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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs.
Revision 1.95 by root, Sun Feb 22 06:55:28 2009 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
63	#ifdef USE_ITHREADS	66	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
65	#else	69	#else
66	# define JSON_SLOW 0	70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
67	#endif	72	#endif
68		73
69	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
70	static SV json_true, json_false;	75	static SV json_true, json_false;
71		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
72	typedef struct {	86	typedef struct {
73	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
74	SV *cb_object;	91	SV *cb_object;
75	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;
76	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
77		107
78	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
79	// utility functions	109	// utility functions
80		110
81	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
82	shrink (SV *sv)	123	shrink (SV *sv)
83	{	124	{
84	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
85	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
86	{	128	{
87	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
88	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
89	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
95	// 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
96	// case of an error.	138	// case of an error.
97	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
98	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
99	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
100	inline UV	142	INLINE UV
101	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
102	{	144	{
103	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
106	{	148	{
107	*clen = 2;	149	*clen = 2;
108	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
109	}	151	}
110	else	152	else
111	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
112	*clen = (STRLEN)-1;	154	}
113	return (UV)-1;	155
114	}	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;
115	}	178	}
116		179
117	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
118	// encoder	181	// encoder
119		182
…		…
123	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
124	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
125	SV *sv; // result scalar	188	SV *sv; // result scalar
126	JSON json;	189	JSON json;
127	U32 indent; // indentation level	190	U32 indent; // indentation level
128	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
129	} enc_t;	192	} enc_t;
130		193
131	inline void	194	INLINE void
132	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
133	{	196	{
134	if (expect_false (enc->cur + len >= enc->end))	197	if (expect_false (enc->cur + len >= enc->end))
135	{	198	{
136	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
137	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
138	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
140	}	203	}
141	}	204	}
142		205
143	inline void	206	INLINE void
144	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
145	{	208	{
146	need (enc, 1);	209	need (enc, 1);
147	*enc->cur++ = ch;	210	*enc->cur++ = ch;
148	}	211	}
…		…
202	{	265	{
203	uch = ch;	266	uch = ch;
204	clen = 1;	267	clen = 1;
205	}	268	}
206		269
207	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
209
210	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
211	{	271	{
212	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
213	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
214	need (enc, len += 11);	277	need (enc, len += 11);
215	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
216	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
217	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
218	enc->cur += 12;	281	enc->cur += 12;
219	}	282	}
220	else	283	else
221	{	284	{
222	static char hexdigit [16] = "0123456789abcdef";
223	need (enc, len += 5);	285	need (enc, len += 5);
224	*enc->cur++ = '\\';	286	*enc->cur++ = '\\';
225	*enc->cur++ = 'u';	287	*enc->cur++ = 'u';
226	*enc->cur++ = hexdigit [ uch >> 12 ];	288	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
227	*enc->cur++ = hexdigit [(uch >> 8) & 15];	289	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
228	*enc->cur++ = hexdigit [(uch >> 4) & 15];	290	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
229	*enc->cur++ = hexdigit [(uch >> 0) & 15];	291	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
230	}	292	}
231		293
232	str += clen;	294	str += clen;
233	}	295	}
234	else if (enc->json.flags & F_LATIN1)	296	else if (enc->json.flags & F_LATIN1)
…		…
246	while (--clen);	308	while (--clen);
247	}	309	}
248	else	310	else
249	{	311	{
250	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
251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	313	enc->cur = encode_utf8 (enc->cur, uch);
252	++str;	314	++str;
253	}	315	}
254	}	316	}
255	}	317	}
256	}	318	}
257		319
258	--len;	320	--len;
259	}	321	}
260	}	322	}
261		323
262	inline void	324	INLINE void
263	encode_indent (enc_t *enc)	325	encode_indent (enc_t *enc)
264	{	326	{
265	if (enc->json.flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
266	{	328	{
267	int spaces = enc->indent * INDENT_STEP;	329	int spaces = enc->indent * INDENT_STEP;
…		…
270	memset (enc->cur, ' ', spaces);	332	memset (enc->cur, ' ', spaces);
271	enc->cur += spaces;	333	enc->cur += spaces;
272	}	334	}
273	}	335	}
274		336
275	inline void	337	INLINE void
276	encode_space (enc_t *enc)	338	encode_space (enc_t *enc)
277	{	339	{
278	need (enc, 1);	340	need (enc, 1);
279	encode_ch (enc, ' ');	341	encode_ch (enc, ' ');
280	}	342	}
281		343
282	inline void	344	INLINE void
283	encode_nl (enc_t *enc)	345	encode_nl (enc_t *enc)
284	{	346	{
285	if (enc->json.flags & F_INDENT)	347	if (enc->json.flags & F_INDENT)
286	{	348	{
287	need (enc, 1);	349	need (enc, 1);
288	encode_ch (enc, '\n');	350	encode_ch (enc, '\n');
289	}	351	}
290	}	352	}
291		353
292	inline void	354	INLINE void
293	encode_comma (enc_t *enc)	355	encode_comma (enc_t *enc)
294	{	356	{
295	encode_ch (enc, ',');	357	encode_ch (enc, ',');
296		358
297	if (enc->json.flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
…		…
305	static void	367	static void
306	encode_av (enc_t enc, AV av)	368	encode_av (enc_t enc, AV av)
307	{	369	{
308	int i, len = av_len (av);	370	int i, len = av_len (av);
309		371
310	if (enc->indent >= enc->maxdepth)	372	if (enc->indent >= enc->json.max_depth)
311	croak ("data structure too deep (hit recursion limit)");	373	croak (ERR_NESTING_EXCEEDED);
312		374
313	encode_ch (enc, '['); encode_nl (enc);	375	encode_ch (enc, '[');
314	++enc->indent;	376
		377	if (len >= 0)
		378	{
		379	encode_nl (enc); ++enc->indent;
315		380
316	for (i = 0; i <= len; ++i)	381	for (i = 0; i <= len; ++i)
317	{	382	{
318	SV **svp = av_fetch (av, i, 0);	383	SV **svp = av_fetch (av, i, 0);
319		384
320	encode_indent (enc);	385	encode_indent (enc);
321		386
322	if (svp)	387	if (svp)
323	encode_sv (enc, *svp);	388	encode_sv (enc, *svp);
324	else	389	else
325	encode_str (enc, "null", 4, 0);	390	encode_str (enc, "null", 4, 0);
326		391
327	if (i < len)	392	if (i < len)
328	encode_comma (enc);	393	encode_comma (enc);
329	}	394	}
330		395
		396	encode_nl (enc); --enc->indent; encode_indent (enc);
		397	}
		398
331	encode_nl (enc);	399	encode_ch (enc, ']');
332
333	--enc->indent;
334	encode_indent (enc); encode_ch (enc, ']');
335	}	400	}
336		401
337	static void	402	static void
338	encode_he (enc_t enc, HE he)	403	encode_hk (enc_t enc, HE he)
339	{	404	{
340	encode_ch (enc, '"');	405	encode_ch (enc, '"');
341		406
342	if (HeKLEN (he) == HEf_SVKEY)	407	if (HeKLEN (he) == HEf_SVKEY)
343	{	408	{
…		…
356	encode_ch (enc, '"');	421	encode_ch (enc, '"');
357		422
358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	423	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
359	encode_ch (enc, ':');	424	encode_ch (enc, ':');
360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	425	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
361	encode_sv (enc, HeVAL (he));
362	}	426	}
363		427
364	// compare hash entries, used when all keys are bytestrings	428	// compare hash entries, used when all keys are bytestrings
365	static int	429	static int
366	he_cmp_fast (const void a_, const void b_)	430	he_cmp_fast (const void a_, const void b_)
…		…
371	HE b = (HE **)b_;	435	HE b = (HE **)b_;
372		436
373	STRLEN la = HeKLEN (a);	437	STRLEN la = HeKLEN (a);
374	STRLEN lb = HeKLEN (b);	438	STRLEN lb = HeKLEN (b);
375		439
376	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	440	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
377	cmp = la - lb;	441	cmp = lb - la;
378		442
379	return cmp;	443	return cmp;
380	}	444	}
381		445
382	// 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
383	static int	447	static int
384	he_cmp_slow (const void a, const void b)	448	he_cmp_slow (const void a, const void b)
385	{	449	{
386	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	450	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
387	}	451	}
388		452
389	static void	453	static void
390	encode_hv (enc_t enc, HV hv)	454	encode_hv (enc_t enc, HV hv)
391	{	455	{
392	int count, i;	456	HE *he;
393		457
394	if (enc->indent >= enc->maxdepth)	458	if (enc->indent >= enc->json.max_depth)
395	croak ("data structure too deep (hit recursion limit)");	459	croak (ERR_NESTING_EXCEEDED);
396		460
397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	461	encode_ch (enc, '{');
398		462
399	if ((count = hv_iterinit (hv)))
400	{
401	// for canonical output we have to sort by keys first	463	// for canonical output we have to sort by keys first
402	// actually, this is mostly due to the stupid so-called	464	// actually, this is mostly due to the stupid so-called
403	// security workaround added somewhere in 5.8.x.	465	// security workaround added somewhere in 5.8.x
404	// that randomises hash orderings	466	// that randomises hash orderings
405	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))
406	{	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	{
407	int fast = 1;	487	int i, fast = 1;
408	HE *he;
409	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	488	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
410	HE *hes = _alloca (count sizeof (HE));	489	HE *hes = _alloca (count sizeof (HE));
411	#else	490	#else
412	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
413	#endif	492	#endif
…		…
440		519
441	FREETMPS;	520	FREETMPS;
442	LEAVE;	521	LEAVE;
443	}	522	}
444		523
445	for (i = 0; i < count; ++i)	524	encode_nl (enc); ++enc->indent;
		525
		526	while (count--)
446	{	527	{
447	encode_indent (enc);	528	encode_indent (enc);
		529	he = hes [count];
448	encode_he (enc, hes [i]);	530	encode_hk (enc, he);
		531	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
449		532
450	if (i < count - 1)	533	if (count)
451	encode_comma (enc);	534	encode_comma (enc);
452	}	535	}
453		536
454	encode_nl (enc);	537	encode_nl (enc); --enc->indent; encode_indent (enc);
455	}	538	}
		539	}
456	else	540	else
		541	{
		542	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		543	if ((he = hv_iternext (hv)))
457	{	544	{
458	HE *he = hv_iternext (hv);	545	encode_nl (enc); ++enc->indent;
459		546
460	for (;;)	547	for (;;)
461	{	548	{
462	encode_indent (enc);	549	encode_indent (enc);
463	encode_he (enc, he);	550	encode_hk (enc, he);
		551	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
464		552
465	if (!(he = hv_iternext (hv)))	553	if (!(he = hv_iternext (hv)))
466	break;	554	break;
467		555
468	encode_comma (enc);	556	encode_comma (enc);
469	}	557	}
470		558
471	encode_nl (enc);	559	encode_nl (enc); --enc->indent; encode_indent (enc);
472	}	560	}
473	}	561	}
474		562
475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	563	encode_ch (enc, '}');
476	}	564	}
477		565
478	// encode objects, arrays and special \0=false and \1=true values.	566	// encode objects, arrays and special \0=false and \1=true values.
479	static void	567	static void
480	encode_rv (enc_t enc, SV sv)	568	encode_rv (enc_t enc, SV sv)
…		…
557		645
558	if (len == 1 && *pv == '1')	646	if (len == 1 && *pv == '1')
559	encode_str (enc, "true", 4, 0);	647	encode_str (enc, "true", 4, 0);
560	else if (len == 1 && *pv == '0')	648	else if (len == 1 && *pv == '0')
561	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);
562	else	652	else
563	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",
564	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	654	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
565	}	655	}
		656	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		657	encode_str (enc, "null", 4, 0);
566	else	658	else
567	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",
568	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	660	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
569	}	661	}
570		662
…		…
588	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	680	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
589	enc->cur += strlen (enc->cur);	681	enc->cur += strlen (enc->cur);
590	}	682	}
591	else if (SvIOKp (sv))	683	else if (SvIOKp (sv))
592	{	684	{
593	// 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
594	if (SvUV (sv) & ~(UV)0x7fff)	686	// we assume two's complement
595	{	687	// we assume no aliasing issues in the union
596	// large integer, use the (rather slow) snprintf way.	688	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
597	need (enc, sizeof (UV) * 3);	689	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
598	enc->cur +=
599	SvIsUV(sv)
600	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
601	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
602	}
603	else
604	{	690	{
605	// optimise the "small number case"	691	// optimise the "small number case"
606	// 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
607	I32 i = SvIV (sv);	694	I32 i = SvIVX (sv);
608	U32 u;	695	U32 u;
609	char digit, nz = 0;	696	char digit, nz = 0;
610		697
611	need (enc, 6);	698	need (enc, 6);
612		699
…		…
618		705
619	// 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
620	// 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,
621	// resulting in a net multiplication by 10.	708	// resulting in a net multiplication by 10.
622	// we always write the digit to memory but conditionally increment	709	// we always write the digit to memory but conditionally increment
623	// the pointer, to ease the usage of conditional move instructions.	710	// the pointer, to enable the use of conditional move instructions.
624	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;
625	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;
626	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;
627	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;
628	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'
629	}	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	}
630	}	726	}
631	else if (SvROK (sv))	727	else if (SvROK (sv))
632	encode_rv (enc, SvRV (sv));	728	encode_rv (enc, SvRV (sv));
633	else if (!SvOK (sv))	729	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
634	encode_str (enc, "null", 4, 0);	730	encode_str (enc, "null", 4, 0);
635	else	731	else
636	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",
637	SvPV_nolen (sv), SvFLAGS (sv));	733	SvPV_nolen (sv), SvFLAGS (sv));
638	}	734	}
…		…
648	enc.json = *json;	744	enc.json = *json;
649	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	745	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
650	enc.cur = SvPVX (enc.sv);	746	enc.cur = SvPVX (enc.sv);
651	enc.end = SvEND (enc.sv);	747	enc.end = SvEND (enc.sv);
652	enc.indent = 0;	748	enc.indent = 0;
653	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;
654		752
655	SvPOK_only (enc.sv);	753	SvPOK_only (enc.sv);
656	encode_sv (&enc, scalar);	754	encode_sv (&enc, scalar);
657		755
658	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	756	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
679	JSON json;	777	JSON json;
680	U32 depth; // recursion depth	778	U32 depth; // recursion depth
681	U32 maxdepth; // recursion depth limit	779	U32 maxdepth; // recursion depth limit
682	} dec_t;	780	} dec_t;
683		781
684	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
685	decode_ws (dec_t *dec)	792	decode_ws (dec_t *dec)
686	{	793	{
687	for (;;)	794	for (;;)
688	{	795	{
689	char ch = *dec->cur;	796	char ch = *dec->cur;
690		797
691	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	}
692	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	810	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
693	break;	811	break; // parse error, but let higher level handle it, gives better error messages
694		812
695	++dec->cur;	813	++dec->cur;
696	}	814	}
697	}	815	}
698		816
…		…
702	if (*dec->cur != ch) \	820	if (*dec->cur != ch) \
703	ERR (# ch " expected"); \	821	ERR (# ch " expected"); \
704	++dec->cur; \	822	++dec->cur; \
705	SE	823	SE
706		824
707	#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)
708	#define DEC_DEC_DEPTH --dec->depth	826	#define DEC_DEC_DEPTH --dec->depth
709		827
710	static SV decode_sv (dec_t dec);	828	static SV decode_sv (dec_t dec);
711		829
712	static signed char decode_hexdigit[256];	830	static signed char decode_hexdigit[256];
…		…
804		922
805	if (hi >= 0x80)	923	if (hi >= 0x80)
806	{	924	{
807	utf8 = 1;	925	utf8 = 1;
808		926
809	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	927	cur = encode_utf8 (cur, hi);
810	}	928	}
811	else	929	else
812	*cur++ = hi;	930	*cur++ = hi;
813	}	931	}
814	break;	932	break;
…		…
816	default:	934	default:
817	--dec_cur;	935	--dec_cur;
818	ERR ("illegal backslash escape sequence in string");	936	ERR ("illegal backslash escape sequence in string");
819	}	937	}
820	}	938	}
821	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	939	else if (expect_true (ch >= 0x20 && ch < 0x80))
822	*cur++ = ch;	940	*cur++ = ch;
823	else if (ch >= 0x80)	941	else if (ch >= 0x80)
824	{	942	{
825	STRLEN clen;	943	STRLEN clen;
826	UV uch;	944	UV uch;
…		…
852	{	970	{
853	STRLEN len = cur - buf;	971	STRLEN len = cur - buf;
854		972
855	if (sv)	973	if (sv)
856	{	974	{
857	SvGROW (sv, SvCUR (sv) + len + 1);	975	STRLEN cur = SvCUR (sv);
		976
		977	if (SvLEN (sv) <= cur + len)
		978	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		979
858	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	980	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
859	SvCUR_set (sv, SvCUR (sv) + len);	981	SvCUR_set (sv, SvCUR (sv) + len);
860	}	982	}
861	else	983	else
862	sv = newSVpvn (buf, len);	984	sv = newSVpvn (buf, len);
…		…
949		1071
950	if (!is_nv)	1072	if (!is_nv)
951	{	1073	{
952	int len = dec->cur - start;	1074	int len = dec->cur - start;
953		1075
954	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1076	// special case the rather common 1..5-digit-int case
955	if (*start == '-')	1077	if (*start == '-')
956	switch (len)	1078	switch (len)
957	{	1079	{
958	case 2: return newSViv (-( start [1] - '0' * 1));	1080	case 2: return newSViv (-( start [1] - '0' * 1));
959	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1081	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
960	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1082	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
961	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1083	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1084	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
962	}	1085	}
963	else	1086	else
964	switch (len)	1087	switch (len)
965	{	1088	{
966	case 1: return newSViv ( start [0] - '0' * 1);	1089	case 1: return newSViv ( start [0] - '0' * 1);
967	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1090	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
968	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1091	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
969	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1092	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1093	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
970	}	1094	}
971		1095
972	{	1096	{
973	UV uv;	1097	UV uv;
974	int numtype = grok_number (start, len, &uv);	1098	int numtype = grok_number (start, len, &uv);
…		…
1035		1159
1036	if (*dec->cur != ',')	1160	if (*dec->cur != ',')
1037	ERR (", or ] expected while parsing array");	1161	ERR (", or ] expected while parsing array");
1038		1162
1039	++dec->cur;	1163	++dec->cur;
		1164
		1165	decode_ws (dec);
		1166
		1167	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1168	{
		1169	++dec->cur;
		1170	break;
		1171	}
1040	}	1172	}
1041		1173
1042	DEC_DEC_DEPTH;	1174	DEC_DEC_DEPTH;
1043	return newRV_noinc ((SV *)av);	1175	return newRV_noinc ((SV *)av);
1044		1176
…		…
1060	if (*dec->cur == '}')	1192	if (*dec->cur == '}')
1061	++dec->cur;	1193	++dec->cur;
1062	else	1194	else
1063	for (;;)	1195	for (;;)
1064	{	1196	{
1065	decode_ws (dec); EXPECT_CH ('"');	1197	EXPECT_CH ('"');
1066		1198
1067	// heuristic: assume that	1199	// heuristic: assume that
1068	// a) decode_str + hv_store_ent are abysmally slow.	1200	// a) decode_str + hv_store_ent are abysmally slow.
1069	// b) most hash keys are short, simple ascii text.	1201	// b) most hash keys are short, simple ascii text.
1070	// => try to "fast-match" such strings to avoid	1202	// => try to "fast-match" such strings to avoid
…		…
1074	char *p = dec->cur;	1206	char *p = dec->cur;
1075	char *e = p + 24; // only try up to 24 bytes	1207	char *e = p + 24; // only try up to 24 bytes
1076		1208
1077	for (;;)	1209	for (;;)
1078	{	1210	{
1079	// the >= 0x80 is true on most architectures	1211	// the >= 0x80 is false on most architectures
1080	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1212	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1081	{	1213	{
1082	// slow path, back up and use decode_str	1214	// slow path, back up and use decode_str
1083	SV *key = decode_str (dec);	1215	SV *key = decode_str (dec);
1084	if (!key)	1216	if (!key)
1085	goto fail;	1217	goto fail;
1086		1218
1087	decode_ws (dec); EXPECT_CH (':');	1219	decode_ws (dec); EXPECT_CH (':');
1088		1220
		1221	decode_ws (dec);
1089	value = decode_sv (dec);	1222	value = decode_sv (dec);
1090	if (!value)	1223	if (!value)
1091	{	1224	{
1092	SvREFCNT_dec (key);	1225	SvREFCNT_dec (key);
1093	goto fail;	1226	goto fail;
…		…
1105	int len = p - key;	1238	int len = p - key;
1106	dec->cur = p + 1;	1239	dec->cur = p + 1;
1107		1240
1108	decode_ws (dec); EXPECT_CH (':');	1241	decode_ws (dec); EXPECT_CH (':');
1109		1242
		1243	decode_ws (dec);
1110	value = decode_sv (dec);	1244	value = decode_sv (dec);
1111	if (!value)	1245	if (!value)
1112	goto fail;	1246	goto fail;
1113		1247
1114	hv_store (hv, key, len, value, 0);	1248	hv_store (hv, key, len, value, 0);
…		…
1130		1264
1131	if (*dec->cur != ',')	1265	if (*dec->cur != ',')
1132	ERR (", or } expected while parsing object/hash");	1266	ERR (", or } expected while parsing object/hash");
1133		1267
1134	++dec->cur;	1268	++dec->cur;
		1269
		1270	decode_ws (dec);
		1271
		1272	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1273	{
		1274	++dec->cur;
		1275	break;
		1276	}
1135	}	1277	}
1136		1278
1137	DEC_DEC_DEPTH;	1279	DEC_DEC_DEPTH;
1138	sv = newRV_noinc ((SV *)hv);	1280	sv = newRV_noinc ((SV *)hv);
1139		1281
…		…
1204	}	1346	}
1205		1347
1206	static SV *	1348	static SV *
1207	decode_sv (dec_t *dec)	1349	decode_sv (dec_t *dec)
1208	{	1350	{
1209	decode_ws (dec);
1210
1211	// the beauty of JSON: you need exactly one character lookahead	1351	// the beauty of JSON: you need exactly one character lookahead
1212	// to parse anything.	1352	// to parse everything.
1213	switch (*dec->cur)	1353	switch (*dec->cur)
1214	{	1354	{
1215	case '"': ++dec->cur; return decode_str (dec);	1355	case '"': ++dec->cur; return decode_str (dec);
1216	case '[': ++dec->cur; return decode_av (dec);	1356	case '[': ++dec->cur; return decode_av (dec);
1217	case '{': ++dec->cur; return decode_hv (dec);	1357	case '{': ++dec->cur; return decode_hv (dec);
1218		1358
1219	case '-':	1359	case '-':
1220	case '0': case '1': case '2': case '3': case '4':	1360	case '0': case '1': case '2': case '3': case '4':
1221	case '5': case '6': case '7': case '8': case '9':	1361	case '5': case '6': case '7': case '8': case '9':
1222	return decode_num (dec);	1362	return decode_num (dec);
1223		1363
1224	case 't':	1364	case 't':
1225	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1365	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1226	{	1366	{
1227	dec->cur += 4;	1367	dec->cur += 4;
		1368	#if JSON_SLOW
		1369	json_true = get_bool ("JSON::XS::true");
		1370	#endif
1228	return SvREFCNT_inc (json_true);	1371	return newSVsv (json_true);
1229	}	1372	}
1230	else	1373	else
1231	ERR ("'true' expected");	1374	ERR ("'true' expected");
1232		1375
1233	break;	1376	break;
1234		1377
1235	case 'f':	1378	case 'f':
1236	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1379	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1237	{	1380	{
1238	dec->cur += 5;	1381	dec->cur += 5;
		1382	#if JSON_SLOW
		1383	json_false = get_bool ("JSON::XS::false");
		1384	#endif
1239	return SvREFCNT_inc (json_false);	1385	return newSVsv (json_false);
1240	}	1386	}
1241	else	1387	else
1242	ERR ("'false' expected");	1388	ERR ("'false' expected");
1243		1389
1244	break;	1390	break;
…		…
1262	fail:	1408	fail:
1263	return 0;	1409	return 0;
1264	}	1410	}
1265		1411
1266	static SV *	1412	static SV *
1267	decode_json (SV string, JSON json, UV *offset_return)	1413	decode_json (SV string, JSON json, STRLEN *offset_return)
1268	{	1414	{
1269	dec_t dec;	1415	dec_t dec;
1270	UV offset;	1416	STRLEN offset;
1271	SV *sv;	1417	SV *sv;
1272		1418
		1419	/* work around bugs in 5.10 where manipulating magic values
		1420	* will perl ignore the magic in subsequent accesses
		1421	*/
1273	SvGETMAGIC (string);	1422	/SvGETMAGIC (string);/
		1423	if (SvMAGICAL (string))
		1424	string = sv_2mortal (newSVsv (string));
		1425
1274	SvUPGRADE (string, SVt_PV);	1426	SvUPGRADE (string, SVt_PV);
1275		1427
1276	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1428	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1429	* assertion with -DDEBUGGING, although SvCUR is documented to
		1430	* return the xpv_cur field which certainly exists after upgrading.
		1431	* according to nicholas clark, calling SvPOK fixes this.
		1432	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1433	* and hope for the best.
		1434	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1435	* assertion business is seriously broken, try yet another workaround
		1436	* for the broken -DDEBUGGING.
		1437	*/
		1438	#ifdef DEBUGGING
		1439	offset = SvOK (string) ? sv_len (string) : 0;
		1440	#else
		1441	offset = SvCUR (string);
		1442	#endif
		1443
		1444	if (offset > json->max_size && json->max_size)
1277	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1445	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1278	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1446	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
1279		1447
1280	if (json->flags & F_UTF8)	1448	if (json->flags & F_UTF8)
1281	sv_utf8_downgrade (string, 0);	1449	sv_utf8_downgrade (string, 0);
1282	else	1450	else
1283	sv_utf8_upgrade (string);	1451	sv_utf8_upgrade (string);
1284		1452
1285	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1453	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1286		1454
1287	dec.json = *json;	1455	dec.json = *json;
1288	dec.cur = SvPVX (string);	1456	dec.cur = SvPVX (string);
1289	dec.end = SvEND (string);	1457	dec.end = SvEND (string);
1290	dec.err = 0;	1458	dec.err = 0;
1291	dec.depth = 0;	1459	dec.depth = 0;
1292	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1293		1460
1294	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1461	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1295	dec.json.flags \|= F_HOOK;	1462	dec.json.flags \|= F_HOOK;
1296		1463
1297	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1464	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1465
		1466	decode_ws (&dec);
1298	sv = decode_sv (&dec);	1467	sv = decode_sv (&dec);
1299		1468
1300	if (!(offset_return \|\| !sv))	1469	if (!(offset_return \|\| !sv))
1301	{	1470	{
1302	// check for trailing garbage	1471	// check for trailing garbage
…		…
1346		1515
1347	return sv;	1516	return sv;
1348	}	1517	}
1349		1518
1350	/////////////////////////////////////////////////////////////////////////////	1519	/////////////////////////////////////////////////////////////////////////////
		1520	// incremental parser
		1521
		1522	static void
		1523	incr_parse (JSON *self)
		1524	{
		1525	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1526
		1527	for (;;)
		1528	{
		1529	//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
		1530	switch (self->incr_mode)
		1531	{
		1532	// only used for intiial whitespace skipping
		1533	case INCR_M_WS:
		1534	for (;;)
		1535	{
		1536	if (*p > 0x20)
		1537	{
		1538	self->incr_mode = INCR_M_JSON;
		1539	goto incr_m_json;
		1540	}
		1541	else if (!*p)
		1542	goto interrupt;
		1543
		1544	++p;
		1545	}
		1546
		1547	// skip a single char inside a string (for \\-processing)
		1548	case INCR_M_BS:
		1549	if (!*p)
		1550	goto interrupt;
		1551
		1552	++p;
		1553	self->incr_mode = INCR_M_STR;
		1554	goto incr_m_str;
		1555
		1556	// inside a string
		1557	case INCR_M_STR:
		1558	incr_m_str:
		1559	for (;;)
		1560	{
		1561	if (*p == '"')
		1562	{
		1563	++p;
		1564	self->incr_mode = INCR_M_JSON;
		1565
		1566	if (!self->incr_nest)
		1567	goto interrupt;
		1568
		1569	goto incr_m_json;
		1570	}
		1571	else if (*p == '\\')
		1572	{
		1573	++p; // "virtually" consumes character after \
		1574
		1575	if (!*p) // if at end of string we have to switch modes
		1576	{
		1577	self->incr_mode = INCR_M_BS;
		1578	goto interrupt;
		1579	}
		1580	}
		1581	else if (!*p)
		1582	goto interrupt;
		1583
		1584	++p;
		1585	}
		1586
		1587	// after initial ws, outside string
		1588	case INCR_M_JSON:
		1589	incr_m_json:
		1590	for (;;)
		1591	{
		1592	switch (*p++)
		1593	{
		1594	case 0:
		1595	--p;
		1596	goto interrupt;
		1597
		1598	case 0x09:
		1599	case 0x0a:
		1600	case 0x0d:
		1601	case 0x20:
		1602	if (!self->incr_nest)
		1603	{
		1604	--p; // do not eat the whitespace, let the next round do it
		1605	goto interrupt;
		1606	}
		1607	break;
		1608
		1609	case '"':
		1610	self->incr_mode = INCR_M_STR;
		1611	goto incr_m_str;
		1612
		1613	case '[':
		1614	case '{':
		1615	if (++self->incr_nest > self->max_depth)
		1616	croak (ERR_NESTING_EXCEEDED);
		1617	break;
		1618
		1619	case ']':
		1620	case '}':
		1621	if (--self->incr_nest <= 0)
		1622	goto interrupt;
		1623	}
		1624	}
		1625	}
		1626
		1627	modechange:
		1628	;
		1629	}
		1630
		1631	interrupt:
		1632	self->incr_pos = p - SvPVX (self->incr_text);
		1633	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1634	}
		1635
		1636	/////////////////////////////////////////////////////////////////////////////
1351	// XS interface functions	1637	// XS interface functions
1352		1638
1353	MODULE = JSON::XS PACKAGE = JSON::XS	1639	MODULE = JSON::XS PACKAGE = JSON::XS
1354		1640
1355	BOOT:	1641	BOOT:
…		…
1364	: -1;	1650	: -1;
1365		1651
1366	json_stash = gv_stashpv ("JSON::XS" , 1);	1652	json_stash = gv_stashpv ("JSON::XS" , 1);
1367	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1653	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1368		1654
1369	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1655	json_true = get_bool ("JSON::XS::true");
1370	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1656	json_false = get_bool ("JSON::XS::false");
1371	}	1657	}
1372		1658
1373	PROTOTYPES: DISABLE	1659	PROTOTYPES: DISABLE
1374		1660
1375	void CLONE (...)	1661	void CLONE (...)
…		…
1378	json_boolean_stash = 0;	1664	json_boolean_stash = 0;
1379		1665
1380	void new (char *klass)	1666	void new (char *klass)
1381	PPCODE:	1667	PPCODE:
1382	{	1668	{
1383	HV *stash = !JSON_SLOW \|\| json_stash
1384	? json_stash
1385	: gv_stashpv ("JSON::XS", 1);
1386	SV *pv = NEWSV (0, sizeof (JSON));	1669	SV *pv = NEWSV (0, sizeof (JSON));
1387	SvPOK_only (pv);	1670	SvPOK_only (pv);
1388	Zero (SvPVX (pv), 1, JSON);	1671	json_init ((JSON *)SvPVX (pv));
1389	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1672	XPUSHs (sv_2mortal (sv_bless (
1390	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));	1673	newRV_noinc (pv),
		1674	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1675	)));
1391	}	1676	}
1392		1677
1393	void ascii (JSON *self, int enable = 1)	1678	void ascii (JSON *self, int enable = 1)
1394	ALIAS:	1679	ALIAS:
1395	ascii = F_ASCII	1680	ascii = F_ASCII
…		…
1402	pretty = F_PRETTY	1687	pretty = F_PRETTY
1403	allow_nonref = F_ALLOW_NONREF	1688	allow_nonref = F_ALLOW_NONREF
1404	shrink = F_SHRINK	1689	shrink = F_SHRINK
1405	allow_blessed = F_ALLOW_BLESSED	1690	allow_blessed = F_ALLOW_BLESSED
1406	convert_blessed = F_CONV_BLESSED	1691	convert_blessed = F_CONV_BLESSED
		1692	relaxed = F_RELAXED
		1693	allow_unknown = F_ALLOW_UNKNOWN
1407	PPCODE:	1694	PPCODE:
1408	{	1695	{
1409	if (enable)	1696	if (enable)
1410	self->flags \|= ix;	1697	self->flags \|= ix;
1411	else	1698	else
1412	self->flags &= ~ix;	1699	self->flags &= ~ix;
1413		1700
1414	XPUSHs (ST (0));	1701	XPUSHs (ST (0));
1415	}	1702	}
1416		1703
1417	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1704	void get_ascii (JSON *self)
		1705	ALIAS:
		1706	get_ascii = F_ASCII
		1707	get_latin1 = F_LATIN1
		1708	get_utf8 = F_UTF8
		1709	get_indent = F_INDENT
		1710	get_canonical = F_CANONICAL
		1711	get_space_before = F_SPACE_BEFORE
		1712	get_space_after = F_SPACE_AFTER
		1713	get_allow_nonref = F_ALLOW_NONREF
		1714	get_shrink = F_SHRINK
		1715	get_allow_blessed = F_ALLOW_BLESSED
		1716	get_convert_blessed = F_CONV_BLESSED
		1717	get_relaxed = F_RELAXED
		1718	get_allow_unknown = F_ALLOW_UNKNOWN
1418	PPCODE:	1719	PPCODE:
1419	{	1720	XPUSHs (boolSV (self->flags & ix));
1420	UV log2 = 0;
1421		1721
1422	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1722	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1423		1723	PPCODE:
1424	while ((1UL << log2) < max_depth)	1724	self->max_depth = max_depth;
1425	++log2;
1426
1427	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1428
1429	XPUSHs (ST (0));	1725	XPUSHs (ST (0));
1430	}
1431		1726
		1727	U32 get_max_depth (JSON *self)
		1728	CODE:
		1729	RETVAL = self->max_depth;
		1730	OUTPUT:
		1731	RETVAL
		1732
1432	void max_size (JSON *self, UV max_size = 0)	1733	void max_size (JSON *self, U32 max_size = 0)
1433	PPCODE:	1734	PPCODE:
1434	{	1735	self->max_size = max_size;
1435	UV log2 = 0;
1436
1437	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1438	if (max_size == 1) max_size = 2;
1439
1440	while ((1UL << log2) < max_size)
1441	++log2;
1442
1443	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1444
1445	XPUSHs (ST (0));	1736	XPUSHs (ST (0));
1446	}	1737
		1738	int get_max_size (JSON *self)
		1739	CODE:
		1740	RETVAL = self->max_size;
		1741	OUTPUT:
		1742	RETVAL
1447		1743
1448	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1744	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1449	PPCODE:	1745	PPCODE:
1450	{	1746	{
1451	SvREFCNT_dec (self->cb_object);	1747	SvREFCNT_dec (self->cb_object);
…		…
1485	XPUSHs (decode_json (jsonstr, self, 0));	1781	XPUSHs (decode_json (jsonstr, self, 0));
1486		1782
1487	void decode_prefix (JSON self, SV jsonstr)	1783	void decode_prefix (JSON self, SV jsonstr)
1488	PPCODE:	1784	PPCODE:
1489	{	1785	{
1490	UV offset;	1786	STRLEN offset;
1491	EXTEND (SP, 2);	1787	EXTEND (SP, 2);
1492	PUSHs (decode_json (jsonstr, self, &offset));	1788	PUSHs (decode_json (jsonstr, self, &offset));
1493	PUSHs (sv_2mortal (newSVuv (offset)));	1789	PUSHs (sv_2mortal (newSVuv (offset)));
		1790	}
		1791
		1792	void incr_parse (JSON self, SV jsonstr = 0)
		1793	PPCODE:
		1794	{
		1795	if (!self->incr_text)
		1796	self->incr_text = newSVpvn ("", 0);
		1797
		1798	// append data, if any
		1799	if (jsonstr)
		1800	{
		1801	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1802	{
		1803	/* utf-8-ness differs, need to upgrade */
		1804	sv_utf8_upgrade (self->incr_text);
		1805
		1806	if (self->incr_pos)
		1807	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1808	- (U8 *)SvPVX (self->incr_text);
		1809	}
		1810
		1811	{
		1812	STRLEN len;
		1813	const char *str = SvPV (jsonstr, len);
		1814	STRLEN cur = SvCUR (self->incr_text);
		1815
		1816	if (SvLEN (self->incr_text) <= cur + len)
		1817	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1818
		1819	Move (str, SvEND (self->incr_text), len, char);
		1820	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1821	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1822	}
		1823	}
		1824
		1825	if (GIMME_V != G_VOID)
		1826	do
		1827	{
		1828	STRLEN offset;
		1829
		1830	if (!INCR_DONE (self))
		1831	{
		1832	incr_parse (self);
		1833
		1834	if (self->incr_pos > self->max_size && self->max_size)
		1835	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1836	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1837
		1838	if (!INCR_DONE (self))
		1839	break;
		1840	}
		1841
		1842	XPUSHs (decode_json (self->incr_text, self, &offset));
		1843
		1844	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1845	self->incr_pos -= offset;
		1846	self->incr_nest = 0;
		1847	self->incr_mode = 0;
		1848	}
		1849	while (GIMME_V == G_ARRAY);
		1850	}
		1851
		1852	SV incr_text (JSON self)
		1853	ATTRS: lvalue
		1854	CODE:
		1855	{
		1856	if (self->incr_pos)
		1857	croak ("incr_text can not be called when the incremental parser already started parsing");
		1858
		1859	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1860	}
		1861	OUTPUT:
		1862	RETVAL
		1863
		1864	void incr_skip (JSON *self)
		1865	CODE:
		1866	{
		1867	if (self->incr_pos)
		1868	{
		1869	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1870	self->incr_pos = 0;
		1871	self->incr_nest = 0;
		1872	self->incr_mode = 0;
		1873	}
		1874	}
		1875
		1876	void incr_reset (JSON *self)
		1877	CODE:
		1878	{
		1879	SvREFCNT_dec (self->incr_text);
		1880	self->incr_text = 0;
		1881	self->incr_pos = 0;
		1882	self->incr_nest = 0;
		1883	self->incr_mode = 0;
1494	}	1884	}
1495		1885
1496	void DESTROY (JSON *self)	1886	void DESTROY (JSON *self)
1497	CODE:	1887	CODE:
1498	SvREFCNT_dec (self->cb_sk_object);	1888	SvREFCNT_dec (self->cb_sk_object);
1499	SvREFCNT_dec (self->cb_object);	1889	SvREFCNT_dec (self->cb_object);
		1890	SvREFCNT_dec (self->incr_text);
1500		1891
1501	PROTOTYPES: ENABLE	1892	PROTOTYPES: ENABLE
1502		1893
1503	void to_json (SV *scalar)	1894	void encode_json (SV *scalar)
		1895	ALIAS:
		1896	to_json_ = 0
		1897	encode_json = F_UTF8
1504	PPCODE:	1898	PPCODE:
1505	{	1899	{
1506	JSON json = { F_DEFAULT \| F_UTF8 };	1900	JSON json;
		1901	json_init (&json);
		1902	json.flags \|= ix;
1507	XPUSHs (encode_json (scalar, &json));	1903	XPUSHs (encode_json (scalar, &json));
1508	}	1904	}
1509		1905
1510	void from_json (SV *jsonstr)	1906	void decode_json (SV *jsonstr)
		1907	ALIAS:
		1908	from_json_ = 0
		1909	decode_json = F_UTF8
1511	PPCODE:	1910	PPCODE:
1512	{	1911	{
1513	JSON json = { F_DEFAULT \| F_UTF8 };	1912	JSON json;
		1913	json_init (&json);
		1914	json.flags \|= ix;
1514	XPUSHs (decode_json (jsonstr, &json, 0));	1915	XPUSHs (decode_json (jsonstr, &json, 0));
1515	}	1916	}
1516		1917
		1918

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs. Revision 1.95 by root, Sun Feb 22 06:55:28 2009 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs.
Revision 1.95 by root, Sun Feb 22 06:55:28 2009 UTC