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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs.
Revision 1.71 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
25	#define F_SPACE_BEFORE 0x00000020UL	26	#define F_SPACE_BEFORE 0x00000020UL
26	#define F_SPACE_AFTER 0x00000040UL	27	#define F_SPACE_AFTER 0x00000040UL
27	#define F_ALLOW_NONREF 0x00000100UL	28	#define F_ALLOW_NONREF 0x00000100UL
28	#define F_SHRINK 0x00000200UL	29	#define F_SHRINK 0x00000200UL
29	#define F_ALLOW_BLESSED 0x00000400UL	30	#define F_ALLOW_BLESSED 0x00000400UL
30	#define F_CONV_BLESSED 0x00000800UL // NYI	31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
31	#define F_MAXDEPTH 0xf8000000UL	34	#define F_MAXDEPTH 0xf8000000UL
32	#define S_MAXDEPTH 27	35	#define S_MAXDEPTH 27
33	#define F_MAXSIZE 0x01f00000UL	36	#define F_MAXSIZE 0x01f00000UL
34	#define S_MAXSIZE 20	37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
35		39
36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))	41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
38		42
39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
…		…
46		50
47	#define SB do {	51	#define SB do {
48	#define SE } while (0)	52	#define SE } while (0)
49		53
50	#if __GNUC__ >= 3	54	#if __GNUC__ >= 3
51	# define expect(expr,value) __builtin_expect ((expr),(value))	55	# define expect(expr,value) __builtin_expect ((expr), (value))
52	# define inline inline	56	# define INLINE static inline
53	#else	57	#else
54	# define expect(expr,value) (expr)	58	# define expect(expr,value) (expr)
55	# define inline static	59	# define INLINE static
56	#endif	60	#endif
57		61
58	#define expect_false(expr) expect ((expr) != 0, 0)	62	#define expect_false(expr) expect ((expr) != 0, 0)
59	#define expect_true(expr) expect ((expr) != 0, 1)	63	#define expect_true(expr) expect ((expr) != 0, 1)
60		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
61	static HV json_stash, json_boolean_stash; // JSON::XS::	73	static HV json_stash, json_boolean_stash; // JSON::XS::
62	static SV json_true, json_false;	74	static SV json_true, json_false;
63		75
64	typedef struct {	76	typedef struct {
65	U32 flags;	77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
66	} JSON;	80	} JSON;
67		81
68	/////////////////////////////////////////////////////////////////////////////	82	/////////////////////////////////////////////////////////////////////////////
69	// utility functions	83	// utility functions
70		84
71	inline void	85	INLINE void
72	shrink (SV *sv)	86	shrink (SV *sv)
73	{	87	{
74	sv_utf8_downgrade (sv, 1);	88	sv_utf8_downgrade (sv, 1);
75	if (SvLEN (sv) > SvCUR (sv) + 1)	89	if (SvLEN (sv) > SvCUR (sv) + 1)
76	{	90	{
…		…
85	// 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
86	// case of an error.	100	// case of an error.
87	// we special-case "safe" characters from U+80 .. U+7FF,	101	// we special-case "safe" characters from U+80 .. U+7FF,
88	// but use the very good perl function to parse anything else.	102	// but use the very good perl function to parse anything else.
89	// note that we never call this function for a ascii codepoints	103	// note that we never call this function for a ascii codepoints
90	inline UV	104	INLINE UV
91	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
92	{	106	{
93	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
94	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
95	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
…		…
114	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
115	SV *sv; // result scalar	129	SV *sv; // result scalar
116	JSON json;	130	JSON json;
117	U32 indent; // indentation level	131	U32 indent; // indentation level
118	U32 maxdepth; // max. indentation/recursion level	132	U32 maxdepth; // max. indentation/recursion level
		133	UV limit; // escape character values >= this value when encoding
119	} enc_t;	134	} enc_t;
120		135
121	inline void	136	INLINE void
122	need (enc_t *enc, STRLEN len)	137	need (enc_t *enc, STRLEN len)
123	{	138	{
124	if (expect_false (enc->cur + len >= enc->end))	139	if (expect_false (enc->cur + len >= enc->end))
125	{	140	{
126	STRLEN cur = enc->cur - SvPVX (enc->sv);	141	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
128	enc->cur = SvPVX (enc->sv) + cur;	143	enc->cur = SvPVX (enc->sv) + cur;
129	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	144	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
130	}	145	}
131	}	146	}
132		147
133	inline void	148	INLINE void
134	encode_ch (enc_t *enc, char ch)	149	encode_ch (enc_t *enc, char ch)
135	{	150	{
136	need (enc, 1);	151	need (enc, 1);
137	*enc->cur++ = ch;	152	*enc->cur++ = ch;
138	}	153	}
…		…
192	{	207	{
193	uch = ch;	208	uch = ch;
194	clen = 1;	209	clen = 1;
195	}	210	}
196		211
197	if (uch > 0x10FFFFUL)	212	if (uch < 0x20 \|\| uch >= enc->limit)
198	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
199
200	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
201	{	213	{
202	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
203	{	215	{
		216	if (uch > 0x10FFFFUL)
		217	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		218
204	need (enc, len += 11);	219	need (enc, len += 11);
205	sprintf (enc->cur, "\\u%04x\\u%04x",	220	sprintf (enc->cur, "\\u%04x\\u%04x",
206	(int)((uch - 0x10000) / 0x400 + 0xD800),	221	(int)((uch - 0x10000) / 0x400 + 0xD800),
207	(int)((uch - 0x10000) % 0x400 + 0xDC00));	222	(int)((uch - 0x10000) % 0x400 + 0xDC00));
208	enc->cur += 12;	223	enc->cur += 12;
…		…
247		262
248	--len;	263	--len;
249	}	264	}
250	}	265	}
251		266
252	inline void	267	INLINE void
253	encode_indent (enc_t *enc)	268	encode_indent (enc_t *enc)
254	{	269	{
255	if (enc->json.flags & F_INDENT)	270	if (enc->json.flags & F_INDENT)
256	{	271	{
257	int spaces = enc->indent * INDENT_STEP;	272	int spaces = enc->indent * INDENT_STEP;
…		…
260	memset (enc->cur, ' ', spaces);	275	memset (enc->cur, ' ', spaces);
261	enc->cur += spaces;	276	enc->cur += spaces;
262	}	277	}
263	}	278	}
264		279
265	inline void	280	INLINE void
266	encode_space (enc_t *enc)	281	encode_space (enc_t *enc)
267	{	282	{
268	need (enc, 1);	283	need (enc, 1);
269	encode_ch (enc, ' ');	284	encode_ch (enc, ' ');
270	}	285	}
271		286
272	inline void	287	INLINE void
273	encode_nl (enc_t *enc)	288	encode_nl (enc_t *enc)
274	{	289	{
275	if (enc->json.flags & F_INDENT)	290	if (enc->json.flags & F_INDENT)
276	{	291	{
277	need (enc, 1);	292	need (enc, 1);
278	encode_ch (enc, '\n');	293	encode_ch (enc, '\n');
279	}	294	}
280	}	295	}
281		296
282	inline void	297	INLINE void
283	encode_comma (enc_t *enc)	298	encode_comma (enc_t *enc)
284	{	299	{
285	encode_ch (enc, ',');	300	encode_ch (enc, ',');
286		301
287	if (enc->json.flags & F_INDENT)	302	if (enc->json.flags & F_INDENT)
…		…
298	int i, len = av_len (av);	313	int i, len = av_len (av);
299		314
300	if (enc->indent >= enc->maxdepth)	315	if (enc->indent >= enc->maxdepth)
301	croak ("data structure too deep (hit recursion limit)");	316	croak ("data structure too deep (hit recursion limit)");
302		317
303	encode_ch (enc, '['); encode_nl (enc);	318	encode_ch (enc, '[');
304	++enc->indent;	319
		320	if (len >= 0)
		321	{
		322	encode_nl (enc); ++enc->indent;
305		323
306	for (i = 0; i <= len; ++i)	324	for (i = 0; i <= len; ++i)
307	{	325	{
		326	SV **svp = av_fetch (av, i, 0);
		327
308	encode_indent (enc);	328	encode_indent (enc);
309	encode_sv (enc, *av_fetch (av, i, 0));
310		329
		330	if (svp)
		331	encode_sv (enc, *svp);
		332	else
		333	encode_str (enc, "null", 4, 0);
		334
311	if (i < len)	335	if (i < len)
312	encode_comma (enc);	336	encode_comma (enc);
313	}	337	}
314		338
		339	encode_nl (enc); --enc->indent; encode_indent (enc);
		340	}
		341
315	encode_nl (enc);	342	encode_ch (enc, ']');
316
317	--enc->indent;
318	encode_indent (enc); encode_ch (enc, ']');
319	}	343	}
320		344
321	static void	345	static void
322	encode_he (enc_t enc, HE he)	346	encode_hk (enc_t enc, HE he)
323	{	347	{
324	encode_ch (enc, '"');	348	encode_ch (enc, '"');
325		349
326	if (HeKLEN (he) == HEf_SVKEY)	350	if (HeKLEN (he) == HEf_SVKEY)
327	{	351	{
…		…
340	encode_ch (enc, '"');	364	encode_ch (enc, '"');
341		365
342	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	366	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
343	encode_ch (enc, ':');	367	encode_ch (enc, ':');
344	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	368	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
345	encode_sv (enc, HeVAL (he));
346	}	369	}
347		370
348	// compare hash entries, used when all keys are bytestrings	371	// compare hash entries, used when all keys are bytestrings
349	static int	372	static int
350	he_cmp_fast (const void a_, const void b_)	373	he_cmp_fast (const void a_, const void b_)
…		…
355	HE b = (HE **)b_;	378	HE b = (HE **)b_;
356		379
357	STRLEN la = HeKLEN (a);	380	STRLEN la = HeKLEN (a);
358	STRLEN lb = HeKLEN (b);	381	STRLEN lb = HeKLEN (b);
359		382
360	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	383	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
361	cmp = la - lb;	384	cmp = lb - la;
362		385
363	return cmp;	386	return cmp;
364	}	387	}
365		388
366	// 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
367	static int	390	static int
368	he_cmp_slow (const void a, const void b)	391	he_cmp_slow (const void a, const void b)
369	{	392	{
370	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	393	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
371	}	394	}
372		395
373	static void	396	static void
374	encode_hv (enc_t enc, HV hv)	397	encode_hv (enc_t enc, HV hv)
375	{	398	{
		399	HE *he;
376	int count, i;	400	int count;
377		401
378	if (enc->indent >= enc->maxdepth)	402	if (enc->indent >= enc->maxdepth)
379	croak ("data structure too deep (hit recursion limit)");	403	croak ("data structure too deep (hit recursion limit)");
380		404
381	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	405	encode_ch (enc, '{');
382		406
383	if ((count = hv_iterinit (hv)))
384	{
385	// for canonical output we have to sort by keys first	407	// for canonical output we have to sort by keys first
386	// actually, this is mostly due to the stupid so-called	408	// actually, this is mostly due to the stupid so-called
387	// security workaround added somewhere in 5.8.x.	409	// security workaround added somewhere in 5.8.x.
388	// that randomises hash orderings	410	// that randomises hash orderings
389	if (enc->json.flags & F_CANONICAL)	411	if (enc->json.flags & F_CANONICAL)
		412	{
		413	int count = hv_iterinit (hv);
		414
		415	if (SvMAGICAL (hv))
390	{	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	{
391	int fast = 1;	431	int i, fast = 1;
392	HE *he;
393	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	432	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
394	HE *hes = _alloca (count sizeof (HE));	433	HE *hes = _alloca (count sizeof (HE));
395	#else	434	#else
396	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
397	#endif	436	#endif
…		…
424		463
425	FREETMPS;	464	FREETMPS;
426	LEAVE;	465	LEAVE;
427	}	466	}
428		467
429	for (i = 0; i < count; ++i)	468	encode_nl (enc); ++enc->indent;
		469
		470	while (count--)
430	{	471	{
431	encode_indent (enc);	472	encode_indent (enc);
		473	he = hes [count];
432	encode_he (enc, hes [i]);	474	encode_hk (enc, he);
		475	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
433		476
434	if (i < count - 1)	477	if (count)
435	encode_comma (enc);	478	encode_comma (enc);
436	}	479	}
437		480
438	encode_nl (enc);	481	encode_nl (enc); --enc->indent; encode_indent (enc);
439	}	482	}
		483	}
440	else	484	else
		485	{
		486	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		487	if ((he = hv_iternext (hv)))
441	{	488	{
442	HE *he = hv_iternext (hv);	489	encode_nl (enc); ++enc->indent;
443		490
444	for (;;)	491	for (;;)
445	{	492	{
446	encode_indent (enc);	493	encode_indent (enc);
447	encode_he (enc, he);	494	encode_hk (enc, he);
		495	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
448		496
449	if (!(he = hv_iternext (hv)))	497	if (!(he = hv_iternext (hv)))
450	break;	498	break;
451		499
452	encode_comma (enc);	500	encode_comma (enc);
453	}	501	}
454		502
455	encode_nl (enc);	503	encode_nl (enc); --enc->indent; encode_indent (enc);
456	}	504	}
457	}	505	}
458		506
459	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	507	encode_ch (enc, '}');
460	}	508	}
461		509
462	// encode objects, arrays and special \0=false and \1=true values.	510	// encode objects, arrays and special \0=false and \1=true values.
463	static void	511	static void
464	encode_rv (enc_t enc, SV sv)	512	encode_rv (enc_t enc, SV sv)
…		…
468	SvGETMAGIC (sv);	516	SvGETMAGIC (sv);
469	svt = SvTYPE (sv);	517	svt = SvTYPE (sv);
470		518
471	if (expect_false (SvOBJECT (sv)))	519	if (expect_false (SvOBJECT (sv)))
472	{	520	{
		521	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		522	? json_boolean_stash
		523	: gv_stashpv ("JSON::XS::Boolean", 1);
		524
473	if (SvSTASH (sv) == json_boolean_stash)	525	if (SvSTASH (sv) == stash)
474	{	526	{
475	if (SvIV (sv) == 0)	527	if (SvIV (sv))
		528	encode_str (enc, "true", 4, 0);
		529	else
476	encode_str (enc, "false", 5, 0);	530	encode_str (enc, "false", 5, 0);
477	else
478	encode_str (enc, "true", 4, 0);
479	}	531	}
480	else	532	else
481	{	533	{
482	#if 0	534	#if 0
483	if (0 && sv_derived_from (rv, "JSON::Literal"))	535	if (0 && sv_derived_from (rv, "JSON::Literal"))
…		…
486	}	538	}
487	#endif	539	#endif
488	if (enc->json.flags & F_CONV_BLESSED)	540	if (enc->json.flags & F_CONV_BLESSED)
489	{	541	{
490	// we re-bless the reference to get overload and other niceties right	542	// we re-bless the reference to get overload and other niceties right
491	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	543	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
492		544
493	if (to_json)	545	if (to_json)
494	{	546	{
495	dSP;	547	dSP;
496	ENTER;	548
497	SAVETMPS;
498	PUSHMARK (SP);	549	ENTER; SAVETMPS; PUSHMARK (SP);
499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	550	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
500		551
501	// calling with G_SCALAR ensures that we always get a 1 reutrn value	552	// calling with G_SCALAR ensures that we always get a 1 return value
502	// check anyways.
503	PUTBACK;	553	PUTBACK;
504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	554	call_sv ((SV *)GvCV (to_json), G_SCALAR);
505	SPAGAIN;	555	SPAGAIN;
506		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
507	encode_sv (enc, POPs);	564	encode_sv (enc, sv);
508		565
509	FREETMPS;	566	FREETMPS; LEAVE;
510	LEAVE;
511	}	567	}
512	else if (enc->json.flags & F_ALLOW_BLESSED)	568	else if (enc->json.flags & F_ALLOW_BLESSED)
513	encode_str (enc, "null", 4, 0);	569	encode_str (enc, "null", 4, 0);
514	else	570	else
515	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",	571	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
…		…
526	encode_hv (enc, (HV *)sv);	582	encode_hv (enc, (HV *)sv);
527	else if (svt == SVt_PVAV)	583	else if (svt == SVt_PVAV)
528	encode_av (enc, (AV *)sv);	584	encode_av (enc, (AV *)sv);
529	else if (svt < SVt_PVAV)	585	else if (svt < SVt_PVAV)
530	{	586	{
531	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')
532	encode_str (enc, "false", 5, 0);	593	encode_str (enc, "false", 5, 0);
533	else if (SvNIOK (sv) && SvIV (sv) == 1)
534	encode_str (enc, "true", 4, 0);
535	else	594	else
536	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",
537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
538	}	597	}
539	else	598	else
…		…
622	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	681	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
623	enc.cur = SvPVX (enc.sv);	682	enc.cur = SvPVX (enc.sv);
624	enc.end = SvEND (enc.sv);	683	enc.end = SvEND (enc.sv);
625	enc.indent = 0;	684	enc.indent = 0;
626	enc.maxdepth = DEC_DEPTH (enc.json.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;
627		689
628	SvPOK_only (enc.sv);	690	SvPOK_only (enc.sv);
629	encode_sv (&enc, scalar);	691	encode_sv (&enc, scalar);
630		692
631	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	693	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
652	JSON json;	714	JSON json;
653	U32 depth; // recursion depth	715	U32 depth; // recursion depth
654	U32 maxdepth; // recursion depth limit	716	U32 maxdepth; // recursion depth limit
655	} dec_t;	717	} dec_t;
656		718
657	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
658	decode_ws (dec_t *dec)	729	decode_ws (dec_t *dec)
659	{	730	{
660	for (;;)	731	for (;;)
661	{	732	{
662	char ch = *dec->cur;	733	char ch = *dec->cur;
663		734
664	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	}
665	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	747	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
666	break;	748	break; // parse error, but let higher level handle it, gives better error messages
667		749
668	++dec->cur;	750	++dec->cur;
669	}	751	}
670	}	752	}
671		753
…		…
920	is_nv = 1;	1002	is_nv = 1;
921	}	1003	}
922		1004
923	if (!is_nv)	1005	if (!is_nv)
924	{	1006	{
		1007	int len = dec->cur - start;
		1008
925	// 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
926	if (*start == '-')	1010	if (*start == '-')
927	switch (dec->cur - start)	1011	switch (len)
928	{	1012	{
929	case 2: return newSViv (-( start [1] - '0' * 1));	1013	case 2: return newSViv (-( start [1] - '0' * 1));
930	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1014	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
931	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1015	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
932	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	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));
933	}	1018	}
934	else	1019	else
935	switch (dec->cur - start)	1020	switch (len)
936	{	1021	{
937	case 1: return newSViv ( start [0] - '0' * 1);	1022	case 1: return newSViv ( start [0] - '0' * 1);
938	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1023	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
939	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1024	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
940	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	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);
941	}	1027	}
942		1028
943	{	1029	{
944	UV uv;	1030	UV uv;
945	int numtype = grok_number (start, dec->cur - start, &uv);	1031	int numtype = grok_number (start, len, &uv);
946	if (numtype & IS_NUMBER_IN_UV)	1032	if (numtype & IS_NUMBER_IN_UV)
947	if (numtype & IS_NUMBER_NEG)	1033	if (numtype & IS_NUMBER_NEG)
948	{	1034	{
949	if (uv < (UV)IV_MIN)	1035	if (uv < (UV)IV_MIN)
950	return newSViv (-(IV)uv);	1036	return newSViv (-(IV)uv);
951	}	1037	}
952	else	1038	else
953	return newSVuv (uv);	1039	return newSVuv (uv);
954
955	// here would likely be the place for bigint support
956	}	1040	}
957	}
958		1041
959	// if we ever support bigint or bigfloat, this is the place for bigfloat	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
960	return newSVnv (Atof (start));	1058	return newSVnv (Atof (start));
961		1059
962	fail:	1060	fail:
963	return 0;	1061	return 0;
964	}	1062	}
…		…
994		1092
995	if (*dec->cur != ',')	1093	if (*dec->cur != ',')
996	ERR (", or ] expected while parsing array");	1094	ERR (", or ] expected while parsing array");
997		1095
998	++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	}
999	}	1105	}
1000		1106
1001	DEC_DEC_DEPTH;	1107	DEC_DEC_DEPTH;
1002	return newRV_noinc ((SV *)av);	1108	return newRV_noinc ((SV *)av);
1003		1109
…		…
1008	}	1114	}
1009		1115
1010	static SV *	1116	static SV *
1011	decode_hv (dec_t *dec)	1117	decode_hv (dec_t *dec)
1012	{	1118	{
		1119	SV *sv;
1013	HV *hv = newHV ();	1120	HV *hv = newHV ();
1014		1121
1015	DEC_INC_DEPTH;	1122	DEC_INC_DEPTH;
1016	decode_ws (dec);	1123	decode_ws (dec);
1017		1124
1018	if (*dec->cur == '}')	1125	if (*dec->cur == '}')
1019	++dec->cur;	1126	++dec->cur;
1020	else	1127	else
1021	for (;;)	1128	for (;;)
1022	{	1129	{
1023	decode_ws (dec); EXPECT_CH ('"');	1130	EXPECT_CH ('"');
1024		1131
1025	// heuristic: assume that	1132	// heuristic: assume that
1026	// a) decode_str + hv_store_ent are abysmally slow.	1133	// a) decode_str + hv_store_ent are abysmally slow.
1027	// b) most hash keys are short, simple ascii text.	1134	// b) most hash keys are short, simple ascii text.
1028	// => try to "fast-match" such strings to avoid	1135	// => try to "fast-match" such strings to avoid
…		…
1042	if (!key)	1149	if (!key)
1043	goto fail;	1150	goto fail;
1044		1151
1045	decode_ws (dec); EXPECT_CH (':');	1152	decode_ws (dec); EXPECT_CH (':');
1046		1153
		1154	decode_ws (dec);
1047	value = decode_sv (dec);	1155	value = decode_sv (dec);
1048	if (!value)	1156	if (!value)
1049	{	1157	{
1050	SvREFCNT_dec (key);	1158	SvREFCNT_dec (key);
1051	goto fail;	1159	goto fail;
…		…
1063	int len = p - key;	1171	int len = p - key;
1064	dec->cur = p + 1;	1172	dec->cur = p + 1;
1065		1173
1066	decode_ws (dec); EXPECT_CH (':');	1174	decode_ws (dec); EXPECT_CH (':');
1067		1175
		1176	decode_ws (dec);
1068	value = decode_sv (dec);	1177	value = decode_sv (dec);
1069	if (!value)	1178	if (!value)
1070	goto fail;	1179	goto fail;
1071		1180
1072	hv_store (hv, key, len, value, 0);	1181	hv_store (hv, key, len, value, 0);
…		…
1088		1197
1089	if (*dec->cur != ',')	1198	if (*dec->cur != ',')
1090	ERR (", or } expected while parsing object/hash");	1199	ERR (", or } expected while parsing object/hash");
1091		1200
1092	++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	}
1093	}	1210	}
1094		1211
1095	DEC_DEC_DEPTH;	1212	DEC_DEC_DEPTH;
1096	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;
1097		1274
1098	fail:	1275	fail:
1099	SvREFCNT_dec (hv);	1276	SvREFCNT_dec (hv);
1100	DEC_DEC_DEPTH;	1277	DEC_DEC_DEPTH;
1101	return 0;	1278	return 0;
1102	}	1279	}
1103		1280
1104	static SV *	1281	static SV *
1105	decode_sv (dec_t *dec)	1282	decode_sv (dec_t *dec)
1106	{	1283	{
1107	decode_ws (dec);
1108
1109	// the beauty of JSON: you need exactly one character lookahead	1284	// the beauty of JSON: you need exactly one character lookahead
1110	// to parse anything.	1285	// to parse anything.
1111	switch (*dec->cur)	1286	switch (*dec->cur)
1112	{	1287	{
1113	case '"': ++dec->cur; return decode_str (dec);	1288	case '"': ++dec->cur; return decode_str (dec);
…		…
1121		1296
1122	case 't':	1297	case 't':
1123	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1298	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1124	{	1299	{
1125	dec->cur += 4;	1300	dec->cur += 4;
		1301	#if JSON_SLOW
		1302	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1303	#endif
1126	return SvREFCNT_inc (json_true);	1304	return SvREFCNT_inc (json_true);
1127	}	1305	}
1128	else	1306	else
1129	ERR ("'true' expected");	1307	ERR ("'true' expected");
1130		1308
…		…
1132		1310
1133	case 'f':	1311	case 'f':
1134	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1312	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1135	{	1313	{
1136	dec->cur += 5;	1314	dec->cur += 5;
		1315	#if JSON_SLOW
		1316	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1317	#endif
1137	return SvREFCNT_inc (json_false);	1318	return SvREFCNT_inc (json_false);
1138	}	1319	}
1139	else	1320	else
1140	ERR ("'false' expected");	1321	ERR ("'false' expected");
1141		1322
…		…
1187	dec.end = SvEND (string);	1368	dec.end = SvEND (string);
1188	dec.err = 0;	1369	dec.err = 0;
1189	dec.depth = 0;	1370	dec.depth = 0;
1190	dec.maxdepth = DEC_DEPTH (dec.json.flags);	1371	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1191		1372
		1373	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1374	dec.json.flags \|= F_HOOK;
		1375
1192	*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);
1193	sv = decode_sv (&dec);	1379	sv = decode_sv (&dec);
1194		1380
1195	if (!(offset_return \|\| !sv))	1381	if (!(offset_return \|\| !sv))
1196	{	1382	{
1197	// check for trailing garbage	1383	// check for trailing garbage
…		…
1265	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1451	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1266	}	1452	}
1267		1453
1268	PROTOTYPES: DISABLE	1454	PROTOTYPES: DISABLE
1269		1455
		1456	void CLONE (...)
		1457	CODE:
		1458	json_stash = 0;
		1459	json_boolean_stash = 0;
		1460
1270	void new (char *klass)	1461	void new (char *klass)
1271	PPCODE:	1462	PPCODE:
1272	{	1463	{
1273	SV *pv = NEWSV (0, sizeof (JSON));	1464	SV *pv = NEWSV (0, sizeof (JSON));
1274	SvPOK_only (pv);	1465	SvPOK_only (pv);
1275	Zero (SvPVX (pv), 1, sizeof (JSON));	1466	Zero (SvPVX (pv), 1, JSON);
1276	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1467	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1277	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1468	XPUSHs (sv_2mortal (sv_bless (
		1469	newRV_noinc (pv),
		1470	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1471	)));
1278	}	1472	}
1279		1473
1280	void ascii (JSON *self, int enable = 1)	1474	void ascii (JSON *self, int enable = 1)
1281	ALIAS:	1475	ALIAS:
1282	ascii = F_ASCII	1476	ascii = F_ASCII
…		…
1289	pretty = F_PRETTY	1483	pretty = F_PRETTY
1290	allow_nonref = F_ALLOW_NONREF	1484	allow_nonref = F_ALLOW_NONREF
1291	shrink = F_SHRINK	1485	shrink = F_SHRINK
1292	allow_blessed = F_ALLOW_BLESSED	1486	allow_blessed = F_ALLOW_BLESSED
1293	convert_blessed = F_CONV_BLESSED	1487	convert_blessed = F_CONV_BLESSED
		1488	relaxed = F_RELAXED
1294	PPCODE:	1489	PPCODE:
1295	{	1490	{
1296	if (enable)	1491	if (enable)
1297	self->flags \|= ix;	1492	self->flags \|= ix;
1298	else	1493	else
1299	self->flags &= ~ix;	1494	self->flags &= ~ix;
1300		1495
1301	XPUSHs (ST (0));	1496	XPUSHs (ST (0));
1302	}	1497	}
1303		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
1304	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1516	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1305	PPCODE:	1517	PPCODE:
1306	{	1518	{
1307	UV log2 = 0;	1519	UV log2 = 0;
1308		1520
…		…
1314	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1526	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1315		1527
1316	XPUSHs (ST (0));	1528	XPUSHs (ST (0));
1317	}	1529	}
1318		1530
		1531	U32 get_max_depth (JSON *self)
		1532	CODE:
		1533	RETVAL = DEC_DEPTH (self->flags);
		1534	OUTPUT:
		1535	RETVAL
		1536
1319	void max_size (JSON *self, UV max_size = 0)	1537	void max_size (JSON *self, UV max_size = 0)
1320	PPCODE:	1538	PPCODE:
1321	{	1539	{
1322	UV log2 = 0;	1540	UV log2 = 0;
1323		1541
…		…
1326		1544
1327	while ((1UL << log2) < max_size)	1545	while ((1UL << log2) < max_size)
1328	++log2;	1546	++log2;
1329		1547
1330	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1548	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1549
		1550	XPUSHs (ST (0));
		1551	}
		1552
		1553	int get_max_size (JSON *self)
		1554	CODE:
		1555	RETVAL = DEC_SIZE (self->flags);
		1556	OUTPUT:
		1557	RETVAL
		1558
		1559	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1560	PPCODE:
		1561	{
		1562	SvREFCNT_dec (self->cb_object);
		1563	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1564
		1565	XPUSHs (ST (0));
		1566	}
		1567
		1568	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		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	}
1331		1586
1332	XPUSHs (ST (0));	1587	XPUSHs (ST (0));
1333	}	1588	}
1334		1589
1335	void encode (JSON self, SV scalar)	1590	void encode (JSON self, SV scalar)
…		…
1347	EXTEND (SP, 2);	1602	EXTEND (SP, 2);
1348	PUSHs (decode_json (jsonstr, self, &offset));	1603	PUSHs (decode_json (jsonstr, self, &offset));
1349	PUSHs (sv_2mortal (newSVuv (offset)));	1604	PUSHs (sv_2mortal (newSVuv (offset)));
1350	}	1605	}
1351		1606
		1607	void DESTROY (JSON *self)
		1608	CODE:
		1609	SvREFCNT_dec (self->cb_sk_object);
		1610	SvREFCNT_dec (self->cb_object);
		1611
1352	PROTOTYPES: ENABLE	1612	PROTOTYPES: ENABLE
1353		1613
1354	void to_json (SV *scalar)	1614	void encode_json (SV *scalar)
1355	PPCODE:	1615	PPCODE:
1356	{	1616	{
1357	JSON json = { F_DEFAULT \| F_UTF8 };	1617	JSON json = { F_DEFAULT \| F_UTF8 };
1358	XPUSHs (encode_json (scalar, &json));	1618	XPUSHs (encode_json (scalar, &json));
1359	}	1619	}
1360		1620
1361	void from_json (SV *jsonstr)	1621	void decode_json (SV *jsonstr)
1362	PPCODE:	1622	PPCODE:
1363	{	1623	{
1364	JSON json = { F_DEFAULT \| F_UTF8 };	1624	JSON json = { F_DEFAULT \| F_UTF8 };
1365	XPUSHs (decode_json (jsonstr, &json, 0));	1625	XPUSHs (decode_json (jsonstr, &json, 0));
1366	}	1626	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs. Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC vs.
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC