[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 1.84 by root, Thu Mar 27 06:37:35 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	#define F_ALLOW_UNKNOWN 0x00002000UL
		37
31	#define F_MAXDEPTH 0xf8000000UL	38	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27	39	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	40	#define F_MAXSIZE 0x01f00000UL
34	#define S_MAXSIZE 20	41	#define S_MAXSIZE 20
		42	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
35		43
36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	44	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))	45	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
38		46
39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	47	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
…		…
46		54
47	#define SB do {	55	#define SB do {
48	#define SE } while (0)	56	#define SE } while (0)
49		57
50	#if __GNUC__ >= 3	58	#if __GNUC__ >= 3
51	# define expect(expr,value) __builtin_expect ((expr),(value))	59	# define expect(expr,value) __builtin_expect ((expr), (value))
52	# define inline inline	60	# define INLINE static inline
53	#else	61	#else
54	# define expect(expr,value) (expr)	62	# define expect(expr,value) (expr)
55	# define inline static	63	# define INLINE static
56	#endif	64	#endif
57		65
58	#define expect_false(expr) expect ((expr) != 0, 0)	66	#define expect_false(expr) expect ((expr) != 0, 0)
59	#define expect_true(expr) expect ((expr) != 0, 1)	67	#define expect_true(expr) expect ((expr) != 0, 1)
60		68
		69	#define IN_RANGE_INC(type,val,beg,end) \
		70	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		71	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		72
		73	#ifdef USE_ITHREADS
		74	# define JSON_SLOW 1
		75	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		76	#else
		77	# define JSON_SLOW 0
		78	# define JSON_STASH json_stash
		79	#endif
		80
61	static HV json_stash, json_boolean_stash; // JSON::XS::	81	static HV json_stash, json_boolean_stash; // JSON::XS::
62	static SV json_true, json_false;	82	static SV json_true, json_false;
63		83
		84	enum {
		85	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		86	INCR_M_STR, // inside string
		87	INCR_M_BS, // inside backslash
		88	INCR_M_JSON // outside anything, count nesting
		89	};
		90
		91	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		92
64	typedef struct {	93	typedef struct {
65	U32 flags;	94	U32 flags;
		95	SV *cb_object;
		96	HV *cb_sk_object;
		97
		98	// for the incremental parser
		99	SV *incr_text; // the source text so far
		100	STRLEN incr_pos; // the current offset into the text
		101	int incr_nest; // {[]}-nesting level
		102	int incr_mode;
66	} JSON;	103	} JSON;
67		104
68	/////////////////////////////////////////////////////////////////////////////	105	/////////////////////////////////////////////////////////////////////////////
69	// utility functions	106	// utility functions
70		107
71	inline void	108	INLINE SV *
		109	get_bool (const char *name)
		110	{
		111	SV *sv = get_sv (name, 1);
		112
		113	SvREADONLY_on (sv);
		114	SvREADONLY_on (SvRV (sv));
		115
		116	return sv;
		117	}
		118
		119	INLINE void
72	shrink (SV *sv)	120	shrink (SV *sv)
73	{	121	{
74	sv_utf8_downgrade (sv, 1);	122	sv_utf8_downgrade (sv, 1);
75	if (SvLEN (sv) > SvCUR (sv) + 1)	123	if (SvLEN (sv) > SvCUR (sv) + 1)
76	{	124	{
…		…
85	// decode an utf-8 character and return it, or (UV)-1 in	133	// decode an utf-8 character and return it, or (UV)-1 in
86	// case of an error.	134	// case of an error.
87	// we special-case "safe" characters from U+80 .. U+7FF,	135	// we special-case "safe" characters from U+80 .. U+7FF,
88	// but use the very good perl function to parse anything else.	136	// but use the very good perl function to parse anything else.
89	// note that we never call this function for a ascii codepoints	137	// note that we never call this function for a ascii codepoints
90	inline UV	138	INLINE UV
91	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	139	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
92	{	140	{
93	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	141	if (expect_true (len >= 2
94	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	142	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
95	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	143	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
96	{	144	{
97	*clen = 2;	145	*clen = 2;
98	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	146	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
99	}	147	}
100	else	148	else
101	{	149	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
102	*clen = (STRLEN)-1;	150	}
103	return (UV)-1;	151
104	}	152	// likewise for encoding, also never called for ascii codepoints
		153	// this function takes advantage of this fact, although current gccs
		154	// seem to optimise the check for >= 0x80 away anyways
		155	INLINE unsigned char *
		156	encode_utf8 (unsigned char *s, UV ch)
		157	{
		158	if (expect_false (ch < 0x000080))
		159	*s++ = ch;
		160	else if (expect_true (ch < 0x000800))
		161	*s++ = 0xc0 \| ( ch >> 6),
		162	*s++ = 0x80 \| ( ch & 0x3f);
		163	else if ( ch < 0x010000)
		164	*s++ = 0xe0 \| ( ch >> 12),
		165	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		166	*s++ = 0x80 \| ( ch & 0x3f);
		167	else if ( ch < 0x110000)
		168	*s++ = 0xf0 \| ( ch >> 18),
		169	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		170	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		171	*s++ = 0x80 \| ( ch & 0x3f);
		172
		173	return s;
105	}	174	}
106		175
107	/////////////////////////////////////////////////////////////////////////////	176	/////////////////////////////////////////////////////////////////////////////
108	// encoder	177	// encoder
109		178
…		…
114	char *end; // SvEND (sv)	183	char *end; // SvEND (sv)
115	SV *sv; // result scalar	184	SV *sv; // result scalar
116	JSON json;	185	JSON json;
117	U32 indent; // indentation level	186	U32 indent; // indentation level
118	U32 maxdepth; // max. indentation/recursion level	187	U32 maxdepth; // max. indentation/recursion level
		188	UV limit; // escape character values >= this value when encoding
119	} enc_t;	189	} enc_t;
120		190
121	inline void	191	INLINE void
122	need (enc_t *enc, STRLEN len)	192	need (enc_t *enc, STRLEN len)
123	{	193	{
124	if (expect_false (enc->cur + len >= enc->end))	194	if (expect_false (enc->cur + len >= enc->end))
125	{	195	{
126	STRLEN cur = enc->cur - SvPVX (enc->sv);	196	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
128	enc->cur = SvPVX (enc->sv) + cur;	198	enc->cur = SvPVX (enc->sv) + cur;
129	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	199	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
130	}	200	}
131	}	201	}
132		202
133	inline void	203	INLINE void
134	encode_ch (enc_t *enc, char ch)	204	encode_ch (enc_t *enc, char ch)
135	{	205	{
136	need (enc, 1);	206	need (enc, 1);
137	*enc->cur++ = ch;	207	*enc->cur++ = ch;
138	}	208	}
…		…
192	{	262	{
193	uch = ch;	263	uch = ch;
194	clen = 1;	264	clen = 1;
195	}	265	}
196		266
197	if (uch > 0x10FFFFUL)	267	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	{	268	{
202	if (uch > 0xFFFFUL)	269	if (uch >= 0x10000UL)
203	{	270	{
		271	if (uch >= 0x110000UL)
		272	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		273
204	need (enc, len += 11);	274	need (enc, len += 11);
205	sprintf (enc->cur, "\\u%04x\\u%04x",	275	sprintf (enc->cur, "\\u%04x\\u%04x",
206	(int)((uch - 0x10000) / 0x400 + 0xD800),	276	(int)((uch - 0x10000) / 0x400 + 0xD800),
207	(int)((uch - 0x10000) % 0x400 + 0xDC00));	277	(int)((uch - 0x10000) % 0x400 + 0xDC00));
208	enc->cur += 12;	278	enc->cur += 12;
…		…
236	while (--clen);	306	while (--clen);
237	}	307	}
238	else	308	else
239	{	309	{
240	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	310	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
241	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	311	enc->cur = encode_utf8 (enc->cur, uch);
242	++str;	312	++str;
243	}	313	}
244	}	314	}
245	}	315	}
246	}	316	}
247		317
248	--len;	318	--len;
249	}	319	}
250	}	320	}
251		321
252	inline void	322	INLINE void
253	encode_indent (enc_t *enc)	323	encode_indent (enc_t *enc)
254	{	324	{
255	if (enc->json.flags & F_INDENT)	325	if (enc->json.flags & F_INDENT)
256	{	326	{
257	int spaces = enc->indent * INDENT_STEP;	327	int spaces = enc->indent * INDENT_STEP;
…		…
260	memset (enc->cur, ' ', spaces);	330	memset (enc->cur, ' ', spaces);
261	enc->cur += spaces;	331	enc->cur += spaces;
262	}	332	}
263	}	333	}
264		334
265	inline void	335	INLINE void
266	encode_space (enc_t *enc)	336	encode_space (enc_t *enc)
267	{	337	{
268	need (enc, 1);	338	need (enc, 1);
269	encode_ch (enc, ' ');	339	encode_ch (enc, ' ');
270	}	340	}
271		341
272	inline void	342	INLINE void
273	encode_nl (enc_t *enc)	343	encode_nl (enc_t *enc)
274	{	344	{
275	if (enc->json.flags & F_INDENT)	345	if (enc->json.flags & F_INDENT)
276	{	346	{
277	need (enc, 1);	347	need (enc, 1);
278	encode_ch (enc, '\n');	348	encode_ch (enc, '\n');
279	}	349	}
280	}	350	}
281		351
282	inline void	352	INLINE void
283	encode_comma (enc_t *enc)	353	encode_comma (enc_t *enc)
284	{	354	{
285	encode_ch (enc, ',');	355	encode_ch (enc, ',');
286		356
287	if (enc->json.flags & F_INDENT)	357	if (enc->json.flags & F_INDENT)
…		…
298	int i, len = av_len (av);	368	int i, len = av_len (av);
299		369
300	if (enc->indent >= enc->maxdepth)	370	if (enc->indent >= enc->maxdepth)
301	croak ("data structure too deep (hit recursion limit)");	371	croak ("data structure too deep (hit recursion limit)");
302		372
303	encode_ch (enc, '['); encode_nl (enc);	373	encode_ch (enc, '[');
304	++enc->indent;	374
		375	if (len >= 0)
		376	{
		377	encode_nl (enc); ++enc->indent;
305		378
306	for (i = 0; i <= len; ++i)	379	for (i = 0; i <= len; ++i)
307	{	380	{
		381	SV **svp = av_fetch (av, i, 0);
		382
308	encode_indent (enc);	383	encode_indent (enc);
309	encode_sv (enc, *av_fetch (av, i, 0));
310		384
		385	if (svp)
		386	encode_sv (enc, *svp);
		387	else
		388	encode_str (enc, "null", 4, 0);
		389
311	if (i < len)	390	if (i < len)
312	encode_comma (enc);	391	encode_comma (enc);
313	}	392	}
314		393
		394	encode_nl (enc); --enc->indent; encode_indent (enc);
		395	}
		396
315	encode_nl (enc);	397	encode_ch (enc, ']');
316
317	--enc->indent;
318	encode_indent (enc); encode_ch (enc, ']');
319	}	398	}
320		399
321	static void	400	static void
322	encode_he (enc_t enc, HE he)	401	encode_hk (enc_t enc, HE he)
323	{	402	{
324	encode_ch (enc, '"');	403	encode_ch (enc, '"');
325		404
326	if (HeKLEN (he) == HEf_SVKEY)	405	if (HeKLEN (he) == HEf_SVKEY)
327	{	406	{
…		…
340	encode_ch (enc, '"');	419	encode_ch (enc, '"');
341		420
342	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	421	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
343	encode_ch (enc, ':');	422	encode_ch (enc, ':');
344	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	423	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
345	encode_sv (enc, HeVAL (he));
346	}	424	}
347		425
348	// compare hash entries, used when all keys are bytestrings	426	// compare hash entries, used when all keys are bytestrings
349	static int	427	static int
350	he_cmp_fast (const void a_, const void b_)	428	he_cmp_fast (const void a_, const void b_)
…		…
355	HE b = (HE **)b_;	433	HE b = (HE **)b_;
356		434
357	STRLEN la = HeKLEN (a);	435	STRLEN la = HeKLEN (a);
358	STRLEN lb = HeKLEN (b);	436	STRLEN lb = HeKLEN (b);
359		437
360	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	438	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
361	cmp = la - lb;	439	cmp = lb - la;
362		440
363	return cmp;	441	return cmp;
364	}	442	}
365		443
366	// compare hash entries, used when some keys are sv's or utf-x	444	// compare hash entries, used when some keys are sv's or utf-x
367	static int	445	static int
368	he_cmp_slow (const void a, const void b)	446	he_cmp_slow (const void a, const void b)
369	{	447	{
370	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	448	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
371	}	449	}
372		450
373	static void	451	static void
374	encode_hv (enc_t enc, HV hv)	452	encode_hv (enc_t enc, HV hv)
375	{	453	{
376	int count, i;	454	HE *he;
377		455
378	if (enc->indent >= enc->maxdepth)	456	if (enc->indent >= enc->maxdepth)
379	croak ("data structure too deep (hit recursion limit)");	457	croak ("data structure too deep (hit recursion limit)");
380		458
381	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	459	encode_ch (enc, '{');
382		460
383	if ((count = hv_iterinit (hv)))
384	{
385	// for canonical output we have to sort by keys first	461	// for canonical output we have to sort by keys first
386	// actually, this is mostly due to the stupid so-called	462	// actually, this is mostly due to the stupid so-called
387	// security workaround added somewhere in 5.8.x.	463	// security workaround added somewhere in 5.8.x.
388	// that randomises hash orderings	464	// that randomises hash orderings
389	if (enc->json.flags & F_CANONICAL)	465	if (enc->json.flags & F_CANONICAL)
		466	{
		467	int count = hv_iterinit (hv);
		468
		469	if (SvMAGICAL (hv))
390	{	470	{
		471	// need to count by iterating. could improve by dynamically building the vector below
		472	// but I don't care for the speed of this special case.
		473	// note also that we will run into undefined behaviour when the two iterations
		474	// do not result in the same count, something I might care for in some later release.
		475
		476	count = 0;
		477	while (hv_iternext (hv))
		478	++count;
		479
		480	hv_iterinit (hv);
		481	}
		482
		483	if (count)
		484	{
391	int fast = 1;	485	int i, fast = 1;
392	HE *he;
393	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	486	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
394	HE *hes = _alloca (count sizeof (HE));	487	HE *hes = _alloca (count sizeof (HE));
395	#else	488	#else
396	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	489	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
397	#endif	490	#endif
…		…
424		517
425	FREETMPS;	518	FREETMPS;
426	LEAVE;	519	LEAVE;
427	}	520	}
428		521
429	for (i = 0; i < count; ++i)	522	encode_nl (enc); ++enc->indent;
		523
		524	while (count--)
430	{	525	{
431	encode_indent (enc);	526	encode_indent (enc);
		527	he = hes [count];
432	encode_he (enc, hes [i]);	528	encode_hk (enc, he);
		529	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
433		530
434	if (i < count - 1)	531	if (count)
435	encode_comma (enc);	532	encode_comma (enc);
436	}	533	}
437		534
438	encode_nl (enc);	535	encode_nl (enc); --enc->indent; encode_indent (enc);
439	}	536	}
		537	}
440	else	538	else
		539	{
		540	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		541	if ((he = hv_iternext (hv)))
441	{	542	{
442	HE *he = hv_iternext (hv);	543	encode_nl (enc); ++enc->indent;
443		544
444	for (;;)	545	for (;;)
445	{	546	{
446	encode_indent (enc);	547	encode_indent (enc);
447	encode_he (enc, he);	548	encode_hk (enc, he);
		549	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
448		550
449	if (!(he = hv_iternext (hv)))	551	if (!(he = hv_iternext (hv)))
450	break;	552	break;
451		553
452	encode_comma (enc);	554	encode_comma (enc);
453	}	555	}
454		556
455	encode_nl (enc);	557	encode_nl (enc); --enc->indent; encode_indent (enc);
456	}	558	}
457	}	559	}
458		560
459	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	561	encode_ch (enc, '}');
460	}	562	}
461		563
462	// encode objects, arrays and special \0=false and \1=true values.	564	// encode objects, arrays and special \0=false and \1=true values.
463	static void	565	static void
464	encode_rv (enc_t enc, SV sv)	566	encode_rv (enc_t enc, SV sv)
…		…
468	SvGETMAGIC (sv);	570	SvGETMAGIC (sv);
469	svt = SvTYPE (sv);	571	svt = SvTYPE (sv);
470		572
471	if (expect_false (SvOBJECT (sv)))	573	if (expect_false (SvOBJECT (sv)))
472	{	574	{
		575	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		576	? json_boolean_stash
		577	: gv_stashpv ("JSON::XS::Boolean", 1);
		578
473	if (SvSTASH (sv) == json_boolean_stash)	579	if (SvSTASH (sv) == stash)
474	{	580	{
475	if (SvIV (sv) == 0)	581	if (SvIV (sv))
		582	encode_str (enc, "true", 4, 0);
		583	else
476	encode_str (enc, "false", 5, 0);	584	encode_str (enc, "false", 5, 0);
477	else
478	encode_str (enc, "true", 4, 0);
479	}	585	}
480	else	586	else
481	{	587	{
482	#if 0	588	#if 0
483	if (0 && sv_derived_from (rv, "JSON::Literal"))	589	if (0 && sv_derived_from (rv, "JSON::Literal"))
…		…
486	}	592	}
487	#endif	593	#endif
488	if (enc->json.flags & F_CONV_BLESSED)	594	if (enc->json.flags & F_CONV_BLESSED)
489	{	595	{
490	// we re-bless the reference to get overload and other niceties right	596	// we re-bless the reference to get overload and other niceties right
491	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	597	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
492		598
493	if (to_json)	599	if (to_json)
494	{	600	{
495	dSP;	601	dSP;
496	ENTER;	602
497	SAVETMPS;
498	PUSHMARK (SP);	603	ENTER; SAVETMPS; PUSHMARK (SP);
499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	604	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
500		605
501	// calling with G_SCALAR ensures that we always get a 1 reutrn value	606	// calling with G_SCALAR ensures that we always get a 1 return value
502	// check anyways.
503	PUTBACK;	607	PUTBACK;
504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	608	call_sv ((SV *)GvCV (to_json), G_SCALAR);
505	SPAGAIN;	609	SPAGAIN;
506		610
		611	// catch this surprisingly common error
		612	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		613	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		614
		615	sv = POPs;
		616	PUTBACK;
		617
507	encode_sv (enc, POPs);	618	encode_sv (enc, sv);
508		619
509	FREETMPS;	620	FREETMPS; LEAVE;
510	LEAVE;
511	}	621	}
512	else if (enc->json.flags & F_ALLOW_BLESSED)	622	else if (enc->json.flags & F_ALLOW_BLESSED)
513	encode_str (enc, "null", 4, 0);	623	encode_str (enc, "null", 4, 0);
514	else	624	else
515	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",	625	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
…		…
526	encode_hv (enc, (HV *)sv);	636	encode_hv (enc, (HV *)sv);
527	else if (svt == SVt_PVAV)	637	else if (svt == SVt_PVAV)
528	encode_av (enc, (AV *)sv);	638	encode_av (enc, (AV *)sv);
529	else if (svt < SVt_PVAV)	639	else if (svt < SVt_PVAV)
530	{	640	{
531	if (SvNIOK (sv) && SvIV (sv) == 0)	641	STRLEN len = 0;
		642	char *pv = svt ? SvPV (sv, len) : 0;
		643
		644	if (len == 1 && *pv == '1')
		645	encode_str (enc, "true", 4, 0);
		646	else if (len == 1 && *pv == '0')
532	encode_str (enc, "false", 5, 0);	647	encode_str (enc, "false", 5, 0);
533	else if (SvNIOK (sv) && SvIV (sv) == 1)	648	else if (enc->json.flags & F_ALLOW_UNKNOWN)
534	encode_str (enc, "true", 4, 0);	649	encode_str (enc, "null", 4, 0);
535	else	650	else
536	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	651	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	652	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
538	}	653	}
		654	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		655	encode_str (enc, "null", 4, 0);
539	else	656	else
540	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	657	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
541	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	658	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
542	}	659	}
543		660
…		…
561	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	678	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
562	enc->cur += strlen (enc->cur);	679	enc->cur += strlen (enc->cur);
563	}	680	}
564	else if (SvIOKp (sv))	681	else if (SvIOKp (sv))
565	{	682	{
566	// we assume we can always read an IV as a UV	683	// we assume we can always read an IV as a UV and vice versa
567	if (SvUV (sv) & ~(UV)0x7fff)	684	// we assume two's complement
568	{	685	// we assume no aliasing issues in the union
569	// large integer, use the (rather slow) snprintf way.	686	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
570	need (enc, sizeof (UV) * 3);	687	: 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	{	688	{
578	// optimise the "small number case"	689	// optimise the "small number case"
579	// code will likely be branchless and use only a single multiplication	690	// code will likely be branchless and use only a single multiplication
		691	// works for numbers up to 59074
580	I32 i = SvIV (sv);	692	I32 i = SvIVX (sv);
581	U32 u;	693	U32 u;
582	char digit, nz = 0;	694	char digit, nz = 0;
583		695
584	need (enc, 6);	696	need (enc, 6);
585		697
…		…
591		703
592	// now output digit by digit, each time masking out the integer part	704	// 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,	705	// and multiplying by 5 while moving the decimal point one to the right,
594	// resulting in a net multiplication by 10.	706	// resulting in a net multiplication by 10.
595	// we always write the digit to memory but conditionally increment	707	// we always write the digit to memory but conditionally increment
596	// the pointer, to ease the usage of conditional move instructions.	708	// 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;	709	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;	710	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;	711	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;	712	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'	713	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
602	}	714	}
		715	else
		716	{
		717	// large integer, use the (rather slow) snprintf way.
		718	need (enc, IVUV_MAXCHARS);
		719	enc->cur +=
		720	SvIsUV(sv)
		721	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		722	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		723	}
603	}	724	}
604	else if (SvROK (sv))	725	else if (SvROK (sv))
605	encode_rv (enc, SvRV (sv));	726	encode_rv (enc, SvRV (sv));
606	else if (!SvOK (sv))	727	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
607	encode_str (enc, "null", 4, 0);	728	encode_str (enc, "null", 4, 0);
608	else	729	else
609	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	730	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
610	SvPV_nolen (sv), SvFLAGS (sv));	731	SvPV_nolen (sv), SvFLAGS (sv));
611	}	732	}
…		…
622	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	743	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
623	enc.cur = SvPVX (enc.sv);	744	enc.cur = SvPVX (enc.sv);
624	enc.end = SvEND (enc.sv);	745	enc.end = SvEND (enc.sv);
625	enc.indent = 0;	746	enc.indent = 0;
626	enc.maxdepth = DEC_DEPTH (enc.json.flags);	747	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		748	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		749	: enc.json.flags & F_LATIN1 ? 0x000100UL
		750	: 0x110000UL;
627		751
628	SvPOK_only (enc.sv);	752	SvPOK_only (enc.sv);
629	encode_sv (&enc, scalar);	753	encode_sv (&enc, scalar);
630		754
631	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	755	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
652	JSON json;	776	JSON json;
653	U32 depth; // recursion depth	777	U32 depth; // recursion depth
654	U32 maxdepth; // recursion depth limit	778	U32 maxdepth; // recursion depth limit
655	} dec_t;	779	} dec_t;
656		780
657	inline void	781	INLINE void
		782	decode_comment (dec_t *dec)
		783	{
		784	// only '#'-style comments allowed a.t.m.
		785
		786	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		787	++dec->cur;
		788	}
		789
		790	INLINE void
658	decode_ws (dec_t *dec)	791	decode_ws (dec_t *dec)
659	{	792	{
660	for (;;)	793	for (;;)
661	{	794	{
662	char ch = *dec->cur;	795	char ch = *dec->cur;
663		796
664	if (ch > 0x20	797	if (ch > 0x20)
		798	{
		799	if (expect_false (ch == '#'))
		800	{
		801	if (dec->json.flags & F_RELAXED)
		802	decode_comment (dec);
		803	else
		804	break;
		805	}
		806	else
		807	break;
		808	}
665	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	809	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
666	break;	810	break; // parse error, but let higher level handle it, gives better error messages
667		811
668	++dec->cur;	812	++dec->cur;
669	}	813	}
670	}	814	}
671		815
…		…
777		921
778	if (hi >= 0x80)	922	if (hi >= 0x80)
779	{	923	{
780	utf8 = 1;	924	utf8 = 1;
781		925
782	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	926	cur = encode_utf8 (cur, hi);
783	}	927	}
784	else	928	else
785	*cur++ = hi;	929	*cur++ = hi;
786	}	930	}
787	break;	931	break;
…		…
789	default:	933	default:
790	--dec_cur;	934	--dec_cur;
791	ERR ("illegal backslash escape sequence in string");	935	ERR ("illegal backslash escape sequence in string");
792	}	936	}
793	}	937	}
794	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	938	else if (expect_true (ch >= 0x20 && ch < 0x80))
795	*cur++ = ch;	939	*cur++ = ch;
796	else if (ch >= 0x80)	940	else if (ch >= 0x80)
797	{	941	{
798	STRLEN clen;	942	STRLEN clen;
799	UV uch;	943	UV uch;
…		…
920	is_nv = 1;	1064	is_nv = 1;
921	}	1065	}
922		1066
923	if (!is_nv)	1067	if (!is_nv)
924	{	1068	{
		1069	int len = dec->cur - start;
		1070
925	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1071	// special case the rather common 1..5-digit-int case
926	if (*start == '-')	1072	if (*start == '-')
927	switch (dec->cur - start)	1073	switch (len)
928	{	1074	{
929	case 2: return newSViv (-( start [1] - '0' * 1));	1075	case 2: return newSViv (-( start [1] - '0' * 1));
930	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1076	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));	1077	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));	1078	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1079	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
933	}	1080	}
934	else	1081	else
935	switch (dec->cur - start)	1082	switch (len)
936	{	1083	{
937	case 1: return newSViv ( start [0] - '0' * 1);	1084	case 1: return newSViv ( start [0] - '0' * 1);
938	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1085	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);	1086	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);	1087	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1088	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
941	}	1089	}
942		1090
943	{	1091	{
944	UV uv;	1092	UV uv;
945	int numtype = grok_number (start, dec->cur - start, &uv);	1093	int numtype = grok_number (start, len, &uv);
946	if (numtype & IS_NUMBER_IN_UV)	1094	if (numtype & IS_NUMBER_IN_UV)
947	if (numtype & IS_NUMBER_NEG)	1095	if (numtype & IS_NUMBER_NEG)
948	{	1096	{
949	if (uv < (UV)IV_MIN)	1097	if (uv < (UV)IV_MIN)
950	return newSViv (-(IV)uv);	1098	return newSViv (-(IV)uv);
951	}	1099	}
952	else	1100	else
953	return newSVuv (uv);	1101	return newSVuv (uv);
954
955	// here would likely be the place for bigint support
956	}	1102	}
957	}
958		1103
959	// if we ever support bigint or bigfloat, this is the place for bigfloat	1104	len -= *start == '-' ? 1 : 0;
		1105
		1106	// does not fit into IV or UV, try NV
		1107	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1108	#if defined (LDBL_DIG)
		1109	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1110	#endif
		1111	)
		1112	// fits into NV without loss of precision
		1113	return newSVnv (Atof (start));
		1114
		1115	// everything else fails, convert it to a string
		1116	return newSVpvn (start, dec->cur - start);
		1117	}
		1118
		1119	// loss of precision here
960	return newSVnv (Atof (start));	1120	return newSVnv (Atof (start));
961		1121
962	fail:	1122	fail:
963	return 0;	1123	return 0;
964	}	1124	}
…		…
994		1154
995	if (*dec->cur != ',')	1155	if (*dec->cur != ',')
996	ERR (", or ] expected while parsing array");	1156	ERR (", or ] expected while parsing array");
997		1157
998	++dec->cur;	1158	++dec->cur;
		1159
		1160	decode_ws (dec);
		1161
		1162	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1163	{
		1164	++dec->cur;
		1165	break;
		1166	}
999	}	1167	}
1000		1168
1001	DEC_DEC_DEPTH;	1169	DEC_DEC_DEPTH;
1002	return newRV_noinc ((SV *)av);	1170	return newRV_noinc ((SV *)av);
1003		1171
…		…
1008	}	1176	}
1009		1177
1010	static SV *	1178	static SV *
1011	decode_hv (dec_t *dec)	1179	decode_hv (dec_t *dec)
1012	{	1180	{
		1181	SV *sv;
1013	HV *hv = newHV ();	1182	HV *hv = newHV ();
1014		1183
1015	DEC_INC_DEPTH;	1184	DEC_INC_DEPTH;
1016	decode_ws (dec);	1185	decode_ws (dec);
1017		1186
1018	if (*dec->cur == '}')	1187	if (*dec->cur == '}')
1019	++dec->cur;	1188	++dec->cur;
1020	else	1189	else
1021	for (;;)	1190	for (;;)
1022	{	1191	{
1023	decode_ws (dec); EXPECT_CH ('"');	1192	EXPECT_CH ('"');
1024		1193
1025	// heuristic: assume that	1194	// heuristic: assume that
1026	// a) decode_str + hv_store_ent are abysmally slow.	1195	// a) decode_str + hv_store_ent are abysmally slow.
1027	// b) most hash keys are short, simple ascii text.	1196	// b) most hash keys are short, simple ascii text.
1028	// => try to "fast-match" such strings to avoid	1197	// => try to "fast-match" such strings to avoid
…		…
1032	char *p = dec->cur;	1201	char *p = dec->cur;
1033	char *e = p + 24; // only try up to 24 bytes	1202	char *e = p + 24; // only try up to 24 bytes
1034		1203
1035	for (;;)	1204	for (;;)
1036	{	1205	{
1037	// the >= 0x80 is true on most architectures	1206	// the >= 0x80 is false on most architectures
1038	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1207	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1039	{	1208	{
1040	// slow path, back up and use decode_str	1209	// slow path, back up and use decode_str
1041	SV *key = decode_str (dec);	1210	SV *key = decode_str (dec);
1042	if (!key)	1211	if (!key)
1043	goto fail;	1212	goto fail;
1044		1213
1045	decode_ws (dec); EXPECT_CH (':');	1214	decode_ws (dec); EXPECT_CH (':');
1046		1215
		1216	decode_ws (dec);
1047	value = decode_sv (dec);	1217	value = decode_sv (dec);
1048	if (!value)	1218	if (!value)
1049	{	1219	{
1050	SvREFCNT_dec (key);	1220	SvREFCNT_dec (key);
1051	goto fail;	1221	goto fail;
…		…
1063	int len = p - key;	1233	int len = p - key;
1064	dec->cur = p + 1;	1234	dec->cur = p + 1;
1065		1235
1066	decode_ws (dec); EXPECT_CH (':');	1236	decode_ws (dec); EXPECT_CH (':');
1067		1237
		1238	decode_ws (dec);
1068	value = decode_sv (dec);	1239	value = decode_sv (dec);
1069	if (!value)	1240	if (!value)
1070	goto fail;	1241	goto fail;
1071		1242
1072	hv_store (hv, key, len, value, 0);	1243	hv_store (hv, key, len, value, 0);
…		…
1088		1259
1089	if (*dec->cur != ',')	1260	if (*dec->cur != ',')
1090	ERR (", or } expected while parsing object/hash");	1261	ERR (", or } expected while parsing object/hash");
1091		1262
1092	++dec->cur;	1263	++dec->cur;
		1264
		1265	decode_ws (dec);
		1266
		1267	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1268	{
		1269	++dec->cur;
		1270	break;
		1271	}
1093	}	1272	}
1094		1273
1095	DEC_DEC_DEPTH;	1274	DEC_DEC_DEPTH;
1096	return newRV_noinc ((SV *)hv);	1275	sv = newRV_noinc ((SV *)hv);
		1276
		1277	// check filter callbacks
		1278	if (dec->json.flags & F_HOOK)
		1279	{
		1280	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1281	{
		1282	HE cb, he;
		1283
		1284	hv_iterinit (hv);
		1285	he = hv_iternext (hv);
		1286	hv_iterinit (hv);
		1287
		1288	// the next line creates a mortal sv each time its called.
		1289	// might want to optimise this for common cases.
		1290	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1291
		1292	if (cb)
		1293	{
		1294	dSP;
		1295	int count;
		1296
		1297	ENTER; SAVETMPS; PUSHMARK (SP);
		1298	XPUSHs (HeVAL (he));
		1299
		1300	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1301
		1302	if (count == 1)
		1303	{
		1304	sv = newSVsv (POPs);
		1305	FREETMPS; LEAVE;
		1306	return sv;
		1307	}
		1308
		1309	FREETMPS; LEAVE;
		1310	}
		1311	}
		1312
		1313	if (dec->json.cb_object)
		1314	{
		1315	dSP;
		1316	int count;
		1317
		1318	ENTER; SAVETMPS; PUSHMARK (SP);
		1319	XPUSHs (sv_2mortal (sv));
		1320
		1321	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1322
		1323	if (count == 1)
		1324	{
		1325	sv = newSVsv (POPs);
		1326	FREETMPS; LEAVE;
		1327	return sv;
		1328	}
		1329
		1330	SvREFCNT_inc (sv);
		1331	FREETMPS; LEAVE;
		1332	}
		1333	}
		1334
		1335	return sv;
1097		1336
1098	fail:	1337	fail:
1099	SvREFCNT_dec (hv);	1338	SvREFCNT_dec (hv);
1100	DEC_DEC_DEPTH;	1339	DEC_DEC_DEPTH;
1101	return 0;	1340	return 0;
1102	}	1341	}
1103		1342
1104	static SV *	1343	static SV *
1105	decode_sv (dec_t *dec)	1344	decode_sv (dec_t *dec)
1106	{	1345	{
1107	decode_ws (dec);
1108
1109	// the beauty of JSON: you need exactly one character lookahead	1346	// the beauty of JSON: you need exactly one character lookahead
1110	// to parse anything.	1347	// to parse everything.
1111	switch (*dec->cur)	1348	switch (*dec->cur)
1112	{	1349	{
1113	case '"': ++dec->cur; return decode_str (dec);	1350	case '"': ++dec->cur; return decode_str (dec);
1114	case '[': ++dec->cur; return decode_av (dec);	1351	case '[': ++dec->cur; return decode_av (dec);
1115	case '{': ++dec->cur; return decode_hv (dec);	1352	case '{': ++dec->cur; return decode_hv (dec);
1116		1353
1117	case '-':	1354	case '-':
1118	case '0': case '1': case '2': case '3': case '4':	1355	case '0': case '1': case '2': case '3': case '4':
1119	case '5': case '6': case '7': case '8': case '9':	1356	case '5': case '6': case '7': case '8': case '9':
1120	return decode_num (dec);	1357	return decode_num (dec);
1121		1358
1122	case 't':	1359	case 't':
1123	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1360	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1124	{	1361	{
1125	dec->cur += 4;	1362	dec->cur += 4;
		1363	#if JSON_SLOW
		1364	json_true = get_bool ("JSON::XS::true");
		1365	#endif
1126	return SvREFCNT_inc (json_true);	1366	return newSVsv (json_true);
1127	}	1367	}
1128	else	1368	else
1129	ERR ("'true' expected");	1369	ERR ("'true' expected");
1130		1370
1131	break;	1371	break;
1132		1372
1133	case 'f':	1373	case 'f':
1134	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1374	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1135	{	1375	{
1136	dec->cur += 5;	1376	dec->cur += 5;
		1377	#if JSON_SLOW
		1378	json_false = get_bool ("JSON::XS::false");
		1379	#endif
1137	return SvREFCNT_inc (json_false);	1380	return newSVsv (json_false);
1138	}	1381	}
1139	else	1382	else
1140	ERR ("'false' expected");	1383	ERR ("'false' expected");
1141		1384
1142	break;	1385	break;
…		…
1160	fail:	1403	fail:
1161	return 0;	1404	return 0;
1162	}	1405	}
1163		1406
1164	static SV *	1407	static SV *
1165	decode_json (SV string, JSON json, UV *offset_return)	1408	decode_json (SV string, JSON json, STRLEN *offset_return)
1166	{	1409	{
1167	dec_t dec;	1410	dec_t dec;
1168	UV offset;	1411	STRLEN offset;
1169	SV *sv;	1412	SV *sv;
1170		1413
1171	SvGETMAGIC (string);	1414	SvGETMAGIC (string);
1172	SvUPGRADE (string, SVt_PV);	1415	SvUPGRADE (string, SVt_PV);
1173		1416
…		…
1187	dec.end = SvEND (string);	1430	dec.end = SvEND (string);
1188	dec.err = 0;	1431	dec.err = 0;
1189	dec.depth = 0;	1432	dec.depth = 0;
1190	dec.maxdepth = DEC_DEPTH (dec.json.flags);	1433	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1191		1434
		1435	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1436	dec.json.flags \|= F_HOOK;
		1437
1192	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1438	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1439
		1440	decode_ws (&dec);
1193	sv = decode_sv (&dec);	1441	sv = decode_sv (&dec);
1194		1442
1195	if (!(offset_return \|\| !sv))	1443	if (!(offset_return \|\| !sv))
1196	{	1444	{
1197	// check for trailing garbage	1445	// check for trailing garbage
…		…
1241		1489
1242	return sv;	1490	return sv;
1243	}	1491	}
1244		1492
1245	/////////////////////////////////////////////////////////////////////////////	1493	/////////////////////////////////////////////////////////////////////////////
		1494	// incremental parser
		1495
		1496	static void
		1497	incr_parse (JSON *self)
		1498	{
		1499	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1500
		1501	for (;;)
		1502	{
		1503	//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
		1504	switch (self->incr_mode)
		1505	{
		1506	// only used for intiial whitespace skipping
		1507	case INCR_M_WS:
		1508	for (;;)
		1509	{
		1510	if (*p > 0x20)
		1511	{
		1512	self->incr_mode = INCR_M_JSON;
		1513	goto incr_m_json;
		1514	}
		1515	else if (!*p)
		1516	goto interrupt;
		1517
		1518	++p;
		1519	}
		1520
		1521	// skip a single char inside a string (for \\-processing)
		1522	case INCR_M_BS:
		1523	if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	self->incr_mode = INCR_M_STR;
		1528	goto incr_m_str;
		1529
		1530	// inside a string
		1531	case INCR_M_STR:
		1532	incr_m_str:
		1533	for (;;)
		1534	{
		1535	if (*p == '"')
		1536	{
		1537	++p;
		1538	self->incr_mode = INCR_M_JSON;
		1539
		1540	if (!self->incr_nest)
		1541	goto interrupt;
		1542
		1543	goto incr_m_json;
		1544	}
		1545	else if (*p == '\\')
		1546	{
		1547	++p; // "virtually" consumes character after \
		1548
		1549	if (!*p) // if at end of string we have to switch modes
		1550	{
		1551	self->incr_mode = INCR_M_BS;
		1552	goto interrupt;
		1553	}
		1554	}
		1555	else if (!*p)
		1556	goto interrupt;
		1557
		1558	++p;
		1559	}
		1560
		1561	// after initial ws, outside string
		1562	case INCR_M_JSON:
		1563	incr_m_json:
		1564	for (;;)
		1565	{
		1566	switch (*p++)
		1567	{
		1568	case 0:
		1569	--p;
		1570	goto interrupt;
		1571
		1572	case 0x09:
		1573	case 0x0a:
		1574	case 0x0d:
		1575	case 0x20:
		1576	if (!self->incr_nest)
		1577	{
		1578	--p; // do not eat the whitespace, let the next round do it
		1579	goto interrupt;
		1580	}
		1581	break;
		1582
		1583	case '"':
		1584	self->incr_mode = INCR_M_STR;
		1585	goto incr_m_str;
		1586
		1587	case '[':
		1588	case '{':
		1589	++self->incr_nest;
		1590	break;
		1591
		1592	case ']':
		1593	case '}':
		1594	if (!--self->incr_nest)
		1595	goto interrupt;
		1596	}
		1597	}
		1598	}
		1599
		1600	modechange:
		1601	;
		1602	}
		1603
		1604	interrupt:
		1605	self->incr_pos = p - SvPVX (self->incr_text);
		1606	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
		1607	}
		1608
		1609	/////////////////////////////////////////////////////////////////////////////
1246	// XS interface functions	1610	// XS interface functions
1247		1611
1248	MODULE = JSON::XS PACKAGE = JSON::XS	1612	MODULE = JSON::XS PACKAGE = JSON::XS
1249		1613
1250	BOOT:	1614	BOOT:
…		…
1259	: -1;	1623	: -1;
1260		1624
1261	json_stash = gv_stashpv ("JSON::XS" , 1);	1625	json_stash = gv_stashpv ("JSON::XS" , 1);
1262	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1626	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1263		1627
1264	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1628	json_true = get_bool ("JSON::XS::true");
1265	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1629	json_false = get_bool ("JSON::XS::false");
1266	}	1630	}
1267		1631
1268	PROTOTYPES: DISABLE	1632	PROTOTYPES: DISABLE
		1633
		1634	void CLONE (...)
		1635	CODE:
		1636	json_stash = 0;
		1637	json_boolean_stash = 0;
1269		1638
1270	void new (char *klass)	1639	void new (char *klass)
1271	PPCODE:	1640	PPCODE:
1272	{	1641	{
1273	SV *pv = NEWSV (0, sizeof (JSON));	1642	SV *pv = NEWSV (0, sizeof (JSON));
1274	SvPOK_only (pv);	1643	SvPOK_only (pv);
1275	Zero (SvPVX (pv), 1, sizeof (JSON));	1644	Zero (SvPVX (pv), 1, JSON);
1276	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1645	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1277	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1646	XPUSHs (sv_2mortal (sv_bless (
		1647	newRV_noinc (pv),
		1648	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1649	)));
1278	}	1650	}
1279		1651
1280	void ascii (JSON *self, int enable = 1)	1652	void ascii (JSON *self, int enable = 1)
1281	ALIAS:	1653	ALIAS:
1282	ascii = F_ASCII	1654	ascii = F_ASCII
…		…
1289	pretty = F_PRETTY	1661	pretty = F_PRETTY
1290	allow_nonref = F_ALLOW_NONREF	1662	allow_nonref = F_ALLOW_NONREF
1291	shrink = F_SHRINK	1663	shrink = F_SHRINK
1292	allow_blessed = F_ALLOW_BLESSED	1664	allow_blessed = F_ALLOW_BLESSED
1293	convert_blessed = F_CONV_BLESSED	1665	convert_blessed = F_CONV_BLESSED
		1666	relaxed = F_RELAXED
		1667	allow_unknown = F_ALLOW_UNKNOWN
1294	PPCODE:	1668	PPCODE:
1295	{	1669	{
1296	if (enable)	1670	if (enable)
1297	self->flags \|= ix;	1671	self->flags \|= ix;
1298	else	1672	else
1299	self->flags &= ~ix;	1673	self->flags &= ~ix;
1300		1674
1301	XPUSHs (ST (0));	1675	XPUSHs (ST (0));
1302	}	1676	}
1303		1677
		1678	void get_ascii (JSON *self)
		1679	ALIAS:
		1680	get_ascii = F_ASCII
		1681	get_latin1 = F_LATIN1
		1682	get_utf8 = F_UTF8
		1683	get_indent = F_INDENT
		1684	get_canonical = F_CANONICAL
		1685	get_space_before = F_SPACE_BEFORE
		1686	get_space_after = F_SPACE_AFTER
		1687	get_allow_nonref = F_ALLOW_NONREF
		1688	get_shrink = F_SHRINK
		1689	get_allow_blessed = F_ALLOW_BLESSED
		1690	get_convert_blessed = F_CONV_BLESSED
		1691	get_relaxed = F_RELAXED
		1692	get_allow_unknown = F_ALLOW_UNKNOWN
		1693	PPCODE:
		1694	XPUSHs (boolSV (self->flags & ix));
		1695
1304	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1696	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1305	PPCODE:	1697	PPCODE:
1306	{	1698	{
1307	UV log2 = 0;	1699	UV log2 = 0;
1308		1700
…		…
1314	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1706	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1315		1707
1316	XPUSHs (ST (0));	1708	XPUSHs (ST (0));
1317	}	1709	}
1318		1710
		1711	U32 get_max_depth (JSON *self)
		1712	CODE:
		1713	RETVAL = DEC_DEPTH (self->flags);
		1714	OUTPUT:
		1715	RETVAL
		1716
1319	void max_size (JSON *self, UV max_size = 0)	1717	void max_size (JSON *self, UV max_size = 0)
1320	PPCODE:	1718	PPCODE:
1321	{	1719	{
1322	UV log2 = 0;	1720	UV log2 = 0;
1323		1721
…		…
1326		1724
1327	while ((1UL << log2) < max_size)	1725	while ((1UL << log2) < max_size)
1328	++log2;	1726	++log2;
1329		1727
1330	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1728	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1729
		1730	XPUSHs (ST (0));
		1731	}
		1732
		1733	int get_max_size (JSON *self)
		1734	CODE:
		1735	RETVAL = DEC_SIZE (self->flags);
		1736	OUTPUT:
		1737	RETVAL
		1738
		1739	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1740	PPCODE:
		1741	{
		1742	SvREFCNT_dec (self->cb_object);
		1743	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1744
		1745	XPUSHs (ST (0));
		1746	}
		1747
		1748	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1749	PPCODE:
		1750	{
		1751	if (!self->cb_sk_object)
		1752	self->cb_sk_object = newHV ();
		1753
		1754	if (SvOK (cb))
		1755	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1756	else
		1757	{
		1758	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1759
		1760	if (!HvKEYS (self->cb_sk_object))
		1761	{
		1762	SvREFCNT_dec (self->cb_sk_object);
		1763	self->cb_sk_object = 0;
		1764	}
		1765	}
1331		1766
1332	XPUSHs (ST (0));	1767	XPUSHs (ST (0));
1333	}	1768	}
1334		1769
1335	void encode (JSON self, SV scalar)	1770	void encode (JSON self, SV scalar)
…		…
1341	XPUSHs (decode_json (jsonstr, self, 0));	1776	XPUSHs (decode_json (jsonstr, self, 0));
1342		1777
1343	void decode_prefix (JSON self, SV jsonstr)	1778	void decode_prefix (JSON self, SV jsonstr)
1344	PPCODE:	1779	PPCODE:
1345	{	1780	{
1346	UV offset;	1781	STRLEN offset;
1347	EXTEND (SP, 2);	1782	EXTEND (SP, 2);
1348	PUSHs (decode_json (jsonstr, self, &offset));	1783	PUSHs (decode_json (jsonstr, self, &offset));
1349	PUSHs (sv_2mortal (newSVuv (offset)));	1784	PUSHs (sv_2mortal (newSVuv (offset)));
1350	}	1785	}
1351		1786
		1787	void incr_parse (JSON self, SV jsonstr = 0)
		1788	PPCODE:
		1789	{
		1790	if (!self->incr_text)
		1791	self->incr_text = newSVpvn ("", 0);
		1792
		1793	// append data, if any
		1794	if (jsonstr)
		1795	{
		1796	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1797	{
		1798	/* utf-8-ness differs, need to upgrade */
		1799	sv_utf8_upgrade (self->incr_text);
		1800
		1801	if (self->incr_pos)
		1802	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1803	- (U8 *)SvPVX (self->incr_text);
		1804	}
		1805
		1806	{
		1807	STRLEN len;
		1808	const char *str = SvPV (jsonstr, len);
		1809	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1810	Move (str, SvEND (self->incr_text), len, char);
		1811	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1812	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1813	}
		1814	}
		1815
		1816	if (GIMME_V != G_VOID)
		1817	do
		1818	{
		1819	STRLEN offset;
		1820
		1821	if (!INCR_DONE (self))
		1822	{
		1823	incr_parse (self);
		1824	if (!INCR_DONE (self))
		1825	break;
		1826	}
		1827
		1828	XPUSHs (decode_json (self->incr_text, self, &offset));
		1829
		1830	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1831	self->incr_pos -= offset;
		1832	self->incr_nest = 0;
		1833	self->incr_mode = 0;
		1834	}
		1835	while (GIMME_V == G_ARRAY);
		1836	}
		1837
		1838	SV incr_text (JSON self)
		1839	ATTRS: lvalue
		1840	CODE:
		1841	{
		1842	if (self->incr_pos)
		1843	croak ("incr_text can not be called when the incremental parser already started parsing");
		1844
		1845	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1846	}
		1847	OUTPUT:
		1848	RETVAL
		1849
		1850	void incr_skip (JSON *self)
		1851	CODE:
		1852	{
		1853	if (self->incr_pos)
		1854	{
		1855	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1856	self->incr_pos = 0;
		1857	self->incr_nest = 0;
		1858	self->incr_mode = 0;
		1859	}
		1860	}
		1861
		1862	void DESTROY (JSON *self)
		1863	CODE:
		1864	SvREFCNT_dec (self->cb_sk_object);
		1865	SvREFCNT_dec (self->cb_object);
		1866	SvREFCNT_dec (self->incr_text);
		1867
1352	PROTOTYPES: ENABLE	1868	PROTOTYPES: ENABLE
1353		1869
1354	void to_json (SV *scalar)	1870	void encode_json (SV *scalar)
		1871	ALIAS:
		1872	to_json_ = 0
		1873	encode_json = F_UTF8
1355	PPCODE:	1874	PPCODE:
1356	{	1875	{
1357	JSON json = { F_DEFAULT \| F_UTF8 };	1876	JSON json = { F_DEFAULT \| ix };
1358	XPUSHs (encode_json (scalar, &json));	1877	XPUSHs (encode_json (scalar, &json));
1359	}	1878	}
1360		1879
1361	void from_json (SV *jsonstr)	1880	void decode_json (SV *jsonstr)
		1881	ALIAS:
		1882	from_json_ = 0
		1883	decode_json = F_UTF8
1362	PPCODE:	1884	PPCODE:
1363	{	1885	{
1364	JSON json = { F_DEFAULT \| F_UTF8 };	1886	JSON json = { F_DEFAULT \| ix };
1365	XPUSHs (decode_json (jsonstr, &json, 0));	1887	XPUSHs (decode_json (jsonstr, &json, 0));
1366	}	1888	}
1367		1889
		1890

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 1.84 by root, Thu Mar 27 06:37:35 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 1.84 by root, Thu Mar 27 06:37:35 2008 UTC