[ViewVC] Diff of: cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.65 by root, Sat Oct 13 01:55:31 2007 UTC vs.
Revision 1.110 by root, Wed Jun 1 13:01:09 2011 UTC

…		…
4		4
5	#include <assert.h>	5	#include <assert.h>
6	#include <string.h>	6	#include <string.h>
7	#include <stdlib.h>	7	#include <stdlib.h>
8	#include <stdio.h>	8	#include <stdio.h>
		9	#include <limits.h>
9	#include <float.h>	10	#include <float.h>
10		11
11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
12	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
13	#endif	14	#endif
14		15
15	// some old perls do not have this, try to make it work, no	16	// some old perls do not have this, try to make it work, no
16	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarantees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
19	#endif	20	#endif
		21
		22	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20		24
21	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
…		…
28	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	34	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	35	#define F_CONV_BLESSED 0x00000800UL
32	#define F_RELAXED 0x00001000UL	36	#define F_RELAXED 0x00001000UL
33		37	#define F_ALLOW_UNKNOWN 0x00002000UL
34	#define F_MAXDEPTH 0xf8000000UL
35	#define S_MAXDEPTH 27
36	#define F_MAXSIZE 0x01f00000UL
37	#define S_MAXSIZE 20
38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
39		39
40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
42
43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
44	#define F_DEFAULT (9UL << S_MAXDEPTH)
45		41
46	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
47	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
48		44
49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
50		46
		47	#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
		48
51	#define SB do {	49	#define SB do {
52	#define SE } while (0)	50	#define SE } while (0)
53		51
54	#if __GNUC__ >= 3	52	#if __GNUC__ >= 3
55	# define expect(expr,value) __builtin_expect ((expr),(value))	53	# define expect(expr,value) __builtin_expect ((expr), (value))
56	# define inline inline	54	# define INLINE static inline
57	#else	55	#else
58	# define expect(expr,value) (expr)	56	# define expect(expr,value) (expr)
59	# define inline static	57	# define INLINE static
60	#endif	58	#endif
61		59
62	#define expect_false(expr) expect ((expr) != 0, 0)	60	#define expect_false(expr) expect ((expr) != 0, 0)
63	#define expect_true(expr) expect ((expr) != 0, 1)	61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#define IN_RANGE_INC(type,val,beg,end) \
		64	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		65	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		66
		67	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
64		68
65	#ifdef USE_ITHREADS	69	#ifdef USE_ITHREADS
66	# define JSON_SLOW 1	70	# define JSON_SLOW 1
67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
68	#else	72	#else
…		…
71	#endif	75	#endif
72		76
73	static HV json_stash, json_boolean_stash; // JSON::XS::	77	static HV json_stash, json_boolean_stash; // JSON::XS::
74	static SV json_true, json_false;	78	static SV json_true, json_false;
75		79
		80	enum {
		81	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		82	INCR_M_STR, // inside string
		83	INCR_M_BS, // inside backslash
		84	INCR_M_C0, // inside comment in initial whitespace sequence
		85	INCR_M_C1, // inside comment in other places
		86	INCR_M_JSON // outside anything, count nesting
		87	};
		88
		89	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		90
76	typedef struct {	91	typedef struct {
77	U32 flags;	92	U32 flags;
		93	U32 max_depth;
		94	STRLEN max_size;
		95
78	SV *cb_object;	96	SV *cb_object;
79	HV *cb_sk_object;	97	HV *cb_sk_object;
		98
		99	// for the incremental parser
		100	SV *incr_text; // the source text so far
		101	STRLEN incr_pos; // the current offset into the text
		102	int incr_nest; // {[]}-nesting level
		103	unsigned char incr_mode;
80	} JSON;	104	} JSON;
		105
		106	INLINE void
		107	json_init (JSON *json)
		108	{
		109	Zero (json, 1, JSON);
		110	json->max_depth = 512;
		111	}
81		112
82	/////////////////////////////////////////////////////////////////////////////	113	/////////////////////////////////////////////////////////////////////////////
83	// utility functions	114	// utility functions
84		115
85	inline void	116	INLINE SV *
		117	get_bool (const char *name)
		118	{
		119	SV *sv = get_sv (name, 1);
		120
		121	SvREADONLY_on (sv);
		122	SvREADONLY_on (SvRV (sv));
		123
		124	return sv;
		125	}
		126
		127	INLINE void
86	shrink (SV *sv)	128	shrink (SV *sv)
87	{	129	{
88	sv_utf8_downgrade (sv, 1);	130	sv_utf8_downgrade (sv, 1);
		131
89	if (SvLEN (sv) > SvCUR (sv) + 1)	132	if (SvLEN (sv) > SvCUR (sv) + 1)
90	{	133	{
91	#ifdef SvPV_shrink_to_cur	134	#ifdef SvPV_shrink_to_cur
92	SvPV_shrink_to_cur (sv);	135	SvPV_shrink_to_cur (sv);
93	#elif defined (SvPV_renew)	136	#elif defined (SvPV_renew)
…		…
99	// decode an utf-8 character and return it, or (UV)-1 in	142	// decode an utf-8 character and return it, or (UV)-1 in
100	// case of an error.	143	// case of an error.
101	// we special-case "safe" characters from U+80 .. U+7FF,	144	// we special-case "safe" characters from U+80 .. U+7FF,
102	// but use the very good perl function to parse anything else.	145	// but use the very good perl function to parse anything else.
103	// note that we never call this function for a ascii codepoints	146	// note that we never call this function for a ascii codepoints
104	inline UV	147	INLINE UV
105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	148	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
106	{	149	{
107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	150	if (expect_true (len >= 2
108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	151	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	152	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
110	{	153	{
111	*clen = 2;	154	*clen = 2;
112	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	155	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
113	}	156	}
114	else	157	else
115	{	158	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
116	*clen = (STRLEN)-1;	159	}
117	return (UV)-1;	160
		161	// likewise for encoding, also never called for ascii codepoints
		162	// this function takes advantage of this fact, although current gccs
		163	// seem to optimise the check for >= 0x80 away anyways
		164	INLINE unsigned char *
		165	encode_utf8 (unsigned char *s, UV ch)
		166	{
		167	if (expect_false (ch < 0x000080))
		168	*s++ = ch;
		169	else if (expect_true (ch < 0x000800))
		170	*s++ = 0xc0 \| ( ch >> 6),
		171	*s++ = 0x80 \| ( ch & 0x3f);
		172	else if ( ch < 0x010000)
		173	*s++ = 0xe0 \| ( ch >> 12),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176	else if ( ch < 0x110000)
		177	*s++ = 0xf0 \| ( ch >> 18),
		178	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		179	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		180	*s++ = 0x80 \| ( ch & 0x3f);
		181
		182	return s;
		183	}
		184
		185	// convert offset pointer to character index, sv must be string
		186	static STRLEN
		187	ptr_to_index (SV sv, char offset)
		188	{
		189	return SvUTF8 (sv)
		190	? utf8_distance (offset, SvPVX (sv))
		191	: offset - SvPVX (sv);
		192	}
		193
		194	/////////////////////////////////////////////////////////////////////////////
		195	// fp hell
		196
		197	// scan a group of digits, and a trailing exponent
		198	static void
		199	json_atof_scan1 (const char s, NV accum, int *expo, int postdp, int maxdepth)
		200	{
		201	UV uaccum = 0;
		202	int eaccum = 0;
		203
		204	// if we recurse too deep, skip all remaining digits
		205	// to avoid a stack overflow attack
		206	if (expect_false (--maxdepth <= 0))
		207	while (((U8)*s - '0') < 10)
		208	++s;
		209
		210	for (;;)
118	}	211	{
119	}	212	U8 dig = (U8)*s - '0';
120		213
		214	if (expect_false (dig >= 10))
		215	{
		216	if (dig == (U8)((U8)'.' - (U8)'0'))
		217	{
		218	++s;
		219	json_atof_scan1 (s, accum, expo, 1, maxdepth);
		220	}
		221	else if ((dig \| ' ') == 'e' - '0')
		222	{
		223	int exp2 = 0;
		224	int neg = 0;
		225
		226	++s;
		227
		228	if (*s == '-')
		229	{
		230	++s;
		231	neg = 1;
		232	}
		233	else if (*s == '+')
		234	++s;
		235
		236	while ((dig = (U8)*s - '0') < 10)
		237	exp2 = exp2 * 10 + *s++ - '0';
		238
		239	*expo += neg ? -exp2 : exp2;
		240	}
		241
		242	break;
		243	}
		244
		245	++s;
		246
		247	uaccum = uaccum * 10 + dig;
		248	++eaccum;
		249
		250	// if we have too many digits, then recurse for more
		251	// we actually do this for rather few digits
		252	if (uaccum >= (UV_MAX - 9) / 10)
		253	{
		254	if (postdp) *expo -= eaccum;
		255	json_atof_scan1 (s, accum, expo, postdp, maxdepth);
		256	if (postdp) *expo += eaccum;
		257
		258	break;
		259	}
		260	}
		261
		262	// this relies greatly on the quality of the pow ()
		263	// implementation of the platform, but a good
		264	// implementation is hard to beat.
		265	// (IEEE 754 conformant ones are required to be exact)
		266	if (postdp) *expo -= eaccum;
		267	accum += uaccum Perl_pow (10., *expo);
		268	*expo += eaccum;
		269	}
		270
		271	static NV
		272	json_atof (const char *s)
		273	{
		274	NV accum = 0.;
		275	int expo = 0;
		276	int neg = 0;
		277
		278	if (*s == '-')
		279	{
		280	++s;
		281	neg = 1;
		282	}
		283
		284	// a recursion depth of ten gives us >>500 bits
		285	json_atof_scan1 (s, &accum, &expo, 0, 10);
		286
		287	return neg ? -accum : accum;
		288	}
121	/////////////////////////////////////////////////////////////////////////////	289	/////////////////////////////////////////////////////////////////////////////
122	// encoder	290	// encoder
123		291
124	// structure used for encoding JSON	292	// structure used for encoding JSON
125	typedef struct	293	typedef struct
…		…
127	char *cur; // SvPVX (sv) + current output position	295	char *cur; // SvPVX (sv) + current output position
128	char *end; // SvEND (sv)	296	char *end; // SvEND (sv)
129	SV *sv; // result scalar	297	SV *sv; // result scalar
130	JSON json;	298	JSON json;
131	U32 indent; // indentation level	299	U32 indent; // indentation level
132	U32 maxdepth; // max. indentation/recursion level	300	UV limit; // escape character values >= this value when encoding
133	} enc_t;	301	} enc_t;
134		302
135	inline void	303	INLINE void
136	need (enc_t *enc, STRLEN len)	304	need (enc_t *enc, STRLEN len)
137	{	305	{
138	if (expect_false (enc->cur + len >= enc->end))	306	if (expect_false (enc->cur + len >= enc->end))
139	{	307	{
140	STRLEN cur = enc->cur - SvPVX (enc->sv);	308	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
141	SvGROW (enc->sv, cur + len + 1);	309	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
142	enc->cur = SvPVX (enc->sv) + cur;	310	enc->cur = SvPVX (enc->sv) + cur;
143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	311	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
144	}	312	}
145	}	313	}
146		314
147	inline void	315	INLINE void
148	encode_ch (enc_t *enc, char ch)	316	encode_ch (enc_t *enc, char ch)
149	{	317	{
150	need (enc, 1);	318	need (enc, 1);
151	*enc->cur++ = ch;	319	*enc->cur++ = ch;
152	}	320	}
…		…
206	{	374	{
207	uch = ch;	375	uch = ch;
208	clen = 1;	376	clen = 1;
209	}	377	}
210		378
211	if (uch > 0x10FFFFUL)	379	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
213
214	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
215	{	380	{
216	if (uch > 0xFFFFUL)	381	if (uch >= 0x10000UL)
217	{	382	{
		383	if (uch >= 0x110000UL)
		384	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		385
218	need (enc, len += 11);	386	need (enc, len += 11);
219	sprintf (enc->cur, "\\u%04x\\u%04x",	387	sprintf (enc->cur, "\\u%04x\\u%04x",
220	(int)((uch - 0x10000) / 0x400 + 0xD800),	388	(int)((uch - 0x10000) / 0x400 + 0xD800),
221	(int)((uch - 0x10000) % 0x400 + 0xDC00));	389	(int)((uch - 0x10000) % 0x400 + 0xDC00));
222	enc->cur += 12;	390	enc->cur += 12;
223	}	391	}
224	else	392	else
225	{	393	{
226	static char hexdigit [16] = "0123456789abcdef";
227	need (enc, len += 5);	394	need (enc, len += 5);
228	*enc->cur++ = '\\';	395	*enc->cur++ = '\\';
229	*enc->cur++ = 'u';	396	*enc->cur++ = 'u';
230	*enc->cur++ = hexdigit [ uch >> 12 ];	397	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
231	*enc->cur++ = hexdigit [(uch >> 8) & 15];	398	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
232	*enc->cur++ = hexdigit [(uch >> 4) & 15];	399	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
233	*enc->cur++ = hexdigit [(uch >> 0) & 15];	400	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
234	}	401	}
235		402
236	str += clen;	403	str += clen;
237	}	404	}
238	else if (enc->json.flags & F_LATIN1)	405	else if (enc->json.flags & F_LATIN1)
…		…
250	while (--clen);	417	while (--clen);
251	}	418	}
252	else	419	else
253	{	420	{
254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	421	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	422	enc->cur = encode_utf8 (enc->cur, uch);
256	++str;	423	++str;
257	}	424	}
258	}	425	}
259	}	426	}
260	}	427	}
261		428
262	--len;	429	--len;
263	}	430	}
264	}	431	}
265		432
266	inline void	433	INLINE void
267	encode_indent (enc_t *enc)	434	encode_indent (enc_t *enc)
268	{	435	{
269	if (enc->json.flags & F_INDENT)	436	if (enc->json.flags & F_INDENT)
270	{	437	{
271	int spaces = enc->indent * INDENT_STEP;	438	int spaces = enc->indent * INDENT_STEP;
…		…
274	memset (enc->cur, ' ', spaces);	441	memset (enc->cur, ' ', spaces);
275	enc->cur += spaces;	442	enc->cur += spaces;
276	}	443	}
277	}	444	}
278		445
279	inline void	446	INLINE void
280	encode_space (enc_t *enc)	447	encode_space (enc_t *enc)
281	{	448	{
282	need (enc, 1);	449	need (enc, 1);
283	encode_ch (enc, ' ');	450	encode_ch (enc, ' ');
284	}	451	}
285		452
286	inline void	453	INLINE void
287	encode_nl (enc_t *enc)	454	encode_nl (enc_t *enc)
288	{	455	{
289	if (enc->json.flags & F_INDENT)	456	if (enc->json.flags & F_INDENT)
290	{	457	{
291	need (enc, 1);	458	need (enc, 1);
292	encode_ch (enc, '\n');	459	encode_ch (enc, '\n');
293	}	460	}
294	}	461	}
295		462
296	inline void	463	INLINE void
297	encode_comma (enc_t *enc)	464	encode_comma (enc_t *enc)
298	{	465	{
299	encode_ch (enc, ',');	466	encode_ch (enc, ',');
300		467
301	if (enc->json.flags & F_INDENT)	468	if (enc->json.flags & F_INDENT)
…		…
309	static void	476	static void
310	encode_av (enc_t enc, AV av)	477	encode_av (enc_t enc, AV av)
311	{	478	{
312	int i, len = av_len (av);	479	int i, len = av_len (av);
313		480
314	if (enc->indent >= enc->maxdepth)	481	if (enc->indent >= enc->json.max_depth)
315	croak ("data structure too deep (hit recursion limit)");	482	croak (ERR_NESTING_EXCEEDED);
316		483
317	encode_ch (enc, '[');	484	encode_ch (enc, '[');
318		485
319	if (len >= 0)	486	if (len >= 0)
320	{	487	{
…		…
394		561
395	static void	562	static void
396	encode_hv (enc_t enc, HV hv)	563	encode_hv (enc_t enc, HV hv)
397	{	564	{
398	HE *he;	565	HE *he;
399	int count;
400		566
401	if (enc->indent >= enc->maxdepth)	567	if (enc->indent >= enc->json.max_depth)
402	croak ("data structure too deep (hit recursion limit)");	568	croak (ERR_NESTING_EXCEEDED);
403		569
404	encode_ch (enc, '{');	570	encode_ch (enc, '{');
405		571
406	// for canonical output we have to sort by keys first	572	// for canonical output we have to sort by keys first
407	// actually, this is mostly due to the stupid so-called	573	// actually, this is mostly due to the stupid so-called
408	// security workaround added somewhere in 5.8.x.	574	// security workaround added somewhere in 5.8.x
409	// that randomises hash orderings	575	// that randomises hash orderings
410	if (enc->json.flags & F_CANONICAL)	576	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
411	{	577	{
412	int count = hv_iterinit (hv);	578	int count = hv_iterinit (hv);
413		579
414	if (SvMAGICAL (hv))	580	if (SvMAGICAL (hv))
415	{	581	{
…		…
588		754
589	if (len == 1 && *pv == '1')	755	if (len == 1 && *pv == '1')
590	encode_str (enc, "true", 4, 0);	756	encode_str (enc, "true", 4, 0);
591	else if (len == 1 && *pv == '0')	757	else if (len == 1 && *pv == '0')
592	encode_str (enc, "false", 5, 0);	758	encode_str (enc, "false", 5, 0);
		759	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		760	encode_str (enc, "null", 4, 0);
593	else	761	else
594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	762	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	763	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
596	}	764	}
		765	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		766	encode_str (enc, "null", 4, 0);
597	else	767	else
598	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	768	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	769	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
600	}	770	}
601		771
…		…
619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	789	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
620	enc->cur += strlen (enc->cur);	790	enc->cur += strlen (enc->cur);
621	}	791	}
622	else if (SvIOKp (sv))	792	else if (SvIOKp (sv))
623	{	793	{
624	// we assume we can always read an IV as a UV	794	// we assume we can always read an IV as a UV and vice versa
625	if (SvUV (sv) & ~(UV)0x7fff)	795	// we assume two's complement
626	{	796	// we assume no aliasing issues in the union
627	// large integer, use the (rather slow) snprintf way.	797	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
628	need (enc, sizeof (UV) * 3);	798	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
629	enc->cur +=
630	SvIsUV(sv)
631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
633	}
634	else
635	{	799	{
636	// optimise the "small number case"	800	// optimise the "small number case"
637	// code will likely be branchless and use only a single multiplication	801	// code will likely be branchless and use only a single multiplication
		802	// works for numbers up to 59074
638	I32 i = SvIV (sv);	803	I32 i = SvIVX (sv);
639	U32 u;	804	U32 u;
640	char digit, nz = 0;	805	char digit, nz = 0;
641		806
642	need (enc, 6);	807	need (enc, 6);
643		808
…		…
649		814
650	// now output digit by digit, each time masking out the integer part	815	// now output digit by digit, each time masking out the integer part
651	// and multiplying by 5 while moving the decimal point one to the right,	816	// and multiplying by 5 while moving the decimal point one to the right,
652	// resulting in a net multiplication by 10.	817	// resulting in a net multiplication by 10.
653	// we always write the digit to memory but conditionally increment	818	// we always write the digit to memory but conditionally increment
654	// the pointer, to ease the usage of conditional move instructions.	819	// the pointer, to enable the use of conditional move instructions.
655	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	820	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
656	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	821	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
657	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	822	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
658	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	823	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	824	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
660	}	825	}
		826	else
		827	{
		828	// large integer, use the (rather slow) snprintf way.
		829	need (enc, IVUV_MAXCHARS);
		830	enc->cur +=
		831	SvIsUV(sv)
		832	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		833	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		834	}
661	}	835	}
662	else if (SvROK (sv))	836	else if (SvROK (sv))
663	encode_rv (enc, SvRV (sv));	837	encode_rv (enc, SvRV (sv));
664	else if (!SvOK (sv))	838	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
665	encode_str (enc, "null", 4, 0);	839	encode_str (enc, "null", 4, 0);
666	else	840	else
667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	841	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
668	SvPV_nolen (sv), SvFLAGS (sv));	842	SvPV_nolen (sv), SvFLAGS (sv));
669	}	843	}
…		…
679	enc.json = *json;	853	enc.json = *json;
680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	854	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
681	enc.cur = SvPVX (enc.sv);	855	enc.cur = SvPVX (enc.sv);
682	enc.end = SvEND (enc.sv);	856	enc.end = SvEND (enc.sv);
683	enc.indent = 0;	857	enc.indent = 0;
684	enc.maxdepth = DEC_DEPTH (enc.json.flags);	858	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		859	: enc.json.flags & F_LATIN1 ? 0x000100UL
		860	: 0x110000UL;
685		861
686	SvPOK_only (enc.sv);	862	SvPOK_only (enc.sv);
687	encode_sv (&enc, scalar);	863	encode_sv (&enc, scalar);
		864	encode_nl (&enc);
688		865
689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	866	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	867	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
691		868
692	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	869	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
710	JSON json;	887	JSON json;
711	U32 depth; // recursion depth	888	U32 depth; // recursion depth
712	U32 maxdepth; // recursion depth limit	889	U32 maxdepth; // recursion depth limit
713	} dec_t;	890	} dec_t;
714		891
715	inline void	892	INLINE void
716	decode_comment (dec_t *dec)	893	decode_comment (dec_t *dec)
717	{	894	{
718	// only '#'-style comments allowed a.t.m.	895	// only '#'-style comments allowed a.t.m.
719		896
720	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)	897	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
721	++dec->cur;	898	++dec->cur;
722	}	899	}
723		900
724	inline void	901	INLINE void
725	decode_ws (dec_t *dec)	902	decode_ws (dec_t *dec)
726	{	903	{
727	for (;;)	904	for (;;)
728	{	905	{
729	char ch = *dec->cur;	906	char ch = *dec->cur;
…		…
753	if (*dec->cur != ch) \	930	if (*dec->cur != ch) \
754	ERR (# ch " expected"); \	931	ERR (# ch " expected"); \
755	++dec->cur; \	932	++dec->cur; \
756	SE	933	SE
757		934
758	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	935	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
759	#define DEC_DEC_DEPTH --dec->depth	936	#define DEC_DEC_DEPTH --dec->depth
760		937
761	static SV decode_sv (dec_t dec);	938	static SV decode_sv (dec_t dec);
762		939
763	static signed char decode_hexdigit[256];	940	static signed char decode_hexdigit[256];
…		…
855		1032
856	if (hi >= 0x80)	1033	if (hi >= 0x80)
857	{	1034	{
858	utf8 = 1;	1035	utf8 = 1;
859		1036
860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	1037	cur = encode_utf8 (cur, hi);
861	}	1038	}
862	else	1039	else
863	*cur++ = hi;	1040	*cur++ = hi;
864	}	1041	}
865	break;	1042	break;
…		…
867	default:	1044	default:
868	--dec_cur;	1045	--dec_cur;
869	ERR ("illegal backslash escape sequence in string");	1046	ERR ("illegal backslash escape sequence in string");
870	}	1047	}
871	}	1048	}
872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	1049	else if (expect_true (ch >= 0x20 && ch < 0x80))
873	*cur++ = ch;	1050	*cur++ = ch;
874	else if (ch >= 0x80)	1051	else if (ch >= 0x80)
875	{	1052	{
876	STRLEN clen;	1053	STRLEN clen;
877	UV uch;
878		1054
879	--dec_cur;	1055	--dec_cur;
880		1056
881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	1057	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
882	if (clen == (STRLEN)-1)	1058	if (clen == (STRLEN)-1)
883	ERR ("malformed UTF-8 character in JSON string");	1059	ERR ("malformed UTF-8 character in JSON string");
884		1060
885	do	1061	do
886	cur++ = dec_cur++;	1062	cur++ = dec_cur++;
…		…
903	{	1079	{
904	STRLEN len = cur - buf;	1080	STRLEN len = cur - buf;
905		1081
906	if (sv)	1082	if (sv)
907	{	1083	{
908	SvGROW (sv, SvCUR (sv) + len + 1);	1084	STRLEN cur = SvCUR (sv);
		1085
		1086	if (SvLEN (sv) <= cur + len)
		1087	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1088
909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	1089	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
910	SvCUR_set (sv, SvCUR (sv) + len);	1090	SvCUR_set (sv, SvCUR (sv) + len);
911	}	1091	}
912	else	1092	else
913	sv = newSVpvn (buf, len);	1093	sv = newSVpvn (buf, len);
…		…
1000		1180
1001	if (!is_nv)	1181	if (!is_nv)
1002	{	1182	{
1003	int len = dec->cur - start;	1183	int len = dec->cur - start;
1004		1184
1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1185	// special case the rather common 1..5-digit-int case
1006	if (*start == '-')	1186	if (*start == '-')
1007	switch (len)	1187	switch (len)
1008	{	1188	{
1009	case 2: return newSViv (-( start [1] - '0' * 1));	1189	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1190	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1011	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1191	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1012	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1192	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1193	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1013	}	1194	}
1014	else	1195	else
1015	switch (len)	1196	switch (len)
1016	{	1197	{
1017	case 1: return newSViv ( start [0] - '0' * 1);	1198	case 1: return newSViv ( start [0] - '0' * 1);
1018	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1199	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1019	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1200	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1020	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1201	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1202	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1021	}	1203	}
1022		1204
1023	{	1205	{
1024	UV uv;	1206	UV uv;
1025	int numtype = grok_number (start, len, &uv);	1207	int numtype = grok_number (start, len, &uv);
…		…
1034	}	1216	}
1035		1217
1036	len -= *start == '-' ? 1 : 0;	1218	len -= *start == '-' ? 1 : 0;
1037		1219
1038	// does not fit into IV or UV, try NV	1220	// does not fit into IV or UV, try NV
1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)	1221	if (len <= NV_DIG)
1040	#if defined (LDBL_DIG)
1041	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1042	#endif
1043	)
1044	// fits into NV without loss of precision	1222	// fits into NV without loss of precision
1045	return newSVnv (Atof (start));	1223	return newSVnv (json_atof (start));
1046		1224
1047	// everything else fails, convert it to a string	1225	// everything else fails, convert it to a string
1048	return newSVpvn (start, dec->cur - start);	1226	return newSVpvn (start, dec->cur - start);
1049	}	1227	}
1050		1228
1051	// loss of precision here	1229	// loss of precision here
1052	return newSVnv (Atof (start));	1230	return newSVnv (json_atof (start));
1053		1231
1054	fail:	1232	fail:
1055	return 0;	1233	return 0;
1056	}	1234	}
1057		1235
…		…
1133	char *p = dec->cur;	1311	char *p = dec->cur;
1134	char *e = p + 24; // only try up to 24 bytes	1312	char *e = p + 24; // only try up to 24 bytes
1135		1313
1136	for (;;)	1314	for (;;)
1137	{	1315	{
1138	// the >= 0x80 is true on most architectures	1316	// the >= 0x80 is false on most architectures
1139	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1317	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1140	{	1318	{
1141	// slow path, back up and use decode_str	1319	// slow path, back up and use decode_str
1142	SV *key = decode_str (dec);	1320	SV *key = decode_str (dec);
1143	if (!key)	1321	if (!key)
…		…
1226	dSP;	1404	dSP;
1227	int count;	1405	int count;
1228		1406
1229	ENTER; SAVETMPS; PUSHMARK (SP);	1407	ENTER; SAVETMPS; PUSHMARK (SP);
1230	XPUSHs (HeVAL (he));	1408	XPUSHs (HeVAL (he));
		1409	sv_2mortal (sv);
1231		1410
1232	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1411	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1233		1412
1234	if (count == 1)	1413	if (count == 1)
1235	{	1414	{
1236	sv = newSVsv (POPs);	1415	sv = newSVsv (POPs);
1237	FREETMPS; LEAVE;	1416	FREETMPS; LEAVE;
1238	return sv;	1417	return sv;
1239	}	1418	}
1240		1419
		1420	SvREFCNT_inc (sv);
1241	FREETMPS; LEAVE;	1421	FREETMPS; LEAVE;
1242	}	1422	}
1243	}	1423	}
1244		1424
1245	if (dec->json.cb_object)	1425	if (dec->json.cb_object)
…		…
1274		1454
1275	static SV *	1455	static SV *
1276	decode_sv (dec_t *dec)	1456	decode_sv (dec_t *dec)
1277	{	1457	{
1278	// the beauty of JSON: you need exactly one character lookahead	1458	// the beauty of JSON: you need exactly one character lookahead
1279	// to parse anything.	1459	// to parse everything.
1280	switch (*dec->cur)	1460	switch (*dec->cur)
1281	{	1461	{
1282	case '"': ++dec->cur; return decode_str (dec);	1462	case '"': ++dec->cur; return decode_str (dec);
1283	case '[': ++dec->cur; return decode_av (dec);	1463	case '[': ++dec->cur; return decode_av (dec);
1284	case '{': ++dec->cur; return decode_hv (dec);	1464	case '{': ++dec->cur; return decode_hv (dec);
1285		1465
1286	case '-':	1466	case '-':
1287	case '0': case '1': case '2': case '3': case '4':	1467	case '0': case '1': case '2': case '3': case '4':
1288	case '5': case '6': case '7': case '8': case '9':	1468	case '5': case '6': case '7': case '8': case '9':
1289	return decode_num (dec);	1469	return decode_num (dec);
…		…
1291	case 't':	1471	case 't':
1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1472	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1293	{	1473	{
1294	dec->cur += 4;	1474	dec->cur += 4;
1295	#if JSON_SLOW	1475	#if JSON_SLOW
1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1476	json_true = get_bool ("JSON::XS::true");
1297	#endif	1477	#endif
1298	return SvREFCNT_inc (json_true);	1478	return newSVsv (json_true);
1299	}	1479	}
1300	else	1480	else
1301	ERR ("'true' expected");	1481	ERR ("'true' expected");
1302		1482
1303	break;	1483	break;
…		…
1305	case 'f':	1485	case 'f':
1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1486	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1307	{	1487	{
1308	dec->cur += 5;	1488	dec->cur += 5;
1309	#if JSON_SLOW	1489	#if JSON_SLOW
1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1490	json_false = get_bool ("JSON::XS::false");
1311	#endif	1491	#endif
1312	return SvREFCNT_inc (json_false);	1492	return newSVsv (json_false);
1313	}	1493	}
1314	else	1494	else
1315	ERR ("'false' expected");	1495	ERR ("'false' expected");
1316		1496
1317	break;	1497	break;
…		…
1335	fail:	1515	fail:
1336	return 0;	1516	return 0;
1337	}	1517	}
1338		1518
1339	static SV *	1519	static SV *
1340	decode_json (SV string, JSON json, UV *offset_return)	1520	decode_json (SV string, JSON json, char **offset_return)
1341	{	1521	{
1342	dec_t dec;	1522	dec_t dec;
1343	UV offset;
1344	SV *sv;	1523	SV *sv;
1345		1524
		1525	/* work around bugs in 5.10 where manipulating magic values
		1526	* will perl ignore the magic in subsequent accesses.
		1527	* also make a copy of non-PV values, to get them into a clean
		1528	* state (SvPV should do that, but it's buggy, see below).
		1529	*/
1346	SvGETMAGIC (string);	1530	/SvGETMAGIC (string);/
		1531	if (SvMAGICAL (string) \|\| !SvPOK (string))
		1532	string = sv_2mortal (newSVsv (string));
		1533
1347	SvUPGRADE (string, SVt_PV);	1534	SvUPGRADE (string, SVt_PV);
1348		1535
1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1536	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1537	* assertion with -DDEBUGGING, although SvCUR is documented to
		1538	* return the xpv_cur field which certainly exists after upgrading.
		1539	* according to nicholas clark, calling SvPOK fixes this.
		1540	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1541	* and hope for the best.
		1542	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1543	* assertion business is seriously broken, try yet another workaround
		1544	* for the broken -DDEBUGGING.
		1545	*/
		1546	{
		1547	#ifdef DEBUGGING
		1548	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1549	#else
		1550	STRLEN offset = SvCUR (string);
		1551	#endif
		1552
		1553	if (offset > json->max_size && json->max_size)
1350	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1554	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1351	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1555	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1556	}
1352		1557
1353	if (json->flags & F_UTF8)	1558	if (DECODE_WANTS_OCTETS (json))
1354	sv_utf8_downgrade (string, 0);	1559	sv_utf8_downgrade (string, 0);
1355	else	1560	else
1356	sv_utf8_upgrade (string);	1561	sv_utf8_upgrade (string);
1357		1562
1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1563	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1359		1564
1360	dec.json = *json;	1565	dec.json = *json;
1361	dec.cur = SvPVX (string);	1566	dec.cur = SvPVX (string);
1362	dec.end = SvEND (string);	1567	dec.end = SvEND (string);
1363	dec.err = 0;	1568	dec.err = 0;
1364	dec.depth = 0;	1569	dec.depth = 0;
1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1366		1570
1367	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1571	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1368	dec.json.flags \|= F_HOOK;	1572	dec.json.flags \|= F_HOOK;
1369		1573
1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1574	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1371		1575
1372	decode_ws (&dec);	1576	decode_ws (&dec);
1373	sv = decode_sv (&dec);	1577	sv = decode_sv (&dec);
		1578
		1579	if (offset_return)
		1580	*offset_return = dec.cur;
1374		1581
1375	if (!(offset_return \|\| !sv))	1582	if (!(offset_return \|\| !sv))
1376	{	1583	{
1377	// check for trailing garbage	1584	// check for trailing garbage
1378	decode_ws (&dec);	1585	decode_ws (&dec);
…		…
1381	{	1588	{
1382	dec.err = "garbage after JSON object";	1589	dec.err = "garbage after JSON object";
1383	SvREFCNT_dec (sv);	1590	SvREFCNT_dec (sv);
1384	sv = 0;	1591	sv = 0;
1385	}	1592	}
1386	}
1387
1388	if (offset_return \|\| !sv)
1389	{
1390	offset = dec.json.flags & F_UTF8
1391	? dec.cur - SvPVX (string)
1392	: utf8_distance (dec.cur, SvPVX (string));
1393
1394	if (offset_return)
1395	*offset_return = offset;
1396	}	1593	}
1397		1594
1398	if (!sv)	1595	if (!sv)
1399	{	1596	{
1400	SV *uni = sv_newmortal ();	1597	SV *uni = sv_newmortal ();
…		…
1406	SAVEVPTR (PL_curcop);	1603	SAVEVPTR (PL_curcop);
1407	PL_curcop = &cop;	1604	PL_curcop = &cop;
1408	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1605	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1409	LEAVE;	1606	LEAVE;
1410		1607
1411	croak ("%s, at character offset %d [\"%s\"]",	1608	croak ("%s, at character offset %d (before \"%s\")",
1412	dec.err,	1609	dec.err,
1413	(int)offset,	1610	ptr_to_index (string, dec.cur),
1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1611	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1415	}	1612	}
1416		1613
1417	sv = sv_2mortal (sv);	1614	sv = sv_2mortal (sv);
1418		1615
1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1616	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1617	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1421		1618
1422	return sv;	1619	return sv;
		1620	}
		1621
		1622	/////////////////////////////////////////////////////////////////////////////
		1623	// incremental parser
		1624
		1625	static void
		1626	incr_parse (JSON *self)
		1627	{
		1628	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1629
		1630	// the state machine here is a bit convoluted and could be simplified a lot
		1631	// but this would make it slower, so...
		1632
		1633	for (;;)
		1634	{
		1635	//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
		1636	switch (self->incr_mode)
		1637	{
		1638	// only used for initial whitespace skipping
		1639	case INCR_M_WS:
		1640	for (;;)
		1641	{
		1642	if (*p > 0x20)
		1643	{
		1644	if (*p == '#')
		1645	{
		1646	self->incr_mode = INCR_M_C0;
		1647	goto incr_m_c;
		1648	}
		1649	else
		1650	{
		1651	self->incr_mode = INCR_M_JSON;
		1652	goto incr_m_json;
		1653	}
		1654	}
		1655	else if (!*p)
		1656	goto interrupt;
		1657
		1658	++p;
		1659	}
		1660
		1661	// skip a single char inside a string (for \\-processing)
		1662	case INCR_M_BS:
		1663	if (!*p)
		1664	goto interrupt;
		1665
		1666	++p;
		1667	self->incr_mode = INCR_M_STR;
		1668	goto incr_m_str;
		1669
		1670	// inside #-style comments
		1671	case INCR_M_C0:
		1672	case INCR_M_C1:
		1673	incr_m_c:
		1674	for (;;)
		1675	{
		1676	if (*p == '\n')
		1677	{
		1678	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1679	break;
		1680	}
		1681	else if (!*p)
		1682	goto interrupt;
		1683
		1684	++p;
		1685	}
		1686
		1687	break;
		1688
		1689	// inside a string
		1690	case INCR_M_STR:
		1691	incr_m_str:
		1692	for (;;)
		1693	{
		1694	if (*p == '"')
		1695	{
		1696	++p;
		1697	self->incr_mode = INCR_M_JSON;
		1698
		1699	if (!self->incr_nest)
		1700	goto interrupt;
		1701
		1702	goto incr_m_json;
		1703	}
		1704	else if (*p == '\\')
		1705	{
		1706	++p; // "virtually" consumes character after \
		1707
		1708	if (!*p) // if at end of string we have to switch modes
		1709	{
		1710	self->incr_mode = INCR_M_BS;
		1711	goto interrupt;
		1712	}
		1713	}
		1714	else if (!*p)
		1715	goto interrupt;
		1716
		1717	++p;
		1718	}
		1719
		1720	// after initial ws, outside string
		1721	case INCR_M_JSON:
		1722	incr_m_json:
		1723	for (;;)
		1724	{
		1725	switch (*p++)
		1726	{
		1727	case 0:
		1728	--p;
		1729	goto interrupt;
		1730
		1731	case 0x09:
		1732	case 0x0a:
		1733	case 0x0d:
		1734	case 0x20:
		1735	if (!self->incr_nest)
		1736	{
		1737	--p; // do not eat the whitespace, let the next round do it
		1738	goto interrupt;
		1739	}
		1740	break;
		1741
		1742	case '"':
		1743	self->incr_mode = INCR_M_STR;
		1744	goto incr_m_str;
		1745
		1746	case '[':
		1747	case '{':
		1748	if (++self->incr_nest > self->max_depth)
		1749	croak (ERR_NESTING_EXCEEDED);
		1750	break;
		1751
		1752	case ']':
		1753	case '}':
		1754	if (--self->incr_nest <= 0)
		1755	goto interrupt;
		1756	break;
		1757
		1758	case '#':
		1759	self->incr_mode = INCR_M_C1;
		1760	goto incr_m_c;
		1761	}
		1762	}
		1763	}
		1764
		1765	modechange:
		1766	;
		1767	}
		1768
		1769	interrupt:
		1770	self->incr_pos = p - SvPVX (self->incr_text);
		1771	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1772	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1423	}	1773	}
1424		1774
1425	/////////////////////////////////////////////////////////////////////////////	1775	/////////////////////////////////////////////////////////////////////////////
1426	// XS interface functions	1776	// XS interface functions
1427		1777
…		…
1439	: -1;	1789	: -1;
1440		1790
1441	json_stash = gv_stashpv ("JSON::XS" , 1);	1791	json_stash = gv_stashpv ("JSON::XS" , 1);
1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1792	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1443		1793
1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1794	json_true = get_bool ("JSON::XS::true");
1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1795	json_false = get_bool ("JSON::XS::false");
		1796
		1797	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1446	}	1798	}
1447		1799
1448	PROTOTYPES: DISABLE	1800	PROTOTYPES: DISABLE
1449		1801
1450	void CLONE (...)	1802	void CLONE (...)
…		…
1455	void new (char *klass)	1807	void new (char *klass)
1456	PPCODE:	1808	PPCODE:
1457	{	1809	{
1458	SV *pv = NEWSV (0, sizeof (JSON));	1810	SV *pv = NEWSV (0, sizeof (JSON));
1459	SvPOK_only (pv);	1811	SvPOK_only (pv);
1460	Zero (SvPVX (pv), 1, JSON);	1812	json_init ((JSON *)SvPVX (pv));
1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1813	XPUSHs (sv_2mortal (sv_bless (
1462	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1814	newRV_noinc (pv),
		1815	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1816	)));
1463	}	1817	}
1464		1818
1465	void ascii (JSON *self, int enable = 1)	1819	void ascii (JSON *self, int enable = 1)
1466	ALIAS:	1820	ALIAS:
1467	ascii = F_ASCII	1821	ascii = F_ASCII
…		…
1475	allow_nonref = F_ALLOW_NONREF	1829	allow_nonref = F_ALLOW_NONREF
1476	shrink = F_SHRINK	1830	shrink = F_SHRINK
1477	allow_blessed = F_ALLOW_BLESSED	1831	allow_blessed = F_ALLOW_BLESSED
1478	convert_blessed = F_CONV_BLESSED	1832	convert_blessed = F_CONV_BLESSED
1479	relaxed = F_RELAXED	1833	relaxed = F_RELAXED
		1834	allow_unknown = F_ALLOW_UNKNOWN
1480	PPCODE:	1835	PPCODE:
1481	{	1836	{
1482	if (enable)	1837	if (enable)
1483	self->flags \|= ix;	1838	self->flags \|= ix;
1484	else	1839	else
1485	self->flags &= ~ix;	1840	self->flags &= ~ix;
1486		1841
1487	XPUSHs (ST (0));	1842	XPUSHs (ST (0));
1488	}	1843	}
1489		1844
1490	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1845	void get_ascii (JSON *self)
		1846	ALIAS:
		1847	get_ascii = F_ASCII
		1848	get_latin1 = F_LATIN1
		1849	get_utf8 = F_UTF8
		1850	get_indent = F_INDENT
		1851	get_canonical = F_CANONICAL
		1852	get_space_before = F_SPACE_BEFORE
		1853	get_space_after = F_SPACE_AFTER
		1854	get_allow_nonref = F_ALLOW_NONREF
		1855	get_shrink = F_SHRINK
		1856	get_allow_blessed = F_ALLOW_BLESSED
		1857	get_convert_blessed = F_CONV_BLESSED
		1858	get_relaxed = F_RELAXED
		1859	get_allow_unknown = F_ALLOW_UNKNOWN
1491	PPCODE:	1860	PPCODE:
1492	{	1861	XPUSHs (boolSV (self->flags & ix));
1493	UV log2 = 0;
1494		1862
1495	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1863	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1496		1864	PPCODE:
1497	while ((1UL << log2) < max_depth)	1865	self->max_depth = max_depth;
1498	++log2;
1499
1500	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1501
1502	XPUSHs (ST (0));	1866	XPUSHs (ST (0));
1503	}
1504		1867
		1868	U32 get_max_depth (JSON *self)
		1869	CODE:
		1870	RETVAL = self->max_depth;
		1871	OUTPUT:
		1872	RETVAL
		1873
1505	void max_size (JSON *self, UV max_size = 0)	1874	void max_size (JSON *self, U32 max_size = 0)
1506	PPCODE:	1875	PPCODE:
1507	{	1876	self->max_size = max_size;
1508	UV log2 = 0;
1509
1510	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1511	if (max_size == 1) max_size = 2;
1512
1513	while ((1UL << log2) < max_size)
1514	++log2;
1515
1516	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1517
1518	XPUSHs (ST (0));	1877	XPUSHs (ST (0));
1519	}	1878
		1879	int get_max_size (JSON *self)
		1880	CODE:
		1881	RETVAL = self->max_size;
		1882	OUTPUT:
		1883	RETVAL
1520		1884
1521	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1885	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1522	PPCODE:	1886	PPCODE:
1523	{	1887	{
1524	SvREFCNT_dec (self->cb_object);	1888	SvREFCNT_dec (self->cb_object);
…		…
1558	XPUSHs (decode_json (jsonstr, self, 0));	1922	XPUSHs (decode_json (jsonstr, self, 0));
1559		1923
1560	void decode_prefix (JSON self, SV jsonstr)	1924	void decode_prefix (JSON self, SV jsonstr)
1561	PPCODE:	1925	PPCODE:
1562	{	1926	{
1563	UV offset;	1927	char *offset;
1564	EXTEND (SP, 2);	1928	EXTEND (SP, 2);
1565	PUSHs (decode_json (jsonstr, self, &offset));	1929	PUSHs (decode_json (jsonstr, self, &offset));
1566	PUSHs (sv_2mortal (newSVuv (offset)));	1930	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1931	}
		1932
		1933	void incr_parse (JSON self, SV jsonstr = 0)
		1934	PPCODE:
		1935	{
		1936	if (!self->incr_text)
		1937	self->incr_text = newSVpvn ("", 0);
		1938
		1939	/* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
		1940	if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
		1941	if (DECODE_WANTS_OCTETS (self))
		1942	{
		1943	if (self->incr_pos)
		1944	self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
		1945	(U8 *)SvPVX (self->incr_text) + self->incr_pos);
		1946
		1947	sv_utf8_downgrade (self->incr_text, 0);
		1948	}
		1949	else
		1950	{
		1951	sv_utf8_upgrade (self->incr_text);
		1952
		1953	if (self->incr_pos)
		1954	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1955	- (U8 *)SvPVX (self->incr_text);
		1956	}
		1957
		1958	// append data, if any
		1959	if (jsonstr)
		1960	{
		1961	/* make sure both strings have same encoding */
		1962	if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
		1963	if (SvUTF8 (jsonstr))
		1964	sv_utf8_downgrade (jsonstr, 0);
		1965	else
		1966	sv_utf8_upgrade (jsonstr);
		1967
		1968	/* and then just blindly append */
		1969	{
		1970	STRLEN len;
		1971	const char *str = SvPV (jsonstr, len);
		1972	STRLEN cur = SvCUR (self->incr_text);
		1973
		1974	if (SvLEN (self->incr_text) <= cur + len)
		1975	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1976
		1977	Move (str, SvEND (self->incr_text), len, char);
		1978	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1979	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1980	}
		1981	}
		1982
		1983	if (GIMME_V != G_VOID)
		1984	do
		1985	{
		1986	char *offset;
		1987
		1988	if (!INCR_DONE (self))
		1989	{
		1990	incr_parse (self);
		1991
		1992	if (self->incr_pos > self->max_size && self->max_size)
		1993	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1994	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1995
		1996	if (!INCR_DONE (self))
		1997	break;
		1998	}
		1999
		2000	XPUSHs (decode_json (self->incr_text, self, &offset));
		2001
		2002	self->incr_pos -= offset - SvPVX (self->incr_text);
		2003	self->incr_nest = 0;
		2004	self->incr_mode = 0;
		2005
		2006	sv_chop (self->incr_text, offset);
		2007	}
		2008	while (GIMME_V == G_ARRAY);
		2009	}
		2010
		2011	SV incr_text (JSON self)
		2012	ATTRS: lvalue
		2013	CODE:
		2014	{
		2015	if (self->incr_pos)
		2016	croak ("incr_text can not be called when the incremental parser already started parsing");
		2017
		2018	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		2019	}
		2020	OUTPUT:
		2021	RETVAL
		2022
		2023	void incr_skip (JSON *self)
		2024	CODE:
		2025	{
		2026	if (self->incr_pos)
		2027	{
		2028	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		2029	self->incr_pos = 0;
		2030	self->incr_nest = 0;
		2031	self->incr_mode = 0;
		2032	}
		2033	}
		2034
		2035	void incr_reset (JSON *self)
		2036	CODE:
		2037	{
		2038	SvREFCNT_dec (self->incr_text);
		2039	self->incr_text = 0;
		2040	self->incr_pos = 0;
		2041	self->incr_nest = 0;
		2042	self->incr_mode = 0;
1567	}	2043	}
1568		2044
1569	void DESTROY (JSON *self)	2045	void DESTROY (JSON *self)
1570	CODE:	2046	CODE:
1571	SvREFCNT_dec (self->cb_sk_object);	2047	SvREFCNT_dec (self->cb_sk_object);
1572	SvREFCNT_dec (self->cb_object);	2048	SvREFCNT_dec (self->cb_object);
		2049	SvREFCNT_dec (self->incr_text);
1573		2050
1574	PROTOTYPES: ENABLE	2051	PROTOTYPES: ENABLE
1575		2052
1576	void to_json (SV *scalar)	2053	void encode_json (SV *scalar)
		2054	ALIAS:
		2055	to_json_ = 0
		2056	encode_json = F_UTF8
1577	PPCODE:	2057	PPCODE:
1578	{	2058	{
1579	JSON json = { F_DEFAULT \| F_UTF8 };	2059	JSON json;
		2060	json_init (&json);
		2061	json.flags \|= ix;
1580	XPUSHs (encode_json (scalar, &json));	2062	XPUSHs (encode_json (scalar, &json));
1581	}	2063	}
1582		2064
1583	void from_json (SV *jsonstr)	2065	void decode_json (SV *jsonstr)
		2066	ALIAS:
		2067	from_json_ = 0
		2068	decode_json = F_UTF8
1584	PPCODE:	2069	PPCODE:
1585	{	2070	{
1586	JSON json = { F_DEFAULT \| F_UTF8 };	2071	JSON json;
		2072	json_init (&json);
		2073	json.flags \|= ix;
1587	XPUSHs (decode_json (jsonstr, &json, 0));	2074	XPUSHs (decode_json (jsonstr, &json, 0));
1588	}	2075	}
1589		2076

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.65 by root, Sat Oct 13 01:55:31 2007 UTC vs. Revision 1.110 by root, Wed Jun 1 13:01:09 2011 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.65 by root, Sat Oct 13 01:55:31 2007 UTC vs.
Revision 1.110 by root, Wed Jun 1 13:01:09 2011 UTC