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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.60 by root, Mon Aug 13 16:19:13 2007 UTC vs.
Revision 1.103 by root, Wed Jan 6 08:01:39 2010 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)
		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?)"
62		66
63	#ifdef USE_ITHREADS	67	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	68	# define JSON_SLOW 1
65	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))	69	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
66	#else	70	#else
…		…
69	#endif	73	#endif
70		74
71	static HV json_stash, json_boolean_stash; // JSON::XS::	75	static HV json_stash, json_boolean_stash; // JSON::XS::
72	static SV json_true, json_false;	76	static SV json_true, json_false;
73		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_C0, // inside comment in initial whitespace sequence
		83	INCR_M_C1, // inside comment in other places
		84	INCR_M_JSON // outside anything, count nesting
		85	};
		86
		87	#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
		88
74	typedef struct {	89	typedef struct {
75	U32 flags;	90	U32 flags;
		91	U32 max_depth;
		92	STRLEN max_size;
		93
76	SV *cb_object;	94	SV *cb_object;
77	HV *cb_sk_object;	95	HV *cb_sk_object;
		96
		97	// for the incremental parser
		98	SV *incr_text; // the source text so far
		99	STRLEN incr_pos; // the current offset into the text
		100	int incr_nest; // {[]}-nesting level
		101	unsigned char incr_mode;
78	} JSON;	102	} JSON;
		103
		104	INLINE void
		105	json_init (JSON *json)
		106	{
		107	Zero (json, 1, JSON);
		108	json->max_depth = 512;
		109	}
79		110
80	/////////////////////////////////////////////////////////////////////////////	111	/////////////////////////////////////////////////////////////////////////////
81	// utility functions	112	// utility functions
82		113
83	inline void	114	INLINE SV *
		115	get_bool (const char *name)
		116	{
		117	SV *sv = get_sv (name, 1);
		118
		119	SvREADONLY_on (sv);
		120	SvREADONLY_on (SvRV (sv));
		121
		122	return sv;
		123	}
		124
		125	INLINE void
84	shrink (SV *sv)	126	shrink (SV *sv)
85	{	127	{
86	sv_utf8_downgrade (sv, 1);	128	sv_utf8_downgrade (sv, 1);
		129
87	if (SvLEN (sv) > SvCUR (sv) + 1)	130	if (SvLEN (sv) > SvCUR (sv) + 1)
88	{	131	{
89	#ifdef SvPV_shrink_to_cur	132	#ifdef SvPV_shrink_to_cur
90	SvPV_shrink_to_cur (sv);	133	SvPV_shrink_to_cur (sv);
91	#elif defined (SvPV_renew)	134	#elif defined (SvPV_renew)
…		…
97	// decode an utf-8 character and return it, or (UV)-1 in	140	// decode an utf-8 character and return it, or (UV)-1 in
98	// case of an error.	141	// case of an error.
99	// we special-case "safe" characters from U+80 .. U+7FF,	142	// we special-case "safe" characters from U+80 .. U+7FF,
100	// but use the very good perl function to parse anything else.	143	// but use the very good perl function to parse anything else.
101	// note that we never call this function for a ascii codepoints	144	// note that we never call this function for a ascii codepoints
102	inline UV	145	INLINE UV
103	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	146	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
104	{	147	{
105	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	148	if (expect_true (len >= 2
106	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	149	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
107	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	150	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
108	{	151	{
109	*clen = 2;	152	*clen = 2;
110	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	153	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
111	}	154	}
112	else	155	else
113	{	156	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
114	*clen = (STRLEN)-1;	157	}
115	return (UV)-1;	158
116	}	159	// likewise for encoding, also never called for ascii codepoints
		160	// this function takes advantage of this fact, although current gccs
		161	// seem to optimise the check for >= 0x80 away anyways
		162	INLINE unsigned char *
		163	encode_utf8 (unsigned char *s, UV ch)
		164	{
		165	if (expect_false (ch < 0x000080))
		166	*s++ = ch;
		167	else if (expect_true (ch < 0x000800))
		168	*s++ = 0xc0 \| ( ch >> 6),
		169	*s++ = 0x80 \| ( ch & 0x3f);
		170	else if ( ch < 0x010000)
		171	*s++ = 0xe0 \| ( ch >> 12),
		172	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		173	*s++ = 0x80 \| ( ch & 0x3f);
		174	else if ( ch < 0x110000)
		175	*s++ = 0xf0 \| ( ch >> 18),
		176	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		177	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		178	*s++ = 0x80 \| ( ch & 0x3f);
		179
		180	return s;
		181	}
		182
		183	// convert offset pointer to character index, sv must be string
		184	static STRLEN
		185	ptr_to_index (SV sv, char offset)
		186	{
		187	return SvUTF8 (sv)
		188	? utf8_distance (offset, SvPVX (sv))
		189	: offset - SvPVX (sv);
117	}	190	}
118		191
119	/////////////////////////////////////////////////////////////////////////////	192	/////////////////////////////////////////////////////////////////////////////
120	// encoder	193	// encoder
121		194
…		…
125	char *cur; // SvPVX (sv) + current output position	198	char *cur; // SvPVX (sv) + current output position
126	char *end; // SvEND (sv)	199	char *end; // SvEND (sv)
127	SV *sv; // result scalar	200	SV *sv; // result scalar
128	JSON json;	201	JSON json;
129	U32 indent; // indentation level	202	U32 indent; // indentation level
130	U32 maxdepth; // max. indentation/recursion level	203	UV limit; // escape character values >= this value when encoding
131	} enc_t;	204	} enc_t;
132		205
133	inline void	206	INLINE void
134	need (enc_t *enc, STRLEN len)	207	need (enc_t *enc, STRLEN len)
135	{	208	{
136	if (expect_false (enc->cur + len >= enc->end))	209	if (expect_false (enc->cur + len >= enc->end))
137	{	210	{
138	STRLEN cur = enc->cur - SvPVX (enc->sv);	211	STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
139	SvGROW (enc->sv, cur + len + 1);	212	SvGROW (enc->sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
140	enc->cur = SvPVX (enc->sv) + cur;	213	enc->cur = SvPVX (enc->sv) + cur;
141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	214	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
142	}	215	}
143	}	216	}
144		217
145	inline void	218	INLINE void
146	encode_ch (enc_t *enc, char ch)	219	encode_ch (enc_t *enc, char ch)
147	{	220	{
148	need (enc, 1);	221	need (enc, 1);
149	*enc->cur++ = ch;	222	*enc->cur++ = ch;
150	}	223	}
…		…
204	{	277	{
205	uch = ch;	278	uch = ch;
206	clen = 1;	279	clen = 1;
207	}	280	}
208		281
209	if (uch > 0x10FFFFUL)	282	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	{	283	{
214	if (uch > 0xFFFFUL)	284	if (uch >= 0x10000UL)
215	{	285	{
		286	if (uch >= 0x110000UL)
		287	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		288
216	need (enc, len += 11);	289	need (enc, len += 11);
217	sprintf (enc->cur, "\\u%04x\\u%04x",	290	sprintf (enc->cur, "\\u%04x\\u%04x",
218	(int)((uch - 0x10000) / 0x400 + 0xD800),	291	(int)((uch - 0x10000) / 0x400 + 0xD800),
219	(int)((uch - 0x10000) % 0x400 + 0xDC00));	292	(int)((uch - 0x10000) % 0x400 + 0xDC00));
220	enc->cur += 12;	293	enc->cur += 12;
221	}	294	}
222	else	295	else
223	{	296	{
224	static char hexdigit [16] = "0123456789abcdef";
225	need (enc, len += 5);	297	need (enc, len += 5);
226	*enc->cur++ = '\\';	298	*enc->cur++ = '\\';
227	*enc->cur++ = 'u';	299	*enc->cur++ = 'u';
228	*enc->cur++ = hexdigit [ uch >> 12 ];	300	*enc->cur++ = PL_hexdigit [ uch >> 12 ];
229	*enc->cur++ = hexdigit [(uch >> 8) & 15];	301	*enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
230	*enc->cur++ = hexdigit [(uch >> 4) & 15];	302	*enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
231	*enc->cur++ = hexdigit [(uch >> 0) & 15];	303	*enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
232	}	304	}
233		305
234	str += clen;	306	str += clen;
235	}	307	}
236	else if (enc->json.flags & F_LATIN1)	308	else if (enc->json.flags & F_LATIN1)
…		…
248	while (--clen);	320	while (--clen);
249	}	321	}
250	else	322	else
251	{	323	{
252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	324	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	325	enc->cur = encode_utf8 (enc->cur, uch);
254	++str;	326	++str;
255	}	327	}
256	}	328	}
257	}	329	}
258	}	330	}
259		331
260	--len;	332	--len;
261	}	333	}
262	}	334	}
263		335
264	inline void	336	INLINE void
265	encode_indent (enc_t *enc)	337	encode_indent (enc_t *enc)
266	{	338	{
267	if (enc->json.flags & F_INDENT)	339	if (enc->json.flags & F_INDENT)
268	{	340	{
269	int spaces = enc->indent * INDENT_STEP;	341	int spaces = enc->indent * INDENT_STEP;
…		…
272	memset (enc->cur, ' ', spaces);	344	memset (enc->cur, ' ', spaces);
273	enc->cur += spaces;	345	enc->cur += spaces;
274	}	346	}
275	}	347	}
276		348
277	inline void	349	INLINE void
278	encode_space (enc_t *enc)	350	encode_space (enc_t *enc)
279	{	351	{
280	need (enc, 1);	352	need (enc, 1);
281	encode_ch (enc, ' ');	353	encode_ch (enc, ' ');
282	}	354	}
283		355
284	inline void	356	INLINE void
285	encode_nl (enc_t *enc)	357	encode_nl (enc_t *enc)
286	{	358	{
287	if (enc->json.flags & F_INDENT)	359	if (enc->json.flags & F_INDENT)
288	{	360	{
289	need (enc, 1);	361	need (enc, 1);
290	encode_ch (enc, '\n');	362	encode_ch (enc, '\n');
291	}	363	}
292	}	364	}
293		365
294	inline void	366	INLINE void
295	encode_comma (enc_t *enc)	367	encode_comma (enc_t *enc)
296	{	368	{
297	encode_ch (enc, ',');	369	encode_ch (enc, ',');
298		370
299	if (enc->json.flags & F_INDENT)	371	if (enc->json.flags & F_INDENT)
…		…
307	static void	379	static void
308	encode_av (enc_t enc, AV av)	380	encode_av (enc_t enc, AV av)
309	{	381	{
310	int i, len = av_len (av);	382	int i, len = av_len (av);
311		383
312	if (enc->indent >= enc->maxdepth)	384	if (enc->indent >= enc->json.max_depth)
313	croak ("data structure too deep (hit recursion limit)");	385	croak (ERR_NESTING_EXCEEDED);
314		386
315	encode_ch (enc, '['); encode_nl (enc);	387	encode_ch (enc, '[');
316	++enc->indent;	388
		389	if (len >= 0)
		390	{
		391	encode_nl (enc); ++enc->indent;
317		392
318	for (i = 0; i <= len; ++i)	393	for (i = 0; i <= len; ++i)
319	{	394	{
320	SV **svp = av_fetch (av, i, 0);	395	SV **svp = av_fetch (av, i, 0);
321		396
322	encode_indent (enc);	397	encode_indent (enc);
323		398
324	if (svp)	399	if (svp)
325	encode_sv (enc, *svp);	400	encode_sv (enc, *svp);
326	else	401	else
327	encode_str (enc, "null", 4, 0);	402	encode_str (enc, "null", 4, 0);
328		403
329	if (i < len)	404	if (i < len)
330	encode_comma (enc);	405	encode_comma (enc);
331	}	406	}
332		407
		408	encode_nl (enc); --enc->indent; encode_indent (enc);
		409	}
		410
333	encode_nl (enc);	411	encode_ch (enc, ']');
334
335	--enc->indent;
336	encode_indent (enc); encode_ch (enc, ']');
337	}	412	}
338		413
339	static void	414	static void
340	encode_he (enc_t enc, HE he)	415	encode_hk (enc_t enc, HE he)
341	{	416	{
342	encode_ch (enc, '"');	417	encode_ch (enc, '"');
343		418
344	if (HeKLEN (he) == HEf_SVKEY)	419	if (HeKLEN (he) == HEf_SVKEY)
345	{	420	{
…		…
358	encode_ch (enc, '"');	433	encode_ch (enc, '"');
359		434
360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	435	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
361	encode_ch (enc, ':');	436	encode_ch (enc, ':');
362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	437	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
363	encode_sv (enc, HeVAL (he));
364	}	438	}
365		439
366	// compare hash entries, used when all keys are bytestrings	440	// compare hash entries, used when all keys are bytestrings
367	static int	441	static int
368	he_cmp_fast (const void a_, const void b_)	442	he_cmp_fast (const void a_, const void b_)
…		…
373	HE b = (HE **)b_;	447	HE b = (HE **)b_;
374		448
375	STRLEN la = HeKLEN (a);	449	STRLEN la = HeKLEN (a);
376	STRLEN lb = HeKLEN (b);	450	STRLEN lb = HeKLEN (b);
377		451
378	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	452	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
379	cmp = la - lb;	453	cmp = lb - la;
380		454
381	return cmp;	455	return cmp;
382	}	456	}
383		457
384	// compare hash entries, used when some keys are sv's or utf-x	458	// compare hash entries, used when some keys are sv's or utf-x
385	static int	459	static int
386	he_cmp_slow (const void a, const void b)	460	he_cmp_slow (const void a, const void b)
387	{	461	{
388	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	462	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
389	}	463	}
390		464
391	static void	465	static void
392	encode_hv (enc_t enc, HV hv)	466	encode_hv (enc_t enc, HV hv)
393	{	467	{
394	int count, i;	468	HE *he;
395		469
396	if (enc->indent >= enc->maxdepth)	470	if (enc->indent >= enc->json.max_depth)
397	croak ("data structure too deep (hit recursion limit)");	471	croak (ERR_NESTING_EXCEEDED);
398		472
399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	473	encode_ch (enc, '{');
400		474
401	if ((count = hv_iterinit (hv)))
402	{
403	// for canonical output we have to sort by keys first	475	// for canonical output we have to sort by keys first
404	// actually, this is mostly due to the stupid so-called	476	// actually, this is mostly due to the stupid so-called
405	// security workaround added somewhere in 5.8.x.	477	// security workaround added somewhere in 5.8.x
406	// that randomises hash orderings	478	// that randomises hash orderings
407	if (enc->json.flags & F_CANONICAL)	479	if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
		480	{
		481	int count = hv_iterinit (hv);
		482
		483	if (SvMAGICAL (hv))
408	{	484	{
		485	// need to count by iterating. could improve by dynamically building the vector below
		486	// but I don't care for the speed of this special case.
		487	// note also that we will run into undefined behaviour when the two iterations
		488	// do not result in the same count, something I might care for in some later release.
		489
		490	count = 0;
		491	while (hv_iternext (hv))
		492	++count;
		493
		494	hv_iterinit (hv);
		495	}
		496
		497	if (count)
		498	{
409	int fast = 1;	499	int i, fast = 1;
410	HE *he;
411	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	500	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
412	HE *hes = _alloca (count sizeof (HE));	501	HE *hes = _alloca (count sizeof (HE));
413	#else	502	#else
414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	503	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
415	#endif	504	#endif
…		…
442		531
443	FREETMPS;	532	FREETMPS;
444	LEAVE;	533	LEAVE;
445	}	534	}
446		535
447	for (i = 0; i < count; ++i)	536	encode_nl (enc); ++enc->indent;
		537
		538	while (count--)
448	{	539	{
449	encode_indent (enc);	540	encode_indent (enc);
		541	he = hes [count];
450	encode_he (enc, hes [i]);	542	encode_hk (enc, he);
		543	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		544
452	if (i < count - 1)	545	if (count)
453	encode_comma (enc);	546	encode_comma (enc);
454	}	547	}
455		548
456	encode_nl (enc);	549	encode_nl (enc); --enc->indent; encode_indent (enc);
457	}	550	}
		551	}
458	else	552	else
		553	{
		554	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		555	if ((he = hv_iternext (hv)))
459	{	556	{
460	HE *he = hv_iternext (hv);	557	encode_nl (enc); ++enc->indent;
461		558
462	for (;;)	559	for (;;)
463	{	560	{
464	encode_indent (enc);	561	encode_indent (enc);
465	encode_he (enc, he);	562	encode_hk (enc, he);
		563	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
466		564
467	if (!(he = hv_iternext (hv)))	565	if (!(he = hv_iternext (hv)))
468	break;	566	break;
469		567
470	encode_comma (enc);	568	encode_comma (enc);
471	}	569	}
472		570
473	encode_nl (enc);	571	encode_nl (enc); --enc->indent; encode_indent (enc);
474	}	572	}
475	}	573	}
476		574
477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	575	encode_ch (enc, '}');
478	}	576	}
479		577
480	// encode objects, arrays and special \0=false and \1=true values.	578	// encode objects, arrays and special \0=false and \1=true values.
481	static void	579	static void
482	encode_rv (enc_t enc, SV sv)	580	encode_rv (enc_t enc, SV sv)
…		…
559		657
560	if (len == 1 && *pv == '1')	658	if (len == 1 && *pv == '1')
561	encode_str (enc, "true", 4, 0);	659	encode_str (enc, "true", 4, 0);
562	else if (len == 1 && *pv == '0')	660	else if (len == 1 && *pv == '0')
563	encode_str (enc, "false", 5, 0);	661	encode_str (enc, "false", 5, 0);
		662	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		663	encode_str (enc, "null", 4, 0);
564	else	664	else
565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	665	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	666	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
567	}	667	}
		668	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		669	encode_str (enc, "null", 4, 0);
568	else	670	else
569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	671	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	672	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
571	}	673	}
572		674
…		…
590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	692	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
591	enc->cur += strlen (enc->cur);	693	enc->cur += strlen (enc->cur);
592	}	694	}
593	else if (SvIOKp (sv))	695	else if (SvIOKp (sv))
594	{	696	{
595	// we assume we can always read an IV as a UV	697	// we assume we can always read an IV as a UV and vice versa
596	if (SvUV (sv) & ~(UV)0x7fff)	698	// we assume two's complement
597	{	699	// we assume no aliasing issues in the union
598	// large integer, use the (rather slow) snprintf way.	700	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
599	need (enc, sizeof (UV) * 3);	701	: 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	{	702	{
607	// optimise the "small number case"	703	// optimise the "small number case"
608	// code will likely be branchless and use only a single multiplication	704	// code will likely be branchless and use only a single multiplication
		705	// works for numbers up to 59074
609	I32 i = SvIV (sv);	706	I32 i = SvIVX (sv);
610	U32 u;	707	U32 u;
611	char digit, nz = 0;	708	char digit, nz = 0;
612		709
613	need (enc, 6);	710	need (enc, 6);
614		711
…		…
620		717
621	// now output digit by digit, each time masking out the integer part	718	// 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,	719	// and multiplying by 5 while moving the decimal point one to the right,
623	// resulting in a net multiplication by 10.	720	// resulting in a net multiplication by 10.
624	// we always write the digit to memory but conditionally increment	721	// we always write the digit to memory but conditionally increment
625	// the pointer, to ease the usage of conditional move instructions.	722	// 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;	723	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;	724	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;	725	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;	726	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'	727	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
631	}	728	}
		729	else
		730	{
		731	// large integer, use the (rather slow) snprintf way.
		732	need (enc, IVUV_MAXCHARS);
		733	enc->cur +=
		734	SvIsUV(sv)
		735	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		736	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		737	}
632	}	738	}
633	else if (SvROK (sv))	739	else if (SvROK (sv))
634	encode_rv (enc, SvRV (sv));	740	encode_rv (enc, SvRV (sv));
635	else if (!SvOK (sv))	741	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
636	encode_str (enc, "null", 4, 0);	742	encode_str (enc, "null", 4, 0);
637	else	743	else
638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	744	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
639	SvPV_nolen (sv), SvFLAGS (sv));	745	SvPV_nolen (sv), SvFLAGS (sv));
640	}	746	}
…		…
650	enc.json = *json;	756	enc.json = *json;
651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	757	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
652	enc.cur = SvPVX (enc.sv);	758	enc.cur = SvPVX (enc.sv);
653	enc.end = SvEND (enc.sv);	759	enc.end = SvEND (enc.sv);
654	enc.indent = 0;	760	enc.indent = 0;
655	enc.maxdepth = DEC_DEPTH (enc.json.flags);	761	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		762	: enc.json.flags & F_LATIN1 ? 0x000100UL
		763	: 0x110000UL;
656		764
657	SvPOK_only (enc.sv);	765	SvPOK_only (enc.sv);
658	encode_sv (&enc, scalar);	766	encode_sv (&enc, scalar);
		767	encode_nl (&enc);
659		768
660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	769	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
661	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	770	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
662		771
663	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	772	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
…		…
681	JSON json;	790	JSON json;
682	U32 depth; // recursion depth	791	U32 depth; // recursion depth
683	U32 maxdepth; // recursion depth limit	792	U32 maxdepth; // recursion depth limit
684	} dec_t;	793	} dec_t;
685		794
686	inline void	795	INLINE void
		796	decode_comment (dec_t *dec)
		797	{
		798	// only '#'-style comments allowed a.t.m.
		799
		800	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		801	++dec->cur;
		802	}
		803
		804	INLINE void
687	decode_ws (dec_t *dec)	805	decode_ws (dec_t *dec)
688	{	806	{
689	for (;;)	807	for (;;)
690	{	808	{
691	char ch = *dec->cur;	809	char ch = *dec->cur;
692		810
693	if (ch > 0x20	811	if (ch > 0x20)
		812	{
		813	if (expect_false (ch == '#'))
		814	{
		815	if (dec->json.flags & F_RELAXED)
		816	decode_comment (dec);
		817	else
		818	break;
		819	}
		820	else
		821	break;
		822	}
694	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	823	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
695	break;	824	break; // parse error, but let higher level handle it, gives better error messages
696		825
697	++dec->cur;	826	++dec->cur;
698	}	827	}
699	}	828	}
700		829
…		…
704	if (*dec->cur != ch) \	833	if (*dec->cur != ch) \
705	ERR (# ch " expected"); \	834	ERR (# ch " expected"); \
706	++dec->cur; \	835	++dec->cur; \
707	SE	836	SE
708		837
709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	838	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
710	#define DEC_DEC_DEPTH --dec->depth	839	#define DEC_DEC_DEPTH --dec->depth
711		840
712	static SV decode_sv (dec_t dec);	841	static SV decode_sv (dec_t dec);
713		842
714	static signed char decode_hexdigit[256];	843	static signed char decode_hexdigit[256];
…		…
806		935
807	if (hi >= 0x80)	936	if (hi >= 0x80)
808	{	937	{
809	utf8 = 1;	938	utf8 = 1;
810		939
811	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	940	cur = encode_utf8 (cur, hi);
812	}	941	}
813	else	942	else
814	*cur++ = hi;	943	*cur++ = hi;
815	}	944	}
816	break;	945	break;
…		…
818	default:	947	default:
819	--dec_cur;	948	--dec_cur;
820	ERR ("illegal backslash escape sequence in string");	949	ERR ("illegal backslash escape sequence in string");
821	}	950	}
822	}	951	}
823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	952	else if (expect_true (ch >= 0x20 && ch < 0x80))
824	*cur++ = ch;	953	*cur++ = ch;
825	else if (ch >= 0x80)	954	else if (ch >= 0x80)
826	{	955	{
827	STRLEN clen;	956	STRLEN clen;
828	UV uch;
829		957
830	--dec_cur;	958	--dec_cur;
831		959
832	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);	960	decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
833	if (clen == (STRLEN)-1)	961	if (clen == (STRLEN)-1)
834	ERR ("malformed UTF-8 character in JSON string");	962	ERR ("malformed UTF-8 character in JSON string");
835		963
836	do	964	do
837	cur++ = dec_cur++;	965	cur++ = dec_cur++;
…		…
854	{	982	{
855	STRLEN len = cur - buf;	983	STRLEN len = cur - buf;
856		984
857	if (sv)	985	if (sv)
858	{	986	{
859	SvGROW (sv, SvCUR (sv) + len + 1);	987	STRLEN cur = SvCUR (sv);
		988
		989	if (SvLEN (sv) <= cur + len)
		990	SvGROW (sv, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		991
860	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	992	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
861	SvCUR_set (sv, SvCUR (sv) + len);	993	SvCUR_set (sv, SvCUR (sv) + len);
862	}	994	}
863	else	995	else
864	sv = newSVpvn (buf, len);	996	sv = newSVpvn (buf, len);
…		…
951		1083
952	if (!is_nv)	1084	if (!is_nv)
953	{	1085	{
954	int len = dec->cur - start;	1086	int len = dec->cur - start;
955		1087
956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1088	// special case the rather common 1..5-digit-int case
957	if (*start == '-')	1089	if (*start == '-')
958	switch (len)	1090	switch (len)
959	{	1091	{
960	case 2: return newSViv (-( start [1] - '0' * 1));	1092	case 2: return newSViv (-(IV)( start [1] - '0' * 1));
961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1093	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));	1094	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));	1095	case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1096	case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
964	}	1097	}
965	else	1098	else
966	switch (len)	1099	switch (len)
967	{	1100	{
968	case 1: return newSViv ( start [0] - '0' * 1);	1101	case 1: return newSViv ( start [0] - '0' * 1);
969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1102	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);	1103	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);	1104	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1105	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
972	}	1106	}
973		1107
974	{	1108	{
975	UV uv;	1109	UV uv;
976	int numtype = grok_number (start, len, &uv);	1110	int numtype = grok_number (start, len, &uv);
…		…
1037		1171
1038	if (*dec->cur != ',')	1172	if (*dec->cur != ',')
1039	ERR (", or ] expected while parsing array");	1173	ERR (", or ] expected while parsing array");
1040		1174
1041	++dec->cur;	1175	++dec->cur;
		1176
		1177	decode_ws (dec);
		1178
		1179	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1180	{
		1181	++dec->cur;
		1182	break;
		1183	}
1042	}	1184	}
1043		1185
1044	DEC_DEC_DEPTH;	1186	DEC_DEC_DEPTH;
1045	return newRV_noinc ((SV *)av);	1187	return newRV_noinc ((SV *)av);
1046		1188
…		…
1062	if (*dec->cur == '}')	1204	if (*dec->cur == '}')
1063	++dec->cur;	1205	++dec->cur;
1064	else	1206	else
1065	for (;;)	1207	for (;;)
1066	{	1208	{
1067	decode_ws (dec); EXPECT_CH ('"');	1209	EXPECT_CH ('"');
1068		1210
1069	// heuristic: assume that	1211	// heuristic: assume that
1070	// a) decode_str + hv_store_ent are abysmally slow.	1212	// a) decode_str + hv_store_ent are abysmally slow.
1071	// b) most hash keys are short, simple ascii text.	1213	// b) most hash keys are short, simple ascii text.
1072	// => try to "fast-match" such strings to avoid	1214	// => try to "fast-match" such strings to avoid
…		…
1076	char *p = dec->cur;	1218	char *p = dec->cur;
1077	char *e = p + 24; // only try up to 24 bytes	1219	char *e = p + 24; // only try up to 24 bytes
1078		1220
1079	for (;;)	1221	for (;;)
1080	{	1222	{
1081	// the >= 0x80 is true on most architectures	1223	// the >= 0x80 is false on most architectures
1082	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1224	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1083	{	1225	{
1084	// slow path, back up and use decode_str	1226	// slow path, back up and use decode_str
1085	SV *key = decode_str (dec);	1227	SV *key = decode_str (dec);
1086	if (!key)	1228	if (!key)
1087	goto fail;	1229	goto fail;
1088		1230
1089	decode_ws (dec); EXPECT_CH (':');	1231	decode_ws (dec); EXPECT_CH (':');
1090		1232
		1233	decode_ws (dec);
1091	value = decode_sv (dec);	1234	value = decode_sv (dec);
1092	if (!value)	1235	if (!value)
1093	{	1236	{
1094	SvREFCNT_dec (key);	1237	SvREFCNT_dec (key);
1095	goto fail;	1238	goto fail;
…		…
1107	int len = p - key;	1250	int len = p - key;
1108	dec->cur = p + 1;	1251	dec->cur = p + 1;
1109		1252
1110	decode_ws (dec); EXPECT_CH (':');	1253	decode_ws (dec); EXPECT_CH (':');
1111		1254
		1255	decode_ws (dec);
1112	value = decode_sv (dec);	1256	value = decode_sv (dec);
1113	if (!value)	1257	if (!value)
1114	goto fail;	1258	goto fail;
1115		1259
1116	hv_store (hv, key, len, value, 0);	1260	hv_store (hv, key, len, value, 0);
…		…
1132		1276
1133	if (*dec->cur != ',')	1277	if (*dec->cur != ',')
1134	ERR (", or } expected while parsing object/hash");	1278	ERR (", or } expected while parsing object/hash");
1135		1279
1136	++dec->cur;	1280	++dec->cur;
		1281
		1282	decode_ws (dec);
		1283
		1284	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1285	{
		1286	++dec->cur;
		1287	break;
		1288	}
1137	}	1289	}
1138		1290
1139	DEC_DEC_DEPTH;	1291	DEC_DEC_DEPTH;
1140	sv = newRV_noinc ((SV *)hv);	1292	sv = newRV_noinc ((SV *)hv);
1141		1293
…		…
1206	}	1358	}
1207		1359
1208	static SV *	1360	static SV *
1209	decode_sv (dec_t *dec)	1361	decode_sv (dec_t *dec)
1210	{	1362	{
1211	decode_ws (dec);
1212
1213	// the beauty of JSON: you need exactly one character lookahead	1363	// the beauty of JSON: you need exactly one character lookahead
1214	// to parse anything.	1364	// to parse everything.
1215	switch (*dec->cur)	1365	switch (*dec->cur)
1216	{	1366	{
1217	case '"': ++dec->cur; return decode_str (dec);	1367	case '"': ++dec->cur; return decode_str (dec);
1218	case '[': ++dec->cur; return decode_av (dec);	1368	case '[': ++dec->cur; return decode_av (dec);
1219	case '{': ++dec->cur; return decode_hv (dec);	1369	case '{': ++dec->cur; return decode_hv (dec);
1220		1370
1221	case '-':	1371	case '-':
1222	case '0': case '1': case '2': case '3': case '4':	1372	case '0': case '1': case '2': case '3': case '4':
1223	case '5': case '6': case '7': case '8': case '9':	1373	case '5': case '6': case '7': case '8': case '9':
1224	return decode_num (dec);	1374	return decode_num (dec);
…		…
1226	case 't':	1376	case 't':
1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1377	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1228	{	1378	{
1229	dec->cur += 4;	1379	dec->cur += 4;
1230	#if JSON_SLOW	1380	#if JSON_SLOW
1231	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);	1381	json_true = get_bool ("JSON::XS::true");
1232	#endif	1382	#endif
1233	return SvREFCNT_inc (json_true);	1383	return newSVsv (json_true);
1234	}	1384	}
1235	else	1385	else
1236	ERR ("'true' expected");	1386	ERR ("'true' expected");
1237		1387
1238	break;	1388	break;
…		…
1240	case 'f':	1390	case 'f':
1241	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1391	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1242	{	1392	{
1243	dec->cur += 5;	1393	dec->cur += 5;
1244	#if JSON_SLOW	1394	#if JSON_SLOW
1245	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1395	json_false = get_bool ("JSON::XS::false");
1246	#endif	1396	#endif
1247	return SvREFCNT_inc (json_false);	1397	return newSVsv (json_false);
1248	}	1398	}
1249	else	1399	else
1250	ERR ("'false' expected");	1400	ERR ("'false' expected");
1251		1401
1252	break;	1402	break;
…		…
1270	fail:	1420	fail:
1271	return 0;	1421	return 0;
1272	}	1422	}
1273		1423
1274	static SV *	1424	static SV *
1275	decode_json (SV string, JSON json, UV *offset_return)	1425	decode_json (SV string, JSON json, char **offset_return)
1276	{	1426	{
1277	dec_t dec;	1427	dec_t dec;
1278	UV offset;
1279	SV *sv;	1428	SV *sv;
1280		1429
		1430	/* work around bugs in 5.10 where manipulating magic values
		1431	* will perl ignore the magic in subsequent accesses
		1432	*/
1281	SvGETMAGIC (string);	1433	/SvGETMAGIC (string);/
		1434	if (SvMAGICAL (string))
		1435	string = sv_2mortal (newSVsv (string));
		1436
1282	SvUPGRADE (string, SVt_PV);	1437	SvUPGRADE (string, SVt_PV);
1283		1438
1284	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))	1439	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1440	* assertion with -DDEBUGGING, although SvCUR is documented to
		1441	* return the xpv_cur field which certainly exists after upgrading.
		1442	* according to nicholas clark, calling SvPOK fixes this.
		1443	* But it doesn't fix it, so try another workaround, call SvPV_nolen
		1444	* and hope for the best.
		1445	* Damnit, SvPV_nolen still trips over yet another assertion. This
		1446	* assertion business is seriously broken, try yet another workaround
		1447	* for the broken -DDEBUGGING.
		1448	*/
		1449	{
		1450	#ifdef DEBUGGING
		1451	STRLEN offset = SvOK (string) ? sv_len (string) : 0;
		1452	#else
		1453	STRLEN offset = SvCUR (string);
		1454	#endif
		1455
		1456	if (offset > json->max_size && json->max_size)
1285	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",	1457	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1286	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));	1458	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1459	}
1287		1460
1288	if (json->flags & F_UTF8)	1461	if (json->flags & F_UTF8)
1289	sv_utf8_downgrade (string, 0);	1462	sv_utf8_downgrade (string, 0);
1290	else	1463	else
1291	sv_utf8_upgrade (string);	1464	sv_utf8_upgrade (string);
1292		1465
1293	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1466	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1294		1467
1295	dec.json = *json;	1468	dec.json = *json;
1296	dec.cur = SvPVX (string);	1469	dec.cur = SvPVX (string);
1297	dec.end = SvEND (string);	1470	dec.end = SvEND (string);
1298	dec.err = 0;	1471	dec.err = 0;
1299	dec.depth = 0;	1472	dec.depth = 0;
1300	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1301		1473
1302	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1474	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1303	dec.json.flags \|= F_HOOK;	1475	dec.json.flags \|= F_HOOK;
1304		1476
1305	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1477	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1478
		1479	decode_ws (&dec);
1306	sv = decode_sv (&dec);	1480	sv = decode_sv (&dec);
		1481
		1482	if (offset_return)
		1483	*offset_return = dec.cur;
1307		1484
1308	if (!(offset_return \|\| !sv))	1485	if (!(offset_return \|\| !sv))
1309	{	1486	{
1310	// check for trailing garbage	1487	// check for trailing garbage
1311	decode_ws (&dec);	1488	decode_ws (&dec);
…		…
1314	{	1491	{
1315	dec.err = "garbage after JSON object";	1492	dec.err = "garbage after JSON object";
1316	SvREFCNT_dec (sv);	1493	SvREFCNT_dec (sv);
1317	sv = 0;	1494	sv = 0;
1318	}	1495	}
1319	}
1320
1321	if (offset_return \|\| !sv)
1322	{
1323	offset = dec.json.flags & F_UTF8
1324	? dec.cur - SvPVX (string)
1325	: utf8_distance (dec.cur, SvPVX (string));
1326
1327	if (offset_return)
1328	*offset_return = offset;
1329	}	1496	}
1330		1497
1331	if (!sv)	1498	if (!sv)
1332	{	1499	{
1333	SV *uni = sv_newmortal ();	1500	SV *uni = sv_newmortal ();
…		…
1339	SAVEVPTR (PL_curcop);	1506	SAVEVPTR (PL_curcop);
1340	PL_curcop = &cop;	1507	PL_curcop = &cop;
1341	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1508	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1342	LEAVE;	1509	LEAVE;
1343		1510
1344	croak ("%s, at character offset %d [\"%s\"]",	1511	croak ("%s, at character offset %d (before \"%s\")",
1345	dec.err,	1512	dec.err,
1346	(int)offset,	1513	ptr_to_index (string, dec.cur),
1347	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1514	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1348	}	1515	}
1349		1516
1350	sv = sv_2mortal (sv);	1517	sv = sv_2mortal (sv);
1351		1518
1352	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))	1519	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1353	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1520	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1354		1521
1355	return sv;	1522	return sv;
		1523	}
		1524
		1525	/////////////////////////////////////////////////////////////////////////////
		1526	// incremental parser
		1527
		1528	static void
		1529	incr_parse (JSON *self)
		1530	{
		1531	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1532
		1533	// the state machine here is a bit convoluted and could be simplified a lot
		1534	// but this would make it slower, so...
		1535
		1536	for (;;)
		1537	{
		1538	//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
		1539	switch (self->incr_mode)
		1540	{
		1541	// only used for initial whitespace skipping
		1542	case INCR_M_WS:
		1543	for (;;)
		1544	{
		1545	if (*p > 0x20)
		1546	{
		1547	if (*p == '#')
		1548	{
		1549	self->incr_mode = INCR_M_C0;
		1550	goto incr_m_c;
		1551	}
		1552	else
		1553	{
		1554	self->incr_mode = INCR_M_JSON;
		1555	goto incr_m_json;
		1556	}
		1557	}
		1558	else if (!*p)
		1559	goto interrupt;
		1560
		1561	++p;
		1562	}
		1563
		1564	// skip a single char inside a string (for \\-processing)
		1565	case INCR_M_BS:
		1566	if (!*p)
		1567	goto interrupt;
		1568
		1569	++p;
		1570	self->incr_mode = INCR_M_STR;
		1571	goto incr_m_str;
		1572
		1573	// inside #-style comments
		1574	case INCR_M_C0:
		1575	case INCR_M_C1:
		1576	incr_m_c:
		1577	for (;;)
		1578	{
		1579	if (*p == '\n')
		1580	{
		1581	self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
		1582	break;
		1583	}
		1584	else if (!*p)
		1585	goto interrupt;
		1586
		1587	++p;
		1588	}
		1589
		1590	break;
		1591
		1592	// inside a string
		1593	case INCR_M_STR:
		1594	incr_m_str:
		1595	for (;;)
		1596	{
		1597	if (*p == '"')
		1598	{
		1599	++p;
		1600	self->incr_mode = INCR_M_JSON;
		1601
		1602	if (!self->incr_nest)
		1603	goto interrupt;
		1604
		1605	goto incr_m_json;
		1606	}
		1607	else if (*p == '\\')
		1608	{
		1609	++p; // "virtually" consumes character after \
		1610
		1611	if (!*p) // if at end of string we have to switch modes
		1612	{
		1613	self->incr_mode = INCR_M_BS;
		1614	goto interrupt;
		1615	}
		1616	}
		1617	else if (!*p)
		1618	goto interrupt;
		1619
		1620	++p;
		1621	}
		1622
		1623	// after initial ws, outside string
		1624	case INCR_M_JSON:
		1625	incr_m_json:
		1626	for (;;)
		1627	{
		1628	switch (*p++)
		1629	{
		1630	case 0:
		1631	--p;
		1632	goto interrupt;
		1633
		1634	case 0x09:
		1635	case 0x0a:
		1636	case 0x0d:
		1637	case 0x20:
		1638	if (!self->incr_nest)
		1639	{
		1640	--p; // do not eat the whitespace, let the next round do it
		1641	goto interrupt;
		1642	}
		1643	break;
		1644
		1645	case '"':
		1646	self->incr_mode = INCR_M_STR;
		1647	goto incr_m_str;
		1648
		1649	case '[':
		1650	case '{':
		1651	if (++self->incr_nest > self->max_depth)
		1652	croak (ERR_NESTING_EXCEEDED);
		1653	break;
		1654
		1655	case ']':
		1656	case '}':
		1657	if (--self->incr_nest <= 0)
		1658	goto interrupt;
		1659	break;
		1660
		1661	case '#':
		1662	self->incr_mode = INCR_M_C1;
		1663	goto incr_m_c;
		1664	}
		1665	}
		1666	}
		1667
		1668	modechange:
		1669	;
		1670	}
		1671
		1672	interrupt:
		1673	self->incr_pos = p - SvPVX (self->incr_text);
		1674	//printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
		1675	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1356	}	1676	}
1357		1677
1358	/////////////////////////////////////////////////////////////////////////////	1678	/////////////////////////////////////////////////////////////////////////////
1359	// XS interface functions	1679	// XS interface functions
1360		1680
…		…
1372	: -1;	1692	: -1;
1373		1693
1374	json_stash = gv_stashpv ("JSON::XS" , 1);	1694	json_stash = gv_stashpv ("JSON::XS" , 1);
1375	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1695	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1376		1696
1377	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1697	json_true = get_bool ("JSON::XS::true");
1378	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1698	json_false = get_bool ("JSON::XS::false");
		1699
		1700	CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1379	}	1701	}
1380		1702
1381	PROTOTYPES: DISABLE	1703	PROTOTYPES: DISABLE
1382		1704
1383	void CLONE (...)	1705	void CLONE (...)
…		…
1388	void new (char *klass)	1710	void new (char *klass)
1389	PPCODE:	1711	PPCODE:
1390	{	1712	{
1391	SV *pv = NEWSV (0, sizeof (JSON));	1713	SV *pv = NEWSV (0, sizeof (JSON));
1392	SvPOK_only (pv);	1714	SvPOK_only (pv);
1393	Zero (SvPVX (pv), 1, JSON);	1715	json_init ((JSON *)SvPVX (pv));
1394	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1716	XPUSHs (sv_2mortal (sv_bless (
1395	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));	1717	newRV_noinc (pv),
		1718	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1719	)));
1396	}	1720	}
1397		1721
1398	void ascii (JSON *self, int enable = 1)	1722	void ascii (JSON *self, int enable = 1)
1399	ALIAS:	1723	ALIAS:
1400	ascii = F_ASCII	1724	ascii = F_ASCII
…		…
1407	pretty = F_PRETTY	1731	pretty = F_PRETTY
1408	allow_nonref = F_ALLOW_NONREF	1732	allow_nonref = F_ALLOW_NONREF
1409	shrink = F_SHRINK	1733	shrink = F_SHRINK
1410	allow_blessed = F_ALLOW_BLESSED	1734	allow_blessed = F_ALLOW_BLESSED
1411	convert_blessed = F_CONV_BLESSED	1735	convert_blessed = F_CONV_BLESSED
		1736	relaxed = F_RELAXED
		1737	allow_unknown = F_ALLOW_UNKNOWN
1412	PPCODE:	1738	PPCODE:
1413	{	1739	{
1414	if (enable)	1740	if (enable)
1415	self->flags \|= ix;	1741	self->flags \|= ix;
1416	else	1742	else
1417	self->flags &= ~ix;	1743	self->flags &= ~ix;
1418		1744
1419	XPUSHs (ST (0));	1745	XPUSHs (ST (0));
1420	}	1746	}
1421		1747
1422	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1748	void get_ascii (JSON *self)
		1749	ALIAS:
		1750	get_ascii = F_ASCII
		1751	get_latin1 = F_LATIN1
		1752	get_utf8 = F_UTF8
		1753	get_indent = F_INDENT
		1754	get_canonical = F_CANONICAL
		1755	get_space_before = F_SPACE_BEFORE
		1756	get_space_after = F_SPACE_AFTER
		1757	get_allow_nonref = F_ALLOW_NONREF
		1758	get_shrink = F_SHRINK
		1759	get_allow_blessed = F_ALLOW_BLESSED
		1760	get_convert_blessed = F_CONV_BLESSED
		1761	get_relaxed = F_RELAXED
		1762	get_allow_unknown = F_ALLOW_UNKNOWN
1423	PPCODE:	1763	PPCODE:
1424	{	1764	XPUSHs (boolSV (self->flags & ix));
1425	UV log2 = 0;
1426		1765
1427	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1766	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1428		1767	PPCODE:
1429	while ((1UL << log2) < max_depth)	1768	self->max_depth = max_depth;
1430	++log2;
1431
1432	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1433
1434	XPUSHs (ST (0));	1769	XPUSHs (ST (0));
1435	}
1436		1770
		1771	U32 get_max_depth (JSON *self)
		1772	CODE:
		1773	RETVAL = self->max_depth;
		1774	OUTPUT:
		1775	RETVAL
		1776
1437	void max_size (JSON *self, UV max_size = 0)	1777	void max_size (JSON *self, U32 max_size = 0)
1438	PPCODE:	1778	PPCODE:
1439	{	1779	self->max_size = max_size;
1440	UV log2 = 0;
1441
1442	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1443	if (max_size == 1) max_size = 2;
1444
1445	while ((1UL << log2) < max_size)
1446	++log2;
1447
1448	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1449
1450	XPUSHs (ST (0));	1780	XPUSHs (ST (0));
1451	}	1781
		1782	int get_max_size (JSON *self)
		1783	CODE:
		1784	RETVAL = self->max_size;
		1785	OUTPUT:
		1786	RETVAL
1452		1787
1453	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1788	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1454	PPCODE:	1789	PPCODE:
1455	{	1790	{
1456	SvREFCNT_dec (self->cb_object);	1791	SvREFCNT_dec (self->cb_object);
…		…
1490	XPUSHs (decode_json (jsonstr, self, 0));	1825	XPUSHs (decode_json (jsonstr, self, 0));
1491		1826
1492	void decode_prefix (JSON self, SV jsonstr)	1827	void decode_prefix (JSON self, SV jsonstr)
1493	PPCODE:	1828	PPCODE:
1494	{	1829	{
1495	UV offset;	1830	char *offset;
1496	EXTEND (SP, 2);	1831	EXTEND (SP, 2);
1497	PUSHs (decode_json (jsonstr, self, &offset));	1832	PUSHs (decode_json (jsonstr, self, &offset));
1498	PUSHs (sv_2mortal (newSVuv (offset)));	1833	PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, offset))));
		1834	}
		1835
		1836	void incr_parse (JSON self, SV jsonstr = 0)
		1837	PPCODE:
		1838	{
		1839	if (!self->incr_text)
		1840	self->incr_text = newSVpvn ("", 0);
		1841
		1842	// append data, if any
		1843	if (jsonstr)
		1844	{
		1845	if (SvUTF8 (jsonstr))
		1846	{
		1847	if (!SvUTF8 (self->incr_text))
		1848	{
		1849	/* utf-8-ness differs, need to upgrade */
		1850	sv_utf8_upgrade (self->incr_text);
		1851
		1852	if (self->incr_pos)
		1853	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1854	- (U8 *)SvPVX (self->incr_text);
		1855	}
		1856	}
		1857	else if (SvUTF8 (self->incr_text))
		1858	sv_utf8_upgrade (jsonstr);
		1859
		1860	{
		1861	STRLEN len;
		1862	const char *str = SvPV (jsonstr, len);
		1863	STRLEN cur = SvCUR (self->incr_text);
		1864
		1865	if (SvLEN (self->incr_text) <= cur + len)
		1866	SvGROW (self->incr_text, cur + (len < (cur >> 2) ? cur >> 2 : len) + 1);
		1867
		1868	Move (str, SvEND (self->incr_text), len, char);
		1869	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1870	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1871	}
		1872	}
		1873
		1874	if (GIMME_V != G_VOID)
		1875	do
		1876	{
		1877	char *offset;
		1878
		1879	if (!INCR_DONE (self))
		1880	{
		1881	incr_parse (self);
		1882
		1883	if (self->incr_pos > self->max_size && self->max_size)
		1884	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1885	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1886
		1887	if (!INCR_DONE (self))
		1888	break;
		1889	}
		1890
		1891	XPUSHs (decode_json (self->incr_text, self, &offset));
		1892
		1893	self->incr_pos -= offset - SvPVX (self->incr_text);
		1894	self->incr_nest = 0;
		1895	self->incr_mode = 0;
		1896
		1897	sv_chop (self->incr_text, offset);
		1898	}
		1899	while (GIMME_V == G_ARRAY);
		1900	}
		1901
		1902	SV incr_text (JSON self)
		1903	ATTRS: lvalue
		1904	CODE:
		1905	{
		1906	if (self->incr_pos)
		1907	croak ("incr_text can not be called when the incremental parser already started parsing");
		1908
		1909	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1910	}
		1911	OUTPUT:
		1912	RETVAL
		1913
		1914	void incr_skip (JSON *self)
		1915	CODE:
		1916	{
		1917	if (self->incr_pos)
		1918	{
		1919	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1920	self->incr_pos = 0;
		1921	self->incr_nest = 0;
		1922	self->incr_mode = 0;
		1923	}
		1924	}
		1925
		1926	void incr_reset (JSON *self)
		1927	CODE:
		1928	{
		1929	SvREFCNT_dec (self->incr_text);
		1930	self->incr_text = 0;
		1931	self->incr_pos = 0;
		1932	self->incr_nest = 0;
		1933	self->incr_mode = 0;
1499	}	1934	}
1500		1935
1501	void DESTROY (JSON *self)	1936	void DESTROY (JSON *self)
1502	CODE:	1937	CODE:
1503	SvREFCNT_dec (self->cb_sk_object);	1938	SvREFCNT_dec (self->cb_sk_object);
1504	SvREFCNT_dec (self->cb_object);	1939	SvREFCNT_dec (self->cb_object);
		1940	SvREFCNT_dec (self->incr_text);
1505		1941
1506	PROTOTYPES: ENABLE	1942	PROTOTYPES: ENABLE
1507		1943
1508	void to_json (SV *scalar)	1944	void encode_json (SV *scalar)
		1945	ALIAS:
		1946	to_json_ = 0
		1947	encode_json = F_UTF8
1509	PPCODE:	1948	PPCODE:
1510	{	1949	{
1511	JSON json = { F_DEFAULT \| F_UTF8 };	1950	JSON json;
		1951	json_init (&json);
		1952	json.flags \|= ix;
1512	XPUSHs (encode_json (scalar, &json));	1953	XPUSHs (encode_json (scalar, &json));
1513	}	1954	}
1514		1955
1515	void from_json (SV *jsonstr)	1956	void decode_json (SV *jsonstr)
		1957	ALIAS:
		1958	from_json_ = 0
		1959	decode_json = F_UTF8
1516	PPCODE:	1960	PPCODE:
1517	{	1961	{
1518	JSON json = { F_DEFAULT \| F_UTF8 };	1962	JSON json;
		1963	json_init (&json);
		1964	json.flags \|= ix;
1519	XPUSHs (decode_json (jsonstr, &json, 0));	1965	XPUSHs (decode_json (jsonstr, &json, 0));
1520	}	1966	}
1521		1967
		1968

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.60 by root, Mon Aug 13 16:19:13 2007 UTC vs. Revision 1.103 by root, Wed Jan 6 08:01:39 2010 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.60 by root, Mon Aug 13 16:19:13 2007 UTC vs.
Revision 1.103 by root, Wed Jan 6 08:01:39 2010 UTC