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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.63 by root, Mon Aug 27 01:49:01 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	{
		616	// large integer, use the (rather slow) snprintf way.
511	need (enc, 32);	617	need (enc, sizeof (UV) * 3);
512	enc->cur +=	618	enc->cur +=
513	SvIsUV(sv)	619	SvIsUV(sv)
514	? snprintf (enc->cur, 32, "%"UVuf, (UV)SvUVX (sv))	620	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
515	: snprintf (enc->cur, 32, "%"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
…		…
599	char ch = *dec->cur;	709	char ch = *dec->cur;
600		710
601	if (ch > 0x20	711	if (ch > 0x20
602	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	712	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
603	break;	713	break;
		714
		715	if (ch == '#' && dec->json.flags & F_RELAXED)
		716	++dec->cur;
604		717
605	++dec->cur;	718	++dec->cur;
606	}	719	}
607	}	720	}
608		721
…		…
625	decode_4hex (dec_t *dec)	738	decode_4hex (dec_t *dec)
626	{	739	{
627	signed char d1, d2, d3, d4;	740	signed char d1, d2, d3, d4;
628	unsigned char cur = (unsigned char )dec->cur;	741	unsigned char cur = (unsigned char )dec->cur;
629		742
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	743	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");	744	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");	745	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");	746	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		747
635	dec->cur += 4;	748	dec->cur += 4;
636		749
637	return ((UV)d1) << 12	750	return ((UV)d1) << 12
638	\| ((UV)d2) << 8	751	\| ((UV)d2) << 8
…		…
646	static SV *	759	static SV *
647	decode_str (dec_t *dec)	760	decode_str (dec_t *dec)
648	{	761	{
649	SV *sv = 0;	762	SV *sv = 0;
650	int utf8 = 0;	763	int utf8 = 0;
		764	char *dec_cur = dec->cur;
651		765
652	do	766	do
653	{	767	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	768	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	769	char *cur = buf;
656		770
657	do	771	do
658	{	772	{
659	unsigned char ch = (unsigned char )dec->cur++;	773	unsigned char ch = (unsigned char )dec_cur++;
660		774
661	if (expect_false (ch == '"'))	775	if (expect_false (ch == '"'))
662	{	776	{
663	--dec->cur;	777	--dec_cur;
664	break;	778	break;
665	}	779	}
666	else if (expect_false (ch == '\\'))	780	else if (expect_false (ch == '\\'))
667	{	781	{
668	switch (*dec->cur)	782	switch (*dec_cur)
669	{	783	{
670	case '\\':	784	case '\\':
671	case '/':	785	case '/':
672	case '"': cur++ = dec->cur++; break;	786	case '"': cur++ = dec_cur++; break;
673		787
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	788	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	789	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	790	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	791	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	792	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		793
680	case 'u':	794	case 'u':
681	{	795	{
682	UV lo, hi;	796	UV lo, hi;
683	++dec->cur;	797	++dec_cur;
684		798
		799	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	800	hi = decode_4hex (dec);
		801	dec_cur = dec->cur;
686	if (hi == (UV)-1)	802	if (hi == (UV)-1)
687	goto fail;	803	goto fail;
688		804
689	// possibly a surrogate pair	805	// possibly a surrogate pair
690	if (hi >= 0xd800)	806	if (hi >= 0xd800)
691	if (hi < 0xdc00)	807	if (hi < 0xdc00)
692	{	808	{
693	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	809	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	810	ERR ("missing low surrogate character in surrogate pair");
695		811
696	dec->cur += 2;	812	dec_cur += 2;
697		813
		814	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	815	lo = decode_4hex (dec);
		816	dec_cur = dec->cur;
699	if (lo == (UV)-1)	817	if (lo == (UV)-1)
700	goto fail;	818	goto fail;
701		819
702	if (lo < 0xdc00 \|\| lo >= 0xe000)	820	if (lo < 0xdc00 \|\| lo >= 0xe000)
703	ERR ("surrogate pair expected");	821	ERR ("surrogate pair expected");
…		…
717	*cur++ = hi;	835	*cur++ = hi;
718	}	836	}
719	break;	837	break;
720		838
721	default:	839	default:
722	--dec->cur;	840	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	841	ERR ("illegal backslash escape sequence in string");
724	}	842	}
725	}	843	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	844	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
727	*cur++ = ch;	845	*cur++ = ch;
728	else if (ch >= 0x80)	846	else if (ch >= 0x80)
729	{	847	{
730	STRLEN clen;	848	STRLEN clen;
731	UV uch;	849	UV uch;
732		850
733	--dec->cur;	851	--dec_cur;
734		852
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	853	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	854	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	855	ERR ("malformed UTF-8 character in JSON string");
738		856
739	do	857	do
740	cur++ = dec->cur++;	858	cur++ = dec_cur++;
741	while (--clen);	859	while (--clen);
742		860
743	utf8 = 1;	861	utf8 = 1;
744	}	862	}
745	else	863	else
746	{	864	{
747	--dec->cur;	865	--dec_cur;
748		866
749	if (!ch)	867	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	868	ERR ("unexpected end of string while parsing JSON string");
751	else	869	else
752	ERR ("invalid character encountered while parsing JSON string");	870	ERR ("invalid character encountered while parsing JSON string");
…		…
765	}	883	}
766	else	884	else
767	sv = newSVpvn (buf, len);	885	sv = newSVpvn (buf, len);
768	}	886	}
769	}	887	}
770	while (*dec->cur != '"');	888	while (*dec_cur != '"');
771		889
772	++dec->cur;	890	++dec_cur;
773		891
774	if (sv)	892	if (sv)
775	{	893	{
776	SvPOK_only (sv);	894	SvPOK_only (sv);
777	*SvEND (sv) = 0;	895	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	898	SvUTF8_on (sv);
781	}	899	}
782	else	900	else
783	sv = newSVpvn ("", 0);	901	sv = newSVpvn ("", 0);
784		902
		903	dec->cur = dec_cur;
785	return sv;	904	return sv;
786		905
787	fail:	906	fail:
		907	dec->cur = dec_cur;
788	return 0;	908	return 0;
789	}	909	}
790		910
791	static SV *	911	static SV *
792	decode_num (dec_t *dec)	912	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	970	is_nv = 1;
851	}	971	}
852		972
853	if (!is_nv)	973	if (!is_nv)
854	{	974	{
		975	int len = dec->cur - start;
		976
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	977	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
856	if (*start == '-')	978	if (*start == '-')
857	switch (dec->cur - start)	979	switch (len)
858	{	980	{
859	case 2: return newSViv (-( start [1] - '0' ));	981	case 2: return newSViv (-( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	982	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));	983	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));	984	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
863	}	985	}
864	else	986	else
865	switch (dec->cur - start)	987	switch (len)
866	{	988	{
867	case 1: return newSViv ( start [0] - '0' );	989	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	990	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);	991	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);	992	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
871	}	993	}
872		994
873	{	995	{
874	UV uv;	996	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	997	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	998	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	999	if (numtype & IS_NUMBER_NEG)
878	{	1000	{
879	if (uv < (UV)IV_MIN)	1001	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1002	return newSViv (-(IV)uv);
881	}	1003	}
882	else	1004	else
883	return newSVuv (uv);	1005	return newSVuv (uv);
884	}	1006	}
885	}
886		1007
		1008	len -= *start == '-' ? 1 : 0;
		1009
		1010	// does not fit into IV or UV, try NV
		1011	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1012	#if defined (LDBL_DIG)
		1013	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1014	#endif
		1015	)
		1016	// fits into NV without loss of precision
		1017	return newSVnv (Atof (start));
		1018
		1019	// everything else fails, convert it to a string
		1020	return newSVpvn (start, dec->cur - start);
		1021	}
		1022
		1023	// loss of precision here
887	return newSVnv (Atof (start));	1024	return newSVnv (Atof (start));
888		1025
889	fail:	1026	fail:
890	return 0;	1027	return 0;
891	}	1028	}
…		…
921		1058
922	if (*dec->cur != ',')	1059	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1060	ERR (", or ] expected while parsing array");
924		1061
925	++dec->cur;	1062	++dec->cur;
		1063
		1064	decode_ws (dec);
		1065
		1066	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1067	{
		1068	++dec->cur;
		1069	break;
		1070	}
926	}	1071	}
927		1072
928	DEC_DEC_DEPTH;	1073	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1074	return newRV_noinc ((SV *)av);
930		1075
…		…
935	}	1080	}
936		1081
937	static SV *	1082	static SV *
938	decode_hv (dec_t *dec)	1083	decode_hv (dec_t *dec)
939	{	1084	{
		1085	SV *sv;
940	HV *hv = newHV ();	1086	HV *hv = newHV ();
941		1087
942	DEC_INC_DEPTH;	1088	DEC_INC_DEPTH;
943	decode_ws (dec);	1089	decode_ws (dec);
944		1090
945	if (*dec->cur == '}')	1091	if (*dec->cur == '}')
946	++dec->cur;	1092	++dec->cur;
947	else	1093	else
948	for (;;)	1094	for (;;)
949	{	1095	{
		1096	EXPECT_CH ('"');
		1097
		1098	// heuristic: assume that
		1099	// a) decode_str + hv_store_ent are abysmally slow.
		1100	// b) most hash keys are short, simple ascii text.
		1101	// => try to "fast-match" such strings to avoid
		1102	// the overhead of decode_str + hv_store_ent.
		1103	{
950	SV key, value;	1104	SV *value;
		1105	char *p = dec->cur;
		1106	char *e = p + 24; // only try up to 24 bytes
951		1107
952	decode_ws (dec); EXPECT_CH ('"');	1108	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	{	1109	{
		1110	// the >= 0x80 is true on most architectures
		1111	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1112	{
		1113	// slow path, back up and use decode_str
		1114	SV *key = decode_str (dec);
		1115	if (!key)
		1116	goto fail;
		1117
		1118	decode_ws (dec); EXPECT_CH (':');
		1119
		1120	decode_ws (dec);
		1121	value = decode_sv (dec);
		1122	if (!value)
		1123	{
		1124	SvREFCNT_dec (key);
		1125	goto fail;
		1126	}
		1127
		1128	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1129	SvREFCNT_dec (key);
		1130
		1131	break;
		1132	}
		1133	else if (*p == '"')
		1134	{
		1135	// fast path, got a simple key
		1136	char *key = dec->cur;
		1137	int len = p - key;
		1138	dec->cur = p + 1;
		1139
		1140	decode_ws (dec); EXPECT_CH (':');
		1141
		1142	decode_ws (dec);
		1143	value = decode_sv (dec);
		1144	if (!value)
964	goto fail;	1145	goto fail;
		1146
		1147	hv_store (hv, key, len, value, 0);
		1148
		1149	break;
		1150	}
		1151
		1152	++p;
965	}	1153	}
966		1154	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1155
970	decode_ws (dec);	1156	decode_ws (dec);
971		1157
972	if (*dec->cur == '}')	1158	if (*dec->cur == '}')
973	{	1159	{
…		…
977		1163
978	if (*dec->cur != ',')	1164	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1165	ERR (", or } expected while parsing object/hash");
980		1166
981	++dec->cur;	1167	++dec->cur;
		1168
		1169	decode_ws (dec);
		1170
		1171	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1172	{
		1173	++dec->cur;
		1174	break;
		1175	}
982	}	1176	}
983		1177
984	DEC_DEC_DEPTH;	1178	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1179	sv = newRV_noinc ((SV *)hv);
		1180
		1181	// check filter callbacks
		1182	if (dec->json.flags & F_HOOK)
		1183	{
		1184	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1185	{
		1186	HE cb, he;
		1187
		1188	hv_iterinit (hv);
		1189	he = hv_iternext (hv);
		1190	hv_iterinit (hv);
		1191
		1192	// the next line creates a mortal sv each time its called.
		1193	// might want to optimise this for common cases.
		1194	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1195
		1196	if (cb)
		1197	{
		1198	dSP;
		1199	int count;
		1200
		1201	ENTER; SAVETMPS; PUSHMARK (SP);
		1202	XPUSHs (HeVAL (he));
		1203
		1204	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1205
		1206	if (count == 1)
		1207	{
		1208	sv = newSVsv (POPs);
		1209	FREETMPS; LEAVE;
		1210	return sv;
		1211	}
		1212
		1213	FREETMPS; LEAVE;
		1214	}
		1215	}
		1216
		1217	if (dec->json.cb_object)
		1218	{
		1219	dSP;
		1220	int count;
		1221
		1222	ENTER; SAVETMPS; PUSHMARK (SP);
		1223	XPUSHs (sv_2mortal (sv));
		1224
		1225	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1226
		1227	if (count == 1)
		1228	{
		1229	sv = newSVsv (POPs);
		1230	FREETMPS; LEAVE;
		1231	return sv;
		1232	}
		1233
		1234	SvREFCNT_inc (sv);
		1235	FREETMPS; LEAVE;
		1236	}
		1237	}
		1238
		1239	return sv;
986		1240
987	fail:	1241	fail:
988	SvREFCNT_dec (hv);	1242	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1243	DEC_DEC_DEPTH;
990	return 0;	1244	return 0;
991	}	1245	}
992		1246
993	static SV *	1247	static SV *
994	decode_sv (dec_t *dec)	1248	decode_sv (dec_t *dec)
995	{	1249	{
996	decode_ws (dec);	1250	// the beauty of JSON: you need exactly one character lookahead
		1251	// to parse anything.
997	switch (*dec->cur)	1252	switch (*dec->cur)
998	{	1253	{
999	case '"': ++dec->cur; return decode_str (dec);	1254	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1255	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1256	case '{': ++dec->cur; return decode_hv (dec);
…		…
1007		1262
1008	case 't':	1263	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1264	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1265	{
1011	dec->cur += 4;	1266	dec->cur += 4;
1012	return newSViv (1);	1267	#if JSON_SLOW
		1268	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1269	#endif
		1270	return SvREFCNT_inc (json_true);
1013	}	1271	}
1014	else	1272	else
1015	ERR ("'true' expected");	1273	ERR ("'true' expected");
1016		1274
1017	break;	1275	break;
1018		1276
1019	case 'f':	1277	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1278	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1279	{
1022	dec->cur += 5;	1280	dec->cur += 5;
1023	return newSViv (0);	1281	#if JSON_SLOW
		1282	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1283	#endif
		1284	return SvREFCNT_inc (json_false);
1024	}	1285	}
1025	else	1286	else
1026	ERR ("'false' expected");	1287	ERR ("'false' expected");
1027		1288
1028	break;	1289	break;
…		…
1046	fail:	1307	fail:
1047	return 0;	1308	return 0;
1048	}	1309	}
1049		1310
1050	static SV *	1311	static SV *
1051	decode_json (SV string, U32 flags, UV offset_return)	1312	decode_json (SV string, JSON json, UV *offset_return)
1052	{	1313	{
1053	dec_t dec;	1314	dec_t dec;
1054	UV offset;	1315	UV offset;
1055	SV *sv;	1316	SV *sv;
1056		1317
1057	SvGETMAGIC (string);	1318	SvGETMAGIC (string);
1058	SvUPGRADE (string, SVt_PV);	1319	SvUPGRADE (string, SVt_PV);
1059		1320
		1321	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1322	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1323	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1324
1060	if (flags & F_UTF8)	1325	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1326	sv_utf8_downgrade (string, 0);
1062	else	1327	else
1063	sv_utf8_upgrade (string);	1328	sv_utf8_upgrade (string);
1064		1329
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1330	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1331
1067	dec.flags = flags;	1332	dec.json = *json;
1068	dec.cur = SvPVX (string);	1333	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1334	dec.end = SvEND (string);
1070	dec.err = 0;	1335	dec.err = 0;
1071	dec.depth = 0;	1336	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1337	dec.maxdepth = DEC_DEPTH (dec.json.flags);
		1338
		1339	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1340	dec.json.flags \|= F_HOOK;
1073		1341
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1342	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1343
		1344	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1345	sv = decode_sv (&dec);
1076		1346
1077	if (!(offset_return \|\| !sv))	1347	if (!(offset_return \|\| !sv))
1078	{	1348	{
1079	// check for trailing garbage	1349	// check for trailing garbage
…		…
1087	}	1357	}
1088	}	1358	}
1089		1359
1090	if (offset_return \|\| !sv)	1360	if (offset_return \|\| !sv)
1091	{	1361	{
1092	offset = dec.flags & F_UTF8	1362	offset = dec.json.flags & F_UTF8
1093	? dec.cur - SvPVX (string)	1363	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));	1364	: utf8_distance (dec.cur, SvPVX (string));
1095		1365
1096	if (offset_return)	1366	if (offset_return)
1097	*offset_return = offset;	1367	*offset_return = offset;
…		…
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1386	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1387	}
1118		1388
1119	sv = sv_2mortal (sv);	1389	sv = sv_2mortal (sv);
1120		1390
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1391	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)");	1392	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1393
1124	return sv;	1394	return sv;
1125	}	1395	}
1126		1396
…		…
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1400	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1401
1132	BOOT:	1402	BOOT:
1133	{	1403	{
1134	int i;	1404	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1405
1138	for (i = 0; i < 256; ++i)	1406	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1407	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1408	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1409	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1410	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1411	: -1;
1144		1412
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1413	json_stash = gv_stashpv ("JSON::XS" , 1);
		1414	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1415
		1416	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1417	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1146	}	1418	}
1147		1419
1148	PROTOTYPES: DISABLE	1420	PROTOTYPES: DISABLE
1149		1421
1150	SV new (char dummy)	1422	void CLONE (...)
1151	CODE:	1423	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1424	json_stash = 0;
1153	OUTPUT:	1425	json_boolean_stash = 0;
1154	RETVAL
1155		1426
		1427	void new (char *klass)
		1428	PPCODE:
		1429	{
		1430	SV *pv = NEWSV (0, sizeof (JSON));
		1431	SvPOK_only (pv);
		1432	Zero (SvPVX (pv), 1, JSON);
		1433	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1434	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1435	}
		1436
1156	SV ascii (SV self, int enable = 1)	1437	void ascii (JSON *self, int enable = 1)
1157	ALIAS:	1438	ALIAS:
1158	ascii = F_ASCII	1439	ascii = F_ASCII
1159	latin1 = F_LATIN1	1440	latin1 = F_LATIN1
1160	utf8 = F_UTF8	1441	utf8 = F_UTF8
1161	indent = F_INDENT	1442	indent = F_INDENT
1162	canonical = F_CANONICAL	1443	canonical = F_CANONICAL
1163	space_before = F_SPACE_BEFORE	1444	space_before = F_SPACE_BEFORE
1164	space_after = F_SPACE_AFTER	1445	space_after = F_SPACE_AFTER
1165	pretty = F_PRETTY	1446	pretty = F_PRETTY
1166	allow_nonref = F_ALLOW_NONREF	1447	allow_nonref = F_ALLOW_NONREF
1167	shrink = F_SHRINK	1448	shrink = F_SHRINK
		1449	allow_blessed = F_ALLOW_BLESSED
		1450	convert_blessed = F_CONV_BLESSED
		1451	relaxed = F_RELAXED
1168	CODE:	1452	PPCODE:
1169	{	1453	{
1170	UV *uv = SvJSON (self);
1171	if (enable)	1454	if (enable)
1172	*uv \|= ix;	1455	self->flags \|= ix;
1173	else	1456	else
1174	*uv &= ~ix;	1457	self->flags &= ~ix;
1175		1458
1176	RETVAL = newSVsv (self);	1459	XPUSHs (ST (0));
1177	}	1460	}
1178	OUTPUT:
1179	RETVAL
1180		1461
1181	SV max_depth (SV self, UV max_depth = 0x80000000UL)	1462	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1182	CODE:	1463	PPCODE:
1183	{	1464	{
1184	UV *uv = SvJSON (self);
1185	UV log2 = 0;	1465	UV log2 = 0;
1186		1466
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1467	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1188		1468
1189	while ((1UL << log2) < max_depth)	1469	while ((1UL << log2) < max_depth)
1190	++log2;	1470	++log2;
1191		1471
1192	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1472	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1193		1473
1194	RETVAL = newSVsv (self);	1474	XPUSHs (ST (0));
1195	}	1475	}
1196	OUTPUT:
1197	RETVAL
1198		1476
1199	void encode (SV self, SV scalar)	1477	void max_size (JSON *self, UV max_size = 0)
1200	PPCODE:	1478	PPCODE:
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	1479	{
		1480	UV log2 = 0;
1202		1481
1203	void decode (SV self, SV jsonstr)	1482	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1483	if (max_size == 1) max_size = 2;
		1484
		1485	while ((1UL << log2) < max_size)
		1486	++log2;
		1487
		1488	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1489
		1490	XPUSHs (ST (0));
		1491	}
		1492
		1493	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1204	PPCODE:	1494	PPCODE:
		1495	{
		1496	SvREFCNT_dec (self->cb_object);
		1497	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1498
		1499	XPUSHs (ST (0));
		1500	}
		1501
		1502	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1503	PPCODE:
		1504	{
		1505	if (!self->cb_sk_object)
		1506	self->cb_sk_object = newHV ();
		1507
		1508	if (SvOK (cb))
		1509	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1510	else
		1511	{
		1512	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1513
		1514	if (!HvKEYS (self->cb_sk_object))
		1515	{
		1516	SvREFCNT_dec (self->cb_sk_object);
		1517	self->cb_sk_object = 0;
		1518	}
		1519	}
		1520
		1521	XPUSHs (ST (0));
		1522	}
		1523
		1524	void encode (JSON self, SV scalar)
		1525	PPCODE:
		1526	XPUSHs (encode_json (scalar, self));
		1527
		1528	void decode (JSON self, SV jsonstr)
		1529	PPCODE:
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1530	XPUSHs (decode_json (jsonstr, self, 0));
1206		1531
1207	void decode_prefix (SV self, SV jsonstr)	1532	void decode_prefix (JSON self, SV jsonstr)
1208	PPCODE:	1533	PPCODE:
1209	{	1534	{
1210	UV offset;	1535	UV offset;
1211	EXTEND (SP, 2);	1536	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	1537	PUSHs (decode_json (jsonstr, self, &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));	1538	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	1539	}
1215		1540
		1541	void DESTROY (JSON *self)
		1542	CODE:
		1543	SvREFCNT_dec (self->cb_sk_object);
		1544	SvREFCNT_dec (self->cb_object);
		1545
1216	PROTOTYPES: ENABLE	1546	PROTOTYPES: ENABLE
1217		1547
1218	void to_json (SV *scalar)	1548	void to_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:	1549	PPCODE:
		1550	{
		1551	JSON json = { F_DEFAULT \| F_UTF8 };
1222	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));	1552	XPUSHs (encode_json (scalar, &json));
		1553	}
1223		1554
1224	void from_json (SV *jsonstr)	1555	void from_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:	1556	PPCODE:
		1557	{
		1558	JSON json = { F_DEFAULT \| F_UTF8 };
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));	1559	XPUSHs (decode_json (jsonstr, &json, 0));
		1560	}
1229		1561

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs. Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.35 by root, Wed Jun 6 14:52:49 2007 UTC vs.
Revision 1.63 by root, Mon Aug 27 01:49:01 2007 UTC