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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs.
Revision 1.75 by root, Wed Mar 19 15:37:54 2008 UTC

…		…
4		4
5	#include <assert.h>	5	#include <assert.h>
6	#include <string.h>	6	#include <string.h>
7	#include <stdlib.h>	7	#include <stdlib.h>
8	#include <stdio.h>	8	#include <stdio.h>
		9	#include <limits.h>
9	#include <float.h>	10	#include <float.h>
10		11
11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	12	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
12	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
13	#endif	14	#endif
…		…
15	// some old perls do not have this, try to make it work, no	16	// some old perls do not have this, try to make it work, no
16	// guarentees, though. if it breaks, you get to keep the pieces.	17	// guarentees, though. if it breaks, you get to keep the pieces.
17	#ifndef UTF8_MAXBYTES	18	#ifndef UTF8_MAXBYTES
18	# define UTF8_MAXBYTES 13	19	# define UTF8_MAXBYTES 13
19	#endif	20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
20		23
21	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
22	#define F_LATIN1 0x00000002UL	25	#define F_LATIN1 0x00000002UL
23	#define F_UTF8 0x00000004UL	26	#define F_UTF8 0x00000004UL
24	#define F_INDENT 0x00000008UL	27	#define F_INDENT 0x00000008UL
…		…
27	#define F_SPACE_AFTER 0x00000040UL	30	#define F_SPACE_AFTER 0x00000040UL
28	#define F_ALLOW_NONREF 0x00000100UL	31	#define F_ALLOW_NONREF 0x00000100UL
29	#define F_SHRINK 0x00000200UL	32	#define F_SHRINK 0x00000200UL
30	#define F_ALLOW_BLESSED 0x00000400UL	33	#define F_ALLOW_BLESSED 0x00000400UL
31	#define F_CONV_BLESSED 0x00000800UL	34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
32	#define F_MAXDEPTH 0xf8000000UL	37	#define F_MAXDEPTH 0xf8000000UL
33	#define S_MAXDEPTH 27	38	#define S_MAXDEPTH 27
34	#define F_MAXSIZE 0x01f00000UL	39	#define F_MAXSIZE 0x01f00000UL
35	#define S_MAXSIZE 20	40	#define S_MAXSIZE 20
36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing	41	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
…		…
48		53
49	#define SB do {	54	#define SB do {
50	#define SE } while (0)	55	#define SE } while (0)
51		56
52	#if __GNUC__ >= 3	57	#if __GNUC__ >= 3
53	# define expect(expr,value) __builtin_expect ((expr),(value))	58	# define expect(expr,value) __builtin_expect ((expr), (value))
54	# define inline inline	59	# define INLINE static inline
55	#else	60	#else
56	# define expect(expr,value) (expr)	61	# define expect(expr,value) (expr)
57	# define inline static	62	# define INLINE static
58	#endif	63	#endif
59		64
60	#define expect_false(expr) expect ((expr) != 0, 0)	65	#define expect_false(expr) expect ((expr) != 0, 0)
61	#define expect_true(expr) expect ((expr) != 0, 1)	66	#define expect_true(expr) expect ((expr) != 0, 1)
62		67
		68	#define IN_RANGE_INC(type,val,beg,end) \
		69	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		70	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		71
63	#ifdef USE_ITHREADS	72	#ifdef USE_ITHREADS
64	# define JSON_SLOW 1	73	# define JSON_SLOW 1
		74	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
65	#else	75	#else
66	# define JSON_SLOW 0	76	# define JSON_SLOW 0
		77	# define JSON_STASH json_stash
67	#endif	78	#endif
68		79
69	static HV json_stash, json_boolean_stash; // JSON::XS::	80	static HV json_stash, json_boolean_stash; // JSON::XS::
70	static SV json_true, json_false;	81	static SV json_true, json_false;
71		82
…		…
76	} JSON;	87	} JSON;
77		88
78	/////////////////////////////////////////////////////////////////////////////	89	/////////////////////////////////////////////////////////////////////////////
79	// utility functions	90	// utility functions
80		91
81	inline void	92	INLINE void
82	shrink (SV *sv)	93	shrink (SV *sv)
83	{	94	{
84	sv_utf8_downgrade (sv, 1);	95	sv_utf8_downgrade (sv, 1);
85	if (SvLEN (sv) > SvCUR (sv) + 1)	96	if (SvLEN (sv) > SvCUR (sv) + 1)
86	{	97	{
…		…
95	// decode an utf-8 character and return it, or (UV)-1 in	106	// decode an utf-8 character and return it, or (UV)-1 in
96	// case of an error.	107	// case of an error.
97	// we special-case "safe" characters from U+80 .. U+7FF,	108	// we special-case "safe" characters from U+80 .. U+7FF,
98	// but use the very good perl function to parse anything else.	109	// but use the very good perl function to parse anything else.
99	// note that we never call this function for a ascii codepoints	110	// note that we never call this function for a ascii codepoints
100	inline UV	111	INLINE UV
101	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	112	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
102	{	113	{
103	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))	114	if (expect_true (len >= 2
104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	115	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	116	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
106	{	117	{
107	*clen = 2;	118	*clen = 2;
108	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	119	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
109	}	120	}
110	else	121	else
111	{	122	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
112	*clen = (STRLEN)-1;	123	}
113	return (UV)-1;	124
114	}	125	// likewise for encoding, also never called for ascii codepoints
		126	// this function takes advantage of this fact, although current gccs
		127	// seem to optimise the check for >= 0x80 away anyways
		128	INLINE unsigned char *
		129	encode_utf8 (unsigned char *s, UV ch)
		130	{
		131	if (expect_false (ch < 0x000080))
		132	*s++ = ch;
		133	else if (expect_true (ch < 0x000800))
		134	*s++ = 0xc0 \| ( ch >> 6),
		135	*s++ = 0x80 \| ( ch & 0x3f);
		136	else if ( ch < 0x010000)
		137	*s++ = 0xe0 \| ( ch >> 12),
		138	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		139	*s++ = 0x80 \| ( ch & 0x3f);
		140	else if ( ch < 0x110000)
		141	*s++ = 0xf0 \| ( ch >> 18),
		142	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		143	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		144	*s++ = 0x80 \| ( ch & 0x3f);
		145
		146	return s;
115	}	147	}
116		148
117	/////////////////////////////////////////////////////////////////////////////	149	/////////////////////////////////////////////////////////////////////////////
118	// encoder	150	// encoder
119		151
…		…
124	char *end; // SvEND (sv)	156	char *end; // SvEND (sv)
125	SV *sv; // result scalar	157	SV *sv; // result scalar
126	JSON json;	158	JSON json;
127	U32 indent; // indentation level	159	U32 indent; // indentation level
128	U32 maxdepth; // max. indentation/recursion level	160	U32 maxdepth; // max. indentation/recursion level
		161	UV limit; // escape character values >= this value when encoding
129	} enc_t;	162	} enc_t;
130		163
131	inline void	164	INLINE void
132	need (enc_t *enc, STRLEN len)	165	need (enc_t *enc, STRLEN len)
133	{	166	{
134	if (expect_false (enc->cur + len >= enc->end))	167	if (expect_false (enc->cur + len >= enc->end))
135	{	168	{
136	STRLEN cur = enc->cur - SvPVX (enc->sv);	169	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
138	enc->cur = SvPVX (enc->sv) + cur;	171	enc->cur = SvPVX (enc->sv) + cur;
139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	172	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
140	}	173	}
141	}	174	}
142		175
143	inline void	176	INLINE void
144	encode_ch (enc_t *enc, char ch)	177	encode_ch (enc_t *enc, char ch)
145	{	178	{
146	need (enc, 1);	179	need (enc, 1);
147	*enc->cur++ = ch;	180	*enc->cur++ = ch;
148	}	181	}
…		…
202	{	235	{
203	uch = ch;	236	uch = ch;
204	clen = 1;	237	clen = 1;
205	}	238	}
206		239
207	if (uch > 0x10FFFFUL)	240	if (uch < 0x80/0x20/ \|\| uch >= enc->limit)
208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
209
210	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
211	{	241	{
212	if (uch > 0xFFFFUL)	242	if (uch >= 0x10000UL)
213	{	243	{
		244	if (uch >= 0x110000UL)
		245	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		246
214	need (enc, len += 11);	247	need (enc, len += 11);
215	sprintf (enc->cur, "\\u%04x\\u%04x",	248	sprintf (enc->cur, "\\u%04x\\u%04x",
216	(int)((uch - 0x10000) / 0x400 + 0xD800),	249	(int)((uch - 0x10000) / 0x400 + 0xD800),
217	(int)((uch - 0x10000) % 0x400 + 0xDC00));	250	(int)((uch - 0x10000) % 0x400 + 0xDC00));
218	enc->cur += 12;	251	enc->cur += 12;
…		…
246	while (--clen);	279	while (--clen);
247	}	280	}
248	else	281	else
249	{	282	{
250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	283	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	284	enc->cur = encode_utf8 (enc->cur, uch);
252	++str;	285	++str;
253	}	286	}
254	}	287	}
255	}	288	}
256	}	289	}
257		290
258	--len;	291	--len;
259	}	292	}
260	}	293	}
261		294
262	inline void	295	INLINE void
263	encode_indent (enc_t *enc)	296	encode_indent (enc_t *enc)
264	{	297	{
265	if (enc->json.flags & F_INDENT)	298	if (enc->json.flags & F_INDENT)
266	{	299	{
267	int spaces = enc->indent * INDENT_STEP;	300	int spaces = enc->indent * INDENT_STEP;
…		…
270	memset (enc->cur, ' ', spaces);	303	memset (enc->cur, ' ', spaces);
271	enc->cur += spaces;	304	enc->cur += spaces;
272	}	305	}
273	}	306	}
274		307
275	inline void	308	INLINE void
276	encode_space (enc_t *enc)	309	encode_space (enc_t *enc)
277	{	310	{
278	need (enc, 1);	311	need (enc, 1);
279	encode_ch (enc, ' ');	312	encode_ch (enc, ' ');
280	}	313	}
281		314
282	inline void	315	INLINE void
283	encode_nl (enc_t *enc)	316	encode_nl (enc_t *enc)
284	{	317	{
285	if (enc->json.flags & F_INDENT)	318	if (enc->json.flags & F_INDENT)
286	{	319	{
287	need (enc, 1);	320	need (enc, 1);
288	encode_ch (enc, '\n');	321	encode_ch (enc, '\n');
289	}	322	}
290	}	323	}
291		324
292	inline void	325	INLINE void
293	encode_comma (enc_t *enc)	326	encode_comma (enc_t *enc)
294	{	327	{
295	encode_ch (enc, ',');	328	encode_ch (enc, ',');
296		329
297	if (enc->json.flags & F_INDENT)	330	if (enc->json.flags & F_INDENT)
…		…
308	int i, len = av_len (av);	341	int i, len = av_len (av);
309		342
310	if (enc->indent >= enc->maxdepth)	343	if (enc->indent >= enc->maxdepth)
311	croak ("data structure too deep (hit recursion limit)");	344	croak ("data structure too deep (hit recursion limit)");
312		345
313	encode_ch (enc, '['); encode_nl (enc);	346	encode_ch (enc, '[');
314	++enc->indent;	347
		348	if (len >= 0)
		349	{
		350	encode_nl (enc); ++enc->indent;
315		351
316	for (i = 0; i <= len; ++i)	352	for (i = 0; i <= len; ++i)
317	{	353	{
318	SV **svp = av_fetch (av, i, 0);	354	SV **svp = av_fetch (av, i, 0);
319		355
320	encode_indent (enc);	356	encode_indent (enc);
321		357
322	if (svp)	358	if (svp)
323	encode_sv (enc, *svp);	359	encode_sv (enc, *svp);
324	else	360	else
325	encode_str (enc, "null", 4, 0);	361	encode_str (enc, "null", 4, 0);
326		362
327	if (i < len)	363	if (i < len)
328	encode_comma (enc);	364	encode_comma (enc);
329	}	365	}
330		366
		367	encode_nl (enc); --enc->indent; encode_indent (enc);
		368	}
		369
331	encode_nl (enc);	370	encode_ch (enc, ']');
332
333	--enc->indent;
334	encode_indent (enc); encode_ch (enc, ']');
335	}	371	}
336		372
337	static void	373	static void
338	encode_he (enc_t enc, HE he)	374	encode_hk (enc_t enc, HE he)
339	{	375	{
340	encode_ch (enc, '"');	376	encode_ch (enc, '"');
341		377
342	if (HeKLEN (he) == HEf_SVKEY)	378	if (HeKLEN (he) == HEf_SVKEY)
343	{	379	{
…		…
356	encode_ch (enc, '"');	392	encode_ch (enc, '"');
357		393
358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	394	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
359	encode_ch (enc, ':');	395	encode_ch (enc, ':');
360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	396	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
361	encode_sv (enc, HeVAL (he));
362	}	397	}
363		398
364	// compare hash entries, used when all keys are bytestrings	399	// compare hash entries, used when all keys are bytestrings
365	static int	400	static int
366	he_cmp_fast (const void a_, const void b_)	401	he_cmp_fast (const void a_, const void b_)
…		…
371	HE b = (HE **)b_;	406	HE b = (HE **)b_;
372		407
373	STRLEN la = HeKLEN (a);	408	STRLEN la = HeKLEN (a);
374	STRLEN lb = HeKLEN (b);	409	STRLEN lb = HeKLEN (b);
375		410
376	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	411	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
377	cmp = la - lb;	412	cmp = lb - la;
378		413
379	return cmp;	414	return cmp;
380	}	415	}
381		416
382	// compare hash entries, used when some keys are sv's or utf-x	417	// compare hash entries, used when some keys are sv's or utf-x
383	static int	418	static int
384	he_cmp_slow (const void a, const void b)	419	he_cmp_slow (const void a, const void b)
385	{	420	{
386	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	421	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
387	}	422	}
388		423
389	static void	424	static void
390	encode_hv (enc_t enc, HV hv)	425	encode_hv (enc_t enc, HV hv)
391	{	426	{
		427	HE *he;
392	int count, i;	428	int count;
393		429
394	if (enc->indent >= enc->maxdepth)	430	if (enc->indent >= enc->maxdepth)
395	croak ("data structure too deep (hit recursion limit)");	431	croak ("data structure too deep (hit recursion limit)");
396		432
397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	433	encode_ch (enc, '{');
398		434
399	if ((count = hv_iterinit (hv)))
400	{
401	// for canonical output we have to sort by keys first	435	// for canonical output we have to sort by keys first
402	// actually, this is mostly due to the stupid so-called	436	// actually, this is mostly due to the stupid so-called
403	// security workaround added somewhere in 5.8.x.	437	// security workaround added somewhere in 5.8.x.
404	// that randomises hash orderings	438	// that randomises hash orderings
405	if (enc->json.flags & F_CANONICAL)	439	if (enc->json.flags & F_CANONICAL)
		440	{
		441	int count = hv_iterinit (hv);
		442
		443	if (SvMAGICAL (hv))
406	{	444	{
		445	// need to count by iterating. could improve by dynamically building the vector below
		446	// but I don't care for the speed of this special case.
		447	// note also that we will run into undefined behaviour when the two iterations
		448	// do not result in the same count, something I might care for in some later release.
		449
		450	count = 0;
		451	while (hv_iternext (hv))
		452	++count;
		453
		454	hv_iterinit (hv);
		455	}
		456
		457	if (count)
		458	{
407	int fast = 1;	459	int i, fast = 1;
408	HE *he;
409	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	460	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
410	HE *hes = _alloca (count sizeof (HE));	461	HE *hes = _alloca (count sizeof (HE));
411	#else	462	#else
412	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	463	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
413	#endif	464	#endif
…		…
440		491
441	FREETMPS;	492	FREETMPS;
442	LEAVE;	493	LEAVE;
443	}	494	}
444		495
445	for (i = 0; i < count; ++i)	496	encode_nl (enc); ++enc->indent;
		497
		498	while (count--)
446	{	499	{
447	encode_indent (enc);	500	encode_indent (enc);
		501	he = hes [count];
448	encode_he (enc, hes [i]);	502	encode_hk (enc, he);
		503	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
449		504
450	if (i < count - 1)	505	if (count)
451	encode_comma (enc);	506	encode_comma (enc);
452	}	507	}
453		508
454	encode_nl (enc);	509	encode_nl (enc); --enc->indent; encode_indent (enc);
455	}	510	}
		511	}
456	else	512	else
		513	{
		514	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		515	if ((he = hv_iternext (hv)))
457	{	516	{
458	HE *he = hv_iternext (hv);	517	encode_nl (enc); ++enc->indent;
459		518
460	for (;;)	519	for (;;)
461	{	520	{
462	encode_indent (enc);	521	encode_indent (enc);
463	encode_he (enc, he);	522	encode_hk (enc, he);
		523	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
464		524
465	if (!(he = hv_iternext (hv)))	525	if (!(he = hv_iternext (hv)))
466	break;	526	break;
467		527
468	encode_comma (enc);	528	encode_comma (enc);
469	}	529	}
470		530
471	encode_nl (enc);	531	encode_nl (enc); --enc->indent; encode_indent (enc);
472	}	532	}
473	}	533	}
474		534
475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	535	encode_ch (enc, '}');
476	}	536	}
477		537
478	// encode objects, arrays and special \0=false and \1=true values.	538	// encode objects, arrays and special \0=false and \1=true values.
479	static void	539	static void
480	encode_rv (enc_t enc, SV sv)	540	encode_rv (enc_t enc, SV sv)
…		…
588	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	648	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
589	enc->cur += strlen (enc->cur);	649	enc->cur += strlen (enc->cur);
590	}	650	}
591	else if (SvIOKp (sv))	651	else if (SvIOKp (sv))
592	{	652	{
593	// we assume we can always read an IV as a UV	653	// we assume we can always read an IV as a UV and vice versa
594	if (SvUV (sv) & ~(UV)0x7fff)	654	// we assume two's complement
595	{	655	// we assume no aliasing issues in the union
596	// large integer, use the (rather slow) snprintf way.	656	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
597	need (enc, sizeof (UV) * 3);	657	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
598	enc->cur +=
599	SvIsUV(sv)
600	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
601	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
602	}
603	else
604	{	658	{
605	// optimise the "small number case"	659	// optimise the "small number case"
606	// code will likely be branchless and use only a single multiplication	660	// code will likely be branchless and use only a single multiplication
		661	// works for numbers up to 59074
607	I32 i = SvIV (sv);	662	I32 i = SvIVX (sv);
608	U32 u;	663	U32 u;
609	char digit, nz = 0;	664	char digit, nz = 0;
610		665
611	need (enc, 6);	666	need (enc, 6);
612		667
…		…
618		673
619	// now output digit by digit, each time masking out the integer part	674	// now output digit by digit, each time masking out the integer part
620	// and multiplying by 5 while moving the decimal point one to the right,	675	// and multiplying by 5 while moving the decimal point one to the right,
621	// resulting in a net multiplication by 10.	676	// resulting in a net multiplication by 10.
622	// we always write the digit to memory but conditionally increment	677	// we always write the digit to memory but conditionally increment
623	// the pointer, to ease the usage of conditional move instructions.	678	// the pointer, to enable the use of conditional move instructions.
624	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	679	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffffUL) 5;
625	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	680	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffffUL) 5;
626	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	681	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffffUL) 5;
627	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	682	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffffUL) 5;
628	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'	683	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		684	}
		685	else
		686	{
		687	// large integer, use the (rather slow) snprintf way.
		688	need (enc, IVUV_MAXCHARS);
		689	enc->cur +=
		690	SvIsUV(sv)
		691	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
		692	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
629	}	693	}
630	}	694	}
631	else if (SvROK (sv))	695	else if (SvROK (sv))
632	encode_rv (enc, SvRV (sv));	696	encode_rv (enc, SvRV (sv));
633	else if (!SvOK (sv))	697	else if (!SvOK (sv))
…		…
649	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	713	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
650	enc.cur = SvPVX (enc.sv);	714	enc.cur = SvPVX (enc.sv);
651	enc.end = SvEND (enc.sv);	715	enc.end = SvEND (enc.sv);
652	enc.indent = 0;	716	enc.indent = 0;
653	enc.maxdepth = DEC_DEPTH (enc.json.flags);	717	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		718	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		719	: enc.json.flags & F_LATIN1 ? 0x000100UL
		720	: 0x110000UL;
654		721
655	SvPOK_only (enc.sv);	722	SvPOK_only (enc.sv);
656	encode_sv (&enc, scalar);	723	encode_sv (&enc, scalar);
657		724
658	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	725	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
679	JSON json;	746	JSON json;
680	U32 depth; // recursion depth	747	U32 depth; // recursion depth
681	U32 maxdepth; // recursion depth limit	748	U32 maxdepth; // recursion depth limit
682	} dec_t;	749	} dec_t;
683		750
684	inline void	751	INLINE void
		752	decode_comment (dec_t *dec)
		753	{
		754	// only '#'-style comments allowed a.t.m.
		755
		756	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		757	++dec->cur;
		758	}
		759
		760	INLINE void
685	decode_ws (dec_t *dec)	761	decode_ws (dec_t *dec)
686	{	762	{
687	for (;;)	763	for (;;)
688	{	764	{
689	char ch = *dec->cur;	765	char ch = *dec->cur;
690		766
691	if (ch > 0x20	767	if (ch > 0x20)
		768	{
		769	if (expect_false (ch == '#'))
		770	{
		771	if (dec->json.flags & F_RELAXED)
		772	decode_comment (dec);
		773	else
		774	break;
		775	}
		776	else
		777	break;
		778	}
692	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	779	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
693	break;	780	break; // parse error, but let higher level handle it, gives better error messages
694		781
695	++dec->cur;	782	++dec->cur;
696	}	783	}
697	}	784	}
698		785
…		…
804		891
805	if (hi >= 0x80)	892	if (hi >= 0x80)
806	{	893	{
807	utf8 = 1;	894	utf8 = 1;
808		895
809	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	896	cur = encode_utf8 (cur, hi);
810	}	897	}
811	else	898	else
812	*cur++ = hi;	899	*cur++ = hi;
813	}	900	}
814	break;	901	break;
…		…
816	default:	903	default:
817	--dec_cur;	904	--dec_cur;
818	ERR ("illegal backslash escape sequence in string");	905	ERR ("illegal backslash escape sequence in string");
819	}	906	}
820	}	907	}
821	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	908	else if (expect_true (ch >= 0x20 && ch < 0x80))
822	*cur++ = ch;	909	*cur++ = ch;
823	else if (ch >= 0x80)	910	else if (ch >= 0x80)
824	{	911	{
825	STRLEN clen;	912	STRLEN clen;
826	UV uch;	913	UV uch;
…		…
949		1036
950	if (!is_nv)	1037	if (!is_nv)
951	{	1038	{
952	int len = dec->cur - start;	1039	int len = dec->cur - start;
953		1040
954	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1041	// special case the rather common 1..5-digit-int case
955	if (*start == '-')	1042	if (*start == '-')
956	switch (len)	1043	switch (len)
957	{	1044	{
958	case 2: return newSViv (-( start [1] - '0' * 1));	1045	case 2: return newSViv (-( start [1] - '0' * 1));
959	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1046	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
960	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1047	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
961	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1048	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1049	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
962	}	1050	}
963	else	1051	else
964	switch (len)	1052	switch (len)
965	{	1053	{
966	case 1: return newSViv ( start [0] - '0' * 1);	1054	case 1: return newSViv ( start [0] - '0' * 1);
967	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1055	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
968	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1056	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
969	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1057	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1058	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
970	}	1059	}
971		1060
972	{	1061	{
973	UV uv;	1062	UV uv;
974	int numtype = grok_number (start, len, &uv);	1063	int numtype = grok_number (start, len, &uv);
…		…
1035		1124
1036	if (*dec->cur != ',')	1125	if (*dec->cur != ',')
1037	ERR (", or ] expected while parsing array");	1126	ERR (", or ] expected while parsing array");
1038		1127
1039	++dec->cur;	1128	++dec->cur;
		1129
		1130	decode_ws (dec);
		1131
		1132	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1133	{
		1134	++dec->cur;
		1135	break;
		1136	}
1040	}	1137	}
1041		1138
1042	DEC_DEC_DEPTH;	1139	DEC_DEC_DEPTH;
1043	return newRV_noinc ((SV *)av);	1140	return newRV_noinc ((SV *)av);
1044		1141
…		…
1060	if (*dec->cur == '}')	1157	if (*dec->cur == '}')
1061	++dec->cur;	1158	++dec->cur;
1062	else	1159	else
1063	for (;;)	1160	for (;;)
1064	{	1161	{
1065	decode_ws (dec); EXPECT_CH ('"');	1162	EXPECT_CH ('"');
1066		1163
1067	// heuristic: assume that	1164	// heuristic: assume that
1068	// a) decode_str + hv_store_ent are abysmally slow.	1165	// a) decode_str + hv_store_ent are abysmally slow.
1069	// b) most hash keys are short, simple ascii text.	1166	// b) most hash keys are short, simple ascii text.
1070	// => try to "fast-match" such strings to avoid	1167	// => try to "fast-match" such strings to avoid
…		…
1084	if (!key)	1181	if (!key)
1085	goto fail;	1182	goto fail;
1086		1183
1087	decode_ws (dec); EXPECT_CH (':');	1184	decode_ws (dec); EXPECT_CH (':');
1088		1185
		1186	decode_ws (dec);
1089	value = decode_sv (dec);	1187	value = decode_sv (dec);
1090	if (!value)	1188	if (!value)
1091	{	1189	{
1092	SvREFCNT_dec (key);	1190	SvREFCNT_dec (key);
1093	goto fail;	1191	goto fail;
…		…
1105	int len = p - key;	1203	int len = p - key;
1106	dec->cur = p + 1;	1204	dec->cur = p + 1;
1107		1205
1108	decode_ws (dec); EXPECT_CH (':');	1206	decode_ws (dec); EXPECT_CH (':');
1109		1207
		1208	decode_ws (dec);
1110	value = decode_sv (dec);	1209	value = decode_sv (dec);
1111	if (!value)	1210	if (!value)
1112	goto fail;	1211	goto fail;
1113		1212
1114	hv_store (hv, key, len, value, 0);	1213	hv_store (hv, key, len, value, 0);
…		…
1130		1229
1131	if (*dec->cur != ',')	1230	if (*dec->cur != ',')
1132	ERR (", or } expected while parsing object/hash");	1231	ERR (", or } expected while parsing object/hash");
1133		1232
1134	++dec->cur;	1233	++dec->cur;
		1234
		1235	decode_ws (dec);
		1236
		1237	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1238	{
		1239	++dec->cur;
		1240	break;
		1241	}
1135	}	1242	}
1136		1243
1137	DEC_DEC_DEPTH;	1244	DEC_DEC_DEPTH;
1138	sv = newRV_noinc ((SV *)hv);	1245	sv = newRV_noinc ((SV *)hv);
1139		1246
…		…
1204	}	1311	}
1205		1312
1206	static SV *	1313	static SV *
1207	decode_sv (dec_t *dec)	1314	decode_sv (dec_t *dec)
1208	{	1315	{
1209	decode_ws (dec);
1210
1211	// the beauty of JSON: you need exactly one character lookahead	1316	// the beauty of JSON: you need exactly one character lookahead
1212	// to parse anything.	1317	// to parse everything.
1213	switch (*dec->cur)	1318	switch (*dec->cur)
1214	{	1319	{
1215	case '"': ++dec->cur; return decode_str (dec);	1320	case '"': ++dec->cur; return decode_str (dec);
1216	case '[': ++dec->cur; return decode_av (dec);	1321	case '[': ++dec->cur; return decode_av (dec);
1217	case '{': ++dec->cur; return decode_hv (dec);	1322	case '{': ++dec->cur; return decode_hv (dec);
1218		1323
1219	case '-':	1324	case '-':
1220	case '0': case '1': case '2': case '3': case '4':	1325	case '0': case '1': case '2': case '3': case '4':
1221	case '5': case '6': case '7': case '8': case '9':	1326	case '5': case '6': case '7': case '8': case '9':
1222	return decode_num (dec);	1327	return decode_num (dec);
1223		1328
1224	case 't':	1329	case 't':
1225	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1330	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1226	{	1331	{
1227	dec->cur += 4;	1332	dec->cur += 4;
		1333	#if JSON_SLOW
		1334	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1335	#endif
1228	return SvREFCNT_inc (json_true);	1336	return SvREFCNT_inc (json_true);
1229	}	1337	}
1230	else	1338	else
1231	ERR ("'true' expected");	1339	ERR ("'true' expected");
1232		1340
…		…
1234		1342
1235	case 'f':	1343	case 'f':
1236	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1344	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1237	{	1345	{
1238	dec->cur += 5;	1346	dec->cur += 5;
		1347	#if JSON_SLOW
		1348	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1349	#endif
1239	return SvREFCNT_inc (json_false);	1350	return SvREFCNT_inc (json_false);
1240	}	1351	}
1241	else	1352	else
1242	ERR ("'false' expected");	1353	ERR ("'false' expected");
1243		1354
…		…
1293		1404
1294	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1405	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1295	dec.json.flags \|= F_HOOK;	1406	dec.json.flags \|= F_HOOK;
1296		1407
1297	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1408	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1409
		1410	decode_ws (&dec);
1298	sv = decode_sv (&dec);	1411	sv = decode_sv (&dec);
1299		1412
1300	if (!(offset_return \|\| !sv))	1413	if (!(offset_return \|\| !sv))
1301	{	1414	{
1302	// check for trailing garbage	1415	// check for trailing garbage
…		…
1378	json_boolean_stash = 0;	1491	json_boolean_stash = 0;
1379		1492
1380	void new (char *klass)	1493	void new (char *klass)
1381	PPCODE:	1494	PPCODE:
1382	{	1495	{
1383	HV *stash = !JSON_SLOW \|\| json_stash
1384	? json_stash
1385	: gv_stashpv ("JSON::XS", 1);
1386	SV *pv = NEWSV (0, sizeof (JSON));	1496	SV *pv = NEWSV (0, sizeof (JSON));
1387	SvPOK_only (pv);	1497	SvPOK_only (pv);
1388	Zero (SvPVX (pv), 1, JSON);	1498	Zero (SvPVX (pv), 1, JSON);
1389	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1499	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1390	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));	1500	XPUSHs (sv_2mortal (sv_bless (
		1501	newRV_noinc (pv),
		1502	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1503	)));
1391	}	1504	}
1392		1505
1393	void ascii (JSON *self, int enable = 1)	1506	void ascii (JSON *self, int enable = 1)
1394	ALIAS:	1507	ALIAS:
1395	ascii = F_ASCII	1508	ascii = F_ASCII
…		…
1402	pretty = F_PRETTY	1515	pretty = F_PRETTY
1403	allow_nonref = F_ALLOW_NONREF	1516	allow_nonref = F_ALLOW_NONREF
1404	shrink = F_SHRINK	1517	shrink = F_SHRINK
1405	allow_blessed = F_ALLOW_BLESSED	1518	allow_blessed = F_ALLOW_BLESSED
1406	convert_blessed = F_CONV_BLESSED	1519	convert_blessed = F_CONV_BLESSED
		1520	relaxed = F_RELAXED
1407	PPCODE:	1521	PPCODE:
1408	{	1522	{
1409	if (enable)	1523	if (enable)
1410	self->flags \|= ix;	1524	self->flags \|= ix;
1411	else	1525	else
1412	self->flags &= ~ix;	1526	self->flags &= ~ix;
1413		1527
1414	XPUSHs (ST (0));	1528	XPUSHs (ST (0));
1415	}	1529	}
1416		1530
		1531	void get_ascii (JSON *self)
		1532	ALIAS:
		1533	get_ascii = F_ASCII
		1534	get_latin1 = F_LATIN1
		1535	get_utf8 = F_UTF8
		1536	get_indent = F_INDENT
		1537	get_canonical = F_CANONICAL
		1538	get_space_before = F_SPACE_BEFORE
		1539	get_space_after = F_SPACE_AFTER
		1540	get_allow_nonref = F_ALLOW_NONREF
		1541	get_shrink = F_SHRINK
		1542	get_allow_blessed = F_ALLOW_BLESSED
		1543	get_convert_blessed = F_CONV_BLESSED
		1544	get_relaxed = F_RELAXED
		1545	PPCODE:
		1546	XPUSHs (boolSV (self->flags & ix));
		1547
1417	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1548	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1418	PPCODE:	1549	PPCODE:
1419	{	1550	{
1420	UV log2 = 0;	1551	UV log2 = 0;
1421		1552
…		…
1427	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1558	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1428		1559
1429	XPUSHs (ST (0));	1560	XPUSHs (ST (0));
1430	}	1561	}
1431		1562
		1563	U32 get_max_depth (JSON *self)
		1564	CODE:
		1565	RETVAL = DEC_DEPTH (self->flags);
		1566	OUTPUT:
		1567	RETVAL
		1568
1432	void max_size (JSON *self, UV max_size = 0)	1569	void max_size (JSON *self, UV max_size = 0)
1433	PPCODE:	1570	PPCODE:
1434	{	1571	{
1435	UV log2 = 0;	1572	UV log2 = 0;
1436		1573
…		…
1442		1579
1443	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1580	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1444		1581
1445	XPUSHs (ST (0));	1582	XPUSHs (ST (0));
1446	}	1583	}
		1584
		1585	int get_max_size (JSON *self)
		1586	CODE:
		1587	RETVAL = DEC_SIZE (self->flags);
		1588	OUTPUT:
		1589	RETVAL
1447		1590
1448	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1591	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1449	PPCODE:	1592	PPCODE:
1450	{	1593	{
1451	SvREFCNT_dec (self->cb_object);	1594	SvREFCNT_dec (self->cb_object);
…		…
1498	SvREFCNT_dec (self->cb_sk_object);	1641	SvREFCNT_dec (self->cb_sk_object);
1499	SvREFCNT_dec (self->cb_object);	1642	SvREFCNT_dec (self->cb_object);
1500		1643
1501	PROTOTYPES: ENABLE	1644	PROTOTYPES: ENABLE
1502		1645
1503	void to_json (SV *scalar)	1646	void encode_json (SV *scalar)
1504	PPCODE:	1647	PPCODE:
1505	{	1648	{
1506	JSON json = { F_DEFAULT \| F_UTF8 };	1649	JSON json = { F_DEFAULT \| F_UTF8 };
1507	XPUSHs (encode_json (scalar, &json));	1650	XPUSHs (encode_json (scalar, &json));
1508	}	1651	}
1509		1652
1510	void from_json (SV *jsonstr)	1653	void decode_json (SV *jsonstr)
1511	PPCODE:	1654	PPCODE:
1512	{	1655	{
1513	JSON json = { F_DEFAULT \| F_UTF8 };	1656	JSON json = { F_DEFAULT \| F_UTF8 };
1514	XPUSHs (decode_json (jsonstr, &json, 0));	1657	XPUSHs (decode_json (jsonstr, &json, 0));
1515	}	1658	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs. Revision 1.75 by root, Wed Mar 19 15:37:54 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC vs.
Revision 1.75 by root, Wed Mar 19 15:37:54 2008 UTC