ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.11 by root, Sat Mar 24 19:42:14 2007 UTC vs.
Revision 1.55 by root, Mon Jul 23 22:57:40 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;
		68	HV *cb_sk_object;
		69	} JSON;
		70
		71	/////////////////////////////////////////////////////////////////////////////
		72	// utility functions
		73
		74	inline void
		75	shrink (SV *sv)
		76	{
		77	sv_utf8_downgrade (sv, 1);
		78	if (SvLEN (sv) > SvCUR (sv) + 1)
		79	{
		80	#ifdef SvPV_shrink_to_cur
		81	SvPV_shrink_to_cur (sv);
		82	#elif defined (SvPV_renew)
		83	SvPV_renew (sv, SvCUR (sv) + 1);
		84	#endif
		85	}
		86	}
		87
		88	// decode an utf-8 character and return it, or (UV)-1 in
		89	// case of an error.
		90	// we special-case "safe" characters from U+80 .. U+7FF,
		91	// but use the very good perl function to parse anything else.
		92	// note that we never call this function for a ascii codepoints
		93	inline UV
		94	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		95	{
		96	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		99	{
		100	*clen = 2;
		101	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		102	}
		103	else
		104	{
		105	*clen = (STRLEN)-1;
		106	return (UV)-1;
		107	}
		108	}
		109
		110	/////////////////////////////////////////////////////////////////////////////
		111	// encoder
27		112
28	// structure used for encoding JSON	113	// structure used for encoding JSON
29	typedef struct	114	typedef struct
30	{	115	{
31	char *cur;	116	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	117	char *end; // SvEND (sv)
34	SV *sv;	118	SV *sv; // result scalar
35	UV flags;	119	JSON json;
36	int max_recurse;	120	U32 indent; // indentation level
37	int indent;	121	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	122	} enc_t;
39		123
40	// structure used for decoding JSON	124	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)	125	need (enc_t *enc, STRLEN len)
71	{	126	{
72	if (enc->cur + len >= enc->end)	127	if (expect_false (enc->cur + len >= enc->end))
73	{	128	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	129	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	130	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	131	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	133	}
79	}	134	}
80		135
81	static void	136	inline void
82	encode_ch (enc_t *enc, char ch)	137	encode_ch (enc_t *enc, char ch)
83	{	138	{
84	need (enc, 1);	139	need (enc, 1);
85	*enc->cur++ = ch;	140	*enc->cur++ = ch;
86	}	141	}
…		…
94		149
95	while (str < end)	150	while (str < end)
96	{	151	{
97	unsigned char ch = *(unsigned char *)str;	152	unsigned char ch = *(unsigned char *)str;
98		153
99	if (ch >= 0x20 && ch < 0x80) // most common case	154	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	155	{
101	if (ch == '"') // but with slow exceptions	156	if (expect_false (ch == '"')) // but with slow exceptions
102	{	157	{
103	need (enc, len += 1);	158	need (enc, len += 1);
104	*enc->cur++ = '\\';	159	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	160	*enc->cur++ = '"';
106	}	161	}
107	else if (ch == '\\')	162	else if (expect_false (ch == '\\'))
108	{	163	{
109	need (enc, len += 1);	164	need (enc, len += 1);
110	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
112	}	167	}
…		…
130	STRLEN clen;	185	STRLEN clen;
131	UV uch;	186	UV uch;
132		187
133	if (is_utf8)	188	if (is_utf8)
134	{	189	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	190	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	191	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	192	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	193	}
139	else	194	else
140	{	195	{
…		…
143	}	198	}
144		199
145	if (uch > 0x10FFFFUL)	200	if (uch > 0x10FFFFUL)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147		202
148	if (uch < 0x80 || enc->flags & F_ASCII)	203	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149	{	204	{
150	if (uch > 0xFFFFUL)	205	if (uch > 0xFFFFUL)
151	{	206	{
152	need (enc, len += 11);	207	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	208	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	222	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	223	}
169		224
170	str += clen;	225	str += clen;
171	}	226	}
		227	else if (enc->json.flags & F_LATIN1)
		228	{
		229	*enc->cur++ = uch;
		230	str += clen;
		231	}
172	else if (is_utf8)	232	else if (is_utf8)
173	{	233	{
174	need (enc, len += clen);	234	need (enc, len += clen);
175	do	235	do
176	{	236	{
…		…
178	}	238	}
179	while (--clen);	239	while (--clen);
180	}	240	}
181	else	241	else
182	{	242	{
183	need (enc, len += 10); // never more than 11 bytes needed	243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185	++str;	245	++str;
186	}	246	}
187	}	247	}
188	}	248	}
…		…
190		250
191	--len;	251	--len;
192	}	252	}
193	}	253	}
194		254
195	#define INDENT SB \	255	inline void
		256	encode_indent (enc_t *enc)
		257	{
196	if (enc->flags & F_INDENT) \	258	if (enc->json.flags & F_INDENT)
197	{ \	259	{
198	int i_; \	260	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	261
200	for (i_ = enc->indent * 3; i_--; )\	262	need (enc, spaces);
		263	memset (enc->cur, ' ', spaces);
		264	enc->cur += spaces;
		265	}
		266	}
		267
		268	inline void
		269	encode_space (enc_t *enc)
		270	{
		271	need (enc, 1);
		272	encode_ch (enc, ' ');
		273	}
		274
		275	inline void
		276	encode_nl (enc_t *enc)
		277	{
		278	if (enc->json.flags & F_INDENT)
		279	{
		280	need (enc, 1);
201	encode_ch (enc, ' '); \	281	encode_ch (enc, '\n');
202	} \	282	}
203	SE	283	}
204		284
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	285	inline void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	286	encode_comma (enc_t *enc)
207	#define COMMA SB \	287	{
208	encode_ch (enc, ','); \	288	encode_ch (enc, ',');
		289
209	if (enc->flags & F_INDENT) \	290	if (enc->json.flags & F_INDENT)
210	NL; \	291	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	292	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	293	encode_space (enc);
213	SE	294	}
214		295
215	static void encode_sv (enc_t *enc, SV *sv);	296	static void encode_sv (enc_t *enc, SV *sv);
216		297
217	static void	298	static void
218	encode_av (enc_t *enc, AV *av)	299	encode_av (enc_t *enc, AV *av)
219	{	300	{
220	int i, len = av_len (av);	301	int i, len = av_len (av);
221		302
222	encode_ch (enc, '['); NL;	303	if (enc->indent >= enc->maxdepth)
		304	croak ("data structure too deep (hit recursion limit)");
		305
		306	encode_ch (enc, '['); encode_nl (enc);
223	++enc->indent;	307	++enc->indent;
224		308
225	for (i = 0; i <= len; ++i)	309	for (i = 0; i <= len; ++i)
226	{	310	{
227	INDENT;	311	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));	312
		313	encode_indent (enc);
		314
		315	if (svp)
		316	encode_sv (enc, *svp);
		317	else
		318	encode_str (enc, "null", 4, 0);
229		319
230	if (i < len)	320	if (i < len)
231	COMMA;	321	encode_comma (enc);
232	}	322	}
233		323
234	NL;	324	encode_nl (enc);
235		325
236	--enc->indent;	326	--enc->indent;
237	INDENT; encode_ch (enc, ']');	327	encode_indent (enc); encode_ch (enc, ']');
238	}	328	}
239		329
240	static void	330	static void
241	encode_he (enc_t *enc, HE *he)	331	encode_he (enc_t *enc, HE *he)
242	{	332	{
…		…
256	else	346	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	347	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		348
259	encode_ch (enc, '"');	349	encode_ch (enc, '"');
260		350
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	352	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));	354	encode_sv (enc, HeVAL (he));
265	}	355	}
266		356
267	// compare hash entries, used when all keys are bytestrings	357	// compare hash entries, used when all keys are bytestrings
268	static int	358	static int
…		…
292	static void	382	static void
293	encode_hv (enc_t *enc, HV *hv)	383	encode_hv (enc_t *enc, HV *hv)
294	{	384	{
295	int count, i;	385	int count, i;
296		386
		387	if (enc->indent >= enc->maxdepth)
		388	croak ("data structure too deep (hit recursion limit)");
		389
297	encode_ch (enc, '{'); NL; ++enc->indent;	390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298		391
299	if ((count = hv_iterinit (hv)))	392	if ((count = hv_iterinit (hv)))
300	{	393	{
301	// for canonical output we have to sort by keys first	394	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	395	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	396	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	397	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	398	if (enc->json.flags & F_CANONICAL)
306	{	399	{
307	HE *he, *hes [count];
308	int fast = 1;	400	int fast = 1;
		401	HE *he;
		402	#if defined(__BORLANDC__) || defined(_MSC_VER)
		403	HE **hes = _alloca (count * sizeof (HE));
		404	#else
		405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		406	#endif
309		407
310	i = 0;	408	i = 0;
311	while ((he = hv_iternext (hv)))	409	while ((he = hv_iternext (hv)))
312	{	410	{
313	hes [i++] = he;	411	hes [i++] = he;
…		…
337	LEAVE;	435	LEAVE;
338	}	436	}
339		437
340	for (i = 0; i < count; ++i)	438	for (i = 0; i < count; ++i)
341	{	439	{
342	INDENT;	440	encode_indent (enc);
343	encode_he (enc, hes [i]);	441	encode_he (enc, hes [i]);
344		442
345	if (i < count - 1)	443	if (i < count - 1)
346	COMMA;	444	encode_comma (enc);
347	}
348
349	NL;	445	}
		446
		447	encode_nl (enc);
350	}	448	}
351	else	449	else
352	{	450	{
353	SV *sv;
354	HE *he = hv_iternext (hv);	451	HE *he = hv_iternext (hv);
355		452
356	for (;;)	453	for (;;)
357	{	454	{
358	INDENT;	455	encode_indent (enc);
359	encode_he (enc, he);	456	encode_he (enc, he);
360		457
361	if (!(he = hv_iternext (hv)))	458	if (!(he = hv_iternext (hv)))
362	break;	459	break;
363		460
364	COMMA;	461	encode_comma (enc);
365	}
366
367	NL;	462	}
		463
		464	encode_nl (enc);
368	}	465	}
369	}	466	}
370		467
371	--enc->indent; INDENT; encode_ch (enc, '}');	468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		469	}
		470
		471	// encode objects, arrays and special \0=false and \1=true values.
		472	static void
		473	encode_rv (enc_t *enc, SV *sv)
		474	{
		475	svtype svt;
		476
		477	SvGETMAGIC (sv);
		478	svt = SvTYPE (sv);
		479
		480	if (expect_false (SvOBJECT (sv)))
		481	{
		482	if (SvSTASH (sv) == json_boolean_stash)
		483	{
		484	if (SvIV (sv))
		485	encode_str (enc, "true", 4, 0);
		486	else
		487	encode_str (enc, "false", 5, 0);
		488	}
		489	else
		490	{
		491	#if 0
		492	if (0 && sv_derived_from (rv, "JSON::Literal"))
		493	{
		494	// not yet
		495	}
		496	#endif
		497	if (enc->json.flags & F_CONV_BLESSED)
		498	{
		499	// we re-bless the reference to get overload and other niceties right
		500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		501
		502	if (to_json)
		503	{
		504	int count;
		505	dSP;
		506
		507	ENTER; SAVETMPS; PUSHMARK (SP);
		508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		509
		510	// calling with G_SCALAR ensures that we always get a 1 return value
		511	PUTBACK;
		512	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		513	SPAGAIN;
		514
		515	// catch this surprisingly common error
		516	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		517	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		518
		519	sv = POPs;
		520	PUTBACK;
		521
		522	encode_sv (enc, sv);
		523
		524	FREETMPS; LEAVE;
		525	}
		526	else if (enc->json.flags & F_ALLOW_BLESSED)
		527	encode_str (enc, "null", 4, 0);
		528	else
		529	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		530	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		531	}
		532	else if (enc->json.flags & F_ALLOW_BLESSED)
		533	encode_str (enc, "null", 4, 0);
		534	else
		535	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		536	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		537	}
		538	}
		539	else if (svt == SVt_PVHV)
		540	encode_hv (enc, (HV *)sv);
		541	else if (svt == SVt_PVAV)
		542	encode_av (enc, (AV *)sv);
		543	else if (svt < SVt_PVAV)
		544	{
		545	STRLEN len = 0;
		546	char *pv = svt ? SvPV (sv, len) : 0;
		547
		548	if (len == 1 && *pv == '1')
		549	encode_str (enc, "true", 4, 0);
		550	else if (len == 1 && *pv == '0')
		551	encode_str (enc, "false", 5, 0);
		552	else
		553	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		554	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		555	}
		556	else
		557	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		558	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	559	}
373		560
374	static void	561	static void
375	encode_sv (enc_t *enc, SV *sv)	562	encode_sv (enc_t *enc, SV *sv)
376	{	563	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	571	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	572	encode_ch (enc, '"');
386	}	573	}
387	else if (SvNOKp (sv))	574	else if (SvNOKp (sv))
388	{	575	{
		576	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	577	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	579	enc->cur += strlen (enc->cur);
392	}	580	}
393	else if (SvIOKp (sv))	581	else if (SvIOKp (sv))
394	{	582	{
395	need (enc, 64);	583	// we assume we can always read an IV as a UV
		584	if (SvUV (sv) & ~(UV)0x7fff)
		585	{
		586	// large integer, use the (rather slow) snprintf way.
		587	need (enc, sizeof (UV) * 3);
396	enc->cur +=	588	enc->cur +=
397	SvIsUV(sv)	589	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	591	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		592	}
		593	else
		594	{
		595	// optimise the "small number case"
		596	// code will likely be branchless and use only a single multiplication
		597	I32 i = SvIV (sv);
		598	U32 u;
		599	char digit, nz = 0;
		600
		601	need (enc, 6);
		602
		603	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		604	u = i < 0 ? -i : i;
		605
		606	// convert to 4.28 fixed-point representation
		607	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		608
		609	// now output digit by digit, each time masking out the integer part
		610	// and multiplying by 5 while moving the decimal point one to the right,
		611	// resulting in a net multiplication by 10.
		612	// we always write the digit to memory but conditionally increment
		613	// the pointer, to ease the usage of conditional move instructions.
		614	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		615	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		616	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		617	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		618	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		619	}
400	}	620	}
401	else if (SvROK (sv))	621	else if (SvROK (sv))
402	{	622	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))	623	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	624	encode_str (enc, "null", 4, 0);
420	else	625	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	626	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	627	SvPV_nolen (sv), SvFLAGS (sv));
423	}	628	}
424		629
425	static SV *	630	static SV *
426	encode_json (SV *scalar, UV flags)	631	encode_json (SV *scalar, JSON *json)
427	{	632	{
		633	enc_t enc;
		634
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	635	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	636	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		637
431	enc_t enc;	638	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	640	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	641	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	642	enc.indent = 0;
		643	enc.maxdepth = DEC_DEPTH (enc.json.flags);
438		644
439	SvPOK_only (enc.sv);	645	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	646	encode_sv (&enc, scalar);
441		647
		648	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		649	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		650
442	if (!(flags & (F_ASCII | F_UTF8)))	651	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	652	SvUTF8_on (enc.sv);
444		653
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	654	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	655	shrink (enc.sv);
449		656
450	return enc.sv;	657	return enc.sv;
451	}	658	}
452		659
453	/////////////////////////////////////////////////////////////////////////////	660	/////////////////////////////////////////////////////////////////////////////
		661	// decoder
454		662
455	#define WS \	663	// structure used for decoding JSON
		664	typedef struct
		665	{
		666	char *cur; // current parser pointer
		667	char *end; // end of input string
		668	const char *err; // parse error, if != 0
		669	JSON json;
		670	U32 depth; // recursion depth
		671	U32 maxdepth; // recursion depth limit
		672	} dec_t;
		673
		674	inline void
		675	decode_ws (dec_t *dec)
		676	{
456	for (;;) \	677	for (;;)
457	{ \	678	{
458	char ch = *dec->cur; \	679	char ch = *dec->cur;
		680
459	if (ch > 0x20 \	681	if (ch > 0x20
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	682	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461	break; \	683	break;
		684
462	++dec->cur; \	685	++dec->cur;
463	}	686	}
		687	}
464		688
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	689	#define ERR(reason) SB dec->err = reason; goto fail; SE
		690
466	#define EXPECT_CH(ch) SB \	691	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	692	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	693	ERR (# ch " expected"); \
469	++dec->cur; \	694	++dec->cur; \
470	SE	695	SE
471		696
		697	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		698	#define DEC_DEC_DEPTH --dec->depth
		699
472	static SV *decode_sv (dec_t *dec);	700	static SV *decode_sv (dec_t *dec);
473		701
474	static signed char decode_hexdigit[256];	702	static signed char decode_hexdigit[256];
475		703
476	static UV	704	static UV
477	decode_4hex (dec_t *dec)	705	decode_4hex (dec_t *dec)
478	{	706	{
479	signed char d1, d2, d3, d4;	707	signed char d1, d2, d3, d4;
		708	unsigned char *cur = (unsigned char *)dec->cur;
480		709
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	710	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");	711	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]];	712	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");	713	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		714
490	dec->cur += 4;	715	dec->cur += 4;
491		716
492	return ((UV)d1) << 12	717	return ((UV)d1) << 12
493	| ((UV)d2) << 8	718	| ((UV)d2) << 8
…		…
496		721
497	fail:	722	fail:
498	return (UV)-1;	723	return (UV)-1;
499	}	724	}
500		725
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 *	726	static SV *
517	decode_str (dec_t *dec)	727	decode_str (dec_t *dec)
518	{	728	{
519	SV *sv = NEWSV (0,2);	729	SV *sv = 0;
520	int utf8 = 0;	730	int utf8 = 0;
521	char *cur = SvPVX (sv);	731	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		732
524	for (;;)	733	do
525	{	734	{
526	unsigned char ch = *(unsigned char *)dec->cur;	735	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		736	char *cur = buf;
527		737
528	if (ch == '"')	738	do
529	break;
530	else if (ch == '\\')
531	{	739	{
532	switch (*++dec->cur)	740	unsigned char ch = *(unsigned char *)dec_cur++;
		741
		742	if (expect_false (ch == '"'))
		743	{
		744	--dec_cur;
		745	break;
533	{	746	}
534	case '\\':	747	else if (expect_false (ch == '\\'))
535	case '/':	748	{
536	case '"': APPEND_CH (*dec->cur++); break;	749	switch (*dec_cur)
537
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;
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	{	750	{
546	UV lo, hi;	751	case '\\':
547	++dec->cur;	752	case '/':
		753	case '"': *cur++ = *dec_cur++; break;
548		754
549	hi = decode_4hex (dec);	755	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	756	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	757	case 'n': ++dec_cur; *cur++ = '\012'; break;
		758	case 'f': ++dec_cur; *cur++ = '\014'; break;
		759	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		760
553	// possibly a surrogate pair	761	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	762	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	763	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	764	++dec_cur;
558		765
559	dec->cur += 2;	766	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	767	hi = decode_4hex (dec);
		768	dec_cur = dec->cur;
562	if (lo == (UV)-1)	769	if (hi == (UV)-1)
563	goto fail;	770	goto fail;
564		771
		772	// possibly a surrogate pair
		773	if (hi >= 0xd800)
		774	if (hi < 0xdc00)
		775	{
		776	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		777	ERR ("missing low surrogate character in surrogate pair");
		778
		779	dec_cur += 2;
		780
		781	dec->cur = dec_cur;
		782	lo = decode_4hex (dec);
		783	dec_cur = dec->cur;
		784	if (lo == (UV)-1)
		785	goto fail;
		786
565	if (lo < 0xdc00 || lo >= 0xe000)	787	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	788	ERR ("surrogate pair expected");
567		789
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	790	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		791	}
		792	else if (hi < 0xe000)
		793	ERR ("missing high surrogate character in surrogate pair");
		794
		795	if (hi >= 0x80)
		796	{
		797	utf8 = 1;
		798
		799	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		800	}
		801	else
		802	*cur++ = hi;
569	}	803	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	804	break;
575	utf8 = 1;
576		805
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	806	default:
581	APPEND_CH (hi);	807	--dec_cur;
		808	ERR ("illegal backslash escape sequence in string");
582	}	809	}
583	break;	810	}
		811	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		812	*cur++ = ch;
		813	else if (ch >= 0x80)
		814	{
		815	STRLEN clen;
		816	UV uch;
584		817
585	default:
586	--dec->cur;	818	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	819
		820	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		821	if (clen == (STRLEN)-1)
		822	ERR ("malformed UTF-8 character in JSON string");
		823
		824	do
		825	*cur++ = *dec_cur++;
		826	while (--clen);
		827
		828	utf8 = 1;
		829	}
		830	else
		831	{
		832	--dec_cur;
		833
		834	if (!ch)
		835	ERR ("unexpected end of string while parsing JSON string");
		836	else
		837	ERR ("invalid character encountered while parsing JSON string");
588	}	838	}
589	}	839	}
590	else if (ch >= 0x20 && ch <= 0x7f)	840	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	841
592	else if (ch >= 0x80)
593	{	842	{
594	STRLEN clen;	843	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		844
599	APPEND_GROW (clen);	845	if (sv)
600	do
601	{	846	{
602	*cur++ = *dec->cur++;	847	SvGROW (sv, SvCUR (sv) + len + 1);
		848	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		849	SvCUR_set (sv, SvCUR (sv) + len);
603	}	850	}
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	851	else
611	ERR ("invalid character encountered");	852	sv = newSVpvn (buf, len);
612	}	853	}
		854	}
		855	while (*dec_cur != '"');
613		856
614	++dec->cur;	857	++dec_cur;
615		858
616	SvCUR_set (sv, cur - SvPVX (sv));	859	if (sv)
617		860	{
618	SvPOK_only (sv);	861	SvPOK_only (sv);
619	*SvEND (sv) = 0;	862	*SvEND (sv) = 0;
620		863
621	if (utf8)	864	if (utf8)
622	SvUTF8_on (sv);	865	SvUTF8_on (sv);
		866	}
		867	else
		868	sv = newSVpvn ("", 0);
623		869
624	if (dec->flags & F_SHRINK)	870	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	871	return sv;
628		872
629	fail:	873	fail:
630	SvREFCNT_dec (sv);	874	dec->cur = dec_cur;
631	return 0;	875	return 0;
632	}	876	}
633		877
634	static SV *	878	static SV *
635	decode_num (dec_t *dec)	879	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	937	is_nv = 1;
694	}	938	}
695		939
696	if (!is_nv)	940	if (!is_nv)
697	{	941	{
698	UV uv;	942	// 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);	943	if (*start == '-')
700	if (numtype & IS_NUMBER_IN_UV)	944	switch (dec->cur - start)
701	if (numtype & IS_NUMBER_NEG)
702	{	945	{
703	if (uv < (UV)IV_MIN)	946	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	947	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		948	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		949	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705	}	950	}
		951	else
		952	switch (dec->cur - start)
		953	{
		954	case 1: return newSViv ( start [0] - '0' * 1);
		955	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		956	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		957	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		958	}
		959
		960	{
		961	UV uv;
		962	int numtype = grok_number (start, dec->cur - start, &uv);
		963	if (numtype & IS_NUMBER_IN_UV)
		964	if (numtype & IS_NUMBER_NEG)
		965	{
		966	if (uv < (UV)IV_MIN)
		967	return newSViv (-(IV)uv);
		968	}
706	else	969	else
707	return newSVuv (uv);	970	return newSVuv (uv);
		971
		972	// here would likely be the place for bigint support
708	}	973	}
		974	}
709		975
		976	// if we ever support bigint or bigfloat, this is the place for bigfloat
710	return newSVnv (Atof (start));	977	return newSVnv (Atof (start));
711		978
712	fail:	979	fail:
713	return 0;	980	return 0;
714	}	981	}
…		…
716	static SV *	983	static SV *
717	decode_av (dec_t *dec)	984	decode_av (dec_t *dec)
718	{	985	{
719	AV *av = newAV ();	986	AV *av = newAV ();
720		987
721	WS;	988	DEC_INC_DEPTH;
		989	decode_ws (dec);
		990
722	if (*dec->cur == ']')	991	if (*dec->cur == ']')
723	++dec->cur;	992	++dec->cur;
724	else	993	else
725	for (;;)	994	for (;;)
726	{	995	{
…		…
730	if (!value)	999	if (!value)
731	goto fail;	1000	goto fail;
732		1001
733	av_push (av, value);	1002	av_push (av, value);
734		1003
735	WS;	1004	decode_ws (dec);
736		1005
737	if (*dec->cur == ']')	1006	if (*dec->cur == ']')
738	{	1007	{
739	++dec->cur;	1008	++dec->cur;
740	break;	1009	break;
…		…
744	ERR (", or ] expected while parsing array");	1013	ERR (", or ] expected while parsing array");
745		1014
746	++dec->cur;	1015	++dec->cur;
747	}	1016	}
748		1017
		1018	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1019	return newRV_noinc ((SV *)av);
750		1020
751	fail:	1021	fail:
752	SvREFCNT_dec (av);	1022	SvREFCNT_dec (av);
		1023	DEC_DEC_DEPTH;
753	return 0;	1024	return 0;
754	}	1025	}
755		1026
756	static SV *	1027	static SV *
757	decode_hv (dec_t *dec)	1028	decode_hv (dec_t *dec)
758	{	1029	{
		1030	SV *sv;
759	HV *hv = newHV ();	1031	HV *hv = newHV ();
760		1032
761	WS;	1033	DEC_INC_DEPTH;
		1034	decode_ws (dec);
		1035
762	if (*dec->cur == '}')	1036	if (*dec->cur == '}')
763	++dec->cur;	1037	++dec->cur;
764	else	1038	else
765	for (;;)	1039	for (;;)
766	{	1040	{
		1041	decode_ws (dec); EXPECT_CH ('"');
		1042
		1043	// heuristic: assume that
		1044	// a) decode_str + hv_store_ent are abysmally slow.
		1045	// b) most hash keys are short, simple ascii text.
		1046	// => try to "fast-match" such strings to avoid
		1047	// the overhead of decode_str + hv_store_ent.
		1048	{
767	SV *key, *value;	1049	SV *value;
		1050	char *p = dec->cur;
		1051	char *e = p + 24; // only try up to 24 bytes
768		1052
769	WS; EXPECT_CH ('"');	1053	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	{	1054	{
		1055	// the >= 0x80 is true on most architectures
		1056	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1057	{
		1058	// slow path, back up and use decode_str
		1059	SV *key = decode_str (dec);
		1060	if (!key)
		1061	goto fail;
		1062
		1063	decode_ws (dec); EXPECT_CH (':');
		1064
		1065	value = decode_sv (dec);
		1066	if (!value)
		1067	{
		1068	SvREFCNT_dec (key);
		1069	goto fail;
		1070	}
		1071
		1072	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1073	SvREFCNT_dec (key);
		1074
		1075	break;
		1076	}
		1077	else if (*p == '"')
		1078	{
		1079	// fast path, got a simple key
		1080	char *key = dec->cur;
		1081	int len = p - key;
		1082	dec->cur = p + 1;
		1083
		1084	decode_ws (dec); EXPECT_CH (':');
		1085
		1086	value = decode_sv (dec);
		1087	if (!value)
781	goto fail;	1088	goto fail;
		1089
		1090	hv_store (hv, key, len, value, 0);
		1091
		1092	break;
		1093	}
		1094
		1095	++p;
782	}	1096	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1097	}
		1098
		1099	decode_ws (dec);
788		1100
789	if (*dec->cur == '}')	1101	if (*dec->cur == '}')
790	{	1102	{
791	++dec->cur;	1103	++dec->cur;
792	break;	1104	break;
…		…
796	ERR (", or } expected while parsing object/hash");	1108	ERR (", or } expected while parsing object/hash");
797		1109
798	++dec->cur;	1110	++dec->cur;
799	}	1111	}
800		1112
		1113	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1114	sv = newRV_noinc ((SV *)hv);
		1115
		1116	// check filter callbacks
		1117	if (dec->json.flags & F_HOOK)
		1118	{
		1119	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1120	{
		1121	HE *cb, *he;
		1122
		1123	hv_iterinit (hv);
		1124	he = hv_iternext (hv);
		1125	hv_iterinit (hv);
		1126
		1127	// the next line creates a mortal sv each time its called.
		1128	// might want to optimise this for common cases.
		1129	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1130
		1131	if (cb)
		1132	{
		1133	dSP;
		1134	int count;
		1135
		1136	ENTER; SAVETMPS; PUSHMARK (SP);
		1137	XPUSHs (HeVAL (he));
		1138
		1139	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1140
		1141	if (count == 1)
		1142	{
		1143	sv = newSVsv (POPs);
		1144	FREETMPS; LEAVE;
		1145	return sv;
		1146	}
		1147
		1148	FREETMPS; LEAVE;
		1149	}
		1150	}
		1151
		1152	if (dec->json.cb_object)
		1153	{
		1154	dSP;
		1155	int count;
		1156
		1157	ENTER; SAVETMPS; PUSHMARK (SP);
		1158	XPUSHs (sv_2mortal (sv));
		1159
		1160	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1161
		1162	if (count == 1)
		1163	{
		1164	sv = newSVsv (POPs);
		1165	FREETMPS; LEAVE;
		1166	return sv;
		1167	}
		1168
		1169	SvREFCNT_inc (sv);
		1170	FREETMPS; LEAVE;
		1171	}
		1172	}
		1173
		1174	return sv;
802		1175
803	fail:	1176	fail:
804	SvREFCNT_dec (hv);	1177	SvREFCNT_dec (hv);
		1178	DEC_DEC_DEPTH;
805	return 0;	1179	return 0;
806	}	1180	}
807		1181
808	static SV *	1182	static SV *
809	decode_sv (dec_t *dec)	1183	decode_sv (dec_t *dec)
810	{	1184	{
811	WS;	1185	decode_ws (dec);
		1186
		1187	// the beauty of JSON: you need exactly one character lookahead
		1188	// to parse anything.
812	switch (*dec->cur)	1189	switch (*dec->cur)
813	{	1190	{
814	case '"': ++dec->cur; return decode_str (dec);	1191	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1192	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1193	case '{': ++dec->cur; return decode_hv (dec);
…		…
822		1199
823	case 't':	1200	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1202	{
826	dec->cur += 4;	1203	dec->cur += 4;
827	return newSViv (1);	1204	return SvREFCNT_inc (json_true);
828	}	1205	}
829	else	1206	else
830	ERR ("'true' expected");	1207	ERR ("'true' expected");
831		1208
832	break;	1209	break;
833		1210
834	case 'f':	1211	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1213	{
837	dec->cur += 5;	1214	dec->cur += 5;
838	return newSViv (0);	1215	return SvREFCNT_inc (json_false);
839	}	1216	}
840	else	1217	else
841	ERR ("'false' expected");	1218	ERR ("'false' expected");
842		1219
843	break;	1220	break;
…		…
852	ERR ("'null' expected");	1229	ERR ("'null' expected");
853		1230
854	break;	1231	break;
855		1232
856	default:	1233	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1234	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1235	break;
859	}	1236	}
860		1237
861	fail:	1238	fail:
862	return 0;	1239	return 0;
863	}	1240	}
864		1241
865	static SV *	1242	static SV *
866	decode_json (SV *string, UV flags)	1243	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1244	{
		1245	dec_t dec;
		1246	UV offset;
868	SV *sv;	1247	SV *sv;
869		1248
		1249	SvGETMAGIC (string);
		1250	SvUPGRADE (string, SVt_PV);
		1251
		1252	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1253	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1254	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1255
870	if (flags & F_UTF8)	1256	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1257	sv_utf8_downgrade (string, 0);
872	else	1258	else
873	sv_utf8_upgrade (string);	1259	sv_utf8_upgrade (string);
874		1260
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1261	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1262
877	dec_t dec;	1263	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1264	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1265	dec.end = SvEND (string);
881	dec.err = 0;	1266	dec.err = 0;
		1267	dec.depth = 0;
		1268	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1269
		1270	if (dec.json.cb_object || dec.json.cb_sk_object)
		1271	dec.json.flags |= F_HOOK;
		1272
		1273	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
883	sv = decode_sv (&dec);	1274	sv = decode_sv (&dec);
884		1275
		1276	if (!(offset_return || !sv))
		1277	{
		1278	// check for trailing garbage
		1279	decode_ws (&dec);
		1280
		1281	if (*dec.cur)
		1282	{
		1283	dec.err = "garbage after JSON object";
		1284	SvREFCNT_dec (sv);
		1285	sv = 0;
		1286	}
		1287	}
		1288
		1289	if (offset_return || !sv)
		1290	{
		1291	offset = dec.json.flags & F_UTF8
		1292	? dec.cur - SvPVX (string)
		1293	: utf8_distance (dec.cur, SvPVX (string));
		1294
		1295	if (offset_return)
		1296	*offset_return = offset;
		1297	}
		1298
885	if (!sv)	1299	if (!sv)
886	{	1300	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1301	SV *uni = sv_newmortal ();
891		1302
892	// horrible hack to silence warning inside pv_uni_display	1303	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1304	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1305	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1307	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1308	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1309	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1310	LEAVE;
900		1311
901	croak ("%s, at character offset %d (%s)",	1312	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1313	dec.err,
903	(int)offset,	1314	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1315	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1316	}
906		1317
907	sv = sv_2mortal (sv);	1318	sv = sv_2mortal (sv);
908		1319
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1320	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)");	1321	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1322
912	return sv;	1323	return sv;
913	}	1324	}
914		1325
		1326	/////////////////////////////////////////////////////////////////////////////
		1327	// XS interface functions
		1328
915	MODULE = JSON::XS PACKAGE = JSON::XS	1329	MODULE = JSON::XS PACKAGE = JSON::XS
916		1330
917	BOOT:	1331	BOOT:
918	{	1332	{
919	int i;	1333	int i;
920		1334
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1335	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1336	decode_hexdigit [i] =
		1337	i >= '0' && i <= '9' ? i - '0'
		1338	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1339	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1340	: -1;
924		1341
925	for (i = 7; i--; )	1342	json_stash = gv_stashpv ("JSON::XS" , 1);
		1343	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1344
		1345	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1346	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1347	}
		1348
		1349	PROTOTYPES: DISABLE
		1350
		1351	void new (char *klass)
		1352	PPCODE:
		1353	{
		1354	SV *pv = NEWSV (0, sizeof (JSON));
		1355	SvPOK_only (pv);
		1356	Zero (SvPVX (pv), 1, JSON);
		1357	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1358	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1359	}
		1360
		1361	void ascii (JSON *self, int enable = 1)
		1362	ALIAS:
		1363	ascii = F_ASCII
		1364	latin1 = F_LATIN1
		1365	utf8 = F_UTF8
		1366	indent = F_INDENT
		1367	canonical = F_CANONICAL
		1368	space_before = F_SPACE_BEFORE
		1369	space_after = F_SPACE_AFTER
		1370	pretty = F_PRETTY
		1371	allow_nonref = F_ALLOW_NONREF
		1372	shrink = F_SHRINK
		1373	allow_blessed = F_ALLOW_BLESSED
		1374	convert_blessed = F_CONV_BLESSED
		1375	PPCODE:
		1376	{
		1377	if (enable)
		1378	self->flags |= ix;
		1379	else
		1380	self->flags &= ~ix;
		1381
		1382	XPUSHs (ST (0));
		1383	}
		1384
		1385	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1386	PPCODE:
		1387	{
		1388	UV log2 = 0;
		1389
		1390	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1391
		1392	while ((1UL << log2) < max_depth)
		1393	++log2;
		1394
		1395	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1396
		1397	XPUSHs (ST (0));
		1398	}
		1399
		1400	void max_size (JSON *self, UV max_size = 0)
		1401	PPCODE:
		1402	{
		1403	UV log2 = 0;
		1404
		1405	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1406	if (max_size == 1) max_size = 2;
		1407
		1408	while ((1UL << log2) < max_size)
		1409	++log2;
		1410
		1411	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1412
		1413	XPUSHs (ST (0));
		1414	}
		1415
		1416	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1417	PPCODE:
		1418	{
		1419	SvREFCNT_dec (self->cb_object);
		1420	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1421
		1422	XPUSHs (ST (0));
		1423	}
		1424
		1425	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1426	PPCODE:
		1427	{
		1428	if (!self->cb_sk_object)
		1429	self->cb_sk_object = newHV ();
		1430
		1431	if (SvOK (cb))
		1432	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1433	else
926	{	1434	{
927	decode_hexdigit ['a' + i] = 10 + i;	1435	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
928	decode_hexdigit ['A' + i] = 10 + i;	1436
		1437	if (!HvKEYS (self->cb_sk_object))
		1438	{
		1439	SvREFCNT_dec (self->cb_sk_object);
		1440	self->cb_sk_object = 0;
		1441	}
929	}	1442	}
930		1443
931	json_stash = gv_stashpv ("JSON::XS", 1);	1444	XPUSHs (ST (0));
932	}	1445	}
933		1446
934	PROTOTYPES: DISABLE	1447	void encode (JSON *self, SV *scalar)
		1448	PPCODE:
		1449	XPUSHs (encode_json (scalar, self));
935		1450
936	SV *new (char *dummy)	1451	void decode (JSON *self, SV *jsonstr)
		1452	PPCODE:
		1453	XPUSHs (decode_json (jsonstr, self, 0));
		1454
		1455	void decode_prefix (JSON *self, SV *jsonstr)
		1456	PPCODE:
		1457	{
		1458	UV offset;
		1459	EXTEND (SP, 2);
		1460	PUSHs (decode_json (jsonstr, self, &offset));
		1461	PUSHs (sv_2mortal (newSVuv (offset)));
		1462	}
		1463
		1464	void DESTROY (JSON *self)
937	CODE:	1465	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1466	SvREFCNT_dec (self->cb_sk_object);
939	OUTPUT:	1467	SvREFCNT_dec (self->cb_object);
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)
967	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));
969
970	void decode (SV *self, SV *jsonstr)
971	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
973		1468
974	PROTOTYPES: ENABLE	1469	PROTOTYPES: ENABLE
975		1470
976	void to_json (SV *scalar)	1471	void to_json (SV *scalar)
977	PPCODE:	1472	PPCODE:
		1473	{
		1474	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1475	XPUSHs (encode_json (scalar, &json));
		1476	}
979		1477
980	void from_json (SV *jsonstr)	1478	void from_json (SV *jsonstr)
981	PPCODE:	1479	PPCODE:
		1480	{
		1481	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1482	XPUSHs (decode_json (jsonstr, &json, 0));
		1483	}
983		1484

Diff Legend

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