ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.3 by root, Thu Mar 22 18:10:29 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 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
16	#define F_ALLOW_NONREF 0x00000080	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 ))
17		38
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		41
21	#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
22		46
23	#define SB do {	47	#define SB do {
24	#define SE } while (0)	48	#define SE } while (0)
25		49
26	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
27		114
28	// structure used for encoding JSON	115	// structure used for encoding JSON
29	typedef struct	116	typedef struct
30	{	117	{
31	char *cur;	118	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	119	char *end; // SvEND (sv)
34	SV *sv;	120	SV *sv; // result scalar
35	UV flags;	121	U32 flags; // F_*
36	int max_recurse;	122	U32 indent; // indentation level
37	int indent;	123	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	124	} enc_t;
39		125
40	// structure used for decoding JSON	126	inline void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	/////////////////////////////////////////////////////////////////////////////
59
60	static void
61	need (enc_t *enc, STRLEN len)	127	need (enc_t *enc, STRLEN len)
62	{	128	{
63	if (enc->cur + len >= enc->end)	129	if (expect_false (enc->cur + len >= enc->end))
64	{	130	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	131	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	132	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	133	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvEND (enc->sv);	134	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	135	}
70	}	136	}
71		137
72	static void	138	inline void
73	encode_ch (enc_t *enc, char ch)	139	encode_ch (enc_t *enc, char ch)
74	{	140	{
75	need (enc, 1);	141	need (enc, 1);
76	*enc->cur++ = ch;	142	*enc->cur++ = ch;
77	}	143	}
…		…
79	static void	145	static void
80	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)
81	{	147	{
82	char *end = str + len;	148	char *end = str + len;
83		149
		150	need (enc, len);
		151
84	while (str < end)	152	while (str < end)
85	{	153	{
86	unsigned char ch = *(unsigned char *)str;	154	unsigned char ch = *(unsigned char *)str;
		155
87	if (ch >= 0x20 && ch < 0x80) // most common case	156	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	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
89	*enc->cur++ = ch;	171	*enc->cur++ = ch;
		172
90	str++;	173	++str;
91	}	174	}
92	else	175	else
93	{	176	{
94	STRLEN clen;	177	switch (ch)
95	UV uch;
96
97	if (is_utf8)
98	{
99	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
100	if (clen < 0)
101	croak ("malformed UTF-8 character in string, cannot convert to JSON");
102	}	178	{
103	else	179	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
104	{	180	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
105	uch = ch;	181	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
106	clen = 1;	182	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
107	}	183	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
108		184
109	need (enc, len += 6);	185	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	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;
115	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	{
116	need (enc, len += 6);	209	need (enc, len += 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	210	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	211	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	212	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	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;
		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	}
		243	else
		244	{
		245	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		246	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		247	++str;
		248	}
121	}	249	}
122	else
123	{
124	sprintf (enc->cur, "\\u%04x", uch);
125	enc->cur += 6;
126	}
127	}
128	else if (is_utf8)
129	{	250	}
130	memcpy (enc->cur, str, clen);
131	enc->cur += clen;
132	}
133	else
134	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
135
136	str += clen;
137	}	251	}
138		252
139	--len;	253	--len;
140	}	254	}
141	}	255	}
142		256
143	#define INDENT SB \	257	inline void
		258	encode_indent (enc_t *enc)
		259	{
144	if (enc->flags & F_INDENT) \	260	if (enc->flags & F_INDENT)
145	{ \	261	{
146	int i_; \	262	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \
148	for (i_ = enc->indent * 3; i_--; )\
149	encode_ch (enc, ' '); \
150	} \
151	SE
152		263
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	264	need (enc, spaces);
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	265	memset (enc->cur, ' ', spaces);
155	#define COMMA SB \	266	enc->cur += spaces;
		267	}
		268	}
		269
		270	inline void
		271	encode_space (enc_t *enc)
		272	{
		273	need (enc, 1);
156	encode_ch (enc, ','); \	274	encode_ch (enc, ' ');
		275	}
		276
		277	inline void
		278	encode_nl (enc_t *enc)
		279	{
157	if (enc->flags & F_INDENT) \	280	if (enc->flags & F_INDENT)
158	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);
159	else if (enc->flags & F_SPACE_AFTER) \	294	else if (enc->flags & F_SPACE_AFTER)
160	SPACE; \	295	encode_space (enc);
161	SE	296	}
162		297
163	static void encode_sv (enc_t *enc, SV *sv);	298	static void encode_sv (enc_t *enc, SV *sv);
164		299
165	static void	300	static void
166	encode_av (enc_t *enc, AV *av)	301	encode_av (enc_t *enc, AV *av)
167	{	302	{
168	int i, len = av_len (av);	303	int i, len = av_len (av);
169		304
170	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);
171	++enc->indent;	309	++enc->indent;
172		310
173	for (i = 0; i <= len; ++i)	311	for (i = 0; i <= len; ++i)
174	{	312	{
175	INDENT;	313	encode_indent (enc);
176	encode_sv (enc, *av_fetch (av, i, 0));	314	encode_sv (enc, *av_fetch (av, i, 0));
177		315
178	if (i < len)	316	if (i < len)
179	COMMA;	317	encode_comma (enc);
180	}	318	}
181		319
182	NL;	320	encode_nl (enc);
183		321
184	--enc->indent;	322	--enc->indent;
185	INDENT; encode_ch (enc, ']');	323	encode_indent (enc); encode_ch (enc, ']');
186	}	324	}
187		325
188	static void	326	static void
189	encode_he (enc_t *enc, HE *he)	327	encode_he (enc_t *enc, HE *he)
190	{	328	{
…		…
192		330
193	if (HeKLEN (he) == HEf_SVKEY)	331	if (HeKLEN (he) == HEf_SVKEY)
194	{	332	{
195	SV *sv = HeSVKEY (he);	333	SV *sv = HeSVKEY (he);
196	STRLEN len;	334	STRLEN len;
		335	char *str;
		336
		337	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	338	str = SvPV (sv, len);
198		339
199	encode_str (enc, str, len, SvUTF8 (sv));	340	encode_str (enc, str, len, SvUTF8 (sv));
200	}	341	}
201	else	342	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	343	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		344
204	encode_ch (enc, '"');	345	encode_ch (enc, '"');
205		346
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	347	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	348	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	349	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));	350	encode_sv (enc, HeVAL (he));
210	}	351	}
211		352
212	// compare hash entries, used when all keys are bytestrings	353	// compare hash entries, used when all keys are bytestrings
213	static int	354	static int
…		…
219	HE *b = *(HE **)b_;	360	HE *b = *(HE **)b_;
220		361
221	STRLEN la = HeKLEN (a);	362	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	363	STRLEN lb = HeKLEN (b);
223		364
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	365	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	366	cmp = la - lb;
226		367
227	return cmp;	368	return cmp;
228	}	369	}
229		370
230	// 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
…		…
237	static void	378	static void
238	encode_hv (enc_t *enc, HV *hv)	379	encode_hv (enc_t *enc, HV *hv)
239	{	380	{
240	int count, i;	381	int count, i;
241		382
		383	if (enc->indent >= enc->maxdepth)
		384	croak ("data structure too deep (hit recursion limit)");
		385
242	encode_ch (enc, '{'); NL; ++enc->indent;	386	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
243		387
244	if ((count = hv_iterinit (hv)))	388	if ((count = hv_iterinit (hv)))
245	{	389	{
246	// for canonical output we have to sort by keys first	390	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	391	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	392	// security workaround added somewhere in 5.8.x.
249	// that randomises hash orderings	393	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	394	if (enc->flags & F_CANONICAL)
251	{	395	{
252	HE *he, *hes [count];
253	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
254		403
255	i = 0;	404	i = 0;
256	while ((he = hv_iternext (hv)))	405	while ((he = hv_iternext (hv)))
257	{	406	{
258	hes [i++] = he;	407	hes [i++] = he;
…		…
264		413
265	if (fast)	414	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	415	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	416	else
268	{	417	{
269	// hack to disable "use bytes"	418	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	419	COP cop = *PL_curcop;
271	cop.op_private = 0;	420	cop.op_private = 0;
		421
		422	ENTER;
		423	SAVETMPS;
		424
		425	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	426	PL_curcop = &cop;
273		427
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	428	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		429
276	FREETMPS;	430	FREETMPS;
277		431	LEAVE;
278	PL_curcop = oldcop;
279	}	432	}
280		433
281	for (i = 0; i < count; ++i)	434	for (i = 0; i < count; ++i)
282	{	435	{
283	INDENT;	436	encode_indent (enc);
284	encode_he (enc, hes [i]);	437	encode_he (enc, hes [i]);
285		438
286	if (i < count - 1)	439	if (i < count - 1)
287	COMMA;	440	encode_comma (enc);
288	}
289
290	NL;	441	}
		442
		443	encode_nl (enc);
291	}	444	}
292	else	445	else
293	{	446	{
294	SV *sv;
295	HE *he = hv_iternext (hv);	447	HE *he = hv_iternext (hv);
296		448
297	for (;;)	449	for (;;)
298	{	450	{
299	INDENT;	451	encode_indent (enc);
300	encode_he (enc, he);	452	encode_he (enc, he);
301		453
302	if (!(he = hv_iternext (hv)))	454	if (!(he = hv_iternext (hv)))
303	break;	455	break;
304		456
305	COMMA;	457	encode_comma (enc);
306	}
307
308	NL;	458	}
		459
		460	encode_nl (enc);
309	}	461	}
310	}	462	}
311		463
312	--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))));
313	}	547	}
314		548
315	static void	549	static void
316	encode_sv (enc_t *enc, SV *sv)	550	encode_sv (enc_t *enc, SV *sv)
317	{	551	{
		552	SvGETMAGIC (sv);
		553
318	if (SvPOKp (sv))	554	if (SvPOKp (sv))
319	{	555	{
320	STRLEN len;	556	STRLEN len;
321	char *str = SvPV (sv, len);	557	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	558	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	559	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	560	encode_ch (enc, '"');
325	}	561	}
326	else if (SvNOKp (sv))	562	else if (SvNOKp (sv))
327	{	563	{
		564	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	565	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	566	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	567	enc->cur += strlen (enc->cur);
331	}	568	}
332	else if (SvIOKp (sv))	569	else if (SvIOKp (sv))
333	{	570	{
334	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);
335	enc->cur +=	576	enc->cur +=
336	SvIsUV(sv)	577	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	578	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338	: 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	}
339	}	608	}
340	else if (SvROK (sv))	609	else if (SvROK (sv))
341	{	610	encode_rv (enc, SvRV (sv));
342	if (!--enc->max_recurse)
343	croak ("data structure too deep (hit recursion limit)");
344
345	sv = SvRV (sv);
346
347	switch (SvTYPE (sv))
348	{
349	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
350	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
351
352	default:
353	croak ("JSON can only represent references to arrays or hashes");
354	}
355	}
356	else if (!SvOK (sv))	611	else if (!SvOK (sv))
357	encode_str (enc, "null", 4, 0);	612	encode_str (enc, "null", 4, 0);
358	else	613	else
359	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));
360	}	616	}
361		617
362	static SV *	618	static SV *
363	encode_json (SV *scalar, UV flags)	619	encode_json (SV *scalar, U32 flags)
364	{	620	{
		621	enc_t enc;
		622
365	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	623	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
366	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");	624	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
367		625
368	enc_t enc;
369	enc.flags = flags;	626	enc.flags = flags;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	627	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	628	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	629	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	630	enc.indent = 0;
		631	enc.maxdepth = DEC_DEPTH (flags);
375		632
376	SvPOK_only (enc.sv);	633	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	634	encode_sv (&enc, scalar);
378		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
379	if (!(flags & (F_ASCII | F_UTF8)))	639	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	640	SvUTF8_on (enc.sv);
381		641
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	642	if (enc.flags & F_SHRINK)
		643	shrink (enc.sv);
		644
383	return enc.sv;	645	return enc.sv;
384	}	646	}
385		647
386	/////////////////////////////////////////////////////////////////////////////	648	/////////////////////////////////////////////////////////////////////////////
		649	// decoder
387		650
388	#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	{
389	for (;;) \	665	for (;;)
390	{ \	666	{
391	char ch = *dec->cur; \	667	char ch = *dec->cur;
		668
392	if (ch > 0x20 \	669	if (ch > 0x20
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	670	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394	break; \	671	break;
		672
395	++dec->cur; \	673	++dec->cur;
396	}	674	}
		675	}
397		676
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	677	#define ERR(reason) SB dec->err = reason; goto fail; SE
		678
399	#define EXPECT_CH(ch) SB \	679	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	680	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	681	ERR (# ch " expected"); \
402	++dec->cur; \	682	++dec->cur; \
403	SE	683	SE
404		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
405	static SV *decode_sv (dec_t *dec);	688	static SV *decode_sv (dec_t *dec);
406
407	#define APPEND_CH(ch) SB \
408	SvGROW (sv, cur + 1 + 1); \
409	SvPVX (sv)[cur++] = (ch); \
410	SE
411		689
412	static signed char decode_hexdigit[256];	690	static signed char decode_hexdigit[256];
413		691
414	static UV	692	static UV
415	decode_4hex (dec_t *dec)	693	decode_4hex (dec_t *dec)
416	{	694	{
417	signed char d1, d2, d3, d4;	695	signed char d1, d2, d3, d4;
		696	unsigned char *cur = (unsigned char *)dec->cur;
418		697
419	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");
420	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");
421	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");
422	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");
423	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
424	if (d3 < 0) ERR ("four hexadecimal digits expected");
425	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
426	if (d4 < 0) ERR ("four hexadecimal digits expected");
427		702
428	dec->cur += 4;	703	dec->cur += 4;
429		704
430	return ((UV)d1) << 12	705	return ((UV)d1) << 12
431	| ((UV)d2) << 8	706	| ((UV)d2) << 8
…		…
437	}	712	}
438		713
439	static SV *	714	static SV *
440	decode_str (dec_t *dec)	715	decode_str (dec_t *dec)
441	{	716	{
442	SV *sv = NEWSV (0,2);	717	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	718	int utf8 = 0;
		719	char *dec_cur = dec->cur;
445		720
446	for (;;)	721	do
447	{	722	{
448	unsigned char ch = *(unsigned char *)dec->cur;	723	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		724	char *cur = buf;
449		725
450	if (ch == '"')	726	do
451	break;
452	else if (ch == '\\')
453	{	727	{
454	switch (*++dec->cur)	728	unsigned char ch = *(unsigned char *)dec_cur++;
		729
		730	if (expect_false (ch == '"'))
		731	{
		732	--dec_cur;
		733	break;
455	{	734	}
456	case '\\':	735	else if (expect_false (ch == '\\'))
457	case '/':	736	{
458	case '"': APPEND_CH (*dec->cur++); break;	737	switch (*dec_cur)
459
460	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
461	case 't': APPEND_CH ('\011'); ++dec->cur; break;
462	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
463	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
464	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
465
466	case 'u':
467	{	738	{
468	UV lo, hi;	739	case '\\':
469	++dec->cur;	740	case '/':
		741	case '"': *cur++ = *dec_cur++; break;
470		742
471	hi = decode_4hex (dec);	743	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	744	case 't': ++dec_cur; *cur++ = '\011'; break;
473	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;
474		748
475	// possibly a surrogate pair	749	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	750	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	751	UV lo, hi;
479	ERR ("illegal surrogate character");	752	++dec_cur;
480		753
481	dec->cur += 2;	754	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	755	hi = decode_4hex (dec);
		756	dec_cur = dec->cur;
484	if (lo == (UV)-1)	757	if (hi == (UV)-1)
485	goto fail;	758	goto fail;
486		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
487	if (lo < 0xdc00 || lo >= 0xe000)	775	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	776	ERR ("surrogate pair expected");
489		777
490	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;
491	}	791	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	792	break;
497	utf8 = 1;
498		793
499	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
500	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
501	}
502	else	794	default:
503	APPEND_CH (hi);	795	--dec_cur;
		796	ERR ("illegal backslash escape sequence in string");
504	}	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");
505	break;	824	else
		825	ERR ("invalid character encountered while parsing JSON string");
506	}	826	}
507	}	827	}
508	else if (ch >= 0x20 && ch <= 0x7f)	828	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	829
510	else if (ch >= 0x80)	830	{
		831	STRLEN len = cur - buf;
		832
		833	if (sv)
511	{	834	{
512	STRLEN clen;	835	SvGROW (sv, SvCUR (sv) + len + 1);
513	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	836	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514	if (clen < 0)	837	SvCUR_set (sv, SvCUR (sv) + len);
515	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
516
517	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
518	memcpy (SvPVX (sv) + cur, dec->cur, clen);
519	dec->cur += clen;
520	}	838	}
521	else	839	else
522	ERR ("invalid character encountered");	840	sv = newSVpvn (buf, len);
523	}	841	}
		842	}
		843	while (*dec_cur != '"');
524		844
525	++dec->cur;	845	++dec_cur;
526		846
		847	if (sv)
		848	{
527	SvPOK_only (sv);	849	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	850	*SvEND (sv) = 0;
531		851
532	if (utf8)	852	if (utf8)
533	SvUTF8_on (sv);	853	SvUTF8_on (sv);
		854	}
		855	else
		856	sv = newSVpvn ("", 0);
534		857
		858	dec->cur = dec_cur;
535	return sv;	859	return sv;
536		860
537	fail:	861	fail:
538	SvREFCNT_dec (sv);	862	dec->cur = dec_cur;
539	return 0;	863	return 0;
540	}	864	}
541		865
542	static SV *	866	static SV *
543	decode_num (dec_t *dec)	867	decode_num (dec_t *dec)
…		…
553	{	877	{
554	++dec->cur;	878	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	879	if (*dec->cur >= '0' && *dec->cur <= '9')
556	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)");
557	}	881	}
558		882	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	883	ERR ("malformed number (no digits after initial minus)");
		884	else
		885	do
		886	{
		887	++dec->cur;
		888	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	889	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		890
563	// [frac]	891	// [frac]
564	if (*dec->cur == '.')	892	if (*dec->cur == '.')
565	{	893	{
566	is_nv = 1;	894	++dec->cur;
		895
		896	if (*dec->cur < '0' || *dec->cur > '9')
		897	ERR ("malformed number (no digits after decimal point)");
567		898
568	do	899	do
569	{	900	{
570	++dec->cur;	901	++dec->cur;
571	}	902	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	903	while (*dec->cur >= '0' && *dec->cur <= '9');
		904
		905	is_nv = 1;
573	}	906	}
574		907
575	// [exp]	908	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	909	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	910	{
578	is_nv = 1;
579
580	++dec->cur;	911	++dec->cur;
		912
581	if (*dec->cur == '-' || *dec->cur == '+')	913	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	914	++dec->cur;
583		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	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	923	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	924
		925	is_nv = 1;
586	}	926	}
587		927
588	if (!is_nv)	928	if (!is_nv)
589	{	929	{
590	UV uv;	930	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
591	int numtype = grok_number (start, dec->cur - start, &uv);	931	if (*start == '-')
592	if (numtype & IS_NUMBER_IN_UV)	932	switch (dec->cur - start)
593	if (numtype & IS_NUMBER_NEG)
594	{	933	{
595	if (uv < (UV)IV_MIN)	934	case 2: return newSViv (-( start [1] - '0' * 1));
596	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));
597	}	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	}
598	else	957	else
599	return newSVuv (uv);	958	return newSVuv (uv);
		959
		960	// here would likely be the place for bigint support
600	}	961	}
		962	}
601		963
		964	// if we ever support bigint or bigfloat, this is the place for bigfloat
602	return newSVnv (Atof (start));	965	return newSVnv (Atof (start));
603		966
604	fail:	967	fail:
605	return 0;	968	return 0;
606	}	969	}
…		…
608	static SV *	971	static SV *
609	decode_av (dec_t *dec)	972	decode_av (dec_t *dec)
610	{	973	{
611	AV *av = newAV ();	974	AV *av = newAV ();
612		975
		976	DEC_INC_DEPTH;
		977	decode_ws (dec);
		978
		979	if (*dec->cur == ']')
		980	++dec->cur;
		981	else
613	for (;;)	982	for (;;)
614	{	983	{
615	SV *value;	984	SV *value;
616		985
617	value = decode_sv (dec);	986	value = decode_sv (dec);
618	if (!value)	987	if (!value)
619	goto fail;	988	goto fail;
620		989
621	av_push (av, value);	990	av_push (av, value);
622		991
623	WS;	992	decode_ws (dec);
624		993
625	if (*dec->cur == ']')	994	if (*dec->cur == ']')
626	{	995	{
627	++dec->cur;	996	++dec->cur;
628	break;	997	break;
		998	}
629	}	999
630
631	if (*dec->cur != ',')	1000	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	1001	ERR (", or ] expected while parsing array");
633		1002
634	++dec->cur;	1003	++dec->cur;
635	}	1004	}
636		1005
		1006	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1007	return newRV_noinc ((SV *)av);
638		1008
639	fail:	1009	fail:
640	SvREFCNT_dec (av);	1010	SvREFCNT_dec (av);
		1011	DEC_DEC_DEPTH;
641	return 0;	1012	return 0;
642	}	1013	}
643		1014
644	static SV *	1015	static SV *
645	decode_hv (dec_t *dec)	1016	decode_hv (dec_t *dec)
646	{	1017	{
647	HV *hv = newHV ();	1018	HV *hv = newHV ();
648		1019
		1020	DEC_INC_DEPTH;
		1021	decode_ws (dec);
		1022
		1023	if (*dec->cur == '}')
		1024	++dec->cur;
		1025	else
649	for (;;)	1026	for (;;)
650	{	1027	{
651	SV *key, *value;	1028	decode_ws (dec); EXPECT_CH ('"');
652		1029
653	WS; EXPECT_CH ('"');	1030	// heuristic: assume that
654		1031	// a) decode_str + hv_store_ent are abysmally slow
655	key = decode_str (dec);	1032	// b) most hash keys are short, simple ascii text
656	if (!key)	1033	// so try to "fast-match" such strings to avoid
657	goto fail;	1034	// the overhead of hv_store_ent.
658
659	WS; EXPECT_CH (':');
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1035	{
		1036	SV *value;
		1037	char *p = dec->cur;
		1038	char *e = p + 24; // only try up to 24 bytes
		1039
		1040	for (;;)
		1041	{
		1042	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1043	{
		1044	// slow path, back up and use decode_str
		1045	SV *key = decode_str (dec);
		1046	if (!key)
		1047	goto fail;
		1048
		1049	decode_ws (dec); EXPECT_CH (':');
		1050
		1051	value = decode_sv (dec);
		1052	if (!value)
		1053	{
		1054	SvREFCNT_dec (key);
		1055	goto fail;
		1056	}
		1057
		1058	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1059	SvREFCNT_dec (key);
		1060
		1061	break;
		1062	}
		1063	else if (*p == '"')
		1064	{
		1065	// fast path, got a simple key
		1066	char *key = dec->cur;
		1067	int len = p - key;
		1068	dec->cur = p + 1;
		1069
		1070	decode_ws (dec); EXPECT_CH (':');
		1071
		1072	value = decode_sv (dec);
		1073	if (!value)
665	goto fail;	1074	goto fail;
		1075
		1076	hv_store (hv, key, len, value, 0);
		1077
		1078	break;
		1079	}
		1080
		1081	++p;
		1082	}
666	}	1083	}
667		1084
668	//TODO: optimise	1085	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1086
671	WS;
672
673	if (*dec->cur == '}')	1087	if (*dec->cur == '}')
674	{	1088	{
675	++dec->cur;	1089	++dec->cur;
676	break;	1090	break;
677	}	1091	}
678		1092
679	if (*dec->cur != ',')	1093	if (*dec->cur != ',')
680	ERR (", or } expected while parsing object/hash");	1094	ERR (", or } expected while parsing object/hash");
681		1095
682	++dec->cur;	1096	++dec->cur;
683	}	1097	}
684		1098
		1099	DEC_DEC_DEPTH;
685	return newRV_noinc ((SV *)hv);	1100	return newRV_noinc ((SV *)hv);
686		1101
687	fail:	1102	fail:
688	SvREFCNT_dec (hv);	1103	SvREFCNT_dec (hv);
		1104	DEC_DEC_DEPTH;
689	return 0;	1105	return 0;
690	}	1106	}
691		1107
692	static SV *	1108	static SV *
693	decode_sv (dec_t *dec)	1109	decode_sv (dec_t *dec)
694	{	1110	{
695	WS;	1111	decode_ws (dec);
		1112
		1113	// the beauty of JSON: you need exactly one character lookahead
		1114	// to parse anything.
696	switch (*dec->cur)	1115	switch (*dec->cur)
697	{	1116	{
698	case '"': ++dec->cur; return decode_str (dec);	1117	case '"': ++dec->cur; return decode_str (dec);
699	case '[': ++dec->cur; return decode_av (dec);	1118	case '[': ++dec->cur; return decode_av (dec);
700	case '{': ++dec->cur; return decode_hv (dec);	1119	case '{': ++dec->cur; return decode_hv (dec);
…		…
706		1125
707	case 't':	1126	case 't':
708	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1127	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709	{	1128	{
710	dec->cur += 4;	1129	dec->cur += 4;
711	return newSViv (1);	1130	return SvREFCNT_inc (json_true);
712	}	1131	}
713	else	1132	else
714	ERR ("'true' expected");	1133	ERR ("'true' expected");
715		1134
716	break;	1135	break;
717		1136
718	case 'f':	1137	case 'f':
719	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1138	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720	{	1139	{
721	dec->cur += 5;	1140	dec->cur += 5;
722	return newSViv (0);	1141	return SvREFCNT_inc (json_false);
723	}	1142	}
724	else	1143	else
725	ERR ("'false' expected");	1144	ERR ("'false' expected");
726		1145
727	break;	1146	break;
728		1147
729	case 'n':	1148	case 'n':
730	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1149	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731	{	1150	{
732	dec->cur += 4;	1151	dec->cur += 4;
733	return newSViv (1);	1152	return newSVsv (&PL_sv_undef);
734	}	1153	}
735	else	1154	else
736	ERR ("'null' expected");	1155	ERR ("'null' expected");
737		1156
738	break;	1157	break;
739		1158
740	default:	1159	default:
741	ERR ("malformed json string");	1160	ERR ("malformed JSON string, neither array, object, number, string or atom");
742	break;	1161	break;
743	}	1162	}
744		1163
745	fail:	1164	fail:
746	return 0;	1165	return 0;
747	}	1166	}
748		1167
749	static SV *	1168	static SV *
750	decode_json (SV *string, UV flags)	1169	decode_json (SV *string, U32 flags, UV *offset_return)
751	{	1170	{
		1171	dec_t dec;
		1172	UV offset;
752	SV *sv;	1173	SV *sv;
753		1174
		1175	SvGETMAGIC (string);
		1176	SvUPGRADE (string, SVt_PV);
		1177
		1178	if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
		1179	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1180	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
		1181
754	if (!(flags & F_UTF8))	1182	if (flags & F_UTF8)
		1183	sv_utf8_downgrade (string, 0);
		1184	else
755	sv_utf8_upgrade (string);	1185	sv_utf8_upgrade (string);
756		1186
757	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1187	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758		1188
759	dec_t dec;
760	dec.flags = flags;	1189	dec.flags = flags;
761	dec.cur = SvPVX (string);	1190	dec.cur = SvPVX (string);
762	dec.end = SvEND (string);	1191	dec.end = SvEND (string);
763	dec.err = 0;	1192	dec.err = 0;
		1193	dec.depth = 0;
		1194	dec.maxdepth = DEC_DEPTH (dec.flags);
764		1195
765	*dec.end = 1; // invalid anywhere	1196	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
766	sv = decode_sv (&dec);	1197	sv = decode_sv (&dec);
767	*dec.end = 0;	1198
		1199	if (!(offset_return || !sv))
		1200	{
		1201	// check for trailing garbage
		1202	decode_ws (&dec);
		1203
		1204	if (*dec.cur)
		1205	{
		1206	dec.err = "garbage after JSON object";
		1207	SvREFCNT_dec (sv);
		1208	sv = 0;
		1209	}
		1210	}
		1211
		1212	if (offset_return || !sv)
		1213	{
		1214	offset = dec.flags & F_UTF8
		1215	? dec.cur - SvPVX (string)
		1216	: utf8_distance (dec.cur, SvPVX (string));
		1217
		1218	if (offset_return)
		1219	*offset_return = offset;
		1220	}
768		1221
769	if (!sv)	1222	if (!sv)
770	{	1223	{
771	IV offset = utf8_distance (dec.cur, SvPVX (string));
772	SV *uni = sv_newmortal ();	1224	SV *uni = sv_newmortal ();
773		1225
		1226	// horrible hack to silence warning inside pv_uni_display
		1227	COP cop = *PL_curcop;
		1228	cop.cop_warnings = pWARN_NONE;
		1229	ENTER;
		1230	SAVEVPTR (PL_curcop);
		1231	PL_curcop = &cop;
774	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1232	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1233	LEAVE;
		1234
775	croak ("%s, at character %d (%s)",	1235	croak ("%s, at character offset %d [\"%s\"]",
776	dec.err,	1236	dec.err,
777	(int)offset,	1237	(int)offset,
778	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1238	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779	}	1239	}
780		1240
781	sv = sv_2mortal (sv);	1241	sv = sv_2mortal (sv);
782		1242
783	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1243	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
784	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1244	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785		1245
786	return sv;	1246	return sv;
787	}	1247	}
788		1248
		1249	/////////////////////////////////////////////////////////////////////////////
		1250	// XS interface functions
		1251
789	MODULE = JSON::XS PACKAGE = JSON::XS	1252	MODULE = JSON::XS PACKAGE = JSON::XS
790		1253
791	BOOT:	1254	BOOT:
792	{	1255	{
793	int i;	1256	int i;
794		1257
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	1258	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	1259	decode_hexdigit [i] =
		1260	i >= '0' && i <= '9' ? i - '0'
		1261	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1262	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1263	: -1;
798		1264
799	for (i = 6; --i; )
800	{
801	decode_hexdigit ['a' + i] = 10 + i;
802	decode_hexdigit ['A' + i] = 10 + i;
803	}
804
805	json_stash = gv_stashpv ("JSON::XS", 1);	1265	json_stash = gv_stashpv ("JSON::XS" , 1);
		1266	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1267
		1268	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1269	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
806	}	1270	}
		1271
		1272	PROTOTYPES: DISABLE
807		1273
808	SV *new (char *dummy)	1274	SV *new (char *dummy)
809	CODE:	1275	CODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1276	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
811	OUTPUT:	1277	OUTPUT:
812	RETVAL	1278	RETVAL
813		1279
814	SV *ascii (SV *self, int enable)	1280	SV *ascii (SV *self, int enable = 1)
815	ALIAS:	1281	ALIAS:
816	ascii = F_ASCII	1282	ascii = F_ASCII
		1283	latin1 = F_LATIN1
817	utf8 = F_UTF8	1284	utf8 = F_UTF8
818	indent = F_INDENT	1285	indent = F_INDENT
819	canonical = F_CANONICAL	1286	canonical = F_CANONICAL
820	space_before = F_SPACE_BEFORE	1287	space_before = F_SPACE_BEFORE
821	space_after = F_SPACE_AFTER	1288	space_after = F_SPACE_AFTER
822	json_rpc = F_JSON_RPC
823	pretty = F_PRETTY	1289	pretty = F_PRETTY
824	allow_nonref = F_ALLOW_NONREF	1290	allow_nonref = F_ALLOW_NONREF
		1291	shrink = F_SHRINK
		1292	allow_blessed = F_ALLOW_BLESSED
		1293	convert_blessed = F_CONV_BLESSED
825	CODE:	1294	CODE:
826	{	1295	{
827	UV *uv = SvJSON (self);	1296	UV *uv = SvJSON (self);
828	if (enable)	1297	if (enable)
829	*uv |= ix;	1298	*uv |= ix;
…		…
833	RETVAL = newSVsv (self);	1302	RETVAL = newSVsv (self);
834	}	1303	}
835	OUTPUT:	1304	OUTPUT:
836	RETVAL	1305	RETVAL
837		1306
		1307	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1308	CODE:
		1309	{
		1310	UV *uv = SvJSON (self);
		1311	UV log2 = 0;
		1312
		1313	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1314
		1315	while ((1UL << log2) < max_depth)
		1316	++log2;
		1317
		1318	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1319
		1320	RETVAL = newSVsv (self);
		1321	}
		1322	OUTPUT:
		1323	RETVAL
		1324
		1325	SV *max_size (SV *self, UV max_size = 0)
		1326	CODE:
		1327	{
		1328	UV *uv = SvJSON (self);
		1329	UV log2 = 0;
		1330
		1331	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1332	if (max_size == 1) max_size = 2;
		1333
		1334	while ((1UL << log2) < max_size)
		1335	++log2;
		1336
		1337	*uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1338
		1339	RETVAL = newSVsv (self);
		1340	}
		1341	OUTPUT:
		1342	RETVAL
		1343
838	void encode (SV *self, SV *scalar)	1344	void encode (SV *self, SV *scalar)
839	PPCODE:	1345	PPCODE:
840	XPUSHs (encode_json (scalar, *SvJSON (self)));	1346	XPUSHs (encode_json (scalar, *SvJSON (self)));
841		1347
842	void decode (SV *self, SV *jsonstr)	1348	void decode (SV *self, SV *jsonstr)
843	PPCODE:	1349	PPCODE:
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1350	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1351
		1352	void decode_prefix (SV *self, SV *jsonstr)
		1353	PPCODE:
		1354	{
		1355	UV offset;
		1356	EXTEND (SP, 2);
		1357	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1358	PUSHs (sv_2mortal (newSVuv (offset)));
		1359	}
		1360
		1361	PROTOTYPES: ENABLE
845		1362
846	void to_json (SV *scalar)	1363	void to_json (SV *scalar)
		1364	ALIAS:
		1365	objToJson = 0
847	PPCODE:	1366	PPCODE:
848	XPUSHs (encode_json (scalar, F_UTF8));	1367	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
849		1368
850	void from_json (SV *jsonstr)	1369	void from_json (SV *jsonstr)
		1370	ALIAS:
		1371	jsonToObj = 0
851	PPCODE:	1372	PPCODE:
852	XPUSHs (decode_json (jsonstr, F_UTF8));	1373	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
853		1374

Diff Legend

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