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 1.44 by root, Mon Jun 25 04:08:17 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
		34	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
16		35
17	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	36	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
18	#define F_DEFAULT 0	37	#define F_DEFAULT (9UL << S_MAXDEPTH)
19		38
20	#define INIT_SIZE 32 // initial scalar size to be allocated	39	#define INIT_SIZE 32 // initial scalar size to be allocated
		40	#define INDENT_STEP 3 // spaces per indentation level
		41
		42	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
21		43
22	#define SB do {	44	#define SB do {
23	#define SE } while (0)	45	#define SE } while (0)
24		46
25	static HV *json_stash;	47	#if __GNUC__ >= 3
		48	# define expect(expr,value) __builtin_expect ((expr),(value))
		49	# define inline inline
		50	#else
		51	# define expect(expr,value) (expr)
		52	# define inline static
		53	#endif
		54
		55	#define expect_false(expr) expect ((expr) != 0, 0)
		56	#define expect_true(expr) expect ((expr) != 0, 1)
		57
		58	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		59	static SV *json_true, *json_false;
		60
		61	/////////////////////////////////////////////////////////////////////////////
		62	// utility functions
		63
		64	static UV *
		65	SvJSON (SV *sv)
		66	{
		67	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		68	croak ("object is not of type JSON::XS");
		69
		70	return &SvUVX (SvRV (sv));
		71	}
		72
		73	static void
		74	shrink (SV *sv)
		75	{
		76	sv_utf8_downgrade (sv, 1);
		77	if (SvLEN (sv) > SvCUR (sv) + 1)
		78	{
		79	#ifdef SvPV_shrink_to_cur
		80	SvPV_shrink_to_cur (sv);
		81	#elif defined (SvPV_renew)
		82	SvPV_renew (sv, SvCUR (sv) + 1);
		83	#endif
		84	}
		85	}
		86
		87	// decode an utf-8 character and return it, or (UV)-1 in
		88	// case of an error.
		89	// we special-case "safe" characters from U+80 .. U+7FF,
		90	// but use the very good perl function to parse anything else.
		91	// note that we never call this function for a ascii codepoints
		92	inline UV
		93	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		94	{
		95	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		98	{
		99	*clen = 2;
		100	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		101	}
		102	else
		103	{
		104	*clen = (STRLEN)-1;
		105	return (UV)-1;
		106	}
		107	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// encoder
26		111
27	// structure used for encoding JSON	112	// structure used for encoding JSON
28	typedef struct	113	typedef struct
29	{	114	{
30	char *cur;	115	char *cur; // SvPVX (sv) + current output position
31	STRLEN len; // SvLEN (sv)
32	char *end; // SvEND (sv)	116	char *end; // SvEND (sv)
33	SV *sv;	117	SV *sv; // result scalar
34	UV flags;	118	U32 flags; // F_*
35	int max_recurse;	119	U32 indent; // indentation level
36	int indent;	120	U32 maxdepth; // max. indentation/recursion level
37	} enc_t;	121	} enc_t;
38		122
39	// structure used for decoding JSON	123	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)	124	need (enc_t *enc, STRLEN len)
61	{	125	{
62	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
63	{	127	{
64	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
65	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
66	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
67	enc->end = SvEND (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
68	}	132	}
69	}	133	}
70		134
71	static void	135	inline void
72	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
73	{	137	{
74	need (enc, 1);	138	need (enc, 1);
75	*enc->cur++ = ch;	139	*enc->cur++ = ch;
76	}	140	}
…		…
78	static void	142	static void
79	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	143	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
80	{	144	{
81	char *end = str + len;	145	char *end = str + len;
82		146
		147	need (enc, len);
		148
83	while (str < end)	149	while (str < end)
84	{	150	{
85	unsigned char ch = *(unsigned char *)str;	151	unsigned char ch = *(unsigned char *)str;
		152
86	if (ch >= 0x20 && ch < 0x80) // most common case	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
87	{	154	{
		155	if (expect_false (ch == '"')) // but with slow exceptions
		156	{
		157	need (enc, len += 1);
		158	*enc->cur++ = '\\';
		159	*enc->cur++ = '"';
		160	}
		161	else if (expect_false (ch == '\\'))
		162	{
		163	need (enc, len += 1);
		164	*enc->cur++ = '\\';
		165	*enc->cur++ = '\\';
		166	}
		167	else
88	*enc->cur++ = ch;	168	*enc->cur++ = ch;
		169
89	str++;	170	++str;
90	}	171	}
91	else	172	else
92	{	173	{
93	STRLEN clen;	174	switch (ch)
94	UV uch;
95
96	if (is_utf8)
97	{	175	{
98	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	176	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
99	if (clen < 0)	177	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");	178	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
101	}	179	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
102	else	180	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
103	{
104	uch = ch;
105	clen = 1;
106	}
107		181
108	need (enc, len += 6);	182	default:
109
110	if (uch < 0xa0 || enc->flags & F_ASCII)
111	{
112	if (uch > 0xFFFFUL)
113	{	183	{
		184	STRLEN clen;
		185	UV uch;
		186
		187	if (is_utf8)
		188	{
		189	uch = decode_utf8 (str, end - str, &clen);
		190	if (clen == (STRLEN)-1)
		191	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		192	}
		193	else
		194	{
		195	uch = ch;
114	len += 6;	196	clen = 1;
		197	}
		198
		199	if (uch > 0x10FFFFUL)
		200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		201
		202	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
		203	{
		204	if (uch > 0xFFFFUL)
		205	{
115	need (enc, len += 6);	206	need (enc, len += 11);
116	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
117	(uch - 0x10000) / 0x400 + 0xD800,	208	(int)((uch - 0x10000) / 0x400 + 0xD800),
118	(uch - 0x10000) % 0x400 + 0xDC00);	209	(int)((uch - 0x10000) % 0x400 + 0xDC00));
119	enc->cur += 12;	210	enc->cur += 12;
		211	}
		212	else
		213	{
		214	static char hexdigit [16] = "0123456789abcdef";
		215	need (enc, len += 5);
		216	*enc->cur++ = '\\';
		217	*enc->cur++ = 'u';
		218	*enc->cur++ = hexdigit [ uch >> 12 ];
		219	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		220	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		221	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		222	}
		223
		224	str += clen;
120	}	225	}
		226	else if (enc->flags & F_LATIN1)
		227	{
		228	*enc->cur++ = uch;
		229	str += clen;
		230	}
		231	else if (is_utf8)
		232	{
		233	need (enc, len += clen);
		234	do
		235	{
		236	*enc->cur++ = *str++;
		237	}
		238	while (--clen);
		239	}
121	else	240	else
122	{	241	{
123	sprintf (enc->cur, "\\u%04x", uch);	242	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
124	enc->cur += 6;	243	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		244	++str;
		245	}
125	}	246	}
126	}	247	}
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	}	248	}
137		249
138	--len;	250	--len;
139	}	251	}
140	}	252	}
141		253
142	#define INDENT SB \	254	inline void
		255	encode_indent (enc_t *enc)
		256	{
143	if (enc->flags & F_INDENT) \	257	if (enc->flags & F_INDENT)
144	{ \	258	{
145	int i_; \	259	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		260
152	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	261	need (enc, spaces);
153	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	262	memset (enc->cur, ' ', spaces);
154	#define COMMA SB \	263	enc->cur += spaces;
		264	}
		265	}
		266
		267	inline void
		268	encode_space (enc_t *enc)
		269	{
		270	need (enc, 1);
155	encode_ch (enc, ','); \	271	encode_ch (enc, ' ');
		272	}
		273
		274	inline void
		275	encode_nl (enc_t *enc)
		276	{
156	if (enc->flags & F_INDENT) \	277	if (enc->flags & F_INDENT)
157	NL; \	278	{
		279	need (enc, 1);
		280	encode_ch (enc, '\n');
		281	}
		282	}
		283
		284	inline void
		285	encode_comma (enc_t *enc)
		286	{
		287	encode_ch (enc, ',');
		288
		289	if (enc->flags & F_INDENT)
		290	encode_nl (enc);
158	else if (enc->flags & F_SPACE_AFTER) \	291	else if (enc->flags & F_SPACE_AFTER)
159	SPACE; \	292	encode_space (enc);
160	SE	293	}
161		294
162	static void encode_sv (enc_t *enc, SV *sv);	295	static void encode_sv (enc_t *enc, SV *sv);
163		296
164	static void	297	static void
165	encode_av (enc_t *enc, AV *av)	298	encode_av (enc_t *enc, AV *av)
166	{	299	{
167	int i, len = av_len (av);	300	int i, len = av_len (av);
168		301
169	encode_ch (enc, '['); NL;	302	if (enc->indent >= enc->maxdepth)
		303	croak ("data structure too deep (hit recursion limit)");
		304
		305	encode_ch (enc, '['); encode_nl (enc);
170	++enc->indent;	306	++enc->indent;
171		307
172	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
173	{	309	{
174	INDENT;	310	encode_indent (enc);
175	encode_sv (enc, *av_fetch (av, i, 0));	311	encode_sv (enc, *av_fetch (av, i, 0));
176		312
177	if (i < len)	313	if (i < len)
178	COMMA;	314	encode_comma (enc);
179	}	315	}
180		316
181	NL;	317	encode_nl (enc);
182		318
183	--enc->indent;	319	--enc->indent;
184	INDENT; encode_ch (enc, ']');	320	encode_indent (enc); encode_ch (enc, ']');
185	}	321	}
186		322
187	static void	323	static void
188	encode_he (enc_t *enc, HE *he)	324	encode_he (enc_t *enc, HE *he)
189	{	325	{
…		…
191		327
192	if (HeKLEN (he) == HEf_SVKEY)	328	if (HeKLEN (he) == HEf_SVKEY)
193	{	329	{
194	SV *sv = HeSVKEY (he);	330	SV *sv = HeSVKEY (he);
195	STRLEN len;	331	STRLEN len;
		332	char *str;
		333
		334	SvGETMAGIC (sv);
196	char *str = SvPV (sv, len);	335	str = SvPV (sv, len);
197		336
198	encode_str (enc, str, len, SvUTF8 (sv));	337	encode_str (enc, str, len, SvUTF8 (sv));
199	}	338	}
200	else	339	else
201	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	340	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
202		341
203	encode_ch (enc, '"');	342	encode_ch (enc, '"');
204		343
205	if (enc->flags & F_SPACE_BEFORE) SPACE;	344	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
206	encode_ch (enc, ':');	345	encode_ch (enc, ':');
207	if (enc->flags & F_SPACE_AFTER ) SPACE;	346	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
208	encode_sv (enc, HeVAL (he));	347	encode_sv (enc, HeVAL (he));
209	}	348	}
210		349
211	// compare hash entries, used when all keys are bytestrings	350	// compare hash entries, used when all keys are bytestrings
212	static int	351	static int
…		…
218	HE *b = *(HE **)b_;	357	HE *b = *(HE **)b_;
219		358
220	STRLEN la = HeKLEN (a);	359	STRLEN la = HeKLEN (a);
221	STRLEN lb = HeKLEN (b);	360	STRLEN lb = HeKLEN (b);
222		361
223	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
224	cmp = la < lb ? -1 : la == lb ? 0 : 1;	363	cmp = la - lb;
225		364
226	return cmp;	365	return cmp;
227	}	366	}
228		367
229	// compare hash entries, used when some keys are sv's or utf-x	368	// compare hash entries, used when some keys are sv's or utf-x
…		…
236	static void	375	static void
237	encode_hv (enc_t *enc, HV *hv)	376	encode_hv (enc_t *enc, HV *hv)
238	{	377	{
239	int count, i;	378	int count, i;
240		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
241	encode_ch (enc, '{'); NL; ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
242		384
243	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
244	{	386	{
245	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
246	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
247	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
248	// that randomises hash orderings	390	// that randomises hash orderings
249	if (enc->flags & F_CANONICAL)	391	if (enc->flags & F_CANONICAL)
250	{	392	{
251	HE *he, *hes [count];
252	int fast = 1;	393	int fast = 1;
		394	HE *he;
		395	#if defined(__BORLANDC__) || defined(_MSC_VER)
		396	HE **hes = _alloca (count * sizeof (HE));
		397	#else
		398	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		399	#endif
253		400
254	i = 0;	401	i = 0;
255	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
256	{	403	{
257	hes [i++] = he;	404	hes [i++] = he;
…		…
263		410
264	if (fast)	411	if (fast)
265	qsort (hes, count, sizeof (HE *), he_cmp_fast);	412	qsort (hes, count, sizeof (HE *), he_cmp_fast);
266	else	413	else
267	{	414	{
268	// hack to disable "use bytes"	415	// hack to forcefully disable "use bytes"
269	COP *oldcop = PL_curcop, cop;	416	COP cop = *PL_curcop;
270	cop.op_private = 0;	417	cop.op_private = 0;
		418
		419	ENTER;
		420	SAVETMPS;
		421
		422	SAVEVPTR (PL_curcop);
271	PL_curcop = &cop;	423	PL_curcop = &cop;
272		424
273	SAVETMPS;
274	qsort (hes, count, sizeof (HE *), he_cmp_slow);	425	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		426
275	FREETMPS;	427	FREETMPS;
276		428	LEAVE;
277	PL_curcop = oldcop;
278	}	429	}
279		430
280	for (i = 0; i < count; ++i)	431	for (i = 0; i < count; ++i)
281	{	432	{
282	INDENT;	433	encode_indent (enc);
283	encode_he (enc, hes [i]);	434	encode_he (enc, hes [i]);
284		435
285	if (i < count - 1)	436	if (i < count - 1)
286	COMMA;	437	encode_comma (enc);
287	}	438	}
288		439
289	NL;	440	encode_nl (enc);
290	}	441	}
291	else	442	else
292	{	443	{
293	SV *sv;
294	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
295		445
296	for (;;)	446	for (;;)
297	{	447	{
298	INDENT;	448	encode_indent (enc);
299	encode_he (enc, he);	449	encode_he (enc, he);
300		450
301	if (!(he = hv_iternext (hv)))	451	if (!(he = hv_iternext (hv)))
302	break;	452	break;
303		453
304	COMMA;	454	encode_comma (enc);
305	}	455	}
306		456
307	NL;	457	encode_nl (enc);
308	}	458	}
309	}	459	}
310		460
311	--enc->indent; INDENT; encode_ch (enc, '}');	461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		462	}
		463
		464	// encode objects, arrays and special \0=false and \1=true values.
		465	static void
		466	encode_rv (enc_t *enc, SV *sv)
		467	{
		468	svtype svt;
		469
		470	SvGETMAGIC (sv);
		471	svt = SvTYPE (sv);
		472
		473	if (expect_false (SvOBJECT (sv)))
		474	{
		475	if (SvSTASH (sv) == json_boolean_stash)
		476	{
		477	if (SvIV (sv) == 0)
		478	encode_str (enc, "false", 5, 0);
		479	else
		480	encode_str (enc, "true", 4, 0);
		481	}
		482	else
		483	{
		484	#if 0
		485	if (0 && sv_derived_from (rv, "JSON::Literal"))
		486	{
		487	// not yet
		488	}
		489	#endif
		490	if (enc->flags & F_CONV_BLESSED)
		491	{
		492	// we re-bless the reference to get overload and other niceties right
		493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		494
		495	if (to_json)
		496	{
		497	dSP;
		498	ENTER;
		499	SAVETMPS;
		500	PUSHMARK (SP);
		501	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		502
		503	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		504	// check anyways.
		505	PUTBACK;
		506	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		507	SPAGAIN;
		508
		509	encode_sv (enc, POPs);
		510
		511	FREETMPS;
		512	LEAVE;
		513	}
		514	else if (enc->flags & F_ALLOW_BLESSED)
		515	encode_str (enc, "null", 4, 0);
		516	else
		517	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		518	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		519	}
		520	else if (enc->flags & F_ALLOW_BLESSED)
		521	encode_str (enc, "null", 4, 0);
		522	else
		523	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		524	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		525	}
		526	}
		527	else if (svt == SVt_PVHV)
		528	encode_hv (enc, (HV *)sv);
		529	else if (svt == SVt_PVAV)
		530	encode_av (enc, (AV *)sv);
		531	else if (svt < SVt_PVAV)
		532	{
		533	if (SvNIOK (sv) && SvIV (sv) == 0)
		534	encode_str (enc, "false", 5, 0);
		535	else if (SvNIOK (sv) && SvIV (sv) == 1)
		536	encode_str (enc, "true", 4, 0);
		537	else
		538	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		539	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		540	}
		541	else
		542	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		543	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
312	}	544	}
313		545
314	static void	546	static void
315	encode_sv (enc_t *enc, SV *sv)	547	encode_sv (enc_t *enc, SV *sv)
316	{	548	{
		549	SvGETMAGIC (sv);
		550
317	if (SvPOKp (sv))	551	if (SvPOKp (sv))
318	{	552	{
319	STRLEN len;	553	STRLEN len;
320	char *str = SvPV (sv, len);	554	char *str = SvPV (sv, len);
321	encode_ch (enc, '"');	555	encode_ch (enc, '"');
322	encode_str (enc, str, len, SvUTF8 (sv));	556	encode_str (enc, str, len, SvUTF8 (sv));
323	encode_ch (enc, '"');	557	encode_ch (enc, '"');
324	}	558	}
325	else if (SvNOKp (sv))	559	else if (SvNOKp (sv))
326	{	560	{
		561	// trust that perl will do the right thing w.r.t. JSON syntax.
327	need (enc, NV_DIG + 32);	562	need (enc, NV_DIG + 32);
328	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	563	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
329	enc->cur += strlen (enc->cur);	564	enc->cur += strlen (enc->cur);
330	}	565	}
331	else if (SvIOKp (sv))	566	else if (SvIOKp (sv))
332	{	567	{
333	need (enc, 64);	568	// we assume we can always read an IV as a UV
		569	if (SvUV (sv) & ~(UV)0x7fff)
		570	{
		571	// large integer, use the (rather slow) snprintf way.
		572	need (enc, sizeof (UV) * 3);
334	enc->cur +=	573	enc->cur +=
335	SvIsUV(sv)	574	SvIsUV(sv)
336	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	575	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
337	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	576	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		577	}
		578	else
		579	{
		580	// optimise the "small number case"
		581	// code will likely be branchless and use only a single multiplication
		582	I32 i = SvIV (sv);
		583	U32 u;
		584	char digit, nz = 0;
		585
		586	need (enc, 6);
		587
		588	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		589	u = i < 0 ? -i : i;
		590
		591	// convert to 4.28 fixed-point representation
		592	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		593
		594	// now output digit by digit, each time masking out the integer part
		595	// and multiplying by 5 while moving the decimal point one to the right,
		596	// resulting in a net multiplication by 10.
		597	// we always write the digit to memory but conditionally increment
		598	// the pointer, to ease the usage of conditional move instructions.
		599	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		600	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		601	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		602	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		603	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		604	}
338	}	605	}
339	else if (SvROK (sv))	606	else if (SvROK (sv))
340	{	607	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))	608	else if (!SvOK (sv))
356	encode_str (enc, "null", 4, 0);	609	encode_str (enc, "null", 4, 0);
357	else	610	else
358	croak ("encountered perl type that JSON cannot handle");	611	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		612	SvPV_nolen (sv), SvFLAGS (sv));
359	}	613	}
360		614
361	static SV *	615	static SV *
362	encode_json (SV *scalar, UV flags)	616	encode_json (SV *scalar, U32 flags)
363	{	617	{
364	enc_t enc;	618	enc_t enc;
		619
		620	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
		621	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		622
365	enc.flags = flags;	623	enc.flags = flags;
366	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	624	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
367	enc.cur = SvPVX (enc.sv);	625	enc.cur = SvPVX (enc.sv);
368	enc.end = SvEND (enc.sv);	626	enc.end = SvEND (enc.sv);
369	enc.max_recurse = 0;
370	enc.indent = 0;	627	enc.indent = 0;
		628	enc.maxdepth = DEC_DEPTH (flags);
371		629
372	SvPOK_only (enc.sv);	630	SvPOK_only (enc.sv);
373	encode_sv (&enc, scalar);	631	encode_sv (&enc, scalar);
374		632
		633	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		634	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		635
375	if (!(flags & (F_ASCII | F_UTF8)))	636	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
376	SvUTF8_on (enc.sv);	637	SvUTF8_on (enc.sv);
377		638
378	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	639	if (enc.flags & F_SHRINK)
		640	shrink (enc.sv);
		641
379	return enc.sv;	642	return enc.sv;
380	}	643	}
381		644
382	/////////////////////////////////////////////////////////////////////////////	645	/////////////////////////////////////////////////////////////////////////////
		646	// decoder
383		647
384	#define WS \	648	// structure used for decoding JSON
		649	typedef struct
		650	{
		651	char *cur; // current parser pointer
		652	char *end; // end of input string
		653	const char *err; // parse error, if != 0
		654	U32 flags; // F_*
		655	U32 depth; // recursion depth
		656	U32 maxdepth; // recursion depth limit
		657	} dec_t;
		658
		659	inline void
		660	decode_ws (dec_t *dec)
		661	{
385	for (;;) \	662	for (;;)
386	{ \	663	{
387	char ch = *dec->cur; \	664	char ch = *dec->cur;
		665
388	if (ch > 0x20 \	666	if (ch > 0x20
389	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	667	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
390	break; \	668	break;
		669
391	++dec->cur; \	670	++dec->cur;
392	}	671	}
		672	}
393		673
394	#define ERR(reason) SB dec->err = reason; goto fail; SE	674	#define ERR(reason) SB dec->err = reason; goto fail; SE
		675
395	#define EXPECT_CH(ch) SB \	676	#define EXPECT_CH(ch) SB \
396	if (*dec->cur != ch) \	677	if (*dec->cur != ch) \
397	ERR (# ch " expected"); \	678	ERR (# ch " expected"); \
398	++dec->cur; \	679	++dec->cur; \
399	SE	680	SE
400		681
		682	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		683	#define DEC_DEC_DEPTH --dec->depth
		684
401	static SV *decode_sv (dec_t *dec);	685	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		686
408	static signed char decode_hexdigit[256];	687	static signed char decode_hexdigit[256];
409		688
410	static UV	689	static UV
411	decode_4hex (dec_t *dec)	690	decode_4hex (dec_t *dec)
412	{	691	{
413	signed char d1, d2, d3, d4;	692	signed char d1, d2, d3, d4;
		693	unsigned char *cur = (unsigned char *)dec->cur;
414		694
415	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	695	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");	696	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]];	697	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");	698	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		699
424	dec->cur += 4;	700	dec->cur += 4;
425		701
426	return ((UV)d1) << 12	702	return ((UV)d1) << 12
427	| ((UV)d2) << 8	703	| ((UV)d2) << 8
…		…
433	}	709	}
434		710
435	static SV *	711	static SV *
436	decode_str (dec_t *dec)	712	decode_str (dec_t *dec)
437	{	713	{
438	SV *sv = NEWSV (0,2);	714	SV *sv = 0;
439	STRLEN cur = 0;
440	int utf8 = 0;	715	int utf8 = 0;
		716	char *dec_cur = dec->cur;
441		717
442	for (;;)	718	do
443	{	719	{
444	unsigned char ch = *(unsigned char *)dec->cur;	720	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		721	char *cur = buf;
445		722
446	if (ch == '"')	723	do
447	break;
448	else if (ch == '\\')
449	{	724	{
450	switch (*++dec->cur)	725	unsigned char ch = *(unsigned char *)dec_cur++;
		726
		727	if (expect_false (ch == '"'))
451	{	728	{
452	case '\\':	729	--dec_cur;
453	case '/':	730	break;
454	case '"': APPEND_CH (*dec->cur++); break;	731	}
455		732	else if (expect_false (ch == '\\'))
456	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	733	{
457	case 't': APPEND_CH ('\011'); ++dec->cur; break;	734	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	{	735	{
464	UV lo, hi;	736	case '\\':
465	++dec->cur;	737	case '/':
		738	case '"': *cur++ = *dec_cur++; break;
466		739
467	hi = decode_4hex (dec);	740	case 'b': ++dec_cur; *cur++ = '\010'; break;
468	if (hi == (UV)-1)	741	case 't': ++dec_cur; *cur++ = '\011'; break;
469	goto fail;	742	case 'n': ++dec_cur; *cur++ = '\012'; break;
		743	case 'f': ++dec_cur; *cur++ = '\014'; break;
		744	case 'r': ++dec_cur; *cur++ = '\015'; break;
470		745
471	// possibly a surrogate pair	746	case 'u':
472	if (hi >= 0xd800 && hi < 0xdc00)
473	{	747	{
474	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	748	UV lo, hi;
475	ERR ("illegal surrogate character");	749	++dec_cur;
476		750
477	dec->cur += 2;	751	dec->cur = dec_cur;
478
479	lo = decode_4hex (dec);	752	hi = decode_4hex (dec);
		753	dec_cur = dec->cur;
480	if (lo == (UV)-1)	754	if (hi == (UV)-1)
481	goto fail;	755	goto fail;
482		756
		757	// possibly a surrogate pair
		758	if (hi >= 0xd800)
		759	if (hi < 0xdc00)
		760	{
		761	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		762	ERR ("missing low surrogate character in surrogate pair");
		763
		764	dec_cur += 2;
		765
		766	dec->cur = dec_cur;
		767	lo = decode_4hex (dec);
		768	dec_cur = dec->cur;
		769	if (lo == (UV)-1)
		770	goto fail;
		771
483	if (lo < 0xdc00 || lo >= 0xe000)	772	if (lo < 0xdc00 || lo >= 0xe000)
484	ERR ("surrogate pair expected");	773	ERR ("surrogate pair expected");
485		774
486	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	775	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		776	}
		777	else if (hi < 0xe000)
		778	ERR ("missing high surrogate character in surrogate pair");
		779
		780	if (hi >= 0x80)
		781	{
		782	utf8 = 1;
		783
		784	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		785	}
		786	else
		787	*cur++ = hi;
487	}	788	}
488	else if (lo >= 0xdc00 && lo < 0xe000)
489	ERR ("illegal surrogate character");
490
491	if (hi >= 0x80)
492	{	789	break;
493	utf8 = 1;
494		790
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	791	default:
499	APPEND_CH (hi);	792	--dec_cur;
		793	ERR ("illegal backslash escape sequence in string");
500	}	794	}
		795	}
		796	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		797	*cur++ = ch;
		798	else if (ch >= 0x80)
		799	{
		800	STRLEN clen;
		801	UV uch;
		802
		803	--dec_cur;
		804
		805	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		806	if (clen == (STRLEN)-1)
		807	ERR ("malformed UTF-8 character in JSON string");
		808
		809	do
		810	*cur++ = *dec_cur++;
		811	while (--clen);
		812
		813	utf8 = 1;
		814	}
		815	else
		816	{
		817	--dec_cur;
		818
		819	if (!ch)
		820	ERR ("unexpected end of string while parsing JSON string");
501	break;	821	else
		822	ERR ("invalid character encountered while parsing JSON string");
502	}	823	}
503	}	824	}
504	else if (ch >= 0x20 && ch <= 0x7f)	825	while (cur < buf + SHORT_STRING_LEN);
505	APPEND_CH (*dec->cur++);	826
506	else if (ch >= 0x80)	827	{
		828	STRLEN len = cur - buf;
		829
		830	if (sv)
507	{	831	{
508	STRLEN clen;	832	SvGROW (sv, SvCUR (sv) + len + 1);
509	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	833	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
510	if (clen < 0)	834	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	}	835	}
517	else	836	else
518	ERR ("invalid character encountered");	837	sv = newSVpvn (buf, len);
519	}	838	}
		839	}
		840	while (*dec_cur != '"');
520		841
521	++dec->cur;	842	++dec_cur;
522		843
		844	if (sv)
		845	{
523	SvPOK_only (sv);	846	SvPOK_only (sv);
524
525	SvCUR_set (sv, cur);
526	*SvEND (sv) = 0;	847	*SvEND (sv) = 0;
527		848
528	if (utf8)	849	if (utf8)
529	SvUTF8_on (sv);	850	SvUTF8_on (sv);
		851	}
		852	else
		853	sv = newSVpvn ("", 0);
530		854
		855	dec->cur = dec_cur;
531	return sv;	856	return sv;
532		857
533	fail:	858	fail:
534	SvREFCNT_dec (sv);	859	dec->cur = dec_cur;
535	return 0;	860	return 0;
536	}	861	}
537		862
538	static SV *	863	static SV *
539	decode_num (dec_t *dec)	864	decode_num (dec_t *dec)
…		…
549	{	874	{
550	++dec->cur;	875	++dec->cur;
551	if (*dec->cur >= '0' && *dec->cur <= '9')	876	if (*dec->cur >= '0' && *dec->cur <= '9')
552	ERR ("malformed number (leading zero must not be followed by another digit)");	877	ERR ("malformed number (leading zero must not be followed by another digit)");
553	}	878	}
554		879	else if (*dec->cur < '0' || *dec->cur > '9')
555	// int	880	ERR ("malformed number (no digits after initial minus)");
		881	else
		882	do
		883	{
		884	++dec->cur;
		885	}
556	while (*dec->cur >= '0' && *dec->cur <= '9')	886	while (*dec->cur >= '0' && *dec->cur <= '9');
557	++dec->cur;
558		887
559	// [frac]	888	// [frac]
560	if (*dec->cur == '.')	889	if (*dec->cur == '.')
561	{	890	{
562	is_nv = 1;	891	++dec->cur;
		892
		893	if (*dec->cur < '0' || *dec->cur > '9')
		894	ERR ("malformed number (no digits after decimal point)");
563		895
564	do	896	do
565	{	897	{
566	++dec->cur;	898	++dec->cur;
567	}	899	}
568	while (*dec->cur >= '0' && *dec->cur <= '9');	900	while (*dec->cur >= '0' && *dec->cur <= '9');
		901
		902	is_nv = 1;
569	}	903	}
570		904
571	// [exp]	905	// [exp]
572	if (*dec->cur == 'e' || *dec->cur == 'E')	906	if (*dec->cur == 'e' || *dec->cur == 'E')
573	{	907	{
574	is_nv = 1;
575
576	++dec->cur;	908	++dec->cur;
		909
577	if (*dec->cur == '-' || *dec->cur == '+')	910	if (*dec->cur == '-' || *dec->cur == '+')
578	++dec->cur;	911	++dec->cur;
579		912
		913	if (*dec->cur < '0' || *dec->cur > '9')
		914	ERR ("malformed number (no digits after exp sign)");
		915
		916	do
		917	{
		918	++dec->cur;
		919	}
580	while (*dec->cur >= '0' && *dec->cur <= '9')	920	while (*dec->cur >= '0' && *dec->cur <= '9');
581	++dec->cur;	921
		922	is_nv = 1;
582	}	923	}
583		924
584	if (!is_nv)	925	if (!is_nv)
585	{	926	{
586	UV uv;	927	// 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);	928	if (*start == '-')
588	if (numtype & IS_NUMBER_IN_UV)	929	switch (dec->cur - start)
589	if (numtype & IS_NUMBER_NEG)
590	{	930	{
591	if (uv < (UV)IV_MIN)	931	case 2: return newSViv (-( start [1] - '0' * 1));
592	return newSViv (-(IV)uv);	932	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		933	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		934	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
593	}	935	}
		936	else
		937	switch (dec->cur - start)
		938	{
		939	case 1: return newSViv ( start [0] - '0' * 1);
		940	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		941	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		942	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		943	}
		944
		945	{
		946	UV uv;
		947	int numtype = grok_number (start, dec->cur - start, &uv);
		948	if (numtype & IS_NUMBER_IN_UV)
		949	if (numtype & IS_NUMBER_NEG)
		950	{
		951	if (uv < (UV)IV_MIN)
		952	return newSViv (-(IV)uv);
		953	}
594	else	954	else
595	return newSVuv (uv);	955	return newSVuv (uv);
		956
		957	// here would likely be the place for bigint support
596	}	958	}
		959	}
597		960
		961	// if we ever support bigint or bigfloat, this is the place for bigfloat
598	return newSVnv (Atof (start));	962	return newSVnv (Atof (start));
599		963
600	fail:	964	fail:
601	return 0;	965	return 0;
602	}	966	}
…		…
604	static SV *	968	static SV *
605	decode_av (dec_t *dec)	969	decode_av (dec_t *dec)
606	{	970	{
607	AV *av = newAV ();	971	AV *av = newAV ();
608		972
		973	DEC_INC_DEPTH;
		974	decode_ws (dec);
		975
		976	if (*dec->cur == ']')
		977	++dec->cur;
		978	else
609	for (;;)	979	for (;;)
610	{	980	{
611	SV *value;	981	SV *value;
612		982
613	value = decode_sv (dec);	983	value = decode_sv (dec);
614	if (!value)	984	if (!value)
615	goto fail;	985	goto fail;
616		986
617	av_push (av, value);	987	av_push (av, value);
618		988
619	WS;	989	decode_ws (dec);
620		990
621	if (*dec->cur == ']')	991	if (*dec->cur == ']')
622	{	992	{
623	++dec->cur;	993	++dec->cur;
624	break;	994	break;
		995	}
625	}	996
626
627	if (*dec->cur != ',')	997	if (*dec->cur != ',')
628	ERR (", or ] expected while parsing array");	998	ERR (", or ] expected while parsing array");
629		999
630	++dec->cur;	1000	++dec->cur;
631	}	1001	}
632		1002
		1003	DEC_DEC_DEPTH;
633	return newRV_noinc ((SV *)av);	1004	return newRV_noinc ((SV *)av);
634		1005
635	fail:	1006	fail:
636	SvREFCNT_dec (av);	1007	SvREFCNT_dec (av);
		1008	DEC_DEC_DEPTH;
637	return 0;	1009	return 0;
638	}	1010	}
639		1011
640	static SV *	1012	static SV *
641	decode_hv (dec_t *dec)	1013	decode_hv (dec_t *dec)
642	{	1014	{
643	HV *hv = newHV ();	1015	HV *hv = newHV ();
644		1016
		1017	DEC_INC_DEPTH;
		1018	decode_ws (dec);
		1019
		1020	if (*dec->cur == '}')
		1021	++dec->cur;
		1022	else
645	for (;;)	1023	for (;;)
646	{	1024	{
647	SV *key, *value;	1025	SV *key, *value;
648		1026
649	WS; EXPECT_CH ('"');	1027	decode_ws (dec); EXPECT_CH ('"');
650		1028
651	key = decode_str (dec);	1029	key = decode_str (dec);
652	if (!key)	1030	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;	1031	goto fail;
		1032
		1033	decode_ws (dec); EXPECT_CH (':');
		1034
		1035	value = decode_sv (dec);
		1036	if (!value)
		1037	{
		1038	SvREFCNT_dec (key);
		1039	goto fail;
662	}	1040	}
663		1041
664	//TODO: optimise
665	hv_store_ent (hv, key, value, 0);	1042	hv_store_ent (hv, key, value, 0);
		1043	SvREFCNT_dec (key);
666		1044
667	WS;	1045	decode_ws (dec);
668		1046
669	if (*dec->cur == '}')	1047	if (*dec->cur == '}')
670	{	1048	{
671	++dec->cur;	1049	++dec->cur;
672	break;	1050	break;
673	}	1051	}
674		1052
675	if (*dec->cur != ',')	1053	if (*dec->cur != ',')
676	ERR (", or } expected while parsing object/hash");	1054	ERR (", or } expected while parsing object/hash");
677		1055
678	++dec->cur;	1056	++dec->cur;
679	}	1057	}
680		1058
		1059	DEC_DEC_DEPTH;
681	return newRV_noinc ((SV *)hv);	1060	return newRV_noinc ((SV *)hv);
682		1061
683	fail:	1062	fail:
684	SvREFCNT_dec (hv);	1063	SvREFCNT_dec (hv);
		1064	DEC_DEC_DEPTH;
685	return 0;	1065	return 0;
686	}	1066	}
687		1067
688	static SV *	1068	static SV *
689	decode_sv (dec_t *dec)	1069	decode_sv (dec_t *dec)
690	{	1070	{
691	WS;	1071	decode_ws (dec);
		1072
		1073	// the beauty of JSON: you need exactly one character lookahead
		1074	// to parse anything.
692	switch (*dec->cur)	1075	switch (*dec->cur)
693	{	1076	{
694	case '"': ++dec->cur; return decode_str (dec);	1077	case '"': ++dec->cur; return decode_str (dec);
695	case '[': ++dec->cur; return decode_av (dec);	1078	case '[': ++dec->cur; return decode_av (dec);
696	case '{': ++dec->cur; return decode_hv (dec);	1079	case '{': ++dec->cur; return decode_hv (dec);
…		…
702		1085
703	case 't':	1086	case 't':
704	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1087	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
705	{	1088	{
706	dec->cur += 4;	1089	dec->cur += 4;
707	return newSViv (1);	1090	return SvREFCNT_inc (json_true);
708	}	1091	}
709	else	1092	else
710	ERR ("'true' expected");	1093	ERR ("'true' expected");
711		1094
712	break;	1095	break;
713		1096
714	case 'f':	1097	case 'f':
715	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1098	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
716	{	1099	{
717	dec->cur += 5;	1100	dec->cur += 5;
718	return newSViv (0);	1101	return SvREFCNT_inc (json_false);
719	}	1102	}
720	else	1103	else
721	ERR ("'false' expected");	1104	ERR ("'false' expected");
722		1105
723	break;	1106	break;
724		1107
725	case 'n':	1108	case 'n':
726	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1109	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
727	{	1110	{
728	dec->cur += 4;	1111	dec->cur += 4;
729	return newSViv (1);	1112	return newSVsv (&PL_sv_undef);
730	}	1113	}
731	else	1114	else
732	ERR ("'null' expected");	1115	ERR ("'null' expected");
733		1116
734	break;	1117	break;
735		1118
736	default:	1119	default:
737	ERR ("malformed json string");	1120	ERR ("malformed JSON string, neither array, object, number, string or atom");
738	break;	1121	break;
739	}	1122	}
740		1123
741	fail:	1124	fail:
742	return 0;	1125	return 0;
743	}	1126	}
744		1127
745	static SV *	1128	static SV *
746	decode_json (SV *string, UV flags)	1129	decode_json (SV *string, U32 flags, UV *offset_return)
747	{	1130	{
		1131	dec_t dec;
		1132	UV offset;
748	SV *sv;	1133	SV *sv;
749		1134
		1135	SvGETMAGIC (string);
		1136	SvUPGRADE (string, SVt_PV);
		1137
750	if (!(flags & F_UTF8))	1138	if (flags & F_UTF8)
		1139	sv_utf8_downgrade (string, 0);
		1140	else
751	sv_utf8_upgrade (string);	1141	sv_utf8_upgrade (string);
752		1142
753	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1143	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
754		1144
755	dec_t dec;
756	dec.flags = flags;	1145	dec.flags = flags;
757	dec.cur = SvPVX (string);	1146	dec.cur = SvPVX (string);
758	dec.end = SvEND (string);	1147	dec.end = SvEND (string);
759	dec.err = 0;	1148	dec.err = 0;
		1149	dec.depth = 0;
		1150	dec.maxdepth = DEC_DEPTH (dec.flags);
760		1151
761	*dec.end = 1; // invalid anywhere	1152	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
762	sv = decode_sv (&dec);	1153	sv = decode_sv (&dec);
763	*dec.end = 0;	1154
		1155	if (!(offset_return || !sv))
		1156	{
		1157	// check for trailing garbage
		1158	decode_ws (&dec);
		1159
		1160	if (*dec.cur)
		1161	{
		1162	dec.err = "garbage after JSON object";
		1163	SvREFCNT_dec (sv);
		1164	sv = 0;
		1165	}
		1166	}
		1167
		1168	if (offset_return || !sv)
		1169	{
		1170	offset = dec.flags & F_UTF8
		1171	? dec.cur - SvPVX (string)
		1172	: utf8_distance (dec.cur, SvPVX (string));
		1173
		1174	if (offset_return)
		1175	*offset_return = offset;
		1176	}
764		1177
765	if (!sv)	1178	if (!sv)
766	{	1179	{
767	IV offset = utf8_distance (dec.cur, SvPVX (string));
768	SV *uni = sv_newmortal ();	1180	SV *uni = sv_newmortal ();
769		1181
		1182	// horrible hack to silence warning inside pv_uni_display
		1183	COP cop = *PL_curcop;
		1184	cop.cop_warnings = pWARN_NONE;
		1185	ENTER;
		1186	SAVEVPTR (PL_curcop);
		1187	PL_curcop = &cop;
770	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1188	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1189	LEAVE;
		1190
771	croak ("%s, at character %d (%s)",	1191	croak ("%s, at character offset %d [\"%s\"]",
772	dec.err,	1192	dec.err,
773	(int)offset,	1193	(int)offset,
774	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1194	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
775	}	1195	}
776		1196
777	sv_dump (sv);//D
778	return sv_2mortal (sv);	1197	sv = sv_2mortal (sv);
		1198
		1199	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
		1200	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
		1201
		1202	return sv;
779	}	1203	}
		1204
		1205	/////////////////////////////////////////////////////////////////////////////
		1206	// XS interface functions
780		1207
781	MODULE = JSON::XS PACKAGE = JSON::XS	1208	MODULE = JSON::XS PACKAGE = JSON::XS
782		1209
783	BOOT:	1210	BOOT:
784	{	1211	{
785	int i;	1212	int i;
786		1213
787	memset (decode_hexdigit, 0xff, 256);
788	for (i = 10; i--; )	1214	for (i = 0; i < 256; ++i)
789	decode_hexdigit ['0' + i] = i;	1215	decode_hexdigit [i] =
		1216	i >= '0' && i <= '9' ? i - '0'
		1217	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1218	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1219	: -1;
790		1220
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);	1221	json_stash = gv_stashpv ("JSON::XS" , 1);
		1222	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1223
		1224	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1225	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
798	}	1226	}
		1227
		1228	PROTOTYPES: DISABLE
799		1229
800	SV *new (char *dummy)	1230	SV *new (char *dummy)
801	CODE:	1231	CODE:
802	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1232	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
803	OUTPUT:	1233	OUTPUT:
804	RETVAL	1234	RETVAL
805		1235
806	SV *ascii (SV *self, int enable)	1236	SV *ascii (SV *self, int enable = 1)
807	ALIAS:	1237	ALIAS:
808	ascii = F_ASCII	1238	ascii = F_ASCII
		1239	latin1 = F_LATIN1
809	utf8 = F_UTF8	1240	utf8 = F_UTF8
810	indent = F_INDENT	1241	indent = F_INDENT
811	canonical = F_CANONICAL	1242	canonical = F_CANONICAL
812	space_before = F_SPACE_BEFORE	1243	space_before = F_SPACE_BEFORE
813	space_after = F_SPACE_AFTER	1244	space_after = F_SPACE_AFTER
814	json_rpc = F_JSON_RPC
815	pretty = F_PRETTY	1245	pretty = F_PRETTY
		1246	allow_nonref = F_ALLOW_NONREF
		1247	shrink = F_SHRINK
		1248	allow_blessed = F_ALLOW_BLESSED
		1249	convert_blessed = F_CONV_BLESSED
816	CODE:	1250	CODE:
817	{	1251	{
818	UV *uv = SvJSON (self);	1252	UV *uv = SvJSON (self);
819	if (enable)	1253	if (enable)
820	*uv |= ix;	1254	*uv |= ix;
…		…
824	RETVAL = newSVsv (self);	1258	RETVAL = newSVsv (self);
825	}	1259	}
826	OUTPUT:	1260	OUTPUT:
827	RETVAL	1261	RETVAL
828		1262
		1263	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1264	CODE:
		1265	{
		1266	UV *uv = SvJSON (self);
		1267	UV log2 = 0;
		1268
		1269	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1270
		1271	while ((1UL << log2) < max_depth)
		1272	++log2;
		1273
		1274	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1275
		1276	RETVAL = newSVsv (self);
		1277	}
		1278	OUTPUT:
		1279	RETVAL
		1280
829	void encode (SV *self, SV *scalar)	1281	void encode (SV *self, SV *scalar)
830	PPCODE:	1282	PPCODE:
831	XPUSHs (encode_json (scalar, *SvJSON (self)));	1283	XPUSHs (encode_json (scalar, *SvJSON (self)));
832		1284
833	void decode (SV *self, SV *jsonstr)	1285	void decode (SV *self, SV *jsonstr)
834	PPCODE:	1286	PPCODE:
835	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1287	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1288
		1289	void decode_prefix (SV *self, SV *jsonstr)
		1290	PPCODE:
		1291	{
		1292	UV offset;
		1293	EXTEND (SP, 2);
		1294	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1295	PUSHs (sv_2mortal (newSVuv (offset)));
		1296	}
		1297
		1298	PROTOTYPES: ENABLE
836		1299
837	void to_json (SV *scalar)	1300	void to_json (SV *scalar)
		1301	ALIAS:
		1302	objToJson = 0
838	PPCODE:	1303	PPCODE:
839	XPUSHs (encode_json (scalar, F_UTF8));	1304	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
840		1305
841	void from_json (SV *jsonstr)	1306	void from_json (SV *jsonstr)
		1307	ALIAS:
		1308	jsonToObj = 0
842	PPCODE:	1309	PPCODE:
843	XPUSHs (decode_json (jsonstr, F_UTF8));	1310	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
844		1311

Diff Legend

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