[ViewVC] Diff of: 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 1.102 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>
		9	#include <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarantees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
		24
9	#define F_ASCII 0x00000001	25	#define F_ASCII 0x00000001UL
		26	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	27	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	28	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	29	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	30	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	31	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	32	#define F_ALLOW_NONREF 0x00000100UL
		33	#define F_SHRINK 0x00000200UL
		34	#define F_ALLOW_BLESSED 0x00000400UL
		35	#define F_CONV_BLESSED 0x00000800UL
		36	#define F_RELAXED 0x00001000UL
		37	#define F_ALLOW_UNKNOWN 0x00002000UL
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		39
18	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
19	#define F_DEFAULT 0
20		41
21	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
22		46
23	#define SB do {	47	#define SB do {
24	#define SE } while (0)	48	#define SE } while (0)
25		49
26	static HV *json_stash;	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
		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	}
		108
		109	/////////////////////////////////////////////////////////////////////////////
		110	// utility functions
		111
		112	INLINE SV *
		113	get_bool (const char *name)
		114	{
		115	SV *sv = get_sv (name, 1);
		116
		117	SvREADONLY_on (sv);
		118	SvREADONLY_on (SvRV (sv));
		119
		120	return sv;
		121	}
		122
		123	INLINE void
		124	shrink (SV *sv)
		125	{
		126	sv_utf8_downgrade (sv, 1);
		127
		128	if (SvLEN (sv) > SvCUR (sv) + 1)
		129	{
		130	#ifdef SvPV_shrink_to_cur
		131	SvPV_shrink_to_cur (sv);
		132	#elif defined (SvPV_renew)
		133	SvPV_renew (sv, SvCUR (sv) + 1);
		134	#endif
		135	}
		136	}
		137
		138	// decode an utf-8 character and return it, or (UV)-1 in
		139	// case of an error.
		140	// we special-case "safe" characters from U+80 .. U+7FF,
		141	// but use the very good perl function to parse anything else.
		142	// note that we never call this function for a ascii codepoints
		143	INLINE UV
		144	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
		145	{
		146	if (expect_true (len >= 2
		147	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		148	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		149	{
		150	*clen = 2;
		151	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
		152	}
		153	else
		154	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		155	}
		156
		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);
		188	}
		189
		190	/////////////////////////////////////////////////////////////////////////////
		191	// encoder
27		192
28	// structure used for encoding JSON	193	// structure used for encoding JSON
29	typedef struct	194	typedef struct
30	{	195	{
31	char *cur;	196	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	197	char *end; // SvEND (sv)
34	SV *sv;	198	SV *sv; // result scalar
35	UV flags;	199	JSON json;
36	int max_recurse;	200	U32 indent; // indentation level
37	int indent;	201	UV limit; // escape character values >= this value when encoding
38	} enc_t;	202	} enc_t;
39		203
40	// structure used for decoding JSON	204	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)	205	need (enc_t *enc, STRLEN len)
62	{	206	{
63	if (enc->cur + len >= enc->end)	207	if (expect_false (enc->cur + len >= enc->end))
64	{	208	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	209	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	210	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	211	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	212	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	213	}
70	}	214	}
71		215
72	static void	216	INLINE void
73	encode_ch (enc_t *enc, char ch)	217	encode_ch (enc_t *enc, char ch)
74	{	218	{
75	need (enc, 1);	219	need (enc, 1);
76	*enc->cur++ = ch;	220	*enc->cur++ = ch;
77	}	221	}
85		229
86	while (str < end)	230	while (str < end)
87	{	231	{
88	unsigned char ch = (unsigned char )str;	232	unsigned char ch = (unsigned char )str;
89		233
90	if (ch == '"')	234	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	235	{
		236	if (expect_false (ch == '"')) // but with slow exceptions
		237	{
92	need (enc, len += 1);	238	need (enc, len += 1);
93	*enc->cur++ = '\\';	239	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	240	*enc->cur++ = '"';
95	++str;
96	}	241	}
97	else if (ch == '\\')	242	else if (expect_false (ch == '\\'))
98	{	243	{
99	need (enc, len += 1);	244	need (enc, len += 1);
100	*enc->cur++ = '\\';	245	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	246	*enc->cur++ = '\\';
102	++str;
103	}	247	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	248	else
105	{
106	*enc->cur++ = ch;	249	*enc->cur++ = ch;
107	++str;	250
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;	251	++str;
122	}	252	}
123	else	253	else
124	{	254	{
125	STRLEN clen;	255	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{	256	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	257	case '\010': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'b'; ++str; break;
131	if (clen < 0)	258	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");	259	case '\012': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'n'; ++str; break;
133	}	260	case '\014': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'f'; ++str; break;
134	else	261	case '\015': need (enc, len += 1); enc->cur++ = '\\'; enc->cur++ = 'r'; ++str; break;
135	{
136	uch = ch;
137	clen = 1;
138	}
139		262
140	if (uch < 0x80 \|\| enc->flags & F_ASCII)	263	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	264	{
		265	STRLEN clen;
		266	UV uch;
		267
		268	if (is_utf8)
		269	{
		270	uch = decode_utf8 (str, end - str, &clen);
		271	if (clen == (STRLEN)-1)
		272	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		273	}
		274	else
		275	{
		276	uch = ch;
		277	clen = 1;
		278	}
		279
		280	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
		281	{
		282	if (uch >= 0x10000UL)
		283	{
		284	if (uch >= 0x110000UL)
		285	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		286
144	need (enc, len += 11);	287	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	288	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	289	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	290	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	291	enc->cur += 12;
		292	}
		293	else
		294	{
		295	need (enc, len += 5);
		296	*enc->cur++ = '\\';
		297	*enc->cur++ = 'u';
		298	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		299	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		300	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		301	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		302	}
		303
		304	str += clen;
		305	}
		306	else if (enc->json.flags & F_LATIN1)
		307	{
		308	*enc->cur++ = uch;
		309	str += clen;
		310	}
		311	else if (is_utf8)
		312	{
		313	need (enc, len += clen);
		314	do
		315	{
		316	enc->cur++ = str++;
		317	}
		318	while (--clen);
		319	}
		320	else
		321	{
		322	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		323	enc->cur = encode_utf8 (enc->cur, uch);
		324	++str;
		325	}
149	}	326	}
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	}	327	}
176	}	328	}
177		329
178	--len;	330	--len;
179	}	331	}
180	}	332	}
181		333
182	#define INDENT SB \	334	INLINE void
		335	encode_indent (enc_t *enc)
		336	{
183	if (enc->flags & F_INDENT) \	337	if (enc->json.flags & F_INDENT)
184	{ \	338	{
185	int i_; \	339	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	340
187	for (i_ = enc->indent * 3; i_--; )\	341	need (enc, spaces);
		342	memset (enc->cur, ' ', spaces);
		343	enc->cur += spaces;
		344	}
		345	}
		346
		347	INLINE void
		348	encode_space (enc_t *enc)
		349	{
		350	need (enc, 1);
		351	encode_ch (enc, ' ');
		352	}
		353
		354	INLINE void
		355	encode_nl (enc_t *enc)
		356	{
		357	if (enc->json.flags & F_INDENT)
		358	{
		359	need (enc, 1);
188	encode_ch (enc, ' '); \	360	encode_ch (enc, '\n');
189	} \	361	}
190	SE	362	}
191		363
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	364	INLINE void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	365	encode_comma (enc_t *enc)
194	#define COMMA SB \	366	{
195	encode_ch (enc, ','); \	367	encode_ch (enc, ',');
		368
196	if (enc->flags & F_INDENT) \	369	if (enc->json.flags & F_INDENT)
197	NL; \	370	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	371	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	372	encode_space (enc);
200	SE	373	}
201		374
202	static void encode_sv (enc_t enc, SV sv);	375	static void encode_sv (enc_t enc, SV sv);
203		376
204	static void	377	static void
205	encode_av (enc_t enc, AV av)	378	encode_av (enc_t enc, AV av)
206	{	379	{
207	int i, len = av_len (av);	380	int i, len = av_len (av);
208		381
		382	if (enc->indent >= enc->json.max_depth)
		383	croak (ERR_NESTING_EXCEEDED);
		384
209	encode_ch (enc, '['); NL;	385	encode_ch (enc, '[');
210	++enc->indent;	386
		387	if (len >= 0)
		388	{
		389	encode_nl (enc); ++enc->indent;
211		390
212	for (i = 0; i <= len; ++i)	391	for (i = 0; i <= len; ++i)
213	{	392	{
214	INDENT;	393	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));
216		394
		395	encode_indent (enc);
		396
		397	if (svp)
		398	encode_sv (enc, *svp);
		399	else
		400	encode_str (enc, "null", 4, 0);
		401
217	if (i < len)	402	if (i < len)
218	COMMA;	403	encode_comma (enc);
219	}	404	}
220		405
221	NL;	406	encode_nl (enc); --enc->indent; encode_indent (enc);
222		407	}
223	--enc->indent;	408
224	INDENT; encode_ch (enc, ']');	409	encode_ch (enc, ']');
225	}	410	}
226		411
227	static void	412	static void
228	encode_he (enc_t enc, HE he)	413	encode_hk (enc_t enc, HE he)
229	{	414	{
230	encode_ch (enc, '"');	415	encode_ch (enc, '"');
231		416
232	if (HeKLEN (he) == HEf_SVKEY)	417	if (HeKLEN (he) == HEf_SVKEY)
233	{	418	{
…		…
243	else	428	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	429	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		430
246	encode_ch (enc, '"');	431	encode_ch (enc, '"');
247		432
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	433	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	434	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	435	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));
252	}	436	}
253		437
254	// compare hash entries, used when all keys are bytestrings	438	// compare hash entries, used when all keys are bytestrings
255	static int	439	static int
256	he_cmp_fast (const void a_, const void b_)	440	he_cmp_fast (const void a_, const void b_)
…		…
261	HE b = (HE **)b_;	445	HE b = (HE **)b_;
262		446
263	STRLEN la = HeKLEN (a);	447	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	448	STRLEN lb = HeKLEN (b);
265		449
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	450	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	451	cmp = lb - la;
268		452
269	return cmp;	453	return cmp;
270	}	454	}
271		455
272	// 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
273	static int	457	static int
274	he_cmp_slow (const void a, const void b)	458	he_cmp_slow (const void a, const void b)
275	{	459	{
276	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	460	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
277	}	461	}
278		462
279	static void	463	static void
280	encode_hv (enc_t enc, HV hv)	464	encode_hv (enc_t enc, HV hv)
281	{	465	{
282	int count, i;	466	HE *he;
283		467
284	encode_ch (enc, '{'); NL; ++enc->indent;	468	if (enc->indent >= enc->json.max_depth)
		469	croak (ERR_NESTING_EXCEEDED);
285		470
286	if ((count = hv_iterinit (hv)))	471	encode_ch (enc, '{');
287	{	472
288	// for canonical output we have to sort by keys first	473	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	474	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	475	// security workaround added somewhere in 5.8.x
291	// that randomises hash orderings	476	// that randomises hash orderings
292	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))
293	{	482	{
294	HE he, hes [count];	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	{
295	int fast = 1;	497	int i, fast = 1;
		498	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		499	HE *hes = _alloca (count sizeof (HE));
		500	#else
		501	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		502	#endif
296		503
297	i = 0;	504	i = 0;
298	while ((he = hv_iternext (hv)))	505	while ((he = hv_iternext (hv)))
299	{	506	{
300	hes [i++] = he;	507	hes [i++] = he;
…		…
306		513
307	if (fast)	514	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	515	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	516	else
310	{	517	{
311	// hack to disable "use bytes"	518	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	519	COP cop = *PL_curcop;
313	cop.op_private = 0;	520	cop.op_private = 0;
		521
		522	ENTER;
		523	SAVETMPS;
		524
		525	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	526	PL_curcop = &cop;
315		527
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	528	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		529
318	FREETMPS;	530	FREETMPS;
319		531	LEAVE;
320	PL_curcop = oldcop;
321	}	532	}
322		533
323	for (i = 0; i < count; ++i)	534	encode_nl (enc); ++enc->indent;
		535
		536	while (count--)
324	{	537	{
325	INDENT;	538	encode_indent (enc);
		539	he = hes [count];
326	encode_he (enc, hes [i]);	540	encode_hk (enc, he);
		541	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327		542
328	if (i < count - 1)	543	if (count)
329	COMMA;	544	encode_comma (enc);
330	}	545	}
331		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)))
332	NL;	554	{
		555	encode_nl (enc); ++enc->indent;
		556
		557	for (;;)
		558	{
		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);
333	}	597	}
334	else	598	else
335	{	599	{
336	SV *sv;	600	#if 0
337	HE *he = hv_iternext (hv);	601	if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339	for (;;)
340	{	602	{
341	INDENT;	603	// not yet
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}	604	}
349		605	#endif
		606	if (enc->json.flags & F_CONV_BLESSED)
350	NL;	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);
		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
		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))));
351	}	645	}
352	}	646	}
		647	else if (svt == SVt_PVHV)
		648	encode_hv (enc, (HV *)sv);
		649	else if (svt == SVt_PVAV)
		650	encode_av (enc, (AV *)sv);
		651	else if (svt < SVt_PVAV)
		652	{
		653	STRLEN len = 0;
		654	char *pv = svt ? SvPV (sv, len) : 0;
353		655
354	--enc->indent; INDENT; encode_ch (enc, '}');	656	if (len == 1 && *pv == '1')
		657	encode_str (enc, "true", 4, 0);
		658	else if (len == 1 && *pv == '0')
		659	encode_str (enc, "false", 5, 0);
		660	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		661	encode_str (enc, "null", 4, 0);
		662	else
		663	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		664	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		665	}
		666	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		667	encode_str (enc, "null", 4, 0);
		668	else
		669	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		670	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	671	}
356		672
357	static void	673	static void
358	encode_sv (enc_t enc, SV sv)	674	encode_sv (enc_t enc, SV sv)
359	{	675	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	683	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	684	encode_ch (enc, '"');
369	}	685	}
370	else if (SvNOKp (sv))	686	else if (SvNOKp (sv))
371	{	687	{
		688	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	689	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	690	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	691	enc->cur += strlen (enc->cur);
375	}	692	}
376	else if (SvIOKp (sv))	693	else if (SvIOKp (sv))
377	{	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
378	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);
379	enc->cur +=	731	enc->cur +=
380	SvIsUV(sv)	732	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	733	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	734	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		735	}
383	}	736	}
384	else if (SvROK (sv))	737	else if (SvROK (sv))
385	{	738	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)	739	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);	740	encode_str (enc, "null", 4, 0);
402	else	741	else
403	croak ("encountered perl type that JSON cannot handle");	742	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		743	SvPV_nolen (sv), SvFLAGS (sv));
404	}	744	}
405		745
406	static SV *	746	static SV *
407	encode_json (SV *scalar, UV flags)	747	encode_json (SV scalar, JSON json)
408	{	748	{
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;	749	enc_t enc;
413	enc.flags = flags;	750
		751	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		752	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		753
		754	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	755	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	756	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	757	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	758	enc.indent = 0;
		759	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		760	: enc.json.flags & F_LATIN1 ? 0x000100UL
		761	: 0x110000UL;
419		762
420	SvPOK_only (enc.sv);	763	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	764	encode_sv (&enc, scalar);
		765	encode_nl (&enc);
422		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
423	if (!(flags & (F_ASCII \| F_UTF8)))	770	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
424	SvUTF8_on (enc.sv);	771	SvUTF8_on (enc.sv);
425		772
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	773	if (enc.json.flags & F_SHRINK)
		774	shrink (enc.sv);
		775
427	return enc.sv;	776	return enc.sv;
428	}	777	}
429		778
430	/////////////////////////////////////////////////////////////////////////////	779	/////////////////////////////////////////////////////////////////////////////
		780	// decoder
431		781
432	#define WS \	782	// structure used for decoding JSON
		783	typedef struct
		784	{
		785	char *cur; // current parser pointer
		786	char *end; // end of input string
		787	const char *err; // parse error, if != 0
		788	JSON json;
		789	U32 depth; // recursion depth
		790	U32 maxdepth; // recursion depth limit
		791	} dec_t;
		792
		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
		803	decode_ws (dec_t *dec)
		804	{
433	for (;;) \	805	for (;;)
434	{ \	806	{
435	char ch = *dec->cur; \	807	char ch = *dec->cur;
		808
436	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	}
437	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	821	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438	break; \	822	break; // parse error, but let higher level handle it, gives better error messages
		823
439	++dec->cur; \	824	++dec->cur;
440	}	825	}
		826	}
441		827
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	828	#define ERR(reason) SB dec->err = reason; goto fail; SE
		829
443	#define EXPECT_CH(ch) SB \	830	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	831	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	832	ERR (# ch " expected"); \
446	++dec->cur; \	833	++dec->cur; \
447	SE	834	SE
448		835
		836	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		837	#define DEC_DEC_DEPTH --dec->depth
		838
449	static SV decode_sv (dec_t dec);	839	static SV decode_sv (dec_t dec);
450		840
451	static signed char decode_hexdigit[256];	841	static signed char decode_hexdigit[256];
452		842
453	static UV	843	static UV
454	decode_4hex (dec_t *dec)	844	decode_4hex (dec_t *dec)
455	{	845	{
456	signed char d1, d2, d3, d4;	846	signed char d1, d2, d3, d4;
		847	unsigned char cur = (unsigned char )dec->cur;
457		848
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	849	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");	850	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]];	851	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");	852	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		853
467	dec->cur += 4;	854	dec->cur += 4;
468		855
469	return ((UV)d1) << 12	856	return ((UV)d1) << 12
470	\| ((UV)d2) << 8	857	\| ((UV)d2) << 8
…		…
473		860
474	fail:	861	fail:
475	return (UV)-1;	862	return (UV)-1;
476	}	863	}
477		864
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 *	865	static SV *
494	decode_str (dec_t *dec)	866	decode_str (dec_t *dec)
495	{	867	{
496	SV *sv = NEWSV (0,2);	868	SV *sv = 0;
497	int utf8 = 0;	869	int utf8 = 0;
498	char *cur = SvPVX (sv);	870	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		871
501	for (;;)	872	do
502	{	873	{
503	unsigned char ch = (unsigned char )dec->cur;	874	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		875	char *cur = buf;
504		876
505	if (ch == '"')	877	do
506	break;
507	else if (ch == '\\')
508	{	878	{
509	switch (*++dec->cur)	879	unsigned char ch = (unsigned char )dec_cur++;
		880
		881	if (expect_false (ch == '"'))
510	{	882	{
511	case '\\':	883	--dec_cur;
512	case '/':	884	break;
513	case '"': APPEND_CH (*dec->cur++); break;	885	}
514		886	else if (expect_false (ch == '\\'))
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	887	{
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;	888	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	{	889	{
523	UV lo, hi;	890	case '\\':
524	++dec->cur;	891	case '/':
		892	case '"': cur++ = dec_cur++; break;
525		893
526	hi = decode_4hex (dec);	894	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	895	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	896	case 'n': ++dec_cur; *cur++ = '\012'; break;
		897	case 'f': ++dec_cur; *cur++ = '\014'; break;
		898	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		899
530	// possibly a surrogate pair	900	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	901	{
533	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	902	UV lo, hi;
534	ERR ("illegal surrogate character");	903	++dec_cur;
535		904
536	dec->cur += 2;	905	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	906	hi = decode_4hex (dec);
		907	dec_cur = dec->cur;
539	if (lo == (UV)-1)	908	if (hi == (UV)-1)
540	goto fail;	909	goto fail;
541		910
		911	// possibly a surrogate pair
		912	if (hi >= 0xd800)
		913	if (hi < 0xdc00)
		914	{
		915	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
		916	ERR ("missing low surrogate character in surrogate pair");
		917
		918	dec_cur += 2;
		919
		920	dec->cur = dec_cur;
		921	lo = decode_4hex (dec);
		922	dec_cur = dec->cur;
		923	if (lo == (UV)-1)
		924	goto fail;
		925
542	if (lo < 0xdc00 \|\| lo >= 0xe000)	926	if (lo < 0xdc00 \|\| lo >= 0xe000)
543	ERR ("surrogate pair expected");	927	ERR ("surrogate pair expected");
544		928
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	929	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		930	}
		931	else if (hi < 0xe000)
		932	ERR ("missing high surrogate character in surrogate pair");
		933
		934	if (hi >= 0x80)
		935	{
		936	utf8 = 1;
		937
		938	cur = encode_utf8 (cur, hi);
		939	}
		940	else
		941	*cur++ = hi;
546	}	942	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	943	break;
552	utf8 = 1;
553		944
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	945	default:
558	APPEND_CH (hi);	946	--dec_cur;
		947	ERR ("illegal backslash escape sequence in string");
559	}	948	}
		949	}
		950	else if (expect_true (ch >= 0x20 && ch < 0x80))
		951	*cur++ = ch;
		952	else if (ch >= 0x80)
		953	{
		954	STRLEN clen;
		955
		956	--dec_cur;
		957
		958	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		959	if (clen == (STRLEN)-1)
		960	ERR ("malformed UTF-8 character in JSON string");
		961
		962	do
		963	cur++ = dec_cur++;
		964	while (--clen);
		965
		966	utf8 = 1;
		967	}
		968	else
		969	{
		970	--dec_cur;
		971
		972	if (!ch)
		973	ERR ("unexpected end of string while parsing JSON string");
560	break;	974	else
		975	ERR ("invalid character encountered while parsing JSON string");
561	}	976	}
562	}	977	}
563	else if (ch >= 0x20 && ch <= 0x7f)	978	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	979
565	else if (ch >= 0x80)	980	{
		981	STRLEN len = cur - buf;
		982
		983	if (sv)
566	{	984	{
567	STRLEN clen;	985	STRLEN cur = SvCUR (sv);
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
569	if (clen < 0)
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571		986
572	APPEND_GROW (clen);	987	if (SvLEN (sv) <= cur + len)
573	memcpy (cur, dec->cur, clen);	988	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574	cur += clen;	989
575	dec->cur += clen;	990	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		991	SvCUR_set (sv, SvCUR (sv) + len);
576	}	992	}
577	else	993	else
578	ERR ("invalid character encountered");	994	sv = newSVpvn (buf, len);
579	}	995	}
		996	}
		997	while (*dec_cur != '"');
580		998
581	++dec->cur;	999	++dec_cur;
582		1000
583	SvCUR_set (sv, cur - SvPVX (sv));	1001	if (sv)
584		1002	{
585	SvPOK_only (sv);	1003	SvPOK_only (sv);
586	*SvEND (sv) = 0;	1004	*SvEND (sv) = 0;
587		1005
588	if (utf8)	1006	if (utf8)
589	SvUTF8_on (sv);	1007	SvUTF8_on (sv);
		1008	}
		1009	else
		1010	sv = newSVpvn ("", 0);
590		1011
		1012	dec->cur = dec_cur;
591	return sv;	1013	return sv;
592		1014
593	fail:	1015	fail:
594	SvREFCNT_dec (sv);	1016	dec->cur = dec_cur;
595	return 0;	1017	return 0;
596	}	1018	}
597		1019
598	static SV *	1020	static SV *
599	decode_num (dec_t *dec)	1021	decode_num (dec_t *dec)
…		…
609	{	1031	{
610	++dec->cur;	1032	++dec->cur;
611	if (dec->cur >= '0' && dec->cur <= '9')	1033	if (dec->cur >= '0' && dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	1034	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	1035	}
614		1036	else if (dec->cur < '0' \|\| dec->cur > '9')
615	// int	1037	ERR ("malformed number (no digits after initial minus)");
		1038	else
		1039	do
		1040	{
		1041	++dec->cur;
		1042	}
616	while (dec->cur >= '0' && dec->cur <= '9')	1043	while (dec->cur >= '0' && dec->cur <= '9');
617	++dec->cur;
618		1044
619	// [frac]	1045	// [frac]
620	if (*dec->cur == '.')	1046	if (*dec->cur == '.')
621	{	1047	{
622	is_nv = 1;	1048	++dec->cur;
		1049
		1050	if (dec->cur < '0' \|\| dec->cur > '9')
		1051	ERR ("malformed number (no digits after decimal point)");
623		1052
624	do	1053	do
625	{	1054	{
626	++dec->cur;	1055	++dec->cur;
627	}	1056	}
628	while (dec->cur >= '0' && dec->cur <= '9');	1057	while (dec->cur >= '0' && dec->cur <= '9');
		1058
		1059	is_nv = 1;
629	}	1060	}
630		1061
631	// [exp]	1062	// [exp]
632	if (dec->cur == 'e' \|\| dec->cur == 'E')	1063	if (dec->cur == 'e' \|\| dec->cur == 'E')
633	{	1064	{
634	is_nv = 1;
635
636	++dec->cur;	1065	++dec->cur;
		1066
637	if (dec->cur == '-' \|\| dec->cur == '+')	1067	if (dec->cur == '-' \|\| dec->cur == '+')
638	++dec->cur;	1068	++dec->cur;
639		1069
		1070	if (dec->cur < '0' \|\| dec->cur > '9')
		1071	ERR ("malformed number (no digits after exp sign)");
		1072
		1073	do
		1074	{
		1075	++dec->cur;
		1076	}
640	while (dec->cur >= '0' && dec->cur <= '9')	1077	while (dec->cur >= '0' && dec->cur <= '9');
641	++dec->cur;	1078
		1079	is_nv = 1;
642	}	1080	}
643		1081
644	if (!is_nv)	1082	if (!is_nv)
645	{	1083	{
646	UV uv;	1084	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	1085
648	if (numtype & IS_NUMBER_IN_UV)	1086	// special case the rather common 1..5-digit-int case
649	if (numtype & IS_NUMBER_NEG)	1087	if (*start == '-')
		1088	switch (len)
650	{	1089	{
651	if (uv < (UV)IV_MIN)	1090	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652	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));
653	}	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	}
654	else	1115	else
655	return newSVuv (uv);	1116	return newSVuv (uv);
656	}	1117	}
657		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
658	return newSVnv (Atof (start));	1135	return newSVnv (Atof (start));
659		1136
660	fail:	1137	fail:
661	return 0;	1138	return 0;
662	}	1139	}
…		…
664	static SV *	1141	static SV *
665	decode_av (dec_t *dec)	1142	decode_av (dec_t *dec)
666	{	1143	{
667	AV *av = newAV ();	1144	AV *av = newAV ();
668		1145
		1146	DEC_INC_DEPTH;
		1147	decode_ws (dec);
		1148
		1149	if (*dec->cur == ']')
		1150	++dec->cur;
		1151	else
669	for (;;)	1152	for (;;)
670	{	1153	{
671	SV *value;	1154	SV *value;
672		1155
673	value = decode_sv (dec);	1156	value = decode_sv (dec);
674	if (!value)	1157	if (!value)
675	goto fail;	1158	goto fail;
676		1159
677	av_push (av, value);	1160	av_push (av, value);
678		1161
679	WS;	1162	decode_ws (dec);
680		1163
681	if (*dec->cur == ']')	1164	if (*dec->cur == ']')
682	{	1165	{
683	++dec->cur;	1166	++dec->cur;
684	break;	1167	break;
		1168	}
685	}	1169
686
687	if (*dec->cur != ',')	1170	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1171	ERR (", or ] expected while parsing array");
689		1172
690	++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	}
691	}	1182	}
692		1183
		1184	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1185	return newRV_noinc ((SV *)av);
694		1186
695	fail:	1187	fail:
696	SvREFCNT_dec (av);	1188	SvREFCNT_dec (av);
		1189	DEC_DEC_DEPTH;
697	return 0;	1190	return 0;
698	}	1191	}
699		1192
700	static SV *	1193	static SV *
701	decode_hv (dec_t *dec)	1194	decode_hv (dec_t *dec)
702	{	1195	{
		1196	SV *sv;
703	HV *hv = newHV ();	1197	HV *hv = newHV ();
704		1198
		1199	DEC_INC_DEPTH;
		1200	decode_ws (dec);
		1201
		1202	if (*dec->cur == '}')
		1203	++dec->cur;
		1204	else
705	for (;;)	1205	for (;;)
706	{	1206	{
707	SV key, value;
708
709	WS; EXPECT_CH ('"');	1207	EXPECT_CH ('"');
710		1208
711	key = decode_str (dec);	1209	// heuristic: assume that
712	if (!key)	1210	// a) decode_str + hv_store_ent are abysmally slow.
713	goto fail;	1211	// b) most hash keys are short, simple ascii text.
714		1212	// => try to "fast-match" such strings to avoid
715	WS; EXPECT_CH (':');	1213	// the overhead of decode_str + hv_store_ent.
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1214	{
		1215	SV *value;
		1216	char *p = dec->cur;
		1217	char *e = p + 24; // only try up to 24 bytes
		1218
		1219	for (;;)
		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);
720	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)
721	goto fail;	1256	goto fail;
		1257
		1258	hv_store (hv, key, len, value, 0);
		1259
		1260	break;
		1261	}
		1262
		1263	++p;
		1264	}
722	}	1265	}
723		1266
724	//TODO: optimise	1267	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1268
727	WS;
728
729	if (*dec->cur == '}')	1269	if (*dec->cur == '}')
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
		1274
		1275	if (*dec->cur != ',')
		1276	ERR (", or } expected while parsing object/hash");
		1277
		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	}
		1287	}
		1288
		1289	DEC_DEC_DEPTH;
		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)
730	{	1296	{
731	++dec->cur;	1297	HE cb, he;
732	break;	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	}
733	}	1326	}
734		1327
735	if (*dec->cur != ',')	1328	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1329	{
		1330	dSP;
		1331	int count;
737		1332
738	++dec->cur;	1333	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1334	XPUSHs (sv_2mortal (sv));
740		1335
741	return newRV_noinc ((SV *)hv);	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;
742		1351
743	fail:	1352	fail:
744	SvREFCNT_dec (hv);	1353	SvREFCNT_dec (hv);
		1354	DEC_DEC_DEPTH;
745	return 0;	1355	return 0;
746	}	1356	}
747		1357
748	static SV *	1358	static SV *
749	decode_sv (dec_t *dec)	1359	decode_sv (dec_t *dec)
750	{	1360	{
751	WS;	1361	// the beauty of JSON: you need exactly one character lookahead
		1362	// to parse everything.
752	switch (*dec->cur)	1363	switch (*dec->cur)
753	{	1364	{
754	case '"': ++dec->cur; return decode_str (dec);	1365	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1366	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1367	case '{': ++dec->cur; return decode_hv (dec);
757		1368
758	case '-':	1369	case '-':
759	case '0': case '1': case '2': case '3': case '4':	1370	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':	1371	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);	1372	return decode_num (dec);
762		1373
763	case 't':	1374	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1375	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1376	{
766	dec->cur += 4;	1377	dec->cur += 4;
		1378	#if JSON_SLOW
		1379	json_true = get_bool ("JSON::XS::true");
		1380	#endif
767	return newSViv (1);	1381	return newSVsv (json_true);
768	}	1382	}
769	else	1383	else
770	ERR ("'true' expected");	1384	ERR ("'true' expected");
771		1385
772	break;	1386	break;
773		1387
774	case 'f':	1388	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1389	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1390	{
777	dec->cur += 5;	1391	dec->cur += 5;
		1392	#if JSON_SLOW
		1393	json_false = get_bool ("JSON::XS::false");
		1394	#endif
778	return newSViv (0);	1395	return newSVsv (json_false);
779	}	1396	}
780	else	1397	else
781	ERR ("'false' expected");	1398	ERR ("'false' expected");
782		1399
783	break;	1400	break;
784		1401
785	case 'n':	1402	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1403	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1404	{
788	dec->cur += 4;	1405	dec->cur += 4;
789	return newSViv (1);	1406	return newSVsv (&PL_sv_undef);
790	}	1407	}
791	else	1408	else
792	ERR ("'null' expected");	1409	ERR ("'null' expected");
793		1410
794	break;	1411	break;
795		1412
796	default:	1413	default:
797	ERR ("malformed json string");	1414	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1415	break;
799	}	1416	}
800		1417
801	fail:	1418	fail:
802	return 0;	1419	return 0;
803	}	1420	}
804		1421
805	static SV *	1422	static SV *
806	decode_json (SV *string, UV flags)	1423	decode_json (SV string, JSON json, char **offset_return)
807	{	1424	{
		1425	dec_t dec;
808	SV *sv;	1426	SV *sv;
809		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
		1435	SvUPGRADE (string, SVt_PV);
		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
810	if (!(flags & F_UTF8))	1459	if (json->flags & F_UTF8)
		1460	sv_utf8_downgrade (string, 0);
		1461	else
811	sv_utf8_upgrade (string);	1462	sv_utf8_upgrade (string);
812		1463
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1464	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1465
815	dec_t dec;	1466	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1467	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1468	dec.end = SvEND (string);
819	dec.err = 0;	1469	dec.err = 0;
		1470	dec.depth = 0;
820		1471
821	*dec.end = 1; // invalid anywhere	1472	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1473	dec.json.flags \|= F_HOOK;
		1474
		1475	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1476
		1477	decode_ws (&dec);
822	sv = decode_sv (&dec);	1478	sv = decode_sv (&dec);
823	*dec.end = 0;	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	}
824		1495
825	if (!sv)	1496	if (!sv)
826	{	1497	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1498	SV *uni = sv_newmortal ();
829		1499
		1500	// horrible hack to silence warning inside pv_uni_display
		1501	COP cop = *PL_curcop;
		1502	cop.cop_warnings = pWARN_NONE;
		1503	ENTER;
		1504	SAVEVPTR (PL_curcop);
		1505	PL_curcop = &cop;
830	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);
		1507	LEAVE;
		1508
831	croak ("%s, at character %d (%s)",	1509	croak ("%s, at character offset %d (before \"%s\")",
832	dec.err,	1510	dec.err,
833	(int)offset,	1511	ptr_to_index (string, dec.cur),
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1512	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1513	}
836		1514
837	sv = sv_2mortal (sv);	1515	sv = sv_2mortal (sv);
838		1516
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1517	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)");	1518	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1519
842	return sv;	1520	return sv;
843	}	1521	}
844		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
		1638	}
		1639
		1640	/////////////////////////////////////////////////////////////////////////////
		1641	// XS interface functions
		1642
845	MODULE = JSON::XS PACKAGE = JSON::XS	1643	MODULE = JSON::XS PACKAGE = JSON::XS
846		1644
847	BOOT:	1645	BOOT:
848	{	1646	{
849	int i;	1647	int i;
850		1648
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1649	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1650	decode_hexdigit [i] =
		1651	i >= '0' && i <= '9' ? i - '0'
		1652	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1653	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1654	: -1;
854		1655
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);	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 */
862	}	1663	}
863		1664
864	PROTOTYPES: DISABLE	1665	PROTOTYPES: DISABLE
865		1666
866	SV new (char dummy)	1667	void CLONE (...)
867	CODE:	1668	CODE:
868	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;
869	OUTPUT:	1736	OUTPUT:
870	RETVAL	1737	RETVAL
871		1738
872	SV ascii (SV self, int enable)	1739	void max_size (JSON *self, U32 max_size = 0)
873	ALIAS:	1740	PPCODE:
874	ascii = F_ASCII	1741	self->max_size = max_size;
875	utf8 = F_UTF8	1742	XPUSHs (ST (0));
876	indent = F_INDENT	1743
877	canonical = F_CANONICAL	1744	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:	1745	CODE:
884	{	1746	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:	1747	OUTPUT:
894	RETVAL	1748	RETVAL
895		1749
896	void encode (SV self, SV scalar)	1750	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
897	PPCODE:	1751	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1752	{
		1753	SvREFCNT_dec (self->cb_object);
		1754	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899		1755
900	void decode (SV self, SV jsonstr)	1756	XPUSHs (ST (0));
		1757	}
		1758
		1759	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
901	PPCODE:	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:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	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
		1866	CODE:
		1867	{
		1868	if (self->incr_pos)
		1869	croak ("incr_text can not be called when the incremental parser already started parsing");
		1870
		1871	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1872	}
		1873	OUTPUT:
		1874	RETVAL
		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);
903		1903
904	PROTOTYPES: ENABLE	1904	PROTOTYPES: ENABLE
905		1905
906	void to_json (SV *scalar)	1906	void encode_json (SV *scalar)
		1907	ALIAS:
		1908	to_json_ = 0
		1909	encode_json = F_UTF8
907	PPCODE:	1910	PPCODE:
		1911	{
		1912	JSON json;
		1913	json_init (&json);
		1914	json.flags \|= ix;
908	XPUSHs (encode_json (scalar, F_UTF8));	1915	XPUSHs (encode_json (scalar, &json));
		1916	}
909		1917
910	void from_json (SV *jsonstr)	1918	void decode_json (SV *jsonstr)
		1919	ALIAS:
		1920	from_json_ = 0
		1921	decode_json = F_UTF8
911	PPCODE:	1922	PPCODE:
		1923	{
		1924	JSON json;
		1925	json_init (&json);
		1926	json.flags \|= ix;
912	XPUSHs (decode_json (jsonstr, F_UTF8));	1927	XPUSHs (decode_json (jsonstr, &json, 0));
		1928	}
913		1929
		1930

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs. Revision 1.102 by root, Sat Oct 10 01:48:50 2009 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.102 by root, Sat Oct 10 01:48:50 2009 UTC