ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.6 by root, Fri Mar 23 15:10:55 2007 UTC vs.
Revision 1.79 by root, Tue Mar 25 23:00:31 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
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>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
		37	#define F_MAXDEPTH 0xf8000000UL
		38	#define S_MAXDEPTH 27
		39	#define F_MAXSIZE 0x01f00000UL
		40	#define S_MAXSIZE 20
		41	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		42
		43	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		44	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		45
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	46	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	47	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		48
22	#define INIT_SIZE 32 // initial scalar size to be allocated	49	#define INIT_SIZE 32 // initial scalar size to be allocated
		50	#define INDENT_STEP 3 // spaces per indentation level
		51
		52	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		53
24	#define SB do {	54	#define SB do {
25	#define SE } while (0)	55	#define SE } while (0)
26		56
27	static HV *json_stash;	57	#if __GNUC__ >= 3
		58	# define expect(expr,value) __builtin_expect ((expr), (value))
		59	# define INLINE static inline
		60	#else
		61	# define expect(expr,value) (expr)
		62	# define INLINE static
		63	#endif
		64
		65	#define expect_false(expr) expect ((expr) != 0, 0)
		66	#define expect_true(expr) expect ((expr) != 0, 1)
		67
		68	#define IN_RANGE_INC(type,val,beg,end) \
		69	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		70	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		71
		72	#ifdef USE_ITHREADS
		73	# define JSON_SLOW 1
		74	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		75	#else
		76	# define JSON_SLOW 0
		77	# define JSON_STASH json_stash
		78	#endif
		79
		80	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		81	static SV *json_true, *json_false;
		82
		83	enum {
		84	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		85	INCR_M_STR, // inside string
		86	INCR_M_BS, // inside backslash
		87	INCR_M_JSON // outside anything, count nesting
		88	};
		89
		90	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		91
		92	typedef struct {
		93	U32 flags;
		94	SV *cb_object;
		95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	int incr_mode;
		102	} JSON;
		103
		104	/////////////////////////////////////////////////////////////////////////////
		105	// utility functions
		106
		107	INLINE void
		108	shrink (SV *sv)
		109	{
		110	sv_utf8_downgrade (sv, 1);
		111	if (SvLEN (sv) > SvCUR (sv) + 1)
		112	{
		113	#ifdef SvPV_shrink_to_cur
		114	SvPV_shrink_to_cur (sv);
		115	#elif defined (SvPV_renew)
		116	SvPV_renew (sv, SvCUR (sv) + 1);
		117	#endif
		118	}
		119	}
		120
		121	// decode an utf-8 character and return it, or (UV)-1 in
		122	// case of an error.
		123	// we special-case "safe" characters from U+80 .. U+7FF,
		124	// but use the very good perl function to parse anything else.
		125	// note that we never call this function for a ascii codepoints
		126	INLINE UV
		127	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		128	{
		129	if (expect_true (len >= 2
		130	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		131	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		132	{
		133	*clen = 2;
		134	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		135	}
		136	else
		137	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		138	}
		139
		140	// likewise for encoding, also never called for ascii codepoints
		141	// this function takes advantage of this fact, although current gccs
		142	// seem to optimise the check for >= 0x80 away anyways
		143	INLINE unsigned char *
		144	encode_utf8 (unsigned char *s, UV ch)
		145	{
		146	if (expect_false (ch < 0x000080))
		147	*s++ = ch;
		148	else if (expect_true (ch < 0x000800))
		149	*s++ = 0xc0 | ( ch >> 6),
		150	*s++ = 0x80 | ( ch & 0x3f);
		151	else if ( ch < 0x010000)
		152	*s++ = 0xe0 | ( ch >> 12),
		153	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		154	*s++ = 0x80 | ( ch & 0x3f);
		155	else if ( ch < 0x110000)
		156	*s++ = 0xf0 | ( ch >> 18),
		157	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		158	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		159	*s++ = 0x80 | ( ch & 0x3f);
		160
		161	return s;
		162	}
		163
		164	/////////////////////////////////////////////////////////////////////////////
		165	// encoder
28		166
29	// structure used for encoding JSON	167	// structure used for encoding JSON
30	typedef struct	168	typedef struct
31	{	169	{
32	char *cur;	170	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	171	char *end; // SvEND (sv)
35	SV *sv;	172	SV *sv; // result scalar
36	UV flags;	173	JSON json;
37	int max_recurse;	174	U32 indent; // indentation level
38	int indent;	175	U32 maxdepth; // max. indentation/recursion level
		176	UV limit; // escape character values >= this value when encoding
39	} enc_t;	177	} enc_t;
40		178
41	// structure used for decoding JSON	179	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	/////////////////////////////////////////////////////////////////////////////
60
61	static void
62	need (enc_t *enc, STRLEN len)	180	need (enc_t *enc, STRLEN len)
63	{	181	{
64	if (enc->cur + len >= enc->end)	182	if (expect_false (enc->cur + len >= enc->end))
65	{	183	{
66	STRLEN cur = enc->cur - SvPVX (enc->sv);	184	STRLEN cur = enc->cur - SvPVX (enc->sv);
67	SvGROW (enc->sv, cur + len + 1);	185	SvGROW (enc->sv, cur + len + 1);
68	enc->cur = SvPVX (enc->sv) + cur;	186	enc->cur = SvPVX (enc->sv) + cur;
69	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	187	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
70	}	188	}
71	}	189	}
72		190
73	static void	191	INLINE void
74	encode_ch (enc_t *enc, char ch)	192	encode_ch (enc_t *enc, char ch)
75	{	193	{
76	need (enc, 1);	194	need (enc, 1);
77	*enc->cur++ = ch;	195	*enc->cur++ = ch;
78	}	196	}
…		…
86		204
87	while (str < end)	205	while (str < end)
88	{	206	{
89	unsigned char ch = *(unsigned char *)str;	207	unsigned char ch = *(unsigned char *)str;
90		208
91	if (ch >= 0x20 && ch < 0x80) // most common case	209	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
92	{	210	{
93	if (ch == '"') // but with slow exceptions	211	if (expect_false (ch == '"')) // but with slow exceptions
94	{	212	{
95	need (enc, len += 1);	213	need (enc, len += 1);
96	*enc->cur++ = '\\';	214	*enc->cur++ = '\\';
97	*enc->cur++ = '"';	215	*enc->cur++ = '"';
98	}	216	}
99	else if (ch == '\\')	217	else if (expect_false (ch == '\\'))
100	{	218	{
101	need (enc, len += 1);	219	need (enc, len += 1);
102	*enc->cur++ = '\\';	220	*enc->cur++ = '\\';
103	*enc->cur++ = '\\';	221	*enc->cur++ = '\\';
104	}	222	}
…		…
122	STRLEN clen;	240	STRLEN clen;
123	UV uch;	241	UV uch;
124		242
125	if (is_utf8)	243	if (is_utf8)
126	{	244	{
127	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	245	uch = decode_utf8 (str, end - str, &clen);
128	if (clen == (STRLEN)-1)	246	if (clen == (STRLEN)-1)
129	croak ("malformed UTF-8 character in string, cannot convert to JSON");	247	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
130	}	248	}
131	else	249	else
132	{	250	{
133	uch = ch;	251	uch = ch;
134	clen = 1;	252	clen = 1;
135	}	253	}
136		254
137	if (uch < 0x80 || enc->flags & F_ASCII)	255	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
138	{	256	{
139	if (uch > 0xFFFFUL)	257	if (uch >= 0x10000UL)
140	{	258	{
		259	if (uch >= 0x110000UL)
		260	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		261
141	need (enc, len += 11);	262	need (enc, len += 11);
142	sprintf (enc->cur, "\\u%04x\\u%04x",	263	sprintf (enc->cur, "\\u%04x\\u%04x",
143	(uch - 0x10000) / 0x400 + 0xD800,	264	(int)((uch - 0x10000) / 0x400 + 0xD800),
144	(uch - 0x10000) % 0x400 + 0xDC00);	265	(int)((uch - 0x10000) % 0x400 + 0xDC00));
145	enc->cur += 12;	266	enc->cur += 12;
146	}	267	}
147	else	268	else
148	{	269	{
149	static char hexdigit [16] = "0123456789abcdef";	270	static char hexdigit [16] = "0123456789abcdef";
…		…
156	*enc->cur++ = hexdigit [(uch >> 0) & 15];	277	*enc->cur++ = hexdigit [(uch >> 0) & 15];
157	}	278	}
158		279
159	str += clen;	280	str += clen;
160	}	281	}
		282	else if (enc->json.flags & F_LATIN1)
		283	{
		284	*enc->cur++ = uch;
		285	str += clen;
		286	}
161	else if (is_utf8)	287	else if (is_utf8)
162	{	288	{
163	need (enc, len += clen);	289	need (enc, len += clen);
164	do	290	do
165	{	291	{
…		…
167	}	293	}
168	while (--clen);	294	while (--clen);
169	}	295	}
170	else	296	else
171	{	297	{
172	need (enc, 10); // never more than 11 bytes needed	298	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	299	enc->cur = encode_utf8 (enc->cur, uch);
174	++str;	300	++str;
175	}	301	}
176	}	302	}
177	}	303	}
178	}	304	}
179		305
180	--len;	306	--len;
181	}	307	}
182	}	308	}
183		309
184	#define INDENT SB \	310	INLINE void
		311	encode_indent (enc_t *enc)
		312	{
185	if (enc->flags & F_INDENT) \	313	if (enc->json.flags & F_INDENT)
186	{ \	314	{
187	int i_; \	315	int spaces = enc->indent * INDENT_STEP;
188	need (enc, enc->indent); \	316
189	for (i_ = enc->indent * 3; i_--; )\	317	need (enc, spaces);
		318	memset (enc->cur, ' ', spaces);
		319	enc->cur += spaces;
		320	}
		321	}
		322
		323	INLINE void
		324	encode_space (enc_t *enc)
		325	{
		326	need (enc, 1);
		327	encode_ch (enc, ' ');
		328	}
		329
		330	INLINE void
		331	encode_nl (enc_t *enc)
		332	{
		333	if (enc->json.flags & F_INDENT)
		334	{
		335	need (enc, 1);
190	encode_ch (enc, ' '); \	336	encode_ch (enc, '\n');
191	} \	337	}
192	SE	338	}
193		339
194	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	340	INLINE void
195	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	341	encode_comma (enc_t *enc)
196	#define COMMA SB \	342	{
197	encode_ch (enc, ','); \	343	encode_ch (enc, ',');
		344
198	if (enc->flags & F_INDENT) \	345	if (enc->json.flags & F_INDENT)
199	NL; \	346	encode_nl (enc);
200	else if (enc->flags & F_SPACE_AFTER) \	347	else if (enc->json.flags & F_SPACE_AFTER)
201	SPACE; \	348	encode_space (enc);
202	SE	349	}
203		350
204	static void encode_sv (enc_t *enc, SV *sv);	351	static void encode_sv (enc_t *enc, SV *sv);
205		352
206	static void	353	static void
207	encode_av (enc_t *enc, AV *av)	354	encode_av (enc_t *enc, AV *av)
208	{	355	{
209	int i, len = av_len (av);	356	int i, len = av_len (av);
210		357
		358	if (enc->indent >= enc->maxdepth)
		359	croak ("data structure too deep (hit recursion limit)");
		360
211	encode_ch (enc, '['); NL;	361	encode_ch (enc, '[');
212	++enc->indent;	362
		363	if (len >= 0)
		364	{
		365	encode_nl (enc); ++enc->indent;
213		366
214	for (i = 0; i <= len; ++i)	367	for (i = 0; i <= len; ++i)
215	{	368	{
216	INDENT;	369	SV **svp = av_fetch (av, i, 0);
217	encode_sv (enc, *av_fetch (av, i, 0));
218		370
		371	encode_indent (enc);
		372
		373	if (svp)
		374	encode_sv (enc, *svp);
		375	else
		376	encode_str (enc, "null", 4, 0);
		377
219	if (i < len)	378	if (i < len)
220	COMMA;	379	encode_comma (enc);
221	}	380	}
222		381
223	NL;	382	encode_nl (enc); --enc->indent; encode_indent (enc);
224		383	}
225	--enc->indent;	384
226	INDENT; encode_ch (enc, ']');	385	encode_ch (enc, ']');
227	}	386	}
228		387
229	static void	388	static void
230	encode_he (enc_t *enc, HE *he)	389	encode_hk (enc_t *enc, HE *he)
231	{	390	{
232	encode_ch (enc, '"');	391	encode_ch (enc, '"');
233		392
234	if (HeKLEN (he) == HEf_SVKEY)	393	if (HeKLEN (he) == HEf_SVKEY)
235	{	394	{
…		…
245	else	404	else
246	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	405	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
247		406
248	encode_ch (enc, '"');	407	encode_ch (enc, '"');
249		408
250	if (enc->flags & F_SPACE_BEFORE) SPACE;	409	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
251	encode_ch (enc, ':');	410	encode_ch (enc, ':');
252	if (enc->flags & F_SPACE_AFTER ) SPACE;	411	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
253	encode_sv (enc, HeVAL (he));
254	}	412	}
255		413
256	// compare hash entries, used when all keys are bytestrings	414	// compare hash entries, used when all keys are bytestrings
257	static int	415	static int
258	he_cmp_fast (const void *a_, const void *b_)	416	he_cmp_fast (const void *a_, const void *b_)
…		…
263	HE *b = *(HE **)b_;	421	HE *b = *(HE **)b_;
264		422
265	STRLEN la = HeKLEN (a);	423	STRLEN la = HeKLEN (a);
266	STRLEN lb = HeKLEN (b);	424	STRLEN lb = HeKLEN (b);
267		425
268	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	426	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
269	cmp = la < lb ? -1 : la == lb ? 0 : 1;	427	cmp = lb - la;
270		428
271	return cmp;	429	return cmp;
272	}	430	}
273		431
274	// compare hash entries, used when some keys are sv's or utf-x	432	// compare hash entries, used when some keys are sv's or utf-x
275	static int	433	static int
276	he_cmp_slow (const void *a, const void *b)	434	he_cmp_slow (const void *a, const void *b)
277	{	435	{
278	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	436	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
279	}	437	}
280		438
281	static void	439	static void
282	encode_hv (enc_t *enc, HV *hv)	440	encode_hv (enc_t *enc, HV *hv)
283	{	441	{
284	int count, i;	442	HE *he;
285		443
286	encode_ch (enc, '{'); NL; ++enc->indent;	444	if (enc->indent >= enc->maxdepth)
		445	croak ("data structure too deep (hit recursion limit)");
287		446
288	if ((count = hv_iterinit (hv)))	447	encode_ch (enc, '{');
289	{	448
290	// for canonical output we have to sort by keys first	449	// for canonical output we have to sort by keys first
291	// actually, this is mostly due to the stupid so-called	450	// actually, this is mostly due to the stupid so-called
292	// security workaround added somewhere in 5.8.x.	451	// security workaround added somewhere in 5.8.x.
293	// that randomises hash orderings	452	// that randomises hash orderings
294	if (enc->flags & F_CANONICAL)	453	if (enc->json.flags & F_CANONICAL)
		454	{
		455	int count = hv_iterinit (hv);
		456
		457	if (SvMAGICAL (hv))
295	{	458	{
296	HE *he, *hes [count];	459	// need to count by iterating. could improve by dynamically building the vector below
		460	// but I don't care for the speed of this special case.
		461	// note also that we will run into undefined behaviour when the two iterations
		462	// do not result in the same count, something I might care for in some later release.
		463
		464	count = 0;
		465	while (hv_iternext (hv))
		466	++count;
		467
		468	hv_iterinit (hv);
		469	}
		470
		471	if (count)
		472	{
297	int fast = 1;	473	int i, fast = 1;
		474	#if defined(__BORLANDC__) || defined(_MSC_VER)
		475	HE **hes = _alloca (count * sizeof (HE));
		476	#else
		477	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		478	#endif
298		479
299	i = 0;	480	i = 0;
300	while ((he = hv_iternext (hv)))	481	while ((he = hv_iternext (hv)))
301	{	482	{
302	hes [i++] = he;	483	hes [i++] = he;
…		…
308		489
309	if (fast)	490	if (fast)
310	qsort (hes, count, sizeof (HE *), he_cmp_fast);	491	qsort (hes, count, sizeof (HE *), he_cmp_fast);
311	else	492	else
312	{	493	{
313	// hack to disable "use bytes"	494	// hack to forcefully disable "use bytes"
314	COP *oldcop = PL_curcop, cop;	495	COP cop = *PL_curcop;
315	cop.op_private = 0;	496	cop.op_private = 0;
		497
		498	ENTER;
		499	SAVETMPS;
		500
		501	SAVEVPTR (PL_curcop);
316	PL_curcop = &cop;	502	PL_curcop = &cop;
317		503
318	SAVETMPS;
319	qsort (hes, count, sizeof (HE *), he_cmp_slow);	504	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		505
320	FREETMPS;	506	FREETMPS;
321		507	LEAVE;
322	PL_curcop = oldcop;
323	}	508	}
324		509
325	for (i = 0; i < count; ++i)	510	encode_nl (enc); ++enc->indent;
		511
		512	while (count--)
326	{	513	{
327	INDENT;	514	encode_indent (enc);
		515	he = hes [count];
328	encode_he (enc, hes [i]);	516	encode_hk (enc, he);
		517	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
329		518
330	if (i < count - 1)	519	if (count)
331	COMMA;	520	encode_comma (enc);
332	}	521	}
333		522
		523	encode_nl (enc); --enc->indent; encode_indent (enc);
		524	}
		525	}
		526	else
		527	{
		528	if (hv_iterinit (hv) || SvMAGICAL (hv))
		529	if ((he = hv_iternext (hv)))
334	NL;	530	{
		531	encode_nl (enc); ++enc->indent;
		532
		533	for (;;)
		534	{
		535	encode_indent (enc);
		536	encode_hk (enc, he);
		537	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		538
		539	if (!(he = hv_iternext (hv)))
		540	break;
		541
		542	encode_comma (enc);
		543	}
		544
		545	encode_nl (enc); --enc->indent; encode_indent (enc);
		546	}
		547	}
		548
		549	encode_ch (enc, '}');
		550	}
		551
		552	// encode objects, arrays and special \0=false and \1=true values.
		553	static void
		554	encode_rv (enc_t *enc, SV *sv)
		555	{
		556	svtype svt;
		557
		558	SvGETMAGIC (sv);
		559	svt = SvTYPE (sv);
		560
		561	if (expect_false (SvOBJECT (sv)))
		562	{
		563	HV *stash = !JSON_SLOW || json_boolean_stash
		564	? json_boolean_stash
		565	: gv_stashpv ("JSON::XS::Boolean", 1);
		566
		567	if (SvSTASH (sv) == stash)
		568	{
		569	if (SvIV (sv))
		570	encode_str (enc, "true", 4, 0);
		571	else
		572	encode_str (enc, "false", 5, 0);
335	}	573	}
336	else	574	else
337	{	575	{
338	SV *sv;	576	#if 0
339	HE *he = hv_iternext (hv);	577	if (0 && sv_derived_from (rv, "JSON::Literal"))
340
341	for (;;)
342	{	578	{
343	INDENT;	579	// not yet
344	encode_he (enc, he);
345
346	if (!(he = hv_iternext (hv)))
347	break;
348
349	COMMA;
350	}	580	}
351		581	#endif
		582	if (enc->json.flags & F_CONV_BLESSED)
352	NL;	583	{
		584	// we re-bless the reference to get overload and other niceties right
		585	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		586
		587	if (to_json)
		588	{
		589	dSP;
		590
		591	ENTER; SAVETMPS; PUSHMARK (SP);
		592	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		593
		594	// calling with G_SCALAR ensures that we always get a 1 return value
		595	PUTBACK;
		596	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		597	SPAGAIN;
		598
		599	// catch this surprisingly common error
		600	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		601	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		602
		603	sv = POPs;
		604	PUTBACK;
		605
		606	encode_sv (enc, sv);
		607
		608	FREETMPS; LEAVE;
		609	}
		610	else if (enc->json.flags & F_ALLOW_BLESSED)
		611	encode_str (enc, "null", 4, 0);
		612	else
		613	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		614	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		615	}
		616	else if (enc->json.flags & F_ALLOW_BLESSED)
		617	encode_str (enc, "null", 4, 0);
		618	else
		619	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		620	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
353	}	621	}
354	}	622	}
		623	else if (svt == SVt_PVHV)
		624	encode_hv (enc, (HV *)sv);
		625	else if (svt == SVt_PVAV)
		626	encode_av (enc, (AV *)sv);
		627	else if (svt < SVt_PVAV)
		628	{
		629	STRLEN len = 0;
		630	char *pv = svt ? SvPV (sv, len) : 0;
355		631
356	--enc->indent; INDENT; encode_ch (enc, '}');	632	if (len == 1 && *pv == '1')
		633	encode_str (enc, "true", 4, 0);
		634	else if (len == 1 && *pv == '0')
		635	encode_str (enc, "false", 5, 0);
		636	else
		637	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		638	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		639	}
		640	else
		641	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		642	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
357	}	643	}
358		644
359	static void	645	static void
360	encode_sv (enc_t *enc, SV *sv)	646	encode_sv (enc_t *enc, SV *sv)
361	{	647	{
…		…
369	encode_str (enc, str, len, SvUTF8 (sv));	655	encode_str (enc, str, len, SvUTF8 (sv));
370	encode_ch (enc, '"');	656	encode_ch (enc, '"');
371	}	657	}
372	else if (SvNOKp (sv))	658	else if (SvNOKp (sv))
373	{	659	{
		660	// trust that perl will do the right thing w.r.t. JSON syntax.
374	need (enc, NV_DIG + 32);	661	need (enc, NV_DIG + 32);
375	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	662	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
376	enc->cur += strlen (enc->cur);	663	enc->cur += strlen (enc->cur);
377	}	664	}
378	else if (SvIOKp (sv))	665	else if (SvIOKp (sv))
379	{	666	{
		667	// we assume we can always read an IV as a UV and vice versa
		668	// we assume two's complement
		669	// we assume no aliasing issues in the union
		670	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		671	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		672	{
		673	// optimise the "small number case"
		674	// code will likely be branchless and use only a single multiplication
		675	// works for numbers up to 59074
		676	I32 i = SvIVX (sv);
		677	U32 u;
		678	char digit, nz = 0;
		679
380	need (enc, 64);	680	need (enc, 6);
		681
		682	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		683	u = i < 0 ? -i : i;
		684
		685	// convert to 4.28 fixed-point representation
		686	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		687
		688	// now output digit by digit, each time masking out the integer part
		689	// and multiplying by 5 while moving the decimal point one to the right,
		690	// resulting in a net multiplication by 10.
		691	// we always write the digit to memory but conditionally increment
		692	// the pointer, to enable the use of conditional move instructions.
		693	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		694	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		695	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		696	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		697	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		698	}
		699	else
		700	{
		701	// large integer, use the (rather slow) snprintf way.
		702	need (enc, IVUV_MAXCHARS);
381	enc->cur +=	703	enc->cur +=
382	SvIsUV(sv)	704	SvIsUV(sv)
383	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	705	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
384	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	706	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		707	}
385	}	708	}
386	else if (SvROK (sv))	709	else if (SvROK (sv))
387	{	710	encode_rv (enc, SvRV (sv));
388	if (!--enc->max_recurse)
389	croak ("data structure too deep (hit recursion limit)");
390
391	sv = SvRV (sv);
392
393	switch (SvTYPE (sv))
394	{
395	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
396	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
397
398	default:
399	croak ("JSON can only represent references to arrays or hashes");
400	}
401	}
402	else if (!SvOK (sv))	711	else if (!SvOK (sv))
403	encode_str (enc, "null", 4, 0);	712	encode_str (enc, "null", 4, 0);
404	else	713	else
405	croak ("encountered perl type that JSON cannot handle");	714	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		715	SvPV_nolen (sv), SvFLAGS (sv));
406	}	716	}
407		717
408	static SV *	718	static SV *
409	encode_json (SV *scalar, UV flags)	719	encode_json (SV *scalar, JSON *json)
410	{	720	{
411	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
412	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
413
414	enc_t enc;	721	enc_t enc;
415	enc.flags = flags;	722
		723	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		724	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		725
		726	enc.json = *json;
416	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	727	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
417	enc.cur = SvPVX (enc.sv);	728	enc.cur = SvPVX (enc.sv);
418	enc.end = SvEND (enc.sv);	729	enc.end = SvEND (enc.sv);
419	enc.max_recurse = 0;
420	enc.indent = 0;	730	enc.indent = 0;
		731	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		732	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		733	: enc.json.flags & F_LATIN1 ? 0x000100UL
		734	: 0x110000UL;
421		735
422	SvPOK_only (enc.sv);	736	SvPOK_only (enc.sv);
423	encode_sv (&enc, scalar);	737	encode_sv (&enc, scalar);
424		738
		739	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		740	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		741
425	if (!(flags & (F_ASCII | F_UTF8)))	742	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
426	SvUTF8_on (enc.sv);	743	SvUTF8_on (enc.sv);
427		744
428	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
429
430	#ifdef SvPV_shrink_to_cur
431	if (enc.flags & F_SHRINK)	745	if (enc.json.flags & F_SHRINK)
432	SvPV_shrink_to_cur (enc.sv);	746	shrink (enc.sv);
433	#endif	747
434	return enc.sv;	748	return enc.sv;
435	}	749	}
436		750
437	/////////////////////////////////////////////////////////////////////////////	751	/////////////////////////////////////////////////////////////////////////////
		752	// decoder
438		753
439	#define WS \	754	// structure used for decoding JSON
		755	typedef struct
		756	{
		757	char *cur; // current parser pointer
		758	char *end; // end of input string
		759	const char *err; // parse error, if != 0
		760	JSON json;
		761	U32 depth; // recursion depth
		762	U32 maxdepth; // recursion depth limit
		763	} dec_t;
		764
		765	INLINE void
		766	decode_comment (dec_t *dec)
		767	{
		768	// only '#'-style comments allowed a.t.m.
		769
		770	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		771	++dec->cur;
		772	}
		773
		774	INLINE void
		775	decode_ws (dec_t *dec)
		776	{
440	for (;;) \	777	for (;;)
441	{ \	778	{
442	char ch = *dec->cur; \	779	char ch = *dec->cur;
		780
443	if (ch > 0x20 \	781	if (ch > 0x20)
		782	{
		783	if (expect_false (ch == '#'))
		784	{
		785	if (dec->json.flags & F_RELAXED)
		786	decode_comment (dec);
		787	else
		788	break;
		789	}
		790	else
		791	break;
		792	}
444	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	793	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
445	break; \	794	break; // parse error, but let higher level handle it, gives better error messages
		795
446	++dec->cur; \	796	++dec->cur;
447	}	797	}
		798	}
448		799
449	#define ERR(reason) SB dec->err = reason; goto fail; SE	800	#define ERR(reason) SB dec->err = reason; goto fail; SE
		801
450	#define EXPECT_CH(ch) SB \	802	#define EXPECT_CH(ch) SB \
451	if (*dec->cur != ch) \	803	if (*dec->cur != ch) \
452	ERR (# ch " expected"); \	804	ERR (# ch " expected"); \
453	++dec->cur; \	805	++dec->cur; \
454	SE	806	SE
455		807
		808	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		809	#define DEC_DEC_DEPTH --dec->depth
		810
456	static SV *decode_sv (dec_t *dec);	811	static SV *decode_sv (dec_t *dec);
457		812
458	static signed char decode_hexdigit[256];	813	static signed char decode_hexdigit[256];
459		814
460	static UV	815	static UV
461	decode_4hex (dec_t *dec)	816	decode_4hex (dec_t *dec)
462	{	817	{
463	signed char d1, d2, d3, d4;	818	signed char d1, d2, d3, d4;
		819	unsigned char *cur = (unsigned char *)dec->cur;
464		820
465	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	821	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
466	if (d1 < 0) ERR ("four hexadecimal digits expected");	822	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
467	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	823	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
468	if (d2 < 0) ERR ("four hexadecimal digits expected");	824	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
469	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
470	if (d3 < 0) ERR ("four hexadecimal digits expected");
471	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
472	if (d4 < 0) ERR ("four hexadecimal digits expected");
473		825
474	dec->cur += 4;	826	dec->cur += 4;
475		827
476	return ((UV)d1) << 12	828	return ((UV)d1) << 12
477	| ((UV)d2) << 8	829	| ((UV)d2) << 8
…		…
480		832
481	fail:	833	fail:
482	return (UV)-1;	834	return (UV)-1;
483	}	835	}
484		836
485	#define APPEND_GROW(n) SB \
486	if (cur + (n) >= end) \
487	{ \
488	STRLEN ofs = cur - SvPVX (sv); \
489	SvGROW (sv, ofs + (n) + 1); \
490	cur = SvPVX (sv) + ofs; \
491	end = SvEND (sv); \
492	} \
493	SE
494
495	#define APPEND_CH(ch) SB \
496	APPEND_GROW (1); \
497	*cur++ = (ch); \
498	SE
499
500	static SV *	837	static SV *
501	decode_str (dec_t *dec)	838	decode_str (dec_t *dec)
502	{	839	{
503	SV *sv = NEWSV (0,2);	840	SV *sv = 0;
504	int utf8 = 0;	841	int utf8 = 0;
505	char *cur = SvPVX (sv);	842	char *dec_cur = dec->cur;
506	char *end = SvEND (sv);
507		843
508	for (;;)	844	do
509	{	845	{
510	unsigned char ch = *(unsigned char *)dec->cur;	846	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		847	char *cur = buf;
511		848
512	if (ch == '"')	849	do
513	break;
514	else if (ch == '\\')
515	{	850	{
516	switch (*++dec->cur)	851	unsigned char ch = *(unsigned char *)dec_cur++;
		852
		853	if (expect_false (ch == '"'))
517	{	854	{
518	case '\\':	855	--dec_cur;
519	case '/':	856	break;
520	case '"': APPEND_CH (*dec->cur++); break;	857	}
521		858	else if (expect_false (ch == '\\'))
522	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	859	{
523	case 't': APPEND_CH ('\011'); ++dec->cur; break;	860	switch (*dec_cur)
524	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
525	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
526	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
527
528	case 'u':
529	{	861	{
530	UV lo, hi;	862	case '\\':
531	++dec->cur;	863	case '/':
		864	case '"': *cur++ = *dec_cur++; break;
532		865
533	hi = decode_4hex (dec);	866	case 'b': ++dec_cur; *cur++ = '\010'; break;
534	if (hi == (UV)-1)	867	case 't': ++dec_cur; *cur++ = '\011'; break;
535	goto fail;	868	case 'n': ++dec_cur; *cur++ = '\012'; break;
		869	case 'f': ++dec_cur; *cur++ = '\014'; break;
		870	case 'r': ++dec_cur; *cur++ = '\015'; break;
536		871
537	// possibly a surrogate pair	872	case 'u':
538	if (hi >= 0xd800 && hi < 0xdc00)
539	{	873	{
540	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	874	UV lo, hi;
541	ERR ("missing low surrogate character in surrogate pair");	875	++dec_cur;
542		876
543	dec->cur += 2;	877	dec->cur = dec_cur;
544
545	lo = decode_4hex (dec);	878	hi = decode_4hex (dec);
		879	dec_cur = dec->cur;
546	if (lo == (UV)-1)	880	if (hi == (UV)-1)
547	goto fail;	881	goto fail;
548		882
		883	// possibly a surrogate pair
		884	if (hi >= 0xd800)
		885	if (hi < 0xdc00)
		886	{
		887	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		888	ERR ("missing low surrogate character in surrogate pair");
		889
		890	dec_cur += 2;
		891
		892	dec->cur = dec_cur;
		893	lo = decode_4hex (dec);
		894	dec_cur = dec->cur;
		895	if (lo == (UV)-1)
		896	goto fail;
		897
549	if (lo < 0xdc00 || lo >= 0xe000)	898	if (lo < 0xdc00 || lo >= 0xe000)
550	ERR ("surrogate pair expected");	899	ERR ("surrogate pair expected");
551		900
552	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	901	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		902	}
		903	else if (hi < 0xe000)
		904	ERR ("missing high surrogate character in surrogate pair");
		905
		906	if (hi >= 0x80)
		907	{
		908	utf8 = 1;
		909
		910	cur = encode_utf8 (cur, hi);
		911	}
		912	else
		913	*cur++ = hi;
553	}	914	}
554	else if (hi >= 0xdc00 && hi < 0xe000)
555	ERR ("missing high surrogate character in surrogate pair");
556
557	if (hi >= 0x80)
558	{	915	break;
559	utf8 = 1;
560		916
561	APPEND_GROW (4); // at most 4 bytes for 21 bits
562	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
563	}
564	else	917	default:
565	APPEND_CH (hi);	918	--dec_cur;
		919	ERR ("illegal backslash escape sequence in string");
566	}	920	}
567	break;	921	}
		922	else if (expect_true (ch >= 0x20 && ch < 0x80))
		923	*cur++ = ch;
		924	else if (ch >= 0x80)
		925	{
		926	STRLEN clen;
		927	UV uch;
568		928
569	default:
570	--dec->cur;	929	--dec_cur;
571	ERR ("illegal backslash escape sequence in string");	930
		931	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		932	if (clen == (STRLEN)-1)
		933	ERR ("malformed UTF-8 character in JSON string");
		934
		935	do
		936	*cur++ = *dec_cur++;
		937	while (--clen);
		938
		939	utf8 = 1;
		940	}
		941	else
		942	{
		943	--dec_cur;
		944
		945	if (!ch)
		946	ERR ("unexpected end of string while parsing JSON string");
		947	else
		948	ERR ("invalid character encountered while parsing JSON string");
572	}	949	}
573	}	950	}
574	else if (ch >= 0x20 && ch <= 0x7f)	951	while (cur < buf + SHORT_STRING_LEN);
575	APPEND_CH (*dec->cur++);	952
576	else if (ch >= 0x80)
577	{	953	{
578	STRLEN clen;	954	STRLEN len = cur - buf;
579	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
580	if (clen == (STRLEN)-1)
581	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
582		955
583	APPEND_GROW (clen);	956	if (sv)
584	do
585	{	957	{
586	*cur++ = *dec->cur++;	958	SvGROW (sv, SvCUR (sv) + len + 1);
		959	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		960	SvCUR_set (sv, SvCUR (sv) + len);
587	}	961	}
588	while (--clen);
589
590	utf8 = 1;
591	}
592	else if (dec->cur == dec->end)
593	ERR ("unexpected end of string while parsing json string");
594	else	962	else
595	ERR ("invalid character encountered");	963	sv = newSVpvn (buf, len);
596	}	964	}
		965	}
		966	while (*dec_cur != '"');
597		967
598	++dec->cur;	968	++dec_cur;
599		969
600	SvCUR_set (sv, cur - SvPVX (sv));	970	if (sv)
601		971	{
602	SvPOK_only (sv);	972	SvPOK_only (sv);
603	*SvEND (sv) = 0;	973	*SvEND (sv) = 0;
604		974
605	if (utf8)	975	if (utf8)
606	SvUTF8_on (sv);	976	SvUTF8_on (sv);
		977	}
		978	else
		979	sv = newSVpvn ("", 0);
607		980
608	#ifdef SvPV_shrink_to_cur	981	dec->cur = dec_cur;
609	if (dec->flags & F_SHRINK)
610	SvPV_shrink_to_cur (sv);
611	#endif
612
613	return sv;	982	return sv;
614		983
615	fail:	984	fail:
616	SvREFCNT_dec (sv);	985	dec->cur = dec_cur;
617	return 0;	986	return 0;
618	}	987	}
619		988
620	static SV *	989	static SV *
621	decode_num (dec_t *dec)	990	decode_num (dec_t *dec)
…		…
679	is_nv = 1;	1048	is_nv = 1;
680	}	1049	}
681		1050
682	if (!is_nv)	1051	if (!is_nv)
683	{	1052	{
684	UV uv;	1053	int len = dec->cur - start;
685	int numtype = grok_number (start, dec->cur - start, &uv);	1054
686	if (numtype & IS_NUMBER_IN_UV)	1055	// special case the rather common 1..5-digit-int case
687	if (numtype & IS_NUMBER_NEG)	1056	if (*start == '-')
		1057	switch (len)
688	{	1058	{
689	if (uv < (UV)IV_MIN)	1059	case 2: return newSViv (-( start [1] - '0' * 1));
690	return newSViv (-(IV)uv);	1060	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1061	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1062	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1063	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
691	}	1064	}
		1065	else
		1066	switch (len)
		1067	{
		1068	case 1: return newSViv ( start [0] - '0' * 1);
		1069	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1070	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1071	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1072	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1073	}
		1074
		1075	{
		1076	UV uv;
		1077	int numtype = grok_number (start, len, &uv);
		1078	if (numtype & IS_NUMBER_IN_UV)
		1079	if (numtype & IS_NUMBER_NEG)
		1080	{
		1081	if (uv < (UV)IV_MIN)
		1082	return newSViv (-(IV)uv);
		1083	}
692	else	1084	else
693	return newSVuv (uv);	1085	return newSVuv (uv);
694	}	1086	}
695		1087
		1088	len -= *start == '-' ? 1 : 0;
		1089
		1090	// does not fit into IV or UV, try NV
		1091	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1092	#if defined (LDBL_DIG)
		1093	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1094	#endif
		1095	)
		1096	// fits into NV without loss of precision
		1097	return newSVnv (Atof (start));
		1098
		1099	// everything else fails, convert it to a string
		1100	return newSVpvn (start, dec->cur - start);
		1101	}
		1102
		1103	// loss of precision here
696	return newSVnv (Atof (start));	1104	return newSVnv (Atof (start));
697		1105
698	fail:	1106	fail:
699	return 0;	1107	return 0;
700	}	1108	}
…		…
702	static SV *	1110	static SV *
703	decode_av (dec_t *dec)	1111	decode_av (dec_t *dec)
704	{	1112	{
705	AV *av = newAV ();	1113	AV *av = newAV ();
706		1114
707	WS;	1115	DEC_INC_DEPTH;
		1116	decode_ws (dec);
		1117
708	if (*dec->cur == ']')	1118	if (*dec->cur == ']')
709	++dec->cur;	1119	++dec->cur;
710	else	1120	else
711	for (;;)	1121	for (;;)
712	{	1122	{
…		…
716	if (!value)	1126	if (!value)
717	goto fail;	1127	goto fail;
718		1128
719	av_push (av, value);	1129	av_push (av, value);
720		1130
721	WS;	1131	decode_ws (dec);
722		1132
723	if (*dec->cur == ']')	1133	if (*dec->cur == ']')
724	{	1134	{
725	++dec->cur;	1135	++dec->cur;
726	break;	1136	break;
…		…
728		1138
729	if (*dec->cur != ',')	1139	if (*dec->cur != ',')
730	ERR (", or ] expected while parsing array");	1140	ERR (", or ] expected while parsing array");
731		1141
732	++dec->cur;	1142	++dec->cur;
		1143
		1144	decode_ws (dec);
		1145
		1146	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1147	{
		1148	++dec->cur;
		1149	break;
		1150	}
733	}	1151	}
734		1152
		1153	DEC_DEC_DEPTH;
735	return newRV_noinc ((SV *)av);	1154	return newRV_noinc ((SV *)av);
736		1155
737	fail:	1156	fail:
738	SvREFCNT_dec (av);	1157	SvREFCNT_dec (av);
		1158	DEC_DEC_DEPTH;
739	return 0;	1159	return 0;
740	}	1160	}
741		1161
742	static SV *	1162	static SV *
743	decode_hv (dec_t *dec)	1163	decode_hv (dec_t *dec)
744	{	1164	{
		1165	SV *sv;
745	HV *hv = newHV ();	1166	HV *hv = newHV ();
746		1167
747	WS;	1168	DEC_INC_DEPTH;
		1169	decode_ws (dec);
		1170
748	if (*dec->cur == '}')	1171	if (*dec->cur == '}')
749	++dec->cur;	1172	++dec->cur;
750	else	1173	else
751	for (;;)	1174	for (;;)
752	{	1175	{
753	SV *key, *value;
754
755	WS; EXPECT_CH ('"');	1176	EXPECT_CH ('"');
756		1177
757	key = decode_str (dec);	1178	// heuristic: assume that
758	if (!key)	1179	// a) decode_str + hv_store_ent are abysmally slow.
759	goto fail;	1180	// b) most hash keys are short, simple ascii text.
		1181	// => try to "fast-match" such strings to avoid
		1182	// the overhead of decode_str + hv_store_ent.
		1183	{
		1184	SV *value;
		1185	char *p = dec->cur;
		1186	char *e = p + 24; // only try up to 24 bytes
760		1187
761	WS; EXPECT_CH (':');	1188	for (;;)
762
763	value = decode_sv (dec);
764	if (!value)
765	{	1189	{
		1190	// the >= 0x80 is false on most architectures
		1191	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1192	{
		1193	// slow path, back up and use decode_str
		1194	SV *key = decode_str (dec);
		1195	if (!key)
		1196	goto fail;
		1197
		1198	decode_ws (dec); EXPECT_CH (':');
		1199
		1200	decode_ws (dec);
		1201	value = decode_sv (dec);
		1202	if (!value)
		1203	{
		1204	SvREFCNT_dec (key);
		1205	goto fail;
		1206	}
		1207
		1208	hv_store_ent (hv, key, value, 0);
766	SvREFCNT_dec (key);	1209	SvREFCNT_dec (key);
		1210
		1211	break;
		1212	}
		1213	else if (*p == '"')
		1214	{
		1215	// fast path, got a simple key
		1216	char *key = dec->cur;
		1217	int len = p - key;
		1218	dec->cur = p + 1;
		1219
		1220	decode_ws (dec); EXPECT_CH (':');
		1221
		1222	decode_ws (dec);
		1223	value = decode_sv (dec);
		1224	if (!value)
767	goto fail;	1225	goto fail;
		1226
		1227	hv_store (hv, key, len, value, 0);
		1228
		1229	break;
		1230	}
		1231
		1232	++p;
768	}	1233	}
769
770	//TODO: optimise
771	hv_store_ent (hv, key, value, 0);
772
773	WS;	1234	}
		1235
		1236	decode_ws (dec);
774		1237
775	if (*dec->cur == '}')	1238	if (*dec->cur == '}')
776	{	1239	{
777	++dec->cur;	1240	++dec->cur;
778	break;	1241	break;
…		…
780		1243
781	if (*dec->cur != ',')	1244	if (*dec->cur != ',')
782	ERR (", or } expected while parsing object/hash");	1245	ERR (", or } expected while parsing object/hash");
783		1246
784	++dec->cur;	1247	++dec->cur;
		1248
		1249	decode_ws (dec);
		1250
		1251	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1252	{
		1253	++dec->cur;
		1254	break;
		1255	}
785	}	1256	}
786		1257
		1258	DEC_DEC_DEPTH;
787	return newRV_noinc ((SV *)hv);	1259	sv = newRV_noinc ((SV *)hv);
		1260
		1261	// check filter callbacks
		1262	if (dec->json.flags & F_HOOK)
		1263	{
		1264	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1265	{
		1266	HE *cb, *he;
		1267
		1268	hv_iterinit (hv);
		1269	he = hv_iternext (hv);
		1270	hv_iterinit (hv);
		1271
		1272	// the next line creates a mortal sv each time its called.
		1273	// might want to optimise this for common cases.
		1274	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1275
		1276	if (cb)
		1277	{
		1278	dSP;
		1279	int count;
		1280
		1281	ENTER; SAVETMPS; PUSHMARK (SP);
		1282	XPUSHs (HeVAL (he));
		1283
		1284	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1285
		1286	if (count == 1)
		1287	{
		1288	sv = newSVsv (POPs);
		1289	FREETMPS; LEAVE;
		1290	return sv;
		1291	}
		1292
		1293	FREETMPS; LEAVE;
		1294	}
		1295	}
		1296
		1297	if (dec->json.cb_object)
		1298	{
		1299	dSP;
		1300	int count;
		1301
		1302	ENTER; SAVETMPS; PUSHMARK (SP);
		1303	XPUSHs (sv_2mortal (sv));
		1304
		1305	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1306
		1307	if (count == 1)
		1308	{
		1309	sv = newSVsv (POPs);
		1310	FREETMPS; LEAVE;
		1311	return sv;
		1312	}
		1313
		1314	SvREFCNT_inc (sv);
		1315	FREETMPS; LEAVE;
		1316	}
		1317	}
		1318
		1319	return sv;
788		1320
789	fail:	1321	fail:
790	SvREFCNT_dec (hv);	1322	SvREFCNT_dec (hv);
		1323	DEC_DEC_DEPTH;
791	return 0;	1324	return 0;
792	}	1325	}
793		1326
794	static SV *	1327	static SV *
795	decode_sv (dec_t *dec)	1328	decode_sv (dec_t *dec)
796	{	1329	{
797	WS;	1330	// the beauty of JSON: you need exactly one character lookahead
		1331	// to parse everything.
798	switch (*dec->cur)	1332	switch (*dec->cur)
799	{	1333	{
800	case '"': ++dec->cur; return decode_str (dec);	1334	case '"': ++dec->cur; return decode_str (dec);
801	case '[': ++dec->cur; return decode_av (dec);	1335	case '[': ++dec->cur; return decode_av (dec);
802	case '{': ++dec->cur; return decode_hv (dec);	1336	case '{': ++dec->cur; return decode_hv (dec);
803		1337
804	case '-':	1338	case '-':
805	case '0': case '1': case '2': case '3': case '4':	1339	case '0': case '1': case '2': case '3': case '4':
806	case '5': case '6': case '7': case '8': case '9':	1340	case '5': case '6': case '7': case '8': case '9':
807	return decode_num (dec);	1341	return decode_num (dec);
808		1342
809	case 't':	1343	case 't':
810	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1344	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
811	{	1345	{
812	dec->cur += 4;	1346	dec->cur += 4;
813	return newSViv (1);	1347	#if JSON_SLOW
		1348	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1349	#endif
		1350	return SvREFCNT_inc (json_true);
814	}	1351	}
815	else	1352	else
816	ERR ("'true' expected");	1353	ERR ("'true' expected");
817		1354
818	break;	1355	break;
819		1356
820	case 'f':	1357	case 'f':
821	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1358	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
822	{	1359	{
823	dec->cur += 5;	1360	dec->cur += 5;
824	return newSViv (0);	1361	#if JSON_SLOW
		1362	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1363	#endif
		1364	return SvREFCNT_inc (json_false);
825	}	1365	}
826	else	1366	else
827	ERR ("'false' expected");	1367	ERR ("'false' expected");
828		1368
829	break;	1369	break;
…		…
838	ERR ("'null' expected");	1378	ERR ("'null' expected");
839		1379
840	break;	1380	break;
841		1381
842	default:	1382	default:
843	ERR ("malformed json string");	1383	ERR ("malformed JSON string, neither array, object, number, string or atom");
844	break;	1384	break;
845	}	1385	}
846		1386
847	fail:	1387	fail:
848	return 0;	1388	return 0;
849	}	1389	}
850		1390
851	static SV *	1391	static SV *
852	decode_json (SV *string, UV flags)	1392	decode_json (SV *string, JSON *json, STRLEN *offset_return)
853	{	1393	{
		1394	dec_t dec;
		1395	STRLEN offset;
854	SV *sv;	1396	SV *sv;
855		1397
		1398	SvGETMAGIC (string);
		1399	SvUPGRADE (string, SVt_PV);
		1400
		1401	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1402	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1403	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1404
856	if (flags & F_UTF8)	1405	if (json->flags & F_UTF8)
857	sv_utf8_downgrade (string, 0);	1406	sv_utf8_downgrade (string, 0);
858	else	1407	else
859	sv_utf8_upgrade (string);	1408	sv_utf8_upgrade (string);
860		1409
861	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1410	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
862		1411
863	dec_t dec;	1412	dec.json = *json;
864	dec.flags = flags;
865	dec.cur = SvPVX (string);	1413	dec.cur = SvPVX (string);
866	dec.end = SvEND (string);	1414	dec.end = SvEND (string);
867	dec.err = 0;	1415	dec.err = 0;
		1416	dec.depth = 0;
		1417	dec.maxdepth = DEC_DEPTH (dec.json.flags);
868		1418
		1419	if (dec.json.cb_object || dec.json.cb_sk_object)
		1420	dec.json.flags |= F_HOOK;
		1421
		1422	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1423
		1424	decode_ws (&dec);
869	sv = decode_sv (&dec);	1425	sv = decode_sv (&dec);
870		1426
		1427	if (!(offset_return || !sv))
		1428	{
		1429	// check for trailing garbage
		1430	decode_ws (&dec);
		1431
		1432	if (*dec.cur)
		1433	{
		1434	dec.err = "garbage after JSON object";
		1435	SvREFCNT_dec (sv);
		1436	sv = 0;
		1437	}
		1438	}
		1439
		1440	if (offset_return || !sv)
		1441	{
		1442	offset = dec.json.flags & F_UTF8
		1443	? dec.cur - SvPVX (string)
		1444	: utf8_distance (dec.cur, SvPVX (string));
		1445
		1446	if (offset_return)
		1447	*offset_return = offset;
		1448	}
		1449
871	if (!sv)	1450	if (!sv)
872	{	1451	{
873	IV offset = utf8_distance (dec.cur, SvPVX (string));
874	SV *uni = sv_newmortal ();	1452	SV *uni = sv_newmortal ();
		1453
875	// horrible hack to silence warning inside pv_uni_display	1454	// horrible hack to silence warning inside pv_uni_display
876	COP cop;	1455	COP cop = *PL_curcop;
877	memset (&cop, 0, sizeof (cop));
878	cop.cop_warnings = pWARN_NONE;	1456	cop.cop_warnings = pWARN_NONE;
		1457	ENTER;
879	SAVEVPTR (PL_curcop);	1458	SAVEVPTR (PL_curcop);
880	PL_curcop = &cop;	1459	PL_curcop = &cop;
881
882	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1460	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1461	LEAVE;
		1462
883	croak ("%s, at character offset %d (%s)",	1463	croak ("%s, at character offset %d [\"%s\"]",
884	dec.err,	1464	dec.err,
885	(int)offset,	1465	(int)offset,
886	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1466	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
887	}	1467	}
888		1468
889	sv = sv_2mortal (sv);	1469	sv = sv_2mortal (sv);
890		1470
891	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1471	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
892	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1472	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
893		1473
894	return sv;	1474	return sv;
895	}	1475	}
896		1476
		1477	/////////////////////////////////////////////////////////////////////////////
		1478	// incremental parser
		1479
		1480	static void
		1481	incr_parse (JSON *self)
		1482	{
		1483	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1484
		1485	for (;;)
		1486	{
		1487	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1488	switch (self->incr_mode)
		1489	{
		1490	// only used for intiial whitespace skipping
		1491	case INCR_M_WS:
		1492	for (;;)
		1493	{
		1494	if (*p > 0x20)
		1495	{
		1496	self->incr_mode = INCR_M_JSON;
		1497	goto incr_m_json;
		1498	}
		1499	else if (!*p)
		1500	goto interrupt;
		1501
		1502	++p;
		1503	}
		1504
		1505	// skip a single char inside a string (for \\-processing)
		1506	case INCR_M_BS:
		1507	if (!*p)
		1508	goto interrupt;
		1509
		1510	++p;
		1511	self->incr_mode = INCR_M_STR;
		1512	goto incr_m_str;
		1513
		1514	// inside a string
		1515	case INCR_M_STR:
		1516	incr_m_str:
		1517	for (;;)
		1518	{
		1519	if (*p == '"')
		1520	{
		1521	++p;
		1522	self->incr_mode = INCR_M_JSON;
		1523
		1524	if (!self->incr_nest)
		1525	goto interrupt;
		1526
		1527	goto incr_m_json;
		1528	}
		1529	else if (*p == '\\')
		1530	{
		1531	++p; // "virtually" consumes character after \
		1532
		1533	if (!*p) // if at end of string we have to switch modes
		1534	{
		1535	self->incr_mode = INCR_M_BS;
		1536	goto interrupt;
		1537	}
		1538	}
		1539	else if (!*p)
		1540	goto interrupt;
		1541
		1542	++p;
		1543	}
		1544
		1545	// after initial ws, outside string
		1546	case INCR_M_JSON:
		1547	incr_m_json:
		1548	for (;;)
		1549	{
		1550	switch (*p++)
		1551	{
		1552	case 0:
		1553	--p;
		1554	goto interrupt;
		1555
		1556	case 0x09:
		1557	case 0x0a:
		1558	case 0x0d:
		1559	case 0x20:
		1560	if (!self->incr_nest)
		1561	{
		1562	--p; // do not eat the whitespace, let the next round do it
		1563	goto interrupt;
		1564	}
		1565	break;
		1566
		1567	case '"':
		1568	self->incr_mode = INCR_M_STR;
		1569	goto incr_m_str;
		1570
		1571	case '[':
		1572	case '{':
		1573	++self->incr_nest;
		1574	break;
		1575
		1576	case ']':
		1577	case '}':
		1578	if (!--self->incr_nest)
		1579	goto interrupt;
		1580	}
		1581	}
		1582	}
		1583
		1584	modechange:
		1585	;
		1586	}
		1587
		1588	interrupt:
		1589	self->incr_pos = p - SvPVX (self->incr_text);
		1590	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1591	}
		1592
		1593	/////////////////////////////////////////////////////////////////////////////
		1594	// XS interface functions
		1595
897	MODULE = JSON::XS PACKAGE = JSON::XS	1596	MODULE = JSON::XS PACKAGE = JSON::XS
898		1597
899	BOOT:	1598	BOOT:
900	{	1599	{
901	int i;	1600	int i;
902		1601
903	memset (decode_hexdigit, 0xff, 256);
904	for (i = 10; i--; )	1602	for (i = 0; i < 256; ++i)
905	decode_hexdigit ['0' + i] = i;	1603	decode_hexdigit [i] =
		1604	i >= '0' && i <= '9' ? i - '0'
		1605	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1606	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1607	: -1;
906		1608
907	for (i = 7; i--; )
908	{
909	decode_hexdigit ['a' + i] = 10 + i;
910	decode_hexdigit ['A' + i] = 10 + i;
911	}
912
913	json_stash = gv_stashpv ("JSON::XS", 1);	1609	json_stash = gv_stashpv ("JSON::XS" , 1);
		1610	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1611
		1612	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1613	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
914	}	1614	}
915		1615
916	PROTOTYPES: DISABLE	1616	PROTOTYPES: DISABLE
917		1617
918	SV *new (char *dummy)	1618	void CLONE (...)
919	CODE:	1619	CODE:
920	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1620	json_stash = 0;
		1621	json_boolean_stash = 0;
		1622
		1623	void new (char *klass)
		1624	PPCODE:
		1625	{
		1626	SV *pv = NEWSV (0, sizeof (JSON));
		1627	SvPOK_only (pv);
		1628	Zero (SvPVX (pv), 1, JSON);
		1629	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1630	XPUSHs (sv_2mortal (sv_bless (
		1631	newRV_noinc (pv),
		1632	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1633	)));
		1634	}
		1635
		1636	void ascii (JSON *self, int enable = 1)
		1637	ALIAS:
		1638	ascii = F_ASCII
		1639	latin1 = F_LATIN1
		1640	utf8 = F_UTF8
		1641	indent = F_INDENT
		1642	canonical = F_CANONICAL
		1643	space_before = F_SPACE_BEFORE
		1644	space_after = F_SPACE_AFTER
		1645	pretty = F_PRETTY
		1646	allow_nonref = F_ALLOW_NONREF
		1647	shrink = F_SHRINK
		1648	allow_blessed = F_ALLOW_BLESSED
		1649	convert_blessed = F_CONV_BLESSED
		1650	relaxed = F_RELAXED
		1651	PPCODE:
		1652	{
		1653	if (enable)
		1654	self->flags |= ix;
		1655	else
		1656	self->flags &= ~ix;
		1657
		1658	XPUSHs (ST (0));
		1659	}
		1660
		1661	void get_ascii (JSON *self)
		1662	ALIAS:
		1663	get_ascii = F_ASCII
		1664	get_latin1 = F_LATIN1
		1665	get_utf8 = F_UTF8
		1666	get_indent = F_INDENT
		1667	get_canonical = F_CANONICAL
		1668	get_space_before = F_SPACE_BEFORE
		1669	get_space_after = F_SPACE_AFTER
		1670	get_allow_nonref = F_ALLOW_NONREF
		1671	get_shrink = F_SHRINK
		1672	get_allow_blessed = F_ALLOW_BLESSED
		1673	get_convert_blessed = F_CONV_BLESSED
		1674	get_relaxed = F_RELAXED
		1675	PPCODE:
		1676	XPUSHs (boolSV (self->flags & ix));
		1677
		1678	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1679	PPCODE:
		1680	{
		1681	UV log2 = 0;
		1682
		1683	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1684
		1685	while ((1UL << log2) < max_depth)
		1686	++log2;
		1687
		1688	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1689
		1690	XPUSHs (ST (0));
		1691	}
		1692
		1693	U32 get_max_depth (JSON *self)
		1694	CODE:
		1695	RETVAL = DEC_DEPTH (self->flags);
921	OUTPUT:	1696	OUTPUT:
922	RETVAL	1697	RETVAL
923		1698
924	SV *ascii (SV *self, int enable = 1)	1699	void max_size (JSON *self, UV max_size = 0)
925	ALIAS:	1700	PPCODE:
926	ascii = F_ASCII	1701	{
927	utf8 = F_UTF8	1702	UV log2 = 0;
928	indent = F_INDENT	1703
929	canonical = F_CANONICAL	1704	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
930	space_before = F_SPACE_BEFORE	1705	if (max_size == 1) max_size = 2;
931	space_after = F_SPACE_AFTER	1706
932	json_rpc = F_JSON_RPC	1707	while ((1UL << log2) < max_size)
933	pretty = F_PRETTY	1708	++log2;
934	allow_nonref = F_ALLOW_NONREF	1709
935	shrink = F_SHRINK	1710	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1711
		1712	XPUSHs (ST (0));
		1713	}
		1714
		1715	int get_max_size (JSON *self)
936	CODE:	1716	CODE:
937	{	1717	RETVAL = DEC_SIZE (self->flags);
938	UV *uv = SvJSON (self);
939	if (enable)
940	*uv |= ix;
941	else
942	*uv &= ~ix;
943
944	RETVAL = newSVsv (self);
945	}
946	OUTPUT:	1718	OUTPUT:
947	RETVAL	1719	RETVAL
948		1720
949	void encode (SV *self, SV *scalar)	1721	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
950	PPCODE:	1722	PPCODE:
951	XPUSHs (encode_json (scalar, *SvJSON (self)));	1723	{
		1724	SvREFCNT_dec (self->cb_object);
		1725	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
952		1726
953	void decode (SV *self, SV *jsonstr)	1727	XPUSHs (ST (0));
		1728	}
		1729
		1730	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
954	PPCODE:	1731	PPCODE:
		1732	{
		1733	if (!self->cb_sk_object)
		1734	self->cb_sk_object = newHV ();
		1735
		1736	if (SvOK (cb))
		1737	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1738	else
		1739	{
		1740	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1741
		1742	if (!HvKEYS (self->cb_sk_object))
		1743	{
		1744	SvREFCNT_dec (self->cb_sk_object);
		1745	self->cb_sk_object = 0;
		1746	}
		1747	}
		1748
		1749	XPUSHs (ST (0));
		1750	}
		1751
		1752	void encode (JSON *self, SV *scalar)
		1753	PPCODE:
		1754	XPUSHs (encode_json (scalar, self));
		1755
		1756	void decode (JSON *self, SV *jsonstr)
		1757	PPCODE:
955	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1758	XPUSHs (decode_json (jsonstr, self, 0));
		1759
		1760	void decode_prefix (JSON *self, SV *jsonstr)
		1761	PPCODE:
		1762	{
		1763	STRLEN offset;
		1764	EXTEND (SP, 2);
		1765	PUSHs (decode_json (jsonstr, self, &offset));
		1766	PUSHs (sv_2mortal (newSVuv (offset)));
		1767	}
		1768
		1769	void incr_parse (JSON *self, SV *jsonstr = 0)
		1770	PPCODE:
		1771	{
		1772	if (!self->incr_text)
		1773	self->incr_text = newSVpvn ("", 0);
		1774
		1775	// append data, if any
		1776	if (jsonstr)
		1777	{
		1778	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1779	{
		1780	/* utf-8-ness differs, need to upgrade */
		1781	sv_utf8_upgrade (self->incr_text);
		1782
		1783	if (self->incr_pos)
		1784	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1785	- (U8 *)SvPVX (self->incr_text);
		1786	}
		1787
		1788	{
		1789	STRLEN len;
		1790	const char *str = SvPV (jsonstr, len);
		1791	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1792	Move (str, SvEND (self->incr_text), len, char);
		1793	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1794	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1795	}
		1796	}
		1797
		1798	if (GIMME_V != G_VOID)
		1799	do
		1800	{
		1801	STRLEN offset;
		1802
		1803	incr_parse (self);
		1804
		1805	if (!INCR_DONE (self))
		1806	break;
		1807
		1808	XPUSHs (decode_json (self->incr_text, self, &offset));
		1809
		1810	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1811	self->incr_pos -= offset;
		1812	self->incr_nest = 0;
		1813	self->incr_mode = 0;
		1814	}
		1815	while (GIMME_V == G_ARRAY);
		1816	}
		1817
		1818	SV *incr_text (JSON *self)
		1819	ATTRS: lvalue
		1820	CODE:
		1821	{
		1822	if (self->incr_pos)
		1823	croak ("incr_text can only be called after a successful incr_parse call in scalar context");
		1824
		1825	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1826	}
		1827	OUTPUT:
		1828	RETVAL
		1829
		1830	void incr_skip (JSON *self)
		1831	CODE:
		1832	{
		1833	if (!self->incr_pos || !INCR_DONE (self))
		1834	croak ("incr_text can only be called after an unsuccessful incr_parse call in scalar context");//D
		1835
		1836	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1837	self->incr_pos = 0;
		1838	self->incr_nest = 0;
		1839	self->incr_mode = 0;
		1840	}
		1841
		1842	void DESTROY (JSON *self)
		1843	CODE:
		1844	SvREFCNT_dec (self->cb_sk_object);
		1845	SvREFCNT_dec (self->cb_object);
		1846	SvREFCNT_dec (self->incr_text);
956		1847
957	PROTOTYPES: ENABLE	1848	PROTOTYPES: ENABLE
958		1849
959	void to_json (SV *scalar)	1850	void encode_json (SV *scalar)
		1851	ALIAS:
		1852	to_json_ = 0
		1853	encode_json = F_UTF8
960	PPCODE:	1854	PPCODE:
		1855	{
		1856	JSON json = { F_DEFAULT | ix };
961	XPUSHs (encode_json (scalar, F_UTF8));	1857	XPUSHs (encode_json (scalar, &json));
		1858	}
962		1859
963	void from_json (SV *jsonstr)	1860	void decode_json (SV *jsonstr)
		1861	ALIAS:
		1862	from_json_ = 0
		1863	decode_json = F_UTF8
964	PPCODE:	1864	PPCODE:
		1865	{
		1866	JSON json = { F_DEFAULT | ix };
965	XPUSHs (decode_json (jsonstr, F_UTF8));	1867	XPUSHs (decode_json (jsonstr, &json, 0));
		1868	}
966		1869
		1870

Diff Legend

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