ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.7 by root, Fri Mar 23 15:57:18 2007 UTC vs.
Revision 1.50 by root, Mon Jul 2 00:29:38 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, *cb_sk_object;
		68	} JSON;
		69
		70	/////////////////////////////////////////////////////////////////////////////
		71	// utility functions
		72
		73	inline void
		74	shrink (SV *sv)
		75	{
		76	sv_utf8_downgrade (sv, 1);
		77	if (SvLEN (sv) > SvCUR (sv) + 1)
		78	{
		79	#ifdef SvPV_shrink_to_cur
		80	SvPV_shrink_to_cur (sv);
		81	#elif defined (SvPV_renew)
		82	SvPV_renew (sv, SvCUR (sv) + 1);
		83	#endif
		84	}
		85	}
		86
		87	// decode an utf-8 character and return it, or (UV)-1 in
		88	// case of an error.
		89	// we special-case "safe" characters from U+80 .. U+7FF,
		90	// but use the very good perl function to parse anything else.
		91	// note that we never call this function for a ascii codepoints
		92	inline UV
		93	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		94	{
		95	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		98	{
		99	*clen = 2;
		100	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		101	}
		102	else
		103	{
		104	*clen = (STRLEN)-1;
		105	return (UV)-1;
		106	}
		107	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// encoder
28		111
29	// structure used for encoding JSON	112	// structure used for encoding JSON
30	typedef struct	113	typedef struct
31	{	114	{
32	char *cur;	115	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	116	char *end; // SvEND (sv)
35	SV *sv;	117	SV *sv; // result scalar
36	UV flags;	118	JSON json;
37	int max_recurse;	119	U32 indent; // indentation level
38	int indent;	120	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	121	} enc_t;
40		122
41	// structure used for decoding JSON	123	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)	124	need (enc_t *enc, STRLEN len)
72	{	125	{
73	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
74	{	127	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	132	}
80	}	133	}
81		134
82	static void	135	inline void
83	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
84	{	137	{
85	need (enc, 1);	138	need (enc, 1);
86	*enc->cur++ = ch;	139	*enc->cur++ = ch;
87	}	140	}
…		…
95		148
96	while (str < end)	149	while (str < end)
97	{	150	{
98	unsigned char ch = *(unsigned char *)str;	151	unsigned char ch = *(unsigned char *)str;
99		152
100	if (ch >= 0x20 && ch < 0x80) // most common case	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	154	{
102	if (ch == '"') // but with slow exceptions	155	if (expect_false (ch == '"')) // but with slow exceptions
103	{	156	{
104	need (enc, len += 1);	157	need (enc, len += 1);
105	*enc->cur++ = '\\';	158	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	159	*enc->cur++ = '"';
107	}	160	}
108	else if (ch == '\\')	161	else if (expect_false (ch == '\\'))
109	{	162	{
110	need (enc, len += 1);	163	need (enc, len += 1);
111	*enc->cur++ = '\\';	164	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
113	}	166	}
…		…
131	STRLEN clen;	184	STRLEN clen;
132	UV uch;	185	UV uch;
133		186
134	if (is_utf8)	187	if (is_utf8)
135	{	188	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	189	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	190	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	191	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	192	}
140	else	193	else
141	{	194	{
142	uch = ch;	195	uch = ch;
143	clen = 1;	196	clen = 1;
144	}	197	}
145		198
146	if (uch < 0x80 || enc->flags & F_ASCII)	199	if (uch > 0x10FFFFUL)
		200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		201
		202	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
147	{	203	{
148	if (uch > 0xFFFFUL)	204	if (uch > 0xFFFFUL)
149	{	205	{
150	need (enc, len += 11);	206	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	208	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	209	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	210	enc->cur += 12;
155	}	211	}
156	else	212	else
157	{	213	{
158	static char hexdigit [16] = "0123456789abcdef";	214	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	221	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	222	}
167		223
168	str += clen;	224	str += clen;
169	}	225	}
		226	else if (enc->json.flags & F_LATIN1)
		227	{
		228	*enc->cur++ = uch;
		229	str += clen;
		230	}
170	else if (is_utf8)	231	else if (is_utf8)
171	{	232	{
172	need (enc, len += clen);	233	need (enc, len += clen);
173	do	234	do
174	{	235	{
…		…
176	}	237	}
177	while (--clen);	238	while (--clen);
178	}	239	}
179	else	240	else
180	{	241	{
181	need (enc, len += 10); // never more than 11 bytes needed	242	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	243	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	244	++str;
184	}	245	}
185	}	246	}
186	}	247	}
…		…
188		249
189	--len;	250	--len;
190	}	251	}
191	}	252	}
192		253
193	#define INDENT SB \	254	inline void
		255	encode_indent (enc_t *enc)
		256	{
194	if (enc->flags & F_INDENT) \	257	if (enc->json.flags & F_INDENT)
195	{ \	258	{
196	int i_; \	259	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	260
198	for (i_ = enc->indent * 3; i_--; )\	261	need (enc, spaces);
		262	memset (enc->cur, ' ', spaces);
		263	enc->cur += spaces;
		264	}
		265	}
		266
		267	inline void
		268	encode_space (enc_t *enc)
		269	{
		270	need (enc, 1);
		271	encode_ch (enc, ' ');
		272	}
		273
		274	inline void
		275	encode_nl (enc_t *enc)
		276	{
		277	if (enc->json.flags & F_INDENT)
		278	{
		279	need (enc, 1);
199	encode_ch (enc, ' '); \	280	encode_ch (enc, '\n');
200	} \	281	}
201	SE	282	}
202		283
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	284	inline void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	285	encode_comma (enc_t *enc)
205	#define COMMA SB \	286	{
206	encode_ch (enc, ','); \	287	encode_ch (enc, ',');
		288
207	if (enc->flags & F_INDENT) \	289	if (enc->json.flags & F_INDENT)
208	NL; \	290	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	291	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	292	encode_space (enc);
211	SE	293	}
212		294
213	static void encode_sv (enc_t *enc, SV *sv);	295	static void encode_sv (enc_t *enc, SV *sv);
214		296
215	static void	297	static void
216	encode_av (enc_t *enc, AV *av)	298	encode_av (enc_t *enc, AV *av)
217	{	299	{
218	int i, len = av_len (av);	300	int i, len = av_len (av);
219		301
220	encode_ch (enc, '['); NL;	302	if (enc->indent >= enc->maxdepth)
		303	croak ("data structure too deep (hit recursion limit)");
		304
		305	encode_ch (enc, '['); encode_nl (enc);
221	++enc->indent;	306	++enc->indent;
222		307
223	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
224	{	309	{
225	INDENT;	310	encode_indent (enc);
226	encode_sv (enc, *av_fetch (av, i, 0));	311	encode_sv (enc, *av_fetch (av, i, 0));
227		312
228	if (i < len)	313	if (i < len)
229	COMMA;	314	encode_comma (enc);
230	}	315	}
231		316
232	NL;	317	encode_nl (enc);
233		318
234	--enc->indent;	319	--enc->indent;
235	INDENT; encode_ch (enc, ']');	320	encode_indent (enc); encode_ch (enc, ']');
236	}	321	}
237		322
238	static void	323	static void
239	encode_he (enc_t *enc, HE *he)	324	encode_he (enc_t *enc, HE *he)
240	{	325	{
…		…
254	else	339	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	340	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		341
257	encode_ch (enc, '"');	342	encode_ch (enc, '"');
258		343
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	344	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	345	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	346	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));	347	encode_sv (enc, HeVAL (he));
263	}	348	}
264		349
265	// compare hash entries, used when all keys are bytestrings	350	// compare hash entries, used when all keys are bytestrings
266	static int	351	static int
…		…
272	HE *b = *(HE **)b_;	357	HE *b = *(HE **)b_;
273		358
274	STRLEN la = HeKLEN (a);	359	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	360	STRLEN lb = HeKLEN (b);
276		361
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	363	cmp = la - lb;
279		364
280	return cmp;	365	return cmp;
281	}	366	}
282		367
283	// compare hash entries, used when some keys are sv's or utf-x	368	// compare hash entries, used when some keys are sv's or utf-x
…		…
290	static void	375	static void
291	encode_hv (enc_t *enc, HV *hv)	376	encode_hv (enc_t *enc, HV *hv)
292	{	377	{
293	int count, i;	378	int count, i;
294		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
295	encode_ch (enc, '{'); NL; ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
296		384
297	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
298	{	386	{
299	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	390	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	391	if (enc->json.flags & F_CANONICAL)
304	{	392	{
305	HE *he, *hes [count];
306	int fast = 1;	393	int fast = 1;
		394	HE *he;
		395	#if defined(__BORLANDC__) || defined(_MSC_VER)
		396	HE **hes = _alloca (count * sizeof (HE));
		397	#else
		398	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		399	#endif
307		400
308	i = 0;	401	i = 0;
309	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
310	{	403	{
311	hes [i++] = he;	404	hes [i++] = he;
…		…
317		410
318	if (fast)	411	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	412	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	413	else
321	{	414	{
322	// hack to disable "use bytes"	415	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	416	COP cop = *PL_curcop;
324	cop.op_private = 0;	417	cop.op_private = 0;
		418
		419	ENTER;
		420	SAVETMPS;
		421
		422	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	423	PL_curcop = &cop;
326		424
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	425	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		426
329	FREETMPS;	427	FREETMPS;
330		428	LEAVE;
331	PL_curcop = oldcop;
332	}	429	}
333		430
334	for (i = 0; i < count; ++i)	431	for (i = 0; i < count; ++i)
335	{	432	{
336	INDENT;	433	encode_indent (enc);
337	encode_he (enc, hes [i]);	434	encode_he (enc, hes [i]);
338		435
339	if (i < count - 1)	436	if (i < count - 1)
340	COMMA;	437	encode_comma (enc);
341	}	438	}
342		439
343	NL;	440	encode_nl (enc);
344	}	441	}
345	else	442	else
346	{	443	{
347	SV *sv;
348	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
349		445
350	for (;;)	446	for (;;)
351	{	447	{
352	INDENT;	448	encode_indent (enc);
353	encode_he (enc, he);	449	encode_he (enc, he);
354		450
355	if (!(he = hv_iternext (hv)))	451	if (!(he = hv_iternext (hv)))
356	break;	452	break;
357		453
358	COMMA;	454	encode_comma (enc);
359	}	455	}
360		456
361	NL;	457	encode_nl (enc);
362	}	458	}
363	}	459	}
364		460
365	--enc->indent; INDENT; encode_ch (enc, '}');	461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		462	}
		463
		464	// encode objects, arrays and special \0=false and \1=true values.
		465	static void
		466	encode_rv (enc_t *enc, SV *sv)
		467	{
		468	svtype svt;
		469
		470	SvGETMAGIC (sv);
		471	svt = SvTYPE (sv);
		472
		473	if (expect_false (SvOBJECT (sv)))
		474	{
		475	if (SvSTASH (sv) == json_boolean_stash)
		476	{
		477	if (SvIV (sv) == 0)
		478	encode_str (enc, "false", 5, 0);
		479	else
		480	encode_str (enc, "true", 4, 0);
		481	}
		482	else
		483	{
		484	#if 0
		485	if (0 && sv_derived_from (rv, "JSON::Literal"))
		486	{
		487	// not yet
		488	}
		489	#endif
		490	if (enc->json.flags & F_CONV_BLESSED)
		491	{
		492	// we re-bless the reference to get overload and other niceties right
		493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		494
		495	if (to_json)
		496	{
		497	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
		498	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		499
		500	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		501	// check anyways.
		502	PUTBACK;
		503	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		504	SPAGAIN;
		505
		506	encode_sv (enc, POPs);
		507
		508	FREETMPS; LEAVE;
		509	}
		510	else if (enc->json.flags & F_ALLOW_BLESSED)
		511	encode_str (enc, "null", 4, 0);
		512	else
		513	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		514	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		515	}
		516	else if (enc->json.flags & F_ALLOW_BLESSED)
		517	encode_str (enc, "null", 4, 0);
		518	else
		519	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		520	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		521	}
		522	}
		523	else if (svt == SVt_PVHV)
		524	encode_hv (enc, (HV *)sv);
		525	else if (svt == SVt_PVAV)
		526	encode_av (enc, (AV *)sv);
		527	else if (svt < SVt_PVAV)
		528	{
		529	STRLEN len = 0;
		530	char *pv = svt ? SvPV (sv, len) : 0;
		531
		532	if (len == 1 && *pv == '0')
		533	encode_str (enc, "false", 5, 0);
		534	else if (len == 1 && *pv == '1')
		535	encode_str (enc, "true", 4, 0);
		536	else
		537	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		538	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		539	}
		540	else
		541	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	543	}
367		544
368	static void	545	static void
369	encode_sv (enc_t *enc, SV *sv)	546	encode_sv (enc_t *enc, SV *sv)
370	{	547	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	555	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	556	encode_ch (enc, '"');
380	}	557	}
381	else if (SvNOKp (sv))	558	else if (SvNOKp (sv))
382	{	559	{
		560	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	561	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	562	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	563	enc->cur += strlen (enc->cur);
386	}	564	}
387	else if (SvIOKp (sv))	565	else if (SvIOKp (sv))
388	{	566	{
389	need (enc, 64);	567	// we assume we can always read an IV as a UV
		568	if (SvUV (sv) & ~(UV)0x7fff)
		569	{
		570	// large integer, use the (rather slow) snprintf way.
		571	need (enc, sizeof (UV) * 3);
390	enc->cur +=	572	enc->cur +=
391	SvIsUV(sv)	573	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	574	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	575	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		576	}
		577	else
		578	{
		579	// optimise the "small number case"
		580	// code will likely be branchless and use only a single multiplication
		581	I32 i = SvIV (sv);
		582	U32 u;
		583	char digit, nz = 0;
		584
		585	need (enc, 6);
		586
		587	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		588	u = i < 0 ? -i : i;
		589
		590	// convert to 4.28 fixed-point representation
		591	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		592
		593	// now output digit by digit, each time masking out the integer part
		594	// and multiplying by 5 while moving the decimal point one to the right,
		595	// resulting in a net multiplication by 10.
		596	// we always write the digit to memory but conditionally increment
		597	// the pointer, to ease the usage of conditional move instructions.
		598	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		599	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		600	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		601	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		602	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		603	}
394	}	604	}
395	else if (SvROK (sv))	605	else if (SvROK (sv))
396	{	606	encode_rv (enc, SvRV (sv));
397	if (!--enc->max_recurse)
398	croak ("data structure too deep (hit recursion limit)");
399
400	sv = SvRV (sv);
401
402	switch (SvTYPE (sv))
403	{
404	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407	default:
408	croak ("JSON can only represent references to arrays or hashes");
409	}
410	}
411	else if (!SvOK (sv))	607	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	608	encode_str (enc, "null", 4, 0);
413	else	609	else
414	croak ("encountered perl type that JSON cannot handle");	610	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		611	SvPV_nolen (sv), SvFLAGS (sv));
415	}	612	}
416		613
417	static SV *	614	static SV *
418	encode_json (SV *scalar, UV flags)	615	encode_json (SV *scalar, JSON *json)
419	{	616	{
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	enc_t enc;	617	enc_t enc;
424	enc.flags = flags;	618
		619	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		620	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		621
		622	enc.json = *json;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	623	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	624	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	625	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	626	enc.indent = 0;
		627	enc.maxdepth = DEC_DEPTH (enc.json.flags);
430		628
431	SvPOK_only (enc.sv);	629	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	630	encode_sv (&enc, scalar);
433		631
		632	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		633	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		634
434	if (!(flags & (F_ASCII | F_UTF8)))	635	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	636	SvUTF8_on (enc.sv);
436		637
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438
439	if (enc.flags & F_SHRINK)	638	if (enc.json.flags & F_SHRINK)
440	shrink (enc.sv);	639	shrink (enc.sv);
441		640
442	return enc.sv;	641	return enc.sv;
443	}	642	}
444		643
445	/////////////////////////////////////////////////////////////////////////////	644	/////////////////////////////////////////////////////////////////////////////
		645	// decoder
446		646
447	#define WS \	647	// structure used for decoding JSON
		648	typedef struct
		649	{
		650	char *cur; // current parser pointer
		651	char *end; // end of input string
		652	const char *err; // parse error, if != 0
		653	JSON json;
		654	U32 depth; // recursion depth
		655	U32 maxdepth; // recursion depth limit
		656	} dec_t;
		657
		658	inline void
		659	decode_ws (dec_t *dec)
		660	{
448	for (;;) \	661	for (;;)
449	{ \	662	{
450	char ch = *dec->cur; \	663	char ch = *dec->cur;
		664
451	if (ch > 0x20 \	665	if (ch > 0x20
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	666	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
453	break; \	667	break;
		668
454	++dec->cur; \	669	++dec->cur;
455	}	670	}
		671	}
456		672
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	673	#define ERR(reason) SB dec->err = reason; goto fail; SE
		674
458	#define EXPECT_CH(ch) SB \	675	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	676	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	677	ERR (# ch " expected"); \
461	++dec->cur; \	678	++dec->cur; \
462	SE	679	SE
463		680
		681	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		682	#define DEC_DEC_DEPTH --dec->depth
		683
464	static SV *decode_sv (dec_t *dec);	684	static SV *decode_sv (dec_t *dec);
465		685
466	static signed char decode_hexdigit[256];	686	static signed char decode_hexdigit[256];
467		687
468	static UV	688	static UV
469	decode_4hex (dec_t *dec)	689	decode_4hex (dec_t *dec)
470	{	690	{
471	signed char d1, d2, d3, d4;	691	signed char d1, d2, d3, d4;
		692	unsigned char *cur = (unsigned char *)dec->cur;
472		693
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	694	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	695	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	696	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	697	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
477	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478	if (d3 < 0) ERR ("four hexadecimal digits expected");
479	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480	if (d4 < 0) ERR ("four hexadecimal digits expected");
481		698
482	dec->cur += 4;	699	dec->cur += 4;
483		700
484	return ((UV)d1) << 12	701	return ((UV)d1) << 12
485	| ((UV)d2) << 8	702	| ((UV)d2) << 8
…		…
488		705
489	fail:	706	fail:
490	return (UV)-1;	707	return (UV)-1;
491	}	708	}
492		709
493	#define APPEND_GROW(n) SB \
494	if (cur + (n) >= end) \
495	{ \
496	STRLEN ofs = cur - SvPVX (sv); \
497	SvGROW (sv, ofs + (n) + 1); \
498	cur = SvPVX (sv) + ofs; \
499	end = SvEND (sv); \
500	} \
501	SE
502
503	#define APPEND_CH(ch) SB \
504	APPEND_GROW (1); \
505	*cur++ = (ch); \
506	SE
507
508	static SV *	710	static SV *
509	decode_str (dec_t *dec)	711	decode_str (dec_t *dec)
510	{	712	{
511	SV *sv = NEWSV (0,2);	713	SV *sv = 0;
512	int utf8 = 0;	714	int utf8 = 0;
513	char *cur = SvPVX (sv);	715	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		716
516	for (;;)	717	do
517	{	718	{
518	unsigned char ch = *(unsigned char *)dec->cur;	719	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		720	char *cur = buf;
519		721
520	if (ch == '"')	722	do
521	break;
522	else if (ch == '\\')
523	{	723	{
524	switch (*++dec->cur)	724	unsigned char ch = *(unsigned char *)dec_cur++;
		725
		726	if (expect_false (ch == '"'))
525	{	727	{
526	case '\\':	728	--dec_cur;
527	case '/':	729	break;
528	case '"': APPEND_CH (*dec->cur++); break;	730	}
529		731	else if (expect_false (ch == '\\'))
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	732	{
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;	733	switch (*dec_cur)
532	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536	case 'u':
537	{	734	{
538	UV lo, hi;	735	case '\\':
539	++dec->cur;	736	case '/':
		737	case '"': *cur++ = *dec_cur++; break;
540		738
541	hi = decode_4hex (dec);	739	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	740	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	741	case 'n': ++dec_cur; *cur++ = '\012'; break;
		742	case 'f': ++dec_cur; *cur++ = '\014'; break;
		743	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		744
545	// possibly a surrogate pair	745	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	746	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	747	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	748	++dec_cur;
550		749
551	dec->cur += 2;	750	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	751	hi = decode_4hex (dec);
		752	dec_cur = dec->cur;
554	if (lo == (UV)-1)	753	if (hi == (UV)-1)
555	goto fail;	754	goto fail;
556		755
		756	// possibly a surrogate pair
		757	if (hi >= 0xd800)
		758	if (hi < 0xdc00)
		759	{
		760	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		761	ERR ("missing low surrogate character in surrogate pair");
		762
		763	dec_cur += 2;
		764
		765	dec->cur = dec_cur;
		766	lo = decode_4hex (dec);
		767	dec_cur = dec->cur;
		768	if (lo == (UV)-1)
		769	goto fail;
		770
557	if (lo < 0xdc00 || lo >= 0xe000)	771	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	772	ERR ("surrogate pair expected");
559		773
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	774	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		775	}
		776	else if (hi < 0xe000)
		777	ERR ("missing high surrogate character in surrogate pair");
		778
		779	if (hi >= 0x80)
		780	{
		781	utf8 = 1;
		782
		783	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		784	}
		785	else
		786	*cur++ = hi;
561	}	787	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	788	break;
567	utf8 = 1;
568		789
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	790	default:
573	APPEND_CH (hi);	791	--dec_cur;
		792	ERR ("illegal backslash escape sequence in string");
574	}	793	}
575	break;	794	}
		795	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		796	*cur++ = ch;
		797	else if (ch >= 0x80)
		798	{
		799	STRLEN clen;
		800	UV uch;
576		801
577	default:
578	--dec->cur;	802	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	803
		804	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		805	if (clen == (STRLEN)-1)
		806	ERR ("malformed UTF-8 character in JSON string");
		807
		808	do
		809	*cur++ = *dec_cur++;
		810	while (--clen);
		811
		812	utf8 = 1;
		813	}
		814	else
		815	{
		816	--dec_cur;
		817
		818	if (!ch)
		819	ERR ("unexpected end of string while parsing JSON string");
		820	else
		821	ERR ("invalid character encountered while parsing JSON string");
580	}	822	}
581	}	823	}
582	else if (ch >= 0x20 && ch <= 0x7f)	824	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	825
584	else if (ch >= 0x80)
585	{	826	{
586	STRLEN clen;	827	STRLEN len = cur - buf;
587	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	if (clen == (STRLEN)-1)
589	ERR ("malformed UTF-8 character in JSON string");
590		828
591	APPEND_GROW (clen);	829	if (sv)
592	do
593	{	830	{
594	*cur++ = *dec->cur++;	831	SvGROW (sv, SvCUR (sv) + len + 1);
		832	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		833	SvCUR_set (sv, SvCUR (sv) + len);
595	}	834	}
596	while (--clen);
597
598	utf8 = 1;
599	}
600	else if (dec->cur == dec->end)
601	ERR ("unexpected end of string while parsing json string");
602	else	835	else
603	ERR ("invalid character encountered");	836	sv = newSVpvn (buf, len);
604	}	837	}
		838	}
		839	while (*dec_cur != '"');
605		840
606	++dec->cur;	841	++dec_cur;
607		842
608	SvCUR_set (sv, cur - SvPVX (sv));	843	if (sv)
609		844	{
610	SvPOK_only (sv);	845	SvPOK_only (sv);
611	*SvEND (sv) = 0;	846	*SvEND (sv) = 0;
612		847
613	if (utf8)	848	if (utf8)
614	SvUTF8_on (sv);	849	SvUTF8_on (sv);
		850	}
		851	else
		852	sv = newSVpvn ("", 0);
615		853
616	if (dec->flags & F_SHRINK)	854	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	855	return sv;
620		856
621	fail:	857	fail:
622	SvREFCNT_dec (sv);	858	dec->cur = dec_cur;
623	return 0;	859	return 0;
624	}	860	}
625		861
626	static SV *	862	static SV *
627	decode_num (dec_t *dec)	863	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	921	is_nv = 1;
686	}	922	}
687		923
688	if (!is_nv)	924	if (!is_nv)
689	{	925	{
690	UV uv;	926	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
691	int numtype = grok_number (start, dec->cur - start, &uv);	927	if (*start == '-')
692	if (numtype & IS_NUMBER_IN_UV)	928	switch (dec->cur - start)
693	if (numtype & IS_NUMBER_NEG)
694	{	929	{
695	if (uv < (UV)IV_MIN)	930	case 2: return newSViv (-( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	931	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		932	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		933	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
697	}	934	}
		935	else
		936	switch (dec->cur - start)
		937	{
		938	case 1: return newSViv ( start [0] - '0' * 1);
		939	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		940	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		941	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		942	}
		943
		944	{
		945	UV uv;
		946	int numtype = grok_number (start, dec->cur - start, &uv);
		947	if (numtype & IS_NUMBER_IN_UV)
		948	if (numtype & IS_NUMBER_NEG)
		949	{
		950	if (uv < (UV)IV_MIN)
		951	return newSViv (-(IV)uv);
		952	}
698	else	953	else
699	return newSVuv (uv);	954	return newSVuv (uv);
		955
		956	// here would likely be the place for bigint support
700	}	957	}
		958	}
701		959
		960	// if we ever support bigint or bigfloat, this is the place for bigfloat
702	return newSVnv (Atof (start));	961	return newSVnv (Atof (start));
703		962
704	fail:	963	fail:
705	return 0;	964	return 0;
706	}	965	}
…		…
708	static SV *	967	static SV *
709	decode_av (dec_t *dec)	968	decode_av (dec_t *dec)
710	{	969	{
711	AV *av = newAV ();	970	AV *av = newAV ();
712		971
713	WS;	972	DEC_INC_DEPTH;
		973	decode_ws (dec);
		974
714	if (*dec->cur == ']')	975	if (*dec->cur == ']')
715	++dec->cur;	976	++dec->cur;
716	else	977	else
717	for (;;)	978	for (;;)
718	{	979	{
…		…
722	if (!value)	983	if (!value)
723	goto fail;	984	goto fail;
724		985
725	av_push (av, value);	986	av_push (av, value);
726		987
727	WS;	988	decode_ws (dec);
728		989
729	if (*dec->cur == ']')	990	if (*dec->cur == ']')
730	{	991	{
731	++dec->cur;	992	++dec->cur;
732	break;	993	break;
…		…
736	ERR (", or ] expected while parsing array");	997	ERR (", or ] expected while parsing array");
737		998
738	++dec->cur;	999	++dec->cur;
739	}	1000	}
740		1001
		1002	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1003	return newRV_noinc ((SV *)av);
742		1004
743	fail:	1005	fail:
744	SvREFCNT_dec (av);	1006	SvREFCNT_dec (av);
		1007	DEC_DEC_DEPTH;
745	return 0;	1008	return 0;
746	}	1009	}
747		1010
748	static SV *	1011	static SV *
749	decode_hv (dec_t *dec)	1012	decode_hv (dec_t *dec)
750	{	1013	{
		1014	SV *sv;
751	HV *hv = newHV ();	1015	HV *hv = newHV ();
752		1016
753	WS;	1017	DEC_INC_DEPTH;
		1018	decode_ws (dec);
		1019
754	if (*dec->cur == '}')	1020	if (*dec->cur == '}')
755	++dec->cur;	1021	++dec->cur;
756	else	1022	else
757	for (;;)	1023	for (;;)
758	{	1024	{
		1025	decode_ws (dec); EXPECT_CH ('"');
		1026
		1027	// heuristic: assume that
		1028	// a) decode_str + hv_store_ent are abysmally slow.
		1029	// b) most hash keys are short, simple ascii text.
		1030	// => try to "fast-match" such strings to avoid
		1031	// the overhead of decode_str + hv_store_ent.
		1032	{
759	SV *key, *value;	1033	SV *value;
		1034	char *p = dec->cur;
		1035	char *e = p + 24; // only try up to 24 bytes
760		1036
761	WS; EXPECT_CH ('"');	1037	for (;;)
762
763	key = decode_str (dec);
764	if (!key)
765	goto fail;
766
767	WS; EXPECT_CH (':');
768
769	value = decode_sv (dec);
770	if (!value)
771	{	1038	{
		1039	// the >= 0x80 is true on most architectures
		1040	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1041	{
		1042	// slow path, back up and use decode_str
		1043	SV *key = decode_str (dec);
		1044	if (!key)
		1045	goto fail;
		1046
		1047	decode_ws (dec); EXPECT_CH (':');
		1048
		1049	value = decode_sv (dec);
		1050	if (!value)
		1051	{
		1052	SvREFCNT_dec (key);
		1053	goto fail;
		1054	}
		1055
		1056	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1057	SvREFCNT_dec (key);
		1058
		1059	break;
		1060	}
		1061	else if (*p == '"')
		1062	{
		1063	// fast path, got a simple key
		1064	char *key = dec->cur;
		1065	int len = p - key;
		1066	dec->cur = p + 1;
		1067
		1068	decode_ws (dec); EXPECT_CH (':');
		1069
		1070	value = decode_sv (dec);
		1071	if (!value)
773	goto fail;	1072	goto fail;
		1073
		1074	hv_store (hv, key, len, value, 0);
		1075
		1076	break;
		1077	}
		1078
		1079	++p;
774	}	1080	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1081	}
		1082
		1083	decode_ws (dec);
780		1084
781	if (*dec->cur == '}')	1085	if (*dec->cur == '}')
782	{	1086	{
783	++dec->cur;	1087	++dec->cur;
784	break;	1088	break;
…		…
788	ERR (", or } expected while parsing object/hash");	1092	ERR (", or } expected while parsing object/hash");
789		1093
790	++dec->cur;	1094	++dec->cur;
791	}	1095	}
792		1096
		1097	DEC_DEC_DEPTH;
		1098	sv = newRV_noinc ((SV *)hv);
		1099
		1100	// check filter callbacks
		1101	if (dec->json.flags & F_HOOK)
		1102	{
		1103	ENTER; SAVETMPS;
		1104
		1105	if (HvKEYS (hv) == 1 && dec->json.cb_sk_object)
		1106	{
		1107	int count;
		1108
		1109	dSP; PUSHMARK (SP);
		1110	XPUSHs (sv_2mortal (sv));
		1111
		1112	PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN;
		1113
		1114	if (count == 1)
		1115	sv = newSVsv (POPs);
		1116	else
		1117	SvREFCNT_inc (sv);
		1118	}
		1119
		1120	if (dec->json.cb_object)
		1121	{
		1122	int count;
		1123
		1124	dSP; ENTER; SAVETMPS; PUSHMARK (SP);
		1125	XPUSHs (sv_2mortal (sv));
		1126
		1127	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1128
		1129	if (count == 1)
		1130	sv = newSVsv (POPs);
		1131	else
		1132	SvREFCNT_inc (sv);
		1133	}
		1134
		1135	FREETMPS; LEAVE;
		1136	}
		1137
793	return newRV_noinc ((SV *)hv);	1138	return newRV_noinc ((SV *)hv);
794		1139
795	fail:	1140	fail:
796	SvREFCNT_dec (hv);	1141	SvREFCNT_dec (hv);
		1142	DEC_DEC_DEPTH;
797	return 0;	1143	return 0;
798	}	1144	}
799		1145
800	static SV *	1146	static SV *
801	decode_sv (dec_t *dec)	1147	decode_sv (dec_t *dec)
802	{	1148	{
803	WS;	1149	decode_ws (dec);
		1150
		1151	// the beauty of JSON: you need exactly one character lookahead
		1152	// to parse anything.
804	switch (*dec->cur)	1153	switch (*dec->cur)
805	{	1154	{
806	case '"': ++dec->cur; return decode_str (dec);	1155	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1156	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1157	case '{': ++dec->cur; return decode_hv (dec);
…		…
814		1163
815	case 't':	1164	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1165	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1166	{
818	dec->cur += 4;	1167	dec->cur += 4;
819	return newSViv (1);	1168	return SvREFCNT_inc (json_true);
820	}	1169	}
821	else	1170	else
822	ERR ("'true' expected");	1171	ERR ("'true' expected");
823		1172
824	break;	1173	break;
825		1174
826	case 'f':	1175	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1176	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1177	{
829	dec->cur += 5;	1178	dec->cur += 5;
830	return newSViv (0);	1179	return SvREFCNT_inc (json_false);
831	}	1180	}
832	else	1181	else
833	ERR ("'false' expected");	1182	ERR ("'false' expected");
834		1183
835	break;	1184	break;
…		…
844	ERR ("'null' expected");	1193	ERR ("'null' expected");
845		1194
846	break;	1195	break;
847		1196
848	default:	1197	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1198	ERR ("malformed JSON string, neither array, object, number, string or atom");
850	break;	1199	break;
851	}	1200	}
852		1201
853	fail:	1202	fail:
854	return 0;	1203	return 0;
855	}	1204	}
856		1205
857	static SV *	1206	static SV *
858	decode_json (SV *string, UV flags)	1207	decode_json (SV *string, JSON *json, UV *offset_return)
859	{	1208	{
		1209	dec_t dec;
		1210	UV offset;
860	SV *sv;	1211	SV *sv;
861		1212
		1213	SvGETMAGIC (string);
		1214	SvUPGRADE (string, SVt_PV);
		1215
		1216	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1217	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1218	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1219
862	if (flags & F_UTF8)	1220	if (json->flags & F_UTF8)
863	sv_utf8_downgrade (string, 0);	1221	sv_utf8_downgrade (string, 0);
864	else	1222	else
865	sv_utf8_upgrade (string);	1223	sv_utf8_upgrade (string);
866		1224
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1225	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1226
869	dec_t dec;	1227	dec.json = *json;
870	dec.flags = flags;
871	dec.cur = SvPVX (string);	1228	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1229	dec.end = SvEND (string);
873	dec.err = 0;	1230	dec.err = 0;
		1231	dec.depth = 0;
		1232	dec.maxdepth = DEC_DEPTH (dec.json.flags);
874		1233
		1234	if (dec.json.cb_object || dec.json.cb_sk_object)
		1235	dec.json.flags |= F_HOOK;
		1236
		1237	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
875	sv = decode_sv (&dec);	1238	sv = decode_sv (&dec);
876		1239
		1240	if (!(offset_return || !sv))
		1241	{
		1242	// check for trailing garbage
		1243	decode_ws (&dec);
		1244
		1245	if (*dec.cur)
		1246	{
		1247	dec.err = "garbage after JSON object";
		1248	SvREFCNT_dec (sv);
		1249	sv = 0;
		1250	}
		1251	}
		1252
		1253	if (offset_return || !sv)
		1254	{
		1255	offset = dec.json.flags & F_UTF8
		1256	? dec.cur - SvPVX (string)
		1257	: utf8_distance (dec.cur, SvPVX (string));
		1258
		1259	if (offset_return)
		1260	*offset_return = offset;
		1261	}
		1262
877	if (!sv)	1263	if (!sv)
878	{	1264	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1265	SV *uni = sv_newmortal ();
		1266
883	// horrible hack to silence warning inside pv_uni_display	1267	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1268	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1269	cop.cop_warnings = pWARN_NONE;
		1270	ENTER;
887	SAVEVPTR (PL_curcop);	1271	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1272	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1273	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1274	LEAVE;
		1275
891	croak ("%s, at character offset %d (%s)",	1276	croak ("%s, at character offset %d [\"%s\"]",
892	dec.err,	1277	dec.err,
893	(int)offset,	1278	(int)offset,
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1279	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1280	}
896		1281
897	sv = sv_2mortal (sv);	1282	sv = sv_2mortal (sv);
898		1283
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1284	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
900	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1285	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1286
902	return sv;	1287	return sv;
903	}	1288	}
904		1289
		1290	/////////////////////////////////////////////////////////////////////////////
		1291	// XS interface functions
		1292
905	MODULE = JSON::XS PACKAGE = JSON::XS	1293	MODULE = JSON::XS PACKAGE = JSON::XS
906		1294
907	BOOT:	1295	BOOT:
908	{	1296	{
909	int i;	1297	int i;
910		1298
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1299	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1300	decode_hexdigit [i] =
		1301	i >= '0' && i <= '9' ? i - '0'
		1302	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1303	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1304	: -1;
914		1305
915	for (i = 7; i--; )	1306	json_stash = gv_stashpv ("JSON::XS" , 1);
		1307	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1308
		1309	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1311	}
		1312
		1313	PROTOTYPES: DISABLE
		1314
		1315	void new (char *klass)
		1316	PPCODE:
		1317	{
		1318	SV *pv = NEWSV (0, sizeof (JSON));
		1319	SvPOK_only (pv);
		1320	Zero (SvPVX (pv), 1, JSON);
		1321	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1322	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1323	}
		1324
		1325	void ascii (JSON *self, int enable = 1)
		1326	ALIAS:
		1327	ascii = F_ASCII
		1328	latin1 = F_LATIN1
		1329	utf8 = F_UTF8
		1330	indent = F_INDENT
		1331	canonical = F_CANONICAL
		1332	space_before = F_SPACE_BEFORE
		1333	space_after = F_SPACE_AFTER
		1334	pretty = F_PRETTY
		1335	allow_nonref = F_ALLOW_NONREF
		1336	shrink = F_SHRINK
		1337	allow_blessed = F_ALLOW_BLESSED
		1338	convert_blessed = F_CONV_BLESSED
		1339	PPCODE:
		1340	{
		1341	if (enable)
		1342	self->flags |= ix;
		1343	else
		1344	self->flags &= ~ix;
		1345
		1346	XPUSHs (ST (0));
		1347	}
		1348
		1349	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1350	PPCODE:
		1351	{
		1352	UV log2 = 0;
		1353
		1354	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1355
		1356	while ((1UL << log2) < max_depth)
		1357	++log2;
		1358
		1359	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1360
		1361	XPUSHs (ST (0));
		1362	}
		1363
		1364	void max_size (JSON *self, UV max_size = 0)
		1365	PPCODE:
		1366	{
		1367	UV log2 = 0;
		1368
		1369	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1370	if (max_size == 1) max_size = 2;
		1371
		1372	while ((1UL << log2) < max_size)
		1373	++log2;
		1374
		1375	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1376
		1377	XPUSHs (ST (0));
		1378	}
		1379
		1380	void filter_json_objects (JSON *self, SV *cb = &PL_sv_undef)
		1381	ALIAS:
		1382	filter_sk_json_objects = 1
		1383	PPCODE:
		1384	{
		1385	if (!SvOK (cb))
		1386	cb = 0;
		1387
		1388	switch (ix)
916	{	1389	{
917	decode_hexdigit ['a' + i] = 10 + i;	1390	case 0: self->cb_object = cb; break;
918	decode_hexdigit ['A' + i] = 10 + i;	1391	case 1: self->cb_sk_object = cb; break;
919	}	1392	}
920		1393
921	json_stash = gv_stashpv ("JSON::XS", 1);	1394	XPUSHs (ST (0));
922	}	1395	}
923		1396
924	PROTOTYPES: DISABLE
925
926	SV *new (char *dummy)
927	CODE:
928	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);
929	OUTPUT:
930	RETVAL
931
932	SV *ascii (SV *self, int enable = 1)
933	ALIAS:
934	ascii = F_ASCII
935	utf8 = F_UTF8
936	indent = F_INDENT
937	canonical = F_CANONICAL
938	space_before = F_SPACE_BEFORE
939	space_after = F_SPACE_AFTER
940	json_rpc = F_JSON_RPC
941	pretty = F_PRETTY
942	allow_nonref = F_ALLOW_NONREF
943	shrink = F_SHRINK
944	CODE:
945	{
946	UV *uv = SvJSON (self);
947	if (enable)
948	*uv |= ix;
949	else
950	*uv &= ~ix;
951
952	RETVAL = newSVsv (self);
953	}
954	OUTPUT:
955	RETVAL
956
957	void encode (SV *self, SV *scalar)	1397	void encode (JSON *self, SV *scalar)
958	PPCODE:	1398	PPCODE:
959	XPUSHs (encode_json (scalar, *SvJSON (self)));	1399	XPUSHs (encode_json (scalar, self));
960		1400
961	void decode (SV *self, SV *jsonstr)	1401	void decode (JSON *self, SV *jsonstr)
962	PPCODE:	1402	PPCODE:
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1403	XPUSHs (decode_json (jsonstr, self, 0));
		1404
		1405	void decode_prefix (JSON *self, SV *jsonstr)
		1406	PPCODE:
		1407	{
		1408	UV offset;
		1409	EXTEND (SP, 2);
		1410	PUSHs (decode_json (jsonstr, self, &offset));
		1411	PUSHs (sv_2mortal (newSVuv (offset)));
		1412	}
964		1413
965	PROTOTYPES: ENABLE	1414	PROTOTYPES: ENABLE
966		1415
967	void to_json (SV *scalar)	1416	void to_json (SV *scalar)
968	PPCODE:	1417	PPCODE:
		1418	{
		1419	JSON json = { F_DEFAULT | F_UTF8 };
969	XPUSHs (encode_json (scalar, F_UTF8));	1420	XPUSHs (encode_json (scalar, &json));
		1421	}
970		1422
971	void from_json (SV *jsonstr)	1423	void from_json (SV *jsonstr)
972	PPCODE:	1424	PPCODE:
		1425	{
		1426	JSON json = { F_DEFAULT | F_UTF8 };
973	XPUSHs (decode_json (jsonstr, F_UTF8));	1427	XPUSHs (decode_json (jsonstr, &json, 0));
		1428	}
974		1429

Diff Legend

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