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 1.53 by root, Tue Jul 10 15:45:34 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
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		36
		37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		39
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	41	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		42
21	#define INIT_SIZE 32 // initial scalar size to be allocated	43	#define INIT_SIZE 32 // initial scalar size to be allocated
		44	#define INDENT_STEP 3 // spaces per indentation level
		45
		46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		47
23	#define SB do {	48	#define SB do {
24	#define SE } while (0)	49	#define SE } while (0)
25		50
26	static HV *json_stash;	51	#if __GNUC__ >= 3
		52	# define expect(expr,value) __builtin_expect ((expr),(value))
		53	# define inline inline
		54	#else
		55	# define expect(expr,value) (expr)
		56	# define inline static
		57	#endif
		58
		59	#define expect_false(expr) expect ((expr) != 0, 0)
		60	#define expect_true(expr) expect ((expr) != 0, 1)
		61
		62	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		63	static SV *json_true, *json_false;
		64
		65	typedef struct {
		66	U32 flags;
		67	SV *cb_object;
		68	HV *cb_sk_object;
		69	} JSON;
		70
		71	/////////////////////////////////////////////////////////////////////////////
		72	// utility functions
		73
		74	inline void
		75	shrink (SV *sv)
		76	{
		77	sv_utf8_downgrade (sv, 1);
		78	if (SvLEN (sv) > SvCUR (sv) + 1)
		79	{
		80	#ifdef SvPV_shrink_to_cur
		81	SvPV_shrink_to_cur (sv);
		82	#elif defined (SvPV_renew)
		83	SvPV_renew (sv, SvCUR (sv) + 1);
		84	#endif
		85	}
		86	}
		87
		88	// decode an utf-8 character and return it, or (UV)-1 in
		89	// case of an error.
		90	// we special-case "safe" characters from U+80 .. U+7FF,
		91	// but use the very good perl function to parse anything else.
		92	// note that we never call this function for a ascii codepoints
		93	inline UV
		94	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		95	{
		96	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		99	{
		100	*clen = 2;
		101	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		102	}
		103	else
		104	{
		105	*clen = (STRLEN)-1;
		106	return (UV)-1;
		107	}
		108	}
		109
		110	/////////////////////////////////////////////////////////////////////////////
		111	// encoder
27		112
28	// structure used for encoding JSON	113	// structure used for encoding JSON
29	typedef struct	114	typedef struct
30	{	115	{
31	char *cur;	116	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	117	char *end; // SvEND (sv)
34	SV *sv;	118	SV *sv; // result scalar
35	UV flags;	119	JSON json;
36	int max_recurse;	120	U32 indent; // indentation level
37	int indent;	121	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	122	} enc_t;
39		123
40	// structure used for decoding JSON	124	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)	125	need (enc_t *enc, STRLEN len)
62	{	126	{
63	if (enc->cur + len >= enc->end)	127	if (expect_false (enc->cur + len >= enc->end))
64	{	128	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	129	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	130	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	131	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvEND (enc->sv);	132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	133	}
70	}	134	}
71		135
72	static void	136	inline void
73	encode_ch (enc_t *enc, char ch)	137	encode_ch (enc_t *enc, char ch)
74	{	138	{
75	need (enc, 1);	139	need (enc, 1);
76	*enc->cur++ = ch;	140	*enc->cur++ = ch;
77	}	141	}
…		…
79	static void	143	static void
80	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	144	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81	{	145	{
82	char *end = str + len;	146	char *end = str + len;
83		147
		148	need (enc, len);
		149
84	while (str < end)	150	while (str < end)
85	{	151	{
86	unsigned char ch = *(unsigned char *)str;	152	unsigned char ch = *(unsigned char *)str;
		153
87	if (ch >= 0x20 && ch < 0x80) // most common case	154	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	155	{
		156	if (expect_false (ch == '"')) // but with slow exceptions
		157	{
		158	need (enc, len += 1);
		159	*enc->cur++ = '\\';
		160	*enc->cur++ = '"';
		161	}
		162	else if (expect_false (ch == '\\'))
		163	{
		164	need (enc, len += 1);
		165	*enc->cur++ = '\\';
		166	*enc->cur++ = '\\';
		167	}
		168	else
89	*enc->cur++ = ch;	169	*enc->cur++ = ch;
		170
90	str++;	171	++str;
91	}	172	}
92	else	173	else
93	{	174	{
94	STRLEN clen;	175	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	}	176	{
103	else	177	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
104	{	178	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
105	uch = ch;	179	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
106	clen = 1;	180	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
107	}	181	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
108		182
109	need (enc, len += 6);	183	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	184	{
		185	STRLEN clen;
		186	UV uch;
		187
		188	if (is_utf8)
		189	{
		190	uch = decode_utf8 (str, end - str, &clen);
		191	if (clen == (STRLEN)-1)
		192	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		193	}
		194	else
		195	{
		196	uch = ch;
115	len += 6;	197	clen = 1;
		198	}
		199
		200	if (uch > 0x10FFFFUL)
		201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		202
		203	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		204	{
		205	if (uch > 0xFFFFUL)
		206	{
116	need (enc, len += 6);	207	need (enc, len += 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	208	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	209	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	210	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	enc->cur += 12;	211	enc->cur += 12;
		212	}
		213	else
		214	{
		215	static char hexdigit [16] = "0123456789abcdef";
		216	need (enc, len += 5);
		217	*enc->cur++ = '\\';
		218	*enc->cur++ = 'u';
		219	*enc->cur++ = hexdigit [ uch >> 12 ];
		220	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		221	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		222	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		223	}
		224
		225	str += clen;
		226	}
		227	else if (enc->json.flags & F_LATIN1)
		228	{
		229	*enc->cur++ = uch;
		230	str += clen;
		231	}
		232	else if (is_utf8)
		233	{
		234	need (enc, len += clen);
		235	do
		236	{
		237	*enc->cur++ = *str++;
		238	}
		239	while (--clen);
		240	}
		241	else
		242	{
		243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		245	++str;
		246	}
121	}	247	}
122	else
123	{
124	sprintf (enc->cur, "\\u%04x", uch);
125	enc->cur += 6;
126	}
127	}
128	else if (is_utf8)
129	{	248	}
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	}	249	}
138		250
139	--len;	251	--len;
140	}	252	}
141	}	253	}
142		254
143	#define INDENT SB \	255	inline void
		256	encode_indent (enc_t *enc)
		257	{
144	if (enc->flags & F_INDENT) \	258	if (enc->json.flags & F_INDENT)
145	{ \	259	{
146	int i_; \	260	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \	261
148	for (i_ = enc->indent * 3; i_--; )\	262	need (enc, spaces);
		263	memset (enc->cur, ' ', spaces);
		264	enc->cur += spaces;
		265	}
		266	}
		267
		268	inline void
		269	encode_space (enc_t *enc)
		270	{
		271	need (enc, 1);
		272	encode_ch (enc, ' ');
		273	}
		274
		275	inline void
		276	encode_nl (enc_t *enc)
		277	{
		278	if (enc->json.flags & F_INDENT)
		279	{
		280	need (enc, 1);
149	encode_ch (enc, ' '); \	281	encode_ch (enc, '\n');
150	} \	282	}
151	SE	283	}
152		284
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	285	inline void
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	286	encode_comma (enc_t *enc)
155	#define COMMA SB \	287	{
156	encode_ch (enc, ','); \	288	encode_ch (enc, ',');
		289
157	if (enc->flags & F_INDENT) \	290	if (enc->json.flags & F_INDENT)
158	NL; \	291	encode_nl (enc);
159	else if (enc->flags & F_SPACE_AFTER) \	292	else if (enc->json.flags & F_SPACE_AFTER)
160	SPACE; \	293	encode_space (enc);
161	SE	294	}
162		295
163	static void encode_sv (enc_t *enc, SV *sv);	296	static void encode_sv (enc_t *enc, SV *sv);
164		297
165	static void	298	static void
166	encode_av (enc_t *enc, AV *av)	299	encode_av (enc_t *enc, AV *av)
167	{	300	{
168	int i, len = av_len (av);	301	int i, len = av_len (av);
169		302
170	encode_ch (enc, '['); NL;	303	if (enc->indent >= enc->maxdepth)
		304	croak ("data structure too deep (hit recursion limit)");
		305
		306	encode_ch (enc, '['); encode_nl (enc);
171	++enc->indent;	307	++enc->indent;
172		308
173	for (i = 0; i <= len; ++i)	309	for (i = 0; i <= len; ++i)
174	{	310	{
175	INDENT;	311	encode_indent (enc);
176	encode_sv (enc, *av_fetch (av, i, 0));	312	encode_sv (enc, *av_fetch (av, i, 0));
177		313
178	if (i < len)	314	if (i < len)
179	COMMA;	315	encode_comma (enc);
180	}	316	}
181		317
182	NL;	318	encode_nl (enc);
183		319
184	--enc->indent;	320	--enc->indent;
185	INDENT; encode_ch (enc, ']');	321	encode_indent (enc); encode_ch (enc, ']');
186	}	322	}
187		323
188	static void	324	static void
189	encode_he (enc_t *enc, HE *he)	325	encode_he (enc_t *enc, HE *he)
190	{	326	{
…		…
192		328
193	if (HeKLEN (he) == HEf_SVKEY)	329	if (HeKLEN (he) == HEf_SVKEY)
194	{	330	{
195	SV *sv = HeSVKEY (he);	331	SV *sv = HeSVKEY (he);
196	STRLEN len;	332	STRLEN len;
		333	char *str;
		334
		335	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	336	str = SvPV (sv, len);
198		337
199	encode_str (enc, str, len, SvUTF8 (sv));	338	encode_str (enc, str, len, SvUTF8 (sv));
200	}	339	}
201	else	340	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	341	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		342
204	encode_ch (enc, '"');	343	encode_ch (enc, '"');
205		344
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	345	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	346	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	347	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));	348	encode_sv (enc, HeVAL (he));
210	}	349	}
211		350
212	// compare hash entries, used when all keys are bytestrings	351	// compare hash entries, used when all keys are bytestrings
213	static int	352	static int
…		…
219	HE *b = *(HE **)b_;	358	HE *b = *(HE **)b_;
220		359
221	STRLEN la = HeKLEN (a);	360	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	361	STRLEN lb = HeKLEN (b);
223		362
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	363	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	364	cmp = la - lb;
226		365
227	return cmp;	366	return cmp;
228	}	367	}
229		368
230	// compare hash entries, used when some keys are sv's or utf-x	369	// compare hash entries, used when some keys are sv's or utf-x
…		…
237	static void	376	static void
238	encode_hv (enc_t *enc, HV *hv)	377	encode_hv (enc_t *enc, HV *hv)
239	{	378	{
240	int count, i;	379	int count, i;
241		380
		381	if (enc->indent >= enc->maxdepth)
		382	croak ("data structure too deep (hit recursion limit)");
		383
242	encode_ch (enc, '{'); NL; ++enc->indent;	384	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
243		385
244	if ((count = hv_iterinit (hv)))	386	if ((count = hv_iterinit (hv)))
245	{	387	{
246	// for canonical output we have to sort by keys first	388	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	389	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	390	// security workaround added somewhere in 5.8.x.
249	// that randomises hash orderings	391	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	392	if (enc->json.flags & F_CANONICAL)
251	{	393	{
252	HE *he, *hes [count];
253	int fast = 1;	394	int fast = 1;
		395	HE *he;
		396	#if defined(__BORLANDC__) || defined(_MSC_VER)
		397	HE **hes = _alloca (count * sizeof (HE));
		398	#else
		399	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		400	#endif
254		401
255	i = 0;	402	i = 0;
256	while ((he = hv_iternext (hv)))	403	while ((he = hv_iternext (hv)))
257	{	404	{
258	hes [i++] = he;	405	hes [i++] = he;
…		…
264		411
265	if (fast)	412	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	413	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	414	else
268	{	415	{
269	// hack to disable "use bytes"	416	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	417	COP cop = *PL_curcop;
271	cop.op_private = 0;	418	cop.op_private = 0;
		419
		420	ENTER;
		421	SAVETMPS;
		422
		423	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	424	PL_curcop = &cop;
273		425
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	426	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		427
276	FREETMPS;	428	FREETMPS;
277		429	LEAVE;
278	PL_curcop = oldcop;
279	}	430	}
280		431
281	for (i = 0; i < count; ++i)	432	for (i = 0; i < count; ++i)
282	{	433	{
283	INDENT;	434	encode_indent (enc);
284	encode_he (enc, hes [i]);	435	encode_he (enc, hes [i]);
285		436
286	if (i < count - 1)	437	if (i < count - 1)
287	COMMA;	438	encode_comma (enc);
288	}
289
290	NL;	439	}
		440
		441	encode_nl (enc);
291	}	442	}
292	else	443	else
293	{	444	{
294	SV *sv;
295	HE *he = hv_iternext (hv);	445	HE *he = hv_iternext (hv);
296		446
297	for (;;)	447	for (;;)
298	{	448	{
299	INDENT;	449	encode_indent (enc);
300	encode_he (enc, he);	450	encode_he (enc, he);
301		451
302	if (!(he = hv_iternext (hv)))	452	if (!(he = hv_iternext (hv)))
303	break;	453	break;
304		454
305	COMMA;	455	encode_comma (enc);
306	}
307
308	NL;	456	}
		457
		458	encode_nl (enc);
309	}	459	}
310	}	460	}
311		461
312	--enc->indent; INDENT; encode_ch (enc, '}');	462	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		463	}
		464
		465	// encode objects, arrays and special \0=false and \1=true values.
		466	static void
		467	encode_rv (enc_t *enc, SV *sv)
		468	{
		469	svtype svt;
		470
		471	SvGETMAGIC (sv);
		472	svt = SvTYPE (sv);
		473
		474	if (expect_false (SvOBJECT (sv)))
		475	{
		476	if (SvSTASH (sv) == json_boolean_stash)
		477	{
		478	if (SvIV (sv))
		479	encode_str (enc, "true", 4, 0);
		480	else
		481	encode_str (enc, "false", 5, 0);
		482	}
		483	else
		484	{
		485	#if 0
		486	if (0 && sv_derived_from (rv, "JSON::Literal"))
		487	{
		488	// not yet
		489	}
		490	#endif
		491	if (enc->json.flags & F_CONV_BLESSED)
		492	{
		493	// we re-bless the reference to get overload and other niceties right
		494	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		495
		496	if (to_json)
		497	{
		498	dSP;
		499
		500	ENTER; SAVETMPS; 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 return value
		504	// check anyways.
		505	PUTBACK;
		506	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		507	SPAGAIN;
		508
		509	// catch this surprisingly common error
		510	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		511	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		512
		513	sv = POPs;
		514	PUTBACK;
		515
		516	encode_sv (enc, sv);
		517
		518	FREETMPS; LEAVE;
		519	}
		520	else if (enc->json.flags & F_ALLOW_BLESSED)
		521	encode_str (enc, "null", 4, 0);
		522	else
		523	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		524	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		525	}
		526	else if (enc->json.flags & F_ALLOW_BLESSED)
		527	encode_str (enc, "null", 4, 0);
		528	else
		529	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		530	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		531	}
		532	}
		533	else if (svt == SVt_PVHV)
		534	encode_hv (enc, (HV *)sv);
		535	else if (svt == SVt_PVAV)
		536	encode_av (enc, (AV *)sv);
		537	else if (svt < SVt_PVAV)
		538	{
		539	STRLEN len = 0;
		540	char *pv = svt ? SvPV (sv, len) : 0;
		541
		542	if (len == 1 && *pv == '1')
		543	encode_str (enc, "true", 4, 0);
		544	else if (len == 1 && *pv == '0')
		545	encode_str (enc, "false", 5, 0);
		546	else
		547	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		548	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		549	}
		550	else
		551	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		552	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313	}	553	}
314		554
315	static void	555	static void
316	encode_sv (enc_t *enc, SV *sv)	556	encode_sv (enc_t *enc, SV *sv)
317	{	557	{
		558	SvGETMAGIC (sv);
		559
318	if (SvPOKp (sv))	560	if (SvPOKp (sv))
319	{	561	{
320	STRLEN len;	562	STRLEN len;
321	char *str = SvPV (sv, len);	563	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	564	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	565	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	566	encode_ch (enc, '"');
325	}	567	}
326	else if (SvNOKp (sv))	568	else if (SvNOKp (sv))
327	{	569	{
		570	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	571	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	572	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	573	enc->cur += strlen (enc->cur);
331	}	574	}
332	else if (SvIOKp (sv))	575	else if (SvIOKp (sv))
333	{	576	{
334	need (enc, 64);	577	// we assume we can always read an IV as a UV
		578	if (SvUV (sv) & ~(UV)0x7fff)
		579	{
		580	// large integer, use the (rather slow) snprintf way.
		581	need (enc, sizeof (UV) * 3);
335	enc->cur +=	582	enc->cur +=
336	SvIsUV(sv)	583	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	584	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	585	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		586	}
		587	else
		588	{
		589	// optimise the "small number case"
		590	// code will likely be branchless and use only a single multiplication
		591	I32 i = SvIV (sv);
		592	U32 u;
		593	char digit, nz = 0;
		594
		595	need (enc, 6);
		596
		597	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		598	u = i < 0 ? -i : i;
		599
		600	// convert to 4.28 fixed-point representation
		601	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		602
		603	// now output digit by digit, each time masking out the integer part
		604	// and multiplying by 5 while moving the decimal point one to the right,
		605	// resulting in a net multiplication by 10.
		606	// we always write the digit to memory but conditionally increment
		607	// the pointer, to ease the usage of conditional move instructions.
		608	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		609	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		610	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		611	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		612	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		613	}
339	}	614	}
340	else if (SvROK (sv))	615	else if (SvROK (sv))
341	{	616	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))	617	else if (!SvOK (sv))
357	encode_str (enc, "null", 4, 0);	618	encode_str (enc, "null", 4, 0);
358	else	619	else
359	croak ("encountered perl type that JSON cannot handle");	620	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		621	SvPV_nolen (sv), SvFLAGS (sv));
360	}	622	}
361		623
362	static SV *	624	static SV *
363	encode_json (SV *scalar, UV flags)	625	encode_json (SV *scalar, JSON *json)
364	{	626	{
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;	627	enc_t enc;
369	enc.flags = flags;	628
		629	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		630	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		631
		632	enc.json = *json;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	633	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	634	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	635	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	636	enc.indent = 0;
		637	enc.maxdepth = DEC_DEPTH (enc.json.flags);
375		638
376	SvPOK_only (enc.sv);	639	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	640	encode_sv (&enc, scalar);
378		641
		642	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		643	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		644
379	if (!(flags & (F_ASCII | F_UTF8)))	645	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	646	SvUTF8_on (enc.sv);
381		647
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	648	if (enc.json.flags & F_SHRINK)
		649	shrink (enc.sv);
		650
383	return enc.sv;	651	return enc.sv;
384	}	652	}
385		653
386	/////////////////////////////////////////////////////////////////////////////	654	/////////////////////////////////////////////////////////////////////////////
		655	// decoder
387		656
388	#define WS \	657	// structure used for decoding JSON
		658	typedef struct
		659	{
		660	char *cur; // current parser pointer
		661	char *end; // end of input string
		662	const char *err; // parse error, if != 0
		663	JSON json;
		664	U32 depth; // recursion depth
		665	U32 maxdepth; // recursion depth limit
		666	} dec_t;
		667
		668	inline void
		669	decode_ws (dec_t *dec)
		670	{
389	for (;;) \	671	for (;;)
390	{ \	672	{
391	char ch = *dec->cur; \	673	char ch = *dec->cur;
		674
392	if (ch > 0x20 \	675	if (ch > 0x20
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	676	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394	break; \	677	break;
		678
395	++dec->cur; \	679	++dec->cur;
396	}	680	}
		681	}
397		682
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	683	#define ERR(reason) SB dec->err = reason; goto fail; SE
		684
399	#define EXPECT_CH(ch) SB \	685	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	686	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	687	ERR (# ch " expected"); \
402	++dec->cur; \	688	++dec->cur; \
403	SE	689	SE
404		690
		691	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		692	#define DEC_DEC_DEPTH --dec->depth
		693
405	static SV *decode_sv (dec_t *dec);	694	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		695
412	static signed char decode_hexdigit[256];	696	static signed char decode_hexdigit[256];
413		697
414	static UV	698	static UV
415	decode_4hex (dec_t *dec)	699	decode_4hex (dec_t *dec)
416	{	700	{
417	signed char d1, d2, d3, d4;	701	signed char d1, d2, d3, d4;
		702	unsigned char *cur = (unsigned char *)dec->cur;
418		703
419	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	704	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");	705	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]];	706	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");	707	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		708
428	dec->cur += 4;	709	dec->cur += 4;
429		710
430	return ((UV)d1) << 12	711	return ((UV)d1) << 12
431	| ((UV)d2) << 8	712	| ((UV)d2) << 8
…		…
437	}	718	}
438		719
439	static SV *	720	static SV *
440	decode_str (dec_t *dec)	721	decode_str (dec_t *dec)
441	{	722	{
442	SV *sv = NEWSV (0,2);	723	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	724	int utf8 = 0;
		725	char *dec_cur = dec->cur;
445		726
446	for (;;)	727	do
447	{	728	{
448	unsigned char ch = *(unsigned char *)dec->cur;	729	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		730	char *cur = buf;
449		731
450	if (ch == '"')	732	do
451	break;
452	else if (ch == '\\')
453	{	733	{
454	switch (*++dec->cur)	734	unsigned char ch = *(unsigned char *)dec_cur++;
		735
		736	if (expect_false (ch == '"'))
		737	{
		738	--dec_cur;
		739	break;
455	{	740	}
456	case '\\':	741	else if (expect_false (ch == '\\'))
457	case '/':	742	{
458	case '"': APPEND_CH (*dec->cur++); break;	743	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	{	744	{
468	UV lo, hi;	745	case '\\':
469	++dec->cur;	746	case '/':
		747	case '"': *cur++ = *dec_cur++; break;
470		748
471	hi = decode_4hex (dec);	749	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	750	case 't': ++dec_cur; *cur++ = '\011'; break;
473	goto fail;	751	case 'n': ++dec_cur; *cur++ = '\012'; break;
		752	case 'f': ++dec_cur; *cur++ = '\014'; break;
		753	case 'r': ++dec_cur; *cur++ = '\015'; break;
474		754
475	// possibly a surrogate pair	755	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	756	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	757	UV lo, hi;
479	ERR ("illegal surrogate character");	758	++dec_cur;
480		759
481	dec->cur += 2;	760	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	761	hi = decode_4hex (dec);
		762	dec_cur = dec->cur;
484	if (lo == (UV)-1)	763	if (hi == (UV)-1)
485	goto fail;	764	goto fail;
486		765
		766	// possibly a surrogate pair
		767	if (hi >= 0xd800)
		768	if (hi < 0xdc00)
		769	{
		770	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		771	ERR ("missing low surrogate character in surrogate pair");
		772
		773	dec_cur += 2;
		774
		775	dec->cur = dec_cur;
		776	lo = decode_4hex (dec);
		777	dec_cur = dec->cur;
		778	if (lo == (UV)-1)
		779	goto fail;
		780
487	if (lo < 0xdc00 || lo >= 0xe000)	781	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	782	ERR ("surrogate pair expected");
489		783
490	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	784	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		785	}
		786	else if (hi < 0xe000)
		787	ERR ("missing high surrogate character in surrogate pair");
		788
		789	if (hi >= 0x80)
		790	{
		791	utf8 = 1;
		792
		793	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		794	}
		795	else
		796	*cur++ = hi;
491	}	797	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	798	break;
497	utf8 = 1;
498		799
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	800	default:
503	APPEND_CH (hi);	801	--dec_cur;
		802	ERR ("illegal backslash escape sequence in string");
504	}	803	}
		804	}
		805	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		806	*cur++ = ch;
		807	else if (ch >= 0x80)
		808	{
		809	STRLEN clen;
		810	UV uch;
		811
		812	--dec_cur;
		813
		814	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		815	if (clen == (STRLEN)-1)
		816	ERR ("malformed UTF-8 character in JSON string");
		817
		818	do
		819	*cur++ = *dec_cur++;
		820	while (--clen);
		821
		822	utf8 = 1;
		823	}
		824	else
		825	{
		826	--dec_cur;
		827
		828	if (!ch)
		829	ERR ("unexpected end of string while parsing JSON string");
505	break;	830	else
		831	ERR ("invalid character encountered while parsing JSON string");
506	}	832	}
507	}	833	}
508	else if (ch >= 0x20 && ch <= 0x7f)	834	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	835
510	else if (ch >= 0x80)	836	{
		837	STRLEN len = cur - buf;
		838
		839	if (sv)
511	{	840	{
512	STRLEN clen;	841	SvGROW (sv, SvCUR (sv) + len + 1);
513	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	842	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514	if (clen < 0)	843	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	}	844	}
521	else	845	else
522	ERR ("invalid character encountered");	846	sv = newSVpvn (buf, len);
523	}	847	}
		848	}
		849	while (*dec_cur != '"');
524		850
525	++dec->cur;	851	++dec_cur;
526		852
		853	if (sv)
		854	{
527	SvPOK_only (sv);	855	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	856	*SvEND (sv) = 0;
531		857
532	if (utf8)	858	if (utf8)
533	SvUTF8_on (sv);	859	SvUTF8_on (sv);
		860	}
		861	else
		862	sv = newSVpvn ("", 0);
534		863
		864	dec->cur = dec_cur;
535	return sv;	865	return sv;
536		866
537	fail:	867	fail:
538	SvREFCNT_dec (sv);	868	dec->cur = dec_cur;
539	return 0;	869	return 0;
540	}	870	}
541		871
542	static SV *	872	static SV *
543	decode_num (dec_t *dec)	873	decode_num (dec_t *dec)
…		…
553	{	883	{
554	++dec->cur;	884	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	885	if (*dec->cur >= '0' && *dec->cur <= '9')
556	ERR ("malformed number (leading zero must not be followed by another digit)");	886	ERR ("malformed number (leading zero must not be followed by another digit)");
557	}	887	}
558		888	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	889	ERR ("malformed number (no digits after initial minus)");
		890	else
		891	do
		892	{
		893	++dec->cur;
		894	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	895	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		896
563	// [frac]	897	// [frac]
564	if (*dec->cur == '.')	898	if (*dec->cur == '.')
565	{	899	{
566	is_nv = 1;	900	++dec->cur;
		901
		902	if (*dec->cur < '0' || *dec->cur > '9')
		903	ERR ("malformed number (no digits after decimal point)");
567		904
568	do	905	do
569	{	906	{
570	++dec->cur;	907	++dec->cur;
571	}	908	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	909	while (*dec->cur >= '0' && *dec->cur <= '9');
		910
		911	is_nv = 1;
573	}	912	}
574		913
575	// [exp]	914	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	915	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	916	{
578	is_nv = 1;
579
580	++dec->cur;	917	++dec->cur;
		918
581	if (*dec->cur == '-' || *dec->cur == '+')	919	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	920	++dec->cur;
583		921
		922	if (*dec->cur < '0' || *dec->cur > '9')
		923	ERR ("malformed number (no digits after exp sign)");
		924
		925	do
		926	{
		927	++dec->cur;
		928	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	929	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	930
		931	is_nv = 1;
586	}	932	}
587		933
588	if (!is_nv)	934	if (!is_nv)
589	{	935	{
590	UV uv;	936	// 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);	937	if (*start == '-')
592	if (numtype & IS_NUMBER_IN_UV)	938	switch (dec->cur - start)
593	if (numtype & IS_NUMBER_NEG)
594	{	939	{
595	if (uv < (UV)IV_MIN)	940	case 2: return newSViv (-( start [1] - '0' * 1));
596	return newSViv (-(IV)uv);	941	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		942	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		943	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
597	}	944	}
		945	else
		946	switch (dec->cur - start)
		947	{
		948	case 1: return newSViv ( start [0] - '0' * 1);
		949	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		950	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		951	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		952	}
		953
		954	{
		955	UV uv;
		956	int numtype = grok_number (start, dec->cur - start, &uv);
		957	if (numtype & IS_NUMBER_IN_UV)
		958	if (numtype & IS_NUMBER_NEG)
		959	{
		960	if (uv < (UV)IV_MIN)
		961	return newSViv (-(IV)uv);
		962	}
598	else	963	else
599	return newSVuv (uv);	964	return newSVuv (uv);
		965
		966	// here would likely be the place for bigint support
600	}	967	}
		968	}
601		969
		970	// if we ever support bigint or bigfloat, this is the place for bigfloat
602	return newSVnv (Atof (start));	971	return newSVnv (Atof (start));
603		972
604	fail:	973	fail:
605	return 0;	974	return 0;
606	}	975	}
…		…
608	static SV *	977	static SV *
609	decode_av (dec_t *dec)	978	decode_av (dec_t *dec)
610	{	979	{
611	AV *av = newAV ();	980	AV *av = newAV ();
612		981
		982	DEC_INC_DEPTH;
		983	decode_ws (dec);
		984
		985	if (*dec->cur == ']')
		986	++dec->cur;
		987	else
613	for (;;)	988	for (;;)
614	{	989	{
615	SV *value;	990	SV *value;
616		991
617	value = decode_sv (dec);	992	value = decode_sv (dec);
618	if (!value)	993	if (!value)
619	goto fail;	994	goto fail;
620		995
621	av_push (av, value);	996	av_push (av, value);
622		997
623	WS;	998	decode_ws (dec);
624		999
625	if (*dec->cur == ']')	1000	if (*dec->cur == ']')
626	{	1001	{
627	++dec->cur;	1002	++dec->cur;
628	break;	1003	break;
		1004	}
629	}	1005
630
631	if (*dec->cur != ',')	1006	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	1007	ERR (", or ] expected while parsing array");
633		1008
634	++dec->cur;	1009	++dec->cur;
635	}	1010	}
636		1011
		1012	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1013	return newRV_noinc ((SV *)av);
638		1014
639	fail:	1015	fail:
640	SvREFCNT_dec (av);	1016	SvREFCNT_dec (av);
		1017	DEC_DEC_DEPTH;
641	return 0;	1018	return 0;
642	}	1019	}
643		1020
644	static SV *	1021	static SV *
645	decode_hv (dec_t *dec)	1022	decode_hv (dec_t *dec)
646	{	1023	{
		1024	SV *sv;
647	HV *hv = newHV ();	1025	HV *hv = newHV ();
648		1026
		1027	DEC_INC_DEPTH;
		1028	decode_ws (dec);
		1029
		1030	if (*dec->cur == '}')
		1031	++dec->cur;
		1032	else
649	for (;;)	1033	for (;;)
650	{	1034	{
651	SV *key, *value;	1035	decode_ws (dec); EXPECT_CH ('"');
652		1036
653	WS; EXPECT_CH ('"');	1037	// heuristic: assume that
654		1038	// a) decode_str + hv_store_ent are abysmally slow.
655	key = decode_str (dec);	1039	// b) most hash keys are short, simple ascii text.
656	if (!key)	1040	// => try to "fast-match" such strings to avoid
657	goto fail;	1041	// the overhead of decode_str + hv_store_ent.
658
659	WS; EXPECT_CH (':');
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1042	{
		1043	SV *value;
		1044	char *p = dec->cur;
		1045	char *e = p + 24; // only try up to 24 bytes
		1046
		1047	for (;;)
		1048	{
		1049	// the >= 0x80 is true on most architectures
		1050	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1051	{
		1052	// slow path, back up and use decode_str
		1053	SV *key = decode_str (dec);
		1054	if (!key)
		1055	goto fail;
		1056
		1057	decode_ws (dec); EXPECT_CH (':');
		1058
		1059	value = decode_sv (dec);
		1060	if (!value)
		1061	{
		1062	SvREFCNT_dec (key);
		1063	goto fail;
		1064	}
		1065
		1066	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1067	SvREFCNT_dec (key);
		1068
		1069	break;
		1070	}
		1071	else if (*p == '"')
		1072	{
		1073	// fast path, got a simple key
		1074	char *key = dec->cur;
		1075	int len = p - key;
		1076	dec->cur = p + 1;
		1077
		1078	decode_ws (dec); EXPECT_CH (':');
		1079
		1080	value = decode_sv (dec);
		1081	if (!value)
665	goto fail;	1082	goto fail;
		1083
		1084	hv_store (hv, key, len, value, 0);
		1085
		1086	break;
		1087	}
		1088
		1089	++p;
		1090	}
666	}	1091	}
667		1092
668	//TODO: optimise	1093	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1094
671	WS;
672
673	if (*dec->cur == '}')	1095	if (*dec->cur == '}')
		1096	{
		1097	++dec->cur;
		1098	break;
		1099	}
		1100
		1101	if (*dec->cur != ',')
		1102	ERR (", or } expected while parsing object/hash");
		1103
		1104	++dec->cur;
		1105	}
		1106
		1107	DEC_DEC_DEPTH;
		1108	sv = newRV_noinc ((SV *)hv);
		1109
		1110	// check filter callbacks
		1111	if (dec->json.flags & F_HOOK)
		1112	{
		1113	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
674	{	1114	{
675	++dec->cur;	1115	HE *cb, *he;
676	break;	1116
		1117	hv_iterinit (hv);
		1118	he = hv_iternext (hv);
		1119	hv_iterinit (hv);
		1120
		1121	// the next line creates a mortal sv each time its called.
		1122	// might want to optimise this for common cases.
		1123	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1124
		1125	if (cb)
		1126	{
		1127	dSP;
		1128	int count;
		1129
		1130	ENTER; SAVETMPS; PUSHMARK (SP);
		1131	XPUSHs (HeVAL (he));
		1132
		1133	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1134
		1135	if (count == 1)
		1136	{
		1137	sv = newSVsv (POPs);
		1138	FREETMPS; LEAVE;
		1139	return sv;
		1140	}
		1141
		1142	FREETMPS; LEAVE;
		1143	}
677	}	1144	}
678		1145
679	if (*dec->cur != ',')	1146	if (dec->json.cb_object)
680	ERR (", or } expected while parsing object/hash");	1147	{
		1148	dSP;
		1149	int count;
681		1150
682	++dec->cur;	1151	ENTER; SAVETMPS; PUSHMARK (SP);
683	}	1152	XPUSHs (sv_2mortal (sv));
684		1153
685	return newRV_noinc ((SV *)hv);	1154	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1155
		1156	if (count == 1)
		1157	{
		1158	sv = newSVsv (POPs);
		1159	FREETMPS; LEAVE;
		1160	return sv;
		1161	}
		1162
		1163	SvREFCNT_inc (sv);
		1164	FREETMPS; LEAVE;
		1165	}
		1166	}
		1167
		1168	return sv;
686		1169
687	fail:	1170	fail:
688	SvREFCNT_dec (hv);	1171	SvREFCNT_dec (hv);
		1172	DEC_DEC_DEPTH;
689	return 0;	1173	return 0;
690	}	1174	}
691		1175
692	static SV *	1176	static SV *
693	decode_sv (dec_t *dec)	1177	decode_sv (dec_t *dec)
694	{	1178	{
695	WS;	1179	decode_ws (dec);
		1180
		1181	// the beauty of JSON: you need exactly one character lookahead
		1182	// to parse anything.
696	switch (*dec->cur)	1183	switch (*dec->cur)
697	{	1184	{
698	case '"': ++dec->cur; return decode_str (dec);	1185	case '"': ++dec->cur; return decode_str (dec);
699	case '[': ++dec->cur; return decode_av (dec);	1186	case '[': ++dec->cur; return decode_av (dec);
700	case '{': ++dec->cur; return decode_hv (dec);	1187	case '{': ++dec->cur; return decode_hv (dec);
…		…
706		1193
707	case 't':	1194	case 't':
708	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1195	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709	{	1196	{
710	dec->cur += 4;	1197	dec->cur += 4;
711	return newSViv (1);	1198	return SvREFCNT_inc (json_true);
712	}	1199	}
713	else	1200	else
714	ERR ("'true' expected");	1201	ERR ("'true' expected");
715		1202
716	break;	1203	break;
717		1204
718	case 'f':	1205	case 'f':
719	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1206	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720	{	1207	{
721	dec->cur += 5;	1208	dec->cur += 5;
722	return newSViv (0);	1209	return SvREFCNT_inc (json_false);
723	}	1210	}
724	else	1211	else
725	ERR ("'false' expected");	1212	ERR ("'false' expected");
726		1213
727	break;	1214	break;
728		1215
729	case 'n':	1216	case 'n':
730	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1217	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731	{	1218	{
732	dec->cur += 4;	1219	dec->cur += 4;
733	return newSViv (1);	1220	return newSVsv (&PL_sv_undef);
734	}	1221	}
735	else	1222	else
736	ERR ("'null' expected");	1223	ERR ("'null' expected");
737		1224
738	break;	1225	break;
739		1226
740	default:	1227	default:
741	ERR ("malformed json string");	1228	ERR ("malformed JSON string, neither array, object, number, string or atom");
742	break;	1229	break;
743	}	1230	}
744		1231
745	fail:	1232	fail:
746	return 0;	1233	return 0;
747	}	1234	}
748		1235
749	static SV *	1236	static SV *
750	decode_json (SV *string, UV flags)	1237	decode_json (SV *string, JSON *json, UV *offset_return)
751	{	1238	{
		1239	dec_t dec;
		1240	UV offset;
752	SV *sv;	1241	SV *sv;
753		1242
		1243	SvGETMAGIC (string);
		1244	SvUPGRADE (string, SVt_PV);
		1245
		1246	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1247	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1248	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1249
754	if (!(flags & F_UTF8))	1250	if (json->flags & F_UTF8)
		1251	sv_utf8_downgrade (string, 0);
		1252	else
755	sv_utf8_upgrade (string);	1253	sv_utf8_upgrade (string);
756		1254
757	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1255	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758		1256
759	dec_t dec;	1257	dec.json = *json;
760	dec.flags = flags;
761	dec.cur = SvPVX (string);	1258	dec.cur = SvPVX (string);
762	dec.end = SvEND (string);	1259	dec.end = SvEND (string);
763	dec.err = 0;	1260	dec.err = 0;
		1261	dec.depth = 0;
		1262	dec.maxdepth = DEC_DEPTH (dec.json.flags);
764		1263
765	*dec.end = 1; // invalid anywhere	1264	if (dec.json.cb_object || dec.json.cb_sk_object)
		1265	dec.json.flags |= F_HOOK;
		1266
		1267	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
766	sv = decode_sv (&dec);	1268	sv = decode_sv (&dec);
767	*dec.end = 0;	1269
		1270	if (!(offset_return || !sv))
		1271	{
		1272	// check for trailing garbage
		1273	decode_ws (&dec);
		1274
		1275	if (*dec.cur)
		1276	{
		1277	dec.err = "garbage after JSON object";
		1278	SvREFCNT_dec (sv);
		1279	sv = 0;
		1280	}
		1281	}
		1282
		1283	if (offset_return || !sv)
		1284	{
		1285	offset = dec.json.flags & F_UTF8
		1286	? dec.cur - SvPVX (string)
		1287	: utf8_distance (dec.cur, SvPVX (string));
		1288
		1289	if (offset_return)
		1290	*offset_return = offset;
		1291	}
768		1292
769	if (!sv)	1293	if (!sv)
770	{	1294	{
771	IV offset = utf8_distance (dec.cur, SvPVX (string));
772	SV *uni = sv_newmortal ();	1295	SV *uni = sv_newmortal ();
773		1296
		1297	// horrible hack to silence warning inside pv_uni_display
		1298	COP cop = *PL_curcop;
		1299	cop.cop_warnings = pWARN_NONE;
		1300	ENTER;
		1301	SAVEVPTR (PL_curcop);
		1302	PL_curcop = &cop;
774	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1303	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1304	LEAVE;
		1305
775	croak ("%s, at character %d (%s)",	1306	croak ("%s, at character offset %d [\"%s\"]",
776	dec.err,	1307	dec.err,
777	(int)offset,	1308	(int)offset,
778	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1309	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779	}	1310	}
780		1311
781	sv = sv_2mortal (sv);	1312	sv = sv_2mortal (sv);
782		1313
783	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1314	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)");	1315	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785		1316
786	return sv;	1317	return sv;
787	}	1318	}
788		1319
		1320	/////////////////////////////////////////////////////////////////////////////
		1321	// XS interface functions
		1322
789	MODULE = JSON::XS PACKAGE = JSON::XS	1323	MODULE = JSON::XS PACKAGE = JSON::XS
790		1324
791	BOOT:	1325	BOOT:
792	{	1326	{
793	int i;	1327	int i;
794		1328
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	1329	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	1330	decode_hexdigit [i] =
		1331	i >= '0' && i <= '9' ? i - '0'
		1332	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1333	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1334	: -1;
798		1335
799	for (i = 6; --i; )	1336	json_stash = gv_stashpv ("JSON::XS" , 1);
		1337	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1338
		1339	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1340	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1341	}
		1342
		1343	PROTOTYPES: DISABLE
		1344
		1345	void new (char *klass)
		1346	PPCODE:
		1347	{
		1348	SV *pv = NEWSV (0, sizeof (JSON));
		1349	SvPOK_only (pv);
		1350	Zero (SvPVX (pv), 1, JSON);
		1351	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1352	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1353	}
		1354
		1355	void ascii (JSON *self, int enable = 1)
		1356	ALIAS:
		1357	ascii = F_ASCII
		1358	latin1 = F_LATIN1
		1359	utf8 = F_UTF8
		1360	indent = F_INDENT
		1361	canonical = F_CANONICAL
		1362	space_before = F_SPACE_BEFORE
		1363	space_after = F_SPACE_AFTER
		1364	pretty = F_PRETTY
		1365	allow_nonref = F_ALLOW_NONREF
		1366	shrink = F_SHRINK
		1367	allow_blessed = F_ALLOW_BLESSED
		1368	convert_blessed = F_CONV_BLESSED
		1369	PPCODE:
		1370	{
		1371	if (enable)
		1372	self->flags |= ix;
		1373	else
		1374	self->flags &= ~ix;
		1375
		1376	XPUSHs (ST (0));
		1377	}
		1378
		1379	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1380	PPCODE:
		1381	{
		1382	UV log2 = 0;
		1383
		1384	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1385
		1386	while ((1UL << log2) < max_depth)
		1387	++log2;
		1388
		1389	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1390
		1391	XPUSHs (ST (0));
		1392	}
		1393
		1394	void max_size (JSON *self, UV max_size = 0)
		1395	PPCODE:
		1396	{
		1397	UV log2 = 0;
		1398
		1399	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1400	if (max_size == 1) max_size = 2;
		1401
		1402	while ((1UL << log2) < max_size)
		1403	++log2;
		1404
		1405	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1406
		1407	XPUSHs (ST (0));
		1408	}
		1409
		1410	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1411	PPCODE:
		1412	{
		1413	SvREFCNT_dec (self->cb_object);
		1414	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1415
		1416	XPUSHs (ST (0));
		1417	}
		1418
		1419	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1420	PPCODE:
		1421	{
		1422	if (!self->cb_sk_object)
		1423	self->cb_sk_object = newHV ();
		1424
		1425	if (SvOK (cb))
		1426	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1427	else
800	{	1428	{
801	decode_hexdigit ['a' + i] = 10 + i;	1429	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
802	decode_hexdigit ['A' + i] = 10 + i;	1430
		1431	if (!HvKEYS (self->cb_sk_object))
		1432	{
		1433	SvREFCNT_dec (self->cb_sk_object);
		1434	self->cb_sk_object = 0;
		1435	}
803	}	1436	}
804		1437
805	json_stash = gv_stashpv ("JSON::XS", 1);	1438	XPUSHs (ST (0));
806	}	1439	}
807		1440
808	SV *new (char *dummy)	1441	void encode (JSON *self, SV *scalar)
		1442	PPCODE:
		1443	XPUSHs (encode_json (scalar, self));
		1444
		1445	void decode (JSON *self, SV *jsonstr)
		1446	PPCODE:
		1447	XPUSHs (decode_json (jsonstr, self, 0));
		1448
		1449	void decode_prefix (JSON *self, SV *jsonstr)
		1450	PPCODE:
		1451	{
		1452	UV offset;
		1453	EXTEND (SP, 2);
		1454	PUSHs (decode_json (jsonstr, self, &offset));
		1455	PUSHs (sv_2mortal (newSVuv (offset)));
		1456	}
		1457
		1458	void DESTROY (JSON *self)
809	CODE:	1459	CODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1460	SvREFCNT_dec (self->cb_sk_object);
811	OUTPUT:	1461	SvREFCNT_dec (self->cb_object);
812	RETVAL
813		1462
814	SV *ascii (SV *self, int enable)	1463	PROTOTYPES: ENABLE
815	ALIAS:
816	ascii = F_ASCII
817	utf8 = F_UTF8
818	indent = F_INDENT
819	canonical = F_CANONICAL
820	space_before = F_SPACE_BEFORE
821	space_after = F_SPACE_AFTER
822	json_rpc = F_JSON_RPC
823	pretty = F_PRETTY
824	allow_nonref = F_ALLOW_NONREF
825	CODE:
826	{
827	UV *uv = SvJSON (self);
828	if (enable)
829	*uv |= ix;
830	else
831	*uv &= ~ix;
832
833	RETVAL = newSVsv (self);
834	}
835	OUTPUT:
836	RETVAL
837
838	void encode (SV *self, SV *scalar)
839	PPCODE:
840	XPUSHs (encode_json (scalar, *SvJSON (self)));
841
842	void decode (SV *self, SV *jsonstr)
843	PPCODE:
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
845		1464
846	void to_json (SV *scalar)	1465	void to_json (SV *scalar)
847	PPCODE:	1466	PPCODE:
		1467	{
		1468	JSON json = { F_DEFAULT | F_UTF8 };
848	XPUSHs (encode_json (scalar, F_UTF8));	1469	XPUSHs (encode_json (scalar, &json));
		1470	}
849		1471
850	void from_json (SV *jsonstr)	1472	void from_json (SV *jsonstr)
851	PPCODE:	1473	PPCODE:
		1474	{
		1475	JSON json = { F_DEFAULT | F_UTF8 };
852	XPUSHs (decode_json (jsonstr, F_UTF8));	1476	XPUSHs (decode_json (jsonstr, &json, 0));
		1477	}
853		1478

Diff Legend

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