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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC vs.
Revision 1.88 by root, Sun Jun 22 12:05:49 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)
62		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
63	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
64	static SV json_true, json_false;	75	static SV json_true, json_false;
65		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
66	typedef struct {	86	typedef struct {
67	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
68	SV *cb_object;	91	SV *cb_object;
69	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;
70	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
71		107
72	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
73	// utility functions	109	// utility functions
74		110
75	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
76	shrink (SV *sv)	123	shrink (SV *sv)
77	{	124	{
78	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
79	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
80	{	128	{
81	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
82	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
83	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
89	// 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
90	// case of an error.	138	// case of an error.
91	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
92	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
93	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
94	inline UV	142	INLINE UV
95	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
96	{	144	{
97	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
98	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
99	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
100	{	148	{
101	*clen = 2;	149	*clen = 2;
102	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
103	}	151	}
104	else	152	else
105	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
106	*clen = (STRLEN)-1;	154	}
107	return (UV)-1;	155
108	}	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;
109	}	178	}
110		179
111	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
112	// encoder	181	// encoder
113		182
…		…
117	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
118	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
119	SV *sv; // result scalar	188	SV *sv; // result scalar
120	JSON json;	189	JSON json;
121	U32 indent; // indentation level	190	U32 indent; // indentation level
122	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
123	} enc_t;	192	} enc_t;
124		193
125	inline void	194	INLINE void
126	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
127	{	196	{
128	if (expect_false (enc->cur + len >= enc->end))	197	if (expect_false (enc->cur + len >= enc->end))
129	{	198	{
130	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
132	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
133	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
134	}	203	}
135	}	204	}
136		205
137	inline void	206	INLINE void
138	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
139	{	208	{
140	need (enc, 1);	209	need (enc, 1);
141	*enc->cur++ = ch;	210	*enc->cur++ = ch;
142	}	211	}
…		…
196	{	265	{
197	uch = ch;	266	uch = ch;
198	clen = 1;	267	clen = 1;
199	}	268	}
200		269
201	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
202	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
203
204	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
205	{	271	{
206	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
207	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
208	need (enc, len += 11);	277	need (enc, len += 11);
209	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
210	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
211	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
212	enc->cur += 12;	281	enc->cur += 12;
…		…
240	while (--clen);	309	while (--clen);
241	}	310	}
242	else	311	else
243	{	312	{
244	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	313	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
245	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	314	enc->cur = encode_utf8 (enc->cur, uch);
246	++str;	315	++str;
247	}	316	}
248	}	317	}
249	}	318	}
250	}	319	}
251		320
252	--len;	321	--len;
253	}	322	}
254	}	323	}
255		324
256	inline void	325	INLINE void
257	encode_indent (enc_t *enc)	326	encode_indent (enc_t *enc)
258	{	327	{
259	if (enc->json.flags & F_INDENT)	328	if (enc->json.flags & F_INDENT)
260	{	329	{
261	int spaces = enc->indent * INDENT_STEP;	330	int spaces = enc->indent * INDENT_STEP;
…		…
264	memset (enc->cur, ' ', spaces);	333	memset (enc->cur, ' ', spaces);
265	enc->cur += spaces;	334	enc->cur += spaces;
266	}	335	}
267	}	336	}
268		337
269	inline void	338	INLINE void
270	encode_space (enc_t *enc)	339	encode_space (enc_t *enc)
271	{	340	{
272	need (enc, 1);	341	need (enc, 1);
273	encode_ch (enc, ' ');	342	encode_ch (enc, ' ');
274	}	343	}
275		344
276	inline void	345	INLINE void
277	encode_nl (enc_t *enc)	346	encode_nl (enc_t *enc)
278	{	347	{
279	if (enc->json.flags & F_INDENT)	348	if (enc->json.flags & F_INDENT)
280	{	349	{
281	need (enc, 1);	350	need (enc, 1);
282	encode_ch (enc, '\n');	351	encode_ch (enc, '\n');
283	}	352	}
284	}	353	}
285		354
286	inline void	355	INLINE void
287	encode_comma (enc_t *enc)	356	encode_comma (enc_t *enc)
288	{	357	{
289	encode_ch (enc, ',');	358	encode_ch (enc, ',');
290		359
291	if (enc->json.flags & F_INDENT)	360	if (enc->json.flags & F_INDENT)
…		…
299	static void	368	static void
300	encode_av (enc_t enc, AV av)	369	encode_av (enc_t enc, AV av)
301	{	370	{
302	int i, len = av_len (av);	371	int i, len = av_len (av);
303		372
304	if (enc->indent >= enc->maxdepth)	373	if (enc->indent >= enc->json.max_depth)
305	croak ("data structure too deep (hit recursion limit)");	374	croak (ERR_NESTING_EXCEEDED);
306		375
307	encode_ch (enc, '['); encode_nl (enc);	376	encode_ch (enc, '[');
308	++enc->indent;	377
		378	if (len >= 0)
		379	{
		380	encode_nl (enc); ++enc->indent;
309		381
310	for (i = 0; i <= len; ++i)	382	for (i = 0; i <= len; ++i)
311	{	383	{
312	SV **svp = av_fetch (av, i, 0);	384	SV **svp = av_fetch (av, i, 0);
313		385
314	encode_indent (enc);	386	encode_indent (enc);
315		387
316	if (svp)	388	if (svp)
317	encode_sv (enc, *svp);	389	encode_sv (enc, *svp);
318	else	390	else
319	encode_str (enc, "null", 4, 0);	391	encode_str (enc, "null", 4, 0);
320		392
321	if (i < len)	393	if (i < len)
322	encode_comma (enc);	394	encode_comma (enc);
323	}	395	}
324		396
		397	encode_nl (enc); --enc->indent; encode_indent (enc);
		398	}
		399
325	encode_nl (enc);	400	encode_ch (enc, ']');
326
327	--enc->indent;
328	encode_indent (enc); encode_ch (enc, ']');
329	}	401	}
330		402
331	static void	403	static void
332	encode_he (enc_t enc, HE he)	404	encode_hk (enc_t enc, HE he)
333	{	405	{
334	encode_ch (enc, '"');	406	encode_ch (enc, '"');
335		407
336	if (HeKLEN (he) == HEf_SVKEY)	408	if (HeKLEN (he) == HEf_SVKEY)
337	{	409	{
…		…
350	encode_ch (enc, '"');	422	encode_ch (enc, '"');
351		423
352	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	424	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
353	encode_ch (enc, ':');	425	encode_ch (enc, ':');
354	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	426	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
355	encode_sv (enc, HeVAL (he));
356	}	427	}
357		428
358	// compare hash entries, used when all keys are bytestrings	429	// compare hash entries, used when all keys are bytestrings
359	static int	430	static int
360	he_cmp_fast (const void a_, const void b_)	431	he_cmp_fast (const void a_, const void b_)
…		…
365	HE b = (HE **)b_;	436	HE b = (HE **)b_;
366		437
367	STRLEN la = HeKLEN (a);	438	STRLEN la = HeKLEN (a);
368	STRLEN lb = HeKLEN (b);	439	STRLEN lb = HeKLEN (b);
369		440
370	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	441	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
371	cmp = la - lb;	442	cmp = lb - la;
372		443
373	return cmp;	444	return cmp;
374	}	445	}
375		446
376	// compare hash entries, used when some keys are sv's or utf-x	447	// compare hash entries, used when some keys are sv's or utf-x
377	static int	448	static int
378	he_cmp_slow (const void a, const void b)	449	he_cmp_slow (const void a, const void b)
379	{	450	{
380	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	451	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
381	}	452	}
382		453
383	static void	454	static void
384	encode_hv (enc_t enc, HV hv)	455	encode_hv (enc_t enc, HV hv)
385	{	456	{
386	int count, i;	457	HE *he;
387		458
388	if (enc->indent >= enc->maxdepth)	459	if (enc->indent >= enc->json.max_depth)
389	croak ("data structure too deep (hit recursion limit)");	460	croak (ERR_NESTING_EXCEEDED);
390		461
391	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	462	encode_ch (enc, '{');
392		463
393	if ((count = hv_iterinit (hv)))
394	{
395	// for canonical output we have to sort by keys first	464	// for canonical output we have to sort by keys first
396	// actually, this is mostly due to the stupid so-called	465	// actually, this is mostly due to the stupid so-called
397	// security workaround added somewhere in 5.8.x.	466	// security workaround added somewhere in 5.8.x.
398	// that randomises hash orderings	467	// that randomises hash orderings
399	if (enc->json.flags & F_CANONICAL)	468	if (enc->json.flags & F_CANONICAL)
		469	{
		470	int count = hv_iterinit (hv);
		471
		472	if (SvMAGICAL (hv))
400	{	473	{
		474	// need to count by iterating. could improve by dynamically building the vector below
		475	// but I don't care for the speed of this special case.
		476	// note also that we will run into undefined behaviour when the two iterations
		477	// do not result in the same count, something I might care for in some later release.
		478
		479	count = 0;
		480	while (hv_iternext (hv))
		481	++count;
		482
		483	hv_iterinit (hv);
		484	}
		485
		486	if (count)
		487	{
401	int fast = 1;	488	int i, fast = 1;
402	HE *he;
403	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	489	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
404	HE *hes = _alloca (count sizeof (HE));	490	HE *hes = _alloca (count sizeof (HE));
405	#else	491	#else
406	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	492	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
407	#endif	493	#endif
…		…
434		520
435	FREETMPS;	521	FREETMPS;
436	LEAVE;	522	LEAVE;
437	}	523	}
438		524
439	for (i = 0; i < count; ++i)	525	encode_nl (enc); ++enc->indent;
		526
		527	while (count--)
440	{	528	{
441	encode_indent (enc);	529	encode_indent (enc);
		530	he = hes [count];
442	encode_he (enc, hes [i]);	531	encode_hk (enc, he);
		532	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
443		533
444	if (i < count - 1)	534	if (count)
445	encode_comma (enc);	535	encode_comma (enc);
446	}	536	}
447		537
448	encode_nl (enc);	538	encode_nl (enc); --enc->indent; encode_indent (enc);
449	}	539	}
		540	}
450	else	541	else
		542	{
		543	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		544	if ((he = hv_iternext (hv)))
451	{	545	{
452	HE *he = hv_iternext (hv);	546	encode_nl (enc); ++enc->indent;
453		547
454	for (;;)	548	for (;;)
455	{	549	{
456	encode_indent (enc);	550	encode_indent (enc);
457	encode_he (enc, he);	551	encode_hk (enc, he);
		552	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
458		553
459	if (!(he = hv_iternext (hv)))	554	if (!(he = hv_iternext (hv)))
460	break;	555	break;
461		556
462	encode_comma (enc);	557	encode_comma (enc);
463	}	558	}
464		559
465	encode_nl (enc);	560	encode_nl (enc); --enc->indent; encode_indent (enc);
466	}	561	}
467	}	562	}
468		563
469	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	564	encode_ch (enc, '}');
470	}	565	}
471		566
472	// encode objects, arrays and special \0=false and \1=true values.	567	// encode objects, arrays and special \0=false and \1=true values.
473	static void	568	static void
474	encode_rv (enc_t enc, SV sv)	569	encode_rv (enc_t enc, SV sv)
…		…
478	SvGETMAGIC (sv);	573	SvGETMAGIC (sv);
479	svt = SvTYPE (sv);	574	svt = SvTYPE (sv);
480		575
481	if (expect_false (SvOBJECT (sv)))	576	if (expect_false (SvOBJECT (sv)))
482	{	577	{
		578	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		579	? json_boolean_stash
		580	: gv_stashpv ("JSON::XS::Boolean", 1);
		581
483	if (SvSTASH (sv) == json_boolean_stash)	582	if (SvSTASH (sv) == stash)
484	{	583	{
485	if (SvIV (sv))	584	if (SvIV (sv))
486	encode_str (enc, "true", 4, 0);	585	encode_str (enc, "true", 4, 0);
487	else	586	else
488	encode_str (enc, "false", 5, 0);	587	encode_str (enc, "false", 5, 0);
…		…
500	// we re-bless the reference to get overload and other niceties right	599	// we re-bless the reference to get overload and other niceties right
501	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);	600	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
502		601
503	if (to_json)	602	if (to_json)
504	{	603	{
505	int count;
506	dSP;	604	dSP;
507		605
508	ENTER; SAVETMPS; PUSHMARK (SP);	606	ENTER; SAVETMPS; PUSHMARK (SP);
509	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	607	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
510		608
…		…
548		646
549	if (len == 1 && *pv == '1')	647	if (len == 1 && *pv == '1')
550	encode_str (enc, "true", 4, 0);	648	encode_str (enc, "true", 4, 0);
551	else if (len == 1 && *pv == '0')	649	else if (len == 1 && *pv == '0')
552	encode_str (enc, "false", 5, 0);	650	encode_str (enc, "false", 5, 0);
		651	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		652	encode_str (enc, "null", 4, 0);
553	else	653	else
554	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	654	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
555	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	655	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
556	}	656	}
		657	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		658	encode_str (enc, "null", 4, 0);
557	else	659	else
558	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	660	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
559	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	661	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
560	}	662	}
561		663
…		…
579	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	681	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
580	enc->cur += strlen (enc->cur);	682	enc->cur += strlen (enc->cur);
581	}	683	}
582	else if (SvIOKp (sv))	684	else if (SvIOKp (sv))
583	{	685	{
584	// we assume we can always read an IV as a UV	686	// we assume we can always read an IV as a UV and vice versa
585	if (SvUV (sv) & ~(UV)0x7fff)	687	// we assume two's complement
586	{	688	// we assume no aliasing issues in the union
587	// large integer, use the (rather slow) snprintf way.	689	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
588	need (enc, sizeof (UV) * 3);	690	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
589	enc->cur +=
590	SvIsUV(sv)
591	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
592	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
593	}
594	else
595	{	691	{
596	// optimise the "small number case"	692	// optimise the "small number case"
597	// code will likely be branchless and use only a single multiplication	693	// code will likely be branchless and use only a single multiplication
		694	// works for numbers up to 59074
598	I32 i = SvIV (sv);	695	I32 i = SvIVX (sv);
599	U32 u;	696	U32 u;
600	char digit, nz = 0;	697	char digit, nz = 0;
601		698
602	need (enc, 6);	699	need (enc, 6);
603		700
…		…
609		706
610	// now output digit by digit, each time masking out the integer part	707	// now output digit by digit, each time masking out the integer part
611	// and multiplying by 5 while moving the decimal point one to the right,	708	// and multiplying by 5 while moving the decimal point one to the right,
612	// resulting in a net multiplication by 10.	709	// resulting in a net multiplication by 10.
613	// we always write the digit to memory but conditionally increment	710	// we always write the digit to memory but conditionally increment
614	// the pointer, to ease the usage of conditional move instructions.	711	// the pointer, to enable the use of conditional move instructions.
615	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	712	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
616	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	713	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
617	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	714	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
618	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	715	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
619	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	716	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
620	}	717	}
		718	else
		719	{
		720	// large integer, use the (rather slow) snprintf way.
		721	need (enc, IVUV_MAXCHARS);
		722	enc->cur +=
		723	SvIsUV(sv)
		724	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		725	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		726	}
621	}	727	}
622	else if (SvROK (sv))	728	else if (SvROK (sv))
623	encode_rv (enc, SvRV (sv));	729	encode_rv (enc, SvRV (sv));
624	else if (!SvOK (sv))	730	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
625	encode_str (enc, "null", 4, 0);	731	encode_str (enc, "null", 4, 0);
626	else	732	else
627	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	733	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
628	SvPV_nolen (sv), SvFLAGS (sv));	734	SvPV_nolen (sv), SvFLAGS (sv));
629	}	735	}
…		…
639	enc.json = *json;	745	enc.json = *json;
640	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	746	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
641	enc.cur = SvPVX (enc.sv);	747	enc.cur = SvPVX (enc.sv);
642	enc.end = SvEND (enc.sv);	748	enc.end = SvEND (enc.sv);
643	enc.indent = 0;	749	enc.indent = 0;
644	enc.maxdepth = DEC_DEPTH (enc.json.flags);	750	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		751	: enc.json.flags & F_LATIN1 ? 0x000100UL
		752	: 0x110000UL;
645		753
646	SvPOK_only (enc.sv);	754	SvPOK_only (enc.sv);
647	encode_sv (&enc, scalar);	755	encode_sv (&enc, scalar);
648		756
649	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	757	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
670	JSON json;	778	JSON json;
671	U32 depth; // recursion depth	779	U32 depth; // recursion depth
672	U32 maxdepth; // recursion depth limit	780	U32 maxdepth; // recursion depth limit
673	} dec_t;	781	} dec_t;
674		782
675	inline void	783	INLINE void
		784	decode_comment (dec_t *dec)
		785	{
		786	// only '#'-style comments allowed a.t.m.
		787
		788	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		789	++dec->cur;
		790	}
		791
		792	INLINE void
676	decode_ws (dec_t *dec)	793	decode_ws (dec_t *dec)
677	{	794	{
678	for (;;)	795	for (;;)
679	{	796	{
680	char ch = *dec->cur;	797	char ch = *dec->cur;
681		798
682	if (ch > 0x20	799	if (ch > 0x20)
		800	{
		801	if (expect_false (ch == '#'))
		802	{
		803	if (dec->json.flags & F_RELAXED)
		804	decode_comment (dec);
		805	else
		806	break;
		807	}
		808	else
		809	break;
		810	}
683	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	811	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
684	break;	812	break; // parse error, but let higher level handle it, gives better error messages
685		813
686	++dec->cur;	814	++dec->cur;
687	}	815	}
688	}	816	}
689		817
…		…
693	if (*dec->cur != ch) \	821	if (*dec->cur != ch) \
694	ERR (# ch " expected"); \	822	ERR (# ch " expected"); \
695	++dec->cur; \	823	++dec->cur; \
696	SE	824	SE
697		825
698	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	826	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
699	#define DEC_DEC_DEPTH --dec->depth	827	#define DEC_DEC_DEPTH --dec->depth
700		828
701	static SV decode_sv (dec_t dec);	829	static SV decode_sv (dec_t dec);
702		830
703	static signed char decode_hexdigit[256];	831	static signed char decode_hexdigit[256];
…		…
795		923
796	if (hi >= 0x80)	924	if (hi >= 0x80)
797	{	925	{
798	utf8 = 1;	926	utf8 = 1;
799		927
800	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	928	cur = encode_utf8 (cur, hi);
801	}	929	}
802	else	930	else
803	*cur++ = hi;	931	*cur++ = hi;
804	}	932	}
805	break;	933	break;
…		…
807	default:	935	default:
808	--dec_cur;	936	--dec_cur;
809	ERR ("illegal backslash escape sequence in string");	937	ERR ("illegal backslash escape sequence in string");
810	}	938	}
811	}	939	}
812	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	940	else if (expect_true (ch >= 0x20 && ch < 0x80))
813	*cur++ = ch;	941	*cur++ = ch;
814	else if (ch >= 0x80)	942	else if (ch >= 0x80)
815	{	943	{
816	STRLEN clen;	944	STRLEN clen;
817	UV uch;	945	UV uch;
…		…
940		1068
941	if (!is_nv)	1069	if (!is_nv)
942	{	1070	{
943	int len = dec->cur - start;	1071	int len = dec->cur - start;
944		1072
945	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1073	// special case the rather common 1..5-digit-int case
946	if (*start == '-')	1074	if (*start == '-')
947	switch (len)	1075	switch (len)
948	{	1076	{
949	case 2: return newSViv (-( start [1] - '0' * 1));	1077	case 2: return newSViv (-( start [1] - '0' * 1));
950	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1078	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
951	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1079	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
952	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1080	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1081	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
953	}	1082	}
954	else	1083	else
955	switch (len)	1084	switch (len)
956	{	1085	{
957	case 1: return newSViv ( start [0] - '0' * 1);	1086	case 1: return newSViv ( start [0] - '0' * 1);
958	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1087	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
959	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1088	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
960	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1089	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1090	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
961	}	1091	}
962		1092
963	{	1093	{
964	UV uv;	1094	UV uv;
965	int numtype = grok_number (start, len, &uv);	1095	int numtype = grok_number (start, len, &uv);
…		…
1026		1156
1027	if (*dec->cur != ',')	1157	if (*dec->cur != ',')
1028	ERR (", or ] expected while parsing array");	1158	ERR (", or ] expected while parsing array");
1029		1159
1030	++dec->cur;	1160	++dec->cur;
		1161
		1162	decode_ws (dec);
		1163
		1164	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1165	{
		1166	++dec->cur;
		1167	break;
		1168	}
1031	}	1169	}
1032		1170
1033	DEC_DEC_DEPTH;	1171	DEC_DEC_DEPTH;
1034	return newRV_noinc ((SV *)av);	1172	return newRV_noinc ((SV *)av);
1035		1173
…		…
1051	if (*dec->cur == '}')	1189	if (*dec->cur == '}')
1052	++dec->cur;	1190	++dec->cur;
1053	else	1191	else
1054	for (;;)	1192	for (;;)
1055	{	1193	{
1056	decode_ws (dec); EXPECT_CH ('"');	1194	EXPECT_CH ('"');
1057		1195
1058	// heuristic: assume that	1196	// heuristic: assume that
1059	// a) decode_str + hv_store_ent are abysmally slow.	1197	// a) decode_str + hv_store_ent are abysmally slow.
1060	// b) most hash keys are short, simple ascii text.	1198	// b) most hash keys are short, simple ascii text.
1061	// => try to "fast-match" such strings to avoid	1199	// => try to "fast-match" such strings to avoid
…		…
1065	char *p = dec->cur;	1203	char *p = dec->cur;
1066	char *e = p + 24; // only try up to 24 bytes	1204	char *e = p + 24; // only try up to 24 bytes
1067		1205
1068	for (;;)	1206	for (;;)
1069	{	1207	{
1070	// the >= 0x80 is true on most architectures	1208	// the >= 0x80 is false on most architectures
1071	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1209	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1072	{	1210	{
1073	// slow path, back up and use decode_str	1211	// slow path, back up and use decode_str
1074	SV *key = decode_str (dec);	1212	SV *key = decode_str (dec);
1075	if (!key)	1213	if (!key)
1076	goto fail;	1214	goto fail;
1077		1215
1078	decode_ws (dec); EXPECT_CH (':');	1216	decode_ws (dec); EXPECT_CH (':');
1079		1217
		1218	decode_ws (dec);
1080	value = decode_sv (dec);	1219	value = decode_sv (dec);
1081	if (!value)	1220	if (!value)
1082	{	1221	{
1083	SvREFCNT_dec (key);	1222	SvREFCNT_dec (key);
1084	goto fail;	1223	goto fail;
…		…
1096	int len = p - key;	1235	int len = p - key;
1097	dec->cur = p + 1;	1236	dec->cur = p + 1;
1098		1237
1099	decode_ws (dec); EXPECT_CH (':');	1238	decode_ws (dec); EXPECT_CH (':');
1100		1239
		1240	decode_ws (dec);
1101	value = decode_sv (dec);	1241	value = decode_sv (dec);
1102	if (!value)	1242	if (!value)
1103	goto fail;	1243	goto fail;
1104		1244
1105	hv_store (hv, key, len, value, 0);	1245	hv_store (hv, key, len, value, 0);
…		…
1121		1261
1122	if (*dec->cur != ',')	1262	if (*dec->cur != ',')
1123	ERR (", or } expected while parsing object/hash");	1263	ERR (", or } expected while parsing object/hash");
1124		1264
1125	++dec->cur;	1265	++dec->cur;
		1266
		1267	decode_ws (dec);
		1268
		1269	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
1126	}	1274	}
1127		1275
1128	DEC_DEC_DEPTH;	1276	DEC_DEC_DEPTH;
1129	sv = newRV_noinc ((SV *)hv);	1277	sv = newRV_noinc ((SV *)hv);
1130		1278
…		…
1195	}	1343	}
1196		1344
1197	static SV *	1345	static SV *
1198	decode_sv (dec_t *dec)	1346	decode_sv (dec_t *dec)
1199	{	1347	{
1200	decode_ws (dec);
1201
1202	// the beauty of JSON: you need exactly one character lookahead	1348	// the beauty of JSON: you need exactly one character lookahead
1203	// to parse anything.	1349	// to parse everything.
1204	switch (*dec->cur)	1350	switch (*dec->cur)
1205	{	1351	{
1206	case '"': ++dec->cur; return decode_str (dec);	1352	case '"': ++dec->cur; return decode_str (dec);
1207	case '[': ++dec->cur; return decode_av (dec);	1353	case '[': ++dec->cur; return decode_av (dec);
1208	case '{': ++dec->cur; return decode_hv (dec);	1354	case '{': ++dec->cur; return decode_hv (dec);
1209		1355
1210	case '-':	1356	case '-':
1211	case '0': case '1': case '2': case '3': case '4':	1357	case '0': case '1': case '2': case '3': case '4':
1212	case '5': case '6': case '7': case '8': case '9':	1358	case '5': case '6': case '7': case '8': case '9':
1213	return decode_num (dec);	1359	return decode_num (dec);
1214		1360
1215	case 't':	1361	case 't':
1216	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1362	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1217	{	1363	{
1218	dec->cur += 4;	1364	dec->cur += 4;
		1365	#if JSON_SLOW
		1366	json_true = get_bool ("JSON::XS::true");
		1367	#endif
1219	return SvREFCNT_inc (json_true);	1368	return newSVsv (json_true);
1220	}	1369	}
1221	else	1370	else
1222	ERR ("'true' expected");	1371	ERR ("'true' expected");
1223		1372
1224	break;	1373	break;
1225		1374
1226	case 'f':	1375	case 'f':
1227	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1376	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1228	{	1377	{
1229	dec->cur += 5;	1378	dec->cur += 5;
		1379	#if JSON_SLOW
		1380	json_false = get_bool ("JSON::XS::false");
		1381	#endif
1230	return SvREFCNT_inc (json_false);	1382	return newSVsv (json_false);
1231	}	1383	}
1232	else	1384	else
1233	ERR ("'false' expected");	1385	ERR ("'false' expected");
1234		1386
1235	break;	1387	break;
…		…
1253	fail:	1405	fail:
1254	return 0;	1406	return 0;
1255	}	1407	}
1256		1408
1257	static SV *	1409	static SV *
1258	decode_json (SV string, JSON json, UV *offset_return)	1410	decode_json (SV string, JSON json, STRLEN *offset_return)
1259	{	1411	{
1260	dec_t dec;	1412	dec_t dec;
1261	UV offset;	1413	STRLEN offset;
1262	SV *sv;	1414	SV *sv;
1263		1415
1264	SvGETMAGIC (string);	1416	SvGETMAGIC (string);
1265	SvUPGRADE (string, SVt_PV);	1417	SvUPGRADE (string, SVt_PV);
1266		1418
1267	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1419	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1420	* assertion with -DDEBUGGING, although SvCUR is documented to
		1421	* return the xpv_cur field which certainly exists after upgrading.
		1422	* according to nicholas clark, calling SvPOK fixes this.
		1423	*/
		1424	SvPOK (string);
		1425
		1426	if (SvCUR (string) > json->max_size && json->max_size)
1268	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1427	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1269	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1428	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
1270		1429
1271	if (json->flags & F_UTF8)	1430	if (json->flags & F_UTF8)
1272	sv_utf8_downgrade (string, 0);	1431	sv_utf8_downgrade (string, 0);
1273	else	1432	else
1274	sv_utf8_upgrade (string);	1433	sv_utf8_upgrade (string);
1275		1434
1276	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1435	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1277		1436
1278	dec.json = *json;	1437	dec.json = *json;
1279	dec.cur = SvPVX (string);	1438	dec.cur = SvPVX (string);
1280	dec.end = SvEND (string);	1439	dec.end = SvEND (string);
1281	dec.err = 0;	1440	dec.err = 0;
1282	dec.depth = 0;	1441	dec.depth = 0;
1283	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1284		1442
1285	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1443	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1286	dec.json.flags \|= F_HOOK;	1444	dec.json.flags \|= F_HOOK;
1287		1445
1288	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1446	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1447
		1448	decode_ws (&dec);
1289	sv = decode_sv (&dec);	1449	sv = decode_sv (&dec);
1290		1450
1291	if (!(offset_return \|\| !sv))	1451	if (!(offset_return \|\| !sv))
1292	{	1452	{
1293	// check for trailing garbage	1453	// check for trailing garbage
…		…
1337		1497
1338	return sv;	1498	return sv;
1339	}	1499	}
1340		1500
1341	/////////////////////////////////////////////////////////////////////////////	1501	/////////////////////////////////////////////////////////////////////////////
		1502	// incremental parser
		1503
		1504	static void
		1505	incr_parse (JSON *self)
		1506	{
		1507	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1508
		1509	for (;;)
		1510	{
		1511	//printf ("loop pod %d p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), p, p, self->incr_mode, self->incr_nest);//D
		1512	switch (self->incr_mode)
		1513	{
		1514	// only used for intiial whitespace skipping
		1515	case INCR_M_WS:
		1516	for (;;)
		1517	{
		1518	if (*p > 0x20)
		1519	{
		1520	self->incr_mode = INCR_M_JSON;
		1521	goto incr_m_json;
		1522	}
		1523	else if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	}
		1528
		1529	// skip a single char inside a string (for \\-processing)
		1530	case INCR_M_BS:
		1531	if (!*p)
		1532	goto interrupt;
		1533
		1534	++p;
		1535	self->incr_mode = INCR_M_STR;
		1536	goto incr_m_str;
		1537
		1538	// inside a string
		1539	case INCR_M_STR:
		1540	incr_m_str:
		1541	for (;;)
		1542	{
		1543	if (*p == '"')
		1544	{
		1545	++p;
		1546	self->incr_mode = INCR_M_JSON;
		1547
		1548	if (!self->incr_nest)
		1549	goto interrupt;
		1550
		1551	goto incr_m_json;
		1552	}
		1553	else if (*p == '\\')
		1554	{
		1555	++p; // "virtually" consumes character after \
		1556
		1557	if (!*p) // if at end of string we have to switch modes
		1558	{
		1559	self->incr_mode = INCR_M_BS;
		1560	goto interrupt;
		1561	}
		1562	}
		1563	else if (!*p)
		1564	goto interrupt;
		1565
		1566	++p;
		1567	}
		1568
		1569	// after initial ws, outside string
		1570	case INCR_M_JSON:
		1571	incr_m_json:
		1572	for (;;)
		1573	{
		1574	switch (*p++)
		1575	{
		1576	case 0:
		1577	--p;
		1578	goto interrupt;
		1579
		1580	case 0x09:
		1581	case 0x0a:
		1582	case 0x0d:
		1583	case 0x20:
		1584	if (!self->incr_nest)
		1585	{
		1586	--p; // do not eat the whitespace, let the next round do it
		1587	goto interrupt;
		1588	}
		1589	break;
		1590
		1591	case '"':
		1592	self->incr_mode = INCR_M_STR;
		1593	goto incr_m_str;
		1594
		1595	case '[':
		1596	case '{':
		1597	if (++self->incr_nest > self->max_depth)
		1598	croak (ERR_NESTING_EXCEEDED);
		1599	break;
		1600
		1601	case ']':
		1602	case '}':
		1603	if (--self->incr_nest <= 0)
		1604	goto interrupt;
		1605	}
		1606	}
		1607	}
		1608
		1609	modechange:
		1610	;
		1611	}
		1612
		1613	interrupt:
		1614	self->incr_pos = p - SvPVX (self->incr_text);
		1615	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1616	}
		1617
		1618	/////////////////////////////////////////////////////////////////////////////
1342	// XS interface functions	1619	// XS interface functions
1343		1620
1344	MODULE = JSON::XS PACKAGE = JSON::XS	1621	MODULE = JSON::XS PACKAGE = JSON::XS
1345		1622
1346	BOOT:	1623	BOOT:
…		…
1355	: -1;	1632	: -1;
1356		1633
1357	json_stash = gv_stashpv ("JSON::XS" , 1);	1634	json_stash = gv_stashpv ("JSON::XS" , 1);
1358	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1635	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1359		1636
1360	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1637	json_true = get_bool ("JSON::XS::true");
1361	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1638	json_false = get_bool ("JSON::XS::false");
1362	}	1639	}
1363		1640
1364	PROTOTYPES: DISABLE	1641	PROTOTYPES: DISABLE
		1642
		1643	void CLONE (...)
		1644	CODE:
		1645	json_stash = 0;
		1646	json_boolean_stash = 0;
1365		1647
1366	void new (char *klass)	1648	void new (char *klass)
1367	PPCODE:	1649	PPCODE:
1368	{	1650	{
1369	SV *pv = NEWSV (0, sizeof (JSON));	1651	SV *pv = NEWSV (0, sizeof (JSON));
1370	SvPOK_only (pv);	1652	SvPOK_only (pv);
1371	Zero (SvPVX (pv), 1, JSON);	1653	json_init ((JSON *)SvPVX (pv));
1372	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1654	XPUSHs (sv_2mortal (sv_bless (
1373	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1655	newRV_noinc (pv),
		1656	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1657	)));
1374	}	1658	}
1375		1659
1376	void ascii (JSON *self, int enable = 1)	1660	void ascii (JSON *self, int enable = 1)
1377	ALIAS:	1661	ALIAS:
1378	ascii = F_ASCII	1662	ascii = F_ASCII
…		…
1385	pretty = F_PRETTY	1669	pretty = F_PRETTY
1386	allow_nonref = F_ALLOW_NONREF	1670	allow_nonref = F_ALLOW_NONREF
1387	shrink = F_SHRINK	1671	shrink = F_SHRINK
1388	allow_blessed = F_ALLOW_BLESSED	1672	allow_blessed = F_ALLOW_BLESSED
1389	convert_blessed = F_CONV_BLESSED	1673	convert_blessed = F_CONV_BLESSED
		1674	relaxed = F_RELAXED
		1675	allow_unknown = F_ALLOW_UNKNOWN
1390	PPCODE:	1676	PPCODE:
1391	{	1677	{
1392	if (enable)	1678	if (enable)
1393	self->flags \|= ix;	1679	self->flags \|= ix;
1394	else	1680	else
1395	self->flags &= ~ix;	1681	self->flags &= ~ix;
1396		1682
1397	XPUSHs (ST (0));	1683	XPUSHs (ST (0));
1398	}	1684	}
1399		1685
1400	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1686	void get_ascii (JSON *self)
		1687	ALIAS:
		1688	get_ascii = F_ASCII
		1689	get_latin1 = F_LATIN1
		1690	get_utf8 = F_UTF8
		1691	get_indent = F_INDENT
		1692	get_canonical = F_CANONICAL
		1693	get_space_before = F_SPACE_BEFORE
		1694	get_space_after = F_SPACE_AFTER
		1695	get_allow_nonref = F_ALLOW_NONREF
		1696	get_shrink = F_SHRINK
		1697	get_allow_blessed = F_ALLOW_BLESSED
		1698	get_convert_blessed = F_CONV_BLESSED
		1699	get_relaxed = F_RELAXED
		1700	get_allow_unknown = F_ALLOW_UNKNOWN
1401	PPCODE:	1701	PPCODE:
1402	{	1702	XPUSHs (boolSV (self->flags & ix));
1403	UV log2 = 0;
1404		1703
1405	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1704	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1406		1705	PPCODE:
1407	while ((1UL << log2) < max_depth)	1706	self->max_depth = max_depth;
1408	++log2;
1409
1410	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1411
1412	XPUSHs (ST (0));	1707	XPUSHs (ST (0));
1413	}
1414		1708
		1709	U32 get_max_depth (JSON *self)
		1710	CODE:
		1711	RETVAL = self->max_depth;
		1712	OUTPUT:
		1713	RETVAL
		1714
1415	void max_size (JSON *self, UV max_size = 0)	1715	void max_size (JSON *self, U32 max_size = 0)
1416	PPCODE:	1716	PPCODE:
1417	{	1717	self->max_size = max_size;
1418	UV log2 = 0;
1419
1420	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1421	if (max_size == 1) max_size = 2;
1422
1423	while ((1UL << log2) < max_size)
1424	++log2;
1425
1426	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1427
1428	XPUSHs (ST (0));	1718	XPUSHs (ST (0));
1429	}	1719
		1720	int get_max_size (JSON *self)
		1721	CODE:
		1722	RETVAL = self->max_size;
		1723	OUTPUT:
		1724	RETVAL
1430		1725
1431	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1726	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1432	PPCODE:	1727	PPCODE:
1433	{	1728	{
1434	SvREFCNT_dec (self->cb_object);	1729	SvREFCNT_dec (self->cb_object);
…		…
1468	XPUSHs (decode_json (jsonstr, self, 0));	1763	XPUSHs (decode_json (jsonstr, self, 0));
1469		1764
1470	void decode_prefix (JSON self, SV jsonstr)	1765	void decode_prefix (JSON self, SV jsonstr)
1471	PPCODE:	1766	PPCODE:
1472	{	1767	{
1473	UV offset;	1768	STRLEN offset;
1474	EXTEND (SP, 2);	1769	EXTEND (SP, 2);
1475	PUSHs (decode_json (jsonstr, self, &offset));	1770	PUSHs (decode_json (jsonstr, self, &offset));
1476	PUSHs (sv_2mortal (newSVuv (offset)));	1771	PUSHs (sv_2mortal (newSVuv (offset)));
		1772	}
		1773
		1774	void incr_parse (JSON self, SV jsonstr = 0)
		1775	PPCODE:
		1776	{
		1777	if (!self->incr_text)
		1778	self->incr_text = newSVpvn ("", 0);
		1779
		1780	// append data, if any
		1781	if (jsonstr)
		1782	{
		1783	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1784	{
		1785	/* utf-8-ness differs, need to upgrade */
		1786	sv_utf8_upgrade (self->incr_text);
		1787
		1788	if (self->incr_pos)
		1789	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1790	- (U8 *)SvPVX (self->incr_text);
		1791	}
		1792
		1793	{
		1794	STRLEN len;
		1795	const char *str = SvPV (jsonstr, len);
		1796	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1797	Move (str, SvEND (self->incr_text), len, char);
		1798	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1799	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1800	}
		1801	}
		1802
		1803	if (GIMME_V != G_VOID)
		1804	do
		1805	{
		1806	STRLEN offset;
		1807
		1808	if (!INCR_DONE (self))
		1809	{
		1810	incr_parse (self);
		1811
		1812	if (self->incr_pos > self->max_size && self->max_size)
		1813	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1814	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1815
		1816	if (!INCR_DONE (self))
		1817	break;
		1818	}
		1819
		1820	XPUSHs (decode_json (self->incr_text, self, &offset));
		1821
		1822	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1823	self->incr_pos -= offset;
		1824	self->incr_nest = 0;
		1825	self->incr_mode = 0;
		1826	}
		1827	while (GIMME_V == G_ARRAY);
		1828	}
		1829
		1830	SV incr_text (JSON self)
		1831	ATTRS: lvalue
		1832	CODE:
		1833	{
		1834	if (self->incr_pos)
		1835	croak ("incr_text can not be called when the incremental parser already started parsing");
		1836
		1837	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1838	}
		1839	OUTPUT:
		1840	RETVAL
		1841
		1842	void incr_skip (JSON *self)
		1843	CODE:
		1844	{
		1845	if (self->incr_pos)
		1846	{
		1847	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1848	self->incr_pos = 0;
		1849	self->incr_nest = 0;
		1850	self->incr_mode = 0;
		1851	}
		1852	}
		1853
		1854	void incr_reset (JSON *self)
		1855	CODE:
		1856	{
		1857	SvREFCNT_dec (self->incr_text);
		1858	self->incr_text = 0;
		1859	self->incr_pos = 0;
		1860	self->incr_nest = 0;
		1861	self->incr_mode = 0;
1477	}	1862	}
1478		1863
1479	void DESTROY (JSON *self)	1864	void DESTROY (JSON *self)
1480	CODE:	1865	CODE:
1481	SvREFCNT_dec (self->cb_sk_object);	1866	SvREFCNT_dec (self->cb_sk_object);
1482	SvREFCNT_dec (self->cb_object);	1867	SvREFCNT_dec (self->cb_object);
		1868	SvREFCNT_dec (self->incr_text);
1483		1869
1484	PROTOTYPES: ENABLE	1870	PROTOTYPES: ENABLE
1485		1871
1486	void to_json (SV *scalar)	1872	void encode_json (SV *scalar)
		1873	ALIAS:
		1874	to_json_ = 0
		1875	encode_json = F_UTF8
1487	PPCODE:	1876	PPCODE:
1488	{	1877	{
1489	JSON json = { F_DEFAULT \| F_UTF8 };	1878	JSON json;
		1879	json_init (&json);
		1880	json.flags \|= ix;
1490	XPUSHs (encode_json (scalar, &json));	1881	XPUSHs (encode_json (scalar, &json));
1491	}	1882	}
1492		1883
1493	void from_json (SV *jsonstr)	1884	void decode_json (SV *jsonstr)
		1885	ALIAS:
		1886	from_json_ = 0
		1887	decode_json = F_UTF8
1494	PPCODE:	1888	PPCODE:
1495	{	1889	{
1496	JSON json = { F_DEFAULT \| F_UTF8 };	1890	JSON json;
		1891	json_init (&json);
		1892	json.flags \|= ix;
1497	XPUSHs (decode_json (jsonstr, &json, 0));	1893	XPUSHs (decode_json (jsonstr, &json, 0));
1498	}	1894	}
1499		1895
		1896

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC vs. Revision 1.88 by root, Sun Jun 22 12:05:49 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.56 by root, Thu Jul 26 11:33:35 2007 UTC vs.
Revision 1.88 by root, Sun Jun 22 12:05:49 2008 UTC