ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.21 by root, Thu Mar 29 02:45:49 2007 UTC vs.
Revision 1.101 by root, Tue Sep 8 18:00:03 2009 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	// guarantees, 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
17	#define F_MAXDEPTH 0xf8000000UL	35	#define F_RELAXED 0x00001000UL
18	#define S_MAXDEPTH 27	36	#define F_ALLOW_UNKNOWN 0x00002000UL
19		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
20	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
21
22	// F_SELFCONVERT? <=> to_json/toJson
23	// F_BLESSED? <=> { $__class__$ => }
24		38
25	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
26	#define F_DEFAULT (12UL << S_MAXDEPTH)
27		40
28	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
29	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
30		43
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	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
33		45
34	#define SB do {	46	#define SB do {
35	#define SE } while (0)	47	#define SE } while (0)
36		48
		49	#if __GNUC__ >= 3
		50	# define expect(expr,value) __builtin_expect ((expr), (value))
		51	# define INLINE static inline
		52	#else
		53	# define expect(expr,value) (expr)
		54	# define INLINE static
		55	#endif
		56
		57	#define expect_false(expr) expect ((expr) != 0, 0)
		58	#define expect_true(expr) expect ((expr) != 0, 1)
		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
37	static HV *json_stash; // JSON::XS::	74	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		75	static SV *json_true, *json_false;
		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		85
		86	typedef struct {
		87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
		91	SV *cb_object;
		92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	int incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
		99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
38		107
39	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
40	// utility functions	109	// utility functions
41		110
42	static UV *	111	INLINE SV *
43	SvJSON (SV *sv)	112	get_bool (const char *name)
44	{	113	{
45	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	114	SV *sv = get_sv (name, 1);
46	croak ("object is not of type JSON::XS");
47		115
48	return &SvUVX (SvRV (sv));	116	SvREADONLY_on (sv);
49	}	117	SvREADONLY_on (SvRV (sv));
50		118
51	static void	119	return sv;
		120	}
		121
		122	INLINE void
52	shrink (SV *sv)	123	shrink (SV *sv)
53	{	124	{
54	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
55	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
56	{	128	{
57	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
58	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
59	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
65	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
66	// case of an error.	138	// case of an error.
67	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
68	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
69	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
70	static UV	142	INLINE UV
71	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	143	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
72	{	144	{
73	if (s[0] > 0xdf || s[0] < 0xc2)	145	if (expect_true (len >= 2
74	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
75	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
76	{	148	{
77	*clen = 2;	149	*clen = 2;
78	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
79	}	151	}
80	else	152	else
81	return (UV)-1;	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		154	}
		155
		156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 | ( ch >> 6),
		166	*s++ = 0x80 | ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 | ( ch >> 12),
		169	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 | ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 | ( ch >> 18),
		173	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176
		177	return s;
		178	}
		179
		180	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV *sv, char *offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
82	}	187	}
83		188
84	/////////////////////////////////////////////////////////////////////////////	189	/////////////////////////////////////////////////////////////////////////////
85	// encoder	190	// encoder
86		191
…		…
88	typedef struct	193	typedef struct
89	{	194	{
90	char *cur; // SvPVX (sv) + current output position	195	char *cur; // SvPVX (sv) + current output position
91	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
92	SV *sv; // result scalar	197	SV *sv; // result scalar
93	U32 flags; // F_*	198	JSON json;
94	U32 indent; // indentation level	199	U32 indent; // indentation level
95	U32 maxdepth; // max. indentation/recursion level	200	UV limit; // escape character values >= this value when encoding
96	} enc_t;	201	} enc_t;
97		202
98	static void	203	INLINE void
99	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
100	{	205	{
101	if (enc->cur + len >= enc->end)	206	if (expect_false (enc->cur + len >= enc->end))
102	{	207	{
103	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
104	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
105	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
106	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
107	}	212	}
108	}	213	}
109		214
110	static void	215	INLINE void
111	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
112	{	217	{
113	need (enc, 1);	218	need (enc, 1);
114	*enc->cur++ = ch;	219	*enc->cur++ = ch;
115	}	220	}
…		…
123		228
124	while (str < end)	229	while (str < end)
125	{	230	{
126	unsigned char ch = *(unsigned char *)str;	231	unsigned char ch = *(unsigned char *)str;
127		232
128	if (ch >= 0x20 && ch < 0x80) // most common case	233	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
129	{	234	{
130	if (ch == '"') // but with slow exceptions	235	if (expect_false (ch == '"')) // but with slow exceptions
131	{	236	{
132	need (enc, len += 1);	237	need (enc, len += 1);
133	*enc->cur++ = '\\';	238	*enc->cur++ = '\\';
134	*enc->cur++ = '"';	239	*enc->cur++ = '"';
135	}	240	}
136	else if (ch == '\\')	241	else if (expect_false (ch == '\\'))
137	{	242	{
138	need (enc, len += 1);	243	need (enc, len += 1);
139	*enc->cur++ = '\\';	244	*enc->cur++ = '\\';
140	*enc->cur++ = '\\';	245	*enc->cur++ = '\\';
141	}	246	}
…		…
159	STRLEN clen;	264	STRLEN clen;
160	UV uch;	265	UV uch;
161		266
162	if (is_utf8)	267	if (is_utf8)
163	{	268	{
164	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
165	uch = decode_utf8 (str, end - str, &clen);	269	uch = decode_utf8 (str, end - str, &clen);
166	if (clen == (STRLEN)-1)	270	if (clen == (STRLEN)-1)
167	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	271	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
168	}	272	}
169	else	273	else
170	{	274	{
171	uch = ch;	275	uch = ch;
172	clen = 1;	276	clen = 1;
173	}	277	}
174		278
175	if (uch > 0x10FFFFUL)	279	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
176	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
177
178	if (uch < 0x80 || enc->flags & F_ASCII)
179	{	280	{
180	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
181	{	282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
182	need (enc, len += 11);	286	need (enc, len += 11);
183	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
184	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
185	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
186	enc->cur += 12;	290	enc->cur += 12;
187	}	291	}
188	else	292	else
189	{	293	{
190	static char hexdigit [16] = "0123456789abcdef";
191	need (enc, len += 5);	294	need (enc, len += 5);
192	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
193	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
194	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
195	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
196	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
197	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
198	}	301	}
199		302
		303	str += clen;
		304	}
		305	else if (enc->json.flags & F_LATIN1)
		306	{
		307	*enc->cur++ = uch;
200	str += clen;	308	str += clen;
201	}	309	}
202	else if (is_utf8)	310	else if (is_utf8)
203	{	311	{
204	need (enc, len += clen);	312	need (enc, len += clen);
…		…
208	}	316	}
209	while (--clen);	317	while (--clen);
210	}	318	}
211	else	319	else
212	{	320	{
213	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
214	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
215	++str;	323	++str;
216	}	324	}
217	}	325	}
218	}	326	}
219	}	327	}
220		328
221	--len;	329	--len;
222	}	330	}
223	}	331	}
224		332
225	static void	333	INLINE void
226	encode_indent (enc_t *enc)	334	encode_indent (enc_t *enc)
227	{	335	{
228	if (enc->flags & F_INDENT)	336	if (enc->json.flags & F_INDENT)
229	{	337	{
230	int spaces = enc->indent * INDENT_STEP;	338	int spaces = enc->indent * INDENT_STEP;
231		339
232	need (enc, spaces);	340	need (enc, spaces);
233	memset (enc->cur, ' ', spaces);	341	memset (enc->cur, ' ', spaces);
234	enc->cur += spaces;	342	enc->cur += spaces;
235	}	343	}
236	}	344	}
237		345
238	static void	346	INLINE void
239	encode_space (enc_t *enc)	347	encode_space (enc_t *enc)
240	{	348	{
241	need (enc, 1);	349	need (enc, 1);
242	encode_ch (enc, ' ');	350	encode_ch (enc, ' ');
243	}	351	}
244		352
245	static void	353	INLINE void
246	encode_nl (enc_t *enc)	354	encode_nl (enc_t *enc)
247	{	355	{
248	if (enc->flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
249	{	357	{
250	need (enc, 1);	358	need (enc, 1);
251	encode_ch (enc, '\n');	359	encode_ch (enc, '\n');
252	}	360	}
253	}	361	}
254		362
255	static void	363	INLINE void
256	encode_comma (enc_t *enc)	364	encode_comma (enc_t *enc)
257	{	365	{
258	encode_ch (enc, ',');	366	encode_ch (enc, ',');
259		367
260	if (enc->flags & F_INDENT)	368	if (enc->json.flags & F_INDENT)
261	encode_nl (enc);	369	encode_nl (enc);
262	else if (enc->flags & F_SPACE_AFTER)	370	else if (enc->json.flags & F_SPACE_AFTER)
263	encode_space (enc);	371	encode_space (enc);
264	}	372	}
265		373
266	static void encode_sv (enc_t *enc, SV *sv);	374	static void encode_sv (enc_t *enc, SV *sv);
267		375
268	static void	376	static void
269	encode_av (enc_t *enc, AV *av)	377	encode_av (enc_t *enc, AV *av)
270	{	378	{
271	int i, len = av_len (av);	379	int i, len = av_len (av);
272		380
273	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
274	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
275		383
276	encode_ch (enc, '['); encode_nl (enc);	384	encode_ch (enc, '[');
277	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
278		389
279	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
280	{	391	{
		392	SV **svp = av_fetch (av, i, 0);
		393
281	encode_indent (enc);	394	encode_indent (enc);
282	encode_sv (enc, *av_fetch (av, i, 0));
283		395
		396	if (svp)
		397	encode_sv (enc, *svp);
		398	else
		399	encode_str (enc, "null", 4, 0);
		400
284	if (i < len)	401	if (i < len)
285	encode_comma (enc);	402	encode_comma (enc);
286	}	403	}
287		404
		405	encode_nl (enc); --enc->indent; encode_indent (enc);
		406	}
		407
288	encode_nl (enc);	408	encode_ch (enc, ']');
289
290	--enc->indent;
291	encode_indent (enc); encode_ch (enc, ']');
292	}	409	}
293		410
294	static void	411	static void
295	encode_he (enc_t *enc, HE *he)	412	encode_hk (enc_t *enc, HE *he)
296	{	413	{
297	encode_ch (enc, '"');	414	encode_ch (enc, '"');
298		415
299	if (HeKLEN (he) == HEf_SVKEY)	416	if (HeKLEN (he) == HEf_SVKEY)
300	{	417	{
…		…
310	else	427	else
311	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	428	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
312		429
313	encode_ch (enc, '"');	430	encode_ch (enc, '"');
314		431
315	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	432	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
316	encode_ch (enc, ':');	433	encode_ch (enc, ':');
317	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	434	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
318	encode_sv (enc, HeVAL (he));
319	}	435	}
320		436
321	// compare hash entries, used when all keys are bytestrings	437	// compare hash entries, used when all keys are bytestrings
322	static int	438	static int
323	he_cmp_fast (const void *a_, const void *b_)	439	he_cmp_fast (const void *a_, const void *b_)
…		…
328	HE *b = *(HE **)b_;	444	HE *b = *(HE **)b_;
329		445
330	STRLEN la = HeKLEN (a);	446	STRLEN la = HeKLEN (a);
331	STRLEN lb = HeKLEN (b);	447	STRLEN lb = HeKLEN (b);
332		448
333	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	449	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
334	cmp = la - lb;	450	cmp = lb - la;
335		451
336	return cmp;	452	return cmp;
337	}	453	}
338		454
339	// compare hash entries, used when some keys are sv's or utf-x	455	// compare hash entries, used when some keys are sv's or utf-x
340	static int	456	static int
341	he_cmp_slow (const void *a, const void *b)	457	he_cmp_slow (const void *a, const void *b)
342	{	458	{
343	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	459	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
344	}	460	}
345		461
346	static void	462	static void
347	encode_hv (enc_t *enc, HV *hv)	463	encode_hv (enc_t *enc, HV *hv)
348	{	464	{
349	int count, i;	465	HE *he;
350		466
351	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
352	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
353		469
354	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	470	encode_ch (enc, '{');
355		471
356	if ((count = hv_iterinit (hv)))
357	{
358	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
359	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
360	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
361	// that randomises hash orderings	475	// that randomises hash orderings
362	if (enc->flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		477	{
		478	int count = hv_iterinit (hv);
		479
		480	if (SvMAGICAL (hv))
363	{	481	{
		482	// need to count by iterating. could improve by dynamically building the vector below
		483	// but I don't care for the speed of this special case.
		484	// note also that we will run into undefined behaviour when the two iterations
		485	// do not result in the same count, something I might care for in some later release.
		486
		487	count = 0;
		488	while (hv_iternext (hv))
		489	++count;
		490
		491	hv_iterinit (hv);
		492	}
		493
		494	if (count)
		495	{
		496	int i, fast = 1;
		497	#if defined(__BORLANDC__) || defined(_MSC_VER)
		498	HE **hes = _alloca (count * sizeof (HE));
		499	#else
364	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	500	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
365	int fast = 1;	501	#endif
366		502
367	i = 0;	503	i = 0;
368	while ((he = hv_iternext (hv)))	504	while ((he = hv_iternext (hv)))
369	{	505	{
370	hes [i++] = he;	506	hes [i++] = he;
…		…
392		528
393	FREETMPS;	529	FREETMPS;
394	LEAVE;	530	LEAVE;
395	}	531	}
396		532
397	for (i = 0; i < count; ++i)	533	encode_nl (enc); ++enc->indent;
		534
		535	while (count--)
398	{	536	{
399	encode_indent (enc);	537	encode_indent (enc);
		538	he = hes [count];
400	encode_he (enc, hes [i]);	539	encode_hk (enc, he);
		540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
401		541
402	if (i < count - 1)	542	if (count)
403	encode_comma (enc);	543	encode_comma (enc);
404	}	544	}
405		545
		546	encode_nl (enc); --enc->indent; encode_indent (enc);
		547	}
		548	}
		549	else
		550	{
		551	if (hv_iterinit (hv) || SvMAGICAL (hv))
		552	if ((he = hv_iternext (hv)))
		553	{
		554	encode_nl (enc); ++enc->indent;
		555
		556	for (;;)
		557	{
406	encode_nl (enc);	558	encode_indent (enc);
		559	encode_hk (enc, he);
		560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		561
		562	if (!(he = hv_iternext (hv)))
		563	break;
		564
		565	encode_comma (enc);
		566	}
		567
		568	encode_nl (enc); --enc->indent; encode_indent (enc);
		569	}
		570	}
		571
		572	encode_ch (enc, '}');
		573	}
		574
		575	// encode objects, arrays and special \0=false and \1=true values.
		576	static void
		577	encode_rv (enc_t *enc, SV *sv)
		578	{
		579	svtype svt;
		580
		581	SvGETMAGIC (sv);
		582	svt = SvTYPE (sv);
		583
		584	if (expect_false (SvOBJECT (sv)))
		585	{
		586	HV *stash = !JSON_SLOW || json_boolean_stash
		587	? json_boolean_stash
		588	: gv_stashpv ("JSON::XS::Boolean", 1);
		589
		590	if (SvSTASH (sv) == stash)
		591	{
		592	if (SvIV (sv))
		593	encode_str (enc, "true", 4, 0);
		594	else
		595	encode_str (enc, "false", 5, 0);
407	}	596	}
408	else	597	else
409	{	598	{
410	SV *sv;	599	#if 0
411	HE *he = hv_iternext (hv);	600	if (0 && sv_derived_from (rv, "JSON::Literal"))
412
413	for (;;)
414	{	601	{
415	encode_indent (enc);	602	// not yet
416	encode_he (enc, he);
417
418	if (!(he = hv_iternext (hv)))
419	break;
420
421	encode_comma (enc);
422	}	603	}
		604	#endif
		605	if (enc->json.flags & F_CONV_BLESSED)
		606	{
		607	// we re-bless the reference to get overload and other niceties right
		608	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
423		609
		610	if (to_json)
		611	{
		612	dSP;
		613
		614	ENTER; SAVETMPS; PUSHMARK (SP);
		615	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		616
		617	// calling with G_SCALAR ensures that we always get a 1 return value
		618	PUTBACK;
		619	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		620	SPAGAIN;
		621
		622	// catch this surprisingly common error
		623	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		624	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		625
		626	sv = POPs;
		627	PUTBACK;
		628
424	encode_nl (enc);	629	encode_sv (enc, sv);
		630
		631	FREETMPS; LEAVE;
		632	}
		633	else if (enc->json.flags & F_ALLOW_BLESSED)
		634	encode_str (enc, "null", 4, 0);
		635	else
		636	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		637	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		638	}
		639	else if (enc->json.flags & F_ALLOW_BLESSED)
		640	encode_str (enc, "null", 4, 0);
		641	else
		642	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		643	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
425	}	644	}
426	}	645	}
427
428	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
429	}
430
431	// encode objects, arrays and special \0=false and \1=true values.
432	static void
433	encode_rv (enc_t *enc, SV *sv)
434	{
435	SvGETMAGIC (sv);
436
437	svtype svt = SvTYPE (sv);
438
439	if (svt == SVt_PVHV)	646	else if (svt == SVt_PVHV)
440	encode_hv (enc, (HV *)sv);	647	encode_hv (enc, (HV *)sv);
441	else if (svt == SVt_PVAV)	648	else if (svt == SVt_PVAV)
442	encode_av (enc, (AV *)sv);	649	encode_av (enc, (AV *)sv);
443	else if (svt < SVt_PVAV)	650	else if (svt < SVt_PVAV)
444	{	651	{
445	if (SvNIOK (sv) && SvIV (sv) == 0)	652	STRLEN len = 0;
		653	char *pv = svt ? SvPV (sv, len) : 0;
		654
		655	if (len == 1 && *pv == '1')
		656	encode_str (enc, "true", 4, 0);
		657	else if (len == 1 && *pv == '0')
446	encode_str (enc, "false", 5, 0);	658	encode_str (enc, "false", 5, 0);
447	else if (SvNIOK (sv) && SvIV (sv) == 1)	659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
448	encode_str (enc, "true", 4, 0);	660	encode_str (enc, "null", 4, 0);
449	else	661	else
450	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
451	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
452	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
453	else	667	else
454	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
455	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
456	}	670	}
457		671
…		…
468	encode_str (enc, str, len, SvUTF8 (sv));	682	encode_str (enc, str, len, SvUTF8 (sv));
469	encode_ch (enc, '"');	683	encode_ch (enc, '"');
470	}	684	}
471	else if (SvNOKp (sv))	685	else if (SvNOKp (sv))
472	{	686	{
		687	// trust that perl will do the right thing w.r.t. JSON syntax.
473	need (enc, NV_DIG + 32);	688	need (enc, NV_DIG + 32);
474	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
475	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
476	}	691	}
477	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
478	{	693	{
		694	// we assume we can always read an IV as a UV and vice versa
		695	// we assume two's complement
		696	// we assume no aliasing issues in the union
		697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		699	{
		700	// optimise the "small number case"
		701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
		703	I32 i = SvIVX (sv);
		704	U32 u;
		705	char digit, nz = 0;
		706
479	need (enc, 64);	707	need (enc, 6);
		708
		709	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		710	u = i < 0 ? -i : i;
		711
		712	// convert to 4.28 fixed-point representation
		713	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		714
		715	// now output digit by digit, each time masking out the integer part
		716	// and multiplying by 5 while moving the decimal point one to the right,
		717	// resulting in a net multiplication by 10.
		718	// we always write the digit to memory but conditionally increment
		719	// the pointer, to enable the use of conditional move instructions.
		720	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		721	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		722	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		723	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
480	enc->cur +=	730	enc->cur +=
481	SvIsUV(sv)	731	SvIsUV(sv)
482	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
483	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
484	}	735	}
485	else if (SvROK (sv))	736	else if (SvROK (sv))
486	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
487	else if (!SvOK (sv))	738	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
488	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
489	else	740	else
490	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
491	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
492	}	743	}
493		744
494	static SV *	745	static SV *
495	encode_json (SV *scalar, U32 flags)	746	encode_json (SV *scalar, JSON *json)
496	{	747	{
		748	enc_t enc;
		749
497	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	750	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
498	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	751	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
499		752
500	enc_t enc;	753	enc.json = *json;
501	enc.flags = flags;
502	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
503	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
504	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
505	enc.indent = 0;	757	enc.indent = 0;
506	enc.maxdepth = DEC_DEPTH (flags);	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
507		761
508	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
509	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
		764	encode_nl (&enc);
510		765
		766	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		767	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		768
511	if (!(flags & (F_ASCII | F_UTF8)))	769	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
512	SvUTF8_on (enc.sv);	770	SvUTF8_on (enc.sv);
513		771
514	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
515
516	if (enc.flags & F_SHRINK)	772	if (enc.json.flags & F_SHRINK)
517	shrink (enc.sv);	773	shrink (enc.sv);
518		774
519	return enc.sv;	775	return enc.sv;
520	}	776	}
521		777
…		…
526	typedef struct	782	typedef struct
527	{	783	{
528	char *cur; // current parser pointer	784	char *cur; // current parser pointer
529	char *end; // end of input string	785	char *end; // end of input string
530	const char *err; // parse error, if != 0	786	const char *err; // parse error, if != 0
531	U32 flags; // F_*	787	JSON json;
532	U32 depth; // recursion depth	788	U32 depth; // recursion depth
533	U32 maxdepth; // recursion depth limit	789	U32 maxdepth; // recursion depth limit
534	} dec_t;	790	} dec_t;
535		791
536	static void	792	INLINE void
		793	decode_comment (dec_t *dec)
		794	{
		795	// only '#'-style comments allowed a.t.m.
		796
		797	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		798	++dec->cur;
		799	}
		800
		801	INLINE void
537	decode_ws (dec_t *dec)	802	decode_ws (dec_t *dec)
538	{	803	{
539	for (;;)	804	for (;;)
540	{	805	{
541	char ch = *dec->cur;	806	char ch = *dec->cur;
542		807
543	if (ch > 0x20	808	if (ch > 0x20)
		809	{
		810	if (expect_false (ch == '#'))
		811	{
		812	if (dec->json.flags & F_RELAXED)
		813	decode_comment (dec);
		814	else
		815	break;
		816	}
		817	else
		818	break;
		819	}
544	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	820	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
545	break;	821	break; // parse error, but let higher level handle it, gives better error messages
546		822
547	++dec->cur;	823	++dec->cur;
548	}	824	}
549	}	825	}
550		826
…		…
554	if (*dec->cur != ch) \	830	if (*dec->cur != ch) \
555	ERR (# ch " expected"); \	831	ERR (# ch " expected"); \
556	++dec->cur; \	832	++dec->cur; \
557	SE	833	SE
558		834
559	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	835	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
560	#define DEC_DEC_DEPTH --dec->depth	836	#define DEC_DEC_DEPTH --dec->depth
561		837
562	static SV *decode_sv (dec_t *dec);	838	static SV *decode_sv (dec_t *dec);
563		839
564	static signed char decode_hexdigit[256];	840	static signed char decode_hexdigit[256];
…		…
567	decode_4hex (dec_t *dec)	843	decode_4hex (dec_t *dec)
568	{	844	{
569	signed char d1, d2, d3, d4;	845	signed char d1, d2, d3, d4;
570	unsigned char *cur = (unsigned char *)dec->cur;	846	unsigned char *cur = (unsigned char *)dec->cur;
571		847
572	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	848	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
573	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	849	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
574	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	850	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
575	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	851	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
576		852
577	dec->cur += 4;	853	dec->cur += 4;
578		854
579	return ((UV)d1) << 12	855	return ((UV)d1) << 12
580	| ((UV)d2) << 8	856	| ((UV)d2) << 8
…		…
588	static SV *	864	static SV *
589	decode_str (dec_t *dec)	865	decode_str (dec_t *dec)
590	{	866	{
591	SV *sv = 0;	867	SV *sv = 0;
592	int utf8 = 0;	868	int utf8 = 0;
		869	char *dec_cur = dec->cur;
593		870
594	do	871	do
595	{	872	{
596	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	873	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
597	char *cur = buf;	874	char *cur = buf;
598		875
599	do	876	do
600	{	877	{
601	unsigned char ch = *(unsigned char *)dec->cur++;	878	unsigned char ch = *(unsigned char *)dec_cur++;
602		879
603	if (ch == '"')	880	if (expect_false (ch == '"'))
604	{	881	{
605	--dec->cur;	882	--dec_cur;
606	break;	883	break;
607	}	884	}
608	else if (ch == '\\')	885	else if (expect_false (ch == '\\'))
609	{	886	{
610	switch (*dec->cur)	887	switch (*dec_cur)
611	{	888	{
612	case '\\':	889	case '\\':
613	case '/':	890	case '/':
614	case '"': *cur++ = *dec->cur++; break;	891	case '"': *cur++ = *dec_cur++; break;
615		892
616	case 'b': ++dec->cur; *cur++ = '\010'; break;	893	case 'b': ++dec_cur; *cur++ = '\010'; break;
617	case 't': ++dec->cur; *cur++ = '\011'; break;	894	case 't': ++dec_cur; *cur++ = '\011'; break;
618	case 'n': ++dec->cur; *cur++ = '\012'; break;	895	case 'n': ++dec_cur; *cur++ = '\012'; break;
619	case 'f': ++dec->cur; *cur++ = '\014'; break;	896	case 'f': ++dec_cur; *cur++ = '\014'; break;
620	case 'r': ++dec->cur; *cur++ = '\015'; break;	897	case 'r': ++dec_cur; *cur++ = '\015'; break;
621		898
622	case 'u':	899	case 'u':
623	{	900	{
624	UV lo, hi;	901	UV lo, hi;
625	++dec->cur;	902	++dec_cur;
626		903
		904	dec->cur = dec_cur;
627	hi = decode_4hex (dec);	905	hi = decode_4hex (dec);
		906	dec_cur = dec->cur;
628	if (hi == (UV)-1)	907	if (hi == (UV)-1)
629	goto fail;	908	goto fail;
630		909
631	// possibly a surrogate pair	910	// possibly a surrogate pair
632	if (hi >= 0xd800)	911	if (hi >= 0xd800)
633	if (hi < 0xdc00)	912	if (hi < 0xdc00)
634	{	913	{
635	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	914	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
636	ERR ("missing low surrogate character in surrogate pair");	915	ERR ("missing low surrogate character in surrogate pair");
637		916
638	dec->cur += 2;	917	dec_cur += 2;
639		918
		919	dec->cur = dec_cur;
640	lo = decode_4hex (dec);	920	lo = decode_4hex (dec);
		921	dec_cur = dec->cur;
641	if (lo == (UV)-1)	922	if (lo == (UV)-1)
642	goto fail;	923	goto fail;
643		924
644	if (lo < 0xdc00 || lo >= 0xe000)	925	if (lo < 0xdc00 || lo >= 0xe000)
645	ERR ("surrogate pair expected");	926	ERR ("surrogate pair expected");
…		…
651		932
652	if (hi >= 0x80)	933	if (hi >= 0x80)
653	{	934	{
654	utf8 = 1;	935	utf8 = 1;
655		936
656	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	937	cur = encode_utf8 (cur, hi);
657	}	938	}
658	else	939	else
659	*cur++ = hi;	940	*cur++ = hi;
660	}	941	}
661	break;	942	break;
662		943
663	default:	944	default:
664	--dec->cur;	945	--dec_cur;
665	ERR ("illegal backslash escape sequence in string");	946	ERR ("illegal backslash escape sequence in string");
666	}	947	}
667	}	948	}
668	else if (ch >= 0x20 && ch <= 0x7f)	949	else if (expect_true (ch >= 0x20 && ch < 0x80))
669	*cur++ = ch;	950	*cur++ = ch;
670	else if (ch >= 0x80)	951	else if (ch >= 0x80)
671	{	952	{
672	--dec->cur;
673
674	STRLEN clen;	953	STRLEN clen;
		954
		955	--dec_cur;
		956
675	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	957	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
676	if (clen == (STRLEN)-1)	958	if (clen == (STRLEN)-1)
677	ERR ("malformed UTF-8 character in JSON string");	959	ERR ("malformed UTF-8 character in JSON string");
678		960
679	do	961	do
680	{
681	*cur++ = *dec->cur++;	962	*cur++ = *dec_cur++;
682	}
683	while (--clen);	963	while (--clen);
684		964
685	utf8 = 1;	965	utf8 = 1;
686	}	966	}
687	else if (!ch)
688	ERR ("unexpected end of string while parsing json string");
689	else	967	else
		968	{
		969	--dec_cur;
		970
		971	if (!ch)
		972	ERR ("unexpected end of string while parsing JSON string");
		973	else
690	ERR ("invalid character encountered");	974	ERR ("invalid character encountered while parsing JSON string");
691		975	}
692	}	976	}
693	while (cur < buf + SHORT_STRING_LEN);	977	while (cur < buf + SHORT_STRING_LEN);
694		978
		979	{
695	STRLEN len = cur - buf;	980	STRLEN len = cur - buf;
696		981
697	if (sv)	982	if (sv)
698	{	983	{
699	SvGROW (sv, SvCUR (sv) + len + 1);	984	STRLEN cur = SvCUR (sv);
		985
		986	if (SvLEN (sv) <= cur + len)
		987	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		988
700	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	989	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
701	SvCUR_set (sv, SvCUR (sv) + len);	990	SvCUR_set (sv, SvCUR (sv) + len);
702	}	991	}
703	else	992	else
704	sv = newSVpvn (buf, len);	993	sv = newSVpvn (buf, len);
705	}	994	}
		995	}
706	while (*dec->cur != '"');	996	while (*dec_cur != '"');
707		997
708	++dec->cur;	998	++dec_cur;
709		999
710	if (sv)	1000	if (sv)
711	{	1001	{
712	SvPOK_only (sv);	1002	SvPOK_only (sv);
713	*SvEND (sv) = 0;	1003	*SvEND (sv) = 0;
…		…
716	SvUTF8_on (sv);	1006	SvUTF8_on (sv);
717	}	1007	}
718	else	1008	else
719	sv = newSVpvn ("", 0);	1009	sv = newSVpvn ("", 0);
720		1010
		1011	dec->cur = dec_cur;
721	return sv;	1012	return sv;
722		1013
723	fail:	1014	fail:
		1015	dec->cur = dec_cur;
724	return 0;	1016	return 0;
725	}	1017	}
726		1018
727	static SV *	1019	static SV *
728	decode_num (dec_t *dec)	1020	decode_num (dec_t *dec)
…		…
786	is_nv = 1;	1078	is_nv = 1;
787	}	1079	}
788		1080
789	if (!is_nv)	1081	if (!is_nv)
790	{	1082	{
791	UV uv;	1083	int len = dec->cur - start;
792	int numtype = grok_number (start, dec->cur - start, &uv);	1084
793	if (numtype & IS_NUMBER_IN_UV)	1085	// special case the rather common 1..5-digit-int case
794	if (numtype & IS_NUMBER_NEG)	1086	if (*start == '-')
		1087	switch (len)
795	{	1088	{
796	if (uv < (UV)IV_MIN)	1089	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
797	return newSViv (-(IV)uv);	1090	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1091	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1092	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1093	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
798	}	1094	}
		1095	else
		1096	switch (len)
		1097	{
		1098	case 1: return newSViv ( start [0] - '0' * 1);
		1099	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1100	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1101	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1102	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1103	}
		1104
		1105	{
		1106	UV uv;
		1107	int numtype = grok_number (start, len, &uv);
		1108	if (numtype & IS_NUMBER_IN_UV)
		1109	if (numtype & IS_NUMBER_NEG)
		1110	{
		1111	if (uv < (UV)IV_MIN)
		1112	return newSViv (-(IV)uv);
		1113	}
799	else	1114	else
800	return newSVuv (uv);	1115	return newSVuv (uv);
801	}	1116	}
802		1117
		1118	len -= *start == '-' ? 1 : 0;
		1119
		1120	// does not fit into IV or UV, try NV
		1121	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1122	#if defined (LDBL_DIG)
		1123	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1124	#endif
		1125	)
		1126	// fits into NV without loss of precision
		1127	return newSVnv (Atof (start));
		1128
		1129	// everything else fails, convert it to a string
		1130	return newSVpvn (start, dec->cur - start);
		1131	}
		1132
		1133	// loss of precision here
803	return newSVnv (Atof (start));	1134	return newSVnv (Atof (start));
804		1135
805	fail:	1136	fail:
806	return 0;	1137	return 0;
807	}	1138	}
…		…
837		1168
838	if (*dec->cur != ',')	1169	if (*dec->cur != ',')
839	ERR (", or ] expected while parsing array");	1170	ERR (", or ] expected while parsing array");
840		1171
841	++dec->cur;	1172	++dec->cur;
		1173
		1174	decode_ws (dec);
		1175
		1176	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1177	{
		1178	++dec->cur;
		1179	break;
		1180	}
842	}	1181	}
843		1182
844	DEC_DEC_DEPTH;	1183	DEC_DEC_DEPTH;
845	return newRV_noinc ((SV *)av);	1184	return newRV_noinc ((SV *)av);
846		1185
…		…
851	}	1190	}
852		1191
853	static SV *	1192	static SV *
854	decode_hv (dec_t *dec)	1193	decode_hv (dec_t *dec)
855	{	1194	{
		1195	SV *sv;
856	HV *hv = newHV ();	1196	HV *hv = newHV ();
857		1197
858	DEC_INC_DEPTH;	1198	DEC_INC_DEPTH;
859	decode_ws (dec);	1199	decode_ws (dec);
860		1200
861	if (*dec->cur == '}')	1201	if (*dec->cur == '}')
862	++dec->cur;	1202	++dec->cur;
863	else	1203	else
864	for (;;)	1204	for (;;)
865	{	1205	{
		1206	EXPECT_CH ('"');
		1207
		1208	// heuristic: assume that
		1209	// a) decode_str + hv_store_ent are abysmally slow.
		1210	// b) most hash keys are short, simple ascii text.
		1211	// => try to "fast-match" such strings to avoid
		1212	// the overhead of decode_str + hv_store_ent.
		1213	{
866	SV *key, *value;	1214	SV *value;
		1215	char *p = dec->cur;
		1216	char *e = p + 24; // only try up to 24 bytes
867		1217
868	decode_ws (dec); EXPECT_CH ('"');	1218	for (;;)
869
870	key = decode_str (dec);
871	if (!key)
872	goto fail;
873
874	decode_ws (dec); EXPECT_CH (':');
875
876	value = decode_sv (dec);
877	if (!value)
878	{	1219	{
		1220	// the >= 0x80 is false on most architectures
		1221	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1222	{
		1223	// slow path, back up and use decode_str
		1224	SV *key = decode_str (dec);
		1225	if (!key)
		1226	goto fail;
		1227
		1228	decode_ws (dec); EXPECT_CH (':');
		1229
		1230	decode_ws (dec);
		1231	value = decode_sv (dec);
		1232	if (!value)
		1233	{
		1234	SvREFCNT_dec (key);
		1235	goto fail;
		1236	}
		1237
		1238	hv_store_ent (hv, key, value, 0);
879	SvREFCNT_dec (key);	1239	SvREFCNT_dec (key);
		1240
		1241	break;
		1242	}
		1243	else if (*p == '"')
		1244	{
		1245	// fast path, got a simple key
		1246	char *key = dec->cur;
		1247	int len = p - key;
		1248	dec->cur = p + 1;
		1249
		1250	decode_ws (dec); EXPECT_CH (':');
		1251
		1252	decode_ws (dec);
		1253	value = decode_sv (dec);
		1254	if (!value)
880	goto fail;	1255	goto fail;
		1256
		1257	hv_store (hv, key, len, value, 0);
		1258
		1259	break;
		1260	}
		1261
		1262	++p;
881	}	1263	}
882		1264	}
883	hv_store_ent (hv, key, value, 0);
884	SvREFCNT_dec (key);
885		1265
886	decode_ws (dec);	1266	decode_ws (dec);
887		1267
888	if (*dec->cur == '}')	1268	if (*dec->cur == '}')
889	{	1269	{
…		…
893		1273
894	if (*dec->cur != ',')	1274	if (*dec->cur != ',')
895	ERR (", or } expected while parsing object/hash");	1275	ERR (", or } expected while parsing object/hash");
896		1276
897	++dec->cur;	1277	++dec->cur;
		1278
		1279	decode_ws (dec);
		1280
		1281	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1282	{
		1283	++dec->cur;
		1284	break;
		1285	}
898	}	1286	}
899		1287
900	DEC_DEC_DEPTH;	1288	DEC_DEC_DEPTH;
901	return newRV_noinc ((SV *)hv);	1289	sv = newRV_noinc ((SV *)hv);
		1290
		1291	// check filter callbacks
		1292	if (dec->json.flags & F_HOOK)
		1293	{
		1294	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1295	{
		1296	HE *cb, *he;
		1297
		1298	hv_iterinit (hv);
		1299	he = hv_iternext (hv);
		1300	hv_iterinit (hv);
		1301
		1302	// the next line creates a mortal sv each time its called.
		1303	// might want to optimise this for common cases.
		1304	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1305
		1306	if (cb)
		1307	{
		1308	dSP;
		1309	int count;
		1310
		1311	ENTER; SAVETMPS; PUSHMARK (SP);
		1312	XPUSHs (HeVAL (he));
		1313
		1314	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1315
		1316	if (count == 1)
		1317	{
		1318	sv = newSVsv (POPs);
		1319	FREETMPS; LEAVE;
		1320	return sv;
		1321	}
		1322
		1323	FREETMPS; LEAVE;
		1324	}
		1325	}
		1326
		1327	if (dec->json.cb_object)
		1328	{
		1329	dSP;
		1330	int count;
		1331
		1332	ENTER; SAVETMPS; PUSHMARK (SP);
		1333	XPUSHs (sv_2mortal (sv));
		1334
		1335	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1336
		1337	if (count == 1)
		1338	{
		1339	sv = newSVsv (POPs);
		1340	FREETMPS; LEAVE;
		1341	return sv;
		1342	}
		1343
		1344	SvREFCNT_inc (sv);
		1345	FREETMPS; LEAVE;
		1346	}
		1347	}
		1348
		1349	return sv;
902		1350
903	fail:	1351	fail:
904	SvREFCNT_dec (hv);	1352	SvREFCNT_dec (hv);
905	DEC_DEC_DEPTH;	1353	DEC_DEC_DEPTH;
906	return 0;	1354	return 0;
907	}	1355	}
908		1356
909	static SV *	1357	static SV *
910	decode_sv (dec_t *dec)	1358	decode_sv (dec_t *dec)
911	{	1359	{
912	decode_ws (dec);	1360	// the beauty of JSON: you need exactly one character lookahead
		1361	// to parse everything.
913	switch (*dec->cur)	1362	switch (*dec->cur)
914	{	1363	{
915	case '"': ++dec->cur; return decode_str (dec);	1364	case '"': ++dec->cur; return decode_str (dec);
916	case '[': ++dec->cur; return decode_av (dec);	1365	case '[': ++dec->cur; return decode_av (dec);
917	case '{': ++dec->cur; return decode_hv (dec);	1366	case '{': ++dec->cur; return decode_hv (dec);
918		1367
919	case '-':	1368	case '-':
920	case '0': case '1': case '2': case '3': case '4':	1369	case '0': case '1': case '2': case '3': case '4':
921	case '5': case '6': case '7': case '8': case '9':	1370	case '5': case '6': case '7': case '8': case '9':
922	return decode_num (dec);	1371	return decode_num (dec);
923		1372
924	case 't':	1373	case 't':
925	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1374	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
926	{	1375	{
927	dec->cur += 4;	1376	dec->cur += 4;
		1377	#if JSON_SLOW
		1378	json_true = get_bool ("JSON::XS::true");
		1379	#endif
928	return newSViv (1);	1380	return newSVsv (json_true);
929	}	1381	}
930	else	1382	else
931	ERR ("'true' expected");	1383	ERR ("'true' expected");
932		1384
933	break;	1385	break;
934		1386
935	case 'f':	1387	case 'f':
936	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1388	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
937	{	1389	{
938	dec->cur += 5;	1390	dec->cur += 5;
		1391	#if JSON_SLOW
		1392	json_false = get_bool ("JSON::XS::false");
		1393	#endif
939	return newSViv (0);	1394	return newSVsv (json_false);
940	}	1395	}
941	else	1396	else
942	ERR ("'false' expected");	1397	ERR ("'false' expected");
943		1398
944	break;	1399	break;
…		…
953	ERR ("'null' expected");	1408	ERR ("'null' expected");
954		1409
955	break;	1410	break;
956		1411
957	default:	1412	default:
958	ERR ("malformed json string, neither array, object, number, string or atom");	1413	ERR ("malformed JSON string, neither array, object, number, string or atom");
959	break;	1414	break;
960	}	1415	}
961		1416
962	fail:	1417	fail:
963	return 0;	1418	return 0;
964	}	1419	}
965		1420
966	static SV *	1421	static SV *
967	decode_json (SV *string, U32 flags)	1422	decode_json (SV *string, JSON *json, char **offset_return)
968	{	1423	{
		1424	dec_t dec;
969	SV *sv;	1425	SV *sv;
970		1426
		1427	/* work around bugs in 5.10 where manipulating magic values
		1428	* will perl ignore the magic in subsequent accesses
		1429	*/
		1430	/*SvGETMAGIC (string);*/
		1431	if (SvMAGICAL (string))
		1432	string = sv_2mortal (newSVsv (string));
		1433
		1434	SvUPGRADE (string, SVt_PV);
		1435
		1436	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1437	* assertion with -DDEBUGGING, although SvCUR is documented to
		1438	* return the xpv_cur field which certainly exists after upgrading.
		1439	* according to nicholas clark, calling SvPOK fixes this.
		1440	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1441	* and hope for the best.
		1442	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1443	* assertion business is seriously broken, try yet another workaround
		1444	* for the broken -DDEBUGGING.
		1445	*/
		1446	{
		1447	#ifdef DEBUGGING
		1448	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1449	#else
		1450	STRLEN offset = SvCUR (string);
		1451	#endif
		1452
		1453	if (offset > json->max_size && json->max_size)
		1454	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1455	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1456	}
		1457
971	if (flags & F_UTF8)	1458	if (json->flags & F_UTF8)
972	sv_utf8_downgrade (string, 0);	1459	sv_utf8_downgrade (string, 0);
973	else	1460	else
974	sv_utf8_upgrade (string);	1461	sv_utf8_upgrade (string);
975		1462
976	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1463	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
977		1464
978	dec_t dec;	1465	dec.json = *json;
979	dec.flags = flags;
980	dec.cur = SvPVX (string);	1466	dec.cur = SvPVX (string);
981	dec.end = SvEND (string);	1467	dec.end = SvEND (string);
982	dec.err = 0;	1468	dec.err = 0;
983	dec.depth = 0;	1469	dec.depth = 0;
984	dec.maxdepth = DEC_DEPTH (dec.flags);
985		1470
		1471	if (dec.json.cb_object || dec.json.cb_sk_object)
		1472	dec.json.flags |= F_HOOK;
		1473
986	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1474	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1475
		1476	decode_ws (&dec);
987	sv = decode_sv (&dec);	1477	sv = decode_sv (&dec);
988		1478
		1479	if (offset_return)
		1480	*offset_return = dec.cur;
		1481
		1482	if (!(offset_return || !sv))
		1483	{
		1484	// check for trailing garbage
		1485	decode_ws (&dec);
		1486
		1487	if (*dec.cur)
		1488	{
		1489	dec.err = "garbage after JSON object";
		1490	SvREFCNT_dec (sv);
		1491	sv = 0;
		1492	}
		1493	}
		1494
989	if (!sv)	1495	if (!sv)
990	{	1496	{
991	IV offset = dec.flags & F_UTF8
992	? dec.cur - SvPVX (string)
993	: utf8_distance (dec.cur, SvPVX (string));
994	SV *uni = sv_newmortal ();	1497	SV *uni = sv_newmortal ();
995		1498
996	// horrible hack to silence warning inside pv_uni_display	1499	// horrible hack to silence warning inside pv_uni_display
997	COP cop = *PL_curcop;	1500	COP cop = *PL_curcop;
998	cop.cop_warnings = pWARN_NONE;	1501	cop.cop_warnings = pWARN_NONE;
…		…
1000	SAVEVPTR (PL_curcop);	1503	SAVEVPTR (PL_curcop);
1001	PL_curcop = &cop;	1504	PL_curcop = &cop;
1002	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1505	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1003	LEAVE;	1506	LEAVE;
1004		1507
1005	croak ("%s, at character offset %d (%s)",	1508	croak ("%s, at character offset %d (before \"%s\")",
1006	dec.err,	1509	dec.err,
1007	(int)offset,	1510	ptr_to_index (string, dec.cur),
1008	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1511	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1009	}	1512	}
1010		1513
1011	sv = sv_2mortal (sv);	1514	sv = sv_2mortal (sv);
1012		1515
1013	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1516	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1014	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1517	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1015		1518
1016	return sv;	1519	return sv;
		1520	}
		1521
		1522	/////////////////////////////////////////////////////////////////////////////
		1523	// incremental parser
		1524
		1525	static void
		1526	incr_parse (JSON *self)
		1527	{
		1528	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1529
		1530	for (;;)
		1531	{
		1532	//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
		1533	switch (self->incr_mode)
		1534	{
		1535	// only used for intiial whitespace skipping
		1536	case INCR_M_WS:
		1537	for (;;)
		1538	{
		1539	if (*p > 0x20)
		1540	{
		1541	self->incr_mode = INCR_M_JSON;
		1542	goto incr_m_json;
		1543	}
		1544	else if (!*p)
		1545	goto interrupt;
		1546
		1547	++p;
		1548	}
		1549
		1550	// skip a single char inside a string (for \\-processing)
		1551	case INCR_M_BS:
		1552	if (!*p)
		1553	goto interrupt;
		1554
		1555	++p;
		1556	self->incr_mode = INCR_M_STR;
		1557	goto incr_m_str;
		1558
		1559	// inside a string
		1560	case INCR_M_STR:
		1561	incr_m_str:
		1562	for (;;)
		1563	{
		1564	if (*p == '"')
		1565	{
		1566	++p;
		1567	self->incr_mode = INCR_M_JSON;
		1568
		1569	if (!self->incr_nest)
		1570	goto interrupt;
		1571
		1572	goto incr_m_json;
		1573	}
		1574	else if (*p == '\\')
		1575	{
		1576	++p; // "virtually" consumes character after \
		1577
		1578	if (!*p) // if at end of string we have to switch modes
		1579	{
		1580	self->incr_mode = INCR_M_BS;
		1581	goto interrupt;
		1582	}
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	// after initial ws, outside string
		1591	case INCR_M_JSON:
		1592	incr_m_json:
		1593	for (;;)
		1594	{
		1595	switch (*p++)
		1596	{
		1597	case 0:
		1598	--p;
		1599	goto interrupt;
		1600
		1601	case 0x09:
		1602	case 0x0a:
		1603	case 0x0d:
		1604	case 0x20:
		1605	if (!self->incr_nest)
		1606	{
		1607	--p; // do not eat the whitespace, let the next round do it
		1608	goto interrupt;
		1609	}
		1610	break;
		1611
		1612	case '"':
		1613	self->incr_mode = INCR_M_STR;
		1614	goto incr_m_str;
		1615
		1616	case '[':
		1617	case '{':
		1618	if (++self->incr_nest > self->max_depth)
		1619	croak (ERR_NESTING_EXCEEDED);
		1620	break;
		1621
		1622	case ']':
		1623	case '}':
		1624	if (--self->incr_nest <= 0)
		1625	goto interrupt;
		1626	}
		1627	}
		1628	}
		1629
		1630	modechange:
		1631	;
		1632	}
		1633
		1634	interrupt:
		1635	self->incr_pos = p - SvPVX (self->incr_text);
		1636	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1017	}	1637	}
1018		1638
1019	/////////////////////////////////////////////////////////////////////////////	1639	/////////////////////////////////////////////////////////////////////////////
1020	// XS interface functions	1640	// XS interface functions
1021		1641
1022	MODULE = JSON::XS PACKAGE = JSON::XS	1642	MODULE = JSON::XS PACKAGE = JSON::XS
1023		1643
1024	BOOT:	1644	BOOT:
1025	{	1645	{
1026	int i;	1646	int i;
1027
1028	memset (decode_hexdigit, 0xff, 256);
1029		1647
1030	for (i = 0; i < 256; ++i)	1648	for (i = 0; i < 256; ++i)
1031	decode_hexdigit [i] =	1649	decode_hexdigit [i] =
1032	i >= '0' && i <= '9' ? i - '0'	1650	i >= '0' && i <= '9' ? i - '0'
1033	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1651	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1034	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1652	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1035	: -1;	1653	: -1;
1036		1654
1037	json_stash = gv_stashpv ("JSON::XS", 1);	1655	json_stash = gv_stashpv ("JSON::XS" , 1);
		1656	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1657
		1658	json_true = get_bool ("JSON::XS::true");
		1659	json_false = get_bool ("JSON::XS::false");
		1660
		1661	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1038	}	1662	}
1039		1663
1040	PROTOTYPES: DISABLE	1664	PROTOTYPES: DISABLE
1041		1665
1042	SV *new (char *dummy)	1666	void CLONE (...)
1043	CODE:	1667	CODE:
1044	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1668	json_stash = 0;
		1669	json_boolean_stash = 0;
		1670
		1671	void new (char *klass)
		1672	PPCODE:
		1673	{
		1674	SV *pv = NEWSV (0, sizeof (JSON));
		1675	SvPOK_only (pv);
		1676	json_init ((JSON *)SvPVX (pv));
		1677	XPUSHs (sv_2mortal (sv_bless (
		1678	newRV_noinc (pv),
		1679	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1680	)));
		1681	}
		1682
		1683	void ascii (JSON *self, int enable = 1)
		1684	ALIAS:
		1685	ascii = F_ASCII
		1686	latin1 = F_LATIN1
		1687	utf8 = F_UTF8
		1688	indent = F_INDENT
		1689	canonical = F_CANONICAL
		1690	space_before = F_SPACE_BEFORE
		1691	space_after = F_SPACE_AFTER
		1692	pretty = F_PRETTY
		1693	allow_nonref = F_ALLOW_NONREF
		1694	shrink = F_SHRINK
		1695	allow_blessed = F_ALLOW_BLESSED
		1696	convert_blessed = F_CONV_BLESSED
		1697	relaxed = F_RELAXED
		1698	allow_unknown = F_ALLOW_UNKNOWN
		1699	PPCODE:
		1700	{
		1701	if (enable)
		1702	self->flags |= ix;
		1703	else
		1704	self->flags &= ~ix;
		1705
		1706	XPUSHs (ST (0));
		1707	}
		1708
		1709	void get_ascii (JSON *self)
		1710	ALIAS:
		1711	get_ascii = F_ASCII
		1712	get_latin1 = F_LATIN1
		1713	get_utf8 = F_UTF8
		1714	get_indent = F_INDENT
		1715	get_canonical = F_CANONICAL
		1716	get_space_before = F_SPACE_BEFORE
		1717	get_space_after = F_SPACE_AFTER
		1718	get_allow_nonref = F_ALLOW_NONREF
		1719	get_shrink = F_SHRINK
		1720	get_allow_blessed = F_ALLOW_BLESSED
		1721	get_convert_blessed = F_CONV_BLESSED
		1722	get_relaxed = F_RELAXED
		1723	get_allow_unknown = F_ALLOW_UNKNOWN
		1724	PPCODE:
		1725	XPUSHs (boolSV (self->flags & ix));
		1726
		1727	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1728	PPCODE:
		1729	self->max_depth = max_depth;
		1730	XPUSHs (ST (0));
		1731
		1732	U32 get_max_depth (JSON *self)
		1733	CODE:
		1734	RETVAL = self->max_depth;
1045	OUTPUT:	1735	OUTPUT:
1046	RETVAL	1736	RETVAL
1047		1737
1048	SV *ascii (SV *self, int enable = 1)	1738	void max_size (JSON *self, U32 max_size = 0)
1049	ALIAS:	1739	PPCODE:
1050	ascii = F_ASCII	1740	self->max_size = max_size;
1051	utf8 = F_UTF8	1741	XPUSHs (ST (0));
1052	indent = F_INDENT	1742
1053	canonical = F_CANONICAL	1743	int get_max_size (JSON *self)
1054	space_before = F_SPACE_BEFORE
1055	space_after = F_SPACE_AFTER
1056	pretty = F_PRETTY
1057	allow_nonref = F_ALLOW_NONREF
1058	shrink = F_SHRINK
1059	CODE:	1744	CODE:
1060	{	1745	RETVAL = self->max_size;
1061	UV *uv = SvJSON (self);
1062	if (enable)
1063	*uv |= ix;
1064	else
1065	*uv &= ~ix;
1066
1067	RETVAL = newSVsv (self);
1068	}
1069	OUTPUT:	1746	OUTPUT:
1070	RETVAL	1747	RETVAL
1071		1748
1072	SV *max_depth (SV *self, int max_depth = 0x80000000UL)	1749	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1750	PPCODE:
		1751	{
		1752	SvREFCNT_dec (self->cb_object);
		1753	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1754
		1755	XPUSHs (ST (0));
		1756	}
		1757
		1758	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1759	PPCODE:
		1760	{
		1761	if (!self->cb_sk_object)
		1762	self->cb_sk_object = newHV ();
		1763
		1764	if (SvOK (cb))
		1765	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1766	else
		1767	{
		1768	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1769
		1770	if (!HvKEYS (self->cb_sk_object))
		1771	{
		1772	SvREFCNT_dec (self->cb_sk_object);
		1773	self->cb_sk_object = 0;
		1774	}
		1775	}
		1776
		1777	XPUSHs (ST (0));
		1778	}
		1779
		1780	void encode (JSON *self, SV *scalar)
		1781	PPCODE:
		1782	XPUSHs (encode_json (scalar, self));
		1783
		1784	void decode (JSON *self, SV *jsonstr)
		1785	PPCODE:
		1786	XPUSHs (decode_json (jsonstr, self, 0));
		1787
		1788	void decode_prefix (JSON *self, SV *jsonstr)
		1789	PPCODE:
		1790	{
		1791	char *offset;
		1792	EXTEND (SP, 2);
		1793	PUSHs (decode_json (jsonstr, self, &offset));
		1794	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1795	}
		1796
		1797	void incr_parse (JSON *self, SV *jsonstr = 0)
		1798	PPCODE:
		1799	{
		1800	if (!self->incr_text)
		1801	self->incr_text = newSVpvn ("", 0);
		1802
		1803	// append data, if any
		1804	if (jsonstr)
		1805	{
		1806	if (SvUTF8 (jsonstr))
		1807	{
		1808	if (!SvUTF8 (self->incr_text))
		1809	{
		1810	/* utf-8-ness differs, need to upgrade */
		1811	sv_utf8_upgrade (self->incr_text);
		1812
		1813	if (self->incr_pos)
		1814	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1815	- (U8 *)SvPVX (self->incr_text);
		1816	}
		1817	}
		1818	else if (SvUTF8 (self->incr_text))
		1819	sv_utf8_upgrade (jsonstr);
		1820
		1821	{
		1822	STRLEN len;
		1823	const char *str = SvPV (jsonstr, len);
		1824	STRLEN cur = SvCUR (self->incr_text);
		1825
		1826	if (SvLEN (self->incr_text) <= cur + len)
		1827	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1828
		1829	Move (str, SvEND (self->incr_text), len, char);
		1830	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1831	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1832	}
		1833	}
		1834
		1835	if (GIMME_V != G_VOID)
		1836	do
		1837	{
		1838	char *offset;
		1839
		1840	if (!INCR_DONE (self))
		1841	{
		1842	incr_parse (self);
		1843
		1844	if (self->incr_pos > self->max_size && self->max_size)
		1845	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1846	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1847
		1848	if (!INCR_DONE (self))
		1849	break;
		1850	}
		1851
		1852	XPUSHs (decode_json (self->incr_text, self, &offset));
		1853
		1854	self->incr_pos -= offset - SvPVX (self->incr_text);
		1855	self->incr_nest = 0;
		1856	self->incr_mode = 0;
		1857
		1858	sv_chop (self->incr_text, offset);
		1859	}
		1860	while (GIMME_V == G_ARRAY);
		1861	}
		1862
		1863	SV *incr_text (JSON *self)
		1864	ATTRS: lvalue
1073	CODE:	1865	CODE:
1074	{	1866	{
1075	UV *uv = SvJSON (self);	1867	if (self->incr_pos)
1076	UV log2 = 0;	1868	croak ("incr_text can not be called when the incremental parser already started parsing");
1077		1869
1078	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1870	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1079
1080	while ((1UL << log2) < max_depth)
1081	++log2;
1082
1083	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1084
1085	RETVAL = newSVsv (self);
1086	}	1871	}
1087	OUTPUT:	1872	OUTPUT:
1088	RETVAL	1873	RETVAL
1089		1874
		1875	void incr_skip (JSON *self)
		1876	CODE:
		1877	{
		1878	if (self->incr_pos)
		1879	{
		1880	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
		1884	}
		1885	}
		1886
		1887	void incr_reset (JSON *self)
		1888	CODE:
		1889	{
		1890	SvREFCNT_dec (self->incr_text);
		1891	self->incr_text = 0;
		1892	self->incr_pos = 0;
		1893	self->incr_nest = 0;
		1894	self->incr_mode = 0;
		1895	}
		1896
		1897	void DESTROY (JSON *self)
		1898	CODE:
		1899	SvREFCNT_dec (self->cb_sk_object);
		1900	SvREFCNT_dec (self->cb_object);
		1901	SvREFCNT_dec (self->incr_text);
		1902
		1903	PROTOTYPES: ENABLE
		1904
1090	void encode (SV *self, SV *scalar)	1905	void encode_json (SV *scalar)
		1906	ALIAS:
		1907	to_json_ = 0
		1908	encode_json = F_UTF8
1091	PPCODE:	1909	PPCODE:
		1910	{
		1911	JSON json;
		1912	json_init (&json);
		1913	json.flags |= ix;
1092	XPUSHs (encode_json (scalar, *SvJSON (self)));	1914	XPUSHs (encode_json (scalar, &json));
		1915	}
1093		1916
1094	void decode (SV *self, SV *jsonstr)	1917	void decode_json (SV *jsonstr)
		1918	ALIAS:
		1919	from_json_ = 0
		1920	decode_json = F_UTF8
1095	PPCODE:	1921	PPCODE:
		1922	{
		1923	JSON json;
		1924	json_init (&json);
		1925	json.flags |= ix;
1096	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1926	XPUSHs (decode_json (jsonstr, &json, 0));
		1927	}
1097		1928
1098	PROTOTYPES: ENABLE
1099		1929
1100	void to_json (SV *scalar)
1101	ALIAS:
1102	objToJson = 0
1103	PPCODE:
1104	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1105
1106	void from_json (SV *jsonstr)
1107	ALIAS:
1108	jsonToObj = 0
1109	PPCODE:
1110	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1111

Diff Legend

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