ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.44 by root, Mon Jun 25 04:08:17 2007 UTC

…		…
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include "assert.h"
6	#include "string.h"	6	#include "string.h"
7	#include "stdlib.h"	7	#include "stdlib.h"
		8	#include "stdio.h"
8		9
		10	#if defined(__BORLANDC__) || defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
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
		34	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
17		35
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	36	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	37	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		38
21	#define INIT_SIZE 32 // initial scalar size to be allocated	39	#define INIT_SIZE 32 // initial scalar size to be allocated
		40	#define INDENT_STEP 3 // spaces per indentation level
		41
		42	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		43
23	#define SB do {	44	#define SB do {
24	#define SE } while (0)	45	#define SE } while (0)
25		46
26	static HV *json_stash;	47	#if __GNUC__ >= 3
		48	# define expect(expr,value) __builtin_expect ((expr),(value))
		49	# define inline inline
		50	#else
		51	# define expect(expr,value) (expr)
		52	# define inline static
		53	#endif
		54
		55	#define expect_false(expr) expect ((expr) != 0, 0)
		56	#define expect_true(expr) expect ((expr) != 0, 1)
		57
		58	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		59	static SV *json_true, *json_false;
		60
		61	/////////////////////////////////////////////////////////////////////////////
		62	// utility functions
		63
		64	static UV *
		65	SvJSON (SV *sv)
		66	{
		67	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		68	croak ("object is not of type JSON::XS");
		69
		70	return &SvUVX (SvRV (sv));
		71	}
		72
		73	static void
		74	shrink (SV *sv)
		75	{
		76	sv_utf8_downgrade (sv, 1);
		77	if (SvLEN (sv) > SvCUR (sv) + 1)
		78	{
		79	#ifdef SvPV_shrink_to_cur
		80	SvPV_shrink_to_cur (sv);
		81	#elif defined (SvPV_renew)
		82	SvPV_renew (sv, SvCUR (sv) + 1);
		83	#endif
		84	}
		85	}
		86
		87	// decode an utf-8 character and return it, or (UV)-1 in
		88	// case of an error.
		89	// we special-case "safe" characters from U+80 .. U+7FF,
		90	// but use the very good perl function to parse anything else.
		91	// note that we never call this function for a ascii codepoints
		92	inline UV
		93	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		94	{
		95	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		98	{
		99	*clen = 2;
		100	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		101	}
		102	else
		103	{
		104	*clen = (STRLEN)-1;
		105	return (UV)-1;
		106	}
		107	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// encoder
27		111
28	// structure used for encoding JSON	112	// structure used for encoding JSON
29	typedef struct	113	typedef struct
30	{	114	{
31	char *cur;	115	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	116	char *end; // SvEND (sv)
34	SV *sv;	117	SV *sv; // result scalar
35	UV flags;	118	U32 flags; // F_*
36	int max_recurse;	119	U32 indent; // indentation level
37	int indent;	120	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	121	} enc_t;
39		122
40	// structure used for decoding JSON	123	inline void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const 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)	124	need (enc_t *enc, STRLEN len)
62	{	125	{
63	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
64	{	127	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	132	}
70	}	133	}
71		134
72	static void	135	inline void
73	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
74	{	137	{
75	need (enc, 1);	138	need (enc, 1);
76	*enc->cur++ = ch;	139	*enc->cur++ = ch;
77	}	140	}
…		…
85		148
86	while (str < end)	149	while (str < end)
87	{	150	{
88	unsigned char ch = *(unsigned char *)str;	151	unsigned char ch = *(unsigned char *)str;
89		152
90	if (ch == '"')	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	154	{
		155	if (expect_false (ch == '"')) // but with slow exceptions
		156	{
92	need (enc, len += 1);	157	need (enc, len += 1);
93	*enc->cur++ = '\\';	158	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	159	*enc->cur++ = '"';
95	++str;
96	}	160	}
97	else if (ch == '\\')	161	else if (expect_false (ch == '\\'))
98	{	162	{
99	need (enc, len += 1);	163	need (enc, len += 1);
100	*enc->cur++ = '\\';	164	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
102	++str;
103	}	166	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	167	else
105	{
106	*enc->cur++ = ch;	168	*enc->cur++ = ch;
107	++str;	169
108	}
109	else if (ch == '\015')
110	{
111	need (enc, len += 1);
112	*enc->cur++ = '\\';
113	*enc->cur++ = 'r';
114	++str;
115	}
116	else if (ch == '\012')
117	{
118	need (enc, len += 1);
119	*enc->cur++ = '\\';
120	*enc->cur++ = 'n';
121	++str;	170	++str;
122	}	171	}
123	else	172	else
124	{	173	{
125	STRLEN clen;	174	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{	175	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	176	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131	if (clen < 0)	177	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");	178	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133	}	179	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134	else	180	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135	{
136	uch = ch;
137	clen = 1;
138	}
139		181
140	if (uch < 0x80 || enc->flags & F_ASCII)	182	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	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;
		196	clen = 1;
		197	}
		198
		199	if (uch > 0x10FFFFUL)
		200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		201
		202	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
		203	{
		204	if (uch > 0xFFFFUL)
		205	{
144	need (enc, len += 11);	206	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	208	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	209	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	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->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	}
149	}	246	}
150	else
151	{
152	static char hexdigit [16] = "0123456789abcdef";
153	need (enc, len += 5);
154	*enc->cur++ = '\\';
155	*enc->cur++ = 'u';
156	*enc->cur++ = hexdigit [ uch >> 12 ];
157	*enc->cur++ = hexdigit [(uch >> 8) & 15];
158	*enc->cur++ = hexdigit [(uch >> 4) & 15];
159	*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	}
161
162	str += clen;
163	}
164	else if (is_utf8)
165	{
166	need (enc, len += clen);
167	while (clen--)
168	*enc->cur++ = *str++;
169	}
170	else
171	{
172	need (enc, 10); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174	++str;
175	}	247	}
176	}	248	}
177		249
178	--len;	250	--len;
179	}	251	}
180	}	252	}
181		253
182	#define INDENT SB \	254	inline void
		255	encode_indent (enc_t *enc)
		256	{
183	if (enc->flags & F_INDENT) \	257	if (enc->flags & F_INDENT)
184	{ \	258	{
185	int i_; \	259	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \
187	for (i_ = enc->indent * 3; i_--; )\
188	encode_ch (enc, ' '); \
189	} \
190	SE
191		260
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	261	need (enc, spaces);
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	262	memset (enc->cur, ' ', spaces);
194	#define COMMA SB \	263	enc->cur += spaces;
		264	}
		265	}
		266
		267	inline void
		268	encode_space (enc_t *enc)
		269	{
		270	need (enc, 1);
195	encode_ch (enc, ','); \	271	encode_ch (enc, ' ');
		272	}
		273
		274	inline void
		275	encode_nl (enc_t *enc)
		276	{
196	if (enc->flags & F_INDENT) \	277	if (enc->flags & F_INDENT)
197	NL; \	278	{
		279	need (enc, 1);
		280	encode_ch (enc, '\n');
		281	}
		282	}
		283
		284	inline void
		285	encode_comma (enc_t *enc)
		286	{
		287	encode_ch (enc, ',');
		288
		289	if (enc->flags & F_INDENT)
		290	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	291	else if (enc->flags & F_SPACE_AFTER)
199	SPACE; \	292	encode_space (enc);
200	SE	293	}
201		294
202	static void encode_sv (enc_t *enc, SV *sv);	295	static void encode_sv (enc_t *enc, SV *sv);
203		296
204	static void	297	static void
205	encode_av (enc_t *enc, AV *av)	298	encode_av (enc_t *enc, AV *av)
206	{	299	{
207	int i, len = av_len (av);	300	int i, len = av_len (av);
208		301
209	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);
210	++enc->indent;	306	++enc->indent;
211		307
212	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
213	{	309	{
214	INDENT;	310	encode_indent (enc);
215	encode_sv (enc, *av_fetch (av, i, 0));	311	encode_sv (enc, *av_fetch (av, i, 0));
216		312
217	if (i < len)	313	if (i < len)
218	COMMA;	314	encode_comma (enc);
219	}	315	}
220		316
221	NL;	317	encode_nl (enc);
222		318
223	--enc->indent;	319	--enc->indent;
224	INDENT; encode_ch (enc, ']');	320	encode_indent (enc); encode_ch (enc, ']');
225	}	321	}
226		322
227	static void	323	static void
228	encode_he (enc_t *enc, HE *he)	324	encode_he (enc_t *enc, HE *he)
229	{	325	{
…		…
243	else	339	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	340	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		341
246	encode_ch (enc, '"');	342	encode_ch (enc, '"');
247		343
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	344	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	345	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	346	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));	347	encode_sv (enc, HeVAL (he));
252	}	348	}
253		349
254	// compare hash entries, used when all keys are bytestrings	350	// compare hash entries, used when all keys are bytestrings
255	static int	351	static int
…		…
261	HE *b = *(HE **)b_;	357	HE *b = *(HE **)b_;
262		358
263	STRLEN la = HeKLEN (a);	359	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	360	STRLEN lb = HeKLEN (b);
265		361
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	363	cmp = la - lb;
268		364
269	return cmp;	365	return cmp;
270	}	366	}
271		367
272	// 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
…		…
279	static void	375	static void
280	encode_hv (enc_t *enc, HV *hv)	376	encode_hv (enc_t *enc, HV *hv)
281	{	377	{
282	int count, i;	378	int count, i;
283		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
284	encode_ch (enc, '{'); NL; ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
285		384
286	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
287	{	386	{
288	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
291	// that randomises hash orderings	390	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)	391	if (enc->flags & F_CANONICAL)
293	{	392	{
294	HE *he, *hes [count];
295	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
296		400
297	i = 0;	401	i = 0;
298	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
299	{	403	{
300	hes [i++] = he;	404	hes [i++] = he;
…		…
306		410
307	if (fast)	411	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	412	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	413	else
310	{	414	{
311	// hack to disable "use bytes"	415	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	416	COP cop = *PL_curcop;
313	cop.op_private = 0;	417	cop.op_private = 0;
		418
		419	ENTER;
		420	SAVETMPS;
		421
		422	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	423	PL_curcop = &cop;
315		424
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	425	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		426
318	FREETMPS;	427	FREETMPS;
319		428	LEAVE;
320	PL_curcop = oldcop;
321	}	429	}
322		430
323	for (i = 0; i < count; ++i)	431	for (i = 0; i < count; ++i)
324	{	432	{
325	INDENT;	433	encode_indent (enc);
326	encode_he (enc, hes [i]);	434	encode_he (enc, hes [i]);
327		435
328	if (i < count - 1)	436	if (i < count - 1)
329	COMMA;	437	encode_comma (enc);
330	}	438	}
331		439
332	NL;	440	encode_nl (enc);
333	}	441	}
334	else	442	else
335	{	443	{
336	SV *sv;
337	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
338		445
339	for (;;)	446	for (;;)
340	{	447	{
341	INDENT;	448	encode_indent (enc);
342	encode_he (enc, he);	449	encode_he (enc, he);
343		450
344	if (!(he = hv_iternext (hv)))	451	if (!(he = hv_iternext (hv)))
345	break;	452	break;
346		453
347	COMMA;	454	encode_comma (enc);
348	}	455	}
349		456
350	NL;	457	encode_nl (enc);
351	}	458	}
352	}	459	}
353		460
354	--enc->indent; INDENT; encode_ch (enc, '}');	461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		462	}
		463
		464	// encode objects, arrays and special \0=false and \1=true values.
		465	static void
		466	encode_rv (enc_t *enc, SV *sv)
		467	{
		468	svtype svt;
		469
		470	SvGETMAGIC (sv);
		471	svt = SvTYPE (sv);
		472
		473	if (expect_false (SvOBJECT (sv)))
		474	{
		475	if (SvSTASH (sv) == json_boolean_stash)
		476	{
		477	if (SvIV (sv) == 0)
		478	encode_str (enc, "false", 5, 0);
		479	else
		480	encode_str (enc, "true", 4, 0);
		481	}
		482	else
		483	{
		484	#if 0
		485	if (0 && sv_derived_from (rv, "JSON::Literal"))
		486	{
		487	// not yet
		488	}
		489	#endif
		490	if (enc->flags & F_CONV_BLESSED)
		491	{
		492	// we re-bless the reference to get overload and other niceties right
		493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		494
		495	if (to_json)
		496	{
		497	dSP;
		498	ENTER;
		499	SAVETMPS;
		500	PUSHMARK (SP);
		501	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		502
		503	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		504	// check anyways.
		505	PUTBACK;
		506	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		507	SPAGAIN;
		508
		509	encode_sv (enc, POPs);
		510
		511	FREETMPS;
		512	LEAVE;
		513	}
		514	else if (enc->flags & F_ALLOW_BLESSED)
		515	encode_str (enc, "null", 4, 0);
		516	else
		517	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		518	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		519	}
		520	else if (enc->flags & F_ALLOW_BLESSED)
		521	encode_str (enc, "null", 4, 0);
		522	else
		523	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		524	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		525	}
		526	}
		527	else if (svt == SVt_PVHV)
		528	encode_hv (enc, (HV *)sv);
		529	else if (svt == SVt_PVAV)
		530	encode_av (enc, (AV *)sv);
		531	else if (svt < SVt_PVAV)
		532	{
		533	if (SvNIOK (sv) && SvIV (sv) == 0)
		534	encode_str (enc, "false", 5, 0);
		535	else if (SvNIOK (sv) && SvIV (sv) == 1)
		536	encode_str (enc, "true", 4, 0);
		537	else
		538	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		539	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		540	}
		541	else
		542	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		543	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	544	}
356		545
357	static void	546	static void
358	encode_sv (enc_t *enc, SV *sv)	547	encode_sv (enc_t *enc, SV *sv)
359	{	548	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	556	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	557	encode_ch (enc, '"');
369	}	558	}
370	else if (SvNOKp (sv))	559	else if (SvNOKp (sv))
371	{	560	{
		561	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	562	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	563	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	564	enc->cur += strlen (enc->cur);
375	}	565	}
376	else if (SvIOKp (sv))	566	else if (SvIOKp (sv))
377	{	567	{
378	need (enc, 64);	568	// we assume we can always read an IV as a UV
		569	if (SvUV (sv) & ~(UV)0x7fff)
		570	{
		571	// large integer, use the (rather slow) snprintf way.
		572	need (enc, sizeof (UV) * 3);
379	enc->cur +=	573	enc->cur +=
380	SvIsUV(sv)	574	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	575	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	576	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		577	}
		578	else
		579	{
		580	// optimise the "small number case"
		581	// code will likely be branchless and use only a single multiplication
		582	I32 i = SvIV (sv);
		583	U32 u;
		584	char digit, nz = 0;
		585
		586	need (enc, 6);
		587
		588	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		589	u = i < 0 ? -i : i;
		590
		591	// convert to 4.28 fixed-point representation
		592	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		593
		594	// now output digit by digit, each time masking out the integer part
		595	// and multiplying by 5 while moving the decimal point one to the right,
		596	// resulting in a net multiplication by 10.
		597	// we always write the digit to memory but conditionally increment
		598	// the pointer, to ease the usage of conditional move instructions.
		599	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		600	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		601	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		602	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		603	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		604	}
383	}	605	}
384	else if (SvROK (sv))	606	else if (SvROK (sv))
385	{	607	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)
387	croak ("data structure too deep (hit recursion limit)");
388
389	sv = SvRV (sv);
390
391	switch (SvTYPE (sv))
392	{
393	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396	default:
397	croak ("JSON can only represent references to arrays or hashes");
398	}
399	}
400	else if (!SvOK (sv))	608	else if (!SvOK (sv))
401	encode_str (enc, "null", 4, 0);	609	encode_str (enc, "null", 4, 0);
402	else	610	else
403	croak ("encountered perl type that JSON cannot handle");	611	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		612	SvPV_nolen (sv), SvFLAGS (sv));
404	}	613	}
405		614
406	static SV *	615	static SV *
407	encode_json (SV *scalar, UV flags)	616	encode_json (SV *scalar, U32 flags)
408	{	617	{
		618	enc_t enc;
		619
409	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	620	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");	621	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
411		622
412	enc_t enc;
413	enc.flags = flags;	623	enc.flags = flags;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	624	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	625	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	626	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	627	enc.indent = 0;
		628	enc.maxdepth = DEC_DEPTH (flags);
419		629
420	SvPOK_only (enc.sv);	630	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	631	encode_sv (&enc, scalar);
422		632
		633	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		634	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		635
423	if (!(flags & (F_ASCII | F_UTF8)))	636	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	637	SvUTF8_on (enc.sv);
425		638
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	639	if (enc.flags & F_SHRINK)
		640	shrink (enc.sv);
		641
427	return enc.sv;	642	return enc.sv;
428	}	643	}
429		644
430	/////////////////////////////////////////////////////////////////////////////	645	/////////////////////////////////////////////////////////////////////////////
		646	// decoder
431		647
432	#define WS \	648	// structure used for decoding JSON
		649	typedef struct
		650	{
		651	char *cur; // current parser pointer
		652	char *end; // end of input string
		653	const char *err; // parse error, if != 0
		654	U32 flags; // F_*
		655	U32 depth; // recursion depth
		656	U32 maxdepth; // recursion depth limit
		657	} dec_t;
		658
		659	inline void
		660	decode_ws (dec_t *dec)
		661	{
433	for (;;) \	662	for (;;)
434	{ \	663	{
435	char ch = *dec->cur; \	664	char ch = *dec->cur;
		665
436	if (ch > 0x20 \	666	if (ch > 0x20
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	667	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
438	break; \	668	break;
		669
439	++dec->cur; \	670	++dec->cur;
440	}	671	}
		672	}
441		673
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	674	#define ERR(reason) SB dec->err = reason; goto fail; SE
		675
443	#define EXPECT_CH(ch) SB \	676	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	677	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	678	ERR (# ch " expected"); \
446	++dec->cur; \	679	++dec->cur; \
447	SE	680	SE
448		681
		682	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		683	#define DEC_DEC_DEPTH --dec->depth
		684
449	static SV *decode_sv (dec_t *dec);	685	static SV *decode_sv (dec_t *dec);
450		686
451	static signed char decode_hexdigit[256];	687	static signed char decode_hexdigit[256];
452		688
453	static UV	689	static UV
454	decode_4hex (dec_t *dec)	690	decode_4hex (dec_t *dec)
455	{	691	{
456	signed char d1, d2, d3, d4;	692	signed char d1, d2, d3, d4;
		693	unsigned char *cur = (unsigned char *)dec->cur;
457		694
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	695	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459	if (d1 < 0) ERR ("four hexadecimal digits expected");	696	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	697	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461	if (d2 < 0) ERR ("four hexadecimal digits expected");	698	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463	if (d3 < 0) ERR ("four hexadecimal digits expected");
464	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465	if (d4 < 0) ERR ("four hexadecimal digits expected");
466		699
467	dec->cur += 4;	700	dec->cur += 4;
468		701
469	return ((UV)d1) << 12	702	return ((UV)d1) << 12
470	| ((UV)d2) << 8	703	| ((UV)d2) << 8
…		…
473		706
474	fail:	707	fail:
475	return (UV)-1;	708	return (UV)-1;
476	}	709	}
477		710
478	#define APPEND_GROW(n) SB \
479	if (cur + (n) >= end) \
480	{ \
481	STRLEN ofs = cur - SvPVX (sv); \
482	SvGROW (sv, ofs + (n) + 1); \
483	cur = SvPVX (sv) + ofs; \
484	end = SvEND (sv); \
485	} \
486	SE
487
488	#define APPEND_CH(ch) SB \
489	APPEND_GROW (1); \
490	*cur++ = (ch); \
491	SE
492
493	static SV *	711	static SV *
494	decode_str (dec_t *dec)	712	decode_str (dec_t *dec)
495	{	713	{
496	SV *sv = NEWSV (0,2);	714	SV *sv = 0;
497	int utf8 = 0;	715	int utf8 = 0;
498	char *cur = SvPVX (sv);	716	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		717
501	for (;;)	718	do
502	{	719	{
503	unsigned char ch = *(unsigned char *)dec->cur;	720	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		721	char *cur = buf;
504		722
505	if (ch == '"')	723	do
506	break;
507	else if (ch == '\\')
508	{	724	{
509	switch (*++dec->cur)	725	unsigned char ch = *(unsigned char *)dec_cur++;
		726
		727	if (expect_false (ch == '"'))
510	{	728	{
511	case '\\':	729	--dec_cur;
512	case '/':	730	break;
513	case '"': APPEND_CH (*dec->cur++); break;	731	}
514		732	else if (expect_false (ch == '\\'))
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	733	{
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;	734	switch (*dec_cur)
517	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521	case 'u':
522	{	735	{
523	UV lo, hi;	736	case '\\':
524	++dec->cur;	737	case '/':
		738	case '"': *cur++ = *dec_cur++; break;
525		739
526	hi = decode_4hex (dec);	740	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	741	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	742	case 'n': ++dec_cur; *cur++ = '\012'; break;
		743	case 'f': ++dec_cur; *cur++ = '\014'; break;
		744	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		745
530	// possibly a surrogate pair	746	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	747	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	748	UV lo, hi;
534	ERR ("illegal surrogate character");	749	++dec_cur;
535		750
536	dec->cur += 2;	751	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	752	hi = decode_4hex (dec);
		753	dec_cur = dec->cur;
539	if (lo == (UV)-1)	754	if (hi == (UV)-1)
540	goto fail;	755	goto fail;
541		756
		757	// possibly a surrogate pair
		758	if (hi >= 0xd800)
		759	if (hi < 0xdc00)
		760	{
		761	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		762	ERR ("missing low surrogate character in surrogate pair");
		763
		764	dec_cur += 2;
		765
		766	dec->cur = dec_cur;
		767	lo = decode_4hex (dec);
		768	dec_cur = dec->cur;
		769	if (lo == (UV)-1)
		770	goto fail;
		771
542	if (lo < 0xdc00 || lo >= 0xe000)	772	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	773	ERR ("surrogate pair expected");
544		774
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	775	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		776	}
		777	else if (hi < 0xe000)
		778	ERR ("missing high surrogate character in surrogate pair");
		779
		780	if (hi >= 0x80)
		781	{
		782	utf8 = 1;
		783
		784	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		785	}
		786	else
		787	*cur++ = hi;
546	}	788	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	789	break;
552	utf8 = 1;
553		790
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	791	default:
558	APPEND_CH (hi);	792	--dec_cur;
		793	ERR ("illegal backslash escape sequence in string");
559	}	794	}
		795	}
		796	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		797	*cur++ = ch;
		798	else if (ch >= 0x80)
		799	{
		800	STRLEN clen;
		801	UV uch;
		802
		803	--dec_cur;
		804
		805	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		806	if (clen == (STRLEN)-1)
		807	ERR ("malformed UTF-8 character in JSON string");
		808
		809	do
		810	*cur++ = *dec_cur++;
		811	while (--clen);
		812
		813	utf8 = 1;
		814	}
		815	else
		816	{
		817	--dec_cur;
		818
		819	if (!ch)
		820	ERR ("unexpected end of string while parsing JSON string");
560	break;	821	else
		822	ERR ("invalid character encountered while parsing JSON string");
561	}	823	}
562	}	824	}
563	else if (ch >= 0x20 && ch <= 0x7f)	825	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	826
565	else if (ch >= 0x80)	827	{
		828	STRLEN len = cur - buf;
		829
		830	if (sv)
566	{	831	{
567	STRLEN clen;	832	SvGROW (sv, SvCUR (sv) + len + 1);
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	833	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569	if (clen < 0)	834	SvCUR_set (sv, SvCUR (sv) + len);
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571
572	APPEND_GROW (clen);
573	memcpy (cur, dec->cur, clen);
574	cur += clen;
575	dec->cur += clen;
576	}	835	}
577	else	836	else
578	ERR ("invalid character encountered");	837	sv = newSVpvn (buf, len);
579	}	838	}
		839	}
		840	while (*dec_cur != '"');
580		841
581	++dec->cur;	842	++dec_cur;
582		843
583	SvCUR_set (sv, cur - SvPVX (sv));	844	if (sv)
584		845	{
585	SvPOK_only (sv);	846	SvPOK_only (sv);
586	*SvEND (sv) = 0;	847	*SvEND (sv) = 0;
587		848
588	if (utf8)	849	if (utf8)
589	SvUTF8_on (sv);	850	SvUTF8_on (sv);
		851	}
		852	else
		853	sv = newSVpvn ("", 0);
590		854
		855	dec->cur = dec_cur;
591	return sv;	856	return sv;
592		857
593	fail:	858	fail:
594	SvREFCNT_dec (sv);	859	dec->cur = dec_cur;
595	return 0;	860	return 0;
596	}	861	}
597		862
598	static SV *	863	static SV *
599	decode_num (dec_t *dec)	864	decode_num (dec_t *dec)
…		…
609	{	874	{
610	++dec->cur;	875	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	876	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	877	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	878	}
614		879	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	880	ERR ("malformed number (no digits after initial minus)");
		881	else
		882	do
		883	{
		884	++dec->cur;
		885	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	886	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		887
619	// [frac]	888	// [frac]
620	if (*dec->cur == '.')	889	if (*dec->cur == '.')
621	{	890	{
622	is_nv = 1;	891	++dec->cur;
		892
		893	if (*dec->cur < '0' || *dec->cur > '9')
		894	ERR ("malformed number (no digits after decimal point)");
623		895
624	do	896	do
625	{	897	{
626	++dec->cur;	898	++dec->cur;
627	}	899	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	900	while (*dec->cur >= '0' && *dec->cur <= '9');
		901
		902	is_nv = 1;
629	}	903	}
630		904
631	// [exp]	905	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	906	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	907	{
634	is_nv = 1;
635
636	++dec->cur;	908	++dec->cur;
		909
637	if (*dec->cur == '-' || *dec->cur == '+')	910	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	911	++dec->cur;
639		912
		913	if (*dec->cur < '0' || *dec->cur > '9')
		914	ERR ("malformed number (no digits after exp sign)");
		915
		916	do
		917	{
		918	++dec->cur;
		919	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	920	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	921
		922	is_nv = 1;
642	}	923	}
643		924
644	if (!is_nv)	925	if (!is_nv)
645	{	926	{
646	UV uv;	927	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
647	int numtype = grok_number (start, dec->cur - start, &uv);	928	if (*start == '-')
648	if (numtype & IS_NUMBER_IN_UV)	929	switch (dec->cur - start)
649	if (numtype & IS_NUMBER_NEG)
650	{	930	{
651	if (uv < (UV)IV_MIN)	931	case 2: return newSViv (-( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	932	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		933	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		934	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
653	}	935	}
		936	else
		937	switch (dec->cur - start)
		938	{
		939	case 1: return newSViv ( start [0] - '0' * 1);
		940	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		941	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		942	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		943	}
		944
		945	{
		946	UV uv;
		947	int numtype = grok_number (start, dec->cur - start, &uv);
		948	if (numtype & IS_NUMBER_IN_UV)
		949	if (numtype & IS_NUMBER_NEG)
		950	{
		951	if (uv < (UV)IV_MIN)
		952	return newSViv (-(IV)uv);
		953	}
654	else	954	else
655	return newSVuv (uv);	955	return newSVuv (uv);
		956
		957	// here would likely be the place for bigint support
656	}	958	}
		959	}
657		960
		961	// if we ever support bigint or bigfloat, this is the place for bigfloat
658	return newSVnv (Atof (start));	962	return newSVnv (Atof (start));
659		963
660	fail:	964	fail:
661	return 0;	965	return 0;
662	}	966	}
…		…
664	static SV *	968	static SV *
665	decode_av (dec_t *dec)	969	decode_av (dec_t *dec)
666	{	970	{
667	AV *av = newAV ();	971	AV *av = newAV ();
668		972
		973	DEC_INC_DEPTH;
		974	decode_ws (dec);
		975
		976	if (*dec->cur == ']')
		977	++dec->cur;
		978	else
669	for (;;)	979	for (;;)
670	{	980	{
671	SV *value;	981	SV *value;
672		982
673	value = decode_sv (dec);	983	value = decode_sv (dec);
674	if (!value)	984	if (!value)
675	goto fail;	985	goto fail;
676		986
677	av_push (av, value);	987	av_push (av, value);
678		988
679	WS;	989	decode_ws (dec);
680		990
681	if (*dec->cur == ']')	991	if (*dec->cur == ']')
682	{	992	{
683	++dec->cur;	993	++dec->cur;
684	break;	994	break;
		995	}
685	}	996
686
687	if (*dec->cur != ',')	997	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	998	ERR (", or ] expected while parsing array");
689		999
690	++dec->cur;	1000	++dec->cur;
691	}	1001	}
692		1002
		1003	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1004	return newRV_noinc ((SV *)av);
694		1005
695	fail:	1006	fail:
696	SvREFCNT_dec (av);	1007	SvREFCNT_dec (av);
		1008	DEC_DEC_DEPTH;
697	return 0;	1009	return 0;
698	}	1010	}
699		1011
700	static SV *	1012	static SV *
701	decode_hv (dec_t *dec)	1013	decode_hv (dec_t *dec)
702	{	1014	{
703	HV *hv = newHV ();	1015	HV *hv = newHV ();
704		1016
		1017	DEC_INC_DEPTH;
		1018	decode_ws (dec);
		1019
		1020	if (*dec->cur == '}')
		1021	++dec->cur;
		1022	else
705	for (;;)	1023	for (;;)
706	{	1024	{
707	SV *key, *value;	1025	SV *key, *value;
708		1026
709	WS; EXPECT_CH ('"');	1027	decode_ws (dec); EXPECT_CH ('"');
710		1028
711	key = decode_str (dec);	1029	key = decode_str (dec);
712	if (!key)	1030	if (!key)
713	goto fail;
714
715	WS; EXPECT_CH (':');
716
717	value = decode_sv (dec);
718	if (!value)
719	{
720	SvREFCNT_dec (key);
721	goto fail;	1031	goto fail;
		1032
		1033	decode_ws (dec); EXPECT_CH (':');
		1034
		1035	value = decode_sv (dec);
		1036	if (!value)
		1037	{
		1038	SvREFCNT_dec (key);
		1039	goto fail;
722	}	1040	}
723		1041
724	//TODO: optimise
725	hv_store_ent (hv, key, value, 0);	1042	hv_store_ent (hv, key, value, 0);
		1043	SvREFCNT_dec (key);
726		1044
727	WS;	1045	decode_ws (dec);
728		1046
729	if (*dec->cur == '}')	1047	if (*dec->cur == '}')
730	{	1048	{
731	++dec->cur;	1049	++dec->cur;
732	break;	1050	break;
733	}	1051	}
734		1052
735	if (*dec->cur != ',')	1053	if (*dec->cur != ',')
736	ERR (", or } expected while parsing object/hash");	1054	ERR (", or } expected while parsing object/hash");
737		1055
738	++dec->cur;	1056	++dec->cur;
739	}	1057	}
740		1058
		1059	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)hv);	1060	return newRV_noinc ((SV *)hv);
742		1061
743	fail:	1062	fail:
744	SvREFCNT_dec (hv);	1063	SvREFCNT_dec (hv);
		1064	DEC_DEC_DEPTH;
745	return 0;	1065	return 0;
746	}	1066	}
747		1067
748	static SV *	1068	static SV *
749	decode_sv (dec_t *dec)	1069	decode_sv (dec_t *dec)
750	{	1070	{
751	WS;	1071	decode_ws (dec);
		1072
		1073	// the beauty of JSON: you need exactly one character lookahead
		1074	// to parse anything.
752	switch (*dec->cur)	1075	switch (*dec->cur)
753	{	1076	{
754	case '"': ++dec->cur; return decode_str (dec);	1077	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1078	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1079	case '{': ++dec->cur; return decode_hv (dec);
…		…
762		1085
763	case 't':	1086	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1087	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1088	{
766	dec->cur += 4;	1089	dec->cur += 4;
767	return newSViv (1);	1090	return SvREFCNT_inc (json_true);
768	}	1091	}
769	else	1092	else
770	ERR ("'true' expected");	1093	ERR ("'true' expected");
771		1094
772	break;	1095	break;
773		1096
774	case 'f':	1097	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1098	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1099	{
777	dec->cur += 5;	1100	dec->cur += 5;
778	return newSViv (0);	1101	return SvREFCNT_inc (json_false);
779	}	1102	}
780	else	1103	else
781	ERR ("'false' expected");	1104	ERR ("'false' expected");
782		1105
783	break;	1106	break;
784		1107
785	case 'n':	1108	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1109	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1110	{
788	dec->cur += 4;	1111	dec->cur += 4;
789	return newSViv (1);	1112	return newSVsv (&PL_sv_undef);
790	}	1113	}
791	else	1114	else
792	ERR ("'null' expected");	1115	ERR ("'null' expected");
793		1116
794	break;	1117	break;
795		1118
796	default:	1119	default:
797	ERR ("malformed json string");	1120	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1121	break;
799	}	1122	}
800		1123
801	fail:	1124	fail:
802	return 0;	1125	return 0;
803	}	1126	}
804		1127
805	static SV *	1128	static SV *
806	decode_json (SV *string, UV flags)	1129	decode_json (SV *string, U32 flags, UV *offset_return)
807	{	1130	{
		1131	dec_t dec;
		1132	UV offset;
808	SV *sv;	1133	SV *sv;
809		1134
		1135	SvGETMAGIC (string);
		1136	SvUPGRADE (string, SVt_PV);
		1137
810	if (!(flags & F_UTF8))	1138	if (flags & F_UTF8)
		1139	sv_utf8_downgrade (string, 0);
		1140	else
811	sv_utf8_upgrade (string);	1141	sv_utf8_upgrade (string);
812		1142
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1143	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1144
815	dec_t dec;
816	dec.flags = flags;	1145	dec.flags = flags;
817	dec.cur = SvPVX (string);	1146	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1147	dec.end = SvEND (string);
819	dec.err = 0;	1148	dec.err = 0;
		1149	dec.depth = 0;
		1150	dec.maxdepth = DEC_DEPTH (dec.flags);
820		1151
821	*dec.end = 1; // invalid anywhere	1152	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
822	sv = decode_sv (&dec);	1153	sv = decode_sv (&dec);
823	*dec.end = 0;	1154
		1155	if (!(offset_return || !sv))
		1156	{
		1157	// check for trailing garbage
		1158	decode_ws (&dec);
		1159
		1160	if (*dec.cur)
		1161	{
		1162	dec.err = "garbage after JSON object";
		1163	SvREFCNT_dec (sv);
		1164	sv = 0;
		1165	}
		1166	}
		1167
		1168	if (offset_return || !sv)
		1169	{
		1170	offset = dec.flags & F_UTF8
		1171	? dec.cur - SvPVX (string)
		1172	: utf8_distance (dec.cur, SvPVX (string));
		1173
		1174	if (offset_return)
		1175	*offset_return = offset;
		1176	}
824		1177
825	if (!sv)	1178	if (!sv)
826	{	1179	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1180	SV *uni = sv_newmortal ();
829		1181
		1182	// horrible hack to silence warning inside pv_uni_display
		1183	COP cop = *PL_curcop;
		1184	cop.cop_warnings = pWARN_NONE;
		1185	ENTER;
		1186	SAVEVPTR (PL_curcop);
		1187	PL_curcop = &cop;
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1188	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1189	LEAVE;
		1190
831	croak ("%s, at character %d (%s)",	1191	croak ("%s, at character offset %d [\"%s\"]",
832	dec.err,	1192	dec.err,
833	(int)offset,	1193	(int)offset,
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1194	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1195	}
836		1196
837	sv = sv_2mortal (sv);	1197	sv = sv_2mortal (sv);
838		1198
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1199	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))
840	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1200	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1201
842	return sv;	1202	return sv;
843	}	1203	}
844		1204
		1205	/////////////////////////////////////////////////////////////////////////////
		1206	// XS interface functions
		1207
845	MODULE = JSON::XS PACKAGE = JSON::XS	1208	MODULE = JSON::XS PACKAGE = JSON::XS
846		1209
847	BOOT:	1210	BOOT:
848	{	1211	{
849	int i;	1212	int i;
850		1213
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1214	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1215	decode_hexdigit [i] =
		1216	i >= '0' && i <= '9' ? i - '0'
		1217	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1218	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1219	: -1;
854		1220
855	for (i = 7; i--; )
856	{
857	decode_hexdigit ['a' + i] = 10 + i;
858	decode_hexdigit ['A' + i] = 10 + i;
859	}
860
861	json_stash = gv_stashpv ("JSON::XS", 1);	1221	json_stash = gv_stashpv ("JSON::XS" , 1);
		1222	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1223
		1224	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1225	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
862	}	1226	}
863		1227
864	PROTOTYPES: DISABLE	1228	PROTOTYPES: DISABLE
865		1229
866	SV *new (char *dummy)	1230	SV *new (char *dummy)
867	CODE:	1231	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1232	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
869	OUTPUT:	1233	OUTPUT:
870	RETVAL	1234	RETVAL
871		1235
872	SV *ascii (SV *self, int enable)	1236	SV *ascii (SV *self, int enable = 1)
873	ALIAS:	1237	ALIAS:
874	ascii = F_ASCII	1238	ascii = F_ASCII
		1239	latin1 = F_LATIN1
875	utf8 = F_UTF8	1240	utf8 = F_UTF8
876	indent = F_INDENT	1241	indent = F_INDENT
877	canonical = F_CANONICAL	1242	canonical = F_CANONICAL
878	space_before = F_SPACE_BEFORE	1243	space_before = F_SPACE_BEFORE
879	space_after = F_SPACE_AFTER	1244	space_after = F_SPACE_AFTER
880	json_rpc = F_JSON_RPC
881	pretty = F_PRETTY	1245	pretty = F_PRETTY
882	allow_nonref = F_ALLOW_NONREF	1246	allow_nonref = F_ALLOW_NONREF
		1247	shrink = F_SHRINK
		1248	allow_blessed = F_ALLOW_BLESSED
		1249	convert_blessed = F_CONV_BLESSED
883	CODE:	1250	CODE:
884	{	1251	{
885	UV *uv = SvJSON (self);	1252	UV *uv = SvJSON (self);
886	if (enable)	1253	if (enable)
887	*uv |= ix;	1254	*uv |= ix;
…		…
891	RETVAL = newSVsv (self);	1258	RETVAL = newSVsv (self);
892	}	1259	}
893	OUTPUT:	1260	OUTPUT:
894	RETVAL	1261	RETVAL
895		1262
		1263	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1264	CODE:
		1265	{
		1266	UV *uv = SvJSON (self);
		1267	UV log2 = 0;
		1268
		1269	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1270
		1271	while ((1UL << log2) < max_depth)
		1272	++log2;
		1273
		1274	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1275
		1276	RETVAL = newSVsv (self);
		1277	}
		1278	OUTPUT:
		1279	RETVAL
		1280
896	void encode (SV *self, SV *scalar)	1281	void encode (SV *self, SV *scalar)
897	PPCODE:	1282	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1283	XPUSHs (encode_json (scalar, *SvJSON (self)));
899		1284
900	void decode (SV *self, SV *jsonstr)	1285	void decode (SV *self, SV *jsonstr)
901	PPCODE:	1286	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1287	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1288
		1289	void decode_prefix (SV *self, SV *jsonstr)
		1290	PPCODE:
		1291	{
		1292	UV offset;
		1293	EXTEND (SP, 2);
		1294	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1295	PUSHs (sv_2mortal (newSVuv (offset)));
		1296	}
903		1297
904	PROTOTYPES: ENABLE	1298	PROTOTYPES: ENABLE
905		1299
906	void to_json (SV *scalar)	1300	void to_json (SV *scalar)
		1301	ALIAS:
		1302	objToJson = 0
907	PPCODE:	1303	PPCODE:
908	XPUSHs (encode_json (scalar, F_UTF8));	1304	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
909		1305
910	void from_json (SV *jsonstr)	1306	void from_json (SV *jsonstr)
		1307	ALIAS:
		1308	jsonToObj = 0
911	PPCODE:	1309	PPCODE:
912	XPUSHs (decode_json (jsonstr, F_UTF8));	1310	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
913		1311

Diff Legend

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