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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs.
Revision 1.74 by root, Wed Mar 19 15:17:53 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
107	}	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 (expect_false (ch < 0x000080))
		129	*s++ = ch;
		130	else if (expect_true (ch < 0x000800))
		131	*s++ = 0xc0 \| ( ch >> 6),
		132	*s++ = 0x80 \| ( ch & 0x3f);
		133	else if ( ch < 0x010000)
		134	*s++ = 0xe0 \| ( ch >> 12),
		135	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		136	*s++ = 0x80 \| ( ch & 0x3f);
		137	else if ( ch < 0x110000)
		138	*s++ = 0xf0 \| ( ch >> 18),
		139	*s++ = 0x80 \| ((ch >> 12) & 0x3f),
		140	*s++ = 0x80 \| ((ch >> 6) & 0x3f),
		141	*s++ = 0x80 \| ( ch & 0x3f);
		142
		143	return s;
108	}	144	}
109		145
110	/////////////////////////////////////////////////////////////////////////////	146	/////////////////////////////////////////////////////////////////////////////
111	// encoder	147	// encoder
112		148
…		…
117	char *end; // SvEND (sv)	153	char *end; // SvEND (sv)
118	SV *sv; // result scalar	154	SV *sv; // result scalar
119	JSON json;	155	JSON json;
120	U32 indent; // indentation level	156	U32 indent; // indentation level
121	U32 maxdepth; // max. indentation/recursion level	157	U32 maxdepth; // max. indentation/recursion level
		158	UV limit; // escape character values >= this value when encoding
122	} enc_t;	159	} enc_t;
123		160
124	inline void	161	INLINE void
125	need (enc_t *enc, STRLEN len)	162	need (enc_t *enc, STRLEN len)
126	{	163	{
127	if (expect_false (enc->cur + len >= enc->end))	164	if (expect_false (enc->cur + len >= enc->end))
128	{	165	{
129	STRLEN cur = enc->cur - SvPVX (enc->sv);	166	STRLEN cur = enc->cur - SvPVX (enc->sv);
…		…
131	enc->cur = SvPVX (enc->sv) + cur;	168	enc->cur = SvPVX (enc->sv) + cur;
132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	169	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
133	}	170	}
134	}	171	}
135		172
136	inline void	173	INLINE void
137	encode_ch (enc_t *enc, char ch)	174	encode_ch (enc_t *enc, char ch)
138	{	175	{
139	need (enc, 1);	176	need (enc, 1);
140	*enc->cur++ = ch;	177	*enc->cur++ = ch;
141	}	178	}
…		…
195	{	232	{
196	uch = ch;	233	uch = ch;
197	clen = 1;	234	clen = 1;
198	}	235	}
199		236
200	if (uch > 0x10FFFFUL)	237	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	{	238	{
205	if (uch > 0xFFFFUL)	239	if (uch >= 0x10000UL)
206	{	240	{
		241	if (uch >= 0x110000UL)
		242	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		243
207	need (enc, len += 11);	244	need (enc, len += 11);
208	sprintf (enc->cur, "\\u%04x\\u%04x",	245	sprintf (enc->cur, "\\u%04x\\u%04x",
209	(int)((uch - 0x10000) / 0x400 + 0xD800),	246	(int)((uch - 0x10000) / 0x400 + 0xD800),
210	(int)((uch - 0x10000) % 0x400 + 0xDC00));	247	(int)((uch - 0x10000) % 0x400 + 0xDC00));
211	enc->cur += 12;	248	enc->cur += 12;
…		…
239	while (--clen);	276	while (--clen);
240	}	277	}
241	else	278	else
242	{	279	{
243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed	280	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	281	enc->cur = encode_utf8 (enc->cur, uch);
245	++str;	282	++str;
246	}	283	}
247	}	284	}
248	}	285	}
249	}	286	}
250		287
251	--len;	288	--len;
252	}	289	}
253	}	290	}
254		291
255	inline void	292	INLINE void
256	encode_indent (enc_t *enc)	293	encode_indent (enc_t *enc)
257	{	294	{
258	if (enc->json.flags & F_INDENT)	295	if (enc->json.flags & F_INDENT)
259	{	296	{
260	int spaces = enc->indent * INDENT_STEP;	297	int spaces = enc->indent * INDENT_STEP;
…		…
263	memset (enc->cur, ' ', spaces);	300	memset (enc->cur, ' ', spaces);
264	enc->cur += spaces;	301	enc->cur += spaces;
265	}	302	}
266	}	303	}
267		304
268	inline void	305	INLINE void
269	encode_space (enc_t *enc)	306	encode_space (enc_t *enc)
270	{	307	{
271	need (enc, 1);	308	need (enc, 1);
272	encode_ch (enc, ' ');	309	encode_ch (enc, ' ');
273	}	310	}
274		311
275	inline void	312	INLINE void
276	encode_nl (enc_t *enc)	313	encode_nl (enc_t *enc)
277	{	314	{
278	if (enc->json.flags & F_INDENT)	315	if (enc->json.flags & F_INDENT)
279	{	316	{
280	need (enc, 1);	317	need (enc, 1);
281	encode_ch (enc, '\n');	318	encode_ch (enc, '\n');
282	}	319	}
283	}	320	}
284		321
285	inline void	322	INLINE void
286	encode_comma (enc_t *enc)	323	encode_comma (enc_t *enc)
287	{	324	{
288	encode_ch (enc, ',');	325	encode_ch (enc, ',');
289		326
290	if (enc->json.flags & F_INDENT)	327	if (enc->json.flags & F_INDENT)
…		…
301	int i, len = av_len (av);	338	int i, len = av_len (av);
302		339
303	if (enc->indent >= enc->maxdepth)	340	if (enc->indent >= enc->maxdepth)
304	croak ("data structure too deep (hit recursion limit)");	341	croak ("data structure too deep (hit recursion limit)");
305		342
306	encode_ch (enc, '['); encode_nl (enc);	343	encode_ch (enc, '[');
307	++enc->indent;	344
		345	if (len >= 0)
		346	{
		347	encode_nl (enc); ++enc->indent;
308		348
309	for (i = 0; i <= len; ++i)	349	for (i = 0; i <= len; ++i)
310	{	350	{
311	SV **svp = av_fetch (av, i, 0);	351	SV **svp = av_fetch (av, i, 0);
312		352
313	encode_indent (enc);	353	encode_indent (enc);
314		354
315	if (svp)	355	if (svp)
316	encode_sv (enc, *svp);	356	encode_sv (enc, *svp);
317	else	357	else
318	encode_str (enc, "null", 4, 0);	358	encode_str (enc, "null", 4, 0);
319		359
320	if (i < len)	360	if (i < len)
321	encode_comma (enc);	361	encode_comma (enc);
322	}	362	}
323		363
		364	encode_nl (enc); --enc->indent; encode_indent (enc);
		365	}
		366
324	encode_nl (enc);	367	encode_ch (enc, ']');
325
326	--enc->indent;
327	encode_indent (enc); encode_ch (enc, ']');
328	}	368	}
329		369
330	static void	370	static void
331	encode_he (enc_t enc, HE he)	371	encode_hk (enc_t enc, HE he)
332	{	372	{
333	encode_ch (enc, '"');	373	encode_ch (enc, '"');
334		374
335	if (HeKLEN (he) == HEf_SVKEY)	375	if (HeKLEN (he) == HEf_SVKEY)
336	{	376	{
…		…
349	encode_ch (enc, '"');	389	encode_ch (enc, '"');
350		390
351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);	391	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
352	encode_ch (enc, ':');	392	encode_ch (enc, ':');
353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);	393	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
354	encode_sv (enc, HeVAL (he));
355	}	394	}
356		395
357	// compare hash entries, used when all keys are bytestrings	396	// compare hash entries, used when all keys are bytestrings
358	static int	397	static int
359	he_cmp_fast (const void a_, const void b_)	398	he_cmp_fast (const void a_, const void b_)
…		…
364	HE b = (HE **)b_;	403	HE b = (HE **)b_;
365		404
366	STRLEN la = HeKLEN (a);	405	STRLEN la = HeKLEN (a);
367	STRLEN lb = HeKLEN (b);	406	STRLEN lb = HeKLEN (b);
368		407
369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	408	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
370	cmp = la - lb;	409	cmp = lb - la;
371		410
372	return cmp;	411	return cmp;
373	}	412	}
374		413
375	// compare hash entries, used when some keys are sv's or utf-x	414	// compare hash entries, used when some keys are sv's or utf-x
376	static int	415	static int
377	he_cmp_slow (const void a, const void b)	416	he_cmp_slow (const void a, const void b)
378	{	417	{
379	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	418	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
380	}	419	}
381		420
382	static void	421	static void
383	encode_hv (enc_t enc, HV hv)	422	encode_hv (enc_t enc, HV hv)
384	{	423	{
		424	HE *he;
385	int count, i;	425	int count;
386		426
387	if (enc->indent >= enc->maxdepth)	427	if (enc->indent >= enc->maxdepth)
388	croak ("data structure too deep (hit recursion limit)");	428	croak ("data structure too deep (hit recursion limit)");
389		429
390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	430	encode_ch (enc, '{');
391		431
392	if ((count = hv_iterinit (hv)))
393	{
394	// for canonical output we have to sort by keys first	432	// for canonical output we have to sort by keys first
395	// actually, this is mostly due to the stupid so-called	433	// actually, this is mostly due to the stupid so-called
396	// security workaround added somewhere in 5.8.x.	434	// security workaround added somewhere in 5.8.x.
397	// that randomises hash orderings	435	// that randomises hash orderings
398	if (enc->json.flags & F_CANONICAL)	436	if (enc->json.flags & F_CANONICAL)
		437	{
		438	int count = hv_iterinit (hv);
		439
		440	if (SvMAGICAL (hv))
399	{	441	{
		442	// need to count by iterating. could improve by dynamically building the vector below
		443	// but I don't care for the speed of this special case.
		444	// note also that we will run into undefined behaviour when the two iterations
		445	// do not result in the same count, something I might care for in some later release.
		446
		447	count = 0;
		448	while (hv_iternext (hv))
		449	++count;
		450
		451	hv_iterinit (hv);
		452	}
		453
		454	if (count)
		455	{
400	int fast = 1;	456	int i, fast = 1;
401	HE *he;
402	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	457	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
403	HE *hes = _alloca (count sizeof (HE));	458	HE *hes = _alloca (count sizeof (HE));
404	#else	459	#else
405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	460	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
406	#endif	461	#endif
…		…
433		488
434	FREETMPS;	489	FREETMPS;
435	LEAVE;	490	LEAVE;
436	}	491	}
437		492
438	for (i = 0; i < count; ++i)	493	encode_nl (enc); ++enc->indent;
		494
		495	while (count--)
439	{	496	{
440	encode_indent (enc);	497	encode_indent (enc);
		498	he = hes [count];
441	encode_he (enc, hes [i]);	499	encode_hk (enc, he);
		500	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
442		501
443	if (i < count - 1)	502	if (count)
444	encode_comma (enc);	503	encode_comma (enc);
445	}	504	}
446		505
447	encode_nl (enc);	506	encode_nl (enc); --enc->indent; encode_indent (enc);
448	}	507	}
		508	}
449	else	509	else
		510	{
		511	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		512	if ((he = hv_iternext (hv)))
450	{	513	{
451	HE *he = hv_iternext (hv);	514	encode_nl (enc); ++enc->indent;
452		515
453	for (;;)	516	for (;;)
454	{	517	{
455	encode_indent (enc);	518	encode_indent (enc);
456	encode_he (enc, he);	519	encode_hk (enc, he);
		520	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
457		521
458	if (!(he = hv_iternext (hv)))	522	if (!(he = hv_iternext (hv)))
459	break;	523	break;
460		524
461	encode_comma (enc);	525	encode_comma (enc);
462	}	526	}
463		527
464	encode_nl (enc);	528	encode_nl (enc); --enc->indent; encode_indent (enc);
465	}	529	}
466	}	530	}
467		531
468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	532	encode_ch (enc, '}');
469	}	533	}
470		534
471	// encode objects, arrays and special \0=false and \1=true values.	535	// encode objects, arrays and special \0=false and \1=true values.
472	static void	536	static void
473	encode_rv (enc_t enc, SV sv)	537	encode_rv (enc_t enc, SV sv)
…		…
477	SvGETMAGIC (sv);	541	SvGETMAGIC (sv);
478	svt = SvTYPE (sv);	542	svt = SvTYPE (sv);
479		543
480	if (expect_false (SvOBJECT (sv)))	544	if (expect_false (SvOBJECT (sv)))
481	{	545	{
		546	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		547	? json_boolean_stash
		548	: gv_stashpv ("JSON::XS::Boolean", 1);
		549
482	if (SvSTASH (sv) == json_boolean_stash)	550	if (SvSTASH (sv) == stash)
483	{	551	{
484	if (SvIV (sv))	552	if (SvIV (sv))
485	encode_str (enc, "true", 4, 0);	553	encode_str (enc, "true", 4, 0);
486	else	554	else
487	encode_str (enc, "false", 5, 0);	555	encode_str (enc, "false", 5, 0);
…		…
499	// we re-bless the reference to get overload and other niceties right	567	// we re-bless the reference to get overload and other niceties right
500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);	568	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
501		569
502	if (to_json)	570	if (to_json)
503	{	571	{
504	int count;
505	dSP;	572	dSP;
506		573
507	ENTER; SAVETMPS; PUSHMARK (SP);	574	ENTER; SAVETMPS; PUSHMARK (SP);
508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));	575	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
509		576
…		…
578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	645	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
579	enc->cur += strlen (enc->cur);	646	enc->cur += strlen (enc->cur);
580	}	647	}
581	else if (SvIOKp (sv))	648	else if (SvIOKp (sv))
582	{	649	{
583	// we assume we can always read an IV as a UV	650	// we assume we can always read an IV as a UV and vice versa
584	if (SvUV (sv) & ~(UV)0x7fff)	651	// we assume two's complement
		652	// we assume no aliasing issues in the union
		653	if (SvIsUV (sv) ? SvUVX (sv) > 59000
		654	: SvIVX (sv) > 59000 \|\| SvIVX (sv) < -59000)
585	{	655	{
586	// large integer, use the (rather slow) snprintf way.	656	// large integer, use the (rather slow) snprintf way.
587	need (enc, sizeof (UV) * 3);	657	need (enc, sizeof (UV) * 5 / 2 + 1); // CHAR_BIT is at least 8
588	enc->cur +=	658	enc->cur +=
589	SvIsUV(sv)	659	SvIsUV(sv)
590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))	660	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
591	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));	661	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
592	}	662	}
593	else	663	else
594	{	664	{
595	// optimise the "small number case"	665	// optimise the "small number case"
596	// code will likely be branchless and use only a single multiplication	666	// code will likely be branchless and use only a single multiplication
		667	// works for numbers up to 59074
597	I32 i = SvIV (sv);	668	I32 i = SvIVX (sv);
598	U32 u;	669	U32 u;
599	char digit, nz = 0;	670	char digit, nz = 0;
600		671
601	need (enc, 6);	672	need (enc, 6);
602		673
…		…
639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	710	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
640	enc.cur = SvPVX (enc.sv);	711	enc.cur = SvPVX (enc.sv);
641	enc.end = SvEND (enc.sv);	712	enc.end = SvEND (enc.sv);
642	enc.indent = 0;	713	enc.indent = 0;
643	enc.maxdepth = DEC_DEPTH (enc.json.flags);	714	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		715	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		716	: enc.json.flags & F_LATIN1 ? 0x000100UL
		717	: 0x110000UL;
644		718
645	SvPOK_only (enc.sv);	719	SvPOK_only (enc.sv);
646	encode_sv (&enc, scalar);	720	encode_sv (&enc, scalar);
647		721
648	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	722	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
…		…
669	JSON json;	743	JSON json;
670	U32 depth; // recursion depth	744	U32 depth; // recursion depth
671	U32 maxdepth; // recursion depth limit	745	U32 maxdepth; // recursion depth limit
672	} dec_t;	746	} dec_t;
673		747
674	inline void	748	INLINE void
		749	decode_comment (dec_t *dec)
		750	{
		751	// only '#'-style comments allowed a.t.m.
		752
		753	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		754	++dec->cur;
		755	}
		756
		757	INLINE void
675	decode_ws (dec_t *dec)	758	decode_ws (dec_t *dec)
676	{	759	{
677	for (;;)	760	for (;;)
678	{	761	{
679	char ch = *dec->cur;	762	char ch = *dec->cur;
680		763
681	if (ch > 0x20	764	if (ch > 0x20)
		765	{
		766	if (expect_false (ch == '#'))
		767	{
		768	if (dec->json.flags & F_RELAXED)
		769	decode_comment (dec);
		770	else
		771	break;
		772	}
		773	else
		774	break;
		775	}
682	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	776	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
683	break;	777	break; // parse error, but let higher level handle it, gives better error messages
684		778
685	++dec->cur;	779	++dec->cur;
686	}	780	}
687	}	781	}
688		782
…		…
794		888
795	if (hi >= 0x80)	889	if (hi >= 0x80)
796	{	890	{
797	utf8 = 1;	891	utf8 = 1;
798		892
799	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	893	cur = encode_utf8 (cur, hi);
800	}	894	}
801	else	895	else
802	*cur++ = hi;	896	*cur++ = hi;
803	}	897	}
804	break;	898	break;
…		…
806	default:	900	default:
807	--dec_cur;	901	--dec_cur;
808	ERR ("illegal backslash escape sequence in string");	902	ERR ("illegal backslash escape sequence in string");
809	}	903	}
810	}	904	}
811	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	905	else if (expect_true (ch >= 0x20 && ch < 0x80))
812	*cur++ = ch;	906	*cur++ = ch;
813	else if (ch >= 0x80)	907	else if (ch >= 0x80)
814	{	908	{
815	STRLEN clen;	909	STRLEN clen;
816	UV uch;	910	UV uch;
…		…
937	is_nv = 1;	1031	is_nv = 1;
938	}	1032	}
939		1033
940	if (!is_nv)	1034	if (!is_nv)
941	{	1035	{
		1036	int len = dec->cur - start;
		1037
942	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1038	// special case the rather common 1..5-digit-int case
943	if (*start == '-')	1039	if (*start == '-')
944	switch (dec->cur - start)	1040	switch (len)
945	{	1041	{
946	case 2: return newSViv (-( start [1] - '0' * 1));	1042	case 2: return newSViv (-( start [1] - '0' * 1));
947	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1043	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
948	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1044	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
949	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1045	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1046	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
950	}	1047	}
951	else	1048	else
952	switch (dec->cur - start)	1049	switch (len)
953	{	1050	{
954	case 1: return newSViv ( start [0] - '0' * 1);	1051	case 1: return newSViv ( start [0] - '0' * 1);
955	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1052	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
956	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1053	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
957	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1054	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1055	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
958	}	1056	}
959		1057
960	{	1058	{
961	UV uv;	1059	UV uv;
962	int numtype = grok_number (start, dec->cur - start, &uv);	1060	int numtype = grok_number (start, len, &uv);
963	if (numtype & IS_NUMBER_IN_UV)	1061	if (numtype & IS_NUMBER_IN_UV)
964	if (numtype & IS_NUMBER_NEG)	1062	if (numtype & IS_NUMBER_NEG)
965	{	1063	{
966	if (uv < (UV)IV_MIN)	1064	if (uv < (UV)IV_MIN)
967	return newSViv (-(IV)uv);	1065	return newSViv (-(IV)uv);
968	}	1066	}
969	else	1067	else
970	return newSVuv (uv);	1068	return newSVuv (uv);
971
972	// here would likely be the place for bigint support
973	}	1069	}
974	}
975		1070
976	// if we ever support bigint or bigfloat, this is the place for bigfloat	1071	len -= *start == '-' ? 1 : 0;
		1072
		1073	// does not fit into IV or UV, try NV
		1074	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1075	#if defined (LDBL_DIG)
		1076	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1077	#endif
		1078	)
		1079	// fits into NV without loss of precision
		1080	return newSVnv (Atof (start));
		1081
		1082	// everything else fails, convert it to a string
		1083	return newSVpvn (start, dec->cur - start);
		1084	}
		1085
		1086	// loss of precision here
977	return newSVnv (Atof (start));	1087	return newSVnv (Atof (start));
978		1088
979	fail:	1089	fail:
980	return 0;	1090	return 0;
981	}	1091	}
…		…
1011		1121
1012	if (*dec->cur != ',')	1122	if (*dec->cur != ',')
1013	ERR (", or ] expected while parsing array");	1123	ERR (", or ] expected while parsing array");
1014		1124
1015	++dec->cur;	1125	++dec->cur;
		1126
		1127	decode_ws (dec);
		1128
		1129	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1130	{
		1131	++dec->cur;
		1132	break;
		1133	}
1016	}	1134	}
1017		1135
1018	DEC_DEC_DEPTH;	1136	DEC_DEC_DEPTH;
1019	return newRV_noinc ((SV *)av);	1137	return newRV_noinc ((SV *)av);
1020		1138
…		…
1036	if (*dec->cur == '}')	1154	if (*dec->cur == '}')
1037	++dec->cur;	1155	++dec->cur;
1038	else	1156	else
1039	for (;;)	1157	for (;;)
1040	{	1158	{
1041	decode_ws (dec); EXPECT_CH ('"');	1159	EXPECT_CH ('"');
1042		1160
1043	// heuristic: assume that	1161	// heuristic: assume that
1044	// a) decode_str + hv_store_ent are abysmally slow.	1162	// a) decode_str + hv_store_ent are abysmally slow.
1045	// b) most hash keys are short, simple ascii text.	1163	// b) most hash keys are short, simple ascii text.
1046	// => try to "fast-match" such strings to avoid	1164	// => try to "fast-match" such strings to avoid
…		…
1060	if (!key)	1178	if (!key)
1061	goto fail;	1179	goto fail;
1062		1180
1063	decode_ws (dec); EXPECT_CH (':');	1181	decode_ws (dec); EXPECT_CH (':');
1064		1182
		1183	decode_ws (dec);
1065	value = decode_sv (dec);	1184	value = decode_sv (dec);
1066	if (!value)	1185	if (!value)
1067	{	1186	{
1068	SvREFCNT_dec (key);	1187	SvREFCNT_dec (key);
1069	goto fail;	1188	goto fail;
…		…
1081	int len = p - key;	1200	int len = p - key;
1082	dec->cur = p + 1;	1201	dec->cur = p + 1;
1083		1202
1084	decode_ws (dec); EXPECT_CH (':');	1203	decode_ws (dec); EXPECT_CH (':');
1085		1204
		1205	decode_ws (dec);
1086	value = decode_sv (dec);	1206	value = decode_sv (dec);
1087	if (!value)	1207	if (!value)
1088	goto fail;	1208	goto fail;
1089		1209
1090	hv_store (hv, key, len, value, 0);	1210	hv_store (hv, key, len, value, 0);
…		…
1106		1226
1107	if (*dec->cur != ',')	1227	if (*dec->cur != ',')
1108	ERR (", or } expected while parsing object/hash");	1228	ERR (", or } expected while parsing object/hash");
1109		1229
1110	++dec->cur;	1230	++dec->cur;
		1231
		1232	decode_ws (dec);
		1233
		1234	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1235	{
		1236	++dec->cur;
		1237	break;
		1238	}
1111	}	1239	}
1112		1240
1113	DEC_DEC_DEPTH;	1241	DEC_DEC_DEPTH;
1114	sv = newRV_noinc ((SV *)hv);	1242	sv = newRV_noinc ((SV *)hv);
1115		1243
…		…
1180	}	1308	}
1181		1309
1182	static SV *	1310	static SV *
1183	decode_sv (dec_t *dec)	1311	decode_sv (dec_t *dec)
1184	{	1312	{
1185	decode_ws (dec);
1186
1187	// the beauty of JSON: you need exactly one character lookahead	1313	// the beauty of JSON: you need exactly one character lookahead
1188	// to parse anything.	1314	// to parse everything.
1189	switch (*dec->cur)	1315	switch (*dec->cur)
1190	{	1316	{
1191	case '"': ++dec->cur; return decode_str (dec);	1317	case '"': ++dec->cur; return decode_str (dec);
1192	case '[': ++dec->cur; return decode_av (dec);	1318	case '[': ++dec->cur; return decode_av (dec);
1193	case '{': ++dec->cur; return decode_hv (dec);	1319	case '{': ++dec->cur; return decode_hv (dec);
1194		1320
1195	case '-':	1321	case '-':
1196	case '0': case '1': case '2': case '3': case '4':	1322	case '0': case '1': case '2': case '3': case '4':
1197	case '5': case '6': case '7': case '8': case '9':	1323	case '5': case '6': case '7': case '8': case '9':
1198	return decode_num (dec);	1324	return decode_num (dec);
1199		1325
1200	case 't':	1326	case 't':
1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1327	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1202	{	1328	{
1203	dec->cur += 4;	1329	dec->cur += 4;
		1330	#if JSON_SLOW
		1331	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1332	#endif
1204	return SvREFCNT_inc (json_true);	1333	return SvREFCNT_inc (json_true);
1205	}	1334	}
1206	else	1335	else
1207	ERR ("'true' expected");	1336	ERR ("'true' expected");
1208		1337
…		…
1210		1339
1211	case 'f':	1340	case 'f':
1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1341	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1213	{	1342	{
1214	dec->cur += 5;	1343	dec->cur += 5;
		1344	#if JSON_SLOW
		1345	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1346	#endif
1215	return SvREFCNT_inc (json_false);	1347	return SvREFCNT_inc (json_false);
1216	}	1348	}
1217	else	1349	else
1218	ERR ("'false' expected");	1350	ERR ("'false' expected");
1219		1351
…		…
1269		1401
1270	if (dec.json.cb_object \|\| dec.json.cb_sk_object)	1402	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
1271	dec.json.flags \|= F_HOOK;	1403	dec.json.flags \|= F_HOOK;
1272		1404
1273	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1405	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1406
		1407	decode_ws (&dec);
1274	sv = decode_sv (&dec);	1408	sv = decode_sv (&dec);
1275		1409
1276	if (!(offset_return \|\| !sv))	1410	if (!(offset_return \|\| !sv))
1277	{	1411	{
1278	// check for trailing garbage	1412	// check for trailing garbage
…		…
1346	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);	1480	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1347	}	1481	}
1348		1482
1349	PROTOTYPES: DISABLE	1483	PROTOTYPES: DISABLE
1350		1484
		1485	void CLONE (...)
		1486	CODE:
		1487	json_stash = 0;
		1488	json_boolean_stash = 0;
		1489
1351	void new (char *klass)	1490	void new (char *klass)
1352	PPCODE:	1491	PPCODE:
1353	{	1492	{
1354	SV *pv = NEWSV (0, sizeof (JSON));	1493	SV *pv = NEWSV (0, sizeof (JSON));
1355	SvPOK_only (pv);	1494	SvPOK_only (pv);
1356	Zero (SvPVX (pv), 1, JSON);	1495	Zero (SvPVX (pv), 1, JSON);
1357	((JSON *)SvPVX (pv))->flags = F_DEFAULT;	1496	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1358	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));	1497	XPUSHs (sv_2mortal (sv_bless (
		1498	newRV_noinc (pv),
		1499	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1500	)));
1359	}	1501	}
1360		1502
1361	void ascii (JSON *self, int enable = 1)	1503	void ascii (JSON *self, int enable = 1)
1362	ALIAS:	1504	ALIAS:
1363	ascii = F_ASCII	1505	ascii = F_ASCII
…		…
1370	pretty = F_PRETTY	1512	pretty = F_PRETTY
1371	allow_nonref = F_ALLOW_NONREF	1513	allow_nonref = F_ALLOW_NONREF
1372	shrink = F_SHRINK	1514	shrink = F_SHRINK
1373	allow_blessed = F_ALLOW_BLESSED	1515	allow_blessed = F_ALLOW_BLESSED
1374	convert_blessed = F_CONV_BLESSED	1516	convert_blessed = F_CONV_BLESSED
		1517	relaxed = F_RELAXED
1375	PPCODE:	1518	PPCODE:
1376	{	1519	{
1377	if (enable)	1520	if (enable)
1378	self->flags \|= ix;	1521	self->flags \|= ix;
1379	else	1522	else
1380	self->flags &= ~ix;	1523	self->flags &= ~ix;
1381		1524
1382	XPUSHs (ST (0));	1525	XPUSHs (ST (0));
1383	}	1526	}
1384		1527
		1528	void get_ascii (JSON *self)
		1529	ALIAS:
		1530	get_ascii = F_ASCII
		1531	get_latin1 = F_LATIN1
		1532	get_utf8 = F_UTF8
		1533	get_indent = F_INDENT
		1534	get_canonical = F_CANONICAL
		1535	get_space_before = F_SPACE_BEFORE
		1536	get_space_after = F_SPACE_AFTER
		1537	get_allow_nonref = F_ALLOW_NONREF
		1538	get_shrink = F_SHRINK
		1539	get_allow_blessed = F_ALLOW_BLESSED
		1540	get_convert_blessed = F_CONV_BLESSED
		1541	get_relaxed = F_RELAXED
		1542	PPCODE:
		1543	XPUSHs (boolSV (self->flags & ix));
		1544
1385	void max_depth (JSON *self, UV max_depth = 0x80000000UL)	1545	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1386	PPCODE:	1546	PPCODE:
1387	{	1547	{
1388	UV log2 = 0;	1548	UV log2 = 0;
1389		1549
…		…
1395	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1555	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1396		1556
1397	XPUSHs (ST (0));	1557	XPUSHs (ST (0));
1398	}	1558	}
1399		1559
		1560	U32 get_max_depth (JSON *self)
		1561	CODE:
		1562	RETVAL = DEC_DEPTH (self->flags);
		1563	OUTPUT:
		1564	RETVAL
		1565
1400	void max_size (JSON *self, UV max_size = 0)	1566	void max_size (JSON *self, UV max_size = 0)
1401	PPCODE:	1567	PPCODE:
1402	{	1568	{
1403	UV log2 = 0;	1569	UV log2 = 0;
1404		1570
…		…
1410		1576
1411	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);	1577	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
1412		1578
1413	XPUSHs (ST (0));	1579	XPUSHs (ST (0));
1414	}	1580	}
		1581
		1582	int get_max_size (JSON *self)
		1583	CODE:
		1584	RETVAL = DEC_SIZE (self->flags);
		1585	OUTPUT:
		1586	RETVAL
1415		1587
1416	void filter_json_object (JSON self, SV cb = &PL_sv_undef)	1588	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1417	PPCODE:	1589	PPCODE:
1418	{	1590	{
1419	SvREFCNT_dec (self->cb_object);	1591	SvREFCNT_dec (self->cb_object);
…		…
1466	SvREFCNT_dec (self->cb_sk_object);	1638	SvREFCNT_dec (self->cb_sk_object);
1467	SvREFCNT_dec (self->cb_object);	1639	SvREFCNT_dec (self->cb_object);
1468		1640
1469	PROTOTYPES: ENABLE	1641	PROTOTYPES: ENABLE
1470		1642
1471	void to_json (SV *scalar)	1643	void encode_json (SV *scalar)
1472	PPCODE:	1644	PPCODE:
1473	{	1645	{
1474	JSON json = { F_DEFAULT \| F_UTF8 };	1646	JSON json = { F_DEFAULT \| F_UTF8 };
1475	XPUSHs (encode_json (scalar, &json));	1647	XPUSHs (encode_json (scalar, &json));
1476	}	1648	}
1477		1649
1478	void from_json (SV *jsonstr)	1650	void decode_json (SV *jsonstr)
1479	PPCODE:	1651	PPCODE:
1480	{	1652	{
1481	JSON json = { F_DEFAULT \| F_UTF8 };	1653	JSON json = { F_DEFAULT \| F_UTF8 };
1482	XPUSHs (decode_json (jsonstr, &json, 0));	1654	XPUSHs (decode_json (jsonstr, &json, 0));
1483	}	1655	}

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs. Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.55 by root, Mon Jul 23 22:57:40 2007 UTC vs.
Revision 1.74 by root, Wed Mar 19 15:17:53 2008 UTC