ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.26 by root, Fri Apr 6 21:17:09 2007 UTC vs.
Revision 1.102 by root, Sat Oct 10 01:48:50 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	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
14	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
15	#define F_UTF8 0x00000002UL	27	#define F_UTF8 0x00000004UL
16	#define F_INDENT 0x00000004UL	28	#define F_INDENT 0x00000008UL
17	#define F_CANONICAL 0x00000008UL	29	#define F_CANONICAL 0x00000010UL
18	#define F_SPACE_BEFORE 0x00000010UL	30	#define F_SPACE_BEFORE 0x00000020UL
19	#define F_SPACE_AFTER 0x00000020UL	31	#define F_SPACE_AFTER 0x00000040UL
20	#define F_ALLOW_NONREF 0x00000080UL	32	#define F_ALLOW_NONREF 0x00000100UL
21	#define F_SHRINK 0x00000100UL	33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
22	#define F_MAXDEPTH 0xf8000000UL	36	#define F_RELAXED 0x00001000UL
23	#define S_MAXDEPTH 27	37	#define F_ALLOW_UNKNOWN 0x00002000UL
24		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
25	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
26
27	// F_SELFCONVERT? <=> to_json/toJson
28	// F_BLESSED? <=> { $__class__$ => }
29		39
30	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	40	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
31	#define F_DEFAULT (9UL << S_MAXDEPTH)
32		41
33	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
34	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
35		44
36	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38		46
39	#define SB do {	47	#define SB do {
40	#define SE } while (0)	48	#define SE } while (0)
41		49
		50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr), (value))
		52	# define INLINE static inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define INLINE static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	#define IN_RANGE_INC(type,val,beg,end) \
		62	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		63	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		64
		65	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		66
		67	#ifdef USE_ITHREADS
		68	# define JSON_SLOW 1
		69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		70	#else
		71	# define JSON_SLOW 0
		72	# define JSON_STASH json_stash
		73	#endif
		74
42	static HV *json_stash; // JSON::XS::	75	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		76	static SV *json_true, *json_false;
		77
		78	enum {
		79	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		80	INCR_M_STR, // inside string
		81	INCR_M_BS, // inside backslash
		82	INCR_M_JSON // outside anything, count nesting
		83	};
		84
		85	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		86
		87	typedef struct {
		88	U32 flags;
		89	U32 max_depth;
		90	STRLEN max_size;
		91
		92	SV *cb_object;
		93	HV *cb_sk_object;
		94
		95	// for the incremental parser
		96	SV *incr_text; // the source text so far
		97	STRLEN incr_pos; // the current offset into the text
		98	int incr_nest; // {[]}-nesting level
		99	unsigned char incr_mode;
		100	} JSON;
		101
		102	INLINE void
		103	json_init (JSON *json)
		104	{
		105	Zero (json, 1, JSON);
		106	json->max_depth = 512;
		107	}
43		108
44	/////////////////////////////////////////////////////////////////////////////	109	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	110	// utility functions
46		111
47	static UV *	112	INLINE SV *
48	SvJSON (SV *sv)	113	get_bool (const char *name)
49	{	114	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	115	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		116
53	return &SvUVX (SvRV (sv));	117	SvREADONLY_on (sv);
54	}	118	SvREADONLY_on (SvRV (sv));
55		119
56	static void	120	return sv;
		121	}
		122
		123	INLINE void
57	shrink (SV *sv)	124	shrink (SV *sv)
58	{	125	{
59	sv_utf8_downgrade (sv, 1);	126	sv_utf8_downgrade (sv, 1);
		127
60	if (SvLEN (sv) > SvCUR (sv) + 1)	128	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	129	{
62	#ifdef SvPV_shrink_to_cur	130	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	131	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	132	#elif defined (SvPV_renew)
…		…
70	// decode an utf-8 character and return it, or (UV)-1 in	138	// decode an utf-8 character and return it, or (UV)-1 in
71	// case of an error.	139	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	140	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	141	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	142	// note that we never call this function for a ascii codepoints
75	static UV	143	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	144	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	145	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	146	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	147	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	148	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	149	{
82	*clen = 2;	150	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	151	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	152	}
85	else	153	else
86	{	154	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
87	*clen = (STRLEN)-1;	155	}
88	return (UV)-1;	156
89	}	157	// likewise for encoding, also never called for ascii codepoints
		158	// this function takes advantage of this fact, although current gccs
		159	// seem to optimise the check for >= 0x80 away anyways
		160	INLINE unsigned char *
		161	encode_utf8 (unsigned char *s, UV ch)
		162	{
		163	if (expect_false (ch < 0x000080))
		164	*s++ = ch;
		165	else if (expect_true (ch < 0x000800))
		166	*s++ = 0xc0 | ( ch >> 6),
		167	*s++ = 0x80 | ( ch & 0x3f);
		168	else if ( ch < 0x010000)
		169	*s++ = 0xe0 | ( ch >> 12),
		170	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172	else if ( ch < 0x110000)
		173	*s++ = 0xf0 | ( ch >> 18),
		174	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		175	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		176	*s++ = 0x80 | ( ch & 0x3f);
		177
		178	return s;
		179	}
		180
		181	// convert offset pointer to character index, sv must be string
		182	static STRLEN
		183	ptr_to_index (SV *sv, char *offset)
		184	{
		185	return SvUTF8 (sv)
		186	? utf8_distance (offset, SvPVX (sv))
		187	: offset - SvPVX (sv);
90	}	188	}
91		189
92	/////////////////////////////////////////////////////////////////////////////	190	/////////////////////////////////////////////////////////////////////////////
93	// encoder	191	// encoder
94		192
…		…
96	typedef struct	194	typedef struct
97	{	195	{
98	char *cur; // SvPVX (sv) + current output position	196	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	197	char *end; // SvEND (sv)
100	SV *sv; // result scalar	198	SV *sv; // result scalar
101	U32 flags; // F_*	199	JSON json;
102	U32 indent; // indentation level	200	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	201	UV limit; // escape character values >= this value when encoding
104	} enc_t;	202	} enc_t;
105		203
106	static void	204	INLINE void
107	need (enc_t *enc, STRLEN len)	205	need (enc_t *enc, STRLEN len)
108	{	206	{
109	if (enc->cur + len >= enc->end)	207	if (expect_false (enc->cur + len >= enc->end))
110	{	208	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	209	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	210	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	211	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	212	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	213	}
116	}	214	}
117		215
118	static void	216	INLINE void
119	encode_ch (enc_t *enc, char ch)	217	encode_ch (enc_t *enc, char ch)
120	{	218	{
121	need (enc, 1);	219	need (enc, 1);
122	*enc->cur++ = ch;	220	*enc->cur++ = ch;
123	}	221	}
…		…
131		229
132	while (str < end)	230	while (str < end)
133	{	231	{
134	unsigned char ch = *(unsigned char *)str;	232	unsigned char ch = *(unsigned char *)str;
135		233
136	if (ch >= 0x20 && ch < 0x80) // most common case	234	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	235	{
138	if (ch == '"') // but with slow exceptions	236	if (expect_false (ch == '"')) // but with slow exceptions
139	{	237	{
140	need (enc, len += 1);	238	need (enc, len += 1);
141	*enc->cur++ = '\\';	239	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	240	*enc->cur++ = '"';
143	}	241	}
144	else if (ch == '\\')	242	else if (expect_false (ch == '\\'))
145	{	243	{
146	need (enc, len += 1);	244	need (enc, len += 1);
147	*enc->cur++ = '\\';	245	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	246	*enc->cur++ = '\\';
149	}	247	}
…		…
167	STRLEN clen;	265	STRLEN clen;
168	UV uch;	266	UV uch;
169		267
170	if (is_utf8)	268	if (is_utf8)
171	{	269	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	270	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	271	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	272	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	273	}
177	else	274	else
178	{	275	{
179	uch = ch;	276	uch = ch;
180	clen = 1;	277	clen = 1;
181	}	278	}
182		279
183	if (uch > 0x10FFFFUL)	280	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)
187	{	281	{
188	if (uch > 0xFFFFUL)	282	if (uch >= 0x10000UL)
189	{	283	{
		284	if (uch >= 0x110000UL)
		285	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		286
190	need (enc, len += 11);	287	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	288	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	289	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	290	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	291	enc->cur += 12;
195	}	292	}
196	else	293	else
197	{	294	{
198	static char hexdigit [16] = "0123456789abcdef";
199	need (enc, len += 5);	295	need (enc, len += 5);
200	*enc->cur++ = '\\';	296	*enc->cur++ = '\\';
201	*enc->cur++ = 'u';	297	*enc->cur++ = 'u';
202	*enc->cur++ = hexdigit [ uch >> 12 ];	298	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
203	*enc->cur++ = hexdigit [(uch >> 8) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
204	*enc->cur++ = hexdigit [(uch >> 4) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	301	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
206	}	302	}
207		303
		304	str += clen;
		305	}
		306	else if (enc->json.flags & F_LATIN1)
		307	{
		308	*enc->cur++ = uch;
208	str += clen;	309	str += clen;
209	}	310	}
210	else if (is_utf8)	311	else if (is_utf8)
211	{	312	{
212	need (enc, len += clen);	313	need (enc, len += clen);
…		…
216	}	317	}
217	while (--clen);	318	while (--clen);
218	}	319	}
219	else	320	else
220	{	321	{
221	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	322	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
222	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	323	enc->cur = encode_utf8 (enc->cur, uch);
223	++str;	324	++str;
224	}	325	}
225	}	326	}
226	}	327	}
227	}	328	}
228		329
229	--len;	330	--len;
230	}	331	}
231	}	332	}
232		333
233	static void	334	INLINE void
234	encode_indent (enc_t *enc)	335	encode_indent (enc_t *enc)
235	{	336	{
236	if (enc->flags & F_INDENT)	337	if (enc->json.flags & F_INDENT)
237	{	338	{
238	int spaces = enc->indent * INDENT_STEP;	339	int spaces = enc->indent * INDENT_STEP;
239		340
240	need (enc, spaces);	341	need (enc, spaces);
241	memset (enc->cur, ' ', spaces);	342	memset (enc->cur, ' ', spaces);
242	enc->cur += spaces;	343	enc->cur += spaces;
243	}	344	}
244	}	345	}
245		346
246	static void	347	INLINE void
247	encode_space (enc_t *enc)	348	encode_space (enc_t *enc)
248	{	349	{
249	need (enc, 1);	350	need (enc, 1);
250	encode_ch (enc, ' ');	351	encode_ch (enc, ' ');
251	}	352	}
252		353
253	static void	354	INLINE void
254	encode_nl (enc_t *enc)	355	encode_nl (enc_t *enc)
255	{	356	{
256	if (enc->flags & F_INDENT)	357	if (enc->json.flags & F_INDENT)
257	{	358	{
258	need (enc, 1);	359	need (enc, 1);
259	encode_ch (enc, '\n');	360	encode_ch (enc, '\n');
260	}	361	}
261	}	362	}
262		363
263	static void	364	INLINE void
264	encode_comma (enc_t *enc)	365	encode_comma (enc_t *enc)
265	{	366	{
266	encode_ch (enc, ',');	367	encode_ch (enc, ',');
267		368
268	if (enc->flags & F_INDENT)	369	if (enc->json.flags & F_INDENT)
269	encode_nl (enc);	370	encode_nl (enc);
270	else if (enc->flags & F_SPACE_AFTER)	371	else if (enc->json.flags & F_SPACE_AFTER)
271	encode_space (enc);	372	encode_space (enc);
272	}	373	}
273		374
274	static void encode_sv (enc_t *enc, SV *sv);	375	static void encode_sv (enc_t *enc, SV *sv);
275		376
276	static void	377	static void
277	encode_av (enc_t *enc, AV *av)	378	encode_av (enc_t *enc, AV *av)
278	{	379	{
279	int i, len = av_len (av);	380	int i, len = av_len (av);
280		381
281	if (enc->indent >= enc->maxdepth)	382	if (enc->indent >= enc->json.max_depth)
282	croak ("data structure too deep (hit recursion limit)");	383	croak (ERR_NESTING_EXCEEDED);
283		384
284	encode_ch (enc, '['); encode_nl (enc);	385	encode_ch (enc, '[');
285	++enc->indent;	386
		387	if (len >= 0)
		388	{
		389	encode_nl (enc); ++enc->indent;
286		390
287	for (i = 0; i <= len; ++i)	391	for (i = 0; i <= len; ++i)
288	{	392	{
		393	SV **svp = av_fetch (av, i, 0);
		394
289	encode_indent (enc);	395	encode_indent (enc);
290	encode_sv (enc, *av_fetch (av, i, 0));
291		396
		397	if (svp)
		398	encode_sv (enc, *svp);
		399	else
		400	encode_str (enc, "null", 4, 0);
		401
292	if (i < len)	402	if (i < len)
293	encode_comma (enc);	403	encode_comma (enc);
294	}	404	}
295		405
		406	encode_nl (enc); --enc->indent; encode_indent (enc);
		407	}
		408
296	encode_nl (enc);	409	encode_ch (enc, ']');
297
298	--enc->indent;
299	encode_indent (enc); encode_ch (enc, ']');
300	}	410	}
301		411
302	static void	412	static void
303	encode_he (enc_t *enc, HE *he)	413	encode_hk (enc_t *enc, HE *he)
304	{	414	{
305	encode_ch (enc, '"');	415	encode_ch (enc, '"');
306		416
307	if (HeKLEN (he) == HEf_SVKEY)	417	if (HeKLEN (he) == HEf_SVKEY)
308	{	418	{
…		…
318	else	428	else
319	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	429	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
320		430
321	encode_ch (enc, '"');	431	encode_ch (enc, '"');
322		432
323	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	433	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
324	encode_ch (enc, ':');	434	encode_ch (enc, ':');
325	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	435	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
326	encode_sv (enc, HeVAL (he));
327	}	436	}
328		437
329	// compare hash entries, used when all keys are bytestrings	438	// compare hash entries, used when all keys are bytestrings
330	static int	439	static int
331	he_cmp_fast (const void *a_, const void *b_)	440	he_cmp_fast (const void *a_, const void *b_)
…		…
336	HE *b = *(HE **)b_;	445	HE *b = *(HE **)b_;
337		446
338	STRLEN la = HeKLEN (a);	447	STRLEN la = HeKLEN (a);
339	STRLEN lb = HeKLEN (b);	448	STRLEN lb = HeKLEN (b);
340		449
341	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	450	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
342	cmp = la - lb;	451	cmp = lb - la;
343		452
344	return cmp;	453	return cmp;
345	}	454	}
346		455
347	// compare hash entries, used when some keys are sv's or utf-x	456	// compare hash entries, used when some keys are sv's or utf-x
348	static int	457	static int
349	he_cmp_slow (const void *a, const void *b)	458	he_cmp_slow (const void *a, const void *b)
350	{	459	{
351	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	460	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
352	}	461	}
353		462
354	static void	463	static void
355	encode_hv (enc_t *enc, HV *hv)	464	encode_hv (enc_t *enc, HV *hv)
356	{	465	{
357	int count, i;	466	HE *he;
358		467
359	if (enc->indent >= enc->maxdepth)	468	if (enc->indent >= enc->json.max_depth)
360	croak ("data structure too deep (hit recursion limit)");	469	croak (ERR_NESTING_EXCEEDED);
361		470
362	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	471	encode_ch (enc, '{');
363		472
364	if ((count = hv_iterinit (hv)))
365	{
366	// for canonical output we have to sort by keys first	473	// for canonical output we have to sort by keys first
367	// actually, this is mostly due to the stupid so-called	474	// actually, this is mostly due to the stupid so-called
368	// security workaround added somewhere in 5.8.x.	475	// security workaround added somewhere in 5.8.x
369	// that randomises hash orderings	476	// that randomises hash orderings
370	if (enc->flags & F_CANONICAL)	477	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		478	{
		479	int count = hv_iterinit (hv);
		480
		481	if (SvMAGICAL (hv))
371	{	482	{
		483	// need to count by iterating. could improve by dynamically building the vector below
		484	// but I don't care for the speed of this special case.
		485	// note also that we will run into undefined behaviour when the two iterations
		486	// do not result in the same count, something I might care for in some later release.
		487
		488	count = 0;
		489	while (hv_iternext (hv))
		490	++count;
		491
		492	hv_iterinit (hv);
		493	}
		494
		495	if (count)
		496	{
		497	int i, fast = 1;
		498	#if defined(__BORLANDC__) || defined(_MSC_VER)
		499	HE **hes = _alloca (count * sizeof (HE));
		500	#else
372	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	501	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
373	int fast = 1;	502	#endif
374		503
375	i = 0;	504	i = 0;
376	while ((he = hv_iternext (hv)))	505	while ((he = hv_iternext (hv)))
377	{	506	{
378	hes [i++] = he;	507	hes [i++] = he;
…		…
400		529
401	FREETMPS;	530	FREETMPS;
402	LEAVE;	531	LEAVE;
403	}	532	}
404		533
405	for (i = 0; i < count; ++i)	534	encode_nl (enc); ++enc->indent;
		535
		536	while (count--)
406	{	537	{
407	encode_indent (enc);	538	encode_indent (enc);
		539	he = hes [count];
408	encode_he (enc, hes [i]);	540	encode_hk (enc, he);
		541	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
409		542
410	if (i < count - 1)	543	if (count)
411	encode_comma (enc);	544	encode_comma (enc);
412	}	545	}
413		546
		547	encode_nl (enc); --enc->indent; encode_indent (enc);
		548	}
		549	}
		550	else
		551	{
		552	if (hv_iterinit (hv) || SvMAGICAL (hv))
		553	if ((he = hv_iternext (hv)))
		554	{
		555	encode_nl (enc); ++enc->indent;
		556
		557	for (;;)
		558	{
414	encode_nl (enc);	559	encode_indent (enc);
		560	encode_hk (enc, he);
		561	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		562
		563	if (!(he = hv_iternext (hv)))
		564	break;
		565
		566	encode_comma (enc);
		567	}
		568
		569	encode_nl (enc); --enc->indent; encode_indent (enc);
		570	}
		571	}
		572
		573	encode_ch (enc, '}');
		574	}
		575
		576	// encode objects, arrays and special \0=false and \1=true values.
		577	static void
		578	encode_rv (enc_t *enc, SV *sv)
		579	{
		580	svtype svt;
		581
		582	SvGETMAGIC (sv);
		583	svt = SvTYPE (sv);
		584
		585	if (expect_false (SvOBJECT (sv)))
		586	{
		587	HV *stash = !JSON_SLOW || json_boolean_stash
		588	? json_boolean_stash
		589	: gv_stashpv ("JSON::XS::Boolean", 1);
		590
		591	if (SvSTASH (sv) == stash)
		592	{
		593	if (SvIV (sv))
		594	encode_str (enc, "true", 4, 0);
		595	else
		596	encode_str (enc, "false", 5, 0);
415	}	597	}
416	else	598	else
417	{	599	{
418	HE *he = hv_iternext (hv);	600	#if 0
419		601	if (0 && sv_derived_from (rv, "JSON::Literal"))
420	for (;;)
421	{	602	{
422	encode_indent (enc);	603	// not yet
423	encode_he (enc, he);
424
425	if (!(he = hv_iternext (hv)))
426	break;
427
428	encode_comma (enc);
429	}	604	}
		605	#endif
		606	if (enc->json.flags & F_CONV_BLESSED)
		607	{
		608	// we re-bless the reference to get overload and other niceties right
		609	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
430		610
		611	if (to_json)
		612	{
		613	dSP;
		614
		615	ENTER; SAVETMPS; PUSHMARK (SP);
		616	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		617
		618	// calling with G_SCALAR ensures that we always get a 1 return value
		619	PUTBACK;
		620	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		621	SPAGAIN;
		622
		623	// catch this surprisingly common error
		624	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		625	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		626
		627	sv = POPs;
		628	PUTBACK;
		629
431	encode_nl (enc);	630	encode_sv (enc, sv);
		631
		632	FREETMPS; LEAVE;
		633	}
		634	else if (enc->json.flags & F_ALLOW_BLESSED)
		635	encode_str (enc, "null", 4, 0);
		636	else
		637	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		638	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		639	}
		640	else if (enc->json.flags & F_ALLOW_BLESSED)
		641	encode_str (enc, "null", 4, 0);
		642	else
		643	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		644	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
432	}	645	}
433	}	646	}
434
435	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
436	}
437
438	// encode objects, arrays and special \0=false and \1=true values.
439	static void
440	encode_rv (enc_t *enc, SV *sv)
441	{
442	svtype svt;
443
444	SvGETMAGIC (sv);
445	svt = SvTYPE (sv);
446
447	if (svt == SVt_PVHV)	647	else if (svt == SVt_PVHV)
448	encode_hv (enc, (HV *)sv);	648	encode_hv (enc, (HV *)sv);
449	else if (svt == SVt_PVAV)	649	else if (svt == SVt_PVAV)
450	encode_av (enc, (AV *)sv);	650	encode_av (enc, (AV *)sv);
451	else if (svt < SVt_PVAV)	651	else if (svt < SVt_PVAV)
452	{	652	{
453	if (SvNIOK (sv) && SvIV (sv) == 0)	653	STRLEN len = 0;
		654	char *pv = svt ? SvPV (sv, len) : 0;
		655
		656	if (len == 1 && *pv == '1')
		657	encode_str (enc, "true", 4, 0);
		658	else if (len == 1 && *pv == '0')
454	encode_str (enc, "false", 5, 0);	659	encode_str (enc, "false", 5, 0);
455	else if (SvNIOK (sv) && SvIV (sv) == 1)	660	else if (enc->json.flags & F_ALLOW_UNKNOWN)
456	encode_str (enc, "true", 4, 0);	661	encode_str (enc, "null", 4, 0);
457	else	662	else
458	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	663	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
459	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	664	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
460	}	665	}
		666	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		667	encode_str (enc, "null", 4, 0);
461	else	668	else
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	669	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	670	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
464	}	671	}
465		672
…		…
476	encode_str (enc, str, len, SvUTF8 (sv));	683	encode_str (enc, str, len, SvUTF8 (sv));
477	encode_ch (enc, '"');	684	encode_ch (enc, '"');
478	}	685	}
479	else if (SvNOKp (sv))	686	else if (SvNOKp (sv))
480	{	687	{
		688	// trust that perl will do the right thing w.r.t. JSON syntax.
481	need (enc, NV_DIG + 32);	689	need (enc, NV_DIG + 32);
482	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	690	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
483	enc->cur += strlen (enc->cur);	691	enc->cur += strlen (enc->cur);
484	}	692	}
485	else if (SvIOKp (sv))	693	else if (SvIOKp (sv))
486	{	694	{
		695	// we assume we can always read an IV as a UV and vice versa
		696	// we assume two's complement
		697	// we assume no aliasing issues in the union
		698	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		699	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		700	{
		701	// optimise the "small number case"
		702	// code will likely be branchless and use only a single multiplication
		703	// works for numbers up to 59074
		704	I32 i = SvIVX (sv);
		705	U32 u;
		706	char digit, nz = 0;
		707
487	need (enc, 64);	708	need (enc, 6);
		709
		710	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		711	u = i < 0 ? -i : i;
		712
		713	// convert to 4.28 fixed-point representation
		714	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		715
		716	// now output digit by digit, each time masking out the integer part
		717	// and multiplying by 5 while moving the decimal point one to the right,
		718	// resulting in a net multiplication by 10.
		719	// we always write the digit to memory but conditionally increment
		720	// the pointer, to enable the use of conditional move instructions.
		721	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		722	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		723	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		724	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		725	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		726	}
		727	else
		728	{
		729	// large integer, use the (rather slow) snprintf way.
		730	need (enc, IVUV_MAXCHARS);
488	enc->cur +=	731	enc->cur +=
489	SvIsUV(sv)	732	SvIsUV(sv)
490	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	733	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
491	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	734	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		735	}
492	}	736	}
493	else if (SvROK (sv))	737	else if (SvROK (sv))
494	encode_rv (enc, SvRV (sv));	738	encode_rv (enc, SvRV (sv));
495	else if (!SvOK (sv))	739	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
496	encode_str (enc, "null", 4, 0);	740	encode_str (enc, "null", 4, 0);
497	else	741	else
498	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	742	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
499	SvPV_nolen (sv), SvFLAGS (sv));	743	SvPV_nolen (sv), SvFLAGS (sv));
500	}	744	}
501		745
502	static SV *	746	static SV *
503	encode_json (SV *scalar, U32 flags)	747	encode_json (SV *scalar, JSON *json)
504	{	748	{
505	enc_t enc;	749	enc_t enc;
506		750
507	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	751	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
508	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	752	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
509		753
510	enc.flags = flags;	754	enc.json = *json;
511	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	755	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
512	enc.cur = SvPVX (enc.sv);	756	enc.cur = SvPVX (enc.sv);
513	enc.end = SvEND (enc.sv);	757	enc.end = SvEND (enc.sv);
514	enc.indent = 0;	758	enc.indent = 0;
515	enc.maxdepth = DEC_DEPTH (flags);	759	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		760	: enc.json.flags & F_LATIN1 ? 0x000100UL
		761	: 0x110000UL;
516		762
517	SvPOK_only (enc.sv);	763	SvPOK_only (enc.sv);
518	encode_sv (&enc, scalar);	764	encode_sv (&enc, scalar);
		765	encode_nl (&enc);
519		766
		767	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		768	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		769
520	if (!(flags & (F_ASCII | F_UTF8)))	770	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
521	SvUTF8_on (enc.sv);	771	SvUTF8_on (enc.sv);
522		772
523	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
524
525	if (enc.flags & F_SHRINK)	773	if (enc.json.flags & F_SHRINK)
526	shrink (enc.sv);	774	shrink (enc.sv);
527		775
528	return enc.sv;	776	return enc.sv;
529	}	777	}
530		778
…		…
535	typedef struct	783	typedef struct
536	{	784	{
537	char *cur; // current parser pointer	785	char *cur; // current parser pointer
538	char *end; // end of input string	786	char *end; // end of input string
539	const char *err; // parse error, if != 0	787	const char *err; // parse error, if != 0
540	U32 flags; // F_*	788	JSON json;
541	U32 depth; // recursion depth	789	U32 depth; // recursion depth
542	U32 maxdepth; // recursion depth limit	790	U32 maxdepth; // recursion depth limit
543	} dec_t;	791	} dec_t;
544		792
545	static void	793	INLINE void
		794	decode_comment (dec_t *dec)
		795	{
		796	// only '#'-style comments allowed a.t.m.
		797
		798	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		799	++dec->cur;
		800	}
		801
		802	INLINE void
546	decode_ws (dec_t *dec)	803	decode_ws (dec_t *dec)
547	{	804	{
548	for (;;)	805	for (;;)
549	{	806	{
550	char ch = *dec->cur;	807	char ch = *dec->cur;
551		808
552	if (ch > 0x20	809	if (ch > 0x20)
		810	{
		811	if (expect_false (ch == '#'))
		812	{
		813	if (dec->json.flags & F_RELAXED)
		814	decode_comment (dec);
		815	else
		816	break;
		817	}
		818	else
		819	break;
		820	}
553	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	821	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
554	break;	822	break; // parse error, but let higher level handle it, gives better error messages
555		823
556	++dec->cur;	824	++dec->cur;
557	}	825	}
558	}	826	}
559		827
…		…
563	if (*dec->cur != ch) \	831	if (*dec->cur != ch) \
564	ERR (# ch " expected"); \	832	ERR (# ch " expected"); \
565	++dec->cur; \	833	++dec->cur; \
566	SE	834	SE
567		835
568	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	836	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
569	#define DEC_DEC_DEPTH --dec->depth	837	#define DEC_DEC_DEPTH --dec->depth
570		838
571	static SV *decode_sv (dec_t *dec);	839	static SV *decode_sv (dec_t *dec);
572		840
573	static signed char decode_hexdigit[256];	841	static signed char decode_hexdigit[256];
…		…
576	decode_4hex (dec_t *dec)	844	decode_4hex (dec_t *dec)
577	{	845	{
578	signed char d1, d2, d3, d4;	846	signed char d1, d2, d3, d4;
579	unsigned char *cur = (unsigned char *)dec->cur;	847	unsigned char *cur = (unsigned char *)dec->cur;
580		848
581	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	849	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
582	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	850	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
583	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	851	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
584	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	852	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
585		853
586	dec->cur += 4;	854	dec->cur += 4;
587		855
588	return ((UV)d1) << 12	856	return ((UV)d1) << 12
589	| ((UV)d2) << 8	857	| ((UV)d2) << 8
…		…
597	static SV *	865	static SV *
598	decode_str (dec_t *dec)	866	decode_str (dec_t *dec)
599	{	867	{
600	SV *sv = 0;	868	SV *sv = 0;
601	int utf8 = 0;	869	int utf8 = 0;
		870	char *dec_cur = dec->cur;
602		871
603	do	872	do
604	{	873	{
605	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	874	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
606	char *cur = buf;	875	char *cur = buf;
607		876
608	do	877	do
609	{	878	{
610	unsigned char ch = *(unsigned char *)dec->cur++;	879	unsigned char ch = *(unsigned char *)dec_cur++;
611		880
612	if (ch == '"')	881	if (expect_false (ch == '"'))
613	{	882	{
614	--dec->cur;	883	--dec_cur;
615	break;	884	break;
616	}	885	}
617	else if (ch == '\\')	886	else if (expect_false (ch == '\\'))
618	{	887	{
619	switch (*dec->cur)	888	switch (*dec_cur)
620	{	889	{
621	case '\\':	890	case '\\':
622	case '/':	891	case '/':
623	case '"': *cur++ = *dec->cur++; break;	892	case '"': *cur++ = *dec_cur++; break;
624		893
625	case 'b': ++dec->cur; *cur++ = '\010'; break;	894	case 'b': ++dec_cur; *cur++ = '\010'; break;
626	case 't': ++dec->cur; *cur++ = '\011'; break;	895	case 't': ++dec_cur; *cur++ = '\011'; break;
627	case 'n': ++dec->cur; *cur++ = '\012'; break;	896	case 'n': ++dec_cur; *cur++ = '\012'; break;
628	case 'f': ++dec->cur; *cur++ = '\014'; break;	897	case 'f': ++dec_cur; *cur++ = '\014'; break;
629	case 'r': ++dec->cur; *cur++ = '\015'; break;	898	case 'r': ++dec_cur; *cur++ = '\015'; break;
630		899
631	case 'u':	900	case 'u':
632	{	901	{
633	UV lo, hi;	902	UV lo, hi;
634	++dec->cur;	903	++dec_cur;
635		904
		905	dec->cur = dec_cur;
636	hi = decode_4hex (dec);	906	hi = decode_4hex (dec);
		907	dec_cur = dec->cur;
637	if (hi == (UV)-1)	908	if (hi == (UV)-1)
638	goto fail;	909	goto fail;
639		910
640	// possibly a surrogate pair	911	// possibly a surrogate pair
641	if (hi >= 0xd800)	912	if (hi >= 0xd800)
642	if (hi < 0xdc00)	913	if (hi < 0xdc00)
643	{	914	{
644	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	915	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
645	ERR ("missing low surrogate character in surrogate pair");	916	ERR ("missing low surrogate character in surrogate pair");
646		917
647	dec->cur += 2;	918	dec_cur += 2;
648		919
		920	dec->cur = dec_cur;
649	lo = decode_4hex (dec);	921	lo = decode_4hex (dec);
		922	dec_cur = dec->cur;
650	if (lo == (UV)-1)	923	if (lo == (UV)-1)
651	goto fail;	924	goto fail;
652		925
653	if (lo < 0xdc00 || lo >= 0xe000)	926	if (lo < 0xdc00 || lo >= 0xe000)
654	ERR ("surrogate pair expected");	927	ERR ("surrogate pair expected");
…		…
660		933
661	if (hi >= 0x80)	934	if (hi >= 0x80)
662	{	935	{
663	utf8 = 1;	936	utf8 = 1;
664		937
665	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	938	cur = encode_utf8 (cur, hi);
666	}	939	}
667	else	940	else
668	*cur++ = hi;	941	*cur++ = hi;
669	}	942	}
670	break;	943	break;
671		944
672	default:	945	default:
673	--dec->cur;	946	--dec_cur;
674	ERR ("illegal backslash escape sequence in string");	947	ERR ("illegal backslash escape sequence in string");
675	}	948	}
676	}	949	}
677	else if (ch >= 0x20 && ch <= 0x7f)	950	else if (expect_true (ch >= 0x20 && ch < 0x80))
678	*cur++ = ch;	951	*cur++ = ch;
679	else if (ch >= 0x80)	952	else if (ch >= 0x80)
680	{	953	{
681	STRLEN clen;	954	STRLEN clen;
682	UV uch;
683		955
684	--dec->cur;	956	--dec_cur;
685		957
686	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	958	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
687	if (clen == (STRLEN)-1)	959	if (clen == (STRLEN)-1)
688	ERR ("malformed UTF-8 character in JSON string");	960	ERR ("malformed UTF-8 character in JSON string");
689		961
690	do	962	do
691	*cur++ = *dec->cur++;	963	*cur++ = *dec_cur++;
692	while (--clen);	964	while (--clen);
693		965
694	utf8 = 1;	966	utf8 = 1;
695	}	967	}
696	else	968	else
697	{	969	{
698	--dec->cur;	970	--dec_cur;
699		971
700	if (!ch)	972	if (!ch)
701	ERR ("unexpected end of string while parsing JSON string");	973	ERR ("unexpected end of string while parsing JSON string");
702	else	974	else
703	ERR ("invalid character encountered while parsing JSON string");	975	ERR ("invalid character encountered while parsing JSON string");
…		…
708	{	980	{
709	STRLEN len = cur - buf;	981	STRLEN len = cur - buf;
710		982
711	if (sv)	983	if (sv)
712	{	984	{
713	SvGROW (sv, SvCUR (sv) + len + 1);	985	STRLEN cur = SvCUR (sv);
		986
		987	if (SvLEN (sv) <= cur + len)
		988	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		989
714	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	990	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
715	SvCUR_set (sv, SvCUR (sv) + len);	991	SvCUR_set (sv, SvCUR (sv) + len);
716	}	992	}
717	else	993	else
718	sv = newSVpvn (buf, len);	994	sv = newSVpvn (buf, len);
719	}	995	}
720	}	996	}
721	while (*dec->cur != '"');	997	while (*dec_cur != '"');
722		998
723	++dec->cur;	999	++dec_cur;
724		1000
725	if (sv)	1001	if (sv)
726	{	1002	{
727	SvPOK_only (sv);	1003	SvPOK_only (sv);
728	*SvEND (sv) = 0;	1004	*SvEND (sv) = 0;
…		…
731	SvUTF8_on (sv);	1007	SvUTF8_on (sv);
732	}	1008	}
733	else	1009	else
734	sv = newSVpvn ("", 0);	1010	sv = newSVpvn ("", 0);
735		1011
		1012	dec->cur = dec_cur;
736	return sv;	1013	return sv;
737		1014
738	fail:	1015	fail:
		1016	dec->cur = dec_cur;
739	return 0;	1017	return 0;
740	}	1018	}
741		1019
742	static SV *	1020	static SV *
743	decode_num (dec_t *dec)	1021	decode_num (dec_t *dec)
…		…
801	is_nv = 1;	1079	is_nv = 1;
802	}	1080	}
803		1081
804	if (!is_nv)	1082	if (!is_nv)
805	{	1083	{
806	UV uv;	1084	int len = dec->cur - start;
807	int numtype = grok_number (start, dec->cur - start, &uv);	1085
808	if (numtype & IS_NUMBER_IN_UV)	1086	// special case the rather common 1..5-digit-int case
809	if (numtype & IS_NUMBER_NEG)	1087	if (*start == '-')
		1088	switch (len)
810	{	1089	{
811	if (uv < (UV)IV_MIN)	1090	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
812	return newSViv (-(IV)uv);	1091	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1092	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1093	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1094	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
813	}	1095	}
		1096	else
		1097	switch (len)
		1098	{
		1099	case 1: return newSViv ( start [0] - '0' * 1);
		1100	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1101	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1102	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1103	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1104	}
		1105
		1106	{
		1107	UV uv;
		1108	int numtype = grok_number (start, len, &uv);
		1109	if (numtype & IS_NUMBER_IN_UV)
		1110	if (numtype & IS_NUMBER_NEG)
		1111	{
		1112	if (uv < (UV)IV_MIN)
		1113	return newSViv (-(IV)uv);
		1114	}
814	else	1115	else
815	return newSVuv (uv);	1116	return newSVuv (uv);
816	}	1117	}
817		1118
		1119	len -= *start == '-' ? 1 : 0;
		1120
		1121	// does not fit into IV or UV, try NV
		1122	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1123	#if defined (LDBL_DIG)
		1124	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1125	#endif
		1126	)
		1127	// fits into NV without loss of precision
		1128	return newSVnv (Atof (start));
		1129
		1130	// everything else fails, convert it to a string
		1131	return newSVpvn (start, dec->cur - start);
		1132	}
		1133
		1134	// loss of precision here
818	return newSVnv (Atof (start));	1135	return newSVnv (Atof (start));
819		1136
820	fail:	1137	fail:
821	return 0;	1138	return 0;
822	}	1139	}
…		…
852		1169
853	if (*dec->cur != ',')	1170	if (*dec->cur != ',')
854	ERR (", or ] expected while parsing array");	1171	ERR (", or ] expected while parsing array");
855		1172
856	++dec->cur;	1173	++dec->cur;
		1174
		1175	decode_ws (dec);
		1176
		1177	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1178	{
		1179	++dec->cur;
		1180	break;
		1181	}
857	}	1182	}
858		1183
859	DEC_DEC_DEPTH;	1184	DEC_DEC_DEPTH;
860	return newRV_noinc ((SV *)av);	1185	return newRV_noinc ((SV *)av);
861		1186
…		…
866	}	1191	}
867		1192
868	static SV *	1193	static SV *
869	decode_hv (dec_t *dec)	1194	decode_hv (dec_t *dec)
870	{	1195	{
		1196	SV *sv;
871	HV *hv = newHV ();	1197	HV *hv = newHV ();
872		1198
873	DEC_INC_DEPTH;	1199	DEC_INC_DEPTH;
874	decode_ws (dec);	1200	decode_ws (dec);
875		1201
876	if (*dec->cur == '}')	1202	if (*dec->cur == '}')
877	++dec->cur;	1203	++dec->cur;
878	else	1204	else
879	for (;;)	1205	for (;;)
880	{	1206	{
		1207	EXPECT_CH ('"');
		1208
		1209	// heuristic: assume that
		1210	// a) decode_str + hv_store_ent are abysmally slow.
		1211	// b) most hash keys are short, simple ascii text.
		1212	// => try to "fast-match" such strings to avoid
		1213	// the overhead of decode_str + hv_store_ent.
		1214	{
881	SV *key, *value;	1215	SV *value;
		1216	char *p = dec->cur;
		1217	char *e = p + 24; // only try up to 24 bytes
882		1218
883	decode_ws (dec); EXPECT_CH ('"');	1219	for (;;)
884
885	key = decode_str (dec);
886	if (!key)
887	goto fail;
888
889	decode_ws (dec); EXPECT_CH (':');
890
891	value = decode_sv (dec);
892	if (!value)
893	{	1220	{
		1221	// the >= 0x80 is false on most architectures
		1222	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1223	{
		1224	// slow path, back up and use decode_str
		1225	SV *key = decode_str (dec);
		1226	if (!key)
		1227	goto fail;
		1228
		1229	decode_ws (dec); EXPECT_CH (':');
		1230
		1231	decode_ws (dec);
		1232	value = decode_sv (dec);
		1233	if (!value)
		1234	{
		1235	SvREFCNT_dec (key);
		1236	goto fail;
		1237	}
		1238
		1239	hv_store_ent (hv, key, value, 0);
894	SvREFCNT_dec (key);	1240	SvREFCNT_dec (key);
		1241
		1242	break;
		1243	}
		1244	else if (*p == '"')
		1245	{
		1246	// fast path, got a simple key
		1247	char *key = dec->cur;
		1248	int len = p - key;
		1249	dec->cur = p + 1;
		1250
		1251	decode_ws (dec); EXPECT_CH (':');
		1252
		1253	decode_ws (dec);
		1254	value = decode_sv (dec);
		1255	if (!value)
895	goto fail;	1256	goto fail;
		1257
		1258	hv_store (hv, key, len, value, 0);
		1259
		1260	break;
		1261	}
		1262
		1263	++p;
896	}	1264	}
897		1265	}
898	hv_store_ent (hv, key, value, 0);
899	SvREFCNT_dec (key);
900		1266
901	decode_ws (dec);	1267	decode_ws (dec);
902		1268
903	if (*dec->cur == '}')	1269	if (*dec->cur == '}')
904	{	1270	{
…		…
908		1274
909	if (*dec->cur != ',')	1275	if (*dec->cur != ',')
910	ERR (", or } expected while parsing object/hash");	1276	ERR (", or } expected while parsing object/hash");
911		1277
912	++dec->cur;	1278	++dec->cur;
		1279
		1280	decode_ws (dec);
		1281
		1282	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1283	{
		1284	++dec->cur;
		1285	break;
		1286	}
913	}	1287	}
914		1288
915	DEC_DEC_DEPTH;	1289	DEC_DEC_DEPTH;
916	return newRV_noinc ((SV *)hv);	1290	sv = newRV_noinc ((SV *)hv);
		1291
		1292	// check filter callbacks
		1293	if (dec->json.flags & F_HOOK)
		1294	{
		1295	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1296	{
		1297	HE *cb, *he;
		1298
		1299	hv_iterinit (hv);
		1300	he = hv_iternext (hv);
		1301	hv_iterinit (hv);
		1302
		1303	// the next line creates a mortal sv each time its called.
		1304	// might want to optimise this for common cases.
		1305	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1306
		1307	if (cb)
		1308	{
		1309	dSP;
		1310	int count;
		1311
		1312	ENTER; SAVETMPS; PUSHMARK (SP);
		1313	XPUSHs (HeVAL (he));
		1314
		1315	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1316
		1317	if (count == 1)
		1318	{
		1319	sv = newSVsv (POPs);
		1320	FREETMPS; LEAVE;
		1321	return sv;
		1322	}
		1323
		1324	FREETMPS; LEAVE;
		1325	}
		1326	}
		1327
		1328	if (dec->json.cb_object)
		1329	{
		1330	dSP;
		1331	int count;
		1332
		1333	ENTER; SAVETMPS; PUSHMARK (SP);
		1334	XPUSHs (sv_2mortal (sv));
		1335
		1336	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1337
		1338	if (count == 1)
		1339	{
		1340	sv = newSVsv (POPs);
		1341	FREETMPS; LEAVE;
		1342	return sv;
		1343	}
		1344
		1345	SvREFCNT_inc (sv);
		1346	FREETMPS; LEAVE;
		1347	}
		1348	}
		1349
		1350	return sv;
917		1351
918	fail:	1352	fail:
919	SvREFCNT_dec (hv);	1353	SvREFCNT_dec (hv);
920	DEC_DEC_DEPTH;	1354	DEC_DEC_DEPTH;
921	return 0;	1355	return 0;
922	}	1356	}
923		1357
924	static SV *	1358	static SV *
925	decode_sv (dec_t *dec)	1359	decode_sv (dec_t *dec)
926	{	1360	{
927	decode_ws (dec);	1361	// the beauty of JSON: you need exactly one character lookahead
		1362	// to parse everything.
928	switch (*dec->cur)	1363	switch (*dec->cur)
929	{	1364	{
930	case '"': ++dec->cur; return decode_str (dec);	1365	case '"': ++dec->cur; return decode_str (dec);
931	case '[': ++dec->cur; return decode_av (dec);	1366	case '[': ++dec->cur; return decode_av (dec);
932	case '{': ++dec->cur; return decode_hv (dec);	1367	case '{': ++dec->cur; return decode_hv (dec);
933		1368
934	case '-':	1369	case '-':
935	case '0': case '1': case '2': case '3': case '4':	1370	case '0': case '1': case '2': case '3': case '4':
936	case '5': case '6': case '7': case '8': case '9':	1371	case '5': case '6': case '7': case '8': case '9':
937	return decode_num (dec);	1372	return decode_num (dec);
938		1373
939	case 't':	1374	case 't':
940	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1375	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
941	{	1376	{
942	dec->cur += 4;	1377	dec->cur += 4;
		1378	#if JSON_SLOW
		1379	json_true = get_bool ("JSON::XS::true");
		1380	#endif
943	return newSViv (1);	1381	return newSVsv (json_true);
944	}	1382	}
945	else	1383	else
946	ERR ("'true' expected");	1384	ERR ("'true' expected");
947		1385
948	break;	1386	break;
949		1387
950	case 'f':	1388	case 'f':
951	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1389	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
952	{	1390	{
953	dec->cur += 5;	1391	dec->cur += 5;
		1392	#if JSON_SLOW
		1393	json_false = get_bool ("JSON::XS::false");
		1394	#endif
954	return newSViv (0);	1395	return newSVsv (json_false);
955	}	1396	}
956	else	1397	else
957	ERR ("'false' expected");	1398	ERR ("'false' expected");
958		1399
959	break;	1400	break;
…		…
977	fail:	1418	fail:
978	return 0;	1419	return 0;
979	}	1420	}
980		1421
981	static SV *	1422	static SV *
982	decode_json (SV *string, U32 flags)	1423	decode_json (SV *string, JSON *json, char **offset_return)
983	{	1424	{
984	dec_t dec;	1425	dec_t dec;
985	SV *sv;	1426	SV *sv;
986		1427
		1428	/* work around bugs in 5.10 where manipulating magic values
		1429	* will perl ignore the magic in subsequent accesses
		1430	*/
		1431	/*SvGETMAGIC (string);*/
		1432	if (SvMAGICAL (string))
		1433	string = sv_2mortal (newSVsv (string));
		1434
987	SvUPGRADE (string, SVt_PV);	1435	SvUPGRADE (string, SVt_PV);
988		1436
		1437	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1438	* assertion with -DDEBUGGING, although SvCUR is documented to
		1439	* return the xpv_cur field which certainly exists after upgrading.
		1440	* according to nicholas clark, calling SvPOK fixes this.
		1441	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1442	* and hope for the best.
		1443	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1444	* assertion business is seriously broken, try yet another workaround
		1445	* for the broken -DDEBUGGING.
		1446	*/
		1447	{
		1448	#ifdef DEBUGGING
		1449	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1450	#else
		1451	STRLEN offset = SvCUR (string);
		1452	#endif
		1453
		1454	if (offset > json->max_size && json->max_size)
		1455	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1456	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1457	}
		1458
989	if (flags & F_UTF8)	1459	if (json->flags & F_UTF8)
990	sv_utf8_downgrade (string, 0);	1460	sv_utf8_downgrade (string, 0);
991	else	1461	else
992	sv_utf8_upgrade (string);	1462	sv_utf8_upgrade (string);
993		1463
994	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1464	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
995		1465
996	dec.flags = flags;	1466	dec.json = *json;
997	dec.cur = SvPVX (string);	1467	dec.cur = SvPVX (string);
998	dec.end = SvEND (string);	1468	dec.end = SvEND (string);
999	dec.err = 0;	1469	dec.err = 0;
1000	dec.depth = 0;	1470	dec.depth = 0;
1001	dec.maxdepth = DEC_DEPTH (dec.flags);
1002		1471
		1472	if (dec.json.cb_object || dec.json.cb_sk_object)
		1473	dec.json.flags |= F_HOOK;
		1474
1003	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1475	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1476
		1477	decode_ws (&dec);
1004	sv = decode_sv (&dec);	1478	sv = decode_sv (&dec);
1005		1479
		1480	if (offset_return)
		1481	*offset_return = dec.cur;
		1482
		1483	if (!(offset_return || !sv))
		1484	{
		1485	// check for trailing garbage
		1486	decode_ws (&dec);
		1487
		1488	if (*dec.cur)
		1489	{
		1490	dec.err = "garbage after JSON object";
		1491	SvREFCNT_dec (sv);
		1492	sv = 0;
		1493	}
		1494	}
		1495
1006	if (!sv)	1496	if (!sv)
1007	{	1497	{
1008	IV offset = dec.flags & F_UTF8
1009	? dec.cur - SvPVX (string)
1010	: utf8_distance (dec.cur, SvPVX (string));
1011	SV *uni = sv_newmortal ();	1498	SV *uni = sv_newmortal ();
1012		1499
1013	// horrible hack to silence warning inside pv_uni_display	1500	// horrible hack to silence warning inside pv_uni_display
1014	COP cop = *PL_curcop;	1501	COP cop = *PL_curcop;
1015	cop.cop_warnings = pWARN_NONE;	1502	cop.cop_warnings = pWARN_NONE;
…		…
1017	SAVEVPTR (PL_curcop);	1504	SAVEVPTR (PL_curcop);
1018	PL_curcop = &cop;	1505	PL_curcop = &cop;
1019	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1506	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1020	LEAVE;	1507	LEAVE;
1021		1508
1022	croak ("%s, at character offset %d [\"%s\"]",	1509	croak ("%s, at character offset %d (before \"%s\")",
1023	dec.err,	1510	dec.err,
1024	(int)offset,	1511	ptr_to_index (string, dec.cur),
1025	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1512	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1026	}	1513	}
1027		1514
1028	sv = sv_2mortal (sv);	1515	sv = sv_2mortal (sv);
1029		1516
1030	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1517	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1031	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1518	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1032		1519
1033	return sv;	1520	return sv;
		1521	}
		1522
		1523	/////////////////////////////////////////////////////////////////////////////
		1524	// incremental parser
		1525
		1526	static void
		1527	incr_parse (JSON *self)
		1528	{
		1529	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1530
		1531	for (;;)
		1532	{
		1533	//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
		1534	switch (self->incr_mode)
		1535	{
		1536	// only used for intiial whitespace skipping
		1537	case INCR_M_WS:
		1538	for (;;)
		1539	{
		1540	if (*p > 0x20)
		1541	{
		1542	self->incr_mode = INCR_M_JSON;
		1543	goto incr_m_json;
		1544	}
		1545	else if (!*p)
		1546	goto interrupt;
		1547
		1548	++p;
		1549	}
		1550
		1551	// skip a single char inside a string (for \\-processing)
		1552	case INCR_M_BS:
		1553	if (!*p)
		1554	goto interrupt;
		1555
		1556	++p;
		1557	self->incr_mode = INCR_M_STR;
		1558	goto incr_m_str;
		1559
		1560	// inside a string
		1561	case INCR_M_STR:
		1562	incr_m_str:
		1563	for (;;)
		1564	{
		1565	if (*p == '"')
		1566	{
		1567	++p;
		1568	self->incr_mode = INCR_M_JSON;
		1569
		1570	if (!self->incr_nest)
		1571	goto interrupt;
		1572
		1573	goto incr_m_json;
		1574	}
		1575	else if (*p == '\\')
		1576	{
		1577	++p; // "virtually" consumes character after \
		1578
		1579	if (!*p) // if at end of string we have to switch modes
		1580	{
		1581	self->incr_mode = INCR_M_BS;
		1582	goto interrupt;
		1583	}
		1584	}
		1585	else if (!*p)
		1586	goto interrupt;
		1587
		1588	++p;
		1589	}
		1590
		1591	// after initial ws, outside string
		1592	case INCR_M_JSON:
		1593	incr_m_json:
		1594	for (;;)
		1595	{
		1596	switch (*p++)
		1597	{
		1598	case 0:
		1599	--p;
		1600	goto interrupt;
		1601
		1602	case 0x09:
		1603	case 0x0a:
		1604	case 0x0d:
		1605	case 0x20:
		1606	if (!self->incr_nest)
		1607	{
		1608	--p; // do not eat the whitespace, let the next round do it
		1609	goto interrupt;
		1610	}
		1611	break;
		1612
		1613	case '"':
		1614	self->incr_mode = INCR_M_STR;
		1615	goto incr_m_str;
		1616
		1617	case '[':
		1618	case '{':
		1619	if (++self->incr_nest > self->max_depth)
		1620	croak (ERR_NESTING_EXCEEDED);
		1621	break;
		1622
		1623	case ']':
		1624	case '}':
		1625	if (--self->incr_nest <= 0)
		1626	goto interrupt;
		1627	}
		1628	}
		1629	}
		1630
		1631	modechange:
		1632	;
		1633	}
		1634
		1635	interrupt:
		1636	self->incr_pos = p - SvPVX (self->incr_text);
		1637	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1034	}	1638	}
1035		1639
1036	/////////////////////////////////////////////////////////////////////////////	1640	/////////////////////////////////////////////////////////////////////////////
1037	// XS interface functions	1641	// XS interface functions
1038		1642
1039	MODULE = JSON::XS PACKAGE = JSON::XS	1643	MODULE = JSON::XS PACKAGE = JSON::XS
1040		1644
1041	BOOT:	1645	BOOT:
1042	{	1646	{
1043	int i;	1647	int i;
1044
1045	memset (decode_hexdigit, 0xff, 256);
1046		1648
1047	for (i = 0; i < 256; ++i)	1649	for (i = 0; i < 256; ++i)
1048	decode_hexdigit [i] =	1650	decode_hexdigit [i] =
1049	i >= '0' && i <= '9' ? i - '0'	1651	i >= '0' && i <= '9' ? i - '0'
1050	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1652	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1051	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1653	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1052	: -1;	1654	: -1;
1053		1655
1054	json_stash = gv_stashpv ("JSON::XS", 1);	1656	json_stash = gv_stashpv ("JSON::XS" , 1);
		1657	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1658
		1659	json_true = get_bool ("JSON::XS::true");
		1660	json_false = get_bool ("JSON::XS::false");
		1661
		1662	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1055	}	1663	}
1056		1664
1057	PROTOTYPES: DISABLE	1665	PROTOTYPES: DISABLE
1058		1666
1059	SV *new (char *dummy)	1667	void CLONE (...)
1060	CODE:	1668	CODE:
1061	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1669	json_stash = 0;
		1670	json_boolean_stash = 0;
		1671
		1672	void new (char *klass)
		1673	PPCODE:
		1674	{
		1675	SV *pv = NEWSV (0, sizeof (JSON));
		1676	SvPOK_only (pv);
		1677	json_init ((JSON *)SvPVX (pv));
		1678	XPUSHs (sv_2mortal (sv_bless (
		1679	newRV_noinc (pv),
		1680	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1681	)));
		1682	}
		1683
		1684	void ascii (JSON *self, int enable = 1)
		1685	ALIAS:
		1686	ascii = F_ASCII
		1687	latin1 = F_LATIN1
		1688	utf8 = F_UTF8
		1689	indent = F_INDENT
		1690	canonical = F_CANONICAL
		1691	space_before = F_SPACE_BEFORE
		1692	space_after = F_SPACE_AFTER
		1693	pretty = F_PRETTY
		1694	allow_nonref = F_ALLOW_NONREF
		1695	shrink = F_SHRINK
		1696	allow_blessed = F_ALLOW_BLESSED
		1697	convert_blessed = F_CONV_BLESSED
		1698	relaxed = F_RELAXED
		1699	allow_unknown = F_ALLOW_UNKNOWN
		1700	PPCODE:
		1701	{
		1702	if (enable)
		1703	self->flags |= ix;
		1704	else
		1705	self->flags &= ~ix;
		1706
		1707	XPUSHs (ST (0));
		1708	}
		1709
		1710	void get_ascii (JSON *self)
		1711	ALIAS:
		1712	get_ascii = F_ASCII
		1713	get_latin1 = F_LATIN1
		1714	get_utf8 = F_UTF8
		1715	get_indent = F_INDENT
		1716	get_canonical = F_CANONICAL
		1717	get_space_before = F_SPACE_BEFORE
		1718	get_space_after = F_SPACE_AFTER
		1719	get_allow_nonref = F_ALLOW_NONREF
		1720	get_shrink = F_SHRINK
		1721	get_allow_blessed = F_ALLOW_BLESSED
		1722	get_convert_blessed = F_CONV_BLESSED
		1723	get_relaxed = F_RELAXED
		1724	get_allow_unknown = F_ALLOW_UNKNOWN
		1725	PPCODE:
		1726	XPUSHs (boolSV (self->flags & ix));
		1727
		1728	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1729	PPCODE:
		1730	self->max_depth = max_depth;
		1731	XPUSHs (ST (0));
		1732
		1733	U32 get_max_depth (JSON *self)
		1734	CODE:
		1735	RETVAL = self->max_depth;
1062	OUTPUT:	1736	OUTPUT:
1063	RETVAL	1737	RETVAL
1064		1738
1065	SV *ascii (SV *self, int enable = 1)	1739	void max_size (JSON *self, U32 max_size = 0)
1066	ALIAS:	1740	PPCODE:
1067	ascii = F_ASCII	1741	self->max_size = max_size;
1068	utf8 = F_UTF8	1742	XPUSHs (ST (0));
1069	indent = F_INDENT	1743
1070	canonical = F_CANONICAL	1744	int get_max_size (JSON *self)
1071	space_before = F_SPACE_BEFORE
1072	space_after = F_SPACE_AFTER
1073	pretty = F_PRETTY
1074	allow_nonref = F_ALLOW_NONREF
1075	shrink = F_SHRINK
1076	CODE:	1745	CODE:
1077	{	1746	RETVAL = self->max_size;
1078	UV *uv = SvJSON (self);
1079	if (enable)
1080	*uv |= ix;
1081	else
1082	*uv &= ~ix;
1083
1084	RETVAL = newSVsv (self);
1085	}
1086	OUTPUT:	1747	OUTPUT:
1087	RETVAL	1748	RETVAL
1088		1749
1089	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1750	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1751	PPCODE:
		1752	{
		1753	SvREFCNT_dec (self->cb_object);
		1754	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1755
		1756	XPUSHs (ST (0));
		1757	}
		1758
		1759	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1760	PPCODE:
		1761	{
		1762	if (!self->cb_sk_object)
		1763	self->cb_sk_object = newHV ();
		1764
		1765	if (SvOK (cb))
		1766	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1767	else
		1768	{
		1769	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1770
		1771	if (!HvKEYS (self->cb_sk_object))
		1772	{
		1773	SvREFCNT_dec (self->cb_sk_object);
		1774	self->cb_sk_object = 0;
		1775	}
		1776	}
		1777
		1778	XPUSHs (ST (0));
		1779	}
		1780
		1781	void encode (JSON *self, SV *scalar)
		1782	PPCODE:
		1783	XPUSHs (encode_json (scalar, self));
		1784
		1785	void decode (JSON *self, SV *jsonstr)
		1786	PPCODE:
		1787	XPUSHs (decode_json (jsonstr, self, 0));
		1788
		1789	void decode_prefix (JSON *self, SV *jsonstr)
		1790	PPCODE:
		1791	{
		1792	char *offset;
		1793	EXTEND (SP, 2);
		1794	PUSHs (decode_json (jsonstr, self, &offset));
		1795	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1796	}
		1797
		1798	void incr_parse (JSON *self, SV *jsonstr = 0)
		1799	PPCODE:
		1800	{
		1801	if (!self->incr_text)
		1802	self->incr_text = newSVpvn ("", 0);
		1803
		1804	// append data, if any
		1805	if (jsonstr)
		1806	{
		1807	if (SvUTF8 (jsonstr))
		1808	{
		1809	if (!SvUTF8 (self->incr_text))
		1810	{
		1811	/* utf-8-ness differs, need to upgrade */
		1812	sv_utf8_upgrade (self->incr_text);
		1813
		1814	if (self->incr_pos)
		1815	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1816	- (U8 *)SvPVX (self->incr_text);
		1817	}
		1818	}
		1819	else if (SvUTF8 (self->incr_text))
		1820	sv_utf8_upgrade (jsonstr);
		1821
		1822	{
		1823	STRLEN len;
		1824	const char *str = SvPV (jsonstr, len);
		1825	STRLEN cur = SvCUR (self->incr_text);
		1826
		1827	if (SvLEN (self->incr_text) <= cur + len)
		1828	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1829
		1830	Move (str, SvEND (self->incr_text), len, char);
		1831	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1832	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1833	}
		1834	}
		1835
		1836	if (GIMME_V != G_VOID)
		1837	do
		1838	{
		1839	char *offset;
		1840
		1841	if (!INCR_DONE (self))
		1842	{
		1843	incr_parse (self);
		1844
		1845	if (self->incr_pos > self->max_size && self->max_size)
		1846	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1847	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1848
		1849	if (!INCR_DONE (self))
		1850	break;
		1851	}
		1852
		1853	XPUSHs (decode_json (self->incr_text, self, &offset));
		1854
		1855	self->incr_pos -= offset - SvPVX (self->incr_text);
		1856	self->incr_nest = 0;
		1857	self->incr_mode = 0;
		1858
		1859	sv_chop (self->incr_text, offset);
		1860	}
		1861	while (GIMME_V == G_ARRAY);
		1862	}
		1863
		1864	SV *incr_text (JSON *self)
		1865	ATTRS: lvalue
1090	CODE:	1866	CODE:
1091	{	1867	{
1092	UV *uv = SvJSON (self);	1868	if (self->incr_pos)
1093	UV log2 = 0;	1869	croak ("incr_text can not be called when the incremental parser already started parsing");
1094		1870
1095	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1871	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1096
1097	while ((1UL << log2) < max_depth)
1098	++log2;
1099
1100	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1101
1102	RETVAL = newSVsv (self);
1103	}	1872	}
1104	OUTPUT:	1873	OUTPUT:
1105	RETVAL	1874	RETVAL
1106		1875
		1876	void incr_skip (JSON *self)
		1877	CODE:
		1878	{
		1879	if (self->incr_pos)
		1880	{
		1881	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1882	self->incr_pos = 0;
		1883	self->incr_nest = 0;
		1884	self->incr_mode = 0;
		1885	}
		1886	}
		1887
		1888	void incr_reset (JSON *self)
		1889	CODE:
		1890	{
		1891	SvREFCNT_dec (self->incr_text);
		1892	self->incr_text = 0;
		1893	self->incr_pos = 0;
		1894	self->incr_nest = 0;
		1895	self->incr_mode = 0;
		1896	}
		1897
		1898	void DESTROY (JSON *self)
		1899	CODE:
		1900	SvREFCNT_dec (self->cb_sk_object);
		1901	SvREFCNT_dec (self->cb_object);
		1902	SvREFCNT_dec (self->incr_text);
		1903
		1904	PROTOTYPES: ENABLE
		1905
1107	void encode (SV *self, SV *scalar)	1906	void encode_json (SV *scalar)
		1907	ALIAS:
		1908	to_json_ = 0
		1909	encode_json = F_UTF8
1108	PPCODE:	1910	PPCODE:
		1911	{
		1912	JSON json;
		1913	json_init (&json);
		1914	json.flags |= ix;
1109	XPUSHs (encode_json (scalar, *SvJSON (self)));	1915	XPUSHs (encode_json (scalar, &json));
		1916	}
1110		1917
1111	void decode (SV *self, SV *jsonstr)	1918	void decode_json (SV *jsonstr)
		1919	ALIAS:
		1920	from_json_ = 0
		1921	decode_json = F_UTF8
1112	PPCODE:	1922	PPCODE:
		1923	{
		1924	JSON json;
		1925	json_init (&json);
		1926	json.flags |= ix;
1113	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1927	XPUSHs (decode_json (jsonstr, &json, 0));
		1928	}
1114		1929
1115	PROTOTYPES: ENABLE
1116		1930
1117	void to_json (SV *scalar)
1118	ALIAS:
1119	objToJson = 0
1120	PPCODE:
1121	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1122
1123	void from_json (SV *jsonstr)
1124	ALIAS:
1125	jsonToObj = 0
1126	PPCODE:
1127	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1128

Diff Legend

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