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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs.
Revision 1.100 by root, Sat Sep 5 23:00:56 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
14		15
15	// some old perls do not have this, try to make it work, no	16	// some old perls do not have this, try to make it work, no
16	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarantees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
19	#endif	20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
20		23
21	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
…		…
27	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
28	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
32	#define F_MAXDEPTH 0xf8000000UL
33	#define S_MAXDEPTH 27
34	#define F_MAXSIZE 0x01f00000UL	35	#define F_RELAXED 0x00001000UL
35	#define S_MAXSIZE 20	36	#define F_ALLOW_UNKNOWN 0x00002000UL
36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
37		38
38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
40
41	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
42	#define F_DEFAULT (9UL << S_MAXDEPTH)
43		40
44	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
45	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
46		43
47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
48		45
49	#define SB do {	46	#define SB do {
50	#define SE } while (0)	47	#define SE } while (0)
51		48
52	#if __GNUC__ >= 3	49	#if __GNUC__ >= 3
53	# define expect(expr,value) __builtin_expect ((expr),(value))	50	# define expect(expr,value) __builtin_expect ((expr), (value))
54	# define inline inline	51	# define INLINE static inline
55	#else	52	#else
56	# define expect(expr,value) (expr)	53	# define expect(expr,value) (expr)
57	# define inline static	54	# define INLINE static
58	#endif	55	#endif
59		56
60	#define expect_false(expr) expect ((expr) != 0, 0)	57	#define expect_false(expr) expect ((expr) != 0, 0)
61	#define expect_true(expr) expect ((expr) != 0, 1)	58	#define expect_true(expr) expect ((expr) != 0, 1)
		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
62		65
63	#ifdef USE_ITHREADS	66	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	67	# define JSON_SLOW 1
65	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
66	#else	69	#else
…		…
69	#endif	72	#endif
70		73
71	static HV json_stash, json_boolean_stash; // JSON::XS::	74	static HV json_stash, json_boolean_stash; // JSON::XS::
72	static SV json_true, json_false;	75	static SV json_true, json_false;
73		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		85
74	typedef struct {	86	typedef struct {
75	U32 flags;	87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
76	SV *cb_object;	91	SV *cb_object;
77	HV *cb_sk_object;	92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	int incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
78	} JSON;	99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
79		107
80	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
81	// utility functions	109	// utility functions
82		110
83	inline void	111	INLINE SV *
		112	get_bool (const char *name)
		113	{
		114	SV *sv = get_sv (name, 1);
		115
		116	SvREADONLY_on (sv);
		117	SvREADONLY_on (SvRV (sv));
		118
		119	return sv;
		120	}
		121
		122	INLINE void
84	shrink (SV *sv)	123	shrink (SV *sv)
85	{	124	{
86	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
87	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
88	{	128	{
89	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
90	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
91	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
97	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
98	// case of an error.	138	// case of an error.
99	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
100	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
101	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
102	inline UV	142	INLINE UV
103	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	143	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
104	{	144	{
105	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	145	if (expect_true (len >= 2
106	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
107	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
108	{	148	{
109	*clen = 2;	149	*clen = 2;
110	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
111	}	151	}
112	else	152	else
113	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
114	*clen = (STRLEN)-1;	154	}
115	return (UV)-1;	155
116	}	156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 \| ( ch >> 6),
		166	*s++ = 0x80 \| ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 \| ( ch >> 12),
		169	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 \| ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 \| ( ch >> 18),
		173	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 \| ( ch & 0x3f);
		176
		177	return s;
		178	}
		179
		180	// convert offset pointer to character index, sv must be string
		181	static STRLEN
		182	ptr_to_index (SV sv, char offset)
		183	{
		184	return SvUTF8 (sv)
		185	? utf8_distance (offset, SvPVX (sv))
		186	: offset - SvPVX (sv);
117	}	187	}
118		188
119	/////////////////////////////////////////////////////////////////////////////	189	/////////////////////////////////////////////////////////////////////////////
120	// encoder	190	// encoder
121		191
…		…
125	char *cur; // SvPVX (sv) + current output position	195	char *cur; // SvPVX (sv) + current output position
126	char *end; // SvEND (sv)	196	char *end; // SvEND (sv)
127	SV *sv; // result scalar	197	SV *sv; // result scalar
128	JSON json;	198	JSON json;
129	U32 indent; // indentation level	199	U32 indent; // indentation level
130	U32 maxdepth; // max. indentation/recursion level	200	UV limit; // escape character values >= this value when encoding
131	} enc_t;	201	} enc_t;
132		202
133	inline void	203	INLINE void
134	need (enc_t *enc, STRLEN len)	204	need (enc_t *enc, STRLEN len)
135	{	205	{
136	if (expect_false (enc->cur + len >= enc->end))	206	if (expect_false (enc->cur + len >= enc->end))
137	{	207	{
138	STRLEN cur = enc->cur - SvPVX (enc->sv);	208	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
139	SvGROW (enc->sv, cur + len + 1);	209	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
140	enc->cur = SvPVX (enc->sv) + cur;	210	enc->cur = SvPVX (enc->sv) + cur;
141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	211	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
142	}	212	}
143	}	213	}
144		214
145	inline void	215	INLINE void
146	encode_ch (enc_t *enc, char ch)	216	encode_ch (enc_t *enc, char ch)
147	{	217	{
148	need (enc, 1);	218	need (enc, 1);
149	*enc->cur++ = ch;	219	*enc->cur++ = ch;
150	}	220	}
…		…
204	{	274	{
205	uch = ch;	275	uch = ch;
206	clen = 1;	276	clen = 1;
207	}	277	}
208		278
209	if (uch > 0x10FFFFUL)	279	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
210	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
211
212	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
213	{	280	{
214	if (uch > 0xFFFFUL)	281	if (uch >= 0x10000UL)
215	{	282	{
		283	if (uch >= 0x110000UL)
		284	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		285
216	need (enc, len += 11);	286	need (enc, len += 11);
217	sprintf (enc->cur, "\\u%04x\\u%04x",	287	sprintf (enc->cur, "\\u%04x\\u%04x",
218	(int)((uch - 0x10000) / 0x400 + 0xD800),	288	(int)((uch - 0x10000) / 0x400 + 0xD800),
219	(int)((uch - 0x10000) % 0x400 + 0xDC00));	289	(int)((uch - 0x10000) % 0x400 + 0xDC00));
220	enc->cur += 12;	290	enc->cur += 12;
221	}	291	}
222	else	292	else
223	{	293	{
224	static char hexdigit [16] = "0123456789abcdef";
225	need (enc, len += 5);	294	need (enc, len += 5);
226	*enc->cur++ = '\\';	295	*enc->cur++ = '\\';
227	*enc->cur++ = 'u';	296	*enc->cur++ = 'u';
228	*enc->cur++ = hexdigit [ uch >> 12 ];	297	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
229	*enc->cur++ = hexdigit [(uch >> 8) & 15];	298	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
230	*enc->cur++ = hexdigit [(uch >> 4) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
231	*enc->cur++ = hexdigit [(uch >> 0) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
232	}	301	}
233		302
234	str += clen;	303	str += clen;
235	}	304	}
236	else if (enc->json.flags & F_LATIN1)	305	else if (enc->json.flags & F_LATIN1)
…		…
248	while (--clen);	317	while (--clen);
249	}	318	}
250	else	319	else
251	{	320	{
252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	321	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	322	enc->cur = encode_utf8 (enc->cur, uch);
254	++str;	323	++str;
255	}	324	}
256	}	325	}
257	}	326	}
258	}	327	}
259		328
260	--len;	329	--len;
261	}	330	}
262	}	331	}
263		332
264	inline void	333	INLINE void
265	encode_indent (enc_t *enc)	334	encode_indent (enc_t *enc)
266	{	335	{
267	if (enc->json.flags & F_INDENT)	336	if (enc->json.flags & F_INDENT)
268	{	337	{
269	int spaces = enc->indent * INDENT_STEP;	338	int spaces = enc->indent * INDENT_STEP;
…		…
272	memset (enc->cur, ' ', spaces);	341	memset (enc->cur, ' ', spaces);
273	enc->cur += spaces;	342	enc->cur += spaces;
274	}	343	}
275	}	344	}
276		345
277	inline void	346	INLINE void
278	encode_space (enc_t *enc)	347	encode_space (enc_t *enc)
279	{	348	{
280	need (enc, 1);	349	need (enc, 1);
281	encode_ch (enc, ' ');	350	encode_ch (enc, ' ');
282	}	351	}
283		352
284	inline void	353	INLINE void
285	encode_nl (enc_t *enc)	354	encode_nl (enc_t *enc)
286	{	355	{
287	if (enc->json.flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
288	{	357	{
289	need (enc, 1);	358	need (enc, 1);
290	encode_ch (enc, '\n');	359	encode_ch (enc, '\n');
291	}	360	}
292	}	361	}
293		362
294	inline void	363	INLINE void
295	encode_comma (enc_t *enc)	364	encode_comma (enc_t *enc)
296	{	365	{
297	encode_ch (enc, ',');	366	encode_ch (enc, ',');
298		367
299	if (enc->json.flags & F_INDENT)	368	if (enc->json.flags & F_INDENT)
…		…
307	static void	376	static void
308	encode_av (enc_t enc, AV av)	377	encode_av (enc_t enc, AV av)
309	{	378	{
310	int i, len = av_len (av);	379	int i, len = av_len (av);
311		380
312	if (enc->indent >= enc->maxdepth)	381	if (enc->indent >= enc->json.max_depth)
313	croak ("data structure too deep (hit recursion limit)");	382	croak (ERR_NESTING_EXCEEDED);
314		383
315	encode_ch (enc, '['); encode_nl (enc);	384	encode_ch (enc, '[');
316	++enc->indent;	385
		386	if (len >= 0)
		387	{
		388	encode_nl (enc); ++enc->indent;
317		389
318	for (i = 0; i <= len; ++i)	390	for (i = 0; i <= len; ++i)
319	{	391	{
320	SV **svp = av_fetch (av, i, 0);	392	SV **svp = av_fetch (av, i, 0);
321		393
322	encode_indent (enc);	394	encode_indent (enc);
323		395
324	if (svp)	396	if (svp)
325	encode_sv (enc, *svp);	397	encode_sv (enc, *svp);
326	else	398	else
327	encode_str (enc, "null", 4, 0);	399	encode_str (enc, "null", 4, 0);
328		400
329	if (i < len)	401	if (i < len)
330	encode_comma (enc);	402	encode_comma (enc);
331	}	403	}
332		404
		405	encode_nl (enc); --enc->indent; encode_indent (enc);
		406	}
		407
333	encode_nl (enc);	408	encode_ch (enc, ']');
334
335	--enc->indent;
336	encode_indent (enc); encode_ch (enc, ']');
337	}	409	}
338		410
339	static void	411	static void
340	encode_he (enc_t enc, HE he)	412	encode_hk (enc_t enc, HE he)
341	{	413	{
342	encode_ch (enc, '"');	414	encode_ch (enc, '"');
343		415
344	if (HeKLEN (he) == HEf_SVKEY)	416	if (HeKLEN (he) == HEf_SVKEY)
345	{	417	{
…		…
358	encode_ch (enc, '"');	430	encode_ch (enc, '"');
359		431
360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	432	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
361	encode_ch (enc, ':');	433	encode_ch (enc, ':');
362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	434	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
363	encode_sv (enc, HeVAL (he));
364	}	435	}
365		436
366	// compare hash entries, used when all keys are bytestrings	437	// compare hash entries, used when all keys are bytestrings
367	static int	438	static int
368	he_cmp_fast (const void a_, const void b_)	439	he_cmp_fast (const void a_, const void b_)
…		…
373	HE b = (HE **)b_;	444	HE b = (HE **)b_;
374		445
375	STRLEN la = HeKLEN (a);	446	STRLEN la = HeKLEN (a);
376	STRLEN lb = HeKLEN (b);	447	STRLEN lb = HeKLEN (b);
377		448
378	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	449	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
379	cmp = la - lb;	450	cmp = lb - la;
380		451
381	return cmp;	452	return cmp;
382	}	453	}
383		454
384	// compare hash entries, used when some keys are sv's or utf-x	455	// compare hash entries, used when some keys are sv's or utf-x
385	static int	456	static int
386	he_cmp_slow (const void a, const void b)	457	he_cmp_slow (const void a, const void b)
387	{	458	{
388	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	459	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
389	}	460	}
390		461
391	static void	462	static void
392	encode_hv (enc_t enc, HV hv)	463	encode_hv (enc_t enc, HV hv)
393	{	464	{
394	int count, i;	465	HE *he;
395		466
396	if (enc->indent >= enc->maxdepth)	467	if (enc->indent >= enc->json.max_depth)
397	croak ("data structure too deep (hit recursion limit)");	468	croak (ERR_NESTING_EXCEEDED);
398		469
399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	470	encode_ch (enc, '{');
400		471
401	if ((count = hv_iterinit (hv)))
402	{
403	// for canonical output we have to sort by keys first	472	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	473	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	474	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	475	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	476	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		477	{
		478	int count = hv_iterinit (hv);
		479
		480	if (SvMAGICAL (hv))
408	{	481	{
		482	// need to count by iterating. could improve by dynamically building the vector below
		483	// but I don't care for the speed of this special case.
		484	// note also that we will run into undefined behaviour when the two iterations
		485	// do not result in the same count, something I might care for in some later release.
		486
		487	count = 0;
		488	while (hv_iternext (hv))
		489	++count;
		490
		491	hv_iterinit (hv);
		492	}
		493
		494	if (count)
		495	{
409	int fast = 1;	496	int i, fast = 1;
410	HE *he;
411	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	497	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
412	HE *hes = _alloca (count sizeof (HE));	498	HE *hes = _alloca (count sizeof (HE));
413	#else	499	#else
414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	500	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
415	#endif	501	#endif
…		…
442		528
443	FREETMPS;	529	FREETMPS;
444	LEAVE;	530	LEAVE;
445	}	531	}
446		532
447	for (i = 0; i < count; ++i)	533	encode_nl (enc); ++enc->indent;
		534
		535	while (count--)
448	{	536	{
449	encode_indent (enc);	537	encode_indent (enc);
		538	he = hes [count];
450	encode_he (enc, hes [i]);	539	encode_hk (enc, he);
		540	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		541
452	if (i < count - 1)	542	if (count)
453	encode_comma (enc);	543	encode_comma (enc);
454	}	544	}
455		545
456	encode_nl (enc);	546	encode_nl (enc); --enc->indent; encode_indent (enc);
457	}	547	}
		548	}
458	else	549	else
		550	{
		551	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		552	if ((he = hv_iternext (hv)))
459	{	553	{
460	HE *he = hv_iternext (hv);	554	encode_nl (enc); ++enc->indent;
461		555
462	for (;;)	556	for (;;)
463	{	557	{
464	encode_indent (enc);	558	encode_indent (enc);
465	encode_he (enc, he);	559	encode_hk (enc, he);
		560	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
466		561
467	if (!(he = hv_iternext (hv)))	562	if (!(he = hv_iternext (hv)))
468	break;	563	break;
469		564
470	encode_comma (enc);	565	encode_comma (enc);
471	}	566	}
472		567
473	encode_nl (enc);	568	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	569	}
475	}	570	}
476		571
477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	572	encode_ch (enc, '}');
478	}	573	}
479		574
480	// encode objects, arrays and special \0=false and \1=true values.	575	// encode objects, arrays and special \0=false and \1=true values.
481	static void	576	static void
482	encode_rv (enc_t enc, SV sv)	577	encode_rv (enc_t enc, SV sv)
…		…
559		654
560	if (len == 1 && *pv == '1')	655	if (len == 1 && *pv == '1')
561	encode_str (enc, "true", 4, 0);	656	encode_str (enc, "true", 4, 0);
562	else if (len == 1 && *pv == '0')	657	else if (len == 1 && *pv == '0')
563	encode_str (enc, "false", 5, 0);	658	encode_str (enc, "false", 5, 0);
		659	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		660	encode_str (enc, "null", 4, 0);
564	else	661	else
565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	662	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	663	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
567	}	664	}
		665	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		666	encode_str (enc, "null", 4, 0);
568	else	667	else
569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	668	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	669	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
571	}	670	}
572		671
…		…
590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	689	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
591	enc->cur += strlen (enc->cur);	690	enc->cur += strlen (enc->cur);
592	}	691	}
593	else if (SvIOKp (sv))	692	else if (SvIOKp (sv))
594	{	693	{
595	// we assume we can always read an IV as a UV	694	// we assume we can always read an IV as a UV and vice versa
596	if (SvUV (sv) & ~(UV)0x7fff)	695	// we assume two's complement
597	{	696	// we assume no aliasing issues in the union
598	// large integer, use the (rather slow) snprintf way.	697	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
599	need (enc, sizeof (UV) * 3);	698	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
600	enc->cur +=
601	SvIsUV(sv)
602	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
603	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
604	}
605	else
606	{	699	{
607	// optimise the "small number case"	700	// optimise the "small number case"
608	// code will likely be branchless and use only a single multiplication	701	// code will likely be branchless and use only a single multiplication
		702	// works for numbers up to 59074
609	I32 i = SvIV (sv);	703	I32 i = SvIVX (sv);
610	U32 u;	704	U32 u;
611	char digit, nz = 0;	705	char digit, nz = 0;
612		706
613	need (enc, 6);	707	need (enc, 6);
614		708
…		…
620		714
621	// now output digit by digit, each time masking out the integer part	715	// now output digit by digit, each time masking out the integer part
622	// and multiplying by 5 while moving the decimal point one to the right,	716	// and multiplying by 5 while moving the decimal point one to the right,
623	// resulting in a net multiplication by 10.	717	// resulting in a net multiplication by 10.
624	// we always write the digit to memory but conditionally increment	718	// we always write the digit to memory but conditionally increment
625	// the pointer, to ease the usage of conditional move instructions.	719	// the pointer, to enable the use of conditional move instructions.
626	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	720	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
627	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	721	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
628	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	722	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
629	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	723	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
630	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	724	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
631	}	725	}
		726	else
		727	{
		728	// large integer, use the (rather slow) snprintf way.
		729	need (enc, IVUV_MAXCHARS);
		730	enc->cur +=
		731	SvIsUV(sv)
		732	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		733	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		734	}
632	}	735	}
633	else if (SvROK (sv))	736	else if (SvROK (sv))
634	encode_rv (enc, SvRV (sv));	737	encode_rv (enc, SvRV (sv));
635	else if (!SvOK (sv))	738	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
636	encode_str (enc, "null", 4, 0);	739	encode_str (enc, "null", 4, 0);
637	else	740	else
638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	741	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
639	SvPV_nolen (sv), SvFLAGS (sv));	742	SvPV_nolen (sv), SvFLAGS (sv));
640	}	743	}
…		…
650	enc.json = *json;	753	enc.json = *json;
651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	754	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
652	enc.cur = SvPVX (enc.sv);	755	enc.cur = SvPVX (enc.sv);
653	enc.end = SvEND (enc.sv);	756	enc.end = SvEND (enc.sv);
654	enc.indent = 0;	757	enc.indent = 0;
655	enc.maxdepth = DEC_DEPTH (enc.json.flags);	758	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		759	: enc.json.flags & F_LATIN1 ? 0x000100UL
		760	: 0x110000UL;
656		761
657	SvPOK_only (enc.sv);	762	SvPOK_only (enc.sv);
658	encode_sv (&enc, scalar);	763	encode_sv (&enc, scalar);
		764	encode_nl (&enc);
659		765
660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	766	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
661	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	767	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
662		768
663	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	769	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
681	JSON json;	787	JSON json;
682	U32 depth; // recursion depth	788	U32 depth; // recursion depth
683	U32 maxdepth; // recursion depth limit	789	U32 maxdepth; // recursion depth limit
684	} dec_t;	790	} dec_t;
685		791
686	inline void	792	INLINE void
		793	decode_comment (dec_t *dec)
		794	{
		795	// only '#'-style comments allowed a.t.m.
		796
		797	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		798	++dec->cur;
		799	}
		800
		801	INLINE void
687	decode_ws (dec_t *dec)	802	decode_ws (dec_t *dec)
688	{	803	{
689	for (;;)	804	for (;;)
690	{	805	{
691	char ch = *dec->cur;	806	char ch = *dec->cur;
692		807
693	if (ch > 0x20	808	if (ch > 0x20)
		809	{
		810	if (expect_false (ch == '#'))
		811	{
		812	if (dec->json.flags & F_RELAXED)
		813	decode_comment (dec);
		814	else
		815	break;
		816	}
		817	else
		818	break;
		819	}
694	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	820	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
695	break;	821	break; // parse error, but let higher level handle it, gives better error messages
696		822
697	++dec->cur;	823	++dec->cur;
698	}	824	}
699	}	825	}
700		826
…		…
704	if (*dec->cur != ch) \	830	if (*dec->cur != ch) \
705	ERR (# ch " expected"); \	831	ERR (# ch " expected"); \
706	++dec->cur; \	832	++dec->cur; \
707	SE	833	SE
708		834
709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	835	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
710	#define DEC_DEC_DEPTH --dec->depth	836	#define DEC_DEC_DEPTH --dec->depth
711		837
712	static SV decode_sv (dec_t dec);	838	static SV decode_sv (dec_t dec);
713		839
714	static signed char decode_hexdigit[256];	840	static signed char decode_hexdigit[256];
…		…
806		932
807	if (hi >= 0x80)	933	if (hi >= 0x80)
808	{	934	{
809	utf8 = 1;	935	utf8 = 1;
810		936
811	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	937	cur = encode_utf8 (cur, hi);
812	}	938	}
813	else	939	else
814	*cur++ = hi;	940	*cur++ = hi;
815	}	941	}
816	break;	942	break;
…		…
818	default:	944	default:
819	--dec_cur;	945	--dec_cur;
820	ERR ("illegal backslash escape sequence in string");	946	ERR ("illegal backslash escape sequence in string");
821	}	947	}
822	}	948	}
823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	949	else if (expect_true (ch >= 0x20 && ch < 0x80))
824	*cur++ = ch;	950	*cur++ = ch;
825	else if (ch >= 0x80)	951	else if (ch >= 0x80)
826	{	952	{
827	STRLEN clen;	953	STRLEN clen;
828	UV uch;	954	UV uch;
…		…
854	{	980	{
855	STRLEN len = cur - buf;	981	STRLEN len = cur - buf;
856		982
857	if (sv)	983	if (sv)
858	{	984	{
859	SvGROW (sv, SvCUR (sv) + len + 1);	985	STRLEN cur = SvCUR (sv);
		986
		987	if (SvLEN (sv) <= cur + len)
		988	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		989
860	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	990	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
861	SvCUR_set (sv, SvCUR (sv) + len);	991	SvCUR_set (sv, SvCUR (sv) + len);
862	}	992	}
863	else	993	else
864	sv = newSVpvn (buf, len);	994	sv = newSVpvn (buf, len);
…		…
951		1081
952	if (!is_nv)	1082	if (!is_nv)
953	{	1083	{
954	int len = dec->cur - start;	1084	int len = dec->cur - start;
955		1085
956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1086	// special case the rather common 1..5-digit-int case
957	if (*start == '-')	1087	if (*start == '-')
958	switch (len)	1088	switch (len)
959	{	1089	{
960	case 2: return newSViv (-( start [1] - '0' * 1));	1090	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1091	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
962	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1092	case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
963	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1093	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1094	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
964	}	1095	}
965	else	1096	else
966	switch (len)	1097	switch (len)
967	{	1098	{
968	case 1: return newSViv ( start [0] - '0' * 1);	1099	case 1: return newSViv ( start [0] - '0' * 1);
969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1100	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
970	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1101	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
971	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1102	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1103	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
972	}	1104	}
973		1105
974	{	1106	{
975	UV uv;	1107	UV uv;
976	int numtype = grok_number (start, len, &uv);	1108	int numtype = grok_number (start, len, &uv);
…		…
1037		1169
1038	if (*dec->cur != ',')	1170	if (*dec->cur != ',')
1039	ERR (", or ] expected while parsing array");	1171	ERR (", or ] expected while parsing array");
1040		1172
1041	++dec->cur;	1173	++dec->cur;
		1174
		1175	decode_ws (dec);
		1176
		1177	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1178	{
		1179	++dec->cur;
		1180	break;
		1181	}
1042	}	1182	}
1043		1183
1044	DEC_DEC_DEPTH;	1184	DEC_DEC_DEPTH;
1045	return newRV_noinc ((SV *)av);	1185	return newRV_noinc ((SV *)av);
1046		1186
…		…
1062	if (*dec->cur == '}')	1202	if (*dec->cur == '}')
1063	++dec->cur;	1203	++dec->cur;
1064	else	1204	else
1065	for (;;)	1205	for (;;)
1066	{	1206	{
1067	decode_ws (dec); EXPECT_CH ('"');	1207	EXPECT_CH ('"');
1068		1208
1069	// heuristic: assume that	1209	// heuristic: assume that
1070	// a) decode_str + hv_store_ent are abysmally slow.	1210	// a) decode_str + hv_store_ent are abysmally slow.
1071	// b) most hash keys are short, simple ascii text.	1211	// b) most hash keys are short, simple ascii text.
1072	// => try to "fast-match" such strings to avoid	1212	// => try to "fast-match" such strings to avoid
…		…
1076	char *p = dec->cur;	1216	char *p = dec->cur;
1077	char *e = p + 24; // only try up to 24 bytes	1217	char *e = p + 24; // only try up to 24 bytes
1078		1218
1079	for (;;)	1219	for (;;)
1080	{	1220	{
1081	// the >= 0x80 is true on most architectures	1221	// the >= 0x80 is false on most architectures
1082	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1222	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1083	{	1223	{
1084	// slow path, back up and use decode_str	1224	// slow path, back up and use decode_str
1085	SV *key = decode_str (dec);	1225	SV *key = decode_str (dec);
1086	if (!key)	1226	if (!key)
1087	goto fail;	1227	goto fail;
1088		1228
1089	decode_ws (dec); EXPECT_CH (':');	1229	decode_ws (dec); EXPECT_CH (':');
1090		1230
		1231	decode_ws (dec);
1091	value = decode_sv (dec);	1232	value = decode_sv (dec);
1092	if (!value)	1233	if (!value)
1093	{	1234	{
1094	SvREFCNT_dec (key);	1235	SvREFCNT_dec (key);
1095	goto fail;	1236	goto fail;
…		…
1107	int len = p - key;	1248	int len = p - key;
1108	dec->cur = p + 1;	1249	dec->cur = p + 1;
1109		1250
1110	decode_ws (dec); EXPECT_CH (':');	1251	decode_ws (dec); EXPECT_CH (':');
1111		1252
		1253	decode_ws (dec);
1112	value = decode_sv (dec);	1254	value = decode_sv (dec);
1113	if (!value)	1255	if (!value)
1114	goto fail;	1256	goto fail;
1115		1257
1116	hv_store (hv, key, len, value, 0);	1258	hv_store (hv, key, len, value, 0);
…		…
1132		1274
1133	if (*dec->cur != ',')	1275	if (*dec->cur != ',')
1134	ERR (", or } expected while parsing object/hash");	1276	ERR (", or } expected while parsing object/hash");
1135		1277
1136	++dec->cur;	1278	++dec->cur;
		1279
		1280	decode_ws (dec);
		1281
		1282	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1283	{
		1284	++dec->cur;
		1285	break;
		1286	}
1137	}	1287	}
1138		1288
1139	DEC_DEC_DEPTH;	1289	DEC_DEC_DEPTH;
1140	sv = newRV_noinc ((SV *)hv);	1290	sv = newRV_noinc ((SV *)hv);
1141		1291
…		…
1206	}	1356	}
1207		1357
1208	static SV *	1358	static SV *
1209	decode_sv (dec_t *dec)	1359	decode_sv (dec_t *dec)
1210	{	1360	{
1211	decode_ws (dec);
1212
1213	// the beauty of JSON: you need exactly one character lookahead	1361	// the beauty of JSON: you need exactly one character lookahead
1214	// to parse anything.	1362	// to parse everything.
1215	switch (*dec->cur)	1363	switch (*dec->cur)
1216	{	1364	{
1217	case '"': ++dec->cur; return decode_str (dec);	1365	case '"': ++dec->cur; return decode_str (dec);
1218	case '[': ++dec->cur; return decode_av (dec);	1366	case '[': ++dec->cur; return decode_av (dec);
1219	case '{': ++dec->cur; return decode_hv (dec);	1367	case '{': ++dec->cur; return decode_hv (dec);
1220		1368
1221	case '-':	1369	case '-':
1222	case '0': case '1': case '2': case '3': case '4':	1370	case '0': case '1': case '2': case '3': case '4':
1223	case '5': case '6': case '7': case '8': case '9':	1371	case '5': case '6': case '7': case '8': case '9':
1224	return decode_num (dec);	1372	return decode_num (dec);
1225		1373
1226	case 't':	1374	case 't':
1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1375	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1228	{	1376	{
1229	dec->cur += 4;	1377	dec->cur += 4;
		1378	#if JSON_SLOW
		1379	json_true = get_bool ("JSON::XS::true");
		1380	#endif
1230	return SvREFCNT_inc (json_true);	1381	return newSVsv (json_true);
1231	}	1382	}
1232	else	1383	else
1233	ERR ("'true' expected");	1384	ERR ("'true' expected");
1234		1385
1235	break;	1386	break;
1236		1387
1237	case 'f':	1388	case 'f':
1238	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1389	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1239	{	1390	{
1240	dec->cur += 5;	1391	dec->cur += 5;
		1392	#if JSON_SLOW
		1393	json_false = get_bool ("JSON::XS::false");
		1394	#endif
1241	return SvREFCNT_inc (json_false);	1395	return newSVsv (json_false);
1242	}	1396	}
1243	else	1397	else
1244	ERR ("'false' expected");	1398	ERR ("'false' expected");
1245		1399
1246	break;	1400	break;
…		…
1264	fail:	1418	fail:
1265	return 0;	1419	return 0;
1266	}	1420	}
1267		1421
1268	static SV *	1422	static SV *
1269	decode_json (SV string, JSON json, UV *offset_return)	1423	decode_json (SV string, JSON json, char **offset_return)
1270	{	1424	{
1271	dec_t dec;	1425	dec_t dec;
1272	UV offset;
1273	SV *sv;	1426	SV *sv;
1274		1427
		1428	/* work around bugs in 5.10 where manipulating magic values
		1429	* will perl ignore the magic in subsequent accesses
		1430	*/
1275	SvGETMAGIC (string);	1431	/SvGETMAGIC (string);/
		1432	if (SvMAGICAL (string))
		1433	string = sv_2mortal (newSVsv (string));
		1434
1276	SvUPGRADE (string, SVt_PV);	1435	SvUPGRADE (string, SVt_PV);
1277		1436
1278	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1437	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1438	* assertion with -DDEBUGGING, although SvCUR is documented to
		1439	* return the xpv_cur field which certainly exists after upgrading.
		1440	* according to nicholas clark, calling SvPOK fixes this.
		1441	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1442	* and hope for the best.
		1443	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1444	* assertion business is seriously broken, try yet another workaround
		1445	* for the broken -DDEBUGGING.
		1446	*/
		1447	{
		1448	#ifdef DEBUGGING
		1449	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1450	#else
		1451	STRLEN offset = SvCUR (string);
		1452	#endif
		1453
		1454	if (offset > json->max_size && json->max_size)
1279	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1455	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1280	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1456	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1457	}
1281		1458
1282	if (json->flags & F_UTF8)	1459	if (json->flags & F_UTF8)
1283	sv_utf8_downgrade (string, 0);	1460	sv_utf8_downgrade (string, 0);
1284	else	1461	else
1285	sv_utf8_upgrade (string);	1462	sv_utf8_upgrade (string);
1286		1463
1287	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1464	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1288		1465
1289	dec.json = *json;	1466	dec.json = *json;
1290	dec.cur = SvPVX (string);	1467	dec.cur = SvPVX (string);
1291	dec.end = SvEND (string);	1468	dec.end = SvEND (string);
1292	dec.err = 0;	1469	dec.err = 0;
1293	dec.depth = 0;	1470	dec.depth = 0;
1294	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1295		1471
1296	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1472	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1297	dec.json.flags \|= F_HOOK;	1473	dec.json.flags \|= F_HOOK;
1298		1474
1299	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1475	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1476
		1477	decode_ws (&dec);
1300	sv = decode_sv (&dec);	1478	sv = decode_sv (&dec);
		1479
		1480	if (offset_return)
		1481	*offset_return = dec.cur;
1301		1482
1302	if (!(offset_return \|\| !sv))	1483	if (!(offset_return \|\| !sv))
1303	{	1484	{
1304	// check for trailing garbage	1485	// check for trailing garbage
1305	decode_ws (&dec);	1486	decode_ws (&dec);
…		…
1308	{	1489	{
1309	dec.err = "garbage after JSON object";	1490	dec.err = "garbage after JSON object";
1310	SvREFCNT_dec (sv);	1491	SvREFCNT_dec (sv);
1311	sv = 0;	1492	sv = 0;
1312	}	1493	}
1313	}
1314
1315	if (offset_return \|\| !sv)
1316	{
1317	offset = dec.json.flags & F_UTF8
1318	? dec.cur - SvPVX (string)
1319	: utf8_distance (dec.cur, SvPVX (string));
1320
1321	if (offset_return)
1322	*offset_return = offset;
1323	}	1494	}
1324		1495
1325	if (!sv)	1496	if (!sv)
1326	{	1497	{
1327	SV *uni = sv_newmortal ();	1498	SV *uni = sv_newmortal ();
…		…
1333	SAVEVPTR (PL_curcop);	1504	SAVEVPTR (PL_curcop);
1334	PL_curcop = &cop;	1505	PL_curcop = &cop;
1335	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1506	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1336	LEAVE;	1507	LEAVE;
1337		1508
1338	croak ("%s, at character offset %d [\"%s\"]",	1509	croak ("%s, at character offset %d (before \"%s\")",
1339	dec.err,	1510	dec.err,
1340	(int)offset,	1511	ptr_to_index (string, dec.cur),
1341	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1512	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1342	}	1513	}
1343		1514
1344	sv = sv_2mortal (sv);	1515	sv = sv_2mortal (sv);
1345		1516
1346	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1517	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1347	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1518	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1348		1519
1349	return sv;	1520	return sv;
		1521	}
		1522
		1523	/////////////////////////////////////////////////////////////////////////////
		1524	// incremental parser
		1525
		1526	static void
		1527	incr_parse (JSON *self)
		1528	{
		1529	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1530
		1531	for (;;)
		1532	{
		1533	//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
		1534	switch (self->incr_mode)
		1535	{
		1536	// only used for intiial whitespace skipping
		1537	case INCR_M_WS:
		1538	for (;;)
		1539	{
		1540	if (*p > 0x20)
		1541	{
		1542	self->incr_mode = INCR_M_JSON;
		1543	goto incr_m_json;
		1544	}
		1545	else if (!*p)
		1546	goto interrupt;
		1547
		1548	++p;
		1549	}
		1550
		1551	// skip a single char inside a string (for \\-processing)
		1552	case INCR_M_BS:
		1553	if (!*p)
		1554	goto interrupt;
		1555
		1556	++p;
		1557	self->incr_mode = INCR_M_STR;
		1558	goto incr_m_str;
		1559
		1560	// inside a string
		1561	case INCR_M_STR:
		1562	incr_m_str:
		1563	for (;;)
		1564	{
		1565	if (*p == '"')
		1566	{
		1567	++p;
		1568	self->incr_mode = INCR_M_JSON;
		1569
		1570	if (!self->incr_nest)
		1571	goto interrupt;
		1572
		1573	goto incr_m_json;
		1574	}
		1575	else if (*p == '\\')
		1576	{
		1577	++p; // "virtually" consumes character after \
		1578
		1579	if (!*p) // if at end of string we have to switch modes
		1580	{
		1581	self->incr_mode = INCR_M_BS;
		1582	goto interrupt;
		1583	}
		1584	}
		1585	else if (!*p)
		1586	goto interrupt;
		1587
		1588	++p;
		1589	}
		1590
		1591	// after initial ws, outside string
		1592	case INCR_M_JSON:
		1593	incr_m_json:
		1594	for (;;)
		1595	{
		1596	switch (*p++)
		1597	{
		1598	case 0:
		1599	--p;
		1600	goto interrupt;
		1601
		1602	case 0x09:
		1603	case 0x0a:
		1604	case 0x0d:
		1605	case 0x20:
		1606	if (!self->incr_nest)
		1607	{
		1608	--p; // do not eat the whitespace, let the next round do it
		1609	goto interrupt;
		1610	}
		1611	break;
		1612
		1613	case '"':
		1614	self->incr_mode = INCR_M_STR;
		1615	goto incr_m_str;
		1616
		1617	case '[':
		1618	case '{':
		1619	if (++self->incr_nest > self->max_depth)
		1620	croak (ERR_NESTING_EXCEEDED);
		1621	break;
		1622
		1623	case ']':
		1624	case '}':
		1625	if (--self->incr_nest <= 0)
		1626	goto interrupt;
		1627	}
		1628	}
		1629	}
		1630
		1631	modechange:
		1632	;
		1633	}
		1634
		1635	interrupt:
		1636	self->incr_pos = p - SvPVX (self->incr_text);
		1637	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1350	}	1638	}
1351		1639
1352	/////////////////////////////////////////////////////////////////////////////	1640	/////////////////////////////////////////////////////////////////////////////
1353	// XS interface functions	1641	// XS interface functions
1354		1642
…		…
1366	: -1;	1654	: -1;
1367		1655
1368	json_stash = gv_stashpv ("JSON::XS" , 1);	1656	json_stash = gv_stashpv ("JSON::XS" , 1);
1369	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1657	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1370		1658
1371	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1659	json_true = get_bool ("JSON::XS::true");
1372	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1660	json_false = get_bool ("JSON::XS::false");
		1661
		1662	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1373	}	1663	}
1374		1664
1375	PROTOTYPES: DISABLE	1665	PROTOTYPES: DISABLE
1376		1666
1377	void CLONE (...)	1667	void CLONE (...)
…		…
1382	void new (char *klass)	1672	void new (char *klass)
1383	PPCODE:	1673	PPCODE:
1384	{	1674	{
1385	SV *pv = NEWSV (0, sizeof (JSON));	1675	SV *pv = NEWSV (0, sizeof (JSON));
1386	SvPOK_only (pv);	1676	SvPOK_only (pv);
1387	Zero (SvPVX (pv), 1, JSON);	1677	json_init ((JSON *)SvPVX (pv));
1388	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1678	XPUSHs (sv_2mortal (sv_bless (
1389	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1679	newRV_noinc (pv),
		1680	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1681	)));
1390	}	1682	}
1391		1683
1392	void ascii (JSON *self, int enable = 1)	1684	void ascii (JSON *self, int enable = 1)
1393	ALIAS:	1685	ALIAS:
1394	ascii = F_ASCII	1686	ascii = F_ASCII
…		…
1401	pretty = F_PRETTY	1693	pretty = F_PRETTY
1402	allow_nonref = F_ALLOW_NONREF	1694	allow_nonref = F_ALLOW_NONREF
1403	shrink = F_SHRINK	1695	shrink = F_SHRINK
1404	allow_blessed = F_ALLOW_BLESSED	1696	allow_blessed = F_ALLOW_BLESSED
1405	convert_blessed = F_CONV_BLESSED	1697	convert_blessed = F_CONV_BLESSED
		1698	relaxed = F_RELAXED
		1699	allow_unknown = F_ALLOW_UNKNOWN
1406	PPCODE:	1700	PPCODE:
1407	{	1701	{
1408	if (enable)	1702	if (enable)
1409	self->flags \|= ix;	1703	self->flags \|= ix;
1410	else	1704	else
1411	self->flags &= ~ix;	1705	self->flags &= ~ix;
1412		1706
1413	XPUSHs (ST (0));	1707	XPUSHs (ST (0));
1414	}	1708	}
1415		1709
1416	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1710	void get_ascii (JSON *self)
		1711	ALIAS:
		1712	get_ascii = F_ASCII
		1713	get_latin1 = F_LATIN1
		1714	get_utf8 = F_UTF8
		1715	get_indent = F_INDENT
		1716	get_canonical = F_CANONICAL
		1717	get_space_before = F_SPACE_BEFORE
		1718	get_space_after = F_SPACE_AFTER
		1719	get_allow_nonref = F_ALLOW_NONREF
		1720	get_shrink = F_SHRINK
		1721	get_allow_blessed = F_ALLOW_BLESSED
		1722	get_convert_blessed = F_CONV_BLESSED
		1723	get_relaxed = F_RELAXED
		1724	get_allow_unknown = F_ALLOW_UNKNOWN
1417	PPCODE:	1725	PPCODE:
1418	{	1726	XPUSHs (boolSV (self->flags & ix));
1419	UV log2 = 0;
1420		1727
1421	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1728	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1422		1729	PPCODE:
1423	while ((1UL << log2) < max_depth)	1730	self->max_depth = max_depth;
1424	++log2;
1425
1426	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1427
1428	XPUSHs (ST (0));	1731	XPUSHs (ST (0));
1429	}
1430		1732
		1733	U32 get_max_depth (JSON *self)
		1734	CODE:
		1735	RETVAL = self->max_depth;
		1736	OUTPUT:
		1737	RETVAL
		1738
1431	void max_size (JSON *self, UV max_size = 0)	1739	void max_size (JSON *self, U32 max_size = 0)
1432	PPCODE:	1740	PPCODE:
1433	{	1741	self->max_size = max_size;
1434	UV log2 = 0;
1435
1436	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1437	if (max_size == 1) max_size = 2;
1438
1439	while ((1UL << log2) < max_size)
1440	++log2;
1441
1442	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1443
1444	XPUSHs (ST (0));	1742	XPUSHs (ST (0));
1445	}	1743
		1744	int get_max_size (JSON *self)
		1745	CODE:
		1746	RETVAL = self->max_size;
		1747	OUTPUT:
		1748	RETVAL
1446		1749
1447	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1750	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1448	PPCODE:	1751	PPCODE:
1449	{	1752	{
1450	SvREFCNT_dec (self->cb_object);	1753	SvREFCNT_dec (self->cb_object);
…		…
1484	XPUSHs (decode_json (jsonstr, self, 0));	1787	XPUSHs (decode_json (jsonstr, self, 0));
1485		1788
1486	void decode_prefix (JSON self, SV jsonstr)	1789	void decode_prefix (JSON self, SV jsonstr)
1487	PPCODE:	1790	PPCODE:
1488	{	1791	{
1489	UV offset;	1792	char *offset;
1490	EXTEND (SP, 2);	1793	EXTEND (SP, 2);
1491	PUSHs (decode_json (jsonstr, self, &offset));	1794	PUSHs (decode_json (jsonstr, self, &offset));
1492	PUSHs (sv_2mortal (newSVuv (offset)));	1795	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1796	}
		1797
		1798	void incr_parse (JSON self, SV jsonstr = 0)
		1799	PPCODE:
		1800	{
		1801	if (!self->incr_text)
		1802	self->incr_text = newSVpvn ("", 0);
		1803
		1804	// append data, if any
		1805	if (jsonstr)
		1806	{
		1807	if (SvUTF8 (jsonstr))
		1808	{
		1809	if (!SvUTF8 (self->incr_text))
		1810	{
		1811	/* utf-8-ness differs, need to upgrade */
		1812	sv_utf8_upgrade (self->incr_text);
		1813
		1814	if (self->incr_pos)
		1815	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1816	- (U8 *)SvPVX (self->incr_text);
		1817	}
		1818	}
		1819	else if (SvUTF8 (self->incr_text))
		1820	sv_utf8_upgrade (jsonstr);
		1821
		1822	{
		1823	STRLEN len;
		1824	const char *str = SvPV (jsonstr, len);
		1825	STRLEN cur = SvCUR (self->incr_text);
		1826
		1827	if (SvLEN (self->incr_text) <= cur + len)
		1828	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1829
		1830	Move (str, SvEND (self->incr_text), len, char);
		1831	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1832	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1833	}
		1834	}
		1835
		1836	if (GIMME_V != G_VOID)
		1837	do
		1838	{
		1839	char *offset;
		1840
		1841	if (!INCR_DONE (self))
		1842	{
		1843	incr_parse (self);
		1844
		1845	if (self->incr_pos > self->max_size && self->max_size)
		1846	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1847	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1848
		1849	if (!INCR_DONE (self))
		1850	break;
		1851	}
		1852
		1853	XPUSHs (decode_json (self->incr_text, self, &offset));
		1854
		1855	self->incr_pos -= offset - SvPVX (self->incr_text);
		1856	self->incr_nest = 0;
		1857	self->incr_mode = 0;
		1858
		1859	sv_chop (self->incr_text, offset);
		1860	}
		1861	while (GIMME_V == G_ARRAY);
		1862	}
		1863
		1864	SV incr_text (JSON self)
		1865	ATTRS: lvalue
		1866	CODE:
		1867	{
		1868	if (self->incr_pos)
		1869	croak ("incr_text can not be called when the incremental parser already started parsing");
		1870
		1871	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1872	}
		1873	OUTPUT:
		1874	RETVAL
		1875
		1876	void incr_skip (JSON *self)
		1877	CODE:
		1878	{
		1879	if (self->incr_pos)
		1880	{
		1881	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1882	self->incr_pos = 0;
		1883	self->incr_nest = 0;
		1884	self->incr_mode = 0;
		1885	}
		1886	}
		1887
		1888	void incr_reset (JSON *self)
		1889	CODE:
		1890	{
		1891	SvREFCNT_dec (self->incr_text);
		1892	self->incr_text = 0;
		1893	self->incr_pos = 0;
		1894	self->incr_nest = 0;
		1895	self->incr_mode = 0;
1493	}	1896	}
1494		1897
1495	void DESTROY (JSON *self)	1898	void DESTROY (JSON *self)
1496	CODE:	1899	CODE:
1497	SvREFCNT_dec (self->cb_sk_object);	1900	SvREFCNT_dec (self->cb_sk_object);
1498	SvREFCNT_dec (self->cb_object);	1901	SvREFCNT_dec (self->cb_object);
		1902	SvREFCNT_dec (self->incr_text);
1499		1903
1500	PROTOTYPES: ENABLE	1904	PROTOTYPES: ENABLE
1501		1905
1502	void to_json (SV *scalar)	1906	void encode_json (SV *scalar)
		1907	ALIAS:
		1908	to_json_ = 0
		1909	encode_json = F_UTF8
1503	PPCODE:	1910	PPCODE:
1504	{	1911	{
1505	JSON json = { F_DEFAULT \| F_UTF8 };	1912	JSON json;
		1913	json_init (&json);
		1914	json.flags \|= ix;
1506	XPUSHs (encode_json (scalar, &json));	1915	XPUSHs (encode_json (scalar, &json));
1507	}	1916	}
1508		1917
1509	void from_json (SV *jsonstr)	1918	void decode_json (SV *jsonstr)
		1919	ALIAS:
		1920	from_json_ = 0
		1921	decode_json = F_UTF8
1510	PPCODE:	1922	PPCODE:
1511	{	1923	{
1512	JSON json = { F_DEFAULT \| F_UTF8 };	1924	JSON json;
		1925	json_init (&json);
		1926	json.flags \|= ix;
1513	XPUSHs (decode_json (jsonstr, &json, 0));	1927	XPUSHs (decode_json (jsonstr, &json, 0));
1514	}	1928	}
1515		1929
		1930

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs. Revision 1.100 by root, Sat Sep 5 23:00:56 2009 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.59 by root, Mon Aug 13 16:14:20 2007 UTC vs.
Revision 1.100 by root, Sat Sep 5 23:00:56 2009 UTC