ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.10 by root, Sat Mar 24 01:15:22 2007 UTC vs.
Revision 1.49 by root, Sun Jul 1 23:40:07 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_ALLOW_NONREF 0x00000080	27	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		36
		37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		39
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	41	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		42
21	#define INIT_SIZE 32 // initial scalar size to be allocated	43	#define INIT_SIZE 32 // initial scalar size to be allocated
		44	#define INDENT_STEP 3 // spaces per indentation level
		45
		46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		47
23	#define SB do {	48	#define SB do {
24	#define SE } while (0)	49	#define SE } while (0)
25		50
26	static HV *json_stash;	51	#if __GNUC__ >= 3
		52	# define expect(expr,value) __builtin_expect ((expr),(value))
		53	# define inline inline
		54	#else
		55	# define expect(expr,value) (expr)
		56	# define inline static
		57	#endif
		58
		59	#define expect_false(expr) expect ((expr) != 0, 0)
		60	#define expect_true(expr) expect ((expr) != 0, 1)
		61
		62	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		63	static SV *json_true, *json_false;
		64
		65	typedef struct {
		66	U32 flags;
		67	SV *cb_object, *cb_sk_object;
		68	} JSON;
		69
		70	/////////////////////////////////////////////////////////////////////////////
		71	// utility functions
		72
		73	inline void
		74	shrink (SV *sv)
		75	{
		76	sv_utf8_downgrade (sv, 1);
		77	if (SvLEN (sv) > SvCUR (sv) + 1)
		78	{
		79	#ifdef SvPV_shrink_to_cur
		80	SvPV_shrink_to_cur (sv);
		81	#elif defined (SvPV_renew)
		82	SvPV_renew (sv, SvCUR (sv) + 1);
		83	#endif
		84	}
		85	}
		86
		87	// decode an utf-8 character and return it, or (UV)-1 in
		88	// case of an error.
		89	// we special-case "safe" characters from U+80 .. U+7FF,
		90	// but use the very good perl function to parse anything else.
		91	// note that we never call this function for a ascii codepoints
		92	inline UV
		93	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		94	{
		95	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		98	{
		99	*clen = 2;
		100	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		101	}
		102	else
		103	{
		104	*clen = (STRLEN)-1;
		105	return (UV)-1;
		106	}
		107	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// encoder
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	JSON json;
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	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	124	need (enc_t *enc, STRLEN len)
71	{	125	{
72	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
73	{	127	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	132	}
79	}	133	}
80		134
81	static void	135	inline void
82	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
83	{	137	{
84	need (enc, 1);	138	need (enc, 1);
85	*enc->cur++ = ch;	139	*enc->cur++ = ch;
86	}	140	}
…		…
94		148
95	while (str < end)	149	while (str < end)
96	{	150	{
97	unsigned char ch = *(unsigned char *)str;	151	unsigned char ch = *(unsigned char *)str;
98		152
99	if (ch >= 0x20 && ch < 0x80) // most common case	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	154	{
101	if (ch == '"') // but with slow exceptions	155	if (expect_false (ch == '"')) // but with slow exceptions
102	{	156	{
103	need (enc, len += 1);	157	need (enc, len += 1);
104	*enc->cur++ = '\\';	158	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	159	*enc->cur++ = '"';
106	}	160	}
107	else if (ch == '\\')	161	else if (expect_false (ch == '\\'))
108	{	162	{
109	need (enc, len += 1);	163	need (enc, len += 1);
110	*enc->cur++ = '\\';	164	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
112	}	166	}
…		…
130	STRLEN clen;	184	STRLEN clen;
131	UV uch;	185	UV uch;
132		186
133	if (is_utf8)	187	if (is_utf8)
134	{	188	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	189	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	190	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	191	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	192	}
139	else	193	else
140	{	194	{
…		…
143	}	197	}
144		198
145	if (uch > 0x10FFFFUL)	199	if (uch > 0x10FFFFUL)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147		201
148	if (uch < 0x80 || enc->flags & F_ASCII)	202	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149	{	203	{
150	if (uch > 0xFFFFUL)	204	if (uch > 0xFFFFUL)
151	{	205	{
152	need (enc, len += 11);	206	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	221	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	222	}
169		223
170	str += clen;	224	str += clen;
171	}	225	}
		226	else if (enc->json.flags & F_LATIN1)
		227	{
		228	*enc->cur++ = uch;
		229	str += clen;
		230	}
172	else if (is_utf8)	231	else if (is_utf8)
173	{	232	{
174	need (enc, len += clen);	233	need (enc, len += clen);
175	do	234	do
176	{	235	{
…		…
178	}	237	}
179	while (--clen);	238	while (--clen);
180	}	239	}
181	else	240	else
182	{	241	{
183	need (enc, len += 10); // never more than 11 bytes needed	242	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	243	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185	++str;	244	++str;
186	}	245	}
187	}	246	}
188	}	247	}
…		…
190		249
191	--len;	250	--len;
192	}	251	}
193	}	252	}
194		253
195	#define INDENT SB \	254	inline void
		255	encode_indent (enc_t *enc)
		256	{
196	if (enc->flags & F_INDENT) \	257	if (enc->json.flags & F_INDENT)
197	{ \	258	{
198	int i_; \	259	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	260
200	for (i_ = enc->indent * 3; i_--; )\	261	need (enc, spaces);
		262	memset (enc->cur, ' ', spaces);
		263	enc->cur += spaces;
		264	}
		265	}
		266
		267	inline void
		268	encode_space (enc_t *enc)
		269	{
		270	need (enc, 1);
		271	encode_ch (enc, ' ');
		272	}
		273
		274	inline void
		275	encode_nl (enc_t *enc)
		276	{
		277	if (enc->json.flags & F_INDENT)
		278	{
		279	need (enc, 1);
201	encode_ch (enc, ' '); \	280	encode_ch (enc, '\n');
202	} \	281	}
203	SE	282	}
204		283
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	284	inline void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	285	encode_comma (enc_t *enc)
207	#define COMMA SB \	286	{
208	encode_ch (enc, ','); \	287	encode_ch (enc, ',');
		288
209	if (enc->flags & F_INDENT) \	289	if (enc->json.flags & F_INDENT)
210	NL; \	290	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	291	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	292	encode_space (enc);
213	SE	293	}
214		294
215	static void encode_sv (enc_t *enc, SV *sv);	295	static void encode_sv (enc_t *enc, SV *sv);
216		296
217	static void	297	static void
218	encode_av (enc_t *enc, AV *av)	298	encode_av (enc_t *enc, AV *av)
219	{	299	{
220	int i, len = av_len (av);	300	int i, len = av_len (av);
221		301
222	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);
223	++enc->indent;	306	++enc->indent;
224		307
225	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
226	{	309	{
227	INDENT;	310	encode_indent (enc);
228	encode_sv (enc, *av_fetch (av, i, 0));	311	encode_sv (enc, *av_fetch (av, i, 0));
229		312
230	if (i < len)	313	if (i < len)
231	COMMA;	314	encode_comma (enc);
232	}	315	}
233		316
234	NL;	317	encode_nl (enc);
235		318
236	--enc->indent;	319	--enc->indent;
237	INDENT; encode_ch (enc, ']');	320	encode_indent (enc); encode_ch (enc, ']');
238	}	321	}
239		322
240	static void	323	static void
241	encode_he (enc_t *enc, HE *he)	324	encode_he (enc_t *enc, HE *he)
242	{	325	{
…		…
256	else	339	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	340	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		341
259	encode_ch (enc, '"');	342	encode_ch (enc, '"');
260		343
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	344	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	345	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	346	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));	347	encode_sv (enc, HeVAL (he));
265	}	348	}
266		349
267	// compare hash entries, used when all keys are bytestrings	350	// compare hash entries, used when all keys are bytestrings
268	static int	351	static int
…		…
274	HE *b = *(HE **)b_;	357	HE *b = *(HE **)b_;
275		358
276	STRLEN la = HeKLEN (a);	359	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	360	STRLEN lb = HeKLEN (b);
278		361
279	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
280	cmp = la < lb ? -1 : la == lb ? 0 : 1;	363	cmp = la - lb;
281		364
282	return cmp;	365	return cmp;
283	}	366	}
284		367
285	// 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
…		…
292	static void	375	static void
293	encode_hv (enc_t *enc, HV *hv)	376	encode_hv (enc_t *enc, HV *hv)
294	{	377	{
295	int count, i;	378	int count, i;
296		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
297	encode_ch (enc, '{'); NL; ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298		384
299	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
300	{	386	{
301	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	390	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	391	if (enc->json.flags & F_CANONICAL)
306	{	392	{
307	HE *he, *hes [count];
308	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
309		400
310	i = 0;	401	i = 0;
311	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
312	{	403	{
313	hes [i++] = he;	404	hes [i++] = he;
…		…
337	LEAVE;	428	LEAVE;
338	}	429	}
339		430
340	for (i = 0; i < count; ++i)	431	for (i = 0; i < count; ++i)
341	{	432	{
342	INDENT;	433	encode_indent (enc);
343	encode_he (enc, hes [i]);	434	encode_he (enc, hes [i]);
344		435
345	if (i < count - 1)	436	if (i < count - 1)
346	COMMA;	437	encode_comma (enc);
347	}	438	}
348		439
349	NL;	440	encode_nl (enc);
350	}	441	}
351	else	442	else
352	{	443	{
353	SV *sv;
354	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
355		445
356	for (;;)	446	for (;;)
357	{	447	{
358	INDENT;	448	encode_indent (enc);
359	encode_he (enc, he);	449	encode_he (enc, he);
360		450
361	if (!(he = hv_iternext (hv)))	451	if (!(he = hv_iternext (hv)))
362	break;	452	break;
363		453
364	COMMA;	454	encode_comma (enc);
365	}	455	}
366		456
367	NL;	457	encode_nl (enc);
368	}	458	}
369	}	459	}
370		460
371	--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->json.flags & F_CONV_BLESSED)
		491	{
		492	// we re-bless the reference to get overload and other niceties right
		493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		494
		495	if (to_json)
		496	{
		497	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
		498	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		499
		500	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		501	// check anyways.
		502	PUTBACK;
		503	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		504	SPAGAIN;
		505
		506	encode_sv (enc, POPs);
		507
		508	FREETMPS; LEAVE;
		509	}
		510	else if (enc->json.flags & F_ALLOW_BLESSED)
		511	encode_str (enc, "null", 4, 0);
		512	else
		513	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		514	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		515	}
		516	else if (enc->json.flags & F_ALLOW_BLESSED)
		517	encode_str (enc, "null", 4, 0);
		518	else
		519	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		520	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		521	}
		522	}
		523	else if (svt == SVt_PVHV)
		524	encode_hv (enc, (HV *)sv);
		525	else if (svt == SVt_PVAV)
		526	encode_av (enc, (AV *)sv);
		527	else if (svt < SVt_PVAV)
		528	{
		529	if (SvNIOK (sv) && SvIV (sv) == 0)
		530	encode_str (enc, "false", 5, 0);
		531	else if (SvNIOK (sv) && SvIV (sv) == 1)
		532	encode_str (enc, "true", 4, 0);
		533	else
		534	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		535	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		536	}
		537	else
		538	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		539	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	540	}
373		541
374	static void	542	static void
375	encode_sv (enc_t *enc, SV *sv)	543	encode_sv (enc_t *enc, SV *sv)
376	{	544	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	552	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	553	encode_ch (enc, '"');
386	}	554	}
387	else if (SvNOKp (sv))	555	else if (SvNOKp (sv))
388	{	556	{
		557	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	558	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	559	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	560	enc->cur += strlen (enc->cur);
392	}	561	}
393	else if (SvIOKp (sv))	562	else if (SvIOKp (sv))
394	{	563	{
395	need (enc, 64);	564	// we assume we can always read an IV as a UV
		565	if (SvUV (sv) & ~(UV)0x7fff)
		566	{
		567	// large integer, use the (rather slow) snprintf way.
		568	need (enc, sizeof (UV) * 3);
396	enc->cur +=	569	enc->cur +=
397	SvIsUV(sv)	570	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	571	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	572	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		573	}
		574	else
		575	{
		576	// optimise the "small number case"
		577	// code will likely be branchless and use only a single multiplication
		578	I32 i = SvIV (sv);
		579	U32 u;
		580	char digit, nz = 0;
		581
		582	need (enc, 6);
		583
		584	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		585	u = i < 0 ? -i : i;
		586
		587	// convert to 4.28 fixed-point representation
		588	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		589
		590	// now output digit by digit, each time masking out the integer part
		591	// and multiplying by 5 while moving the decimal point one to the right,
		592	// resulting in a net multiplication by 10.
		593	// we always write the digit to memory but conditionally increment
		594	// the pointer, to ease the usage of conditional move instructions.
		595	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		596	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		597	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		598	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		599	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		600	}
400	}	601	}
401	else if (SvROK (sv))	602	else if (SvROK (sv))
402	{	603	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))	604	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	605	encode_str (enc, "null", 4, 0);
420	else	606	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	607	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	608	SvPV_nolen (sv), SvFLAGS (sv));
423	}	609	}
424		610
425	static SV *	611	static SV *
426	encode_json (SV *scalar, UV flags)	612	encode_json (SV *scalar, JSON *json)
427	{	613	{
		614	enc_t enc;
		615
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	616	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	617	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		618
431	enc_t enc;	619	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	620	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	621	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	622	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	623	enc.indent = 0;
		624	enc.maxdepth = DEC_DEPTH (enc.json.flags);
438		625
439	SvPOK_only (enc.sv);	626	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	627	encode_sv (&enc, scalar);
441		628
		629	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		630	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		631
442	if (!(flags & (F_ASCII | F_UTF8)))	632	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	633	SvUTF8_on (enc.sv);
444		634
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	635	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	636	shrink (enc.sv);
449		637
450	return enc.sv;	638	return enc.sv;
451	}	639	}
452		640
453	/////////////////////////////////////////////////////////////////////////////	641	/////////////////////////////////////////////////////////////////////////////
		642	// decoder
454		643
455	#define WS \	644	// structure used for decoding JSON
		645	typedef struct
		646	{
		647	char *cur; // current parser pointer
		648	char *end; // end of input string
		649	const char *err; // parse error, if != 0
		650	JSON json;
		651	U32 depth; // recursion depth
		652	U32 maxdepth; // recursion depth limit
		653	} dec_t;
		654
		655	inline void
		656	decode_ws (dec_t *dec)
		657	{
456	for (;;) \	658	for (;;)
457	{ \	659	{
458	char ch = *dec->cur; \	660	char ch = *dec->cur;
		661
459	if (ch > 0x20 \	662	if (ch > 0x20
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	663	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461	break; \	664	break;
		665
462	++dec->cur; \	666	++dec->cur;
463	}	667	}
		668	}
464		669
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	670	#define ERR(reason) SB dec->err = reason; goto fail; SE
		671
466	#define EXPECT_CH(ch) SB \	672	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	673	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	674	ERR (# ch " expected"); \
469	++dec->cur; \	675	++dec->cur; \
470	SE	676	SE
471		677
		678	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		679	#define DEC_DEC_DEPTH --dec->depth
		680
472	static SV *decode_sv (dec_t *dec);	681	static SV *decode_sv (dec_t *dec);
473		682
474	static signed char decode_hexdigit[256];	683	static signed char decode_hexdigit[256];
475		684
476	static UV	685	static UV
477	decode_4hex (dec_t *dec)	686	decode_4hex (dec_t *dec)
478	{	687	{
479	signed char d1, d2, d3, d4;	688	signed char d1, d2, d3, d4;
		689	unsigned char *cur = (unsigned char *)dec->cur;
480		690
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	691	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	692	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	693	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	694	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		695
490	dec->cur += 4;	696	dec->cur += 4;
491		697
492	return ((UV)d1) << 12	698	return ((UV)d1) << 12
493	| ((UV)d2) << 8	699	| ((UV)d2) << 8
…		…
496		702
497	fail:	703	fail:
498	return (UV)-1;	704	return (UV)-1;
499	}	705	}
500		706
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	707	static SV *
517	decode_str (dec_t *dec)	708	decode_str (dec_t *dec)
518	{	709	{
519	SV *sv = NEWSV (0,2);	710	SV *sv = 0;
520	int utf8 = 0;	711	int utf8 = 0;
521	char *cur = SvPVX (sv);	712	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		713
524	for (;;)	714	do
525	{	715	{
526	unsigned char ch = *(unsigned char *)dec->cur;	716	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		717	char *cur = buf;
527		718
528	if (ch == '"')	719	do
529	break;
530	else if (ch == '\\')
531	{	720	{
532	switch (*++dec->cur)	721	unsigned char ch = *(unsigned char *)dec_cur++;
		722
		723	if (expect_false (ch == '"'))
533	{	724	{
534	case '\\':	725	--dec_cur;
535	case '/':	726	break;
536	case '"': APPEND_CH (*dec->cur++); break;	727	}
537		728	else if (expect_false (ch == '\\'))
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	729	{
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;	730	switch (*dec_cur)
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	731	{
546	UV lo, hi;	732	case '\\':
547	++dec->cur;	733	case '/':
		734	case '"': *cur++ = *dec_cur++; break;
548		735
549	hi = decode_4hex (dec);	736	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	737	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	738	case 'n': ++dec_cur; *cur++ = '\012'; break;
		739	case 'f': ++dec_cur; *cur++ = '\014'; break;
		740	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		741
553	// possibly a surrogate pair	742	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	743	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	744	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	745	++dec_cur;
558		746
559	dec->cur += 2;	747	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	748	hi = decode_4hex (dec);
		749	dec_cur = dec->cur;
562	if (lo == (UV)-1)	750	if (hi == (UV)-1)
563	goto fail;	751	goto fail;
564		752
		753	// possibly a surrogate pair
		754	if (hi >= 0xd800)
		755	if (hi < 0xdc00)
		756	{
		757	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		758	ERR ("missing low surrogate character in surrogate pair");
		759
		760	dec_cur += 2;
		761
		762	dec->cur = dec_cur;
		763	lo = decode_4hex (dec);
		764	dec_cur = dec->cur;
		765	if (lo == (UV)-1)
		766	goto fail;
		767
565	if (lo < 0xdc00 || lo >= 0xe000)	768	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	769	ERR ("surrogate pair expected");
567		770
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	771	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		772	}
		773	else if (hi < 0xe000)
		774	ERR ("missing high surrogate character in surrogate pair");
		775
		776	if (hi >= 0x80)
		777	{
		778	utf8 = 1;
		779
		780	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		781	}
		782	else
		783	*cur++ = hi;
569	}	784	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	785	break;
575	utf8 = 1;
576		786
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	787	default:
581	APPEND_CH (hi);	788	--dec_cur;
		789	ERR ("illegal backslash escape sequence in string");
582	}	790	}
583	break;	791	}
		792	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		793	*cur++ = ch;
		794	else if (ch >= 0x80)
		795	{
		796	STRLEN clen;
		797	UV uch;
584		798
585	default:
586	--dec->cur;	799	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	800
		801	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		802	if (clen == (STRLEN)-1)
		803	ERR ("malformed UTF-8 character in JSON string");
		804
		805	do
		806	*cur++ = *dec_cur++;
		807	while (--clen);
		808
		809	utf8 = 1;
		810	}
		811	else
		812	{
		813	--dec_cur;
		814
		815	if (!ch)
		816	ERR ("unexpected end of string while parsing JSON string");
		817	else
		818	ERR ("invalid character encountered while parsing JSON string");
588	}	819	}
589	}	820	}
590	else if (ch >= 0x20 && ch <= 0x7f)	821	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	822
592	else if (ch >= 0x80)
593	{	823	{
594	STRLEN clen;	824	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		825
599	APPEND_GROW (clen);	826	if (sv)
600	do
601	{	827	{
602	*cur++ = *dec->cur++;	828	SvGROW (sv, SvCUR (sv) + len + 1);
		829	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		830	SvCUR_set (sv, SvCUR (sv) + len);
603	}	831	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	832	else
611	ERR ("invalid character encountered");	833	sv = newSVpvn (buf, len);
612	}	834	}
		835	}
		836	while (*dec_cur != '"');
613		837
614	++dec->cur;	838	++dec_cur;
615		839
616	SvCUR_set (sv, cur - SvPVX (sv));	840	if (sv)
617		841	{
618	SvPOK_only (sv);	842	SvPOK_only (sv);
619	*SvEND (sv) = 0;	843	*SvEND (sv) = 0;
620		844
621	if (utf8)	845	if (utf8)
622	SvUTF8_on (sv);	846	SvUTF8_on (sv);
		847	}
		848	else
		849	sv = newSVpvn ("", 0);
623		850
624	if (dec->flags & F_SHRINK)	851	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	852	return sv;
628		853
629	fail:	854	fail:
630	SvREFCNT_dec (sv);	855	dec->cur = dec_cur;
631	return 0;	856	return 0;
632	}	857	}
633		858
634	static SV *	859	static SV *
635	decode_num (dec_t *dec)	860	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	918	is_nv = 1;
694	}	919	}
695		920
696	if (!is_nv)	921	if (!is_nv)
697	{	922	{
698	UV uv;	923	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
699	int numtype = grok_number (start, dec->cur - start, &uv);	924	if (*start == '-')
700	if (numtype & IS_NUMBER_IN_UV)	925	switch (dec->cur - start)
701	if (numtype & IS_NUMBER_NEG)
702	{	926	{
703	if (uv < (UV)IV_MIN)	927	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	928	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		929	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		930	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705	}	931	}
		932	else
		933	switch (dec->cur - start)
		934	{
		935	case 1: return newSViv ( start [0] - '0' * 1);
		936	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		937	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		938	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		939	}
		940
		941	{
		942	UV uv;
		943	int numtype = grok_number (start, dec->cur - start, &uv);
		944	if (numtype & IS_NUMBER_IN_UV)
		945	if (numtype & IS_NUMBER_NEG)
		946	{
		947	if (uv < (UV)IV_MIN)
		948	return newSViv (-(IV)uv);
		949	}
706	else	950	else
707	return newSVuv (uv);	951	return newSVuv (uv);
		952
		953	// here would likely be the place for bigint support
708	}	954	}
		955	}
709		956
		957	// if we ever support bigint or bigfloat, this is the place for bigfloat
710	return newSVnv (Atof (start));	958	return newSVnv (Atof (start));
711		959
712	fail:	960	fail:
713	return 0;	961	return 0;
714	}	962	}
…		…
716	static SV *	964	static SV *
717	decode_av (dec_t *dec)	965	decode_av (dec_t *dec)
718	{	966	{
719	AV *av = newAV ();	967	AV *av = newAV ();
720		968
721	WS;	969	DEC_INC_DEPTH;
		970	decode_ws (dec);
		971
722	if (*dec->cur == ']')	972	if (*dec->cur == ']')
723	++dec->cur;	973	++dec->cur;
724	else	974	else
725	for (;;)	975	for (;;)
726	{	976	{
…		…
730	if (!value)	980	if (!value)
731	goto fail;	981	goto fail;
732		982
733	av_push (av, value);	983	av_push (av, value);
734		984
735	WS;	985	decode_ws (dec);
736		986
737	if (*dec->cur == ']')	987	if (*dec->cur == ']')
738	{	988	{
739	++dec->cur;	989	++dec->cur;
740	break;	990	break;
…		…
744	ERR (", or ] expected while parsing array");	994	ERR (", or ] expected while parsing array");
745		995
746	++dec->cur;	996	++dec->cur;
747	}	997	}
748		998
		999	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1000	return newRV_noinc ((SV *)av);
750		1001
751	fail:	1002	fail:
752	SvREFCNT_dec (av);	1003	SvREFCNT_dec (av);
		1004	DEC_DEC_DEPTH;
753	return 0;	1005	return 0;
754	}	1006	}
755		1007
756	static SV *	1008	static SV *
757	decode_hv (dec_t *dec)	1009	decode_hv (dec_t *dec)
758	{	1010	{
		1011	SV *sv;
759	HV *hv = newHV ();	1012	HV *hv = newHV ();
760		1013
761	WS;	1014	DEC_INC_DEPTH;
		1015	decode_ws (dec);
		1016
762	if (*dec->cur == '}')	1017	if (*dec->cur == '}')
763	++dec->cur;	1018	++dec->cur;
764	else	1019	else
765	for (;;)	1020	for (;;)
766	{	1021	{
		1022	decode_ws (dec); EXPECT_CH ('"');
		1023
		1024	// heuristic: assume that
		1025	// a) decode_str + hv_store_ent are abysmally slow.
		1026	// b) most hash keys are short, simple ascii text.
		1027	// => try to "fast-match" such strings to avoid
		1028	// the overhead of decode_str + hv_store_ent.
		1029	{
767	SV *key, *value;	1030	SV *value;
		1031	char *p = dec->cur;
		1032	char *e = p + 24; // only try up to 24 bytes
768		1033
769	WS; EXPECT_CH ('"');	1034	for (;;)
770
771	key = decode_str (dec);
772	if (!key)
773	goto fail;
774
775	WS; EXPECT_CH (':');
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1035	{
		1036	// the >= 0x80 is true on most architectures
		1037	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1038	{
		1039	// slow path, back up and use decode_str
		1040	SV *key = decode_str (dec);
		1041	if (!key)
		1042	goto fail;
		1043
		1044	decode_ws (dec); EXPECT_CH (':');
		1045
		1046	value = decode_sv (dec);
		1047	if (!value)
		1048	{
		1049	SvREFCNT_dec (key);
		1050	goto fail;
		1051	}
		1052
		1053	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1054	SvREFCNT_dec (key);
		1055
		1056	break;
		1057	}
		1058	else if (*p == '"')
		1059	{
		1060	// fast path, got a simple key
		1061	char *key = dec->cur;
		1062	int len = p - key;
		1063	dec->cur = p + 1;
		1064
		1065	decode_ws (dec); EXPECT_CH (':');
		1066
		1067	value = decode_sv (dec);
		1068	if (!value)
781	goto fail;	1069	goto fail;
		1070
		1071	hv_store (hv, key, len, value, 0);
		1072
		1073	break;
		1074	}
		1075
		1076	++p;
782	}	1077	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1078	}
		1079
		1080	decode_ws (dec);
788		1081
789	if (*dec->cur == '}')	1082	if (*dec->cur == '}')
790	{	1083	{
791	++dec->cur;	1084	++dec->cur;
792	break;	1085	break;
…		…
796	ERR (", or } expected while parsing object/hash");	1089	ERR (", or } expected while parsing object/hash");
797		1090
798	++dec->cur;	1091	++dec->cur;
799	}	1092	}
800		1093
		1094	DEC_DEC_DEPTH;
		1095	sv = newRV_noinc ((SV *)hv);
		1096
		1097	// check filter callbacks
		1098	if (dec->json.flags & F_HOOK)
		1099	{
		1100	ENTER; SAVETMPS;
		1101
		1102	if (HvKEYS (hv) == 1 && dec->json.cb_sk_object)
		1103	{
		1104	int count;
		1105
		1106	dSP; PUSHMARK (SP);
		1107	XPUSHs (sv_2mortal (sv));
		1108
		1109	PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN;
		1110
		1111	if (count == 1)
		1112	sv = newSVsv (POPs);
		1113	else
		1114	SvREFCNT_inc (sv);
		1115	}
		1116
		1117	if (dec->json.cb_object)
		1118	{
		1119	int count;
		1120
		1121	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
		1122	XPUSHs (sv_2mortal (sv));
		1123
		1124	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1125
		1126	if (count == 1)
		1127	sv = newSVsv (POPs);
		1128	else
		1129	SvREFCNT_inc (sv);
		1130	}
		1131
		1132	FREETMPS; LEAVE;
		1133	}
		1134
801	return newRV_noinc ((SV *)hv);	1135	return newRV_noinc ((SV *)hv);
802		1136
803	fail:	1137	fail:
804	SvREFCNT_dec (hv);	1138	SvREFCNT_dec (hv);
		1139	DEC_DEC_DEPTH;
805	return 0;	1140	return 0;
806	}	1141	}
807		1142
808	static SV *	1143	static SV *
809	decode_sv (dec_t *dec)	1144	decode_sv (dec_t *dec)
810	{	1145	{
811	WS;	1146	decode_ws (dec);
		1147
		1148	// the beauty of JSON: you need exactly one character lookahead
		1149	// to parse anything.
812	switch (*dec->cur)	1150	switch (*dec->cur)
813	{	1151	{
814	case '"': ++dec->cur; return decode_str (dec);	1152	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1153	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1154	case '{': ++dec->cur; return decode_hv (dec);
…		…
822		1160
823	case 't':	1161	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1162	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1163	{
826	dec->cur += 4;	1164	dec->cur += 4;
827	return newSViv (1);	1165	return SvREFCNT_inc (json_true);
828	}	1166	}
829	else	1167	else
830	ERR ("'true' expected");	1168	ERR ("'true' expected");
831		1169
832	break;	1170	break;
833		1171
834	case 'f':	1172	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1173	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1174	{
837	dec->cur += 5;	1175	dec->cur += 5;
838	return newSViv (0);	1176	return SvREFCNT_inc (json_false);
839	}	1177	}
840	else	1178	else
841	ERR ("'false' expected");	1179	ERR ("'false' expected");
842		1180
843	break;	1181	break;
…		…
852	ERR ("'null' expected");	1190	ERR ("'null' expected");
853		1191
854	break;	1192	break;
855		1193
856	default:	1194	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1195	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1196	break;
859	}	1197	}
860		1198
861	fail:	1199	fail:
862	return 0;	1200	return 0;
863	}	1201	}
864		1202
865	static SV *	1203	static SV *
866	decode_json (SV *string, UV flags)	1204	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1205	{
		1206	dec_t dec;
		1207	UV offset;
868	SV *sv;	1208	SV *sv;
869		1209
		1210	SvGETMAGIC (string);
		1211	SvUPGRADE (string, SVt_PV);
		1212
		1213	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1214	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1215	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1216
870	if (flags & F_UTF8)	1217	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1218	sv_utf8_downgrade (string, 0);
872	else	1219	else
873	sv_utf8_upgrade (string);	1220	sv_utf8_upgrade (string);
874		1221
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1222	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1223
877	dec_t dec;	1224	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1225	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1226	dec.end = SvEND (string);
881	dec.err = 0;	1227	dec.err = 0;
		1228	dec.depth = 0;
		1229	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1230
		1231	if (dec.json.cb_object || dec.json.cb_sk_object)
		1232	dec.json.flags |= F_HOOK;
		1233
		1234	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
883	sv = decode_sv (&dec);	1235	sv = decode_sv (&dec);
884		1236
		1237	if (!(offset_return || !sv))
		1238	{
		1239	// check for trailing garbage
		1240	decode_ws (&dec);
		1241
		1242	if (*dec.cur)
		1243	{
		1244	dec.err = "garbage after JSON object";
		1245	SvREFCNT_dec (sv);
		1246	sv = 0;
		1247	}
		1248	}
		1249
		1250	if (offset_return || !sv)
		1251	{
		1252	offset = dec.json.flags & F_UTF8
		1253	? dec.cur - SvPVX (string)
		1254	: utf8_distance (dec.cur, SvPVX (string));
		1255
		1256	if (offset_return)
		1257	*offset_return = offset;
		1258	}
		1259
885	if (!sv)	1260	if (!sv)
886	{	1261	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1262	SV *uni = sv_newmortal ();
891		1263
892	// horrible hack to silence warning inside pv_uni_display	1264	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1265	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1266	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1268	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1269	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1270	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1271	LEAVE;
900		1272
901	croak ("%s, at character offset %d (%s)",	1273	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1274	dec.err,
903	(int)offset,	1275	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1276	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1277	}
906		1278
907	sv = sv_2mortal (sv);	1279	sv = sv_2mortal (sv);
908		1280
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1281	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1282	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1283
912	return sv;	1284	return sv;
913	}	1285	}
914		1286
		1287	/////////////////////////////////////////////////////////////////////////////
		1288	// XS interface functions
		1289
915	MODULE = JSON::XS PACKAGE = JSON::XS	1290	MODULE = JSON::XS PACKAGE = JSON::XS
916		1291
917	BOOT:	1292	BOOT:
918	{	1293	{
919	int i;	1294	int i;
920		1295
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1296	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1297	decode_hexdigit [i] =
		1298	i >= '0' && i <= '9' ? i - '0'
		1299	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1300	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1301	: -1;
924		1302
925	for (i = 7; i--; )	1303	json_stash = gv_stashpv ("JSON::XS" , 1);
		1304	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1305
		1306	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1307	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1308	}
		1309
		1310	PROTOTYPES: DISABLE
		1311
		1312	void new (char *klass)
		1313	PPCODE:
		1314	{
		1315	SV *pv = NEWSV (0, sizeof (JSON));
		1316	SvPOK_only (pv);
		1317	Zero (SvPVX (pv), 1, JSON);
		1318	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1319	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1320	}
		1321
		1322	void ascii (JSON *self, int enable = 1)
		1323	ALIAS:
		1324	ascii = F_ASCII
		1325	latin1 = F_LATIN1
		1326	utf8 = F_UTF8
		1327	indent = F_INDENT
		1328	canonical = F_CANONICAL
		1329	space_before = F_SPACE_BEFORE
		1330	space_after = F_SPACE_AFTER
		1331	pretty = F_PRETTY
		1332	allow_nonref = F_ALLOW_NONREF
		1333	shrink = F_SHRINK
		1334	allow_blessed = F_ALLOW_BLESSED
		1335	convert_blessed = F_CONV_BLESSED
		1336	PPCODE:
		1337	{
		1338	if (enable)
		1339	self->flags |= ix;
		1340	else
		1341	self->flags &= ~ix;
		1342
		1343	XPUSHs (ST (0));
		1344	}
		1345
		1346	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1347	PPCODE:
		1348	{
		1349	UV log2 = 0;
		1350
		1351	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1352
		1353	while ((1UL << log2) < max_depth)
		1354	++log2;
		1355
		1356	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1357
		1358	XPUSHs (ST (0));
		1359	}
		1360
		1361	void max_size (JSON *self, UV max_size = 0)
		1362	PPCODE:
		1363	{
		1364	UV log2 = 0;
		1365
		1366	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1367	if (max_size == 1) max_size = 2;
		1368
		1369	while ((1UL << log2) < max_size)
		1370	++log2;
		1371
		1372	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1373
		1374	XPUSHs (ST (0));
		1375	}
		1376
		1377	void filter_json_objects (JSON *self, SV *cb = &PL_sv_undef)
		1378	ALIAS:
		1379	filter_sk_json_objects = 1
		1380	PPCODE:
		1381	{
		1382	if (!SvOK (cb))
		1383	cb = 0;
		1384
		1385	switch (ix)
926	{	1386	{
927	decode_hexdigit ['a' + i] = 10 + i;	1387	case 0: self->cb_object = cb; break;
928	decode_hexdigit ['A' + i] = 10 + i;	1388	case 1: self->cb_sk_object = cb; break;
929	}	1389	}
930		1390
931	json_stash = gv_stashpv ("JSON::XS", 1);	1391	XPUSHs (ST (0));
932	}	1392	}
933		1393
934	PROTOTYPES: DISABLE
935
936	SV *new (char *dummy)
937	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
939	OUTPUT:
940	RETVAL
941
942	SV *ascii (SV *self, int enable = 1)
943	ALIAS:
944	ascii = F_ASCII
945	utf8 = F_UTF8
946	indent = F_INDENT
947	canonical = F_CANONICAL
948	space_before = F_SPACE_BEFORE
949	space_after = F_SPACE_AFTER
950	pretty = F_PRETTY
951	allow_nonref = F_ALLOW_NONREF
952	shrink = F_SHRINK
953	CODE:
954	{
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:
964	RETVAL
965
966	void encode (SV *self, SV *scalar)	1394	void encode (JSON *self, SV *scalar)
967	PPCODE:	1395	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));	1396	XPUSHs (encode_json (scalar, self));
969		1397
970	void decode (SV *self, SV *jsonstr)	1398	void decode (JSON *self, SV *jsonstr)
971	PPCODE:	1399	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1400	XPUSHs (decode_json (jsonstr, self, 0));
		1401
		1402	void decode_prefix (JSON *self, SV *jsonstr)
		1403	PPCODE:
		1404	{
		1405	UV offset;
		1406	EXTEND (SP, 2);
		1407	PUSHs (decode_json (jsonstr, self, &offset));
		1408	PUSHs (sv_2mortal (newSVuv (offset)));
		1409	}
973		1410
974	PROTOTYPES: ENABLE	1411	PROTOTYPES: ENABLE
975		1412
976	void to_json (SV *scalar)	1413	void to_json (SV *scalar)
977	PPCODE:	1414	PPCODE:
		1415	{
		1416	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1417	XPUSHs (encode_json (scalar, &json));
		1418	}
979		1419
980	void from_json (SV *jsonstr)	1420	void from_json (SV *jsonstr)
981	PPCODE:	1421	PPCODE:
		1422	{
		1423	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1424	XPUSHs (decode_json (jsonstr, &json, 0));
		1425	}
983		1426

Diff Legend

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