ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.88 by root, Sun Jun 22 12:05:49 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
		32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36	#define F_ALLOW_UNKNOWN 0x00002000UL
		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		38
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0
20		40
21	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
		42	#define INDENT_STEP 3 // spaces per indentation level
		43
		44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		45
23	#define SB do {	46	#define SB do {
24	#define SE } while (0)	47	#define SE } while (0)
25		48
26	static HV *json_stash;	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
		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	}
		107
		108	/////////////////////////////////////////////////////////////////////////////
		109	// utility functions
		110
		111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
		123	shrink (SV *sv)
		124	{
		125	sv_utf8_downgrade (sv, 1);
		126
		127	if (SvLEN (sv) > SvCUR (sv) + 1)
		128	{
		129	#ifdef SvPV_shrink_to_cur
		130	SvPV_shrink_to_cur (sv);
		131	#elif defined (SvPV_renew)
		132	SvPV_renew (sv, SvCUR (sv) + 1);
		133	#endif
		134	}
		135	}
		136
		137	// decode an utf-8 character and return it, or (UV)-1 in
		138	// case of an error.
		139	// we special-case "safe" characters from U+80 .. U+7FF,
		140	// but use the very good perl function to parse anything else.
		141	// note that we never call this function for a ascii codepoints
		142	INLINE UV
		143	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		144	{
		145	if (expect_true (len >= 2
		146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		148	{
		149	*clen = 2;
		150	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		151	}
		152	else
		153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		154	}
		155
		156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 | ( ch >> 6),
		166	*s++ = 0x80 | ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 | ( ch >> 12),
		169	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 | ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 | ( ch >> 18),
		173	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176
		177	return s;
		178	}
		179
		180	/////////////////////////////////////////////////////////////////////////////
		181	// encoder
27		182
28	// structure used for encoding JSON	183	// structure used for encoding JSON
29	typedef struct	184	typedef struct
30	{	185	{
31	char *cur;	186	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
34	SV *sv;	188	SV *sv; // result scalar
35	UV flags;	189	JSON json;
36	int max_recurse;	190	U32 indent; // indentation level
37	int indent;	191	UV limit; // escape character values >= this value when encoding
38	} enc_t;	192	} enc_t;
39		193
40	// structure used for decoding JSON	194	INLINE void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	/////////////////////////////////////////////////////////////////////////////
59
60	static void
61	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
62	{	196	{
63	if (enc->cur + len >= enc->end)	197	if (expect_false (enc->cur + len >= enc->end))
64	{	198	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	203	}
70	}	204	}
71		205
72	static void	206	INLINE void
73	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
74	{	208	{
75	need (enc, 1);	209	need (enc, 1);
76	*enc->cur++ = ch;	210	*enc->cur++ = ch;
77	}	211	}
…		…
85		219
86	while (str < end)	220	while (str < end)
87	{	221	{
88	unsigned char ch = *(unsigned char *)str;	222	unsigned char ch = *(unsigned char *)str;
89		223
90	if (ch == '"')	224	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	225	{
		226	if (expect_false (ch == '"')) // but with slow exceptions
		227	{
92	need (enc, len += 1);	228	need (enc, len += 1);
93	*enc->cur++ = '\\';	229	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	230	*enc->cur++ = '"';
95	++str;
96	}	231	}
97	else if (ch == '\\')	232	else if (expect_false (ch == '\\'))
98	{	233	{
99	need (enc, len += 1);	234	need (enc, len += 1);
100	*enc->cur++ = '\\';	235	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	236	*enc->cur++ = '\\';
102	++str;
103	}	237	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	238	else
105	{
106	*enc->cur++ = ch;	239	*enc->cur++ = ch;
107	++str;	240
108	}
109	else if (ch == '\015')
110	{
111	need (enc, len += 1);
112	*enc->cur++ = '\\';
113	*enc->cur++ = 'r';
114	++str;
115	}
116	else if (ch == '\012')
117	{
118	need (enc, len += 1);
119	*enc->cur++ = '\\';
120	*enc->cur++ = 'n';
121	++str;	241	++str;
122	}	242	}
123	else	243	else
124	{	244	{
125	STRLEN clen;	245	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{	246	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	247	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131	if (clen < 0)	248	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");	249	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133	}	250	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134	else	251	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135	{
136	uch = ch;
137	clen = 1;
138	}
139		252
140	if (uch < 0x80 || enc->flags & F_ASCII)	253	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	254	{
		255	STRLEN clen;
		256	UV uch;
		257
		258	if (is_utf8)
		259	{
		260	uch = decode_utf8 (str, end - str, &clen);
		261	if (clen == (STRLEN)-1)
		262	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		263	}
		264	else
		265	{
		266	uch = ch;
		267	clen = 1;
		268	}
		269
		270	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		271	{
		272	if (uch >= 0x10000UL)
		273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
144	need (enc, len += 11);	277	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	281	enc->cur += 12;
		282	}
		283	else
		284	{
		285	static char hexdigit [16] = "0123456789abcdef";
		286	need (enc, len += 5);
		287	*enc->cur++ = '\\';
		288	*enc->cur++ = 'u';
		289	*enc->cur++ = hexdigit [ uch >> 12 ];
		290	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		291	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		292	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		293	}
		294
		295	str += clen;
		296	}
		297	else if (enc->json.flags & F_LATIN1)
		298	{
		299	*enc->cur++ = uch;
		300	str += clen;
		301	}
		302	else if (is_utf8)
		303	{
		304	need (enc, len += clen);
		305	do
		306	{
		307	*enc->cur++ = *str++;
		308	}
		309	while (--clen);
		310	}
		311	else
		312	{
		313	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		314	enc->cur = encode_utf8 (enc->cur, uch);
		315	++str;
		316	}
149	}	317	}
150	else
151	{
152	static char hexdigit [16] = "0123456789abcdef";
153	need (enc, len += 5);
154	*enc->cur++ = '\\';
155	*enc->cur++ = 'u';
156	*enc->cur++ = hexdigit [ uch >> 12 ];
157	*enc->cur++ = hexdigit [(uch >> 8) & 15];
158	*enc->cur++ = hexdigit [(uch >> 4) & 15];
159	*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	}
161
162	str += clen;
163	}
164	else if (is_utf8)
165	{
166	need (enc, len += clen);
167	while (clen--)
168	*enc->cur++ = *str++;
169	}
170	else
171	{
172	need (enc, 10); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174	++str;
175	}	318	}
176	}	319	}
177		320
178	--len;	321	--len;
179	}	322	}
180	}	323	}
181		324
182	#define INDENT SB \	325	INLINE void
		326	encode_indent (enc_t *enc)
		327	{
183	if (enc->flags & F_INDENT) \	328	if (enc->json.flags & F_INDENT)
184	{ \	329	{
185	int i_; \	330	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	331
187	for (i_ = enc->indent * 3; i_--; )\	332	need (enc, spaces);
		333	memset (enc->cur, ' ', spaces);
		334	enc->cur += spaces;
		335	}
		336	}
		337
		338	INLINE void
		339	encode_space (enc_t *enc)
		340	{
		341	need (enc, 1);
		342	encode_ch (enc, ' ');
		343	}
		344
		345	INLINE void
		346	encode_nl (enc_t *enc)
		347	{
		348	if (enc->json.flags & F_INDENT)
		349	{
		350	need (enc, 1);
188	encode_ch (enc, ' '); \	351	encode_ch (enc, '\n');
189	} \	352	}
190	SE	353	}
191		354
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	355	INLINE void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	356	encode_comma (enc_t *enc)
194	#define COMMA SB \	357	{
195	encode_ch (enc, ','); \	358	encode_ch (enc, ',');
		359
196	if (enc->flags & F_INDENT) \	360	if (enc->json.flags & F_INDENT)
197	NL; \	361	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	362	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	363	encode_space (enc);
200	SE	364	}
201		365
202	static void encode_sv (enc_t *enc, SV *sv);	366	static void encode_sv (enc_t *enc, SV *sv);
203		367
204	static void	368	static void
205	encode_av (enc_t *enc, AV *av)	369	encode_av (enc_t *enc, AV *av)
206	{	370	{
207	int i, len = av_len (av);	371	int i, len = av_len (av);
208		372
		373	if (enc->indent >= enc->json.max_depth)
		374	croak (ERR_NESTING_EXCEEDED);
		375
209	encode_ch (enc, '['); NL;	376	encode_ch (enc, '[');
210	++enc->indent;	377
		378	if (len >= 0)
		379	{
		380	encode_nl (enc); ++enc->indent;
211		381
212	for (i = 0; i <= len; ++i)	382	for (i = 0; i <= len; ++i)
213	{	383	{
214	INDENT;	384	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));
216		385
		386	encode_indent (enc);
		387
		388	if (svp)
		389	encode_sv (enc, *svp);
		390	else
		391	encode_str (enc, "null", 4, 0);
		392
217	if (i < len)	393	if (i < len)
218	COMMA;	394	encode_comma (enc);
219	}	395	}
220		396
221	NL;	397	encode_nl (enc); --enc->indent; encode_indent (enc);
222		398	}
223	--enc->indent;	399
224	INDENT; encode_ch (enc, ']');	400	encode_ch (enc, ']');
225	}	401	}
226		402
227	static void	403	static void
228	encode_he (enc_t *enc, HE *he)	404	encode_hk (enc_t *enc, HE *he)
229	{	405	{
230	encode_ch (enc, '"');	406	encode_ch (enc, '"');
231		407
232	if (HeKLEN (he) == HEf_SVKEY)	408	if (HeKLEN (he) == HEf_SVKEY)
233	{	409	{
…		…
243	else	419	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	420	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		421
246	encode_ch (enc, '"');	422	encode_ch (enc, '"');
247		423
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	424	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	425	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	426	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));
252	}	427	}
253		428
254	// compare hash entries, used when all keys are bytestrings	429	// compare hash entries, used when all keys are bytestrings
255	static int	430	static int
256	he_cmp_fast (const void *a_, const void *b_)	431	he_cmp_fast (const void *a_, const void *b_)
…		…
261	HE *b = *(HE **)b_;	436	HE *b = *(HE **)b_;
262		437
263	STRLEN la = HeKLEN (a);	438	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	439	STRLEN lb = HeKLEN (b);
265		440
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	441	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	442	cmp = lb - la;
268		443
269	return cmp;	444	return cmp;
270	}	445	}
271		446
272	// compare hash entries, used when some keys are sv's or utf-x	447	// compare hash entries, used when some keys are sv's or utf-x
273	static int	448	static int
274	he_cmp_slow (const void *a, const void *b)	449	he_cmp_slow (const void *a, const void *b)
275	{	450	{
276	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	451	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277	}	452	}
278		453
279	static void	454	static void
280	encode_hv (enc_t *enc, HV *hv)	455	encode_hv (enc_t *enc, HV *hv)
281	{	456	{
282	int count, i;	457	HE *he;
283		458
284	encode_ch (enc, '{'); NL; ++enc->indent;	459	if (enc->indent >= enc->json.max_depth)
		460	croak (ERR_NESTING_EXCEEDED);
285		461
286	if ((count = hv_iterinit (hv)))	462	encode_ch (enc, '{');
287	{	463
288	// for canonical output we have to sort by keys first	464	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	465	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	466	// security workaround added somewhere in 5.8.x.
291	// that randomises hash orderings	467	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)	468	if (enc->json.flags & F_CANONICAL)
		469	{
		470	int count = hv_iterinit (hv);
		471
		472	if (SvMAGICAL (hv))
293	{	473	{
294	HE *he, *hes [count];	474	// need to count by iterating. could improve by dynamically building the vector below
		475	// but I don't care for the speed of this special case.
		476	// note also that we will run into undefined behaviour when the two iterations
		477	// do not result in the same count, something I might care for in some later release.
		478
		479	count = 0;
		480	while (hv_iternext (hv))
		481	++count;
		482
		483	hv_iterinit (hv);
		484	}
		485
		486	if (count)
		487	{
295	int fast = 1;	488	int i, fast = 1;
		489	#if defined(__BORLANDC__) || defined(_MSC_VER)
		490	HE **hes = _alloca (count * sizeof (HE));
		491	#else
		492	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		493	#endif
296		494
297	i = 0;	495	i = 0;
298	while ((he = hv_iternext (hv)))	496	while ((he = hv_iternext (hv)))
299	{	497	{
300	hes [i++] = he;	498	hes [i++] = he;
…		…
306		504
307	if (fast)	505	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	506	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	507	else
310	{	508	{
311	// hack to disable "use bytes"	509	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	510	COP cop = *PL_curcop;
313	cop.op_private = 0;	511	cop.op_private = 0;
		512
		513	ENTER;
		514	SAVETMPS;
		515
		516	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	517	PL_curcop = &cop;
315		518
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	519	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		520
318	FREETMPS;	521	FREETMPS;
319		522	LEAVE;
320	PL_curcop = oldcop;
321	}	523	}
322		524
323	for (i = 0; i < count; ++i)	525	encode_nl (enc); ++enc->indent;
		526
		527	while (count--)
324	{	528	{
325	INDENT;	529	encode_indent (enc);
		530	he = hes [count];
326	encode_he (enc, hes [i]);	531	encode_hk (enc, he);
		532	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327		533
328	if (i < count - 1)	534	if (count)
329	COMMA;	535	encode_comma (enc);
330	}	536	}
331		537
		538	encode_nl (enc); --enc->indent; encode_indent (enc);
		539	}
		540	}
		541	else
		542	{
		543	if (hv_iterinit (hv) || SvMAGICAL (hv))
		544	if ((he = hv_iternext (hv)))
332	NL;	545	{
		546	encode_nl (enc); ++enc->indent;
		547
		548	for (;;)
		549	{
		550	encode_indent (enc);
		551	encode_hk (enc, he);
		552	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		553
		554	if (!(he = hv_iternext (hv)))
		555	break;
		556
		557	encode_comma (enc);
		558	}
		559
		560	encode_nl (enc); --enc->indent; encode_indent (enc);
		561	}
		562	}
		563
		564	encode_ch (enc, '}');
		565	}
		566
		567	// encode objects, arrays and special \0=false and \1=true values.
		568	static void
		569	encode_rv (enc_t *enc, SV *sv)
		570	{
		571	svtype svt;
		572
		573	SvGETMAGIC (sv);
		574	svt = SvTYPE (sv);
		575
		576	if (expect_false (SvOBJECT (sv)))
		577	{
		578	HV *stash = !JSON_SLOW || json_boolean_stash
		579	? json_boolean_stash
		580	: gv_stashpv ("JSON::XS::Boolean", 1);
		581
		582	if (SvSTASH (sv) == stash)
		583	{
		584	if (SvIV (sv))
		585	encode_str (enc, "true", 4, 0);
		586	else
		587	encode_str (enc, "false", 5, 0);
333	}	588	}
334	else	589	else
335	{	590	{
336	SV *sv;	591	#if 0
337	HE *he = hv_iternext (hv);	592	if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339	for (;;)
340	{	593	{
341	INDENT;	594	// not yet
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}	595	}
349		596	#endif
		597	if (enc->json.flags & F_CONV_BLESSED)
350	NL;	598	{
		599	// we re-bless the reference to get overload and other niceties right
		600	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		601
		602	if (to_json)
		603	{
		604	dSP;
		605
		606	ENTER; SAVETMPS; PUSHMARK (SP);
		607	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		608
		609	// calling with G_SCALAR ensures that we always get a 1 return value
		610	PUTBACK;
		611	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		612	SPAGAIN;
		613
		614	// catch this surprisingly common error
		615	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		616	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		617
		618	sv = POPs;
		619	PUTBACK;
		620
		621	encode_sv (enc, sv);
		622
		623	FREETMPS; LEAVE;
		624	}
		625	else if (enc->json.flags & F_ALLOW_BLESSED)
		626	encode_str (enc, "null", 4, 0);
		627	else
		628	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		629	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		630	}
		631	else if (enc->json.flags & F_ALLOW_BLESSED)
		632	encode_str (enc, "null", 4, 0);
		633	else
		634	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		635	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
351	}	636	}
352	}	637	}
		638	else if (svt == SVt_PVHV)
		639	encode_hv (enc, (HV *)sv);
		640	else if (svt == SVt_PVAV)
		641	encode_av (enc, (AV *)sv);
		642	else if (svt < SVt_PVAV)
		643	{
		644	STRLEN len = 0;
		645	char *pv = svt ? SvPV (sv, len) : 0;
353		646
354	--enc->indent; INDENT; encode_ch (enc, '}');	647	if (len == 1 && *pv == '1')
		648	encode_str (enc, "true", 4, 0);
		649	else if (len == 1 && *pv == '0')
		650	encode_str (enc, "false", 5, 0);
		651	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		652	encode_str (enc, "null", 4, 0);
		653	else
		654	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		655	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		656	}
		657	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		658	encode_str (enc, "null", 4, 0);
		659	else
		660	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		661	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	662	}
356		663
357	static void	664	static void
358	encode_sv (enc_t *enc, SV *sv)	665	encode_sv (enc_t *enc, SV *sv)
359	{	666	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	674	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	675	encode_ch (enc, '"');
369	}	676	}
370	else if (SvNOKp (sv))	677	else if (SvNOKp (sv))
371	{	678	{
		679	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	680	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	681	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	682	enc->cur += strlen (enc->cur);
375	}	683	}
376	else if (SvIOKp (sv))	684	else if (SvIOKp (sv))
377	{	685	{
		686	// we assume we can always read an IV as a UV and vice versa
		687	// we assume two's complement
		688	// we assume no aliasing issues in the union
		689	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		690	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		691	{
		692	// optimise the "small number case"
		693	// code will likely be branchless and use only a single multiplication
		694	// works for numbers up to 59074
		695	I32 i = SvIVX (sv);
		696	U32 u;
		697	char digit, nz = 0;
		698
378	need (enc, 64);	699	need (enc, 6);
		700
		701	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		702	u = i < 0 ? -i : i;
		703
		704	// convert to 4.28 fixed-point representation
		705	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		706
		707	// now output digit by digit, each time masking out the integer part
		708	// and multiplying by 5 while moving the decimal point one to the right,
		709	// resulting in a net multiplication by 10.
		710	// we always write the digit to memory but conditionally increment
		711	// the pointer, to enable the use of conditional move instructions.
		712	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		713	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		714	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		715	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		716	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		717	}
		718	else
		719	{
		720	// large integer, use the (rather slow) snprintf way.
		721	need (enc, IVUV_MAXCHARS);
379	enc->cur +=	722	enc->cur +=
380	SvIsUV(sv)	723	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	724	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	725	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		726	}
383	}	727	}
384	else if (SvROK (sv))	728	else if (SvROK (sv))
385	{	729	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)	730	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
387	croak ("data structure too deep (hit recursion limit)");
388
389	sv = SvRV (sv);
390
391	switch (SvTYPE (sv))
392	{
393	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396	default:
397	croak ("JSON can only represent references to arrays or hashes");
398	}
399	}
400	else if (!SvOK (sv))
401	encode_str (enc, "null", 4, 0);	731	encode_str (enc, "null", 4, 0);
402	else	732	else
403	croak ("encountered perl type that JSON cannot handle");	733	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		734	SvPV_nolen (sv), SvFLAGS (sv));
404	}	735	}
405		736
406	static SV *	737	static SV *
407	encode_json (SV *scalar, UV flags)	738	encode_json (SV *scalar, JSON *json)
408	{	739	{
409	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412	enc_t enc;	740	enc_t enc;
413	enc.flags = flags;	741
		742	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		743	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		744
		745	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	746	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	747	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	748	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	749	enc.indent = 0;
		750	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		751	: enc.json.flags & F_LATIN1 ? 0x000100UL
		752	: 0x110000UL;
419		753
420	SvPOK_only (enc.sv);	754	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	755	encode_sv (&enc, scalar);
422		756
		757	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		758	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		759
423	if (!(flags & (F_ASCII | F_UTF8)))	760	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	761	SvUTF8_on (enc.sv);
425		762
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	763	if (enc.json.flags & F_SHRINK)
		764	shrink (enc.sv);
		765
427	return enc.sv;	766	return enc.sv;
428	}	767	}
429		768
430	/////////////////////////////////////////////////////////////////////////////	769	/////////////////////////////////////////////////////////////////////////////
		770	// decoder
431		771
432	#define WS \	772	// structure used for decoding JSON
		773	typedef struct
		774	{
		775	char *cur; // current parser pointer
		776	char *end; // end of input string
		777	const char *err; // parse error, if != 0
		778	JSON json;
		779	U32 depth; // recursion depth
		780	U32 maxdepth; // recursion depth limit
		781	} dec_t;
		782
		783	INLINE void
		784	decode_comment (dec_t *dec)
		785	{
		786	// only '#'-style comments allowed a.t.m.
		787
		788	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		789	++dec->cur;
		790	}
		791
		792	INLINE void
		793	decode_ws (dec_t *dec)
		794	{
433	for (;;) \	795	for (;;)
434	{ \	796	{
435	char ch = *dec->cur; \	797	char ch = *dec->cur;
		798
436	if (ch > 0x20 \	799	if (ch > 0x20)
		800	{
		801	if (expect_false (ch == '#'))
		802	{
		803	if (dec->json.flags & F_RELAXED)
		804	decode_comment (dec);
		805	else
		806	break;
		807	}
		808	else
		809	break;
		810	}
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	811	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438	break; \	812	break; // parse error, but let higher level handle it, gives better error messages
		813
439	++dec->cur; \	814	++dec->cur;
440	}	815	}
		816	}
441		817
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	818	#define ERR(reason) SB dec->err = reason; goto fail; SE
		819
443	#define EXPECT_CH(ch) SB \	820	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	821	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	822	ERR (# ch " expected"); \
446	++dec->cur; \	823	++dec->cur; \
447	SE	824	SE
448		825
		826	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		827	#define DEC_DEC_DEPTH --dec->depth
		828
449	static SV *decode_sv (dec_t *dec);	829	static SV *decode_sv (dec_t *dec);
450		830
451	static signed char decode_hexdigit[256];	831	static signed char decode_hexdigit[256];
452		832
453	static UV	833	static UV
454	decode_4hex (dec_t *dec)	834	decode_4hex (dec_t *dec)
455	{	835	{
456	signed char d1, d2, d3, d4;	836	signed char d1, d2, d3, d4;
		837	unsigned char *cur = (unsigned char *)dec->cur;
457		838
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	839	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459	if (d1 < 0) ERR ("four hexadecimal digits expected");	840	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	841	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461	if (d2 < 0) ERR ("four hexadecimal digits expected");	842	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463	if (d3 < 0) ERR ("four hexadecimal digits expected");
464	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465	if (d4 < 0) ERR ("four hexadecimal digits expected");
466		843
467	dec->cur += 4;	844	dec->cur += 4;
468		845
469	return ((UV)d1) << 12	846	return ((UV)d1) << 12
470	| ((UV)d2) << 8	847	| ((UV)d2) << 8
…		…
473		850
474	fail:	851	fail:
475	return (UV)-1;	852	return (UV)-1;
476	}	853	}
477		854
478	#define APPEND_GROW(n) SB \
479	if (cur + (n) >= end) \
480	{ \
481	STRLEN ofs = cur - SvPVX (sv); \
482	SvGROW (sv, ofs + (n) + 1); \
483	cur = SvPVX (sv) + ofs; \
484	end = SvEND (sv); \
485	} \
486	SE
487
488	#define APPEND_CH(ch) SB \
489	APPEND_GROW (1); \
490	*cur++ = (ch); \
491	SE
492
493	static SV *	855	static SV *
494	decode_str (dec_t *dec)	856	decode_str (dec_t *dec)
495	{	857	{
496	SV *sv = NEWSV (0,2);	858	SV *sv = 0;
497	int utf8 = 0;	859	int utf8 = 0;
498	char *cur = SvPVX (sv);	860	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		861
501	for (;;)	862	do
502	{	863	{
503	unsigned char ch = *(unsigned char *)dec->cur;	864	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		865	char *cur = buf;
504		866
505	if (ch == '"')	867	do
506	break;
507	else if (ch == '\\')
508	{	868	{
509	switch (*++dec->cur)	869	unsigned char ch = *(unsigned char *)dec_cur++;
		870
		871	if (expect_false (ch == '"'))
510	{	872	{
511	case '\\':	873	--dec_cur;
512	case '/':	874	break;
513	case '"': APPEND_CH (*dec->cur++); break;	875	}
514		876	else if (expect_false (ch == '\\'))
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	877	{
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;	878	switch (*dec_cur)
517	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521	case 'u':
522	{	879	{
523	UV lo, hi;	880	case '\\':
524	++dec->cur;	881	case '/':
		882	case '"': *cur++ = *dec_cur++; break;
525		883
526	hi = decode_4hex (dec);	884	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	885	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	886	case 'n': ++dec_cur; *cur++ = '\012'; break;
		887	case 'f': ++dec_cur; *cur++ = '\014'; break;
		888	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		889
530	// possibly a surrogate pair	890	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	891	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	892	UV lo, hi;
534	ERR ("illegal surrogate character");	893	++dec_cur;
535		894
536	dec->cur += 2;	895	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	896	hi = decode_4hex (dec);
		897	dec_cur = dec->cur;
539	if (lo == (UV)-1)	898	if (hi == (UV)-1)
540	goto fail;	899	goto fail;
541		900
		901	// possibly a surrogate pair
		902	if (hi >= 0xd800)
		903	if (hi < 0xdc00)
		904	{
		905	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		906	ERR ("missing low surrogate character in surrogate pair");
		907
		908	dec_cur += 2;
		909
		910	dec->cur = dec_cur;
		911	lo = decode_4hex (dec);
		912	dec_cur = dec->cur;
		913	if (lo == (UV)-1)
		914	goto fail;
		915
542	if (lo < 0xdc00 || lo >= 0xe000)	916	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	917	ERR ("surrogate pair expected");
544		918
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	919	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		920	}
		921	else if (hi < 0xe000)
		922	ERR ("missing high surrogate character in surrogate pair");
		923
		924	if (hi >= 0x80)
		925	{
		926	utf8 = 1;
		927
		928	cur = encode_utf8 (cur, hi);
		929	}
		930	else
		931	*cur++ = hi;
546	}	932	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	933	break;
552	utf8 = 1;
553		934
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	935	default:
558	APPEND_CH (hi);	936	--dec_cur;
		937	ERR ("illegal backslash escape sequence in string");
559	}	938	}
		939	}
		940	else if (expect_true (ch >= 0x20 && ch < 0x80))
		941	*cur++ = ch;
		942	else if (ch >= 0x80)
		943	{
		944	STRLEN clen;
		945	UV uch;
		946
		947	--dec_cur;
		948
		949	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		950	if (clen == (STRLEN)-1)
		951	ERR ("malformed UTF-8 character in JSON string");
		952
		953	do
		954	*cur++ = *dec_cur++;
		955	while (--clen);
		956
		957	utf8 = 1;
		958	}
		959	else
		960	{
		961	--dec_cur;
		962
		963	if (!ch)
		964	ERR ("unexpected end of string while parsing JSON string");
560	break;	965	else
		966	ERR ("invalid character encountered while parsing JSON string");
561	}	967	}
562	}	968	}
563	else if (ch >= 0x20 && ch <= 0x7f)	969	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	970
565	else if (ch >= 0x80)	971	{
		972	STRLEN len = cur - buf;
		973
		974	if (sv)
566	{	975	{
567	STRLEN clen;	976	SvGROW (sv, SvCUR (sv) + len + 1);
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	977	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569	if (clen < 0)	978	SvCUR_set (sv, SvCUR (sv) + len);
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571
572	APPEND_GROW (clen);
573	memcpy (cur, dec->cur, clen);
574	cur += clen;
575	dec->cur += clen;
576	}	979	}
577	else	980	else
578	ERR ("invalid character encountered");	981	sv = newSVpvn (buf, len);
579	}	982	}
		983	}
		984	while (*dec_cur != '"');
580		985
581	++dec->cur;	986	++dec_cur;
582		987
583	SvCUR_set (sv, cur - SvPVX (sv));	988	if (sv)
584		989	{
585	SvPOK_only (sv);	990	SvPOK_only (sv);
586	*SvEND (sv) = 0;	991	*SvEND (sv) = 0;
587		992
588	if (utf8)	993	if (utf8)
589	SvUTF8_on (sv);	994	SvUTF8_on (sv);
		995	}
		996	else
		997	sv = newSVpvn ("", 0);
590		998
		999	dec->cur = dec_cur;
591	return sv;	1000	return sv;
592		1001
593	fail:	1002	fail:
594	SvREFCNT_dec (sv);	1003	dec->cur = dec_cur;
595	return 0;	1004	return 0;
596	}	1005	}
597		1006
598	static SV *	1007	static SV *
599	decode_num (dec_t *dec)	1008	decode_num (dec_t *dec)
…		…
609	{	1018	{
610	++dec->cur;	1019	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	1020	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	1021	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	1022	}
614		1023	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	1024	ERR ("malformed number (no digits after initial minus)");
		1025	else
		1026	do
		1027	{
		1028	++dec->cur;
		1029	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	1030	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		1031
619	// [frac]	1032	// [frac]
620	if (*dec->cur == '.')	1033	if (*dec->cur == '.')
621	{	1034	{
622	is_nv = 1;	1035	++dec->cur;
		1036
		1037	if (*dec->cur < '0' || *dec->cur > '9')
		1038	ERR ("malformed number (no digits after decimal point)");
623		1039
624	do	1040	do
625	{	1041	{
626	++dec->cur;	1042	++dec->cur;
627	}	1043	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	1044	while (*dec->cur >= '0' && *dec->cur <= '9');
		1045
		1046	is_nv = 1;
629	}	1047	}
630		1048
631	// [exp]	1049	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	1050	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	1051	{
634	is_nv = 1;
635
636	++dec->cur;	1052	++dec->cur;
		1053
637	if (*dec->cur == '-' || *dec->cur == '+')	1054	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	1055	++dec->cur;
639		1056
		1057	if (*dec->cur < '0' || *dec->cur > '9')
		1058	ERR ("malformed number (no digits after exp sign)");
		1059
		1060	do
		1061	{
		1062	++dec->cur;
		1063	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	1064	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	1065
		1066	is_nv = 1;
642	}	1067	}
643		1068
644	if (!is_nv)	1069	if (!is_nv)
645	{	1070	{
646	UV uv;	1071	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	1072
648	if (numtype & IS_NUMBER_IN_UV)	1073	// special case the rather common 1..5-digit-int case
649	if (numtype & IS_NUMBER_NEG)	1074	if (*start == '-')
		1075	switch (len)
650	{	1076	{
651	if (uv < (UV)IV_MIN)	1077	case 2: return newSViv (-( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	1078	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1079	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1080	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1081	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
653	}	1082	}
		1083	else
		1084	switch (len)
		1085	{
		1086	case 1: return newSViv ( start [0] - '0' * 1);
		1087	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1088	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1089	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1090	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1091	}
		1092
		1093	{
		1094	UV uv;
		1095	int numtype = grok_number (start, len, &uv);
		1096	if (numtype & IS_NUMBER_IN_UV)
		1097	if (numtype & IS_NUMBER_NEG)
		1098	{
		1099	if (uv < (UV)IV_MIN)
		1100	return newSViv (-(IV)uv);
		1101	}
654	else	1102	else
655	return newSVuv (uv);	1103	return newSVuv (uv);
656	}	1104	}
657		1105
		1106	len -= *start == '-' ? 1 : 0;
		1107
		1108	// does not fit into IV or UV, try NV
		1109	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1110	#if defined (LDBL_DIG)
		1111	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1112	#endif
		1113	)
		1114	// fits into NV without loss of precision
		1115	return newSVnv (Atof (start));
		1116
		1117	// everything else fails, convert it to a string
		1118	return newSVpvn (start, dec->cur - start);
		1119	}
		1120
		1121	// loss of precision here
658	return newSVnv (Atof (start));	1122	return newSVnv (Atof (start));
659		1123
660	fail:	1124	fail:
661	return 0;	1125	return 0;
662	}	1126	}
…		…
664	static SV *	1128	static SV *
665	decode_av (dec_t *dec)	1129	decode_av (dec_t *dec)
666	{	1130	{
667	AV *av = newAV ();	1131	AV *av = newAV ();
668		1132
		1133	DEC_INC_DEPTH;
		1134	decode_ws (dec);
		1135
		1136	if (*dec->cur == ']')
		1137	++dec->cur;
		1138	else
669	for (;;)	1139	for (;;)
670	{	1140	{
671	SV *value;	1141	SV *value;
672		1142
673	value = decode_sv (dec);	1143	value = decode_sv (dec);
674	if (!value)	1144	if (!value)
675	goto fail;	1145	goto fail;
676		1146
677	av_push (av, value);	1147	av_push (av, value);
678		1148
679	WS;	1149	decode_ws (dec);
680		1150
681	if (*dec->cur == ']')	1151	if (*dec->cur == ']')
682	{	1152	{
683	++dec->cur;	1153	++dec->cur;
684	break;	1154	break;
		1155	}
685	}	1156
686
687	if (*dec->cur != ',')	1157	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1158	ERR (", or ] expected while parsing array");
689		1159
690	++dec->cur;	1160	++dec->cur;
		1161
		1162	decode_ws (dec);
		1163
		1164	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1165	{
		1166	++dec->cur;
		1167	break;
		1168	}
691	}	1169	}
692		1170
		1171	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1172	return newRV_noinc ((SV *)av);
694		1173
695	fail:	1174	fail:
696	SvREFCNT_dec (av);	1175	SvREFCNT_dec (av);
		1176	DEC_DEC_DEPTH;
697	return 0;	1177	return 0;
698	}	1178	}
699		1179
700	static SV *	1180	static SV *
701	decode_hv (dec_t *dec)	1181	decode_hv (dec_t *dec)
702	{	1182	{
		1183	SV *sv;
703	HV *hv = newHV ();	1184	HV *hv = newHV ();
704		1185
		1186	DEC_INC_DEPTH;
		1187	decode_ws (dec);
		1188
		1189	if (*dec->cur == '}')
		1190	++dec->cur;
		1191	else
705	for (;;)	1192	for (;;)
706	{	1193	{
707	SV *key, *value;
708
709	WS; EXPECT_CH ('"');	1194	EXPECT_CH ('"');
710		1195
711	key = decode_str (dec);	1196	// heuristic: assume that
712	if (!key)	1197	// a) decode_str + hv_store_ent are abysmally slow.
713	goto fail;	1198	// b) most hash keys are short, simple ascii text.
714		1199	// => try to "fast-match" such strings to avoid
715	WS; EXPECT_CH (':');	1200	// the overhead of decode_str + hv_store_ent.
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1201	{
		1202	SV *value;
		1203	char *p = dec->cur;
		1204	char *e = p + 24; // only try up to 24 bytes
		1205
		1206	for (;;)
		1207	{
		1208	// the >= 0x80 is false on most architectures
		1209	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1210	{
		1211	// slow path, back up and use decode_str
		1212	SV *key = decode_str (dec);
		1213	if (!key)
		1214	goto fail;
		1215
		1216	decode_ws (dec); EXPECT_CH (':');
		1217
		1218	decode_ws (dec);
		1219	value = decode_sv (dec);
		1220	if (!value)
		1221	{
		1222	SvREFCNT_dec (key);
		1223	goto fail;
		1224	}
		1225
		1226	hv_store_ent (hv, key, value, 0);
720	SvREFCNT_dec (key);	1227	SvREFCNT_dec (key);
		1228
		1229	break;
		1230	}
		1231	else if (*p == '"')
		1232	{
		1233	// fast path, got a simple key
		1234	char *key = dec->cur;
		1235	int len = p - key;
		1236	dec->cur = p + 1;
		1237
		1238	decode_ws (dec); EXPECT_CH (':');
		1239
		1240	decode_ws (dec);
		1241	value = decode_sv (dec);
		1242	if (!value)
721	goto fail;	1243	goto fail;
		1244
		1245	hv_store (hv, key, len, value, 0);
		1246
		1247	break;
		1248	}
		1249
		1250	++p;
		1251	}
722	}	1252	}
723		1253
724	//TODO: optimise	1254	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1255
727	WS;
728
729	if (*dec->cur == '}')	1256	if (*dec->cur == '}')
		1257	{
		1258	++dec->cur;
		1259	break;
		1260	}
		1261
		1262	if (*dec->cur != ',')
		1263	ERR (", or } expected while parsing object/hash");
		1264
		1265	++dec->cur;
		1266
		1267	decode_ws (dec);
		1268
		1269	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
		1274	}
		1275
		1276	DEC_DEC_DEPTH;
		1277	sv = newRV_noinc ((SV *)hv);
		1278
		1279	// check filter callbacks
		1280	if (dec->json.flags & F_HOOK)
		1281	{
		1282	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730	{	1283	{
731	++dec->cur;	1284	HE *cb, *he;
732	break;	1285
		1286	hv_iterinit (hv);
		1287	he = hv_iternext (hv);
		1288	hv_iterinit (hv);
		1289
		1290	// the next line creates a mortal sv each time its called.
		1291	// might want to optimise this for common cases.
		1292	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1293
		1294	if (cb)
		1295	{
		1296	dSP;
		1297	int count;
		1298
		1299	ENTER; SAVETMPS; PUSHMARK (SP);
		1300	XPUSHs (HeVAL (he));
		1301
		1302	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1303
		1304	if (count == 1)
		1305	{
		1306	sv = newSVsv (POPs);
		1307	FREETMPS; LEAVE;
		1308	return sv;
		1309	}
		1310
		1311	FREETMPS; LEAVE;
		1312	}
733	}	1313	}
734		1314
735	if (*dec->cur != ',')	1315	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1316	{
		1317	dSP;
		1318	int count;
737		1319
738	++dec->cur;	1320	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1321	XPUSHs (sv_2mortal (sv));
740		1322
741	return newRV_noinc ((SV *)hv);	1323	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1324
		1325	if (count == 1)
		1326	{
		1327	sv = newSVsv (POPs);
		1328	FREETMPS; LEAVE;
		1329	return sv;
		1330	}
		1331
		1332	SvREFCNT_inc (sv);
		1333	FREETMPS; LEAVE;
		1334	}
		1335	}
		1336
		1337	return sv;
742		1338
743	fail:	1339	fail:
744	SvREFCNT_dec (hv);	1340	SvREFCNT_dec (hv);
		1341	DEC_DEC_DEPTH;
745	return 0;	1342	return 0;
746	}	1343	}
747		1344
748	static SV *	1345	static SV *
749	decode_sv (dec_t *dec)	1346	decode_sv (dec_t *dec)
750	{	1347	{
751	WS;	1348	// the beauty of JSON: you need exactly one character lookahead
		1349	// to parse everything.
752	switch (*dec->cur)	1350	switch (*dec->cur)
753	{	1351	{
754	case '"': ++dec->cur; return decode_str (dec);	1352	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1353	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1354	case '{': ++dec->cur; return decode_hv (dec);
757		1355
758	case '-':	1356	case '-':
759	case '0': case '1': case '2': case '3': case '4':	1357	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':	1358	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);	1359	return decode_num (dec);
762		1360
763	case 't':	1361	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1362	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1363	{
766	dec->cur += 4;	1364	dec->cur += 4;
		1365	#if JSON_SLOW
		1366	json_true = get_bool ("JSON::XS::true");
		1367	#endif
767	return newSViv (1);	1368	return newSVsv (json_true);
768	}	1369	}
769	else	1370	else
770	ERR ("'true' expected");	1371	ERR ("'true' expected");
771		1372
772	break;	1373	break;
773		1374
774	case 'f':	1375	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1376	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1377	{
777	dec->cur += 5;	1378	dec->cur += 5;
		1379	#if JSON_SLOW
		1380	json_false = get_bool ("JSON::XS::false");
		1381	#endif
778	return newSViv (0);	1382	return newSVsv (json_false);
779	}	1383	}
780	else	1384	else
781	ERR ("'false' expected");	1385	ERR ("'false' expected");
782		1386
783	break;	1387	break;
784		1388
785	case 'n':	1389	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1390	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1391	{
788	dec->cur += 4;	1392	dec->cur += 4;
789	return newSViv (1);	1393	return newSVsv (&PL_sv_undef);
790	}	1394	}
791	else	1395	else
792	ERR ("'null' expected");	1396	ERR ("'null' expected");
793		1397
794	break;	1398	break;
795		1399
796	default:	1400	default:
797	ERR ("malformed json string");	1401	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1402	break;
799	}	1403	}
800		1404
801	fail:	1405	fail:
802	return 0;	1406	return 0;
803	}	1407	}
804		1408
805	static SV *	1409	static SV *
806	decode_json (SV *string, UV flags)	1410	decode_json (SV *string, JSON *json, STRLEN *offset_return)
807	{	1411	{
		1412	dec_t dec;
		1413	STRLEN offset;
808	SV *sv;	1414	SV *sv;
809		1415
		1416	SvGETMAGIC (string);
		1417	SvUPGRADE (string, SVt_PV);
		1418
		1419	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1420	* assertion with -DDEBUGGING, although SvCUR is documented to
		1421	* return the xpv_cur field which certainly exists after upgrading.
		1422	* according to nicholas clark, calling SvPOK fixes this.
		1423	*/
		1424	SvPOK (string);
		1425
		1426	if (SvCUR (string) > json->max_size && json->max_size)
		1427	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1428	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1429
810	if (!(flags & F_UTF8))	1430	if (json->flags & F_UTF8)
		1431	sv_utf8_downgrade (string, 0);
		1432	else
811	sv_utf8_upgrade (string);	1433	sv_utf8_upgrade (string);
812		1434
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1435	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1436
815	dec_t dec;	1437	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1438	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1439	dec.end = SvEND (string);
819	dec.err = 0;	1440	dec.err = 0;
		1441	dec.depth = 0;
820		1442
821	*dec.end = 1; // invalid anywhere	1443	if (dec.json.cb_object || dec.json.cb_sk_object)
		1444	dec.json.flags |= F_HOOK;
		1445
		1446	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1447
		1448	decode_ws (&dec);
822	sv = decode_sv (&dec);	1449	sv = decode_sv (&dec);
823	*dec.end = 0;	1450
		1451	if (!(offset_return || !sv))
		1452	{
		1453	// check for trailing garbage
		1454	decode_ws (&dec);
		1455
		1456	if (*dec.cur)
		1457	{
		1458	dec.err = "garbage after JSON object";
		1459	SvREFCNT_dec (sv);
		1460	sv = 0;
		1461	}
		1462	}
		1463
		1464	if (offset_return || !sv)
		1465	{
		1466	offset = dec.json.flags & F_UTF8
		1467	? dec.cur - SvPVX (string)
		1468	: utf8_distance (dec.cur, SvPVX (string));
		1469
		1470	if (offset_return)
		1471	*offset_return = offset;
		1472	}
824		1473
825	if (!sv)	1474	if (!sv)
826	{	1475	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1476	SV *uni = sv_newmortal ();
829		1477
		1478	// horrible hack to silence warning inside pv_uni_display
		1479	COP cop = *PL_curcop;
		1480	cop.cop_warnings = pWARN_NONE;
		1481	ENTER;
		1482	SAVEVPTR (PL_curcop);
		1483	PL_curcop = &cop;
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1484	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1485	LEAVE;
		1486
831	croak ("%s, at character %d (%s)",	1487	croak ("%s, at character offset %d [\"%s\"]",
832	dec.err,	1488	dec.err,
833	(int)offset,	1489	(int)offset,
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1490	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1491	}
836		1492
837	sv = sv_2mortal (sv);	1493	sv = sv_2mortal (sv);
838		1494
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1495	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
840	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1496	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1497
842	return sv;	1498	return sv;
843	}	1499	}
844		1500
		1501	/////////////////////////////////////////////////////////////////////////////
		1502	// incremental parser
		1503
		1504	static void
		1505	incr_parse (JSON *self)
		1506	{
		1507	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1508
		1509	for (;;)
		1510	{
		1511	//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
		1512	switch (self->incr_mode)
		1513	{
		1514	// only used for intiial whitespace skipping
		1515	case INCR_M_WS:
		1516	for (;;)
		1517	{
		1518	if (*p > 0x20)
		1519	{
		1520	self->incr_mode = INCR_M_JSON;
		1521	goto incr_m_json;
		1522	}
		1523	else if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	}
		1528
		1529	// skip a single char inside a string (for \\-processing)
		1530	case INCR_M_BS:
		1531	if (!*p)
		1532	goto interrupt;
		1533
		1534	++p;
		1535	self->incr_mode = INCR_M_STR;
		1536	goto incr_m_str;
		1537
		1538	// inside a string
		1539	case INCR_M_STR:
		1540	incr_m_str:
		1541	for (;;)
		1542	{
		1543	if (*p == '"')
		1544	{
		1545	++p;
		1546	self->incr_mode = INCR_M_JSON;
		1547
		1548	if (!self->incr_nest)
		1549	goto interrupt;
		1550
		1551	goto incr_m_json;
		1552	}
		1553	else if (*p == '\\')
		1554	{
		1555	++p; // "virtually" consumes character after \
		1556
		1557	if (!*p) // if at end of string we have to switch modes
		1558	{
		1559	self->incr_mode = INCR_M_BS;
		1560	goto interrupt;
		1561	}
		1562	}
		1563	else if (!*p)
		1564	goto interrupt;
		1565
		1566	++p;
		1567	}
		1568
		1569	// after initial ws, outside string
		1570	case INCR_M_JSON:
		1571	incr_m_json:
		1572	for (;;)
		1573	{
		1574	switch (*p++)
		1575	{
		1576	case 0:
		1577	--p;
		1578	goto interrupt;
		1579
		1580	case 0x09:
		1581	case 0x0a:
		1582	case 0x0d:
		1583	case 0x20:
		1584	if (!self->incr_nest)
		1585	{
		1586	--p; // do not eat the whitespace, let the next round do it
		1587	goto interrupt;
		1588	}
		1589	break;
		1590
		1591	case '"':
		1592	self->incr_mode = INCR_M_STR;
		1593	goto incr_m_str;
		1594
		1595	case '[':
		1596	case '{':
		1597	if (++self->incr_nest > self->max_depth)
		1598	croak (ERR_NESTING_EXCEEDED);
		1599	break;
		1600
		1601	case ']':
		1602	case '}':
		1603	if (--self->incr_nest <= 0)
		1604	goto interrupt;
		1605	}
		1606	}
		1607	}
		1608
		1609	modechange:
		1610	;
		1611	}
		1612
		1613	interrupt:
		1614	self->incr_pos = p - SvPVX (self->incr_text);
		1615	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1616	}
		1617
		1618	/////////////////////////////////////////////////////////////////////////////
		1619	// XS interface functions
		1620
845	MODULE = JSON::XS PACKAGE = JSON::XS	1621	MODULE = JSON::XS PACKAGE = JSON::XS
846		1622
847	BOOT:	1623	BOOT:
848	{	1624	{
849	int i;	1625	int i;
850		1626
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1627	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1628	decode_hexdigit [i] =
		1629	i >= '0' && i <= '9' ? i - '0'
		1630	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1631	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1632	: -1;
854		1633
855	for (i = 7; i--; )
856	{
857	decode_hexdigit ['a' + i] = 10 + i;
858	decode_hexdigit ['A' + i] = 10 + i;
859	}
860
861	json_stash = gv_stashpv ("JSON::XS", 1);	1634	json_stash = gv_stashpv ("JSON::XS" , 1);
		1635	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1636
		1637	json_true = get_bool ("JSON::XS::true");
		1638	json_false = get_bool ("JSON::XS::false");
862	}	1639	}
863		1640
864	PROTOTYPES: DISABLE	1641	PROTOTYPES: DISABLE
865		1642
866	SV *new (char *dummy)	1643	void CLONE (...)
867	CODE:	1644	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1645	json_stash = 0;
		1646	json_boolean_stash = 0;
		1647
		1648	void new (char *klass)
		1649	PPCODE:
		1650	{
		1651	SV *pv = NEWSV (0, sizeof (JSON));
		1652	SvPOK_only (pv);
		1653	json_init ((JSON *)SvPVX (pv));
		1654	XPUSHs (sv_2mortal (sv_bless (
		1655	newRV_noinc (pv),
		1656	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1657	)));
		1658	}
		1659
		1660	void ascii (JSON *self, int enable = 1)
		1661	ALIAS:
		1662	ascii = F_ASCII
		1663	latin1 = F_LATIN1
		1664	utf8 = F_UTF8
		1665	indent = F_INDENT
		1666	canonical = F_CANONICAL
		1667	space_before = F_SPACE_BEFORE
		1668	space_after = F_SPACE_AFTER
		1669	pretty = F_PRETTY
		1670	allow_nonref = F_ALLOW_NONREF
		1671	shrink = F_SHRINK
		1672	allow_blessed = F_ALLOW_BLESSED
		1673	convert_blessed = F_CONV_BLESSED
		1674	relaxed = F_RELAXED
		1675	allow_unknown = F_ALLOW_UNKNOWN
		1676	PPCODE:
		1677	{
		1678	if (enable)
		1679	self->flags |= ix;
		1680	else
		1681	self->flags &= ~ix;
		1682
		1683	XPUSHs (ST (0));
		1684	}
		1685
		1686	void get_ascii (JSON *self)
		1687	ALIAS:
		1688	get_ascii = F_ASCII
		1689	get_latin1 = F_LATIN1
		1690	get_utf8 = F_UTF8
		1691	get_indent = F_INDENT
		1692	get_canonical = F_CANONICAL
		1693	get_space_before = F_SPACE_BEFORE
		1694	get_space_after = F_SPACE_AFTER
		1695	get_allow_nonref = F_ALLOW_NONREF
		1696	get_shrink = F_SHRINK
		1697	get_allow_blessed = F_ALLOW_BLESSED
		1698	get_convert_blessed = F_CONV_BLESSED
		1699	get_relaxed = F_RELAXED
		1700	get_allow_unknown = F_ALLOW_UNKNOWN
		1701	PPCODE:
		1702	XPUSHs (boolSV (self->flags & ix));
		1703
		1704	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1705	PPCODE:
		1706	self->max_depth = max_depth;
		1707	XPUSHs (ST (0));
		1708
		1709	U32 get_max_depth (JSON *self)
		1710	CODE:
		1711	RETVAL = self->max_depth;
869	OUTPUT:	1712	OUTPUT:
870	RETVAL	1713	RETVAL
871		1714
872	SV *ascii (SV *self, int enable)	1715	void max_size (JSON *self, U32 max_size = 0)
873	ALIAS:	1716	PPCODE:
874	ascii = F_ASCII	1717	self->max_size = max_size;
875	utf8 = F_UTF8	1718	XPUSHs (ST (0));
876	indent = F_INDENT	1719
877	canonical = F_CANONICAL	1720	int get_max_size (JSON *self)
878	space_before = F_SPACE_BEFORE
879	space_after = F_SPACE_AFTER
880	json_rpc = F_JSON_RPC
881	pretty = F_PRETTY
882	allow_nonref = F_ALLOW_NONREF
883	CODE:	1721	CODE:
884	{	1722	RETVAL = self->max_size;
885	UV *uv = SvJSON (self);
886	if (enable)
887	*uv |= ix;
888	else
889	*uv &= ~ix;
890
891	RETVAL = newSVsv (self);
892	}
893	OUTPUT:	1723	OUTPUT:
894	RETVAL	1724	RETVAL
895		1725
896	void encode (SV *self, SV *scalar)	1726	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897	PPCODE:	1727	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1728	{
		1729	SvREFCNT_dec (self->cb_object);
		1730	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899		1731
900	void decode (SV *self, SV *jsonstr)	1732	XPUSHs (ST (0));
		1733	}
		1734
		1735	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901	PPCODE:	1736	PPCODE:
		1737	{
		1738	if (!self->cb_sk_object)
		1739	self->cb_sk_object = newHV ();
		1740
		1741	if (SvOK (cb))
		1742	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1743	else
		1744	{
		1745	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1746
		1747	if (!HvKEYS (self->cb_sk_object))
		1748	{
		1749	SvREFCNT_dec (self->cb_sk_object);
		1750	self->cb_sk_object = 0;
		1751	}
		1752	}
		1753
		1754	XPUSHs (ST (0));
		1755	}
		1756
		1757	void encode (JSON *self, SV *scalar)
		1758	PPCODE:
		1759	XPUSHs (encode_json (scalar, self));
		1760
		1761	void decode (JSON *self, SV *jsonstr)
		1762	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1763	XPUSHs (decode_json (jsonstr, self, 0));
		1764
		1765	void decode_prefix (JSON *self, SV *jsonstr)
		1766	PPCODE:
		1767	{
		1768	STRLEN offset;
		1769	EXTEND (SP, 2);
		1770	PUSHs (decode_json (jsonstr, self, &offset));
		1771	PUSHs (sv_2mortal (newSVuv (offset)));
		1772	}
		1773
		1774	void incr_parse (JSON *self, SV *jsonstr = 0)
		1775	PPCODE:
		1776	{
		1777	if (!self->incr_text)
		1778	self->incr_text = newSVpvn ("", 0);
		1779
		1780	// append data, if any
		1781	if (jsonstr)
		1782	{
		1783	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1784	{
		1785	/* utf-8-ness differs, need to upgrade */
		1786	sv_utf8_upgrade (self->incr_text);
		1787
		1788	if (self->incr_pos)
		1789	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1790	- (U8 *)SvPVX (self->incr_text);
		1791	}
		1792
		1793	{
		1794	STRLEN len;
		1795	const char *str = SvPV (jsonstr, len);
		1796	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1797	Move (str, SvEND (self->incr_text), len, char);
		1798	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1799	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1800	}
		1801	}
		1802
		1803	if (GIMME_V != G_VOID)
		1804	do
		1805	{
		1806	STRLEN offset;
		1807
		1808	if (!INCR_DONE (self))
		1809	{
		1810	incr_parse (self);
		1811
		1812	if (self->incr_pos > self->max_size && self->max_size)
		1813	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1814	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1815
		1816	if (!INCR_DONE (self))
		1817	break;
		1818	}
		1819
		1820	XPUSHs (decode_json (self->incr_text, self, &offset));
		1821
		1822	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1823	self->incr_pos -= offset;
		1824	self->incr_nest = 0;
		1825	self->incr_mode = 0;
		1826	}
		1827	while (GIMME_V == G_ARRAY);
		1828	}
		1829
		1830	SV *incr_text (JSON *self)
		1831	ATTRS: lvalue
		1832	CODE:
		1833	{
		1834	if (self->incr_pos)
		1835	croak ("incr_text can not be called when the incremental parser already started parsing");
		1836
		1837	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1838	}
		1839	OUTPUT:
		1840	RETVAL
		1841
		1842	void incr_skip (JSON *self)
		1843	CODE:
		1844	{
		1845	if (self->incr_pos)
		1846	{
		1847	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1848	self->incr_pos = 0;
		1849	self->incr_nest = 0;
		1850	self->incr_mode = 0;
		1851	}
		1852	}
		1853
		1854	void incr_reset (JSON *self)
		1855	CODE:
		1856	{
		1857	SvREFCNT_dec (self->incr_text);
		1858	self->incr_text = 0;
		1859	self->incr_pos = 0;
		1860	self->incr_nest = 0;
		1861	self->incr_mode = 0;
		1862	}
		1863
		1864	void DESTROY (JSON *self)
		1865	CODE:
		1866	SvREFCNT_dec (self->cb_sk_object);
		1867	SvREFCNT_dec (self->cb_object);
		1868	SvREFCNT_dec (self->incr_text);
903		1869
904	PROTOTYPES: ENABLE	1870	PROTOTYPES: ENABLE
905		1871
906	void to_json (SV *scalar)	1872	void encode_json (SV *scalar)
		1873	ALIAS:
		1874	to_json_ = 0
		1875	encode_json = F_UTF8
907	PPCODE:	1876	PPCODE:
		1877	{
		1878	JSON json;
		1879	json_init (&json);
		1880	json.flags |= ix;
908	XPUSHs (encode_json (scalar, F_UTF8));	1881	XPUSHs (encode_json (scalar, &json));
		1882	}
909		1883
910	void from_json (SV *jsonstr)	1884	void decode_json (SV *jsonstr)
		1885	ALIAS:
		1886	from_json_ = 0
		1887	decode_json = F_UTF8
911	PPCODE:	1888	PPCODE:
		1889	{
		1890	JSON json;
		1891	json_init (&json);
		1892	json.flags |= ix;
912	XPUSHs (decode_json (jsonstr, F_UTF8));	1893	XPUSHs (decode_json (jsonstr, &json, 0));
		1894	}
913		1895
		1896

Diff Legend

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