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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs.
Revision 1.94 by root, Thu Nov 20 03:59:53 2008 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
…		…
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	// guarentees, 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	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
20		23
21	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
…		…
27	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
28	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
32	#define F_MAXDEPTH 0xf8000000UL
33	#define S_MAXDEPTH 27
34	#define F_MAXSIZE 0x01f00000UL	35	#define F_RELAXED 0x00001000UL
35	#define S_MAXSIZE 20	36	#define F_ALLOW_UNKNOWN 0x00002000UL
36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
37		38
38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
40
41	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
42	#define F_DEFAULT (9UL << S_MAXDEPTH)
43		40
44	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
45	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
46		43
47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
48		45
49	#define SB do {	46	#define SB do {
50	#define SE } while (0)	47	#define SE } while (0)
51		48
52	#if __GNUC__ >= 3	49	#if __GNUC__ >= 3
53	# define expect(expr,value) __builtin_expect ((expr),(value))	50	# define expect(expr,value) __builtin_expect ((expr), (value))
54	# define inline inline	51	# define INLINE static inline
55	#else	52	#else
56	# define expect(expr,value) (expr)	53	# define expect(expr,value) (expr)
57	# define inline static	54	# define INLINE static
58	#endif	55	#endif
59		56
60	#define expect_false(expr) expect ((expr) != 0, 0)	57	#define expect_false(expr) expect ((expr) != 0, 0)
61	#define expect_true(expr) expect ((expr) != 0, 1)	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?)"
62		65
63	#ifdef USE_ITHREADS	66	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	67	# define JSON_SLOW 1
65	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
66	#else	69	#else
…		…
69	#endif	72	#endif
70		73
71	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
72	static SV json_true, json_false;	75	static SV json_true, json_false;
73		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
74	typedef struct {	86	typedef struct {
75	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
76	SV *cb_object;	91	SV *cb_object;
77	HV *cb_sk_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;
78	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
79		107
80	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
81	// utility functions	109	// utility functions
82		110
83	inline void	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
84	shrink (SV *sv)	123	shrink (SV *sv)
85	{	124	{
86	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
87	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
88	{	128	{
89	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
90	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
91	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
97	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
98	// case of an error.	138	// case of an error.
99	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
100	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
101	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
102	inline UV	142	INLINE UV
103	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
104	{	144	{
105	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
106	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
107	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
108	{	148	{
109	*clen = 2;	149	*clen = 2;
110	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
111	}	151	}
112	else	152	else
113	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
114	*clen = (STRLEN)-1;	154	}
115	return (UV)-1;	155
116	}	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;
117	}	178	}
118		179
119	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
120	// encoder	181	// encoder
121		182
…		…
125	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
126	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
127	SV *sv; // result scalar	188	SV *sv; // result scalar
128	JSON json;	189	JSON json;
129	U32 indent; // indentation level	190	U32 indent; // indentation level
130	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
131	} enc_t;	192	} enc_t;
132		193
133	inline void	194	INLINE void
134	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
135	{	196	{
136	if (expect_false (enc->cur + len >= enc->end))	197	if (expect_false (enc->cur + len >= enc->end))
137	{	198	{
138	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
139	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + len + 1);
140	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
142	}	203	}
143	}	204	}
144		205
145	inline void	206	INLINE void
146	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
147	{	208	{
148	need (enc, 1);	209	need (enc, 1);
149	*enc->cur++ = ch;	210	*enc->cur++ = ch;
150	}	211	}
…		…
204	{	265	{
205	uch = ch;	266	uch = ch;
206	clen = 1;	267	clen = 1;
207	}	268	}
208		269
209	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
210	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
211
212	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
213	{	271	{
214	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
215	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
216	need (enc, len += 11);	277	need (enc, len += 11);
217	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
218	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
219	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
220	enc->cur += 12;	281	enc->cur += 12;
221	}	282	}
222	else	283	else
223	{	284	{
224	static char hexdigit [16] = "0123456789abcdef";
225	need (enc, len += 5);	285	need (enc, len += 5);
226	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
227	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
228	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
229	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
230	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
231	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
232	}	292	}
233		293
234	str += clen;	294	str += clen;
235	}	295	}
236	else if (enc->json.flags & F_LATIN1)	296	else if (enc->json.flags & F_LATIN1)
…		…
248	while (--clen);	308	while (--clen);
249	}	309	}
250	else	310	else
251	{	311	{
252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	312	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	313	enc->cur = encode_utf8 (enc->cur, uch);
254	++str;	314	++str;
255	}	315	}
256	}	316	}
257	}	317	}
258	}	318	}
259		319
260	--len;	320	--len;
261	}	321	}
262	}	322	}
263		323
264	inline void	324	INLINE void
265	encode_indent (enc_t *enc)	325	encode_indent (enc_t *enc)
266	{	326	{
267	if (enc->json.flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
268	{	328	{
269	int spaces = enc->indent * INDENT_STEP;	329	int spaces = enc->indent * INDENT_STEP;
…		…
272	memset (enc->cur, ' ', spaces);	332	memset (enc->cur, ' ', spaces);
273	enc->cur += spaces;	333	enc->cur += spaces;
274	}	334	}
275	}	335	}
276		336
277	inline void	337	INLINE void
278	encode_space (enc_t *enc)	338	encode_space (enc_t *enc)
279	{	339	{
280	need (enc, 1);	340	need (enc, 1);
281	encode_ch (enc, ' ');	341	encode_ch (enc, ' ');
282	}	342	}
283		343
284	inline void	344	INLINE void
285	encode_nl (enc_t *enc)	345	encode_nl (enc_t *enc)
286	{	346	{
287	if (enc->json.flags & F_INDENT)	347	if (enc->json.flags & F_INDENT)
288	{	348	{
289	need (enc, 1);	349	need (enc, 1);
290	encode_ch (enc, '\n');	350	encode_ch (enc, '\n');
291	}	351	}
292	}	352	}
293		353
294	inline void	354	INLINE void
295	encode_comma (enc_t *enc)	355	encode_comma (enc_t *enc)
296	{	356	{
297	encode_ch (enc, ',');	357	encode_ch (enc, ',');
298		358
299	if (enc->json.flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
…		…
307	static void	367	static void
308	encode_av (enc_t enc, AV av)	368	encode_av (enc_t enc, AV av)
309	{	369	{
310	int i, len = av_len (av);	370	int i, len = av_len (av);
311		371
312	if (enc->indent >= enc->maxdepth)	372	if (enc->indent >= enc->json.max_depth)
313	croak ("data structure too deep (hit recursion limit)");	373	croak (ERR_NESTING_EXCEEDED);
314		374
315	encode_ch (enc, '['); encode_nl (enc);	375	encode_ch (enc, '[');
316	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
317		380
318	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
319	{	382	{
320	SV **svp = av_fetch (av, i, 0);	383	SV **svp = av_fetch (av, i, 0);
321		384
322	encode_indent (enc);	385	encode_indent (enc);
323		386
324	if (svp)	387	if (svp)
325	encode_sv (enc, *svp);	388	encode_sv (enc, *svp);
326	else	389	else
327	encode_str (enc, "null", 4, 0);	390	encode_str (enc, "null", 4, 0);
328		391
329	if (i < len)	392	if (i < len)
330	encode_comma (enc);	393	encode_comma (enc);
331	}	394	}
332		395
		396	encode_nl (enc); --enc->indent; encode_indent (enc);
		397	}
		398
333	encode_nl (enc);	399	encode_ch (enc, ']');
334
335	--enc->indent;
336	encode_indent (enc); encode_ch (enc, ']');
337	}	400	}
338		401
339	static void	402	static void
340	encode_he (enc_t enc, HE he)	403	encode_hk (enc_t enc, HE he)
341	{	404	{
342	encode_ch (enc, '"');	405	encode_ch (enc, '"');
343		406
344	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
345	{	408	{
…		…
358	encode_ch (enc, '"');	421	encode_ch (enc, '"');
359		422
360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
361	encode_ch (enc, ':');	424	encode_ch (enc, ':');
362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
363	encode_sv (enc, HeVAL (he));
364	}	426	}
365		427
366	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
367	static int	429	static int
368	he_cmp_fast (const void a_, const void b_)	430	he_cmp_fast (const void a_, const void b_)
…		…
373	HE b = (HE **)b_;	435	HE b = (HE **)b_;
374		436
375	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
376	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
377		439
378	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
379	cmp = la - lb;	441	cmp = lb - la;
380		442
381	return cmp;	443	return cmp;
382	}	444	}
383		445
384	// compare hash entries, used when some keys are sv's or utf-x	446	// compare hash entries, used when some keys are sv's or utf-x
385	static int	447	static int
386	he_cmp_slow (const void a, const void b)	448	he_cmp_slow (const void a, const void b)
387	{	449	{
388	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	450	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
389	}	451	}
390		452
391	static void	453	static void
392	encode_hv (enc_t enc, HV hv)	454	encode_hv (enc_t enc, HV hv)
393	{	455	{
394	int count, i;	456	HE *he;
395		457
396	if (enc->indent >= enc->maxdepth)	458	if (enc->indent >= enc->json.max_depth)
397	croak ("data structure too deep (hit recursion limit)");	459	croak (ERR_NESTING_EXCEEDED);
398		460
399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	461	encode_ch (enc, '{');
400		462
401	if ((count = hv_iterinit (hv)))
402	{
403	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	466	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	467	if (enc->json.flags & F_CANONICAL)
		468	{
		469	int count = hv_iterinit (hv);
		470
		471	if (SvMAGICAL (hv))
408	{	472	{
		473	// need to count by iterating. could improve by dynamically building the vector below
		474	// but I don't care for the speed of this special case.
		475	// note also that we will run into undefined behaviour when the two iterations
		476	// do not result in the same count, something I might care for in some later release.
		477
		478	count = 0;
		479	while (hv_iternext (hv))
		480	++count;
		481
		482	hv_iterinit (hv);
		483	}
		484
		485	if (count)
		486	{
409	int fast = 1;	487	int i, fast = 1;
410	HE *he;
411	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	488	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
412	HE *hes = _alloca (count sizeof (HE));	489	HE *hes = _alloca (count sizeof (HE));
413	#else	490	#else
414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	491	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
415	#endif	492	#endif
…		…
442		519
443	FREETMPS;	520	FREETMPS;
444	LEAVE;	521	LEAVE;
445	}	522	}
446		523
447	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
448	{	527	{
449	encode_indent (enc);	528	encode_indent (enc);
		529	he = hes [count];
450	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		532
452	if (i < count - 1)	533	if (count)
453	encode_comma (enc);	534	encode_comma (enc);
454	}	535	}
455		536
456	encode_nl (enc);	537	encode_nl (enc); --enc->indent; encode_indent (enc);
457	}	538	}
		539	}
458	else	540	else
		541	{
		542	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
459	{	544	{
460	HE *he = hv_iternext (hv);	545	encode_nl (enc); ++enc->indent;
461		546
462	for (;;)	547	for (;;)
463	{	548	{
464	encode_indent (enc);	549	encode_indent (enc);
465	encode_he (enc, he);	550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
466		552
467	if (!(he = hv_iternext (hv)))	553	if (!(he = hv_iternext (hv)))
468	break;	554	break;
469		555
470	encode_comma (enc);	556	encode_comma (enc);
471	}	557	}
472		558
473	encode_nl (enc);	559	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	560	}
475	}	561	}
476		562
477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	563	encode_ch (enc, '}');
478	}	564	}
479		565
480	// encode objects, arrays and special \0=false and \1=true values.	566	// encode objects, arrays and special \0=false and \1=true values.
481	static void	567	static void
482	encode_rv (enc_t enc, SV sv)	568	encode_rv (enc_t enc, SV sv)
…		…
559		645
560	if (len == 1 && *pv == '1')	646	if (len == 1 && *pv == '1')
561	encode_str (enc, "true", 4, 0);	647	encode_str (enc, "true", 4, 0);
562	else if (len == 1 && *pv == '0')	648	else if (len == 1 && *pv == '0')
563	encode_str (enc, "false", 5, 0);	649	encode_str (enc, "false", 5, 0);
		650	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		651	encode_str (enc, "null", 4, 0);
564	else	652	else
565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	653	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
567	}	655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
568	else	658	else
569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	659	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
571	}	661	}
572		662
…		…
590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
591	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
592	}	682	}
593	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
594	{	684	{
595	// we assume we can always read an IV as a UV	685	// we assume we can always read an IV as a UV and vice versa
596	if (SvUV (sv) & ~(UV)0x7fff)	686	// we assume two's complement
597	{	687	// we assume no aliasing issues in the union
598	// large integer, use the (rather slow) snprintf way.	688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
599	need (enc, sizeof (UV) * 3);	689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
600	enc->cur +=
601	SvIsUV(sv)
602	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
603	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
604	}
605	else
606	{	690	{
607	// optimise the "small number case"	691	// optimise the "small number case"
608	// code will likely be branchless and use only a single multiplication	692	// code will likely be branchless and use only a single multiplication
		693	// works for numbers up to 59074
609	I32 i = SvIV (sv);	694	I32 i = SvIVX (sv);
610	U32 u;	695	U32 u;
611	char digit, nz = 0;	696	char digit, nz = 0;
612		697
613	need (enc, 6);	698	need (enc, 6);
614		699
…		…
620		705
621	// now output digit by digit, each time masking out the integer part	706	// now output digit by digit, each time masking out the integer part
622	// and multiplying by 5 while moving the decimal point one to the right,	707	// and multiplying by 5 while moving the decimal point one to the right,
623	// resulting in a net multiplication by 10.	708	// resulting in a net multiplication by 10.
624	// we always write the digit to memory but conditionally increment	709	// we always write the digit to memory but conditionally increment
625	// the pointer, to ease the usage of conditional move instructions.	710	// the pointer, to enable the use of conditional move instructions.
626	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	711	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
627	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	712	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
628	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	713	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
629	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	714	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
630	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	715	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
631	}	716	}
		717	else
		718	{
		719	// large integer, use the (rather slow) snprintf way.
		720	need (enc, IVUV_MAXCHARS);
		721	enc->cur +=
		722	SvIsUV(sv)
		723	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		724	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		725	}
632	}	726	}
633	else if (SvROK (sv))	727	else if (SvROK (sv))
634	encode_rv (enc, SvRV (sv));	728	encode_rv (enc, SvRV (sv));
635	else if (!SvOK (sv))	729	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
636	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
637	else	731	else
638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	732	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
639	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
640	}	734	}
…		…
650	enc.json = *json;	744	enc.json = *json;
651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
652	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
653	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
654	enc.indent = 0;	748	enc.indent = 0;
655	enc.maxdepth = DEC_DEPTH (enc.json.flags);	749	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		750	: enc.json.flags & F_LATIN1 ? 0x000100UL
		751	: 0x110000UL;
656		752
657	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
658	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
659		755
660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
681	JSON json;	777	JSON json;
682	U32 depth; // recursion depth	778	U32 depth; // recursion depth
683	U32 maxdepth; // recursion depth limit	779	U32 maxdepth; // recursion depth limit
684	} dec_t;	780	} dec_t;
685		781
686	inline void	782	INLINE void
		783	decode_comment (dec_t *dec)
		784	{
		785	// only '#'-style comments allowed a.t.m.
		786
		787	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		788	++dec->cur;
		789	}
		790
		791	INLINE void
687	decode_ws (dec_t *dec)	792	decode_ws (dec_t *dec)
688	{	793	{
689	for (;;)	794	for (;;)
690	{	795	{
691	char ch = *dec->cur;	796	char ch = *dec->cur;
692		797
693	if (ch > 0x20	798	if (ch > 0x20)
		799	{
		800	if (expect_false (ch == '#'))
		801	{
		802	if (dec->json.flags & F_RELAXED)
		803	decode_comment (dec);
		804	else
		805	break;
		806	}
		807	else
		808	break;
		809	}
694	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
695	break;	811	break; // parse error, but let higher level handle it, gives better error messages
696		812
697	++dec->cur;	813	++dec->cur;
698	}	814	}
699	}	815	}
700		816
…		…
704	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
705	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
706	++dec->cur; \	822	++dec->cur; \
707	SE	823	SE
708		824
709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	825	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
710	#define DEC_DEC_DEPTH --dec->depth	826	#define DEC_DEC_DEPTH --dec->depth
711		827
712	static SV decode_sv (dec_t dec);	828	static SV decode_sv (dec_t dec);
713		829
714	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
…		…
806		922
807	if (hi >= 0x80)	923	if (hi >= 0x80)
808	{	924	{
809	utf8 = 1;	925	utf8 = 1;
810		926
811	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	927	cur = encode_utf8 (cur, hi);
812	}	928	}
813	else	929	else
814	*cur++ = hi;	930	*cur++ = hi;
815	}	931	}
816	break;	932	break;
…		…
818	default:	934	default:
819	--dec_cur;	935	--dec_cur;
820	ERR ("illegal backslash escape sequence in string");	936	ERR ("illegal backslash escape sequence in string");
821	}	937	}
822	}	938	}
823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	939	else if (expect_true (ch >= 0x20 && ch < 0x80))
824	*cur++ = ch;	940	*cur++ = ch;
825	else if (ch >= 0x80)	941	else if (ch >= 0x80)
826	{	942	{
827	STRLEN clen;	943	STRLEN clen;
828	UV uch;	944	UV uch;
…		…
951		1067
952	if (!is_nv)	1068	if (!is_nv)
953	{	1069	{
954	int len = dec->cur - start;	1070	int len = dec->cur - start;
955		1071
956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1072	// special case the rather common 1..5-digit-int case
957	if (*start == '-')	1073	if (*start == '-')
958	switch (len)	1074	switch (len)
959	{	1075	{
960	case 2: return newSViv (-( start [1] - '0' * 1));	1076	case 2: return newSViv (-( start [1] - '0' * 1));
961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1077	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
962	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1078	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
963	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1079	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1080	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
964	}	1081	}
965	else	1082	else
966	switch (len)	1083	switch (len)
967	{	1084	{
968	case 1: return newSViv ( start [0] - '0' * 1);	1085	case 1: return newSViv ( start [0] - '0' * 1);
969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1086	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
970	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1087	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
971	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1088	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1089	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
972	}	1090	}
973		1091
974	{	1092	{
975	UV uv;	1093	UV uv;
976	int numtype = grok_number (start, len, &uv);	1094	int numtype = grok_number (start, len, &uv);
…		…
1037		1155
1038	if (*dec->cur != ',')	1156	if (*dec->cur != ',')
1039	ERR (", or ] expected while parsing array");	1157	ERR (", or ] expected while parsing array");
1040		1158
1041	++dec->cur;	1159	++dec->cur;
		1160
		1161	decode_ws (dec);
		1162
		1163	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1164	{
		1165	++dec->cur;
		1166	break;
		1167	}
1042	}	1168	}
1043		1169
1044	DEC_DEC_DEPTH;	1170	DEC_DEC_DEPTH;
1045	return newRV_noinc ((SV *)av);	1171	return newRV_noinc ((SV *)av);
1046		1172
…		…
1062	if (*dec->cur == '}')	1188	if (*dec->cur == '}')
1063	++dec->cur;	1189	++dec->cur;
1064	else	1190	else
1065	for (;;)	1191	for (;;)
1066	{	1192	{
1067	decode_ws (dec); EXPECT_CH ('"');	1193	EXPECT_CH ('"');
1068		1194
1069	// heuristic: assume that	1195	// heuristic: assume that
1070	// a) decode_str + hv_store_ent are abysmally slow.	1196	// a) decode_str + hv_store_ent are abysmally slow.
1071	// b) most hash keys are short, simple ascii text.	1197	// b) most hash keys are short, simple ascii text.
1072	// => try to "fast-match" such strings to avoid	1198	// => try to "fast-match" such strings to avoid
…		…
1076	char *p = dec->cur;	1202	char *p = dec->cur;
1077	char *e = p + 24; // only try up to 24 bytes	1203	char *e = p + 24; // only try up to 24 bytes
1078		1204
1079	for (;;)	1205	for (;;)
1080	{	1206	{
1081	// the >= 0x80 is true on most architectures	1207	// the >= 0x80 is false on most architectures
1082	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1208	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1083	{	1209	{
1084	// slow path, back up and use decode_str	1210	// slow path, back up and use decode_str
1085	SV *key = decode_str (dec);	1211	SV *key = decode_str (dec);
1086	if (!key)	1212	if (!key)
1087	goto fail;	1213	goto fail;
1088		1214
1089	decode_ws (dec); EXPECT_CH (':');	1215	decode_ws (dec); EXPECT_CH (':');
1090		1216
		1217	decode_ws (dec);
1091	value = decode_sv (dec);	1218	value = decode_sv (dec);
1092	if (!value)	1219	if (!value)
1093	{	1220	{
1094	SvREFCNT_dec (key);	1221	SvREFCNT_dec (key);
1095	goto fail;	1222	goto fail;
…		…
1107	int len = p - key;	1234	int len = p - key;
1108	dec->cur = p + 1;	1235	dec->cur = p + 1;
1109		1236
1110	decode_ws (dec); EXPECT_CH (':');	1237	decode_ws (dec); EXPECT_CH (':');
1111		1238
		1239	decode_ws (dec);
1112	value = decode_sv (dec);	1240	value = decode_sv (dec);
1113	if (!value)	1241	if (!value)
1114	goto fail;	1242	goto fail;
1115		1243
1116	hv_store (hv, key, len, value, 0);	1244	hv_store (hv, key, len, value, 0);
…		…
1132		1260
1133	if (*dec->cur != ',')	1261	if (*dec->cur != ',')
1134	ERR (", or } expected while parsing object/hash");	1262	ERR (", or } expected while parsing object/hash");
1135		1263
1136	++dec->cur;	1264	++dec->cur;
		1265
		1266	decode_ws (dec);
		1267
		1268	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1269	{
		1270	++dec->cur;
		1271	break;
		1272	}
1137	}	1273	}
1138		1274
1139	DEC_DEC_DEPTH;	1275	DEC_DEC_DEPTH;
1140	sv = newRV_noinc ((SV *)hv);	1276	sv = newRV_noinc ((SV *)hv);
1141		1277
…		…
1206	}	1342	}
1207		1343
1208	static SV *	1344	static SV *
1209	decode_sv (dec_t *dec)	1345	decode_sv (dec_t *dec)
1210	{	1346	{
1211	decode_ws (dec);
1212
1213	// the beauty of JSON: you need exactly one character lookahead	1347	// the beauty of JSON: you need exactly one character lookahead
1214	// to parse anything.	1348	// to parse everything.
1215	switch (*dec->cur)	1349	switch (*dec->cur)
1216	{	1350	{
1217	case '"': ++dec->cur; return decode_str (dec);	1351	case '"': ++dec->cur; return decode_str (dec);
1218	case '[': ++dec->cur; return decode_av (dec);	1352	case '[': ++dec->cur; return decode_av (dec);
1219	case '{': ++dec->cur; return decode_hv (dec);	1353	case '{': ++dec->cur; return decode_hv (dec);
1220		1354
1221	case '-':	1355	case '-':
1222	case '0': case '1': case '2': case '3': case '4':	1356	case '0': case '1': case '2': case '3': case '4':
1223	case '5': case '6': case '7': case '8': case '9':	1357	case '5': case '6': case '7': case '8': case '9':
1224	return decode_num (dec);	1358	return decode_num (dec);
1225		1359
1226	case 't':	1360	case 't':
1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1361	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1228	{	1362	{
1229	dec->cur += 4;	1363	dec->cur += 4;
		1364	#if JSON_SLOW
		1365	json_true = get_bool ("JSON::XS::true");
		1366	#endif
1230	return SvREFCNT_inc (json_true);	1367	return newSVsv (json_true);
1231	}	1368	}
1232	else	1369	else
1233	ERR ("'true' expected");	1370	ERR ("'true' expected");
1234		1371
1235	break;	1372	break;
1236		1373
1237	case 'f':	1374	case 'f':
1238	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1375	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1239	{	1376	{
1240	dec->cur += 5;	1377	dec->cur += 5;
		1378	#if JSON_SLOW
		1379	json_false = get_bool ("JSON::XS::false");
		1380	#endif
1241	return SvREFCNT_inc (json_false);	1381	return newSVsv (json_false);
1242	}	1382	}
1243	else	1383	else
1244	ERR ("'false' expected");	1384	ERR ("'false' expected");
1245		1385
1246	break;	1386	break;
…		…
1264	fail:	1404	fail:
1265	return 0;	1405	return 0;
1266	}	1406	}
1267		1407
1268	static SV *	1408	static SV *
1269	decode_json (SV string, JSON json, UV *offset_return)	1409	decode_json (SV string, JSON json, STRLEN *offset_return)
1270	{	1410	{
1271	dec_t dec;	1411	dec_t dec;
1272	UV offset;	1412	STRLEN offset;
1273	SV *sv;	1413	SV *sv;
1274		1414
		1415	/* work around bugs in 5.10 where manipulating magic values
		1416	* will perl ignore the magic in subsequent accesses
		1417	*/
1275	SvGETMAGIC (string);	1418	/SvGETMAGIC (string);/
		1419	if (SvMAGICAL (string))
		1420	string = sv_2mortal (newSVsv (string));
		1421
1276	SvUPGRADE (string, SVt_PV);	1422	SvUPGRADE (string, SVt_PV);
1277		1423
1278	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1424	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1425	* assertion with -DDEBUGGING, although SvCUR is documented to
		1426	* return the xpv_cur field which certainly exists after upgrading.
		1427	* according to nicholas clark, calling SvPOK fixes this.
		1428	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1429	* and hope for the best.
		1430	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1431	* assertion business is seriously broken, try yet another workaround
		1432	* for the broken -DDEBUGGING.
		1433	*/
		1434	#ifdef DEBUGGING
		1435	offset = SvOK (string) ? sv_len (string) : 0;
		1436	#else
		1437	offset = SvCUR (string);
		1438	#endif
		1439
		1440	if (offset > json->max_size && json->max_size)
1279	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1441	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1280	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1442	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
1281		1443
1282	if (json->flags & F_UTF8)	1444	if (json->flags & F_UTF8)
1283	sv_utf8_downgrade (string, 0);	1445	sv_utf8_downgrade (string, 0);
1284	else	1446	else
1285	sv_utf8_upgrade (string);	1447	sv_utf8_upgrade (string);
1286		1448
1287	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1449	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1288		1450
1289	dec.json = *json;	1451	dec.json = *json;
1290	dec.cur = SvPVX (string);	1452	dec.cur = SvPVX (string);
1291	dec.end = SvEND (string);	1453	dec.end = SvEND (string);
1292	dec.err = 0;	1454	dec.err = 0;
1293	dec.depth = 0;	1455	dec.depth = 0;
1294	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1295		1456
1296	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1457	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1297	dec.json.flags \|= F_HOOK;	1458	dec.json.flags \|= F_HOOK;
1298		1459
1299	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1460	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1461
		1462	decode_ws (&dec);
1300	sv = decode_sv (&dec);	1463	sv = decode_sv (&dec);
1301		1464
1302	if (!(offset_return \|\| !sv))	1465	if (!(offset_return \|\| !sv))
1303	{	1466	{
1304	// check for trailing garbage	1467	// check for trailing garbage
…		…
1348		1511
1349	return sv;	1512	return sv;
1350	}	1513	}
1351		1514
1352	/////////////////////////////////////////////////////////////////////////////	1515	/////////////////////////////////////////////////////////////////////////////
		1516	// incremental parser
		1517
		1518	static void
		1519	incr_parse (JSON *self)
		1520	{
		1521	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1522
		1523	for (;;)
		1524	{
		1525	//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
		1526	switch (self->incr_mode)
		1527	{
		1528	// only used for intiial whitespace skipping
		1529	case INCR_M_WS:
		1530	for (;;)
		1531	{
		1532	if (*p > 0x20)
		1533	{
		1534	self->incr_mode = INCR_M_JSON;
		1535	goto incr_m_json;
		1536	}
		1537	else if (!*p)
		1538	goto interrupt;
		1539
		1540	++p;
		1541	}
		1542
		1543	// skip a single char inside a string (for \\-processing)
		1544	case INCR_M_BS:
		1545	if (!*p)
		1546	goto interrupt;
		1547
		1548	++p;
		1549	self->incr_mode = INCR_M_STR;
		1550	goto incr_m_str;
		1551
		1552	// inside a string
		1553	case INCR_M_STR:
		1554	incr_m_str:
		1555	for (;;)
		1556	{
		1557	if (*p == '"')
		1558	{
		1559	++p;
		1560	self->incr_mode = INCR_M_JSON;
		1561
		1562	if (!self->incr_nest)
		1563	goto interrupt;
		1564
		1565	goto incr_m_json;
		1566	}
		1567	else if (*p == '\\')
		1568	{
		1569	++p; // "virtually" consumes character after \
		1570
		1571	if (!*p) // if at end of string we have to switch modes
		1572	{
		1573	self->incr_mode = INCR_M_BS;
		1574	goto interrupt;
		1575	}
		1576	}
		1577	else if (!*p)
		1578	goto interrupt;
		1579
		1580	++p;
		1581	}
		1582
		1583	// after initial ws, outside string
		1584	case INCR_M_JSON:
		1585	incr_m_json:
		1586	for (;;)
		1587	{
		1588	switch (*p++)
		1589	{
		1590	case 0:
		1591	--p;
		1592	goto interrupt;
		1593
		1594	case 0x09:
		1595	case 0x0a:
		1596	case 0x0d:
		1597	case 0x20:
		1598	if (!self->incr_nest)
		1599	{
		1600	--p; // do not eat the whitespace, let the next round do it
		1601	goto interrupt;
		1602	}
		1603	break;
		1604
		1605	case '"':
		1606	self->incr_mode = INCR_M_STR;
		1607	goto incr_m_str;
		1608
		1609	case '[':
		1610	case '{':
		1611	if (++self->incr_nest > self->max_depth)
		1612	croak (ERR_NESTING_EXCEEDED);
		1613	break;
		1614
		1615	case ']':
		1616	case '}':
		1617	if (--self->incr_nest <= 0)
		1618	goto interrupt;
		1619	}
		1620	}
		1621	}
		1622
		1623	modechange:
		1624	;
		1625	}
		1626
		1627	interrupt:
		1628	self->incr_pos = p - SvPVX (self->incr_text);
		1629	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1630	}
		1631
		1632	/////////////////////////////////////////////////////////////////////////////
1353	// XS interface functions	1633	// XS interface functions
1354		1634
1355	MODULE = JSON::XS PACKAGE = JSON::XS	1635	MODULE = JSON::XS PACKAGE = JSON::XS
1356		1636
1357	BOOT:	1637	BOOT:
…		…
1366	: -1;	1646	: -1;
1367		1647
1368	json_stash = gv_stashpv ("JSON::XS" , 1);	1648	json_stash = gv_stashpv ("JSON::XS" , 1);
1369	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1649	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1370		1650
1371	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1651	json_true = get_bool ("JSON::XS::true");
1372	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1652	json_false = get_bool ("JSON::XS::false");
1373	}	1653	}
1374		1654
1375	PROTOTYPES: DISABLE	1655	PROTOTYPES: DISABLE
1376		1656
1377	void CLONE (...)	1657	void CLONE (...)
…		…
1382	void new (char *klass)	1662	void new (char *klass)
1383	PPCODE:	1663	PPCODE:
1384	{	1664	{
1385	SV *pv = NEWSV (0, sizeof (JSON));	1665	SV *pv = NEWSV (0, sizeof (JSON));
1386	SvPOK_only (pv);	1666	SvPOK_only (pv);
1387	Zero (SvPVX (pv), 1, JSON);	1667	json_init ((JSON *)SvPVX (pv));
1388	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1668	XPUSHs (sv_2mortal (sv_bless (
1389	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1669	newRV_noinc (pv),
		1670	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1671	)));
1390	}	1672	}
1391		1673
1392	void ascii (JSON *self, int enable = 1)	1674	void ascii (JSON *self, int enable = 1)
1393	ALIAS:	1675	ALIAS:
1394	ascii = F_ASCII	1676	ascii = F_ASCII
…		…
1401	pretty = F_PRETTY	1683	pretty = F_PRETTY
1402	allow_nonref = F_ALLOW_NONREF	1684	allow_nonref = F_ALLOW_NONREF
1403	shrink = F_SHRINK	1685	shrink = F_SHRINK
1404	allow_blessed = F_ALLOW_BLESSED	1686	allow_blessed = F_ALLOW_BLESSED
1405	convert_blessed = F_CONV_BLESSED	1687	convert_blessed = F_CONV_BLESSED
		1688	relaxed = F_RELAXED
		1689	allow_unknown = F_ALLOW_UNKNOWN
1406	PPCODE:	1690	PPCODE:
1407	{	1691	{
1408	if (enable)	1692	if (enable)
1409	self->flags \|= ix;	1693	self->flags \|= ix;
1410	else	1694	else
1411	self->flags &= ~ix;	1695	self->flags &= ~ix;
1412		1696
1413	XPUSHs (ST (0));	1697	XPUSHs (ST (0));
1414	}	1698	}
1415		1699
1416	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1700	void get_ascii (JSON *self)
		1701	ALIAS:
		1702	get_ascii = F_ASCII
		1703	get_latin1 = F_LATIN1
		1704	get_utf8 = F_UTF8
		1705	get_indent = F_INDENT
		1706	get_canonical = F_CANONICAL
		1707	get_space_before = F_SPACE_BEFORE
		1708	get_space_after = F_SPACE_AFTER
		1709	get_allow_nonref = F_ALLOW_NONREF
		1710	get_shrink = F_SHRINK
		1711	get_allow_blessed = F_ALLOW_BLESSED
		1712	get_convert_blessed = F_CONV_BLESSED
		1713	get_relaxed = F_RELAXED
		1714	get_allow_unknown = F_ALLOW_UNKNOWN
1417	PPCODE:	1715	PPCODE:
1418	{	1716	XPUSHs (boolSV (self->flags & ix));
1419	UV log2 = 0;
1420		1717
1421	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1718	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1422		1719	PPCODE:
1423	while ((1UL << log2) < max_depth)	1720	self->max_depth = max_depth;
1424	++log2;
1425
1426	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1427
1428	XPUSHs (ST (0));	1721	XPUSHs (ST (0));
1429	}
1430		1722
		1723	U32 get_max_depth (JSON *self)
		1724	CODE:
		1725	RETVAL = self->max_depth;
		1726	OUTPUT:
		1727	RETVAL
		1728
1431	void max_size (JSON *self, UV max_size = 0)	1729	void max_size (JSON *self, U32 max_size = 0)
1432	PPCODE:	1730	PPCODE:
1433	{	1731	self->max_size = max_size;
1434	UV log2 = 0;
1435
1436	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1437	if (max_size == 1) max_size = 2;
1438
1439	while ((1UL << log2) < max_size)
1440	++log2;
1441
1442	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1443
1444	XPUSHs (ST (0));	1732	XPUSHs (ST (0));
1445	}	1733
		1734	int get_max_size (JSON *self)
		1735	CODE:
		1736	RETVAL = self->max_size;
		1737	OUTPUT:
		1738	RETVAL
1446		1739
1447	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1740	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1448	PPCODE:	1741	PPCODE:
1449	{	1742	{
1450	SvREFCNT_dec (self->cb_object);	1743	SvREFCNT_dec (self->cb_object);
…		…
1484	XPUSHs (decode_json (jsonstr, self, 0));	1777	XPUSHs (decode_json (jsonstr, self, 0));
1485		1778
1486	void decode_prefix (JSON self, SV jsonstr)	1779	void decode_prefix (JSON self, SV jsonstr)
1487	PPCODE:	1780	PPCODE:
1488	{	1781	{
1489	UV offset;	1782	STRLEN offset;
1490	EXTEND (SP, 2);	1783	EXTEND (SP, 2);
1491	PUSHs (decode_json (jsonstr, self, &offset));	1784	PUSHs (decode_json (jsonstr, self, &offset));
1492	PUSHs (sv_2mortal (newSVuv (offset)));	1785	PUSHs (sv_2mortal (newSVuv (offset)));
		1786	}
		1787
		1788	void incr_parse (JSON self, SV jsonstr = 0)
		1789	PPCODE:
		1790	{
		1791	if (!self->incr_text)
		1792	self->incr_text = newSVpvn ("", 0);
		1793
		1794	// append data, if any
		1795	if (jsonstr)
		1796	{
		1797	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1798	{
		1799	/* utf-8-ness differs, need to upgrade */
		1800	sv_utf8_upgrade (self->incr_text);
		1801
		1802	if (self->incr_pos)
		1803	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1804	- (U8 *)SvPVX (self->incr_text);
		1805	}
		1806
		1807	{
		1808	STRLEN len;
		1809	const char *str = SvPV (jsonstr, len);
		1810	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1811	Move (str, SvEND (self->incr_text), len, char);
		1812	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1813	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1814	}
		1815	}
		1816
		1817	if (GIMME_V != G_VOID)
		1818	do
		1819	{
		1820	STRLEN offset;
		1821
		1822	if (!INCR_DONE (self))
		1823	{
		1824	incr_parse (self);
		1825
		1826	if (self->incr_pos > self->max_size && self->max_size)
		1827	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1828	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1829
		1830	if (!INCR_DONE (self))
		1831	break;
		1832	}
		1833
		1834	XPUSHs (decode_json (self->incr_text, self, &offset));
		1835
		1836	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1837	self->incr_pos -= offset;
		1838	self->incr_nest = 0;
		1839	self->incr_mode = 0;
		1840	}
		1841	while (GIMME_V == G_ARRAY);
		1842	}
		1843
		1844	SV incr_text (JSON self)
		1845	ATTRS: lvalue
		1846	CODE:
		1847	{
		1848	if (self->incr_pos)
		1849	croak ("incr_text can not be called when the incremental parser already started parsing");
		1850
		1851	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1852	}
		1853	OUTPUT:
		1854	RETVAL
		1855
		1856	void incr_skip (JSON *self)
		1857	CODE:
		1858	{
		1859	if (self->incr_pos)
		1860	{
		1861	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1862	self->incr_pos = 0;
		1863	self->incr_nest = 0;
		1864	self->incr_mode = 0;
		1865	}
		1866	}
		1867
		1868	void incr_reset (JSON *self)
		1869	CODE:
		1870	{
		1871	SvREFCNT_dec (self->incr_text);
		1872	self->incr_text = 0;
		1873	self->incr_pos = 0;
		1874	self->incr_nest = 0;
		1875	self->incr_mode = 0;
1493	}	1876	}
1494		1877
1495	void DESTROY (JSON *self)	1878	void DESTROY (JSON *self)
1496	CODE:	1879	CODE:
1497	SvREFCNT_dec (self->cb_sk_object);	1880	SvREFCNT_dec (self->cb_sk_object);
1498	SvREFCNT_dec (self->cb_object);	1881	SvREFCNT_dec (self->cb_object);
		1882	SvREFCNT_dec (self->incr_text);
1499		1883
1500	PROTOTYPES: ENABLE	1884	PROTOTYPES: ENABLE
1501		1885
1502	void to_json (SV *scalar)	1886	void encode_json (SV *scalar)
		1887	ALIAS:
		1888	to_json_ = 0
		1889	encode_json = F_UTF8
1503	PPCODE:	1890	PPCODE:
1504	{	1891	{
1505	JSON json = { F_DEFAULT \| F_UTF8 };	1892	JSON json;
		1893	json_init (&json);
		1894	json.flags \|= ix;
1506	XPUSHs (encode_json (scalar, &json));	1895	XPUSHs (encode_json (scalar, &json));
1507	}	1896	}
1508		1897
1509	void from_json (SV *jsonstr)	1898	void decode_json (SV *jsonstr)
		1899	ALIAS:
		1900	from_json_ = 0
		1901	decode_json = F_UTF8
1510	PPCODE:	1902	PPCODE:
1511	{	1903	{
1512	JSON json = { F_DEFAULT \| F_UTF8 };	1904	JSON json;
		1905	json_init (&json);
		1906	json.flags \|= ix;
1513	XPUSHs (decode_json (jsonstr, &json, 0));	1907	XPUSHs (decode_json (jsonstr, &json, 0));
1514	}	1908	}
1515		1909
		1910

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs. Revision 1.94 by root, Thu Nov 20 03:59:53 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs.
Revision 1.94 by root, Thu Nov 20 03:59:53 2008 UTC