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 1.46 by root, Mon Jun 25 22:11:39 2007 UTC

…		…
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include "assert.h"
6	#include "string.h"	6	#include "string.h"
7	#include "stdlib.h"	7	#include "stdlib.h"
		8	#include "stdio.h"
8		9
		10	#if defined(__BORLANDC__) || defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	27	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35
		36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		38
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		41
22	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		46
24	#define SB do {	47	#define SB do {
25	#define SE } while (0)	48	#define SE } while (0)
26		49
27	static HV *json_stash;	50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr),(value))
		52	# define inline inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define inline static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		62	static SV *json_true, *json_false;
		63
		64	/////////////////////////////////////////////////////////////////////////////
		65	// utility functions
		66
		67	static UV *
		68	SvJSON (SV *sv)
		69	{
		70	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		71	croak ("object is not of type JSON::XS");
		72
		73	return &SvUVX (SvRV (sv));
		74	}
		75
		76	static void
		77	shrink (SV *sv)
		78	{
		79	sv_utf8_downgrade (sv, 1);
		80	if (SvLEN (sv) > SvCUR (sv) + 1)
		81	{
		82	#ifdef SvPV_shrink_to_cur
		83	SvPV_shrink_to_cur (sv);
		84	#elif defined (SvPV_renew)
		85	SvPV_renew (sv, SvCUR (sv) + 1);
		86	#endif
		87	}
		88	}
		89
		90	// decode an utf-8 character and return it, or (UV)-1 in
		91	// case of an error.
		92	// we special-case "safe" characters from U+80 .. U+7FF,
		93	// but use the very good perl function to parse anything else.
		94	// note that we never call this function for a ascii codepoints
		95	inline UV
		96	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		97	{
		98	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		99	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		100	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		101	{
		102	*clen = 2;
		103	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		104	}
		105	else
		106	{
		107	*clen = (STRLEN)-1;
		108	return (UV)-1;
		109	}
		110	}
		111
		112	/////////////////////////////////////////////////////////////////////////////
		113	// encoder
28		114
29	// structure used for encoding JSON	115	// structure used for encoding JSON
30	typedef struct	116	typedef struct
31	{	117	{
32	char *cur;	118	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	119	char *end; // SvEND (sv)
35	SV *sv;	120	SV *sv; // result scalar
36	UV flags;	121	U32 flags; // F_*
37	int max_recurse;	122	U32 indent; // indentation level
38	int indent;	123	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	124	} enc_t;
40		125
41	// structure used for decoding JSON	126	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	127	need (enc_t *enc, STRLEN len)
72	{	128	{
73	if (enc->cur + len >= enc->end)	129	if (expect_false (enc->cur + len >= enc->end))
74	{	130	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	131	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	132	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	133	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	134	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	135	}
80	}	136	}
81		137
82	static void	138	inline void
83	encode_ch (enc_t *enc, char ch)	139	encode_ch (enc_t *enc, char ch)
84	{	140	{
85	need (enc, 1);	141	need (enc, 1);
86	*enc->cur++ = ch;	142	*enc->cur++ = ch;
87	}	143	}
…		…
95		151
96	while (str < end)	152	while (str < end)
97	{	153	{
98	unsigned char ch = *(unsigned char *)str;	154	unsigned char ch = *(unsigned char *)str;
99		155
100	if (ch >= 0x20 && ch < 0x80) // most common case	156	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	157	{
102	if (ch == '"') // but with slow exceptions	158	if (expect_false (ch == '"')) // but with slow exceptions
103	{	159	{
104	need (enc, len += 1);	160	need (enc, len += 1);
105	*enc->cur++ = '\\';	161	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	162	*enc->cur++ = '"';
107	}	163	}
108	else if (ch == '\\')	164	else if (expect_false (ch == '\\'))
109	{	165	{
110	need (enc, len += 1);	166	need (enc, len += 1);
111	*enc->cur++ = '\\';	167	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
113	}	169	}
…		…
131	STRLEN clen;	187	STRLEN clen;
132	UV uch;	188	UV uch;
133		189
134	if (is_utf8)	190	if (is_utf8)
135	{	191	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	192	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	193	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	194	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	195	}
140	else	196	else
141	{	197	{
142	uch = ch;	198	uch = ch;
143	clen = 1;	199	clen = 1;
144	}	200	}
145		201
146	if (uch < 0x80 || enc->flags & F_ASCII)	202	if (uch > 0x10FFFFUL)
		203	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		204
		205	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
147	{	206	{
148	if (uch > 0xFFFFUL)	207	if (uch > 0xFFFFUL)
149	{	208	{
150	need (enc, len += 11);	209	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	210	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	211	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	212	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	213	enc->cur += 12;
155	}	214	}
156	else	215	else
157	{	216	{
158	static char hexdigit [16] = "0123456789abcdef";	217	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	224	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	225	}
167		226
168	str += clen;	227	str += clen;
169	}	228	}
		229	else if (enc->flags & F_LATIN1)
		230	{
		231	*enc->cur++ = uch;
		232	str += clen;
		233	}
170	else if (is_utf8)	234	else if (is_utf8)
171	{	235	{
172	need (enc, len += clen);	236	need (enc, len += clen);
173	do	237	do
174	{	238	{
…		…
176	}	240	}
177	while (--clen);	241	while (--clen);
178	}	242	}
179	else	243	else
180	{	244	{
181	need (enc, len += 10); // never more than 11 bytes needed	245	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	246	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	247	++str;
184	}	248	}
185	}	249	}
186	}	250	}
…		…
188		252
189	--len;	253	--len;
190	}	254	}
191	}	255	}
192		256
193	#define INDENT SB \	257	inline void
		258	encode_indent (enc_t *enc)
		259	{
194	if (enc->flags & F_INDENT) \	260	if (enc->flags & F_INDENT)
195	{ \	261	{
196	int i_; \	262	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \
198	for (i_ = enc->indent * 3; i_--; )\
199	encode_ch (enc, ' '); \
200	} \
201	SE
202		263
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	264	need (enc, spaces);
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	265	memset (enc->cur, ' ', spaces);
205	#define COMMA SB \	266	enc->cur += spaces;
		267	}
		268	}
		269
		270	inline void
		271	encode_space (enc_t *enc)
		272	{
		273	need (enc, 1);
206	encode_ch (enc, ','); \	274	encode_ch (enc, ' ');
		275	}
		276
		277	inline void
		278	encode_nl (enc_t *enc)
		279	{
207	if (enc->flags & F_INDENT) \	280	if (enc->flags & F_INDENT)
208	NL; \	281	{
		282	need (enc, 1);
		283	encode_ch (enc, '\n');
		284	}
		285	}
		286
		287	inline void
		288	encode_comma (enc_t *enc)
		289	{
		290	encode_ch (enc, ',');
		291
		292	if (enc->flags & F_INDENT)
		293	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	294	else if (enc->flags & F_SPACE_AFTER)
210	SPACE; \	295	encode_space (enc);
211	SE	296	}
212		297
213	static void encode_sv (enc_t *enc, SV *sv);	298	static void encode_sv (enc_t *enc, SV *sv);
214		299
215	static void	300	static void
216	encode_av (enc_t *enc, AV *av)	301	encode_av (enc_t *enc, AV *av)
217	{	302	{
218	int i, len = av_len (av);	303	int i, len = av_len (av);
219		304
220	encode_ch (enc, '['); NL;	305	if (enc->indent >= enc->maxdepth)
		306	croak ("data structure too deep (hit recursion limit)");
		307
		308	encode_ch (enc, '['); encode_nl (enc);
221	++enc->indent;	309	++enc->indent;
222		310
223	for (i = 0; i <= len; ++i)	311	for (i = 0; i <= len; ++i)
224	{	312	{
225	INDENT;	313	encode_indent (enc);
226	encode_sv (enc, *av_fetch (av, i, 0));	314	encode_sv (enc, *av_fetch (av, i, 0));
227		315
228	if (i < len)	316	if (i < len)
229	COMMA;	317	encode_comma (enc);
230	}	318	}
231		319
232	NL;	320	encode_nl (enc);
233		321
234	--enc->indent;	322	--enc->indent;
235	INDENT; encode_ch (enc, ']');	323	encode_indent (enc); encode_ch (enc, ']');
236	}	324	}
237		325
238	static void	326	static void
239	encode_he (enc_t *enc, HE *he)	327	encode_he (enc_t *enc, HE *he)
240	{	328	{
…		…
254	else	342	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	343	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		344
257	encode_ch (enc, '"');	345	encode_ch (enc, '"');
258		346
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	347	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	348	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	349	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));	350	encode_sv (enc, HeVAL (he));
263	}	351	}
264		352
265	// compare hash entries, used when all keys are bytestrings	353	// compare hash entries, used when all keys are bytestrings
266	static int	354	static int
…		…
272	HE *b = *(HE **)b_;	360	HE *b = *(HE **)b_;
273		361
274	STRLEN la = HeKLEN (a);	362	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	363	STRLEN lb = HeKLEN (b);
276		364
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	365	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	366	cmp = la - lb;
279		367
280	return cmp;	368	return cmp;
281	}	369	}
282		370
283	// compare hash entries, used when some keys are sv's or utf-x	371	// compare hash entries, used when some keys are sv's or utf-x
…		…
290	static void	378	static void
291	encode_hv (enc_t *enc, HV *hv)	379	encode_hv (enc_t *enc, HV *hv)
292	{	380	{
293	int count, i;	381	int count, i;
294		382
		383	if (enc->indent >= enc->maxdepth)
		384	croak ("data structure too deep (hit recursion limit)");
		385
295	encode_ch (enc, '{'); NL; ++enc->indent;	386	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
296		387
297	if ((count = hv_iterinit (hv)))	388	if ((count = hv_iterinit (hv)))
298	{	389	{
299	// for canonical output we have to sort by keys first	390	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	391	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	392	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	393	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	394	if (enc->flags & F_CANONICAL)
304	{	395	{
305	HE *he, *hes [count];
306	int fast = 1;	396	int fast = 1;
		397	HE *he;
		398	#if defined(__BORLANDC__) || defined(_MSC_VER)
		399	HE **hes = _alloca (count * sizeof (HE));
		400	#else
		401	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		402	#endif
307		403
308	i = 0;	404	i = 0;
309	while ((he = hv_iternext (hv)))	405	while ((he = hv_iternext (hv)))
310	{	406	{
311	hes [i++] = he;	407	hes [i++] = he;
…		…
317		413
318	if (fast)	414	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	415	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	416	else
321	{	417	{
322	// hack to disable "use bytes"	418	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	419	COP cop = *PL_curcop;
324	cop.op_private = 0;	420	cop.op_private = 0;
		421
		422	ENTER;
		423	SAVETMPS;
		424
		425	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	426	PL_curcop = &cop;
326		427
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	428	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		429
329	FREETMPS;	430	FREETMPS;
330		431	LEAVE;
331	PL_curcop = oldcop;
332	}	432	}
333		433
334	for (i = 0; i < count; ++i)	434	for (i = 0; i < count; ++i)
335	{	435	{
336	INDENT;	436	encode_indent (enc);
337	encode_he (enc, hes [i]);	437	encode_he (enc, hes [i]);
338		438
339	if (i < count - 1)	439	if (i < count - 1)
340	COMMA;	440	encode_comma (enc);
341	}
342
343	NL;	441	}
		442
		443	encode_nl (enc);
344	}	444	}
345	else	445	else
346	{	446	{
347	SV *sv;
348	HE *he = hv_iternext (hv);	447	HE *he = hv_iternext (hv);
349		448
350	for (;;)	449	for (;;)
351	{	450	{
352	INDENT;	451	encode_indent (enc);
353	encode_he (enc, he);	452	encode_he (enc, he);
354		453
355	if (!(he = hv_iternext (hv)))	454	if (!(he = hv_iternext (hv)))
356	break;	455	break;
357		456
358	COMMA;	457	encode_comma (enc);
359	}
360
361	NL;	458	}
		459
		460	encode_nl (enc);
362	}	461	}
363	}	462	}
364		463
365	--enc->indent; INDENT; encode_ch (enc, '}');	464	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		465	}
		466
		467	// encode objects, arrays and special \0=false and \1=true values.
		468	static void
		469	encode_rv (enc_t *enc, SV *sv)
		470	{
		471	svtype svt;
		472
		473	SvGETMAGIC (sv);
		474	svt = SvTYPE (sv);
		475
		476	if (expect_false (SvOBJECT (sv)))
		477	{
		478	if (SvSTASH (sv) == json_boolean_stash)
		479	{
		480	if (SvIV (sv) == 0)
		481	encode_str (enc, "false", 5, 0);
		482	else
		483	encode_str (enc, "true", 4, 0);
		484	}
		485	else
		486	{
		487	#if 0
		488	if (0 && sv_derived_from (rv, "JSON::Literal"))
		489	{
		490	// not yet
		491	}
		492	#endif
		493	if (enc->flags & F_CONV_BLESSED)
		494	{
		495	// we re-bless the reference to get overload and other niceties right
		496	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		497
		498	if (to_json)
		499	{
		500	dSP;
		501	ENTER;
		502	SAVETMPS;
		503	PUSHMARK (SP);
		504	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		505
		506	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		507	// check anyways.
		508	PUTBACK;
		509	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		510	SPAGAIN;
		511
		512	encode_sv (enc, POPs);
		513
		514	FREETMPS;
		515	LEAVE;
		516	}
		517	else if (enc->flags & F_ALLOW_BLESSED)
		518	encode_str (enc, "null", 4, 0);
		519	else
		520	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		521	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		522	}
		523	else if (enc->flags & F_ALLOW_BLESSED)
		524	encode_str (enc, "null", 4, 0);
		525	else
		526	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		527	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		528	}
		529	}
		530	else if (svt == SVt_PVHV)
		531	encode_hv (enc, (HV *)sv);
		532	else if (svt == SVt_PVAV)
		533	encode_av (enc, (AV *)sv);
		534	else if (svt < SVt_PVAV)
		535	{
		536	if (SvNIOK (sv) && SvIV (sv) == 0)
		537	encode_str (enc, "false", 5, 0);
		538	else if (SvNIOK (sv) && SvIV (sv) == 1)
		539	encode_str (enc, "true", 4, 0);
		540	else
		541	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else
		545	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	547	}
367		548
368	static void	549	static void
369	encode_sv (enc_t *enc, SV *sv)	550	encode_sv (enc_t *enc, SV *sv)
370	{	551	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	559	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	560	encode_ch (enc, '"');
380	}	561	}
381	else if (SvNOKp (sv))	562	else if (SvNOKp (sv))
382	{	563	{
		564	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	565	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	566	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	567	enc->cur += strlen (enc->cur);
386	}	568	}
387	else if (SvIOKp (sv))	569	else if (SvIOKp (sv))
388	{	570	{
389	need (enc, 64);	571	// we assume we can always read an IV as a UV
		572	if (SvUV (sv) & ~(UV)0x7fff)
		573	{
		574	// large integer, use the (rather slow) snprintf way.
		575	need (enc, sizeof (UV) * 3);
390	enc->cur +=	576	enc->cur +=
391	SvIsUV(sv)	577	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	578	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	579	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		580	}
		581	else
		582	{
		583	// optimise the "small number case"
		584	// code will likely be branchless and use only a single multiplication
		585	I32 i = SvIV (sv);
		586	U32 u;
		587	char digit, nz = 0;
		588
		589	need (enc, 6);
		590
		591	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		592	u = i < 0 ? -i : i;
		593
		594	// convert to 4.28 fixed-point representation
		595	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		596
		597	// now output digit by digit, each time masking out the integer part
		598	// and multiplying by 5 while moving the decimal point one to the right,
		599	// resulting in a net multiplication by 10.
		600	// we always write the digit to memory but conditionally increment
		601	// the pointer, to ease the usage of conditional move instructions.
		602	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		603	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		604	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		605	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		606	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		607	}
394	}	608	}
395	else if (SvROK (sv))	609	else if (SvROK (sv))
396	{	610	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))	611	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	612	encode_str (enc, "null", 4, 0);
413	else	613	else
414	croak ("encountered perl type that JSON cannot handle");	614	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		615	SvPV_nolen (sv), SvFLAGS (sv));
415	}	616	}
416		617
417	static SV *	618	static SV *
418	encode_json (SV *scalar, UV flags)	619	encode_json (SV *scalar, U32 flags)
419	{	620	{
		621	enc_t enc;
		622
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	623	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");	624	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
422		625
423	enc_t enc;
424	enc.flags = flags;	626	enc.flags = flags;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	627	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	628	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	629	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	630	enc.indent = 0;
		631	enc.maxdepth = DEC_DEPTH (flags);
430		632
431	SvPOK_only (enc.sv);	633	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	634	encode_sv (&enc, scalar);
433		635
		636	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		637	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		638
434	if (!(flags & (F_ASCII | F_UTF8)))	639	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	640	SvUTF8_on (enc.sv);
436
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438		641
439	if (enc.flags & F_SHRINK)	642	if (enc.flags & F_SHRINK)
440	shrink (enc.sv);	643	shrink (enc.sv);
441		644
442	return enc.sv;	645	return enc.sv;
443	}	646	}
444		647
445	/////////////////////////////////////////////////////////////////////////////	648	/////////////////////////////////////////////////////////////////////////////
		649	// decoder
446		650
447	#define WS \	651	// structure used for decoding JSON
		652	typedef struct
		653	{
		654	char *cur; // current parser pointer
		655	char *end; // end of input string
		656	const char *err; // parse error, if != 0
		657	U32 flags; // F_*
		658	U32 depth; // recursion depth
		659	U32 maxdepth; // recursion depth limit
		660	} dec_t;
		661
		662	inline void
		663	decode_ws (dec_t *dec)
		664	{
448	for (;;) \	665	for (;;)
449	{ \	666	{
450	char ch = *dec->cur; \	667	char ch = *dec->cur;
		668
451	if (ch > 0x20 \	669	if (ch > 0x20
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	670	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
453	break; \	671	break;
		672
454	++dec->cur; \	673	++dec->cur;
455	}	674	}
		675	}
456		676
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	677	#define ERR(reason) SB dec->err = reason; goto fail; SE
		678
458	#define EXPECT_CH(ch) SB \	679	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	680	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	681	ERR (# ch " expected"); \
461	++dec->cur; \	682	++dec->cur; \
462	SE	683	SE
463		684
		685	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		686	#define DEC_DEC_DEPTH --dec->depth
		687
464	static SV *decode_sv (dec_t *dec);	688	static SV *decode_sv (dec_t *dec);
465		689
466	static signed char decode_hexdigit[256];	690	static signed char decode_hexdigit[256];
467		691
468	static UV	692	static UV
469	decode_4hex (dec_t *dec)	693	decode_4hex (dec_t *dec)
470	{	694	{
471	signed char d1, d2, d3, d4;	695	signed char d1, d2, d3, d4;
		696	unsigned char *cur = (unsigned char *)dec->cur;
472		697
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	698	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	699	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	700	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	701	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
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		702
482	dec->cur += 4;	703	dec->cur += 4;
483		704
484	return ((UV)d1) << 12	705	return ((UV)d1) << 12
485	| ((UV)d2) << 8	706	| ((UV)d2) << 8
…		…
488		709
489	fail:	710	fail:
490	return (UV)-1;	711	return (UV)-1;
491	}	712	}
492		713
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 *	714	static SV *
509	decode_str (dec_t *dec)	715	decode_str (dec_t *dec)
510	{	716	{
511	SV *sv = NEWSV (0,2);	717	SV *sv = 0;
512	int utf8 = 0;	718	int utf8 = 0;
513	char *cur = SvPVX (sv);	719	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		720
516	for (;;)	721	do
517	{	722	{
518	unsigned char ch = *(unsigned char *)dec->cur;	723	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		724	char *cur = buf;
519		725
520	if (ch == '"')	726	do
521	break;
522	else if (ch == '\\')
523	{	727	{
524	switch (*++dec->cur)	728	unsigned char ch = *(unsigned char *)dec_cur++;
		729
		730	if (expect_false (ch == '"'))
		731	{
		732	--dec_cur;
		733	break;
525	{	734	}
526	case '\\':	735	else if (expect_false (ch == '\\'))
527	case '/':	736	{
528	case '"': APPEND_CH (*dec->cur++); break;	737	switch (*dec_cur)
529
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;
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	{	738	{
538	UV lo, hi;	739	case '\\':
539	++dec->cur;	740	case '/':
		741	case '"': *cur++ = *dec_cur++; break;
540		742
541	hi = decode_4hex (dec);	743	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	744	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	745	case 'n': ++dec_cur; *cur++ = '\012'; break;
		746	case 'f': ++dec_cur; *cur++ = '\014'; break;
		747	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		748
545	// possibly a surrogate pair	749	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	750	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	751	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	752	++dec_cur;
550		753
551	dec->cur += 2;	754	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	755	hi = decode_4hex (dec);
		756	dec_cur = dec->cur;
554	if (lo == (UV)-1)	757	if (hi == (UV)-1)
555	goto fail;	758	goto fail;
556		759
		760	// possibly a surrogate pair
		761	if (hi >= 0xd800)
		762	if (hi < 0xdc00)
		763	{
		764	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		765	ERR ("missing low surrogate character in surrogate pair");
		766
		767	dec_cur += 2;
		768
		769	dec->cur = dec_cur;
		770	lo = decode_4hex (dec);
		771	dec_cur = dec->cur;
		772	if (lo == (UV)-1)
		773	goto fail;
		774
557	if (lo < 0xdc00 || lo >= 0xe000)	775	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	776	ERR ("surrogate pair expected");
559		777
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	778	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		779	}
		780	else if (hi < 0xe000)
		781	ERR ("missing high surrogate character in surrogate pair");
		782
		783	if (hi >= 0x80)
		784	{
		785	utf8 = 1;
		786
		787	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		788	}
		789	else
		790	*cur++ = hi;
561	}	791	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	792	break;
567	utf8 = 1;
568		793
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	794	default:
573	APPEND_CH (hi);	795	--dec_cur;
		796	ERR ("illegal backslash escape sequence in string");
574	}	797	}
575	break;	798	}
		799	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		800	*cur++ = ch;
		801	else if (ch >= 0x80)
		802	{
		803	STRLEN clen;
		804	UV uch;
576		805
577	default:
578	--dec->cur;	806	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	807
		808	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		809	if (clen == (STRLEN)-1)
		810	ERR ("malformed UTF-8 character in JSON string");
		811
		812	do
		813	*cur++ = *dec_cur++;
		814	while (--clen);
		815
		816	utf8 = 1;
		817	}
		818	else
		819	{
		820	--dec_cur;
		821
		822	if (!ch)
		823	ERR ("unexpected end of string while parsing JSON string");
		824	else
		825	ERR ("invalid character encountered while parsing JSON string");
580	}	826	}
581	}	827	}
582	else if (ch >= 0x20 && ch <= 0x7f)	828	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	829
584	else if (ch >= 0x80)
585	{	830	{
586	STRLEN clen;	831	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		832
591	APPEND_GROW (clen);	833	if (sv)
592	do
593	{	834	{
594	*cur++ = *dec->cur++;	835	SvGROW (sv, SvCUR (sv) + len + 1);
		836	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		837	SvCUR_set (sv, SvCUR (sv) + len);
595	}	838	}
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	839	else
603	ERR ("invalid character encountered");	840	sv = newSVpvn (buf, len);
604	}	841	}
		842	}
		843	while (*dec_cur != '"');
605		844
606	++dec->cur;	845	++dec_cur;
607		846
608	SvCUR_set (sv, cur - SvPVX (sv));	847	if (sv)
609		848	{
610	SvPOK_only (sv);	849	SvPOK_only (sv);
611	*SvEND (sv) = 0;	850	*SvEND (sv) = 0;
612		851
613	if (utf8)	852	if (utf8)
614	SvUTF8_on (sv);	853	SvUTF8_on (sv);
		854	}
		855	else
		856	sv = newSVpvn ("", 0);
615		857
616	if (dec->flags & F_SHRINK)	858	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	859	return sv;
620		860
621	fail:	861	fail:
622	SvREFCNT_dec (sv);	862	dec->cur = dec_cur;
623	return 0;	863	return 0;
624	}	864	}
625		865
626	static SV *	866	static SV *
627	decode_num (dec_t *dec)	867	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	925	is_nv = 1;
686	}	926	}
687		927
688	if (!is_nv)	928	if (!is_nv)
689	{	929	{
690	UV uv;	930	// 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);	931	if (*start == '-')
692	if (numtype & IS_NUMBER_IN_UV)	932	switch (dec->cur - start)
693	if (numtype & IS_NUMBER_NEG)
694	{	933	{
695	if (uv < (UV)IV_MIN)	934	case 2: return newSViv (-( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	935	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		936	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		937	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
697	}	938	}
		939	else
		940	switch (dec->cur - start)
		941	{
		942	case 1: return newSViv ( start [0] - '0' * 1);
		943	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		944	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		945	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		946	}
		947
		948	{
		949	UV uv;
		950	int numtype = grok_number (start, dec->cur - start, &uv);
		951	if (numtype & IS_NUMBER_IN_UV)
		952	if (numtype & IS_NUMBER_NEG)
		953	{
		954	if (uv < (UV)IV_MIN)
		955	return newSViv (-(IV)uv);
		956	}
698	else	957	else
699	return newSVuv (uv);	958	return newSVuv (uv);
		959
		960	// here would likely be the place for bigint support
700	}	961	}
		962	}
701		963
		964	// if we ever support bigint or bigfloat, this is the place for bigfloat
702	return newSVnv (Atof (start));	965	return newSVnv (Atof (start));
703		966
704	fail:	967	fail:
705	return 0;	968	return 0;
706	}	969	}
…		…
708	static SV *	971	static SV *
709	decode_av (dec_t *dec)	972	decode_av (dec_t *dec)
710	{	973	{
711	AV *av = newAV ();	974	AV *av = newAV ();
712		975
713	WS;	976	DEC_INC_DEPTH;
		977	decode_ws (dec);
		978
714	if (*dec->cur == ']')	979	if (*dec->cur == ']')
715	++dec->cur;	980	++dec->cur;
716	else	981	else
717	for (;;)	982	for (;;)
718	{	983	{
…		…
722	if (!value)	987	if (!value)
723	goto fail;	988	goto fail;
724		989
725	av_push (av, value);	990	av_push (av, value);
726		991
727	WS;	992	decode_ws (dec);
728		993
729	if (*dec->cur == ']')	994	if (*dec->cur == ']')
730	{	995	{
731	++dec->cur;	996	++dec->cur;
732	break;	997	break;
…		…
736	ERR (", or ] expected while parsing array");	1001	ERR (", or ] expected while parsing array");
737		1002
738	++dec->cur;	1003	++dec->cur;
739	}	1004	}
740		1005
		1006	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1007	return newRV_noinc ((SV *)av);
742		1008
743	fail:	1009	fail:
744	SvREFCNT_dec (av);	1010	SvREFCNT_dec (av);
		1011	DEC_DEC_DEPTH;
745	return 0;	1012	return 0;
746	}	1013	}
747		1014
748	static SV *	1015	static SV *
749	decode_hv (dec_t *dec)	1016	decode_hv (dec_t *dec)
750	{	1017	{
751	HV *hv = newHV ();	1018	HV *hv = newHV ();
752		1019
753	WS;	1020	DEC_INC_DEPTH;
		1021	decode_ws (dec);
		1022
754	if (*dec->cur == '}')	1023	if (*dec->cur == '}')
755	++dec->cur;	1024	++dec->cur;
756	else	1025	else
757	for (;;)	1026	for (;;)
758	{	1027	{
		1028	decode_ws (dec); EXPECT_CH ('"');
		1029
		1030	// heuristic: assume that
		1031	// a) decode_str + hv_store_ent are abysmally slow
		1032	// b) most hash keys are short, simple ascii text
		1033	// so try to "fast-match" such strings to avoid
		1034	// the overhead of hv_store_ent.
		1035	{
759	SV *key, *value;	1036	SV *value;
		1037	char *p = dec->cur;
		1038	char *e = p + 24; // only try up to 24 bytes
760		1039
761	WS; EXPECT_CH ('"');	1040	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	{	1041	{
		1042	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1043	{
		1044	// slow path, back up and use decode_str
		1045	SV *key = decode_str (dec);
		1046	if (!key)
		1047	goto fail;
		1048
		1049	decode_ws (dec); EXPECT_CH (':');
		1050
		1051	value = decode_sv (dec);
		1052	if (!value)
		1053	{
		1054	SvREFCNT_dec (key);
		1055	goto fail;
		1056	}
		1057
		1058	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1059	SvREFCNT_dec (key);
		1060
		1061	break;
		1062	}
		1063	else if (*p == '"')
		1064	{
		1065	// fast path, got a simple key
		1066	char *key = dec->cur;
		1067	int len = p - key;
		1068	dec->cur = p + 1;
		1069
		1070	decode_ws (dec); EXPECT_CH (':');
		1071
		1072	value = decode_sv (dec);
		1073	if (!value)
773	goto fail;	1074	goto fail;
		1075
		1076	hv_store (hv, key, len, value, 0);
		1077
		1078	break;
		1079	}
		1080
		1081	++p;
774	}	1082	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1083	}
		1084
		1085	decode_ws (dec);
780		1086
781	if (*dec->cur == '}')	1087	if (*dec->cur == '}')
782	{	1088	{
783	++dec->cur;	1089	++dec->cur;
784	break;	1090	break;
…		…
788	ERR (", or } expected while parsing object/hash");	1094	ERR (", or } expected while parsing object/hash");
789		1095
790	++dec->cur;	1096	++dec->cur;
791	}	1097	}
792		1098
		1099	DEC_DEC_DEPTH;
793	return newRV_noinc ((SV *)hv);	1100	return newRV_noinc ((SV *)hv);
794		1101
795	fail:	1102	fail:
796	SvREFCNT_dec (hv);	1103	SvREFCNT_dec (hv);
		1104	DEC_DEC_DEPTH;
797	return 0;	1105	return 0;
798	}	1106	}
799		1107
800	static SV *	1108	static SV *
801	decode_sv (dec_t *dec)	1109	decode_sv (dec_t *dec)
802	{	1110	{
803	WS;	1111	decode_ws (dec);
		1112
		1113	// the beauty of JSON: you need exactly one character lookahead
		1114	// to parse anything.
804	switch (*dec->cur)	1115	switch (*dec->cur)
805	{	1116	{
806	case '"': ++dec->cur; return decode_str (dec);	1117	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1118	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1119	case '{': ++dec->cur; return decode_hv (dec);
…		…
814		1125
815	case 't':	1126	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1127	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1128	{
818	dec->cur += 4;	1129	dec->cur += 4;
819	return newSViv (1);	1130	return SvREFCNT_inc (json_true);
820	}	1131	}
821	else	1132	else
822	ERR ("'true' expected");	1133	ERR ("'true' expected");
823		1134
824	break;	1135	break;
825		1136
826	case 'f':	1137	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1138	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1139	{
829	dec->cur += 5;	1140	dec->cur += 5;
830	return newSViv (0);	1141	return SvREFCNT_inc (json_false);
831	}	1142	}
832	else	1143	else
833	ERR ("'false' expected");	1144	ERR ("'false' expected");
834		1145
835	break;	1146	break;
…		…
844	ERR ("'null' expected");	1155	ERR ("'null' expected");
845		1156
846	break;	1157	break;
847		1158
848	default:	1159	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1160	ERR ("malformed JSON string, neither array, object, number, string or atom");
850	break;	1161	break;
851	}	1162	}
852		1163
853	fail:	1164	fail:
854	return 0;	1165	return 0;
855	}	1166	}
856		1167
857	static SV *	1168	static SV *
858	decode_json (SV *string, UV flags)	1169	decode_json (SV *string, U32 flags, UV *offset_return)
859	{	1170	{
		1171	dec_t dec;
		1172	UV offset;
860	SV *sv;	1173	SV *sv;
		1174
		1175	SvGETMAGIC (string);
		1176	SvUPGRADE (string, SVt_PV);
		1177
		1178	if (flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (flags))
		1179	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1180	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (flags));
861		1181
862	if (flags & F_UTF8)	1182	if (flags & F_UTF8)
863	sv_utf8_downgrade (string, 0);	1183	sv_utf8_downgrade (string, 0);
864	else	1184	else
865	sv_utf8_upgrade (string);	1185	sv_utf8_upgrade (string);
866		1186
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1187	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1188
869	dec_t dec;
870	dec.flags = flags;	1189	dec.flags = flags;
871	dec.cur = SvPVX (string);	1190	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1191	dec.end = SvEND (string);
873	dec.err = 0;	1192	dec.err = 0;
		1193	dec.depth = 0;
		1194	dec.maxdepth = DEC_DEPTH (dec.flags);
874		1195
		1196	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
875	sv = decode_sv (&dec);	1197	sv = decode_sv (&dec);
876		1198
		1199	if (!(offset_return || !sv))
		1200	{
		1201	// check for trailing garbage
		1202	decode_ws (&dec);
		1203
		1204	if (*dec.cur)
		1205	{
		1206	dec.err = "garbage after JSON object";
		1207	SvREFCNT_dec (sv);
		1208	sv = 0;
		1209	}
		1210	}
		1211
		1212	if (offset_return || !sv)
		1213	{
		1214	offset = dec.flags & F_UTF8
		1215	? dec.cur - SvPVX (string)
		1216	: utf8_distance (dec.cur, SvPVX (string));
		1217
		1218	if (offset_return)
		1219	*offset_return = offset;
		1220	}
		1221
877	if (!sv)	1222	if (!sv)
878	{	1223	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1224	SV *uni = sv_newmortal ();
		1225
883	// horrible hack to silence warning inside pv_uni_display	1226	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1227	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1228	cop.cop_warnings = pWARN_NONE;
		1229	ENTER;
887	SAVEVPTR (PL_curcop);	1230	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1231	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1232	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1233	LEAVE;
		1234
891	croak ("%s, at character offset %d (%s)",	1235	croak ("%s, at character offset %d [\"%s\"]",
892	dec.err,	1236	dec.err,
893	(int)offset,	1237	(int)offset,
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1238	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1239	}
896		1240
897	sv = sv_2mortal (sv);	1241	sv = sv_2mortal (sv);
898		1242
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1243	if (!(dec.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)");	1244	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1245
902	return sv;	1246	return sv;
903	}	1247	}
904		1248
		1249	/////////////////////////////////////////////////////////////////////////////
		1250	// XS interface functions
		1251
905	MODULE = JSON::XS PACKAGE = JSON::XS	1252	MODULE = JSON::XS PACKAGE = JSON::XS
906		1253
907	BOOT:	1254	BOOT:
908	{	1255	{
909	int i;	1256	int i;
910		1257
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1258	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1259	decode_hexdigit [i] =
		1260	i >= '0' && i <= '9' ? i - '0'
		1261	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1262	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1263	: -1;
914		1264
915	for (i = 7; i--; )
916	{
917	decode_hexdigit ['a' + i] = 10 + i;
918	decode_hexdigit ['A' + i] = 10 + i;
919	}
920
921	json_stash = gv_stashpv ("JSON::XS", 1);	1265	json_stash = gv_stashpv ("JSON::XS" , 1);
		1266	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1267
		1268	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1269	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
922	}	1270	}
923		1271
924	PROTOTYPES: DISABLE	1272	PROTOTYPES: DISABLE
925		1273
926	SV *new (char *dummy)	1274	SV *new (char *dummy)
…		…
929	OUTPUT:	1277	OUTPUT:
930	RETVAL	1278	RETVAL
931		1279
932	SV *ascii (SV *self, int enable = 1)	1280	SV *ascii (SV *self, int enable = 1)
933	ALIAS:	1281	ALIAS:
934	ascii = F_ASCII	1282	ascii = F_ASCII
		1283	latin1 = F_LATIN1
935	utf8 = F_UTF8	1284	utf8 = F_UTF8
936	indent = F_INDENT	1285	indent = F_INDENT
937	canonical = F_CANONICAL	1286	canonical = F_CANONICAL
938	space_before = F_SPACE_BEFORE	1287	space_before = F_SPACE_BEFORE
939	space_after = F_SPACE_AFTER	1288	space_after = F_SPACE_AFTER
940	json_rpc = F_JSON_RPC
941	pretty = F_PRETTY	1289	pretty = F_PRETTY
942	allow_nonref = F_ALLOW_NONREF	1290	allow_nonref = F_ALLOW_NONREF
943	shrink = F_SHRINK	1291	shrink = F_SHRINK
		1292	allow_blessed = F_ALLOW_BLESSED
		1293	convert_blessed = F_CONV_BLESSED
944	CODE:	1294	CODE:
945	{	1295	{
946	UV *uv = SvJSON (self);	1296	UV *uv = SvJSON (self);
947	if (enable)	1297	if (enable)
948	*uv |= ix;	1298	*uv |= ix;
…		…
952	RETVAL = newSVsv (self);	1302	RETVAL = newSVsv (self);
953	}	1303	}
954	OUTPUT:	1304	OUTPUT:
955	RETVAL	1305	RETVAL
956		1306
		1307	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1308	CODE:
		1309	{
		1310	UV *uv = SvJSON (self);
		1311	UV log2 = 0;
		1312
		1313	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1314
		1315	while ((1UL << log2) < max_depth)
		1316	++log2;
		1317
		1318	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1319
		1320	RETVAL = newSVsv (self);
		1321	}
		1322	OUTPUT:
		1323	RETVAL
		1324
		1325	SV *max_size (SV *self, UV max_size = 0)
		1326	CODE:
		1327	{
		1328	UV *uv = SvJSON (self);
		1329	UV log2 = 0;
		1330
		1331	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1332	if (max_size == 1) max_size = 2;
		1333
		1334	while ((1UL << log2) < max_size)
		1335	++log2;
		1336
		1337	*uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1338
		1339	RETVAL = newSVsv (self);
		1340	}
		1341	OUTPUT:
		1342	RETVAL
		1343
957	void encode (SV *self, SV *scalar)	1344	void encode (SV *self, SV *scalar)
958	PPCODE:	1345	PPCODE:
959	XPUSHs (encode_json (scalar, *SvJSON (self)));	1346	XPUSHs (encode_json (scalar, *SvJSON (self)));
960		1347
961	void decode (SV *self, SV *jsonstr)	1348	void decode (SV *self, SV *jsonstr)
962	PPCODE:	1349	PPCODE:
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1350	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1351
		1352	void decode_prefix (SV *self, SV *jsonstr)
		1353	PPCODE:
		1354	{
		1355	UV offset;
		1356	EXTEND (SP, 2);
		1357	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1358	PUSHs (sv_2mortal (newSVuv (offset)));
		1359	}
964		1360
965	PROTOTYPES: ENABLE	1361	PROTOTYPES: ENABLE
966		1362
967	void to_json (SV *scalar)	1363	void to_json (SV *scalar)
		1364	ALIAS:
		1365	objToJson = 0
968	PPCODE:	1366	PPCODE:
969	XPUSHs (encode_json (scalar, F_UTF8));	1367	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
970		1368
971	void from_json (SV *jsonstr)	1369	void from_json (SV *jsonstr)
		1370	ALIAS:
		1371	jsonToObj = 0
972	PPCODE:	1372	PPCODE:
973	XPUSHs (decode_json (jsonstr, F_UTF8));	1373	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
974		1374

Diff Legend

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