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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 2008 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
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>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		15
14	// 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
15	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarentees, though. if it breaks, you get to keep the pieces.
16	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
17	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
18	#endif	20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
19		23
20	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
21	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
22	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
23	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
25	#define F_SPACE_BEFORE 0x00000020UL	29	#define F_SPACE_BEFORE 0x00000020UL
26	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
27	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
28	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
29	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
30	#define F_CONV_BLESSED 0x00000800UL // NYI	34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
31	#define F_MAXDEPTH 0xf8000000UL	37	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27	38	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	39	#define F_MAXSIZE 0x01f00000UL
34	#define S_MAXSIZE 20	40	#define S_MAXSIZE 20
		41	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
35		42
36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	43	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))	44	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
38		45
39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	46	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
…		…
46		53
47	#define SB do {	54	#define SB do {
48	#define SE } while (0)	55	#define SE } while (0)
49		56
50	#if __GNUC__ >= 3	57	#if __GNUC__ >= 3
51	# define expect(expr,value) __builtin_expect ((expr),(value))	58	# define expect(expr,value) __builtin_expect ((expr), (value))
52	# define inline inline	59	# define INLINE static inline
53	#else	60	#else
54	# define expect(expr,value) (expr)	61	# define expect(expr,value) (expr)
55	# define inline static	62	# define INLINE static
56	#endif	63	#endif
57		64
58	#define expect_false(expr) expect ((expr) != 0, 0)	65	#define expect_false(expr) expect ((expr) != 0, 0)
59	#define expect_true(expr) expect ((expr) != 0, 1)	66	#define expect_true(expr) expect ((expr) != 0, 1)
60		67
		68	#define IN_RANGE_INC(type,val,beg,end) \
		69	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		70	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		71
		72	#ifdef USE_ITHREADS
		73	# define JSON_SLOW 1
		74	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		75	#else
		76	# define JSON_SLOW 0
		77	# define JSON_STASH json_stash
		78	#endif
		79
61	static HV json_stash, json_boolean_stash; // JSON::XS::	80	static HV json_stash, json_boolean_stash; // JSON::XS::
62	static SV json_true, json_false;	81	static SV json_true, json_false;
63		82
		83	enum {
		84	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		85	INCR_M_STR, // inside string
		86	INCR_M_BS, // inside backslash
		87	INCR_M_JSON // outside anything, count nesting
		88	};
		89
		90	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		91
64	typedef struct {	92	typedef struct {
65	U32 flags;	93	U32 flags;
		94	SV *cb_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	int incr_mode;
66	} JSON;	102	} JSON;
67		103
68	/////////////////////////////////////////////////////////////////////////////	104	/////////////////////////////////////////////////////////////////////////////
69	// utility functions	105	// utility functions
70		106
71	inline void	107	INLINE SV *
		108	get_bool (const char *name)
		109	{
		110	SV *sv = get_sv (name, 1);
		111
		112	SvREADONLY_on (sv);
		113	SvREADONLY_on (SvRV (sv));
		114
		115	return sv;
		116	}
		117
		118	INLINE void
72	shrink (SV *sv)	119	shrink (SV *sv)
73	{	120	{
74	sv_utf8_downgrade (sv, 1);	121	sv_utf8_downgrade (sv, 1);
75	if (SvLEN (sv) > SvCUR (sv) + 1)	122	if (SvLEN (sv) > SvCUR (sv) + 1)
76	{	123	{
…		…
85	// decode an utf-8 character and return it, or (UV)-1 in	132	// decode an utf-8 character and return it, or (UV)-1 in
86	// case of an error.	133	// case of an error.
87	// we special-case "safe" characters from U+80 .. U+7FF,	134	// we special-case "safe" characters from U+80 .. U+7FF,
88	// but use the very good perl function to parse anything else.	135	// but use the very good perl function to parse anything else.
89	// note that we never call this function for a ascii codepoints	136	// note that we never call this function for a ascii codepoints
90	inline UV	137	INLINE UV
91	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	138	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
92	{	139	{
93	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	140	if (expect_true (len >= 2
94	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	141	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
95	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	142	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
96	{	143	{
97	*clen = 2;	144	*clen = 2;
98	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	145	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
99	}	146	}
100	else	147	else
101	{	148	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
102	*clen = (STRLEN)-1;	149	}
103	return (UV)-1;	150
104	}	151	// likewise for encoding, also never called for ascii codepoints
		152	// this function takes advantage of this fact, although current gccs
		153	// seem to optimise the check for >= 0x80 away anyways
		154	INLINE unsigned char *
		155	encode_utf8 (unsigned char *s, UV ch)
		156	{
		157	if (expect_false (ch < 0x000080))
		158	*s++ = ch;
		159	else if (expect_true (ch < 0x000800))
		160	*s++ = 0xc0 \| ( ch >> 6),
		161	*s++ = 0x80 \| ( ch & 0x3f);
		162	else if ( ch < 0x010000)
		163	*s++ = 0xe0 \| ( ch >> 12),
		164	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		165	*s++ = 0x80 \| ( ch & 0x3f);
		166	else if ( ch < 0x110000)
		167	*s++ = 0xf0 \| ( ch >> 18),
		168	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		169	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 \| ( ch & 0x3f);
		171
		172	return s;
105	}	173	}
106		174
107	/////////////////////////////////////////////////////////////////////////////	175	/////////////////////////////////////////////////////////////////////////////
108	// encoder	176	// encoder
109		177
…		…
114	char *end; // SvEND (sv)	182	char *end; // SvEND (sv)
115	SV *sv; // result scalar	183	SV *sv; // result scalar
116	JSON json;	184	JSON json;
117	U32 indent; // indentation level	185	U32 indent; // indentation level
118	U32 maxdepth; // max. indentation/recursion level	186	U32 maxdepth; // max. indentation/recursion level
		187	UV limit; // escape character values >= this value when encoding
119	} enc_t;	188	} enc_t;
120		189
121	inline void	190	INLINE void
122	need (enc_t *enc, STRLEN len)	191	need (enc_t *enc, STRLEN len)
123	{	192	{
124	if (expect_false (enc->cur + len >= enc->end))	193	if (expect_false (enc->cur + len >= enc->end))
125	{	194	{
126	STRLEN cur = enc->cur - SvPVX (enc->sv);	195	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
128	enc->cur = SvPVX (enc->sv) + cur;	197	enc->cur = SvPVX (enc->sv) + cur;
129	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	198	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
130	}	199	}
131	}	200	}
132		201
133	inline void	202	INLINE void
134	encode_ch (enc_t *enc, char ch)	203	encode_ch (enc_t *enc, char ch)
135	{	204	{
136	need (enc, 1);	205	need (enc, 1);
137	*enc->cur++ = ch;	206	*enc->cur++ = ch;
138	}	207	}
…		…
192	{	261	{
193	uch = ch;	262	uch = ch;
194	clen = 1;	263	clen = 1;
195	}	264	}
196		265
197	if (uch > 0x10FFFFUL)	266	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
198	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
199
200	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
201	{	267	{
202	if (uch > 0xFFFFUL)	268	if (uch >= 0x10000UL)
203	{	269	{
		270	if (uch >= 0x110000UL)
		271	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		272
204	need (enc, len += 11);	273	need (enc, len += 11);
205	sprintf (enc->cur, "\\u%04x\\u%04x",	274	sprintf (enc->cur, "\\u%04x\\u%04x",
206	(int)((uch - 0x10000) / 0x400 + 0xD800),	275	(int)((uch - 0x10000) / 0x400 + 0xD800),
207	(int)((uch - 0x10000) % 0x400 + 0xDC00));	276	(int)((uch - 0x10000) % 0x400 + 0xDC00));
208	enc->cur += 12;	277	enc->cur += 12;
…		…
236	while (--clen);	305	while (--clen);
237	}	306	}
238	else	307	else
239	{	308	{
240	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	309	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
241	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	310	enc->cur = encode_utf8 (enc->cur, uch);
242	++str;	311	++str;
243	}	312	}
244	}	313	}
245	}	314	}
246	}	315	}
247		316
248	--len;	317	--len;
249	}	318	}
250	}	319	}
251		320
252	inline void	321	INLINE void
253	encode_indent (enc_t *enc)	322	encode_indent (enc_t *enc)
254	{	323	{
255	if (enc->json.flags & F_INDENT)	324	if (enc->json.flags & F_INDENT)
256	{	325	{
257	int spaces = enc->indent * INDENT_STEP;	326	int spaces = enc->indent * INDENT_STEP;
…		…
260	memset (enc->cur, ' ', spaces);	329	memset (enc->cur, ' ', spaces);
261	enc->cur += spaces;	330	enc->cur += spaces;
262	}	331	}
263	}	332	}
264		333
265	inline void	334	INLINE void
266	encode_space (enc_t *enc)	335	encode_space (enc_t *enc)
267	{	336	{
268	need (enc, 1);	337	need (enc, 1);
269	encode_ch (enc, ' ');	338	encode_ch (enc, ' ');
270	}	339	}
271		340
272	inline void	341	INLINE void
273	encode_nl (enc_t *enc)	342	encode_nl (enc_t *enc)
274	{	343	{
275	if (enc->json.flags & F_INDENT)	344	if (enc->json.flags & F_INDENT)
276	{	345	{
277	need (enc, 1);	346	need (enc, 1);
278	encode_ch (enc, '\n');	347	encode_ch (enc, '\n');
279	}	348	}
280	}	349	}
281		350
282	inline void	351	INLINE void
283	encode_comma (enc_t *enc)	352	encode_comma (enc_t *enc)
284	{	353	{
285	encode_ch (enc, ',');	354	encode_ch (enc, ',');
286		355
287	if (enc->json.flags & F_INDENT)	356	if (enc->json.flags & F_INDENT)
…		…
298	int i, len = av_len (av);	367	int i, len = av_len (av);
299		368
300	if (enc->indent >= enc->maxdepth)	369	if (enc->indent >= enc->maxdepth)
301	croak ("data structure too deep (hit recursion limit)");	370	croak ("data structure too deep (hit recursion limit)");
302		371
303	encode_ch (enc, '['); encode_nl (enc);	372	encode_ch (enc, '[');
304	++enc->indent;	373
		374	if (len >= 0)
		375	{
		376	encode_nl (enc); ++enc->indent;
305		377
306	for (i = 0; i <= len; ++i)	378	for (i = 0; i <= len; ++i)
307	{	379	{
		380	SV **svp = av_fetch (av, i, 0);
		381
308	encode_indent (enc);	382	encode_indent (enc);
309	encode_sv (enc, *av_fetch (av, i, 0));
310		383
		384	if (svp)
		385	encode_sv (enc, *svp);
		386	else
		387	encode_str (enc, "null", 4, 0);
		388
311	if (i < len)	389	if (i < len)
312	encode_comma (enc);	390	encode_comma (enc);
313	}	391	}
314		392
		393	encode_nl (enc); --enc->indent; encode_indent (enc);
		394	}
		395
315	encode_nl (enc);	396	encode_ch (enc, ']');
316
317	--enc->indent;
318	encode_indent (enc); encode_ch (enc, ']');
319	}	397	}
320		398
321	static void	399	static void
322	encode_he (enc_t enc, HE he)	400	encode_hk (enc_t enc, HE he)
323	{	401	{
324	encode_ch (enc, '"');	402	encode_ch (enc, '"');
325		403
326	if (HeKLEN (he) == HEf_SVKEY)	404	if (HeKLEN (he) == HEf_SVKEY)
327	{	405	{
…		…
340	encode_ch (enc, '"');	418	encode_ch (enc, '"');
341		419
342	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	420	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
343	encode_ch (enc, ':');	421	encode_ch (enc, ':');
344	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	422	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
345	encode_sv (enc, HeVAL (he));
346	}	423	}
347		424
348	// compare hash entries, used when all keys are bytestrings	425	// compare hash entries, used when all keys are bytestrings
349	static int	426	static int
350	he_cmp_fast (const void a_, const void b_)	427	he_cmp_fast (const void a_, const void b_)
…		…
355	HE b = (HE **)b_;	432	HE b = (HE **)b_;
356		433
357	STRLEN la = HeKLEN (a);	434	STRLEN la = HeKLEN (a);
358	STRLEN lb = HeKLEN (b);	435	STRLEN lb = HeKLEN (b);
359		436
360	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	437	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
361	cmp = la - lb;	438	cmp = lb - la;
362		439
363	return cmp;	440	return cmp;
364	}	441	}
365		442
366	// compare hash entries, used when some keys are sv's or utf-x	443	// compare hash entries, used when some keys are sv's or utf-x
367	static int	444	static int
368	he_cmp_slow (const void a, const void b)	445	he_cmp_slow (const void a, const void b)
369	{	446	{
370	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	447	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
371	}	448	}
372		449
373	static void	450	static void
374	encode_hv (enc_t enc, HV hv)	451	encode_hv (enc_t enc, HV hv)
375	{	452	{
376	int count, i;	453	HE *he;
377		454
378	if (enc->indent >= enc->maxdepth)	455	if (enc->indent >= enc->maxdepth)
379	croak ("data structure too deep (hit recursion limit)");	456	croak ("data structure too deep (hit recursion limit)");
380		457
381	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	458	encode_ch (enc, '{');
382		459
383	if ((count = hv_iterinit (hv)))
384	{
385	// for canonical output we have to sort by keys first	460	// for canonical output we have to sort by keys first
386	// actually, this is mostly due to the stupid so-called	461	// actually, this is mostly due to the stupid so-called
387	// security workaround added somewhere in 5.8.x.	462	// security workaround added somewhere in 5.8.x.
388	// that randomises hash orderings	463	// that randomises hash orderings
389	if (enc->json.flags & F_CANONICAL)	464	if (enc->json.flags & F_CANONICAL)
		465	{
		466	int count = hv_iterinit (hv);
		467
		468	if (SvMAGICAL (hv))
390	{	469	{
		470	// need to count by iterating. could improve by dynamically building the vector below
		471	// but I don't care for the speed of this special case.
		472	// note also that we will run into undefined behaviour when the two iterations
		473	// do not result in the same count, something I might care for in some later release.
		474
		475	count = 0;
		476	while (hv_iternext (hv))
		477	++count;
		478
		479	hv_iterinit (hv);
		480	}
		481
		482	if (count)
		483	{
391	int fast = 1;	484	int i, fast = 1;
392	HE *he;
393	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	485	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
394	HE *hes = _alloca (count sizeof (HE));	486	HE *hes = _alloca (count sizeof (HE));
395	#else	487	#else
396	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	488	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
397	#endif	489	#endif
…		…
424		516
425	FREETMPS;	517	FREETMPS;
426	LEAVE;	518	LEAVE;
427	}	519	}
428		520
429	for (i = 0; i < count; ++i)	521	encode_nl (enc); ++enc->indent;
		522
		523	while (count--)
430	{	524	{
431	encode_indent (enc);	525	encode_indent (enc);
		526	he = hes [count];
432	encode_he (enc, hes [i]);	527	encode_hk (enc, he);
		528	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
433		529
434	if (i < count - 1)	530	if (count)
435	encode_comma (enc);	531	encode_comma (enc);
436	}	532	}
437		533
438	encode_nl (enc);	534	encode_nl (enc); --enc->indent; encode_indent (enc);
439	}	535	}
		536	}
440	else	537	else
		538	{
		539	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		540	if ((he = hv_iternext (hv)))
441	{	541	{
442	HE *he = hv_iternext (hv);	542	encode_nl (enc); ++enc->indent;
443		543
444	for (;;)	544	for (;;)
445	{	545	{
446	encode_indent (enc);	546	encode_indent (enc);
447	encode_he (enc, he);	547	encode_hk (enc, he);
		548	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
448		549
449	if (!(he = hv_iternext (hv)))	550	if (!(he = hv_iternext (hv)))
450	break;	551	break;
451		552
452	encode_comma (enc);	553	encode_comma (enc);
453	}	554	}
454		555
455	encode_nl (enc);	556	encode_nl (enc); --enc->indent; encode_indent (enc);
456	}	557	}
457	}	558	}
458		559
459	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	560	encode_ch (enc, '}');
460	}	561	}
461		562
462	// encode objects, arrays and special \0=false and \1=true values.	563	// encode objects, arrays and special \0=false and \1=true values.
463	static void	564	static void
464	encode_rv (enc_t enc, SV sv)	565	encode_rv (enc_t enc, SV sv)
…		…
468	SvGETMAGIC (sv);	569	SvGETMAGIC (sv);
469	svt = SvTYPE (sv);	570	svt = SvTYPE (sv);
470		571
471	if (expect_false (SvOBJECT (sv)))	572	if (expect_false (SvOBJECT (sv)))
472	{	573	{
		574	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		575	? json_boolean_stash
		576	: gv_stashpv ("JSON::XS::Boolean", 1);
		577
473	if (SvSTASH (sv) == json_boolean_stash)	578	if (SvSTASH (sv) == stash)
474	{	579	{
475	if (SvIV (sv) == 0)	580	if (SvIV (sv))
		581	encode_str (enc, "true", 4, 0);
		582	else
476	encode_str (enc, "false", 5, 0);	583	encode_str (enc, "false", 5, 0);
477	else
478	encode_str (enc, "true", 4, 0);
479	}	584	}
480	else	585	else
481	{	586	{
482	#if 0	587	#if 0
483	if (0 && sv_derived_from (rv, "JSON::Literal"))	588	if (0 && sv_derived_from (rv, "JSON::Literal"))
…		…
486	}	591	}
487	#endif	592	#endif
488	if (enc->json.flags & F_CONV_BLESSED)	593	if (enc->json.flags & F_CONV_BLESSED)
489	{	594	{
490	// we re-bless the reference to get overload and other niceties right	595	// we re-bless the reference to get overload and other niceties right
491	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	596	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
492		597
493	if (to_json)	598	if (to_json)
494	{	599	{
495	dSP;	600	dSP;
496	ENTER;	601
497	SAVETMPS;
498	PUSHMARK (SP);	602	ENTER; SAVETMPS; PUSHMARK (SP);
499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	603	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
500		604
501	// calling with G_SCALAR ensures that we always get a 1 reutrn value	605	// calling with G_SCALAR ensures that we always get a 1 return value
502	// check anyways.
503	PUTBACK;	606	PUTBACK;
504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	607	call_sv ((SV *)GvCV (to_json), G_SCALAR);
505	SPAGAIN;	608	SPAGAIN;
506		609
		610	// catch this surprisingly common error
		611	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		612	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		613
		614	sv = POPs;
		615	PUTBACK;
		616
507	encode_sv (enc, POPs);	617	encode_sv (enc, sv);
508		618
509	FREETMPS;	619	FREETMPS; LEAVE;
510	LEAVE;
511	}	620	}
512	else if (enc->json.flags & F_ALLOW_BLESSED)	621	else if (enc->json.flags & F_ALLOW_BLESSED)
513	encode_str (enc, "null", 4, 0);	622	encode_str (enc, "null", 4, 0);
514	else	623	else
515	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",	624	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
…		…
526	encode_hv (enc, (HV *)sv);	635	encode_hv (enc, (HV *)sv);
527	else if (svt == SVt_PVAV)	636	else if (svt == SVt_PVAV)
528	encode_av (enc, (AV *)sv);	637	encode_av (enc, (AV *)sv);
529	else if (svt < SVt_PVAV)	638	else if (svt < SVt_PVAV)
530	{	639	{
531	if (SvNIOK (sv) && SvIV (sv) == 0)	640	STRLEN len = 0;
		641	char *pv = svt ? SvPV (sv, len) : 0;
		642
		643	if (len == 1 && *pv == '1')
		644	encode_str (enc, "true", 4, 0);
		645	else if (len == 1 && *pv == '0')
532	encode_str (enc, "false", 5, 0);	646	encode_str (enc, "false", 5, 0);
533	else if (SvNIOK (sv) && SvIV (sv) == 1)
534	encode_str (enc, "true", 4, 0);
535	else	647	else
536	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	648	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	649	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
538	}	650	}
539	else	651	else
…		…
561	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	673	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
562	enc->cur += strlen (enc->cur);	674	enc->cur += strlen (enc->cur);
563	}	675	}
564	else if (SvIOKp (sv))	676	else if (SvIOKp (sv))
565	{	677	{
566	// we assume we can always read an IV as a UV	678	// we assume we can always read an IV as a UV and vice versa
567	if (SvUV (sv) & ~(UV)0x7fff)	679	// we assume two's complement
568	{	680	// we assume no aliasing issues in the union
569	// large integer, use the (rather slow) snprintf way.	681	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
570	need (enc, sizeof (UV) * 3);	682	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
571	enc->cur +=
572	SvIsUV(sv)
573	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
574	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
575	}
576	else
577	{	683	{
578	// optimise the "small number case"	684	// optimise the "small number case"
579	// code will likely be branchless and use only a single multiplication	685	// code will likely be branchless and use only a single multiplication
		686	// works for numbers up to 59074
580	I32 i = SvIV (sv);	687	I32 i = SvIVX (sv);
581	U32 u;	688	U32 u;
582	char digit, nz = 0;	689	char digit, nz = 0;
583		690
584	need (enc, 6);	691	need (enc, 6);
585		692
…		…
591		698
592	// now output digit by digit, each time masking out the integer part	699	// now output digit by digit, each time masking out the integer part
593	// and multiplying by 5 while moving the decimal point one to the right,	700	// and multiplying by 5 while moving the decimal point one to the right,
594	// resulting in a net multiplication by 10.	701	// resulting in a net multiplication by 10.
595	// we always write the digit to memory but conditionally increment	702	// we always write the digit to memory but conditionally increment
596	// the pointer, to ease the usage of conditional move instructions.	703	// the pointer, to enable the use of conditional move instructions.
597	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	704	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
598	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	705	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
599	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	706	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
600	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	707	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
601	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	708	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		709	}
		710	else
		711	{
		712	// large integer, use the (rather slow) snprintf way.
		713	need (enc, IVUV_MAXCHARS);
		714	enc->cur +=
		715	SvIsUV(sv)
		716	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		717	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
602	}	718	}
603	}	719	}
604	else if (SvROK (sv))	720	else if (SvROK (sv))
605	encode_rv (enc, SvRV (sv));	721	encode_rv (enc, SvRV (sv));
606	else if (!SvOK (sv))	722	else if (!SvOK (sv))
…		…
622	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	738	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
623	enc.cur = SvPVX (enc.sv);	739	enc.cur = SvPVX (enc.sv);
624	enc.end = SvEND (enc.sv);	740	enc.end = SvEND (enc.sv);
625	enc.indent = 0;	741	enc.indent = 0;
626	enc.maxdepth = DEC_DEPTH (enc.json.flags);	742	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		743	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		744	: enc.json.flags & F_LATIN1 ? 0x000100UL
		745	: 0x110000UL;
627		746
628	SvPOK_only (enc.sv);	747	SvPOK_only (enc.sv);
629	encode_sv (&enc, scalar);	748	encode_sv (&enc, scalar);
630		749
631	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	750	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
652	JSON json;	771	JSON json;
653	U32 depth; // recursion depth	772	U32 depth; // recursion depth
654	U32 maxdepth; // recursion depth limit	773	U32 maxdepth; // recursion depth limit
655	} dec_t;	774	} dec_t;
656		775
657	inline void	776	INLINE void
		777	decode_comment (dec_t *dec)
		778	{
		779	// only '#'-style comments allowed a.t.m.
		780
		781	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		782	++dec->cur;
		783	}
		784
		785	INLINE void
658	decode_ws (dec_t *dec)	786	decode_ws (dec_t *dec)
659	{	787	{
660	for (;;)	788	for (;;)
661	{	789	{
662	char ch = *dec->cur;	790	char ch = *dec->cur;
663		791
664	if (ch > 0x20	792	if (ch > 0x20)
		793	{
		794	if (expect_false (ch == '#'))
		795	{
		796	if (dec->json.flags & F_RELAXED)
		797	decode_comment (dec);
		798	else
		799	break;
		800	}
		801	else
		802	break;
		803	}
665	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	804	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
666	break;	805	break; // parse error, but let higher level handle it, gives better error messages
667		806
668	++dec->cur;	807	++dec->cur;
669	}	808	}
670	}	809	}
671		810
…		…
777		916
778	if (hi >= 0x80)	917	if (hi >= 0x80)
779	{	918	{
780	utf8 = 1;	919	utf8 = 1;
781		920
782	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	921	cur = encode_utf8 (cur, hi);
783	}	922	}
784	else	923	else
785	*cur++ = hi;	924	*cur++ = hi;
786	}	925	}
787	break;	926	break;
…		…
789	default:	928	default:
790	--dec_cur;	929	--dec_cur;
791	ERR ("illegal backslash escape sequence in string");	930	ERR ("illegal backslash escape sequence in string");
792	}	931	}
793	}	932	}
794	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	933	else if (expect_true (ch >= 0x20 && ch < 0x80))
795	*cur++ = ch;	934	*cur++ = ch;
796	else if (ch >= 0x80)	935	else if (ch >= 0x80)
797	{	936	{
798	STRLEN clen;	937	STRLEN clen;
799	UV uch;	938	UV uch;
…		…
920	is_nv = 1;	1059	is_nv = 1;
921	}	1060	}
922		1061
923	if (!is_nv)	1062	if (!is_nv)
924	{	1063	{
		1064	int len = dec->cur - start;
		1065
925	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1066	// special case the rather common 1..5-digit-int case
926	if (*start == '-')	1067	if (*start == '-')
927	switch (dec->cur - start)	1068	switch (len)
928	{	1069	{
929	case 2: return newSViv (-( start [1] - '0' * 1));	1070	case 2: return newSViv (-( start [1] - '0' * 1));
930	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1071	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
931	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1072	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
932	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1073	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1074	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
933	}	1075	}
934	else	1076	else
935	switch (dec->cur - start)	1077	switch (len)
936	{	1078	{
937	case 1: return newSViv ( start [0] - '0' * 1);	1079	case 1: return newSViv ( start [0] - '0' * 1);
938	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1080	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
939	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1081	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
940	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1082	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1083	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
941	}	1084	}
942		1085
943	{	1086	{
944	UV uv;	1087	UV uv;
945	int numtype = grok_number (start, dec->cur - start, &uv);	1088	int numtype = grok_number (start, len, &uv);
946	if (numtype & IS_NUMBER_IN_UV)	1089	if (numtype & IS_NUMBER_IN_UV)
947	if (numtype & IS_NUMBER_NEG)	1090	if (numtype & IS_NUMBER_NEG)
948	{	1091	{
949	if (uv < (UV)IV_MIN)	1092	if (uv < (UV)IV_MIN)
950	return newSViv (-(IV)uv);	1093	return newSViv (-(IV)uv);
951	}	1094	}
952	else	1095	else
953	return newSVuv (uv);	1096	return newSVuv (uv);
954
955	// here would likely be the place for bigint support
956	}	1097	}
957	}
958		1098
959	// if we ever support bigint or bigfloat, this is the place for bigfloat	1099	len -= *start == '-' ? 1 : 0;
		1100
		1101	// does not fit into IV or UV, try NV
		1102	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1103	#if defined (LDBL_DIG)
		1104	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1105	#endif
		1106	)
		1107	// fits into NV without loss of precision
		1108	return newSVnv (Atof (start));
		1109
		1110	// everything else fails, convert it to a string
		1111	return newSVpvn (start, dec->cur - start);
		1112	}
		1113
		1114	// loss of precision here
960	return newSVnv (Atof (start));	1115	return newSVnv (Atof (start));
961		1116
962	fail:	1117	fail:
963	return 0;	1118	return 0;
964	}	1119	}
…		…
994		1149
995	if (*dec->cur != ',')	1150	if (*dec->cur != ',')
996	ERR (", or ] expected while parsing array");	1151	ERR (", or ] expected while parsing array");
997		1152
998	++dec->cur;	1153	++dec->cur;
		1154
		1155	decode_ws (dec);
		1156
		1157	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1158	{
		1159	++dec->cur;
		1160	break;
		1161	}
999	}	1162	}
1000		1163
1001	DEC_DEC_DEPTH;	1164	DEC_DEC_DEPTH;
1002	return newRV_noinc ((SV *)av);	1165	return newRV_noinc ((SV *)av);
1003		1166
…		…
1008	}	1171	}
1009		1172
1010	static SV *	1173	static SV *
1011	decode_hv (dec_t *dec)	1174	decode_hv (dec_t *dec)
1012	{	1175	{
		1176	SV *sv;
1013	HV *hv = newHV ();	1177	HV *hv = newHV ();
1014		1178
1015	DEC_INC_DEPTH;	1179	DEC_INC_DEPTH;
1016	decode_ws (dec);	1180	decode_ws (dec);
1017		1181
1018	if (*dec->cur == '}')	1182	if (*dec->cur == '}')
1019	++dec->cur;	1183	++dec->cur;
1020	else	1184	else
1021	for (;;)	1185	for (;;)
1022	{	1186	{
1023	decode_ws (dec); EXPECT_CH ('"');	1187	EXPECT_CH ('"');
1024		1188
1025	// heuristic: assume that	1189	// heuristic: assume that
1026	// a) decode_str + hv_store_ent are abysmally slow.	1190	// a) decode_str + hv_store_ent are abysmally slow.
1027	// b) most hash keys are short, simple ascii text.	1191	// b) most hash keys are short, simple ascii text.
1028	// => try to "fast-match" such strings to avoid	1192	// => try to "fast-match" such strings to avoid
…		…
1032	char *p = dec->cur;	1196	char *p = dec->cur;
1033	char *e = p + 24; // only try up to 24 bytes	1197	char *e = p + 24; // only try up to 24 bytes
1034		1198
1035	for (;;)	1199	for (;;)
1036	{	1200	{
1037	// the >= 0x80 is true on most architectures	1201	// the >= 0x80 is false on most architectures
1038	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1202	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1039	{	1203	{
1040	// slow path, back up and use decode_str	1204	// slow path, back up and use decode_str
1041	SV *key = decode_str (dec);	1205	SV *key = decode_str (dec);
1042	if (!key)	1206	if (!key)
1043	goto fail;	1207	goto fail;
1044		1208
1045	decode_ws (dec); EXPECT_CH (':');	1209	decode_ws (dec); EXPECT_CH (':');
1046		1210
		1211	decode_ws (dec);
1047	value = decode_sv (dec);	1212	value = decode_sv (dec);
1048	if (!value)	1213	if (!value)
1049	{	1214	{
1050	SvREFCNT_dec (key);	1215	SvREFCNT_dec (key);
1051	goto fail;	1216	goto fail;
…		…
1063	int len = p - key;	1228	int len = p - key;
1064	dec->cur = p + 1;	1229	dec->cur = p + 1;
1065		1230
1066	decode_ws (dec); EXPECT_CH (':');	1231	decode_ws (dec); EXPECT_CH (':');
1067		1232
		1233	decode_ws (dec);
1068	value = decode_sv (dec);	1234	value = decode_sv (dec);
1069	if (!value)	1235	if (!value)
1070	goto fail;	1236	goto fail;
1071		1237
1072	hv_store (hv, key, len, value, 0);	1238	hv_store (hv, key, len, value, 0);
…		…
1088		1254
1089	if (*dec->cur != ',')	1255	if (*dec->cur != ',')
1090	ERR (", or } expected while parsing object/hash");	1256	ERR (", or } expected while parsing object/hash");
1091		1257
1092	++dec->cur;	1258	++dec->cur;
		1259
		1260	decode_ws (dec);
		1261
		1262	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1263	{
		1264	++dec->cur;
		1265	break;
		1266	}
1093	}	1267	}
1094		1268
1095	DEC_DEC_DEPTH;	1269	DEC_DEC_DEPTH;
1096	return newRV_noinc ((SV *)hv);	1270	sv = newRV_noinc ((SV *)hv);
		1271
		1272	// check filter callbacks
		1273	if (dec->json.flags & F_HOOK)
		1274	{
		1275	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1276	{
		1277	HE cb, he;
		1278
		1279	hv_iterinit (hv);
		1280	he = hv_iternext (hv);
		1281	hv_iterinit (hv);
		1282
		1283	// the next line creates a mortal sv each time its called.
		1284	// might want to optimise this for common cases.
		1285	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1286
		1287	if (cb)
		1288	{
		1289	dSP;
		1290	int count;
		1291
		1292	ENTER; SAVETMPS; PUSHMARK (SP);
		1293	XPUSHs (HeVAL (he));
		1294
		1295	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1296
		1297	if (count == 1)
		1298	{
		1299	sv = newSVsv (POPs);
		1300	FREETMPS; LEAVE;
		1301	return sv;
		1302	}
		1303
		1304	FREETMPS; LEAVE;
		1305	}
		1306	}
		1307
		1308	if (dec->json.cb_object)
		1309	{
		1310	dSP;
		1311	int count;
		1312
		1313	ENTER; SAVETMPS; PUSHMARK (SP);
		1314	XPUSHs (sv_2mortal (sv));
		1315
		1316	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1317
		1318	if (count == 1)
		1319	{
		1320	sv = newSVsv (POPs);
		1321	FREETMPS; LEAVE;
		1322	return sv;
		1323	}
		1324
		1325	SvREFCNT_inc (sv);
		1326	FREETMPS; LEAVE;
		1327	}
		1328	}
		1329
		1330	return sv;
1097		1331
1098	fail:	1332	fail:
1099	SvREFCNT_dec (hv);	1333	SvREFCNT_dec (hv);
1100	DEC_DEC_DEPTH;	1334	DEC_DEC_DEPTH;
1101	return 0;	1335	return 0;
1102	}	1336	}
1103		1337
1104	static SV *	1338	static SV *
1105	decode_sv (dec_t *dec)	1339	decode_sv (dec_t *dec)
1106	{	1340	{
1107	decode_ws (dec);
1108
1109	// the beauty of JSON: you need exactly one character lookahead	1341	// the beauty of JSON: you need exactly one character lookahead
1110	// to parse anything.	1342	// to parse everything.
1111	switch (*dec->cur)	1343	switch (*dec->cur)
1112	{	1344	{
1113	case '"': ++dec->cur; return decode_str (dec);	1345	case '"': ++dec->cur; return decode_str (dec);
1114	case '[': ++dec->cur; return decode_av (dec);	1346	case '[': ++dec->cur; return decode_av (dec);
1115	case '{': ++dec->cur; return decode_hv (dec);	1347	case '{': ++dec->cur; return decode_hv (dec);
1116		1348
1117	case '-':	1349	case '-':
1118	case '0': case '1': case '2': case '3': case '4':	1350	case '0': case '1': case '2': case '3': case '4':
1119	case '5': case '6': case '7': case '8': case '9':	1351	case '5': case '6': case '7': case '8': case '9':
1120	return decode_num (dec);	1352	return decode_num (dec);
1121		1353
1122	case 't':	1354	case 't':
1123	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1355	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1124	{	1356	{
1125	dec->cur += 4;	1357	dec->cur += 4;
		1358	#if JSON_SLOW
		1359	json_true = get_bool ("JSON::XS::true");
		1360	#endif
1126	return SvREFCNT_inc (json_true);	1361	return newSVsv (json_true);
1127	}	1362	}
1128	else	1363	else
1129	ERR ("'true' expected");	1364	ERR ("'true' expected");
1130		1365
1131	break;	1366	break;
1132		1367
1133	case 'f':	1368	case 'f':
1134	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1369	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1135	{	1370	{
1136	dec->cur += 5;	1371	dec->cur += 5;
		1372	#if JSON_SLOW
		1373	json_false = get_bool ("JSON::XS::false");
		1374	#endif
1137	return SvREFCNT_inc (json_false);	1375	return newSVsv (json_false);
1138	}	1376	}
1139	else	1377	else
1140	ERR ("'false' expected");	1378	ERR ("'false' expected");
1141		1379
1142	break;	1380	break;
…		…
1160	fail:	1398	fail:
1161	return 0;	1399	return 0;
1162	}	1400	}
1163		1401
1164	static SV *	1402	static SV *
1165	decode_json (SV string, JSON json, UV *offset_return)	1403	decode_json (SV string, JSON json, STRLEN *offset_return)
1166	{	1404	{
1167	dec_t dec;	1405	dec_t dec;
1168	UV offset;	1406	STRLEN offset;
1169	SV *sv;	1407	SV *sv;
1170		1408
1171	SvGETMAGIC (string);	1409	SvGETMAGIC (string);
1172	SvUPGRADE (string, SVt_PV);	1410	SvUPGRADE (string, SVt_PV);
1173		1411
…		…
1187	dec.end = SvEND (string);	1425	dec.end = SvEND (string);
1188	dec.err = 0;	1426	dec.err = 0;
1189	dec.depth = 0;	1427	dec.depth = 0;
1190	dec.maxdepth = DEC_DEPTH (dec.json.flags);	1428	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1191		1429
		1430	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1431	dec.json.flags \|= F_HOOK;
		1432
1192	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1433	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1434
		1435	decode_ws (&dec);
1193	sv = decode_sv (&dec);	1436	sv = decode_sv (&dec);
1194		1437
1195	if (!(offset_return \|\| !sv))	1438	if (!(offset_return \|\| !sv))
1196	{	1439	{
1197	// check for trailing garbage	1440	// check for trailing garbage
…		…
1241		1484
1242	return sv;	1485	return sv;
1243	}	1486	}
1244		1487
1245	/////////////////////////////////////////////////////////////////////////////	1488	/////////////////////////////////////////////////////////////////////////////
		1489	// incremental parser
		1490
		1491	static void
		1492	incr_parse (JSON *self)
		1493	{
		1494	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1495
		1496	for (;;)
		1497	{
		1498	//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
		1499	switch (self->incr_mode)
		1500	{
		1501	// only used for intiial whitespace skipping
		1502	case INCR_M_WS:
		1503	for (;;)
		1504	{
		1505	if (*p > 0x20)
		1506	{
		1507	self->incr_mode = INCR_M_JSON;
		1508	goto incr_m_json;
		1509	}
		1510	else if (!*p)
		1511	goto interrupt;
		1512
		1513	++p;
		1514	}
		1515
		1516	// skip a single char inside a string (for \\-processing)
		1517	case INCR_M_BS:
		1518	if (!*p)
		1519	goto interrupt;
		1520
		1521	++p;
		1522	self->incr_mode = INCR_M_STR;
		1523	goto incr_m_str;
		1524
		1525	// inside a string
		1526	case INCR_M_STR:
		1527	incr_m_str:
		1528	for (;;)
		1529	{
		1530	if (*p == '"')
		1531	{
		1532	++p;
		1533	self->incr_mode = INCR_M_JSON;
		1534
		1535	if (!self->incr_nest)
		1536	goto interrupt;
		1537
		1538	goto incr_m_json;
		1539	}
		1540	else if (*p == '\\')
		1541	{
		1542	++p; // "virtually" consumes character after \
		1543
		1544	if (!*p) // if at end of string we have to switch modes
		1545	{
		1546	self->incr_mode = INCR_M_BS;
		1547	goto interrupt;
		1548	}
		1549	}
		1550	else if (!*p)
		1551	goto interrupt;
		1552
		1553	++p;
		1554	}
		1555
		1556	// after initial ws, outside string
		1557	case INCR_M_JSON:
		1558	incr_m_json:
		1559	for (;;)
		1560	{
		1561	switch (*p++)
		1562	{
		1563	case 0:
		1564	--p;
		1565	goto interrupt;
		1566
		1567	case 0x09:
		1568	case 0x0a:
		1569	case 0x0d:
		1570	case 0x20:
		1571	if (!self->incr_nest)
		1572	{
		1573	--p; // do not eat the whitespace, let the next round do it
		1574	goto interrupt;
		1575	}
		1576	break;
		1577
		1578	case '"':
		1579	self->incr_mode = INCR_M_STR;
		1580	goto incr_m_str;
		1581
		1582	case '[':
		1583	case '{':
		1584	++self->incr_nest;
		1585	break;
		1586
		1587	case ']':
		1588	case '}':
		1589	if (!--self->incr_nest)
		1590	goto interrupt;
		1591	}
		1592	}
		1593	}
		1594
		1595	modechange:
		1596	;
		1597	}
		1598
		1599	interrupt:
		1600	self->incr_pos = p - SvPVX (self->incr_text);
		1601	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1602	}
		1603
		1604	/////////////////////////////////////////////////////////////////////////////
1246	// XS interface functions	1605	// XS interface functions
1247		1606
1248	MODULE = JSON::XS PACKAGE = JSON::XS	1607	MODULE = JSON::XS PACKAGE = JSON::XS
1249		1608
1250	BOOT:	1609	BOOT:
…		…
1259	: -1;	1618	: -1;
1260		1619
1261	json_stash = gv_stashpv ("JSON::XS" , 1);	1620	json_stash = gv_stashpv ("JSON::XS" , 1);
1262	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1621	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1263		1622
1264	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1623	json_true = get_bool ("JSON::XS::true");
1265	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1624	json_false = get_bool ("JSON::XS::false");
1266	}	1625	}
1267		1626
1268	PROTOTYPES: DISABLE	1627	PROTOTYPES: DISABLE
		1628
		1629	void CLONE (...)
		1630	CODE:
		1631	json_stash = 0;
		1632	json_boolean_stash = 0;
1269		1633
1270	void new (char *klass)	1634	void new (char *klass)
1271	PPCODE:	1635	PPCODE:
1272	{	1636	{
1273	SV *pv = NEWSV (0, sizeof (JSON));	1637	SV *pv = NEWSV (0, sizeof (JSON));
1274	SvPOK_only (pv);	1638	SvPOK_only (pv);
1275	Zero (SvPVX (pv), 1, sizeof (JSON));	1639	Zero (SvPVX (pv), 1, JSON);
1276	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1640	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1277	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1641	XPUSHs (sv_2mortal (sv_bless (
		1642	newRV_noinc (pv),
		1643	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1644	)));
1278	}	1645	}
1279		1646
1280	void ascii (JSON *self, int enable = 1)	1647	void ascii (JSON *self, int enable = 1)
1281	ALIAS:	1648	ALIAS:
1282	ascii = F_ASCII	1649	ascii = F_ASCII
…		…
1289	pretty = F_PRETTY	1656	pretty = F_PRETTY
1290	allow_nonref = F_ALLOW_NONREF	1657	allow_nonref = F_ALLOW_NONREF
1291	shrink = F_SHRINK	1658	shrink = F_SHRINK
1292	allow_blessed = F_ALLOW_BLESSED	1659	allow_blessed = F_ALLOW_BLESSED
1293	convert_blessed = F_CONV_BLESSED	1660	convert_blessed = F_CONV_BLESSED
		1661	relaxed = F_RELAXED
1294	PPCODE:	1662	PPCODE:
1295	{	1663	{
1296	if (enable)	1664	if (enable)
1297	self->flags \|= ix;	1665	self->flags \|= ix;
1298	else	1666	else
1299	self->flags &= ~ix;	1667	self->flags &= ~ix;
1300		1668
1301	XPUSHs (ST (0));	1669	XPUSHs (ST (0));
1302	}	1670	}
1303		1671
		1672	void get_ascii (JSON *self)
		1673	ALIAS:
		1674	get_ascii = F_ASCII
		1675	get_latin1 = F_LATIN1
		1676	get_utf8 = F_UTF8
		1677	get_indent = F_INDENT
		1678	get_canonical = F_CANONICAL
		1679	get_space_before = F_SPACE_BEFORE
		1680	get_space_after = F_SPACE_AFTER
		1681	get_allow_nonref = F_ALLOW_NONREF
		1682	get_shrink = F_SHRINK
		1683	get_allow_blessed = F_ALLOW_BLESSED
		1684	get_convert_blessed = F_CONV_BLESSED
		1685	get_relaxed = F_RELAXED
		1686	PPCODE:
		1687	XPUSHs (boolSV (self->flags & ix));
		1688
1304	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1689	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1305	PPCODE:	1690	PPCODE:
1306	{	1691	{
1307	UV log2 = 0;	1692	UV log2 = 0;
1308		1693
…		…
1314	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1699	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1315		1700
1316	XPUSHs (ST (0));	1701	XPUSHs (ST (0));
1317	}	1702	}
1318		1703
		1704	U32 get_max_depth (JSON *self)
		1705	CODE:
		1706	RETVAL = DEC_DEPTH (self->flags);
		1707	OUTPUT:
		1708	RETVAL
		1709
1319	void max_size (JSON *self, UV max_size = 0)	1710	void max_size (JSON *self, UV max_size = 0)
1320	PPCODE:	1711	PPCODE:
1321	{	1712	{
1322	UV log2 = 0;	1713	UV log2 = 0;
1323		1714
…		…
1326		1717
1327	while ((1UL << log2) < max_size)	1718	while ((1UL << log2) < max_size)
1328	++log2;	1719	++log2;
1329		1720
1330	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1721	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1722
		1723	XPUSHs (ST (0));
		1724	}
		1725
		1726	int get_max_size (JSON *self)
		1727	CODE:
		1728	RETVAL = DEC_SIZE (self->flags);
		1729	OUTPUT:
		1730	RETVAL
		1731
		1732	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1733	PPCODE:
		1734	{
		1735	SvREFCNT_dec (self->cb_object);
		1736	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1737
		1738	XPUSHs (ST (0));
		1739	}
		1740
		1741	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1742	PPCODE:
		1743	{
		1744	if (!self->cb_sk_object)
		1745	self->cb_sk_object = newHV ();
		1746
		1747	if (SvOK (cb))
		1748	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1749	else
		1750	{
		1751	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1752
		1753	if (!HvKEYS (self->cb_sk_object))
		1754	{
		1755	SvREFCNT_dec (self->cb_sk_object);
		1756	self->cb_sk_object = 0;
		1757	}
		1758	}
1331		1759
1332	XPUSHs (ST (0));	1760	XPUSHs (ST (0));
1333	}	1761	}
1334		1762
1335	void encode (JSON self, SV scalar)	1763	void encode (JSON self, SV scalar)
…		…
1341	XPUSHs (decode_json (jsonstr, self, 0));	1769	XPUSHs (decode_json (jsonstr, self, 0));
1342		1770
1343	void decode_prefix (JSON self, SV jsonstr)	1771	void decode_prefix (JSON self, SV jsonstr)
1344	PPCODE:	1772	PPCODE:
1345	{	1773	{
1346	UV offset;	1774	STRLEN offset;
1347	EXTEND (SP, 2);	1775	EXTEND (SP, 2);
1348	PUSHs (decode_json (jsonstr, self, &offset));	1776	PUSHs (decode_json (jsonstr, self, &offset));
1349	PUSHs (sv_2mortal (newSVuv (offset)));	1777	PUSHs (sv_2mortal (newSVuv (offset)));
1350	}	1778	}
1351		1779
		1780	void incr_parse (JSON self, SV jsonstr = 0)
		1781	PPCODE:
		1782	{
		1783	if (!self->incr_text)
		1784	self->incr_text = newSVpvn ("", 0);
		1785
		1786	// append data, if any
		1787	if (jsonstr)
		1788	{
		1789	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1790	{
		1791	/* utf-8-ness differs, need to upgrade */
		1792	sv_utf8_upgrade (self->incr_text);
		1793
		1794	if (self->incr_pos)
		1795	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1796	- (U8 *)SvPVX (self->incr_text);
		1797	}
		1798
		1799	{
		1800	STRLEN len;
		1801	const char *str = SvPV (jsonstr, len);
		1802	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1803	Move (str, SvEND (self->incr_text), len, char);
		1804	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1805	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1806	}
		1807	}
		1808
		1809	if (GIMME_V != G_VOID)
		1810	do
		1811	{
		1812	STRLEN offset;
		1813
		1814	if (!INCR_DONE (self))
		1815	{
		1816	incr_parse (self);
		1817	if (!INCR_DONE (self))
		1818	break;
		1819	}
		1820
		1821	XPUSHs (decode_json (self->incr_text, self, &offset));
		1822
		1823	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1824	self->incr_pos -= offset;
		1825	self->incr_nest = 0;
		1826	self->incr_mode = 0;
		1827	}
		1828	while (GIMME_V == G_ARRAY);
		1829	}
		1830
		1831	SV incr_text (JSON self)
		1832	ATTRS: lvalue
		1833	CODE:
		1834	{
		1835	if (self->incr_pos)
		1836	croak ("incr_text can not be called when the incremental parser already started parsing");
		1837
		1838	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1839	}
		1840	OUTPUT:
		1841	RETVAL
		1842
		1843	void incr_skip (JSON *self)
		1844	CODE:
		1845	{
		1846	if (self->incr_pos)
		1847	{
		1848	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1849	self->incr_pos = 0;
		1850	self->incr_nest = 0;
		1851	self->incr_mode = 0;
		1852	}
		1853	}
		1854
		1855	void DESTROY (JSON *self)
		1856	CODE:
		1857	SvREFCNT_dec (self->cb_sk_object);
		1858	SvREFCNT_dec (self->cb_object);
		1859	SvREFCNT_dec (self->incr_text);
		1860
1352	PROTOTYPES: ENABLE	1861	PROTOTYPES: ENABLE
1353		1862
1354	void to_json (SV *scalar)	1863	void encode_json (SV *scalar)
		1864	ALIAS:
		1865	to_json_ = 0
		1866	encode_json = F_UTF8
1355	PPCODE:	1867	PPCODE:
1356	{	1868	{
1357	JSON json = { F_DEFAULT \| F_UTF8 };	1869	JSON json = { F_DEFAULT \| ix };
1358	XPUSHs (encode_json (scalar, &json));	1870	XPUSHs (encode_json (scalar, &json));
1359	}	1871	}
1360		1872
1361	void from_json (SV *jsonstr)	1873	void decode_json (SV *jsonstr)
		1874	ALIAS:
		1875	from_json_ = 0
		1876	decode_json = F_UTF8
1362	PPCODE:	1877	PPCODE:
1363	{	1878	{
1364	JSON json = { F_DEFAULT \| F_UTF8 };	1879	JSON json = { F_DEFAULT \| ix };
1365	XPUSHs (decode_json (jsonstr, &json, 0));	1880	XPUSHs (decode_json (jsonstr, &json, 0));
1366	}	1881	}
1367		1882
		1883

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs. Revision 1.83 by root, Wed Mar 26 22:54:38 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs.
Revision 1.83 by root, Wed Mar 26 22:54:38 2008 UTC