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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.49 by root, Sun Jul 1 23:40:07 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
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	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
35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
…		…
47		50
48	#define SB do {	51	#define SB do {
49	#define SE } while (0)	52	#define SE } while (0)
50		53
51	#if __GNUC__ >= 3	54	#if __GNUC__ >= 3
52	# define expect(expr,value) __builtin_expect ((expr),(value))	55	# define expect(expr,value) __builtin_expect ((expr), (value))
53	# define inline inline	56	# define INLINE static inline
54	#else	57	#else
55	# define expect(expr,value) (expr)	58	# define expect(expr,value) (expr)
56	# define inline static	59	# define INLINE static
57	#endif	60	#endif
58		61
59	#define expect_false(expr) expect ((expr) != 0, 0)	62	#define expect_false(expr) expect ((expr) != 0, 0)
60	#define expect_true(expr) expect ((expr) != 0, 1)	63	#define expect_true(expr) expect ((expr) != 0, 1)
61		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
62	static HV json_stash, json_boolean_stash; // JSON::XS::	73	static HV json_stash, json_boolean_stash; // JSON::XS::
63	static SV json_true, json_false;	74	static SV json_true, json_false;
64		75
65	typedef struct {	76	typedef struct {
66	U32 flags;	77	U32 flags;
67	SV cb_object, cb_sk_object;	78	SV *cb_object;
		79	HV *cb_sk_object;
68	} JSON;	80	} JSON;
69		81
70	/////////////////////////////////////////////////////////////////////////////	82	/////////////////////////////////////////////////////////////////////////////
71	// utility functions	83	// utility functions
72		84
73	inline void	85	INLINE void
74	shrink (SV *sv)	86	shrink (SV *sv)
75	{	87	{
76	sv_utf8_downgrade (sv, 1);	88	sv_utf8_downgrade (sv, 1);
77	if (SvLEN (sv) > SvCUR (sv) + 1)	89	if (SvLEN (sv) > SvCUR (sv) + 1)
78	{	90	{
…		…
87	// 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
88	// case of an error.	100	// case of an error.
89	// we special-case "safe" characters from U+80 .. U+7FF,	101	// we special-case "safe" characters from U+80 .. U+7FF,
90	// but use the very good perl function to parse anything else.	102	// but use the very good perl function to parse anything else.
91	// note that we never call this function for a ascii codepoints	103	// note that we never call this function for a ascii codepoints
92	inline UV	104	INLINE UV
93	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	105	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
94	{	106	{
95	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	107	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
…		…
116	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
117	SV *sv; // result scalar	129	SV *sv; // result scalar
118	JSON json;	130	JSON json;
119	U32 indent; // indentation level	131	U32 indent; // indentation level
120	U32 maxdepth; // max. indentation/recursion level	132	U32 maxdepth; // max. indentation/recursion level
		133	UV limit; // escape character values >= this value when encoding
121	} enc_t;	134	} enc_t;
122		135
123	inline void	136	INLINE void
124	need (enc_t *enc, STRLEN len)	137	need (enc_t *enc, STRLEN len)
125	{	138	{
126	if (expect_false (enc->cur + len >= enc->end))	139	if (expect_false (enc->cur + len >= enc->end))
127	{	140	{
128	STRLEN cur = enc->cur - SvPVX (enc->sv);	141	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
130	enc->cur = SvPVX (enc->sv) + cur;	143	enc->cur = SvPVX (enc->sv) + cur;
131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	144	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
132	}	145	}
133	}	146	}
134		147
135	inline void	148	INLINE void
136	encode_ch (enc_t *enc, char ch)	149	encode_ch (enc_t *enc, char ch)
137	{	150	{
138	need (enc, 1);	151	need (enc, 1);
139	*enc->cur++ = ch;	152	*enc->cur++ = ch;
140	}	153	}
…		…
194	{	207	{
195	uch = ch;	208	uch = ch;
196	clen = 1;	209	clen = 1;
197	}	210	}
198		211
199	if (uch > 0x10FFFFUL)	212	if (uch < 0x20 \|\| uch >= enc->limit)
200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
201
202	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
203	{	213	{
204	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
205	{	215	{
		216	if (uch > 0x10FFFFUL)
		217	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		218
206	need (enc, len += 11);	219	need (enc, len += 11);
207	sprintf (enc->cur, "\\u%04x\\u%04x",	220	sprintf (enc->cur, "\\u%04x\\u%04x",
208	(int)((uch - 0x10000) / 0x400 + 0xD800),	221	(int)((uch - 0x10000) / 0x400 + 0xD800),
209	(int)((uch - 0x10000) % 0x400 + 0xDC00));	222	(int)((uch - 0x10000) % 0x400 + 0xDC00));
210	enc->cur += 12;	223	enc->cur += 12;
…		…
249		262
250	--len;	263	--len;
251	}	264	}
252	}	265	}
253		266
254	inline void	267	INLINE void
255	encode_indent (enc_t *enc)	268	encode_indent (enc_t *enc)
256	{	269	{
257	if (enc->json.flags & F_INDENT)	270	if (enc->json.flags & F_INDENT)
258	{	271	{
259	int spaces = enc->indent * INDENT_STEP;	272	int spaces = enc->indent * INDENT_STEP;
…		…
262	memset (enc->cur, ' ', spaces);	275	memset (enc->cur, ' ', spaces);
263	enc->cur += spaces;	276	enc->cur += spaces;
264	}	277	}
265	}	278	}
266		279
267	inline void	280	INLINE void
268	encode_space (enc_t *enc)	281	encode_space (enc_t *enc)
269	{	282	{
270	need (enc, 1);	283	need (enc, 1);
271	encode_ch (enc, ' ');	284	encode_ch (enc, ' ');
272	}	285	}
273		286
274	inline void	287	INLINE void
275	encode_nl (enc_t *enc)	288	encode_nl (enc_t *enc)
276	{	289	{
277	if (enc->json.flags & F_INDENT)	290	if (enc->json.flags & F_INDENT)
278	{	291	{
279	need (enc, 1);	292	need (enc, 1);
280	encode_ch (enc, '\n');	293	encode_ch (enc, '\n');
281	}	294	}
282	}	295	}
283		296
284	inline void	297	INLINE void
285	encode_comma (enc_t *enc)	298	encode_comma (enc_t *enc)
286	{	299	{
287	encode_ch (enc, ',');	300	encode_ch (enc, ',');
288		301
289	if (enc->json.flags & F_INDENT)	302	if (enc->json.flags & F_INDENT)
…		…
300	int i, len = av_len (av);	313	int i, len = av_len (av);
301		314
302	if (enc->indent >= enc->maxdepth)	315	if (enc->indent >= enc->maxdepth)
303	croak ("data structure too deep (hit recursion limit)");	316	croak ("data structure too deep (hit recursion limit)");
304		317
305	encode_ch (enc, '['); encode_nl (enc);	318	encode_ch (enc, '[');
306	++enc->indent;	319
		320	if (len >= 0)
		321	{
		322	encode_nl (enc); ++enc->indent;
307		323
308	for (i = 0; i <= len; ++i)	324	for (i = 0; i <= len; ++i)
309	{	325	{
		326	SV **svp = av_fetch (av, i, 0);
		327
310	encode_indent (enc);	328	encode_indent (enc);
311	encode_sv (enc, *av_fetch (av, i, 0));
312		329
		330	if (svp)
		331	encode_sv (enc, *svp);
		332	else
		333	encode_str (enc, "null", 4, 0);
		334
313	if (i < len)	335	if (i < len)
314	encode_comma (enc);	336	encode_comma (enc);
315	}	337	}
316		338
		339	encode_nl (enc); --enc->indent; encode_indent (enc);
		340	}
		341
317	encode_nl (enc);	342	encode_ch (enc, ']');
318
319	--enc->indent;
320	encode_indent (enc); encode_ch (enc, ']');
321	}	343	}
322		344
323	static void	345	static void
324	encode_he (enc_t enc, HE he)	346	encode_hk (enc_t enc, HE he)
325	{	347	{
326	encode_ch (enc, '"');	348	encode_ch (enc, '"');
327		349
328	if (HeKLEN (he) == HEf_SVKEY)	350	if (HeKLEN (he) == HEf_SVKEY)
329	{	351	{
…		…
342	encode_ch (enc, '"');	364	encode_ch (enc, '"');
343		365
344	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	366	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
345	encode_ch (enc, ':');	367	encode_ch (enc, ':');
346	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	368	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
347	encode_sv (enc, HeVAL (he));
348	}	369	}
349		370
350	// compare hash entries, used when all keys are bytestrings	371	// compare hash entries, used when all keys are bytestrings
351	static int	372	static int
352	he_cmp_fast (const void a_, const void b_)	373	he_cmp_fast (const void a_, const void b_)
…		…
357	HE b = (HE **)b_;	378	HE b = (HE **)b_;
358		379
359	STRLEN la = HeKLEN (a);	380	STRLEN la = HeKLEN (a);
360	STRLEN lb = HeKLEN (b);	381	STRLEN lb = HeKLEN (b);
361		382
362	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	383	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
363	cmp = la - lb;	384	cmp = lb - la;
364		385
365	return cmp;	386	return cmp;
366	}	387	}
367		388
368	// 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
369	static int	390	static int
370	he_cmp_slow (const void a, const void b)	391	he_cmp_slow (const void a, const void b)
371	{	392	{
372	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	393	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
373	}	394	}
374		395
375	static void	396	static void
376	encode_hv (enc_t enc, HV hv)	397	encode_hv (enc_t enc, HV hv)
377	{	398	{
		399	HE *he;
378	int count, i;	400	int count;
379		401
380	if (enc->indent >= enc->maxdepth)	402	if (enc->indent >= enc->maxdepth)
381	croak ("data structure too deep (hit recursion limit)");	403	croak ("data structure too deep (hit recursion limit)");
382		404
383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	405	encode_ch (enc, '{');
384		406
385	if ((count = hv_iterinit (hv)))
386	{
387	// for canonical output we have to sort by keys first	407	// for canonical output we have to sort by keys first
388	// actually, this is mostly due to the stupid so-called	408	// actually, this is mostly due to the stupid so-called
389	// security workaround added somewhere in 5.8.x.	409	// security workaround added somewhere in 5.8.x.
390	// that randomises hash orderings	410	// that randomises hash orderings
391	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))
392	{	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	{
393	int fast = 1;	431	int i, fast = 1;
394	HE *he;
395	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	432	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
396	HE *hes = _alloca (count sizeof (HE));	433	HE *hes = _alloca (count sizeof (HE));
397	#else	434	#else
398	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
399	#endif	436	#endif
…		…
426		463
427	FREETMPS;	464	FREETMPS;
428	LEAVE;	465	LEAVE;
429	}	466	}
430		467
431	for (i = 0; i < count; ++i)	468	encode_nl (enc); ++enc->indent;
		469
		470	while (count--)
432	{	471	{
433	encode_indent (enc);	472	encode_indent (enc);
		473	he = hes [count];
434	encode_he (enc, hes [i]);	474	encode_hk (enc, he);
		475	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
435		476
436	if (i < count - 1)	477	if (count)
437	encode_comma (enc);	478	encode_comma (enc);
438	}	479	}
439		480
440	encode_nl (enc);	481	encode_nl (enc); --enc->indent; encode_indent (enc);
441	}	482	}
		483	}
442	else	484	else
		485	{
		486	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		487	if ((he = hv_iternext (hv)))
443	{	488	{
444	HE *he = hv_iternext (hv);	489	encode_nl (enc); ++enc->indent;
445		490
446	for (;;)	491	for (;;)
447	{	492	{
448	encode_indent (enc);	493	encode_indent (enc);
449	encode_he (enc, he);	494	encode_hk (enc, he);
		495	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
450		496
451	if (!(he = hv_iternext (hv)))	497	if (!(he = hv_iternext (hv)))
452	break;	498	break;
453		499
454	encode_comma (enc);	500	encode_comma (enc);
455	}	501	}
456		502
457	encode_nl (enc);	503	encode_nl (enc); --enc->indent; encode_indent (enc);
458	}	504	}
459	}	505	}
460		506
461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	507	encode_ch (enc, '}');
462	}	508	}
463		509
464	// encode objects, arrays and special \0=false and \1=true values.	510	// encode objects, arrays and special \0=false and \1=true values.
465	static void	511	static void
466	encode_rv (enc_t enc, SV sv)	512	encode_rv (enc_t enc, SV sv)
…		…
470	SvGETMAGIC (sv);	516	SvGETMAGIC (sv);
471	svt = SvTYPE (sv);	517	svt = SvTYPE (sv);
472		518
473	if (expect_false (SvOBJECT (sv)))	519	if (expect_false (SvOBJECT (sv)))
474	{	520	{
		521	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		522	? json_boolean_stash
		523	: gv_stashpv ("JSON::XS::Boolean", 1);
		524
475	if (SvSTASH (sv) == json_boolean_stash)	525	if (SvSTASH (sv) == stash)
476	{	526	{
477	if (SvIV (sv) == 0)	527	if (SvIV (sv))
		528	encode_str (enc, "true", 4, 0);
		529	else
478	encode_str (enc, "false", 5, 0);	530	encode_str (enc, "false", 5, 0);
479	else
480	encode_str (enc, "true", 4, 0);
481	}	531	}
482	else	532	else
483	{	533	{
484	#if 0	534	#if 0
485	if (0 && sv_derived_from (rv, "JSON::Literal"))	535	if (0 && sv_derived_from (rv, "JSON::Literal"))
…		…
488	}	538	}
489	#endif	539	#endif
490	if (enc->json.flags & F_CONV_BLESSED)	540	if (enc->json.flags & F_CONV_BLESSED)
491	{	541	{
492	// 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
493	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	543	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
494		544
495	if (to_json)	545	if (to_json)
496	{	546	{
		547	dSP;
		548
497	dSP; ENTER; SAVETMPS; PUSHMARK (SP);	549	ENTER; SAVETMPS; PUSHMARK (SP);
498	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	550	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
499		551
500	// 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
501	// check anyways.
502	PUTBACK;	553	PUTBACK;
503	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	554	call_sv ((SV *)GvCV (to_json), G_SCALAR);
504	SPAGAIN;	555	SPAGAIN;
505		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
506	encode_sv (enc, POPs);	564	encode_sv (enc, sv);
507		565
508	FREETMPS; LEAVE;	566	FREETMPS; LEAVE;
509	}	567	}
510	else if (enc->json.flags & F_ALLOW_BLESSED)	568	else if (enc->json.flags & F_ALLOW_BLESSED)
511	encode_str (enc, "null", 4, 0);	569	encode_str (enc, "null", 4, 0);
…		…
524	encode_hv (enc, (HV *)sv);	582	encode_hv (enc, (HV *)sv);
525	else if (svt == SVt_PVAV)	583	else if (svt == SVt_PVAV)
526	encode_av (enc, (AV *)sv);	584	encode_av (enc, (AV *)sv);
527	else if (svt < SVt_PVAV)	585	else if (svt < SVt_PVAV)
528	{	586	{
529	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')
530	encode_str (enc, "false", 5, 0);	593	encode_str (enc, "false", 5, 0);
531	else if (SvNIOK (sv) && SvIV (sv) == 1)
532	encode_str (enc, "true", 4, 0);
533	else	594	else
534	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",
535	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
536	}	597	}
537	else	598	else
…		…
620	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	681	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
621	enc.cur = SvPVX (enc.sv);	682	enc.cur = SvPVX (enc.sv);
622	enc.end = SvEND (enc.sv);	683	enc.end = SvEND (enc.sv);
623	enc.indent = 0;	684	enc.indent = 0;
624	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;
625		689
626	SvPOK_only (enc.sv);	690	SvPOK_only (enc.sv);
627	encode_sv (&enc, scalar);	691	encode_sv (&enc, scalar);
628		692
629	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	693	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
650	JSON json;	714	JSON json;
651	U32 depth; // recursion depth	715	U32 depth; // recursion depth
652	U32 maxdepth; // recursion depth limit	716	U32 maxdepth; // recursion depth limit
653	} dec_t;	717	} dec_t;
654		718
655	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
656	decode_ws (dec_t *dec)	729	decode_ws (dec_t *dec)
657	{	730	{
658	for (;;)	731	for (;;)
659	{	732	{
660	char ch = *dec->cur;	733	char ch = *dec->cur;
661		734
662	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	}
663	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	747	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
664	break;	748	break; // parse error, but let higher level handle it, gives better error messages
665		749
666	++dec->cur;	750	++dec->cur;
667	}	751	}
668	}	752	}
669		753
…		…
918	is_nv = 1;	1002	is_nv = 1;
919	}	1003	}
920		1004
921	if (!is_nv)	1005	if (!is_nv)
922	{	1006	{
		1007	int len = dec->cur - start;
		1008
923	// 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
924	if (*start == '-')	1010	if (*start == '-')
925	switch (dec->cur - start)	1011	switch (len)
926	{	1012	{
927	case 2: return newSViv (-( start [1] - '0' * 1));	1013	case 2: return newSViv (-( start [1] - '0' * 1));
928	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1014	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
929	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));
930	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));
931	}	1018	}
932	else	1019	else
933	switch (dec->cur - start)	1020	switch (len)
934	{	1021	{
935	case 1: return newSViv ( start [0] - '0' * 1);	1022	case 1: return newSViv ( start [0] - '0' * 1);
936	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1023	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
937	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);
938	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);
939	}	1027	}
940		1028
941	{	1029	{
942	UV uv;	1030	UV uv;
943	int numtype = grok_number (start, dec->cur - start, &uv);	1031	int numtype = grok_number (start, len, &uv);
944	if (numtype & IS_NUMBER_IN_UV)	1032	if (numtype & IS_NUMBER_IN_UV)
945	if (numtype & IS_NUMBER_NEG)	1033	if (numtype & IS_NUMBER_NEG)
946	{	1034	{
947	if (uv < (UV)IV_MIN)	1035	if (uv < (UV)IV_MIN)
948	return newSViv (-(IV)uv);	1036	return newSViv (-(IV)uv);
949	}	1037	}
950	else	1038	else
951	return newSVuv (uv);	1039	return newSVuv (uv);
952
953	// here would likely be the place for bigint support
954	}	1040	}
955	}
956		1041
957	// 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
958	return newSVnv (Atof (start));	1058	return newSVnv (Atof (start));
959		1059
960	fail:	1060	fail:
961	return 0;	1061	return 0;
962	}	1062	}
…		…
992		1092
993	if (*dec->cur != ',')	1093	if (*dec->cur != ',')
994	ERR (", or ] expected while parsing array");	1094	ERR (", or ] expected while parsing array");
995		1095
996	++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	}
997	}	1105	}
998		1106
999	DEC_DEC_DEPTH;	1107	DEC_DEC_DEPTH;
1000	return newRV_noinc ((SV *)av);	1108	return newRV_noinc ((SV *)av);
1001		1109
…		…
1017	if (*dec->cur == '}')	1125	if (*dec->cur == '}')
1018	++dec->cur;	1126	++dec->cur;
1019	else	1127	else
1020	for (;;)	1128	for (;;)
1021	{	1129	{
1022	decode_ws (dec); EXPECT_CH ('"');	1130	EXPECT_CH ('"');
1023		1131
1024	// heuristic: assume that	1132	// heuristic: assume that
1025	// a) decode_str + hv_store_ent are abysmally slow.	1133	// a) decode_str + hv_store_ent are abysmally slow.
1026	// b) most hash keys are short, simple ascii text.	1134	// b) most hash keys are short, simple ascii text.
1027	// => try to "fast-match" such strings to avoid	1135	// => try to "fast-match" such strings to avoid
…		…
1041	if (!key)	1149	if (!key)
1042	goto fail;	1150	goto fail;
1043		1151
1044	decode_ws (dec); EXPECT_CH (':');	1152	decode_ws (dec); EXPECT_CH (':');
1045		1153
		1154	decode_ws (dec);
1046	value = decode_sv (dec);	1155	value = decode_sv (dec);
1047	if (!value)	1156	if (!value)
1048	{	1157	{
1049	SvREFCNT_dec (key);	1158	SvREFCNT_dec (key);
1050	goto fail;	1159	goto fail;
…		…
1062	int len = p - key;	1171	int len = p - key;
1063	dec->cur = p + 1;	1172	dec->cur = p + 1;
1064		1173
1065	decode_ws (dec); EXPECT_CH (':');	1174	decode_ws (dec); EXPECT_CH (':');
1066		1175
		1176	decode_ws (dec);
1067	value = decode_sv (dec);	1177	value = decode_sv (dec);
1068	if (!value)	1178	if (!value)
1069	goto fail;	1179	goto fail;
1070		1180
1071	hv_store (hv, key, len, value, 0);	1181	hv_store (hv, key, len, value, 0);
…		…
1087		1197
1088	if (*dec->cur != ',')	1198	if (*dec->cur != ',')
1089	ERR (", or } expected while parsing object/hash");	1199	ERR (", or } expected while parsing object/hash");
1090		1200
1091	++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	}
1092	}	1210	}
1093		1211
1094	DEC_DEC_DEPTH;	1212	DEC_DEC_DEPTH;
1095	sv = newRV_noinc ((SV *)hv);	1213	sv = newRV_noinc ((SV *)hv);
1096		1214
1097	// check filter callbacks	1215	// check filter callbacks
1098	if (dec->json.flags & F_HOOK)	1216	if (dec->json.flags & F_HOOK)
1099	{	1217	{
1100	ENTER; SAVETMPS;	1218	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
1101
1102	if (HvKEYS (hv) == 1 && dec->json.cb_sk_object)
1103	{	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;
1104	int count;	1233	int count;
1105		1234
1106	dSP; PUSHMARK (SP);	1235	ENTER; SAVETMPS; PUSHMARK (SP);
1107	XPUSHs (sv_2mortal (sv));	1236	XPUSHs (HeVAL (he));
1108		1237
1109	PUTBACK; count = call_sv (dec->json.cb_sk_object, G_ARRAY); SPAGAIN;	1238	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1110		1239
1111	if (count == 1)	1240	if (count == 1)
		1241	{
1112	sv = newSVsv (POPs);	1242	sv = newSVsv (POPs);
1113	else	1243	FREETMPS; LEAVE;
1114	SvREFCNT_inc (sv);	1244	return sv;
		1245	}
		1246
		1247	FREETMPS; LEAVE;
		1248	}
1115	}	1249	}
1116		1250
1117	if (dec->json.cb_object)	1251	if (dec->json.cb_object)
1118	{	1252	{
		1253	dSP;
1119	int count;	1254	int count;
1120		1255
1121	dSP; ENTER; SAVETMPS; PUSHMARK (SP);	1256	ENTER; SAVETMPS; PUSHMARK (SP);
1122	XPUSHs (sv_2mortal (sv));	1257	XPUSHs (sv_2mortal (sv));
1123		1258
1124	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1259	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1125		1260
1126	if (count == 1)	1261	if (count == 1)
		1262	{
1127	sv = newSVsv (POPs);	1263	sv = newSVsv (POPs);
1128	else	1264	FREETMPS; LEAVE;
		1265	return sv;
		1266	}
		1267
1129	SvREFCNT_inc (sv);	1268	SvREFCNT_inc (sv);
		1269	FREETMPS; LEAVE;
1130	}	1270	}
1131
1132	FREETMPS; LEAVE;
1133	}	1271	}
1134		1272
1135	return newRV_noinc ((SV *)hv);	1273	return sv;
1136		1274
1137	fail:	1275	fail:
1138	SvREFCNT_dec (hv);	1276	SvREFCNT_dec (hv);
1139	DEC_DEC_DEPTH;	1277	DEC_DEC_DEPTH;
1140	return 0;	1278	return 0;
1141	}	1279	}
1142		1280
1143	static SV *	1281	static SV *
1144	decode_sv (dec_t *dec)	1282	decode_sv (dec_t *dec)
1145	{	1283	{
1146	decode_ws (dec);
1147
1148	// the beauty of JSON: you need exactly one character lookahead	1284	// the beauty of JSON: you need exactly one character lookahead
1149	// to parse anything.	1285	// to parse anything.
1150	switch (*dec->cur)	1286	switch (*dec->cur)
1151	{	1287	{
1152	case '"': ++dec->cur; return decode_str (dec);	1288	case '"': ++dec->cur; return decode_str (dec);
…		…
1160		1296
1161	case 't':	1297	case 't':
1162	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1298	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1163	{	1299	{
1164	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
1165	return SvREFCNT_inc (json_true);	1304	return SvREFCNT_inc (json_true);
1166	}	1305	}
1167	else	1306	else
1168	ERR ("'true' expected");	1307	ERR ("'true' expected");
1169		1308
…		…
1171		1310
1172	case 'f':	1311	case 'f':
1173	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1312	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1174	{	1313	{
1175	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
1176	return SvREFCNT_inc (json_false);	1318	return SvREFCNT_inc (json_false);
1177	}	1319	}
1178	else	1320	else
1179	ERR ("'false' expected");	1321	ERR ("'false' expected");
1180		1322
…		…
1230		1372
1231	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1373	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1232	dec.json.flags \|= F_HOOK;	1374	dec.json.flags \|= F_HOOK;
1233		1375
1234	*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);
1235	sv = decode_sv (&dec);	1379	sv = decode_sv (&dec);
1236		1380
1237	if (!(offset_return \|\| !sv))	1381	if (!(offset_return \|\| !sv))
1238	{	1382	{
1239	// check for trailing garbage	1383	// check for trailing garbage
…		…
1307	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);
1308	}	1452	}
1309		1453
1310	PROTOTYPES: DISABLE	1454	PROTOTYPES: DISABLE
1311		1455
		1456	void CLONE (...)
		1457	CODE:
		1458	json_stash = 0;
		1459	json_boolean_stash = 0;
		1460
1312	void new (char *klass)	1461	void new (char *klass)
1313	PPCODE:	1462	PPCODE:
1314	{	1463	{
1315	SV *pv = NEWSV (0, sizeof (JSON));	1464	SV *pv = NEWSV (0, sizeof (JSON));
1316	SvPOK_only (pv);	1465	SvPOK_only (pv);
1317	Zero (SvPVX (pv), 1, JSON);	1466	Zero (SvPVX (pv), 1, JSON);
1318	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1467	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1319	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	)));
1320	}	1472	}
1321		1473
1322	void ascii (JSON *self, int enable = 1)	1474	void ascii (JSON *self, int enable = 1)
1323	ALIAS:	1475	ALIAS:
1324	ascii = F_ASCII	1476	ascii = F_ASCII
…		…
1331	pretty = F_PRETTY	1483	pretty = F_PRETTY
1332	allow_nonref = F_ALLOW_NONREF	1484	allow_nonref = F_ALLOW_NONREF
1333	shrink = F_SHRINK	1485	shrink = F_SHRINK
1334	allow_blessed = F_ALLOW_BLESSED	1486	allow_blessed = F_ALLOW_BLESSED
1335	convert_blessed = F_CONV_BLESSED	1487	convert_blessed = F_CONV_BLESSED
		1488	relaxed = F_RELAXED
1336	PPCODE:	1489	PPCODE:
1337	{	1490	{
1338	if (enable)	1491	if (enable)
1339	self->flags \|= ix;	1492	self->flags \|= ix;
1340	else	1493	else
1341	self->flags &= ~ix;	1494	self->flags &= ~ix;
1342		1495
1343	XPUSHs (ST (0));	1496	XPUSHs (ST (0));
1344	}	1497	}
1345		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
1346	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1516	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1347	PPCODE:	1517	PPCODE:
1348	{	1518	{
1349	UV log2 = 0;	1519	UV log2 = 0;
1350		1520
…		…
1356	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1526	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1357		1527
1358	XPUSHs (ST (0));	1528	XPUSHs (ST (0));
1359	}	1529	}
1360		1530
		1531	U32 get_max_depth (JSON *self)
		1532	CODE:
		1533	RETVAL = DEC_DEPTH (self->flags);
		1534	OUTPUT:
		1535	RETVAL
		1536
1361	void max_size (JSON *self, UV max_size = 0)	1537	void max_size (JSON *self, UV max_size = 0)
1362	PPCODE:	1538	PPCODE:
1363	{	1539	{
1364	UV log2 = 0;	1540	UV log2 = 0;
1365		1541
…		…
1372	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1548	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1373		1549
1374	XPUSHs (ST (0));	1550	XPUSHs (ST (0));
1375	}	1551	}
1376		1552
		1553	int get_max_size (JSON *self)
		1554	CODE:
		1555	RETVAL = DEC_SIZE (self->flags);
		1556	OUTPUT:
		1557	RETVAL
		1558
1377	void filter_json_objects (JSON self, SV cb = &PL_sv_undef)	1559	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1378	ALIAS:
1379	filter_sk_json_objects = 1
1380	PPCODE:	1560	PPCODE:
1381	{	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
1382	if (!SvOK (cb))	1574	if (SvOK (cb))
1383	cb = 0;	1575	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1384		1576	else
1385	switch (ix)
1386	{	1577	{
1387	case 0: self->cb_object = cb; break;	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);
1388	case 1: self->cb_sk_object = cb; break;	1583	self->cb_sk_object = 0;
		1584	}
1389	}	1585	}
1390		1586
1391	XPUSHs (ST (0));	1587	XPUSHs (ST (0));
1392	}	1588	}
1393		1589
…		…
1406	EXTEND (SP, 2);	1602	EXTEND (SP, 2);
1407	PUSHs (decode_json (jsonstr, self, &offset));	1603	PUSHs (decode_json (jsonstr, self, &offset));
1408	PUSHs (sv_2mortal (newSVuv (offset)));	1604	PUSHs (sv_2mortal (newSVuv (offset)));
1409	}	1605	}
1410		1606
		1607	void DESTROY (JSON *self)
		1608	CODE:
		1609	SvREFCNT_dec (self->cb_sk_object);
		1610	SvREFCNT_dec (self->cb_object);
		1611
1411	PROTOTYPES: ENABLE	1612	PROTOTYPES: ENABLE
1412		1613
1413	void to_json (SV *scalar)	1614	void encode_json (SV *scalar)
1414	PPCODE:	1615	PPCODE:
1415	{	1616	{
1416	JSON json = { F_DEFAULT \| F_UTF8 };	1617	JSON json = { F_DEFAULT \| F_UTF8 };
1417	XPUSHs (encode_json (scalar, &json));	1618	XPUSHs (encode_json (scalar, &json));
1418	}	1619	}
1419		1620
1420	void from_json (SV *jsonstr)	1621	void decode_json (SV *jsonstr)
1421	PPCODE:	1622	PPCODE:
1422	{	1623	{
1423	JSON json = { F_DEFAULT \| F_UTF8 };	1624	JSON json = { F_DEFAULT \| F_UTF8 };
1424	XPUSHs (decode_json (jsonstr, &json, 0));	1625	XPUSHs (decode_json (jsonstr, &json, 0));
1425	}	1626	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.49 by root, Sun Jul 1 23:40:07 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 1.49 by root, Sun Jul 1 23:40:07 2007 UTC vs.
Revision 1.71 by root, Wed Mar 19 03:17:38 2008 UTC