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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.52 by root, Mon Jul 2 02:57:11 2007 UTC vs.
Revision 1.73 by root, Wed Mar 19 13:44:43 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)
		64
		65	#define IN_RANGE_INC(type,val,beg,end) \
		66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
61		76
62	static HV json_stash, json_boolean_stash; // JSON::XS::	77	static HV json_stash, json_boolean_stash; // JSON::XS::
63	static SV json_true, json_false;	78	static SV json_true, json_false;
64		79
65	typedef struct {	80	typedef struct {
…		…
69	} JSON;	84	} JSON;
70		85
71	/////////////////////////////////////////////////////////////////////////////	86	/////////////////////////////////////////////////////////////////////////////
72	// utility functions	87	// utility functions
73		88
74	inline void	89	INLINE void
75	shrink (SV *sv)	90	shrink (SV *sv)
76	{	91	{
77	sv_utf8_downgrade (sv, 1);	92	sv_utf8_downgrade (sv, 1);
78	if (SvLEN (sv) > SvCUR (sv) + 1)	93	if (SvLEN (sv) > SvCUR (sv) + 1)
79	{	94	{
…		…
88	// decode an utf-8 character and return it, or (UV)-1 in	103	// decode an utf-8 character and return it, or (UV)-1 in
89	// case of an error.	104	// case of an error.
90	// we special-case "safe" characters from U+80 .. U+7FF,	105	// we special-case "safe" characters from U+80 .. U+7FF,
91	// but use the very good perl function to parse anything else.	106	// but use the very good perl function to parse anything else.
92	// note that we never call this function for a ascii codepoints	107	// note that we never call this function for a ascii codepoints
93	inline UV	108	INLINE UV
94	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	109	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
95	{	110	{
96	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	111	if (expect_true (len >= 2
97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	112	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	113	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
99	{	114	{
100	*clen = 2;	115	*clen = 2;
101	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	116	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
102	}	117	}
103	else	118	else
104	{	119	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
105	*clen = (STRLEN)-1;	120	}
106	return (UV)-1;	121
		122	// likewise for encoding, also never called for ascii codepoints
		123	// this function takes advantage of this fact, although current gccs
		124	// seem to optimise the check for >= 0x80 away anyways
		125	INLINE unsigned char *
		126	encode_utf8 (unsigned char *s, UV ch)
		127	{
		128	if (ch <= 0x7FF)
107	}	129	{
		130	*s++ = (ch >> 6) \| 0xc0;
		131	*s++ = (ch & 0x3f) \| 0x80;
		132	}
		133	else
		134	s = uvuni_to_utf8_flags (s, ch, 0);
		135
		136	return s;
108	}	137	}
109		138
110	/////////////////////////////////////////////////////////////////////////////	139	/////////////////////////////////////////////////////////////////////////////
111	// encoder	140	// encoder
112		141
…		…
117	char *end; // SvEND (sv)	146	char *end; // SvEND (sv)
118	SV *sv; // result scalar	147	SV *sv; // result scalar
119	JSON json;	148	JSON json;
120	U32 indent; // indentation level	149	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	150	U32 maxdepth; // max. indentation/recursion level
		151	UV limit; // escape character values >= this value when encoding
122	} enc_t;	152	} enc_t;
123		153
124	inline void	154	INLINE void
125	need (enc_t *enc, STRLEN len)	155	need (enc_t *enc, STRLEN len)
126	{	156	{
127	if (expect_false (enc->cur + len >= enc->end))	157	if (expect_false (enc->cur + len >= enc->end))
128	{	158	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	159	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
131	enc->cur = SvPVX (enc->sv) + cur;	161	enc->cur = SvPVX (enc->sv) + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	162	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
133	}	163	}
134	}	164	}
135		165
136	inline void	166	INLINE void
137	encode_ch (enc_t *enc, char ch)	167	encode_ch (enc_t *enc, char ch)
138	{	168	{
139	need (enc, 1);	169	need (enc, 1);
140	*enc->cur++ = ch;	170	*enc->cur++ = ch;
141	}	171	}
…		…
195	{	225	{
196	uch = ch;	226	uch = ch;
197	clen = 1;	227	clen = 1;
198	}	228	}
199		229
200	if (uch > 0x10FFFFUL)	230	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
202
203	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
204	{	231	{
205	if (uch > 0xFFFFUL)	232	if (uch > 0xFFFFUL)
206	{	233	{
		234	if (uch > 0x10FFFFUL)
		235	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		236
207	need (enc, len += 11);	237	need (enc, len += 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	238	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	239	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	240	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	241	enc->cur += 12;
…		…
239	while (--clen);	269	while (--clen);
240	}	270	}
241	else	271	else
242	{	272	{
243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	273	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	274	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	275	++str;
246	}	276	}
247	}	277	}
248	}	278	}
249	}	279	}
250		280
251	--len;	281	--len;
252	}	282	}
253	}	283	}
254		284
255	inline void	285	INLINE void
256	encode_indent (enc_t *enc)	286	encode_indent (enc_t *enc)
257	{	287	{
258	if (enc->json.flags & F_INDENT)	288	if (enc->json.flags & F_INDENT)
259	{	289	{
260	int spaces = enc->indent * INDENT_STEP;	290	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	293	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	294	enc->cur += spaces;
265	}	295	}
266	}	296	}
267		297
268	inline void	298	INLINE void
269	encode_space (enc_t *enc)	299	encode_space (enc_t *enc)
270	{	300	{
271	need (enc, 1);	301	need (enc, 1);
272	encode_ch (enc, ' ');	302	encode_ch (enc, ' ');
273	}	303	}
274		304
275	inline void	305	INLINE void
276	encode_nl (enc_t *enc)	306	encode_nl (enc_t *enc)
277	{	307	{
278	if (enc->json.flags & F_INDENT)	308	if (enc->json.flags & F_INDENT)
279	{	309	{
280	need (enc, 1);	310	need (enc, 1);
281	encode_ch (enc, '\n');	311	encode_ch (enc, '\n');
282	}	312	}
283	}	313	}
284		314
285	inline void	315	INLINE void
286	encode_comma (enc_t *enc)	316	encode_comma (enc_t *enc)
287	{	317	{
288	encode_ch (enc, ',');	318	encode_ch (enc, ',');
289		319
290	if (enc->json.flags & F_INDENT)	320	if (enc->json.flags & F_INDENT)
…		…
301	int i, len = av_len (av);	331	int i, len = av_len (av);
302		332
303	if (enc->indent >= enc->maxdepth)	333	if (enc->indent >= enc->maxdepth)
304	croak ("data structure too deep (hit recursion limit)");	334	croak ("data structure too deep (hit recursion limit)");
305		335
306	encode_ch (enc, '['); encode_nl (enc);	336	encode_ch (enc, '[');
307	++enc->indent;	337
		338	if (len >= 0)
		339	{
		340	encode_nl (enc); ++enc->indent;
308		341
309	for (i = 0; i <= len; ++i)	342	for (i = 0; i <= len; ++i)
310	{	343	{
		344	SV **svp = av_fetch (av, i, 0);
		345
311	encode_indent (enc);	346	encode_indent (enc);
312	encode_sv (enc, *av_fetch (av, i, 0));
313		347
		348	if (svp)
		349	encode_sv (enc, *svp);
		350	else
		351	encode_str (enc, "null", 4, 0);
		352
314	if (i < len)	353	if (i < len)
315	encode_comma (enc);	354	encode_comma (enc);
316	}	355	}
317		356
		357	encode_nl (enc); --enc->indent; encode_indent (enc);
		358	}
		359
318	encode_nl (enc);	360	encode_ch (enc, ']');
319
320	--enc->indent;
321	encode_indent (enc); encode_ch (enc, ']');
322	}	361	}
323		362
324	static void	363	static void
325	encode_he (enc_t enc, HE he)	364	encode_hk (enc_t enc, HE he)
326	{	365	{
327	encode_ch (enc, '"');	366	encode_ch (enc, '"');
328		367
329	if (HeKLEN (he) == HEf_SVKEY)	368	if (HeKLEN (he) == HEf_SVKEY)
330	{	369	{
…		…
343	encode_ch (enc, '"');	382	encode_ch (enc, '"');
344		383
345	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	384	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
346	encode_ch (enc, ':');	385	encode_ch (enc, ':');
347	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	386	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
348	encode_sv (enc, HeVAL (he));
349	}	387	}
350		388
351	// compare hash entries, used when all keys are bytestrings	389	// compare hash entries, used when all keys are bytestrings
352	static int	390	static int
353	he_cmp_fast (const void a_, const void b_)	391	he_cmp_fast (const void a_, const void b_)
…		…
358	HE b = (HE **)b_;	396	HE b = (HE **)b_;
359		397
360	STRLEN la = HeKLEN (a);	398	STRLEN la = HeKLEN (a);
361	STRLEN lb = HeKLEN (b);	399	STRLEN lb = HeKLEN (b);
362		400
363	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	401	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
364	cmp = la - lb;	402	cmp = lb - la;
365		403
366	return cmp;	404	return cmp;
367	}	405	}
368		406
369	// compare hash entries, used when some keys are sv's or utf-x	407	// compare hash entries, used when some keys are sv's or utf-x
370	static int	408	static int
371	he_cmp_slow (const void a, const void b)	409	he_cmp_slow (const void a, const void b)
372	{	410	{
373	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	411	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
374	}	412	}
375		413
376	static void	414	static void
377	encode_hv (enc_t enc, HV hv)	415	encode_hv (enc_t enc, HV hv)
378	{	416	{
		417	HE *he;
379	int count, i;	418	int count;
380		419
381	if (enc->indent >= enc->maxdepth)	420	if (enc->indent >= enc->maxdepth)
382	croak ("data structure too deep (hit recursion limit)");	421	croak ("data structure too deep (hit recursion limit)");
383		422
384	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	423	encode_ch (enc, '{');
385		424
386	if ((count = hv_iterinit (hv)))
387	{
388	// for canonical output we have to sort by keys first	425	// for canonical output we have to sort by keys first
389	// actually, this is mostly due to the stupid so-called	426	// actually, this is mostly due to the stupid so-called
390	// security workaround added somewhere in 5.8.x.	427	// security workaround added somewhere in 5.8.x.
391	// that randomises hash orderings	428	// that randomises hash orderings
392	if (enc->json.flags & F_CANONICAL)	429	if (enc->json.flags & F_CANONICAL)
		430	{
		431	int count = hv_iterinit (hv);
		432
		433	if (SvMAGICAL (hv))
393	{	434	{
		435	// need to count by iterating. could improve by dynamically building the vector below
		436	// but I don't care for the speed of this special case.
		437	// note also that we will run into undefined behaviour when the two iterations
		438	// do not result in the same count, something I might care for in some later release.
		439
		440	count = 0;
		441	while (hv_iternext (hv))
		442	++count;
		443
		444	hv_iterinit (hv);
		445	}
		446
		447	if (count)
		448	{
394	int fast = 1;	449	int i, fast = 1;
395	HE *he;
396	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	450	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
397	HE *hes = _alloca (count sizeof (HE));	451	HE *hes = _alloca (count sizeof (HE));
398	#else	452	#else
399	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	453	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
400	#endif	454	#endif
…		…
427		481
428	FREETMPS;	482	FREETMPS;
429	LEAVE;	483	LEAVE;
430	}	484	}
431		485
432	for (i = 0; i < count; ++i)	486	encode_nl (enc); ++enc->indent;
		487
		488	while (count--)
433	{	489	{
434	encode_indent (enc);	490	encode_indent (enc);
		491	he = hes [count];
435	encode_he (enc, hes [i]);	492	encode_hk (enc, he);
		493	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
436		494
437	if (i < count - 1)	495	if (count)
438	encode_comma (enc);	496	encode_comma (enc);
439	}	497	}
440		498
441	encode_nl (enc);	499	encode_nl (enc); --enc->indent; encode_indent (enc);
442	}	500	}
		501	}
443	else	502	else
		503	{
		504	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		505	if ((he = hv_iternext (hv)))
444	{	506	{
445	HE *he = hv_iternext (hv);	507	encode_nl (enc); ++enc->indent;
446		508
447	for (;;)	509	for (;;)
448	{	510	{
449	encode_indent (enc);	511	encode_indent (enc);
450	encode_he (enc, he);	512	encode_hk (enc, he);
		513	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
451		514
452	if (!(he = hv_iternext (hv)))	515	if (!(he = hv_iternext (hv)))
453	break;	516	break;
454		517
455	encode_comma (enc);	518	encode_comma (enc);
456	}	519	}
457		520
458	encode_nl (enc);	521	encode_nl (enc); --enc->indent; encode_indent (enc);
459	}	522	}
460	}	523	}
461		524
462	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	525	encode_ch (enc, '}');
463	}	526	}
464		527
465	// encode objects, arrays and special \0=false and \1=true values.	528	// encode objects, arrays and special \0=false and \1=true values.
466	static void	529	static void
467	encode_rv (enc_t enc, SV sv)	530	encode_rv (enc_t enc, SV sv)
…		…
471	SvGETMAGIC (sv);	534	SvGETMAGIC (sv);
472	svt = SvTYPE (sv);	535	svt = SvTYPE (sv);
473		536
474	if (expect_false (SvOBJECT (sv)))	537	if (expect_false (SvOBJECT (sv)))
475	{	538	{
		539	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		540	? json_boolean_stash
		541	: gv_stashpv ("JSON::XS::Boolean", 1);
		542
476	if (SvSTASH (sv) == json_boolean_stash)	543	if (SvSTASH (sv) == stash)
477	{	544	{
478	if (SvIV (sv))	545	if (SvIV (sv))
479	encode_str (enc, "true", 4, 0);	546	encode_str (enc, "true", 4, 0);
480	else	547	else
481	encode_str (enc, "false", 5, 0);	548	encode_str (enc, "false", 5, 0);
…		…
489	}	556	}
490	#endif	557	#endif
491	if (enc->json.flags & F_CONV_BLESSED)	558	if (enc->json.flags & F_CONV_BLESSED)
492	{	559	{
493	// we re-bless the reference to get overload and other niceties right	560	// we re-bless the reference to get overload and other niceties right
494	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);	561	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
495		562
496	if (to_json)	563	if (to_json)
497	{	564	{
		565	dSP;
		566
498	dSP; ENTER; SAVETMPS; PUSHMARK (SP);	567	ENTER; SAVETMPS; PUSHMARK (SP);
499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	568	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
500		569
501	// calling with G_SCALAR ensures that we always get a 1 reutrn value	570	// calling with G_SCALAR ensures that we always get a 1 return value
502	// check anyways.
503	PUTBACK;	571	PUTBACK;
504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));	572	call_sv ((SV *)GvCV (to_json), G_SCALAR);
505	SPAGAIN;	573	SPAGAIN;
506		574
		575	// catch this surprisingly common error
		576	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		577	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		578
		579	sv = POPs;
		580	PUTBACK;
		581
507	encode_sv (enc, POPs);	582	encode_sv (enc, sv);
508		583
509	FREETMPS; LEAVE;	584	FREETMPS; LEAVE;
510	}	585	}
511	else if (enc->json.flags & F_ALLOW_BLESSED)	586	else if (enc->json.flags & F_ALLOW_BLESSED)
512	encode_str (enc, "null", 4, 0);	587	encode_str (enc, "null", 4, 0);
…		…
624	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	699	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
625	enc.cur = SvPVX (enc.sv);	700	enc.cur = SvPVX (enc.sv);
626	enc.end = SvEND (enc.sv);	701	enc.end = SvEND (enc.sv);
627	enc.indent = 0;	702	enc.indent = 0;
628	enc.maxdepth = DEC_DEPTH (enc.json.flags);	703	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		704	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		705	: enc.json.flags & F_LATIN1 ? 0x000100UL
		706	: 0x10FFFFUL;
629		707
630	SvPOK_only (enc.sv);	708	SvPOK_only (enc.sv);
631	encode_sv (&enc, scalar);	709	encode_sv (&enc, scalar);
632		710
633	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	711	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
654	JSON json;	732	JSON json;
655	U32 depth; // recursion depth	733	U32 depth; // recursion depth
656	U32 maxdepth; // recursion depth limit	734	U32 maxdepth; // recursion depth limit
657	} dec_t;	735	} dec_t;
658		736
659	inline void	737	INLINE void
		738	decode_comment (dec_t *dec)
		739	{
		740	// only '#'-style comments allowed a.t.m.
		741
		742	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		743	++dec->cur;
		744	}
		745
		746	INLINE void
660	decode_ws (dec_t *dec)	747	decode_ws (dec_t *dec)
661	{	748	{
662	for (;;)	749	for (;;)
663	{	750	{
664	char ch = *dec->cur;	751	char ch = *dec->cur;
665		752
666	if (ch > 0x20	753	if (ch > 0x20)
		754	{
		755	if (expect_false (ch == '#'))
		756	{
		757	if (dec->json.flags & F_RELAXED)
		758	decode_comment (dec);
		759	else
		760	break;
		761	}
		762	else
		763	break;
		764	}
667	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	765	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
668	break;	766	break; // parse error, but let higher level handle it, gives better error messages
669		767
670	++dec->cur;	768	++dec->cur;
671	}	769	}
672	}	770	}
673		771
…		…
779		877
780	if (hi >= 0x80)	878	if (hi >= 0x80)
781	{	879	{
782	utf8 = 1;	880	utf8 = 1;
783		881
784	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	882	cur = encode_utf8 (cur, hi);
785	}	883	}
786	else	884	else
787	*cur++ = hi;	885	*cur++ = hi;
788	}	886	}
789	break;	887	break;
…		…
922	is_nv = 1;	1020	is_nv = 1;
923	}	1021	}
924		1022
925	if (!is_nv)	1023	if (!is_nv)
926	{	1024	{
		1025	int len = dec->cur - start;
		1026
927	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1027	// special case the rather common 1..5-digit-int case
928	if (*start == '-')	1028	if (*start == '-')
929	switch (dec->cur - start)	1029	switch (len)
930	{	1030	{
931	case 2: return newSViv (-( start [1] - '0' * 1));	1031	case 2: return newSViv (-( start [1] - '0' * 1));
932	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1032	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
933	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1033	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
934	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1034	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1035	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
935	}	1036	}
936	else	1037	else
937	switch (dec->cur - start)	1038	switch (len)
938	{	1039	{
939	case 1: return newSViv ( start [0] - '0' * 1);	1040	case 1: return newSViv ( start [0] - '0' * 1);
940	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1041	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
941	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1042	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
942	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1043	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1044	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
943	}	1045	}
944		1046
945	{	1047	{
946	UV uv;	1048	UV uv;
947	int numtype = grok_number (start, dec->cur - start, &uv);	1049	int numtype = grok_number (start, len, &uv);
948	if (numtype & IS_NUMBER_IN_UV)	1050	if (numtype & IS_NUMBER_IN_UV)
949	if (numtype & IS_NUMBER_NEG)	1051	if (numtype & IS_NUMBER_NEG)
950	{	1052	{
951	if (uv < (UV)IV_MIN)	1053	if (uv < (UV)IV_MIN)
952	return newSViv (-(IV)uv);	1054	return newSViv (-(IV)uv);
953	}	1055	}
954	else	1056	else
955	return newSVuv (uv);	1057	return newSVuv (uv);
956
957	// here would likely be the place for bigint support
958	}	1058	}
959	}
960		1059
961	// if we ever support bigint or bigfloat, this is the place for bigfloat	1060	len -= *start == '-' ? 1 : 0;
		1061
		1062	// does not fit into IV or UV, try NV
		1063	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1064	#if defined (LDBL_DIG)
		1065	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1066	#endif
		1067	)
		1068	// fits into NV without loss of precision
		1069	return newSVnv (Atof (start));
		1070
		1071	// everything else fails, convert it to a string
		1072	return newSVpvn (start, dec->cur - start);
		1073	}
		1074
		1075	// loss of precision here
962	return newSVnv (Atof (start));	1076	return newSVnv (Atof (start));
963		1077
964	fail:	1078	fail:
965	return 0;	1079	return 0;
966	}	1080	}
…		…
996		1110
997	if (*dec->cur != ',')	1111	if (*dec->cur != ',')
998	ERR (", or ] expected while parsing array");	1112	ERR (", or ] expected while parsing array");
999		1113
1000	++dec->cur;	1114	++dec->cur;
		1115
		1116	decode_ws (dec);
		1117
		1118	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1119	{
		1120	++dec->cur;
		1121	break;
		1122	}
1001	}	1123	}
1002		1124
1003	DEC_DEC_DEPTH;	1125	DEC_DEC_DEPTH;
1004	return newRV_noinc ((SV *)av);	1126	return newRV_noinc ((SV *)av);
1005		1127
…		…
1021	if (*dec->cur == '}')	1143	if (*dec->cur == '}')
1022	++dec->cur;	1144	++dec->cur;
1023	else	1145	else
1024	for (;;)	1146	for (;;)
1025	{	1147	{
1026	decode_ws (dec); EXPECT_CH ('"');	1148	EXPECT_CH ('"');
1027		1149
1028	// heuristic: assume that	1150	// heuristic: assume that
1029	// a) decode_str + hv_store_ent are abysmally slow.	1151	// a) decode_str + hv_store_ent are abysmally slow.
1030	// b) most hash keys are short, simple ascii text.	1152	// b) most hash keys are short, simple ascii text.
1031	// => try to "fast-match" such strings to avoid	1153	// => try to "fast-match" such strings to avoid
…		…
1045	if (!key)	1167	if (!key)
1046	goto fail;	1168	goto fail;
1047		1169
1048	decode_ws (dec); EXPECT_CH (':');	1170	decode_ws (dec); EXPECT_CH (':');
1049		1171
		1172	decode_ws (dec);
1050	value = decode_sv (dec);	1173	value = decode_sv (dec);
1051	if (!value)	1174	if (!value)
1052	{	1175	{
1053	SvREFCNT_dec (key);	1176	SvREFCNT_dec (key);
1054	goto fail;	1177	goto fail;
…		…
1066	int len = p - key;	1189	int len = p - key;
1067	dec->cur = p + 1;	1190	dec->cur = p + 1;
1068		1191
1069	decode_ws (dec); EXPECT_CH (':');	1192	decode_ws (dec); EXPECT_CH (':');
1070		1193
		1194	decode_ws (dec);
1071	value = decode_sv (dec);	1195	value = decode_sv (dec);
1072	if (!value)	1196	if (!value)
1073	goto fail;	1197	goto fail;
1074		1198
1075	hv_store (hv, key, len, value, 0);	1199	hv_store (hv, key, len, value, 0);
…		…
1091		1215
1092	if (*dec->cur != ',')	1216	if (*dec->cur != ',')
1093	ERR (", or } expected while parsing object/hash");	1217	ERR (", or } expected while parsing object/hash");
1094		1218
1095	++dec->cur;	1219	++dec->cur;
		1220
		1221	decode_ws (dec);
		1222
		1223	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1224	{
		1225	++dec->cur;
		1226	break;
		1227	}
1096	}	1228	}
1097		1229
1098	DEC_DEC_DEPTH;	1230	DEC_DEC_DEPTH;
1099	sv = newRV_noinc ((SV *)hv);	1231	sv = newRV_noinc ((SV *)hv);
1100		1232
…		…
1113	// might want to optimise this for common cases.	1245	// might want to optimise this for common cases.
1114	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);	1246	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1115		1247
1116	if (cb)	1248	if (cb)
1117	{	1249	{
		1250	dSP;
1118	int count;	1251	int count;
1119	ENTER; SAVETMPS;
1120		1252
1121	dSP; PUSHMARK (SP);	1253	ENTER; SAVETMPS; PUSHMARK (SP);
1122	XPUSHs (HeVAL (he));	1254	XPUSHs (HeVAL (he));
1123		1255
1124	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;	1256	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1125		1257
1126	if (count == 1)	1258	if (count == 1)
…		…
1134	}	1266	}
1135	}	1267	}
1136		1268
1137	if (dec->json.cb_object)	1269	if (dec->json.cb_object)
1138	{	1270	{
		1271	dSP;
1139	int count;	1272	int count;
1140	ENTER; SAVETMPS;
1141		1273
1142	dSP; ENTER; SAVETMPS; PUSHMARK (SP);	1274	ENTER; SAVETMPS; PUSHMARK (SP);
1143	XPUSHs (sv_2mortal (sv));	1275	XPUSHs (sv_2mortal (sv));
1144		1276
1145	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;	1277	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1146		1278
1147	if (count == 1)	1279	if (count == 1)
…		…
1165	}	1297	}
1166		1298
1167	static SV *	1299	static SV *
1168	decode_sv (dec_t *dec)	1300	decode_sv (dec_t *dec)
1169	{	1301	{
1170	decode_ws (dec);
1171
1172	// the beauty of JSON: you need exactly one character lookahead	1302	// the beauty of JSON: you need exactly one character lookahead
1173	// to parse anything.	1303	// to parse everything.
1174	switch (*dec->cur)	1304	switch (*dec->cur)
1175	{	1305	{
1176	case '"': ++dec->cur; return decode_str (dec);	1306	case '"': ++dec->cur; return decode_str (dec);
1177	case '[': ++dec->cur; return decode_av (dec);	1307	case '[': ++dec->cur; return decode_av (dec);
1178	case '{': ++dec->cur; return decode_hv (dec);	1308	case '{': ++dec->cur; return decode_hv (dec);
1179		1309
1180	case '-':	1310	case '-':
1181	case '0': case '1': case '2': case '3': case '4':	1311	case '0': case '1': case '2': case '3': case '4':
1182	case '5': case '6': case '7': case '8': case '9':	1312	case '5': case '6': case '7': case '8': case '9':
1183	return decode_num (dec);	1313	return decode_num (dec);
1184		1314
1185	case 't':	1315	case 't':
1186	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1316	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1187	{	1317	{
1188	dec->cur += 4;	1318	dec->cur += 4;
		1319	#if JSON_SLOW
		1320	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1321	#endif
1189	return SvREFCNT_inc (json_true);	1322	return SvREFCNT_inc (json_true);
1190	}	1323	}
1191	else	1324	else
1192	ERR ("'true' expected");	1325	ERR ("'true' expected");
1193		1326
…		…
1195		1328
1196	case 'f':	1329	case 'f':
1197	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1330	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1198	{	1331	{
1199	dec->cur += 5;	1332	dec->cur += 5;
		1333	#if JSON_SLOW
		1334	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1335	#endif
1200	return SvREFCNT_inc (json_false);	1336	return SvREFCNT_inc (json_false);
1201	}	1337	}
1202	else	1338	else
1203	ERR ("'false' expected");	1339	ERR ("'false' expected");
1204		1340
…		…
1254		1390
1255	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1391	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1256	dec.json.flags \|= F_HOOK;	1392	dec.json.flags \|= F_HOOK;
1257		1393
1258	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1394	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1395
		1396	decode_ws (&dec);
1259	sv = decode_sv (&dec);	1397	sv = decode_sv (&dec);
1260		1398
1261	if (!(offset_return \|\| !sv))	1399	if (!(offset_return \|\| !sv))
1262	{	1400	{
1263	// check for trailing garbage	1401	// check for trailing garbage
…		…
1331	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1469	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1332	}	1470	}
1333		1471
1334	PROTOTYPES: DISABLE	1472	PROTOTYPES: DISABLE
1335		1473
		1474	void CLONE (...)
		1475	CODE:
		1476	json_stash = 0;
		1477	json_boolean_stash = 0;
		1478
1336	void new (char *klass)	1479	void new (char *klass)
1337	PPCODE:	1480	PPCODE:
1338	{	1481	{
1339	SV *pv = NEWSV (0, sizeof (JSON));	1482	SV *pv = NEWSV (0, sizeof (JSON));
1340	SvPOK_only (pv);	1483	SvPOK_only (pv);
1341	Zero (SvPVX (pv), 1, JSON);	1484	Zero (SvPVX (pv), 1, JSON);
1342	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1485	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1343	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1486	XPUSHs (sv_2mortal (sv_bless (
		1487	newRV_noinc (pv),
		1488	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1489	)));
1344	}	1490	}
1345		1491
1346	void ascii (JSON *self, int enable = 1)	1492	void ascii (JSON *self, int enable = 1)
1347	ALIAS:	1493	ALIAS:
1348	ascii = F_ASCII	1494	ascii = F_ASCII
…		…
1355	pretty = F_PRETTY	1501	pretty = F_PRETTY
1356	allow_nonref = F_ALLOW_NONREF	1502	allow_nonref = F_ALLOW_NONREF
1357	shrink = F_SHRINK	1503	shrink = F_SHRINK
1358	allow_blessed = F_ALLOW_BLESSED	1504	allow_blessed = F_ALLOW_BLESSED
1359	convert_blessed = F_CONV_BLESSED	1505	convert_blessed = F_CONV_BLESSED
		1506	relaxed = F_RELAXED
1360	PPCODE:	1507	PPCODE:
1361	{	1508	{
1362	if (enable)	1509	if (enable)
1363	self->flags \|= ix;	1510	self->flags \|= ix;
1364	else	1511	else
1365	self->flags &= ~ix;	1512	self->flags &= ~ix;
1366		1513
1367	XPUSHs (ST (0));	1514	XPUSHs (ST (0));
1368	}	1515	}
1369		1516
		1517	void get_ascii (JSON *self)
		1518	ALIAS:
		1519	get_ascii = F_ASCII
		1520	get_latin1 = F_LATIN1
		1521	get_utf8 = F_UTF8
		1522	get_indent = F_INDENT
		1523	get_canonical = F_CANONICAL
		1524	get_space_before = F_SPACE_BEFORE
		1525	get_space_after = F_SPACE_AFTER
		1526	get_allow_nonref = F_ALLOW_NONREF
		1527	get_shrink = F_SHRINK
		1528	get_allow_blessed = F_ALLOW_BLESSED
		1529	get_convert_blessed = F_CONV_BLESSED
		1530	get_relaxed = F_RELAXED
		1531	PPCODE:
		1532	XPUSHs (boolSV (self->flags & ix));
		1533
1370	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1534	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1371	PPCODE:	1535	PPCODE:
1372	{	1536	{
1373	UV log2 = 0;	1537	UV log2 = 0;
1374		1538
…		…
1380	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1544	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1381		1545
1382	XPUSHs (ST (0));	1546	XPUSHs (ST (0));
1383	}	1547	}
1384		1548
		1549	U32 get_max_depth (JSON *self)
		1550	CODE:
		1551	RETVAL = DEC_DEPTH (self->flags);
		1552	OUTPUT:
		1553	RETVAL
		1554
1385	void max_size (JSON *self, UV max_size = 0)	1555	void max_size (JSON *self, UV max_size = 0)
1386	PPCODE:	1556	PPCODE:
1387	{	1557	{
1388	UV log2 = 0;	1558	UV log2 = 0;
1389		1559
…		…
1395		1565
1396	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1566	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1397		1567
1398	XPUSHs (ST (0));	1568	XPUSHs (ST (0));
1399	}	1569	}
		1570
		1571	int get_max_size (JSON *self)
		1572	CODE:
		1573	RETVAL = DEC_SIZE (self->flags);
		1574	OUTPUT:
		1575	RETVAL
1400		1576
1401	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1577	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1402	PPCODE:	1578	PPCODE:
1403	{	1579	{
1404	SvREFCNT_dec (self->cb_object);	1580	SvREFCNT_dec (self->cb_object);
…		…
1451	SvREFCNT_dec (self->cb_sk_object);	1627	SvREFCNT_dec (self->cb_sk_object);
1452	SvREFCNT_dec (self->cb_object);	1628	SvREFCNT_dec (self->cb_object);
1453		1629
1454	PROTOTYPES: ENABLE	1630	PROTOTYPES: ENABLE
1455		1631
1456	void to_json (SV *scalar)	1632	void encode_json (SV *scalar)
1457	PPCODE:	1633	PPCODE:
1458	{	1634	{
1459	JSON json = { F_DEFAULT \| F_UTF8 };	1635	JSON json = { F_DEFAULT \| F_UTF8 };
1460	XPUSHs (encode_json (scalar, &json));	1636	XPUSHs (encode_json (scalar, &json));
1461	}	1637	}
1462		1638
1463	void from_json (SV *jsonstr)	1639	void decode_json (SV *jsonstr)
1464	PPCODE:	1640	PPCODE:
1465	{	1641	{
1466	JSON json = { F_DEFAULT \| F_UTF8 };	1642	JSON json = { F_DEFAULT \| F_UTF8 };
1467	XPUSHs (decode_json (jsonstr, &json, 0));	1643	XPUSHs (decode_json (jsonstr, &json, 0));
1468	}	1644	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.52 by root, Mon Jul 2 02:57:11 2007 UTC vs. Revision 1.73 by root, Wed Mar 19 13:44:43 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.52 by root, Mon Jul 2 02:57:11 2007 UTC vs.
Revision 1.73 by root, Wed Mar 19 13:44:43 2008 UTC