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 1.47 by root, Sun Jul 1 14:08:03 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 json {
		65	U32 flags;
		66	} JSON__XS;
		67
		68	/////////////////////////////////////////////////////////////////////////////
		69	// utility functions
		70
		71	static UV *
		72	SvJSON (SV *sv)
		73	{
		74	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		75	croak ("object is not of type JSON::XS");
		76
		77	return &SvUVX (SvRV (sv));
		78	}
		79
		80	static void
		81	shrink (SV *sv)
		82	{
		83	sv_utf8_downgrade (sv, 1);
		84	if (SvLEN (sv) > SvCUR (sv) + 1)
		85	{
		86	#ifdef SvPV_shrink_to_cur
		87	SvPV_shrink_to_cur (sv);
		88	#elif defined (SvPV_renew)
		89	SvPV_renew (sv, SvCUR (sv) + 1);
		90	#endif
		91	}
		92	}
		93
		94	// decode an utf-8 character and return it, or (UV)-1 in
		95	// case of an error.
		96	// we special-case "safe" characters from U+80 .. U+7FF,
		97	// but use the very good perl function to parse anything else.
		98	// note that we never call this function for a ascii codepoints
		99	inline UV
		100	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		101	{
		102	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		103	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		104	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		105	{
		106	*clen = 2;
		107	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		108	}
		109	else
		110	{
		111	*clen = (STRLEN)-1;
		112	return (UV)-1;
		113	}
		114	}
		115
		116	/////////////////////////////////////////////////////////////////////////////
		117	// encoder
27		118
28	// structure used for encoding JSON	119	// structure used for encoding JSON
29	typedef struct	120	typedef struct
30	{	121	{
31	char *cur;	122	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	123	char *end; // SvEND (sv)
34	SV *sv;	124	SV *sv; // result scalar
35	UV flags;	125	struct json json;
36	int max_recurse;	126	U32 indent; // indentation level
37	int indent;	127	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	128	} enc_t;
39		129
40	// structure used for decoding JSON	130	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)	131	need (enc_t *enc, STRLEN len)
62	{	132	{
63	if (enc->cur + len >= enc->end)	133	if (expect_false (enc->cur + len >= enc->end))
64	{	134	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	135	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	136	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	137	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvEND (enc->sv);	138	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	139	}
70	}	140	}
71		141
72	static void	142	inline void
73	encode_ch (enc_t *enc, char ch)	143	encode_ch (enc_t *enc, char ch)
74	{	144	{
75	need (enc, 1);	145	need (enc, 1);
76	*enc->cur++ = ch;	146	*enc->cur++ = ch;
77	}	147	}
…		…
79	static void	149	static void
80	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	150	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
81	{	151	{
82	char *end = str + len;	152	char *end = str + len;
83		153
		154	need (enc, len);
		155
84	while (str < end)	156	while (str < end)
85	{	157	{
86	unsigned char ch = *(unsigned char *)str;	158	unsigned char ch = *(unsigned char *)str;
		159
87	if (ch >= 0x20 && ch < 0x80) // most common case	160	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
88	{	161	{
		162	if (expect_false (ch == '"')) // but with slow exceptions
		163	{
		164	need (enc, len += 1);
		165	*enc->cur++ = '\\';
		166	*enc->cur++ = '"';
		167	}
		168	else if (expect_false (ch == '\\'))
		169	{
		170	need (enc, len += 1);
		171	*enc->cur++ = '\\';
		172	*enc->cur++ = '\\';
		173	}
		174	else
89	*enc->cur++ = ch;	175	*enc->cur++ = ch;
		176
90	str++;	177	++str;
91	}	178	}
92	else	179	else
93	{	180	{
94	STRLEN clen;	181	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	}	182	{
103	else	183	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
104	{	184	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
105	uch = ch;	185	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
106	clen = 1;	186	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
107	}	187	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
108		188
109	need (enc, len += 6);	189	default:
110
111	if (uch < 0xa0 || enc->flags & F_ASCII)
112	{
113	if (uch > 0xFFFFUL)
114	{	190	{
		191	STRLEN clen;
		192	UV uch;
		193
		194	if (is_utf8)
		195	{
		196	uch = decode_utf8 (str, end - str, &clen);
		197	if (clen == (STRLEN)-1)
		198	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		199	}
		200	else
		201	{
		202	uch = ch;
115	len += 6;	203	clen = 1;
		204	}
		205
		206	if (uch > 0x10FFFFUL)
		207	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		208
		209	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		210	{
		211	if (uch > 0xFFFFUL)
		212	{
116	need (enc, len += 6);	213	need (enc, len += 11);
117	sprintf (enc->cur, "\\u%04x\\u%04x",	214	sprintf (enc->cur, "\\u%04x\\u%04x",
118	(uch - 0x10000) / 0x400 + 0xD800,	215	(int)((uch - 0x10000) / 0x400 + 0xD800),
119	(uch - 0x10000) % 0x400 + 0xDC00);	216	(int)((uch - 0x10000) % 0x400 + 0xDC00));
120	enc->cur += 12;	217	enc->cur += 12;
		218	}
		219	else
		220	{
		221	static char hexdigit [16] = "0123456789abcdef";
		222	need (enc, len += 5);
		223	*enc->cur++ = '\\';
		224	*enc->cur++ = 'u';
		225	*enc->cur++ = hexdigit [ uch >> 12 ];
		226	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		227	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		228	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		229	}
		230
		231	str += clen;
		232	}
		233	else if (enc->json.flags & F_LATIN1)
		234	{
		235	*enc->cur++ = uch;
		236	str += clen;
		237	}
		238	else if (is_utf8)
		239	{
		240	need (enc, len += clen);
		241	do
		242	{
		243	*enc->cur++ = *str++;
		244	}
		245	while (--clen);
		246	}
		247	else
		248	{
		249	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		250	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		251	++str;
		252	}
121	}	253	}
122	else
123	{
124	sprintf (enc->cur, "\\u%04x", uch);
125	enc->cur += 6;
126	}
127	}
128	else if (is_utf8)
129	{	254	}
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	}	255	}
138		256
139	--len;	257	--len;
140	}	258	}
141	}	259	}
142		260
143	#define INDENT SB \	261	inline void
		262	encode_indent (enc_t *enc)
		263	{
144	if (enc->flags & F_INDENT) \	264	if (enc->json.flags & F_INDENT)
145	{ \	265	{
146	int i_; \	266	int spaces = enc->indent * INDENT_STEP;
147	need (enc, enc->indent); \	267
148	for (i_ = enc->indent * 3; i_--; )\	268	need (enc, spaces);
		269	memset (enc->cur, ' ', spaces);
		270	enc->cur += spaces;
		271	}
		272	}
		273
		274	inline void
		275	encode_space (enc_t *enc)
		276	{
		277	need (enc, 1);
		278	encode_ch (enc, ' ');
		279	}
		280
		281	inline void
		282	encode_nl (enc_t *enc)
		283	{
		284	if (enc->json.flags & F_INDENT)
		285	{
		286	need (enc, 1);
149	encode_ch (enc, ' '); \	287	encode_ch (enc, '\n');
150	} \	288	}
151	SE	289	}
152		290
153	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	291	inline void
154	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	292	encode_comma (enc_t *enc)
155	#define COMMA SB \	293	{
156	encode_ch (enc, ','); \	294	encode_ch (enc, ',');
		295
157	if (enc->flags & F_INDENT) \	296	if (enc->json.flags & F_INDENT)
158	NL; \	297	encode_nl (enc);
159	else if (enc->flags & F_SPACE_AFTER) \	298	else if (enc->json.flags & F_SPACE_AFTER)
160	SPACE; \	299	encode_space (enc);
161	SE	300	}
162		301
163	static void encode_sv (enc_t *enc, SV *sv);	302	static void encode_sv (enc_t *enc, SV *sv);
164		303
165	static void	304	static void
166	encode_av (enc_t *enc, AV *av)	305	encode_av (enc_t *enc, AV *av)
167	{	306	{
168	int i, len = av_len (av);	307	int i, len = av_len (av);
169		308
170	encode_ch (enc, '['); NL;	309	if (enc->indent >= enc->maxdepth)
		310	croak ("data structure too deep (hit recursion limit)");
		311
		312	encode_ch (enc, '['); encode_nl (enc);
171	++enc->indent;	313	++enc->indent;
172		314
173	for (i = 0; i <= len; ++i)	315	for (i = 0; i <= len; ++i)
174	{	316	{
175	INDENT;	317	encode_indent (enc);
176	encode_sv (enc, *av_fetch (av, i, 0));	318	encode_sv (enc, *av_fetch (av, i, 0));
177		319
178	if (i < len)	320	if (i < len)
179	COMMA;	321	encode_comma (enc);
180	}	322	}
181		323
182	NL;	324	encode_nl (enc);
183		325
184	--enc->indent;	326	--enc->indent;
185	INDENT; encode_ch (enc, ']');	327	encode_indent (enc); encode_ch (enc, ']');
186	}	328	}
187		329
188	static void	330	static void
189	encode_he (enc_t *enc, HE *he)	331	encode_he (enc_t *enc, HE *he)
190	{	332	{
…		…
192		334
193	if (HeKLEN (he) == HEf_SVKEY)	335	if (HeKLEN (he) == HEf_SVKEY)
194	{	336	{
195	SV *sv = HeSVKEY (he);	337	SV *sv = HeSVKEY (he);
196	STRLEN len;	338	STRLEN len;
		339	char *str;
		340
		341	SvGETMAGIC (sv);
197	char *str = SvPV (sv, len);	342	str = SvPV (sv, len);
198		343
199	encode_str (enc, str, len, SvUTF8 (sv));	344	encode_str (enc, str, len, SvUTF8 (sv));
200	}	345	}
201	else	346	else
202	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	347	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
203		348
204	encode_ch (enc, '"');	349	encode_ch (enc, '"');
205		350
206	if (enc->flags & F_SPACE_BEFORE) SPACE;	351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
207	encode_ch (enc, ':');	352	encode_ch (enc, ':');
208	if (enc->flags & F_SPACE_AFTER ) SPACE;	353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
209	encode_sv (enc, HeVAL (he));	354	encode_sv (enc, HeVAL (he));
210	}	355	}
211		356
212	// compare hash entries, used when all keys are bytestrings	357	// compare hash entries, used when all keys are bytestrings
213	static int	358	static int
…		…
219	HE *b = *(HE **)b_;	364	HE *b = *(HE **)b_;
220		365
221	STRLEN la = HeKLEN (a);	366	STRLEN la = HeKLEN (a);
222	STRLEN lb = HeKLEN (b);	367	STRLEN lb = HeKLEN (b);
223		368
224	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
225	cmp = la < lb ? -1 : la == lb ? 0 : 1;	370	cmp = la - lb;
226		371
227	return cmp;	372	return cmp;
228	}	373	}
229		374
230	// compare hash entries, used when some keys are sv's or utf-x	375	// compare hash entries, used when some keys are sv's or utf-x
…		…
237	static void	382	static void
238	encode_hv (enc_t *enc, HV *hv)	383	encode_hv (enc_t *enc, HV *hv)
239	{	384	{
240	int count, i;	385	int count, i;
241		386
		387	if (enc->indent >= enc->maxdepth)
		388	croak ("data structure too deep (hit recursion limit)");
		389
242	encode_ch (enc, '{'); NL; ++enc->indent;	390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
243		391
244	if ((count = hv_iterinit (hv)))	392	if ((count = hv_iterinit (hv)))
245	{	393	{
246	// for canonical output we have to sort by keys first	394	// for canonical output we have to sort by keys first
247	// actually, this is mostly due to the stupid so-called	395	// actually, this is mostly due to the stupid so-called
248	// security workaround added somewhere in 5.8.x.	396	// security workaround added somewhere in 5.8.x.
249	// that randomises hash orderings	397	// that randomises hash orderings
250	if (enc->flags & F_CANONICAL)	398	if (enc->json.flags & F_CANONICAL)
251	{	399	{
252	HE *he, *hes [count];
253	int fast = 1;	400	int fast = 1;
		401	HE *he;
		402	#if defined(__BORLANDC__) || defined(_MSC_VER)
		403	HE **hes = _alloca (count * sizeof (HE));
		404	#else
		405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		406	#endif
254		407
255	i = 0;	408	i = 0;
256	while ((he = hv_iternext (hv)))	409	while ((he = hv_iternext (hv)))
257	{	410	{
258	hes [i++] = he;	411	hes [i++] = he;
…		…
264		417
265	if (fast)	418	if (fast)
266	qsort (hes, count, sizeof (HE *), he_cmp_fast);	419	qsort (hes, count, sizeof (HE *), he_cmp_fast);
267	else	420	else
268	{	421	{
269	// hack to disable "use bytes"	422	// hack to forcefully disable "use bytes"
270	COP *oldcop = PL_curcop, cop;	423	COP cop = *PL_curcop;
271	cop.op_private = 0;	424	cop.op_private = 0;
		425
		426	ENTER;
		427	SAVETMPS;
		428
		429	SAVEVPTR (PL_curcop);
272	PL_curcop = &cop;	430	PL_curcop = &cop;
273		431
274	SAVETMPS;
275	qsort (hes, count, sizeof (HE *), he_cmp_slow);	432	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		433
276	FREETMPS;	434	FREETMPS;
277		435	LEAVE;
278	PL_curcop = oldcop;
279	}	436	}
280		437
281	for (i = 0; i < count; ++i)	438	for (i = 0; i < count; ++i)
282	{	439	{
283	INDENT;	440	encode_indent (enc);
284	encode_he (enc, hes [i]);	441	encode_he (enc, hes [i]);
285		442
286	if (i < count - 1)	443	if (i < count - 1)
287	COMMA;	444	encode_comma (enc);
288	}
289
290	NL;	445	}
		446
		447	encode_nl (enc);
291	}	448	}
292	else	449	else
293	{	450	{
294	SV *sv;
295	HE *he = hv_iternext (hv);	451	HE *he = hv_iternext (hv);
296		452
297	for (;;)	453	for (;;)
298	{	454	{
299	INDENT;	455	encode_indent (enc);
300	encode_he (enc, he);	456	encode_he (enc, he);
301		457
302	if (!(he = hv_iternext (hv)))	458	if (!(he = hv_iternext (hv)))
303	break;	459	break;
304		460
305	COMMA;	461	encode_comma (enc);
306	}
307
308	NL;	462	}
		463
		464	encode_nl (enc);
309	}	465	}
310	}	466	}
311		467
312	--enc->indent; INDENT; encode_ch (enc, '}');	468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		469	}
		470
		471	// encode objects, arrays and special \0=false and \1=true values.
		472	static void
		473	encode_rv (enc_t *enc, SV *sv)
		474	{
		475	svtype svt;
		476
		477	SvGETMAGIC (sv);
		478	svt = SvTYPE (sv);
		479
		480	if (expect_false (SvOBJECT (sv)))
		481	{
		482	if (SvSTASH (sv) == json_boolean_stash)
		483	{
		484	if (SvIV (sv) == 0)
		485	encode_str (enc, "false", 5, 0);
		486	else
		487	encode_str (enc, "true", 4, 0);
		488	}
		489	else
		490	{
		491	#if 0
		492	if (0 && sv_derived_from (rv, "JSON::Literal"))
		493	{
		494	// not yet
		495	}
		496	#endif
		497	if (enc->json.flags & F_CONV_BLESSED)
		498	{
		499	// we re-bless the reference to get overload and other niceties right
		500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		501
		502	if (to_json)
		503	{
		504	dSP;
		505	ENTER;
		506	SAVETMPS;
		507	PUSHMARK (SP);
		508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		509
		510	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		511	// check anyways.
		512	PUTBACK;
		513	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		514	SPAGAIN;
		515
		516	encode_sv (enc, POPs);
		517
		518	FREETMPS;
		519	LEAVE;
		520	}
		521	else if (enc->json.flags & F_ALLOW_BLESSED)
		522	encode_str (enc, "null", 4, 0);
		523	else
		524	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		525	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		526	}
		527	else if (enc->json.flags & F_ALLOW_BLESSED)
		528	encode_str (enc, "null", 4, 0);
		529	else
		530	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		531	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		532	}
		533	}
		534	else if (svt == SVt_PVHV)
		535	encode_hv (enc, (HV *)sv);
		536	else if (svt == SVt_PVAV)
		537	encode_av (enc, (AV *)sv);
		538	else if (svt < SVt_PVAV)
		539	{
		540	if (SvNIOK (sv) && SvIV (sv) == 0)
		541	encode_str (enc, "false", 5, 0);
		542	else if (SvNIOK (sv) && SvIV (sv) == 1)
		543	encode_str (enc, "true", 4, 0);
		544	else
		545	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		547	}
		548	else
		549	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		550	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
313	}	551	}
314		552
315	static void	553	static void
316	encode_sv (enc_t *enc, SV *sv)	554	encode_sv (enc_t *enc, SV *sv)
317	{	555	{
		556	SvGETMAGIC (sv);
		557
318	if (SvPOKp (sv))	558	if (SvPOKp (sv))
319	{	559	{
320	STRLEN len;	560	STRLEN len;
321	char *str = SvPV (sv, len);	561	char *str = SvPV (sv, len);
322	encode_ch (enc, '"');	562	encode_ch (enc, '"');
323	encode_str (enc, str, len, SvUTF8 (sv));	563	encode_str (enc, str, len, SvUTF8 (sv));
324	encode_ch (enc, '"');	564	encode_ch (enc, '"');
325	}	565	}
326	else if (SvNOKp (sv))	566	else if (SvNOKp (sv))
327	{	567	{
		568	// trust that perl will do the right thing w.r.t. JSON syntax.
328	need (enc, NV_DIG + 32);	569	need (enc, NV_DIG + 32);
329	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	570	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
330	enc->cur += strlen (enc->cur);	571	enc->cur += strlen (enc->cur);
331	}	572	}
332	else if (SvIOKp (sv))	573	else if (SvIOKp (sv))
333	{	574	{
334	need (enc, 64);	575	// we assume we can always read an IV as a UV
		576	if (SvUV (sv) & ~(UV)0x7fff)
		577	{
		578	// large integer, use the (rather slow) snprintf way.
		579	need (enc, sizeof (UV) * 3);
335	enc->cur +=	580	enc->cur +=
336	SvIsUV(sv)	581	SvIsUV(sv)
337	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	582	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
338	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	583	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		584	}
		585	else
		586	{
		587	// optimise the "small number case"
		588	// code will likely be branchless and use only a single multiplication
		589	I32 i = SvIV (sv);
		590	U32 u;
		591	char digit, nz = 0;
		592
		593	need (enc, 6);
		594
		595	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		596	u = i < 0 ? -i : i;
		597
		598	// convert to 4.28 fixed-point representation
		599	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		600
		601	// now output digit by digit, each time masking out the integer part
		602	// and multiplying by 5 while moving the decimal point one to the right,
		603	// resulting in a net multiplication by 10.
		604	// we always write the digit to memory but conditionally increment
		605	// the pointer, to ease the usage of conditional move instructions.
		606	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		607	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		608	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		609	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		610	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		611	}
339	}	612	}
340	else if (SvROK (sv))	613	else if (SvROK (sv))
341	{	614	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))	615	else if (!SvOK (sv))
357	encode_str (enc, "null", 4, 0);	616	encode_str (enc, "null", 4, 0);
358	else	617	else
359	croak ("encountered perl type that JSON cannot handle");	618	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		619	SvPV_nolen (sv), SvFLAGS (sv));
360	}	620	}
361		621
362	static SV *	622	static SV *
363	encode_json (SV *scalar, UV flags)	623	encode_json (SV *scalar, struct json *json)
364	{	624	{
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;	625	enc_t enc;
369	enc.flags = flags;	626
		627	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		628	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		629
		630	enc.json = *json;
370	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	631	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
371	enc.cur = SvPVX (enc.sv);	632	enc.cur = SvPVX (enc.sv);
372	enc.end = SvEND (enc.sv);	633	enc.end = SvEND (enc.sv);
373	enc.max_recurse = 0;
374	enc.indent = 0;	634	enc.indent = 0;
		635	enc.maxdepth = DEC_DEPTH (enc.json.flags);
375		636
376	SvPOK_only (enc.sv);	637	SvPOK_only (enc.sv);
377	encode_sv (&enc, scalar);	638	encode_sv (&enc, scalar);
378		639
		640	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		641	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		642
379	if (!(flags & (F_ASCII | F_UTF8)))	643	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
380	SvUTF8_on (enc.sv);	644	SvUTF8_on (enc.sv);
381		645
382	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	646	if (enc.json.flags & F_SHRINK)
		647	shrink (enc.sv);
		648
383	return enc.sv;	649	return enc.sv;
384	}	650	}
385		651
386	/////////////////////////////////////////////////////////////////////////////	652	/////////////////////////////////////////////////////////////////////////////
		653	// decoder
387		654
388	#define WS \	655	// structure used for decoding JSON
		656	typedef struct
		657	{
		658	char *cur; // current parser pointer
		659	char *end; // end of input string
		660	const char *err; // parse error, if != 0
		661	struct json json;
		662	U32 depth; // recursion depth
		663	U32 maxdepth; // recursion depth limit
		664	} dec_t;
		665
		666	inline void
		667	decode_ws (dec_t *dec)
		668	{
389	for (;;) \	669	for (;;)
390	{ \	670	{
391	char ch = *dec->cur; \	671	char ch = *dec->cur;
		672
392	if (ch > 0x20 \	673	if (ch > 0x20
393	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	674	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
394	break; \	675	break;
		676
395	++dec->cur; \	677	++dec->cur;
396	}	678	}
		679	}
397		680
398	#define ERR(reason) SB dec->err = reason; goto fail; SE	681	#define ERR(reason) SB dec->err = reason; goto fail; SE
		682
399	#define EXPECT_CH(ch) SB \	683	#define EXPECT_CH(ch) SB \
400	if (*dec->cur != ch) \	684	if (*dec->cur != ch) \
401	ERR (# ch " expected"); \	685	ERR (# ch " expected"); \
402	++dec->cur; \	686	++dec->cur; \
403	SE	687	SE
404		688
		689	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		690	#define DEC_DEC_DEPTH --dec->depth
		691
405	static SV *decode_sv (dec_t *dec);	692	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		693
412	static signed char decode_hexdigit[256];	694	static signed char decode_hexdigit[256];
413		695
414	static UV	696	static UV
415	decode_4hex (dec_t *dec)	697	decode_4hex (dec_t *dec)
416	{	698	{
417	signed char d1, d2, d3, d4;	699	signed char d1, d2, d3, d4;
		700	unsigned char *cur = (unsigned char *)dec->cur;
418		701
419	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	702	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");	703	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]];	704	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");	705	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		706
428	dec->cur += 4;	707	dec->cur += 4;
429		708
430	return ((UV)d1) << 12	709	return ((UV)d1) << 12
431	| ((UV)d2) << 8	710	| ((UV)d2) << 8
…		…
437	}	716	}
438		717
439	static SV *	718	static SV *
440	decode_str (dec_t *dec)	719	decode_str (dec_t *dec)
441	{	720	{
442	SV *sv = NEWSV (0,2);	721	SV *sv = 0;
443	STRLEN cur = 0;
444	int utf8 = 0;	722	int utf8 = 0;
		723	char *dec_cur = dec->cur;
445		724
446	for (;;)	725	do
447	{	726	{
448	unsigned char ch = *(unsigned char *)dec->cur;	727	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		728	char *cur = buf;
449		729
450	if (ch == '"')	730	do
451	break;
452	else if (ch == '\\')
453	{	731	{
454	switch (*++dec->cur)	732	unsigned char ch = *(unsigned char *)dec_cur++;
		733
		734	if (expect_false (ch == '"'))
		735	{
		736	--dec_cur;
		737	break;
455	{	738	}
456	case '\\':	739	else if (expect_false (ch == '\\'))
457	case '/':	740	{
458	case '"': APPEND_CH (*dec->cur++); break;	741	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	{	742	{
468	UV lo, hi;	743	case '\\':
469	++dec->cur;	744	case '/':
		745	case '"': *cur++ = *dec_cur++; break;
470		746
471	hi = decode_4hex (dec);	747	case 'b': ++dec_cur; *cur++ = '\010'; break;
472	if (hi == (UV)-1)	748	case 't': ++dec_cur; *cur++ = '\011'; break;
473	goto fail;	749	case 'n': ++dec_cur; *cur++ = '\012'; break;
		750	case 'f': ++dec_cur; *cur++ = '\014'; break;
		751	case 'r': ++dec_cur; *cur++ = '\015'; break;
474		752
475	// possibly a surrogate pair	753	case 'u':
476	if (hi >= 0xd800 && hi < 0xdc00)
477	{	754	{
478	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	755	UV lo, hi;
479	ERR ("illegal surrogate character");	756	++dec_cur;
480		757
481	dec->cur += 2;	758	dec->cur = dec_cur;
482
483	lo = decode_4hex (dec);	759	hi = decode_4hex (dec);
		760	dec_cur = dec->cur;
484	if (lo == (UV)-1)	761	if (hi == (UV)-1)
485	goto fail;	762	goto fail;
486		763
		764	// possibly a surrogate pair
		765	if (hi >= 0xd800)
		766	if (hi < 0xdc00)
		767	{
		768	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		769	ERR ("missing low surrogate character in surrogate pair");
		770
		771	dec_cur += 2;
		772
		773	dec->cur = dec_cur;
		774	lo = decode_4hex (dec);
		775	dec_cur = dec->cur;
		776	if (lo == (UV)-1)
		777	goto fail;
		778
487	if (lo < 0xdc00 || lo >= 0xe000)	779	if (lo < 0xdc00 || lo >= 0xe000)
488	ERR ("surrogate pair expected");	780	ERR ("surrogate pair expected");
489		781
490	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	782	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		783	}
		784	else if (hi < 0xe000)
		785	ERR ("missing high surrogate character in surrogate pair");
		786
		787	if (hi >= 0x80)
		788	{
		789	utf8 = 1;
		790
		791	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		792	}
		793	else
		794	*cur++ = hi;
491	}	795	}
492	else if (lo >= 0xdc00 && lo < 0xe000)
493	ERR ("illegal surrogate character");
494
495	if (hi >= 0x80)
496	{	796	break;
497	utf8 = 1;
498		797
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	798	default:
503	APPEND_CH (hi);	799	--dec_cur;
		800	ERR ("illegal backslash escape sequence in string");
504	}	801	}
		802	}
		803	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		804	*cur++ = ch;
		805	else if (ch >= 0x80)
		806	{
		807	STRLEN clen;
		808	UV uch;
		809
		810	--dec_cur;
		811
		812	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		813	if (clen == (STRLEN)-1)
		814	ERR ("malformed UTF-8 character in JSON string");
		815
		816	do
		817	*cur++ = *dec_cur++;
		818	while (--clen);
		819
		820	utf8 = 1;
		821	}
		822	else
		823	{
		824	--dec_cur;
		825
		826	if (!ch)
		827	ERR ("unexpected end of string while parsing JSON string");
505	break;	828	else
		829	ERR ("invalid character encountered while parsing JSON string");
506	}	830	}
507	}	831	}
508	else if (ch >= 0x20 && ch <= 0x7f)	832	while (cur < buf + SHORT_STRING_LEN);
509	APPEND_CH (*dec->cur++);	833
510	else if (ch >= 0x80)	834	{
		835	STRLEN len = cur - buf;
		836
		837	if (sv)
511	{	838	{
512	STRLEN clen;	839	SvGROW (sv, SvCUR (sv) + len + 1);
513	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	840	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
514	if (clen < 0)	841	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	}	842	}
521	else	843	else
522	ERR ("invalid character encountered");	844	sv = newSVpvn (buf, len);
523	}	845	}
		846	}
		847	while (*dec_cur != '"');
524		848
525	++dec->cur;	849	++dec_cur;
526		850
		851	if (sv)
		852	{
527	SvPOK_only (sv);	853	SvPOK_only (sv);
528
529	SvCUR_set (sv, cur);
530	*SvEND (sv) = 0;	854	*SvEND (sv) = 0;
531		855
532	if (utf8)	856	if (utf8)
533	SvUTF8_on (sv);	857	SvUTF8_on (sv);
		858	}
		859	else
		860	sv = newSVpvn ("", 0);
534		861
		862	dec->cur = dec_cur;
535	return sv;	863	return sv;
536		864
537	fail:	865	fail:
538	SvREFCNT_dec (sv);	866	dec->cur = dec_cur;
539	return 0;	867	return 0;
540	}	868	}
541		869
542	static SV *	870	static SV *
543	decode_num (dec_t *dec)	871	decode_num (dec_t *dec)
…		…
553	{	881	{
554	++dec->cur;	882	++dec->cur;
555	if (*dec->cur >= '0' && *dec->cur <= '9')	883	if (*dec->cur >= '0' && *dec->cur <= '9')
556	ERR ("malformed number (leading zero must not be followed by another digit)");	884	ERR ("malformed number (leading zero must not be followed by another digit)");
557	}	885	}
558		886	else if (*dec->cur < '0' || *dec->cur > '9')
559	// int	887	ERR ("malformed number (no digits after initial minus)");
		888	else
		889	do
		890	{
		891	++dec->cur;
		892	}
560	while (*dec->cur >= '0' && *dec->cur <= '9')	893	while (*dec->cur >= '0' && *dec->cur <= '9');
561	++dec->cur;
562		894
563	// [frac]	895	// [frac]
564	if (*dec->cur == '.')	896	if (*dec->cur == '.')
565	{	897	{
566	is_nv = 1;	898	++dec->cur;
		899
		900	if (*dec->cur < '0' || *dec->cur > '9')
		901	ERR ("malformed number (no digits after decimal point)");
567		902
568	do	903	do
569	{	904	{
570	++dec->cur;	905	++dec->cur;
571	}	906	}
572	while (*dec->cur >= '0' && *dec->cur <= '9');	907	while (*dec->cur >= '0' && *dec->cur <= '9');
		908
		909	is_nv = 1;
573	}	910	}
574		911
575	// [exp]	912	// [exp]
576	if (*dec->cur == 'e' || *dec->cur == 'E')	913	if (*dec->cur == 'e' || *dec->cur == 'E')
577	{	914	{
578	is_nv = 1;
579
580	++dec->cur;	915	++dec->cur;
		916
581	if (*dec->cur == '-' || *dec->cur == '+')	917	if (*dec->cur == '-' || *dec->cur == '+')
582	++dec->cur;	918	++dec->cur;
583		919
		920	if (*dec->cur < '0' || *dec->cur > '9')
		921	ERR ("malformed number (no digits after exp sign)");
		922
		923	do
		924	{
		925	++dec->cur;
		926	}
584	while (*dec->cur >= '0' && *dec->cur <= '9')	927	while (*dec->cur >= '0' && *dec->cur <= '9');
585	++dec->cur;	928
		929	is_nv = 1;
586	}	930	}
587		931
588	if (!is_nv)	932	if (!is_nv)
589	{	933	{
590	UV uv;	934	// 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);	935	if (*start == '-')
592	if (numtype & IS_NUMBER_IN_UV)	936	switch (dec->cur - start)
593	if (numtype & IS_NUMBER_NEG)
594	{	937	{
595	if (uv < (UV)IV_MIN)	938	case 2: return newSViv (-( start [1] - '0' * 1));
596	return newSViv (-(IV)uv);	939	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		940	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		941	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
597	}	942	}
		943	else
		944	switch (dec->cur - start)
		945	{
		946	case 1: return newSViv ( start [0] - '0' * 1);
		947	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		948	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		949	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		950	}
		951
		952	{
		953	UV uv;
		954	int numtype = grok_number (start, dec->cur - start, &uv);
		955	if (numtype & IS_NUMBER_IN_UV)
		956	if (numtype & IS_NUMBER_NEG)
		957	{
		958	if (uv < (UV)IV_MIN)
		959	return newSViv (-(IV)uv);
		960	}
598	else	961	else
599	return newSVuv (uv);	962	return newSVuv (uv);
		963
		964	// here would likely be the place for bigint support
600	}	965	}
		966	}
601		967
		968	// if we ever support bigint or bigfloat, this is the place for bigfloat
602	return newSVnv (Atof (start));	969	return newSVnv (Atof (start));
603		970
604	fail:	971	fail:
605	return 0;	972	return 0;
606	}	973	}
…		…
608	static SV *	975	static SV *
609	decode_av (dec_t *dec)	976	decode_av (dec_t *dec)
610	{	977	{
611	AV *av = newAV ();	978	AV *av = newAV ();
612		979
		980	DEC_INC_DEPTH;
		981	decode_ws (dec);
		982
		983	if (*dec->cur == ']')
		984	++dec->cur;
		985	else
613	for (;;)	986	for (;;)
614	{	987	{
615	SV *value;	988	SV *value;
616		989
617	value = decode_sv (dec);	990	value = decode_sv (dec);
618	if (!value)	991	if (!value)
619	goto fail;	992	goto fail;
620		993
621	av_push (av, value);	994	av_push (av, value);
622		995
623	WS;	996	decode_ws (dec);
624		997
625	if (*dec->cur == ']')	998	if (*dec->cur == ']')
626	{	999	{
627	++dec->cur;	1000	++dec->cur;
628	break;	1001	break;
		1002	}
629	}	1003
630
631	if (*dec->cur != ',')	1004	if (*dec->cur != ',')
632	ERR (", or ] expected while parsing array");	1005	ERR (", or ] expected while parsing array");
633		1006
634	++dec->cur;	1007	++dec->cur;
635	}	1008	}
636		1009
		1010	DEC_DEC_DEPTH;
637	return newRV_noinc ((SV *)av);	1011	return newRV_noinc ((SV *)av);
638		1012
639	fail:	1013	fail:
640	SvREFCNT_dec (av);	1014	SvREFCNT_dec (av);
		1015	DEC_DEC_DEPTH;
641	return 0;	1016	return 0;
642	}	1017	}
643		1018
644	static SV *	1019	static SV *
645	decode_hv (dec_t *dec)	1020	decode_hv (dec_t *dec)
646	{	1021	{
647	HV *hv = newHV ();	1022	HV *hv = newHV ();
648		1023
		1024	DEC_INC_DEPTH;
		1025	decode_ws (dec);
		1026
		1027	if (*dec->cur == '}')
		1028	++dec->cur;
		1029	else
649	for (;;)	1030	for (;;)
650	{	1031	{
651	SV *key, *value;	1032	decode_ws (dec); EXPECT_CH ('"');
652		1033
653	WS; EXPECT_CH ('"');	1034	// heuristic: assume that
654		1035	// a) decode_str + hv_store_ent are abysmally slow.
655	key = decode_str (dec);	1036	// b) most hash keys are short, simple ascii text.
656	if (!key)	1037	// => try to "fast-match" such strings to avoid
657	goto fail;	1038	// the overhead of decode_str + hv_store_ent.
658
659	WS; EXPECT_CH (':');
660
661	value = decode_sv (dec);
662	if (!value)
663	{	1039	{
		1040	SV *value;
		1041	char *p = dec->cur;
		1042	char *e = p + 24; // only try up to 24 bytes
		1043
		1044	for (;;)
		1045	{
		1046	// the >= 0x80 is true on most architectures
		1047	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1048	{
		1049	// slow path, back up and use decode_str
		1050	SV *key = decode_str (dec);
		1051	if (!key)
		1052	goto fail;
		1053
		1054	decode_ws (dec); EXPECT_CH (':');
		1055
		1056	value = decode_sv (dec);
		1057	if (!value)
		1058	{
		1059	SvREFCNT_dec (key);
		1060	goto fail;
		1061	}
		1062
		1063	hv_store_ent (hv, key, value, 0);
664	SvREFCNT_dec (key);	1064	SvREFCNT_dec (key);
		1065
		1066	break;
		1067	}
		1068	else if (*p == '"')
		1069	{
		1070	// fast path, got a simple key
		1071	char *key = dec->cur;
		1072	int len = p - key;
		1073	dec->cur = p + 1;
		1074
		1075	decode_ws (dec); EXPECT_CH (':');
		1076
		1077	value = decode_sv (dec);
		1078	if (!value)
665	goto fail;	1079	goto fail;
		1080
		1081	hv_store (hv, key, len, value, 0);
		1082
		1083	break;
		1084	}
		1085
		1086	++p;
		1087	}
666	}	1088	}
667		1089
668	//TODO: optimise	1090	decode_ws (dec);
669	hv_store_ent (hv, key, value, 0);
670		1091
671	WS;
672
673	if (*dec->cur == '}')	1092	if (*dec->cur == '}')
674	{	1093	{
675	++dec->cur;	1094	++dec->cur;
676	break;	1095	break;
677	}	1096	}
678		1097
679	if (*dec->cur != ',')	1098	if (*dec->cur != ',')
680	ERR (", or } expected while parsing object/hash");	1099	ERR (", or } expected while parsing object/hash");
681		1100
682	++dec->cur;	1101	++dec->cur;
683	}	1102	}
684		1103
		1104	DEC_DEC_DEPTH;
685	return newRV_noinc ((SV *)hv);	1105	return newRV_noinc ((SV *)hv);
686		1106
687	fail:	1107	fail:
688	SvREFCNT_dec (hv);	1108	SvREFCNT_dec (hv);
		1109	DEC_DEC_DEPTH;
689	return 0;	1110	return 0;
690	}	1111	}
691		1112
692	static SV *	1113	static SV *
693	decode_sv (dec_t *dec)	1114	decode_sv (dec_t *dec)
694	{	1115	{
695	WS;	1116	decode_ws (dec);
		1117
		1118	// the beauty of JSON: you need exactly one character lookahead
		1119	// to parse anything.
696	switch (*dec->cur)	1120	switch (*dec->cur)
697	{	1121	{
698	case '"': ++dec->cur; return decode_str (dec);	1122	case '"': ++dec->cur; return decode_str (dec);
699	case '[': ++dec->cur; return decode_av (dec);	1123	case '[': ++dec->cur; return decode_av (dec);
700	case '{': ++dec->cur; return decode_hv (dec);	1124	case '{': ++dec->cur; return decode_hv (dec);
…		…
706		1130
707	case 't':	1131	case 't':
708	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1132	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
709	{	1133	{
710	dec->cur += 4;	1134	dec->cur += 4;
711	return newSViv (1);	1135	return SvREFCNT_inc (json_true);
712	}	1136	}
713	else	1137	else
714	ERR ("'true' expected");	1138	ERR ("'true' expected");
715		1139
716	break;	1140	break;
717		1141
718	case 'f':	1142	case 'f':
719	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1143	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
720	{	1144	{
721	dec->cur += 5;	1145	dec->cur += 5;
722	return newSViv (0);	1146	return SvREFCNT_inc (json_false);
723	}	1147	}
724	else	1148	else
725	ERR ("'false' expected");	1149	ERR ("'false' expected");
726		1150
727	break;	1151	break;
728		1152
729	case 'n':	1153	case 'n':
730	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1154	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
731	{	1155	{
732	dec->cur += 4;	1156	dec->cur += 4;
733	return newSViv (1);	1157	return newSVsv (&PL_sv_undef);
734	}	1158	}
735	else	1159	else
736	ERR ("'null' expected");	1160	ERR ("'null' expected");
737		1161
738	break;	1162	break;
739		1163
740	default:	1164	default:
741	ERR ("malformed json string");	1165	ERR ("malformed JSON string, neither array, object, number, string or atom");
742	break;	1166	break;
743	}	1167	}
744		1168
745	fail:	1169	fail:
746	return 0;	1170	return 0;
747	}	1171	}
748		1172
749	static SV *	1173	static SV *
750	decode_json (SV *string, UV flags)	1174	decode_json (SV *string, struct json *json, UV *offset_return)
751	{	1175	{
		1176	dec_t dec;
		1177	UV offset;
752	SV *sv;	1178	SV *sv;
753		1179
		1180	SvGETMAGIC (string);
		1181	SvUPGRADE (string, SVt_PV);
		1182
		1183	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1184	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1185	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1186
754	if (!(flags & F_UTF8))	1187	if (json->flags & F_UTF8)
		1188	sv_utf8_downgrade (string, 0);
		1189	else
755	sv_utf8_upgrade (string);	1190	sv_utf8_upgrade (string);
756		1191
757	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1192	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
758		1193
759	dec_t dec;	1194	dec.json = *json;
760	dec.flags = flags;
761	dec.cur = SvPVX (string);	1195	dec.cur = SvPVX (string);
762	dec.end = SvEND (string);	1196	dec.end = SvEND (string);
763	dec.err = 0;	1197	dec.err = 0;
		1198	dec.depth = 0;
		1199	dec.maxdepth = DEC_DEPTH (dec.json.flags);
764		1200
765	*dec.end = 1; // invalid anywhere	1201	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
766	sv = decode_sv (&dec);	1202	sv = decode_sv (&dec);
767	*dec.end = 0;	1203
		1204	if (!(offset_return || !sv))
		1205	{
		1206	// check for trailing garbage
		1207	decode_ws (&dec);
		1208
		1209	if (*dec.cur)
		1210	{
		1211	dec.err = "garbage after JSON object";
		1212	SvREFCNT_dec (sv);
		1213	sv = 0;
		1214	}
		1215	}
		1216
		1217	if (offset_return || !sv)
		1218	{
		1219	offset = dec.json.flags & F_UTF8
		1220	? dec.cur - SvPVX (string)
		1221	: utf8_distance (dec.cur, SvPVX (string));
		1222
		1223	if (offset_return)
		1224	*offset_return = offset;
		1225	}
768		1226
769	if (!sv)	1227	if (!sv)
770	{	1228	{
771	IV offset = utf8_distance (dec.cur, SvPVX (string));
772	SV *uni = sv_newmortal ();	1229	SV *uni = sv_newmortal ();
773		1230
		1231	// horrible hack to silence warning inside pv_uni_display
		1232	COP cop = *PL_curcop;
		1233	cop.cop_warnings = pWARN_NONE;
		1234	ENTER;
		1235	SAVEVPTR (PL_curcop);
		1236	PL_curcop = &cop;
774	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1237	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1238	LEAVE;
		1239
775	croak ("%s, at character %d (%s)",	1240	croak ("%s, at character offset %d [\"%s\"]",
776	dec.err,	1241	dec.err,
777	(int)offset,	1242	(int)offset,
778	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1243	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
779	}	1244	}
780		1245
781	sv = sv_2mortal (sv);	1246	sv = sv_2mortal (sv);
782		1247
783	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1248	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)");	1249	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
785		1250
786	return sv;	1251	return sv;
787	}	1252	}
788		1253
		1254	/////////////////////////////////////////////////////////////////////////////
		1255	// XS interface functions
		1256
789	MODULE = JSON::XS PACKAGE = JSON::XS	1257	MODULE = JSON::XS PACKAGE = JSON::XS
790		1258
791	BOOT:	1259	BOOT:
792	{	1260	{
793	int i;	1261	int i;
794		1262
795	memset (decode_hexdigit, 0xff, 256);
796	for (i = 10; i--; )	1263	for (i = 0; i < 256; ++i)
797	decode_hexdigit ['0' + i] = i;	1264	decode_hexdigit [i] =
		1265	i >= '0' && i <= '9' ? i - '0'
		1266	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1267	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1268	: -1;
798		1269
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);	1270	json_stash = gv_stashpv ("JSON::XS" , 1);
		1271	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1272
		1273	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1274	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
806	}	1275	}
		1276
		1277	PROTOTYPES: DISABLE
807		1278
808	SV *new (char *dummy)	1279	SV *new (char *dummy)
809	CODE:	1280	CODE:
810	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1281	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
811	OUTPUT:	1282	OUTPUT:
812	RETVAL	1283	RETVAL
813		1284
814	SV *ascii (SV *self, int enable)	1285	SV *ascii (SV *self, int enable = 1)
815	ALIAS:	1286	ALIAS:
816	ascii = F_ASCII	1287	ascii = F_ASCII
		1288	latin1 = F_LATIN1
817	utf8 = F_UTF8	1289	utf8 = F_UTF8
818	indent = F_INDENT	1290	indent = F_INDENT
819	canonical = F_CANONICAL	1291	canonical = F_CANONICAL
820	space_before = F_SPACE_BEFORE	1292	space_before = F_SPACE_BEFORE
821	space_after = F_SPACE_AFTER	1293	space_after = F_SPACE_AFTER
822	json_rpc = F_JSON_RPC
823	pretty = F_PRETTY	1294	pretty = F_PRETTY
824	allow_nonref = F_ALLOW_NONREF	1295	allow_nonref = F_ALLOW_NONREF
		1296	shrink = F_SHRINK
		1297	allow_blessed = F_ALLOW_BLESSED
		1298	convert_blessed = F_CONV_BLESSED
825	CODE:	1299	CODE:
826	{	1300	{
827	UV *uv = SvJSON (self);	1301	UV *uv = SvJSON (self);
828	if (enable)	1302	if (enable)
829	*uv |= ix;	1303	*uv |= ix;
…		…
833	RETVAL = newSVsv (self);	1307	RETVAL = newSVsv (self);
834	}	1308	}
835	OUTPUT:	1309	OUTPUT:
836	RETVAL	1310	RETVAL
837		1311
		1312	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1313	CODE:
		1314	{
		1315	UV *uv = SvJSON (self);
		1316	UV log2 = 0;
		1317
		1318	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1319
		1320	while ((1UL << log2) < max_depth)
		1321	++log2;
		1322
		1323	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1324
		1325	RETVAL = newSVsv (self);
		1326	}
		1327	OUTPUT:
		1328	RETVAL
		1329
		1330	SV *max_size (SV *self, UV max_size = 0)
		1331	CODE:
		1332	{
		1333	UV *uv = SvJSON (self);
		1334	UV log2 = 0;
		1335
		1336	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1337	if (max_size == 1) max_size = 2;
		1338
		1339	while ((1UL << log2) < max_size)
		1340	++log2;
		1341
		1342	*uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1343
		1344	RETVAL = newSVsv (self);
		1345	}
		1346	OUTPUT:
		1347	RETVAL
		1348
838	void encode (SV *self, SV *scalar)	1349	void encode (SV *self, SV *scalar)
839	PPCODE:	1350	PPCODE:
		1351	{
		1352	struct json json = { *SvJSON (self) };
840	XPUSHs (encode_json (scalar, *SvJSON (self)));	1353	XPUSHs (encode_json (scalar, &json));
		1354	}
841		1355
842	void decode (SV *self, SV *jsonstr)	1356	void decode (SV *self, SV *jsonstr)
843	PPCODE:	1357	PPCODE:
		1358	{
		1359	struct json json = { *SvJSON (self) };
844	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1360	XPUSHs (decode_json (jsonstr, &json, 0));
		1361	}
		1362
		1363	void decode_prefix (SV *self, SV *jsonstr)
		1364	PPCODE:
		1365	{
		1366	UV offset;
		1367	struct json json = { *SvJSON (self) };
		1368	EXTEND (SP, 2);
		1369	PUSHs (decode_json (jsonstr, &json, &offset));
		1370	PUSHs (sv_2mortal (newSVuv (offset)));
		1371	}
		1372
		1373	PROTOTYPES: ENABLE
845		1374
846	void to_json (SV *scalar)	1375	void to_json (SV *scalar)
847	PPCODE:	1376	PPCODE:
		1377	{
		1378	struct json json = { F_DEFAULT | F_UTF8 };
848	XPUSHs (encode_json (scalar, F_UTF8));	1379	XPUSHs (encode_json (scalar, &json));
		1380	}
849		1381
850	void from_json (SV *jsonstr)	1382	void from_json (SV *jsonstr)
851	PPCODE:	1383	PPCODE:
		1384	{
		1385	struct json json = { F_DEFAULT | F_UTF8 };
852	XPUSHs (decode_json (jsonstr, F_UTF8));	1386	XPUSHs (decode_json (jsonstr, &json, 0));
		1387	}
853		1388

Diff Legend

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