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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.36 by root, Wed Jun 6 17:27:33 2007 UTC vs.
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 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
		15	// 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	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
14	#define F_ASCII 0x00000001UL	21	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	22	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	23	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	24	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	25	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	26	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	27	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	28	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
23	#define F_MAXDEPTH 0xf8000000UL	34	#define F_MAXDEPTH 0xf8000000UL
24	#define S_MAXDEPTH 27	35	#define S_MAXDEPTH 27
		36	#define F_MAXSIZE 0x01f00000UL
		37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
25		39
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		42
31	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	43	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		45
34	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
35	#define INDENT_STEP 3 // spaces per indentation level	47	#define INDENT_STEP 3 // spaces per indentation level
36		48
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38		50
39	#define SB do {	51	#define SB do {
40	#define SE } while (0)	52	#define SE } while (0)
41		53
42	#if __GNUC__ >= 3	54	#if __GNUC__ >= 3
48	#endif	60	#endif
49		61
50	#define expect_false(expr) expect ((expr) != 0, 0)	62	#define expect_false(expr) expect ((expr) != 0, 0)
51	#define expect_true(expr) expect ((expr) != 0, 1)	63	#define expect_true(expr) expect ((expr) != 0, 1)
52		64
		65	#ifdef USE_ITHREADS
		66	# define JSON_SLOW 1
		67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		68	#else
		69	# define JSON_SLOW 0
		70	# define JSON_STASH json_stash
		71	#endif
		72
53	static HV *json_stash; // JSON::XS::	73	static HV json_stash, json_boolean_stash; // JSON::XS::
		74	static SV json_true, json_false;
		75
		76	typedef struct {
		77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
		80	} JSON;
54		81
55	/////////////////////////////////////////////////////////////////////////////	82	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	83	// utility functions
57		84
58	static UV *	85	inline void
59	SvJSON (SV *sv)
60	{
61	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
62	croak ("object is not of type JSON::XS");
63
64	return &SvUVX (SvRV (sv));
65	}
66
67	static void
68	shrink (SV *sv)	86	shrink (SV *sv)
69	{	87	{
70	sv_utf8_downgrade (sv, 1);	88	sv_utf8_downgrade (sv, 1);
71	if (SvLEN (sv) > SvCUR (sv) + 1)	89	if (SvLEN (sv) > SvCUR (sv) + 1)
72	{	90	{
…		…
107	typedef struct	125	typedef struct
108	{	126	{
109	char *cur; // SvPVX (sv) + current output position	127	char *cur; // SvPVX (sv) + current output position
110	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
111	SV *sv; // result scalar	129	SV *sv; // result scalar
112	U32 flags; // F_*	130	JSON json;
113	U32 indent; // indentation level	131	U32 indent; // indentation level
114	U32 maxdepth; // max. indentation/recursion level	132	U32 maxdepth; // max. indentation/recursion level
115	} enc_t;	133	} enc_t;
116		134
117	inline void	135	inline void
…		…
178	STRLEN clen;	196	STRLEN clen;
179	UV uch;	197	UV uch;
180		198
181	if (is_utf8)	199	if (is_utf8)
182	{	200	{
183	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
184	uch = decode_utf8 (str, end - str, &clen);	201	uch = decode_utf8 (str, end - str, &clen);
185	if (clen == (STRLEN)-1)	202	if (clen == (STRLEN)-1)
186	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	203	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
187	}	204	}
188	else	205	else
…		…
192	}	209	}
193		210
194	if (uch > 0x10FFFFUL)	211	if (uch > 0x10FFFFUL)
195	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
196		213
197	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))	214	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
198	{	215	{
199	if (uch > 0xFFFFUL)	216	if (uch > 0xFFFFUL)
200	{	217	{
201	need (enc, len += 11);	218	need (enc, len += 11);
202	sprintf (enc->cur, "\\u%04x\\u%04x",	219	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
216	*enc->cur++ = hexdigit [(uch >> 0) & 15];	233	*enc->cur++ = hexdigit [(uch >> 0) & 15];
217	}	234	}
218		235
219	str += clen;	236	str += clen;
220	}	237	}
221	else if (enc->flags & F_LATIN1)	238	else if (enc->json.flags & F_LATIN1)
222	{	239	{
223	*enc->cur++ = uch;	240	*enc->cur++ = uch;
224	str += clen;	241	str += clen;
225	}	242	}
226	else if (is_utf8)	243	else if (is_utf8)
…		…
247	}	264	}
248		265
249	inline void	266	inline void
250	encode_indent (enc_t *enc)	267	encode_indent (enc_t *enc)
251	{	268	{
252	if (enc->flags & F_INDENT)	269	if (enc->json.flags & F_INDENT)
253	{	270	{
254	int spaces = enc->indent * INDENT_STEP;	271	int spaces = enc->indent * INDENT_STEP;
255		272
256	need (enc, spaces);	273	need (enc, spaces);
257	memset (enc->cur, ' ', spaces);	274	memset (enc->cur, ' ', spaces);
…		…
267	}	284	}
268		285
269	inline void	286	inline void
270	encode_nl (enc_t *enc)	287	encode_nl (enc_t *enc)
271	{	288	{
272	if (enc->flags & F_INDENT)	289	if (enc->json.flags & F_INDENT)
273	{	290	{
274	need (enc, 1);	291	need (enc, 1);
275	encode_ch (enc, '\n');	292	encode_ch (enc, '\n');
276	}	293	}
277	}	294	}
…		…
279	inline void	296	inline void
280	encode_comma (enc_t *enc)	297	encode_comma (enc_t *enc)
281	{	298	{
282	encode_ch (enc, ',');	299	encode_ch (enc, ',');
283		300
284	if (enc->flags & F_INDENT)	301	if (enc->json.flags & F_INDENT)
285	encode_nl (enc);	302	encode_nl (enc);
286	else if (enc->flags & F_SPACE_AFTER)	303	else if (enc->json.flags & F_SPACE_AFTER)
287	encode_space (enc);	304	encode_space (enc);
288	}	305	}
289		306
290	static void encode_sv (enc_t enc, SV sv);	307	static void encode_sv (enc_t enc, SV sv);
291		308
…		…
300	encode_ch (enc, '['); encode_nl (enc);	317	encode_ch (enc, '['); encode_nl (enc);
301	++enc->indent;	318	++enc->indent;
302		319
303	for (i = 0; i <= len; ++i)	320	for (i = 0; i <= len; ++i)
304	{	321	{
		322	SV **svp = av_fetch (av, i, 0);
		323
305	encode_indent (enc);	324	encode_indent (enc);
306	encode_sv (enc, *av_fetch (av, i, 0));	325
		326	if (svp)
		327	encode_sv (enc, *svp);
		328	else
		329	encode_str (enc, "null", 4, 0);
307		330
308	if (i < len)	331	if (i < len)
309	encode_comma (enc);	332	encode_comma (enc);
310	}	333	}
311		334
…		…
314	--enc->indent;	337	--enc->indent;
315	encode_indent (enc); encode_ch (enc, ']');	338	encode_indent (enc); encode_ch (enc, ']');
316	}	339	}
317		340
318	static void	341	static void
319	encode_he (enc_t enc, HE he)	342	encode_hk (enc_t enc, HE he)
320	{	343	{
321	encode_ch (enc, '"');	344	encode_ch (enc, '"');
322		345
323	if (HeKLEN (he) == HEf_SVKEY)	346	if (HeKLEN (he) == HEf_SVKEY)
324	{	347	{
…		…
334	else	357	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	358	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		359
337	encode_ch (enc, '"');	360	encode_ch (enc, '"');
338		361
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	362	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	363	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	364	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));
343	}	365	}
344		366
345	// compare hash entries, used when all keys are bytestrings	367	// compare hash entries, used when all keys are bytestrings
346	static int	368	static int
347	he_cmp_fast (const void a_, const void b_)	369	he_cmp_fast (const void a_, const void b_)
…		…
352	HE b = (HE **)b_;	374	HE b = (HE **)b_;
353		375
354	STRLEN la = HeKLEN (a);	376	STRLEN la = HeKLEN (a);
355	STRLEN lb = HeKLEN (b);	377	STRLEN lb = HeKLEN (b);
356		378
357	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	379	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
358	cmp = la - lb;	380	cmp = lb - la;
359		381
360	return cmp;	382	return cmp;
361	}	383	}
362		384
363	// compare hash entries, used when some keys are sv's or utf-x	385	// compare hash entries, used when some keys are sv's or utf-x
364	static int	386	static int
365	he_cmp_slow (const void a, const void b)	387	he_cmp_slow (const void a, const void b)
366	{	388	{
367	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	389	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
368	}	390	}
369		391
370	static void	392	static void
371	encode_hv (enc_t enc, HV hv)	393	encode_hv (enc_t enc, HV hv)
372	{	394	{
		395	HE *he;
373	int count, i;	396	int count;
374		397
375	if (enc->indent >= enc->maxdepth)	398	if (enc->indent >= enc->maxdepth)
376	croak ("data structure too deep (hit recursion limit)");	399	croak ("data structure too deep (hit recursion limit)");
377		400
378	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	401	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
379		402
380	if ((count = hv_iterinit (hv)))
381	{
382	// for canonical output we have to sort by keys first	403	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	404	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	405	// security workaround added somewhere in 5.8.x.
385	// that randomises hash orderings	406	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	407	if (enc->json.flags & F_CANONICAL)
		408	{
		409	int count = hv_iterinit (hv);
		410
		411	if (SvMAGICAL (hv))
387	{	412	{
		413	// need to count by iterating. could improve by dynamically building the vector below
		414	// but I don't care for the speed of this special case.
		415	// note also that we will run into undefined behaviour when the two iterations
		416	// do not result in the same count, something I might care for in some later release.
		417
		418	count = 0;
		419	while (hv_iternext (hv))
		420	++count;
		421
		422	hv_iterinit (hv);
		423	}
		424
		425	if (count)
		426	{
388	int fast = 1;	427	int i, fast = 1;
389	HE *he;
390	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	428	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
391	HE *hes = _alloca (count sizeof (HE));	429	HE *hes = _alloca (count sizeof (HE));
392	#else	430	#else
393	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	431	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394	#endif	432	#endif
…		…
421		459
422	FREETMPS;	460	FREETMPS;
423	LEAVE;	461	LEAVE;
424	}	462	}
425		463
426	for (i = 0; i < count; ++i)	464	while (count--)
427	{	465	{
428	encode_indent (enc);	466	encode_indent (enc);
		467	he = hes [count];
429	encode_he (enc, hes [i]);	468	encode_hk (enc, he);
		469	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
430		470
431	if (i < count - 1)	471	if (count)
432	encode_comma (enc);	472	encode_comma (enc);
433	}	473	}
434		474	}
		475	}
		476	else
		477	{
		478	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		479	if ((he = hv_iternext (hv)))
		480	for (;;)
		481	{
435	encode_nl (enc);	482	encode_indent (enc);
		483	encode_hk (enc, he);
		484	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		485
		486	if (!(he = hv_iternext (hv)))
		487	break;
		488
		489	encode_comma (enc);
		490	}
		491	}
		492
		493	encode_nl (enc);
		494
		495	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		496	}
		497
		498	// encode objects, arrays and special \0=false and \1=true values.
		499	static void
		500	encode_rv (enc_t enc, SV sv)
		501	{
		502	svtype svt;
		503
		504	SvGETMAGIC (sv);
		505	svt = SvTYPE (sv);
		506
		507	if (expect_false (SvOBJECT (sv)))
		508	{
		509	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		510	? json_boolean_stash
		511	: gv_stashpv ("JSON::XS::Boolean", 1);
		512
		513	if (SvSTASH (sv) == stash)
		514	{
		515	if (SvIV (sv))
		516	encode_str (enc, "true", 4, 0);
		517	else
		518	encode_str (enc, "false", 5, 0);
436	}	519	}
437	else	520	else
438	{	521	{
439	HE *he = hv_iternext (hv);	522	#if 0
440		523	if (0 && sv_derived_from (rv, "JSON::Literal"))
441	for (;;)
442	{
443	encode_indent (enc);
444	encode_he (enc, he);
445
446	if (!(he = hv_iternext (hv)))
447	break;
448
449	encode_comma (enc);
450	}	524	{
		525	// not yet
		526	}
		527	#endif
		528	if (enc->json.flags & F_CONV_BLESSED)
		529	{
		530	// we re-bless the reference to get overload and other niceties right
		531	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
451		532
		533	if (to_json)
		534	{
		535	dSP;
		536
		537	ENTER; SAVETMPS; PUSHMARK (SP);
		538	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		539
		540	// calling with G_SCALAR ensures that we always get a 1 return value
		541	PUTBACK;
		542	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		543	SPAGAIN;
		544
		545	// catch this surprisingly common error
		546	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		547	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		548
		549	sv = POPs;
		550	PUTBACK;
		551
452	encode_nl (enc);	552	encode_sv (enc, sv);
		553
		554	FREETMPS; LEAVE;
		555	}
		556	else if (enc->json.flags & F_ALLOW_BLESSED)
		557	encode_str (enc, "null", 4, 0);
		558	else
		559	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		560	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		561	}
		562	else if (enc->json.flags & F_ALLOW_BLESSED)
		563	encode_str (enc, "null", 4, 0);
		564	else
		565	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
453	}	567	}
454	}	568	}
455
456	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
457	}
458
459	// encode objects, arrays and special \0=false and \1=true values.
460	static void
461	encode_rv (enc_t enc, SV sv)
462	{
463	svtype svt;
464
465	SvGETMAGIC (sv);
466	svt = SvTYPE (sv);
467
468	if (svt == SVt_PVHV)	569	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	570	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	571	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	572	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	573	else if (svt < SVt_PVAV)
473	{	574	{
474	if (SvNIOK (sv) && SvIV (sv) == 0)	575	STRLEN len = 0;
		576	char *pv = svt ? SvPV (sv, len) : 0;
		577
		578	if (len == 1 && *pv == '1')
		579	encode_str (enc, "true", 4, 0);
		580	else if (len == 1 && *pv == '0')
475	encode_str (enc, "false", 5, 0);	581	encode_str (enc, "false", 5, 0);
476	else if (SvNIOK (sv) && SvIV (sv) == 1)
477	encode_str (enc, "true", 4, 0);
478	else	582	else
479	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	583	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
480	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	584	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
481	}	585	}
482	else	586	else
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	601	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	602	encode_ch (enc, '"');
499	}	603	}
500	else if (SvNOKp (sv))	604	else if (SvNOKp (sv))
501	{	605	{
		606	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	607	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	608	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	609	enc->cur += strlen (enc->cur);
505	}	610	}
506	else if (SvIOKp (sv))	611	else if (SvIOKp (sv))
507	{	612	{
508	// we assume we can always read an IV as a UV	613	// we assume we can always read an IV as a UV
509	if (SvUV (sv) & ~(UV)0x7fff)	614	if (SvUV (sv) & ~(UV)0x7fff)
510	{	615	{
511	need (enc, 64);	616	// large integer, use the (rather slow) snprintf way.
		617	need (enc, sizeof (UV) * 3);
512	enc->cur +=	618	enc->cur +=
513	SvIsUV(sv)	619	SvIsUV(sv)
514	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	620	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
515	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	621	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
516	}	622	}
517	else	623	else
518	{	624	{
519	// optimise the "small number case"	625	// optimise the "small number case"
520	// code will likely be branchless and use only a single multiplication	626	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);	627	I32 i = SvIV (sv);
522	U32 u;	628	U32 u;
		629	char digit, nz = 0;
523		630
524	need (enc, 6);	631	need (enc, 6);
525		632
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	633	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
527	u = i < 0 ? -i : i;	634	u = i < 0 ? -i : i;
528		635
529	// convert to 4.28 fixed-point representation	636	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits	637	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531		638
532	char digit, nz = 0;	639	// now output digit by digit, each time masking out the integer part
533		640	// and multiplying by 5 while moving the decimal point one to the right,
		641	// resulting in a net multiplication by 10.
		642	// we always write the digit to memory but conditionally increment
		643	// the pointer, to ease the usage of conditional move instructions.
534	digit = u >> 28; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0xfffffff) 5;	644	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
535	digit = u >> 27; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x7ffffff) 5;	645	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
536	digit = u >> 26; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x3ffffff) 5;	646	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
537	digit = u >> 25; enc->cur = digit + '0'; nz \|= digit; enc->cur += nz ? 1 : 0; u = (u & 0x1ffffff) 5;	647	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
538	digit = u >> 24; *enc->cur = digit + '0'; nz \|= digit; enc->cur += 1;	648	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
539	}	649	}
540	}	650	}
541	else if (SvROK (sv))	651	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	652	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	653	else if (!SvOK (sv))
…		…
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	656	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	657	SvPV_nolen (sv), SvFLAGS (sv));
548	}	658	}
549		659
550	static SV *	660	static SV *
551	encode_json (SV *scalar, U32 flags)	661	encode_json (SV scalar, JSON json)
552	{	662	{
553	enc_t enc;	663	enc_t enc;
554		664
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	665	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	666	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		667
558	enc.flags = flags;	668	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	669	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	670	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	671	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	672	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	673	enc.maxdepth = DEC_DEPTH (enc.json.flags);
564		674
565	SvPOK_only (enc.sv);	675	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	676	encode_sv (&enc, scalar);
567		677
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	678	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	679	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		680
571	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	681	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
572	SvUTF8_on (enc.sv);	682	SvUTF8_on (enc.sv);
573		683
574	if (enc.flags & F_SHRINK)	684	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	685	shrink (enc.sv);
576		686
577	return enc.sv;	687	return enc.sv;
578	}	688	}
579		689
…		…
584	typedef struct	694	typedef struct
585	{	695	{
586	char *cur; // current parser pointer	696	char *cur; // current parser pointer
587	char *end; // end of input string	697	char *end; // end of input string
588	const char *err; // parse error, if != 0	698	const char *err; // parse error, if != 0
589	U32 flags; // F_*	699	JSON json;
590	U32 depth; // recursion depth	700	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	701	U32 maxdepth; // recursion depth limit
592	} dec_t;	702	} dec_t;
593		703
594	inline void	704	inline void
		705	decode_comment (dec_t *dec)
		706	{
		707	// only '#'-style comments allowed a.t.m.
		708
		709	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		710	++dec->cur;
		711	}
		712
		713	inline void
595	decode_ws (dec_t *dec)	714	decode_ws (dec_t *dec)
596	{	715	{
597	for (;;)	716	for (;;)
598	{	717	{
599	char ch = *dec->cur;	718	char ch = *dec->cur;
600		719
601	if (ch > 0x20	720	if (ch > 0x20)
		721	{
		722	if (expect_false (ch == '#'))
		723	{
		724	if (dec->json.flags & F_RELAXED)
		725	decode_comment (dec);
		726	else
		727	break;
		728	}
		729	else
		730	break;
		731	}
602	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	732	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
603	break;	733	break; // parse error, but let higher level handle it, gives better error messages
604		734
605	++dec->cur;	735	++dec->cur;
606	}	736	}
607	}	737	}
608		738
…		…
625	decode_4hex (dec_t *dec)	755	decode_4hex (dec_t *dec)
626	{	756	{
627	signed char d1, d2, d3, d4;	757	signed char d1, d2, d3, d4;
628	unsigned char cur = (unsigned char )dec->cur;	758	unsigned char cur = (unsigned char )dec->cur;
629		759
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	760	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
631	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("four hexadecimal digits expected");	761	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
632	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("four hexadecimal digits expected");	762	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
633	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("four hexadecimal digits expected");	763	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		764
635	dec->cur += 4;	765	dec->cur += 4;
636		766
637	return ((UV)d1) << 12	767	return ((UV)d1) << 12
638	\| ((UV)d2) << 8	768	\| ((UV)d2) << 8
…		…
646	static SV *	776	static SV *
647	decode_str (dec_t *dec)	777	decode_str (dec_t *dec)
648	{	778	{
649	SV *sv = 0;	779	SV *sv = 0;
650	int utf8 = 0;	780	int utf8 = 0;
		781	char *dec_cur = dec->cur;
651		782
652	do	783	do
653	{	784	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	785	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	786	char *cur = buf;
656		787
657	do	788	do
658	{	789	{
659	unsigned char ch = (unsigned char )dec->cur++;	790	unsigned char ch = (unsigned char )dec_cur++;
660		791
661	if (expect_false (ch == '"'))	792	if (expect_false (ch == '"'))
662	{	793	{
663	--dec->cur;	794	--dec_cur;
664	break;	795	break;
665	}	796	}
666	else if (expect_false (ch == '\\'))	797	else if (expect_false (ch == '\\'))
667	{	798	{
668	switch (*dec->cur)	799	switch (*dec_cur)
669	{	800	{
670	case '\\':	801	case '\\':
671	case '/':	802	case '/':
672	case '"': cur++ = dec->cur++; break;	803	case '"': cur++ = dec_cur++; break;
673		804
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	805	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	806	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	807	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	808	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	809	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		810
680	case 'u':	811	case 'u':
681	{	812	{
682	UV lo, hi;	813	UV lo, hi;
683	++dec->cur;	814	++dec_cur;
684		815
		816	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	817	hi = decode_4hex (dec);
		818	dec_cur = dec->cur;
686	if (hi == (UV)-1)	819	if (hi == (UV)-1)
687	goto fail;	820	goto fail;
688		821
689	// possibly a surrogate pair	822	// possibly a surrogate pair
690	if (hi >= 0xd800)	823	if (hi >= 0xd800)
691	if (hi < 0xdc00)	824	if (hi < 0xdc00)
692	{	825	{
693	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	826	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	827	ERR ("missing low surrogate character in surrogate pair");
695		828
696	dec->cur += 2;	829	dec_cur += 2;
697		830
		831	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	832	lo = decode_4hex (dec);
		833	dec_cur = dec->cur;
699	if (lo == (UV)-1)	834	if (lo == (UV)-1)
700	goto fail;	835	goto fail;
701		836
702	if (lo < 0xdc00 \|\| lo >= 0xe000)	837	if (lo < 0xdc00 \|\| lo >= 0xe000)
703	ERR ("surrogate pair expected");	838	ERR ("surrogate pair expected");
…		…
717	*cur++ = hi;	852	*cur++ = hi;
718	}	853	}
719	break;	854	break;
720		855
721	default:	856	default:
722	--dec->cur;	857	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	858	ERR ("illegal backslash escape sequence in string");
724	}	859	}
725	}	860	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	861	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
727	*cur++ = ch;	862	*cur++ = ch;
728	else if (ch >= 0x80)	863	else if (ch >= 0x80)
729	{	864	{
730	STRLEN clen;	865	STRLEN clen;
731	UV uch;	866	UV uch;
732		867
733	--dec->cur;	868	--dec_cur;
734		869
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	870	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	871	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	872	ERR ("malformed UTF-8 character in JSON string");
738		873
739	do	874	do
740	cur++ = dec->cur++;	875	cur++ = dec_cur++;
741	while (--clen);	876	while (--clen);
742		877
743	utf8 = 1;	878	utf8 = 1;
744	}	879	}
745	else	880	else
746	{	881	{
747	--dec->cur;	882	--dec_cur;
748		883
749	if (!ch)	884	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	885	ERR ("unexpected end of string while parsing JSON string");
751	else	886	else
752	ERR ("invalid character encountered while parsing JSON string");	887	ERR ("invalid character encountered while parsing JSON string");
…		…
765	}	900	}
766	else	901	else
767	sv = newSVpvn (buf, len);	902	sv = newSVpvn (buf, len);
768	}	903	}
769	}	904	}
770	while (*dec->cur != '"');	905	while (*dec_cur != '"');
771		906
772	++dec->cur;	907	++dec_cur;
773		908
774	if (sv)	909	if (sv)
775	{	910	{
776	SvPOK_only (sv);	911	SvPOK_only (sv);
777	*SvEND (sv) = 0;	912	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	915	SvUTF8_on (sv);
781	}	916	}
782	else	917	else
783	sv = newSVpvn ("", 0);	918	sv = newSVpvn ("", 0);
784		919
		920	dec->cur = dec_cur;
785	return sv;	921	return sv;
786		922
787	fail:	923	fail:
		924	dec->cur = dec_cur;
788	return 0;	925	return 0;
789	}	926	}
790		927
791	static SV *	928	static SV *
792	decode_num (dec_t *dec)	929	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	987	is_nv = 1;
851	}	988	}
852		989
853	if (!is_nv)	990	if (!is_nv)
854	{	991	{
		992	int len = dec->cur - start;
		993
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	994	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
856	if (*start == '-')	995	if (*start == '-')
857	switch (dec->cur - start)	996	switch (len)
858	{	997	{
859	case 2: return newSViv (-( start [1] - '0' ));	998	case 2: return newSViv (-( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	999	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
861	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	1000	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
862	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	1001	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
863	}	1002	}
864	else	1003	else
865	switch (dec->cur - start)	1004	switch (len)
866	{	1005	{
867	case 1: return newSViv ( start [0] - '0' );	1006	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1007	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
869	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	1008	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
870	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	1009	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
871	}	1010	}
872		1011
873	{	1012	{
874	UV uv;	1013	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	1014	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	1015	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	1016	if (numtype & IS_NUMBER_NEG)
878	{	1017	{
879	if (uv < (UV)IV_MIN)	1018	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1019	return newSViv (-(IV)uv);
881	}	1020	}
882	else	1021	else
883	return newSVuv (uv);	1022	return newSVuv (uv);
884	}	1023	}
885	}
886		1024
		1025	len -= *start == '-' ? 1 : 0;
		1026
		1027	// does not fit into IV or UV, try NV
		1028	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1029	#if defined (LDBL_DIG)
		1030	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1031	#endif
		1032	)
		1033	// fits into NV without loss of precision
		1034	return newSVnv (Atof (start));
		1035
		1036	// everything else fails, convert it to a string
		1037	return newSVpvn (start, dec->cur - start);
		1038	}
		1039
		1040	// loss of precision here
887	return newSVnv (Atof (start));	1041	return newSVnv (Atof (start));
888		1042
889	fail:	1043	fail:
890	return 0;	1044	return 0;
891	}	1045	}
…		…
921		1075
922	if (*dec->cur != ',')	1076	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1077	ERR (", or ] expected while parsing array");
924		1078
925	++dec->cur;	1079	++dec->cur;
		1080
		1081	decode_ws (dec);
		1082
		1083	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1084	{
		1085	++dec->cur;
		1086	break;
		1087	}
926	}	1088	}
927		1089
928	DEC_DEC_DEPTH;	1090	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1091	return newRV_noinc ((SV *)av);
930		1092
…		…
935	}	1097	}
936		1098
937	static SV *	1099	static SV *
938	decode_hv (dec_t *dec)	1100	decode_hv (dec_t *dec)
939	{	1101	{
		1102	SV *sv;
940	HV *hv = newHV ();	1103	HV *hv = newHV ();
941		1104
942	DEC_INC_DEPTH;	1105	DEC_INC_DEPTH;
943	decode_ws (dec);	1106	decode_ws (dec);
944		1107
945	if (*dec->cur == '}')	1108	if (*dec->cur == '}')
946	++dec->cur;	1109	++dec->cur;
947	else	1110	else
948	for (;;)	1111	for (;;)
949	{	1112	{
		1113	EXPECT_CH ('"');
		1114
		1115	// heuristic: assume that
		1116	// a) decode_str + hv_store_ent are abysmally slow.
		1117	// b) most hash keys are short, simple ascii text.
		1118	// => try to "fast-match" such strings to avoid
		1119	// the overhead of decode_str + hv_store_ent.
		1120	{
950	SV key, value;	1121	SV *value;
		1122	char *p = dec->cur;
		1123	char *e = p + 24; // only try up to 24 bytes
951		1124
952	decode_ws (dec); EXPECT_CH ('"');	1125	for (;;)
953
954	key = decode_str (dec);
955	if (!key)
956	goto fail;
957
958	decode_ws (dec); EXPECT_CH (':');
959
960	value = decode_sv (dec);
961	if (!value)
962	{	1126	{
		1127	// the >= 0x80 is true on most architectures
		1128	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1129	{
		1130	// slow path, back up and use decode_str
		1131	SV *key = decode_str (dec);
		1132	if (!key)
		1133	goto fail;
		1134
		1135	decode_ws (dec); EXPECT_CH (':');
		1136
		1137	decode_ws (dec);
		1138	value = decode_sv (dec);
		1139	if (!value)
		1140	{
		1141	SvREFCNT_dec (key);
		1142	goto fail;
		1143	}
		1144
		1145	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1146	SvREFCNT_dec (key);
		1147
		1148	break;
		1149	}
		1150	else if (*p == '"')
		1151	{
		1152	// fast path, got a simple key
		1153	char *key = dec->cur;
		1154	int len = p - key;
		1155	dec->cur = p + 1;
		1156
		1157	decode_ws (dec); EXPECT_CH (':');
		1158
		1159	decode_ws (dec);
		1160	value = decode_sv (dec);
		1161	if (!value)
964	goto fail;	1162	goto fail;
		1163
		1164	hv_store (hv, key, len, value, 0);
		1165
		1166	break;
		1167	}
		1168
		1169	++p;
965	}	1170	}
966		1171	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1172
970	decode_ws (dec);	1173	decode_ws (dec);
971		1174
972	if (*dec->cur == '}')	1175	if (*dec->cur == '}')
973	{	1176	{
…		…
977		1180
978	if (*dec->cur != ',')	1181	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1182	ERR (", or } expected while parsing object/hash");
980		1183
981	++dec->cur;	1184	++dec->cur;
		1185
		1186	decode_ws (dec);
		1187
		1188	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1189	{
		1190	++dec->cur;
		1191	break;
		1192	}
982	}	1193	}
983		1194
984	DEC_DEC_DEPTH;	1195	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1196	sv = newRV_noinc ((SV *)hv);
		1197
		1198	// check filter callbacks
		1199	if (dec->json.flags & F_HOOK)
		1200	{
		1201	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1202	{
		1203	HE cb, he;
		1204
		1205	hv_iterinit (hv);
		1206	he = hv_iternext (hv);
		1207	hv_iterinit (hv);
		1208
		1209	// the next line creates a mortal sv each time its called.
		1210	// might want to optimise this for common cases.
		1211	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1212
		1213	if (cb)
		1214	{
		1215	dSP;
		1216	int count;
		1217
		1218	ENTER; SAVETMPS; PUSHMARK (SP);
		1219	XPUSHs (HeVAL (he));
		1220
		1221	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1222
		1223	if (count == 1)
		1224	{
		1225	sv = newSVsv (POPs);
		1226	FREETMPS; LEAVE;
		1227	return sv;
		1228	}
		1229
		1230	FREETMPS; LEAVE;
		1231	}
		1232	}
		1233
		1234	if (dec->json.cb_object)
		1235	{
		1236	dSP;
		1237	int count;
		1238
		1239	ENTER; SAVETMPS; PUSHMARK (SP);
		1240	XPUSHs (sv_2mortal (sv));
		1241
		1242	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1243
		1244	if (count == 1)
		1245	{
		1246	sv = newSVsv (POPs);
		1247	FREETMPS; LEAVE;
		1248	return sv;
		1249	}
		1250
		1251	SvREFCNT_inc (sv);
		1252	FREETMPS; LEAVE;
		1253	}
		1254	}
		1255
		1256	return sv;
986		1257
987	fail:	1258	fail:
988	SvREFCNT_dec (hv);	1259	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1260	DEC_DEC_DEPTH;
990	return 0;	1261	return 0;
991	}	1262	}
992		1263
993	static SV *	1264	static SV *
994	decode_sv (dec_t *dec)	1265	decode_sv (dec_t *dec)
995	{	1266	{
996	decode_ws (dec);	1267	// the beauty of JSON: you need exactly one character lookahead
		1268	// to parse anything.
997	switch (*dec->cur)	1269	switch (*dec->cur)
998	{	1270	{
999	case '"': ++dec->cur; return decode_str (dec);	1271	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1272	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1273	case '{': ++dec->cur; return decode_hv (dec);
…		…
1007		1279
1008	case 't':	1280	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1281	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1282	{
1011	dec->cur += 4;	1283	dec->cur += 4;
1012	return newSViv (1);	1284	#if JSON_SLOW
		1285	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1286	#endif
		1287	return SvREFCNT_inc (json_true);
1013	}	1288	}
1014	else	1289	else
1015	ERR ("'true' expected");	1290	ERR ("'true' expected");
1016		1291
1017	break;	1292	break;
1018		1293
1019	case 'f':	1294	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1295	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1296	{
1022	dec->cur += 5;	1297	dec->cur += 5;
1023	return newSViv (0);	1298	#if JSON_SLOW
		1299	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1300	#endif
		1301	return SvREFCNT_inc (json_false);
1024	}	1302	}
1025	else	1303	else
1026	ERR ("'false' expected");	1304	ERR ("'false' expected");
1027		1305
1028	break;	1306	break;
…		…
1046	fail:	1324	fail:
1047	return 0;	1325	return 0;
1048	}	1326	}
1049		1327
1050	static SV *	1328	static SV *
1051	decode_json (SV string, U32 flags, UV offset_return)	1329	decode_json (SV string, JSON json, UV *offset_return)
1052	{	1330	{
1053	dec_t dec;	1331	dec_t dec;
1054	UV offset;	1332	UV offset;
1055	SV *sv;	1333	SV *sv;
1056		1334
1057	SvGETMAGIC (string);	1335	SvGETMAGIC (string);
1058	SvUPGRADE (string, SVt_PV);	1336	SvUPGRADE (string, SVt_PV);
1059		1337
		1338	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1339	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1340	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1341
1060	if (flags & F_UTF8)	1342	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1343	sv_utf8_downgrade (string, 0);
1062	else	1344	else
1063	sv_utf8_upgrade (string);	1345	sv_utf8_upgrade (string);
1064		1346
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1347	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1348
1067	dec.flags = flags;	1349	dec.json = *json;
1068	dec.cur = SvPVX (string);	1350	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1351	dec.end = SvEND (string);
1070	dec.err = 0;	1352	dec.err = 0;
1071	dec.depth = 0;	1353	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1354	dec.maxdepth = DEC_DEPTH (dec.json.flags);
		1355
		1356	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1357	dec.json.flags \|= F_HOOK;
1073		1358
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1359	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1360
		1361	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1362	sv = decode_sv (&dec);
1076		1363
1077	if (!(offset_return \|\| !sv))	1364	if (!(offset_return \|\| !sv))
1078	{	1365	{
1079	// check for trailing garbage	1366	// check for trailing garbage
…		…
1087	}	1374	}
1088	}	1375	}
1089		1376
1090	if (offset_return \|\| !sv)	1377	if (offset_return \|\| !sv)
1091	{	1378	{
1092	offset = dec.flags & F_UTF8	1379	offset = dec.json.flags & F_UTF8
1093	? dec.cur - SvPVX (string)	1380	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));	1381	: utf8_distance (dec.cur, SvPVX (string));
1095		1382
1096	if (offset_return)	1383	if (offset_return)
1097	*offset_return = offset;	1384	*offset_return = offset;
…		…
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1403	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1404	}
1118		1405
1119	sv = sv_2mortal (sv);	1406	sv = sv_2mortal (sv);
1120		1407
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1408	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1122	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1409	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1410
1124	return sv;	1411	return sv;
1125	}	1412	}
1126		1413
…		…
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1417	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1418
1132	BOOT:	1419	BOOT:
1133	{	1420	{
1134	int i;	1421	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1422
1138	for (i = 0; i < 256; ++i)	1423	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1424	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1425	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1426	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1427	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1428	: -1;
1144		1429
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1430	json_stash = gv_stashpv ("JSON::XS" , 1);
		1431	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1432
		1433	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1434	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1146	}	1435	}
1147		1436
1148	PROTOTYPES: DISABLE	1437	PROTOTYPES: DISABLE
1149		1438
1150	SV new (char dummy)	1439	void CLONE (...)
1151	CODE:	1440	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1441	json_stash = 0;
1153	OUTPUT:	1442	json_boolean_stash = 0;
1154	RETVAL
1155		1443
		1444	void new (char *klass)
		1445	PPCODE:
		1446	{
		1447	SV *pv = NEWSV (0, sizeof (JSON));
		1448	SvPOK_only (pv);
		1449	Zero (SvPVX (pv), 1, JSON);
		1450	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1451	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1452	}
		1453
1156	SV ascii (SV self, int enable = 1)	1454	void ascii (JSON *self, int enable = 1)
1157	ALIAS:	1455	ALIAS:
1158	ascii = F_ASCII	1456	ascii = F_ASCII
1159	latin1 = F_LATIN1	1457	latin1 = F_LATIN1
1160	utf8 = F_UTF8	1458	utf8 = F_UTF8
1161	indent = F_INDENT	1459	indent = F_INDENT
1162	canonical = F_CANONICAL	1460	canonical = F_CANONICAL
1163	space_before = F_SPACE_BEFORE	1461	space_before = F_SPACE_BEFORE
1164	space_after = F_SPACE_AFTER	1462	space_after = F_SPACE_AFTER
1165	pretty = F_PRETTY	1463	pretty = F_PRETTY
1166	allow_nonref = F_ALLOW_NONREF	1464	allow_nonref = F_ALLOW_NONREF
1167	shrink = F_SHRINK	1465	shrink = F_SHRINK
		1466	allow_blessed = F_ALLOW_BLESSED
		1467	convert_blessed = F_CONV_BLESSED
		1468	relaxed = F_RELAXED
1168	CODE:	1469	PPCODE:
1169	{	1470	{
1170	UV *uv = SvJSON (self);
1171	if (enable)	1471	if (enable)
1172	*uv \|= ix;	1472	self->flags \|= ix;
1173	else	1473	else
1174	*uv &= ~ix;	1474	self->flags &= ~ix;
1175		1475
1176	RETVAL = newSVsv (self);	1476	XPUSHs (ST (0));
1177	}	1477	}
1178	OUTPUT:
1179	RETVAL
1180		1478
1181	SV max_depth (SV self, UV max_depth = 0x80000000UL)	1479	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1182	CODE:	1480	PPCODE:
1183	{	1481	{
1184	UV *uv = SvJSON (self);
1185	UV log2 = 0;	1482	UV log2 = 0;
1186		1483
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1484	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1188		1485
1189	while ((1UL << log2) < max_depth)	1486	while ((1UL << log2) < max_depth)
1190	++log2;	1487	++log2;
1191		1488
1192	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1489	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1193		1490
1194	RETVAL = newSVsv (self);	1491	XPUSHs (ST (0));
1195	}	1492	}
1196	OUTPUT:
1197	RETVAL
1198		1493
1199	void encode (SV self, SV scalar)	1494	void max_size (JSON *self, UV max_size = 0)
1200	PPCODE:	1495	PPCODE:
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	1496	{
		1497	UV log2 = 0;
1202		1498
1203	void decode (SV self, SV jsonstr)	1499	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1500	if (max_size == 1) max_size = 2;
		1501
		1502	while ((1UL << log2) < max_size)
		1503	++log2;
		1504
		1505	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1506
		1507	XPUSHs (ST (0));
		1508	}
		1509
		1510	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1204	PPCODE:	1511	PPCODE:
		1512	{
		1513	SvREFCNT_dec (self->cb_object);
		1514	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1515
		1516	XPUSHs (ST (0));
		1517	}
		1518
		1519	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1520	PPCODE:
		1521	{
		1522	if (!self->cb_sk_object)
		1523	self->cb_sk_object = newHV ();
		1524
		1525	if (SvOK (cb))
		1526	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1527	else
		1528	{
		1529	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1530
		1531	if (!HvKEYS (self->cb_sk_object))
		1532	{
		1533	SvREFCNT_dec (self->cb_sk_object);
		1534	self->cb_sk_object = 0;
		1535	}
		1536	}
		1537
		1538	XPUSHs (ST (0));
		1539	}
		1540
		1541	void encode (JSON self, SV scalar)
		1542	PPCODE:
		1543	XPUSHs (encode_json (scalar, self));
		1544
		1545	void decode (JSON self, SV jsonstr)
		1546	PPCODE:
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1547	XPUSHs (decode_json (jsonstr, self, 0));
1206		1548
1207	void decode_prefix (SV self, SV jsonstr)	1549	void decode_prefix (JSON self, SV jsonstr)
1208	PPCODE:	1550	PPCODE:
1209	{	1551	{
1210	UV offset;	1552	UV offset;
1211	EXTEND (SP, 2);	1553	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	1554	PUSHs (decode_json (jsonstr, self, &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));	1555	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	1556	}
1215		1557
		1558	void DESTROY (JSON *self)
		1559	CODE:
		1560	SvREFCNT_dec (self->cb_sk_object);
		1561	SvREFCNT_dec (self->cb_object);
		1562
1216	PROTOTYPES: ENABLE	1563	PROTOTYPES: ENABLE
1217		1564
1218	void to_json (SV *scalar)	1565	void to_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:	1566	PPCODE:
		1567	{
		1568	JSON json = { F_DEFAULT \| F_UTF8 };
1222	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));	1569	XPUSHs (encode_json (scalar, &json));
		1570	}
1223		1571
1224	void from_json (SV *jsonstr)	1572	void from_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:	1573	PPCODE:
		1574	{
		1575	JSON json = { F_DEFAULT \| F_UTF8 };
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));	1576	XPUSHs (decode_json (jsonstr, &json, 0));
		1577	}
1229		1578

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.36 by root, Wed Jun 6 17:27:33 2007 UTC vs. Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.36 by root, Wed Jun 6 17:27:33 2007 UTC vs.
Revision 1.64 by root, Mon Aug 27 02:03:23 2007 UTC