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 1.48 by root, Sun Jul 1 22:20:00 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	typedef struct {
		65	U32 flags;
		66	} JSON;
		67
		68	/////////////////////////////////////////////////////////////////////////////
		69	// utility functions
		70
		71	inline void
		72	shrink (SV *sv)
		73	{
		74	sv_utf8_downgrade (sv, 1);
		75	if (SvLEN (sv) > SvCUR (sv) + 1)
		76	{
		77	#ifdef SvPV_shrink_to_cur
		78	SvPV_shrink_to_cur (sv);
		79	#elif defined (SvPV_renew)
		80	SvPV_renew (sv, SvCUR (sv) + 1);
		81	#endif
		82	}
		83	}
		84
		85	// decode an utf-8 character and return it, or (UV)-1 in
		86	// case of an error.
		87	// we special-case "safe" characters from U+80 .. U+7FF,
		88	// but use the very good perl function to parse anything else.
		89	// note that we never call this function for a ascii codepoints
		90	inline UV
		91	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		92	{
		93	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		94	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		95	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		96	{
		97	*clen = 2;
		98	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		99	}
		100	else
		101	{
		102	*clen = (STRLEN)-1;
		103	return (UV)-1;
		104	}
		105	}
		106
		107	/////////////////////////////////////////////////////////////////////////////
		108	// encoder
27		109
28	// structure used for encoding JSON	110	// structure used for encoding JSON
29	typedef struct	111	typedef struct
30	{	112	{
31	char *cur;	113	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	114	char *end; // SvEND (sv)
34	SV *sv;	115	SV *sv; // result scalar
35	UV flags;	116	JSON json;
36	int max_recurse;	117	U32 indent; // indentation level
37	int indent;	118	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	119	} enc_t;
39		120
40	// structure used for decoding JSON	121	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)	122	need (enc_t *enc, STRLEN len)
62	{	123	{
63	if (enc->cur + len >= enc->end)	124	if (expect_false (enc->cur + len >= enc->end))
64	{	125	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	126	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	127	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	128	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvEND (enc->sv);	129	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	130	}
70	}	131	}
71		132
72	static void	133	inline void
73	encode_ch (enc_t *enc, char ch)	134	encode_ch (enc_t *enc, char ch)
74	{	135	{
75	need (enc, 1);	136	need (enc, 1);
76	*enc->cur++ = ch;	137	*enc->cur++ = ch;
77	}	138	}
…		…
79	static void	140	static void
80	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	141	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81	{	142	{
82	char *end = str + len;	143	char *end = str + len;
83		144
		145	need (enc, len);
		146
84	while (str < end)	147	while (str < end)
85	{	148	{
86	unsigned char ch = *(unsigned char *)str;	149	unsigned char ch = *(unsigned char *)str;
		150
87	if (ch >= 0x20 && ch < 0x80) // most common case	151	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	152	{
		153	if (expect_false (ch == '"')) // but with slow exceptions
		154	{
		155	need (enc, len += 1);
		156	*enc->cur++ = '\\';
		157	*enc->cur++ = '"';
		158	}
		159	else if (expect_false (ch == '\\'))
		160	{
		161	need (enc, len += 1);
		162	*enc->cur++ = '\\';
		163	*enc->cur++ = '\\';
		164	}
		165	else
89	*enc->cur++ = ch;	166	*enc->cur++ = ch;
		167
90	str++;	168	++str;
91	}	169	}
92	else	170	else
93	{	171	{
94	STRLEN clen;	172	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	}	173	{
103	else	174	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
104	{	175	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
105	uch = ch;	176	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
106	clen = 1;	177	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
107	}	178	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
108		179
109	need (enc, len += 6);	180	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	181	{
		182	STRLEN clen;
		183	UV uch;
		184
		185	if (is_utf8)
		186	{
		187	uch = decode_utf8 (str, end - str, &clen);
		188	if (clen == (STRLEN)-1)
		189	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		190	}
		191	else
		192	{
		193	uch = ch;
115	len += 6;	194	clen = 1;
		195	}
		196
		197	if (uch > 0x10FFFFUL)
		198	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		199
		200	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		201	{
		202	if (uch > 0xFFFFUL)
		203	{
116	need (enc, len += 6);	204	need (enc, len += 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	205	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	206	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	207	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	enc->cur += 12;	208	enc->cur += 12;
		209	}
		210	else
		211	{
		212	static char hexdigit [16] = "0123456789abcdef";
		213	need (enc, len += 5);
		214	*enc->cur++ = '\\';
		215	*enc->cur++ = 'u';
		216	*enc->cur++ = hexdigit [ uch >> 12 ];
		217	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		218	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		219	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		220	}
		221
		222	str += clen;
		223	}
		224	else if (enc->json.flags & F_LATIN1)
		225	{
		226	*enc->cur++ = uch;
		227	str += clen;
		228	}
		229	else if (is_utf8)
		230	{
		231	need (enc, len += clen);
		232	do
		233	{
		234	*enc->cur++ = *str++;
		235	}
		236	while (--clen);
		237	}
		238	else
		239	{
		240	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		241	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		242	++str;
		243	}
121	}	244	}
122	else
123	{
124	sprintf (enc->cur, "\\u%04x", uch);
125	enc->cur += 6;
126	}
127	}
128	else if (is_utf8)
129	{	245	}
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	}	246	}
138		247
139	--len;	248	--len;
140	}	249	}
141	}	250	}
142		251
143	#define INDENT SB \	252	inline void
		253	encode_indent (enc_t *enc)
		254	{
144	if (enc->flags & F_INDENT) \	255	if (enc->json.flags & F_INDENT)
145	{ \	256	{
146	int i_; \	257	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \	258
148	for (i_ = enc->indent * 3; i_--; )\	259	need (enc, spaces);
		260	memset (enc->cur, ' ', spaces);
		261	enc->cur += spaces;
		262	}
		263	}
		264
		265	inline void
		266	encode_space (enc_t *enc)
		267	{
		268	need (enc, 1);
		269	encode_ch (enc, ' ');
		270	}
		271
		272	inline void
		273	encode_nl (enc_t *enc)
		274	{
		275	if (enc->json.flags & F_INDENT)
		276	{
		277	need (enc, 1);
149	encode_ch (enc, ' '); \	278	encode_ch (enc, '\n');
150	} \	279	}
151	SE	280	}
152		281
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	282	inline void
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	283	encode_comma (enc_t *enc)
155	#define COMMA SB \	284	{
156	encode_ch (enc, ','); \	285	encode_ch (enc, ',');
		286
157	if (enc->flags & F_INDENT) \	287	if (enc->json.flags & F_INDENT)
158	NL; \	288	encode_nl (enc);
159	else if (enc->flags & F_SPACE_AFTER) \	289	else if (enc->json.flags & F_SPACE_AFTER)
160	SPACE; \	290	encode_space (enc);
161	SE	291	}
162		292
163	static void encode_sv (enc_t *enc, SV *sv);	293	static void encode_sv (enc_t *enc, SV *sv);
164		294
165	static void	295	static void
166	encode_av (enc_t *enc, AV *av)	296	encode_av (enc_t *enc, AV *av)
167	{	297	{
168	int i, len = av_len (av);	298	int i, len = av_len (av);
169		299
170	encode_ch (enc, '['); NL;	300	if (enc->indent >= enc->maxdepth)
		301	croak ("data structure too deep (hit recursion limit)");
		302
		303	encode_ch (enc, '['); encode_nl (enc);
171	++enc->indent;	304	++enc->indent;
172		305
173	for (i = 0; i <= len; ++i)	306	for (i = 0; i <= len; ++i)
174	{	307	{
175	INDENT;	308	encode_indent (enc);
176	encode_sv (enc, *av_fetch (av, i, 0));	309	encode_sv (enc, *av_fetch (av, i, 0));
177		310
178	if (i < len)	311	if (i < len)
179	COMMA;	312	encode_comma (enc);
180	}	313	}
181		314
182	NL;	315	encode_nl (enc);
183		316
184	--enc->indent;	317	--enc->indent;
185	INDENT; encode_ch (enc, ']');	318	encode_indent (enc); encode_ch (enc, ']');
186	}	319	}
187		320
188	static void	321	static void
189	encode_he (enc_t *enc, HE *he)	322	encode_he (enc_t *enc, HE *he)
190	{	323	{
…		…
192		325
193	if (HeKLEN (he) == HEf_SVKEY)	326	if (HeKLEN (he) == HEf_SVKEY)
194	{	327	{
195	SV *sv = HeSVKEY (he);	328	SV *sv = HeSVKEY (he);
196	STRLEN len;	329	STRLEN len;
		330	char *str;
		331
		332	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	333	str = SvPV (sv, len);
198		334
199	encode_str (enc, str, len, SvUTF8 (sv));	335	encode_str (enc, str, len, SvUTF8 (sv));
200	}	336	}
201	else	337	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	338	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		339
204	encode_ch (enc, '"');	340	encode_ch (enc, '"');
205		341
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	342	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	343	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	344	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));	345	encode_sv (enc, HeVAL (he));
210	}	346	}
211		347
212	// compare hash entries, used when all keys are bytestrings	348	// compare hash entries, used when all keys are bytestrings
213	static int	349	static int
…		…
219	HE *b = *(HE **)b_;	355	HE *b = *(HE **)b_;
220		356
221	STRLEN la = HeKLEN (a);	357	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	358	STRLEN lb = HeKLEN (b);
223		359
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	360	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	361	cmp = la - lb;
226		362
227	return cmp;	363	return cmp;
228	}	364	}
229		365
230	// compare hash entries, used when some keys are sv's or utf-x	366	// compare hash entries, used when some keys are sv's or utf-x
…		…
237	static void	373	static void
238	encode_hv (enc_t *enc, HV *hv)	374	encode_hv (enc_t *enc, HV *hv)
239	{	375	{
240	int count, i;	376	int count, i;
241		377
		378	if (enc->indent >= enc->maxdepth)
		379	croak ("data structure too deep (hit recursion limit)");
		380
242	encode_ch (enc, '{'); NL; ++enc->indent;	381	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
243		382
244	if ((count = hv_iterinit (hv)))	383	if ((count = hv_iterinit (hv)))
245	{	384	{
246	// for canonical output we have to sort by keys first	385	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	386	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	387	// security workaround added somewhere in 5.8.x.
249	// that randomises hash orderings	388	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	389	if (enc->json.flags & F_CANONICAL)
251	{	390	{
252	HE *he, *hes [count];
253	int fast = 1;	391	int fast = 1;
		392	HE *he;
		393	#if defined(__BORLANDC__) || defined(_MSC_VER)
		394	HE **hes = _alloca (count * sizeof (HE));
		395	#else
		396	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		397	#endif
254		398
255	i = 0;	399	i = 0;
256	while ((he = hv_iternext (hv)))	400	while ((he = hv_iternext (hv)))
257	{	401	{
258	hes [i++] = he;	402	hes [i++] = he;
…		…
264		408
265	if (fast)	409	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	410	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	411	else
268	{	412	{
269	// hack to disable "use bytes"	413	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	414	COP cop = *PL_curcop;
271	cop.op_private = 0;	415	cop.op_private = 0;
		416
		417	ENTER;
		418	SAVETMPS;
		419
		420	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	421	PL_curcop = &cop;
273		422
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	423	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		424
276	FREETMPS;	425	FREETMPS;
277		426	LEAVE;
278	PL_curcop = oldcop;
279	}	427	}
280		428
281	for (i = 0; i < count; ++i)	429	for (i = 0; i < count; ++i)
282	{	430	{
283	INDENT;	431	encode_indent (enc);
284	encode_he (enc, hes [i]);	432	encode_he (enc, hes [i]);
285		433
286	if (i < count - 1)	434	if (i < count - 1)
287	COMMA;	435	encode_comma (enc);
288	}
289
290	NL;	436	}
		437
		438	encode_nl (enc);
291	}	439	}
292	else	440	else
293	{	441	{
294	SV *sv;
295	HE *he = hv_iternext (hv);	442	HE *he = hv_iternext (hv);
296		443
297	for (;;)	444	for (;;)
298	{	445	{
299	INDENT;	446	encode_indent (enc);
300	encode_he (enc, he);	447	encode_he (enc, he);
301		448
302	if (!(he = hv_iternext (hv)))	449	if (!(he = hv_iternext (hv)))
303	break;	450	break;
304		451
305	COMMA;	452	encode_comma (enc);
306	}
307
308	NL;	453	}
		454
		455	encode_nl (enc);
309	}	456	}
310	}	457	}
311		458
312	--enc->indent; INDENT; encode_ch (enc, '}');	459	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		460	}
		461
		462	// encode objects, arrays and special \0=false and \1=true values.
		463	static void
		464	encode_rv (enc_t *enc, SV *sv)
		465	{
		466	svtype svt;
		467
		468	SvGETMAGIC (sv);
		469	svt = SvTYPE (sv);
		470
		471	if (expect_false (SvOBJECT (sv)))
		472	{
		473	if (SvSTASH (sv) == json_boolean_stash)
		474	{
		475	if (SvIV (sv) == 0)
		476	encode_str (enc, "false", 5, 0);
		477	else
		478	encode_str (enc, "true", 4, 0);
		479	}
		480	else
		481	{
		482	#if 0
		483	if (0 && sv_derived_from (rv, "JSON::Literal"))
		484	{
		485	// not yet
		486	}
		487	#endif
		488	if (enc->json.flags & F_CONV_BLESSED)
		489	{
		490	// we re-bless the reference to get overload and other niceties right
		491	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		492
		493	if (to_json)
		494	{
		495	dSP;
		496	ENTER;
		497	SAVETMPS;
		498	PUSHMARK (SP);
		499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		500
		501	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		502	// check anyways.
		503	PUTBACK;
		504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		505	SPAGAIN;
		506
		507	encode_sv (enc, POPs);
		508
		509	FREETMPS;
		510	LEAVE;
		511	}
		512	else if (enc->json.flags & F_ALLOW_BLESSED)
		513	encode_str (enc, "null", 4, 0);
		514	else
		515	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		516	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		517	}
		518	else if (enc->json.flags & F_ALLOW_BLESSED)
		519	encode_str (enc, "null", 4, 0);
		520	else
		521	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		522	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		523	}
		524	}
		525	else if (svt == SVt_PVHV)
		526	encode_hv (enc, (HV *)sv);
		527	else if (svt == SVt_PVAV)
		528	encode_av (enc, (AV *)sv);
		529	else if (svt < SVt_PVAV)
		530	{
		531	if (SvNIOK (sv) && SvIV (sv) == 0)
		532	encode_str (enc, "false", 5, 0);
		533	else if (SvNIOK (sv) && SvIV (sv) == 1)
		534	encode_str (enc, "true", 4, 0);
		535	else
		536	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		538	}
		539	else
		540	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		541	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313	}	542	}
314		543
315	static void	544	static void
316	encode_sv (enc_t *enc, SV *sv)	545	encode_sv (enc_t *enc, SV *sv)
317	{	546	{
		547	SvGETMAGIC (sv);
		548
318	if (SvPOKp (sv))	549	if (SvPOKp (sv))
319	{	550	{
320	STRLEN len;	551	STRLEN len;
321	char *str = SvPV (sv, len);	552	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	553	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	554	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	555	encode_ch (enc, '"');
325	}	556	}
326	else if (SvNOKp (sv))	557	else if (SvNOKp (sv))
327	{	558	{
		559	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	560	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	561	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	562	enc->cur += strlen (enc->cur);
331	}	563	}
332	else if (SvIOKp (sv))	564	else if (SvIOKp (sv))
333	{	565	{
334	need (enc, 64);	566	// we assume we can always read an IV as a UV
		567	if (SvUV (sv) & ~(UV)0x7fff)
		568	{
		569	// large integer, use the (rather slow) snprintf way.
		570	need (enc, sizeof (UV) * 3);
335	enc->cur +=	571	enc->cur +=
336	SvIsUV(sv)	572	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	573	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	574	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		575	}
		576	else
		577	{
		578	// optimise the "small number case"
		579	// code will likely be branchless and use only a single multiplication
		580	I32 i = SvIV (sv);
		581	U32 u;
		582	char digit, nz = 0;
		583
		584	need (enc, 6);
		585
		586	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		587	u = i < 0 ? -i : i;
		588
		589	// convert to 4.28 fixed-point representation
		590	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		591
		592	// now output digit by digit, each time masking out the integer part
		593	// and multiplying by 5 while moving the decimal point one to the right,
		594	// resulting in a net multiplication by 10.
		595	// we always write the digit to memory but conditionally increment
		596	// the pointer, to ease the usage of conditional move instructions.
		597	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		598	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		599	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		600	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		601	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		602	}
339	}	603	}
340	else if (SvROK (sv))	604	else if (SvROK (sv))
341	{	605	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))	606	else if (!SvOK (sv))
357	encode_str (enc, "null", 4, 0);	607	encode_str (enc, "null", 4, 0);
358	else	608	else
359	croak ("encountered perl type that JSON cannot handle");	609	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		610	SvPV_nolen (sv), SvFLAGS (sv));
360	}	611	}
361		612
362	static SV *	613	static SV *
363	encode_json (SV *scalar, UV flags)	614	encode_json (SV *scalar, JSON *json)
364	{	615	{
365	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
366	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
367
368	enc_t enc;	616	enc_t enc;
369	enc.flags = flags;	617
		618	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		619	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		620
		621	enc.json = *json;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	622	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	623	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	624	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	625	enc.indent = 0;
		626	enc.maxdepth = DEC_DEPTH (enc.json.flags);
375		627
376	SvPOK_only (enc.sv);	628	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	629	encode_sv (&enc, scalar);
378		630
		631	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		632	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		633
379	if (!(flags & (F_ASCII | F_UTF8)))	634	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	635	SvUTF8_on (enc.sv);
381		636
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	637	if (enc.json.flags & F_SHRINK)
		638	shrink (enc.sv);
		639
383	return enc.sv;	640	return enc.sv;
384	}	641	}
385		642
386	/////////////////////////////////////////////////////////////////////////////	643	/////////////////////////////////////////////////////////////////////////////
		644	// decoder
387		645
388	#define WS \	646	// structure used for decoding JSON
		647	typedef struct
		648	{
		649	char *cur; // current parser pointer
		650	char *end; // end of input string
		651	const char *err; // parse error, if != 0
		652	JSON json;
		653	U32 depth; // recursion depth
		654	U32 maxdepth; // recursion depth limit
		655	} dec_t;
		656
		657	inline void
		658	decode_ws (dec_t *dec)
		659	{
389	for (;;) \	660	for (;;)
390	{ \	661	{
391	char ch = *dec->cur; \	662	char ch = *dec->cur;
		663
392	if (ch > 0x20 \	664	if (ch > 0x20
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	665	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394	break; \	666	break;
		667
395	++dec->cur; \	668	++dec->cur;
396	}	669	}
		670	}
397		671
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	672	#define ERR(reason) SB dec->err = reason; goto fail; SE
		673
399	#define EXPECT_CH(ch) SB \	674	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	675	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	676	ERR (# ch " expected"); \
402	++dec->cur; \	677	++dec->cur; \
403	SE	678	SE
404		679
		680	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		681	#define DEC_DEC_DEPTH --dec->depth
		682
405	static SV *decode_sv (dec_t *dec);	683	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		684
412	static signed char decode_hexdigit[256];	685	static signed char decode_hexdigit[256];
413		686
414	static UV	687	static UV
415	decode_4hex (dec_t *dec)	688	decode_4hex (dec_t *dec)
416	{	689	{
417	signed char d1, d2, d3, d4;	690	signed char d1, d2, d3, d4;
		691	unsigned char *cur = (unsigned char *)dec->cur;
418		692
419	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	693	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");	694	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]];	695	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");	696	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		697
428	dec->cur += 4;	698	dec->cur += 4;
429		699
430	return ((UV)d1) << 12	700	return ((UV)d1) << 12
431	| ((UV)d2) << 8	701	| ((UV)d2) << 8
…		…
437	}	707	}
438		708
439	static SV *	709	static SV *
440	decode_str (dec_t *dec)	710	decode_str (dec_t *dec)
441	{	711	{
442	SV *sv = NEWSV (0,2);	712	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	713	int utf8 = 0;
		714	char *dec_cur = dec->cur;
445		715
446	for (;;)	716	do
447	{	717	{
448	unsigned char ch = *(unsigned char *)dec->cur;	718	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		719	char *cur = buf;
449		720
450	if (ch == '"')	721	do
451	break;
452	else if (ch == '\\')
453	{	722	{
454	switch (*++dec->cur)	723	unsigned char ch = *(unsigned char *)dec_cur++;
		724
		725	if (expect_false (ch == '"'))
		726	{
		727	--dec_cur;
		728	break;
455	{	729	}
456	case '\\':	730	else if (expect_false (ch == '\\'))
457	case '/':	731	{
458	case '"': APPEND_CH (*dec->cur++); break;	732	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	{	733	{
468	UV lo, hi;	734	case '\\':
469	++dec->cur;	735	case '/':
		736	case '"': *cur++ = *dec_cur++; break;
470		737
471	hi = decode_4hex (dec);	738	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	739	case 't': ++dec_cur; *cur++ = '\011'; break;
473	goto fail;	740	case 'n': ++dec_cur; *cur++ = '\012'; break;
		741	case 'f': ++dec_cur; *cur++ = '\014'; break;
		742	case 'r': ++dec_cur; *cur++ = '\015'; break;
474		743
475	// possibly a surrogate pair	744	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	745	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	746	UV lo, hi;
479	ERR ("illegal surrogate character");	747	++dec_cur;
480		748
481	dec->cur += 2;	749	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	750	hi = decode_4hex (dec);
		751	dec_cur = dec->cur;
484	if (lo == (UV)-1)	752	if (hi == (UV)-1)
485	goto fail;	753	goto fail;
486		754
		755	// possibly a surrogate pair
		756	if (hi >= 0xd800)
		757	if (hi < 0xdc00)
		758	{
		759	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		760	ERR ("missing low surrogate character in surrogate pair");
		761
		762	dec_cur += 2;
		763
		764	dec->cur = dec_cur;
		765	lo = decode_4hex (dec);
		766	dec_cur = dec->cur;
		767	if (lo == (UV)-1)
		768	goto fail;
		769
487	if (lo < 0xdc00 || lo >= 0xe000)	770	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	771	ERR ("surrogate pair expected");
489		772
490	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	773	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		774	}
		775	else if (hi < 0xe000)
		776	ERR ("missing high surrogate character in surrogate pair");
		777
		778	if (hi >= 0x80)
		779	{
		780	utf8 = 1;
		781
		782	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		783	}
		784	else
		785	*cur++ = hi;
491	}	786	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	787	break;
497	utf8 = 1;
498		788
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	789	default:
503	APPEND_CH (hi);	790	--dec_cur;
		791	ERR ("illegal backslash escape sequence in string");
504	}	792	}
		793	}
		794	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		795	*cur++ = ch;
		796	else if (ch >= 0x80)
		797	{
		798	STRLEN clen;
		799	UV uch;
		800
		801	--dec_cur;
		802
		803	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		804	if (clen == (STRLEN)-1)
		805	ERR ("malformed UTF-8 character in JSON string");
		806
		807	do
		808	*cur++ = *dec_cur++;
		809	while (--clen);
		810
		811	utf8 = 1;
		812	}
		813	else
		814	{
		815	--dec_cur;
		816
		817	if (!ch)
		818	ERR ("unexpected end of string while parsing JSON string");
505	break;	819	else
		820	ERR ("invalid character encountered while parsing JSON string");
506	}	821	}
507	}	822	}
508	else if (ch >= 0x20 && ch <= 0x7f)	823	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	824
510	else if (ch >= 0x80)	825	{
		826	STRLEN len = cur - buf;
		827
		828	if (sv)
511	{	829	{
512	STRLEN clen;	830	SvGROW (sv, SvCUR (sv) + len + 1);
513	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	831	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514	if (clen < 0)	832	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	}	833	}
521	else	834	else
522	ERR ("invalid character encountered");	835	sv = newSVpvn (buf, len);
523	}	836	}
		837	}
		838	while (*dec_cur != '"');
524		839
525	++dec->cur;	840	++dec_cur;
526		841
		842	if (sv)
		843	{
527	SvPOK_only (sv);	844	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	845	*SvEND (sv) = 0;
531		846
532	if (utf8)	847	if (utf8)
533	SvUTF8_on (sv);	848	SvUTF8_on (sv);
		849	}
		850	else
		851	sv = newSVpvn ("", 0);
534		852
		853	dec->cur = dec_cur;
535	return sv;	854	return sv;
536		855
537	fail:	856	fail:
538	SvREFCNT_dec (sv);	857	dec->cur = dec_cur;
539	return 0;	858	return 0;
540	}	859	}
541		860
542	static SV *	861	static SV *
543	decode_num (dec_t *dec)	862	decode_num (dec_t *dec)
…		…
553	{	872	{
554	++dec->cur;	873	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	874	if (*dec->cur >= '0' && *dec->cur <= '9')
556	ERR ("malformed number (leading zero must not be followed by another digit)");	875	ERR ("malformed number (leading zero must not be followed by another digit)");
557	}	876	}
558		877	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	878	ERR ("malformed number (no digits after initial minus)");
		879	else
		880	do
		881	{
		882	++dec->cur;
		883	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	884	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		885
563	// [frac]	886	// [frac]
564	if (*dec->cur == '.')	887	if (*dec->cur == '.')
565	{	888	{
566	is_nv = 1;	889	++dec->cur;
		890
		891	if (*dec->cur < '0' || *dec->cur > '9')
		892	ERR ("malformed number (no digits after decimal point)");
567		893
568	do	894	do
569	{	895	{
570	++dec->cur;	896	++dec->cur;
571	}	897	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	898	while (*dec->cur >= '0' && *dec->cur <= '9');
		899
		900	is_nv = 1;
573	}	901	}
574		902
575	// [exp]	903	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	904	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	905	{
578	is_nv = 1;
579
580	++dec->cur;	906	++dec->cur;
		907
581	if (*dec->cur == '-' || *dec->cur == '+')	908	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	909	++dec->cur;
583		910
		911	if (*dec->cur < '0' || *dec->cur > '9')
		912	ERR ("malformed number (no digits after exp sign)");
		913
		914	do
		915	{
		916	++dec->cur;
		917	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	918	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	919
		920	is_nv = 1;
586	}	921	}
587		922
588	if (!is_nv)	923	if (!is_nv)
589	{	924	{
590	UV uv;	925	// 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);	926	if (*start == '-')
592	if (numtype & IS_NUMBER_IN_UV)	927	switch (dec->cur - start)
593	if (numtype & IS_NUMBER_NEG)
594	{	928	{
595	if (uv < (UV)IV_MIN)	929	case 2: return newSViv (-( start [1] - '0' * 1));
596	return newSViv (-(IV)uv);	930	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		931	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		932	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
597	}	933	}
		934	else
		935	switch (dec->cur - start)
		936	{
		937	case 1: return newSViv ( start [0] - '0' * 1);
		938	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		939	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		940	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		941	}
		942
		943	{
		944	UV uv;
		945	int numtype = grok_number (start, dec->cur - start, &uv);
		946	if (numtype & IS_NUMBER_IN_UV)
		947	if (numtype & IS_NUMBER_NEG)
		948	{
		949	if (uv < (UV)IV_MIN)
		950	return newSViv (-(IV)uv);
		951	}
598	else	952	else
599	return newSVuv (uv);	953	return newSVuv (uv);
		954
		955	// here would likely be the place for bigint support
600	}	956	}
		957	}
601		958
		959	// if we ever support bigint or bigfloat, this is the place for bigfloat
602	return newSVnv (Atof (start));	960	return newSVnv (Atof (start));
603		961
604	fail:	962	fail:
605	return 0;	963	return 0;
606	}	964	}
…		…
608	static SV *	966	static SV *
609	decode_av (dec_t *dec)	967	decode_av (dec_t *dec)
610	{	968	{
611	AV *av = newAV ();	969	AV *av = newAV ();
612		970
		971	DEC_INC_DEPTH;
		972	decode_ws (dec);
		973
		974	if (*dec->cur == ']')
		975	++dec->cur;
		976	else
613	for (;;)	977	for (;;)
614	{	978	{
615	SV *value;	979	SV *value;
616		980
617	value = decode_sv (dec);	981	value = decode_sv (dec);
618	if (!value)	982	if (!value)
619	goto fail;	983	goto fail;
620		984
621	av_push (av, value);	985	av_push (av, value);
622		986
623	WS;	987	decode_ws (dec);
624		988
625	if (*dec->cur == ']')	989	if (*dec->cur == ']')
626	{	990	{
627	++dec->cur;	991	++dec->cur;
628	break;	992	break;
		993	}
629	}	994
630
631	if (*dec->cur != ',')	995	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	996	ERR (", or ] expected while parsing array");
633		997
634	++dec->cur;	998	++dec->cur;
635	}	999	}
636		1000
		1001	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1002	return newRV_noinc ((SV *)av);
638		1003
639	fail:	1004	fail:
640	SvREFCNT_dec (av);	1005	SvREFCNT_dec (av);
		1006	DEC_DEC_DEPTH;
641	return 0;	1007	return 0;
642	}	1008	}
643		1009
644	static SV *	1010	static SV *
645	decode_hv (dec_t *dec)	1011	decode_hv (dec_t *dec)
646	{	1012	{
647	HV *hv = newHV ();	1013	HV *hv = newHV ();
648		1014
		1015	DEC_INC_DEPTH;
		1016	decode_ws (dec);
		1017
		1018	if (*dec->cur == '}')
		1019	++dec->cur;
		1020	else
649	for (;;)	1021	for (;;)
650	{	1022	{
651	SV *key, *value;	1023	decode_ws (dec); EXPECT_CH ('"');
652		1024
653	WS; EXPECT_CH ('"');	1025	// heuristic: assume that
654		1026	// a) decode_str + hv_store_ent are abysmally slow.
655	key = decode_str (dec);	1027	// b) most hash keys are short, simple ascii text.
656	if (!key)	1028	// => try to "fast-match" such strings to avoid
657	goto fail;	1029	// the overhead of decode_str + hv_store_ent.
658
659	WS; EXPECT_CH (':');
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1030	{
		1031	SV *value;
		1032	char *p = dec->cur;
		1033	char *e = p + 24; // only try up to 24 bytes
		1034
		1035	for (;;)
		1036	{
		1037	// the >= 0x80 is true on most architectures
		1038	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1039	{
		1040	// slow path, back up and use decode_str
		1041	SV *key = decode_str (dec);
		1042	if (!key)
		1043	goto fail;
		1044
		1045	decode_ws (dec); EXPECT_CH (':');
		1046
		1047	value = decode_sv (dec);
		1048	if (!value)
		1049	{
		1050	SvREFCNT_dec (key);
		1051	goto fail;
		1052	}
		1053
		1054	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1055	SvREFCNT_dec (key);
		1056
		1057	break;
		1058	}
		1059	else if (*p == '"')
		1060	{
		1061	// fast path, got a simple key
		1062	char *key = dec->cur;
		1063	int len = p - key;
		1064	dec->cur = p + 1;
		1065
		1066	decode_ws (dec); EXPECT_CH (':');
		1067
		1068	value = decode_sv (dec);
		1069	if (!value)
665	goto fail;	1070	goto fail;
		1071
		1072	hv_store (hv, key, len, value, 0);
		1073
		1074	break;
		1075	}
		1076
		1077	++p;
		1078	}
666	}	1079	}
667		1080
668	//TODO: optimise	1081	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1082
671	WS;
672
673	if (*dec->cur == '}')	1083	if (*dec->cur == '}')
674	{	1084	{
675	++dec->cur;	1085	++dec->cur;
676	break;	1086	break;
677	}	1087	}
678		1088
679	if (*dec->cur != ',')	1089	if (*dec->cur != ',')
680	ERR (", or } expected while parsing object/hash");	1090	ERR (", or } expected while parsing object/hash");
681		1091
682	++dec->cur;	1092	++dec->cur;
683	}	1093	}
684		1094
		1095	DEC_DEC_DEPTH;
685	return newRV_noinc ((SV *)hv);	1096	return newRV_noinc ((SV *)hv);
686		1097
687	fail:	1098	fail:
688	SvREFCNT_dec (hv);	1099	SvREFCNT_dec (hv);
		1100	DEC_DEC_DEPTH;
689	return 0;	1101	return 0;
690	}	1102	}
691		1103
692	static SV *	1104	static SV *
693	decode_sv (dec_t *dec)	1105	decode_sv (dec_t *dec)
694	{	1106	{
695	WS;	1107	decode_ws (dec);
		1108
		1109	// the beauty of JSON: you need exactly one character lookahead
		1110	// to parse anything.
696	switch (*dec->cur)	1111	switch (*dec->cur)
697	{	1112	{
698	case '"': ++dec->cur; return decode_str (dec);	1113	case '"': ++dec->cur; return decode_str (dec);
699	case '[': ++dec->cur; return decode_av (dec);	1114	case '[': ++dec->cur; return decode_av (dec);
700	case '{': ++dec->cur; return decode_hv (dec);	1115	case '{': ++dec->cur; return decode_hv (dec);
…		…
706		1121
707	case 't':	1122	case 't':
708	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1123	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709	{	1124	{
710	dec->cur += 4;	1125	dec->cur += 4;
711	return newSViv (1);	1126	return SvREFCNT_inc (json_true);
712	}	1127	}
713	else	1128	else
714	ERR ("'true' expected");	1129	ERR ("'true' expected");
715		1130
716	break;	1131	break;
717		1132
718	case 'f':	1133	case 'f':
719	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1134	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720	{	1135	{
721	dec->cur += 5;	1136	dec->cur += 5;
722	return newSViv (0);	1137	return SvREFCNT_inc (json_false);
723	}	1138	}
724	else	1139	else
725	ERR ("'false' expected");	1140	ERR ("'false' expected");
726		1141
727	break;	1142	break;
728		1143
729	case 'n':	1144	case 'n':
730	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1145	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731	{	1146	{
732	dec->cur += 4;	1147	dec->cur += 4;
733	return newSViv (1);	1148	return newSVsv (&PL_sv_undef);
734	}	1149	}
735	else	1150	else
736	ERR ("'null' expected");	1151	ERR ("'null' expected");
737		1152
738	break;	1153	break;
739		1154
740	default:	1155	default:
741	ERR ("malformed json string");	1156	ERR ("malformed JSON string, neither array, object, number, string or atom");
742	break;	1157	break;
743	}	1158	}
744		1159
745	fail:	1160	fail:
746	return 0;	1161	return 0;
747	}	1162	}
748		1163
749	static SV *	1164	static SV *
750	decode_json (SV *string, UV flags)	1165	decode_json (SV *string, JSON *json, UV *offset_return)
751	{	1166	{
		1167	dec_t dec;
		1168	UV offset;
752	SV *sv;	1169	SV *sv;
753		1170
		1171	SvGETMAGIC (string);
		1172	SvUPGRADE (string, SVt_PV);
		1173
		1174	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1175	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1176	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1177
754	if (!(flags & F_UTF8))	1178	if (json->flags & F_UTF8)
		1179	sv_utf8_downgrade (string, 0);
		1180	else
755	sv_utf8_upgrade (string);	1181	sv_utf8_upgrade (string);
756		1182
757	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1183	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758		1184
759	dec_t dec;	1185	dec.json = *json;
760	dec.flags = flags;
761	dec.cur = SvPVX (string);	1186	dec.cur = SvPVX (string);
762	dec.end = SvEND (string);	1187	dec.end = SvEND (string);
763	dec.err = 0;	1188	dec.err = 0;
		1189	dec.depth = 0;
		1190	dec.maxdepth = DEC_DEPTH (dec.json.flags);
764		1191
765	*dec.end = 1; // invalid anywhere	1192	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
766	sv = decode_sv (&dec);	1193	sv = decode_sv (&dec);
767	*dec.end = 0;	1194
		1195	if (!(offset_return || !sv))
		1196	{
		1197	// check for trailing garbage
		1198	decode_ws (&dec);
		1199
		1200	if (*dec.cur)
		1201	{
		1202	dec.err = "garbage after JSON object";
		1203	SvREFCNT_dec (sv);
		1204	sv = 0;
		1205	}
		1206	}
		1207
		1208	if (offset_return || !sv)
		1209	{
		1210	offset = dec.json.flags & F_UTF8
		1211	? dec.cur - SvPVX (string)
		1212	: utf8_distance (dec.cur, SvPVX (string));
		1213
		1214	if (offset_return)
		1215	*offset_return = offset;
		1216	}
768		1217
769	if (!sv)	1218	if (!sv)
770	{	1219	{
771	IV offset = utf8_distance (dec.cur, SvPVX (string));
772	SV *uni = sv_newmortal ();	1220	SV *uni = sv_newmortal ();
773		1221
		1222	// horrible hack to silence warning inside pv_uni_display
		1223	COP cop = *PL_curcop;
		1224	cop.cop_warnings = pWARN_NONE;
		1225	ENTER;
		1226	SAVEVPTR (PL_curcop);
		1227	PL_curcop = &cop;
774	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1228	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1229	LEAVE;
		1230
775	croak ("%s, at character %d (%s)",	1231	croak ("%s, at character offset %d [\"%s\"]",
776	dec.err,	1232	dec.err,
777	(int)offset,	1233	(int)offset,
778	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1234	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779	}	1235	}
780		1236
781	sv = sv_2mortal (sv);	1237	sv = sv_2mortal (sv);
782		1238
783	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1239	if (!(dec.json.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)");	1240	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785		1241
786	return sv;	1242	return sv;
787	}	1243	}
788		1244
		1245	/////////////////////////////////////////////////////////////////////////////
		1246	// XS interface functions
		1247
789	MODULE = JSON::XS PACKAGE = JSON::XS	1248	MODULE = JSON::XS PACKAGE = JSON::XS
790		1249
791	BOOT:	1250	BOOT:
792	{	1251	{
793	int i;	1252	int i;
794		1253
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	1254	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	1255	decode_hexdigit [i] =
		1256	i >= '0' && i <= '9' ? i - '0'
		1257	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1258	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1259	: -1;
798		1260
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);	1261	json_stash = gv_stashpv ("JSON::XS" , 1);
806	}	1262	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
807		1263
808	SV *new (char *dummy)	1264	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1265	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1266	}
		1267
		1268	PROTOTYPES: DISABLE
		1269
		1270	void new (char *klass)
809	CODE:	1271	PPCODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1272	{
811	OUTPUT:	1273	SV *pv = NEWSV (0, sizeof (JSON));
812	RETVAL	1274	SvPOK_only (pv);
		1275	Zero (SvPVX (pv), 1, sizeof (JSON));
		1276	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1277	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1278	}
813		1279
814	SV *ascii (SV *self, int enable)	1280	void ascii (JSON *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	PPCODE:
826	{	1295	{
827	UV *uv = SvJSON (self);
828	if (enable)	1296	if (enable)
829	*uv |= ix;	1297	self->flags |= ix;
830	else	1298	else
831	*uv &= ~ix;	1299	self->flags &= ~ix;
832		1300
833	RETVAL = newSVsv (self);	1301	XPUSHs (ST (0));
834	}	1302	}
835	OUTPUT:
836	RETVAL
837		1303
838	void encode (SV *self, SV *scalar)	1304	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
839	PPCODE:	1305	PPCODE:
840	XPUSHs (encode_json (scalar, *SvJSON (self)));	1306	{
		1307	UV log2 = 0;
841		1308
842	void decode (SV *self, SV *jsonstr)	1309	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1310
		1311	while ((1UL << log2) < max_depth)
		1312	++log2;
		1313
		1314	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1315
		1316	XPUSHs (ST (0));
		1317	}
		1318
		1319	void max_size (JSON *self, UV max_size = 0)
843	PPCODE:	1320	PPCODE:
		1321	{
		1322	UV log2 = 0;
		1323
		1324	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1325	if (max_size == 1) max_size = 2;
		1326
		1327	while ((1UL << log2) < max_size)
		1328	++log2;
		1329
		1330	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1331
		1332	XPUSHs (ST (0));
		1333	}
		1334
		1335	void encode (JSON *self, SV *scalar)
		1336	PPCODE:
		1337	XPUSHs (encode_json (scalar, self));
		1338
		1339	void decode (JSON *self, SV *jsonstr)
		1340	PPCODE:
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1341	XPUSHs (decode_json (jsonstr, self, 0));
		1342
		1343	void decode_prefix (JSON *self, SV *jsonstr)
		1344	PPCODE:
		1345	{
		1346	UV offset;
		1347	EXTEND (SP, 2);
		1348	PUSHs (decode_json (jsonstr, self, &offset));
		1349	PUSHs (sv_2mortal (newSVuv (offset)));
		1350	}
		1351
		1352	PROTOTYPES: ENABLE
845		1353
846	void to_json (SV *scalar)	1354	void to_json (SV *scalar)
847	PPCODE:	1355	PPCODE:
		1356	{
		1357	JSON json = { F_DEFAULT | F_UTF8 };
848	XPUSHs (encode_json (scalar, F_UTF8));	1358	XPUSHs (encode_json (scalar, &json));
		1359	}
849		1360
850	void from_json (SV *jsonstr)	1361	void from_json (SV *jsonstr)
851	PPCODE:	1362	PPCODE:
		1363	{
		1364	JSON json = { F_DEFAULT | F_UTF8 };
852	XPUSHs (decode_json (jsonstr, F_UTF8));	1365	XPUSHs (decode_json (jsonstr, &json, 0));
		1366	}
853		1367

Diff Legend

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