ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.23 by root, Tue Apr 3 23:34:17 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 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 0x00000001UL	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002UL	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004UL	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008UL	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010UL	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020UL	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080UL	31	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100UL	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
17	#define F_MAXDEPTH 0xf8000000UL	37	#define F_MAXDEPTH 0xf8000000UL
18	#define S_MAXDEPTH 27	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
19		42
20	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	43	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
21		44	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
22	// F_SELFCONVERT? <=> to_json/toJson
23	// F_BLESSED? <=> { $__class__$ => }
24		45
25	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	46	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
26	#define F_DEFAULT (12UL << S_MAXDEPTH)	47	#define F_DEFAULT (9UL << S_MAXDEPTH)
27		48
28	#define INIT_SIZE 32 // initial scalar size to be allocated	49	#define INIT_SIZE 32 // initial scalar size to be allocated
29	#define INDENT_STEP 3 // spaces per indentation level	50	#define INDENT_STEP 3 // spaces per indentation level
30		51
31	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
32	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	52	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
33		53
34	#define SB do {	54	#define SB do {
35	#define SE } while (0)	55	#define SE } while (0)
36		56
		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
37	static HV *json_stash; // JSON::XS::	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;
38		103
39	/////////////////////////////////////////////////////////////////////////////	104	/////////////////////////////////////////////////////////////////////////////
40	// utility functions	105	// utility functions
41		106
42	static UV *	107	INLINE SV *
43	SvJSON (SV *sv)	108	get_bool (const char *name)
44	{	109	{
45	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	110	SV *sv = get_sv (name, 1);
46	croak ("object is not of type JSON::XS");
47		111
48	return &SvUVX (SvRV (sv));	112	SvREADONLY_on (sv);
49	}	113	SvREADONLY_on (SvRV (sv));
50		114
51	static void	115	return sv;
		116	}
		117
		118	INLINE void
52	shrink (SV *sv)	119	shrink (SV *sv)
53	{	120	{
54	sv_utf8_downgrade (sv, 1);	121	sv_utf8_downgrade (sv, 1);
55	if (SvLEN (sv) > SvCUR (sv) + 1)	122	if (SvLEN (sv) > SvCUR (sv) + 1)
56	{	123	{
…		…
65	// decode an utf-8 character and return it, or (UV)-1 in	132	// decode an utf-8 character and return it, or (UV)-1 in
66	// case of an error.	133	// case of an error.
67	// we special-case "safe" characters from U+80 .. U+7FF,	134	// we special-case "safe" characters from U+80 .. U+7FF,
68	// but use the very good perl function to parse anything else.	135	// but use the very good perl function to parse anything else.
69	// note that we never call this function for a ascii codepoints	136	// note that we never call this function for a ascii codepoints
70	static UV	137	INLINE UV
71	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	138	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
72	{	139	{
73	if (s[0] > 0xdf || s[0] < 0xc2)	140	if (expect_true (len >= 2
74	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	141	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
75	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	142	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
76	{	143	{
77	*clen = 2;	144	*clen = 2;
78	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	145	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
79	}	146	}
80	else	147	else
81	{	148	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
82	*clen = (STRLEN)-1;	149	}
83	return (UV)-1;	150
84	}	151	// likewise for encoding, also never called for ascii codepoints
		152	// this function takes advantage of this fact, although current gccs
		153	// seem to optimise the check for >= 0x80 away anyways
		154	INLINE unsigned char *
		155	encode_utf8 (unsigned char *s, UV ch)
		156	{
		157	if (expect_false (ch < 0x000080))
		158	*s++ = ch;
		159	else if (expect_true (ch < 0x000800))
		160	*s++ = 0xc0 | ( ch >> 6),
		161	*s++ = 0x80 | ( ch & 0x3f);
		162	else if ( ch < 0x010000)
		163	*s++ = 0xe0 | ( ch >> 12),
		164	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		165	*s++ = 0x80 | ( ch & 0x3f);
		166	else if ( ch < 0x110000)
		167	*s++ = 0xf0 | ( ch >> 18),
		168	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		169	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 | ( ch & 0x3f);
		171
		172	return s;
85	}	173	}
86		174
87	/////////////////////////////////////////////////////////////////////////////	175	/////////////////////////////////////////////////////////////////////////////
88	// encoder	176	// encoder
89		177
…		…
91	typedef struct	179	typedef struct
92	{	180	{
93	char *cur; // SvPVX (sv) + current output position	181	char *cur; // SvPVX (sv) + current output position
94	char *end; // SvEND (sv)	182	char *end; // SvEND (sv)
95	SV *sv; // result scalar	183	SV *sv; // result scalar
96	U32 flags; // F_*	184	JSON json;
97	U32 indent; // indentation level	185	U32 indent; // indentation level
98	U32 maxdepth; // max. indentation/recursion level	186	U32 maxdepth; // max. indentation/recursion level
		187	UV limit; // escape character values >= this value when encoding
99	} enc_t;	188	} enc_t;
100		189
101	static void	190	INLINE void
102	need (enc_t *enc, STRLEN len)	191	need (enc_t *enc, STRLEN len)
103	{	192	{
104	if (enc->cur + len >= enc->end)	193	if (expect_false (enc->cur + len >= enc->end))
105	{	194	{
106	STRLEN cur = enc->cur - SvPVX (enc->sv);	195	STRLEN cur = enc->cur - SvPVX (enc->sv);
107	SvGROW (enc->sv, cur + len + 1);	196	SvGROW (enc->sv, cur + len + 1);
108	enc->cur = SvPVX (enc->sv) + cur;	197	enc->cur = SvPVX (enc->sv) + cur;
109	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	198	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
110	}	199	}
111	}	200	}
112		201
113	static void	202	INLINE void
114	encode_ch (enc_t *enc, char ch)	203	encode_ch (enc_t *enc, char ch)
115	{	204	{
116	need (enc, 1);	205	need (enc, 1);
117	*enc->cur++ = ch;	206	*enc->cur++ = ch;
118	}	207	}
…		…
126		215
127	while (str < end)	216	while (str < end)
128	{	217	{
129	unsigned char ch = *(unsigned char *)str;	218	unsigned char ch = *(unsigned char *)str;
130		219
131	if (ch >= 0x20 && ch < 0x80) // most common case	220	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
132	{	221	{
133	if (ch == '"') // but with slow exceptions	222	if (expect_false (ch == '"')) // but with slow exceptions
134	{	223	{
135	need (enc, len += 1);	224	need (enc, len += 1);
136	*enc->cur++ = '\\';	225	*enc->cur++ = '\\';
137	*enc->cur++ = '"';	226	*enc->cur++ = '"';
138	}	227	}
139	else if (ch == '\\')	228	else if (expect_false (ch == '\\'))
140	{	229	{
141	need (enc, len += 1);	230	need (enc, len += 1);
142	*enc->cur++ = '\\';	231	*enc->cur++ = '\\';
143	*enc->cur++ = '\\';	232	*enc->cur++ = '\\';
144	}	233	}
…		…
162	STRLEN clen;	251	STRLEN clen;
163	UV uch;	252	UV uch;
164		253
165	if (is_utf8)	254	if (is_utf8)
166	{	255	{
167	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
168	uch = decode_utf8 (str, end - str, &clen);	256	uch = decode_utf8 (str, end - str, &clen);
169	if (clen == (STRLEN)-1)	257	if (clen == (STRLEN)-1)
170	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	258	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
171	}	259	}
172	else	260	else
173	{	261	{
174	uch = ch;	262	uch = ch;
175	clen = 1;	263	clen = 1;
176	}	264	}
177		265
178	if (uch > 0x10FFFFUL)	266	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
179	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
180
181	if (uch < 0x80 || enc->flags & F_ASCII)
182	{	267	{
183	if (uch > 0xFFFFUL)	268	if (uch >= 0x10000UL)
184	{	269	{
		270	if (uch >= 0x110000UL)
		271	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		272
185	need (enc, len += 11);	273	need (enc, len += 11);
186	sprintf (enc->cur, "\\u%04x\\u%04x",	274	sprintf (enc->cur, "\\u%04x\\u%04x",
187	(int)((uch - 0x10000) / 0x400 + 0xD800),	275	(int)((uch - 0x10000) / 0x400 + 0xD800),
188	(int)((uch - 0x10000) % 0x400 + 0xDC00));	276	(int)((uch - 0x10000) % 0x400 + 0xDC00));
189	enc->cur += 12;	277	enc->cur += 12;
…		…
200	*enc->cur++ = hexdigit [(uch >> 0) & 15];	288	*enc->cur++ = hexdigit [(uch >> 0) & 15];
201	}	289	}
202		290
203	str += clen;	291	str += clen;
204	}	292	}
		293	else if (enc->json.flags & F_LATIN1)
		294	{
		295	*enc->cur++ = uch;
		296	str += clen;
		297	}
205	else if (is_utf8)	298	else if (is_utf8)
206	{	299	{
207	need (enc, len += clen);	300	need (enc, len += clen);
208	do	301	do
209	{	302	{
…		…
211	}	304	}
212	while (--clen);	305	while (--clen);
213	}	306	}
214	else	307	else
215	{	308	{
216	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	309	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
217	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	310	enc->cur = encode_utf8 (enc->cur, uch);
218	++str;	311	++str;
219	}	312	}
220	}	313	}
221	}	314	}
222	}	315	}
223		316
224	--len;	317	--len;
225	}	318	}
226	}	319	}
227		320
228	static void	321	INLINE void
229	encode_indent (enc_t *enc)	322	encode_indent (enc_t *enc)
230	{	323	{
231	if (enc->flags & F_INDENT)	324	if (enc->json.flags & F_INDENT)
232	{	325	{
233	int spaces = enc->indent * INDENT_STEP;	326	int spaces = enc->indent * INDENT_STEP;
234		327
235	need (enc, spaces);	328	need (enc, spaces);
236	memset (enc->cur, ' ', spaces);	329	memset (enc->cur, ' ', spaces);
237	enc->cur += spaces;	330	enc->cur += spaces;
238	}	331	}
239	}	332	}
240		333
241	static void	334	INLINE void
242	encode_space (enc_t *enc)	335	encode_space (enc_t *enc)
243	{	336	{
244	need (enc, 1);	337	need (enc, 1);
245	encode_ch (enc, ' ');	338	encode_ch (enc, ' ');
246	}	339	}
247		340
248	static void	341	INLINE void
249	encode_nl (enc_t *enc)	342	encode_nl (enc_t *enc)
250	{	343	{
251	if (enc->flags & F_INDENT)	344	if (enc->json.flags & F_INDENT)
252	{	345	{
253	need (enc, 1);	346	need (enc, 1);
254	encode_ch (enc, '\n');	347	encode_ch (enc, '\n');
255	}	348	}
256	}	349	}
257		350
258	static void	351	INLINE void
259	encode_comma (enc_t *enc)	352	encode_comma (enc_t *enc)
260	{	353	{
261	encode_ch (enc, ',');	354	encode_ch (enc, ',');
262		355
263	if (enc->flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
264	encode_nl (enc);	357	encode_nl (enc);
265	else if (enc->flags & F_SPACE_AFTER)	358	else if (enc->json.flags & F_SPACE_AFTER)
266	encode_space (enc);	359	encode_space (enc);
267	}	360	}
268		361
269	static void encode_sv (enc_t *enc, SV *sv);	362	static void encode_sv (enc_t *enc, SV *sv);
270		363
…		…
274	int i, len = av_len (av);	367	int i, len = av_len (av);
275		368
276	if (enc->indent >= enc->maxdepth)	369	if (enc->indent >= enc->maxdepth)
277	croak ("data structure too deep (hit recursion limit)");	370	croak ("data structure too deep (hit recursion limit)");
278		371
279	encode_ch (enc, '['); encode_nl (enc);	372	encode_ch (enc, '[');
280	++enc->indent;	373
		374	if (len >= 0)
		375	{
		376	encode_nl (enc); ++enc->indent;
281		377
282	for (i = 0; i <= len; ++i)	378	for (i = 0; i <= len; ++i)
283	{	379	{
		380	SV **svp = av_fetch (av, i, 0);
		381
284	encode_indent (enc);	382	encode_indent (enc);
285	encode_sv (enc, *av_fetch (av, i, 0));
286		383
		384	if (svp)
		385	encode_sv (enc, *svp);
		386	else
		387	encode_str (enc, "null", 4, 0);
		388
287	if (i < len)	389	if (i < len)
288	encode_comma (enc);	390	encode_comma (enc);
289	}	391	}
290		392
		393	encode_nl (enc); --enc->indent; encode_indent (enc);
		394	}
		395
291	encode_nl (enc);	396	encode_ch (enc, ']');
292
293	--enc->indent;
294	encode_indent (enc); encode_ch (enc, ']');
295	}	397	}
296		398
297	static void	399	static void
298	encode_he (enc_t *enc, HE *he)	400	encode_hk (enc_t *enc, HE *he)
299	{	401	{
300	encode_ch (enc, '"');	402	encode_ch (enc, '"');
301		403
302	if (HeKLEN (he) == HEf_SVKEY)	404	if (HeKLEN (he) == HEf_SVKEY)
303	{	405	{
…		…
313	else	415	else
314	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	416	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
315		417
316	encode_ch (enc, '"');	418	encode_ch (enc, '"');
317		419
318	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	420	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
319	encode_ch (enc, ':');	421	encode_ch (enc, ':');
320	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	422	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
321	encode_sv (enc, HeVAL (he));
322	}	423	}
323		424
324	// compare hash entries, used when all keys are bytestrings	425	// compare hash entries, used when all keys are bytestrings
325	static int	426	static int
326	he_cmp_fast (const void *a_, const void *b_)	427	he_cmp_fast (const void *a_, const void *b_)
…		…
331	HE *b = *(HE **)b_;	432	HE *b = *(HE **)b_;
332		433
333	STRLEN la = HeKLEN (a);	434	STRLEN la = HeKLEN (a);
334	STRLEN lb = HeKLEN (b);	435	STRLEN lb = HeKLEN (b);
335		436
336	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	437	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
337	cmp = la - lb;	438	cmp = lb - la;
338		439
339	return cmp;	440	return cmp;
340	}	441	}
341		442
342	// compare hash entries, used when some keys are sv's or utf-x	443	// compare hash entries, used when some keys are sv's or utf-x
343	static int	444	static int
344	he_cmp_slow (const void *a, const void *b)	445	he_cmp_slow (const void *a, const void *b)
345	{	446	{
346	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	447	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
347	}	448	}
348		449
349	static void	450	static void
350	encode_hv (enc_t *enc, HV *hv)	451	encode_hv (enc_t *enc, HV *hv)
351	{	452	{
352	int count, i;	453	HE *he;
353		454
354	if (enc->indent >= enc->maxdepth)	455	if (enc->indent >= enc->maxdepth)
355	croak ("data structure too deep (hit recursion limit)");	456	croak ("data structure too deep (hit recursion limit)");
356		457
357	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	458	encode_ch (enc, '{');
358		459
359	if ((count = hv_iterinit (hv)))
360	{
361	// for canonical output we have to sort by keys first	460	// for canonical output we have to sort by keys first
362	// actually, this is mostly due to the stupid so-called	461	// actually, this is mostly due to the stupid so-called
363	// security workaround added somewhere in 5.8.x.	462	// security workaround added somewhere in 5.8.x.
364	// that randomises hash orderings	463	// that randomises hash orderings
365	if (enc->flags & F_CANONICAL)	464	if (enc->json.flags & F_CANONICAL)
		465	{
		466	int count = hv_iterinit (hv);
		467
		468	if (SvMAGICAL (hv))
366	{	469	{
		470	// need to count by iterating. could improve by dynamically building the vector below
		471	// but I don't care for the speed of this special case.
		472	// note also that we will run into undefined behaviour when the two iterations
		473	// do not result in the same count, something I might care for in some later release.
		474
		475	count = 0;
		476	while (hv_iternext (hv))
		477	++count;
		478
		479	hv_iterinit (hv);
		480	}
		481
		482	if (count)
		483	{
		484	int i, fast = 1;
		485	#if defined(__BORLANDC__) || defined(_MSC_VER)
		486	HE **hes = _alloca (count * sizeof (HE));
		487	#else
367	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	488	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
368	int fast = 1;	489	#endif
369		490
370	i = 0;	491	i = 0;
371	while ((he = hv_iternext (hv)))	492	while ((he = hv_iternext (hv)))
372	{	493	{
373	hes [i++] = he;	494	hes [i++] = he;
…		…
395		516
396	FREETMPS;	517	FREETMPS;
397	LEAVE;	518	LEAVE;
398	}	519	}
399		520
400	for (i = 0; i < count; ++i)	521	encode_nl (enc); ++enc->indent;
		522
		523	while (count--)
401	{	524	{
402	encode_indent (enc);	525	encode_indent (enc);
		526	he = hes [count];
403	encode_he (enc, hes [i]);	527	encode_hk (enc, he);
		528	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
404		529
405	if (i < count - 1)	530	if (count)
406	encode_comma (enc);	531	encode_comma (enc);
407	}	532	}
408		533
		534	encode_nl (enc); --enc->indent; encode_indent (enc);
		535	}
		536	}
		537	else
		538	{
		539	if (hv_iterinit (hv) || SvMAGICAL (hv))
		540	if ((he = hv_iternext (hv)))
		541	{
		542	encode_nl (enc); ++enc->indent;
		543
		544	for (;;)
		545	{
409	encode_nl (enc);	546	encode_indent (enc);
		547	encode_hk (enc, he);
		548	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		549
		550	if (!(he = hv_iternext (hv)))
		551	break;
		552
		553	encode_comma (enc);
		554	}
		555
		556	encode_nl (enc); --enc->indent; encode_indent (enc);
		557	}
		558	}
		559
		560	encode_ch (enc, '}');
		561	}
		562
		563	// encode objects, arrays and special \0=false and \1=true values.
		564	static void
		565	encode_rv (enc_t *enc, SV *sv)
		566	{
		567	svtype svt;
		568
		569	SvGETMAGIC (sv);
		570	svt = SvTYPE (sv);
		571
		572	if (expect_false (SvOBJECT (sv)))
		573	{
		574	HV *stash = !JSON_SLOW || json_boolean_stash
		575	? json_boolean_stash
		576	: gv_stashpv ("JSON::XS::Boolean", 1);
		577
		578	if (SvSTASH (sv) == stash)
		579	{
		580	if (SvIV (sv))
		581	encode_str (enc, "true", 4, 0);
		582	else
		583	encode_str (enc, "false", 5, 0);
410	}	584	}
411	else	585	else
412	{	586	{
413	SV *sv;	587	#if 0
414	HE *he = hv_iternext (hv);	588	if (0 && sv_derived_from (rv, "JSON::Literal"))
415
416	for (;;)
417	{	589	{
418	encode_indent (enc);	590	// not yet
419	encode_he (enc, he);
420
421	if (!(he = hv_iternext (hv)))
422	break;
423
424	encode_comma (enc);
425	}	591	}
		592	#endif
		593	if (enc->json.flags & F_CONV_BLESSED)
		594	{
		595	// we re-bless the reference to get overload and other niceties right
		596	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
426		597
		598	if (to_json)
		599	{
		600	dSP;
		601
		602	ENTER; SAVETMPS; PUSHMARK (SP);
		603	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		604
		605	// calling with G_SCALAR ensures that we always get a 1 return value
		606	PUTBACK;
		607	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		608	SPAGAIN;
		609
		610	// catch this surprisingly common error
		611	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		612	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		613
		614	sv = POPs;
		615	PUTBACK;
		616
427	encode_nl (enc);	617	encode_sv (enc, sv);
		618
		619	FREETMPS; LEAVE;
		620	}
		621	else if (enc->json.flags & F_ALLOW_BLESSED)
		622	encode_str (enc, "null", 4, 0);
		623	else
		624	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		625	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		626	}
		627	else if (enc->json.flags & F_ALLOW_BLESSED)
		628	encode_str (enc, "null", 4, 0);
		629	else
		630	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		631	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
428	}	632	}
429	}	633	}
430
431	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
432	}
433
434	// encode objects, arrays and special \0=false and \1=true values.
435	static void
436	encode_rv (enc_t *enc, SV *sv)
437	{
438	SvGETMAGIC (sv);
439
440	svtype svt = SvTYPE (sv);
441
442	if (svt == SVt_PVHV)	634	else if (svt == SVt_PVHV)
443	encode_hv (enc, (HV *)sv);	635	encode_hv (enc, (HV *)sv);
444	else if (svt == SVt_PVAV)	636	else if (svt == SVt_PVAV)
445	encode_av (enc, (AV *)sv);	637	encode_av (enc, (AV *)sv);
446	else if (svt < SVt_PVAV)	638	else if (svt < SVt_PVAV)
447	{	639	{
448	if (SvNIOK (sv) && SvIV (sv) == 0)	640	STRLEN len = 0;
		641	char *pv = svt ? SvPV (sv, len) : 0;
		642
		643	if (len == 1 && *pv == '1')
		644	encode_str (enc, "true", 4, 0);
		645	else if (len == 1 && *pv == '0')
449	encode_str (enc, "false", 5, 0);	646	encode_str (enc, "false", 5, 0);
450	else if (SvNIOK (sv) && SvIV (sv) == 1)
451	encode_str (enc, "true", 4, 0);
452	else	647	else
453	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	648	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
454	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	649	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
455	}	650	}
456	else	651	else
…		…
471	encode_str (enc, str, len, SvUTF8 (sv));	666	encode_str (enc, str, len, SvUTF8 (sv));
472	encode_ch (enc, '"');	667	encode_ch (enc, '"');
473	}	668	}
474	else if (SvNOKp (sv))	669	else if (SvNOKp (sv))
475	{	670	{
		671	// trust that perl will do the right thing w.r.t. JSON syntax.
476	need (enc, NV_DIG + 32);	672	need (enc, NV_DIG + 32);
477	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	673	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
478	enc->cur += strlen (enc->cur);	674	enc->cur += strlen (enc->cur);
479	}	675	}
480	else if (SvIOKp (sv))	676	else if (SvIOKp (sv))
481	{	677	{
		678	// we assume we can always read an IV as a UV and vice versa
		679	// we assume two's complement
		680	// we assume no aliasing issues in the union
		681	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		682	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		683	{
		684	// optimise the "small number case"
		685	// code will likely be branchless and use only a single multiplication
		686	// works for numbers up to 59074
		687	I32 i = SvIVX (sv);
		688	U32 u;
		689	char digit, nz = 0;
		690
482	need (enc, 64);	691	need (enc, 6);
		692
		693	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		694	u = i < 0 ? -i : i;
		695
		696	// convert to 4.28 fixed-point representation
		697	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		698
		699	// now output digit by digit, each time masking out the integer part
		700	// and multiplying by 5 while moving the decimal point one to the right,
		701	// resulting in a net multiplication by 10.
		702	// we always write the digit to memory but conditionally increment
		703	// the pointer, to enable the use of conditional move instructions.
		704	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		705	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		706	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		707	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		708	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		709	}
		710	else
		711	{
		712	// large integer, use the (rather slow) snprintf way.
		713	need (enc, IVUV_MAXCHARS);
483	enc->cur +=	714	enc->cur +=
484	SvIsUV(sv)	715	SvIsUV(sv)
485	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	716	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
486	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	717	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		718	}
487	}	719	}
488	else if (SvROK (sv))	720	else if (SvROK (sv))
489	encode_rv (enc, SvRV (sv));	721	encode_rv (enc, SvRV (sv));
490	else if (!SvOK (sv))	722	else if (!SvOK (sv))
491	encode_str (enc, "null", 4, 0);	723	encode_str (enc, "null", 4, 0);
…		…
493	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	725	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
494	SvPV_nolen (sv), SvFLAGS (sv));	726	SvPV_nolen (sv), SvFLAGS (sv));
495	}	727	}
496		728
497	static SV *	729	static SV *
498	encode_json (SV *scalar, U32 flags)	730	encode_json (SV *scalar, JSON *json)
499	{	731	{
		732	enc_t enc;
		733
500	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	734	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
501	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	735	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
502		736
503	enc_t enc;	737	enc.json = *json;
504	enc.flags = flags;
505	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	738	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
506	enc.cur = SvPVX (enc.sv);	739	enc.cur = SvPVX (enc.sv);
507	enc.end = SvEND (enc.sv);	740	enc.end = SvEND (enc.sv);
508	enc.indent = 0;	741	enc.indent = 0;
509	enc.maxdepth = DEC_DEPTH (flags);	742	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		743	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		744	: enc.json.flags & F_LATIN1 ? 0x000100UL
		745	: 0x110000UL;
510		746
511	SvPOK_only (enc.sv);	747	SvPOK_only (enc.sv);
512	encode_sv (&enc, scalar);	748	encode_sv (&enc, scalar);
513		749
		750	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		751	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		752
514	if (!(flags & (F_ASCII | F_UTF8)))	753	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
515	SvUTF8_on (enc.sv);	754	SvUTF8_on (enc.sv);
516		755
517	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
518
519	if (enc.flags & F_SHRINK)	756	if (enc.json.flags & F_SHRINK)
520	shrink (enc.sv);	757	shrink (enc.sv);
521		758
522	return enc.sv;	759	return enc.sv;
523	}	760	}
524		761
…		…
529	typedef struct	766	typedef struct
530	{	767	{
531	char *cur; // current parser pointer	768	char *cur; // current parser pointer
532	char *end; // end of input string	769	char *end; // end of input string
533	const char *err; // parse error, if != 0	770	const char *err; // parse error, if != 0
534	U32 flags; // F_*	771	JSON json;
535	U32 depth; // recursion depth	772	U32 depth; // recursion depth
536	U32 maxdepth; // recursion depth limit	773	U32 maxdepth; // recursion depth limit
537	} dec_t;	774	} dec_t;
538		775
539	static void	776	INLINE void
		777	decode_comment (dec_t *dec)
		778	{
		779	// only '#'-style comments allowed a.t.m.
		780
		781	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		782	++dec->cur;
		783	}
		784
		785	INLINE void
540	decode_ws (dec_t *dec)	786	decode_ws (dec_t *dec)
541	{	787	{
542	for (;;)	788	for (;;)
543	{	789	{
544	char ch = *dec->cur;	790	char ch = *dec->cur;
545		791
546	if (ch > 0x20	792	if (ch > 0x20)
		793	{
		794	if (expect_false (ch == '#'))
		795	{
		796	if (dec->json.flags & F_RELAXED)
		797	decode_comment (dec);
		798	else
		799	break;
		800	}
		801	else
		802	break;
		803	}
547	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	804	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
548	break;	805	break; // parse error, but let higher level handle it, gives better error messages
549		806
550	++dec->cur;	807	++dec->cur;
551	}	808	}
552	}	809	}
553		810
…		…
570	decode_4hex (dec_t *dec)	827	decode_4hex (dec_t *dec)
571	{	828	{
572	signed char d1, d2, d3, d4;	829	signed char d1, d2, d3, d4;
573	unsigned char *cur = (unsigned char *)dec->cur;	830	unsigned char *cur = (unsigned char *)dec->cur;
574		831
575	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	832	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
576	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	833	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
577	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	834	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
578	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	835	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
579		836
580	dec->cur += 4;	837	dec->cur += 4;
581		838
582	return ((UV)d1) << 12	839	return ((UV)d1) << 12
583	| ((UV)d2) << 8	840	| ((UV)d2) << 8
…		…
591	static SV *	848	static SV *
592	decode_str (dec_t *dec)	849	decode_str (dec_t *dec)
593	{	850	{
594	SV *sv = 0;	851	SV *sv = 0;
595	int utf8 = 0;	852	int utf8 = 0;
		853	char *dec_cur = dec->cur;
596		854
597	do	855	do
598	{	856	{
599	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	857	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
600	char *cur = buf;	858	char *cur = buf;
601		859
602	do	860	do
603	{	861	{
604	unsigned char ch = *(unsigned char *)dec->cur++;	862	unsigned char ch = *(unsigned char *)dec_cur++;
605		863
606	if (ch == '"')	864	if (expect_false (ch == '"'))
607	{	865	{
608	--dec->cur;	866	--dec_cur;
609	break;	867	break;
610	}	868	}
611	else if (ch == '\\')	869	else if (expect_false (ch == '\\'))
612	{	870	{
613	switch (*dec->cur)	871	switch (*dec_cur)
614	{	872	{
615	case '\\':	873	case '\\':
616	case '/':	874	case '/':
617	case '"': *cur++ = *dec->cur++; break;	875	case '"': *cur++ = *dec_cur++; break;
618		876
619	case 'b': ++dec->cur; *cur++ = '\010'; break;	877	case 'b': ++dec_cur; *cur++ = '\010'; break;
620	case 't': ++dec->cur; *cur++ = '\011'; break;	878	case 't': ++dec_cur; *cur++ = '\011'; break;
621	case 'n': ++dec->cur; *cur++ = '\012'; break;	879	case 'n': ++dec_cur; *cur++ = '\012'; break;
622	case 'f': ++dec->cur; *cur++ = '\014'; break;	880	case 'f': ++dec_cur; *cur++ = '\014'; break;
623	case 'r': ++dec->cur; *cur++ = '\015'; break;	881	case 'r': ++dec_cur; *cur++ = '\015'; break;
624		882
625	case 'u':	883	case 'u':
626	{	884	{
627	UV lo, hi;	885	UV lo, hi;
628	++dec->cur;	886	++dec_cur;
629		887
		888	dec->cur = dec_cur;
630	hi = decode_4hex (dec);	889	hi = decode_4hex (dec);
		890	dec_cur = dec->cur;
631	if (hi == (UV)-1)	891	if (hi == (UV)-1)
632	goto fail;	892	goto fail;
633		893
634	// possibly a surrogate pair	894	// possibly a surrogate pair
635	if (hi >= 0xd800)	895	if (hi >= 0xd800)
636	if (hi < 0xdc00)	896	if (hi < 0xdc00)
637	{	897	{
638	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	898	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
639	ERR ("missing low surrogate character in surrogate pair");	899	ERR ("missing low surrogate character in surrogate pair");
640		900
641	dec->cur += 2;	901	dec_cur += 2;
642		902
		903	dec->cur = dec_cur;
643	lo = decode_4hex (dec);	904	lo = decode_4hex (dec);
		905	dec_cur = dec->cur;
644	if (lo == (UV)-1)	906	if (lo == (UV)-1)
645	goto fail;	907	goto fail;
646		908
647	if (lo < 0xdc00 || lo >= 0xe000)	909	if (lo < 0xdc00 || lo >= 0xe000)
648	ERR ("surrogate pair expected");	910	ERR ("surrogate pair expected");
…		…
654		916
655	if (hi >= 0x80)	917	if (hi >= 0x80)
656	{	918	{
657	utf8 = 1;	919	utf8 = 1;
658		920
659	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	921	cur = encode_utf8 (cur, hi);
660	}	922	}
661	else	923	else
662	*cur++ = hi;	924	*cur++ = hi;
663	}	925	}
664	break;	926	break;
665		927
666	default:	928	default:
667	--dec->cur;	929	--dec_cur;
668	ERR ("illegal backslash escape sequence in string");	930	ERR ("illegal backslash escape sequence in string");
669	}	931	}
670	}	932	}
671	else if (ch >= 0x20 && ch <= 0x7f)	933	else if (expect_true (ch >= 0x20 && ch < 0x80))
672	*cur++ = ch;	934	*cur++ = ch;
673	else if (ch >= 0x80)	935	else if (ch >= 0x80)
674	{	936	{
675	--dec->cur;
676
677	STRLEN clen;	937	STRLEN clen;
		938	UV uch;
		939
		940	--dec_cur;
		941
678	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	942	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
679	if (clen == (STRLEN)-1)	943	if (clen == (STRLEN)-1)
680	ERR ("malformed UTF-8 character in JSON string");	944	ERR ("malformed UTF-8 character in JSON string");
681		945
682	do	946	do
683	*cur++ = *dec->cur++;	947	*cur++ = *dec_cur++;
684	while (--clen);	948	while (--clen);
685		949
686	utf8 = 1;	950	utf8 = 1;
687	}	951	}
688	else	952	else
689	{	953	{
690	--dec->cur;	954	--dec_cur;
691		955
692	if (!ch)	956	if (!ch)
693	ERR ("unexpected end of string while parsing json string");	957	ERR ("unexpected end of string while parsing JSON string");
694	else	958	else
695	ERR ("invalid character encountered while parsing json string");	959	ERR ("invalid character encountered while parsing JSON string");
696	}	960	}
697	}	961	}
698	while (cur < buf + SHORT_STRING_LEN);	962	while (cur < buf + SHORT_STRING_LEN);
699		963
		964	{
700	STRLEN len = cur - buf;	965	STRLEN len = cur - buf;
701		966
702	if (sv)	967	if (sv)
703	{	968	{
704	SvGROW (sv, SvCUR (sv) + len + 1);	969	SvGROW (sv, SvCUR (sv) + len + 1);
705	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	970	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
706	SvCUR_set (sv, SvCUR (sv) + len);	971	SvCUR_set (sv, SvCUR (sv) + len);
707	}	972	}
708	else	973	else
709	sv = newSVpvn (buf, len);	974	sv = newSVpvn (buf, len);
710	}	975	}
		976	}
711	while (*dec->cur != '"');	977	while (*dec_cur != '"');
712		978
713	++dec->cur;	979	++dec_cur;
714		980
715	if (sv)	981	if (sv)
716	{	982	{
717	SvPOK_only (sv);	983	SvPOK_only (sv);
718	*SvEND (sv) = 0;	984	*SvEND (sv) = 0;
…		…
721	SvUTF8_on (sv);	987	SvUTF8_on (sv);
722	}	988	}
723	else	989	else
724	sv = newSVpvn ("", 0);	990	sv = newSVpvn ("", 0);
725		991
		992	dec->cur = dec_cur;
726	return sv;	993	return sv;
727		994
728	fail:	995	fail:
		996	dec->cur = dec_cur;
729	return 0;	997	return 0;
730	}	998	}
731		999
732	static SV *	1000	static SV *
733	decode_num (dec_t *dec)	1001	decode_num (dec_t *dec)
…		…
791	is_nv = 1;	1059	is_nv = 1;
792	}	1060	}
793		1061
794	if (!is_nv)	1062	if (!is_nv)
795	{	1063	{
796	UV uv;	1064	int len = dec->cur - start;
797	int numtype = grok_number (start, dec->cur - start, &uv);	1065
798	if (numtype & IS_NUMBER_IN_UV)	1066	// special case the rather common 1..5-digit-int case
799	if (numtype & IS_NUMBER_NEG)	1067	if (*start == '-')
		1068	switch (len)
800	{	1069	{
801	if (uv < (UV)IV_MIN)	1070	case 2: return newSViv (-( start [1] - '0' * 1));
802	return newSViv (-(IV)uv);	1071	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1072	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1073	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1074	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
803	}	1075	}
		1076	else
		1077	switch (len)
		1078	{
		1079	case 1: return newSViv ( start [0] - '0' * 1);
		1080	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1081	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1082	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1083	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1084	}
		1085
		1086	{
		1087	UV uv;
		1088	int numtype = grok_number (start, len, &uv);
		1089	if (numtype & IS_NUMBER_IN_UV)
		1090	if (numtype & IS_NUMBER_NEG)
		1091	{
		1092	if (uv < (UV)IV_MIN)
		1093	return newSViv (-(IV)uv);
		1094	}
804	else	1095	else
805	return newSVuv (uv);	1096	return newSVuv (uv);
806	}	1097	}
807		1098
		1099	len -= *start == '-' ? 1 : 0;
		1100
		1101	// does not fit into IV or UV, try NV
		1102	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1103	#if defined (LDBL_DIG)
		1104	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1105	#endif
		1106	)
		1107	// fits into NV without loss of precision
		1108	return newSVnv (Atof (start));
		1109
		1110	// everything else fails, convert it to a string
		1111	return newSVpvn (start, dec->cur - start);
		1112	}
		1113
		1114	// loss of precision here
808	return newSVnv (Atof (start));	1115	return newSVnv (Atof (start));
809		1116
810	fail:	1117	fail:
811	return 0;	1118	return 0;
812	}	1119	}
…		…
842		1149
843	if (*dec->cur != ',')	1150	if (*dec->cur != ',')
844	ERR (", or ] expected while parsing array");	1151	ERR (", or ] expected while parsing array");
845		1152
846	++dec->cur;	1153	++dec->cur;
		1154
		1155	decode_ws (dec);
		1156
		1157	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1158	{
		1159	++dec->cur;
		1160	break;
		1161	}
847	}	1162	}
848		1163
849	DEC_DEC_DEPTH;	1164	DEC_DEC_DEPTH;
850	return newRV_noinc ((SV *)av);	1165	return newRV_noinc ((SV *)av);
851		1166
…		…
856	}	1171	}
857		1172
858	static SV *	1173	static SV *
859	decode_hv (dec_t *dec)	1174	decode_hv (dec_t *dec)
860	{	1175	{
		1176	SV *sv;
861	HV *hv = newHV ();	1177	HV *hv = newHV ();
862		1178
863	DEC_INC_DEPTH;	1179	DEC_INC_DEPTH;
864	decode_ws (dec);	1180	decode_ws (dec);
865		1181
866	if (*dec->cur == '}')	1182	if (*dec->cur == '}')
867	++dec->cur;	1183	++dec->cur;
868	else	1184	else
869	for (;;)	1185	for (;;)
870	{	1186	{
		1187	EXPECT_CH ('"');
		1188
		1189	// heuristic: assume that
		1190	// a) decode_str + hv_store_ent are abysmally slow.
		1191	// b) most hash keys are short, simple ascii text.
		1192	// => try to "fast-match" such strings to avoid
		1193	// the overhead of decode_str + hv_store_ent.
		1194	{
871	SV *key, *value;	1195	SV *value;
		1196	char *p = dec->cur;
		1197	char *e = p + 24; // only try up to 24 bytes
872		1198
873	decode_ws (dec); EXPECT_CH ('"');	1199	for (;;)
874
875	key = decode_str (dec);
876	if (!key)
877	goto fail;
878
879	decode_ws (dec); EXPECT_CH (':');
880
881	value = decode_sv (dec);
882	if (!value)
883	{	1200	{
		1201	// the >= 0x80 is false on most architectures
		1202	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1203	{
		1204	// slow path, back up and use decode_str
		1205	SV *key = decode_str (dec);
		1206	if (!key)
		1207	goto fail;
		1208
		1209	decode_ws (dec); EXPECT_CH (':');
		1210
		1211	decode_ws (dec);
		1212	value = decode_sv (dec);
		1213	if (!value)
		1214	{
		1215	SvREFCNT_dec (key);
		1216	goto fail;
		1217	}
		1218
		1219	hv_store_ent (hv, key, value, 0);
884	SvREFCNT_dec (key);	1220	SvREFCNT_dec (key);
		1221
		1222	break;
		1223	}
		1224	else if (*p == '"')
		1225	{
		1226	// fast path, got a simple key
		1227	char *key = dec->cur;
		1228	int len = p - key;
		1229	dec->cur = p + 1;
		1230
		1231	decode_ws (dec); EXPECT_CH (':');
		1232
		1233	decode_ws (dec);
		1234	value = decode_sv (dec);
		1235	if (!value)
885	goto fail;	1236	goto fail;
		1237
		1238	hv_store (hv, key, len, value, 0);
		1239
		1240	break;
		1241	}
		1242
		1243	++p;
886	}	1244	}
887		1245	}
888	hv_store_ent (hv, key, value, 0);
889	SvREFCNT_dec (key);
890		1246
891	decode_ws (dec);	1247	decode_ws (dec);
892		1248
893	if (*dec->cur == '}')	1249	if (*dec->cur == '}')
894	{	1250	{
…		…
898		1254
899	if (*dec->cur != ',')	1255	if (*dec->cur != ',')
900	ERR (", or } expected while parsing object/hash");	1256	ERR (", or } expected while parsing object/hash");
901		1257
902	++dec->cur;	1258	++dec->cur;
		1259
		1260	decode_ws (dec);
		1261
		1262	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1263	{
		1264	++dec->cur;
		1265	break;
		1266	}
903	}	1267	}
904		1268
905	DEC_DEC_DEPTH;	1269	DEC_DEC_DEPTH;
906	return newRV_noinc ((SV *)hv);	1270	sv = newRV_noinc ((SV *)hv);
		1271
		1272	// check filter callbacks
		1273	if (dec->json.flags & F_HOOK)
		1274	{
		1275	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1276	{
		1277	HE *cb, *he;
		1278
		1279	hv_iterinit (hv);
		1280	he = hv_iternext (hv);
		1281	hv_iterinit (hv);
		1282
		1283	// the next line creates a mortal sv each time its called.
		1284	// might want to optimise this for common cases.
		1285	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1286
		1287	if (cb)
		1288	{
		1289	dSP;
		1290	int count;
		1291
		1292	ENTER; SAVETMPS; PUSHMARK (SP);
		1293	XPUSHs (HeVAL (he));
		1294
		1295	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1296
		1297	if (count == 1)
		1298	{
		1299	sv = newSVsv (POPs);
		1300	FREETMPS; LEAVE;
		1301	return sv;
		1302	}
		1303
		1304	FREETMPS; LEAVE;
		1305	}
		1306	}
		1307
		1308	if (dec->json.cb_object)
		1309	{
		1310	dSP;
		1311	int count;
		1312
		1313	ENTER; SAVETMPS; PUSHMARK (SP);
		1314	XPUSHs (sv_2mortal (sv));
		1315
		1316	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1317
		1318	if (count == 1)
		1319	{
		1320	sv = newSVsv (POPs);
		1321	FREETMPS; LEAVE;
		1322	return sv;
		1323	}
		1324
		1325	SvREFCNT_inc (sv);
		1326	FREETMPS; LEAVE;
		1327	}
		1328	}
		1329
		1330	return sv;
907		1331
908	fail:	1332	fail:
909	SvREFCNT_dec (hv);	1333	SvREFCNT_dec (hv);
910	DEC_DEC_DEPTH;	1334	DEC_DEC_DEPTH;
911	return 0;	1335	return 0;
912	}	1336	}
913		1337
914	static SV *	1338	static SV *
915	decode_sv (dec_t *dec)	1339	decode_sv (dec_t *dec)
916	{	1340	{
917	decode_ws (dec);	1341	// the beauty of JSON: you need exactly one character lookahead
		1342	// to parse everything.
918	switch (*dec->cur)	1343	switch (*dec->cur)
919	{	1344	{
920	case '"': ++dec->cur; return decode_str (dec);	1345	case '"': ++dec->cur; return decode_str (dec);
921	case '[': ++dec->cur; return decode_av (dec);	1346	case '[': ++dec->cur; return decode_av (dec);
922	case '{': ++dec->cur; return decode_hv (dec);	1347	case '{': ++dec->cur; return decode_hv (dec);
923		1348
924	case '-':	1349	case '-':
925	case '0': case '1': case '2': case '3': case '4':	1350	case '0': case '1': case '2': case '3': case '4':
926	case '5': case '6': case '7': case '8': case '9':	1351	case '5': case '6': case '7': case '8': case '9':
927	return decode_num (dec);	1352	return decode_num (dec);
928		1353
929	case 't':	1354	case 't':
930	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1355	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
931	{	1356	{
932	dec->cur += 4;	1357	dec->cur += 4;
		1358	#if JSON_SLOW
		1359	json_true = get_bool ("JSON::XS::true");
		1360	#endif
933	return newSViv (1);	1361	return newSVsv (json_true);
934	}	1362	}
935	else	1363	else
936	ERR ("'true' expected");	1364	ERR ("'true' expected");
937		1365
938	break;	1366	break;
939		1367
940	case 'f':	1368	case 'f':
941	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1369	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
942	{	1370	{
943	dec->cur += 5;	1371	dec->cur += 5;
		1372	#if JSON_SLOW
		1373	json_false = get_bool ("JSON::XS::false");
		1374	#endif
944	return newSViv (0);	1375	return newSVsv (json_false);
945	}	1376	}
946	else	1377	else
947	ERR ("'false' expected");	1378	ERR ("'false' expected");
948		1379
949	break;	1380	break;
…		…
958	ERR ("'null' expected");	1389	ERR ("'null' expected");
959		1390
960	break;	1391	break;
961		1392
962	default:	1393	default:
963	ERR ("malformed json string, neither array, object, number, string or atom");	1394	ERR ("malformed JSON string, neither array, object, number, string or atom");
964	break;	1395	break;
965	}	1396	}
966		1397
967	fail:	1398	fail:
968	return 0;	1399	return 0;
969	}	1400	}
970		1401
971	static SV *	1402	static SV *
972	decode_json (SV *string, U32 flags)	1403	decode_json (SV *string, JSON *json, STRLEN *offset_return)
973	{	1404	{
		1405	dec_t dec;
		1406	STRLEN offset;
974	SV *sv;	1407	SV *sv;
975		1408
		1409	SvGETMAGIC (string);
976	SvUPGRADE (string, SVt_PV);	1410	SvUPGRADE (string, SVt_PV);
977		1411
		1412	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1413	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1414	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1415
978	if (flags & F_UTF8)	1416	if (json->flags & F_UTF8)
979	sv_utf8_downgrade (string, 0);	1417	sv_utf8_downgrade (string, 0);
980	else	1418	else
981	sv_utf8_upgrade (string);	1419	sv_utf8_upgrade (string);
982		1420
983	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1421	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
984		1422
985	dec_t dec;	1423	dec.json = *json;
986	dec.flags = flags;
987	dec.cur = SvPVX (string);	1424	dec.cur = SvPVX (string);
988	dec.end = SvEND (string);	1425	dec.end = SvEND (string);
989	dec.err = 0;	1426	dec.err = 0;
990	dec.depth = 0;	1427	dec.depth = 0;
991	dec.maxdepth = DEC_DEPTH (dec.flags);	1428	dec.maxdepth = DEC_DEPTH (dec.json.flags);
992		1429
		1430	if (dec.json.cb_object || dec.json.cb_sk_object)
		1431	dec.json.flags |= F_HOOK;
		1432
993	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1433	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1434
		1435	decode_ws (&dec);
994	sv = decode_sv (&dec);	1436	sv = decode_sv (&dec);
995		1437
		1438	if (!(offset_return || !sv))
		1439	{
		1440	// check for trailing garbage
		1441	decode_ws (&dec);
		1442
		1443	if (*dec.cur)
		1444	{
		1445	dec.err = "garbage after JSON object";
		1446	SvREFCNT_dec (sv);
		1447	sv = 0;
		1448	}
		1449	}
		1450
		1451	if (offset_return || !sv)
		1452	{
		1453	offset = dec.json.flags & F_UTF8
		1454	? dec.cur - SvPVX (string)
		1455	: utf8_distance (dec.cur, SvPVX (string));
		1456
		1457	if (offset_return)
		1458	*offset_return = offset;
		1459	}
		1460
996	if (!sv)	1461	if (!sv)
997	{	1462	{
998	IV offset = dec.flags & F_UTF8
999	? dec.cur - SvPVX (string)
1000	: utf8_distance (dec.cur, SvPVX (string));
1001	SV *uni = sv_newmortal ();	1463	SV *uni = sv_newmortal ();
1002		1464
1003	// horrible hack to silence warning inside pv_uni_display	1465	// horrible hack to silence warning inside pv_uni_display
1004	COP cop = *PL_curcop;	1466	COP cop = *PL_curcop;
1005	cop.cop_warnings = pWARN_NONE;	1467	cop.cop_warnings = pWARN_NONE;
…		…
1015	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1477	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1016	}	1478	}
1017		1479
1018	sv = sv_2mortal (sv);	1480	sv = sv_2mortal (sv);
1019		1481
1020	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1482	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1021	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1483	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1022		1484
1023	return sv;	1485	return sv;
		1486	}
		1487
		1488	/////////////////////////////////////////////////////////////////////////////
		1489	// incremental parser
		1490
		1491	static void
		1492	incr_parse (JSON *self)
		1493	{
		1494	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1495
		1496	for (;;)
		1497	{
		1498	//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
		1499	switch (self->incr_mode)
		1500	{
		1501	// only used for intiial whitespace skipping
		1502	case INCR_M_WS:
		1503	for (;;)
		1504	{
		1505	if (*p > 0x20)
		1506	{
		1507	self->incr_mode = INCR_M_JSON;
		1508	goto incr_m_json;
		1509	}
		1510	else if (!*p)
		1511	goto interrupt;
		1512
		1513	++p;
		1514	}
		1515
		1516	// skip a single char inside a string (for \\-processing)
		1517	case INCR_M_BS:
		1518	if (!*p)
		1519	goto interrupt;
		1520
		1521	++p;
		1522	self->incr_mode = INCR_M_STR;
		1523	goto incr_m_str;
		1524
		1525	// inside a string
		1526	case INCR_M_STR:
		1527	incr_m_str:
		1528	for (;;)
		1529	{
		1530	if (*p == '"')
		1531	{
		1532	++p;
		1533	self->incr_mode = INCR_M_JSON;
		1534
		1535	if (!self->incr_nest)
		1536	goto interrupt;
		1537
		1538	goto incr_m_json;
		1539	}
		1540	else if (*p == '\\')
		1541	{
		1542	++p; // "virtually" consumes character after \
		1543
		1544	if (!*p) // if at end of string we have to switch modes
		1545	{
		1546	self->incr_mode = INCR_M_BS;
		1547	goto interrupt;
		1548	}
		1549	}
		1550	else if (!*p)
		1551	goto interrupt;
		1552
		1553	++p;
		1554	}
		1555
		1556	// after initial ws, outside string
		1557	case INCR_M_JSON:
		1558	incr_m_json:
		1559	for (;;)
		1560	{
		1561	switch (*p++)
		1562	{
		1563	case 0:
		1564	--p;
		1565	goto interrupt;
		1566
		1567	case 0x09:
		1568	case 0x0a:
		1569	case 0x0d:
		1570	case 0x20:
		1571	if (!self->incr_nest)
		1572	{
		1573	--p; // do not eat the whitespace, let the next round do it
		1574	goto interrupt;
		1575	}
		1576	break;
		1577
		1578	case '"':
		1579	self->incr_mode = INCR_M_STR;
		1580	goto incr_m_str;
		1581
		1582	case '[':
		1583	case '{':
		1584	++self->incr_nest;
		1585	break;
		1586
		1587	case ']':
		1588	case '}':
		1589	if (!--self->incr_nest)
		1590	goto interrupt;
		1591	}
		1592	}
		1593	}
		1594
		1595	modechange:
		1596	;
		1597	}
		1598
		1599	interrupt:
		1600	self->incr_pos = p - SvPVX (self->incr_text);
		1601	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1024	}	1602	}
1025		1603
1026	/////////////////////////////////////////////////////////////////////////////	1604	/////////////////////////////////////////////////////////////////////////////
1027	// XS interface functions	1605	// XS interface functions
1028		1606
1029	MODULE = JSON::XS PACKAGE = JSON::XS	1607	MODULE = JSON::XS PACKAGE = JSON::XS
1030		1608
1031	BOOT:	1609	BOOT:
1032	{	1610	{
1033	int i;	1611	int i;
1034
1035	memset (decode_hexdigit, 0xff, 256);
1036		1612
1037	for (i = 0; i < 256; ++i)	1613	for (i = 0; i < 256; ++i)
1038	decode_hexdigit [i] =	1614	decode_hexdigit [i] =
1039	i >= '0' && i <= '9' ? i - '0'	1615	i >= '0' && i <= '9' ? i - '0'
1040	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1616	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1041	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1617	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1042	: -1;	1618	: -1;
1043		1619
1044	json_stash = gv_stashpv ("JSON::XS", 1);	1620	json_stash = gv_stashpv ("JSON::XS" , 1);
		1621	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1622
		1623	json_true = get_bool ("JSON::XS::true");
		1624	json_false = get_bool ("JSON::XS::false");
1045	}	1625	}
1046		1626
1047	PROTOTYPES: DISABLE	1627	PROTOTYPES: DISABLE
1048		1628
1049	SV *new (char *dummy)	1629	void CLONE (...)
1050	CODE:	1630	CODE:
1051	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1631	json_stash = 0;
		1632	json_boolean_stash = 0;
		1633
		1634	void new (char *klass)
		1635	PPCODE:
		1636	{
		1637	SV *pv = NEWSV (0, sizeof (JSON));
		1638	SvPOK_only (pv);
		1639	Zero (SvPVX (pv), 1, JSON);
		1640	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1641	XPUSHs (sv_2mortal (sv_bless (
		1642	newRV_noinc (pv),
		1643	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1644	)));
		1645	}
		1646
		1647	void ascii (JSON *self, int enable = 1)
		1648	ALIAS:
		1649	ascii = F_ASCII
		1650	latin1 = F_LATIN1
		1651	utf8 = F_UTF8
		1652	indent = F_INDENT
		1653	canonical = F_CANONICAL
		1654	space_before = F_SPACE_BEFORE
		1655	space_after = F_SPACE_AFTER
		1656	pretty = F_PRETTY
		1657	allow_nonref = F_ALLOW_NONREF
		1658	shrink = F_SHRINK
		1659	allow_blessed = F_ALLOW_BLESSED
		1660	convert_blessed = F_CONV_BLESSED
		1661	relaxed = F_RELAXED
		1662	PPCODE:
		1663	{
		1664	if (enable)
		1665	self->flags |= ix;
		1666	else
		1667	self->flags &= ~ix;
		1668
		1669	XPUSHs (ST (0));
		1670	}
		1671
		1672	void get_ascii (JSON *self)
		1673	ALIAS:
		1674	get_ascii = F_ASCII
		1675	get_latin1 = F_LATIN1
		1676	get_utf8 = F_UTF8
		1677	get_indent = F_INDENT
		1678	get_canonical = F_CANONICAL
		1679	get_space_before = F_SPACE_BEFORE
		1680	get_space_after = F_SPACE_AFTER
		1681	get_allow_nonref = F_ALLOW_NONREF
		1682	get_shrink = F_SHRINK
		1683	get_allow_blessed = F_ALLOW_BLESSED
		1684	get_convert_blessed = F_CONV_BLESSED
		1685	get_relaxed = F_RELAXED
		1686	PPCODE:
		1687	XPUSHs (boolSV (self->flags & ix));
		1688
		1689	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1690	PPCODE:
		1691	{
		1692	UV log2 = 0;
		1693
		1694	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1695
		1696	while ((1UL << log2) < max_depth)
		1697	++log2;
		1698
		1699	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1700
		1701	XPUSHs (ST (0));
		1702	}
		1703
		1704	U32 get_max_depth (JSON *self)
		1705	CODE:
		1706	RETVAL = DEC_DEPTH (self->flags);
1052	OUTPUT:	1707	OUTPUT:
1053	RETVAL	1708	RETVAL
1054		1709
1055	SV *ascii (SV *self, int enable = 1)	1710	void max_size (JSON *self, UV max_size = 0)
1056	ALIAS:	1711	PPCODE:
1057	ascii = F_ASCII	1712	{
1058	utf8 = F_UTF8	1713	UV log2 = 0;
1059	indent = F_INDENT	1714
1060	canonical = F_CANONICAL	1715	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1061	space_before = F_SPACE_BEFORE	1716	if (max_size == 1) max_size = 2;
1062	space_after = F_SPACE_AFTER	1717
1063	pretty = F_PRETTY	1718	while ((1UL << log2) < max_size)
1064	allow_nonref = F_ALLOW_NONREF	1719	++log2;
1065	shrink = F_SHRINK	1720
		1721	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1722
		1723	XPUSHs (ST (0));
		1724	}
		1725
		1726	int get_max_size (JSON *self)
1066	CODE:	1727	CODE:
1067	{	1728	RETVAL = DEC_SIZE (self->flags);
1068	UV *uv = SvJSON (self);
1069	if (enable)
1070	*uv |= ix;
1071	else
1072	*uv &= ~ix;
1073
1074	RETVAL = newSVsv (self);
1075	}
1076	OUTPUT:	1729	OUTPUT:
1077	RETVAL	1730	RETVAL
1078		1731
1079	SV *max_depth (SV *self, int max_depth = 0x80000000UL)	1732	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1733	PPCODE:
		1734	{
		1735	SvREFCNT_dec (self->cb_object);
		1736	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1737
		1738	XPUSHs (ST (0));
		1739	}
		1740
		1741	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1742	PPCODE:
		1743	{
		1744	if (!self->cb_sk_object)
		1745	self->cb_sk_object = newHV ();
		1746
		1747	if (SvOK (cb))
		1748	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1749	else
		1750	{
		1751	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1752
		1753	if (!HvKEYS (self->cb_sk_object))
		1754	{
		1755	SvREFCNT_dec (self->cb_sk_object);
		1756	self->cb_sk_object = 0;
		1757	}
		1758	}
		1759
		1760	XPUSHs (ST (0));
		1761	}
		1762
		1763	void encode (JSON *self, SV *scalar)
		1764	PPCODE:
		1765	XPUSHs (encode_json (scalar, self));
		1766
		1767	void decode (JSON *self, SV *jsonstr)
		1768	PPCODE:
		1769	XPUSHs (decode_json (jsonstr, self, 0));
		1770
		1771	void decode_prefix (JSON *self, SV *jsonstr)
		1772	PPCODE:
		1773	{
		1774	STRLEN offset;
		1775	EXTEND (SP, 2);
		1776	PUSHs (decode_json (jsonstr, self, &offset));
		1777	PUSHs (sv_2mortal (newSVuv (offset)));
		1778	}
		1779
		1780	void incr_parse (JSON *self, SV *jsonstr = 0)
		1781	PPCODE:
		1782	{
		1783	if (!self->incr_text)
		1784	self->incr_text = newSVpvn ("", 0);
		1785
		1786	// append data, if any
		1787	if (jsonstr)
		1788	{
		1789	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1790	{
		1791	/* utf-8-ness differs, need to upgrade */
		1792	sv_utf8_upgrade (self->incr_text);
		1793
		1794	if (self->incr_pos)
		1795	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1796	- (U8 *)SvPVX (self->incr_text);
		1797	}
		1798
		1799	{
		1800	STRLEN len;
		1801	const char *str = SvPV (jsonstr, len);
		1802	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1803	Move (str, SvEND (self->incr_text), len, char);
		1804	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1805	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1806	}
		1807	}
		1808
		1809	if (GIMME_V != G_VOID)
		1810	do
		1811	{
		1812	STRLEN offset;
		1813
		1814	if (!INCR_DONE (self))
		1815	{
		1816	incr_parse (self);
		1817	if (!INCR_DONE (self))
		1818	break;
		1819	}
		1820
		1821	XPUSHs (decode_json (self->incr_text, self, &offset));
		1822
		1823	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1824	self->incr_pos -= offset;
		1825	self->incr_nest = 0;
		1826	self->incr_mode = 0;
		1827	}
		1828	while (GIMME_V == G_ARRAY);
		1829	}
		1830
		1831	SV *incr_text (JSON *self)
		1832	ATTRS: lvalue
1080	CODE:	1833	CODE:
1081	{	1834	{
1082	UV *uv = SvJSON (self);	1835	if (self->incr_pos)
1083	UV log2 = 0;	1836	croak ("incr_text can not be called when the incremental parser already started parsing");
1084		1837
1085	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1838	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1086
1087	while ((1UL << log2) < max_depth)
1088	++log2;
1089
1090	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1091
1092	RETVAL = newSVsv (self);
1093	}	1839	}
1094	OUTPUT:	1840	OUTPUT:
1095	RETVAL	1841	RETVAL
1096		1842
		1843	void incr_skip (JSON *self)
		1844	CODE:
		1845	{
		1846	if (self->incr_pos)
		1847	{
		1848	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1849	self->incr_pos = 0;
		1850	self->incr_nest = 0;
		1851	self->incr_mode = 0;
		1852	}
		1853	}
		1854
		1855	void DESTROY (JSON *self)
		1856	CODE:
		1857	SvREFCNT_dec (self->cb_sk_object);
		1858	SvREFCNT_dec (self->cb_object);
		1859	SvREFCNT_dec (self->incr_text);
		1860
		1861	PROTOTYPES: ENABLE
		1862
1097	void encode (SV *self, SV *scalar)	1863	void encode_json (SV *scalar)
		1864	ALIAS:
		1865	to_json_ = 0
		1866	encode_json = F_UTF8
1098	PPCODE:	1867	PPCODE:
		1868	{
		1869	JSON json = { F_DEFAULT | ix };
1099	XPUSHs (encode_json (scalar, *SvJSON (self)));	1870	XPUSHs (encode_json (scalar, &json));
		1871	}
1100		1872
1101	void decode (SV *self, SV *jsonstr)	1873	void decode_json (SV *jsonstr)
		1874	ALIAS:
		1875	from_json_ = 0
		1876	decode_json = F_UTF8
1102	PPCODE:	1877	PPCODE:
		1878	{
		1879	JSON json = { F_DEFAULT | ix };
1103	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1880	XPUSHs (decode_json (jsonstr, &json, 0));
		1881	}
1104		1882
1105	PROTOTYPES: ENABLE
1106		1883
1107	void to_json (SV *scalar)
1108	ALIAS:
1109	objToJson = 0
1110	PPCODE:
1111	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1112
1113	void from_json (SV *jsonstr)
1114	ALIAS:
1115	jsonToObj = 0
1116	PPCODE:
1117	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1118

Diff Legend

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