ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.1 by root, Thu Mar 22 16:40:16 2007 UTC vs.
Revision 1.51 by root, Mon Jul 2 00:48:18 2007 UTC

…		…
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include "assert.h"
6	#include "string.h"	6	#include "string.h"
7	#include "stdlib.h"	7	#include "stdlib.h"
		8	#include "stdio.h"
8		9
		10	#if defined(__BORLANDC__) || defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040	27	#define F_ALLOW_NONREF 0x00000100UL
		28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL
		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
16		36
17	#define F_DEFAULT 0	37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
		39
		40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
		41	#define F_DEFAULT (9UL << S_MAXDEPTH)
18		42
19	#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
20		47
21	#define SB do {	48	#define SB do {
22	#define SE } while (0)	49	#define SE } while (0)
23		50
24	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, *cb_sk_object;
		68	} JSON;
		69
		70	/////////////////////////////////////////////////////////////////////////////
		71	// utility functions
		72
		73	inline void
		74	shrink (SV *sv)
		75	{
		76	sv_utf8_downgrade (sv, 1);
		77	if (SvLEN (sv) > SvCUR (sv) + 1)
		78	{
		79	#ifdef SvPV_shrink_to_cur
		80	SvPV_shrink_to_cur (sv);
		81	#elif defined (SvPV_renew)
		82	SvPV_renew (sv, SvCUR (sv) + 1);
		83	#endif
		84	}
		85	}
		86
		87	// decode an utf-8 character and return it, or (UV)-1 in
		88	// case of an error.
		89	// we special-case "safe" characters from U+80 .. U+7FF,
		90	// but use the very good perl function to parse anything else.
		91	// note that we never call this function for a ascii codepoints
		92	inline UV
		93	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		94	{
		95	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		98	{
		99	*clen = 2;
		100	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		101	}
		102	else
		103	{
		104	*clen = (STRLEN)-1;
		105	return (UV)-1;
		106	}
		107	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// encoder
25		111
26	// structure used for encoding JSON	112	// structure used for encoding JSON
27	typedef struct	113	typedef struct
28	{	114	{
29	char *cur;	115	char *cur; // SvPVX (sv) + current output position
30	STRLEN len; // SvLEN (sv)
31	char *end; // SvEND (sv)	116	char *end; // SvEND (sv)
32	SV *sv;	117	SV *sv; // result scalar
33	UV flags;	118	JSON json;
34	int max_recurse;	119	U32 indent; // indentation level
35	int indent;	120	U32 maxdepth; // max. indentation/recursion level
36	} enc_t;	121	} enc_t;
37		122
38	// structure used for decoding JSON	123	inline void
39	typedef struct
40	{
41	char *cur;
42	char *end;
43	char *err;
44	UV flags;
45	} dec_t;
46
47	static UV *
48	SvJSON (SV *sv)
49	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
51	croak ("object is not of type JSON::XS");
52
53	return &SvUVX (SvRV (sv));
54	}
55
56	/////////////////////////////////////////////////////////////////////////////
57
58	static void
59	need (enc_t *enc, STRLEN len)	124	need (enc_t *enc, STRLEN len)
60	{	125	{
61	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
62	{	127	{
63	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
64	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
65	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
66	enc->end = SvEND (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
67	}	132	}
68	}	133	}
69		134
70	static void	135	inline void
71	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
72	{	137	{
73	need (enc, 1);	138	need (enc, 1);
74	*enc->cur++ = ch;	139	*enc->cur++ = ch;
75	}	140	}
…		…
77	static void	142	static void
78	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	143	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
79	{	144	{
80	char *end = str + len;	145	char *end = str + len;
81		146
		147	need (enc, len);
		148
82	while (str < end)	149	while (str < end)
83	{	150	{
84	unsigned char ch = *(unsigned char *)str;	151	unsigned char ch = *(unsigned char *)str;
		152
85	if (ch >= 0x20 && ch < 0x80) // most common case	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
86	{	154	{
		155	if (expect_false (ch == '"')) // but with slow exceptions
		156	{
		157	need (enc, len += 1);
		158	*enc->cur++ = '\\';
		159	*enc->cur++ = '"';
		160	}
		161	else if (expect_false (ch == '\\'))
		162	{
		163	need (enc, len += 1);
		164	*enc->cur++ = '\\';
		165	*enc->cur++ = '\\';
		166	}
		167	else
87	*enc->cur++ = ch;	168	*enc->cur++ = ch;
		169
88	str++;	170	++str;
89	}	171	}
90	else	172	else
91	{	173	{
92	STRLEN clen;	174	switch (ch)
93	UV uch;
94
95	if (is_utf8)
96	{
97	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
98	if (clen < 0)
99	croak ("malformed UTF-8 character in string, cannot convert to JSON");
100	}	175	{
101	else	176	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
102	{	177	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
103	uch = ch;	178	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
104	clen = 1;	179	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
105	}	180	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
106		181
107	need (enc, len += 6);	182	default:
108
109	if (uch < 0xa0 || enc->flags & F_ASCII)
110	{
111	if (uch > 0xFFFFUL)
112	{	183	{
		184	STRLEN clen;
		185	UV uch;
		186
		187	if (is_utf8)
		188	{
		189	uch = decode_utf8 (str, end - str, &clen);
		190	if (clen == (STRLEN)-1)
		191	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		192	}
		193	else
		194	{
		195	uch = ch;
113	len += 6;	196	clen = 1;
		197	}
		198
		199	if (uch > 0x10FFFFUL)
		200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		201
		202	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		203	{
		204	if (uch > 0xFFFFUL)
		205	{
114	need (enc, len += 6);	206	need (enc, len += 11);
115	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
116	(uch - 0x10000) / 0x400 + 0xD800,	208	(int)((uch - 0x10000) / 0x400 + 0xD800),
117	(uch - 0x10000) % 0x400 + 0xDC00);	209	(int)((uch - 0x10000) % 0x400 + 0xDC00));
118	enc->cur += 12;	210	enc->cur += 12;
		211	}
		212	else
		213	{
		214	static char hexdigit [16] = "0123456789abcdef";
		215	need (enc, len += 5);
		216	*enc->cur++ = '\\';
		217	*enc->cur++ = 'u';
		218	*enc->cur++ = hexdigit [ uch >> 12 ];
		219	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		220	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		221	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		222	}
		223
		224	str += clen;
		225	}
		226	else if (enc->json.flags & F_LATIN1)
		227	{
		228	*enc->cur++ = uch;
		229	str += clen;
		230	}
		231	else if (is_utf8)
		232	{
		233	need (enc, len += clen);
		234	do
		235	{
		236	*enc->cur++ = *str++;
		237	}
		238	while (--clen);
		239	}
		240	else
		241	{
		242	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		243	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		244	++str;
		245	}
119	}	246	}
120	else
121	{
122	sprintf (enc->cur, "\\u%04x", uch);
123	enc->cur += 6;
124	}
125	}
126	else if (is_utf8)
127	{	247	}
128	memcpy (enc->cur, str, clen);
129	enc->cur += clen;
130	}
131	else
132	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
133
134	str += clen;
135	}	248	}
136		249
137	--len;	250	--len;
138	}	251	}
139	}	252	}
140		253
141	#define INDENT SB \	254	inline void
		255	encode_indent (enc_t *enc)
		256	{
142	if (enc->flags & F_INDENT) \	257	if (enc->json.flags & F_INDENT)
143	{ \	258	{
144	int i_; \	259	int spaces = enc->indent * INDENT_STEP;
145	need (enc, enc->indent); \	260
146	for (i_ = enc->indent * 3; i_--; )\	261	need (enc, spaces);
		262	memset (enc->cur, ' ', spaces);
		263	enc->cur += spaces;
		264	}
		265	}
		266
		267	inline void
		268	encode_space (enc_t *enc)
		269	{
		270	need (enc, 1);
		271	encode_ch (enc, ' ');
		272	}
		273
		274	inline void
		275	encode_nl (enc_t *enc)
		276	{
		277	if (enc->json.flags & F_INDENT)
		278	{
		279	need (enc, 1);
147	encode_ch (enc, ' '); \	280	encode_ch (enc, '\n');
148	} \	281	}
149	SE	282	}
150		283
151	#define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE	284	inline void
152	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	285	encode_comma (enc_t *enc)
153	#define COMMA SB \	286	{
154	encode_ch (enc, ','); \	287	encode_ch (enc, ',');
		288
155	if (enc->flags & F_INDENT) \	289	if (enc->json.flags & F_INDENT)
156	NL; \	290	encode_nl (enc);
157	else if (enc->flags & F_SPACE_AFTER) \	291	else if (enc->json.flags & F_SPACE_AFTER)
158	SPACE; \	292	encode_space (enc);
159	SE	293	}
160		294
161	static void encode_sv (enc_t *enc, SV *sv);	295	static void encode_sv (enc_t *enc, SV *sv);
162		296
163	static void	297	static void
164	encode_av (enc_t *enc, AV *av)	298	encode_av (enc_t *enc, AV *av)
165	{	299	{
166	int i, len = av_len (av);	300	int i, len = av_len (av);
167		301
168	encode_ch (enc, '['); NL;	302	if (enc->indent >= enc->maxdepth)
		303	croak ("data structure too deep (hit recursion limit)");
		304
		305	encode_ch (enc, '['); encode_nl (enc);
169	++enc->indent;	306	++enc->indent;
170		307
171	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
172	{	309	{
173	INDENT;	310	encode_indent (enc);
174	encode_sv (enc, *av_fetch (av, i, 0));	311	encode_sv (enc, *av_fetch (av, i, 0));
175		312
176	if (i < len)	313	if (i < len)
177	COMMA;	314	encode_comma (enc);
178	}	315	}
179		316
180	NL;	317	encode_nl (enc);
181		318
182	--enc->indent;	319	--enc->indent;
183	INDENT; encode_ch (enc, ']');	320	encode_indent (enc); encode_ch (enc, ']');
184	}	321	}
185		322
186	static void	323	static void
187	encode_he (enc_t *enc, HE *he)	324	encode_he (enc_t *enc, HE *he)
188	{	325	{
…		…
190		327
191	if (HeKLEN (he) == HEf_SVKEY)	328	if (HeKLEN (he) == HEf_SVKEY)
192	{	329	{
193	SV *sv = HeSVKEY (he);	330	SV *sv = HeSVKEY (he);
194	STRLEN len;	331	STRLEN len;
		332	char *str;
		333
		334	SvGETMAGIC (sv);
195	char *str = SvPV (sv, len);	335	str = SvPV (sv, len);
196		336
197	encode_str (enc, str, len, SvUTF8 (sv));	337	encode_str (enc, str, len, SvUTF8 (sv));
198	}	338	}
199	else	339	else
200	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	340	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
201		341
202	encode_ch (enc, '"');	342	encode_ch (enc, '"');
203		343
204	if (enc->flags & F_SPACE_BEFORE) SPACE;	344	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
205	encode_ch (enc, ':');	345	encode_ch (enc, ':');
206	if (enc->flags & F_SPACE_AFTER ) SPACE;	346	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
207	encode_sv (enc, HeVAL (he));	347	encode_sv (enc, HeVAL (he));
208	}	348	}
209		349
210	// compare hash entries, used when all keys are bytestrings	350	// compare hash entries, used when all keys are bytestrings
211	static int	351	static int
…		…
217	HE *b = *(HE **)b_;	357	HE *b = *(HE **)b_;
218		358
219	STRLEN la = HeKLEN (a);	359	STRLEN la = HeKLEN (a);
220	STRLEN lb = HeKLEN (b);	360	STRLEN lb = HeKLEN (b);
221		361
222	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
223	cmp = la < lb ? -1 : la == lb ? 0 : 1;	363	cmp = la - lb;
224		364
225	return cmp;	365	return cmp;
226	}	366	}
227		367
228	// compare hash entries, used when some keys are sv's or utf-x	368	// compare hash entries, used when some keys are sv's or utf-x
…		…
235	static void	375	static void
236	encode_hv (enc_t *enc, HV *hv)	376	encode_hv (enc_t *enc, HV *hv)
237	{	377	{
238	int count, i;	378	int count, i;
239		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
240	encode_ch (enc, '{'); NL; ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
241		384
242	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
243	{	386	{
244	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
245	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
246	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
247	// that randomises hash orderings	390	// that randomises hash orderings
248	if (enc->flags & F_CANONICAL)	391	if (enc->json.flags & F_CANONICAL)
249	{	392	{
250	HE *he, *hes [count];
251	int fast = 1;	393	int fast = 1;
		394	HE *he;
		395	#if defined(__BORLANDC__) || defined(_MSC_VER)
		396	HE **hes = _alloca (count * sizeof (HE));
		397	#else
		398	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		399	#endif
252		400
253	i = 0;	401	i = 0;
254	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
255	{	403	{
256	hes [i++] = he;	404	hes [i++] = he;
…		…
262		410
263	if (fast)	411	if (fast)
264	qsort (hes, count, sizeof (HE *), he_cmp_fast);	412	qsort (hes, count, sizeof (HE *), he_cmp_fast);
265	else	413	else
266	{	414	{
267	// hack to disable "use bytes"	415	// hack to forcefully disable "use bytes"
268	COP *oldcop = PL_curcop, cop;	416	COP cop = *PL_curcop;
269	cop.op_private = 0;	417	cop.op_private = 0;
		418
		419	ENTER;
		420	SAVETMPS;
		421
		422	SAVEVPTR (PL_curcop);
270	PL_curcop = &cop;	423	PL_curcop = &cop;
271		424
272	SAVETMPS;
273	qsort (hes, count, sizeof (HE *), he_cmp_slow);	425	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		426
274	FREETMPS;	427	FREETMPS;
275		428	LEAVE;
276	PL_curcop = oldcop;
277	}	429	}
278		430
279	for (i = 0; i < count; ++i)	431	for (i = 0; i < count; ++i)
280	{	432	{
281	INDENT;	433	encode_indent (enc);
282	encode_he (enc, hes [i]);	434	encode_he (enc, hes [i]);
283		435
284	if (i < count - 1)	436	if (i < count - 1)
285	COMMA;	437	encode_comma (enc);
286	}
287
288	NL;	438	}
		439
		440	encode_nl (enc);
289	}	441	}
290	else	442	else
291	{	443	{
292	SV *sv;
293	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
294		445
295	for (;;)	446	for (;;)
296	{	447	{
297	INDENT;	448	encode_indent (enc);
298	encode_he (enc, he);	449	encode_he (enc, he);
299		450
300	if (!(he = hv_iternext (hv)))	451	if (!(he = hv_iternext (hv)))
301	break;	452	break;
302		453
303	COMMA;	454	encode_comma (enc);
304	}
305
306	NL;	455	}
		456
		457	encode_nl (enc);
307	}	458	}
308	}	459	}
309		460
310	--enc->indent; INDENT; encode_ch (enc, '}');	461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		462	}
		463
		464	// encode objects, arrays and special \0=false and \1=true values.
		465	static void
		466	encode_rv (enc_t *enc, SV *sv)
		467	{
		468	svtype svt;
		469
		470	SvGETMAGIC (sv);
		471	svt = SvTYPE (sv);
		472
		473	if (expect_false (SvOBJECT (sv)))
		474	{
		475	if (SvSTASH (sv) == json_boolean_stash)
		476	{
		477	if (SvIV (sv))
		478	encode_str (enc, "true", 4, 0);
		479	else
		480	encode_str (enc, "false", 5, 0);
		481	}
		482	else
		483	{
		484	#if 0
		485	if (0 && sv_derived_from (rv, "JSON::Literal"))
		486	{
		487	// not yet
		488	}
		489	#endif
		490	if (enc->json.flags & F_CONV_BLESSED)
		491	{
		492	// we re-bless the reference to get overload and other niceties right
		493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		494
		495	if (to_json)
		496	{
		497	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
		498	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		499
		500	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		501	// check anyways.
		502	PUTBACK;
		503	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		504	SPAGAIN;
		505
		506	encode_sv (enc, POPs);
		507
		508	FREETMPS; LEAVE;
		509	}
		510	else if (enc->json.flags & F_ALLOW_BLESSED)
		511	encode_str (enc, "null", 4, 0);
		512	else
		513	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		514	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		515	}
		516	else if (enc->json.flags & F_ALLOW_BLESSED)
		517	encode_str (enc, "null", 4, 0);
		518	else
		519	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		520	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		521	}
		522	}
		523	else if (svt == SVt_PVHV)
		524	encode_hv (enc, (HV *)sv);
		525	else if (svt == SVt_PVAV)
		526	encode_av (enc, (AV *)sv);
		527	else if (svt < SVt_PVAV)
		528	{
		529	STRLEN len = 0;
		530	char *pv = svt ? SvPV (sv, len) : 0;
		531
		532	if (len == 1 && *pv == '1')
		533	encode_str (enc, "true", 4, 0);
		534	else if (len == 1 && *pv == '0')
		535	encode_str (enc, "false", 5, 0);
		536	else
		537	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		538	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		539	}
		540	else
		541	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
311	}	543	}
312		544
313	static void	545	static void
314	encode_sv (enc_t *enc, SV *sv)	546	encode_sv (enc_t *enc, SV *sv)
315	{	547	{
		548	SvGETMAGIC (sv);
		549
316	if (SvPOKp (sv))	550	if (SvPOKp (sv))
317	{	551	{
318	STRLEN len;	552	STRLEN len;
319	char *str = SvPV (sv, len);	553	char *str = SvPV (sv, len);
320	encode_ch (enc, '"');	554	encode_ch (enc, '"');
321	encode_str (enc, str, len, SvUTF8 (sv));	555	encode_str (enc, str, len, SvUTF8 (sv));
322	encode_ch (enc, '"');	556	encode_ch (enc, '"');
323	}	557	}
324	else if (SvNOKp (sv))	558	else if (SvNOKp (sv))
325	{	559	{
		560	// trust that perl will do the right thing w.r.t. JSON syntax.
326	need (enc, NV_DIG + 32);	561	need (enc, NV_DIG + 32);
327	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	562	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
328	enc->cur += strlen (enc->cur);	563	enc->cur += strlen (enc->cur);
329	}	564	}
330	else if (SvIOKp (sv))	565	else if (SvIOKp (sv))
331	{	566	{
332	need (enc, 64);	567	// we assume we can always read an IV as a UV
		568	if (SvUV (sv) & ~(UV)0x7fff)
		569	{
		570	// large integer, use the (rather slow) snprintf way.
		571	need (enc, sizeof (UV) * 3);
333	enc->cur +=	572	enc->cur +=
334	SvIsUV(sv)	573	SvIsUV(sv)
335	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	574	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
336	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	575	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		576	}
		577	else
		578	{
		579	// optimise the "small number case"
		580	// code will likely be branchless and use only a single multiplication
		581	I32 i = SvIV (sv);
		582	U32 u;
		583	char digit, nz = 0;
		584
		585	need (enc, 6);
		586
		587	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		588	u = i < 0 ? -i : i;
		589
		590	// convert to 4.28 fixed-point representation
		591	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		592
		593	// now output digit by digit, each time masking out the integer part
		594	// and multiplying by 5 while moving the decimal point one to the right,
		595	// resulting in a net multiplication by 10.
		596	// we always write the digit to memory but conditionally increment
		597	// the pointer, to ease the usage of conditional move instructions.
		598	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		599	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		600	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		601	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		602	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		603	}
337	}	604	}
338	else if (SvROK (sv))	605	else if (SvROK (sv))
339	{	606	encode_rv (enc, SvRV (sv));
340	if (!--enc->max_recurse)
341	croak ("data structure too deep (hit recursion limit)");
342
343	sv = SvRV (sv);
344
345	switch (SvTYPE (sv))
346	{
347	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
348	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
349
350	default:
351	croak ("JSON can only represent references to arrays or hashes");
352	}
353	}
354	else if (!SvOK (sv))	607	else if (!SvOK (sv))
355	encode_str (enc, "null", 4, 0);	608	encode_str (enc, "null", 4, 0);
356	else	609	else
357	croak ("encountered perl type that JSON cannot handle");	610	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		611	SvPV_nolen (sv), SvFLAGS (sv));
358	}	612	}
359		613
360	static SV *	614	static SV *
361	encode_json (SV *scalar, UV flags)	615	encode_json (SV *scalar, JSON *json)
362	{	616	{
363	enc_t enc;	617	enc_t enc;
364	enc.flags = flags;	618
		619	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		620	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		621
		622	enc.json = *json;
365	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	623	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
366	enc.cur = SvPVX (enc.sv);	624	enc.cur = SvPVX (enc.sv);
367	enc.end = SvEND (enc.sv);	625	enc.end = SvEND (enc.sv);
368	enc.max_recurse = 0;
369	enc.indent = 0;	626	enc.indent = 0;
		627	enc.maxdepth = DEC_DEPTH (enc.json.flags);
370		628
371	SvPOK_only (enc.sv);	629	SvPOK_only (enc.sv);
372	encode_sv (&enc, scalar);	630	encode_sv (&enc, scalar);
373		631
		632	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		633	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		634
374	if (!(flags & (F_ASCII | F_UTF8)))	635	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
375	SvUTF8_on (enc.sv);	636	SvUTF8_on (enc.sv);
376		637
377	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	638	if (enc.json.flags & F_SHRINK)
		639	shrink (enc.sv);
		640
378	return enc.sv;	641	return enc.sv;
379	}	642	}
380		643
381	/////////////////////////////////////////////////////////////////////////////	644	/////////////////////////////////////////////////////////////////////////////
		645	// decoder
382		646
383	#define WS \	647	// structure used for decoding JSON
		648	typedef struct
		649	{
		650	char *cur; // current parser pointer
		651	char *end; // end of input string
		652	const char *err; // parse error, if != 0
		653	JSON json;
		654	U32 depth; // recursion depth
		655	U32 maxdepth; // recursion depth limit
		656	} dec_t;
		657
		658	inline void
		659	decode_ws (dec_t *dec)
		660	{
384	for (;;) \	661	for (;;)
385	{ \	662	{
386	char ch = *dec->cur; \	663	char ch = *dec->cur;
		664
387	if (ch > 0x20 \	665	if (ch > 0x20
388	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	666	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
389	break; \	667	break;
		668
390	++dec->cur; \	669	++dec->cur;
391	}	670	}
		671	}
392		672
393	#define ERR(reason) SB dec->err = reason; goto fail; SE	673	#define ERR(reason) SB dec->err = reason; goto fail; SE
		674
394	#define EXPECT_CH(ch) SB \	675	#define EXPECT_CH(ch) SB \
395	if (*dec->cur != ch) \	676	if (*dec->cur != ch) \
396	ERR (# ch " expected"); \	677	ERR (# ch " expected"); \
397	++dec->cur; \	678	++dec->cur; \
398	SE	679	SE
399		680
		681	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		682	#define DEC_DEC_DEPTH --dec->depth
		683
400	static SV *decode_sv (dec_t *dec);	684	static SV *decode_sv (dec_t *dec);
401
402	#define APPEND_CH(ch) SB \
403	SvGROW (sv, cur + 1 + 1); \
404	SvPVX (sv)[cur++] = (ch); \
405	SE
406		685
407	static signed char decode_hexdigit[256];	686	static signed char decode_hexdigit[256];
408		687
409	static UV	688	static UV
410	decode_4hex (dec_t *dec)	689	decode_4hex (dec_t *dec)
411	{	690	{
412	signed char d1, d2, d3, d4;	691	signed char d1, d2, d3, d4;
		692	unsigned char *cur = (unsigned char *)dec->cur;
413		693
414	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	694	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
415	if (d1 < 0) ERR ("four hexadecimal digits expected");	695	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
416	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	696	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
417	if (d2 < 0) ERR ("four hexadecimal digits expected");	697	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
418	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
419	if (d3 < 0) ERR ("four hexadecimal digits expected");
420	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
421	if (d4 < 0) ERR ("four hexadecimal digits expected");
422		698
423	dec->cur += 4;	699	dec->cur += 4;
424		700
425	return ((UV)d1) << 12	701	return ((UV)d1) << 12
426	| ((UV)d2) << 8	702	| ((UV)d2) << 8
…		…
432	}	708	}
433		709
434	static SV *	710	static SV *
435	decode_str (dec_t *dec)	711	decode_str (dec_t *dec)
436	{	712	{
437	SV *sv = NEWSV (0,2);	713	SV *sv = 0;
438	STRLEN cur = 0;
439	int utf8 = 0;	714	int utf8 = 0;
		715	char *dec_cur = dec->cur;
440		716
441	for (;;)	717	do
442	{	718	{
443	unsigned char ch = *(unsigned char *)dec->cur;	719	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		720	char *cur = buf;
444		721
445	if (ch == '"')	722	do
446	break;
447	else if (ch == '\\')
448	{	723	{
449	switch (*++dec->cur)	724	unsigned char ch = *(unsigned char *)dec_cur++;
		725
		726	if (expect_false (ch == '"'))
		727	{
		728	--dec_cur;
		729	break;
450	{	730	}
451	case '\\':	731	else if (expect_false (ch == '\\'))
452	case '/':	732	{
453	case '"': APPEND_CH (*dec->cur++); break;	733	switch (*dec_cur)
454
455	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
456	case 't': APPEND_CH ('\011'); ++dec->cur; break;
457	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
458	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
459	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
460
461	case 'u':
462	{	734	{
463	UV lo, hi;	735	case '\\':
464	++dec->cur;	736	case '/':
		737	case '"': *cur++ = *dec_cur++; break;
465		738
466	hi = decode_4hex (dec);	739	case 'b': ++dec_cur; *cur++ = '\010'; break;
467	if (hi == (UV)-1)	740	case 't': ++dec_cur; *cur++ = '\011'; break;
468	goto fail;	741	case 'n': ++dec_cur; *cur++ = '\012'; break;
		742	case 'f': ++dec_cur; *cur++ = '\014'; break;
		743	case 'r': ++dec_cur; *cur++ = '\015'; break;
469		744
470	// possibly a surrogate pair	745	case 'u':
471	if (hi >= 0xd800 && hi < 0xdc00)
472	{	746	{
473	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	747	UV lo, hi;
474	ERR ("illegal surrogate character");	748	++dec_cur;
475		749
476	dec->cur += 2;	750	dec->cur = dec_cur;
477
478	lo = decode_4hex (dec);	751	hi = decode_4hex (dec);
		752	dec_cur = dec->cur;
479	if (lo == (UV)-1)	753	if (hi == (UV)-1)
480	goto fail;	754	goto fail;
481		755
		756	// possibly a surrogate pair
		757	if (hi >= 0xd800)
		758	if (hi < 0xdc00)
		759	{
		760	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		761	ERR ("missing low surrogate character in surrogate pair");
		762
		763	dec_cur += 2;
		764
		765	dec->cur = dec_cur;
		766	lo = decode_4hex (dec);
		767	dec_cur = dec->cur;
		768	if (lo == (UV)-1)
		769	goto fail;
		770
482	if (lo < 0xdc00 || lo >= 0xe000)	771	if (lo < 0xdc00 || lo >= 0xe000)
483	ERR ("surrogate pair expected");	772	ERR ("surrogate pair expected");
484		773
485	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	774	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		775	}
		776	else if (hi < 0xe000)
		777	ERR ("missing high surrogate character in surrogate pair");
		778
		779	if (hi >= 0x80)
		780	{
		781	utf8 = 1;
		782
		783	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		784	}
		785	else
		786	*cur++ = hi;
486	}	787	}
487	else if (lo >= 0xdc00 && lo < 0xe000)
488	ERR ("illegal surrogate character");
489
490	if (hi >= 0x80)
491	{	788	break;
492	utf8 = 1;
493		789
494	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
495	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
496	}
497	else	790	default:
498	APPEND_CH (hi);	791	--dec_cur;
		792	ERR ("illegal backslash escape sequence in string");
499	}	793	}
		794	}
		795	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		796	*cur++ = ch;
		797	else if (ch >= 0x80)
		798	{
		799	STRLEN clen;
		800	UV uch;
		801
		802	--dec_cur;
		803
		804	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		805	if (clen == (STRLEN)-1)
		806	ERR ("malformed UTF-8 character in JSON string");
		807
		808	do
		809	*cur++ = *dec_cur++;
		810	while (--clen);
		811
		812	utf8 = 1;
		813	}
		814	else
		815	{
		816	--dec_cur;
		817
		818	if (!ch)
		819	ERR ("unexpected end of string while parsing JSON string");
500	break;	820	else
		821	ERR ("invalid character encountered while parsing JSON string");
501	}	822	}
502	}	823	}
503	else if (ch >= 0x20 && ch <= 0x7f)	824	while (cur < buf + SHORT_STRING_LEN);
504	APPEND_CH (*dec->cur++);	825
505	else if (ch >= 0x80)	826	{
		827	STRLEN len = cur - buf;
		828
		829	if (sv)
506	{	830	{
507	STRLEN clen;	831	SvGROW (sv, SvCUR (sv) + len + 1);
508	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	832	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
509	if (clen < 0)	833	SvCUR_set (sv, SvCUR (sv) + len);
510	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
511
512	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits
513	memcpy (SvPVX (sv) + cur, dec->cur, clen);
514	dec->cur += clen;
515	}	834	}
516	else	835	else
517	ERR ("invalid character encountered");	836	sv = newSVpvn (buf, len);
518	}	837	}
		838	}
		839	while (*dec_cur != '"');
519		840
520	++dec->cur;	841	++dec_cur;
521		842
		843	if (sv)
		844	{
522	SvPOK_only (sv);	845	SvPOK_only (sv);
523
524	SvCUR_set (sv, cur);
525	*SvEND (sv) = 0;	846	*SvEND (sv) = 0;
526		847
527	if (utf8)	848	if (utf8)
528	SvUTF8_on (sv);	849	SvUTF8_on (sv);
		850	}
		851	else
		852	sv = newSVpvn ("", 0);
529		853
		854	dec->cur = dec_cur;
530	return sv;	855	return sv;
531		856
532	fail:	857	fail:
533	SvREFCNT_dec (sv);	858	dec->cur = dec_cur;
534	return 0;	859	return 0;
535	}	860	}
536		861
537	static SV *	862	static SV *
538	decode_num (dec_t *dec)	863	decode_num (dec_t *dec)
…		…
548	{	873	{
549	++dec->cur;	874	++dec->cur;
550	if (*dec->cur >= '0' && *dec->cur <= '9')	875	if (*dec->cur >= '0' && *dec->cur <= '9')
551	ERR ("malformed number (leading zero must not be followed by another digit)");	876	ERR ("malformed number (leading zero must not be followed by another digit)");
552	}	877	}
553		878	else if (*dec->cur < '0' || *dec->cur > '9')
554	// int	879	ERR ("malformed number (no digits after initial minus)");
		880	else
		881	do
		882	{
		883	++dec->cur;
		884	}
555	while (*dec->cur >= '0' && *dec->cur <= '9')	885	while (*dec->cur >= '0' && *dec->cur <= '9');
556	++dec->cur;
557		886
558	// [frac]	887	// [frac]
559	if (*dec->cur == '.')	888	if (*dec->cur == '.')
560	{	889	{
561	is_nv = 1;	890	++dec->cur;
		891
		892	if (*dec->cur < '0' || *dec->cur > '9')
		893	ERR ("malformed number (no digits after decimal point)");
562		894
563	do	895	do
564	{	896	{
565	++dec->cur;	897	++dec->cur;
566	}	898	}
567	while (*dec->cur >= '0' && *dec->cur <= '9');	899	while (*dec->cur >= '0' && *dec->cur <= '9');
		900
		901	is_nv = 1;
568	}	902	}
569		903
570	// [exp]	904	// [exp]
571	if (*dec->cur == 'e' || *dec->cur == 'E')	905	if (*dec->cur == 'e' || *dec->cur == 'E')
572	{	906	{
573	is_nv = 1;
574
575	++dec->cur;	907	++dec->cur;
		908
576	if (*dec->cur == '-' || *dec->cur == '+')	909	if (*dec->cur == '-' || *dec->cur == '+')
577	++dec->cur;	910	++dec->cur;
578		911
		912	if (*dec->cur < '0' || *dec->cur > '9')
		913	ERR ("malformed number (no digits after exp sign)");
		914
		915	do
		916	{
		917	++dec->cur;
		918	}
579	while (*dec->cur >= '0' && *dec->cur <= '9')	919	while (*dec->cur >= '0' && *dec->cur <= '9');
580	++dec->cur;	920
		921	is_nv = 1;
581	}	922	}
582		923
583	if (!is_nv)	924	if (!is_nv)
584	{	925	{
585	UV uv;	926	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
586	int numtype = grok_number (start, dec->cur - start, &uv);	927	if (*start == '-')
587	if (numtype & IS_NUMBER_IN_UV)	928	switch (dec->cur - start)
588	if (numtype & IS_NUMBER_NEG)
589	{	929	{
590	if (uv < (UV)IV_MIN)	930	case 2: return newSViv (-( start [1] - '0' * 1));
591	return newSViv (-(IV)uv);	931	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		932	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		933	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
592	}	934	}
		935	else
		936	switch (dec->cur - start)
		937	{
		938	case 1: return newSViv ( start [0] - '0' * 1);
		939	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		940	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		941	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		942	}
		943
		944	{
		945	UV uv;
		946	int numtype = grok_number (start, dec->cur - start, &uv);
		947	if (numtype & IS_NUMBER_IN_UV)
		948	if (numtype & IS_NUMBER_NEG)
		949	{
		950	if (uv < (UV)IV_MIN)
		951	return newSViv (-(IV)uv);
		952	}
593	else	953	else
594	return newSVuv (uv);	954	return newSVuv (uv);
		955
		956	// here would likely be the place for bigint support
595	}	957	}
		958	}
596		959
		960	// if we ever support bigint or bigfloat, this is the place for bigfloat
597	return newSVnv (Atof (start));	961	return newSVnv (Atof (start));
598		962
599	fail:	963	fail:
600	return 0;	964	return 0;
601	}	965	}
…		…
603	static SV *	967	static SV *
604	decode_av (dec_t *dec)	968	decode_av (dec_t *dec)
605	{	969	{
606	AV *av = newAV ();	970	AV *av = newAV ();
607		971
		972	DEC_INC_DEPTH;
		973	decode_ws (dec);
		974
		975	if (*dec->cur == ']')
		976	++dec->cur;
		977	else
608	for (;;)	978	for (;;)
609	{	979	{
610	SV *value;	980	SV *value;
611		981
612	value = decode_sv (dec);	982	value = decode_sv (dec);
613	if (!value)	983	if (!value)
614	goto fail;	984	goto fail;
615		985
616	av_push (av, value);	986	av_push (av, value);
617		987
618	WS;	988	decode_ws (dec);
619		989
620	if (*dec->cur == ']')	990	if (*dec->cur == ']')
621	{	991	{
622	++dec->cur;	992	++dec->cur;
623	break;	993	break;
		994	}
624	}	995
625
626	if (*dec->cur != ',')	996	if (*dec->cur != ',')
627	ERR (", or ] expected while parsing array");	997	ERR (", or ] expected while parsing array");
628		998
629	++dec->cur;	999	++dec->cur;
630	}	1000	}
631		1001
		1002	DEC_DEC_DEPTH;
632	return newRV_noinc ((SV *)av);	1003	return newRV_noinc ((SV *)av);
633		1004
634	fail:	1005	fail:
635	SvREFCNT_dec (av);	1006	SvREFCNT_dec (av);
		1007	DEC_DEC_DEPTH;
636	return 0;	1008	return 0;
637	}	1009	}
638		1010
639	static SV *	1011	static SV *
640	decode_hv (dec_t *dec)	1012	decode_hv (dec_t *dec)
641	{	1013	{
		1014	SV *sv;
642	HV *hv = newHV ();	1015	HV *hv = newHV ();
643		1016
		1017	DEC_INC_DEPTH;
		1018	decode_ws (dec);
		1019
		1020	if (*dec->cur == '}')
		1021	++dec->cur;
		1022	else
644	for (;;)	1023	for (;;)
645	{	1024	{
646	SV *key, *value;	1025	decode_ws (dec); EXPECT_CH ('"');
647		1026
648	WS; EXPECT_CH ('"');	1027	// heuristic: assume that
649		1028	// a) decode_str + hv_store_ent are abysmally slow.
650	key = decode_str (dec);	1029	// b) most hash keys are short, simple ascii text.
651	if (!key)	1030	// => try to "fast-match" such strings to avoid
652	goto fail;	1031	// the overhead of decode_str + hv_store_ent.
653
654	WS; EXPECT_CH (':');
655
656	value = decode_sv (dec);
657	if (!value)
658	{	1032	{
		1033	SV *value;
		1034	char *p = dec->cur;
		1035	char *e = p + 24; // only try up to 24 bytes
		1036
		1037	for (;;)
		1038	{
		1039	// the >= 0x80 is true on most architectures
		1040	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1041	{
		1042	// slow path, back up and use decode_str
		1043	SV *key = decode_str (dec);
		1044	if (!key)
		1045	goto fail;
		1046
		1047	decode_ws (dec); EXPECT_CH (':');
		1048
		1049	value = decode_sv (dec);
		1050	if (!value)
		1051	{
		1052	SvREFCNT_dec (key);
		1053	goto fail;
		1054	}
		1055
		1056	hv_store_ent (hv, key, value, 0);
659	SvREFCNT_dec (key);	1057	SvREFCNT_dec (key);
		1058
		1059	break;
		1060	}
		1061	else if (*p == '"')
		1062	{
		1063	// fast path, got a simple key
		1064	char *key = dec->cur;
		1065	int len = p - key;
		1066	dec->cur = p + 1;
		1067
		1068	decode_ws (dec); EXPECT_CH (':');
		1069
		1070	value = decode_sv (dec);
		1071	if (!value)
660	goto fail;	1072	goto fail;
		1073
		1074	hv_store (hv, key, len, value, 0);
		1075
		1076	break;
		1077	}
		1078
		1079	++p;
		1080	}
661	}	1081	}
662		1082
663	//TODO: optimise	1083	decode_ws (dec);
664	hv_store_ent (hv, key, value, 0);
665		1084
666	WS;
667
668	if (*dec->cur == '}')	1085	if (*dec->cur == '}')
		1086	{
		1087	++dec->cur;
		1088	break;
		1089	}
		1090
		1091	if (*dec->cur != ',')
		1092	ERR (", or } expected while parsing object/hash");
		1093
		1094	++dec->cur;
		1095	}
		1096
		1097	DEC_DEC_DEPTH;
		1098	sv = newRV_noinc ((SV *)hv);
		1099
		1100	// check filter callbacks
		1101	if (dec->json.flags & F_HOOK)
		1102	{
		1103	ENTER; SAVETMPS;
		1104
		1105	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
669	{	1106	{
670	++dec->cur;	1107	int count;
671	break;	1108
		1109	dSP; PUSHMARK (SP);
		1110	XPUSHs (sv_2mortal (sv));
		1111
		1112	PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN;
		1113
		1114	if (count == 1)
		1115	{
		1116	sv = newSVsv (POPs);
		1117	goto filter_ok;
		1118	}
		1119
		1120	SvREFCNT_inc (sv);
672	}	1121	}
673		1122
674	if (*dec->cur != ',')	1123	if (dec->json.cb_object)
675	ERR (", or } expected while parsing object/hash");	1124	{
		1125	int count;
676		1126
677	++dec->cur;	1127	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
678	}	1128	XPUSHs (sv_2mortal (sv));
679		1129
680	return newRV_noinc ((SV *)hv);	1130	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1131
		1132	if (count == 1)
		1133	{
		1134	sv = newSVsv (POPs);
		1135	goto filter_ok;
		1136	}
		1137
		1138	SvREFCNT_inc (sv);
		1139	}
		1140
		1141	filter_ok:
		1142	FREETMPS; LEAVE;
		1143	}
		1144
		1145	return sv;
681		1146
682	fail:	1147	fail:
683	SvREFCNT_dec (hv);	1148	SvREFCNT_dec (hv);
		1149	DEC_DEC_DEPTH;
684	return 0;	1150	return 0;
685	}	1151	}
686		1152
687	static SV *	1153	static SV *
688	decode_sv (dec_t *dec)	1154	decode_sv (dec_t *dec)
689	{	1155	{
690	WS;	1156	decode_ws (dec);
		1157
		1158	// the beauty of JSON: you need exactly one character lookahead
		1159	// to parse anything.
691	switch (*dec->cur)	1160	switch (*dec->cur)
692	{	1161	{
693	case '"': ++dec->cur; return decode_str (dec);	1162	case '"': ++dec->cur; return decode_str (dec);
694	case '[': ++dec->cur; return decode_av (dec);	1163	case '[': ++dec->cur; return decode_av (dec);
695	case '{': ++dec->cur; return decode_hv (dec);	1164	case '{': ++dec->cur; return decode_hv (dec);
…		…
701		1170
702	case 't':	1171	case 't':
703	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1172	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
704	{	1173	{
705	dec->cur += 4;	1174	dec->cur += 4;
706	return newSViv (1);	1175	return SvREFCNT_inc (json_true);
707	}	1176	}
708	else	1177	else
709	ERR ("'true' expected");	1178	ERR ("'true' expected");
710		1179
711	break;	1180	break;
712		1181
713	case 'f':	1182	case 'f':
714	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1183	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
715	{	1184	{
716	dec->cur += 5;	1185	dec->cur += 5;
717	return newSViv (0);	1186	return SvREFCNT_inc (json_false);
718	}	1187	}
719	else	1188	else
720	ERR ("'false' expected");	1189	ERR ("'false' expected");
721		1190
722	break;	1191	break;
723		1192
724	case 'n':	1193	case 'n':
725	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1194	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
726	{	1195	{
727	dec->cur += 4;	1196	dec->cur += 4;
728	return newSViv (1);	1197	return newSVsv (&PL_sv_undef);
729	}	1198	}
730	else	1199	else
731	ERR ("'null' expected");	1200	ERR ("'null' expected");
732		1201
733	break;	1202	break;
734		1203
735	default:	1204	default:
736	ERR ("malformed json string");	1205	ERR ("malformed JSON string, neither array, object, number, string or atom");
737	break;	1206	break;
738	}	1207	}
739		1208
740	fail:	1209	fail:
741	return 0;	1210	return 0;
742	}	1211	}
743		1212
744	static SV *	1213	static SV *
745	decode_json (SV *string, UV flags)	1214	decode_json (SV *string, JSON *json, UV *offset_return)
746	{	1215	{
		1216	dec_t dec;
		1217	UV offset;
747	SV *sv;	1218	SV *sv;
748		1219
		1220	SvGETMAGIC (string);
		1221	SvUPGRADE (string, SVt_PV);
		1222
		1223	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1224	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1225	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1226
749	if (!(flags & F_UTF8))	1227	if (json->flags & F_UTF8)
		1228	sv_utf8_downgrade (string, 0);
		1229	else
750	sv_utf8_upgrade (string);	1230	sv_utf8_upgrade (string);
751		1231
752	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1232	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
753		1233
754	dec_t dec;	1234	dec.json = *json;
755	dec.flags = flags;
756	dec.cur = SvPVX (string);	1235	dec.cur = SvPVX (string);
757	dec.end = SvEND (string);	1236	dec.end = SvEND (string);
758	dec.err = 0;	1237	dec.err = 0;
		1238	dec.depth = 0;
		1239	dec.maxdepth = DEC_DEPTH (dec.json.flags);
759		1240
760	*dec.end = 1; // invalid anywhere	1241	if (dec.json.cb_object || dec.json.cb_sk_object)
		1242	dec.json.flags |= F_HOOK;
		1243
		1244	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
761	sv = decode_sv (&dec);	1245	sv = decode_sv (&dec);
762	*dec.end = 0;	1246
		1247	if (!(offset_return || !sv))
		1248	{
		1249	// check for trailing garbage
		1250	decode_ws (&dec);
		1251
		1252	if (*dec.cur)
		1253	{
		1254	dec.err = "garbage after JSON object";
		1255	SvREFCNT_dec (sv);
		1256	sv = 0;
		1257	}
		1258	}
		1259
		1260	if (offset_return || !sv)
		1261	{
		1262	offset = dec.json.flags & F_UTF8
		1263	? dec.cur - SvPVX (string)
		1264	: utf8_distance (dec.cur, SvPVX (string));
		1265
		1266	if (offset_return)
		1267	*offset_return = offset;
		1268	}
763		1269
764	if (!sv)	1270	if (!sv)
765	{	1271	{
766	IV offset = utf8_distance (dec.cur, SvPVX (string));
767	SV *uni = sv_newmortal ();	1272	SV *uni = sv_newmortal ();
768		1273
		1274	// horrible hack to silence warning inside pv_uni_display
		1275	COP cop = *PL_curcop;
		1276	cop.cop_warnings = pWARN_NONE;
		1277	ENTER;
		1278	SAVEVPTR (PL_curcop);
		1279	PL_curcop = &cop;
769	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1280	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1281	LEAVE;
		1282
770	croak ("%s, at character %d (%s)",	1283	croak ("%s, at character offset %d [\"%s\"]",
771	dec.err,	1284	dec.err,
772	(int)offset,	1285	(int)offset,
773	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1286	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
774	}	1287	}
775		1288
776	sv_dump (sv);//D
777	return sv_2mortal (sv);	1289	sv = sv_2mortal (sv);
		1290
		1291	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
		1292	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
		1293
		1294	return sv;
778	}	1295	}
		1296
		1297	/////////////////////////////////////////////////////////////////////////////
		1298	// XS interface functions
779		1299
780	MODULE = JSON::XS PACKAGE = JSON::XS	1300	MODULE = JSON::XS PACKAGE = JSON::XS
781		1301
782	BOOT:	1302	BOOT:
783	{	1303	{
784	int i;	1304	int i;
785		1305
786	memset (decode_hexdigit, 0xff, 256);
787	for (i = 10; i--; )	1306	for (i = 0; i < 256; ++i)
788	decode_hexdigit ['0' + i] = i;	1307	decode_hexdigit [i] =
		1308	i >= '0' && i <= '9' ? i - '0'
		1309	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1310	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1311	: -1;
789		1312
790	for (i = 6; --i; )	1313	json_stash = gv_stashpv ("JSON::XS" , 1);
		1314	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1315
		1316	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1317	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1318	}
		1319
		1320	PROTOTYPES: DISABLE
		1321
		1322	void new (char *klass)
		1323	PPCODE:
		1324	{
		1325	SV *pv = NEWSV (0, sizeof (JSON));
		1326	SvPOK_only (pv);
		1327	Zero (SvPVX (pv), 1, JSON);
		1328	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1329	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1330	}
		1331
		1332	void ascii (JSON *self, int enable = 1)
		1333	ALIAS:
		1334	ascii = F_ASCII
		1335	latin1 = F_LATIN1
		1336	utf8 = F_UTF8
		1337	indent = F_INDENT
		1338	canonical = F_CANONICAL
		1339	space_before = F_SPACE_BEFORE
		1340	space_after = F_SPACE_AFTER
		1341	pretty = F_PRETTY
		1342	allow_nonref = F_ALLOW_NONREF
		1343	shrink = F_SHRINK
		1344	allow_blessed = F_ALLOW_BLESSED
		1345	convert_blessed = F_CONV_BLESSED
		1346	PPCODE:
		1347	{
		1348	if (enable)
		1349	self->flags |= ix;
		1350	else
		1351	self->flags &= ~ix;
		1352
		1353	XPUSHs (ST (0));
		1354	}
		1355
		1356	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1357	PPCODE:
		1358	{
		1359	UV log2 = 0;
		1360
		1361	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1362
		1363	while ((1UL << log2) < max_depth)
		1364	++log2;
		1365
		1366	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1367
		1368	XPUSHs (ST (0));
		1369	}
		1370
		1371	void max_size (JSON *self, UV max_size = 0)
		1372	PPCODE:
		1373	{
		1374	UV log2 = 0;
		1375
		1376	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1377	if (max_size == 1) max_size = 2;
		1378
		1379	while ((1UL << log2) < max_size)
		1380	++log2;
		1381
		1382	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1383
		1384	XPUSHs (ST (0));
		1385	}
		1386
		1387	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1388	ALIAS:
		1389	filter_json_single_key_object = 1
		1390	PPCODE:
		1391	{
		1392	SV **svp;
		1393
		1394	if (!SvOK (cb))
		1395	cb = 0;
		1396	else
		1397	cb = newSVsv (cb);
		1398
		1399	switch (ix)
791	{	1400	{
792	decode_hexdigit ['a' + i] = 10 + i;	1401	case 0: svp = &self->cb_object ; break;
793	decode_hexdigit ['A' + i] = 10 + i;	1402	case 1: svp = &self->cb_sk_object; break;
794	}	1403	}
795		1404
796	json_stash = gv_stashpv ("JSON::XS", 1);	1405	if (*svp)
797	}	1406	SvREFCNT_dec (*svp);
798		1407
799	SV *new (char *dummy)	1408	*svp = cb;
800	CODE:
801	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
802	OUTPUT:
803	RETVAL
804		1409
805	SV *ascii (SV *self, int enable)	1410	XPUSHs (ST (0));
806	ALIAS:
807	ascii = F_ASCII
808	utf8 = F_UTF8
809	indent = F_INDENT
810	canonical = F_CANONICAL
811	space_before = F_SPACE_BEFORE
812	space_after = F_SPACE_AFTER
813	json_rpc = F_JSON_RPC
814	CODE:
815	{
816	UV *uv = SvJSON (self);
817	if (enable)
818	*uv |= ix;
819	else
820	*uv &= ~ix;
821
822	RETVAL = newSVsv (self);
823	}	1411	}
824	OUTPUT:
825	RETVAL
826		1412
827	void encode (SV *self, SV *scalar)	1413	void encode (JSON *self, SV *scalar)
828	PPCODE:	1414	PPCODE:
829	XPUSHs (encode_json (scalar, *SvJSON (self)));	1415	XPUSHs (encode_json (scalar, self));
830		1416
831	void decode (SV *self, SV *jsondata)	1417	void decode (JSON *self, SV *jsonstr)
832	PPCODE:	1418	PPCODE:
833	XPUSHs (decode_json (jsondata, *SvJSON (self)));	1419	XPUSHs (decode_json (jsonstr, self, 0));
834		1420
		1421	void decode_prefix (JSON *self, SV *jsonstr)
		1422	PPCODE:
		1423	{
		1424	UV offset;
		1425	EXTEND (SP, 2);
		1426	PUSHs (decode_json (jsonstr, self, &offset));
		1427	PUSHs (sv_2mortal (newSVuv (offset)));
		1428	}
		1429
		1430	PROTOTYPES: ENABLE
		1431
		1432	void to_json (SV *scalar)
		1433	PPCODE:
		1434	{
		1435	JSON json = { F_DEFAULT | F_UTF8 };
		1436	XPUSHs (encode_json (scalar, &json));
		1437	}
		1438
		1439	void from_json (SV *jsonstr)
		1440	PPCODE:
		1441	{
		1442	JSON json = { F_DEFAULT | F_UTF8 };
		1443	XPUSHs (decode_json (jsonstr, &json, 0));
		1444	}
		1445

Diff Legend

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