[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 1.68 by root, Thu Nov 29 08:37:11 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
…		…
295	int i, len = av_len (av);	312	int i, len = av_len (av);
296		313
297	if (enc->indent >= enc->maxdepth)	314	if (enc->indent >= enc->maxdepth)
298	croak ("data structure too deep (hit recursion limit)");	315	croak ("data structure too deep (hit recursion limit)");
299		316
300	encode_ch (enc, '['); encode_nl (enc);	317	encode_ch (enc, '[');
301	++enc->indent;	318
		319	if (len >= 0)
		320	{
		321	encode_nl (enc); ++enc->indent;
302		322
303	for (i = 0; i <= len; ++i)	323	for (i = 0; i <= len; ++i)
304	{	324	{
		325	SV **svp = av_fetch (av, i, 0);
		326
305	encode_indent (enc);	327	encode_indent (enc);
306	encode_sv (enc, *av_fetch (av, i, 0));
307		328
		329	if (svp)
		330	encode_sv (enc, *svp);
		331	else
		332	encode_str (enc, "null", 4, 0);
		333
308	if (i < len)	334	if (i < len)
309	encode_comma (enc);	335	encode_comma (enc);
310	}	336	}
311		337
		338	encode_nl (enc); --enc->indent; encode_indent (enc);
		339	}
		340
312	encode_nl (enc);	341	encode_ch (enc, ']');
313
314	--enc->indent;
315	encode_indent (enc); encode_ch (enc, ']');
316	}	342	}
317		343
318	static void	344	static void
319	encode_he (enc_t enc, HE he)	345	encode_hk (enc_t enc, HE he)
320	{	346	{
321	encode_ch (enc, '"');	347	encode_ch (enc, '"');
322		348
323	if (HeKLEN (he) == HEf_SVKEY)	349	if (HeKLEN (he) == HEf_SVKEY)
324	{	350	{
…		…
334	else	360	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	361	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		362
337	encode_ch (enc, '"');	363	encode_ch (enc, '"');
338		364
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	365	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	366	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	367	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));
343	}	368	}
344		369
345	// compare hash entries, used when all keys are bytestrings	370	// compare hash entries, used when all keys are bytestrings
346	static int	371	static int
347	he_cmp_fast (const void a_, const void b_)	372	he_cmp_fast (const void a_, const void b_)
…		…
352	HE b = (HE **)b_;	377	HE b = (HE **)b_;
353		378
354	STRLEN la = HeKLEN (a);	379	STRLEN la = HeKLEN (a);
355	STRLEN lb = HeKLEN (b);	380	STRLEN lb = HeKLEN (b);
356		381
357	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	382	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
358	cmp = la - lb;	383	cmp = lb - la;
359		384
360	return cmp;	385	return cmp;
361	}	386	}
362		387
363	// compare hash entries, used when some keys are sv's or utf-x	388	// compare hash entries, used when some keys are sv's or utf-x
364	static int	389	static int
365	he_cmp_slow (const void a, const void b)	390	he_cmp_slow (const void a, const void b)
366	{	391	{
367	return sv_cmp (HeSVKEY_force ((HE )a), HeSVKEY_force ((HE **)b));	392	return sv_cmp (HeSVKEY_force ((HE )b), HeSVKEY_force ((HE **)a));
368	}	393	}
369		394
370	static void	395	static void
371	encode_hv (enc_t enc, HV hv)	396	encode_hv (enc_t enc, HV hv)
372	{	397	{
		398	HE *he;
373	int count, i;	399	int count;
374		400
375	if (enc->indent >= enc->maxdepth)	401	if (enc->indent >= enc->maxdepth)
376	croak ("data structure too deep (hit recursion limit)");	402	croak ("data structure too deep (hit recursion limit)");
377		403
378	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	404	encode_ch (enc, '{');
379		405
380	if ((count = hv_iterinit (hv)))
381	{
382	// for canonical output we have to sort by keys first	406	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	407	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	408	// security workaround added somewhere in 5.8.x.
385	// that randomises hash orderings	409	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	410	if (enc->json.flags & F_CANONICAL)
		411	{
		412	int count = hv_iterinit (hv);
		413
		414	if (SvMAGICAL (hv))
387	{	415	{
		416	// need to count by iterating. could improve by dynamically building the vector below
		417	// but I don't care for the speed of this special case.
		418	// note also that we will run into undefined behaviour when the two iterations
		419	// do not result in the same count, something I might care for in some later release.
		420
		421	count = 0;
		422	while (hv_iternext (hv))
		423	++count;
		424
		425	hv_iterinit (hv);
		426	}
		427
		428	if (count)
		429	{
388	int fast = 1;	430	int i, fast = 1;
389	HE *he;
390	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	431	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
391	HE *hes = _alloca (count sizeof (HE));	432	HE *hes = _alloca (count sizeof (HE));
392	#else	433	#else
393	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode	434	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
394	#endif	435	#endif
…		…
421		462
422	FREETMPS;	463	FREETMPS;
423	LEAVE;	464	LEAVE;
424	}	465	}
425		466
426	for (i = 0; i < count; ++i)	467	encode_nl (enc); ++enc->indent;
		468
		469	while (count--)
427	{	470	{
428	encode_indent (enc);	471	encode_indent (enc);
		472	he = hes [count];
429	encode_he (enc, hes [i]);	473	encode_hk (enc, he);
		474	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
430		475
431	if (i < count - 1)	476	if (count)
432	encode_comma (enc);	477	encode_comma (enc);
433	}	478	}
434		479
		480	encode_nl (enc); --enc->indent; encode_indent (enc);
		481	}
		482	}
		483	else
		484	{
		485	if (hv_iterinit (hv) \|\| SvMAGICAL (hv))
		486	if ((he = hv_iternext (hv)))
		487	{
		488	encode_nl (enc); ++enc->indent;
		489
		490	for (;;)
		491	{
435	encode_nl (enc);	492	encode_indent (enc);
		493	encode_hk (enc, he);
		494	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		495
		496	if (!(he = hv_iternext (hv)))
		497	break;
		498
		499	encode_comma (enc);
		500	}
		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		503	}
		504	}
		505
		506	encode_ch (enc, '}');
		507	}
		508
		509	// encode objects, arrays and special \0=false and \1=true values.
		510	static void
		511	encode_rv (enc_t enc, SV sv)
		512	{
		513	svtype svt;
		514
		515	SvGETMAGIC (sv);
		516	svt = SvTYPE (sv);
		517
		518	if (expect_false (SvOBJECT (sv)))
		519	{
		520	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		521	? json_boolean_stash
		522	: gv_stashpv ("JSON::XS::Boolean", 1);
		523
		524	if (SvSTASH (sv) == stash)
		525	{
		526	if (SvIV (sv))
		527	encode_str (enc, "true", 4, 0);
		528	else
		529	encode_str (enc, "false", 5, 0);
436	}	530	}
437	else	531	else
438	{	532	{
439	HE *he = hv_iternext (hv);	533	#if 0
440		534	if (0 && sv_derived_from (rv, "JSON::Literal"))
441	for (;;)
442	{	535	{
443	encode_indent (enc);	536	// not yet
444	encode_he (enc, he);
445
446	if (!(he = hv_iternext (hv)))
447	break;
448
449	encode_comma (enc);
450	}	537	}
		538	#endif
		539	if (enc->json.flags & F_CONV_BLESSED)
		540	{
		541	// we re-bless the reference to get overload and other niceties right
		542	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
451		543
		544	if (to_json)
		545	{
		546	dSP;
		547
		548	ENTER; SAVETMPS; PUSHMARK (SP);
		549	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		550
		551	// calling with G_SCALAR ensures that we always get a 1 return value
		552	PUTBACK;
		553	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		554	SPAGAIN;
		555
		556	// catch this surprisingly common error
		557	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		558	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		559
		560	sv = POPs;
		561	PUTBACK;
		562
452	encode_nl (enc);	563	encode_sv (enc, sv);
		564
		565	FREETMPS; LEAVE;
		566	}
		567	else if (enc->json.flags & F_ALLOW_BLESSED)
		568	encode_str (enc, "null", 4, 0);
		569	else
		570	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		571	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		572	}
		573	else if (enc->json.flags & F_ALLOW_BLESSED)
		574	encode_str (enc, "null", 4, 0);
		575	else
		576	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		577	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
453	}	578	}
454	}	579	}
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)	580	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	581	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	582	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	583	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	584	else if (svt < SVt_PVAV)
473	{	585	{
474	if (SvNIOK (sv) && SvIV (sv) == 0)	586	STRLEN len = 0;
		587	char *pv = svt ? SvPV (sv, len) : 0;
		588
		589	if (len == 1 && *pv == '1')
		590	encode_str (enc, "true", 4, 0);
		591	else if (len == 1 && *pv == '0')
475	encode_str (enc, "false", 5, 0);	592	encode_str (enc, "false", 5, 0);
476	else if (SvNIOK (sv) && SvIV (sv) == 1)
477	encode_str (enc, "true", 4, 0);
478	else	593	else
479	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
480	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
481	}	596	}
482	else	597	else
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	612	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	613	encode_ch (enc, '"');
499	}	614	}
500	else if (SvNOKp (sv))	615	else if (SvNOKp (sv))
501	{	616	{
		617	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	618	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	620	enc->cur += strlen (enc->cur);
505	}	621	}
506	else if (SvIOKp (sv))	622	else if (SvIOKp (sv))
507	{	623	{
508	// we assume we can always read an IV as a UV	624	// we assume we can always read an IV as a UV
509	if (SvUV (sv) & ~(UV)0x7fff)	625	if (SvUV (sv) & ~(UV)0x7fff)
510	{	626	{
511	need (enc, 64);	627	// large integer, use the (rather slow) snprintf way.
		628	need (enc, sizeof (UV) * 3);
512	enc->cur +=	629	enc->cur +=
513	SvIsUV(sv)	630	SvIsUV(sv)
514	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
515	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
516	}	633	}
517	else	634	else
518	{	635	{
519	// optimise the "small number case"	636	// optimise the "small number case"
520	// code will likely be branchless and use only a single multiplication	637	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);	638	I32 i = SvIV (sv);
522	U32 u;	639	U32 u;
		640	char digit, nz = 0;
523		641
524	need (enc, 6);	642	need (enc, 6);
525		643
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	644	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
527	u = i < 0 ? -i : i;	645	u = i < 0 ? -i : i;
528		646
529	// convert to 4.28 fixed-point representation	647	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits	648	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531		649
532	char digit, nz = 0;	650	// now output digit by digit, each time masking out the integer part
533		651	// and multiplying by 5 while moving the decimal point one to the right,
		652	// resulting in a net multiplication by 10.
		653	// we always write the digit to memory but conditionally increment
		654	// 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;	655	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;	656	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;	657	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;	658	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;	659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
539	}	660	}
540	}	661	}
541	else if (SvROK (sv))	662	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	663	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	664	else if (!SvOK (sv))
…		…
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	668	SvPV_nolen (sv), SvFLAGS (sv));
548	}	669	}
549		670
550	static SV *	671	static SV *
551	encode_json (SV *scalar, U32 flags)	672	encode_json (SV scalar, JSON json)
552	{	673	{
553	enc_t enc;	674	enc_t enc;
554		675
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	676	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	677	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		678
558	enc.flags = flags;	679	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	681	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	682	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	683	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	684	enc.maxdepth = DEC_DEPTH (enc.json.flags);
564		685
565	SvPOK_only (enc.sv);	686	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	687	encode_sv (&enc, scalar);
567		688
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		691
571	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	692	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
572	SvUTF8_on (enc.sv);	693	SvUTF8_on (enc.sv);
573		694
574	if (enc.flags & F_SHRINK)	695	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	696	shrink (enc.sv);
576		697
577	return enc.sv;	698	return enc.sv;
578	}	699	}
579		700
…		…
584	typedef struct	705	typedef struct
585	{	706	{
586	char *cur; // current parser pointer	707	char *cur; // current parser pointer
587	char *end; // end of input string	708	char *end; // end of input string
588	const char *err; // parse error, if != 0	709	const char *err; // parse error, if != 0
589	U32 flags; // F_*	710	JSON json;
590	U32 depth; // recursion depth	711	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	712	U32 maxdepth; // recursion depth limit
592	} dec_t;	713	} dec_t;
593		714
594	inline void	715	inline void
		716	decode_comment (dec_t *dec)
		717	{
		718	// only '#'-style comments allowed a.t.m.
		719
		720	while (dec->cur && dec->cur != 0x0a && *dec->cur != 0x0d)
		721	++dec->cur;
		722	}
		723
		724	inline void
595	decode_ws (dec_t *dec)	725	decode_ws (dec_t *dec)
596	{	726	{
597	for (;;)	727	for (;;)
598	{	728	{
599	char ch = *dec->cur;	729	char ch = *dec->cur;
600		730
601	if (ch > 0x20	731	if (ch > 0x20)
		732	{
		733	if (expect_false (ch == '#'))
		734	{
		735	if (dec->json.flags & F_RELAXED)
		736	decode_comment (dec);
		737	else
		738	break;
		739	}
		740	else
		741	break;
		742	}
602	\|\| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	743	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
603	break;	744	break; // parse error, but let higher level handle it, gives better error messages
604		745
605	++dec->cur;	746	++dec->cur;
606	}	747	}
607	}	748	}
608		749
…		…
625	decode_4hex (dec_t *dec)	766	decode_4hex (dec_t *dec)
626	{	767	{
627	signed char d1, d2, d3, d4;	768	signed char d1, d2, d3, d4;
628	unsigned char cur = (unsigned char )dec->cur;	769	unsigned char cur = (unsigned char )dec->cur;
629		770
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	771	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");	772	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");	773	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");	774	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		775
635	dec->cur += 4;	776	dec->cur += 4;
636		777
637	return ((UV)d1) << 12	778	return ((UV)d1) << 12
638	\| ((UV)d2) << 8	779	\| ((UV)d2) << 8
…		…
646	static SV *	787	static SV *
647	decode_str (dec_t *dec)	788	decode_str (dec_t *dec)
648	{	789	{
649	SV *sv = 0;	790	SV *sv = 0;
650	int utf8 = 0;	791	int utf8 = 0;
		792	char *dec_cur = dec->cur;
651		793
652	do	794	do
653	{	795	{
654	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];	796	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
655	char *cur = buf;	797	char *cur = buf;
656		798
657	do	799	do
658	{	800	{
659	unsigned char ch = (unsigned char )dec->cur++;	801	unsigned char ch = (unsigned char )dec_cur++;
660		802
661	if (expect_false (ch == '"'))	803	if (expect_false (ch == '"'))
662	{	804	{
663	--dec->cur;	805	--dec_cur;
664	break;	806	break;
665	}	807	}
666	else if (expect_false (ch == '\\'))	808	else if (expect_false (ch == '\\'))
667	{	809	{
668	switch (*dec->cur)	810	switch (*dec_cur)
669	{	811	{
670	case '\\':	812	case '\\':
671	case '/':	813	case '/':
672	case '"': cur++ = dec->cur++; break;	814	case '"': cur++ = dec_cur++; break;
673		815
674	case 'b': ++dec->cur; *cur++ = '\010'; break;	816	case 'b': ++dec_cur; *cur++ = '\010'; break;
675	case 't': ++dec->cur; *cur++ = '\011'; break;	817	case 't': ++dec_cur; *cur++ = '\011'; break;
676	case 'n': ++dec->cur; *cur++ = '\012'; break;	818	case 'n': ++dec_cur; *cur++ = '\012'; break;
677	case 'f': ++dec->cur; *cur++ = '\014'; break;	819	case 'f': ++dec_cur; *cur++ = '\014'; break;
678	case 'r': ++dec->cur; *cur++ = '\015'; break;	820	case 'r': ++dec_cur; *cur++ = '\015'; break;
679		821
680	case 'u':	822	case 'u':
681	{	823	{
682	UV lo, hi;	824	UV lo, hi;
683	++dec->cur;	825	++dec_cur;
684		826
		827	dec->cur = dec_cur;
685	hi = decode_4hex (dec);	828	hi = decode_4hex (dec);
		829	dec_cur = dec->cur;
686	if (hi == (UV)-1)	830	if (hi == (UV)-1)
687	goto fail;	831	goto fail;
688		832
689	// possibly a surrogate pair	833	// possibly a surrogate pair
690	if (hi >= 0xd800)	834	if (hi >= 0xd800)
691	if (hi < 0xdc00)	835	if (hi < 0xdc00)
692	{	836	{
693	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	837	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
694	ERR ("missing low surrogate character in surrogate pair");	838	ERR ("missing low surrogate character in surrogate pair");
695		839
696	dec->cur += 2;	840	dec_cur += 2;
697		841
		842	dec->cur = dec_cur;
698	lo = decode_4hex (dec);	843	lo = decode_4hex (dec);
		844	dec_cur = dec->cur;
699	if (lo == (UV)-1)	845	if (lo == (UV)-1)
700	goto fail;	846	goto fail;
701		847
702	if (lo < 0xdc00 \|\| lo >= 0xe000)	848	if (lo < 0xdc00 \|\| lo >= 0xe000)
703	ERR ("surrogate pair expected");	849	ERR ("surrogate pair expected");
…		…
717	*cur++ = hi;	863	*cur++ = hi;
718	}	864	}
719	break;	865	break;
720		866
721	default:	867	default:
722	--dec->cur;	868	--dec_cur;
723	ERR ("illegal backslash escape sequence in string");	869	ERR ("illegal backslash escape sequence in string");
724	}	870	}
725	}	871	}
726	else if (expect_true (ch >= 0x20 && ch <= 0x7f))	872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
727	*cur++ = ch;	873	*cur++ = ch;
728	else if (ch >= 0x80)	874	else if (ch >= 0x80)
729	{	875	{
730	STRLEN clen;	876	STRLEN clen;
731	UV uch;	877	UV uch;
732		878
733	--dec->cur;	879	--dec_cur;
734		880
735	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
736	if (clen == (STRLEN)-1)	882	if (clen == (STRLEN)-1)
737	ERR ("malformed UTF-8 character in JSON string");	883	ERR ("malformed UTF-8 character in JSON string");
738		884
739	do	885	do
740	cur++ = dec->cur++;	886	cur++ = dec_cur++;
741	while (--clen);	887	while (--clen);
742		888
743	utf8 = 1;	889	utf8 = 1;
744	}	890	}
745	else	891	else
746	{	892	{
747	--dec->cur;	893	--dec_cur;
748		894
749	if (!ch)	895	if (!ch)
750	ERR ("unexpected end of string while parsing JSON string");	896	ERR ("unexpected end of string while parsing JSON string");
751	else	897	else
752	ERR ("invalid character encountered while parsing JSON string");	898	ERR ("invalid character encountered while parsing JSON string");
…		…
765	}	911	}
766	else	912	else
767	sv = newSVpvn (buf, len);	913	sv = newSVpvn (buf, len);
768	}	914	}
769	}	915	}
770	while (*dec->cur != '"');	916	while (*dec_cur != '"');
771		917
772	++dec->cur;	918	++dec_cur;
773		919
774	if (sv)	920	if (sv)
775	{	921	{
776	SvPOK_only (sv);	922	SvPOK_only (sv);
777	*SvEND (sv) = 0;	923	*SvEND (sv) = 0;
…		…
780	SvUTF8_on (sv);	926	SvUTF8_on (sv);
781	}	927	}
782	else	928	else
783	sv = newSVpvn ("", 0);	929	sv = newSVpvn ("", 0);
784		930
		931	dec->cur = dec_cur;
785	return sv;	932	return sv;
786		933
787	fail:	934	fail:
		935	dec->cur = dec_cur;
788	return 0;	936	return 0;
789	}	937	}
790		938
791	static SV *	939	static SV *
792	decode_num (dec_t *dec)	940	decode_num (dec_t *dec)
…		…
850	is_nv = 1;	998	is_nv = 1;
851	}	999	}
852		1000
853	if (!is_nv)	1001	if (!is_nv)
854	{	1002	{
		1003	int len = dec->cur - start;
		1004
855	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
856	if (*start == '-')	1006	if (*start == '-')
857	switch (dec->cur - start)	1007	switch (len)
858	{	1008	{
859	case 2: return newSViv (-( start [1] - '0' ));	1009	case 2: return newSViv (-( start [1] - '0' * 1));
860	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	1010	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));	1011	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));	1012	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
863	}	1013	}
864	else	1014	else
865	switch (dec->cur - start)	1015	switch (len)
866	{	1016	{
867	case 1: return newSViv ( start [0] - '0' );	1017	case 1: return newSViv ( start [0] - '0' * 1);
868	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	1018	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);	1019	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);	1020	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
871	}	1021	}
872		1022
873	{	1023	{
874	UV uv;	1024	UV uv;
875	int numtype = grok_number (start, dec->cur - start, &uv);	1025	int numtype = grok_number (start, len, &uv);
876	if (numtype & IS_NUMBER_IN_UV)	1026	if (numtype & IS_NUMBER_IN_UV)
877	if (numtype & IS_NUMBER_NEG)	1027	if (numtype & IS_NUMBER_NEG)
878	{	1028	{
879	if (uv < (UV)IV_MIN)	1029	if (uv < (UV)IV_MIN)
880	return newSViv (-(IV)uv);	1030	return newSViv (-(IV)uv);
881	}	1031	}
882	else	1032	else
883	return newSVuv (uv);	1033	return newSVuv (uv);
884	}	1034	}
885	}
886		1035
		1036	len -= *start == '-' ? 1 : 0;
		1037
		1038	// does not fit into IV or UV, try NV
		1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1040	#if defined (LDBL_DIG)
		1041	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1042	#endif
		1043	)
		1044	// fits into NV without loss of precision
		1045	return newSVnv (Atof (start));
		1046
		1047	// everything else fails, convert it to a string
		1048	return newSVpvn (start, dec->cur - start);
		1049	}
		1050
		1051	// loss of precision here
887	return newSVnv (Atof (start));	1052	return newSVnv (Atof (start));
888		1053
889	fail:	1054	fail:
890	return 0;	1055	return 0;
891	}	1056	}
…		…
921		1086
922	if (*dec->cur != ',')	1087	if (*dec->cur != ',')
923	ERR (", or ] expected while parsing array");	1088	ERR (", or ] expected while parsing array");
924		1089
925	++dec->cur;	1090	++dec->cur;
		1091
		1092	decode_ws (dec);
		1093
		1094	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1095	{
		1096	++dec->cur;
		1097	break;
		1098	}
926	}	1099	}
927		1100
928	DEC_DEC_DEPTH;	1101	DEC_DEC_DEPTH;
929	return newRV_noinc ((SV *)av);	1102	return newRV_noinc ((SV *)av);
930		1103
…		…
935	}	1108	}
936		1109
937	static SV *	1110	static SV *
938	decode_hv (dec_t *dec)	1111	decode_hv (dec_t *dec)
939	{	1112	{
		1113	SV *sv;
940	HV *hv = newHV ();	1114	HV *hv = newHV ();
941		1115
942	DEC_INC_DEPTH;	1116	DEC_INC_DEPTH;
943	decode_ws (dec);	1117	decode_ws (dec);
944		1118
945	if (*dec->cur == '}')	1119	if (*dec->cur == '}')
946	++dec->cur;	1120	++dec->cur;
947	else	1121	else
948	for (;;)	1122	for (;;)
949	{	1123	{
		1124	EXPECT_CH ('"');
		1125
		1126	// heuristic: assume that
		1127	// a) decode_str + hv_store_ent are abysmally slow.
		1128	// b) most hash keys are short, simple ascii text.
		1129	// => try to "fast-match" such strings to avoid
		1130	// the overhead of decode_str + hv_store_ent.
		1131	{
950	SV key, value;	1132	SV *value;
		1133	char *p = dec->cur;
		1134	char *e = p + 24; // only try up to 24 bytes
951		1135
952	decode_ws (dec); EXPECT_CH ('"');	1136	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	{	1137	{
		1138	// the >= 0x80 is true on most architectures
		1139	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1140	{
		1141	// slow path, back up and use decode_str
		1142	SV *key = decode_str (dec);
		1143	if (!key)
		1144	goto fail;
		1145
		1146	decode_ws (dec); EXPECT_CH (':');
		1147
		1148	decode_ws (dec);
		1149	value = decode_sv (dec);
		1150	if (!value)
		1151	{
		1152	SvREFCNT_dec (key);
		1153	goto fail;
		1154	}
		1155
		1156	hv_store_ent (hv, key, value, 0);
963	SvREFCNT_dec (key);	1157	SvREFCNT_dec (key);
		1158
		1159	break;
		1160	}
		1161	else if (*p == '"')
		1162	{
		1163	// fast path, got a simple key
		1164	char *key = dec->cur;
		1165	int len = p - key;
		1166	dec->cur = p + 1;
		1167
		1168	decode_ws (dec); EXPECT_CH (':');
		1169
		1170	decode_ws (dec);
		1171	value = decode_sv (dec);
		1172	if (!value)
964	goto fail;	1173	goto fail;
		1174
		1175	hv_store (hv, key, len, value, 0);
		1176
		1177	break;
		1178	}
		1179
		1180	++p;
965	}	1181	}
966		1182	}
967	hv_store_ent (hv, key, value, 0);
968	SvREFCNT_dec (key);
969		1183
970	decode_ws (dec);	1184	decode_ws (dec);
971		1185
972	if (*dec->cur == '}')	1186	if (*dec->cur == '}')
973	{	1187	{
…		…
977		1191
978	if (*dec->cur != ',')	1192	if (*dec->cur != ',')
979	ERR (", or } expected while parsing object/hash");	1193	ERR (", or } expected while parsing object/hash");
980		1194
981	++dec->cur;	1195	++dec->cur;
		1196
		1197	decode_ws (dec);
		1198
		1199	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1200	{
		1201	++dec->cur;
		1202	break;
		1203	}
982	}	1204	}
983		1205
984	DEC_DEC_DEPTH;	1206	DEC_DEC_DEPTH;
985	return newRV_noinc ((SV *)hv);	1207	sv = newRV_noinc ((SV *)hv);
		1208
		1209	// check filter callbacks
		1210	if (dec->json.flags & F_HOOK)
		1211	{
		1212	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1213	{
		1214	HE cb, he;
		1215
		1216	hv_iterinit (hv);
		1217	he = hv_iternext (hv);
		1218	hv_iterinit (hv);
		1219
		1220	// the next line creates a mortal sv each time its called.
		1221	// might want to optimise this for common cases.
		1222	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1223
		1224	if (cb)
		1225	{
		1226	dSP;
		1227	int count;
		1228
		1229	ENTER; SAVETMPS; PUSHMARK (SP);
		1230	XPUSHs (HeVAL (he));
		1231
		1232	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1233
		1234	if (count == 1)
		1235	{
		1236	sv = newSVsv (POPs);
		1237	FREETMPS; LEAVE;
		1238	return sv;
		1239	}
		1240
		1241	FREETMPS; LEAVE;
		1242	}
		1243	}
		1244
		1245	if (dec->json.cb_object)
		1246	{
		1247	dSP;
		1248	int count;
		1249
		1250	ENTER; SAVETMPS; PUSHMARK (SP);
		1251	XPUSHs (sv_2mortal (sv));
		1252
		1253	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1254
		1255	if (count == 1)
		1256	{
		1257	sv = newSVsv (POPs);
		1258	FREETMPS; LEAVE;
		1259	return sv;
		1260	}
		1261
		1262	SvREFCNT_inc (sv);
		1263	FREETMPS; LEAVE;
		1264	}
		1265	}
		1266
		1267	return sv;
986		1268
987	fail:	1269	fail:
988	SvREFCNT_dec (hv);	1270	SvREFCNT_dec (hv);
989	DEC_DEC_DEPTH;	1271	DEC_DEC_DEPTH;
990	return 0;	1272	return 0;
991	}	1273	}
992		1274
993	static SV *	1275	static SV *
994	decode_sv (dec_t *dec)	1276	decode_sv (dec_t *dec)
995	{	1277	{
996	decode_ws (dec);	1278	// the beauty of JSON: you need exactly one character lookahead
		1279	// to parse anything.
997	switch (*dec->cur)	1280	switch (*dec->cur)
998	{	1281	{
999	case '"': ++dec->cur; return decode_str (dec);	1282	case '"': ++dec->cur; return decode_str (dec);
1000	case '[': ++dec->cur; return decode_av (dec);	1283	case '[': ++dec->cur; return decode_av (dec);
1001	case '{': ++dec->cur; return decode_hv (dec);	1284	case '{': ++dec->cur; return decode_hv (dec);
…		…
1007		1290
1008	case 't':	1291	case 't':
1009	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1010	{	1293	{
1011	dec->cur += 4;	1294	dec->cur += 4;
1012	return newSViv (1);	1295	#if JSON_SLOW
		1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1297	#endif
		1298	return SvREFCNT_inc (json_true);
1013	}	1299	}
1014	else	1300	else
1015	ERR ("'true' expected");	1301	ERR ("'true' expected");
1016		1302
1017	break;	1303	break;
1018		1304
1019	case 'f':	1305	case 'f':
1020	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1021	{	1307	{
1022	dec->cur += 5;	1308	dec->cur += 5;
1023	return newSViv (0);	1309	#if JSON_SLOW
		1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1311	#endif
		1312	return SvREFCNT_inc (json_false);
1024	}	1313	}
1025	else	1314	else
1026	ERR ("'false' expected");	1315	ERR ("'false' expected");
1027		1316
1028	break;	1317	break;
…		…
1046	fail:	1335	fail:
1047	return 0;	1336	return 0;
1048	}	1337	}
1049		1338
1050	static SV *	1339	static SV *
1051	decode_json (SV string, U32 flags, UV offset_return)	1340	decode_json (SV string, JSON json, UV *offset_return)
1052	{	1341	{
1053	dec_t dec;	1342	dec_t dec;
1054	UV offset;	1343	UV offset;
1055	SV *sv;	1344	SV *sv;
1056		1345
1057	SvGETMAGIC (string);	1346	SvGETMAGIC (string);
1058	SvUPGRADE (string, SVt_PV);	1347	SvUPGRADE (string, SVt_PV);
1059		1348
		1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1350	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1351	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1352
1060	if (flags & F_UTF8)	1353	if (json->flags & F_UTF8)
1061	sv_utf8_downgrade (string, 0);	1354	sv_utf8_downgrade (string, 0);
1062	else	1355	else
1063	sv_utf8_upgrade (string);	1356	sv_utf8_upgrade (string);
1064		1357
1065	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1066		1359
1067	dec.flags = flags;	1360	dec.json = *json;
1068	dec.cur = SvPVX (string);	1361	dec.cur = SvPVX (string);
1069	dec.end = SvEND (string);	1362	dec.end = SvEND (string);
1070	dec.err = 0;	1363	dec.err = 0;
1071	dec.depth = 0;	1364	dec.depth = 0;
1072	dec.maxdepth = DEC_DEPTH (dec.flags);	1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
		1366
		1367	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1368	dec.json.flags \|= F_HOOK;
1073		1369
1074	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1371
		1372	decode_ws (&dec);
1075	sv = decode_sv (&dec);	1373	sv = decode_sv (&dec);
1076		1374
1077	if (!(offset_return \|\| !sv))	1375	if (!(offset_return \|\| !sv))
1078	{	1376	{
1079	// check for trailing garbage	1377	// check for trailing garbage
…		…
1087	}	1385	}
1088	}	1386	}
1089		1387
1090	if (offset_return \|\| !sv)	1388	if (offset_return \|\| !sv)
1091	{	1389	{
1092	offset = dec.flags & F_UTF8	1390	offset = dec.json.flags & F_UTF8
1093	? dec.cur - SvPVX (string)	1391	? dec.cur - SvPVX (string)
1094	: utf8_distance (dec.cur, SvPVX (string));	1392	: utf8_distance (dec.cur, SvPVX (string));
1095		1393
1096	if (offset_return)	1394	if (offset_return)
1097	*offset_return = offset;	1395	*offset_return = offset;
…		…
1116	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1117	}	1415	}
1118		1416
1119	sv = sv_2mortal (sv);	1417	sv = sv_2mortal (sv);
1120		1418
1121	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1419	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)");	1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1123		1421
1124	return sv;	1422	return sv;
1125	}	1423	}
1126		1424
…		…
1130	MODULE = JSON::XS PACKAGE = JSON::XS	1428	MODULE = JSON::XS PACKAGE = JSON::XS
1131		1429
1132	BOOT:	1430	BOOT:
1133	{	1431	{
1134	int i;	1432	int i;
1135
1136	memset (decode_hexdigit, 0xff, 256);
1137		1433
1138	for (i = 0; i < 256; ++i)	1434	for (i = 0; i < 256; ++i)
1139	decode_hexdigit [i] =	1435	decode_hexdigit [i] =
1140	i >= '0' && i <= '9' ? i - '0'	1436	i >= '0' && i <= '9' ? i - '0'
1141	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1437	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1142	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1438	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1143	: -1;	1439	: -1;
1144		1440
1145	json_stash = gv_stashpv ("JSON::XS", 1);	1441	json_stash = gv_stashpv ("JSON::XS" , 1);
		1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1443
		1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1146	}	1446	}
1147		1447
1148	PROTOTYPES: DISABLE	1448	PROTOTYPES: DISABLE
1149		1449
1150	SV new (char dummy)	1450	void CLONE (...)
1151	CODE:	1451	CODE:
1152	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1452	json_stash = 0;
		1453	json_boolean_stash = 0;
		1454
		1455	void new (char *klass)
		1456	PPCODE:
		1457	{
		1458	SV *pv = NEWSV (0, sizeof (JSON));
		1459	SvPOK_only (pv);
		1460	Zero (SvPVX (pv), 1, JSON);
		1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1462	XPUSHs (sv_2mortal (sv_bless (
		1463	newRV_noinc (pv),
		1464	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1465	)));
		1466	}
		1467
		1468	void ascii (JSON *self, int enable = 1)
		1469	ALIAS:
		1470	ascii = F_ASCII
		1471	latin1 = F_LATIN1
		1472	utf8 = F_UTF8
		1473	indent = F_INDENT
		1474	canonical = F_CANONICAL
		1475	space_before = F_SPACE_BEFORE
		1476	space_after = F_SPACE_AFTER
		1477	pretty = F_PRETTY
		1478	allow_nonref = F_ALLOW_NONREF
		1479	shrink = F_SHRINK
		1480	allow_blessed = F_ALLOW_BLESSED
		1481	convert_blessed = F_CONV_BLESSED
		1482	relaxed = F_RELAXED
		1483	PPCODE:
		1484	{
		1485	if (enable)
		1486	self->flags \|= ix;
		1487	else
		1488	self->flags &= ~ix;
		1489
		1490	XPUSHs (ST (0));
		1491	}
		1492
		1493	void get_ascii (JSON *self)
		1494	ALIAS:
		1495	get_ascii = F_ASCII
		1496	get_latin1 = F_LATIN1
		1497	get_utf8 = F_UTF8
		1498	get_indent = F_INDENT
		1499	get_canonical = F_CANONICAL
		1500	get_space_before = F_SPACE_BEFORE
		1501	get_space_after = F_SPACE_AFTER
		1502	get_allow_nonref = F_ALLOW_NONREF
		1503	get_shrink = F_SHRINK
		1504	get_allow_blessed = F_ALLOW_BLESSED
		1505	get_convert_blessed = F_CONV_BLESSED
		1506	get_relaxed = F_RELAXED
		1507	PPCODE:
		1508	XPUSHs (boolSV (self->flags & ix));
		1509
		1510	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1511	PPCODE:
		1512	{
		1513	UV log2 = 0;
		1514
		1515	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1516
		1517	while ((1UL << log2) < max_depth)
		1518	++log2;
		1519
		1520	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1521
		1522	XPUSHs (ST (0));
		1523	}
		1524
		1525	U32 get_max_depth (JSON *self)
		1526	CODE:
		1527	RETVAL = DEC_DEPTH (self->flags);
1153	OUTPUT:	1528	OUTPUT:
1154	RETVAL	1529	RETVAL
1155		1530
1156	SV ascii (SV self, int enable = 1)	1531	void max_size (JSON *self, UV max_size = 0)
1157	ALIAS:	1532	PPCODE:
1158	ascii = F_ASCII	1533	{
1159	latin1 = F_LATIN1	1534	UV log2 = 0;
1160	utf8 = F_UTF8	1535
1161	indent = F_INDENT	1536	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1162	canonical = F_CANONICAL	1537	if (max_size == 1) max_size = 2;
1163	space_before = F_SPACE_BEFORE	1538
1164	space_after = F_SPACE_AFTER	1539	while ((1UL << log2) < max_size)
1165	pretty = F_PRETTY	1540	++log2;
1166	allow_nonref = F_ALLOW_NONREF	1541
1167	shrink = F_SHRINK	1542	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1543
		1544	XPUSHs (ST (0));
		1545	}
		1546
		1547	int get_max_size (JSON *self)
1168	CODE:	1548	CODE:
1169	{	1549	RETVAL = DEC_SIZE (self->flags);
1170	UV *uv = SvJSON (self);
1171	if (enable)
1172	*uv \|= ix;
1173	else
1174	*uv &= ~ix;
1175
1176	RETVAL = newSVsv (self);
1177	}
1178	OUTPUT:	1550	OUTPUT:
1179	RETVAL	1551	RETVAL
1180		1552
1181	SV max_depth (SV self, UV max_depth = 0x80000000UL)	1553	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
1182	CODE:
1183	{
1184	UV *uv = SvJSON (self);
1185	UV log2 = 0;
1186
1187	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1188
1189	while ((1UL << log2) < max_depth)
1190	++log2;
1191
1192	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1193
1194	RETVAL = newSVsv (self);
1195	}
1196	OUTPUT:
1197	RETVAL
1198
1199	void encode (SV self, SV scalar)
1200	PPCODE:	1554	PPCODE:
1201	XPUSHs (encode_json (scalar, *SvJSON (self)));	1555	{
		1556	SvREFCNT_dec (self->cb_object);
		1557	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1202		1558
1203	void decode (SV self, SV jsonstr)	1559	XPUSHs (ST (0));
		1560	}
		1561
		1562	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
1204	PPCODE:	1563	PPCODE:
		1564	{
		1565	if (!self->cb_sk_object)
		1566	self->cb_sk_object = newHV ();
		1567
		1568	if (SvOK (cb))
		1569	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1570	else
		1571	{
		1572	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1573
		1574	if (!HvKEYS (self->cb_sk_object))
		1575	{
		1576	SvREFCNT_dec (self->cb_sk_object);
		1577	self->cb_sk_object = 0;
		1578	}
		1579	}
		1580
		1581	XPUSHs (ST (0));
		1582	}
		1583
		1584	void encode (JSON self, SV scalar)
		1585	PPCODE:
		1586	XPUSHs (encode_json (scalar, self));
		1587
		1588	void decode (JSON self, SV jsonstr)
		1589	PPCODE:
1205	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1590	XPUSHs (decode_json (jsonstr, self, 0));
1206		1591
1207	void decode_prefix (SV self, SV jsonstr)	1592	void decode_prefix (JSON self, SV jsonstr)
1208	PPCODE:	1593	PPCODE:
1209	{	1594	{
1210	UV offset;	1595	UV offset;
1211	EXTEND (SP, 2);	1596	EXTEND (SP, 2);
1212	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	1597	PUSHs (decode_json (jsonstr, self, &offset));
1213	PUSHs (sv_2mortal (newSVuv (offset)));	1598	PUSHs (sv_2mortal (newSVuv (offset)));
1214	}	1599	}
1215		1600
		1601	void DESTROY (JSON *self)
		1602	CODE:
		1603	SvREFCNT_dec (self->cb_sk_object);
		1604	SvREFCNT_dec (self->cb_object);
		1605
1216	PROTOTYPES: ENABLE	1606	PROTOTYPES: ENABLE
1217		1607
1218	void to_json (SV *scalar)	1608	void to_json (SV *scalar)
1219	ALIAS:
1220	objToJson = 0
1221	PPCODE:	1609	PPCODE:
		1610	{
		1611	JSON json = { F_DEFAULT \| F_UTF8 };
1222	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));	1612	XPUSHs (encode_json (scalar, &json));
		1613	}
1223		1614
1224	void from_json (SV *jsonstr)	1615	void from_json (SV *jsonstr)
1225	ALIAS:
1226	jsonToObj = 0
1227	PPCODE:	1616	PPCODE:
		1617	{
		1618	JSON json = { F_DEFAULT \| F_UTF8 };
1228	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));	1619	XPUSHs (decode_json (jsonstr, &json, 0));
		1620	}
1229		1621

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 1.68 by root, Thu Nov 29 08:37:11 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 1.68 by root, Thu Nov 29 08:37:11 2007 UTC