ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.9 by root, Fri Mar 23 17:40:29 2007 UTC vs.
Revision 1.55 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_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	27	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL
		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 ))
18		39
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	41	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		42
22	#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
23		47
24	#define SB do {	48	#define SB do {
25	#define SE } while (0)	49	#define SE } while (0)
26		50
27	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
28		112
29	// structure used for encoding JSON	113	// structure used for encoding JSON
30	typedef struct	114	typedef struct
31	{	115	{
32	char *cur;	116	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	117	char *end; // SvEND (sv)
35	SV *sv;	118	SV *sv; // result scalar
36	UV flags;	119	JSON json;
37	int max_recurse;	120	U32 indent; // indentation level
38	int indent;	121	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	122	} enc_t;
40		123
41	// structure used for decoding JSON	124	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	125	need (enc_t *enc, STRLEN len)
72	{	126	{
73	if (enc->cur + len >= enc->end)	127	if (expect_false (enc->cur + len >= enc->end))
74	{	128	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	129	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	130	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	131	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	133	}
80	}	134	}
81		135
82	static void	136	inline void
83	encode_ch (enc_t *enc, char ch)	137	encode_ch (enc_t *enc, char ch)
84	{	138	{
85	need (enc, 1);	139	need (enc, 1);
86	*enc->cur++ = ch;	140	*enc->cur++ = ch;
87	}	141	}
…		…
95		149
96	while (str < end)	150	while (str < end)
97	{	151	{
98	unsigned char ch = *(unsigned char *)str;	152	unsigned char ch = *(unsigned char *)str;
99		153
100	if (ch >= 0x20 && ch < 0x80) // most common case	154	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	155	{
102	if (ch == '"') // but with slow exceptions	156	if (expect_false (ch == '"')) // but with slow exceptions
103	{	157	{
104	need (enc, len += 1);	158	need (enc, len += 1);
105	*enc->cur++ = '\\';	159	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	160	*enc->cur++ = '"';
107	}	161	}
108	else if (ch == '\\')	162	else if (expect_false (ch == '\\'))
109	{	163	{
110	need (enc, len += 1);	164	need (enc, len += 1);
111	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
113	}	167	}
…		…
131	STRLEN clen;	185	STRLEN clen;
132	UV uch;	186	UV uch;
133		187
134	if (is_utf8)	188	if (is_utf8)
135	{	189	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	190	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	191	if (clen == (STRLEN)-1)
138	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);
139	}	193	}
140	else	194	else
141	{	195	{
…		…
144	}	198	}
145		199
146	if (uch > 0x10FFFFUL)	200	if (uch > 0x10FFFFUL)
147	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);
148		202
149	if (uch < 0x80 || enc->flags & F_ASCII)	203	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
150	{	204	{
151	if (uch > 0xFFFFUL)	205	if (uch > 0xFFFFUL)
152	{	206	{
153	need (enc, len += 11);	207	need (enc, len += 11);
154	sprintf (enc->cur, "\\u%04x\\u%04x",	208	sprintf (enc->cur, "\\u%04x\\u%04x",
155	(uch - 0x10000) / 0x400 + 0xD800,	209	(int)((uch - 0x10000) / 0x400 + 0xD800),
156	(uch - 0x10000) % 0x400 + 0xDC00);	210	(int)((uch - 0x10000) % 0x400 + 0xDC00));
157	enc->cur += 12;	211	enc->cur += 12;
158	}	212	}
159	else	213	else
160	{	214	{
161	static char hexdigit [16] = "0123456789abcdef";	215	static char hexdigit [16] = "0123456789abcdef";
…		…
168	*enc->cur++ = hexdigit [(uch >> 0) & 15];	222	*enc->cur++ = hexdigit [(uch >> 0) & 15];
169	}	223	}
170		224
171	str += clen;	225	str += clen;
172	}	226	}
		227	else if (enc->json.flags & F_LATIN1)
		228	{
		229	*enc->cur++ = uch;
		230	str += clen;
		231	}
173	else if (is_utf8)	232	else if (is_utf8)
174	{	233	{
175	need (enc, len += clen);	234	need (enc, len += clen);
176	do	235	do
177	{	236	{
…		…
179	}	238	}
180	while (--clen);	239	while (--clen);
181	}	240	}
182	else	241	else
183	{	242	{
184	need (enc, len += 10); // never more than 11 bytes needed	243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
185	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
186	++str;	245	++str;
187	}	246	}
188	}	247	}
189	}	248	}
…		…
191		250
192	--len;	251	--len;
193	}	252	}
194	}	253	}
195		254
196	#define INDENT SB \	255	inline void
		256	encode_indent (enc_t *enc)
		257	{
197	if (enc->flags & F_INDENT) \	258	if (enc->json.flags & F_INDENT)
198	{ \	259	{
199	int i_; \	260	int spaces = enc->indent * INDENT_STEP;
200	need (enc, enc->indent); \	261
201	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);
202	encode_ch (enc, ' '); \	281	encode_ch (enc, '\n');
203	} \	282	}
204	SE	283	}
205		284
206	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	285	inline void
207	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	286	encode_comma (enc_t *enc)
208	#define COMMA SB \	287	{
209	encode_ch (enc, ','); \	288	encode_ch (enc, ',');
		289
210	if (enc->flags & F_INDENT) \	290	if (enc->json.flags & F_INDENT)
211	NL; \	291	encode_nl (enc);
212	else if (enc->flags & F_SPACE_AFTER) \	292	else if (enc->json.flags & F_SPACE_AFTER)
213	SPACE; \	293	encode_space (enc);
214	SE	294	}
215		295
216	static void encode_sv (enc_t *enc, SV *sv);	296	static void encode_sv (enc_t *enc, SV *sv);
217		297
218	static void	298	static void
219	encode_av (enc_t *enc, AV *av)	299	encode_av (enc_t *enc, AV *av)
220	{	300	{
221	int i, len = av_len (av);	301	int i, len = av_len (av);
222		302
223	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);
224	++enc->indent;	307	++enc->indent;
225		308
226	for (i = 0; i <= len; ++i)	309	for (i = 0; i <= len; ++i)
227	{	310	{
228	INDENT;	311	SV **svp = av_fetch (av, i, 0);
229	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);
230		319
231	if (i < len)	320	if (i < len)
232	COMMA;	321	encode_comma (enc);
233	}	322	}
234		323
235	NL;	324	encode_nl (enc);
236		325
237	--enc->indent;	326	--enc->indent;
238	INDENT; encode_ch (enc, ']');	327	encode_indent (enc); encode_ch (enc, ']');
239	}	328	}
240		329
241	static void	330	static void
242	encode_he (enc_t *enc, HE *he)	331	encode_he (enc_t *enc, HE *he)
243	{	332	{
…		…
257	else	346	else
258	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	347	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259		348
260	encode_ch (enc, '"');	349	encode_ch (enc, '"');
261		350
262	if (enc->flags & F_SPACE_BEFORE) SPACE;	351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
263	encode_ch (enc, ':');	352	encode_ch (enc, ':');
264	if (enc->flags & F_SPACE_AFTER ) SPACE;	353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
265	encode_sv (enc, HeVAL (he));	354	encode_sv (enc, HeVAL (he));
266	}	355	}
267		356
268	// compare hash entries, used when all keys are bytestrings	357	// compare hash entries, used when all keys are bytestrings
269	static int	358	static int
…		…
275	HE *b = *(HE **)b_;	364	HE *b = *(HE **)b_;
276		365
277	STRLEN la = HeKLEN (a);	366	STRLEN la = HeKLEN (a);
278	STRLEN lb = HeKLEN (b);	367	STRLEN lb = HeKLEN (b);
279		368
280	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
281	cmp = la < lb ? -1 : la == lb ? 0 : 1;	370	cmp = la - lb;
282		371
283	return cmp;	372	return cmp;
284	}	373	}
285		374
286	// compare hash entries, used when some keys are sv's or utf-x	375	// compare hash entries, used when some keys are sv's or utf-x
…		…
293	static void	382	static void
294	encode_hv (enc_t *enc, HV *hv)	383	encode_hv (enc_t *enc, HV *hv)
295	{	384	{
296	int count, i;	385	int count, i;
297		386
		387	if (enc->indent >= enc->maxdepth)
		388	croak ("data structure too deep (hit recursion limit)");
		389
298	encode_ch (enc, '{'); NL; ++enc->indent;	390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
299		391
300	if ((count = hv_iterinit (hv)))	392	if ((count = hv_iterinit (hv)))
301	{	393	{
302	// for canonical output we have to sort by keys first	394	// for canonical output we have to sort by keys first
303	// actually, this is mostly due to the stupid so-called	395	// actually, this is mostly due to the stupid so-called
304	// security workaround added somewhere in 5.8.x.	396	// security workaround added somewhere in 5.8.x.
305	// that randomises hash orderings	397	// that randomises hash orderings
306	if (enc->flags & F_CANONICAL)	398	if (enc->json.flags & F_CANONICAL)
307	{	399	{
308	HE *he, *hes [count];
309	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
310		407
311	i = 0;	408	i = 0;
312	while ((he = hv_iternext (hv)))	409	while ((he = hv_iternext (hv)))
313	{	410	{
314	hes [i++] = he;	411	hes [i++] = he;
…		…
338	LEAVE;	435	LEAVE;
339	}	436	}
340		437
341	for (i = 0; i < count; ++i)	438	for (i = 0; i < count; ++i)
342	{	439	{
343	INDENT;	440	encode_indent (enc);
344	encode_he (enc, hes [i]);	441	encode_he (enc, hes [i]);
345		442
346	if (i < count - 1)	443	if (i < count - 1)
347	COMMA;	444	encode_comma (enc);
348	}
349
350	NL;	445	}
		446
		447	encode_nl (enc);
351	}	448	}
352	else	449	else
353	{	450	{
354	SV *sv;
355	HE *he = hv_iternext (hv);	451	HE *he = hv_iternext (hv);
356		452
357	for (;;)	453	for (;;)
358	{	454	{
359	INDENT;	455	encode_indent (enc);
360	encode_he (enc, he);	456	encode_he (enc, he);
361		457
362	if (!(he = hv_iternext (hv)))	458	if (!(he = hv_iternext (hv)))
363	break;	459	break;
364		460
365	COMMA;	461	encode_comma (enc);
366	}
367
368	NL;	462	}
		463
		464	encode_nl (enc);
369	}	465	}
370	}	466	}
371		467
372	--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))));
373	}	559	}
374		560
375	static void	561	static void
376	encode_sv (enc_t *enc, SV *sv)	562	encode_sv (enc_t *enc, SV *sv)
377	{	563	{
…		…
385	encode_str (enc, str, len, SvUTF8 (sv));	571	encode_str (enc, str, len, SvUTF8 (sv));
386	encode_ch (enc, '"');	572	encode_ch (enc, '"');
387	}	573	}
388	else if (SvNOKp (sv))	574	else if (SvNOKp (sv))
389	{	575	{
		576	// trust that perl will do the right thing w.r.t. JSON syntax.
390	need (enc, NV_DIG + 32);	577	need (enc, NV_DIG + 32);
391	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392	enc->cur += strlen (enc->cur);	579	enc->cur += strlen (enc->cur);
393	}	580	}
394	else if (SvIOKp (sv))	581	else if (SvIOKp (sv))
395	{	582	{
396	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);
397	enc->cur +=	588	enc->cur +=
398	SvIsUV(sv)	589	SvIsUV(sv)
399	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
400	: 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	}
401	}	620	}
402	else if (SvROK (sv))	621	else if (SvROK (sv))
403	{	622	encode_rv (enc, SvRV (sv));
404	SV *rv = SvRV (sv);
405
406	if (!--enc->max_recurse)
407	croak ("data structure too deep (hit recursion limit)");
408
409	switch (SvTYPE (rv))
410	{
411	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
412	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
413
414	default:
415	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
416	SvPV_nolen (sv));
417	}
418	}
419	else if (!SvOK (sv))	623	else if (!SvOK (sv))
420	encode_str (enc, "null", 4, 0);	624	encode_str (enc, "null", 4, 0);
421	else	625	else
422	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",
423	SvPV_nolen (sv), SvFLAGS (sv));	627	SvPV_nolen (sv), SvFLAGS (sv));
424	}	628	}
425		629
426	static SV *	630	static SV *
427	encode_json (SV *scalar, UV flags)	631	encode_json (SV *scalar, JSON *json)
428	{	632	{
		633	enc_t enc;
		634
429	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	635	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	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)");
431		637
432	enc_t enc;	638	enc.json = *json;
433	enc.flags = flags;
434	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435	enc.cur = SvPVX (enc.sv);	640	enc.cur = SvPVX (enc.sv);
436	enc.end = SvEND (enc.sv);	641	enc.end = SvEND (enc.sv);
437	enc.max_recurse = 0;
438	enc.indent = 0;	642	enc.indent = 0;
		643	enc.maxdepth = DEC_DEPTH (enc.json.flags);
439		644
440	SvPOK_only (enc.sv);	645	SvPOK_only (enc.sv);
441	encode_sv (&enc, scalar);	646	encode_sv (&enc, scalar);
442		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
443	if (!(flags & (F_ASCII | F_UTF8)))	651	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
444	SvUTF8_on (enc.sv);	652	SvUTF8_on (enc.sv);
445		653
446	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447
448	if (enc.flags & F_SHRINK)	654	if (enc.json.flags & F_SHRINK)
449	shrink (enc.sv);	655	shrink (enc.sv);
450		656
451	return enc.sv;	657	return enc.sv;
452	}	658	}
453		659
454	/////////////////////////////////////////////////////////////////////////////	660	/////////////////////////////////////////////////////////////////////////////
		661	// decoder
455		662
456	#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	{
457	for (;;) \	677	for (;;)
458	{ \	678	{
459	char ch = *dec->cur; \	679	char ch = *dec->cur;
		680
460	if (ch > 0x20 \	681	if (ch > 0x20
461	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	682	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
462	break; \	683	break;
		684
463	++dec->cur; \	685	++dec->cur;
464	}	686	}
		687	}
465		688
466	#define ERR(reason) SB dec->err = reason; goto fail; SE	689	#define ERR(reason) SB dec->err = reason; goto fail; SE
		690
467	#define EXPECT_CH(ch) SB \	691	#define EXPECT_CH(ch) SB \
468	if (*dec->cur != ch) \	692	if (*dec->cur != ch) \
469	ERR (# ch " expected"); \	693	ERR (# ch " expected"); \
470	++dec->cur; \	694	++dec->cur; \
471	SE	695	SE
472		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
473	static SV *decode_sv (dec_t *dec);	700	static SV *decode_sv (dec_t *dec);
474		701
475	static signed char decode_hexdigit[256];	702	static signed char decode_hexdigit[256];
476		703
477	static UV	704	static UV
478	decode_4hex (dec_t *dec)	705	decode_4hex (dec_t *dec)
479	{	706	{
480	signed char d1, d2, d3, d4;	707	signed char d1, d2, d3, d4;
		708	unsigned char *cur = (unsigned char *)dec->cur;
481		709
482	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");
483	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");
484	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");
485	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");
486	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
487	if (d3 < 0) ERR ("four hexadecimal digits expected");
488	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
489	if (d4 < 0) ERR ("four hexadecimal digits expected");
490		714
491	dec->cur += 4;	715	dec->cur += 4;
492		716
493	return ((UV)d1) << 12	717	return ((UV)d1) << 12
494	| ((UV)d2) << 8	718	| ((UV)d2) << 8
…		…
497		721
498	fail:	722	fail:
499	return (UV)-1;	723	return (UV)-1;
500	}	724	}
501		725
502	#define APPEND_GROW(n) SB \
503	if (cur + (n) >= end) \
504	{ \
505	STRLEN ofs = cur - SvPVX (sv); \
506	SvGROW (sv, ofs + (n) + 1); \
507	cur = SvPVX (sv) + ofs; \
508	end = SvEND (sv); \
509	} \
510	SE
511
512	#define APPEND_CH(ch) SB \
513	APPEND_GROW (1); \
514	*cur++ = (ch); \
515	SE
516
517	static SV *	726	static SV *
518	decode_str (dec_t *dec)	727	decode_str (dec_t *dec)
519	{	728	{
520	SV *sv = NEWSV (0,2);	729	SV *sv = 0;
521	int utf8 = 0;	730	int utf8 = 0;
522	char *cur = SvPVX (sv);	731	char *dec_cur = dec->cur;
523	char *end = SvEND (sv);
524		732
525	for (;;)	733	do
526	{	734	{
527	unsigned char ch = *(unsigned char *)dec->cur;	735	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		736	char *cur = buf;
528		737
529	if (ch == '"')	738	do
530	break;
531	else if (ch == '\\')
532	{	739	{
533	switch (*++dec->cur)	740	unsigned char ch = *(unsigned char *)dec_cur++;
		741
		742	if (expect_false (ch == '"'))
		743	{
		744	--dec_cur;
		745	break;
534	{	746	}
535	case '\\':	747	else if (expect_false (ch == '\\'))
536	case '/':	748	{
537	case '"': APPEND_CH (*dec->cur++); break;	749	switch (*dec_cur)
538
539	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
540	case 't': APPEND_CH ('\011'); ++dec->cur; break;
541	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
542	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
543	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
544
545	case 'u':
546	{	750	{
547	UV lo, hi;	751	case '\\':
548	++dec->cur;	752	case '/':
		753	case '"': *cur++ = *dec_cur++; break;
549		754
550	hi = decode_4hex (dec);	755	case 'b': ++dec_cur; *cur++ = '\010'; break;
551	if (hi == (UV)-1)	756	case 't': ++dec_cur; *cur++ = '\011'; break;
552	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;
553		760
554	// possibly a surrogate pair	761	case 'u':
555	if (hi >= 0xd800 && hi < 0xdc00)
556	{	762	{
557	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	763	UV lo, hi;
558	ERR ("missing low surrogate character in surrogate pair");	764	++dec_cur;
559		765
560	dec->cur += 2;	766	dec->cur = dec_cur;
561
562	lo = decode_4hex (dec);	767	hi = decode_4hex (dec);
		768	dec_cur = dec->cur;
563	if (lo == (UV)-1)	769	if (hi == (UV)-1)
564	goto fail;	770	goto fail;
565		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
566	if (lo < 0xdc00 || lo >= 0xe000)	787	if (lo < 0xdc00 || lo >= 0xe000)
567	ERR ("surrogate pair expected");	788	ERR ("surrogate pair expected");
568		789
569	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;
570	}	803	}
571	else if (hi >= 0xdc00 && hi < 0xe000)
572	ERR ("missing high surrogate character in surrogate pair");
573
574	if (hi >= 0x80)
575	{	804	break;
576	utf8 = 1;
577		805
578	APPEND_GROW (4); // at most 4 bytes for 21 bits
579	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	}
581	else	806	default:
582	APPEND_CH (hi);	807	--dec_cur;
		808	ERR ("illegal backslash escape sequence in string");
583	}	809	}
584	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;
585		817
586	default:
587	--dec->cur;	818	--dec_cur;
588	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");
589	}	838	}
590	}	839	}
591	else if (ch >= 0x20 && ch <= 0x7f)	840	while (cur < buf + SHORT_STRING_LEN);
592	APPEND_CH (*dec->cur++);	841
593	else if (ch >= 0x80)
594	{	842	{
595	STRLEN clen;	843	STRLEN len = cur - buf;
596	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
597	if (clen == (STRLEN)-1)
598	ERR ("malformed UTF-8 character in JSON string");
599		844
600	APPEND_GROW (clen);	845	if (sv)
601	do
602	{	846	{
603	*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);
604	}	850	}
605	while (--clen);
606
607	utf8 = 1;
608	}
609	else if (dec->cur == dec->end)
610	ERR ("unexpected end of string while parsing json string");
611	else	851	else
612	ERR ("invalid character encountered");	852	sv = newSVpvn (buf, len);
613	}	853	}
		854	}
		855	while (*dec_cur != '"');
614		856
615	++dec->cur;	857	++dec_cur;
616		858
617	SvCUR_set (sv, cur - SvPVX (sv));	859	if (sv)
618		860	{
619	SvPOK_only (sv);	861	SvPOK_only (sv);
620	*SvEND (sv) = 0;	862	*SvEND (sv) = 0;
621		863
622	if (utf8)	864	if (utf8)
623	SvUTF8_on (sv);	865	SvUTF8_on (sv);
		866	}
		867	else
		868	sv = newSVpvn ("", 0);
624		869
625	if (dec->flags & F_SHRINK)	870	dec->cur = dec_cur;
626	shrink (sv);
627
628	return sv;	871	return sv;
629		872
630	fail:	873	fail:
631	SvREFCNT_dec (sv);	874	dec->cur = dec_cur;
632	return 0;	875	return 0;
633	}	876	}
634		877
635	static SV *	878	static SV *
636	decode_num (dec_t *dec)	879	decode_num (dec_t *dec)
…		…
694	is_nv = 1;	937	is_nv = 1;
695	}	938	}
696		939
697	if (!is_nv)	940	if (!is_nv)
698	{	941	{
699	UV uv;	942	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
700	int numtype = grok_number (start, dec->cur - start, &uv);	943	if (*start == '-')
701	if (numtype & IS_NUMBER_IN_UV)	944	switch (dec->cur - start)
702	if (numtype & IS_NUMBER_NEG)
703	{	945	{
704	if (uv < (UV)IV_MIN)	946	case 2: return newSViv (-( start [1] - '0' * 1));
705	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));
706	}	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	}
707	else	969	else
708	return newSVuv (uv);	970	return newSVuv (uv);
		971
		972	// here would likely be the place for bigint support
709	}	973	}
		974	}
710		975
		976	// if we ever support bigint or bigfloat, this is the place for bigfloat
711	return newSVnv (Atof (start));	977	return newSVnv (Atof (start));
712		978
713	fail:	979	fail:
714	return 0;	980	return 0;
715	}	981	}
…		…
717	static SV *	983	static SV *
718	decode_av (dec_t *dec)	984	decode_av (dec_t *dec)
719	{	985	{
720	AV *av = newAV ();	986	AV *av = newAV ();
721		987
722	WS;	988	DEC_INC_DEPTH;
		989	decode_ws (dec);
		990
723	if (*dec->cur == ']')	991	if (*dec->cur == ']')
724	++dec->cur;	992	++dec->cur;
725	else	993	else
726	for (;;)	994	for (;;)
727	{	995	{
…		…
731	if (!value)	999	if (!value)
732	goto fail;	1000	goto fail;
733		1001
734	av_push (av, value);	1002	av_push (av, value);
735		1003
736	WS;	1004	decode_ws (dec);
737		1005
738	if (*dec->cur == ']')	1006	if (*dec->cur == ']')
739	{	1007	{
740	++dec->cur;	1008	++dec->cur;
741	break;	1009	break;
…		…
745	ERR (", or ] expected while parsing array");	1013	ERR (", or ] expected while parsing array");
746		1014
747	++dec->cur;	1015	++dec->cur;
748	}	1016	}
749		1017
		1018	DEC_DEC_DEPTH;
750	return newRV_noinc ((SV *)av);	1019	return newRV_noinc ((SV *)av);
751		1020
752	fail:	1021	fail:
753	SvREFCNT_dec (av);	1022	SvREFCNT_dec (av);
		1023	DEC_DEC_DEPTH;
754	return 0;	1024	return 0;
755	}	1025	}
756		1026
757	static SV *	1027	static SV *
758	decode_hv (dec_t *dec)	1028	decode_hv (dec_t *dec)
759	{	1029	{
		1030	SV *sv;
760	HV *hv = newHV ();	1031	HV *hv = newHV ();
761		1032
762	WS;	1033	DEC_INC_DEPTH;
		1034	decode_ws (dec);
		1035
763	if (*dec->cur == '}')	1036	if (*dec->cur == '}')
764	++dec->cur;	1037	++dec->cur;
765	else	1038	else
766	for (;;)	1039	for (;;)
767	{	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	{
768	SV *key, *value;	1049	SV *value;
		1050	char *p = dec->cur;
		1051	char *e = p + 24; // only try up to 24 bytes
769		1052
770	WS; EXPECT_CH ('"');	1053	for (;;)
771
772	key = decode_str (dec);
773	if (!key)
774	goto fail;
775
776	WS; EXPECT_CH (':');
777
778	value = decode_sv (dec);
779	if (!value)
780	{	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);
781	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)
782	goto fail;	1088	goto fail;
		1089
		1090	hv_store (hv, key, len, value, 0);
		1091
		1092	break;
		1093	}
		1094
		1095	++p;
783	}	1096	}
784
785	//TODO: optimise
786	hv_store_ent (hv, key, value, 0);
787
788	WS;	1097	}
		1098
		1099	decode_ws (dec);
789		1100
790	if (*dec->cur == '}')	1101	if (*dec->cur == '}')
791	{	1102	{
792	++dec->cur;	1103	++dec->cur;
793	break;	1104	break;
…		…
797	ERR (", or } expected while parsing object/hash");	1108	ERR (", or } expected while parsing object/hash");
798		1109
799	++dec->cur;	1110	++dec->cur;
800	}	1111	}
801		1112
		1113	DEC_DEC_DEPTH;
802	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;
803		1175
804	fail:	1176	fail:
805	SvREFCNT_dec (hv);	1177	SvREFCNT_dec (hv);
		1178	DEC_DEC_DEPTH;
806	return 0;	1179	return 0;
807	}	1180	}
808		1181
809	static SV *	1182	static SV *
810	decode_sv (dec_t *dec)	1183	decode_sv (dec_t *dec)
811	{	1184	{
812	WS;	1185	decode_ws (dec);
		1186
		1187	// the beauty of JSON: you need exactly one character lookahead
		1188	// to parse anything.
813	switch (*dec->cur)	1189	switch (*dec->cur)
814	{	1190	{
815	case '"': ++dec->cur; return decode_str (dec);	1191	case '"': ++dec->cur; return decode_str (dec);
816	case '[': ++dec->cur; return decode_av (dec);	1192	case '[': ++dec->cur; return decode_av (dec);
817	case '{': ++dec->cur; return decode_hv (dec);	1193	case '{': ++dec->cur; return decode_hv (dec);
…		…
823		1199
824	case 't':	1200	case 't':
825	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826	{	1202	{
827	dec->cur += 4;	1203	dec->cur += 4;
828	return newSViv (1);	1204	return SvREFCNT_inc (json_true);
829	}	1205	}
830	else	1206	else
831	ERR ("'true' expected");	1207	ERR ("'true' expected");
832		1208
833	break;	1209	break;
834		1210
835	case 'f':	1211	case 'f':
836	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837	{	1213	{
838	dec->cur += 5;	1214	dec->cur += 5;
839	return newSViv (0);	1215	return SvREFCNT_inc (json_false);
840	}	1216	}
841	else	1217	else
842	ERR ("'false' expected");	1218	ERR ("'false' expected");
843		1219
844	break;	1220	break;
…		…
853	ERR ("'null' expected");	1229	ERR ("'null' expected");
854		1230
855	break;	1231	break;
856		1232
857	default:	1233	default:
858	ERR ("malformed json string, neither array, object, number, string or atom");	1234	ERR ("malformed JSON string, neither array, object, number, string or atom");
859	break;	1235	break;
860	}	1236	}
861		1237
862	fail:	1238	fail:
863	return 0;	1239	return 0;
864	}	1240	}
865		1241
866	static SV *	1242	static SV *
867	decode_json (SV *string, UV flags)	1243	decode_json (SV *string, JSON *json, UV *offset_return)
868	{	1244	{
		1245	dec_t dec;
		1246	UV offset;
869	SV *sv;	1247	SV *sv;
870		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
871	if (flags & F_UTF8)	1256	if (json->flags & F_UTF8)
872	sv_utf8_downgrade (string, 0);	1257	sv_utf8_downgrade (string, 0);
873	else	1258	else
874	sv_utf8_upgrade (string);	1259	sv_utf8_upgrade (string);
875		1260
876	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1261	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877		1262
878	dec_t dec;	1263	dec.json = *json;
879	dec.flags = flags;
880	dec.cur = SvPVX (string);	1264	dec.cur = SvPVX (string);
881	dec.end = SvEND (string);	1265	dec.end = SvEND (string);
882	dec.err = 0;	1266	dec.err = 0;
		1267	dec.depth = 0;
		1268	dec.maxdepth = DEC_DEPTH (dec.json.flags);
883		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
884	sv = decode_sv (&dec);	1274	sv = decode_sv (&dec);
885		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
886	if (!sv)	1299	if (!sv)
887	{	1300	{
888	IV offset = dec.flags & F_UTF8
889	? dec.cur - SvPVX (string)
890	: utf8_distance (dec.cur, SvPVX (string));
891	SV *uni = sv_newmortal ();	1301	SV *uni = sv_newmortal ();
892		1302
893	// horrible hack to silence warning inside pv_uni_display	1303	// horrible hack to silence warning inside pv_uni_display
894	COP cop = *PL_curcop;	1304	COP cop = *PL_curcop;
895	cop.cop_warnings = pWARN_NONE;	1305	cop.cop_warnings = pWARN_NONE;
…		…
897	SAVEVPTR (PL_curcop);	1307	SAVEVPTR (PL_curcop);
898	PL_curcop = &cop;	1308	PL_curcop = &cop;
899	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);
900	LEAVE;	1310	LEAVE;
901		1311
902	croak ("%s, at character offset %d (%s)",	1312	croak ("%s, at character offset %d [\"%s\"]",
903	dec.err,	1313	dec.err,
904	(int)offset,	1314	(int)offset,
905	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1315	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906	}	1316	}
907		1317
908	sv = sv_2mortal (sv);	1318	sv = sv_2mortal (sv);
909		1319
910	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1320	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
911	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)");
912		1322
913	return sv;	1323	return sv;
914	}	1324	}
915		1325
		1326	/////////////////////////////////////////////////////////////////////////////
		1327	// XS interface functions
		1328
916	MODULE = JSON::XS PACKAGE = JSON::XS	1329	MODULE = JSON::XS PACKAGE = JSON::XS
917		1330
918	BOOT:	1331	BOOT:
919	{	1332	{
920	int i;	1333	int i;
921		1334
922	memset (decode_hexdigit, 0xff, 256);
923	for (i = 10; i--; )	1335	for (i = 0; i < 256; ++i)
924	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;
925		1341
926	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
927	{	1434	{
928	decode_hexdigit ['a' + i] = 10 + i;	1435	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
929	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	}
930	}	1442	}
931		1443
932	json_stash = gv_stashpv ("JSON::XS", 1);	1444	XPUSHs (ST (0));
933	}	1445	}
934		1446
935	PROTOTYPES: DISABLE	1447	void encode (JSON *self, SV *scalar)
		1448	PPCODE:
		1449	XPUSHs (encode_json (scalar, self));
936		1450
937	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)
938	CODE:	1465	CODE:
939	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1466	SvREFCNT_dec (self->cb_sk_object);
940	OUTPUT:	1467	SvREFCNT_dec (self->cb_object);
941	RETVAL
942
943	SV *ascii (SV *self, int enable = 1)
944	ALIAS:
945	ascii = F_ASCII
946	utf8 = F_UTF8
947	indent = F_INDENT
948	canonical = F_CANONICAL
949	space_before = F_SPACE_BEFORE
950	space_after = F_SPACE_AFTER
951	json_rpc = F_JSON_RPC
952	pretty = F_PRETTY
953	allow_nonref = F_ALLOW_NONREF
954	shrink = F_SHRINK
955	CODE:
956	{
957	UV *uv = SvJSON (self);
958	if (enable)
959	*uv |= ix;
960	else
961	*uv &= ~ix;
962
963	RETVAL = newSVsv (self);
964	}
965	OUTPUT:
966	RETVAL
967
968	void encode (SV *self, SV *scalar)
969	PPCODE:
970	XPUSHs (encode_json (scalar, *SvJSON (self)));
971
972	void decode (SV *self, SV *jsonstr)
973	PPCODE:
974	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
975		1468
976	PROTOTYPES: ENABLE	1469	PROTOTYPES: ENABLE
977		1470
978	void to_json (SV *scalar)	1471	void to_json (SV *scalar)
979	PPCODE:	1472	PPCODE:
		1473	{
		1474	JSON json = { F_DEFAULT | F_UTF8 };
980	XPUSHs (encode_json (scalar, F_UTF8));	1475	XPUSHs (encode_json (scalar, &json));
		1476	}
981		1477
982	void from_json (SV *jsonstr)	1478	void from_json (SV *jsonstr)
983	PPCODE:	1479	PPCODE:
		1480	{
		1481	JSON json = { F_DEFAULT | F_UTF8 };
984	XPUSHs (decode_json (jsonstr, F_UTF8));	1482	XPUSHs (decode_json (jsonstr, &json, 0));
		1483	}
985		1484

Diff Legend

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