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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
8	#include "stdio.h"	8	#include <stdio.h>
		9	#include <float.h>
9		10
10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	12	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	13	#endif
13		14
		15	// some old perls do not have this, try to make it work, no
		16	// guarentees, though. if it breaks, you get to keep the pieces.
		17	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
14	#define F_ASCII 0x00000001UL	21	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	22	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	23	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	24	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	25	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	26	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	27	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	28	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
23	#define F_MAXDEPTH 0xf8000000UL	34	#define F_MAXDEPTH 0xf8000000UL
24	#define S_MAXDEPTH 27	35	#define S_MAXDEPTH 27
		36	#define F_MAXSIZE 0x01f00000UL
		37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
25		39
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		42
31	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		45
34	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	47	#define INDENT_STEP 3 // spaces per indentation level
36		48
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38		50
39	#define SB do {	51	#define SB do {
40	#define SE } while (0)	52	#define SE } while (0)
41		53
42	#if __GNUC__ >= 3	54	#if __GNUC__ >= 3
43	# define expect(expr,value) __builtin_expect ((expr),(value))	55	# define expect(expr,value) __builtin_expect ((expr), (value))
44	# define inline inline	56	# define INLINE static inline
45	#else	57	#else
46	# define expect(expr,value) (expr)	58	# define expect(expr,value) (expr)
47	# define inline static	59	# define INLINE static
48	#endif	60	#endif
49		61
50	#define expect_false(expr) expect ((expr) != 0, 0)	62	#define expect_false(expr) expect ((expr) != 0, 0)
51	#define expect_true(expr) expect ((expr) != 0, 1)	63	#define expect_true(expr) expect ((expr) != 0, 1)
52		64
		65	#ifdef USE_ITHREADS
		66	# define JSON_SLOW 1
		67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		68	#else
		69	# define JSON_SLOW 0
		70	# define JSON_STASH json_stash
		71	#endif
		72
53	static HV *json_stash; // JSON::XS::	73	static HV json_stash, json_boolean_stash; // JSON::XS::
		74	static SV json_true, json_false;
		75
		76	typedef struct {
		77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
		80	} JSON;
54		81
55	/////////////////////////////////////////////////////////////////////////////	82	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	83	// utility functions
57		84
58	static UV *	85	INLINE void
59	SvJSON (SV *sv)
60	{
61	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
62	croak ("object is not of type JSON::XS");
63
64	return &SvUVX (SvRV (sv));
65	}
66
67	static void
68	shrink (SV *sv)	86	shrink (SV *sv)
69	{	87	{
70	sv_utf8_downgrade (sv, 1);	88	sv_utf8_downgrade (sv, 1);
71	if (SvLEN (sv) > SvCUR (sv) + 1)	89	if (SvLEN (sv) > SvCUR (sv) + 1)
72	{	90	{
81	// decode an utf-8 character and return it, or (UV)-1 in	99	// decode an utf-8 character and return it, or (UV)-1 in
82	// case of an error.	100	// case of an error.
83	// we special-case "safe" characters from U+80 .. U+7FF,	101	// we special-case "safe" characters from U+80 .. U+7FF,
84	// but use the very good perl function to parse anything else.	102	// but use the very good perl function to parse anything else.
85	// note that we never call this function for a ascii codepoints	103	// note that we never call this function for a ascii codepoints
86	inline UV	104	INLINE UV
87	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
88	{	106	{
89	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
90	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
91	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
…		…
107	typedef struct	125	typedef struct
108	{	126	{
109	char *cur; // SvPVX (sv) + current output position	127	char *cur; // SvPVX (sv) + current output position
110	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
111	SV *sv; // result scalar	129	SV *sv; // result scalar
112	U32 flags; // F_*	130	JSON json;
113	U32 indent; // indentation level	131	U32 indent; // indentation level
114	U32 maxdepth; // max. indentation/recursion level	132	U32 maxdepth; // max. indentation/recursion level
		133	UV limit; // escape character values >= this value when encoding
115	} enc_t;	134	} enc_t;
116		135
117	inline void	136	INLINE void
118	need (enc_t *enc, STRLEN len)	137	need (enc_t *enc, STRLEN len)
119	{	138	{
120	if (expect_false (enc->cur + len >= enc->end))	139	if (expect_false (enc->cur + len >= enc->end))
121	{	140	{
122	STRLEN cur = enc->cur - SvPVX (enc->sv);	141	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
124	enc->cur = SvPVX (enc->sv) + cur;	143	enc->cur = SvPVX (enc->sv) + cur;
125	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	144	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
126	}	145	}
127	}	146	}
128		147
129	inline void	148	INLINE void
130	encode_ch (enc_t *enc, char ch)	149	encode_ch (enc_t *enc, char ch)
131	{	150	{
132	need (enc, 1);	151	need (enc, 1);
133	*enc->cur++ = ch;	152	*enc->cur++ = ch;
134	}	153	}
…		…
178	STRLEN clen;	197	STRLEN clen;
179	UV uch;	198	UV uch;
180		199
181	if (is_utf8)	200	if (is_utf8)
182	{	201	{
183	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
184	uch = decode_utf8 (str, end - str, &clen);	202	uch = decode_utf8 (str, end - str, &clen);
185	if (clen == (STRLEN)-1)	203	if (clen == (STRLEN)-1)
186	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	204	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
187	}	205	}
188	else	206	else
189	{	207	{
190	uch = ch;	208	uch = ch;
191	clen = 1;	209	clen = 1;
192	}	210	}
193		211
194	if (uch > 0x10FFFFUL)	212	if (uch < 0x20 \|\| uch >= enc->limit)
195	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
196
197	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))
198	{	213	{
199	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
200	{	215	{
		216	if (uch > 0x10FFFFUL)
		217	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		218
201	need (enc, len += 11);	219	need (enc, len += 11);
202	sprintf (enc->cur, "\\u%04x\\u%04x",	220	sprintf (enc->cur, "\\u%04x\\u%04x",
203	(int)((uch - 0x10000) / 0x400 + 0xD800),	221	(int)((uch - 0x10000) / 0x400 + 0xD800),
204	(int)((uch - 0x10000) % 0x400 + 0xDC00));	222	(int)((uch - 0x10000) % 0x400 + 0xDC00));
205	enc->cur += 12;	223	enc->cur += 12;
…		…
216	*enc->cur++ = hexdigit [(uch >> 0) & 15];	234	*enc->cur++ = hexdigit [(uch >> 0) & 15];
217	}	235	}
218		236
219	str += clen;	237	str += clen;
220	}	238	}
221	else if (enc->flags & F_LATIN1)	239	else if (enc->json.flags & F_LATIN1)
222	{	240	{
223	*enc->cur++ = uch;	241	*enc->cur++ = uch;
224	str += clen;	242	str += clen;
225	}	243	}
226	else if (is_utf8)	244	else if (is_utf8)
…		…
244		262
245	--len;	263	--len;
246	}	264	}
247	}	265	}
248		266
249	inline void	267	INLINE void
250	encode_indent (enc_t *enc)	268	encode_indent (enc_t *enc)
251	{	269	{
252	if (enc->flags & F_INDENT)	270	if (enc->json.flags & F_INDENT)
253	{	271	{
254	int spaces = enc->indent * INDENT_STEP;	272	int spaces = enc->indent * INDENT_STEP;
255		273
256	need (enc, spaces);	274	need (enc, spaces);
257	memset (enc->cur, ' ', spaces);	275	memset (enc->cur, ' ', spaces);
258	enc->cur += spaces;	276	enc->cur += spaces;
259	}	277	}
260	}	278	}
261		279
262	inline void	280	INLINE void
263	encode_space (enc_t *enc)	281	encode_space (enc_t *enc)
264	{	282	{
265	need (enc, 1);	283	need (enc, 1);
266	encode_ch (enc, ' ');	284	encode_ch (enc, ' ');
267	}	285	}
268		286
269	inline void	287	INLINE void
270	encode_nl (enc_t *enc)	288	encode_nl (enc_t *enc)
271	{	289	{
272	if (enc->flags & F_INDENT)	290	if (enc->json.flags & F_INDENT)
273	{	291	{
274	need (enc, 1);	292	need (enc, 1);
275	encode_ch (enc, '\n');	293	encode_ch (enc, '\n');
276	}	294	}
277	}	295	}
278		296
279	inline void	297	INLINE void
280	encode_comma (enc_t *enc)	298	encode_comma (enc_t *enc)
281	{	299	{
282	encode_ch (enc, ',');	300	encode_ch (enc, ',');
283		301
284	if (enc->flags & F_INDENT)	302	if (enc->json.flags & F_INDENT)
285	encode_nl (enc);	303	encode_nl (enc);
286	else if (enc->flags & F_SPACE_AFTER)	304	else if (enc->json.flags & F_SPACE_AFTER)
287	encode_space (enc);	305	encode_space (enc);
288	}	306	}
289		307
290	static void encode_sv (enc_t enc, SV sv);	308	static void encode_sv (enc_t enc, SV sv);
291		309
…		…
295	int i, len = av_len (av);	313	int i, len = av_len (av);
296		314
297	if (enc->indent >= enc->maxdepth)	315	if (enc->indent >= enc->maxdepth)
298	croak ("data structure too deep (hit recursion limit)");	316	croak ("data structure too deep (hit recursion limit)");
299		317
300	encode_ch (enc, '['); encode_nl (enc);	318	encode_ch (enc, '[');
301	++enc->indent;	319
		320	if (len >= 0)
		321	{
		322	encode_nl (enc); ++enc->indent;
302		323
303	for (i = 0; i <= len; ++i)	324	for (i = 0; i <= len; ++i)
304	{	325	{
		326	SV **svp = av_fetch (av, i, 0);
		327
305	encode_indent (enc);	328	encode_indent (enc);
306	encode_sv (enc, *av_fetch (av, i, 0));
307		329
		330	if (svp)
		331	encode_sv (enc, *svp);
		332	else
		333	encode_str (enc, "null", 4, 0);
		334
308	if (i < len)	335	if (i < len)
309	encode_comma (enc);	336	encode_comma (enc);
310	}	337	}
311		338
		339	encode_nl (enc); --enc->indent; encode_indent (enc);
		340	}
		341
312	encode_nl (enc);	342	encode_ch (enc, ']');
313
314	--enc->indent;
315	encode_indent (enc); encode_ch (enc, ']');
316	}	343	}
317		344
318	static void	345	static void
319	encode_he (enc_t enc, HE he)	346	encode_hk (enc_t enc, HE he)
320	{	347	{
321	encode_ch (enc, '"');	348	encode_ch (enc, '"');
322		349
323	if (HeKLEN (he) == HEf_SVKEY)	350	if (HeKLEN (he) == HEf_SVKEY)
324	{	351	{
…		…
334	else	361	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	362	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		363
337	encode_ch (enc, '"');	364	encode_ch (enc, '"');
338		365
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	366	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	367	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	368	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));
343	}	369	}
344		370
345	// compare hash entries, used when all keys are bytestrings	371	// compare hash entries, used when all keys are bytestrings
346	static int	372	static int
347	he_cmp_fast (const void a_, const void b_)	373	he_cmp_fast (const void a_, const void b_)
…		…
352	HE b = (HE **)b_;	378	HE b = (HE **)b_;
353		379
354	STRLEN la = HeKLEN (a);	380	STRLEN la = HeKLEN (a);
355	STRLEN lb = HeKLEN (b);	381	STRLEN lb = HeKLEN (b);
356		382
357	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	383	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
358	cmp = la - lb;	384	cmp = lb - la;
359		385
360	return cmp;	386	return cmp;
361	}	387	}
362		388
363	// compare hash entries, used when some keys are sv's or utf-x	389	// compare hash entries, used when some keys are sv's or utf-x
364	static int	390	static int
365	he_cmp_slow (const void a, const void b)	391	he_cmp_slow (const void a, const void b)
366	{	392	{
367	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	393	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
368	}	394	}
369		395
370	static void	396	static void
371	encode_hv (enc_t enc, HV hv)	397	encode_hv (enc_t enc, HV hv)
372	{	398	{
		399	HE *he;
373	int count, i;	400	int count;
374		401
375	if (enc->indent >= enc->maxdepth)	402	if (enc->indent >= enc->maxdepth)
376	croak ("data structure too deep (hit recursion limit)");	403	croak ("data structure too deep (hit recursion limit)");
377		404
378	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	405	encode_ch (enc, '{');
379		406
380	if ((count = hv_iterinit (hv)))
381	{
382	// for canonical output we have to sort by keys first	407	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	408	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	409	// security workaround added somewhere in 5.8.x.
385	// that randomises hash orderings	410	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	411	if (enc->json.flags & F_CANONICAL)
		412	{
		413	int count = hv_iterinit (hv);
		414
		415	if (SvMAGICAL (hv))
387	{	416	{
		417	// need to count by iterating. could improve by dynamically building the vector below
		418	// but I don't care for the speed of this special case.
		419	// note also that we will run into undefined behaviour when the two iterations
		420	// do not result in the same count, something I might care for in some later release.
		421
		422	count = 0;
		423	while (hv_iternext (hv))
		424	++count;
		425
		426	hv_iterinit (hv);
		427	}
		428
		429	if (count)
		430	{
388	int fast = 1;	431	int i, fast = 1;
389	HE *he;
390	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	432	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
391	HE *hes = _alloca (count sizeof (HE));	433	HE *hes = _alloca (count sizeof (HE));
392	#else	434	#else
393	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	435	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394	#endif	436	#endif
…		…
421		463
422	FREETMPS;	464	FREETMPS;
423	LEAVE;	465	LEAVE;
424	}	466	}
425		467
426	for (i = 0; i < count; ++i)	468	encode_nl (enc); ++enc->indent;
		469
		470	while (count--)
427	{	471	{
428	encode_indent (enc);	472	encode_indent (enc);
		473	he = hes [count];
429	encode_he (enc, hes [i]);	474	encode_hk (enc, he);
		475	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
430		476
431	if (i < count - 1)	477	if (count)
432	encode_comma (enc);	478	encode_comma (enc);
433	}	479	}
434		480
		481	encode_nl (enc); --enc->indent; encode_indent (enc);
		482	}
		483	}
		484	else
		485	{
		486	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		487	if ((he = hv_iternext (hv)))
		488	{
		489	encode_nl (enc); ++enc->indent;
		490
		491	for (;;)
		492	{
435	encode_nl (enc);	493	encode_indent (enc);
		494	encode_hk (enc, he);
		495	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		496
		497	if (!(he = hv_iternext (hv)))
		498	break;
		499
		500	encode_comma (enc);
		501	}
		502
		503	encode_nl (enc); --enc->indent; encode_indent (enc);
		504	}
		505	}
		506
		507	encode_ch (enc, '}');
		508	}
		509
		510	// encode objects, arrays and special \0=false and \1=true values.
		511	static void
		512	encode_rv (enc_t enc, SV sv)
		513	{
		514	svtype svt;
		515
		516	SvGETMAGIC (sv);
		517	svt = SvTYPE (sv);
		518
		519	if (expect_false (SvOBJECT (sv)))
		520	{
		521	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		522	? json_boolean_stash
		523	: gv_stashpv ("JSON::XS::Boolean", 1);
		524
		525	if (SvSTASH (sv) == stash)
		526	{
		527	if (SvIV (sv))
		528	encode_str (enc, "true", 4, 0);
		529	else
		530	encode_str (enc, "false", 5, 0);
436	}	531	}
437	else	532	else
438	{	533	{
439	HE *he = hv_iternext (hv);	534	#if 0
440		535	if (0 && sv_derived_from (rv, "JSON::Literal"))
441	for (;;)
442	{	536	{
443	encode_indent (enc);	537	// not yet
444	encode_he (enc, he);
445
446	if (!(he = hv_iternext (hv)))
447	break;
448
449	encode_comma (enc);
450	}	538	}
		539	#endif
		540	if (enc->json.flags & F_CONV_BLESSED)
		541	{
		542	// we re-bless the reference to get overload and other niceties right
		543	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
451		544
		545	if (to_json)
		546	{
		547	dSP;
		548
		549	ENTER; SAVETMPS; PUSHMARK (SP);
		550	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		551
		552	// calling with G_SCALAR ensures that we always get a 1 return value
		553	PUTBACK;
		554	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		555	SPAGAIN;
		556
		557	// catch this surprisingly common error
		558	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		559	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		560
		561	sv = POPs;
		562	PUTBACK;
		563
452	encode_nl (enc);	564	encode_sv (enc, sv);
		565
		566	FREETMPS; LEAVE;
		567	}
		568	else if (enc->json.flags & F_ALLOW_BLESSED)
		569	encode_str (enc, "null", 4, 0);
		570	else
		571	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		572	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		573	}
		574	else if (enc->json.flags & F_ALLOW_BLESSED)
		575	encode_str (enc, "null", 4, 0);
		576	else
		577	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		578	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
453	}	579	}
454	}	580	}
455
456	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
457	}
458
459	// encode objects, arrays and special \0=false and \1=true values.
460	static void
461	encode_rv (enc_t enc, SV sv)
462	{
463	svtype svt;
464
465	SvGETMAGIC (sv);
466	svt = SvTYPE (sv);
467
468	if (svt == SVt_PVHV)	581	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	582	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	583	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	584	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	585	else if (svt < SVt_PVAV)
473	{	586	{
474	if (SvNIOK (sv) && SvIV (sv) == 0)	587	STRLEN len = 0;
		588	char *pv = svt ? SvPV (sv, len) : 0;
		589
		590	if (len == 1 && *pv == '1')
		591	encode_str (enc, "true", 4, 0);
		592	else if (len == 1 && *pv == '0')
475	encode_str (enc, "false", 5, 0);	593	encode_str (enc, "false", 5, 0);
476	else if (SvNIOK (sv) && SvIV (sv) == 1)
477	encode_str (enc, "true", 4, 0);
478	else	594	else
479	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	595	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
480	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
481	}	597	}
482	else	598	else
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	613	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	614	encode_ch (enc, '"');
499	}	615	}
500	else if (SvNOKp (sv))	616	else if (SvNOKp (sv))
501	{	617	{
		618	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	619	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	620	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	621	enc->cur += strlen (enc->cur);
505	}	622	}
506	else if (SvIOKp (sv))	623	else if (SvIOKp (sv))
507	{	624	{
508	// we assume we can always read an IV as a UV	625	// we assume we can always read an IV as a UV
509	if (SvUV (sv) & ~(UV)0x7fff)	626	if (SvUV (sv) & ~(UV)0x7fff)
510	{	627	{
		628	// large integer, use the (rather slow) snprintf way.
511	need (enc, 32);	629	need (enc, sizeof (UV) * 3);
512	enc->cur +=	630	enc->cur +=
513	SvIsUV(sv)	631	SvIsUV(sv)
514	? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv))	632	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
515	: snprintf (enc->cur, 32, "%"IVdf, (IV)SvIVX (sv));	633	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
516	}	634	}
517	else	635	else
518	{	636	{
519	// optimise the "small number case"	637	// optimise the "small number case"
520	// code will likely be branchless and use only a single multiplication	638	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);	639	I32 i = SvIV (sv);
522	U32 u;	640	U32 u;
		641	char digit, nz = 0;
523		642
524	need (enc, 6);	643	need (enc, 6);
525		644
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	645	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
527	u = i < 0 ? -i : i;	646	u = i < 0 ? -i : i;
528		647
529	// convert to 4.28 fixed-point representation	648	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits	649	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531		650
532	char digit, nz = 0;	651	// now output digit by digit, each time masking out the integer part
533		652	// and multiplying by 5 while moving the decimal point one to the right,
		653	// resulting in a net multiplication by 10.
		654	// we always write the digit to memory but conditionally increment
		655	// the pointer, to ease the usage of conditional move instructions.
534	digit = u >> 28; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) 5;	656	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
535	digit = u >> 27; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) 5;	657	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
536	digit = u >> 26; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) 5;	658	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
537	digit = u >> 25; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) 5;	659	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
538	digit = u >> 24; *enc->cur = digit + '0'; nz \|= digit; enc->cur += 1;	660	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
539	}	661	}
540	}	662	}
541	else if (SvROK (sv))	663	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	664	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	665	else if (!SvOK (sv))
…		…
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	668	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	669	SvPV_nolen (sv), SvFLAGS (sv));
548	}	670	}
549		671
550	static SV *	672	static SV *
551	encode_json (SV *scalar, U32 flags)	673	encode_json (SV scalar, JSON json)
552	{	674	{
553	enc_t enc;	675	enc_t enc;
554		676
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	677	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	678	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		679
558	enc.flags = flags;	680	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	681	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	682	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	683	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	684	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	685	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		686	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		687	: enc.json.flags & F_LATIN1 ? 0x000100UL
		688	: 0x10FFFFUL;
564		689
565	SvPOK_only (enc.sv);	690	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	691	encode_sv (&enc, scalar);
567		692
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	693	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	694	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		695
571	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	696	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
572	SvUTF8_on (enc.sv);	697	SvUTF8_on (enc.sv);
573		698
574	if (enc.flags & F_SHRINK)	699	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	700	shrink (enc.sv);
576		701
577	return enc.sv;	702	return enc.sv;
578	}	703	}
579		704
…		…
584	typedef struct	709	typedef struct
585	{	710	{
586	char *cur; // current parser pointer	711	char *cur; // current parser pointer
587	char *end; // end of input string	712	char *end; // end of input string
588	const char *err; // parse error, if != 0	713	const char *err; // parse error, if != 0
589	U32 flags; // F_*	714	JSON json;
590	U32 depth; // recursion depth	715	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	716	U32 maxdepth; // recursion depth limit
592	} dec_t;	717	} dec_t;
593		718
594	inline void	719	INLINE void
		720	decode_comment (dec_t *dec)
		721	{
		722	// only '#'-style comments allowed a.t.m.
		723
		724	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		725	++dec->cur;
		726	}
		727
		728	INLINE void
595	decode_ws (dec_t *dec)	729	decode_ws (dec_t *dec)
596	{	730	{
597	for (;;)	731	for (;;)
598	{	732	{
599	char ch = *dec->cur;	733	char ch = *dec->cur;
600		734
601	if (ch > 0x20	735	if (ch > 0x20)
		736	{
		737	if (expect_false (ch == '#'))
		738	{
		739	if (dec->json.flags & F_RELAXED)
		740	decode_comment (dec);
		741	else
		742	break;
		743	}
		744	else
		745	break;
		746	}
602	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	747	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
603	break;	748	break; // parse error, but let higher level handle it, gives better error messages
604		749
605	++dec->cur;	750	++dec->cur;
606	}	751	}
607	}	752	}
608		753
…		…
625	decode_4hex (dec_t *dec)	770	decode_4hex (dec_t *dec)
626	{	771	{
627	signed char d1, d2, d3, d4;	772	signed char d1, d2, d3, d4;
628	unsigned char cur = (unsigned char )dec->cur;	773	unsigned char cur = (unsigned char )dec->cur;
629		774
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	775	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
631	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected");	776	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
632	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected");	777	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
633	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected");	778	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		779
635	dec->cur += 4;	780	dec->cur += 4;
636		781
637	return ((UV)d1) << 12	782	return ((UV)d1) << 12
638	\| ((UV)d2) << 8	783	\| ((UV)d2) << 8
…		…
646	static SV *	791	static SV *
647	decode_str (dec_t *dec)	792	decode_str (dec_t *dec)
648	{	793	{
649	SV *sv = 0;	794	SV *sv = 0;
650	int utf8 = 0;	795	int utf8 = 0;
		796	char *dec_cur = dec->cur;
651		797
652	do	798	do
653	{	799	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	800	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	801	char *cur = buf;
656		802
657	do	803	do
658	{	804	{
659	unsigned char ch = (unsigned char )dec->cur++;	805	unsigned char ch = (unsigned char )dec_cur++;
660		806
661	if (expect_false (ch == '"'))	807	if (expect_false (ch == '"'))
662	{	808	{
663	--dec->cur;	809	--dec_cur;
664	break;	810	break;
665	}	811	}
666	else if (expect_false (ch == '\\'))	812	else if (expect_false (ch == '\\'))
667	{	813	{
668	switch (*dec->cur)	814	switch (*dec_cur)
669	{	815	{
670	case '\\':	816	case '\\':
671	case '/':	817	case '/':
672	case '"': cur++ = dec->cur++; break;	818	case '"': cur++ = dec_cur++; break;
673		819
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	820	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	821	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	822	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	823	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	824	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		825
680	case 'u':	826	case 'u':
681	{	827	{
682	UV lo, hi;	828	UV lo, hi;
683	++dec->cur;	829	++dec_cur;
684		830
		831	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	832	hi = decode_4hex (dec);
		833	dec_cur = dec->cur;
686	if (hi == (UV)-1)	834	if (hi == (UV)-1)
687	goto fail;	835	goto fail;
688		836
689	// possibly a surrogate pair	837	// possibly a surrogate pair
690	if (hi >= 0xd800)	838	if (hi >= 0xd800)
691	if (hi < 0xdc00)	839	if (hi < 0xdc00)
692	{	840	{
693	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	841	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	842	ERR ("missing low surrogate character in surrogate pair");
695		843
696	dec->cur += 2;	844	dec_cur += 2;
697		845
		846	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	847	lo = decode_4hex (dec);
		848	dec_cur = dec->cur;
699	if (lo == (UV)-1)	849	if (lo == (UV)-1)
700	goto fail;	850	goto fail;
701		851
702	if (lo < 0xdc00 \|\| lo >= 0xe000)	852	if (lo < 0xdc00 \|\| lo >= 0xe000)
703	ERR ("surrogate pair expected");	853	ERR ("surrogate pair expected");
…		…
717	*cur++ = hi;	867	*cur++ = hi;
718	}	868	}
719	break;	869	break;
720		870
721	default:	871	default:
722	--dec->cur;	872	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	873	ERR ("illegal backslash escape sequence in string");
724	}	874	}
725	}	875	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	876	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
727	*cur++ = ch;	877	*cur++ = ch;
728	else if (ch >= 0x80)	878	else if (ch >= 0x80)
729	{	879	{
730	STRLEN clen;	880	STRLEN clen;
731	UV uch;	881	UV uch;
732		882
733	--dec->cur;	883	--dec_cur;
734		884
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	885	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	886	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	887	ERR ("malformed UTF-8 character in JSON string");
738		888
739	do	889	do
740	cur++ = dec->cur++;	890	cur++ = dec_cur++;
741	while (--clen);	891	while (--clen);
742		892
743	utf8 = 1;	893	utf8 = 1;
744	}	894	}
745	else	895	else
746	{	896	{
747	--dec->cur;	897	--dec_cur;
748		898
749	if (!ch)	899	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	900	ERR ("unexpected end of string while parsing JSON string");
751	else	901	else
752	ERR ("invalid character encountered while parsing JSON string");	902	ERR ("invalid character encountered while parsing JSON string");
…		…
765	}	915	}
766	else	916	else
767	sv = newSVpvn (buf, len);	917	sv = newSVpvn (buf, len);
768	}	918	}
769	}	919	}
770	while (*dec->cur != '"');	920	while (*dec_cur != '"');
771		921
772	++dec->cur;	922	++dec_cur;
773		923
774	if (sv)	924	if (sv)
775	{	925	{
776	SvPOK_only (sv);	926	SvPOK_only (sv);
777	*SvEND (sv) = 0;	927	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	930	SvUTF8_on (sv);
781	}	931	}
782	else	932	else
783	sv = newSVpvn ("", 0);	933	sv = newSVpvn ("", 0);
784		934
		935	dec->cur = dec_cur;
785	return sv;	936	return sv;
786		937
787	fail:	938	fail:
		939	dec->cur = dec_cur;
788	return 0;	940	return 0;
789	}	941	}
790		942
791	static SV *	943	static SV *
792	decode_num (dec_t *dec)	944	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	1002	is_nv = 1;
851	}	1003	}
852		1004
853	if (!is_nv)	1005	if (!is_nv)
854	{	1006	{
		1007	int len = dec->cur - start;
		1008
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1009	// special case the rather common 1..5-digit-int case
856	if (*start == '-')	1010	if (*start == '-')
857	switch (dec->cur - start)	1011	switch (len)
858	{	1012	{
859	case 2: return newSViv (-( start [1] - '0' ));	1013	case 2: return newSViv (-( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1014	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
861	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1015	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
862	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1016	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1017	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
863	}	1018	}
864	else	1019	else
865	switch (dec->cur - start)	1020	switch (len)
866	{	1021	{
867	case 1: return newSViv ( start [0] - '0' );	1022	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1023	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
869	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1024	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
870	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1025	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1026	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
871	}	1027	}
872		1028
873	{	1029	{
874	UV uv;	1030	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	1031	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	1032	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	1033	if (numtype & IS_NUMBER_NEG)
878	{	1034	{
879	if (uv < (UV)IV_MIN)	1035	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1036	return newSViv (-(IV)uv);
881	}	1037	}
882	else	1038	else
883	return newSVuv (uv);	1039	return newSVuv (uv);
884	}	1040	}
885	}
886		1041
		1042	len -= *start == '-' ? 1 : 0;
		1043
		1044	// does not fit into IV or UV, try NV
		1045	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1046	#if defined (LDBL_DIG)
		1047	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1048	#endif
		1049	)
		1050	// fits into NV without loss of precision
		1051	return newSVnv (Atof (start));
		1052
		1053	// everything else fails, convert it to a string
		1054	return newSVpvn (start, dec->cur - start);
		1055	}
		1056
		1057	// loss of precision here
887	return newSVnv (Atof (start));	1058	return newSVnv (Atof (start));
888		1059
889	fail:	1060	fail:
890	return 0;	1061	return 0;
891	}	1062	}
…		…
921		1092
922	if (*dec->cur != ',')	1093	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1094	ERR (", or ] expected while parsing array");
924		1095
925	++dec->cur;	1096	++dec->cur;
		1097
		1098	decode_ws (dec);
		1099
		1100	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1101	{
		1102	++dec->cur;
		1103	break;
		1104	}
926	}	1105	}
927		1106
928	DEC_DEC_DEPTH;	1107	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1108	return newRV_noinc ((SV *)av);
930		1109
…		…
935	}	1114	}
936		1115
937	static SV *	1116	static SV *
938	decode_hv (dec_t *dec)	1117	decode_hv (dec_t *dec)
939	{	1118	{
		1119	SV *sv;
940	HV *hv = newHV ();	1120	HV *hv = newHV ();
941		1121
942	DEC_INC_DEPTH;	1122	DEC_INC_DEPTH;
943	decode_ws (dec);	1123	decode_ws (dec);
944		1124
945	if (*dec->cur == '}')	1125	if (*dec->cur == '}')
946	++dec->cur;	1126	++dec->cur;
947	else	1127	else
948	for (;;)	1128	for (;;)
949	{	1129	{
		1130	EXPECT_CH ('"');
		1131
		1132	// heuristic: assume that
		1133	// a) decode_str + hv_store_ent are abysmally slow.
		1134	// b) most hash keys are short, simple ascii text.
		1135	// => try to "fast-match" such strings to avoid
		1136	// the overhead of decode_str + hv_store_ent.
		1137	{
950	SV key, value;	1138	SV *value;
		1139	char *p = dec->cur;
		1140	char *e = p + 24; // only try up to 24 bytes
951		1141
952	decode_ws (dec); EXPECT_CH ('"');	1142	for (;;)
953
954	key = decode_str (dec);
955	if (!key)
956	goto fail;
957
958	decode_ws (dec); EXPECT_CH (':');
959
960	value = decode_sv (dec);
961	if (!value)
962	{	1143	{
		1144	// the >= 0x80 is true on most architectures
		1145	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1146	{
		1147	// slow path, back up and use decode_str
		1148	SV *key = decode_str (dec);
		1149	if (!key)
		1150	goto fail;
		1151
		1152	decode_ws (dec); EXPECT_CH (':');
		1153
		1154	decode_ws (dec);
		1155	value = decode_sv (dec);
		1156	if (!value)
		1157	{
		1158	SvREFCNT_dec (key);
		1159	goto fail;
		1160	}
		1161
		1162	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1163	SvREFCNT_dec (key);
		1164
		1165	break;
		1166	}
		1167	else if (*p == '"')
		1168	{
		1169	// fast path, got a simple key
		1170	char *key = dec->cur;
		1171	int len = p - key;
		1172	dec->cur = p + 1;
		1173
		1174	decode_ws (dec); EXPECT_CH (':');
		1175
		1176	decode_ws (dec);
		1177	value = decode_sv (dec);
		1178	if (!value)
964	goto fail;	1179	goto fail;
		1180
		1181	hv_store (hv, key, len, value, 0);
		1182
		1183	break;
		1184	}
		1185
		1186	++p;
965	}	1187	}
966		1188	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1189
970	decode_ws (dec);	1190	decode_ws (dec);
971		1191
972	if (*dec->cur == '}')	1192	if (*dec->cur == '}')
973	{	1193	{
…		…
977		1197
978	if (*dec->cur != ',')	1198	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1199	ERR (", or } expected while parsing object/hash");
980		1200
981	++dec->cur;	1201	++dec->cur;
		1202
		1203	decode_ws (dec);
		1204
		1205	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1206	{
		1207	++dec->cur;
		1208	break;
		1209	}
982	}	1210	}
983		1211
984	DEC_DEC_DEPTH;	1212	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1213	sv = newRV_noinc ((SV *)hv);
		1214
		1215	// check filter callbacks
		1216	if (dec->json.flags & F_HOOK)
		1217	{
		1218	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1219	{
		1220	HE cb, he;
		1221
		1222	hv_iterinit (hv);
		1223	he = hv_iternext (hv);
		1224	hv_iterinit (hv);
		1225
		1226	// the next line creates a mortal sv each time its called.
		1227	// might want to optimise this for common cases.
		1228	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1229
		1230	if (cb)
		1231	{
		1232	dSP;
		1233	int count;
		1234
		1235	ENTER; SAVETMPS; PUSHMARK (SP);
		1236	XPUSHs (HeVAL (he));
		1237
		1238	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1239
		1240	if (count == 1)
		1241	{
		1242	sv = newSVsv (POPs);
		1243	FREETMPS; LEAVE;
		1244	return sv;
		1245	}
		1246
		1247	FREETMPS; LEAVE;
		1248	}
		1249	}
		1250
		1251	if (dec->json.cb_object)
		1252	{
		1253	dSP;
		1254	int count;
		1255
		1256	ENTER; SAVETMPS; PUSHMARK (SP);
		1257	XPUSHs (sv_2mortal (sv));
		1258
		1259	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1260
		1261	if (count == 1)
		1262	{
		1263	sv = newSVsv (POPs);
		1264	FREETMPS; LEAVE;
		1265	return sv;
		1266	}
		1267
		1268	SvREFCNT_inc (sv);
		1269	FREETMPS; LEAVE;
		1270	}
		1271	}
		1272
		1273	return sv;
986		1274
987	fail:	1275	fail:
988	SvREFCNT_dec (hv);	1276	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1277	DEC_DEC_DEPTH;
990	return 0;	1278	return 0;
991	}	1279	}
992		1280
993	static SV *	1281	static SV *
994	decode_sv (dec_t *dec)	1282	decode_sv (dec_t *dec)
995	{	1283	{
996	decode_ws (dec);	1284	// the beauty of JSON: you need exactly one character lookahead
		1285	// to parse anything.
997	switch (*dec->cur)	1286	switch (*dec->cur)
998	{	1287	{
999	case '"': ++dec->cur; return decode_str (dec);	1288	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1289	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1290	case '{': ++dec->cur; return decode_hv (dec);
…		…
1007		1296
1008	case 't':	1297	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1298	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1299	{
1011	dec->cur += 4;	1300	dec->cur += 4;
1012	return newSViv (1);	1301	#if JSON_SLOW
		1302	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1303	#endif
		1304	return SvREFCNT_inc (json_true);
1013	}	1305	}
1014	else	1306	else
1015	ERR ("'true' expected");	1307	ERR ("'true' expected");
1016		1308
1017	break;	1309	break;
1018		1310
1019	case 'f':	1311	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1312	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1313	{
1022	dec->cur += 5;	1314	dec->cur += 5;
1023	return newSViv (0);	1315	#if JSON_SLOW
		1316	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1317	#endif
		1318	return SvREFCNT_inc (json_false);
1024	}	1319	}
1025	else	1320	else
1026	ERR ("'false' expected");	1321	ERR ("'false' expected");
1027		1322
1028	break;	1323	break;
…		…
1046	fail:	1341	fail:
1047	return 0;	1342	return 0;
1048	}	1343	}
1049		1344
1050	static SV *	1345	static SV *
1051	decode_json (SV string, U32 flags, UV offset_return)	1346	decode_json (SV string, JSON json, UV *offset_return)
1052	{	1347	{
1053	dec_t dec;	1348	dec_t dec;
1054	UV offset;	1349	UV offset;
1055	SV *sv;	1350	SV *sv;
1056		1351
1057	SvGETMAGIC (string);	1352	SvGETMAGIC (string);
1058	SvUPGRADE (string, SVt_PV);	1353	SvUPGRADE (string, SVt_PV);
1059		1354
		1355	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1356	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1357	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1358
1060	if (flags & F_UTF8)	1359	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1360	sv_utf8_downgrade (string, 0);
1062	else	1361	else
1063	sv_utf8_upgrade (string);	1362	sv_utf8_upgrade (string);
1064		1363
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1364	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1365
1067	dec.flags = flags;	1366	dec.json = *json;
1068	dec.cur = SvPVX (string);	1367	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1368	dec.end = SvEND (string);
1070	dec.err = 0;	1369	dec.err = 0;
1071	dec.depth = 0;	1370	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1371	dec.maxdepth = DEC_DEPTH (dec.json.flags);
		1372
		1373	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1374	dec.json.flags \|= F_HOOK;
1073		1375
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1376	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1377
		1378	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1379	sv = decode_sv (&dec);
1076		1380
1077	if (!(offset_return \|\| !sv))	1381	if (!(offset_return \|\| !sv))
1078	{	1382	{
1079	// check for trailing garbage	1383	// check for trailing garbage
…		…
1087	}	1391	}
1088	}	1392	}
1089		1393
1090	if (offset_return \|\| !sv)	1394	if (offset_return \|\| !sv)
1091	{	1395	{
1092	offset = dec.flags & F_UTF8	1396	offset = dec.json.flags & F_UTF8
1093	? dec.cur - SvPVX (string)	1397	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));	1398	: utf8_distance (dec.cur, SvPVX (string));
1095		1399
1096	if (offset_return)	1400	if (offset_return)
1097	*offset_return = offset;	1401	*offset_return = offset;
…		…
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1420	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1421	}
1118		1422
1119	sv = sv_2mortal (sv);	1423	sv = sv_2mortal (sv);
1120		1424
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1425	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1122	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1426	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1427
1124	return sv;	1428	return sv;
1125	}	1429	}
1126		1430
…		…
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1434	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1435
1132	BOOT:	1436	BOOT:
1133	{	1437	{
1134	int i;	1438	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1439
1138	for (i = 0; i < 256; ++i)	1440	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1441	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1442	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1443	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1444	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1445	: -1;
1144		1446
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1447	json_stash = gv_stashpv ("JSON::XS" , 1);
		1448	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1449
		1450	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1451	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1146	}	1452	}
1147		1453
1148	PROTOTYPES: DISABLE	1454	PROTOTYPES: DISABLE
1149		1455
1150	SV new (char dummy)	1456	void CLONE (...)
1151	CODE:	1457	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1458	json_stash = 0;
		1459	json_boolean_stash = 0;
		1460
		1461	void new (char *klass)
		1462	PPCODE:
		1463	{
		1464	SV *pv = NEWSV (0, sizeof (JSON));
		1465	SvPOK_only (pv);
		1466	Zero (SvPVX (pv), 1, JSON);
		1467	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1468	XPUSHs (sv_2mortal (sv_bless (
		1469	newRV_noinc (pv),
		1470	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1471	)));
		1472	}
		1473
		1474	void ascii (JSON *self, int enable = 1)
		1475	ALIAS:
		1476	ascii = F_ASCII
		1477	latin1 = F_LATIN1
		1478	utf8 = F_UTF8
		1479	indent = F_INDENT
		1480	canonical = F_CANONICAL
		1481	space_before = F_SPACE_BEFORE
		1482	space_after = F_SPACE_AFTER
		1483	pretty = F_PRETTY
		1484	allow_nonref = F_ALLOW_NONREF
		1485	shrink = F_SHRINK
		1486	allow_blessed = F_ALLOW_BLESSED
		1487	convert_blessed = F_CONV_BLESSED
		1488	relaxed = F_RELAXED
		1489	PPCODE:
		1490	{
		1491	if (enable)
		1492	self->flags \|= ix;
		1493	else
		1494	self->flags &= ~ix;
		1495
		1496	XPUSHs (ST (0));
		1497	}
		1498
		1499	void get_ascii (JSON *self)
		1500	ALIAS:
		1501	get_ascii = F_ASCII
		1502	get_latin1 = F_LATIN1
		1503	get_utf8 = F_UTF8
		1504	get_indent = F_INDENT
		1505	get_canonical = F_CANONICAL
		1506	get_space_before = F_SPACE_BEFORE
		1507	get_space_after = F_SPACE_AFTER
		1508	get_allow_nonref = F_ALLOW_NONREF
		1509	get_shrink = F_SHRINK
		1510	get_allow_blessed = F_ALLOW_BLESSED
		1511	get_convert_blessed = F_CONV_BLESSED
		1512	get_relaxed = F_RELAXED
		1513	PPCODE:
		1514	XPUSHs (boolSV (self->flags & ix));
		1515
		1516	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1517	PPCODE:
		1518	{
		1519	UV log2 = 0;
		1520
		1521	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1522
		1523	while ((1UL << log2) < max_depth)
		1524	++log2;
		1525
		1526	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1527
		1528	XPUSHs (ST (0));
		1529	}
		1530
		1531	U32 get_max_depth (JSON *self)
		1532	CODE:
		1533	RETVAL = DEC_DEPTH (self->flags);
1153	OUTPUT:	1534	OUTPUT:
1154	RETVAL	1535	RETVAL
1155		1536
1156	SV ascii (SV self, int enable = 1)	1537	void max_size (JSON *self, UV max_size = 0)
1157	ALIAS:	1538	PPCODE:
1158	ascii = F_ASCII	1539	{
1159	latin1 = F_LATIN1	1540	UV log2 = 0;
1160	utf8 = F_UTF8	1541
1161	indent = F_INDENT	1542	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1162	canonical = F_CANONICAL	1543	if (max_size == 1) max_size = 2;
1163	space_before = F_SPACE_BEFORE	1544
1164	space_after = F_SPACE_AFTER	1545	while ((1UL << log2) < max_size)
1165	pretty = F_PRETTY	1546	++log2;
1166	allow_nonref = F_ALLOW_NONREF	1547
1167	shrink = F_SHRINK	1548	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1549
		1550	XPUSHs (ST (0));
		1551	}
		1552
		1553	int get_max_size (JSON *self)
1168	CODE:	1554	CODE:
1169	{	1555	RETVAL = DEC_SIZE (self->flags);
1170	UV *uv = SvJSON (self);
1171	if (enable)
1172	*uv \|= ix;
1173	else
1174	*uv &= ~ix;
1175
1176	RETVAL = newSVsv (self);
1177	}
1178	OUTPUT:	1556	OUTPUT:
1179	RETVAL	1557	RETVAL
1180		1558
1181	SV max_depth (SV self, UV max_depth = 0x80000000UL)	1559	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1182	CODE:
1183	{
1184	UV *uv = SvJSON (self);
1185	UV log2 = 0;
1186
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1188
1189	while ((1UL << log2) < max_depth)
1190	++log2;
1191
1192	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1193
1194	RETVAL = newSVsv (self);
1195	}
1196	OUTPUT:
1197	RETVAL
1198
1199	void encode (SV self, SV scalar)
1200	PPCODE:	1560	PPCODE:
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	1561	{
		1562	SvREFCNT_dec (self->cb_object);
		1563	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1202		1564
1203	void decode (SV self, SV jsonstr)	1565	XPUSHs (ST (0));
		1566	}
		1567
		1568	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
1204	PPCODE:	1569	PPCODE:
		1570	{
		1571	if (!self->cb_sk_object)
		1572	self->cb_sk_object = newHV ();
		1573
		1574	if (SvOK (cb))
		1575	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1576	else
		1577	{
		1578	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1579
		1580	if (!HvKEYS (self->cb_sk_object))
		1581	{
		1582	SvREFCNT_dec (self->cb_sk_object);
		1583	self->cb_sk_object = 0;
		1584	}
		1585	}
		1586
		1587	XPUSHs (ST (0));
		1588	}
		1589
		1590	void encode (JSON self, SV scalar)
		1591	PPCODE:
		1592	XPUSHs (encode_json (scalar, self));
		1593
		1594	void decode (JSON self, SV jsonstr)
		1595	PPCODE:
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1596	XPUSHs (decode_json (jsonstr, self, 0));
1206		1597
1207	void decode_prefix (SV self, SV jsonstr)	1598	void decode_prefix (JSON self, SV jsonstr)
1208	PPCODE:	1599	PPCODE:
1209	{	1600	{
1210	UV offset;	1601	UV offset;
1211	EXTEND (SP, 2);	1602	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	1603	PUSHs (decode_json (jsonstr, self, &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));	1604	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	1605	}
1215		1606
		1607	void DESTROY (JSON *self)
		1608	CODE:
		1609	SvREFCNT_dec (self->cb_sk_object);
		1610	SvREFCNT_dec (self->cb_object);
		1611
1216	PROTOTYPES: ENABLE	1612	PROTOTYPES: ENABLE
1217		1613
1218	void to_json (SV *scalar)	1614	void encode_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:	1615	PPCODE:
		1616	{
		1617	JSON json = { F_DEFAULT \| F_UTF8 };
1222	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));	1618	XPUSHs (encode_json (scalar, &json));
		1619	}
1223		1620
1224	void from_json (SV *jsonstr)	1621	void decode_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:	1622	PPCODE:
		1623	{
		1624	JSON json = { F_DEFAULT \| F_UTF8 };
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));	1625	XPUSHs (decode_json (jsonstr, &json, 0));
		1626	}
1229		1627

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs. Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC