1419	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1420	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1421
750	if (!(flags & F_UTF8))	1422	if (json->flags & F_UTF8)
		1423	sv_utf8_downgrade (string, 0);
		1424	else
751	sv_utf8_upgrade (string);	1425	sv_utf8_upgrade (string);
752		1426
753	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1427	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
754		1428
755	dec_t dec;	1429	dec.json = *json;
756	dec.flags = flags;
757	dec.cur = SvPVX (string);	1430	dec.cur = SvPVX (string);
758	dec.end = SvEND (string);	1431	dec.end = SvEND (string);
759	dec.err = 0;	1432	dec.err = 0;
		1433	dec.depth = 0;
760		1434
761	*dec.end = 1; // invalid anywhere	1435	if (dec.json.cb_object || dec.json.cb_sk_object)
		1436	dec.json.flags |= F_HOOK;
		1437
		1438	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1439
		1440	decode_ws (&dec);
762	sv = decode_sv (&dec);	1441	sv = decode_sv (&dec);
763	*dec.end = 0;	1442
		1443	if (!(offset_return || !sv))
		1444	{
		1445	// check for trailing garbage
		1446	decode_ws (&dec);
		1447
		1448	if (*dec.cur)
		1449	{
		1450	dec.err = "garbage after JSON object";
		1451	SvREFCNT_dec (sv);
		1452	sv = 0;
		1453	}
		1454	}
		1455
		1456	if (offset_return || !sv)
		1457	{
		1458	offset = dec.json.flags & F_UTF8
		1459	? dec.cur - SvPVX (string)
		1460	: utf8_distance (dec.cur, SvPVX (string));
		1461
		1462	if (offset_return)
		1463	*offset_return = offset;
		1464	}
764		1465
765	if (!sv)	1466	if (!sv)
766	{	1467	{
767	IV offset = utf8_distance (dec.cur, SvPVX (string));
768	SV *uni = sv_newmortal ();	1468	SV *uni = sv_newmortal ();
769		1469
		1470	// horrible hack to silence warning inside pv_uni_display
		1471	COP cop = *PL_curcop;
		1472	cop.cop_warnings = pWARN_NONE;
		1473	ENTER;
		1474	SAVEVPTR (PL_curcop);
		1475	PL_curcop = &cop;
770	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1476	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1477	LEAVE;
		1478
771	croak ("%s, at character %d (%s)",	1479	croak ("%s, at character offset %d [\"%s\"]",
772	dec.err,	1480	dec.err,
773	(int)offset,	1481	(int)offset,
774	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1482	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
775	}	1483	}
776		1484
777	sv_dump (sv);//D
778	return sv_2mortal (sv);	1485	sv = sv_2mortal (sv);
		1486
		1487	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
		1488	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
		1489
		1490	return sv;
779	}	1491	}
		1492
		1493	/////////////////////////////////////////////////////////////////////////////
		1494	// incremental parser
		1495
		1496	static void
		1497	incr_parse (JSON *self)
		1498	{
		1499	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1500
		1501	for (;;)
		1502	{
		1503	//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
		1504	switch (self->incr_mode)
		1505	{
		1506	// only used for intiial whitespace skipping
		1507	case INCR_M_WS:
		1508	for (;;)
		1509	{
		1510	if (*p > 0x20)
		1511	{
		1512	self->incr_mode = INCR_M_JSON;
		1513	goto incr_m_json;
		1514	}
		1515	else if (!*p)
		1516	goto interrupt;
		1517
		1518	++p;
		1519	}
		1520
		1521	// skip a single char inside a string (for \\-processing)
		1522	case INCR_M_BS:
		1523	if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	self->incr_mode = INCR_M_STR;
		1528	goto incr_m_str;
		1529
		1530	// inside a string
		1531	case INCR_M_STR:
		1532	incr_m_str:
		1533	for (;;)
		1534	{
		1535	if (*p == '"')
		1536	{
		1537	++p;
		1538	self->incr_mode = INCR_M_JSON;
		1539
		1540	if (!self->incr_nest)
		1541	goto interrupt;
		1542
		1543	goto incr_m_json;
		1544	}
		1545	else if (*p == '\\')
		1546	{
		1547	++p; // "virtually" consumes character after \
		1548
		1549	if (!*p) // if at end of string we have to switch modes
		1550	{
		1551	self->incr_mode = INCR_M_BS;
		1552	goto interrupt;
		1553	}
		1554	}
		1555	else if (!*p)
		1556	goto interrupt;
		1557
		1558	++p;
		1559	}
		1560
		1561	// after initial ws, outside string
		1562	case INCR_M_JSON:
		1563	incr_m_json:
		1564	for (;;)
		1565	{
		1566	switch (*p++)
		1567	{
		1568	case 0:
		1569	--p;
		1570	goto interrupt;
		1571
		1572	case 0x09:
		1573	case 0x0a:
		1574	case 0x0d:
		1575	case 0x20:
		1576	if (!self->incr_nest)
		1577	{
		1578	--p; // do not eat the whitespace, let the next round do it
		1579	goto interrupt;
		1580	}
		1581	break;
		1582
		1583	case '"':
		1584	self->incr_mode = INCR_M_STR;
		1585	goto incr_m_str;
		1586
		1587	case '[':
		1588	case '{':
		1589	if (++self->incr_nest > self->max_depth)
		1590	croak (ERR_NESTING_EXCEEDED);
		1591	break;
		1592
		1593	case ']':
		1594	case '}':
		1595	if (!--self->incr_nest)
		1596	goto interrupt;
		1597	}
		1598	}
		1599	}
		1600
		1601	modechange:
		1602	;
		1603	}
		1604
		1605	interrupt:
		1606	self->incr_pos = p - SvPVX (self->incr_text);
		1607	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1608	}
		1609
		1610	/////////////////////////////////////////////////////////////////////////////
		1611	// XS interface functions
780		1612
781	MODULE = JSON::XS PACKAGE = JSON::XS	1613	MODULE = JSON::XS PACKAGE = JSON::XS
782		1614
783	BOOT:	1615	BOOT:
784	{	1616	{
785	int i;	1617	int i;
786		1618
787	memset (decode_hexdigit, 0xff, 256);
788	for (i = 10; i--; )	1619	for (i = 0; i < 256; ++i)
789	decode_hexdigit ['0' + i] = i;	1620	decode_hexdigit [i] =
		1621	i >= '0' && i <= '9' ? i - '0'
		1622	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1623	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1624	: -1;
790		1625
791	for (i = 6; --i; )
792	{
793	decode_hexdigit ['a' + i] = 10 + i;
794	decode_hexdigit ['A' + i] = 10 + i;
795	}
796
797	json_stash = gv_stashpv ("JSON::XS", 1);	1626	json_stash = gv_stashpv ("JSON::XS" , 1);
798	}	1627	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
799		1628
800	SV *new (char *dummy)	1629	json_true = get_bool ("JSON::XS::true");
		1630	json_false = get_bool ("JSON::XS::false");
		1631	}
		1632
		1633	PROTOTYPES: DISABLE
		1634
		1635	void CLONE (...)
801	CODE:	1636	CODE:
802	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1637	json_stash = 0;
		1638	json_boolean_stash = 0;
		1639
		1640	void new (char *klass)
		1641	PPCODE:
		1642	{
		1643	SV *pv = NEWSV (0, sizeof (JSON));
		1644	SvPOK_only (pv);
		1645	json_init ((JSON *)SvPVX (pv));
		1646	XPUSHs (sv_2mortal (sv_bless (
		1647	newRV_noinc (pv),
		1648	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1649	)));
		1650	}
		1651
		1652	void ascii (JSON *self, int enable = 1)
		1653	ALIAS:
		1654	ascii = F_ASCII
		1655	latin1 = F_LATIN1
		1656	utf8 = F_UTF8
		1657	indent = F_INDENT
		1658	canonical = F_CANONICAL
		1659	space_before = F_SPACE_BEFORE
		1660	space_after = F_SPACE_AFTER
		1661	pretty = F_PRETTY
		1662	allow_nonref = F_ALLOW_NONREF
		1663	shrink = F_SHRINK
		1664	allow_blessed = F_ALLOW_BLESSED
		1665	convert_blessed = F_CONV_BLESSED
		1666	relaxed = F_RELAXED
		1667	allow_unknown = F_ALLOW_UNKNOWN
		1668	PPCODE:
		1669	{
		1670	if (enable)
		1671	self->flags |= ix;
		1672	else
		1673	self->flags &= ~ix;
		1674
		1675	XPUSHs (ST (0));
		1676	}
		1677
		1678	void get_ascii (JSON *self)
		1679	ALIAS:
		1680	get_ascii = F_ASCII
		1681	get_latin1 = F_LATIN1
		1682	get_utf8 = F_UTF8
		1683	get_indent = F_INDENT
		1684	get_canonical = F_CANONICAL
		1685	get_space_before = F_SPACE_BEFORE
		1686	get_space_after = F_SPACE_AFTER
		1687	get_allow_nonref = F_ALLOW_NONREF
		1688	get_shrink = F_SHRINK
		1689	get_allow_blessed = F_ALLOW_BLESSED
		1690	get_convert_blessed = F_CONV_BLESSED
		1691	get_relaxed = F_RELAXED
		1692	get_allow_unknown = F_ALLOW_UNKNOWN
		1693	PPCODE:
		1694	XPUSHs (boolSV (self->flags & ix));
		1695
		1696	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1697	PPCODE:
		1698	self->max_depth = max_depth;
		1699	XPUSHs (ST (0));
		1700
		1701	U32 get_max_depth (JSON *self)
		1702	CODE:
		1703	RETVAL = self->max_depth;
803	OUTPUT:	1704	OUTPUT:
804	RETVAL	1705	RETVAL
805		1706
806	SV *ascii (SV *self, int enable)	1707	void max_size (JSON *self, U32 max_size = 0)
807	ALIAS:	1708	PPCODE:
808	ascii = F_ASCII	1709	self->max_size = max_size;
809	utf8 = F_UTF8	1710	XPUSHs (ST (0));
810	indent = F_INDENT	1711
811	canonical = F_CANONICAL	1712	int get_max_size (JSON *self)
812	space_before = F_SPACE_BEFORE
813	space_after = F_SPACE_AFTER
814	json_rpc = F_JSON_RPC
815	pretty = F_PRETTY
816	CODE:	1713	CODE:
817	{	1714	RETVAL = self->max_size;
818	UV *uv = SvJSON (self);
819	if (enable)
820	*uv |= ix;
821	else
822	*uv &= ~ix;
823
824	RETVAL = newSVsv (self);
825	}
826	OUTPUT:	1715	OUTPUT:
827	RETVAL	1716	RETVAL
828		1717
829	void encode (SV *self, SV *scalar)	1718	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
830	PPCODE:	1719	PPCODE:
831	XPUSHs (encode_json (scalar, *SvJSON (self)));	1720	{
		1721	SvREFCNT_dec (self->cb_object);
		1722	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
832		1723
833	void decode (SV *self, SV *jsonstr)	1724	XPUSHs (ST (0));
		1725	}
		1726
		1727	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
834	PPCODE:	1728	PPCODE:
835	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1729	{
		1730	if (!self->cb_sk_object)
		1731	self->cb_sk_object = newHV ();
836		1732
837	void to_json (SV *scalar)	1733	if (SvOK (cb))
		1734	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1735	else
		1736	{
		1737	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1738
		1739	if (!HvKEYS (self->cb_sk_object))
		1740	{
		1741	SvREFCNT_dec (self->cb_sk_object);
		1742	self->cb_sk_object = 0;
		1743	}
		1744	}
		1745
		1746	XPUSHs (ST (0));
		1747	}
		1748
		1749	void encode (JSON *self, SV *scalar)
838	PPCODE:	1750	PPCODE:
839	XPUSHs (encode_json (scalar, F_UTF8));	1751	XPUSHs (encode_json (scalar, self));
840		1752
841	void from_json (SV *jsonstr)	1753	void decode (JSON *self, SV *jsonstr)
842	PPCODE:	1754	PPCODE:
843	XPUSHs (decode_json (jsonstr, F_UTF8));	1755	XPUSHs (decode_json (jsonstr, self, 0));
844		1756
		1757	void decode_prefix (JSON *self, SV *jsonstr)
		1758	PPCODE:
		1759	{
		1760	STRLEN offset;
		1761	EXTEND (SP, 2);
		1762	PUSHs (decode_json (jsonstr, self, &offset));
		1763	PUSHs (sv_2mortal (newSVuv (offset)));
		1764	}
		1765
		1766	void incr_parse (JSON *self, SV *jsonstr = 0)
		1767	PPCODE:
		1768	{
		1769	if (!self->incr_text)
		1770	self->incr_text = newSVpvn ("", 0);
		1771
		1772	// append data, if any
		1773	if (jsonstr)
		1774	{
		1775	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1776	{
		1777	/* utf-8-ness differs, need to upgrade */
		1778	sv_utf8_upgrade (self->incr_text);
		1779
		1780	if (self->incr_pos)
		1781	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1782	- (U8 *)SvPVX (self->incr_text);
		1783	}
		1784
		1785	{
		1786	STRLEN len;
		1787	const char *str = SvPV (jsonstr, len);
		1788	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1789	Move (str, SvEND (self->incr_text), len, char);
		1790	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1791	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1792	}
		1793	}
		1794
		1795	if (GIMME_V != G_VOID)
		1796	do
		1797	{
		1798	STRLEN offset;
		1799
		1800	if (!INCR_DONE (self))
		1801	{
		1802	incr_parse (self);
		1803
		1804	if (self->incr_pos > self->max_size && self->max_size)
		1805	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1806	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1807
		1808	if (!INCR_DONE (self))
		1809	break;
		1810	}
		1811
		1812	XPUSHs (decode_json (self->incr_text, self, &offset));
		1813
		1814	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1815	self->incr_pos -= offset;
		1816	self->incr_nest = 0;
		1817	self->incr_mode = 0;
		1818	}
		1819	while (GIMME_V == G_ARRAY);
		1820	}
		1821
		1822	SV *incr_text (JSON *self)
		1823	ATTRS: lvalue
		1824	CODE:
		1825	{
		1826	if (self->incr_pos)
		1827	croak ("incr_text can not be called when the incremental parser already started parsing");
		1828
		1829	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1830	}
		1831	OUTPUT:
		1832	RETVAL
		1833
		1834	void incr_skip (JSON *self)
		1835	CODE:
		1836	{
		1837	if (self->incr_pos)
		1838	{
		1839	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1840	self->incr_pos = 0;
		1841	self->incr_nest = 0;
		1842	self->incr_mode = 0;
		1843	}
		1844	}
		1845
		1846	void DESTROY (JSON *self)
		1847	CODE:
		1848	SvREFCNT_dec (self->cb_sk_object);
		1849	SvREFCNT_dec (self->cb_object);
		1850	SvREFCNT_dec (self->incr_text);
		1851
		1852	PROTOTYPES: ENABLE
		1853
		1854	void encode_json (SV *scalar)
		1855	ALIAS:
		1856	to_json_ = 0
		1857	encode_json = F_UTF8
		1858	PPCODE:
		1859	{
		1860	JSON json;
		1861	json_init (&json);
		1862	json.flags |= ix;
		1863	XPUSHs (encode_json (scalar, &json));
		1864	}
		1865
		1866	void decode_json (SV *jsonstr)
		1867	ALIAS:
		1868	from_json_ = 0
		1869	decode_json = F_UTF8
		1870	PPCODE:
		1871	{
		1872	JSON json;
		1873	json_init (&json);
		1874	json.flags |= ix;
		1875	XPUSHs (decode_json (jsonstr, &json, 0));
		1876	}
		1877
		1878

Diff Legend

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