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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.38 by root, Mon Jun 11 03:18:07 2007 UTC vs.
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC

…		…
9		9
10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	11	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	12	#endif
13		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
14	#define F_ASCII 0x00000001UL	20	#define F_ASCII 0x00000001UL
15	#define F_LATIN1 0x00000002UL	21	#define F_LATIN1 0x00000002UL
16	#define F_UTF8 0x00000004UL	22	#define F_UTF8 0x00000004UL
17	#define F_INDENT 0x00000008UL	23	#define F_INDENT 0x00000008UL
18	#define F_CANONICAL 0x00000010UL	24	#define F_CANONICAL 0x00000010UL
19	#define F_SPACE_BEFORE 0x00000020UL	25	#define F_SPACE_BEFORE 0x00000020UL
20	#define F_SPACE_AFTER 0x00000040UL	26	#define F_SPACE_AFTER 0x00000040UL
21	#define F_ALLOW_NONREF 0x00000100UL	27	#define F_ALLOW_NONREF 0x00000100UL
22	#define F_SHRINK 0x00000200UL	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
23	#define F_MAXDEPTH 0xf8000000UL	31	#define F_MAXDEPTH 0xf8000000UL
24	#define S_MAXDEPTH 27	32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
25		35
26	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))	36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
27		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
28	// F_SELFCONVERT? <=> to_json/toJson
29	// F_BLESSED? <=> { $__class__$ => }
30		38
31	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	39	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
32	#define F_DEFAULT (9UL << S_MAXDEPTH)	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
33		41
34	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
…		…
48	#endif	56	#endif
49		57
50	#define expect_false(expr) expect ((expr) != 0, 0)	58	#define expect_false(expr) expect ((expr) != 0, 0)
51	#define expect_true(expr) expect ((expr) != 0, 1)	59	#define expect_true(expr) expect ((expr) != 0, 1)
52		60
53	static HV *json_stash; // JSON::XS::	61	static HV json_stash, json_boolean_stash; // JSON::XS::
		62	static SV json_true, json_false;
		63
		64	typedef struct {
		65	U32 flags;
		66	} JSON;
54		67
55	/////////////////////////////////////////////////////////////////////////////	68	/////////////////////////////////////////////////////////////////////////////
56	// utility functions	69	// utility functions
57		70
58	static UV *	71	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)	72	shrink (SV *sv)
69	{	73	{
70	sv_utf8_downgrade (sv, 1);	74	sv_utf8_downgrade (sv, 1);
71	if (SvLEN (sv) > SvCUR (sv) + 1)	75	if (SvLEN (sv) > SvCUR (sv) + 1)
72	{	76	{
…		…
107	typedef struct	111	typedef struct
108	{	112	{
109	char *cur; // SvPVX (sv) + current output position	113	char *cur; // SvPVX (sv) + current output position
110	char *end; // SvEND (sv)	114	char *end; // SvEND (sv)
111	SV *sv; // result scalar	115	SV *sv; // result scalar
112	U32 flags; // F_*	116	JSON json;
113	U32 indent; // indentation level	117	U32 indent; // indentation level
114	U32 maxdepth; // max. indentation/recursion level	118	U32 maxdepth; // max. indentation/recursion level
115	} enc_t;	119	} enc_t;
116		120
117	inline void	121	inline void
…		…
178	STRLEN clen;	182	STRLEN clen;
179	UV uch;	183	UV uch;
180		184
181	if (is_utf8)	185	if (is_utf8)
182	{	186	{
183	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
184	uch = decode_utf8 (str, end - str, &clen);	187	uch = decode_utf8 (str, end - str, &clen);
185	if (clen == (STRLEN)-1)	188	if (clen == (STRLEN)-1)
186	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	189	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
187	}	190	}
188	else	191	else
…		…
192	}	195	}
193		196
194	if (uch > 0x10FFFFUL)	197	if (uch > 0x10FFFFUL)
195	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	198	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
196		199
197	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))	200	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
198	{	201	{
199	if (uch > 0xFFFFUL)	202	if (uch > 0xFFFFUL)
200	{	203	{
201	need (enc, len += 11);	204	need (enc, len += 11);
202	sprintf (enc->cur, "\\u%04x\\u%04x",	205	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
216	*enc->cur++ = hexdigit [(uch >> 0) & 15];	219	*enc->cur++ = hexdigit [(uch >> 0) & 15];
217	}	220	}
218		221
219	str += clen;	222	str += clen;
220	}	223	}
221	else if (enc->flags & F_LATIN1)	224	else if (enc->json.flags & F_LATIN1)
222	{	225	{
223	*enc->cur++ = uch;	226	*enc->cur++ = uch;
224	str += clen;	227	str += clen;
225	}	228	}
226	else if (is_utf8)	229	else if (is_utf8)
…		…
247	}	250	}
248		251
249	inline void	252	inline void
250	encode_indent (enc_t *enc)	253	encode_indent (enc_t *enc)
251	{	254	{
252	if (enc->flags & F_INDENT)	255	if (enc->json.flags & F_INDENT)
253	{	256	{
254	int spaces = enc->indent * INDENT_STEP;	257	int spaces = enc->indent * INDENT_STEP;
255		258
256	need (enc, spaces);	259	need (enc, spaces);
257	memset (enc->cur, ' ', spaces);	260	memset (enc->cur, ' ', spaces);
…		…
267	}	270	}
268		271
269	inline void	272	inline void
270	encode_nl (enc_t *enc)	273	encode_nl (enc_t *enc)
271	{	274	{
272	if (enc->flags & F_INDENT)	275	if (enc->json.flags & F_INDENT)
273	{	276	{
274	need (enc, 1);	277	need (enc, 1);
275	encode_ch (enc, '\n');	278	encode_ch (enc, '\n');
276	}	279	}
277	}	280	}
…		…
279	inline void	282	inline void
280	encode_comma (enc_t *enc)	283	encode_comma (enc_t *enc)
281	{	284	{
282	encode_ch (enc, ',');	285	encode_ch (enc, ',');
283		286
284	if (enc->flags & F_INDENT)	287	if (enc->json.flags & F_INDENT)
285	encode_nl (enc);	288	encode_nl (enc);
286	else if (enc->flags & F_SPACE_AFTER)	289	else if (enc->json.flags & F_SPACE_AFTER)
287	encode_space (enc);	290	encode_space (enc);
288	}	291	}
289		292
290	static void encode_sv (enc_t enc, SV sv);	293	static void encode_sv (enc_t enc, SV sv);
291		294
…		…
334	else	337	else
335	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	338	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
336		339
337	encode_ch (enc, '"');	340	encode_ch (enc, '"');
338		341
339	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	342	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
340	encode_ch (enc, ':');	343	encode_ch (enc, ':');
341	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	344	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
342	encode_sv (enc, HeVAL (he));	345	encode_sv (enc, HeVAL (he));
343	}	346	}
344		347
345	// compare hash entries, used when all keys are bytestrings	348	// compare hash entries, used when all keys are bytestrings
346	static int	349	static int
…		…
381	{	384	{
382	// for canonical output we have to sort by keys first	385	// for canonical output we have to sort by keys first
383	// actually, this is mostly due to the stupid so-called	386	// actually, this is mostly due to the stupid so-called
384	// security workaround added somewhere in 5.8.x.	387	// security workaround added somewhere in 5.8.x.
385	// that randomises hash orderings	388	// that randomises hash orderings
386	if (enc->flags & F_CANONICAL)	389	if (enc->json.flags & F_CANONICAL)
387	{	390	{
388	int fast = 1;	391	int fast = 1;
389	HE *he;	392	HE *he;
390	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)	393	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
391	HE *hes = _alloca (count sizeof (HE));	394	HE *hes = _alloca (count sizeof (HE));
…		…
463	svtype svt;	466	svtype svt;
464		467
465	SvGETMAGIC (sv);	468	SvGETMAGIC (sv);
466	svt = SvTYPE (sv);	469	svt = SvTYPE (sv);
467		470
		471	if (expect_false (SvOBJECT (sv)))
		472	{
		473	if (SvSTASH (sv) == json_boolean_stash)
		474	{
		475	if (SvIV (sv) == 0)
		476	encode_str (enc, "false", 5, 0);
		477	else
		478	encode_str (enc, "true", 4, 0);
		479	}
		480	else
		481	{
		482	#if 0
		483	if (0 && sv_derived_from (rv, "JSON::Literal"))
		484	{
		485	// not yet
		486	}
		487	#endif
		488	if (enc->json.flags & F_CONV_BLESSED)
		489	{
		490	// we re-bless the reference to get overload and other niceties right
		491	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		492
		493	if (to_json)
		494	{
		495	dSP;
		496	ENTER;
		497	SAVETMPS;
		498	PUSHMARK (SP);
		499	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		500
		501	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		502	// check anyways.
		503	PUTBACK;
		504	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		505	SPAGAIN;
		506
		507	encode_sv (enc, POPs);
		508
		509	FREETMPS;
		510	LEAVE;
		511	}
		512	else if (enc->json.flags & F_ALLOW_BLESSED)
		513	encode_str (enc, "null", 4, 0);
		514	else
		515	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		516	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		517	}
		518	else if (enc->json.flags & F_ALLOW_BLESSED)
		519	encode_str (enc, "null", 4, 0);
		520	else
		521	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		522	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		523	}
		524	}
468	if (svt == SVt_PVHV)	525	else if (svt == SVt_PVHV)
469	encode_hv (enc, (HV *)sv);	526	encode_hv (enc, (HV *)sv);
470	else if (svt == SVt_PVAV)	527	else if (svt == SVt_PVAV)
471	encode_av (enc, (AV *)sv);	528	encode_av (enc, (AV *)sv);
472	else if (svt < SVt_PVAV)	529	else if (svt < SVt_PVAV)
473	{	530	{
…		…
497	encode_str (enc, str, len, SvUTF8 (sv));	554	encode_str (enc, str, len, SvUTF8 (sv));
498	encode_ch (enc, '"');	555	encode_ch (enc, '"');
499	}	556	}
500	else if (SvNOKp (sv))	557	else if (SvNOKp (sv))
501	{	558	{
		559	// trust that perl will do the right thing w.r.t. JSON syntax.
502	need (enc, NV_DIG + 32);	560	need (enc, NV_DIG + 32);
503	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	561	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
504	enc->cur += strlen (enc->cur);	562	enc->cur += strlen (enc->cur);
505	}	563	}
506	else if (SvIOKp (sv))	564	else if (SvIOKp (sv))
507	{	565	{
508	// we assume we can always read an IV as a UV	566	// we assume we can always read an IV as a UV
509	if (SvUV (sv) & ~(UV)0x7fff)	567	if (SvUV (sv) & ~(UV)0x7fff)
510	{	568	{
		569	// large integer, use the (rather slow) snprintf way.
511	need (enc, sizeof (UV) * 3);	570	need (enc, sizeof (UV) * 3);
512	enc->cur +=	571	enc->cur +=
513	SvIsUV(sv)	572	SvIsUV(sv)
514	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))	573	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
515	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));	574	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
…		…
518	{	577	{
519	// optimise the "small number case"	578	// optimise the "small number case"
520	// code will likely be branchless and use only a single multiplication	579	// code will likely be branchless and use only a single multiplication
521	I32 i = SvIV (sv);	580	I32 i = SvIV (sv);
522	U32 u;	581	U32 u;
		582	char digit, nz = 0;
523		583
524	need (enc, 6);	584	need (enc, 6);
525		585
526	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;	586	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
527	u = i < 0 ? -i : i;	587	u = i < 0 ? -i : i;
528		588
529	// convert to 4.28 fixed-point representation	589	// convert to 4.28 fixed-point representation
530	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits	590	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
531		591
532	char digit, nz = 0;	592	// now output digit by digit, each time masking out the integer part
533		593	// and multiplying by 5 while moving the decimal point one to the right,
		594	// resulting in a net multiplication by 10.
		595	// we always write the digit to memory but conditionally increment
		596	// the pointer, to ease the usage of conditional move instructions.
534	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;	597	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
535	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;	598	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
536	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;	599	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
537	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;	600	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
538	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1;	601	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
539	}	602	}
540	}	603	}
541	else if (SvROK (sv))	604	else if (SvROK (sv))
542	encode_rv (enc, SvRV (sv));	605	encode_rv (enc, SvRV (sv));
543	else if (!SvOK (sv))	606	else if (!SvOK (sv))
…		…
546	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	609	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
547	SvPV_nolen (sv), SvFLAGS (sv));	610	SvPV_nolen (sv), SvFLAGS (sv));
548	}	611	}
549		612
550	static SV *	613	static SV *
551	encode_json (SV *scalar, U32 flags)	614	encode_json (SV scalar, JSON json)
552	{	615	{
553	enc_t enc;	616	enc_t enc;
554		617
555	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	618	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
556	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	619	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
557		620
558	enc.flags = flags;	621	enc.json = *json;
559	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	622	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
560	enc.cur = SvPVX (enc.sv);	623	enc.cur = SvPVX (enc.sv);
561	enc.end = SvEND (enc.sv);	624	enc.end = SvEND (enc.sv);
562	enc.indent = 0;	625	enc.indent = 0;
563	enc.maxdepth = DEC_DEPTH (flags);	626	enc.maxdepth = DEC_DEPTH (enc.json.flags);
564		627
565	SvPOK_only (enc.sv);	628	SvPOK_only (enc.sv);
566	encode_sv (&enc, scalar);	629	encode_sv (&enc, scalar);
567		630
568	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	631	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
569	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	632	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
570		633
571	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))	634	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
572	SvUTF8_on (enc.sv);	635	SvUTF8_on (enc.sv);
573		636
574	if (enc.flags & F_SHRINK)	637	if (enc.json.flags & F_SHRINK)
575	shrink (enc.sv);	638	shrink (enc.sv);
576		639
577	return enc.sv;	640	return enc.sv;
578	}	641	}
579		642
…		…
584	typedef struct	647	typedef struct
585	{	648	{
586	char *cur; // current parser pointer	649	char *cur; // current parser pointer
587	char *end; // end of input string	650	char *end; // end of input string
588	const char *err; // parse error, if != 0	651	const char *err; // parse error, if != 0
589	U32 flags; // F_*	652	JSON json;
590	U32 depth; // recursion depth	653	U32 depth; // recursion depth
591	U32 maxdepth; // recursion depth limit	654	U32 maxdepth; // recursion depth limit
592	} dec_t;	655	} dec_t;
593		656
594	inline void	657	inline void
…		…
625	decode_4hex (dec_t *dec)	688	decode_4hex (dec_t *dec)
626	{	689	{
627	signed char d1, d2, d3, d4;	690	signed char d1, d2, d3, d4;
628	unsigned char cur = (unsigned char )dec->cur;	691	unsigned char cur = (unsigned char )dec->cur;
629		692
630	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("four hexadecimal digits expected");	693	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");	694	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");	695	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");	696	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
634		697
635	dec->cur += 4;	698	dec->cur += 4;
636		699
637	return ((UV)d1) << 12	700	return ((UV)d1) << 12
638	\| ((UV)d2) << 8	701	\| ((UV)d2) << 8
…		…
861	{	924	{
862	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so	925	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
863	if (*start == '-')	926	if (*start == '-')
864	switch (dec->cur - start)	927	switch (dec->cur - start)
865	{	928	{
866	case 2: return newSViv (-( start [1] - '0' ));	929	case 2: return newSViv (-( start [1] - '0' * 1));
867	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));	930	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
868	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));	931	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
869	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));	932	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
870	}	933	}
871	else	934	else
872	switch (dec->cur - start)	935	switch (dec->cur - start)
873	{	936	{
874	case 1: return newSViv ( start [0] - '0' );	937	case 1: return newSViv ( start [0] - '0' * 1);
875	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);	938	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
876	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);	939	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
877	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);	940	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
878	}	941	}
879		942
…		…
886	if (uv < (UV)IV_MIN)	949	if (uv < (UV)IV_MIN)
887	return newSViv (-(IV)uv);	950	return newSViv (-(IV)uv);
888	}	951	}
889	else	952	else
890	return newSVuv (uv);	953	return newSVuv (uv);
		954
		955	// here would likely be the place for bigint support
891	}	956	}
892	}	957	}
893		958
		959	// if we ever support bigint or bigfloat, this is the place for bigfloat
894	return newSVnv (Atof (start));	960	return newSVnv (Atof (start));
895		961
896	fail:	962	fail:
897	return 0;	963	return 0;
898	}	964	}
…		…
952	if (*dec->cur == '}')	1018	if (*dec->cur == '}')
953	++dec->cur;	1019	++dec->cur;
954	else	1020	else
955	for (;;)	1021	for (;;)
956	{	1022	{
957	SV key, value;
958
959	decode_ws (dec); EXPECT_CH ('"');	1023	decode_ws (dec); EXPECT_CH ('"');
960		1024
961	key = decode_str (dec);	1025	// heuristic: assume that
962	if (!key)	1026	// a) decode_str + hv_store_ent are abysmally slow.
963	goto fail;	1027	// b) most hash keys are short, simple ascii text.
		1028	// => try to "fast-match" such strings to avoid
		1029	// the overhead of decode_str + hv_store_ent.
		1030	{
		1031	SV *value;
		1032	char *p = dec->cur;
		1033	char *e = p + 24; // only try up to 24 bytes
964		1034
965	decode_ws (dec); EXPECT_CH (':');	1035	for (;;)
966
967	value = decode_sv (dec);
968	if (!value)
969	{	1036	{
		1037	// the >= 0x80 is true on most architectures
		1038	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1039	{
		1040	// slow path, back up and use decode_str
		1041	SV *key = decode_str (dec);
		1042	if (!key)
		1043	goto fail;
		1044
		1045	decode_ws (dec); EXPECT_CH (':');
		1046
		1047	value = decode_sv (dec);
		1048	if (!value)
		1049	{
		1050	SvREFCNT_dec (key);
		1051	goto fail;
		1052	}
		1053
		1054	hv_store_ent (hv, key, value, 0);
970	SvREFCNT_dec (key);	1055	SvREFCNT_dec (key);
		1056
		1057	break;
		1058	}
		1059	else if (*p == '"')
		1060	{
		1061	// fast path, got a simple key
		1062	char *key = dec->cur;
		1063	int len = p - key;
		1064	dec->cur = p + 1;
		1065
		1066	decode_ws (dec); EXPECT_CH (':');
		1067
		1068	value = decode_sv (dec);
		1069	if (!value)
971	goto fail;	1070	goto fail;
		1071
		1072	hv_store (hv, key, len, value, 0);
		1073
		1074	break;
		1075	}
		1076
		1077	++p;
972	}	1078	}
973		1079	}
974	hv_store_ent (hv, key, value, 0);
975	SvREFCNT_dec (key);
976		1080
977	decode_ws (dec);	1081	decode_ws (dec);
978		1082
979	if (*dec->cur == '}')	1083	if (*dec->cur == '}')
980	{	1084	{
…		…
999		1103
1000	static SV *	1104	static SV *
1001	decode_sv (dec_t *dec)	1105	decode_sv (dec_t *dec)
1002	{	1106	{
1003	decode_ws (dec);	1107	decode_ws (dec);
		1108
		1109	// the beauty of JSON: you need exactly one character lookahead
		1110	// to parse anything.
1004	switch (*dec->cur)	1111	switch (*dec->cur)
1005	{	1112	{
1006	case '"': ++dec->cur; return decode_str (dec);	1113	case '"': ++dec->cur; return decode_str (dec);
1007	case '[': ++dec->cur; return decode_av (dec);	1114	case '[': ++dec->cur; return decode_av (dec);
1008	case '{': ++dec->cur; return decode_hv (dec);	1115	case '{': ++dec->cur; return decode_hv (dec);
…		…
1014		1121
1015	case 't':	1122	case 't':
1016	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1123	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1017	{	1124	{
1018	dec->cur += 4;	1125	dec->cur += 4;
1019	return newSViv (1);	1126	return SvREFCNT_inc (json_true);
1020	}	1127	}
1021	else	1128	else
1022	ERR ("'true' expected");	1129	ERR ("'true' expected");
1023		1130
1024	break;	1131	break;
1025		1132
1026	case 'f':	1133	case 'f':
1027	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1134	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1028	{	1135	{
1029	dec->cur += 5;	1136	dec->cur += 5;
1030	return newSViv (0);	1137	return SvREFCNT_inc (json_false);
1031	}	1138	}
1032	else	1139	else
1033	ERR ("'false' expected");	1140	ERR ("'false' expected");
1034		1141
1035	break;	1142	break;
…		…
1053	fail:	1160	fail:
1054	return 0;	1161	return 0;
1055	}	1162	}
1056		1163
1057	static SV *	1164	static SV *
1058	decode_json (SV string, U32 flags, UV offset_return)	1165	decode_json (SV string, JSON json, UV *offset_return)
1059	{	1166	{
1060	dec_t dec;	1167	dec_t dec;
1061	UV offset;	1168	UV offset;
1062	SV *sv;	1169	SV *sv;
1063		1170
1064	SvGETMAGIC (string);	1171	SvGETMAGIC (string);
1065	SvUPGRADE (string, SVt_PV);	1172	SvUPGRADE (string, SVt_PV);
1066		1173
		1174	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1175	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1176	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1177
1067	if (flags & F_UTF8)	1178	if (json->flags & F_UTF8)
1068	sv_utf8_downgrade (string, 0);	1179	sv_utf8_downgrade (string, 0);
1069	else	1180	else
1070	sv_utf8_upgrade (string);	1181	sv_utf8_upgrade (string);
1071		1182
1072	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1183	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1073		1184
1074	dec.flags = flags;	1185	dec.json = *json;
1075	dec.cur = SvPVX (string);	1186	dec.cur = SvPVX (string);
1076	dec.end = SvEND (string);	1187	dec.end = SvEND (string);
1077	dec.err = 0;	1188	dec.err = 0;
1078	dec.depth = 0;	1189	dec.depth = 0;
1079	dec.maxdepth = DEC_DEPTH (dec.flags);	1190	dec.maxdepth = DEC_DEPTH (dec.json.flags);
1080		1191
1081	*dec.end = 0; // this should basically be a nop, too, but make sure it's there	1192	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
1082	sv = decode_sv (&dec);	1193	sv = decode_sv (&dec);
1083		1194
1084	if (!(offset_return \|\| !sv))	1195	if (!(offset_return \|\| !sv))
…		…
1094	}	1205	}
1095	}	1206	}
1096		1207
1097	if (offset_return \|\| !sv)	1208	if (offset_return \|\| !sv)
1098	{	1209	{
1099	offset = dec.flags & F_UTF8	1210	offset = dec.json.flags & F_UTF8
1100	? dec.cur - SvPVX (string)	1211	? dec.cur - SvPVX (string)
1101	: utf8_distance (dec.cur, SvPVX (string));	1212	: utf8_distance (dec.cur, SvPVX (string));
1102		1213
1103	if (offset_return)	1214	if (offset_return)
1104	*offset_return = offset;	1215	*offset_return = offset;
…		…
1123	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1234	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1124	}	1235	}
1125		1236
1126	sv = sv_2mortal (sv);	1237	sv = sv_2mortal (sv);
1127		1238
1128	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1239	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1129	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1240	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1130		1241
1131	return sv;	1242	return sv;
1132	}	1243	}
1133		1244
…		…
1137	MODULE = JSON::XS PACKAGE = JSON::XS	1248	MODULE = JSON::XS PACKAGE = JSON::XS
1138		1249
1139	BOOT:	1250	BOOT:
1140	{	1251	{
1141	int i;	1252	int i;
1142
1143	memset (decode_hexdigit, 0xff, 256);
1144		1253
1145	for (i = 0; i < 256; ++i)	1254	for (i = 0; i < 256; ++i)
1146	decode_hexdigit [i] =	1255	decode_hexdigit [i] =
1147	i >= '0' && i <= '9' ? i - '0'	1256	i >= '0' && i <= '9' ? i - '0'
1148	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1257	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1149	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1258	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1150	: -1;	1259	: -1;
1151		1260
1152	json_stash = gv_stashpv ("JSON::XS", 1);	1261	json_stash = gv_stashpv ("JSON::XS" , 1);
		1262	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1263
		1264	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1265	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
1153	}	1266	}
1154		1267
1155	PROTOTYPES: DISABLE	1268	PROTOTYPES: DISABLE
1156		1269
1157	SV new (char dummy)	1270	void new (char *klass)
1158	CODE:	1271	PPCODE:
1159	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1272	{
1160	OUTPUT:	1273	SV *pv = NEWSV (0, sizeof (JSON));
1161	RETVAL	1274	SvPOK_only (pv);
		1275	Zero (SvPVX (pv), 1, sizeof (JSON));
		1276	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1277	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1278	}
1162		1279
1163	SV ascii (SV self, int enable = 1)	1280	void ascii (JSON *self, int enable = 1)
1164	ALIAS:	1281	ALIAS:
1165	ascii = F_ASCII	1282	ascii = F_ASCII
1166	latin1 = F_LATIN1	1283	latin1 = F_LATIN1
1167	utf8 = F_UTF8	1284	utf8 = F_UTF8
1168	indent = F_INDENT	1285	indent = F_INDENT
1169	canonical = F_CANONICAL	1286	canonical = F_CANONICAL
1170	space_before = F_SPACE_BEFORE	1287	space_before = F_SPACE_BEFORE
1171	space_after = F_SPACE_AFTER	1288	space_after = F_SPACE_AFTER
1172	pretty = F_PRETTY	1289	pretty = F_PRETTY
1173	allow_nonref = F_ALLOW_NONREF	1290	allow_nonref = F_ALLOW_NONREF
1174	shrink = F_SHRINK	1291	shrink = F_SHRINK
		1292	allow_blessed = F_ALLOW_BLESSED
		1293	convert_blessed = F_CONV_BLESSED
1175	CODE:	1294	PPCODE:
1176	{	1295	{
1177	UV *uv = SvJSON (self);
1178	if (enable)	1296	if (enable)
1179	*uv \|= ix;	1297	self->flags \|= ix;
1180	else	1298	else
1181	*uv &= ~ix;	1299	self->flags &= ~ix;
1182		1300
1183	RETVAL = newSVsv (self);	1301	XPUSHs (ST (0));
1184	}	1302	}
1185	OUTPUT:
1186	RETVAL
1187		1303
1188	SV max_depth (SV self, UV max_depth = 0x80000000UL)	1304	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
1189	CODE:	1305	PPCODE:
1190	{	1306	{
1191	UV *uv = SvJSON (self);
1192	UV log2 = 0;	1307	UV log2 = 0;
1193		1308
1194	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1309	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1195		1310
1196	while ((1UL << log2) < max_depth)	1311	while ((1UL << log2) < max_depth)
1197	++log2;	1312	++log2;
1198		1313
1199	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);	1314	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
1200		1315
1201	RETVAL = newSVsv (self);	1316	XPUSHs (ST (0));
1202	}	1317	}
1203	OUTPUT:
1204	RETVAL
1205		1318
1206	void encode (SV self, SV scalar)	1319	void max_size (JSON *self, UV max_size = 0)
1207	PPCODE:	1320	PPCODE:
1208	XPUSHs (encode_json (scalar, *SvJSON (self)));	1321	{
		1322	UV log2 = 0;
1209		1323
		1324	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1325	if (max_size == 1) max_size = 2;
		1326
		1327	while ((1UL << log2) < max_size)
		1328	++log2;
		1329
		1330	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1331
		1332	XPUSHs (ST (0));
		1333	}
		1334
1210	void decode (SV self, SV jsonstr)	1335	void encode (JSON self, SV scalar)
1211	PPCODE:	1336	PPCODE:
		1337	XPUSHs (encode_json (scalar, self));
		1338
		1339	void decode (JSON self, SV jsonstr)
		1340	PPCODE:
1212	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));	1341	XPUSHs (decode_json (jsonstr, self, 0));
1213		1342
1214	void decode_prefix (SV self, SV jsonstr)	1343	void decode_prefix (JSON self, SV jsonstr)
1215	PPCODE:	1344	PPCODE:
1216	{	1345	{
1217	UV offset;	1346	UV offset;
1218	EXTEND (SP, 2);	1347	EXTEND (SP, 2);
1219	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));	1348	PUSHs (decode_json (jsonstr, self, &offset));
1220	PUSHs (sv_2mortal (newSVuv (offset)));	1349	PUSHs (sv_2mortal (newSVuv (offset)));
1221	}	1350	}
1222		1351
1223	PROTOTYPES: ENABLE	1352	PROTOTYPES: ENABLE
1224		1353
1225	void to_json (SV *scalar)	1354	void to_json (SV *scalar)
1226	ALIAS:
1227	objToJson = 0
1228	PPCODE:	1355	PPCODE:
		1356	{
		1357	JSON json = { F_DEFAULT \| F_UTF8 };
1229	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));	1358	XPUSHs (encode_json (scalar, &json));
		1359	}
1230		1360
1231	void from_json (SV *jsonstr)	1361	void from_json (SV *jsonstr)
1232	ALIAS:
1233	jsonToObj = 0
1234	PPCODE:	1362	PPCODE:
		1363	{
		1364	JSON json = { F_DEFAULT \| F_UTF8 };
1235	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));	1365	XPUSHs (decode_json (jsonstr, &json, 0));
		1366	}
1236		1367

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.38 by root, Mon Jun 11 03:18:07 2007 UTC vs. Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.38 by root, Mon Jun 11 03:18:07 2007 UTC vs.
Revision 1.48 by root, Sun Jul 1 22:20:00 2007 UTC