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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.52 by root, Mon Jul 2 02:57:11 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
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	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
35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	42	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
…		…
47		54
48	#define SB do {	55	#define SB do {
49	#define SE } while (0)	56	#define SE } while (0)
50		57
51	#if __GNUC__ >= 3	58	#if __GNUC__ >= 3
52	# define expect(expr,value) __builtin_expect ((expr),(value))	59	# define expect(expr,value) __builtin_expect ((expr), (value))
53	# define inline inline	60	# define INLINE static inline
54	#else	61	#else
55	# define expect(expr,value) (expr)	62	# define expect(expr,value) (expr)
56	# define inline static	63	# define INLINE static
57	#endif	64	#endif
58		65
59	#define expect_false(expr) expect ((expr) != 0, 0)	66	#define expect_false(expr) expect ((expr) != 0, 0)
60	#define expect_true(expr) expect ((expr) != 0, 1)	67	#define expect_true(expr) expect ((expr) != 0, 1)
61		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
62	static HV json_stash, json_boolean_stash; // JSON::XS::	81	static HV json_stash, json_boolean_stash; // JSON::XS::
63	static SV json_true, json_false;	82	static SV json_true, json_false;
		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)
64		92
65	typedef struct {	93	typedef struct {
66	U32 flags;	94	U32 flags;
67	SV *cb_object;	95	SV *cb_object;
68	HV *cb_sk_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;
69	} JSON;	103	} JSON;
70		104
71	/////////////////////////////////////////////////////////////////////////////	105	/////////////////////////////////////////////////////////////////////////////
72	// utility functions	106	// utility functions
73		107
74	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
75	shrink (SV *sv)	120	shrink (SV *sv)
76	{	121	{
77	sv_utf8_downgrade (sv, 1);	122	sv_utf8_downgrade (sv, 1);
78	if (SvLEN (sv) > SvCUR (sv) + 1)	123	if (SvLEN (sv) > SvCUR (sv) + 1)
79	{	124	{
…		…
88	// 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
89	// case of an error.	134	// case of an error.
90	// we special-case "safe" characters from U+80 .. U+7FF,	135	// we special-case "safe" characters from U+80 .. U+7FF,
91	// but use the very good perl function to parse anything else.	136	// but use the very good perl function to parse anything else.
92	// note that we never call this function for a ascii codepoints	137	// note that we never call this function for a ascii codepoints
93	inline UV	138	INLINE UV
94	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	139	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
95	{	140	{
96	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	141	if (expect_true (len >= 2
97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	142	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	143	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
99	{	144	{
100	*clen = 2;	145	*clen = 2;
101	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	146	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
102	}	147	}
103	else	148	else
104	{	149	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
105	*clen = (STRLEN)-1;	150	}
106	return (UV)-1;	151
107	}	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;
108	}	174	}
109		175
110	/////////////////////////////////////////////////////////////////////////////	176	/////////////////////////////////////////////////////////////////////////////
111	// encoder	177	// encoder
112		178
…		…
117	char *end; // SvEND (sv)	183	char *end; // SvEND (sv)
118	SV *sv; // result scalar	184	SV *sv; // result scalar
119	JSON json;	185	JSON json;
120	U32 indent; // indentation level	186	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	187	U32 maxdepth; // max. indentation/recursion level
		188	UV limit; // escape character values >= this value when encoding
122	} enc_t;	189	} enc_t;
123		190
124	inline void	191	INLINE void
125	need (enc_t *enc, STRLEN len)	192	need (enc_t *enc, STRLEN len)
126	{	193	{
127	if (expect_false (enc->cur + len >= enc->end))	194	if (expect_false (enc->cur + len >= enc->end))
128	{	195	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	196	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
131	enc->cur = SvPVX (enc->sv) + cur;	198	enc->cur = SvPVX (enc->sv) + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	199	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
133	}	200	}
134	}	201	}
135		202
136	inline void	203	INLINE void
137	encode_ch (enc_t *enc, char ch)	204	encode_ch (enc_t *enc, char ch)
138	{	205	{
139	need (enc, 1);	206	need (enc, 1);
140	*enc->cur++ = ch;	207	*enc->cur++ = ch;
141	}	208	}
…		…
195	{	262	{
196	uch = ch;	263	uch = ch;
197	clen = 1;	264	clen = 1;
198	}	265	}
199		266
200	if (uch > 0x10FFFFUL)	267	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
204	{	268	{
205	if (uch > 0xFFFFUL)	269	if (uch >= 0x10000UL)
206	{	270	{
		271	if (uch >= 0x110000UL)
		272	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		273
207	need (enc, len += 11);	274	need (enc, len += 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	275	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	276	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	277	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	278	enc->cur += 12;
…		…
239	while (--clen);	306	while (--clen);
240	}	307	}
241	else	308	else
242	{	309	{
243	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
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	311	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	312	++str;
246	}	313	}
247	}	314	}
248	}	315	}
249	}	316	}
250		317
251	--len;	318	--len;
252	}	319	}
253	}	320	}
254		321
255	inline void	322	INLINE void
256	encode_indent (enc_t *enc)	323	encode_indent (enc_t *enc)
257	{	324	{
258	if (enc->json.flags & F_INDENT)	325	if (enc->json.flags & F_INDENT)
259	{	326	{
260	int spaces = enc->indent * INDENT_STEP;	327	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	330	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	331	enc->cur += spaces;
265	}	332	}
266	}	333	}
267		334
268	inline void	335	INLINE void
269	encode_space (enc_t *enc)	336	encode_space (enc_t *enc)
270	{	337	{
271	need (enc, 1);	338	need (enc, 1);
272	encode_ch (enc, ' ');	339	encode_ch (enc, ' ');
273	}	340	}
274		341
275	inline void	342	INLINE void
276	encode_nl (enc_t *enc)	343	encode_nl (enc_t *enc)
277	{	344	{
278	if (enc->json.flags & F_INDENT)	345	if (enc->json.flags & F_INDENT)
279	{	346	{
280	need (enc, 1);	347	need (enc, 1);
281	encode_ch (enc, '\n');	348	encode_ch (enc, '\n');
282	}	349	}
283	}	350	}
284		351
285	inline void	352	INLINE void
286	encode_comma (enc_t *enc)	353	encode_comma (enc_t *enc)
287	{	354	{
288	encode_ch (enc, ',');	355	encode_ch (enc, ',');
289		356
290	if (enc->json.flags & F_INDENT)	357	if (enc->json.flags & F_INDENT)
…		…
301	int i, len = av_len (av);	368	int i, len = av_len (av);
302		369
303	if (enc->indent >= enc->maxdepth)	370	if (enc->indent >= enc->maxdepth)
304	croak ("data structure too deep (hit recursion limit)");	371	croak ("data structure too deep (hit recursion limit)");
305		372
306	encode_ch (enc, '['); encode_nl (enc);	373	encode_ch (enc, '[');
307	++enc->indent;	374
		375	if (len >= 0)
		376	{
		377	encode_nl (enc); ++enc->indent;
308		378
309	for (i = 0; i <= len; ++i)	379	for (i = 0; i <= len; ++i)
310	{	380	{
		381	SV **svp = av_fetch (av, i, 0);
		382
311	encode_indent (enc);	383	encode_indent (enc);
312	encode_sv (enc, *av_fetch (av, i, 0));
313		384
		385	if (svp)
		386	encode_sv (enc, *svp);
		387	else
		388	encode_str (enc, "null", 4, 0);
		389
314	if (i < len)	390	if (i < len)
315	encode_comma (enc);	391	encode_comma (enc);
316	}	392	}
317		393
		394	encode_nl (enc); --enc->indent; encode_indent (enc);
		395	}
		396
318	encode_nl (enc);	397	encode_ch (enc, ']');
319
320	--enc->indent;
321	encode_indent (enc); encode_ch (enc, ']');
322	}	398	}
323		399
324	static void	400	static void
325	encode_he (enc_t enc, HE he)	401	encode_hk (enc_t enc, HE he)
326	{	402	{
327	encode_ch (enc, '"');	403	encode_ch (enc, '"');
328		404
329	if (HeKLEN (he) == HEf_SVKEY)	405	if (HeKLEN (he) == HEf_SVKEY)
330	{	406	{
…		…
343	encode_ch (enc, '"');	419	encode_ch (enc, '"');
344		420
345	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	421	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
346	encode_ch (enc, ':');	422	encode_ch (enc, ':');
347	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	423	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
348	encode_sv (enc, HeVAL (he));
349	}	424	}
350		425
351	// compare hash entries, used when all keys are bytestrings	426	// compare hash entries, used when all keys are bytestrings
352	static int	427	static int
353	he_cmp_fast (const void a_, const void b_)	428	he_cmp_fast (const void a_, const void b_)
…		…
358	HE b = (HE **)b_;	433	HE b = (HE **)b_;
359		434
360	STRLEN la = HeKLEN (a);	435	STRLEN la = HeKLEN (a);
361	STRLEN lb = HeKLEN (b);	436	STRLEN lb = HeKLEN (b);
362		437
363	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	438	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
364	cmp = la - lb;	439	cmp = lb - la;
365		440
366	return cmp;	441	return cmp;
367	}	442	}
368		443
369	// 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
370	static int	445	static int
371	he_cmp_slow (const void a, const void b)	446	he_cmp_slow (const void a, const void b)
372	{	447	{
373	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	448	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
374	}	449	}
375		450
376	static void	451	static void
377	encode_hv (enc_t enc, HV hv)	452	encode_hv (enc_t enc, HV hv)
378	{	453	{
379	int count, i;	454	HE *he;
380		455
381	if (enc->indent >= enc->maxdepth)	456	if (enc->indent >= enc->maxdepth)
382	croak ("data structure too deep (hit recursion limit)");	457	croak ("data structure too deep (hit recursion limit)");
383		458
384	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	459	encode_ch (enc, '{');
385		460
386	if ((count = hv_iterinit (hv)))
387	{
388	// for canonical output we have to sort by keys first	461	// for canonical output we have to sort by keys first
389	// actually, this is mostly due to the stupid so-called	462	// actually, this is mostly due to the stupid so-called
390	// security workaround added somewhere in 5.8.x.	463	// security workaround added somewhere in 5.8.x.
391	// that randomises hash orderings	464	// that randomises hash orderings
392	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))
393	{	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	{
394	int fast = 1;	485	int i, fast = 1;
395	HE *he;
396	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	486	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
397	HE *hes = _alloca (count sizeof (HE));	487	HE *hes = _alloca (count sizeof (HE));
398	#else	488	#else
399	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
400	#endif	490	#endif
…		…
427		517
428	FREETMPS;	518	FREETMPS;
429	LEAVE;	519	LEAVE;
430	}	520	}
431		521
432	for (i = 0; i < count; ++i)	522	encode_nl (enc); ++enc->indent;
		523
		524	while (count--)
433	{	525	{
434	encode_indent (enc);	526	encode_indent (enc);
		527	he = hes [count];
435	encode_he (enc, hes [i]);	528	encode_hk (enc, he);
		529	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
436		530
437	if (i < count - 1)	531	if (count)
438	encode_comma (enc);	532	encode_comma (enc);
439	}	533	}
440		534
441	encode_nl (enc);	535	encode_nl (enc); --enc->indent; encode_indent (enc);
442	}	536	}
		537	}
443	else	538	else
		539	{
		540	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		541	if ((he = hv_iternext (hv)))
444	{	542	{
445	HE *he = hv_iternext (hv);	543	encode_nl (enc); ++enc->indent;
446		544
447	for (;;)	545	for (;;)
448	{	546	{
449	encode_indent (enc);	547	encode_indent (enc);
450	encode_he (enc, he);	548	encode_hk (enc, he);
		549	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		550
452	if (!(he = hv_iternext (hv)))	551	if (!(he = hv_iternext (hv)))
453	break;	552	break;
454		553
455	encode_comma (enc);	554	encode_comma (enc);
456	}	555	}
457		556
458	encode_nl (enc);	557	encode_nl (enc); --enc->indent; encode_indent (enc);
459	}	558	}
460	}	559	}
461		560
462	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	561	encode_ch (enc, '}');
463	}	562	}
464		563
465	// encode objects, arrays and special \0=false and \1=true values.	564	// encode objects, arrays and special \0=false and \1=true values.
466	static void	565	static void
467	encode_rv (enc_t enc, SV sv)	566	encode_rv (enc_t enc, SV sv)
…		…
471	SvGETMAGIC (sv);	570	SvGETMAGIC (sv);
472	svt = SvTYPE (sv);	571	svt = SvTYPE (sv);
473		572
474	if (expect_false (SvOBJECT (sv)))	573	if (expect_false (SvOBJECT (sv)))
475	{	574	{
		575	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		576	? json_boolean_stash
		577	: gv_stashpv ("JSON::XS::Boolean", 1);
		578
476	if (SvSTASH (sv) == json_boolean_stash)	579	if (SvSTASH (sv) == stash)
477	{	580	{
478	if (SvIV (sv))	581	if (SvIV (sv))
479	encode_str (enc, "true", 4, 0);	582	encode_str (enc, "true", 4, 0);
480	else	583	else
481	encode_str (enc, "false", 5, 0);	584	encode_str (enc, "false", 5, 0);
…		…
489	}	592	}
490	#endif	593	#endif
491	if (enc->json.flags & F_CONV_BLESSED)	594	if (enc->json.flags & F_CONV_BLESSED)
492	{	595	{
493	// 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
494	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	597	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
495		598
496	if (to_json)	599	if (to_json)
497	{	600	{
		601	dSP;
		602
498	dSP; ENTER; SAVETMPS; 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; LEAVE;	620	FREETMPS; LEAVE;
510	}	621	}
511	else if (enc->json.flags & F_ALLOW_BLESSED)	622	else if (enc->json.flags & F_ALLOW_BLESSED)
512	encode_str (enc, "null", 4, 0);	623	encode_str (enc, "null", 4, 0);
…		…
532		643
533	if (len == 1 && *pv == '1')	644	if (len == 1 && *pv == '1')
534	encode_str (enc, "true", 4, 0);	645	encode_str (enc, "true", 4, 0);
535	else if (len == 1 && *pv == '0')	646	else if (len == 1 && *pv == '0')
536	encode_str (enc, "false", 5, 0);	647	encode_str (enc, "false", 5, 0);
		648	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		649	encode_str (enc, "null", 4, 0);
537	else	650	else
538	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",
539	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	652	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
540	}	653	}
		654	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		655	encode_str (enc, "null", 4, 0);
541	else	656	else
542	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",
543	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	658	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
544	}	659	}
545		660
…		…
563	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	678	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
564	enc->cur += strlen (enc->cur);	679	enc->cur += strlen (enc->cur);
565	}	680	}
566	else if (SvIOKp (sv))	681	else if (SvIOKp (sv))
567	{	682	{
568	// 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
569	if (SvUV (sv) & ~(UV)0x7fff)	684	// we assume two's complement
570	{	685	// we assume no aliasing issues in the union
571	// large integer, use the (rather slow) snprintf way.	686	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
572	need (enc, sizeof (UV) * 3);	687	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
573	enc->cur +=
574	SvIsUV(sv)
575	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
576	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
577	}
578	else
579	{	688	{
580	// optimise the "small number case"	689	// optimise the "small number case"
581	// 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
582	I32 i = SvIV (sv);	692	I32 i = SvIVX (sv);
583	U32 u;	693	U32 u;
584	char digit, nz = 0;	694	char digit, nz = 0;
585		695
586	need (enc, 6);	696	need (enc, 6);
587		697
…		…
593		703
594	// 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
595	// 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,
596	// resulting in a net multiplication by 10.	706	// resulting in a net multiplication by 10.
597	// we always write the digit to memory but conditionally increment	707	// we always write the digit to memory but conditionally increment
598	// the pointer, to ease the usage of conditional move instructions.	708	// the pointer, to enable the use of conditional move instructions.
599	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;
600	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;
601	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;
602	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;
603	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'
604	}	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	}
605	}	724	}
606	else if (SvROK (sv))	725	else if (SvROK (sv))
607	encode_rv (enc, SvRV (sv));	726	encode_rv (enc, SvRV (sv));
608	else if (!SvOK (sv))	727	else if (!SvOK (sv) \|\| enc->json.flags & F_ALLOW_UNKNOWN)
609	encode_str (enc, "null", 4, 0);	728	encode_str (enc, "null", 4, 0);
610	else	729	else
611	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",
612	SvPV_nolen (sv), SvFLAGS (sv));	731	SvPV_nolen (sv), SvFLAGS (sv));
613	}	732	}
…		…
624	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	743	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
625	enc.cur = SvPVX (enc.sv);	744	enc.cur = SvPVX (enc.sv);
626	enc.end = SvEND (enc.sv);	745	enc.end = SvEND (enc.sv);
627	enc.indent = 0;	746	enc.indent = 0;
628	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;
629		751
630	SvPOK_only (enc.sv);	752	SvPOK_only (enc.sv);
631	encode_sv (&enc, scalar);	753	encode_sv (&enc, scalar);
632		754
633	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	755	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
654	JSON json;	776	JSON json;
655	U32 depth; // recursion depth	777	U32 depth; // recursion depth
656	U32 maxdepth; // recursion depth limit	778	U32 maxdepth; // recursion depth limit
657	} dec_t;	779	} dec_t;
658		780
659	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
660	decode_ws (dec_t *dec)	791	decode_ws (dec_t *dec)
661	{	792	{
662	for (;;)	793	for (;;)
663	{	794	{
664	char ch = *dec->cur;	795	char ch = *dec->cur;
665		796
666	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	}
667	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	809	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
668	break;	810	break; // parse error, but let higher level handle it, gives better error messages
669		811
670	++dec->cur;	812	++dec->cur;
671	}	813	}
672	}	814	}
673		815
…		…
779		921
780	if (hi >= 0x80)	922	if (hi >= 0x80)
781	{	923	{
782	utf8 = 1;	924	utf8 = 1;
783		925
784	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	926	cur = encode_utf8 (cur, hi);
785	}	927	}
786	else	928	else
787	*cur++ = hi;	929	*cur++ = hi;
788	}	930	}
789	break;	931	break;
…		…
791	default:	933	default:
792	--dec_cur;	934	--dec_cur;
793	ERR ("illegal backslash escape sequence in string");	935	ERR ("illegal backslash escape sequence in string");
794	}	936	}
795	}	937	}
796	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	938	else if (expect_true (ch >= 0x20 && ch < 0x80))
797	*cur++ = ch;	939	*cur++ = ch;
798	else if (ch >= 0x80)	940	else if (ch >= 0x80)
799	{	941	{
800	STRLEN clen;	942	STRLEN clen;
801	UV uch;	943	UV uch;
…		…
922	is_nv = 1;	1064	is_nv = 1;
923	}	1065	}
924		1066
925	if (!is_nv)	1067	if (!is_nv)
926	{	1068	{
		1069	int len = dec->cur - start;
		1070
927	// 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
928	if (*start == '-')	1072	if (*start == '-')
929	switch (dec->cur - start)	1073	switch (len)
930	{	1074	{
931	case 2: return newSViv (-( start [1] - '0' * 1));	1075	case 2: return newSViv (-( start [1] - '0' * 1));
932	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1076	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
933	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));
934	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));
935	}	1080	}
936	else	1081	else
937	switch (dec->cur - start)	1082	switch (len)
938	{	1083	{
939	case 1: return newSViv ( start [0] - '0' * 1);	1084	case 1: return newSViv ( start [0] - '0' * 1);
940	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1085	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
941	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);
942	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);
943	}	1089	}
944		1090
945	{	1091	{
946	UV uv;	1092	UV uv;
947	int numtype = grok_number (start, dec->cur - start, &uv);	1093	int numtype = grok_number (start, len, &uv);
948	if (numtype & IS_NUMBER_IN_UV)	1094	if (numtype & IS_NUMBER_IN_UV)
949	if (numtype & IS_NUMBER_NEG)	1095	if (numtype & IS_NUMBER_NEG)
950	{	1096	{
951	if (uv < (UV)IV_MIN)	1097	if (uv < (UV)IV_MIN)
952	return newSViv (-(IV)uv);	1098	return newSViv (-(IV)uv);
953	}	1099	}
954	else	1100	else
955	return newSVuv (uv);	1101	return newSVuv (uv);
956
957	// here would likely be the place for bigint support
958	}	1102	}
959	}
960		1103
961	// 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
962	return newSVnv (Atof (start));	1120	return newSVnv (Atof (start));
963		1121
964	fail:	1122	fail:
965	return 0;	1123	return 0;
966	}	1124	}
…		…
996		1154
997	if (*dec->cur != ',')	1155	if (*dec->cur != ',')
998	ERR (", or ] expected while parsing array");	1156	ERR (", or ] expected while parsing array");
999		1157
1000	++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	}
1001	}	1167	}
1002		1168
1003	DEC_DEC_DEPTH;	1169	DEC_DEC_DEPTH;
1004	return newRV_noinc ((SV *)av);	1170	return newRV_noinc ((SV *)av);
1005		1171
…		…
1021	if (*dec->cur == '}')	1187	if (*dec->cur == '}')
1022	++dec->cur;	1188	++dec->cur;
1023	else	1189	else
1024	for (;;)	1190	for (;;)
1025	{	1191	{
1026	decode_ws (dec); EXPECT_CH ('"');	1192	EXPECT_CH ('"');
1027		1193
1028	// heuristic: assume that	1194	// heuristic: assume that
1029	// a) decode_str + hv_store_ent are abysmally slow.	1195	// a) decode_str + hv_store_ent are abysmally slow.
1030	// b) most hash keys are short, simple ascii text.	1196	// b) most hash keys are short, simple ascii text.
1031	// => try to "fast-match" such strings to avoid	1197	// => try to "fast-match" such strings to avoid
…		…
1035	char *p = dec->cur;	1201	char *p = dec->cur;
1036	char *e = p + 24; // only try up to 24 bytes	1202	char *e = p + 24; // only try up to 24 bytes
1037		1203
1038	for (;;)	1204	for (;;)
1039	{	1205	{
1040	// the >= 0x80 is true on most architectures	1206	// the >= 0x80 is false on most architectures
1041	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')	1207	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
1042	{	1208	{
1043	// slow path, back up and use decode_str	1209	// slow path, back up and use decode_str
1044	SV *key = decode_str (dec);	1210	SV *key = decode_str (dec);
1045	if (!key)	1211	if (!key)
1046	goto fail;	1212	goto fail;
1047		1213
1048	decode_ws (dec); EXPECT_CH (':');	1214	decode_ws (dec); EXPECT_CH (':');
1049		1215
		1216	decode_ws (dec);
1050	value = decode_sv (dec);	1217	value = decode_sv (dec);
1051	if (!value)	1218	if (!value)
1052	{	1219	{
1053	SvREFCNT_dec (key);	1220	SvREFCNT_dec (key);
1054	goto fail;	1221	goto fail;
…		…
1066	int len = p - key;	1233	int len = p - key;
1067	dec->cur = p + 1;	1234	dec->cur = p + 1;
1068		1235
1069	decode_ws (dec); EXPECT_CH (':');	1236	decode_ws (dec); EXPECT_CH (':');
1070		1237
		1238	decode_ws (dec);
1071	value = decode_sv (dec);	1239	value = decode_sv (dec);
1072	if (!value)	1240	if (!value)
1073	goto fail;	1241	goto fail;
1074		1242
1075	hv_store (hv, key, len, value, 0);	1243	hv_store (hv, key, len, value, 0);
…		…
1091		1259
1092	if (*dec->cur != ',')	1260	if (*dec->cur != ',')
1093	ERR (", or } expected while parsing object/hash");	1261	ERR (", or } expected while parsing object/hash");
1094		1262
1095	++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	}
1096	}	1272	}
1097		1273
1098	DEC_DEC_DEPTH;	1274	DEC_DEC_DEPTH;
1099	sv = newRV_noinc ((SV *)hv);	1275	sv = newRV_noinc ((SV *)hv);
1100		1276
…		…
1113	// might want to optimise this for common cases.	1289	// might want to optimise this for common cases.
1114	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);	1290	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1115		1291
1116	if (cb)	1292	if (cb)
1117	{	1293	{
		1294	dSP;
1118	int count;	1295	int count;
1119	ENTER; SAVETMPS;
1120		1296
1121	dSP; PUSHMARK (SP);	1297	ENTER; SAVETMPS; PUSHMARK (SP);
1122	XPUSHs (HeVAL (he));	1298	XPUSHs (HeVAL (he));
1123		1299
1124	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1300	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1125		1301
1126	if (count == 1)	1302	if (count == 1)
…		…
1134	}	1310	}
1135	}	1311	}
1136		1312
1137	if (dec->json.cb_object)	1313	if (dec->json.cb_object)
1138	{	1314	{
		1315	dSP;
1139	int count;	1316	int count;
1140	ENTER; SAVETMPS;
1141		1317
1142	dSP; ENTER; SAVETMPS; PUSHMARK (SP);	1318	ENTER; SAVETMPS; PUSHMARK (SP);
1143	XPUSHs (sv_2mortal (sv));	1319	XPUSHs (sv_2mortal (sv));
1144		1320
1145	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1321	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1146		1322
1147	if (count == 1)	1323	if (count == 1)
…		…
1165	}	1341	}
1166		1342
1167	static SV *	1343	static SV *
1168	decode_sv (dec_t *dec)	1344	decode_sv (dec_t *dec)
1169	{	1345	{
1170	decode_ws (dec);
1171
1172	// the beauty of JSON: you need exactly one character lookahead	1346	// the beauty of JSON: you need exactly one character lookahead
1173	// to parse anything.	1347	// to parse everything.
1174	switch (*dec->cur)	1348	switch (*dec->cur)
1175	{	1349	{
1176	case '"': ++dec->cur; return decode_str (dec);	1350	case '"': ++dec->cur; return decode_str (dec);
1177	case '[': ++dec->cur; return decode_av (dec);	1351	case '[': ++dec->cur; return decode_av (dec);
1178	case '{': ++dec->cur; return decode_hv (dec);	1352	case '{': ++dec->cur; return decode_hv (dec);
1179		1353
1180	case '-':	1354	case '-':
1181	case '0': case '1': case '2': case '3': case '4':	1355	case '0': case '1': case '2': case '3': case '4':
1182	case '5': case '6': case '7': case '8': case '9':	1356	case '5': case '6': case '7': case '8': case '9':
1183	return decode_num (dec);	1357	return decode_num (dec);
1184		1358
1185	case 't':	1359	case 't':
1186	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1360	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1187	{	1361	{
1188	dec->cur += 4;	1362	dec->cur += 4;
		1363	#if JSON_SLOW
		1364	json_true = get_bool ("JSON::XS::true");
		1365	#endif
1189	return SvREFCNT_inc (json_true);	1366	return newSVsv (json_true);
1190	}	1367	}
1191	else	1368	else
1192	ERR ("'true' expected");	1369	ERR ("'true' expected");
1193		1370
1194	break;	1371	break;
1195		1372
1196	case 'f':	1373	case 'f':
1197	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1374	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1198	{	1375	{
1199	dec->cur += 5;	1376	dec->cur += 5;
		1377	#if JSON_SLOW
		1378	json_false = get_bool ("JSON::XS::false");
		1379	#endif
1200	return SvREFCNT_inc (json_false);	1380	return newSVsv (json_false);
1201	}	1381	}
1202	else	1382	else
1203	ERR ("'false' expected");	1383	ERR ("'false' expected");
1204		1384
1205	break;	1385	break;
…		…
1223	fail:	1403	fail:
1224	return 0;	1404	return 0;
1225	}	1405	}
1226		1406
1227	static SV *	1407	static SV *
1228	decode_json (SV string, JSON json, UV *offset_return)	1408	decode_json (SV string, JSON json, STRLEN *offset_return)
1229	{	1409	{
1230	dec_t dec;	1410	dec_t dec;
1231	UV offset;	1411	STRLEN offset;
1232	SV *sv;	1412	SV *sv;
1233		1413
1234	SvGETMAGIC (string);	1414	SvGETMAGIC (string);
1235	SvUPGRADE (string, SVt_PV);	1415	SvUPGRADE (string, SVt_PV);
1236		1416
…		…
1254		1434
1255	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1435	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1256	dec.json.flags \|= F_HOOK;	1436	dec.json.flags \|= F_HOOK;
1257		1437
1258	*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);
1259	sv = decode_sv (&dec);	1441	sv = decode_sv (&dec);
1260		1442
1261	if (!(offset_return \|\| !sv))	1443	if (!(offset_return \|\| !sv))
1262	{	1444	{
1263	// check for trailing garbage	1445	// check for trailing garbage
…		…
1307		1489
1308	return sv;	1490	return sv;
1309	}	1491	}
1310		1492
1311	/////////////////////////////////////////////////////////////////////////////	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	/////////////////////////////////////////////////////////////////////////////
1312	// XS interface functions	1610	// XS interface functions
1313		1611
1314	MODULE = JSON::XS PACKAGE = JSON::XS	1612	MODULE = JSON::XS PACKAGE = JSON::XS
1315		1613
1316	BOOT:	1614	BOOT:
…		…
1325	: -1;	1623	: -1;
1326		1624
1327	json_stash = gv_stashpv ("JSON::XS" , 1);	1625	json_stash = gv_stashpv ("JSON::XS" , 1);
1328	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);	1626	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
1329		1627
1330	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );	1628	json_true = get_bool ("JSON::XS::true");
1331	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1629	json_false = get_bool ("JSON::XS::false");
1332	}	1630	}
1333		1631
1334	PROTOTYPES: DISABLE	1632	PROTOTYPES: DISABLE
		1633
		1634	void CLONE (...)
		1635	CODE:
		1636	json_stash = 0;
		1637	json_boolean_stash = 0;
1335		1638
1336	void new (char *klass)	1639	void new (char *klass)
1337	PPCODE:	1640	PPCODE:
1338	{	1641	{
1339	SV *pv = NEWSV (0, sizeof (JSON));	1642	SV *pv = NEWSV (0, sizeof (JSON));
1340	SvPOK_only (pv);	1643	SvPOK_only (pv);
1341	Zero (SvPVX (pv), 1, JSON);	1644	Zero (SvPVX (pv), 1, JSON);
1342	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1645	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1343	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	)));
1344	}	1650	}
1345		1651
1346	void ascii (JSON *self, int enable = 1)	1652	void ascii (JSON *self, int enable = 1)
1347	ALIAS:	1653	ALIAS:
1348	ascii = F_ASCII	1654	ascii = F_ASCII
…		…
1355	pretty = F_PRETTY	1661	pretty = F_PRETTY
1356	allow_nonref = F_ALLOW_NONREF	1662	allow_nonref = F_ALLOW_NONREF
1357	shrink = F_SHRINK	1663	shrink = F_SHRINK
1358	allow_blessed = F_ALLOW_BLESSED	1664	allow_blessed = F_ALLOW_BLESSED
1359	convert_blessed = F_CONV_BLESSED	1665	convert_blessed = F_CONV_BLESSED
		1666	relaxed = F_RELAXED
		1667	allow_unknown = F_ALLOW_UNKNOWN
1360	PPCODE:	1668	PPCODE:
1361	{	1669	{
1362	if (enable)	1670	if (enable)
1363	self->flags \|= ix;	1671	self->flags \|= ix;
1364	else	1672	else
1365	self->flags &= ~ix;	1673	self->flags &= ~ix;
1366		1674
1367	XPUSHs (ST (0));	1675	XPUSHs (ST (0));
1368	}	1676	}
1369		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
1370	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1696	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1371	PPCODE:	1697	PPCODE:
1372	{	1698	{
1373	UV log2 = 0;	1699	UV log2 = 0;
1374		1700
…		…
1380	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1706	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1381		1707
1382	XPUSHs (ST (0));	1708	XPUSHs (ST (0));
1383	}	1709	}
1384		1710
		1711	U32 get_max_depth (JSON *self)
		1712	CODE:
		1713	RETVAL = DEC_DEPTH (self->flags);
		1714	OUTPUT:
		1715	RETVAL
		1716
1385	void max_size (JSON *self, UV max_size = 0)	1717	void max_size (JSON *self, UV max_size = 0)
1386	PPCODE:	1718	PPCODE:
1387	{	1719	{
1388	UV log2 = 0;	1720	UV log2 = 0;
1389		1721
…		…
1395		1727
1396	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1728	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1397		1729
1398	XPUSHs (ST (0));	1730	XPUSHs (ST (0));
1399	}	1731	}
		1732
		1733	int get_max_size (JSON *self)
		1734	CODE:
		1735	RETVAL = DEC_SIZE (self->flags);
		1736	OUTPUT:
		1737	RETVAL
1400		1738
1401	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1739	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1402	PPCODE:	1740	PPCODE:
1403	{	1741	{
1404	SvREFCNT_dec (self->cb_object);	1742	SvREFCNT_dec (self->cb_object);
…		…
1438	XPUSHs (decode_json (jsonstr, self, 0));	1776	XPUSHs (decode_json (jsonstr, self, 0));
1439		1777
1440	void decode_prefix (JSON self, SV jsonstr)	1778	void decode_prefix (JSON self, SV jsonstr)
1441	PPCODE:	1779	PPCODE:
1442	{	1780	{
1443	UV offset;	1781	STRLEN offset;
1444	EXTEND (SP, 2);	1782	EXTEND (SP, 2);
1445	PUSHs (decode_json (jsonstr, self, &offset));	1783	PUSHs (decode_json (jsonstr, self, &offset));
1446	PUSHs (sv_2mortal (newSVuv (offset)));	1784	PUSHs (sv_2mortal (newSVuv (offset)));
		1785	}
		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	}
1447	}	1860	}
1448		1861
1449	void DESTROY (JSON *self)	1862	void DESTROY (JSON *self)
1450	CODE:	1863	CODE:
1451	SvREFCNT_dec (self->cb_sk_object);	1864	SvREFCNT_dec (self->cb_sk_object);
1452	SvREFCNT_dec (self->cb_object);	1865	SvREFCNT_dec (self->cb_object);
		1866	SvREFCNT_dec (self->incr_text);
1453		1867
1454	PROTOTYPES: ENABLE	1868	PROTOTYPES: ENABLE
1455		1869
1456	void to_json (SV *scalar)	1870	void encode_json (SV *scalar)
		1871	ALIAS:
		1872	to_json_ = 0
		1873	encode_json = F_UTF8
1457	PPCODE:	1874	PPCODE:
1458	{	1875	{
1459	JSON json = { F_DEFAULT \| F_UTF8 };	1876	JSON json = { F_DEFAULT \| ix };
1460	XPUSHs (encode_json (scalar, &json));	1877	XPUSHs (encode_json (scalar, &json));
1461	}	1878	}
1462		1879
1463	void from_json (SV *jsonstr)	1880	void decode_json (SV *jsonstr)
		1881	ALIAS:
		1882	from_json_ = 0
		1883	decode_json = F_UTF8
1464	PPCODE:	1884	PPCODE:
1465	{	1885	{
1466	JSON json = { F_DEFAULT \| F_UTF8 };	1886	JSON json = { F_DEFAULT \| ix };
1467	XPUSHs (decode_json (jsonstr, &json, 0));	1887	XPUSHs (decode_json (jsonstr, &json, 0));
1468	}	1888	}
1469		1889
		1890

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.52 by root, Mon Jul 2 02:57:11 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 1.52 by root, Mon Jul 2 02:57:11 2007 UTC vs.
Revision 1.84 by root, Thu Mar 27 06:37:35 2008 UTC