[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 1.102 by root, Sat Oct 10 01:48:50 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	// three extra for rounding, sign, and end of string
		23	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
20		24
21	#define F_ASCII 0x00000001UL	25	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	26	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	27	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	28	#define F_INDENT 0x00000008UL
…		…
27	#define F_SPACE_AFTER 0x00000040UL	31	#define F_SPACE_AFTER 0x00000040UL
28	#define F_ALLOW_NONREF 0x00000100UL	32	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	33	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	34	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	35	#define F_CONV_BLESSED 0x00000800UL
32	#define F_MAXDEPTH 0xf8000000UL
33	#define S_MAXDEPTH 27
34	#define F_MAXSIZE 0x01f00000UL	36	#define F_RELAXED 0x00001000UL
35	#define S_MAXSIZE 20	37	#define F_ALLOW_UNKNOWN 0x00002000UL
36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
37		39
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	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
42	#define F_DEFAULT (9UL << S_MAXDEPTH)
43		41
44	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
45	#define INDENT_STEP 3 // spaces per indentation level	43	#define INDENT_STEP 3 // spaces per indentation level
46		44
47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size	45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
48		46
49	#define SB do {	47	#define SB do {
50	#define SE } while (0)	48	#define SE } while (0)
51		49
52	#if __GNUC__ >= 3	50	#if __GNUC__ >= 3
53	# define expect(expr,value) __builtin_expect ((expr),(value))	51	# define expect(expr,value) __builtin_expect ((expr), (value))
54	# define inline inline	52	# define INLINE static inline
55	#else	53	#else
56	# define expect(expr,value) (expr)	54	# define expect(expr,value) (expr)
57	# define inline static	55	# define INLINE static
58	#endif	56	#endif
59		57
60	#define expect_false(expr) expect ((expr) != 0, 0)	58	#define expect_false(expr) expect ((expr) != 0, 0)
61	#define expect_true(expr) expect ((expr) != 0, 1)	59	#define expect_true(expr) expect ((expr) != 0, 1)
62		60
		61	#define IN_RANGE_INC(type,val,beg,end) \
		62	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		63	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		64
		65	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		66
63	#ifdef USE_ITHREADS	67	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	68	# define JSON_SLOW 1
		69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
65	#else	70	#else
66	# define JSON_SLOW 0	71	# define JSON_SLOW 0
		72	# define JSON_STASH json_stash
67	#endif	73	#endif
68		74
69	static HV json_stash, json_boolean_stash; // JSON::XS::	75	static HV json_stash, json_boolean_stash; // JSON::XS::
70	static SV json_true, json_false;	76	static SV json_true, json_false;
71		77
		78	enum {
		79	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		80	INCR_M_STR, // inside string
		81	INCR_M_BS, // inside backslash
		82	INCR_M_JSON // outside anything, count nesting
		83	};
		84
		85	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		86
72	typedef struct {	87	typedef struct {
73	U32 flags;	88	U32 flags;
		89	U32 max_depth;
		90	STRLEN max_size;
		91
74	SV *cb_object;	92	SV *cb_object;
75	HV *cb_sk_object;	93	HV *cb_sk_object;
		94
		95	// for the incremental parser
		96	SV *incr_text; // the source text so far
		97	STRLEN incr_pos; // the current offset into the text
		98	int incr_nest; // {[]}-nesting level
		99	unsigned char incr_mode;
76	} JSON;	100	} JSON;
		101
		102	INLINE void
		103	json_init (JSON *json)
		104	{
		105	Zero (json, 1, JSON);
		106	json->max_depth = 512;
		107	}
77		108
78	/////////////////////////////////////////////////////////////////////////////	109	/////////////////////////////////////////////////////////////////////////////
79	// utility functions	110	// utility functions
80		111
81	inline void	112	INLINE SV *
		113	get_bool (const char *name)
		114	{
		115	SV *sv = get_sv (name, 1);
		116
		117	SvREADONLY_on (sv);
		118	SvREADONLY_on (SvRV (sv));
		119
		120	return sv;
		121	}
		122
		123	INLINE void
82	shrink (SV *sv)	124	shrink (SV *sv)
83	{	125	{
84	sv_utf8_downgrade (sv, 1);	126	sv_utf8_downgrade (sv, 1);
		127
85	if (SvLEN (sv) > SvCUR (sv) + 1)	128	if (SvLEN (sv) > SvCUR (sv) + 1)
86	{	129	{
87	#ifdef SvPV_shrink_to_cur	130	#ifdef SvPV_shrink_to_cur
88	SvPV_shrink_to_cur (sv);	131	SvPV_shrink_to_cur (sv);
89	#elif defined (SvPV_renew)	132	#elif defined (SvPV_renew)
…		…
95	// decode an utf-8 character and return it, or (UV)-1 in	138	// decode an utf-8 character and return it, or (UV)-1 in
96	// case of an error.	139	// case of an error.
97	// we special-case "safe" characters from U+80 .. U+7FF,	140	// we special-case "safe" characters from U+80 .. U+7FF,
98	// but use the very good perl function to parse anything else.	141	// but use the very good perl function to parse anything else.
99	// note that we never call this function for a ascii codepoints	142	// note that we never call this function for a ascii codepoints
100	inline UV	143	INLINE UV
101	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	144	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
102	{	145	{
103	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	146	if (expect_true (len >= 2
104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	147	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	148	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
106	{	149	{
107	*clen = 2;	150	*clen = 2;
108	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	151	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
109	}	152	}
110	else	153	else
111	{	154	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
112	*clen = (STRLEN)-1;	155	}
113	return (UV)-1;	156
114	}	157	// likewise for encoding, also never called for ascii codepoints
		158	// this function takes advantage of this fact, although current gccs
		159	// seem to optimise the check for >= 0x80 away anyways
		160	INLINE unsigned char *
		161	encode_utf8 (unsigned char *s, UV ch)
		162	{
		163	if (expect_false (ch < 0x000080))
		164	*s++ = ch;
		165	else if (expect_true (ch < 0x000800))
		166	*s++ = 0xc0 \| ( ch >> 6),
		167	*s++ = 0x80 \| ( ch & 0x3f);
		168	else if ( ch < 0x010000)
		169	*s++ = 0xe0 \| ( ch >> 12),
		170	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		171	*s++ = 0x80 \| ( ch & 0x3f);
		172	else if ( ch < 0x110000)
		173	*s++ = 0xf0 \| ( ch >> 18),
		174	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		175	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		176	*s++ = 0x80 \| ( ch & 0x3f);
		177
		178	return s;
		179	}
		180
		181	// convert offset pointer to character index, sv must be string
		182	static STRLEN
		183	ptr_to_index (SV sv, char offset)
		184	{
		185	return SvUTF8 (sv)
		186	? utf8_distance (offset, SvPVX (sv))
		187	: offset - SvPVX (sv);
115	}	188	}
116		189
117	/////////////////////////////////////////////////////////////////////////////	190	/////////////////////////////////////////////////////////////////////////////
118	// encoder	191	// encoder
119		192
…		…
123	char *cur; // SvPVX (sv) + current output position	196	char *cur; // SvPVX (sv) + current output position
124	char *end; // SvEND (sv)	197	char *end; // SvEND (sv)
125	SV *sv; // result scalar	198	SV *sv; // result scalar
126	JSON json;	199	JSON json;
127	U32 indent; // indentation level	200	U32 indent; // indentation level
128	U32 maxdepth; // max. indentation/recursion level	201	UV limit; // escape character values >= this value when encoding
129	} enc_t;	202	} enc_t;
130		203
131	inline void	204	INLINE void
132	need (enc_t *enc, STRLEN len)	205	need (enc_t *enc, STRLEN len)
133	{	206	{
134	if (expect_false (enc->cur + len >= enc->end))	207	if (expect_false (enc->cur + len >= enc->end))
135	{	208	{
136	STRLEN cur = enc->cur - SvPVX (enc->sv);	209	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
137	SvGROW (enc->sv, cur + len + 1);	210	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
138	enc->cur = SvPVX (enc->sv) + cur;	211	enc->cur = SvPVX (enc->sv) + cur;
139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	212	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
140	}	213	}
141	}	214	}
142		215
143	inline void	216	INLINE void
144	encode_ch (enc_t *enc, char ch)	217	encode_ch (enc_t *enc, char ch)
145	{	218	{
146	need (enc, 1);	219	need (enc, 1);
147	*enc->cur++ = ch;	220	*enc->cur++ = ch;
148	}	221	}
…		…
202	{	275	{
203	uch = ch;	276	uch = ch;
204	clen = 1;	277	clen = 1;
205	}	278	}
206		279
207	if (uch > 0x10FFFFUL)	280	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	{	281	{
212	if (uch > 0xFFFFUL)	282	if (uch >= 0x10000UL)
213	{	283	{
		284	if (uch >= 0x110000UL)
		285	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		286
214	need (enc, len += 11);	287	need (enc, len += 11);
215	sprintf (enc->cur, "\\u%04x\\u%04x",	288	sprintf (enc->cur, "\\u%04x\\u%04x",
216	(int)((uch - 0x10000) / 0x400 + 0xD800),	289	(int)((uch - 0x10000) / 0x400 + 0xD800),
217	(int)((uch - 0x10000) % 0x400 + 0xDC00));	290	(int)((uch - 0x10000) % 0x400 + 0xDC00));
218	enc->cur += 12;	291	enc->cur += 12;
219	}	292	}
220	else	293	else
221	{	294	{
222	static char hexdigit [16] = "0123456789abcdef";
223	need (enc, len += 5);	295	need (enc, len += 5);
224	*enc->cur++ = '\\';	296	*enc->cur++ = '\\';
225	*enc->cur++ = 'u';	297	*enc->cur++ = 'u';
226	*enc->cur++ = hexdigit [ uch >> 12 ];	298	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
227	*enc->cur++ = hexdigit [(uch >> 8) & 15];	299	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
228	*enc->cur++ = hexdigit [(uch >> 4) & 15];	300	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
229	*enc->cur++ = hexdigit [(uch >> 0) & 15];	301	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
230	}	302	}
231		303
232	str += clen;	304	str += clen;
233	}	305	}
234	else if (enc->json.flags & F_LATIN1)	306	else if (enc->json.flags & F_LATIN1)
…		…
246	while (--clen);	318	while (--clen);
247	}	319	}
248	else	320	else
249	{	321	{
250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	322	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	323	enc->cur = encode_utf8 (enc->cur, uch);
252	++str;	324	++str;
253	}	325	}
254	}	326	}
255	}	327	}
256	}	328	}
257		329
258	--len;	330	--len;
259	}	331	}
260	}	332	}
261		333
262	inline void	334	INLINE void
263	encode_indent (enc_t *enc)	335	encode_indent (enc_t *enc)
264	{	336	{
265	if (enc->json.flags & F_INDENT)	337	if (enc->json.flags & F_INDENT)
266	{	338	{
267	int spaces = enc->indent * INDENT_STEP;	339	int spaces = enc->indent * INDENT_STEP;
…		…
270	memset (enc->cur, ' ', spaces);	342	memset (enc->cur, ' ', spaces);
271	enc->cur += spaces;	343	enc->cur += spaces;
272	}	344	}
273	}	345	}
274		346
275	inline void	347	INLINE void
276	encode_space (enc_t *enc)	348	encode_space (enc_t *enc)
277	{	349	{
278	need (enc, 1);	350	need (enc, 1);
279	encode_ch (enc, ' ');	351	encode_ch (enc, ' ');
280	}	352	}
281		353
282	inline void	354	INLINE void
283	encode_nl (enc_t *enc)	355	encode_nl (enc_t *enc)
284	{	356	{
285	if (enc->json.flags & F_INDENT)	357	if (enc->json.flags & F_INDENT)
286	{	358	{
287	need (enc, 1);	359	need (enc, 1);
288	encode_ch (enc, '\n');	360	encode_ch (enc, '\n');
289	}	361	}
290	}	362	}
291		363
292	inline void	364	INLINE void
293	encode_comma (enc_t *enc)	365	encode_comma (enc_t *enc)
294	{	366	{
295	encode_ch (enc, ',');	367	encode_ch (enc, ',');
296		368
297	if (enc->json.flags & F_INDENT)	369	if (enc->json.flags & F_INDENT)
…		…
305	static void	377	static void
306	encode_av (enc_t enc, AV av)	378	encode_av (enc_t enc, AV av)
307	{	379	{
308	int i, len = av_len (av);	380	int i, len = av_len (av);
309		381
310	if (enc->indent >= enc->maxdepth)	382	if (enc->indent >= enc->json.max_depth)
311	croak ("data structure too deep (hit recursion limit)");	383	croak (ERR_NESTING_EXCEEDED);
312		384
313	encode_ch (enc, '['); encode_nl (enc);	385	encode_ch (enc, '[');
314	++enc->indent;	386
		387	if (len >= 0)
		388	{
		389	encode_nl (enc); ++enc->indent;
315		390
316	for (i = 0; i <= len; ++i)	391	for (i = 0; i <= len; ++i)
317	{	392	{
318	SV **svp = av_fetch (av, i, 0);	393	SV **svp = av_fetch (av, i, 0);
319		394
320	encode_indent (enc);	395	encode_indent (enc);
321		396
322	if (svp)	397	if (svp)
323	encode_sv (enc, *svp);	398	encode_sv (enc, *svp);
324	else	399	else
325	encode_str (enc, "null", 4, 0);	400	encode_str (enc, "null", 4, 0);
326		401
327	if (i < len)	402	if (i < len)
328	encode_comma (enc);	403	encode_comma (enc);
329	}	404	}
330		405
		406	encode_nl (enc); --enc->indent; encode_indent (enc);
		407	}
		408
331	encode_nl (enc);	409	encode_ch (enc, ']');
332
333	--enc->indent;
334	encode_indent (enc); encode_ch (enc, ']');
335	}	410	}
336		411
337	static void	412	static void
338	encode_he (enc_t enc, HE he)	413	encode_hk (enc_t enc, HE he)
339	{	414	{
340	encode_ch (enc, '"');	415	encode_ch (enc, '"');
341		416
342	if (HeKLEN (he) == HEf_SVKEY)	417	if (HeKLEN (he) == HEf_SVKEY)
343	{	418	{
…		…
356	encode_ch (enc, '"');	431	encode_ch (enc, '"');
357		432
358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	433	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
359	encode_ch (enc, ':');	434	encode_ch (enc, ':');
360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	435	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
361	encode_sv (enc, HeVAL (he));
362	}	436	}
363		437
364	// compare hash entries, used when all keys are bytestrings	438	// compare hash entries, used when all keys are bytestrings
365	static int	439	static int
366	he_cmp_fast (const void a_, const void b_)	440	he_cmp_fast (const void a_, const void b_)
…		…
371	HE b = (HE **)b_;	445	HE b = (HE **)b_;
372		446
373	STRLEN la = HeKLEN (a);	447	STRLEN la = HeKLEN (a);
374	STRLEN lb = HeKLEN (b);	448	STRLEN lb = HeKLEN (b);
375		449
376	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	450	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
377	cmp = la - lb;	451	cmp = lb - la;
378		452
379	return cmp;	453	return cmp;
380	}	454	}
381		455
382	// compare hash entries, used when some keys are sv's or utf-x	456	// compare hash entries, used when some keys are sv's or utf-x
383	static int	457	static int
384	he_cmp_slow (const void a, const void b)	458	he_cmp_slow (const void a, const void b)
385	{	459	{
386	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	460	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
387	}	461	}
388		462
389	static void	463	static void
390	encode_hv (enc_t enc, HV hv)	464	encode_hv (enc_t enc, HV hv)
391	{	465	{
392	int count, i;	466	HE *he;
393		467
394	if (enc->indent >= enc->maxdepth)	468	if (enc->indent >= enc->json.max_depth)
395	croak ("data structure too deep (hit recursion limit)");	469	croak (ERR_NESTING_EXCEEDED);
396		470
397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	471	encode_ch (enc, '{');
398		472
399	if ((count = hv_iterinit (hv)))
400	{
401	// for canonical output we have to sort by keys first	473	// for canonical output we have to sort by keys first
402	// actually, this is mostly due to the stupid so-called	474	// actually, this is mostly due to the stupid so-called
403	// security workaround added somewhere in 5.8.x.	475	// security workaround added somewhere in 5.8.x
404	// that randomises hash orderings	476	// that randomises hash orderings
405	if (enc->json.flags & F_CANONICAL)	477	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		478	{
		479	int count = hv_iterinit (hv);
		480
		481	if (SvMAGICAL (hv))
406	{	482	{
		483	// need to count by iterating. could improve by dynamically building the vector below
		484	// but I don't care for the speed of this special case.
		485	// note also that we will run into undefined behaviour when the two iterations
		486	// do not result in the same count, something I might care for in some later release.
		487
		488	count = 0;
		489	while (hv_iternext (hv))
		490	++count;
		491
		492	hv_iterinit (hv);
		493	}
		494
		495	if (count)
		496	{
407	int fast = 1;	497	int i, fast = 1;
408	HE *he;
409	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	498	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
410	HE *hes = _alloca (count sizeof (HE));	499	HE *hes = _alloca (count sizeof (HE));
411	#else	500	#else
412	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	501	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
413	#endif	502	#endif
…		…
440		529
441	FREETMPS;	530	FREETMPS;
442	LEAVE;	531	LEAVE;
443	}	532	}
444		533
445	for (i = 0; i < count; ++i)	534	encode_nl (enc); ++enc->indent;
		535
		536	while (count--)
446	{	537	{
447	encode_indent (enc);	538	encode_indent (enc);
		539	he = hes [count];
448	encode_he (enc, hes [i]);	540	encode_hk (enc, he);
		541	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
449		542
450	if (i < count - 1)	543	if (count)
451	encode_comma (enc);	544	encode_comma (enc);
452	}	545	}
453		546
454	encode_nl (enc);	547	encode_nl (enc); --enc->indent; encode_indent (enc);
455	}	548	}
		549	}
456	else	550	else
		551	{
		552	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		553	if ((he = hv_iternext (hv)))
457	{	554	{
458	HE *he = hv_iternext (hv);	555	encode_nl (enc); ++enc->indent;
459		556
460	for (;;)	557	for (;;)
461	{	558	{
462	encode_indent (enc);	559	encode_indent (enc);
463	encode_he (enc, he);	560	encode_hk (enc, he);
		561	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
464		562
465	if (!(he = hv_iternext (hv)))	563	if (!(he = hv_iternext (hv)))
466	break;	564	break;
467		565
468	encode_comma (enc);	566	encode_comma (enc);
469	}	567	}
470		568
471	encode_nl (enc);	569	encode_nl (enc); --enc->indent; encode_indent (enc);
472	}	570	}
473	}	571	}
474		572
475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	573	encode_ch (enc, '}');
476	}	574	}
477		575
478	// encode objects, arrays and special \0=false and \1=true values.	576	// encode objects, arrays and special \0=false and \1=true values.
479	static void	577	static void
480	encode_rv (enc_t enc, SV sv)	578	encode_rv (enc_t enc, SV sv)
…		…
557		655
558	if (len == 1 && *pv == '1')	656	if (len == 1 && *pv == '1')
559	encode_str (enc, "true", 4, 0);	657	encode_str (enc, "true", 4, 0);
560	else if (len == 1 && *pv == '0')	658	else if (len == 1 && *pv == '0')
561	encode_str (enc, "false", 5, 0);	659	encode_str (enc, "false", 5, 0);
		660	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		661	encode_str (enc, "null", 4, 0);
562	else	662	else
563	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	663	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
564	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	664	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
565	}	665	}
		666	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		667	encode_str (enc, "null", 4, 0);
566	else	668	else
567	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	669	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
568	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	670	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
569	}	671	}
570		672
…		…
588	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	690	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
589	enc->cur += strlen (enc->cur);	691	enc->cur += strlen (enc->cur);
590	}	692	}
591	else if (SvIOKp (sv))	693	else if (SvIOKp (sv))
592	{	694	{
593	// we assume we can always read an IV as a UV	695	// we assume we can always read an IV as a UV and vice versa
594	if (SvUV (sv) & ~(UV)0x7fff)	696	// we assume two's complement
595	{	697	// we assume no aliasing issues in the union
596	// large integer, use the (rather slow) snprintf way.	698	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
597	need (enc, sizeof (UV) * 3);	699	: 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	{	700	{
605	// optimise the "small number case"	701	// optimise the "small number case"
606	// code will likely be branchless and use only a single multiplication	702	// code will likely be branchless and use only a single multiplication
		703	// works for numbers up to 59074
607	I32 i = SvIV (sv);	704	I32 i = SvIVX (sv);
608	U32 u;	705	U32 u;
609	char digit, nz = 0;	706	char digit, nz = 0;
610		707
611	need (enc, 6);	708	need (enc, 6);
612		709
…		…
618		715
619	// now output digit by digit, each time masking out the integer part	716	// 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,	717	// and multiplying by 5 while moving the decimal point one to the right,
621	// resulting in a net multiplication by 10.	718	// resulting in a net multiplication by 10.
622	// we always write the digit to memory but conditionally increment	719	// we always write the digit to memory but conditionally increment
623	// the pointer, to ease the usage of conditional move instructions.	720	// 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;	721	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;	722	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;	723	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;	724	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'	725	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
629	}	726	}
		727	else
		728	{
		729	// large integer, use the (rather slow) snprintf way.
		730	need (enc, IVUV_MAXCHARS);
		731	enc->cur +=
		732	SvIsUV(sv)
		733	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		734	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		735	}
630	}	736	}
631	else if (SvROK (sv))	737	else if (SvROK (sv))
632	encode_rv (enc, SvRV (sv));	738	encode_rv (enc, SvRV (sv));
633	else if (!SvOK (sv))	739	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
634	encode_str (enc, "null", 4, 0);	740	encode_str (enc, "null", 4, 0);
635	else	741	else
636	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	742	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
637	SvPV_nolen (sv), SvFLAGS (sv));	743	SvPV_nolen (sv), SvFLAGS (sv));
638	}	744	}
…		…
648	enc.json = *json;	754	enc.json = *json;
649	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	755	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
650	enc.cur = SvPVX (enc.sv);	756	enc.cur = SvPVX (enc.sv);
651	enc.end = SvEND (enc.sv);	757	enc.end = SvEND (enc.sv);
652	enc.indent = 0;	758	enc.indent = 0;
653	enc.maxdepth = DEC_DEPTH (enc.json.flags);	759	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		760	: enc.json.flags & F_LATIN1 ? 0x000100UL
		761	: 0x110000UL;
654		762
655	SvPOK_only (enc.sv);	763	SvPOK_only (enc.sv);
656	encode_sv (&enc, scalar);	764	encode_sv (&enc, scalar);
		765	encode_nl (&enc);
657		766
658	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	767	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
659	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	768	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
660		769
661	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	770	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
679	JSON json;	788	JSON json;
680	U32 depth; // recursion depth	789	U32 depth; // recursion depth
681	U32 maxdepth; // recursion depth limit	790	U32 maxdepth; // recursion depth limit
682	} dec_t;	791	} dec_t;
683		792
684	inline void	793	INLINE void
		794	decode_comment (dec_t *dec)
		795	{
		796	// only '#'-style comments allowed a.t.m.
		797
		798	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		799	++dec->cur;
		800	}
		801
		802	INLINE void
685	decode_ws (dec_t *dec)	803	decode_ws (dec_t *dec)
686	{	804	{
687	for (;;)	805	for (;;)
688	{	806	{
689	char ch = *dec->cur;	807	char ch = *dec->cur;
690		808
691	if (ch > 0x20	809	if (ch > 0x20)
		810	{
		811	if (expect_false (ch == '#'))
		812	{
		813	if (dec->json.flags & F_RELAXED)
		814	decode_comment (dec);
		815	else
		816	break;
		817	}
		818	else
		819	break;
		820	}
692	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	821	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
693	break;	822	break; // parse error, but let higher level handle it, gives better error messages
694		823
695	++dec->cur;	824	++dec->cur;
696	}	825	}
697	}	826	}
698		827
…		…
702	if (*dec->cur != ch) \	831	if (*dec->cur != ch) \
703	ERR (# ch " expected"); \	832	ERR (# ch " expected"); \
704	++dec->cur; \	833	++dec->cur; \
705	SE	834	SE
706		835
707	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	836	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
708	#define DEC_DEC_DEPTH --dec->depth	837	#define DEC_DEC_DEPTH --dec->depth
709		838
710	static SV decode_sv (dec_t dec);	839	static SV decode_sv (dec_t dec);
711		840
712	static signed char decode_hexdigit[256];	841	static signed char decode_hexdigit[256];
…		…
804		933
805	if (hi >= 0x80)	934	if (hi >= 0x80)
806	{	935	{
807	utf8 = 1;	936	utf8 = 1;
808		937
809	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	938	cur = encode_utf8 (cur, hi);
810	}	939	}
811	else	940	else
812	*cur++ = hi;	941	*cur++ = hi;
813	}	942	}
814	break;	943	break;
…		…
816	default:	945	default:
817	--dec_cur;	946	--dec_cur;
818	ERR ("illegal backslash escape sequence in string");	947	ERR ("illegal backslash escape sequence in string");
819	}	948	}
820	}	949	}
821	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	950	else if (expect_true (ch >= 0x20 && ch < 0x80))
822	*cur++ = ch;	951	*cur++ = ch;
823	else if (ch >= 0x80)	952	else if (ch >= 0x80)
824	{	953	{
825	STRLEN clen;	954	STRLEN clen;
826	UV uch;
827		955
828	--dec_cur;	956	--dec_cur;
829		957
830	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	958	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
831	if (clen == (STRLEN)-1)	959	if (clen == (STRLEN)-1)
832	ERR ("malformed UTF-8 character in JSON string");	960	ERR ("malformed UTF-8 character in JSON string");
833		961
834	do	962	do
835	cur++ = dec_cur++;	963	cur++ = dec_cur++;
…		…
852	{	980	{
853	STRLEN len = cur - buf;	981	STRLEN len = cur - buf;
854		982
855	if (sv)	983	if (sv)
856	{	984	{
857	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
858	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	990	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
859	SvCUR_set (sv, SvCUR (sv) + len);	991	SvCUR_set (sv, SvCUR (sv) + len);
860	}	992	}
861	else	993	else
862	sv = newSVpvn (buf, len);	994	sv = newSVpvn (buf, len);
…		…
949		1081
950	if (!is_nv)	1082	if (!is_nv)
951	{	1083	{
952	int len = dec->cur - start;	1084	int len = dec->cur - start;
953		1085
954	// 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
955	if (*start == '-')	1087	if (*start == '-')
956	switch (len)	1088	switch (len)
957	{	1089	{
958	case 2: return newSViv (-( start [1] - '0' * 1));	1090	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
959	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1091	case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
960	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));
961	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));
962	}	1095	}
963	else	1096	else
964	switch (len)	1097	switch (len)
965	{	1098	{
966	case 1: return newSViv ( start [0] - '0' * 1);	1099	case 1: return newSViv ( start [0] - '0' * 1);
967	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1100	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);	1101	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);	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);
970	}	1104	}
971		1105
972	{	1106	{
973	UV uv;	1107	UV uv;
974	int numtype = grok_number (start, len, &uv);	1108	int numtype = grok_number (start, len, &uv);
…		…
1035		1169
1036	if (*dec->cur != ',')	1170	if (*dec->cur != ',')
1037	ERR (", or ] expected while parsing array");	1171	ERR (", or ] expected while parsing array");
1038		1172
1039	++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	}
1040	}	1182	}
1041		1183
1042	DEC_DEC_DEPTH;	1184	DEC_DEC_DEPTH;
1043	return newRV_noinc ((SV *)av);	1185	return newRV_noinc ((SV *)av);
1044		1186
…		…
1060	if (*dec->cur == '}')	1202	if (*dec->cur == '}')
1061	++dec->cur;	1203	++dec->cur;
1062	else	1204	else
1063	for (;;)	1205	for (;;)
1064	{	1206	{
1065	decode_ws (dec); EXPECT_CH ('"');	1207	EXPECT_CH ('"');
1066		1208
1067	// heuristic: assume that	1209	// heuristic: assume that
1068	// a) decode_str + hv_store_ent are abysmally slow.	1210	// a) decode_str + hv_store_ent are abysmally slow.
1069	// b) most hash keys are short, simple ascii text.	1211	// b) most hash keys are short, simple ascii text.
1070	// => try to "fast-match" such strings to avoid	1212	// => try to "fast-match" such strings to avoid
…		…
1074	char *p = dec->cur;	1216	char *p = dec->cur;
1075	char *e = p + 24; // only try up to 24 bytes	1217	char *e = p + 24; // only try up to 24 bytes
1076		1218
1077	for (;;)	1219	for (;;)
1078	{	1220	{
1079	// the >= 0x80 is true on most architectures	1221	// the >= 0x80 is false on most architectures
1080	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1222	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1081	{	1223	{
1082	// slow path, back up and use decode_str	1224	// slow path, back up and use decode_str
1083	SV *key = decode_str (dec);	1225	SV *key = decode_str (dec);
1084	if (!key)	1226	if (!key)
1085	goto fail;	1227	goto fail;
1086		1228
1087	decode_ws (dec); EXPECT_CH (':');	1229	decode_ws (dec); EXPECT_CH (':');
1088		1230
		1231	decode_ws (dec);
1089	value = decode_sv (dec);	1232	value = decode_sv (dec);
1090	if (!value)	1233	if (!value)
1091	{	1234	{
1092	SvREFCNT_dec (key);	1235	SvREFCNT_dec (key);
1093	goto fail;	1236	goto fail;
…		…
1105	int len = p - key;	1248	int len = p - key;
1106	dec->cur = p + 1;	1249	dec->cur = p + 1;
1107		1250
1108	decode_ws (dec); EXPECT_CH (':');	1251	decode_ws (dec); EXPECT_CH (':');
1109		1252
		1253	decode_ws (dec);
1110	value = decode_sv (dec);	1254	value = decode_sv (dec);
1111	if (!value)	1255	if (!value)
1112	goto fail;	1256	goto fail;
1113		1257
1114	hv_store (hv, key, len, value, 0);	1258	hv_store (hv, key, len, value, 0);
…		…
1130		1274
1131	if (*dec->cur != ',')	1275	if (*dec->cur != ',')
1132	ERR (", or } expected while parsing object/hash");	1276	ERR (", or } expected while parsing object/hash");
1133		1277
1134	++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	}
1135	}	1287	}
1136		1288
1137	DEC_DEC_DEPTH;	1289	DEC_DEC_DEPTH;
1138	sv = newRV_noinc ((SV *)hv);	1290	sv = newRV_noinc ((SV *)hv);
1139		1291
…		…
1204	}	1356	}
1205		1357
1206	static SV *	1358	static SV *
1207	decode_sv (dec_t *dec)	1359	decode_sv (dec_t *dec)
1208	{	1360	{
1209	decode_ws (dec);
1210
1211	// the beauty of JSON: you need exactly one character lookahead	1361	// the beauty of JSON: you need exactly one character lookahead
1212	// to parse anything.	1362	// to parse everything.
1213	switch (*dec->cur)	1363	switch (*dec->cur)
1214	{	1364	{
1215	case '"': ++dec->cur; return decode_str (dec);	1365	case '"': ++dec->cur; return decode_str (dec);
1216	case '[': ++dec->cur; return decode_av (dec);	1366	case '[': ++dec->cur; return decode_av (dec);
1217	case '{': ++dec->cur; return decode_hv (dec);	1367	case '{': ++dec->cur; return decode_hv (dec);
1218		1368
1219	case '-':	1369	case '-':
1220	case '0': case '1': case '2': case '3': case '4':	1370	case '0': case '1': case '2': case '3': case '4':
1221	case '5': case '6': case '7': case '8': case '9':	1371	case '5': case '6': case '7': case '8': case '9':
1222	return decode_num (dec);	1372	return decode_num (dec);
1223		1373
1224	case 't':	1374	case 't':
1225	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1375	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1226	{	1376	{
1227	dec->cur += 4;	1377	dec->cur += 4;
		1378	#if JSON_SLOW
		1379	json_true = get_bool ("JSON::XS::true");
		1380	#endif
1228	return SvREFCNT_inc (json_true);	1381	return newSVsv (json_true);
1229	}	1382	}
1230	else	1383	else
1231	ERR ("'true' expected");	1384	ERR ("'true' expected");
1232		1385
1233	break;	1386	break;
1234		1387
1235	case 'f':	1388	case 'f':
1236	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1389	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1237	{	1390	{
1238	dec->cur += 5;	1391	dec->cur += 5;
		1392	#if JSON_SLOW
		1393	json_false = get_bool ("JSON::XS::false");
		1394	#endif
1239	return SvREFCNT_inc (json_false);	1395	return newSVsv (json_false);
1240	}	1396	}
1241	else	1397	else
1242	ERR ("'false' expected");	1398	ERR ("'false' expected");
1243		1399
1244	break;	1400	break;
…		…
1262	fail:	1418	fail:
1263	return 0;	1419	return 0;
1264	}	1420	}
1265		1421
1266	static SV *	1422	static SV *
1267	decode_json (SV string, JSON json, UV *offset_return)	1423	decode_json (SV string, JSON json, char **offset_return)
1268	{	1424	{
1269	dec_t dec;	1425	dec_t dec;
1270	UV offset;
1271	SV *sv;	1426	SV *sv;
1272		1427
		1428	/* work around bugs in 5.10 where manipulating magic values
		1429	* will perl ignore the magic in subsequent accesses
		1430	*/
1273	SvGETMAGIC (string);	1431	/SvGETMAGIC (string);/
		1432	if (SvMAGICAL (string))
		1433	string = sv_2mortal (newSVsv (string));
		1434
1274	SvUPGRADE (string, SVt_PV);	1435	SvUPGRADE (string, SVt_PV);
1275		1436
1276	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)
1277	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",
1278	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1456	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1457	}
1279		1458
1280	if (json->flags & F_UTF8)	1459	if (json->flags & F_UTF8)
1281	sv_utf8_downgrade (string, 0);	1460	sv_utf8_downgrade (string, 0);
1282	else	1461	else
1283	sv_utf8_upgrade (string);	1462	sv_utf8_upgrade (string);
1284		1463
1285	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1464	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1286		1465
1287	dec.json = *json;	1466	dec.json = *json;
1288	dec.cur = SvPVX (string);	1467	dec.cur = SvPVX (string);
1289	dec.end = SvEND (string);	1468	dec.end = SvEND (string);
1290	dec.err = 0;	1469	dec.err = 0;
1291	dec.depth = 0;	1470	dec.depth = 0;
1292	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1293		1471
1294	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1472	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1295	dec.json.flags \|= F_HOOK;	1473	dec.json.flags \|= F_HOOK;
1296		1474
1297	*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);
1298	sv = decode_sv (&dec);	1478	sv = decode_sv (&dec);
		1479
		1480	if (offset_return)
		1481	*offset_return = dec.cur;
1299		1482
1300	if (!(offset_return \|\| !sv))	1483	if (!(offset_return \|\| !sv))
1301	{	1484	{
1302	// check for trailing garbage	1485	// check for trailing garbage
1303	decode_ws (&dec);	1486	decode_ws (&dec);
…		…
1306	{	1489	{
1307	dec.err = "garbage after JSON object";	1490	dec.err = "garbage after JSON object";
1308	SvREFCNT_dec (sv);	1491	SvREFCNT_dec (sv);
1309	sv = 0;	1492	sv = 0;
1310	}	1493	}
1311	}
1312
1313	if (offset_return \|\| !sv)
1314	{
1315	offset = dec.json.flags & F_UTF8
1316	? dec.cur - SvPVX (string)
1317	: utf8_distance (dec.cur, SvPVX (string));
1318
1319	if (offset_return)
1320	*offset_return = offset;
1321	}	1494	}
1322		1495
1323	if (!sv)	1496	if (!sv)
1324	{	1497	{
1325	SV *uni = sv_newmortal ();	1498	SV *uni = sv_newmortal ();
…		…
1331	SAVEVPTR (PL_curcop);	1504	SAVEVPTR (PL_curcop);
1332	PL_curcop = &cop;	1505	PL_curcop = &cop;
1333	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);
1334	LEAVE;	1507	LEAVE;
1335		1508
1336	croak ("%s, at character offset %d [\"%s\"]",	1509	croak ("%s, at character offset %d (before \"%s\")",
1337	dec.err,	1510	dec.err,
1338	(int)offset,	1511	ptr_to_index (string, dec.cur),
1339	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1512	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1340	}	1513	}
1341		1514
1342	sv = sv_2mortal (sv);	1515	sv = sv_2mortal (sv);
1343		1516
1344	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1517	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1345	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)");
1346		1519
1347	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
1348	}	1638	}
1349		1639
1350	/////////////////////////////////////////////////////////////////////////////	1640	/////////////////////////////////////////////////////////////////////////////
1351	// XS interface functions	1641	// XS interface functions
1352		1642
…		…
1364	: -1;	1654	: -1;
1365		1655
1366	json_stash = gv_stashpv ("JSON::XS" , 1);	1656	json_stash = gv_stashpv ("JSON::XS" , 1);
1367	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1657	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1368		1658
1369	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1659	json_true = get_bool ("JSON::XS::true");
1370	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 */
1371	}	1663	}
1372		1664
1373	PROTOTYPES: DISABLE	1665	PROTOTYPES: DISABLE
1374		1666
1375	void CLONE (...)	1667	void CLONE (...)
…		…
1378	json_boolean_stash = 0;	1670	json_boolean_stash = 0;
1379		1671
1380	void new (char *klass)	1672	void new (char *klass)
1381	PPCODE:	1673	PPCODE:
1382	{	1674	{
1383	HV *stash = !JSON_SLOW \|\| json_stash
1384	? json_stash
1385	: gv_stashpv ("JSON::XS", 1);
1386	SV *pv = NEWSV (0, sizeof (JSON));	1675	SV *pv = NEWSV (0, sizeof (JSON));
1387	SvPOK_only (pv);	1676	SvPOK_only (pv);
1388	Zero (SvPVX (pv), 1, JSON);	1677	json_init ((JSON *)SvPVX (pv));
1389	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1678	XPUSHs (sv_2mortal (sv_bless (
1390	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));	1679	newRV_noinc (pv),
		1680	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1681	)));
1391	}	1682	}
1392		1683
1393	void ascii (JSON *self, int enable = 1)	1684	void ascii (JSON *self, int enable = 1)
1394	ALIAS:	1685	ALIAS:
1395	ascii = F_ASCII	1686	ascii = F_ASCII
…		…
1402	pretty = F_PRETTY	1693	pretty = F_PRETTY
1403	allow_nonref = F_ALLOW_NONREF	1694	allow_nonref = F_ALLOW_NONREF
1404	shrink = F_SHRINK	1695	shrink = F_SHRINK
1405	allow_blessed = F_ALLOW_BLESSED	1696	allow_blessed = F_ALLOW_BLESSED
1406	convert_blessed = F_CONV_BLESSED	1697	convert_blessed = F_CONV_BLESSED
		1698	relaxed = F_RELAXED
		1699	allow_unknown = F_ALLOW_UNKNOWN
1407	PPCODE:	1700	PPCODE:
1408	{	1701	{
1409	if (enable)	1702	if (enable)
1410	self->flags \|= ix;	1703	self->flags \|= ix;
1411	else	1704	else
1412	self->flags &= ~ix;	1705	self->flags &= ~ix;
1413		1706
1414	XPUSHs (ST (0));	1707	XPUSHs (ST (0));
1415	}	1708	}
1416		1709
1417	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
1418	PPCODE:	1725	PPCODE:
1419	{	1726	XPUSHs (boolSV (self->flags & ix));
1420	UV log2 = 0;
1421		1727
1422	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1728	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1423		1729	PPCODE:
1424	while ((1UL << log2) < max_depth)	1730	self->max_depth = max_depth;
1425	++log2;
1426
1427	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1428
1429	XPUSHs (ST (0));	1731	XPUSHs (ST (0));
1430	}
1431		1732
		1733	U32 get_max_depth (JSON *self)
		1734	CODE:
		1735	RETVAL = self->max_depth;
		1736	OUTPUT:
		1737	RETVAL
		1738
1432	void max_size (JSON *self, UV max_size = 0)	1739	void max_size (JSON *self, U32 max_size = 0)
1433	PPCODE:	1740	PPCODE:
1434	{	1741	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));	1742	XPUSHs (ST (0));
1446	}	1743
		1744	int get_max_size (JSON *self)
		1745	CODE:
		1746	RETVAL = self->max_size;
		1747	OUTPUT:
		1748	RETVAL
1447		1749
1448	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1750	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1449	PPCODE:	1751	PPCODE:
1450	{	1752	{
1451	SvREFCNT_dec (self->cb_object);	1753	SvREFCNT_dec (self->cb_object);
…		…
1485	XPUSHs (decode_json (jsonstr, self, 0));	1787	XPUSHs (decode_json (jsonstr, self, 0));
1486		1788
1487	void decode_prefix (JSON self, SV jsonstr)	1789	void decode_prefix (JSON self, SV jsonstr)
1488	PPCODE:	1790	PPCODE:
1489	{	1791	{
1490	UV offset;	1792	char *offset;
1491	EXTEND (SP, 2);	1793	EXTEND (SP, 2);
1492	PUSHs (decode_json (jsonstr, self, &offset));	1794	PUSHs (decode_json (jsonstr, self, &offset));
1493	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;
1494	}	1896	}
1495		1897
1496	void DESTROY (JSON *self)	1898	void DESTROY (JSON *self)
1497	CODE:	1899	CODE:
1498	SvREFCNT_dec (self->cb_sk_object);	1900	SvREFCNT_dec (self->cb_sk_object);
1499	SvREFCNT_dec (self->cb_object);	1901	SvREFCNT_dec (self->cb_object);
		1902	SvREFCNT_dec (self->incr_text);
1500		1903
1501	PROTOTYPES: ENABLE	1904	PROTOTYPES: ENABLE
1502		1905
1503	void to_json (SV *scalar)	1906	void encode_json (SV *scalar)
		1907	ALIAS:
		1908	to_json_ = 0
		1909	encode_json = F_UTF8
1504	PPCODE:	1910	PPCODE:
1505	{	1911	{
1506	JSON json = { F_DEFAULT \| F_UTF8 };	1912	JSON json;
		1913	json_init (&json);
		1914	json.flags \|= ix;
1507	XPUSHs (encode_json (scalar, &json));	1915	XPUSHs (encode_json (scalar, &json));
1508	}	1916	}
1509		1917
1510	void from_json (SV *jsonstr)	1918	void decode_json (SV *jsonstr)
		1919	ALIAS:
		1920	from_json_ = 0
		1921	decode_json = F_UTF8
1511	PPCODE:	1922	PPCODE:
1512	{	1923	{
1513	JSON json = { F_DEFAULT \| F_UTF8 };	1924	JSON json;
		1925	json_init (&json);
		1926	json.flags \|= ix;
1514	XPUSHs (decode_json (jsonstr, &json, 0));	1927	XPUSHs (decode_json (jsonstr, &json, 0));
1515	}	1928	}
1516		1929
		1930

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 1.102 by root, Sat Oct 10 01:48:50 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 1.102 by root, Sat Oct 10 01:48:50 2009 UTC