[ViewVC] Diff of: cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.9 by root, Fri Mar 23 17:40:29 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 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
17	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35
		36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		38
19	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
20	#define F_DEFAULT 0	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		41
22	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		46
24	#define SB do {	47	#define SB do {
25	#define SE } while (0)	48	#define SE } while (0)
26		49
27	static HV *json_stash;	50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr),(value))
		52	# define inline inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define inline static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	static HV json_stash, json_boolean_stash; // JSON::XS::
		62	static SV json_true, json_false;
		63
		64	/////////////////////////////////////////////////////////////////////////////
		65	// utility functions
		66
		67	static UV *
		68	SvJSON (SV *sv)
		69	{
		70	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		71	croak ("object is not of type JSON::XS");
		72
		73	return &SvUVX (SvRV (sv));
		74	}
		75
		76	static void
		77	shrink (SV *sv)
		78	{
		79	sv_utf8_downgrade (sv, 1);
		80	if (SvLEN (sv) > SvCUR (sv) + 1)
		81	{
		82	#ifdef SvPV_shrink_to_cur
		83	SvPV_shrink_to_cur (sv);
		84	#elif defined (SvPV_renew)
		85	SvPV_renew (sv, SvCUR (sv) + 1);
		86	#endif
		87	}
		88	}
		89
		90	// decode an utf-8 character and return it, or (UV)-1 in
		91	// case of an error.
		92	// we special-case "safe" characters from U+80 .. U+7FF,
		93	// but use the very good perl function to parse anything else.
		94	// note that we never call this function for a ascii codepoints
		95	inline UV
		96	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
		97	{
		98	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
		99	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		100	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		101	{
		102	*clen = 2;
		103	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
		104	}
		105	else
		106	{
		107	*clen = (STRLEN)-1;
		108	return (UV)-1;
		109	}
		110	}
		111
		112	/////////////////////////////////////////////////////////////////////////////
		113	// encoder
28		114
29	// structure used for encoding JSON	115	// structure used for encoding JSON
30	typedef struct	116	typedef struct
31	{	117	{
32	char *cur;	118	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	119	char *end; // SvEND (sv)
35	SV *sv;	120	SV *sv; // result scalar
36	UV flags;	121	U32 flags; // F_*
37	int max_recurse;	122	U32 indent; // indentation level
38	int indent;	123	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	124	} enc_t;
40		125
41	// structure used for decoding JSON	126	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	127	need (enc_t *enc, STRLEN len)
72	{	128	{
73	if (enc->cur + len >= enc->end)	129	if (expect_false (enc->cur + len >= enc->end))
74	{	130	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	131	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	132	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	133	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	134	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	135	}
80	}	136	}
81		137
82	static void	138	inline void
83	encode_ch (enc_t *enc, char ch)	139	encode_ch (enc_t *enc, char ch)
84	{	140	{
85	need (enc, 1);	141	need (enc, 1);
86	*enc->cur++ = ch;	142	*enc->cur++ = ch;
87	}	143	}
…		…
95		151
96	while (str < end)	152	while (str < end)
97	{	153	{
98	unsigned char ch = (unsigned char )str;	154	unsigned char ch = (unsigned char )str;
99		155
100	if (ch >= 0x20 && ch < 0x80) // most common case	156	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	157	{
102	if (ch == '"') // but with slow exceptions	158	if (expect_false (ch == '"')) // but with slow exceptions
103	{	159	{
104	need (enc, len += 1);	160	need (enc, len += 1);
105	*enc->cur++ = '\\';	161	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	162	*enc->cur++ = '"';
107	}	163	}
108	else if (ch == '\\')	164	else if (expect_false (ch == '\\'))
109	{	165	{
110	need (enc, len += 1);	166	need (enc, len += 1);
111	*enc->cur++ = '\\';	167	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
113	}	169	}
…		…
131	STRLEN clen;	187	STRLEN clen;
132	UV uch;	188	UV uch;
133		189
134	if (is_utf8)	190	if (is_utf8)
135	{	191	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	192	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	193	if (clen == (STRLEN)-1)
138	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	194	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	195	}
140	else	196	else
141	{	197	{
…		…
144	}	200	}
145		201
146	if (uch > 0x10FFFFUL)	202	if (uch > 0x10FFFFUL)
147	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	203	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
148		204
149	if (uch < 0x80 \|\| enc->flags & F_ASCII)	205	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))
150	{	206	{
151	if (uch > 0xFFFFUL)	207	if (uch > 0xFFFFUL)
152	{	208	{
153	need (enc, len += 11);	209	need (enc, len += 11);
154	sprintf (enc->cur, "\\u%04x\\u%04x",	210	sprintf (enc->cur, "\\u%04x\\u%04x",
155	(uch - 0x10000) / 0x400 + 0xD800,	211	(int)((uch - 0x10000) / 0x400 + 0xD800),
156	(uch - 0x10000) % 0x400 + 0xDC00);	212	(int)((uch - 0x10000) % 0x400 + 0xDC00));
157	enc->cur += 12;	213	enc->cur += 12;
158	}	214	}
159	else	215	else
160	{	216	{
161	static char hexdigit [16] = "0123456789abcdef";	217	static char hexdigit [16] = "0123456789abcdef";
…		…
168	*enc->cur++ = hexdigit [(uch >> 0) & 15];	224	*enc->cur++ = hexdigit [(uch >> 0) & 15];
169	}	225	}
170		226
171	str += clen;	227	str += clen;
172	}	228	}
		229	else if (enc->flags & F_LATIN1)
		230	{
		231	*enc->cur++ = uch;
		232	str += clen;
		233	}
173	else if (is_utf8)	234	else if (is_utf8)
174	{	235	{
175	need (enc, len += clen);	236	need (enc, len += clen);
176	do	237	do
177	{	238	{
…		…
179	}	240	}
180	while (--clen);	241	while (--clen);
181	}	242	}
182	else	243	else
183	{	244	{
184	need (enc, len += 10); // never more than 11 bytes needed	245	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
185	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	246	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
186	++str;	247	++str;
187	}	248	}
188	}	249	}
189	}	250	}
…		…
191		252
192	--len;	253	--len;
193	}	254	}
194	}	255	}
195		256
196	#define INDENT SB \	257	inline void
		258	encode_indent (enc_t *enc)
		259	{
197	if (enc->flags & F_INDENT) \	260	if (enc->flags & F_INDENT)
198	{ \	261	{
199	int i_; \	262	int spaces = enc->indent * INDENT_STEP;
200	need (enc, enc->indent); \
201	for (i_ = enc->indent * 3; i_--; )\
202	encode_ch (enc, ' '); \
203	} \
204	SE
205		263
206	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	264	need (enc, spaces);
207	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	265	memset (enc->cur, ' ', spaces);
208	#define COMMA SB \	266	enc->cur += spaces;
		267	}
		268	}
		269
		270	inline void
		271	encode_space (enc_t *enc)
		272	{
		273	need (enc, 1);
209	encode_ch (enc, ','); \	274	encode_ch (enc, ' ');
		275	}
		276
		277	inline void
		278	encode_nl (enc_t *enc)
		279	{
210	if (enc->flags & F_INDENT) \	280	if (enc->flags & F_INDENT)
211	NL; \	281	{
		282	need (enc, 1);
		283	encode_ch (enc, '\n');
		284	}
		285	}
		286
		287	inline void
		288	encode_comma (enc_t *enc)
		289	{
		290	encode_ch (enc, ',');
		291
		292	if (enc->flags & F_INDENT)
		293	encode_nl (enc);
212	else if (enc->flags & F_SPACE_AFTER) \	294	else if (enc->flags & F_SPACE_AFTER)
213	SPACE; \	295	encode_space (enc);
214	SE	296	}
215		297
216	static void encode_sv (enc_t enc, SV sv);	298	static void encode_sv (enc_t enc, SV sv);
217		299
218	static void	300	static void
219	encode_av (enc_t enc, AV av)	301	encode_av (enc_t enc, AV av)
220	{	302	{
221	int i, len = av_len (av);	303	int i, len = av_len (av);
222		304
223	encode_ch (enc, '['); NL;	305	if (enc->indent >= enc->maxdepth)
		306	croak ("data structure too deep (hit recursion limit)");
		307
		308	encode_ch (enc, '['); encode_nl (enc);
224	++enc->indent;	309	++enc->indent;
225		310
226	for (i = 0; i <= len; ++i)	311	for (i = 0; i <= len; ++i)
227	{	312	{
228	INDENT;	313	encode_indent (enc);
229	encode_sv (enc, *av_fetch (av, i, 0));	314	encode_sv (enc, *av_fetch (av, i, 0));
230		315
231	if (i < len)	316	if (i < len)
232	COMMA;	317	encode_comma (enc);
233	}	318	}
234		319
235	NL;	320	encode_nl (enc);
236		321
237	--enc->indent;	322	--enc->indent;
238	INDENT; encode_ch (enc, ']');	323	encode_indent (enc); encode_ch (enc, ']');
239	}	324	}
240		325
241	static void	326	static void
242	encode_he (enc_t enc, HE he)	327	encode_he (enc_t enc, HE he)
243	{	328	{
…		…
257	else	342	else
258	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	343	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259		344
260	encode_ch (enc, '"');	345	encode_ch (enc, '"');
261		346
262	if (enc->flags & F_SPACE_BEFORE) SPACE;	347	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
263	encode_ch (enc, ':');	348	encode_ch (enc, ':');
264	if (enc->flags & F_SPACE_AFTER ) SPACE;	349	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
265	encode_sv (enc, HeVAL (he));	350	encode_sv (enc, HeVAL (he));
266	}	351	}
267		352
268	// compare hash entries, used when all keys are bytestrings	353	// compare hash entries, used when all keys are bytestrings
269	static int	354	static int
…		…
275	HE b = (HE **)b_;	360	HE b = (HE **)b_;
276		361
277	STRLEN la = HeKLEN (a);	362	STRLEN la = HeKLEN (a);
278	STRLEN lb = HeKLEN (b);	363	STRLEN lb = HeKLEN (b);
279		364
280	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	365	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
281	cmp = la < lb ? -1 : la == lb ? 0 : 1;	366	cmp = la - lb;
282		367
283	return cmp;	368	return cmp;
284	}	369	}
285		370
286	// compare hash entries, used when some keys are sv's or utf-x	371	// compare hash entries, used when some keys are sv's or utf-x
…		…
293	static void	378	static void
294	encode_hv (enc_t enc, HV hv)	379	encode_hv (enc_t enc, HV hv)
295	{	380	{
296	int count, i;	381	int count, i;
297		382
		383	if (enc->indent >= enc->maxdepth)
		384	croak ("data structure too deep (hit recursion limit)");
		385
298	encode_ch (enc, '{'); NL; ++enc->indent;	386	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
299		387
300	if ((count = hv_iterinit (hv)))	388	if ((count = hv_iterinit (hv)))
301	{	389	{
302	// for canonical output we have to sort by keys first	390	// for canonical output we have to sort by keys first
303	// actually, this is mostly due to the stupid so-called	391	// actually, this is mostly due to the stupid so-called
304	// security workaround added somewhere in 5.8.x.	392	// security workaround added somewhere in 5.8.x.
305	// that randomises hash orderings	393	// that randomises hash orderings
306	if (enc->flags & F_CANONICAL)	394	if (enc->flags & F_CANONICAL)
307	{	395	{
308	HE he, hes [count];
309	int fast = 1;	396	int fast = 1;
		397	HE *he;
		398	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		399	HE *hes = _alloca (count sizeof (HE));
		400	#else
		401	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		402	#endif
310		403
311	i = 0;	404	i = 0;
312	while ((he = hv_iternext (hv)))	405	while ((he = hv_iternext (hv)))
313	{	406	{
314	hes [i++] = he;	407	hes [i++] = he;
…		…
338	LEAVE;	431	LEAVE;
339	}	432	}
340		433
341	for (i = 0; i < count; ++i)	434	for (i = 0; i < count; ++i)
342	{	435	{
343	INDENT;	436	encode_indent (enc);
344	encode_he (enc, hes [i]);	437	encode_he (enc, hes [i]);
345		438
346	if (i < count - 1)	439	if (i < count - 1)
347	COMMA;	440	encode_comma (enc);
348	}
349
350	NL;	441	}
		442
		443	encode_nl (enc);
351	}	444	}
352	else	445	else
353	{	446	{
354	SV *sv;
355	HE *he = hv_iternext (hv);	447	HE *he = hv_iternext (hv);
356		448
357	for (;;)	449	for (;;)
358	{	450	{
359	INDENT;	451	encode_indent (enc);
360	encode_he (enc, he);	452	encode_he (enc, he);
361		453
362	if (!(he = hv_iternext (hv)))	454	if (!(he = hv_iternext (hv)))
363	break;	455	break;
364		456
365	COMMA;	457	encode_comma (enc);
366	}
367
368	NL;	458	}
		459
		460	encode_nl (enc);
369	}	461	}
370	}	462	}
371		463
372	--enc->indent; INDENT; encode_ch (enc, '}');	464	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		465	}
		466
		467	// encode objects, arrays and special \0=false and \1=true values.
		468	static void
		469	encode_rv (enc_t enc, SV sv)
		470	{
		471	svtype svt;
		472
		473	SvGETMAGIC (sv);
		474	svt = SvTYPE (sv);
		475
		476	if (expect_false (SvOBJECT (sv)))
		477	{
		478	if (SvSTASH (sv) == json_boolean_stash)
		479	{
		480	if (SvIV (sv) == 0)
		481	encode_str (enc, "false", 5, 0);
		482	else
		483	encode_str (enc, "true", 4, 0);
		484	}
		485	else
		486	{
		487	#if 0
		488	if (0 && sv_derived_from (rv, "JSON::Literal"))
		489	{
		490	// not yet
		491	}
		492	#endif
		493	if (enc->flags & F_CONV_BLESSED)
		494	{
		495	// we re-bless the reference to get overload and other niceties right
		496	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		497
		498	if (to_json)
		499	{
		500	dSP;
		501	ENTER;
		502	SAVETMPS;
		503	PUSHMARK (SP);
		504	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		505
		506	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		507	// check anyways.
		508	PUTBACK;
		509	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		510	SPAGAIN;
		511
		512	encode_sv (enc, POPs);
		513
		514	FREETMPS;
		515	LEAVE;
		516	}
		517	else if (enc->flags & F_ALLOW_BLESSED)
		518	encode_str (enc, "null", 4, 0);
		519	else
		520	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		521	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		522	}
		523	else if (enc->flags & F_ALLOW_BLESSED)
		524	encode_str (enc, "null", 4, 0);
		525	else
		526	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		527	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		528	}
		529	}
		530	else if (svt == SVt_PVHV)
		531	encode_hv (enc, (HV *)sv);
		532	else if (svt == SVt_PVAV)
		533	encode_av (enc, (AV *)sv);
		534	else if (svt < SVt_PVAV)
		535	{
		536	if (SvNIOK (sv) && SvIV (sv) == 0)
		537	encode_str (enc, "false", 5, 0);
		538	else if (SvNIOK (sv) && SvIV (sv) == 1)
		539	encode_str (enc, "true", 4, 0);
		540	else
		541	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else
		545	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
373	}	547	}
374		548
375	static void	549	static void
376	encode_sv (enc_t enc, SV sv)	550	encode_sv (enc_t enc, SV sv)
377	{	551	{
…		…
385	encode_str (enc, str, len, SvUTF8 (sv));	559	encode_str (enc, str, len, SvUTF8 (sv));
386	encode_ch (enc, '"');	560	encode_ch (enc, '"');
387	}	561	}
388	else if (SvNOKp (sv))	562	else if (SvNOKp (sv))
389	{	563	{
		564	// trust that perl will do the right thing w.r.t. JSON syntax.
390	need (enc, NV_DIG + 32);	565	need (enc, NV_DIG + 32);
391	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	566	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392	enc->cur += strlen (enc->cur);	567	enc->cur += strlen (enc->cur);
393	}	568	}
394	else if (SvIOKp (sv))	569	else if (SvIOKp (sv))
395	{	570	{
396	need (enc, 64);	571	// we assume we can always read an IV as a UV
		572	if (SvUV (sv) & ~(UV)0x7fff)
		573	{
		574	// large integer, use the (rather slow) snprintf way.
		575	need (enc, sizeof (UV) * 3);
397	enc->cur +=	576	enc->cur +=
398	SvIsUV(sv)	577	SvIsUV(sv)
399	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	578	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
400	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	579	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		580	}
		581	else
		582	{
		583	// optimise the "small number case"
		584	// code will likely be branchless and use only a single multiplication
		585	I32 i = SvIV (sv);
		586	U32 u;
		587	char digit, nz = 0;
		588
		589	need (enc, 6);
		590
		591	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		592	u = i < 0 ? -i : i;
		593
		594	// convert to 4.28 fixed-point representation
		595	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		596
		597	// now output digit by digit, each time masking out the integer part
		598	// and multiplying by 5 while moving the decimal point one to the right,
		599	// resulting in a net multiplication by 10.
		600	// we always write the digit to memory but conditionally increment
		601	// the pointer, to ease the usage of conditional move instructions.
		602	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		603	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		604	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		605	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		606	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		607	}
401	}	608	}
402	else if (SvROK (sv))	609	else if (SvROK (sv))
403	{	610	encode_rv (enc, SvRV (sv));
404	SV *rv = SvRV (sv);
405
406	if (!--enc->max_recurse)
407	croak ("data structure too deep (hit recursion limit)");
408
409	switch (SvTYPE (rv))
410	{
411	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
412	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
413
414	default:
415	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
416	SvPV_nolen (sv));
417	}
418	}
419	else if (!SvOK (sv))	611	else if (!SvOK (sv))
420	encode_str (enc, "null", 4, 0);	612	encode_str (enc, "null", 4, 0);
421	else	613	else
422	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	614	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
423	SvPV_nolen (sv), SvFLAGS (sv));	615	SvPV_nolen (sv), SvFLAGS (sv));
424	}	616	}
425		617
426	static SV *	618	static SV *
427	encode_json (SV *scalar, UV flags)	619	encode_json (SV *scalar, U32 flags)
428	{	620	{
		621	enc_t enc;
		622
429	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	623	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	624	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
431		625
432	enc_t enc;
433	enc.flags = flags;	626	enc.flags = flags;
434	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	627	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435	enc.cur = SvPVX (enc.sv);	628	enc.cur = SvPVX (enc.sv);
436	enc.end = SvEND (enc.sv);	629	enc.end = SvEND (enc.sv);
437	enc.max_recurse = 0;
438	enc.indent = 0;	630	enc.indent = 0;
		631	enc.maxdepth = DEC_DEPTH (flags);
439		632
440	SvPOK_only (enc.sv);	633	SvPOK_only (enc.sv);
441	encode_sv (&enc, scalar);	634	encode_sv (&enc, scalar);
442		635
		636	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		637	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		638
443	if (!(flags & (F_ASCII \| F_UTF8)))	639	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
444	SvUTF8_on (enc.sv);	640	SvUTF8_on (enc.sv);
445
446	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447		641
448	if (enc.flags & F_SHRINK)	642	if (enc.flags & F_SHRINK)
449	shrink (enc.sv);	643	shrink (enc.sv);
450		644
451	return enc.sv;	645	return enc.sv;
452	}	646	}
453		647
454	/////////////////////////////////////////////////////////////////////////////	648	/////////////////////////////////////////////////////////////////////////////
		649	// decoder
455		650
456	#define WS \	651	// structure used for decoding JSON
		652	typedef struct
		653	{
		654	char *cur; // current parser pointer
		655	char *end; // end of input string
		656	const char *err; // parse error, if != 0
		657	U32 flags; // F_*
		658	U32 depth; // recursion depth
		659	U32 maxdepth; // recursion depth limit
		660	} dec_t;
		661
		662	inline void
		663	decode_ws (dec_t *dec)
		664	{
457	for (;;) \	665	for (;;)
458	{ \	666	{
459	char ch = *dec->cur; \	667	char ch = *dec->cur;
		668
460	if (ch > 0x20 \	669	if (ch > 0x20
461	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	670	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
462	break; \	671	break;
		672
463	++dec->cur; \	673	++dec->cur;
464	}	674	}
		675	}
465		676
466	#define ERR(reason) SB dec->err = reason; goto fail; SE	677	#define ERR(reason) SB dec->err = reason; goto fail; SE
		678
467	#define EXPECT_CH(ch) SB \	679	#define EXPECT_CH(ch) SB \
468	if (*dec->cur != ch) \	680	if (*dec->cur != ch) \
469	ERR (# ch " expected"); \	681	ERR (# ch " expected"); \
470	++dec->cur; \	682	++dec->cur; \
471	SE	683	SE
472		684
		685	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		686	#define DEC_DEC_DEPTH --dec->depth
		687
473	static SV decode_sv (dec_t dec);	688	static SV decode_sv (dec_t dec);
474		689
475	static signed char decode_hexdigit[256];	690	static signed char decode_hexdigit[256];
476		691
477	static UV	692	static UV
478	decode_4hex (dec_t *dec)	693	decode_4hex (dec_t *dec)
479	{	694	{
480	signed char d1, d2, d3, d4;	695	signed char d1, d2, d3, d4;
		696	unsigned char cur = (unsigned char )dec->cur;
481		697
482	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	698	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
483	if (d1 < 0) ERR ("four hexadecimal digits expected");	699	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
484	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	700	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
485	if (d2 < 0) ERR ("four hexadecimal digits expected");	701	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
486	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
487	if (d3 < 0) ERR ("four hexadecimal digits expected");
488	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
489	if (d4 < 0) ERR ("four hexadecimal digits expected");
490		702
491	dec->cur += 4;	703	dec->cur += 4;
492		704
493	return ((UV)d1) << 12	705	return ((UV)d1) << 12
494	\| ((UV)d2) << 8	706	\| ((UV)d2) << 8
…		…
497		709
498	fail:	710	fail:
499	return (UV)-1;	711	return (UV)-1;
500	}	712	}
501		713
502	#define APPEND_GROW(n) SB \
503	if (cur + (n) >= end) \
504	{ \
505	STRLEN ofs = cur - SvPVX (sv); \
506	SvGROW (sv, ofs + (n) + 1); \
507	cur = SvPVX (sv) + ofs; \
508	end = SvEND (sv); \
509	} \
510	SE
511
512	#define APPEND_CH(ch) SB \
513	APPEND_GROW (1); \
514	*cur++ = (ch); \
515	SE
516
517	static SV *	714	static SV *
518	decode_str (dec_t *dec)	715	decode_str (dec_t *dec)
519	{	716	{
520	SV *sv = NEWSV (0,2);	717	SV *sv = 0;
521	int utf8 = 0;	718	int utf8 = 0;
522	char *cur = SvPVX (sv);	719	char *dec_cur = dec->cur;
523	char *end = SvEND (sv);
524		720
525	for (;;)	721	do
526	{	722	{
527	unsigned char ch = (unsigned char )dec->cur;	723	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		724	char *cur = buf;
528		725
529	if (ch == '"')	726	do
530	break;
531	else if (ch == '\\')
532	{	727	{
533	switch (*++dec->cur)	728	unsigned char ch = (unsigned char )dec_cur++;
		729
		730	if (expect_false (ch == '"'))
		731	{
		732	--dec_cur;
		733	break;
534	{	734	}
535	case '\\':	735	else if (expect_false (ch == '\\'))
536	case '/':	736	{
537	case '"': APPEND_CH (*dec->cur++); break;	737	switch (*dec_cur)
538
539	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
540	case 't': APPEND_CH ('\011'); ++dec->cur; break;
541	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
542	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
543	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
544
545	case 'u':
546	{	738	{
547	UV lo, hi;	739	case '\\':
548	++dec->cur;	740	case '/':
		741	case '"': cur++ = dec_cur++; break;
549		742
550	hi = decode_4hex (dec);	743	case 'b': ++dec_cur; *cur++ = '\010'; break;
551	if (hi == (UV)-1)	744	case 't': ++dec_cur; *cur++ = '\011'; break;
552	goto fail;	745	case 'n': ++dec_cur; *cur++ = '\012'; break;
		746	case 'f': ++dec_cur; *cur++ = '\014'; break;
		747	case 'r': ++dec_cur; *cur++ = '\015'; break;
553		748
554	// possibly a surrogate pair	749	case 'u':
555	if (hi >= 0xd800 && hi < 0xdc00)
556	{	750	{
557	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	751	UV lo, hi;
558	ERR ("missing low surrogate character in surrogate pair");	752	++dec_cur;
559		753
560	dec->cur += 2;	754	dec->cur = dec_cur;
561
562	lo = decode_4hex (dec);	755	hi = decode_4hex (dec);
		756	dec_cur = dec->cur;
563	if (lo == (UV)-1)	757	if (hi == (UV)-1)
564	goto fail;	758	goto fail;
565		759
		760	// possibly a surrogate pair
		761	if (hi >= 0xd800)
		762	if (hi < 0xdc00)
		763	{
		764	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
		765	ERR ("missing low surrogate character in surrogate pair");
		766
		767	dec_cur += 2;
		768
		769	dec->cur = dec_cur;
		770	lo = decode_4hex (dec);
		771	dec_cur = dec->cur;
		772	if (lo == (UV)-1)
		773	goto fail;
		774
566	if (lo < 0xdc00 \|\| lo >= 0xe000)	775	if (lo < 0xdc00 \|\| lo >= 0xe000)
567	ERR ("surrogate pair expected");	776	ERR ("surrogate pair expected");
568		777
569	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	778	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		779	}
		780	else if (hi < 0xe000)
		781	ERR ("missing high surrogate character in surrogate pair");
		782
		783	if (hi >= 0x80)
		784	{
		785	utf8 = 1;
		786
		787	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		788	}
		789	else
		790	*cur++ = hi;
570	}	791	}
571	else if (hi >= 0xdc00 && hi < 0xe000)
572	ERR ("missing high surrogate character in surrogate pair");
573
574	if (hi >= 0x80)
575	{	792	break;
576	utf8 = 1;
577		793
578	APPEND_GROW (4); // at most 4 bytes for 21 bits
579	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	}
581	else	794	default:
582	APPEND_CH (hi);	795	--dec_cur;
		796	ERR ("illegal backslash escape sequence in string");
583	}	797	}
584	break;	798	}
		799	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		800	*cur++ = ch;
		801	else if (ch >= 0x80)
		802	{
		803	STRLEN clen;
		804	UV uch;
585		805
586	default:
587	--dec->cur;	806	--dec_cur;
588	ERR ("illegal backslash escape sequence in string");	807
		808	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		809	if (clen == (STRLEN)-1)
		810	ERR ("malformed UTF-8 character in JSON string");
		811
		812	do
		813	cur++ = dec_cur++;
		814	while (--clen);
		815
		816	utf8 = 1;
		817	}
		818	else
		819	{
		820	--dec_cur;
		821
		822	if (!ch)
		823	ERR ("unexpected end of string while parsing JSON string");
		824	else
		825	ERR ("invalid character encountered while parsing JSON string");
589	}	826	}
590	}	827	}
591	else if (ch >= 0x20 && ch <= 0x7f)	828	while (cur < buf + SHORT_STRING_LEN);
592	APPEND_CH (*dec->cur++);	829
593	else if (ch >= 0x80)
594	{	830	{
595	STRLEN clen;	831	STRLEN len = cur - buf;
596	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
597	if (clen == (STRLEN)-1)
598	ERR ("malformed UTF-8 character in JSON string");
599		832
600	APPEND_GROW (clen);	833	if (sv)
601	do
602	{	834	{
603	cur++ = dec->cur++;	835	SvGROW (sv, SvCUR (sv) + len + 1);
		836	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		837	SvCUR_set (sv, SvCUR (sv) + len);
604	}	838	}
605	while (--clen);
606
607	utf8 = 1;
608	}
609	else if (dec->cur == dec->end)
610	ERR ("unexpected end of string while parsing json string");
611	else	839	else
612	ERR ("invalid character encountered");	840	sv = newSVpvn (buf, len);
613	}	841	}
		842	}
		843	while (*dec_cur != '"');
614		844
615	++dec->cur;	845	++dec_cur;
616		846
617	SvCUR_set (sv, cur - SvPVX (sv));	847	if (sv)
618		848	{
619	SvPOK_only (sv);	849	SvPOK_only (sv);
620	*SvEND (sv) = 0;	850	*SvEND (sv) = 0;
621		851
622	if (utf8)	852	if (utf8)
623	SvUTF8_on (sv);	853	SvUTF8_on (sv);
		854	}
		855	else
		856	sv = newSVpvn ("", 0);
624		857
625	if (dec->flags & F_SHRINK)	858	dec->cur = dec_cur;
626	shrink (sv);
627
628	return sv;	859	return sv;
629		860
630	fail:	861	fail:
631	SvREFCNT_dec (sv);	862	dec->cur = dec_cur;
632	return 0;	863	return 0;
633	}	864	}
634		865
635	static SV *	866	static SV *
636	decode_num (dec_t *dec)	867	decode_num (dec_t *dec)
…		…
694	is_nv = 1;	925	is_nv = 1;
695	}	926	}
696		927
697	if (!is_nv)	928	if (!is_nv)
698	{	929	{
699	UV uv;	930	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
700	int numtype = grok_number (start, dec->cur - start, &uv);	931	if (*start == '-')
701	if (numtype & IS_NUMBER_IN_UV)	932	switch (dec->cur - start)
702	if (numtype & IS_NUMBER_NEG)
703	{	933	{
704	if (uv < (UV)IV_MIN)	934	case 2: return newSViv (-( start [1] - '0' * 1));
705	return newSViv (-(IV)uv);	935	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		936	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		937	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
706	}	938	}
		939	else
		940	switch (dec->cur - start)
		941	{
		942	case 1: return newSViv ( start [0] - '0' * 1);
		943	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		944	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		945	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		946	}
		947
		948	{
		949	UV uv;
		950	int numtype = grok_number (start, dec->cur - start, &uv);
		951	if (numtype & IS_NUMBER_IN_UV)
		952	if (numtype & IS_NUMBER_NEG)
		953	{
		954	if (uv < (UV)IV_MIN)
		955	return newSViv (-(IV)uv);
		956	}
707	else	957	else
708	return newSVuv (uv);	958	return newSVuv (uv);
		959
		960	// here would likely be the place for bigint support
709	}	961	}
		962	}
710		963
		964	// if we ever support bigint or bigfloat, this is the place for bigfloat
711	return newSVnv (Atof (start));	965	return newSVnv (Atof (start));
712		966
713	fail:	967	fail:
714	return 0;	968	return 0;
715	}	969	}
…		…
717	static SV *	971	static SV *
718	decode_av (dec_t *dec)	972	decode_av (dec_t *dec)
719	{	973	{
720	AV *av = newAV ();	974	AV *av = newAV ();
721		975
722	WS;	976	DEC_INC_DEPTH;
		977	decode_ws (dec);
		978
723	if (*dec->cur == ']')	979	if (*dec->cur == ']')
724	++dec->cur;	980	++dec->cur;
725	else	981	else
726	for (;;)	982	for (;;)
727	{	983	{
…		…
731	if (!value)	987	if (!value)
732	goto fail;	988	goto fail;
733		989
734	av_push (av, value);	990	av_push (av, value);
735		991
736	WS;	992	decode_ws (dec);
737		993
738	if (*dec->cur == ']')	994	if (*dec->cur == ']')
739	{	995	{
740	++dec->cur;	996	++dec->cur;
741	break;	997	break;
…		…
745	ERR (", or ] expected while parsing array");	1001	ERR (", or ] expected while parsing array");
746		1002
747	++dec->cur;	1003	++dec->cur;
748	}	1004	}
749		1005
		1006	DEC_DEC_DEPTH;
750	return newRV_noinc ((SV *)av);	1007	return newRV_noinc ((SV *)av);
751		1008
752	fail:	1009	fail:
753	SvREFCNT_dec (av);	1010	SvREFCNT_dec (av);
		1011	DEC_DEC_DEPTH;
754	return 0;	1012	return 0;
755	}	1013	}
756		1014
757	static SV *	1015	static SV *
758	decode_hv (dec_t *dec)	1016	decode_hv (dec_t *dec)
759	{	1017	{
760	HV *hv = newHV ();	1018	HV *hv = newHV ();
761		1019
762	WS;	1020	DEC_INC_DEPTH;
		1021	decode_ws (dec);
		1022
763	if (*dec->cur == '}')	1023	if (*dec->cur == '}')
764	++dec->cur;	1024	++dec->cur;
765	else	1025	else
766	for (;;)	1026	for (;;)
767	{	1027	{
		1028	decode_ws (dec); EXPECT_CH ('"');
		1029
		1030	// heuristic: assume that
		1031	// a) decode_str + hv_store_ent are abysmally slow
		1032	// b) most hash keys are short, simple ascii text
		1033	// so try to "fast-match" such strings to avoid
		1034	// the overhead of hv_store_ent.
		1035	{
768	SV key, value;	1036	SV *value;
		1037	char *p = dec->cur;
		1038	char *e = p + 24; // only try up to 24 bytes
769		1039
770	WS; EXPECT_CH ('"');	1040	for (;;)
771
772	key = decode_str (dec);
773	if (!key)
774	goto fail;
775
776	WS; EXPECT_CH (':');
777
778	value = decode_sv (dec);
779	if (!value)
780	{	1041	{
		1042	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1043	{
		1044	// slow path, back up and use decode_str
		1045	SV *key = decode_str (dec);
		1046	if (!key)
		1047	goto fail;
		1048
		1049	decode_ws (dec); EXPECT_CH (':');
		1050
		1051	value = decode_sv (dec);
		1052	if (!value)
		1053	{
		1054	SvREFCNT_dec (key);
		1055	goto fail;
		1056	}
		1057
		1058	hv_store_ent (hv, key, value, 0);
781	SvREFCNT_dec (key);	1059	SvREFCNT_dec (key);
		1060
		1061	break;
		1062	}
		1063	else if (*p == '"')
		1064	{
		1065	// fast path, got a simple key
		1066	char *key = dec->cur;
		1067	int len = p - key;
		1068	dec->cur = p + 1;
		1069
		1070	decode_ws (dec); EXPECT_CH (':');
		1071
		1072	value = decode_sv (dec);
		1073	if (!value)
782	goto fail;	1074	goto fail;
		1075
		1076	hv_store (hv, key, len, value, 0);
		1077
		1078	break;
		1079	}
		1080
		1081	++p;
783	}	1082	}
784
785	//TODO: optimise
786	hv_store_ent (hv, key, value, 0);
787
788	WS;	1083	}
		1084
		1085	decode_ws (dec);
789		1086
790	if (*dec->cur == '}')	1087	if (*dec->cur == '}')
791	{	1088	{
792	++dec->cur;	1089	++dec->cur;
793	break;	1090	break;
…		…
797	ERR (", or } expected while parsing object/hash");	1094	ERR (", or } expected while parsing object/hash");
798		1095
799	++dec->cur;	1096	++dec->cur;
800	}	1097	}
801		1098
		1099	DEC_DEC_DEPTH;
802	return newRV_noinc ((SV *)hv);	1100	return newRV_noinc ((SV *)hv);
803		1101
804	fail:	1102	fail:
805	SvREFCNT_dec (hv);	1103	SvREFCNT_dec (hv);
		1104	DEC_DEC_DEPTH;
806	return 0;	1105	return 0;
807	}	1106	}
808		1107
809	static SV *	1108	static SV *
810	decode_sv (dec_t *dec)	1109	decode_sv (dec_t *dec)
811	{	1110	{
812	WS;	1111	decode_ws (dec);
		1112
		1113	// the beauty of JSON: you need exactly one character lookahead
		1114	// to parse anything.
813	switch (*dec->cur)	1115	switch (*dec->cur)
814	{	1116	{
815	case '"': ++dec->cur; return decode_str (dec);	1117	case '"': ++dec->cur; return decode_str (dec);
816	case '[': ++dec->cur; return decode_av (dec);	1118	case '[': ++dec->cur; return decode_av (dec);
817	case '{': ++dec->cur; return decode_hv (dec);	1119	case '{': ++dec->cur; return decode_hv (dec);
…		…
823		1125
824	case 't':	1126	case 't':
825	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1127	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826	{	1128	{
827	dec->cur += 4;	1129	dec->cur += 4;
828	return newSViv (1);	1130	return SvREFCNT_inc (json_true);
829	}	1131	}
830	else	1132	else
831	ERR ("'true' expected");	1133	ERR ("'true' expected");
832		1134
833	break;	1135	break;
834		1136
835	case 'f':	1137	case 'f':
836	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1138	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837	{	1139	{
838	dec->cur += 5;	1140	dec->cur += 5;
839	return newSViv (0);	1141	return SvREFCNT_inc (json_false);
840	}	1142	}
841	else	1143	else
842	ERR ("'false' expected");	1144	ERR ("'false' expected");
843		1145
844	break;	1146	break;
…		…
853	ERR ("'null' expected");	1155	ERR ("'null' expected");
854		1156
855	break;	1157	break;
856		1158
857	default:	1159	default:
858	ERR ("malformed json string, neither array, object, number, string or atom");	1160	ERR ("malformed JSON string, neither array, object, number, string or atom");
859	break;	1161	break;
860	}	1162	}
861		1163
862	fail:	1164	fail:
863	return 0;	1165	return 0;
864	}	1166	}
865		1167
866	static SV *	1168	static SV *
867	decode_json (SV *string, UV flags)	1169	decode_json (SV string, U32 flags, UV offset_return)
868	{	1170	{
		1171	dec_t dec;
		1172	UV offset;
869	SV *sv;	1173	SV *sv;
		1174
		1175	SvGETMAGIC (string);
		1176	SvUPGRADE (string, SVt_PV);
		1177
		1178	if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
		1179	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1180	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
870		1181
871	if (flags & F_UTF8)	1182	if (flags & F_UTF8)
872	sv_utf8_downgrade (string, 0);	1183	sv_utf8_downgrade (string, 0);
873	else	1184	else
874	sv_utf8_upgrade (string);	1185	sv_utf8_upgrade (string);
875		1186
876	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1187	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877		1188
878	dec_t dec;
879	dec.flags = flags;	1189	dec.flags = flags;
880	dec.cur = SvPVX (string);	1190	dec.cur = SvPVX (string);
881	dec.end = SvEND (string);	1191	dec.end = SvEND (string);
882	dec.err = 0;	1192	dec.err = 0;
		1193	dec.depth = 0;
		1194	dec.maxdepth = DEC_DEPTH (dec.flags);
883		1195
		1196	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
884	sv = decode_sv (&dec);	1197	sv = decode_sv (&dec);
885		1198
		1199	if (!(offset_return \|\| !sv))
		1200	{
		1201	// check for trailing garbage
		1202	decode_ws (&dec);
		1203
		1204	if (*dec.cur)
		1205	{
		1206	dec.err = "garbage after JSON object";
		1207	SvREFCNT_dec (sv);
		1208	sv = 0;
		1209	}
		1210	}
		1211
		1212	if (offset_return \|\| !sv)
		1213	{
		1214	offset = dec.flags & F_UTF8
		1215	? dec.cur - SvPVX (string)
		1216	: utf8_distance (dec.cur, SvPVX (string));
		1217
		1218	if (offset_return)
		1219	*offset_return = offset;
		1220	}
		1221
886	if (!sv)	1222	if (!sv)
887	{	1223	{
888	IV offset = dec.flags & F_UTF8
889	? dec.cur - SvPVX (string)
890	: utf8_distance (dec.cur, SvPVX (string));
891	SV *uni = sv_newmortal ();	1224	SV *uni = sv_newmortal ();
892		1225
893	// horrible hack to silence warning inside pv_uni_display	1226	// horrible hack to silence warning inside pv_uni_display
894	COP cop = *PL_curcop;	1227	COP cop = *PL_curcop;
895	cop.cop_warnings = pWARN_NONE;	1228	cop.cop_warnings = pWARN_NONE;
…		…
897	SAVEVPTR (PL_curcop);	1230	SAVEVPTR (PL_curcop);
898	PL_curcop = &cop;	1231	PL_curcop = &cop;
899	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1232	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
900	LEAVE;	1233	LEAVE;
901		1234
902	croak ("%s, at character offset %d (%s)",	1235	croak ("%s, at character offset %d [\"%s\"]",
903	dec.err,	1236	dec.err,
904	(int)offset,	1237	(int)offset,
905	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1238	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906	}	1239	}
907		1240
…		…
911	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1244	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
912		1245
913	return sv;	1246	return sv;
914	}	1247	}
915		1248
		1249	/////////////////////////////////////////////////////////////////////////////
		1250	// XS interface functions
		1251
916	MODULE = JSON::XS PACKAGE = JSON::XS	1252	MODULE = JSON::XS PACKAGE = JSON::XS
917		1253
918	BOOT:	1254	BOOT:
919	{	1255	{
920	int i;	1256	int i;
921		1257
922	memset (decode_hexdigit, 0xff, 256);
923	for (i = 10; i--; )	1258	for (i = 0; i < 256; ++i)
924	decode_hexdigit ['0' + i] = i;	1259	decode_hexdigit [i] =
		1260	i >= '0' && i <= '9' ? i - '0'
		1261	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1262	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1263	: -1;
925		1264
926	for (i = 7; i--; )
927	{
928	decode_hexdigit ['a' + i] = 10 + i;
929	decode_hexdigit ['A' + i] = 10 + i;
930	}
931
932	json_stash = gv_stashpv ("JSON::XS", 1);	1265	json_stash = gv_stashpv ("JSON::XS" , 1);
		1266	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1267
		1268	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1269	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
933	}	1270	}
934		1271
935	PROTOTYPES: DISABLE	1272	PROTOTYPES: DISABLE
936		1273
937	SV new (char dummy)	1274	SV new (char dummy)
…		…
940	OUTPUT:	1277	OUTPUT:
941	RETVAL	1278	RETVAL
942		1279
943	SV ascii (SV self, int enable = 1)	1280	SV ascii (SV self, int enable = 1)
944	ALIAS:	1281	ALIAS:
945	ascii = F_ASCII	1282	ascii = F_ASCII
		1283	latin1 = F_LATIN1
946	utf8 = F_UTF8	1284	utf8 = F_UTF8
947	indent = F_INDENT	1285	indent = F_INDENT
948	canonical = F_CANONICAL	1286	canonical = F_CANONICAL
949	space_before = F_SPACE_BEFORE	1287	space_before = F_SPACE_BEFORE
950	space_after = F_SPACE_AFTER	1288	space_after = F_SPACE_AFTER
951	json_rpc = F_JSON_RPC
952	pretty = F_PRETTY	1289	pretty = F_PRETTY
953	allow_nonref = F_ALLOW_NONREF	1290	allow_nonref = F_ALLOW_NONREF
954	shrink = F_SHRINK	1291	shrink = F_SHRINK
		1292	allow_blessed = F_ALLOW_BLESSED
		1293	convert_blessed = F_CONV_BLESSED
955	CODE:	1294	CODE:
956	{	1295	{
957	UV *uv = SvJSON (self);	1296	UV *uv = SvJSON (self);
958	if (enable)	1297	if (enable)
959	*uv \|= ix;	1298	*uv \|= ix;
…		…
963	RETVAL = newSVsv (self);	1302	RETVAL = newSVsv (self);
964	}	1303	}
965	OUTPUT:	1304	OUTPUT:
966	RETVAL	1305	RETVAL
967		1306
		1307	SV max_depth (SV self, UV max_depth = 0x80000000UL)
		1308	CODE:
		1309	{
		1310	UV *uv = SvJSON (self);
		1311	UV log2 = 0;
		1312
		1313	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1314
		1315	while ((1UL << log2) < max_depth)
		1316	++log2;
		1317
		1318	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1319
		1320	RETVAL = newSVsv (self);
		1321	}
		1322	OUTPUT:
		1323	RETVAL
		1324
		1325	SV max_size (SV self, UV max_size = 0)
		1326	CODE:
		1327	{
		1328	UV *uv = SvJSON (self);
		1329	UV log2 = 0;
		1330
		1331	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1332	if (max_size == 1) max_size = 2;
		1333
		1334	while ((1UL << log2) < max_size)
		1335	++log2;
		1336
		1337	uv = uv & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1338
		1339	RETVAL = newSVsv (self);
		1340	}
		1341	OUTPUT:
		1342	RETVAL
		1343
968	void encode (SV self, SV scalar)	1344	void encode (SV self, SV scalar)
969	PPCODE:	1345	PPCODE:
970	XPUSHs (encode_json (scalar, *SvJSON (self)));	1346	XPUSHs (encode_json (scalar, *SvJSON (self)));
971		1347
972	void decode (SV self, SV jsonstr)	1348	void decode (SV self, SV jsonstr)
973	PPCODE:	1349	PPCODE:
974	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1350	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1351
		1352	void decode_prefix (SV self, SV jsonstr)
		1353	PPCODE:
		1354	{
		1355	UV offset;
		1356	EXTEND (SP, 2);
		1357	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1358	PUSHs (sv_2mortal (newSVuv (offset)));
		1359	}
975		1360
976	PROTOTYPES: ENABLE	1361	PROTOTYPES: ENABLE
977		1362
978	void to_json (SV *scalar)	1363	void to_json (SV *scalar)
		1364	ALIAS:
		1365	objToJson = 0
979	PPCODE:	1366	PPCODE:
980	XPUSHs (encode_json (scalar, F_UTF8));	1367	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));
981		1368
982	void from_json (SV *jsonstr)	1369	void from_json (SV *jsonstr)
		1370	ALIAS:
		1371	jsonToObj = 0
983	PPCODE:	1372	PPCODE:
984	XPUSHs (decode_json (jsonstr, F_UTF8));	1373	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));
985		1374

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.9 by root, Fri Mar 23 17:40:29 2007 UTC vs. Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.9 by root, Fri Mar 23 17:40:29 2007 UTC vs.
Revision 1.46 by root, Mon Jun 25 22:11:39 2007 UTC