ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.32 by root, Wed May 9 16:41:12 2007 UTC vs.
Revision 1.95 by root, Sun Feb 22 06:55:28 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"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) || defined(_MSC_VER)	12	#if defined(__BORLANDC__) || defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		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
14	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	28	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	29	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
23	#define F_MAXDEPTH 0xf8000000UL	35	#define F_RELAXED 0x00001000UL
24	#define S_MAXDEPTH 27	36	#define F_ALLOW_UNKNOWN 0x00002000UL
25		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		38
31	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		40
34	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
36		43
37	#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
38		45
39	#define SB do {	46	#define SB do {
40	#define SE } while (0)	47	#define SE } while (0)
41		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
42	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	}
43		107
44	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	109	// utility functions
46		110
47	static UV *	111	INLINE SV *
48	SvJSON (SV *sv)	112	get_bool (const char *name)
49	{	113	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	114	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		115
53	return &SvUVX (SvRV (sv));	116	SvREADONLY_on (sv);
54	}	117	SvREADONLY_on (SvRV (sv));
55		118
56	static void	119	return sv;
		120	}
		121
		122	INLINE void
57	shrink (SV *sv)	123	shrink (SV *sv)
58	{	124	{
59	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
60	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	128	{
62	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
70	// 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
71	// case of an error.	138	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
75	static UV	142	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	143	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	144	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	145	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	148	{
82	*clen = 2;	149	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	151	}
85	else	152	else
86	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
87	*clen = (STRLEN)-1;	154	}
88	return (UV)-1;	155
89	}	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;
90	}	178	}
91		179
92	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
93	// encoder	181	// encoder
94		182
…		…
96	typedef struct	184	typedef struct
97	{	185	{
98	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
100	SV *sv; // result scalar	188	SV *sv; // result scalar
101	U32 flags; // F_*	189	JSON json;
102	U32 indent; // indentation level	190	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
104	} enc_t;	192	} enc_t;
105		193
106	static void	194	INLINE void
107	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
108	{	196	{
109	if (enc->cur + len >= enc->end)	197	if (expect_false (enc->cur + len >= enc->end))
110	{	198	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	203	}
116	}	204	}
117		205
118	static void	206	INLINE void
119	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
120	{	208	{
121	need (enc, 1);	209	need (enc, 1);
122	*enc->cur++ = ch;	210	*enc->cur++ = ch;
123	}	211	}
…		…
131		219
132	while (str < end)	220	while (str < end)
133	{	221	{
134	unsigned char ch = *(unsigned char *)str;	222	unsigned char ch = *(unsigned char *)str;
135		223
136	if (ch >= 0x20 && ch < 0x80) // most common case	224	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	225	{
138	if (ch == '"') // but with slow exceptions	226	if (expect_false (ch == '"')) // but with slow exceptions
139	{	227	{
140	need (enc, len += 1);	228	need (enc, len += 1);
141	*enc->cur++ = '\\';	229	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	230	*enc->cur++ = '"';
143	}	231	}
144	else if (ch == '\\')	232	else if (expect_false (ch == '\\'))
145	{	233	{
146	need (enc, len += 1);	234	need (enc, len += 1);
147	*enc->cur++ = '\\';	235	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	236	*enc->cur++ = '\\';
149	}	237	}
…		…
167	STRLEN clen;	255	STRLEN clen;
168	UV uch;	256	UV uch;
169		257
170	if (is_utf8)	258	if (is_utf8)
171	{	259	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	260	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	261	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	262	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	263	}
177	else	264	else
178	{	265	{
179	uch = ch;	266	uch = ch;
180	clen = 1;	267	clen = 1;
181	}	268	}
182		269
183	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
184	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185
186	if (uch < 0x80 || enc->flags & F_ASCII || (enc->flags & F_LATIN1 && uch > 0xFF))
187	{	271	{
188	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
189	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
190	need (enc, len += 11);	277	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	281	enc->cur += 12;
195	}	282	}
196	else	283	else
197	{	284	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	285	need (enc, len += 5);
200	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	292	}
207		293
208	str += clen;	294	str += clen;
209	}	295	}
210	else if (enc->flags & F_LATIN1)	296	else if (enc->json.flags & F_LATIN1)
211	{	297	{
212	*enc->cur++ = uch;	298	*enc->cur++ = uch;
213	str += clen;	299	str += clen;
214	}	300	}
215	else if (is_utf8)	301	else if (is_utf8)
…		…
222	while (--clen);	308	while (--clen);
223	}	309	}
224	else	310	else
225	{	311	{
226	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	312	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
227	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	313	enc->cur = encode_utf8 (enc->cur, uch);
228	++str;	314	++str;
229	}	315	}
230	}	316	}
231	}	317	}
232	}	318	}
233		319
234	--len;	320	--len;
235	}	321	}
236	}	322	}
237		323
238	static void	324	INLINE void
239	encode_indent (enc_t *enc)	325	encode_indent (enc_t *enc)
240	{	326	{
241	if (enc->flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
242	{	328	{
243	int spaces = enc->indent * INDENT_STEP;	329	int spaces = enc->indent * INDENT_STEP;
244		330
245	need (enc, spaces);	331	need (enc, spaces);
246	memset (enc->cur, ' ', spaces);	332	memset (enc->cur, ' ', spaces);
247	enc->cur += spaces;	333	enc->cur += spaces;
248	}	334	}
249	}	335	}
250		336
251	static void	337	INLINE void
252	encode_space (enc_t *enc)	338	encode_space (enc_t *enc)
253	{	339	{
254	need (enc, 1);	340	need (enc, 1);
255	encode_ch (enc, ' ');	341	encode_ch (enc, ' ');
256	}	342	}
257		343
258	static void	344	INLINE void
259	encode_nl (enc_t *enc)	345	encode_nl (enc_t *enc)
260	{	346	{
261	if (enc->flags & F_INDENT)	347	if (enc->json.flags & F_INDENT)
262	{	348	{
263	need (enc, 1);	349	need (enc, 1);
264	encode_ch (enc, '\n');	350	encode_ch (enc, '\n');
265	}	351	}
266	}	352	}
267		353
268	static void	354	INLINE void
269	encode_comma (enc_t *enc)	355	encode_comma (enc_t *enc)
270	{	356	{
271	encode_ch (enc, ',');	357	encode_ch (enc, ',');
272		358
273	if (enc->flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
274	encode_nl (enc);	360	encode_nl (enc);
275	else if (enc->flags & F_SPACE_AFTER)	361	else if (enc->json.flags & F_SPACE_AFTER)
276	encode_space (enc);	362	encode_space (enc);
277	}	363	}
278		364
279	static void encode_sv (enc_t *enc, SV *sv);	365	static void encode_sv (enc_t *enc, SV *sv);
280		366
281	static void	367	static void
282	encode_av (enc_t *enc, AV *av)	368	encode_av (enc_t *enc, AV *av)
283	{	369	{
284	int i, len = av_len (av);	370	int i, len = av_len (av);
285		371
286	if (enc->indent >= enc->maxdepth)	372	if (enc->indent >= enc->json.max_depth)
287	croak ("data structure too deep (hit recursion limit)");	373	croak (ERR_NESTING_EXCEEDED);
288		374
289	encode_ch (enc, '['); encode_nl (enc);	375	encode_ch (enc, '[');
290	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
291		380
292	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
293	{	382	{
		383	SV **svp = av_fetch (av, i, 0);
		384
294	encode_indent (enc);	385	encode_indent (enc);
295	encode_sv (enc, *av_fetch (av, i, 0));
296		386
		387	if (svp)
		388	encode_sv (enc, *svp);
		389	else
		390	encode_str (enc, "null", 4, 0);
		391
297	if (i < len)	392	if (i < len)
298	encode_comma (enc);	393	encode_comma (enc);
299	}	394	}
300		395
		396	encode_nl (enc); --enc->indent; encode_indent (enc);
		397	}
		398
301	encode_nl (enc);	399	encode_ch (enc, ']');
302
303	--enc->indent;
304	encode_indent (enc); encode_ch (enc, ']');
305	}	400	}
306		401
307	static void	402	static void
308	encode_he (enc_t *enc, HE *he)	403	encode_hk (enc_t *enc, HE *he)
309	{	404	{
310	encode_ch (enc, '"');	405	encode_ch (enc, '"');
311		406
312	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
313	{	408	{
…		…
323	else	418	else
324	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	419	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
325		420
326	encode_ch (enc, '"');	421	encode_ch (enc, '"');
327		422
328	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
329	encode_ch (enc, ':');	424	encode_ch (enc, ':');
330	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
331	encode_sv (enc, HeVAL (he));
332	}	426	}
333		427
334	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
335	static int	429	static int
336	he_cmp_fast (const void *a_, const void *b_)	430	he_cmp_fast (const void *a_, const void *b_)
…		…
341	HE *b = *(HE **)b_;	435	HE *b = *(HE **)b_;
342		436
343	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
344	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
345		439
346	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
347	cmp = la - lb;	441	cmp = lb - la;
348		442
349	return cmp;	443	return cmp;
350	}	444	}
351		445
352	// compare hash entries, used when some keys are sv's or utf-x	446	// compare hash entries, used when some keys are sv's or utf-x
353	static int	447	static int
354	he_cmp_slow (const void *a, const void *b)	448	he_cmp_slow (const void *a, const void *b)
355	{	449	{
356	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	450	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
357	}	451	}
358		452
359	static void	453	static void
360	encode_hv (enc_t *enc, HV *hv)	454	encode_hv (enc_t *enc, HV *hv)
361	{	455	{
362	int count, i;	456	HE *he;
363		457
364	if (enc->indent >= enc->maxdepth)	458	if (enc->indent >= enc->json.max_depth)
365	croak ("data structure too deep (hit recursion limit)");	459	croak (ERR_NESTING_EXCEEDED);
366		460
367	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	461	encode_ch (enc, '{');
368		462
369	if ((count = hv_iterinit (hv)))
370	{
371	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
372	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
373	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
374	// that randomises hash orderings	466	// that randomises hash orderings
375	if (enc->flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
		468	{
		469	int count = hv_iterinit (hv);
		470
		471	if (SvMAGICAL (hv))
376	{	472	{
		473	// need to count by iterating. could improve by dynamically building the vector below
		474	// but I don't care for the speed of this special case.
		475	// note also that we will run into undefined behaviour when the two iterations
		476	// do not result in the same count, something I might care for in some later release.
		477
		478	count = 0;
		479	while (hv_iternext (hv))
		480	++count;
		481
		482	hv_iterinit (hv);
		483	}
		484
		485	if (count)
		486	{
		487	int i, fast = 1;
		488	#if defined(__BORLANDC__) || defined(_MSC_VER)
		489	HE **hes = _alloca (count * sizeof (HE));
		490	#else
377	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	491	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
378	int fast = 1;	492	#endif
379		493
380	i = 0;	494	i = 0;
381	while ((he = hv_iternext (hv)))	495	while ((he = hv_iternext (hv)))
382	{	496	{
383	hes [i++] = he;	497	hes [i++] = he;
…		…
405		519
406	FREETMPS;	520	FREETMPS;
407	LEAVE;	521	LEAVE;
408	}	522	}
409		523
410	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
411	{	527	{
412	encode_indent (enc);	528	encode_indent (enc);
		529	he = hes [count];
413	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
414		532
415	if (i < count - 1)	533	if (count)
416	encode_comma (enc);	534	encode_comma (enc);
417	}	535	}
418		536
		537	encode_nl (enc); --enc->indent; encode_indent (enc);
		538	}
		539	}
		540	else
		541	{
		542	if (hv_iterinit (hv) || SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
		544	{
		545	encode_nl (enc); ++enc->indent;
		546
		547	for (;;)
		548	{
419	encode_nl (enc);	549	encode_indent (enc);
		550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		552
		553	if (!(he = hv_iternext (hv)))
		554	break;
		555
		556	encode_comma (enc);
		557	}
		558
		559	encode_nl (enc); --enc->indent; encode_indent (enc);
		560	}
		561	}
		562
		563	encode_ch (enc, '}');
		564	}
		565
		566	// encode objects, arrays and special \0=false and \1=true values.
		567	static void
		568	encode_rv (enc_t *enc, SV *sv)
		569	{
		570	svtype svt;
		571
		572	SvGETMAGIC (sv);
		573	svt = SvTYPE (sv);
		574
		575	if (expect_false (SvOBJECT (sv)))
		576	{
		577	HV *stash = !JSON_SLOW || json_boolean_stash
		578	? json_boolean_stash
		579	: gv_stashpv ("JSON::XS::Boolean", 1);
		580
		581	if (SvSTASH (sv) == stash)
		582	{
		583	if (SvIV (sv))
		584	encode_str (enc, "true", 4, 0);
		585	else
		586	encode_str (enc, "false", 5, 0);
420	}	587	}
421	else	588	else
422	{	589	{
423	HE *he = hv_iternext (hv);	590	#if 0
424		591	if (0 && sv_derived_from (rv, "JSON::Literal"))
425	for (;;)
426	{	592	{
427	encode_indent (enc);	593	// not yet
428	encode_he (enc, he);
429
430	if (!(he = hv_iternext (hv)))
431	break;
432
433	encode_comma (enc);
434	}	594	}
		595	#endif
		596	if (enc->json.flags & F_CONV_BLESSED)
		597	{
		598	// we re-bless the reference to get overload and other niceties right
		599	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
435		600
		601	if (to_json)
		602	{
		603	dSP;
		604
		605	ENTER; SAVETMPS; PUSHMARK (SP);
		606	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		607
		608	// calling with G_SCALAR ensures that we always get a 1 return value
		609	PUTBACK;
		610	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		611	SPAGAIN;
		612
		613	// catch this surprisingly common error
		614	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		615	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		616
		617	sv = POPs;
		618	PUTBACK;
		619
436	encode_nl (enc);	620	encode_sv (enc, sv);
		621
		622	FREETMPS; LEAVE;
		623	}
		624	else if (enc->json.flags & F_ALLOW_BLESSED)
		625	encode_str (enc, "null", 4, 0);
		626	else
		627	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		628	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		629	}
		630	else if (enc->json.flags & F_ALLOW_BLESSED)
		631	encode_str (enc, "null", 4, 0);
		632	else
		633	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		634	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
437	}	635	}
438	}	636	}
439
440	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
441	}
442
443	// encode objects, arrays and special \0=false and \1=true values.
444	static void
445	encode_rv (enc_t *enc, SV *sv)
446	{
447	svtype svt;
448
449	SvGETMAGIC (sv);
450	svt = SvTYPE (sv);
451
452	if (svt == SVt_PVHV)	637	else if (svt == SVt_PVHV)
453	encode_hv (enc, (HV *)sv);	638	encode_hv (enc, (HV *)sv);
454	else if (svt == SVt_PVAV)	639	else if (svt == SVt_PVAV)
455	encode_av (enc, (AV *)sv);	640	encode_av (enc, (AV *)sv);
456	else if (svt < SVt_PVAV)	641	else if (svt < SVt_PVAV)
457	{	642	{
458	if (SvNIOK (sv) && SvIV (sv) == 0)	643	STRLEN len = 0;
		644	char *pv = svt ? SvPV (sv, len) : 0;
		645
		646	if (len == 1 && *pv == '1')
		647	encode_str (enc, "true", 4, 0);
		648	else if (len == 1 && *pv == '0')
459	encode_str (enc, "false", 5, 0);	649	encode_str (enc, "false", 5, 0);
460	else if (SvNIOK (sv) && SvIV (sv) == 1)	650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
461	encode_str (enc, "true", 4, 0);	651	encode_str (enc, "null", 4, 0);
462	else	652	else
463	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
464	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
465	}	655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
466	else	658	else
467	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
468	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
469	}	661	}
470		662
…		…
481	encode_str (enc, str, len, SvUTF8 (sv));	673	encode_str (enc, str, len, SvUTF8 (sv));
482	encode_ch (enc, '"');	674	encode_ch (enc, '"');
483	}	675	}
484	else if (SvNOKp (sv))	676	else if (SvNOKp (sv))
485	{	677	{
		678	// trust that perl will do the right thing w.r.t. JSON syntax.
486	need (enc, NV_DIG + 32);	679	need (enc, NV_DIG + 32);
487	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
488	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
489	}	682	}
490	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
491	{	684	{
		685	// we assume we can always read an IV as a UV and vice versa
		686	// we assume two's complement
		687	// we assume no aliasing issues in the union
		688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		690	{
		691	// optimise the "small number case"
		692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
		694	I32 i = SvIVX (sv);
		695	U32 u;
		696	char digit, nz = 0;
		697
492	need (enc, 64);	698	need (enc, 6);
		699
		700	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		701	u = i < 0 ? -i : i;
		702
		703	// convert to 4.28 fixed-point representation
		704	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		705
		706	// now output digit by digit, each time masking out the integer part
		707	// and multiplying by 5 while moving the decimal point one to the right,
		708	// resulting in a net multiplication by 10.
		709	// we always write the digit to memory but conditionally increment
		710	// the pointer, to enable the use of conditional move instructions.
		711	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		712	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		713	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		714	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
493	enc->cur +=	721	enc->cur +=
494	SvIsUV(sv)	722	SvIsUV(sv)
495	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
496	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
497	}	726	}
498	else if (SvROK (sv))	727	else if (SvROK (sv))
499	encode_rv (enc, SvRV (sv));	728	encode_rv (enc, SvRV (sv));
500	else if (!SvOK (sv))	729	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
501	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
502	else	731	else
503	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
504	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
505	}	734	}
506		735
507	static SV *	736	static SV *
508	encode_json (SV *scalar, U32 flags)	737	encode_json (SV *scalar, JSON *json)
509	{	738	{
510	enc_t enc;	739	enc_t enc;
511		740
512	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	741	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
513	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	742	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
514		743
515	enc.flags = flags;	744	enc.json = *json;
516	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
517	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
518	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
519	enc.indent = 0;	748	enc.indent = 0;
520	enc.maxdepth = DEC_DEPTH (flags);	749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		750	: enc.json.flags & F_LATIN1 ? 0x000100UL
		751	: 0x110000UL;
521		752
522	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
523	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
524		755
525	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
526	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	757	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
527		758
528	if (!(flags & (F_ASCII | F_LATIN1 | F_UTF8)))	759	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
529	SvUTF8_on (enc.sv);	760	SvUTF8_on (enc.sv);
530		761
531	if (enc.flags & F_SHRINK)	762	if (enc.json.flags & F_SHRINK)
532	shrink (enc.sv);	763	shrink (enc.sv);
533		764
534	return enc.sv;	765	return enc.sv;
535	}	766	}
536		767
…		…
541	typedef struct	772	typedef struct
542	{	773	{
543	char *cur; // current parser pointer	774	char *cur; // current parser pointer
544	char *end; // end of input string	775	char *end; // end of input string
545	const char *err; // parse error, if != 0	776	const char *err; // parse error, if != 0
546	U32 flags; // F_*	777	JSON json;
547	U32 depth; // recursion depth	778	U32 depth; // recursion depth
548	U32 maxdepth; // recursion depth limit	779	U32 maxdepth; // recursion depth limit
549	} dec_t;	780	} dec_t;
550		781
551	static void	782	INLINE void
		783	decode_comment (dec_t *dec)
		784	{
		785	// only '#'-style comments allowed a.t.m.
		786
		787	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		788	++dec->cur;
		789	}
		790
		791	INLINE void
552	decode_ws (dec_t *dec)	792	decode_ws (dec_t *dec)
553	{	793	{
554	for (;;)	794	for (;;)
555	{	795	{
556	char ch = *dec->cur;	796	char ch = *dec->cur;
557		797
558	if (ch > 0x20	798	if (ch > 0x20)
		799	{
		800	if (expect_false (ch == '#'))
		801	{
		802	if (dec->json.flags & F_RELAXED)
		803	decode_comment (dec);
		804	else
		805	break;
		806	}
		807	else
		808	break;
		809	}
559	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
560	break;	811	break; // parse error, but let higher level handle it, gives better error messages
561		812
562	++dec->cur;	813	++dec->cur;
563	}	814	}
564	}	815	}
565		816
…		…
569	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
570	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
571	++dec->cur; \	822	++dec->cur; \
572	SE	823	SE
573		824
574	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
575	#define DEC_DEC_DEPTH --dec->depth	826	#define DEC_DEC_DEPTH --dec->depth
576		827
577	static SV *decode_sv (dec_t *dec);	828	static SV *decode_sv (dec_t *dec);
578		829
579	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
…		…
582	decode_4hex (dec_t *dec)	833	decode_4hex (dec_t *dec)
583	{	834	{
584	signed char d1, d2, d3, d4;	835	signed char d1, d2, d3, d4;
585	unsigned char *cur = (unsigned char *)dec->cur;	836	unsigned char *cur = (unsigned char *)dec->cur;
586		837
587	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	838	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
588	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	839	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
589	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	840	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
590	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	841	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
591		842
592	dec->cur += 4;	843	dec->cur += 4;
593		844
594	return ((UV)d1) << 12	845	return ((UV)d1) << 12
595	| ((UV)d2) << 8	846	| ((UV)d2) << 8
…		…
603	static SV *	854	static SV *
604	decode_str (dec_t *dec)	855	decode_str (dec_t *dec)
605	{	856	{
606	SV *sv = 0;	857	SV *sv = 0;
607	int utf8 = 0;	858	int utf8 = 0;
		859	char *dec_cur = dec->cur;
608		860
609	do	861	do
610	{	862	{
611	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	863	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
612	char *cur = buf;	864	char *cur = buf;
613		865
614	do	866	do
615	{	867	{
616	unsigned char ch = *(unsigned char *)dec->cur++;	868	unsigned char ch = *(unsigned char *)dec_cur++;
617		869
618	if (ch == '"')	870	if (expect_false (ch == '"'))
619	{	871	{
620	--dec->cur;	872	--dec_cur;
621	break;	873	break;
622	}	874	}
623	else if (ch == '\\')	875	else if (expect_false (ch == '\\'))
624	{	876	{
625	switch (*dec->cur)	877	switch (*dec_cur)
626	{	878	{
627	case '\\':	879	case '\\':
628	case '/':	880	case '/':
629	case '"': *cur++ = *dec->cur++; break;	881	case '"': *cur++ = *dec_cur++; break;
630		882
631	case 'b': ++dec->cur; *cur++ = '\010'; break;	883	case 'b': ++dec_cur; *cur++ = '\010'; break;
632	case 't': ++dec->cur; *cur++ = '\011'; break;	884	case 't': ++dec_cur; *cur++ = '\011'; break;
633	case 'n': ++dec->cur; *cur++ = '\012'; break;	885	case 'n': ++dec_cur; *cur++ = '\012'; break;
634	case 'f': ++dec->cur; *cur++ = '\014'; break;	886	case 'f': ++dec_cur; *cur++ = '\014'; break;
635	case 'r': ++dec->cur; *cur++ = '\015'; break;	887	case 'r': ++dec_cur; *cur++ = '\015'; break;
636		888
637	case 'u':	889	case 'u':
638	{	890	{
639	UV lo, hi;	891	UV lo, hi;
640	++dec->cur;	892	++dec_cur;
641		893
		894	dec->cur = dec_cur;
642	hi = decode_4hex (dec);	895	hi = decode_4hex (dec);
		896	dec_cur = dec->cur;
643	if (hi == (UV)-1)	897	if (hi == (UV)-1)
644	goto fail;	898	goto fail;
645		899
646	// possibly a surrogate pair	900	// possibly a surrogate pair
647	if (hi >= 0xd800)	901	if (hi >= 0xd800)
648	if (hi < 0xdc00)	902	if (hi < 0xdc00)
649	{	903	{
650	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	904	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
651	ERR ("missing low surrogate character in surrogate pair");	905	ERR ("missing low surrogate character in surrogate pair");
652		906
653	dec->cur += 2;	907	dec_cur += 2;
654		908
		909	dec->cur = dec_cur;
655	lo = decode_4hex (dec);	910	lo = decode_4hex (dec);
		911	dec_cur = dec->cur;
656	if (lo == (UV)-1)	912	if (lo == (UV)-1)
657	goto fail;	913	goto fail;
658		914
659	if (lo < 0xdc00 || lo >= 0xe000)	915	if (lo < 0xdc00 || lo >= 0xe000)
660	ERR ("surrogate pair expected");	916	ERR ("surrogate pair expected");
…		…
666		922
667	if (hi >= 0x80)	923	if (hi >= 0x80)
668	{	924	{
669	utf8 = 1;	925	utf8 = 1;
670		926
671	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	927	cur = encode_utf8 (cur, hi);
672	}	928	}
673	else	929	else
674	*cur++ = hi;	930	*cur++ = hi;
675	}	931	}
676	break;	932	break;
677		933
678	default:	934	default:
679	--dec->cur;	935	--dec_cur;
680	ERR ("illegal backslash escape sequence in string");	936	ERR ("illegal backslash escape sequence in string");
681	}	937	}
682	}	938	}
683	else if (ch >= 0x20 && ch <= 0x7f)	939	else if (expect_true (ch >= 0x20 && ch < 0x80))
684	*cur++ = ch;	940	*cur++ = ch;
685	else if (ch >= 0x80)	941	else if (ch >= 0x80)
686	{	942	{
687	STRLEN clen;	943	STRLEN clen;
688	UV uch;	944	UV uch;
689		945
690	--dec->cur;	946	--dec_cur;
691		947
692	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	948	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
693	if (clen == (STRLEN)-1)	949	if (clen == (STRLEN)-1)
694	ERR ("malformed UTF-8 character in JSON string");	950	ERR ("malformed UTF-8 character in JSON string");
695		951
696	do	952	do
697	*cur++ = *dec->cur++;	953	*cur++ = *dec_cur++;
698	while (--clen);	954	while (--clen);
699		955
700	utf8 = 1;	956	utf8 = 1;
701	}	957	}
702	else	958	else
703	{	959	{
704	--dec->cur;	960	--dec_cur;
705		961
706	if (!ch)	962	if (!ch)
707	ERR ("unexpected end of string while parsing JSON string");	963	ERR ("unexpected end of string while parsing JSON string");
708	else	964	else
709	ERR ("invalid character encountered while parsing JSON string");	965	ERR ("invalid character encountered while parsing JSON string");
…		…
714	{	970	{
715	STRLEN len = cur - buf;	971	STRLEN len = cur - buf;
716		972
717	if (sv)	973	if (sv)
718	{	974	{
719	SvGROW (sv, SvCUR (sv) + len + 1);	975	STRLEN cur = SvCUR (sv);
		976
		977	if (SvLEN (sv) <= cur + len)
		978	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		979
720	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	980	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
721	SvCUR_set (sv, SvCUR (sv) + len);	981	SvCUR_set (sv, SvCUR (sv) + len);
722	}	982	}
723	else	983	else
724	sv = newSVpvn (buf, len);	984	sv = newSVpvn (buf, len);
725	}	985	}
726	}	986	}
727	while (*dec->cur != '"');	987	while (*dec_cur != '"');
728		988
729	++dec->cur;	989	++dec_cur;
730		990
731	if (sv)	991	if (sv)
732	{	992	{
733	SvPOK_only (sv);	993	SvPOK_only (sv);
734	*SvEND (sv) = 0;	994	*SvEND (sv) = 0;
…		…
737	SvUTF8_on (sv);	997	SvUTF8_on (sv);
738	}	998	}
739	else	999	else
740	sv = newSVpvn ("", 0);	1000	sv = newSVpvn ("", 0);
741		1001
		1002	dec->cur = dec_cur;
742	return sv;	1003	return sv;
743		1004
744	fail:	1005	fail:
		1006	dec->cur = dec_cur;
745	return 0;	1007	return 0;
746	}	1008	}
747		1009
748	static SV *	1010	static SV *
749	decode_num (dec_t *dec)	1011	decode_num (dec_t *dec)
…		…
807	is_nv = 1;	1069	is_nv = 1;
808	}	1070	}
809		1071
810	if (!is_nv)	1072	if (!is_nv)
811	{	1073	{
812	UV uv;	1074	int len = dec->cur - start;
813	int numtype = grok_number (start, dec->cur - start, &uv);	1075
814	if (numtype & IS_NUMBER_IN_UV)	1076	// special case the rather common 1..5-digit-int case
815	if (numtype & IS_NUMBER_NEG)	1077	if (*start == '-')
		1078	switch (len)
816	{	1079	{
817	if (uv < (UV)IV_MIN)	1080	case 2: return newSViv (-( start [1] - '0' * 1));
818	return newSViv (-(IV)uv);	1081	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1082	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1083	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1084	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
819	}	1085	}
		1086	else
		1087	switch (len)
		1088	{
		1089	case 1: return newSViv ( start [0] - '0' * 1);
		1090	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1091	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1092	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1093	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1094	}
		1095
		1096	{
		1097	UV uv;
		1098	int numtype = grok_number (start, len, &uv);
		1099	if (numtype & IS_NUMBER_IN_UV)
		1100	if (numtype & IS_NUMBER_NEG)
		1101	{
		1102	if (uv < (UV)IV_MIN)
		1103	return newSViv (-(IV)uv);
		1104	}
820	else	1105	else
821	return newSVuv (uv);	1106	return newSVuv (uv);
822	}	1107	}
823		1108
		1109	len -= *start == '-' ? 1 : 0;
		1110
		1111	// does not fit into IV or UV, try NV
		1112	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1113	#if defined (LDBL_DIG)
		1114	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1115	#endif
		1116	)
		1117	// fits into NV without loss of precision
		1118	return newSVnv (Atof (start));
		1119
		1120	// everything else fails, convert it to a string
		1121	return newSVpvn (start, dec->cur - start);
		1122	}
		1123
		1124	// loss of precision here
824	return newSVnv (Atof (start));	1125	return newSVnv (Atof (start));
825		1126
826	fail:	1127	fail:
827	return 0;	1128	return 0;
828	}	1129	}
…		…
858		1159
859	if (*dec->cur != ',')	1160	if (*dec->cur != ',')
860	ERR (", or ] expected while parsing array");	1161	ERR (", or ] expected while parsing array");
861		1162
862	++dec->cur;	1163	++dec->cur;
		1164
		1165	decode_ws (dec);
		1166
		1167	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1168	{
		1169	++dec->cur;
		1170	break;
		1171	}
863	}	1172	}
864		1173
865	DEC_DEC_DEPTH;	1174	DEC_DEC_DEPTH;
866	return newRV_noinc ((SV *)av);	1175	return newRV_noinc ((SV *)av);
867		1176
…		…
872	}	1181	}
873		1182
874	static SV *	1183	static SV *
875	decode_hv (dec_t *dec)	1184	decode_hv (dec_t *dec)
876	{	1185	{
		1186	SV *sv;
877	HV *hv = newHV ();	1187	HV *hv = newHV ();
878		1188
879	DEC_INC_DEPTH;	1189	DEC_INC_DEPTH;
880	decode_ws (dec);	1190	decode_ws (dec);
881		1191
882	if (*dec->cur == '}')	1192	if (*dec->cur == '}')
883	++dec->cur;	1193	++dec->cur;
884	else	1194	else
885	for (;;)	1195	for (;;)
886	{	1196	{
		1197	EXPECT_CH ('"');
		1198
		1199	// heuristic: assume that
		1200	// a) decode_str + hv_store_ent are abysmally slow.
		1201	// b) most hash keys are short, simple ascii text.
		1202	// => try to "fast-match" such strings to avoid
		1203	// the overhead of decode_str + hv_store_ent.
		1204	{
887	SV *key, *value;	1205	SV *value;
		1206	char *p = dec->cur;
		1207	char *e = p + 24; // only try up to 24 bytes
888		1208
889	decode_ws (dec); EXPECT_CH ('"');	1209	for (;;)
890
891	key = decode_str (dec);
892	if (!key)
893	goto fail;
894
895	decode_ws (dec); EXPECT_CH (':');
896
897	value = decode_sv (dec);
898	if (!value)
899	{	1210	{
		1211	// the >= 0x80 is false on most architectures
		1212	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1213	{
		1214	// slow path, back up and use decode_str
		1215	SV *key = decode_str (dec);
		1216	if (!key)
		1217	goto fail;
		1218
		1219	decode_ws (dec); EXPECT_CH (':');
		1220
		1221	decode_ws (dec);
		1222	value = decode_sv (dec);
		1223	if (!value)
		1224	{
		1225	SvREFCNT_dec (key);
		1226	goto fail;
		1227	}
		1228
		1229	hv_store_ent (hv, key, value, 0);
900	SvREFCNT_dec (key);	1230	SvREFCNT_dec (key);
		1231
		1232	break;
		1233	}
		1234	else if (*p == '"')
		1235	{
		1236	// fast path, got a simple key
		1237	char *key = dec->cur;
		1238	int len = p - key;
		1239	dec->cur = p + 1;
		1240
		1241	decode_ws (dec); EXPECT_CH (':');
		1242
		1243	decode_ws (dec);
		1244	value = decode_sv (dec);
		1245	if (!value)
901	goto fail;	1246	goto fail;
		1247
		1248	hv_store (hv, key, len, value, 0);
		1249
		1250	break;
		1251	}
		1252
		1253	++p;
902	}	1254	}
903		1255	}
904	hv_store_ent (hv, key, value, 0);
905	SvREFCNT_dec (key);
906		1256
907	decode_ws (dec);	1257	decode_ws (dec);
908		1258
909	if (*dec->cur == '}')	1259	if (*dec->cur == '}')
910	{	1260	{
…		…
914		1264
915	if (*dec->cur != ',')	1265	if (*dec->cur != ',')
916	ERR (", or } expected while parsing object/hash");	1266	ERR (", or } expected while parsing object/hash");
917		1267
918	++dec->cur;	1268	++dec->cur;
		1269
		1270	decode_ws (dec);
		1271
		1272	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1273	{
		1274	++dec->cur;
		1275	break;
		1276	}
919	}	1277	}
920		1278
921	DEC_DEC_DEPTH;	1279	DEC_DEC_DEPTH;
922	return newRV_noinc ((SV *)hv);	1280	sv = newRV_noinc ((SV *)hv);
		1281
		1282	// check filter callbacks
		1283	if (dec->json.flags & F_HOOK)
		1284	{
		1285	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1286	{
		1287	HE *cb, *he;
		1288
		1289	hv_iterinit (hv);
		1290	he = hv_iternext (hv);
		1291	hv_iterinit (hv);
		1292
		1293	// the next line creates a mortal sv each time its called.
		1294	// might want to optimise this for common cases.
		1295	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1296
		1297	if (cb)
		1298	{
		1299	dSP;
		1300	int count;
		1301
		1302	ENTER; SAVETMPS; PUSHMARK (SP);
		1303	XPUSHs (HeVAL (he));
		1304
		1305	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1306
		1307	if (count == 1)
		1308	{
		1309	sv = newSVsv (POPs);
		1310	FREETMPS; LEAVE;
		1311	return sv;
		1312	}
		1313
		1314	FREETMPS; LEAVE;
		1315	}
		1316	}
		1317
		1318	if (dec->json.cb_object)
		1319	{
		1320	dSP;
		1321	int count;
		1322
		1323	ENTER; SAVETMPS; PUSHMARK (SP);
		1324	XPUSHs (sv_2mortal (sv));
		1325
		1326	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1327
		1328	if (count == 1)
		1329	{
		1330	sv = newSVsv (POPs);
		1331	FREETMPS; LEAVE;
		1332	return sv;
		1333	}
		1334
		1335	SvREFCNT_inc (sv);
		1336	FREETMPS; LEAVE;
		1337	}
		1338	}
		1339
		1340	return sv;
923		1341
924	fail:	1342	fail:
925	SvREFCNT_dec (hv);	1343	SvREFCNT_dec (hv);
926	DEC_DEC_DEPTH;	1344	DEC_DEC_DEPTH;
927	return 0;	1345	return 0;
928	}	1346	}
929		1347
930	static SV *	1348	static SV *
931	decode_sv (dec_t *dec)	1349	decode_sv (dec_t *dec)
932	{	1350	{
933	decode_ws (dec);	1351	// the beauty of JSON: you need exactly one character lookahead
		1352	// to parse everything.
934	switch (*dec->cur)	1353	switch (*dec->cur)
935	{	1354	{
936	case '"': ++dec->cur; return decode_str (dec);	1355	case '"': ++dec->cur; return decode_str (dec);
937	case '[': ++dec->cur; return decode_av (dec);	1356	case '[': ++dec->cur; return decode_av (dec);
938	case '{': ++dec->cur; return decode_hv (dec);	1357	case '{': ++dec->cur; return decode_hv (dec);
939		1358
940	case '-':	1359	case '-':
941	case '0': case '1': case '2': case '3': case '4':	1360	case '0': case '1': case '2': case '3': case '4':
942	case '5': case '6': case '7': case '8': case '9':	1361	case '5': case '6': case '7': case '8': case '9':
943	return decode_num (dec);	1362	return decode_num (dec);
944		1363
945	case 't':	1364	case 't':
946	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1365	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
947	{	1366	{
948	dec->cur += 4;	1367	dec->cur += 4;
		1368	#if JSON_SLOW
		1369	json_true = get_bool ("JSON::XS::true");
		1370	#endif
949	return newSViv (1);	1371	return newSVsv (json_true);
950	}	1372	}
951	else	1373	else
952	ERR ("'true' expected");	1374	ERR ("'true' expected");
953		1375
954	break;	1376	break;
955		1377
956	case 'f':	1378	case 'f':
957	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1379	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
958	{	1380	{
959	dec->cur += 5;	1381	dec->cur += 5;
		1382	#if JSON_SLOW
		1383	json_false = get_bool ("JSON::XS::false");
		1384	#endif
960	return newSViv (0);	1385	return newSVsv (json_false);
961	}	1386	}
962	else	1387	else
963	ERR ("'false' expected");	1388	ERR ("'false' expected");
964		1389
965	break;	1390	break;
…		…
983	fail:	1408	fail:
984	return 0;	1409	return 0;
985	}	1410	}
986		1411
987	static SV *	1412	static SV *
988	decode_json (SV *string, U32 flags, UV *offset_return)	1413	decode_json (SV *string, JSON *json, STRLEN *offset_return)
989	{	1414	{
990	dec_t dec;	1415	dec_t dec;
991	UV offset;	1416	STRLEN offset;
992	SV *sv;	1417	SV *sv;
993		1418
		1419	/* work around bugs in 5.10 where manipulating magic values
		1420	* will perl ignore the magic in subsequent accesses
		1421	*/
994	SvGETMAGIC (string);	1422	/*SvGETMAGIC (string);*/
		1423	if (SvMAGICAL (string))
		1424	string = sv_2mortal (newSVsv (string));
		1425
995	SvUPGRADE (string, SVt_PV);	1426	SvUPGRADE (string, SVt_PV);
996		1427
		1428	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1429	* assertion with -DDEBUGGING, although SvCUR is documented to
		1430	* return the xpv_cur field which certainly exists after upgrading.
		1431	* according to nicholas clark, calling SvPOK fixes this.
		1432	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1433	* and hope for the best.
		1434	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1435	* assertion business is seriously broken, try yet another workaround
		1436	* for the broken -DDEBUGGING.
		1437	*/
		1438	#ifdef DEBUGGING
		1439	offset = SvOK (string) ? sv_len (string) : 0;
		1440	#else
		1441	offset = SvCUR (string);
		1442	#endif
		1443
		1444	if (offset > json->max_size && json->max_size)
		1445	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1446	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1447
997	if (flags & F_UTF8)	1448	if (json->flags & F_UTF8)
998	sv_utf8_downgrade (string, 0);	1449	sv_utf8_downgrade (string, 0);
999	else	1450	else
1000	sv_utf8_upgrade (string);	1451	sv_utf8_upgrade (string);
1001		1452
1002	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1453	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1003		1454
1004	dec.flags = flags;	1455	dec.json = *json;
1005	dec.cur = SvPVX (string);	1456	dec.cur = SvPVX (string);
1006	dec.end = SvEND (string);	1457	dec.end = SvEND (string);
1007	dec.err = 0;	1458	dec.err = 0;
1008	dec.depth = 0;	1459	dec.depth = 0;
1009	dec.maxdepth = DEC_DEPTH (dec.flags);	1460
		1461	if (dec.json.cb_object || dec.json.cb_sk_object)
		1462	dec.json.flags |= F_HOOK;
1010		1463
1011	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1464	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1465
		1466	decode_ws (&dec);
1012	sv = decode_sv (&dec);	1467	sv = decode_sv (&dec);
1013		1468
1014	if (!(offset_return || !sv))	1469	if (!(offset_return || !sv))
1015	{	1470	{
1016	// check for trailing garbage	1471	// check for trailing garbage
…		…
1024	}	1479	}
1025	}	1480	}
1026		1481
1027	if (offset_return || !sv)	1482	if (offset_return || !sv)
1028	{	1483	{
1029	offset = dec.flags & F_UTF8	1484	offset = dec.json.flags & F_UTF8
1030	? dec.cur - SvPVX (string)	1485	? dec.cur - SvPVX (string)
1031	: utf8_distance (dec.cur, SvPVX (string));	1486	: utf8_distance (dec.cur, SvPVX (string));
1032		1487
1033	if (offset_return)	1488	if (offset_return)
1034	*offset_return = offset;	1489	*offset_return = offset;
…		…
1053	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1508	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1054	}	1509	}
1055		1510
1056	sv = sv_2mortal (sv);	1511	sv = sv_2mortal (sv);
1057		1512
1058	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1513	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1059	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1514	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1060		1515
1061	return sv;	1516	return sv;
		1517	}
		1518
		1519	/////////////////////////////////////////////////////////////////////////////
		1520	// incremental parser
		1521
		1522	static void
		1523	incr_parse (JSON *self)
		1524	{
		1525	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1526
		1527	for (;;)
		1528	{
		1529	//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
		1530	switch (self->incr_mode)
		1531	{
		1532	// only used for intiial whitespace skipping
		1533	case INCR_M_WS:
		1534	for (;;)
		1535	{
		1536	if (*p > 0x20)
		1537	{
		1538	self->incr_mode = INCR_M_JSON;
		1539	goto incr_m_json;
		1540	}
		1541	else if (!*p)
		1542	goto interrupt;
		1543
		1544	++p;
		1545	}
		1546
		1547	// skip a single char inside a string (for \\-processing)
		1548	case INCR_M_BS:
		1549	if (!*p)
		1550	goto interrupt;
		1551
		1552	++p;
		1553	self->incr_mode = INCR_M_STR;
		1554	goto incr_m_str;
		1555
		1556	// inside a string
		1557	case INCR_M_STR:
		1558	incr_m_str:
		1559	for (;;)
		1560	{
		1561	if (*p == '"')
		1562	{
		1563	++p;
		1564	self->incr_mode = INCR_M_JSON;
		1565
		1566	if (!self->incr_nest)
		1567	goto interrupt;
		1568
		1569	goto incr_m_json;
		1570	}
		1571	else if (*p == '\\')
		1572	{
		1573	++p; // "virtually" consumes character after \
		1574
		1575	if (!*p) // if at end of string we have to switch modes
		1576	{
		1577	self->incr_mode = INCR_M_BS;
		1578	goto interrupt;
		1579	}
		1580	}
		1581	else if (!*p)
		1582	goto interrupt;
		1583
		1584	++p;
		1585	}
		1586
		1587	// after initial ws, outside string
		1588	case INCR_M_JSON:
		1589	incr_m_json:
		1590	for (;;)
		1591	{
		1592	switch (*p++)
		1593	{
		1594	case 0:
		1595	--p;
		1596	goto interrupt;
		1597
		1598	case 0x09:
		1599	case 0x0a:
		1600	case 0x0d:
		1601	case 0x20:
		1602	if (!self->incr_nest)
		1603	{
		1604	--p; // do not eat the whitespace, let the next round do it
		1605	goto interrupt;
		1606	}
		1607	break;
		1608
		1609	case '"':
		1610	self->incr_mode = INCR_M_STR;
		1611	goto incr_m_str;
		1612
		1613	case '[':
		1614	case '{':
		1615	if (++self->incr_nest > self->max_depth)
		1616	croak (ERR_NESTING_EXCEEDED);
		1617	break;
		1618
		1619	case ']':
		1620	case '}':
		1621	if (--self->incr_nest <= 0)
		1622	goto interrupt;
		1623	}
		1624	}
		1625	}
		1626
		1627	modechange:
		1628	;
		1629	}
		1630
		1631	interrupt:
		1632	self->incr_pos = p - SvPVX (self->incr_text);
		1633	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1062	}	1634	}
1063		1635
1064	/////////////////////////////////////////////////////////////////////////////	1636	/////////////////////////////////////////////////////////////////////////////
1065	// XS interface functions	1637	// XS interface functions
1066		1638
1067	MODULE = JSON::XS PACKAGE = JSON::XS	1639	MODULE = JSON::XS PACKAGE = JSON::XS
1068		1640
1069	BOOT:	1641	BOOT:
1070	{	1642	{
1071	int i;	1643	int i;
1072
1073	memset (decode_hexdigit, 0xff, 256);
1074		1644
1075	for (i = 0; i < 256; ++i)	1645	for (i = 0; i < 256; ++i)
1076	decode_hexdigit [i] =	1646	decode_hexdigit [i] =
1077	i >= '0' && i <= '9' ? i - '0'	1647	i >= '0' && i <= '9' ? i - '0'
1078	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1648	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1079	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1649	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1080	: -1;	1650	: -1;
1081		1651
1082	json_stash = gv_stashpv ("JSON::XS", 1);	1652	json_stash = gv_stashpv ("JSON::XS" , 1);
		1653	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1654
		1655	json_true = get_bool ("JSON::XS::true");
		1656	json_false = get_bool ("JSON::XS::false");
1083	}	1657	}
1084		1658
1085	PROTOTYPES: DISABLE	1659	PROTOTYPES: DISABLE
1086		1660
1087	SV *new (char *dummy)	1661	void CLONE (...)
1088	CODE:	1662	CODE:
1089	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1663	json_stash = 0;
		1664	json_boolean_stash = 0;
		1665
		1666	void new (char *klass)
		1667	PPCODE:
		1668	{
		1669	SV *pv = NEWSV (0, sizeof (JSON));
		1670	SvPOK_only (pv);
		1671	json_init ((JSON *)SvPVX (pv));
		1672	XPUSHs (sv_2mortal (sv_bless (
		1673	newRV_noinc (pv),
		1674	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1675	)));
		1676	}
		1677
		1678	void ascii (JSON *self, int enable = 1)
		1679	ALIAS:
		1680	ascii = F_ASCII
		1681	latin1 = F_LATIN1
		1682	utf8 = F_UTF8
		1683	indent = F_INDENT
		1684	canonical = F_CANONICAL
		1685	space_before = F_SPACE_BEFORE
		1686	space_after = F_SPACE_AFTER
		1687	pretty = F_PRETTY
		1688	allow_nonref = F_ALLOW_NONREF
		1689	shrink = F_SHRINK
		1690	allow_blessed = F_ALLOW_BLESSED
		1691	convert_blessed = F_CONV_BLESSED
		1692	relaxed = F_RELAXED
		1693	allow_unknown = F_ALLOW_UNKNOWN
		1694	PPCODE:
		1695	{
		1696	if (enable)
		1697	self->flags |= ix;
		1698	else
		1699	self->flags &= ~ix;
		1700
		1701	XPUSHs (ST (0));
		1702	}
		1703
		1704	void get_ascii (JSON *self)
		1705	ALIAS:
		1706	get_ascii = F_ASCII
		1707	get_latin1 = F_LATIN1
		1708	get_utf8 = F_UTF8
		1709	get_indent = F_INDENT
		1710	get_canonical = F_CANONICAL
		1711	get_space_before = F_SPACE_BEFORE
		1712	get_space_after = F_SPACE_AFTER
		1713	get_allow_nonref = F_ALLOW_NONREF
		1714	get_shrink = F_SHRINK
		1715	get_allow_blessed = F_ALLOW_BLESSED
		1716	get_convert_blessed = F_CONV_BLESSED
		1717	get_relaxed = F_RELAXED
		1718	get_allow_unknown = F_ALLOW_UNKNOWN
		1719	PPCODE:
		1720	XPUSHs (boolSV (self->flags & ix));
		1721
		1722	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1723	PPCODE:
		1724	self->max_depth = max_depth;
		1725	XPUSHs (ST (0));
		1726
		1727	U32 get_max_depth (JSON *self)
		1728	CODE:
		1729	RETVAL = self->max_depth;
1090	OUTPUT:	1730	OUTPUT:
1091	RETVAL	1731	RETVAL
1092		1732
1093	SV *ascii (SV *self, int enable = 1)	1733	void max_size (JSON *self, U32 max_size = 0)
1094	ALIAS:	1734	PPCODE:
1095	ascii = F_ASCII	1735	self->max_size = max_size;
1096	latin1 = F_LATIN1	1736	XPUSHs (ST (0));
1097	utf8 = F_UTF8	1737
1098	indent = F_INDENT	1738	int get_max_size (JSON *self)
1099	canonical = F_CANONICAL
1100	space_before = F_SPACE_BEFORE
1101	space_after = F_SPACE_AFTER
1102	pretty = F_PRETTY
1103	allow_nonref = F_ALLOW_NONREF
1104	shrink = F_SHRINK
1105	CODE:	1739	CODE:
1106	{	1740	RETVAL = self->max_size;
1107	UV *uv = SvJSON (self);
1108	if (enable)
1109	*uv |= ix;
1110	else
1111	*uv &= ~ix;
1112
1113	RETVAL = newSVsv (self);
1114	}
1115	OUTPUT:	1741	OUTPUT:
1116	RETVAL	1742	RETVAL
1117		1743
1118	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1744	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1745	PPCODE:
		1746	{
		1747	SvREFCNT_dec (self->cb_object);
		1748	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1749
		1750	XPUSHs (ST (0));
		1751	}
		1752
		1753	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1754	PPCODE:
		1755	{
		1756	if (!self->cb_sk_object)
		1757	self->cb_sk_object = newHV ();
		1758
		1759	if (SvOK (cb))
		1760	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1761	else
		1762	{
		1763	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1764
		1765	if (!HvKEYS (self->cb_sk_object))
		1766	{
		1767	SvREFCNT_dec (self->cb_sk_object);
		1768	self->cb_sk_object = 0;
		1769	}
		1770	}
		1771
		1772	XPUSHs (ST (0));
		1773	}
		1774
		1775	void encode (JSON *self, SV *scalar)
		1776	PPCODE:
		1777	XPUSHs (encode_json (scalar, self));
		1778
		1779	void decode (JSON *self, SV *jsonstr)
		1780	PPCODE:
		1781	XPUSHs (decode_json (jsonstr, self, 0));
		1782
		1783	void decode_prefix (JSON *self, SV *jsonstr)
		1784	PPCODE:
		1785	{
		1786	STRLEN offset;
		1787	EXTEND (SP, 2);
		1788	PUSHs (decode_json (jsonstr, self, &offset));
		1789	PUSHs (sv_2mortal (newSVuv (offset)));
		1790	}
		1791
		1792	void incr_parse (JSON *self, SV *jsonstr = 0)
		1793	PPCODE:
		1794	{
		1795	if (!self->incr_text)
		1796	self->incr_text = newSVpvn ("", 0);
		1797
		1798	// append data, if any
		1799	if (jsonstr)
		1800	{
		1801	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1802	{
		1803	/* utf-8-ness differs, need to upgrade */
		1804	sv_utf8_upgrade (self->incr_text);
		1805
		1806	if (self->incr_pos)
		1807	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1808	- (U8 *)SvPVX (self->incr_text);
		1809	}
		1810
		1811	{
		1812	STRLEN len;
		1813	const char *str = SvPV (jsonstr, len);
		1814	STRLEN cur = SvCUR (self->incr_text);
		1815
		1816	if (SvLEN (self->incr_text) <= cur + len)
		1817	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1818
		1819	Move (str, SvEND (self->incr_text), len, char);
		1820	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1821	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1822	}
		1823	}
		1824
		1825	if (GIMME_V != G_VOID)
		1826	do
		1827	{
		1828	STRLEN offset;
		1829
		1830	if (!INCR_DONE (self))
		1831	{
		1832	incr_parse (self);
		1833
		1834	if (self->incr_pos > self->max_size && self->max_size)
		1835	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1836	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1837
		1838	if (!INCR_DONE (self))
		1839	break;
		1840	}
		1841
		1842	XPUSHs (decode_json (self->incr_text, self, &offset));
		1843
		1844	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1845	self->incr_pos -= offset;
		1846	self->incr_nest = 0;
		1847	self->incr_mode = 0;
		1848	}
		1849	while (GIMME_V == G_ARRAY);
		1850	}
		1851
		1852	SV *incr_text (JSON *self)
		1853	ATTRS: lvalue
1119	CODE:	1854	CODE:
1120	{	1855	{
1121	UV *uv = SvJSON (self);	1856	if (self->incr_pos)
1122	UV log2 = 0;	1857	croak ("incr_text can not be called when the incremental parser already started parsing");
1123		1858
1124	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1859	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1125
1126	while ((1UL << log2) < max_depth)
1127	++log2;
1128
1129	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1130
1131	RETVAL = newSVsv (self);
1132	}	1860	}
1133	OUTPUT:	1861	OUTPUT:
1134	RETVAL	1862	RETVAL
1135		1863
		1864	void incr_skip (JSON *self)
		1865	CODE:
		1866	{
		1867	if (self->incr_pos)
		1868	{
		1869	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1870	self->incr_pos = 0;
		1871	self->incr_nest = 0;
		1872	self->incr_mode = 0;
		1873	}
		1874	}
		1875
		1876	void incr_reset (JSON *self)
		1877	CODE:
		1878	{
		1879	SvREFCNT_dec (self->incr_text);
		1880	self->incr_text = 0;
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
		1884	}
		1885
		1886	void DESTROY (JSON *self)
		1887	CODE:
		1888	SvREFCNT_dec (self->cb_sk_object);
		1889	SvREFCNT_dec (self->cb_object);
		1890	SvREFCNT_dec (self->incr_text);
		1891
		1892	PROTOTYPES: ENABLE
		1893
1136	void encode (SV *self, SV *scalar)	1894	void encode_json (SV *scalar)
		1895	ALIAS:
		1896	to_json_ = 0
		1897	encode_json = F_UTF8
1137	PPCODE:	1898	PPCODE:
		1899	{
		1900	JSON json;
		1901	json_init (&json);
		1902	json.flags |= ix;
1138	XPUSHs (encode_json (scalar, *SvJSON (self)));	1903	XPUSHs (encode_json (scalar, &json));
		1904	}
1139		1905
1140	void decode (SV *self, SV *jsonstr)	1906	void decode_json (SV *jsonstr)
		1907	ALIAS:
		1908	from_json_ = 0
		1909	decode_json = F_UTF8
1141	PPCODE:	1910	PPCODE:
		1911	{
		1912	JSON json;
		1913	json_init (&json);
		1914	json.flags |= ix;
1142	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1915	XPUSHs (decode_json (jsonstr, &json, 0));
1143
1144	void decode_prefix (SV *self, SV *jsonstr)
1145	PPCODE:
1146	{
1147	UV offset;
1148	EXTEND (SP, 2);
1149	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
1150	PUSHs (sv_2mortal (newSVuv (offset)));
1151	}	1916	}
1152		1917
1153	PROTOTYPES: ENABLE
1154		1918
1155	void to_json (SV *scalar)
1156	ALIAS:
1157	objToJson = 0
1158	PPCODE:
1159	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1160
1161	void from_json (SV *jsonstr)
1162	ALIAS:
1163	jsonToObj = 0
1164	PPCODE:
1165	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8, 0));
1166

Diff Legend

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