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 1.97 by root, Fri Jul 3 09:38:30 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	// 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	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV *sv, char *offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
		187	}
		188
		189	/////////////////////////////////////////////////////////////////////////////
		190	// encoder
27		191
28	// structure used for encoding JSON	192	// structure used for encoding JSON
29	typedef struct	193	typedef struct
30	{	194	{
31	char *cur;	195	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
34	SV *sv;	197	SV *sv; // result scalar
35	UV flags;	198	JSON json;
36	int max_recurse;	199	U32 indent; // indentation level
37	int indent;	200	UV limit; // escape character values >= this value when encoding
38	} enc_t;	201	} enc_t;
39		202
40	// structure used for decoding JSON	203	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)	204	need (enc_t *enc, STRLEN len)
62	{	205	{
63	if (enc->cur + len >= enc->end)	206	if (expect_false (enc->cur + len >= enc->end))
64	{	207	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	212	}
70	}	213	}
71		214
72	static void	215	INLINE void
73	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
74	{	217	{
75	need (enc, 1);	218	need (enc, 1);
76	*enc->cur++ = ch;	219	*enc->cur++ = ch;
77	}	220	}
…		…
85		228
86	while (str < end)	229	while (str < end)
87	{	230	{
88	unsigned char ch = *(unsigned char *)str;	231	unsigned char ch = *(unsigned char *)str;
89		232
90	if (ch == '"')	233	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	234	{
		235	if (expect_false (ch == '"')) // but with slow exceptions
		236	{
92	need (enc, len += 1);	237	need (enc, len += 1);
93	*enc->cur++ = '\\';	238	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	239	*enc->cur++ = '"';
95	++str;
96	}	240	}
97	else if (ch == '\\')	241	else if (expect_false (ch == '\\'))
98	{	242	{
99	need (enc, len += 1);	243	need (enc, len += 1);
100	*enc->cur++ = '\\';	244	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	245	*enc->cur++ = '\\';
102	++str;
103	}	246	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	247	else
105	{
106	*enc->cur++ = ch;	248	*enc->cur++ = ch;
107	++str;	249
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;	250	++str;
122	}	251	}
123	else	252	else
124	{	253	{
125	STRLEN clen;	254	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{	255	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	256	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
131	if (clen < 0)	257	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");	258	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
133	}	259	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
134	else	260	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
135	{
136	uch = ch;
137	clen = 1;
138	}
139		261
140	if (uch < 0x80 || enc->flags & F_ASCII)	262	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	263	{
		264	STRLEN clen;
		265	UV uch;
		266
		267	if (is_utf8)
		268	{
		269	uch = decode_utf8 (str, end - str, &clen);
		270	if (clen == (STRLEN)-1)
		271	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		272	}
		273	else
		274	{
		275	uch = ch;
		276	clen = 1;
		277	}
		278
		279	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
		280	{
		281	if (uch >= 0x10000UL)
		282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
144	need (enc, len += 11);	286	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	290	enc->cur += 12;
		291	}
		292	else
		293	{
		294	need (enc, len += 5);
		295	*enc->cur++ = '\\';
		296	*enc->cur++ = 'u';
		297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
		298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
		299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
		300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
		301	}
		302
		303	str += clen;
		304	}
		305	else if (enc->json.flags & F_LATIN1)
		306	{
		307	*enc->cur++ = uch;
		308	str += clen;
		309	}
		310	else if (is_utf8)
		311	{
		312	need (enc, len += clen);
		313	do
		314	{
		315	*enc->cur++ = *str++;
		316	}
		317	while (--clen);
		318	}
		319	else
		320	{
		321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		322	enc->cur = encode_utf8 (enc->cur, uch);
		323	++str;
		324	}
149	}	325	}
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	}	326	}
176	}	327	}
177		328
178	--len;	329	--len;
179	}	330	}
180	}	331	}
181		332
182	#define INDENT SB \	333	INLINE void
		334	encode_indent (enc_t *enc)
		335	{
183	if (enc->flags & F_INDENT) \	336	if (enc->json.flags & F_INDENT)
184	{ \	337	{
185	int i_; \	338	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	339
187	for (i_ = enc->indent * 3; i_--; )\	340	need (enc, spaces);
		341	memset (enc->cur, ' ', spaces);
		342	enc->cur += spaces;
		343	}
		344	}
		345
		346	INLINE void
		347	encode_space (enc_t *enc)
		348	{
		349	need (enc, 1);
		350	encode_ch (enc, ' ');
		351	}
		352
		353	INLINE void
		354	encode_nl (enc_t *enc)
		355	{
		356	if (enc->json.flags & F_INDENT)
		357	{
		358	need (enc, 1);
188	encode_ch (enc, ' '); \	359	encode_ch (enc, '\n');
189	} \	360	}
190	SE	361	}
191		362
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	363	INLINE void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	364	encode_comma (enc_t *enc)
194	#define COMMA SB \	365	{
195	encode_ch (enc, ','); \	366	encode_ch (enc, ',');
		367
196	if (enc->flags & F_INDENT) \	368	if (enc->json.flags & F_INDENT)
197	NL; \	369	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	370	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	371	encode_space (enc);
200	SE	372	}
201		373
202	static void encode_sv (enc_t *enc, SV *sv);	374	static void encode_sv (enc_t *enc, SV *sv);
203		375
204	static void	376	static void
205	encode_av (enc_t *enc, AV *av)	377	encode_av (enc_t *enc, AV *av)
206	{	378	{
207	int i, len = av_len (av);	379	int i, len = av_len (av);
208		380
		381	if (enc->indent >= enc->json.max_depth)
		382	croak (ERR_NESTING_EXCEEDED);
		383
209	encode_ch (enc, '['); NL;	384	encode_ch (enc, '[');
210	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
211		389
212	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
213	{	391	{
214	INDENT;	392	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));
216		393
		394	encode_indent (enc);
		395
		396	if (svp)
		397	encode_sv (enc, *svp);
		398	else
		399	encode_str (enc, "null", 4, 0);
		400
217	if (i < len)	401	if (i < len)
218	COMMA;	402	encode_comma (enc);
219	}	403	}
220		404
221	NL;	405	encode_nl (enc); --enc->indent; encode_indent (enc);
222		406	}
223	--enc->indent;	407
224	INDENT; encode_ch (enc, ']');	408	encode_ch (enc, ']');
225	}	409	}
226		410
227	static void	411	static void
228	encode_he (enc_t *enc, HE *he)	412	encode_hk (enc_t *enc, HE *he)
229	{	413	{
230	encode_ch (enc, '"');	414	encode_ch (enc, '"');
231		415
232	if (HeKLEN (he) == HEf_SVKEY)	416	if (HeKLEN (he) == HEf_SVKEY)
233	{	417	{
…		…
243	else	427	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	428	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		429
246	encode_ch (enc, '"');	430	encode_ch (enc, '"');
247		431
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	432	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	433	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	434	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));
252	}	435	}
253		436
254	// compare hash entries, used when all keys are bytestrings	437	// compare hash entries, used when all keys are bytestrings
255	static int	438	static int
256	he_cmp_fast (const void *a_, const void *b_)	439	he_cmp_fast (const void *a_, const void *b_)
…		…
261	HE *b = *(HE **)b_;	444	HE *b = *(HE **)b_;
262		445
263	STRLEN la = HeKLEN (a);	446	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	447	STRLEN lb = HeKLEN (b);
265		448
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	449	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	450	cmp = lb - la;
268		451
269	return cmp;	452	return cmp;
270	}	453	}
271		454
272	// compare hash entries, used when some keys are sv's or utf-x	455	// compare hash entries, used when some keys are sv's or utf-x
273	static int	456	static int
274	he_cmp_slow (const void *a, const void *b)	457	he_cmp_slow (const void *a, const void *b)
275	{	458	{
276	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	459	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
277	}	460	}
278		461
279	static void	462	static void
280	encode_hv (enc_t *enc, HV *hv)	463	encode_hv (enc_t *enc, HV *hv)
281	{	464	{
282	int count, i;	465	HE *he;
283		466
284	encode_ch (enc, '{'); NL; ++enc->indent;	467	if (enc->indent >= enc->json.max_depth)
		468	croak (ERR_NESTING_EXCEEDED);
285		469
286	if ((count = hv_iterinit (hv)))	470	encode_ch (enc, '{');
287	{	471
288	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
291	// that randomises hash orderings	475	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL)
		477	{
		478	int count = hv_iterinit (hv);
		479
		480	if (SvMAGICAL (hv))
293	{	481	{
294	HE *he, *hes [count];	482	// need to count by iterating. could improve by dynamically building the vector below
		483	// but I don't care for the speed of this special case.
		484	// note also that we will run into undefined behaviour when the two iterations
		485	// do not result in the same count, something I might care for in some later release.
		486
		487	count = 0;
		488	while (hv_iternext (hv))
		489	++count;
		490
		491	hv_iterinit (hv);
		492	}
		493
		494	if (count)
		495	{
295	int fast = 1;	496	int i, fast = 1;
		497	#if defined(__BORLANDC__) || defined(_MSC_VER)
		498	HE **hes = _alloca (count * sizeof (HE));
		499	#else
		500	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		501	#endif
296		502
297	i = 0;	503	i = 0;
298	while ((he = hv_iternext (hv)))	504	while ((he = hv_iternext (hv)))
299	{	505	{
300	hes [i++] = he;	506	hes [i++] = he;
…		…
306		512
307	if (fast)	513	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	514	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	515	else
310	{	516	{
311	// hack to disable "use bytes"	517	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	518	COP cop = *PL_curcop;
313	cop.op_private = 0;	519	cop.op_private = 0;
		520
		521	ENTER;
		522	SAVETMPS;
		523
		524	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	525	PL_curcop = &cop;
315		526
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	527	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		528
318	FREETMPS;	529	FREETMPS;
319		530	LEAVE;
320	PL_curcop = oldcop;
321	}	531	}
322		532
323	for (i = 0; i < count; ++i)	533	encode_nl (enc); ++enc->indent;
		534
		535	while (count--)
324	{	536	{
325	INDENT;	537	encode_indent (enc);
		538	he = hes [count];
326	encode_he (enc, hes [i]);	539	encode_hk (enc, he);
		540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
327		541
328	if (i < count - 1)	542	if (count)
329	COMMA;	543	encode_comma (enc);
330	}	544	}
331		545
		546	encode_nl (enc); --enc->indent; encode_indent (enc);
		547	}
		548	}
		549	else
		550	{
		551	if (hv_iterinit (hv) || SvMAGICAL (hv))
		552	if ((he = hv_iternext (hv)))
332	NL;	553	{
		554	encode_nl (enc); ++enc->indent;
		555
		556	for (;;)
		557	{
		558	encode_indent (enc);
		559	encode_hk (enc, he);
		560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		561
		562	if (!(he = hv_iternext (hv)))
		563	break;
		564
		565	encode_comma (enc);
		566	}
		567
		568	encode_nl (enc); --enc->indent; encode_indent (enc);
		569	}
		570	}
		571
		572	encode_ch (enc, '}');
		573	}
		574
		575	// encode objects, arrays and special \0=false and \1=true values.
		576	static void
		577	encode_rv (enc_t *enc, SV *sv)
		578	{
		579	svtype svt;
		580
		581	SvGETMAGIC (sv);
		582	svt = SvTYPE (sv);
		583
		584	if (expect_false (SvOBJECT (sv)))
		585	{
		586	HV *stash = !JSON_SLOW || json_boolean_stash
		587	? json_boolean_stash
		588	: gv_stashpv ("JSON::XS::Boolean", 1);
		589
		590	if (SvSTASH (sv) == stash)
		591	{
		592	if (SvIV (sv))
		593	encode_str (enc, "true", 4, 0);
		594	else
		595	encode_str (enc, "false", 5, 0);
333	}	596	}
334	else	597	else
335	{	598	{
336	SV *sv;	599	#if 0
337	HE *he = hv_iternext (hv);	600	if (0 && sv_derived_from (rv, "JSON::Literal"))
338
339	for (;;)
340	{	601	{
341	INDENT;	602	// not yet
342	encode_he (enc, he);
343
344	if (!(he = hv_iternext (hv)))
345	break;
346
347	COMMA;
348	}	603	}
349		604	#endif
		605	if (enc->json.flags & F_CONV_BLESSED)
350	NL;	606	{
		607	// we re-bless the reference to get overload and other niceties right
		608	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		609
		610	if (to_json)
		611	{
		612	dSP;
		613
		614	ENTER; SAVETMPS; PUSHMARK (SP);
		615	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		616
		617	// calling with G_SCALAR ensures that we always get a 1 return value
		618	PUTBACK;
		619	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		620	SPAGAIN;
		621
		622	// catch this surprisingly common error
		623	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		624	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		625
		626	sv = POPs;
		627	PUTBACK;
		628
		629	encode_sv (enc, sv);
		630
		631	FREETMPS; LEAVE;
		632	}
		633	else if (enc->json.flags & F_ALLOW_BLESSED)
		634	encode_str (enc, "null", 4, 0);
		635	else
		636	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		637	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		638	}
		639	else if (enc->json.flags & F_ALLOW_BLESSED)
		640	encode_str (enc, "null", 4, 0);
		641	else
		642	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		643	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
351	}	644	}
352	}	645	}
		646	else if (svt == SVt_PVHV)
		647	encode_hv (enc, (HV *)sv);
		648	else if (svt == SVt_PVAV)
		649	encode_av (enc, (AV *)sv);
		650	else if (svt < SVt_PVAV)
		651	{
		652	STRLEN len = 0;
		653	char *pv = svt ? SvPV (sv, len) : 0;
353		654
354	--enc->indent; INDENT; encode_ch (enc, '}');	655	if (len == 1 && *pv == '1')
		656	encode_str (enc, "true", 4, 0);
		657	else if (len == 1 && *pv == '0')
		658	encode_str (enc, "false", 5, 0);
		659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		660	encode_str (enc, "null", 4, 0);
		661	else
		662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
		667	else
		668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	670	}
356		671
357	static void	672	static void
358	encode_sv (enc_t *enc, SV *sv)	673	encode_sv (enc_t *enc, SV *sv)
359	{	674	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	682	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	683	encode_ch (enc, '"');
369	}	684	}
370	else if (SvNOKp (sv))	685	else if (SvNOKp (sv))
371	{	686	{
		687	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	688	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
375	}	691	}
376	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
377	{	693	{
		694	// we assume we can always read an IV as a UV and vice versa
		695	// we assume two's complement
		696	// we assume no aliasing issues in the union
		697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		699	{
		700	// optimise the "small number case"
		701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
		703	I32 i = SvIVX (sv);
		704	U32 u;
		705	char digit, nz = 0;
		706
378	need (enc, 64);	707	need (enc, 6);
		708
		709	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		710	u = i < 0 ? -i : i;
		711
		712	// convert to 4.28 fixed-point representation
		713	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		714
		715	// now output digit by digit, each time masking out the integer part
		716	// and multiplying by 5 while moving the decimal point one to the right,
		717	// resulting in a net multiplication by 10.
		718	// we always write the digit to memory but conditionally increment
		719	// the pointer, to enable the use of conditional move instructions.
		720	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		721	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		722	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		723	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
379	enc->cur +=	730	enc->cur +=
380	SvIsUV(sv)	731	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
383	}	735	}
384	else if (SvROK (sv))	736	else if (SvROK (sv))
385	{	737	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)	738	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);	739	encode_str (enc, "null", 4, 0);
402	else	740	else
403	croak ("encountered perl type that JSON cannot handle");	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		742	SvPV_nolen (sv), SvFLAGS (sv));
404	}	743	}
405		744
406	static SV *	745	static SV *
407	encode_json (SV *scalar, UV flags)	746	encode_json (SV *scalar, JSON *json)
408	{	747	{
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;	748	enc_t enc;
413	enc.flags = flags;	749
		750	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		751	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		752
		753	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	757	enc.indent = 0;
		758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
419		761
420	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
422		764
		765	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		766	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		767
423	if (!(flags & (F_ASCII | F_UTF8)))	768	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	769	SvUTF8_on (enc.sv);
425		770
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	771	if (enc.json.flags & F_SHRINK)
		772	shrink (enc.sv);
		773
427	return enc.sv;	774	return enc.sv;
428	}	775	}
429		776
430	/////////////////////////////////////////////////////////////////////////////	777	/////////////////////////////////////////////////////////////////////////////
		778	// decoder
431		779
432	#define WS \	780	// structure used for decoding JSON
		781	typedef struct
		782	{
		783	char *cur; // current parser pointer
		784	char *end; // end of input string
		785	const char *err; // parse error, if != 0
		786	JSON json;
		787	U32 depth; // recursion depth
		788	U32 maxdepth; // recursion depth limit
		789	} dec_t;
		790
		791	INLINE void
		792	decode_comment (dec_t *dec)
		793	{
		794	// only '#'-style comments allowed a.t.m.
		795
		796	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		797	++dec->cur;
		798	}
		799
		800	INLINE void
		801	decode_ws (dec_t *dec)
		802	{
433	for (;;) \	803	for (;;)
434	{ \	804	{
435	char ch = *dec->cur; \	805	char ch = *dec->cur;
		806
436	if (ch > 0x20 \	807	if (ch > 0x20)
		808	{
		809	if (expect_false (ch == '#'))
		810	{
		811	if (dec->json.flags & F_RELAXED)
		812	decode_comment (dec);
		813	else
		814	break;
		815	}
		816	else
		817	break;
		818	}
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	819	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
438	break; \	820	break; // parse error, but let higher level handle it, gives better error messages
		821
439	++dec->cur; \	822	++dec->cur;
440	}	823	}
		824	}
441		825
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	826	#define ERR(reason) SB dec->err = reason; goto fail; SE
		827
443	#define EXPECT_CH(ch) SB \	828	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	829	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	830	ERR (# ch " expected"); \
446	++dec->cur; \	831	++dec->cur; \
447	SE	832	SE
448		833
		834	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
		835	#define DEC_DEC_DEPTH --dec->depth
		836
449	static SV *decode_sv (dec_t *dec);	837	static SV *decode_sv (dec_t *dec);
450		838
451	static signed char decode_hexdigit[256];	839	static signed char decode_hexdigit[256];
452		840
453	static UV	841	static UV
454	decode_4hex (dec_t *dec)	842	decode_4hex (dec_t *dec)
455	{	843	{
456	signed char d1, d2, d3, d4;	844	signed char d1, d2, d3, d4;
		845	unsigned char *cur = (unsigned char *)dec->cur;
457		846
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	847	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");	848	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]];	849	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");	850	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		851
467	dec->cur += 4;	852	dec->cur += 4;
468		853
469	return ((UV)d1) << 12	854	return ((UV)d1) << 12
470	| ((UV)d2) << 8	855	| ((UV)d2) << 8
…		…
473		858
474	fail:	859	fail:
475	return (UV)-1;	860	return (UV)-1;
476	}	861	}
477		862
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 *	863	static SV *
494	decode_str (dec_t *dec)	864	decode_str (dec_t *dec)
495	{	865	{
496	SV *sv = NEWSV (0,2);	866	SV *sv = 0;
497	int utf8 = 0;	867	int utf8 = 0;
498	char *cur = SvPVX (sv);	868	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		869
501	for (;;)	870	do
502	{	871	{
503	unsigned char ch = *(unsigned char *)dec->cur;	872	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		873	char *cur = buf;
504		874
505	if (ch == '"')	875	do
506	break;
507	else if (ch == '\\')
508	{	876	{
509	switch (*++dec->cur)	877	unsigned char ch = *(unsigned char *)dec_cur++;
		878
		879	if (expect_false (ch == '"'))
510	{	880	{
511	case '\\':	881	--dec_cur;
512	case '/':	882	break;
513	case '"': APPEND_CH (*dec->cur++); break;	883	}
514		884	else if (expect_false (ch == '\\'))
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	885	{
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;	886	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	{	887	{
523	UV lo, hi;	888	case '\\':
524	++dec->cur;	889	case '/':
		890	case '"': *cur++ = *dec_cur++; break;
525		891
526	hi = decode_4hex (dec);	892	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	893	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	894	case 'n': ++dec_cur; *cur++ = '\012'; break;
		895	case 'f': ++dec_cur; *cur++ = '\014'; break;
		896	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		897
530	// possibly a surrogate pair	898	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	899	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	900	UV lo, hi;
534	ERR ("illegal surrogate character");	901	++dec_cur;
535		902
536	dec->cur += 2;	903	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	904	hi = decode_4hex (dec);
		905	dec_cur = dec->cur;
539	if (lo == (UV)-1)	906	if (hi == (UV)-1)
540	goto fail;	907	goto fail;
541		908
		909	// possibly a surrogate pair
		910	if (hi >= 0xd800)
		911	if (hi < 0xdc00)
		912	{
		913	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		914	ERR ("missing low surrogate character in surrogate pair");
		915
		916	dec_cur += 2;
		917
		918	dec->cur = dec_cur;
		919	lo = decode_4hex (dec);
		920	dec_cur = dec->cur;
		921	if (lo == (UV)-1)
		922	goto fail;
		923
542	if (lo < 0xdc00 || lo >= 0xe000)	924	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	925	ERR ("surrogate pair expected");
544		926
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	927	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		928	}
		929	else if (hi < 0xe000)
		930	ERR ("missing high surrogate character in surrogate pair");
		931
		932	if (hi >= 0x80)
		933	{
		934	utf8 = 1;
		935
		936	cur = encode_utf8 (cur, hi);
		937	}
		938	else
		939	*cur++ = hi;
546	}	940	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	941	break;
552	utf8 = 1;
553		942
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	943	default:
558	APPEND_CH (hi);	944	--dec_cur;
		945	ERR ("illegal backslash escape sequence in string");
559	}	946	}
		947	}
		948	else if (expect_true (ch >= 0x20 && ch < 0x80))
		949	*cur++ = ch;
		950	else if (ch >= 0x80)
		951	{
		952	STRLEN clen;
		953	UV uch;
		954
		955	--dec_cur;
		956
		957	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		958	if (clen == (STRLEN)-1)
		959	ERR ("malformed UTF-8 character in JSON string");
		960
		961	do
		962	*cur++ = *dec_cur++;
		963	while (--clen);
		964
		965	utf8 = 1;
		966	}
		967	else
		968	{
		969	--dec_cur;
		970
		971	if (!ch)
		972	ERR ("unexpected end of string while parsing JSON string");
560	break;	973	else
		974	ERR ("invalid character encountered while parsing JSON string");
561	}	975	}
562	}	976	}
563	else if (ch >= 0x20 && ch <= 0x7f)	977	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	978
565	else if (ch >= 0x80)	979	{
		980	STRLEN len = cur - buf;
		981
		982	if (sv)
566	{	983	{
567	STRLEN clen;	984	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		985
572	APPEND_GROW (clen);	986	if (SvLEN (sv) <= cur + len)
573	memcpy (cur, dec->cur, clen);	987	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
574	cur += clen;	988
575	dec->cur += clen;	989	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		990	SvCUR_set (sv, SvCUR (sv) + len);
576	}	991	}
577	else	992	else
578	ERR ("invalid character encountered");	993	sv = newSVpvn (buf, len);
579	}	994	}
		995	}
		996	while (*dec_cur != '"');
580		997
581	++dec->cur;	998	++dec_cur;
582		999
583	SvCUR_set (sv, cur - SvPVX (sv));	1000	if (sv)
584		1001	{
585	SvPOK_only (sv);	1002	SvPOK_only (sv);
586	*SvEND (sv) = 0;	1003	*SvEND (sv) = 0;
587		1004
588	if (utf8)	1005	if (utf8)
589	SvUTF8_on (sv);	1006	SvUTF8_on (sv);
		1007	}
		1008	else
		1009	sv = newSVpvn ("", 0);
590		1010
		1011	dec->cur = dec_cur;
591	return sv;	1012	return sv;
592		1013
593	fail:	1014	fail:
594	SvREFCNT_dec (sv);	1015	dec->cur = dec_cur;
595	return 0;	1016	return 0;
596	}	1017	}
597		1018
598	static SV *	1019	static SV *
599	decode_num (dec_t *dec)	1020	decode_num (dec_t *dec)
…		…
609	{	1030	{
610	++dec->cur;	1031	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	1032	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	1033	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	1034	}
614		1035	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	1036	ERR ("malformed number (no digits after initial minus)");
		1037	else
		1038	do
		1039	{
		1040	++dec->cur;
		1041	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	1042	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		1043
619	// [frac]	1044	// [frac]
620	if (*dec->cur == '.')	1045	if (*dec->cur == '.')
621	{	1046	{
622	is_nv = 1;	1047	++dec->cur;
		1048
		1049	if (*dec->cur < '0' || *dec->cur > '9')
		1050	ERR ("malformed number (no digits after decimal point)");
623		1051
624	do	1052	do
625	{	1053	{
626	++dec->cur;	1054	++dec->cur;
627	}	1055	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	1056	while (*dec->cur >= '0' && *dec->cur <= '9');
		1057
		1058	is_nv = 1;
629	}	1059	}
630		1060
631	// [exp]	1061	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	1062	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	1063	{
634	is_nv = 1;
635
636	++dec->cur;	1064	++dec->cur;
		1065
637	if (*dec->cur == '-' || *dec->cur == '+')	1066	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	1067	++dec->cur;
639		1068
		1069	if (*dec->cur < '0' || *dec->cur > '9')
		1070	ERR ("malformed number (no digits after exp sign)");
		1071
		1072	do
		1073	{
		1074	++dec->cur;
		1075	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	1076	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	1077
		1078	is_nv = 1;
642	}	1079	}
643		1080
644	if (!is_nv)	1081	if (!is_nv)
645	{	1082	{
646	UV uv;	1083	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	1084
648	if (numtype & IS_NUMBER_IN_UV)	1085	// special case the rather common 1..5-digit-int case
649	if (numtype & IS_NUMBER_NEG)	1086	if (*start == '-')
		1087	switch (len)
650	{	1088	{
651	if (uv < (UV)IV_MIN)	1089	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	1090	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
		1091	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1092	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1093	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
653	}	1094	}
		1095	else
		1096	switch (len)
		1097	{
		1098	case 1: return newSViv ( start [0] - '0' * 1);
		1099	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1100	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1101	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1102	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1103	}
		1104
		1105	{
		1106	UV uv;
		1107	int numtype = grok_number (start, len, &uv);
		1108	if (numtype & IS_NUMBER_IN_UV)
		1109	if (numtype & IS_NUMBER_NEG)
		1110	{
		1111	if (uv < (UV)IV_MIN)
		1112	return newSViv (-(IV)uv);
		1113	}
654	else	1114	else
655	return newSVuv (uv);	1115	return newSVuv (uv);
656	}	1116	}
657		1117
		1118	len -= *start == '-' ? 1 : 0;
		1119
		1120	// does not fit into IV or UV, try NV
		1121	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1122	#if defined (LDBL_DIG)
		1123	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1124	#endif
		1125	)
		1126	// fits into NV without loss of precision
		1127	return newSVnv (Atof (start));
		1128
		1129	// everything else fails, convert it to a string
		1130	return newSVpvn (start, dec->cur - start);
		1131	}
		1132
		1133	// loss of precision here
658	return newSVnv (Atof (start));	1134	return newSVnv (Atof (start));
659		1135
660	fail:	1136	fail:
661	return 0;	1137	return 0;
662	}	1138	}
…		…
664	static SV *	1140	static SV *
665	decode_av (dec_t *dec)	1141	decode_av (dec_t *dec)
666	{	1142	{
667	AV *av = newAV ();	1143	AV *av = newAV ();
668		1144
		1145	DEC_INC_DEPTH;
		1146	decode_ws (dec);
		1147
		1148	if (*dec->cur == ']')
		1149	++dec->cur;
		1150	else
669	for (;;)	1151	for (;;)
670	{	1152	{
671	SV *value;	1153	SV *value;
672		1154
673	value = decode_sv (dec);	1155	value = decode_sv (dec);
674	if (!value)	1156	if (!value)
675	goto fail;	1157	goto fail;
676		1158
677	av_push (av, value);	1159	av_push (av, value);
678		1160
679	WS;	1161	decode_ws (dec);
680		1162
681	if (*dec->cur == ']')	1163	if (*dec->cur == ']')
682	{	1164	{
683	++dec->cur;	1165	++dec->cur;
684	break;	1166	break;
		1167	}
685	}	1168
686
687	if (*dec->cur != ',')	1169	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1170	ERR (", or ] expected while parsing array");
689		1171
690	++dec->cur;	1172	++dec->cur;
		1173
		1174	decode_ws (dec);
		1175
		1176	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1177	{
		1178	++dec->cur;
		1179	break;
		1180	}
691	}	1181	}
692		1182
		1183	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1184	return newRV_noinc ((SV *)av);
694		1185
695	fail:	1186	fail:
696	SvREFCNT_dec (av);	1187	SvREFCNT_dec (av);
		1188	DEC_DEC_DEPTH;
697	return 0;	1189	return 0;
698	}	1190	}
699		1191
700	static SV *	1192	static SV *
701	decode_hv (dec_t *dec)	1193	decode_hv (dec_t *dec)
702	{	1194	{
		1195	SV *sv;
703	HV *hv = newHV ();	1196	HV *hv = newHV ();
704		1197
		1198	DEC_INC_DEPTH;
		1199	decode_ws (dec);
		1200
		1201	if (*dec->cur == '}')
		1202	++dec->cur;
		1203	else
705	for (;;)	1204	for (;;)
706	{	1205	{
707	SV *key, *value;
708
709	WS; EXPECT_CH ('"');	1206	EXPECT_CH ('"');
710		1207
711	key = decode_str (dec);	1208	// heuristic: assume that
712	if (!key)	1209	// a) decode_str + hv_store_ent are abysmally slow.
713	goto fail;	1210	// b) most hash keys are short, simple ascii text.
714		1211	// => try to "fast-match" such strings to avoid
715	WS; EXPECT_CH (':');	1212	// the overhead of decode_str + hv_store_ent.
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1213	{
		1214	SV *value;
		1215	char *p = dec->cur;
		1216	char *e = p + 24; // only try up to 24 bytes
		1217
		1218	for (;;)
		1219	{
		1220	// the >= 0x80 is false on most architectures
		1221	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1222	{
		1223	// slow path, back up and use decode_str
		1224	SV *key = decode_str (dec);
		1225	if (!key)
		1226	goto fail;
		1227
		1228	decode_ws (dec); EXPECT_CH (':');
		1229
		1230	decode_ws (dec);
		1231	value = decode_sv (dec);
		1232	if (!value)
		1233	{
		1234	SvREFCNT_dec (key);
		1235	goto fail;
		1236	}
		1237
		1238	hv_store_ent (hv, key, value, 0);
720	SvREFCNT_dec (key);	1239	SvREFCNT_dec (key);
		1240
		1241	break;
		1242	}
		1243	else if (*p == '"')
		1244	{
		1245	// fast path, got a simple key
		1246	char *key = dec->cur;
		1247	int len = p - key;
		1248	dec->cur = p + 1;
		1249
		1250	decode_ws (dec); EXPECT_CH (':');
		1251
		1252	decode_ws (dec);
		1253	value = decode_sv (dec);
		1254	if (!value)
721	goto fail;	1255	goto fail;
		1256
		1257	hv_store (hv, key, len, value, 0);
		1258
		1259	break;
		1260	}
		1261
		1262	++p;
		1263	}
722	}	1264	}
723		1265
724	//TODO: optimise	1266	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1267
727	WS;
728
729	if (*dec->cur == '}')	1268	if (*dec->cur == '}')
		1269	{
		1270	++dec->cur;
		1271	break;
		1272	}
		1273
		1274	if (*dec->cur != ',')
		1275	ERR (", or } expected while parsing object/hash");
		1276
		1277	++dec->cur;
		1278
		1279	decode_ws (dec);
		1280
		1281	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1282	{
		1283	++dec->cur;
		1284	break;
		1285	}
		1286	}
		1287
		1288	DEC_DEC_DEPTH;
		1289	sv = newRV_noinc ((SV *)hv);
		1290
		1291	// check filter callbacks
		1292	if (dec->json.flags & F_HOOK)
		1293	{
		1294	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730	{	1295	{
731	++dec->cur;	1296	HE *cb, *he;
732	break;	1297
		1298	hv_iterinit (hv);
		1299	he = hv_iternext (hv);
		1300	hv_iterinit (hv);
		1301
		1302	// the next line creates a mortal sv each time its called.
		1303	// might want to optimise this for common cases.
		1304	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1305
		1306	if (cb)
		1307	{
		1308	dSP;
		1309	int count;
		1310
		1311	ENTER; SAVETMPS; PUSHMARK (SP);
		1312	XPUSHs (HeVAL (he));
		1313
		1314	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1315
		1316	if (count == 1)
		1317	{
		1318	sv = newSVsv (POPs);
		1319	FREETMPS; LEAVE;
		1320	return sv;
		1321	}
		1322
		1323	FREETMPS; LEAVE;
		1324	}
733	}	1325	}
734		1326
735	if (*dec->cur != ',')	1327	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1328	{
		1329	dSP;
		1330	int count;
737		1331
738	++dec->cur;	1332	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1333	XPUSHs (sv_2mortal (sv));
740		1334
741	return newRV_noinc ((SV *)hv);	1335	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1336
		1337	if (count == 1)
		1338	{
		1339	sv = newSVsv (POPs);
		1340	FREETMPS; LEAVE;
		1341	return sv;
		1342	}
		1343
		1344	SvREFCNT_inc (sv);
		1345	FREETMPS; LEAVE;
		1346	}
		1347	}
		1348
		1349	return sv;
742		1350
743	fail:	1351	fail:
744	SvREFCNT_dec (hv);	1352	SvREFCNT_dec (hv);
		1353	DEC_DEC_DEPTH;
745	return 0;	1354	return 0;
746	}	1355	}
747		1356
748	static SV *	1357	static SV *
749	decode_sv (dec_t *dec)	1358	decode_sv (dec_t *dec)
750	{	1359	{
751	WS;	1360	// the beauty of JSON: you need exactly one character lookahead
		1361	// to parse everything.
752	switch (*dec->cur)	1362	switch (*dec->cur)
753	{	1363	{
754	case '"': ++dec->cur; return decode_str (dec);	1364	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1365	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1366	case '{': ++dec->cur; return decode_hv (dec);
757		1367
758	case '-':	1368	case '-':
759	case '0': case '1': case '2': case '3': case '4':	1369	case '0': case '1': case '2': case '3': case '4':
760	case '5': case '6': case '7': case '8': case '9':	1370	case '5': case '6': case '7': case '8': case '9':
761	return decode_num (dec);	1371	return decode_num (dec);
762		1372
763	case 't':	1373	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1374	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1375	{
766	dec->cur += 4;	1376	dec->cur += 4;
		1377	#if JSON_SLOW
		1378	json_true = get_bool ("JSON::XS::true");
		1379	#endif
767	return newSViv (1);	1380	return newSVsv (json_true);
768	}	1381	}
769	else	1382	else
770	ERR ("'true' expected");	1383	ERR ("'true' expected");
771		1384
772	break;	1385	break;
773		1386
774	case 'f':	1387	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1388	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1389	{
777	dec->cur += 5;	1390	dec->cur += 5;
		1391	#if JSON_SLOW
		1392	json_false = get_bool ("JSON::XS::false");
		1393	#endif
778	return newSViv (0);	1394	return newSVsv (json_false);
779	}	1395	}
780	else	1396	else
781	ERR ("'false' expected");	1397	ERR ("'false' expected");
782		1398
783	break;	1399	break;
784		1400
785	case 'n':	1401	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1402	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1403	{
788	dec->cur += 4;	1404	dec->cur += 4;
789	return newSViv (1);	1405	return newSVsv (&PL_sv_undef);
790	}	1406	}
791	else	1407	else
792	ERR ("'null' expected");	1408	ERR ("'null' expected");
793		1409
794	break;	1410	break;
795		1411
796	default:	1412	default:
797	ERR ("malformed json string");	1413	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1414	break;
799	}	1415	}
800		1416
801	fail:	1417	fail:
802	return 0;	1418	return 0;
803	}	1419	}
804		1420
805	static SV *	1421	static SV *
806	decode_json (SV *string, UV flags)	1422	decode_json (SV *string, JSON *json, char **offset_return)
807	{	1423	{
		1424	dec_t dec;
808	SV *sv;	1425	SV *sv;
809		1426
		1427	/* work around bugs in 5.10 where manipulating magic values
		1428	* will perl ignore the magic in subsequent accesses
		1429	*/
		1430	/*SvGETMAGIC (string);*/
		1431	if (SvMAGICAL (string))
		1432	string = sv_2mortal (newSVsv (string));
		1433
		1434	SvUPGRADE (string, SVt_PV);
		1435
		1436	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1437	* assertion with -DDEBUGGING, although SvCUR is documented to
		1438	* return the xpv_cur field which certainly exists after upgrading.
		1439	* according to nicholas clark, calling SvPOK fixes this.
		1440	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1441	* and hope for the best.
		1442	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1443	* assertion business is seriously broken, try yet another workaround
		1444	* for the broken -DDEBUGGING.
		1445	*/
		1446	{
		1447	#ifdef DEBUGGING
		1448	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1449	#else
		1450	STRLEN offset = SvCUR (string);
		1451	#endif
		1452
		1453	if (offset > json->max_size && json->max_size)
		1454	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1455	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1456	}
		1457
810	if (!(flags & F_UTF8))	1458	if (json->flags & F_UTF8)
		1459	sv_utf8_downgrade (string, 0);
		1460	else
811	sv_utf8_upgrade (string);	1461	sv_utf8_upgrade (string);
812		1462
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1463	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1464
815	dec_t dec;	1465	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1466	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1467	dec.end = SvEND (string);
819	dec.err = 0;	1468	dec.err = 0;
		1469	dec.depth = 0;
820		1470
821	*dec.end = 1; // invalid anywhere	1471	if (dec.json.cb_object || dec.json.cb_sk_object)
		1472	dec.json.flags |= F_HOOK;
		1473
		1474	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1475
		1476	decode_ws (&dec);
822	sv = decode_sv (&dec);	1477	sv = decode_sv (&dec);
823	*dec.end = 0;	1478
		1479	if (offset_return)
		1480	*offset_return = dec.cur;
		1481
		1482	if (!(offset_return || !sv))
		1483	{
		1484	// check for trailing garbage
		1485	decode_ws (&dec);
		1486
		1487	if (*dec.cur)
		1488	{
		1489	dec.err = "garbage after JSON object";
		1490	SvREFCNT_dec (sv);
		1491	sv = 0;
		1492	}
		1493	}
824		1494
825	if (!sv)	1495	if (!sv)
826	{	1496	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1497	SV *uni = sv_newmortal ();
829		1498
		1499	// horrible hack to silence warning inside pv_uni_display
		1500	COP cop = *PL_curcop;
		1501	cop.cop_warnings = pWARN_NONE;
		1502	ENTER;
		1503	SAVEVPTR (PL_curcop);
		1504	PL_curcop = &cop;
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1505	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1506	LEAVE;
		1507
831	croak ("%s, at character %d (%s)",	1508	croak ("%s, at character offset %d (before \"%s\")",
832	dec.err,	1509	dec.err,
833	(int)offset,	1510	ptr_to_index (string, dec.cur),
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1511	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1512	}
836		1513
837	sv = sv_2mortal (sv);	1514	sv = sv_2mortal (sv);
838		1515
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1516	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)");	1517	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1518
842	return sv;	1519	return sv;
843	}	1520	}
844		1521
		1522	/////////////////////////////////////////////////////////////////////////////
		1523	// incremental parser
		1524
		1525	static void
		1526	incr_parse (JSON *self)
		1527	{
		1528	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1529
		1530	for (;;)
		1531	{
		1532	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1533	switch (self->incr_mode)
		1534	{
		1535	// only used for intiial whitespace skipping
		1536	case INCR_M_WS:
		1537	for (;;)
		1538	{
		1539	if (*p > 0x20)
		1540	{
		1541	self->incr_mode = INCR_M_JSON;
		1542	goto incr_m_json;
		1543	}
		1544	else if (!*p)
		1545	goto interrupt;
		1546
		1547	++p;
		1548	}
		1549
		1550	// skip a single char inside a string (for \\-processing)
		1551	case INCR_M_BS:
		1552	if (!*p)
		1553	goto interrupt;
		1554
		1555	++p;
		1556	self->incr_mode = INCR_M_STR;
		1557	goto incr_m_str;
		1558
		1559	// inside a string
		1560	case INCR_M_STR:
		1561	incr_m_str:
		1562	for (;;)
		1563	{
		1564	if (*p == '"')
		1565	{
		1566	++p;
		1567	self->incr_mode = INCR_M_JSON;
		1568
		1569	if (!self->incr_nest)
		1570	goto interrupt;
		1571
		1572	goto incr_m_json;
		1573	}
		1574	else if (*p == '\\')
		1575	{
		1576	++p; // "virtually" consumes character after \
		1577
		1578	if (!*p) // if at end of string we have to switch modes
		1579	{
		1580	self->incr_mode = INCR_M_BS;
		1581	goto interrupt;
		1582	}
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	// after initial ws, outside string
		1591	case INCR_M_JSON:
		1592	incr_m_json:
		1593	for (;;)
		1594	{
		1595	switch (*p++)
		1596	{
		1597	case 0:
		1598	--p;
		1599	goto interrupt;
		1600
		1601	case 0x09:
		1602	case 0x0a:
		1603	case 0x0d:
		1604	case 0x20:
		1605	if (!self->incr_nest)
		1606	{
		1607	--p; // do not eat the whitespace, let the next round do it
		1608	goto interrupt;
		1609	}
		1610	break;
		1611
		1612	case '"':
		1613	self->incr_mode = INCR_M_STR;
		1614	goto incr_m_str;
		1615
		1616	case '[':
		1617	case '{':
		1618	if (++self->incr_nest > self->max_depth)
		1619	croak (ERR_NESTING_EXCEEDED);
		1620	break;
		1621
		1622	case ']':
		1623	case '}':
		1624	if (--self->incr_nest <= 0)
		1625	goto interrupt;
		1626	}
		1627	}
		1628	}
		1629
		1630	modechange:
		1631	;
		1632	}
		1633
		1634	interrupt:
		1635	self->incr_pos = p - SvPVX (self->incr_text);
		1636	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1637	}
		1638
		1639	/////////////////////////////////////////////////////////////////////////////
		1640	// XS interface functions
		1641
845	MODULE = JSON::XS PACKAGE = JSON::XS	1642	MODULE = JSON::XS PACKAGE = JSON::XS
846		1643
847	BOOT:	1644	BOOT:
848	{	1645	{
849	int i;	1646	int i;
850		1647
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1648	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1649	decode_hexdigit [i] =
		1650	i >= '0' && i <= '9' ? i - '0'
		1651	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1652	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1653	: -1;
854		1654
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);	1655	json_stash = gv_stashpv ("JSON::XS" , 1);
		1656	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1657
		1658	json_true = get_bool ("JSON::XS::true");
		1659	json_false = get_bool ("JSON::XS::false");
862	}	1660	}
863		1661
864	PROTOTYPES: DISABLE	1662	PROTOTYPES: DISABLE
865		1663
866	SV *new (char *dummy)	1664	void CLONE (...)
867	CODE:	1665	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1666	json_stash = 0;
		1667	json_boolean_stash = 0;
		1668
		1669	void new (char *klass)
		1670	PPCODE:
		1671	{
		1672	SV *pv = NEWSV (0, sizeof (JSON));
		1673	SvPOK_only (pv);
		1674	json_init ((JSON *)SvPVX (pv));
		1675	XPUSHs (sv_2mortal (sv_bless (
		1676	newRV_noinc (pv),
		1677	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1678	)));
		1679	}
		1680
		1681	void ascii (JSON *self, int enable = 1)
		1682	ALIAS:
		1683	ascii = F_ASCII
		1684	latin1 = F_LATIN1
		1685	utf8 = F_UTF8
		1686	indent = F_INDENT
		1687	canonical = F_CANONICAL
		1688	space_before = F_SPACE_BEFORE
		1689	space_after = F_SPACE_AFTER
		1690	pretty = F_PRETTY
		1691	allow_nonref = F_ALLOW_NONREF
		1692	shrink = F_SHRINK
		1693	allow_blessed = F_ALLOW_BLESSED
		1694	convert_blessed = F_CONV_BLESSED
		1695	relaxed = F_RELAXED
		1696	allow_unknown = F_ALLOW_UNKNOWN
		1697	PPCODE:
		1698	{
		1699	if (enable)
		1700	self->flags |= ix;
		1701	else
		1702	self->flags &= ~ix;
		1703
		1704	XPUSHs (ST (0));
		1705	}
		1706
		1707	void get_ascii (JSON *self)
		1708	ALIAS:
		1709	get_ascii = F_ASCII
		1710	get_latin1 = F_LATIN1
		1711	get_utf8 = F_UTF8
		1712	get_indent = F_INDENT
		1713	get_canonical = F_CANONICAL
		1714	get_space_before = F_SPACE_BEFORE
		1715	get_space_after = F_SPACE_AFTER
		1716	get_allow_nonref = F_ALLOW_NONREF
		1717	get_shrink = F_SHRINK
		1718	get_allow_blessed = F_ALLOW_BLESSED
		1719	get_convert_blessed = F_CONV_BLESSED
		1720	get_relaxed = F_RELAXED
		1721	get_allow_unknown = F_ALLOW_UNKNOWN
		1722	PPCODE:
		1723	XPUSHs (boolSV (self->flags & ix));
		1724
		1725	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1726	PPCODE:
		1727	self->max_depth = max_depth;
		1728	XPUSHs (ST (0));
		1729
		1730	U32 get_max_depth (JSON *self)
		1731	CODE:
		1732	RETVAL = self->max_depth;
869	OUTPUT:	1733	OUTPUT:
870	RETVAL	1734	RETVAL
871		1735
872	SV *ascii (SV *self, int enable)	1736	void max_size (JSON *self, U32 max_size = 0)
873	ALIAS:	1737	PPCODE:
874	ascii = F_ASCII	1738	self->max_size = max_size;
875	utf8 = F_UTF8	1739	XPUSHs (ST (0));
876	indent = F_INDENT	1740
877	canonical = F_CANONICAL	1741	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:	1742	CODE:
884	{	1743	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:	1744	OUTPUT:
894	RETVAL	1745	RETVAL
895		1746
896	void encode (SV *self, SV *scalar)	1747	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
897	PPCODE:	1748	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));	1749	{
		1750	SvREFCNT_dec (self->cb_object);
		1751	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
899		1752
900	void decode (SV *self, SV *jsonstr)	1753	XPUSHs (ST (0));
		1754	}
		1755
		1756	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
901	PPCODE:	1757	PPCODE:
		1758	{
		1759	if (!self->cb_sk_object)
		1760	self->cb_sk_object = newHV ();
		1761
		1762	if (SvOK (cb))
		1763	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1764	else
		1765	{
		1766	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1767
		1768	if (!HvKEYS (self->cb_sk_object))
		1769	{
		1770	SvREFCNT_dec (self->cb_sk_object);
		1771	self->cb_sk_object = 0;
		1772	}
		1773	}
		1774
		1775	XPUSHs (ST (0));
		1776	}
		1777
		1778	void encode (JSON *self, SV *scalar)
		1779	PPCODE:
		1780	XPUSHs (encode_json (scalar, self));
		1781
		1782	void decode (JSON *self, SV *jsonstr)
		1783	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1784	XPUSHs (decode_json (jsonstr, self, 0));
		1785
		1786	void decode_prefix (JSON *self, SV *jsonstr)
		1787	PPCODE:
		1788	{
		1789	char *offset;
		1790	EXTEND (SP, 2);
		1791	PUSHs (decode_json (jsonstr, self, &offset));
		1792	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1793	}
		1794
		1795	void incr_parse (JSON *self, SV *jsonstr = 0)
		1796	PPCODE:
		1797	{
		1798	if (!self->incr_text)
		1799	self->incr_text = newSVpvn ("", 0);
		1800
		1801	// append data, if any
		1802	if (jsonstr)
		1803	{
		1804	if (SvUTF8 (jsonstr))
		1805	{
		1806	if (!SvUTF8 (self->incr_text))
		1807	{
		1808	/* utf-8-ness differs, need to upgrade */
		1809	sv_utf8_upgrade (self->incr_text);
		1810
		1811	if (self->incr_pos)
		1812	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1813	- (U8 *)SvPVX (self->incr_text);
		1814	}
		1815	}
		1816	else if (SvUTF8 (self->incr_text))
		1817	sv_utf8_upgrade (jsonstr);
		1818
		1819	{
		1820	STRLEN len;
		1821	const char *str = SvPV (jsonstr, len);
		1822	STRLEN cur = SvCUR (self->incr_text);
		1823
		1824	if (SvLEN (self->incr_text) <= cur + len)
		1825	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1826
		1827	Move (str, SvEND (self->incr_text), len, char);
		1828	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1829	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1830	}
		1831	}
		1832
		1833	if (GIMME_V != G_VOID)
		1834	do
		1835	{
		1836	char *offset;
		1837
		1838	if (!INCR_DONE (self))
		1839	{
		1840	incr_parse (self);
		1841
		1842	if (self->incr_pos > self->max_size && self->max_size)
		1843	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1844	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1845
		1846	if (!INCR_DONE (self))
		1847	break;
		1848	}
		1849
		1850	XPUSHs (decode_json (self->incr_text, self, &offset));
		1851
		1852	self->incr_pos -= offset - SvPVX (self->incr_text);
		1853	self->incr_nest = 0;
		1854	self->incr_mode = 0;
		1855
		1856	sv_chop (self->incr_text, offset);
		1857	}
		1858	while (GIMME_V == G_ARRAY);
		1859	}
		1860
		1861	SV *incr_text (JSON *self)
		1862	ATTRS: lvalue
		1863	CODE:
		1864	{
		1865	if (self->incr_pos)
		1866	croak ("incr_text can not be called when the incremental parser already started parsing");
		1867
		1868	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1869	}
		1870	OUTPUT:
		1871	RETVAL
		1872
		1873	void incr_skip (JSON *self)
		1874	CODE:
		1875	{
		1876	if (self->incr_pos)
		1877	{
		1878	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1879	self->incr_pos = 0;
		1880	self->incr_nest = 0;
		1881	self->incr_mode = 0;
		1882	}
		1883	}
		1884
		1885	void incr_reset (JSON *self)
		1886	CODE:
		1887	{
		1888	SvREFCNT_dec (self->incr_text);
		1889	self->incr_text = 0;
		1890	self->incr_pos = 0;
		1891	self->incr_nest = 0;
		1892	self->incr_mode = 0;
		1893	}
		1894
		1895	void DESTROY (JSON *self)
		1896	CODE:
		1897	SvREFCNT_dec (self->cb_sk_object);
		1898	SvREFCNT_dec (self->cb_object);
		1899	SvREFCNT_dec (self->incr_text);
903		1900
904	PROTOTYPES: ENABLE	1901	PROTOTYPES: ENABLE
905		1902
906	void to_json (SV *scalar)	1903	void encode_json (SV *scalar)
		1904	ALIAS:
		1905	to_json_ = 0
		1906	encode_json = F_UTF8
907	PPCODE:	1907	PPCODE:
		1908	{
		1909	JSON json;
		1910	json_init (&json);
		1911	json.flags |= ix;
908	XPUSHs (encode_json (scalar, F_UTF8));	1912	XPUSHs (encode_json (scalar, &json));
		1913	}
909		1914
910	void from_json (SV *jsonstr)	1915	void decode_json (SV *jsonstr)
		1916	ALIAS:
		1917	from_json_ = 0
		1918	decode_json = F_UTF8
911	PPCODE:	1919	PPCODE:
		1920	{
		1921	JSON json;
		1922	json_init (&json);
		1923	json.flags |= ix;
912	XPUSHs (decode_json (jsonstr, F_UTF8));	1924	XPUSHs (decode_json (jsonstr, &json, 0));
		1925	}
913		1926
		1927

Diff Legend

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