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 1.45 by root, Mon Jun 25 06:57:42 2007 UTC

…		…
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"
8		9
		10	#if defined(__BORLANDC__) || defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040	27	#define F_ALLOW_NONREF 0x00000100UL
		28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35
		36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
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	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
19		41
20	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
21		46
22	#define SB do {	47	#define SB do {
23	#define SE } while (0)	48	#define SE } while (0)
24		49
25	static HV *json_stash;	50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr),(value))
		52	# define inline inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define inline static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		62	static SV *json_true, *json_false;
		63
		64	/////////////////////////////////////////////////////////////////////////////
		65	// utility functions
		66
		67	static UV *
		68	SvJSON (SV *sv)
		69	{
		70	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		71	croak ("object is not of type JSON::XS");
		72
		73	return &SvUVX (SvRV (sv));
		74	}
		75
		76	static void
		77	shrink (SV *sv)
		78	{
		79	sv_utf8_downgrade (sv, 1);
		80	if (SvLEN (sv) > SvCUR (sv) + 1)
		81	{
		82	#ifdef SvPV_shrink_to_cur
		83	SvPV_shrink_to_cur (sv);
		84	#elif defined (SvPV_renew)
		85	SvPV_renew (sv, SvCUR (sv) + 1);
		86	#endif
		87	}
		88	}
		89
		90	// decode an utf-8 character and return it, or (UV)-1 in
		91	// case of an error.
		92	// we special-case "safe" characters from U+80 .. U+7FF,
		93	// but use the very good perl function to parse anything else.
		94	// note that we never call this function for a ascii codepoints
		95	inline UV
		96	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		97	{
		98	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		99	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		100	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		101	{
		102	*clen = 2;
		103	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		104	}
		105	else
		106	{
		107	*clen = (STRLEN)-1;
		108	return (UV)-1;
		109	}
		110	}
		111
		112	/////////////////////////////////////////////////////////////////////////////
		113	// encoder
26		114
27	// structure used for encoding JSON	115	// structure used for encoding JSON
28	typedef struct	116	typedef struct
29	{	117	{
30	char *cur;	118	char *cur; // SvPVX (sv) + current output position
31	STRLEN len; // SvLEN (sv)
32	char *end; // SvEND (sv)	119	char *end; // SvEND (sv)
33	SV *sv;	120	SV *sv; // result scalar
34	UV flags;	121	U32 flags; // F_*
35	int max_recurse;	122	U32 indent; // indentation level
36	int indent;	123	U32 maxdepth; // max. indentation/recursion level
37	} enc_t;	124	} enc_t;
38		125
39	// structure used for decoding JSON	126	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)	127	need (enc_t *enc, STRLEN len)
61	{	128	{
62	if (enc->cur + len >= enc->end)	129	if (expect_false (enc->cur + len >= enc->end))
63	{	130	{
64	STRLEN cur = enc->cur - SvPVX (enc->sv);	131	STRLEN cur = enc->cur - SvPVX (enc->sv);
65	SvGROW (enc->sv, cur + len + 1);	132	SvGROW (enc->sv, cur + len + 1);
66	enc->cur = SvPVX (enc->sv) + cur;	133	enc->cur = SvPVX (enc->sv) + cur;
67	enc->end = SvEND (enc->sv);	134	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
68	}	135	}
69	}	136	}
70		137
71	static void	138	inline void
72	encode_ch (enc_t *enc, char ch)	139	encode_ch (enc_t *enc, char ch)
73	{	140	{
74	need (enc, 1);	141	need (enc, 1);
75	*enc->cur++ = ch;	142	*enc->cur++ = ch;
76	}	143	}
…		…
78	static void	145	static void
79	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	146	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
80	{	147	{
81	char *end = str + len;	148	char *end = str + len;
82		149
		150	need (enc, len);
		151
83	while (str < end)	152	while (str < end)
84	{	153	{
85	unsigned char ch = *(unsigned char *)str;	154	unsigned char ch = *(unsigned char *)str;
		155
86	if (ch >= 0x20 && ch < 0x80) // most common case	156	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
87	{	157	{
		158	if (expect_false (ch == '"')) // but with slow exceptions
		159	{
		160	need (enc, len += 1);
		161	*enc->cur++ = '\\';
		162	*enc->cur++ = '"';
		163	}
		164	else if (expect_false (ch == '\\'))
		165	{
		166	need (enc, len += 1);
		167	*enc->cur++ = '\\';
		168	*enc->cur++ = '\\';
		169	}
		170	else
88	*enc->cur++ = ch;	171	*enc->cur++ = ch;
		172
89	str++;	173	++str;
90	}	174	}
91	else	175	else
92	{	176	{
93	STRLEN clen;	177	switch (ch)
94	UV uch;
95
96	if (is_utf8)
97	{	178	{
98	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	179	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
99	if (clen < 0)	180	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");	181	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
101	}	182	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
102	else	183	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
103	{
104	uch = ch;
105	clen = 1;
106	}
107		184
108	need (enc, len += 6);	185	default:
109
110	if (uch < 0xa0 || enc->flags & F_ASCII)
111	{
112	if (uch > 0xFFFFUL)
113	{	186	{
		187	STRLEN clen;
		188	UV uch;
		189
		190	if (is_utf8)
		191	{
		192	uch = decode_utf8 (str, end - str, &clen);
		193	if (clen == (STRLEN)-1)
		194	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		195	}
		196	else
		197	{
		198	uch = ch;
114	len += 6;	199	clen = 1;
		200	}
		201
		202	if (uch > 0x10FFFFUL)
		203	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		204
		205	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
		206	{
		207	if (uch > 0xFFFFUL)
		208	{
115	need (enc, len += 6);	209	need (enc, len += 11);
116	sprintf (enc->cur, "\\u%04x\\u%04x",	210	sprintf (enc->cur, "\\u%04x\\u%04x",
117	(uch - 0x10000) / 0x400 + 0xD800,	211	(int)((uch - 0x10000) / 0x400 + 0xD800),
118	(uch - 0x10000) % 0x400 + 0xDC00);	212	(int)((uch - 0x10000) % 0x400 + 0xDC00));
119	enc->cur += 12;	213	enc->cur += 12;
		214	}
		215	else
		216	{
		217	static char hexdigit [16] = "0123456789abcdef";
		218	need (enc, len += 5);
		219	*enc->cur++ = '\\';
		220	*enc->cur++ = 'u';
		221	*enc->cur++ = hexdigit [ uch >> 12 ];
		222	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		223	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		224	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		225	}
		226
		227	str += clen;
120	}	228	}
		229	else if (enc->flags & F_LATIN1)
		230	{
		231	*enc->cur++ = uch;
		232	str += clen;
		233	}
		234	else if (is_utf8)
		235	{
		236	need (enc, len += clen);
		237	do
		238	{
		239	*enc->cur++ = *str++;
		240	}
		241	while (--clen);
		242	}
121	else	243	else
122	{	244	{
123	sprintf (enc->cur, "\\u%04x", uch);	245	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
124	enc->cur += 6;	246	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		247	++str;
		248	}
125	}	249	}
126	}	250	}
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	}	251	}
137		252
138	--len;	253	--len;
139	}	254	}
140	}	255	}
141		256
142	#define INDENT SB \	257	inline void
		258	encode_indent (enc_t *enc)
		259	{
143	if (enc->flags & F_INDENT) \	260	if (enc->flags & F_INDENT)
144	{ \	261	{
145	int i_; \	262	int spaces = enc->indent * INDENT_STEP;
146	need (enc, enc->indent); \
147	for (i_ = enc->indent * 3; i_--; )\
148	encode_ch (enc, ' '); \
149	} \
150	SE
151		263
152	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	264	need (enc, spaces);
153	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	265	memset (enc->cur, ' ', spaces);
154	#define COMMA SB \	266	enc->cur += spaces;
		267	}
		268	}
		269
		270	inline void
		271	encode_space (enc_t *enc)
		272	{
		273	need (enc, 1);
155	encode_ch (enc, ','); \	274	encode_ch (enc, ' ');
		275	}
		276
		277	inline void
		278	encode_nl (enc_t *enc)
		279	{
156	if (enc->flags & F_INDENT) \	280	if (enc->flags & F_INDENT)
157	NL; \	281	{
		282	need (enc, 1);
		283	encode_ch (enc, '\n');
		284	}
		285	}
		286
		287	inline void
		288	encode_comma (enc_t *enc)
		289	{
		290	encode_ch (enc, ',');
		291
		292	if (enc->flags & F_INDENT)
		293	encode_nl (enc);
158	else if (enc->flags & F_SPACE_AFTER) \	294	else if (enc->flags & F_SPACE_AFTER)
159	SPACE; \	295	encode_space (enc);
160	SE	296	}
161		297
162	static void encode_sv (enc_t *enc, SV *sv);	298	static void encode_sv (enc_t *enc, SV *sv);
163		299
164	static void	300	static void
165	encode_av (enc_t *enc, AV *av)	301	encode_av (enc_t *enc, AV *av)
166	{	302	{
167	int i, len = av_len (av);	303	int i, len = av_len (av);
168		304
169	encode_ch (enc, '['); NL;	305	if (enc->indent >= enc->maxdepth)
		306	croak ("data structure too deep (hit recursion limit)");
		307
		308	encode_ch (enc, '['); encode_nl (enc);
170	++enc->indent;	309	++enc->indent;
171		310
172	for (i = 0; i <= len; ++i)	311	for (i = 0; i <= len; ++i)
173	{	312	{
174	INDENT;	313	encode_indent (enc);
175	encode_sv (enc, *av_fetch (av, i, 0));	314	encode_sv (enc, *av_fetch (av, i, 0));
176		315
177	if (i < len)	316	if (i < len)
178	COMMA;	317	encode_comma (enc);
179	}	318	}
180		319
181	NL;	320	encode_nl (enc);
182		321
183	--enc->indent;	322	--enc->indent;
184	INDENT; encode_ch (enc, ']');	323	encode_indent (enc); encode_ch (enc, ']');
185	}	324	}
186		325
187	static void	326	static void
188	encode_he (enc_t *enc, HE *he)	327	encode_he (enc_t *enc, HE *he)
189	{	328	{
…		…
191		330
192	if (HeKLEN (he) == HEf_SVKEY)	331	if (HeKLEN (he) == HEf_SVKEY)
193	{	332	{
194	SV *sv = HeSVKEY (he);	333	SV *sv = HeSVKEY (he);
195	STRLEN len;	334	STRLEN len;
		335	char *str;
		336
		337	SvGETMAGIC (sv);
196	char *str = SvPV (sv, len);	338	str = SvPV (sv, len);
197		339
198	encode_str (enc, str, len, SvUTF8 (sv));	340	encode_str (enc, str, len, SvUTF8 (sv));
199	}	341	}
200	else	342	else
201	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	343	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
202		344
203	encode_ch (enc, '"');	345	encode_ch (enc, '"');
204		346
205	if (enc->flags & F_SPACE_BEFORE) SPACE;	347	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
206	encode_ch (enc, ':');	348	encode_ch (enc, ':');
207	if (enc->flags & F_SPACE_AFTER ) SPACE;	349	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
208	encode_sv (enc, HeVAL (he));	350	encode_sv (enc, HeVAL (he));
209	}	351	}
210		352
211	// compare hash entries, used when all keys are bytestrings	353	// compare hash entries, used when all keys are bytestrings
212	static int	354	static int
…		…
218	HE *b = *(HE **)b_;	360	HE *b = *(HE **)b_;
219		361
220	STRLEN la = HeKLEN (a);	362	STRLEN la = HeKLEN (a);
221	STRLEN lb = HeKLEN (b);	363	STRLEN lb = HeKLEN (b);
222		364
223	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	365	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
224	cmp = la < lb ? -1 : la == lb ? 0 : 1;	366	cmp = la - lb;
225		367
226	return cmp;	368	return cmp;
227	}	369	}
228		370
229	// compare hash entries, used when some keys are sv's or utf-x	371	// compare hash entries, used when some keys are sv's or utf-x
…		…
236	static void	378	static void
237	encode_hv (enc_t *enc, HV *hv)	379	encode_hv (enc_t *enc, HV *hv)
238	{	380	{
239	int count, i;	381	int count, i;
240		382
		383	if (enc->indent >= enc->maxdepth)
		384	croak ("data structure too deep (hit recursion limit)");
		385
241	encode_ch (enc, '{'); NL; ++enc->indent;	386	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
242		387
243	if ((count = hv_iterinit (hv)))	388	if ((count = hv_iterinit (hv)))
244	{	389	{
245	// for canonical output we have to sort by keys first	390	// for canonical output we have to sort by keys first
246	// actually, this is mostly due to the stupid so-called	391	// actually, this is mostly due to the stupid so-called
247	// security workaround added somewhere in 5.8.x.	392	// security workaround added somewhere in 5.8.x.
248	// that randomises hash orderings	393	// that randomises hash orderings
249	if (enc->flags & F_CANONICAL)	394	if (enc->flags & F_CANONICAL)
250	{	395	{
251	HE *he, *hes [count];
252	int fast = 1;	396	int fast = 1;
		397	HE *he;
		398	#if defined(__BORLANDC__) || defined(_MSC_VER)
		399	HE **hes = _alloca (count * sizeof (HE));
		400	#else
		401	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		402	#endif
253		403
254	i = 0;	404	i = 0;
255	while ((he = hv_iternext (hv)))	405	while ((he = hv_iternext (hv)))
256	{	406	{
257	hes [i++] = he;	407	hes [i++] = he;
…		…
263		413
264	if (fast)	414	if (fast)
265	qsort (hes, count, sizeof (HE *), he_cmp_fast);	415	qsort (hes, count, sizeof (HE *), he_cmp_fast);
266	else	416	else
267	{	417	{
268	// hack to disable "use bytes"	418	// hack to forcefully disable "use bytes"
269	COP *oldcop = PL_curcop, cop;	419	COP cop = *PL_curcop;
270	cop.op_private = 0;	420	cop.op_private = 0;
		421
		422	ENTER;
		423	SAVETMPS;
		424
		425	SAVEVPTR (PL_curcop);
271	PL_curcop = &cop;	426	PL_curcop = &cop;
272		427
273	SAVETMPS;
274	qsort (hes, count, sizeof (HE *), he_cmp_slow);	428	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		429
275	FREETMPS;	430	FREETMPS;
276		431	LEAVE;
277	PL_curcop = oldcop;
278	}	432	}
279		433
280	for (i = 0; i < count; ++i)	434	for (i = 0; i < count; ++i)
281	{	435	{
282	INDENT;	436	encode_indent (enc);
283	encode_he (enc, hes [i]);	437	encode_he (enc, hes [i]);
284		438
285	if (i < count - 1)	439	if (i < count - 1)
286	COMMA;	440	encode_comma (enc);
287	}	441	}
288		442
289	NL;	443	encode_nl (enc);
290	}	444	}
291	else	445	else
292	{	446	{
293	SV *sv;
294	HE *he = hv_iternext (hv);	447	HE *he = hv_iternext (hv);
295		448
296	for (;;)	449	for (;;)
297	{	450	{
298	INDENT;	451	encode_indent (enc);
299	encode_he (enc, he);	452	encode_he (enc, he);
300		453
301	if (!(he = hv_iternext (hv)))	454	if (!(he = hv_iternext (hv)))
302	break;	455	break;
303		456
304	COMMA;	457	encode_comma (enc);
305	}	458	}
306		459
307	NL;	460	encode_nl (enc);
308	}	461	}
309	}	462	}
310		463
311	--enc->indent; INDENT; encode_ch (enc, '}');	464	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		465	}
		466
		467	// encode objects, arrays and special \0=false and \1=true values.
		468	static void
		469	encode_rv (enc_t *enc, SV *sv)
		470	{
		471	svtype svt;
		472
		473	SvGETMAGIC (sv);
		474	svt = SvTYPE (sv);
		475
		476	if (expect_false (SvOBJECT (sv)))
		477	{
		478	if (SvSTASH (sv) == json_boolean_stash)
		479	{
		480	if (SvIV (sv) == 0)
		481	encode_str (enc, "false", 5, 0);
		482	else
		483	encode_str (enc, "true", 4, 0);
		484	}
		485	else
		486	{
		487	#if 0
		488	if (0 && sv_derived_from (rv, "JSON::Literal"))
		489	{
		490	// not yet
		491	}
		492	#endif
		493	if (enc->flags & F_CONV_BLESSED)
		494	{
		495	// we re-bless the reference to get overload and other niceties right
		496	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		497
		498	if (to_json)
		499	{
		500	dSP;
		501	ENTER;
		502	SAVETMPS;
		503	PUSHMARK (SP);
		504	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		505
		506	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		507	// check anyways.
		508	PUTBACK;
		509	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		510	SPAGAIN;
		511
		512	encode_sv (enc, POPs);
		513
		514	FREETMPS;
		515	LEAVE;
		516	}
		517	else if (enc->flags & F_ALLOW_BLESSED)
		518	encode_str (enc, "null", 4, 0);
		519	else
		520	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		521	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		522	}
		523	else if (enc->flags & F_ALLOW_BLESSED)
		524	encode_str (enc, "null", 4, 0);
		525	else
		526	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		527	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		528	}
		529	}
		530	else if (svt == SVt_PVHV)
		531	encode_hv (enc, (HV *)sv);
		532	else if (svt == SVt_PVAV)
		533	encode_av (enc, (AV *)sv);
		534	else if (svt < SVt_PVAV)
		535	{
		536	if (SvNIOK (sv) && SvIV (sv) == 0)
		537	encode_str (enc, "false", 5, 0);
		538	else if (SvNIOK (sv) && SvIV (sv) == 1)
		539	encode_str (enc, "true", 4, 0);
		540	else
		541	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else
		545	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
312	}	547	}
313		548
314	static void	549	static void
315	encode_sv (enc_t *enc, SV *sv)	550	encode_sv (enc_t *enc, SV *sv)
316	{	551	{
		552	SvGETMAGIC (sv);
		553
317	if (SvPOKp (sv))	554	if (SvPOKp (sv))
318	{	555	{
319	STRLEN len;	556	STRLEN len;
320	char *str = SvPV (sv, len);	557	char *str = SvPV (sv, len);
321	encode_ch (enc, '"');	558	encode_ch (enc, '"');
322	encode_str (enc, str, len, SvUTF8 (sv));	559	encode_str (enc, str, len, SvUTF8 (sv));
323	encode_ch (enc, '"');	560	encode_ch (enc, '"');
324	}	561	}
325	else if (SvNOKp (sv))	562	else if (SvNOKp (sv))
326	{	563	{
		564	// trust that perl will do the right thing w.r.t. JSON syntax.
327	need (enc, NV_DIG + 32);	565	need (enc, NV_DIG + 32);
328	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	566	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
329	enc->cur += strlen (enc->cur);	567	enc->cur += strlen (enc->cur);
330	}	568	}
331	else if (SvIOKp (sv))	569	else if (SvIOKp (sv))
332	{	570	{
333	need (enc, 64);	571	// we assume we can always read an IV as a UV
		572	if (SvUV (sv) & ~(UV)0x7fff)
		573	{
		574	// large integer, use the (rather slow) snprintf way.
		575	need (enc, sizeof (UV) * 3);
334	enc->cur +=	576	enc->cur +=
335	SvIsUV(sv)	577	SvIsUV(sv)
336	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	578	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
337	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	579	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		580	}
		581	else
		582	{
		583	// optimise the "small number case"
		584	// code will likely be branchless and use only a single multiplication
		585	I32 i = SvIV (sv);
		586	U32 u;
		587	char digit, nz = 0;
		588
		589	need (enc, 6);
		590
		591	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		592	u = i < 0 ? -i : i;
		593
		594	// convert to 4.28 fixed-point representation
		595	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		596
		597	// now output digit by digit, each time masking out the integer part
		598	// and multiplying by 5 while moving the decimal point one to the right,
		599	// resulting in a net multiplication by 10.
		600	// we always write the digit to memory but conditionally increment
		601	// the pointer, to ease the usage of conditional move instructions.
		602	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		603	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		604	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		605	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		606	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		607	}
338	}	608	}
339	else if (SvROK (sv))	609	else if (SvROK (sv))
340	{	610	encode_rv (enc, SvRV (sv));
341	if (!--enc->max_recurse)
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))	611	else if (!SvOK (sv))
356	encode_str (enc, "null", 4, 0);	612	encode_str (enc, "null", 4, 0);
357	else	613	else
358	croak ("encountered perl type that JSON cannot handle");	614	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		615	SvPV_nolen (sv), SvFLAGS (sv));
359	}	616	}
360		617
361	static SV *	618	static SV *
362	encode_json (SV *scalar, UV flags)	619	encode_json (SV *scalar, U32 flags)
363	{	620	{
364	enc_t enc;	621	enc_t enc;
		622
		623	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
		624	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		625
365	enc.flags = flags;	626	enc.flags = flags;
366	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	627	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
367	enc.cur = SvPVX (enc.sv);	628	enc.cur = SvPVX (enc.sv);
368	enc.end = SvEND (enc.sv);	629	enc.end = SvEND (enc.sv);
369	enc.max_recurse = 0;
370	enc.indent = 0;	630	enc.indent = 0;
		631	enc.maxdepth = DEC_DEPTH (flags);
371		632
372	SvPOK_only (enc.sv);	633	SvPOK_only (enc.sv);
373	encode_sv (&enc, scalar);	634	encode_sv (&enc, scalar);
374		635
		636	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		637	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		638
375	if (!(flags & (F_ASCII | F_UTF8)))	639	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
376	SvUTF8_on (enc.sv);	640	SvUTF8_on (enc.sv);
377		641
378	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	642	if (enc.flags & F_SHRINK)
		643	shrink (enc.sv);
		644
379	return enc.sv;	645	return enc.sv;
380	}	646	}
381		647
382	/////////////////////////////////////////////////////////////////////////////	648	/////////////////////////////////////////////////////////////////////////////
		649	// decoder
383		650
384	#define WS \	651	// structure used for decoding JSON
		652	typedef struct
		653	{
		654	char *cur; // current parser pointer
		655	char *end; // end of input string
		656	const char *err; // parse error, if != 0
		657	U32 flags; // F_*
		658	U32 depth; // recursion depth
		659	U32 maxdepth; // recursion depth limit
		660	} dec_t;
		661
		662	inline void
		663	decode_ws (dec_t *dec)
		664	{
385	for (;;) \	665	for (;;)
386	{ \	666	{
387	char ch = *dec->cur; \	667	char ch = *dec->cur;
		668
388	if (ch > 0x20 \	669	if (ch > 0x20
389	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	670	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
390	break; \	671	break;
		672
391	++dec->cur; \	673	++dec->cur;
392	}	674	}
		675	}
393		676
394	#define ERR(reason) SB dec->err = reason; goto fail; SE	677	#define ERR(reason) SB dec->err = reason; goto fail; SE
		678
395	#define EXPECT_CH(ch) SB \	679	#define EXPECT_CH(ch) SB \
396	if (*dec->cur != ch) \	680	if (*dec->cur != ch) \
397	ERR (# ch " expected"); \	681	ERR (# ch " expected"); \
398	++dec->cur; \	682	++dec->cur; \
399	SE	683	SE
400		684
		685	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		686	#define DEC_DEC_DEPTH --dec->depth
		687
401	static SV *decode_sv (dec_t *dec);	688	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		689
408	static signed char decode_hexdigit[256];	690	static signed char decode_hexdigit[256];
409		691
410	static UV	692	static UV
411	decode_4hex (dec_t *dec)	693	decode_4hex (dec_t *dec)
412	{	694	{
413	signed char d1, d2, d3, d4;	695	signed char d1, d2, d3, d4;
		696	unsigned char *cur = (unsigned char *)dec->cur;
414		697
415	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	698	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");	699	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]];	700	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");	701	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		702
424	dec->cur += 4;	703	dec->cur += 4;
425		704
426	return ((UV)d1) << 12	705	return ((UV)d1) << 12
427	| ((UV)d2) << 8	706	| ((UV)d2) << 8
…		…
433	}	712	}
434		713
435	static SV *	714	static SV *
436	decode_str (dec_t *dec)	715	decode_str (dec_t *dec)
437	{	716	{
438	SV *sv = NEWSV (0,2);	717	SV *sv = 0;
439	STRLEN cur = 0;
440	int utf8 = 0;	718	int utf8 = 0;
		719	char *dec_cur = dec->cur;
441		720
442	for (;;)	721	do
443	{	722	{
444	unsigned char ch = *(unsigned char *)dec->cur;	723	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		724	char *cur = buf;
445		725
446	if (ch == '"')	726	do
447	break;
448	else if (ch == '\\')
449	{	727	{
450	switch (*++dec->cur)	728	unsigned char ch = *(unsigned char *)dec_cur++;
		729
		730	if (expect_false (ch == '"'))
451	{	731	{
452	case '\\':	732	--dec_cur;
453	case '/':	733	break;
454	case '"': APPEND_CH (*dec->cur++); break;	734	}
455		735	else if (expect_false (ch == '\\'))
456	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	736	{
457	case 't': APPEND_CH ('\011'); ++dec->cur; break;	737	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	{	738	{
464	UV lo, hi;	739	case '\\':
465	++dec->cur;	740	case '/':
		741	case '"': *cur++ = *dec_cur++; break;
466		742
467	hi = decode_4hex (dec);	743	case 'b': ++dec_cur; *cur++ = '\010'; break;
468	if (hi == (UV)-1)	744	case 't': ++dec_cur; *cur++ = '\011'; break;
469	goto fail;	745	case 'n': ++dec_cur; *cur++ = '\012'; break;
		746	case 'f': ++dec_cur; *cur++ = '\014'; break;
		747	case 'r': ++dec_cur; *cur++ = '\015'; break;
470		748
471	// possibly a surrogate pair	749	case 'u':
472	if (hi >= 0xd800 && hi < 0xdc00)
473	{	750	{
474	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	751	UV lo, hi;
475	ERR ("illegal surrogate character");	752	++dec_cur;
476		753
477	dec->cur += 2;	754	dec->cur = dec_cur;
478
479	lo = decode_4hex (dec);	755	hi = decode_4hex (dec);
		756	dec_cur = dec->cur;
480	if (lo == (UV)-1)	757	if (hi == (UV)-1)
481	goto fail;	758	goto fail;
482		759
		760	// possibly a surrogate pair
		761	if (hi >= 0xd800)
		762	if (hi < 0xdc00)
		763	{
		764	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		765	ERR ("missing low surrogate character in surrogate pair");
		766
		767	dec_cur += 2;
		768
		769	dec->cur = dec_cur;
		770	lo = decode_4hex (dec);
		771	dec_cur = dec->cur;
		772	if (lo == (UV)-1)
		773	goto fail;
		774
483	if (lo < 0xdc00 || lo >= 0xe000)	775	if (lo < 0xdc00 || lo >= 0xe000)
484	ERR ("surrogate pair expected");	776	ERR ("surrogate pair expected");
485		777
486	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	778	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		779	}
		780	else if (hi < 0xe000)
		781	ERR ("missing high surrogate character in surrogate pair");
		782
		783	if (hi >= 0x80)
		784	{
		785	utf8 = 1;
		786
		787	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		788	}
		789	else
		790	*cur++ = hi;
487	}	791	}
488	else if (lo >= 0xdc00 && lo < 0xe000)
489	ERR ("illegal surrogate character");
490
491	if (hi >= 0x80)
492	{	792	break;
493	utf8 = 1;
494		793
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	794	default:
499	APPEND_CH (hi);	795	--dec_cur;
		796	ERR ("illegal backslash escape sequence in string");
500	}	797	}
		798	}
		799	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		800	*cur++ = ch;
		801	else if (ch >= 0x80)
		802	{
		803	STRLEN clen;
		804	UV uch;
		805
		806	--dec_cur;
		807
		808	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		809	if (clen == (STRLEN)-1)
		810	ERR ("malformed UTF-8 character in JSON string");
		811
		812	do
		813	*cur++ = *dec_cur++;
		814	while (--clen);
		815
		816	utf8 = 1;
		817	}
		818	else
		819	{
		820	--dec_cur;
		821
		822	if (!ch)
		823	ERR ("unexpected end of string while parsing JSON string");
501	break;	824	else
		825	ERR ("invalid character encountered while parsing JSON string");
502	}	826	}
503	}	827	}
504	else if (ch >= 0x20 && ch <= 0x7f)	828	while (cur < buf + SHORT_STRING_LEN);
505	APPEND_CH (*dec->cur++);	829
506	else if (ch >= 0x80)	830	{
		831	STRLEN len = cur - buf;
		832
		833	if (sv)
507	{	834	{
508	STRLEN clen;	835	SvGROW (sv, SvCUR (sv) + len + 1);
509	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	836	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
510	if (clen < 0)	837	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	}	838	}
517	else	839	else
518	ERR ("invalid character encountered");	840	sv = newSVpvn (buf, len);
519	}	841	}
		842	}
		843	while (*dec_cur != '"');
520		844
521	++dec->cur;	845	++dec_cur;
522		846
		847	if (sv)
		848	{
523	SvPOK_only (sv);	849	SvPOK_only (sv);
524
525	SvCUR_set (sv, cur);
526	*SvEND (sv) = 0;	850	*SvEND (sv) = 0;
527		851
528	if (utf8)	852	if (utf8)
529	SvUTF8_on (sv);	853	SvUTF8_on (sv);
		854	}
		855	else
		856	sv = newSVpvn ("", 0);
530		857
		858	dec->cur = dec_cur;
531	return sv;	859	return sv;
532		860
533	fail:	861	fail:
534	SvREFCNT_dec (sv);	862	dec->cur = dec_cur;
535	return 0;	863	return 0;
536	}	864	}
537		865
538	static SV *	866	static SV *
539	decode_num (dec_t *dec)	867	decode_num (dec_t *dec)
…		…
549	{	877	{
550	++dec->cur;	878	++dec->cur;
551	if (*dec->cur >= '0' && *dec->cur <= '9')	879	if (*dec->cur >= '0' && *dec->cur <= '9')
552	ERR ("malformed number (leading zero must not be followed by another digit)");	880	ERR ("malformed number (leading zero must not be followed by another digit)");
553	}	881	}
554		882	else if (*dec->cur < '0' || *dec->cur > '9')
555	// int	883	ERR ("malformed number (no digits after initial minus)");
		884	else
		885	do
		886	{
		887	++dec->cur;
		888	}
556	while (*dec->cur >= '0' && *dec->cur <= '9')	889	while (*dec->cur >= '0' && *dec->cur <= '9');
557	++dec->cur;
558		890
559	// [frac]	891	// [frac]
560	if (*dec->cur == '.')	892	if (*dec->cur == '.')
561	{	893	{
562	is_nv = 1;	894	++dec->cur;
		895
		896	if (*dec->cur < '0' || *dec->cur > '9')
		897	ERR ("malformed number (no digits after decimal point)");
563		898
564	do	899	do
565	{	900	{
566	++dec->cur;	901	++dec->cur;
567	}	902	}
568	while (*dec->cur >= '0' && *dec->cur <= '9');	903	while (*dec->cur >= '0' && *dec->cur <= '9');
		904
		905	is_nv = 1;
569	}	906	}
570		907
571	// [exp]	908	// [exp]
572	if (*dec->cur == 'e' || *dec->cur == 'E')	909	if (*dec->cur == 'e' || *dec->cur == 'E')
573	{	910	{
574	is_nv = 1;
575
576	++dec->cur;	911	++dec->cur;
		912
577	if (*dec->cur == '-' || *dec->cur == '+')	913	if (*dec->cur == '-' || *dec->cur == '+')
578	++dec->cur;	914	++dec->cur;
579		915
		916	if (*dec->cur < '0' || *dec->cur > '9')
		917	ERR ("malformed number (no digits after exp sign)");
		918
		919	do
		920	{
		921	++dec->cur;
		922	}
580	while (*dec->cur >= '0' && *dec->cur <= '9')	923	while (*dec->cur >= '0' && *dec->cur <= '9');
581	++dec->cur;	924
		925	is_nv = 1;
582	}	926	}
583		927
584	if (!is_nv)	928	if (!is_nv)
585	{	929	{
586	UV uv;	930	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
587	int numtype = grok_number (start, dec->cur - start, &uv);	931	if (*start == '-')
588	if (numtype & IS_NUMBER_IN_UV)	932	switch (dec->cur - start)
589	if (numtype & IS_NUMBER_NEG)
590	{	933	{
591	if (uv < (UV)IV_MIN)	934	case 2: return newSViv (-( start [1] - '0' * 1));
592	return newSViv (-(IV)uv);	935	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		936	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		937	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
593	}	938	}
		939	else
		940	switch (dec->cur - start)
		941	{
		942	case 1: return newSViv ( start [0] - '0' * 1);
		943	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		944	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		945	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		946	}
		947
		948	{
		949	UV uv;
		950	int numtype = grok_number (start, dec->cur - start, &uv);
		951	if (numtype & IS_NUMBER_IN_UV)
		952	if (numtype & IS_NUMBER_NEG)
		953	{
		954	if (uv < (UV)IV_MIN)
		955	return newSViv (-(IV)uv);
		956	}
594	else	957	else
595	return newSVuv (uv);	958	return newSVuv (uv);
		959
		960	// here would likely be the place for bigint support
596	}	961	}
		962	}
597		963
		964	// if we ever support bigint or bigfloat, this is the place for bigfloat
598	return newSVnv (Atof (start));	965	return newSVnv (Atof (start));
599		966
600	fail:	967	fail:
601	return 0;	968	return 0;
602	}	969	}
…		…
604	static SV *	971	static SV *
605	decode_av (dec_t *dec)	972	decode_av (dec_t *dec)
606	{	973	{
607	AV *av = newAV ();	974	AV *av = newAV ();
608		975
		976	DEC_INC_DEPTH;
		977	decode_ws (dec);
		978
		979	if (*dec->cur == ']')
		980	++dec->cur;
		981	else
609	for (;;)	982	for (;;)
610	{	983	{
611	SV *value;	984	SV *value;
612		985
613	value = decode_sv (dec);	986	value = decode_sv (dec);
614	if (!value)	987	if (!value)
615	goto fail;	988	goto fail;
616		989
617	av_push (av, value);	990	av_push (av, value);
618		991
619	WS;	992	decode_ws (dec);
620		993
621	if (*dec->cur == ']')	994	if (*dec->cur == ']')
622	{	995	{
623	++dec->cur;	996	++dec->cur;
624	break;	997	break;
		998	}
625	}	999
626
627	if (*dec->cur != ',')	1000	if (*dec->cur != ',')
628	ERR (", or ] expected while parsing array");	1001	ERR (", or ] expected while parsing array");
629		1002
630	++dec->cur;	1003	++dec->cur;
631	}	1004	}
632		1005
		1006	DEC_DEC_DEPTH;
633	return newRV_noinc ((SV *)av);	1007	return newRV_noinc ((SV *)av);
634		1008
635	fail:	1009	fail:
636	SvREFCNT_dec (av);	1010	SvREFCNT_dec (av);
		1011	DEC_DEC_DEPTH;
637	return 0;	1012	return 0;
638	}	1013	}
639		1014
640	static SV *	1015	static SV *
641	decode_hv (dec_t *dec)	1016	decode_hv (dec_t *dec)
642	{	1017	{
643	HV *hv = newHV ();	1018	HV *hv = newHV ();
644		1019
		1020	DEC_INC_DEPTH;
		1021	decode_ws (dec);
		1022
		1023	if (*dec->cur == '}')
		1024	++dec->cur;
		1025	else
645	for (;;)	1026	for (;;)
646	{	1027	{
647	SV *key, *value;	1028	SV *key, *value;
648		1029
649	WS; EXPECT_CH ('"');	1030	decode_ws (dec); EXPECT_CH ('"');
650		1031
651	key = decode_str (dec);	1032	key = decode_str (dec);
652	if (!key)	1033	if (!key)
653	goto fail;
654
655	WS; EXPECT_CH (':');
656
657	value = decode_sv (dec);
658	if (!value)
659	{
660	SvREFCNT_dec (key);
661	goto fail;	1034	goto fail;
		1035
		1036	decode_ws (dec); EXPECT_CH (':');
		1037
		1038	value = decode_sv (dec);
		1039	if (!value)
		1040	{
		1041	SvREFCNT_dec (key);
		1042	goto fail;
662	}	1043	}
663		1044
664	//TODO: optimise
665	hv_store_ent (hv, key, value, 0);	1045	hv_store_ent (hv, key, value, 0);
		1046	SvREFCNT_dec (key);
666		1047
667	WS;	1048	decode_ws (dec);
668		1049
669	if (*dec->cur == '}')	1050	if (*dec->cur == '}')
670	{	1051	{
671	++dec->cur;	1052	++dec->cur;
672	break;	1053	break;
673	}	1054	}
674		1055
675	if (*dec->cur != ',')	1056	if (*dec->cur != ',')
676	ERR (", or } expected while parsing object/hash");	1057	ERR (", or } expected while parsing object/hash");
677		1058
678	++dec->cur;	1059	++dec->cur;
679	}	1060	}
680		1061
		1062	DEC_DEC_DEPTH;
681	return newRV_noinc ((SV *)hv);	1063	return newRV_noinc ((SV *)hv);
682		1064
683	fail:	1065	fail:
684	SvREFCNT_dec (hv);	1066	SvREFCNT_dec (hv);
		1067	DEC_DEC_DEPTH;
685	return 0;	1068	return 0;
686	}	1069	}
687		1070
688	static SV *	1071	static SV *
689	decode_sv (dec_t *dec)	1072	decode_sv (dec_t *dec)
690	{	1073	{
691	WS;	1074	decode_ws (dec);
		1075
		1076	// the beauty of JSON: you need exactly one character lookahead
		1077	// to parse anything.
692	switch (*dec->cur)	1078	switch (*dec->cur)
693	{	1079	{
694	case '"': ++dec->cur; return decode_str (dec);	1080	case '"': ++dec->cur; return decode_str (dec);
695	case '[': ++dec->cur; return decode_av (dec);	1081	case '[': ++dec->cur; return decode_av (dec);
696	case '{': ++dec->cur; return decode_hv (dec);	1082	case '{': ++dec->cur; return decode_hv (dec);
…		…
702		1088
703	case 't':	1089	case 't':
704	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1090	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
705	{	1091	{
706	dec->cur += 4;	1092	dec->cur += 4;
707	return newSViv (1);	1093	return SvREFCNT_inc (json_true);
708	}	1094	}
709	else	1095	else
710	ERR ("'true' expected");	1096	ERR ("'true' expected");
711		1097
712	break;	1098	break;
713		1099
714	case 'f':	1100	case 'f':
715	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1101	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
716	{	1102	{
717	dec->cur += 5;	1103	dec->cur += 5;
718	return newSViv (0);	1104	return SvREFCNT_inc (json_false);
719	}	1105	}
720	else	1106	else
721	ERR ("'false' expected");	1107	ERR ("'false' expected");
722		1108
723	break;	1109	break;
724		1110
725	case 'n':	1111	case 'n':
726	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1112	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
727	{	1113	{
728	dec->cur += 4;	1114	dec->cur += 4;
729	return newSViv (1);	1115	return newSVsv (&PL_sv_undef);
730	}	1116	}
731	else	1117	else
732	ERR ("'null' expected");	1118	ERR ("'null' expected");
733		1119
734	break;	1120	break;
735		1121
736	default:	1122	default:
737	ERR ("malformed json string");	1123	ERR ("malformed JSON string, neither array, object, number, string or atom");
738	break;	1124	break;
739	}	1125	}
740		1126
741	fail:	1127	fail:
742	return 0;	1128	return 0;
743	}	1129	}
744		1130
745	static SV *	1131	static SV *
746	decode_json (SV *string, UV flags)	1132	decode_json (SV *string, U32 flags, UV *offset_return)
747	{	1133	{
		1134	dec_t dec;
		1135	UV offset;
748	SV *sv;	1136	SV *sv;
749		1137
		1138	SvGETMAGIC (string);
		1139	SvUPGRADE (string, SVt_PV);
		1140
		1141	if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
		1142	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1143	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
		1144
750	if (!(flags & F_UTF8))	1145	if (flags & F_UTF8)
		1146	sv_utf8_downgrade (string, 0);
		1147	else
751	sv_utf8_upgrade (string);	1148	sv_utf8_upgrade (string);
752		1149
753	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1150	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
754		1151
755	dec_t dec;
756	dec.flags = flags;	1152	dec.flags = flags;
757	dec.cur = SvPVX (string);	1153	dec.cur = SvPVX (string);
758	dec.end = SvEND (string);	1154	dec.end = SvEND (string);
759	dec.err = 0;	1155	dec.err = 0;
		1156	dec.depth = 0;
		1157	dec.maxdepth = DEC_DEPTH (dec.flags);
760		1158
761	*dec.end = 1; // invalid anywhere	1159	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
762	sv = decode_sv (&dec);	1160	sv = decode_sv (&dec);
763	*dec.end = 0;	1161
		1162	if (!(offset_return || !sv))
		1163	{
		1164	// check for trailing garbage
		1165	decode_ws (&dec);
		1166
		1167	if (*dec.cur)
		1168	{
		1169	dec.err = "garbage after JSON object";
		1170	SvREFCNT_dec (sv);
		1171	sv = 0;
		1172	}
		1173	}
		1174
		1175	if (offset_return || !sv)
		1176	{
		1177	offset = dec.flags & F_UTF8
		1178	? dec.cur - SvPVX (string)
		1179	: utf8_distance (dec.cur, SvPVX (string));
		1180
		1181	if (offset_return)
		1182	*offset_return = offset;
		1183	}
764		1184
765	if (!sv)	1185	if (!sv)
766	{	1186	{
767	IV offset = utf8_distance (dec.cur, SvPVX (string));
768	SV *uni = sv_newmortal ();	1187	SV *uni = sv_newmortal ();
769		1188
		1189	// horrible hack to silence warning inside pv_uni_display
		1190	COP cop = *PL_curcop;
		1191	cop.cop_warnings = pWARN_NONE;
		1192	ENTER;
		1193	SAVEVPTR (PL_curcop);
		1194	PL_curcop = &cop;
770	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1195	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1196	LEAVE;
		1197
771	croak ("%s, at character %d (%s)",	1198	croak ("%s, at character offset %d [\"%s\"]",
772	dec.err,	1199	dec.err,
773	(int)offset,	1200	(int)offset,
774	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1201	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
775	}	1202	}
776		1203
777	sv_dump (sv);//D
778	return sv_2mortal (sv);	1204	sv = sv_2mortal (sv);
		1205
		1206	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
		1207	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
		1208
		1209	return sv;
779	}	1210	}
		1211
		1212	/////////////////////////////////////////////////////////////////////////////
		1213	// XS interface functions
780		1214
781	MODULE = JSON::XS PACKAGE = JSON::XS	1215	MODULE = JSON::XS PACKAGE = JSON::XS
782		1216
783	BOOT:	1217	BOOT:
784	{	1218	{
785	int i;	1219	int i;
786		1220
787	memset (decode_hexdigit, 0xff, 256);
788	for (i = 10; i--; )	1221	for (i = 0; i < 256; ++i)
789	decode_hexdigit ['0' + i] = i;	1222	decode_hexdigit [i] =
		1223	i >= '0' && i <= '9' ? i - '0'
		1224	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1225	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1226	: -1;
790		1227
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);	1228	json_stash = gv_stashpv ("JSON::XS" , 1);
		1229	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1230
		1231	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1232	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
798	}	1233	}
		1234
		1235	PROTOTYPES: DISABLE
799		1236
800	SV *new (char *dummy)	1237	SV *new (char *dummy)
801	CODE:	1238	CODE:
802	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1239	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
803	OUTPUT:	1240	OUTPUT:
804	RETVAL	1241	RETVAL
805		1242
806	SV *ascii (SV *self, int enable)	1243	SV *ascii (SV *self, int enable = 1)
807	ALIAS:	1244	ALIAS:
808	ascii = F_ASCII	1245	ascii = F_ASCII
		1246	latin1 = F_LATIN1
809	utf8 = F_UTF8	1247	utf8 = F_UTF8
810	indent = F_INDENT	1248	indent = F_INDENT
811	canonical = F_CANONICAL	1249	canonical = F_CANONICAL
812	space_before = F_SPACE_BEFORE	1250	space_before = F_SPACE_BEFORE
813	space_after = F_SPACE_AFTER	1251	space_after = F_SPACE_AFTER
814	json_rpc = F_JSON_RPC
815	pretty = F_PRETTY	1252	pretty = F_PRETTY
		1253	allow_nonref = F_ALLOW_NONREF
		1254	shrink = F_SHRINK
		1255	allow_blessed = F_ALLOW_BLESSED
		1256	convert_blessed = F_CONV_BLESSED
816	CODE:	1257	CODE:
817	{	1258	{
818	UV *uv = SvJSON (self);	1259	UV *uv = SvJSON (self);
819	if (enable)	1260	if (enable)
820	*uv |= ix;	1261	*uv |= ix;
…		…
824	RETVAL = newSVsv (self);	1265	RETVAL = newSVsv (self);
825	}	1266	}
826	OUTPUT:	1267	OUTPUT:
827	RETVAL	1268	RETVAL
828		1269
		1270	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1271	CODE:
		1272	{
		1273	UV *uv = SvJSON (self);
		1274	UV log2 = 0;
		1275
		1276	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1277
		1278	while ((1UL << log2) < max_depth)
		1279	++log2;
		1280
		1281	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1282
		1283	RETVAL = newSVsv (self);
		1284	}
		1285	OUTPUT:
		1286	RETVAL
		1287
		1288	SV *max_size (SV *self, UV max_size = 0)
		1289	CODE:
		1290	{
		1291	UV *uv = SvJSON (self);
		1292	UV log2 = 0;
		1293
		1294	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1295	if (max_size == 1) max_size = 2;
		1296
		1297	while ((1UL << log2) < max_size)
		1298	++log2;
		1299
		1300	*uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1301
		1302	RETVAL = newSVsv (self);
		1303	}
		1304	OUTPUT:
		1305	RETVAL
		1306
829	void encode (SV *self, SV *scalar)	1307	void encode (SV *self, SV *scalar)
830	PPCODE:	1308	PPCODE:
831	XPUSHs (encode_json (scalar, *SvJSON (self)));	1309	XPUSHs (encode_json (scalar, *SvJSON (self)));
832		1310
833	void decode (SV *self, SV *jsonstr)	1311	void decode (SV *self, SV *jsonstr)
834	PPCODE:	1312	PPCODE:
835	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1313	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1314
		1315	void decode_prefix (SV *self, SV *jsonstr)
		1316	PPCODE:
		1317	{
		1318	UV offset;
		1319	EXTEND (SP, 2);
		1320	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1321	PUSHs (sv_2mortal (newSVuv (offset)));
		1322	}
		1323
		1324	PROTOTYPES: ENABLE
836		1325
837	void to_json (SV *scalar)	1326	void to_json (SV *scalar)
		1327	ALIAS:
		1328	objToJson = 0
838	PPCODE:	1329	PPCODE:
839	XPUSHs (encode_json (scalar, F_UTF8));	1330	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
840		1331
841	void from_json (SV *jsonstr)	1332	void from_json (SV *jsonstr)
		1333	ALIAS:
		1334	jsonToObj = 0
842	PPCODE:	1335	PPCODE:
843	XPUSHs (decode_json (jsonstr, F_UTF8));	1336	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
844		1337

Diff Legend

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