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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.17 by root, Sun Mar 25 02:46:41 2007 UTC vs.
Revision 1.42 by root, Sat Jun 23 22:56:08 2007 UTC

…		…
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		9
		10	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		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
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	27	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_MAXDEPTH 0xf8000000UL
		30	#define S_MAXDEPTH 27
17		31
18	// F_SKIPINVALID?	32	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
19	// F_EXECCODEREF?	33
20	// F_SELFCONVERT? <=> { &__class__ => }	34	// F_SELFCONVERT? <=> to_json/toJson
		35	// F_BLESSED? <=> { $__class__$ => }
21		36
22	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	37	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
23	#define F_DEFAULT 0	38	#define F_DEFAULT (9UL << S_MAXDEPTH)
24		39
25	#define INIT_SIZE 32 // initial scalar size to be allocated	40	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	41	#define INDENT_STEP 3 // spaces per indentation level
27		42
28	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
29	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	43	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
30		44
31	#define SB do {	45	#define SB do {
32	#define SE } while (0)	46	#define SE } while (0)
		47
		48	#if __GNUC__ >= 3
		49	# define expect(expr,value) __builtin_expect ((expr),(value))
		50	# define inline inline
		51	#else
		52	# define expect(expr,value) (expr)
		53	# define inline static
		54	#endif
		55
		56	#define expect_false(expr) expect ((expr) != 0, 0)
		57	#define expect_true(expr) expect ((expr) != 0, 1)
33		58
34	static HV *json_stash; // JSON::XS::	59	static HV *json_stash; // JSON::XS::
35		60
36	/////////////////////////////////////////////////////////////////////////////	61	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	62	// utility functions
…		…
62	// decode an utf-8 character and return it, or (UV)-1 in	87	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	88	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	89	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	90	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	91	// note that we never call this function for a ascii codepoints
67	static UV	92	inline UV
68	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	93	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
69	{	94	{
70	if (s[0] > 0xdf \|\| s[0] < 0xc2)	95	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	96	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	97	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
73	{	98	{
74	*clen = 2;	99	*clen = 2;
75	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	100	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
76	}	101	}
77	else	102	else
		103	{
		104	*clen = (STRLEN)-1;
78	return (UV)-1;	105	return (UV)-1;
		106	}
79	}	107	}
80		108
81	/////////////////////////////////////////////////////////////////////////////	109	/////////////////////////////////////////////////////////////////////////////
82	// encoder	110	// encoder
83		111
…		…
85	typedef struct	113	typedef struct
86	{	114	{
87	char *cur; // SvPVX (sv) + current output position	115	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	116	char *end; // SvEND (sv)
89	SV *sv; // result scalar	117	SV *sv; // result scalar
90	UV flags; // F_*	118	U32 flags; // F_*
91	int indent; // indentation level	119	U32 indent; // indentation level
92	int max_depth; // max. recursion level	120	U32 maxdepth; // max. indentation/recursion level
93	} enc_t;	121	} enc_t;
94		122
95	static void	123	inline void
96	need (enc_t *enc, STRLEN len)	124	need (enc_t *enc, STRLEN len)
97	{	125	{
98	if (enc->cur + len >= enc->end)	126	if (expect_false (enc->cur + len >= enc->end))
99	{	127	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	128	STRLEN cur = enc->cur - SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	129	SvGROW (enc->sv, cur + len + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	130	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	131	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	132	}
105	}	133	}
106		134
107	static void	135	inline void
108	encode_ch (enc_t *enc, char ch)	136	encode_ch (enc_t *enc, char ch)
109	{	137	{
110	need (enc, 1);	138	need (enc, 1);
111	*enc->cur++ = ch;	139	*enc->cur++ = ch;
112	}	140	}
…		…
120		148
121	while (str < end)	149	while (str < end)
122	{	150	{
123	unsigned char ch = (unsigned char )str;	151	unsigned char ch = (unsigned char )str;
124		152
125	if (ch >= 0x20 && ch < 0x80) // most common case	153	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	154	{
127	if (ch == '"') // but with slow exceptions	155	if (expect_false (ch == '"')) // but with slow exceptions
128	{	156	{
129	need (enc, len += 1);	157	need (enc, len += 1);
130	*enc->cur++ = '\\';	158	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	159	*enc->cur++ = '"';
132	}	160	}
133	else if (ch == '\\')	161	else if (expect_false (ch == '\\'))
134	{	162	{
135	need (enc, len += 1);	163	need (enc, len += 1);
136	*enc->cur++ = '\\';	164	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
138	}	166	}
…		…
156	STRLEN clen;	184	STRLEN clen;
157	UV uch;	185	UV uch;
158		186
159	if (is_utf8)	187	if (is_utf8)
160	{	188	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	189	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	190	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	191	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	192	}
166	else	193	else
…		…
170	}	197	}
171		198
172	if (uch > 0x10FFFFUL)	199	if (uch > 0x10FFFFUL)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	200	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174		201
175	if (uch < 0x80 \|\| enc->flags & F_ASCII)	202	if (uch < 0x80 \|\| enc->flags & F_ASCII \|\| (enc->flags & F_LATIN1 && uch > 0xFF))
176	{	203	{
177	if (uch > 0xFFFFUL)	204	if (uch > 0xFFFFUL)
178	{	205	{
179	need (enc, len += 11);	206	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	207	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	221	*enc->cur++ = hexdigit [(uch >> 0) & 15];
195	}	222	}
196		223
197	str += clen;	224	str += clen;
198	}	225	}
		226	else if (enc->flags & F_LATIN1)
		227	{
		228	*enc->cur++ = uch;
		229	str += clen;
		230	}
199	else if (is_utf8)	231	else if (is_utf8)
200	{	232	{
201	need (enc, len += clen);	233	need (enc, len += clen);
202	do	234	do
203	{	235	{
…		…
205	}	237	}
206	while (--clen);	238	while (--clen);
207	}	239	}
208	else	240	else
209	{	241	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	242	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	243	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
212	++str;	244	++str;
213	}	245	}
214	}	246	}
215	}	247	}
…		…
217		249
218	--len;	250	--len;
219	}	251	}
220	}	252	}
221		253
222	static void	254	inline void
223	encode_indent (enc_t *enc)	255	encode_indent (enc_t *enc)
224	{	256	{
225	if (enc->flags & F_INDENT)	257	if (enc->flags & F_INDENT)
226	{	258	{
227	int spaces = enc->indent * INDENT_STEP;	259	int spaces = enc->indent * INDENT_STEP;
…		…
230	memset (enc->cur, ' ', spaces);	262	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	263	enc->cur += spaces;
232	}	264	}
233	}	265	}
234		266
235	static void	267	inline void
236	encode_space (enc_t *enc)	268	encode_space (enc_t *enc)
237	{	269	{
238	need (enc, 1);	270	need (enc, 1);
239	encode_ch (enc, ' ');	271	encode_ch (enc, ' ');
240	}	272	}
241		273
242	static void	274	inline void
243	encode_nl (enc_t *enc)	275	encode_nl (enc_t *enc)
244	{	276	{
245	if (enc->flags & F_INDENT)	277	if (enc->flags & F_INDENT)
246	{	278	{
247	need (enc, 1);	279	need (enc, 1);
248	encode_ch (enc, '\n');	280	encode_ch (enc, '\n');
249	}	281	}
250	}	282	}
251		283
252	static void	284	inline void
253	encode_comma (enc_t *enc)	285	encode_comma (enc_t *enc)
254	{	286	{
255	encode_ch (enc, ',');	287	encode_ch (enc, ',');
256		288
257	if (enc->flags & F_INDENT)	289	if (enc->flags & F_INDENT)
…		…
264		296
265	static void	297	static void
266	encode_av (enc_t enc, AV av)	298	encode_av (enc_t enc, AV av)
267	{	299	{
268	int i, len = av_len (av);	300	int i, len = av_len (av);
		301
		302	if (enc->indent >= enc->maxdepth)
		303	croak ("data structure too deep (hit recursion limit)");
269		304
270	encode_ch (enc, '['); encode_nl (enc);	305	encode_ch (enc, '['); encode_nl (enc);
271	++enc->indent;	306	++enc->indent;
272		307
273	for (i = 0; i <= len; ++i)	308	for (i = 0; i <= len; ++i)
…		…
340	static void	375	static void
341	encode_hv (enc_t enc, HV hv)	376	encode_hv (enc_t enc, HV hv)
342	{	377	{
343	int count, i;	378	int count, i;
344		379
		380	if (enc->indent >= enc->maxdepth)
		381	croak ("data structure too deep (hit recursion limit)");
		382
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	383	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
346		384
347	if ((count = hv_iterinit (hv)))	385	if ((count = hv_iterinit (hv)))
348	{	386	{
349	// for canonical output we have to sort by keys first	387	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	388	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	389	// security workaround added somewhere in 5.8.x.
352	// that randomises hash orderings	390	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	391	if (enc->flags & F_CANONICAL)
354	{	392	{
355	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	393	int fast = 1;
		394	HE *he;
		395	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		396	HE *hes = _alloca (count sizeof (HE));
		397	#else
		398	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		399	#endif
357		400
358	i = 0;	401	i = 0;
359	while ((he = hv_iternext (hv)))	402	while ((he = hv_iternext (hv)))
360	{	403	{
361	hes [i++] = he;	404	hes [i++] = he;
…		…
396		439
397	encode_nl (enc);	440	encode_nl (enc);
398	}	441	}
399	else	442	else
400	{	443	{
401	SV *sv;
402	HE *he = hv_iternext (hv);	444	HE *he = hv_iternext (hv);
403		445
404	for (;;)	446	for (;;)
405	{	447	{
406	encode_indent (enc);	448	encode_indent (enc);
…		…
415	encode_nl (enc);	457	encode_nl (enc);
416	}	458	}
417	}	459	}
418		460
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	461	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		462	}
		463
		464	// encode objects, arrays and special \0=false and \1=true values.
		465	static void
		466	encode_rv (enc_t enc, SV sv)
		467	{
		468	svtype svt;
		469
		470	SvGETMAGIC (sv);
		471	svt = SvTYPE (sv);
		472
		473	if (svt == SVt_PVHV)
		474	encode_hv (enc, (HV *)sv);
		475	else if (svt == SVt_PVAV)
		476	encode_av (enc, (AV *)sv);
		477	else if (svt < SVt_PVAV)
		478	{
		479	if (SvNIOK (sv) && SvIV (sv) == 0)
		480	encode_str (enc, "false", 5, 0);
		481	else if (SvNIOK (sv) && SvIV (sv) == 1)
		482	encode_str (enc, "true", 4, 0);
		483	else
		484	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		485	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		486	}
		487	else
		488	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		489	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	490	}
421		491
422	static void	492	static void
423	encode_sv (enc_t enc, SV sv)	493	encode_sv (enc_t enc, SV sv)
424	{	494	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	502	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	503	encode_ch (enc, '"');
434	}	504	}
435	else if (SvNOKp (sv))	505	else if (SvNOKp (sv))
436	{	506	{
		507	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	508	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	509	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	510	enc->cur += strlen (enc->cur);
440	}	511	}
441	else if (SvIOKp (sv))	512	else if (SvIOKp (sv))
442	{	513	{
443	need (enc, 64);	514	// we assume we can always read an IV as a UV
		515	if (SvUV (sv) & ~(UV)0x7fff)
		516	{
		517	// large integer, use the (rather slow) snprintf way.
		518	need (enc, sizeof (UV) * 3);
444	enc->cur +=	519	enc->cur +=
445	SvIsUV(sv)	520	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	521	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	522	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		523	}
		524	else
		525	{
		526	// optimise the "small number case"
		527	// code will likely be branchless and use only a single multiplication
		528	I32 i = SvIV (sv);
		529	U32 u;
		530	char digit, nz = 0;
		531
		532	need (enc, 6);
		533
		534	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		535	u = i < 0 ? -i : i;
		536
		537	// convert to 4.28 fixed-point representation
		538	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		539
		540	// now output digit by digit, each time masking out the integer part
		541	// and multiplying by 5 while moving the decimal point one to the right,
		542	// resulting in a net multiplication by 10.
		543	// we always write the digit to memory but conditionally increment
		544	// the pointer, to ease the usage of conditional move instructions.
		545	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		546	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		547	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		548	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		549	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		550	}
448	}	551	}
449	else if (SvROK (sv))	552	else if (SvROK (sv))
450	{	553	encode_rv (enc, SvRV (sv));
451	SV *rv = SvRV (sv);
452
453	if (enc->indent >= enc->max_depth)
454	croak ("data structure too deep (hit recursion limit)");
455
456	switch (SvTYPE (rv))
457	{
458	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
459	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
460
461	default:
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv));
464	}
465	}
466	else if (!SvOK (sv))	554	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	555	encode_str (enc, "null", 4, 0);
468	else	556	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	557	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	558	SvPV_nolen (sv), SvFLAGS (sv));
471	}	559	}
472		560
473	static SV *	561	static SV *
474	encode_json (SV *scalar, UV flags)	562	encode_json (SV *scalar, U32 flags)
475	{	563	{
		564	enc_t enc;
		565
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	566	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	567	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		568
479	enc_t enc;
480	enc.flags = flags;	569	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	570	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	571	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	572	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	573	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	574	enc.maxdepth = DEC_DEPTH (flags);
486		575
487	SvPOK_only (enc.sv);	576	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	577	encode_sv (&enc, scalar);
489		578
		579	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		580	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		581
490	if (!(flags & (F_ASCII \| F_UTF8)))	582	if (!(flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
491	SvUTF8_on (enc.sv);	583	SvUTF8_on (enc.sv);
492
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494		584
495	if (enc.flags & F_SHRINK)	585	if (enc.flags & F_SHRINK)
496	shrink (enc.sv);	586	shrink (enc.sv);
497		587
498	return enc.sv;	588	return enc.sv;
…		…
505	typedef struct	595	typedef struct
506	{	596	{
507	char *cur; // current parser pointer	597	char *cur; // current parser pointer
508	char *end; // end of input string	598	char *end; // end of input string
509	const char *err; // parse error, if != 0	599	const char *err; // parse error, if != 0
510	UV flags; // F_*	600	U32 flags; // F_*
		601	U32 depth; // recursion depth
		602	U32 maxdepth; // recursion depth limit
511	} dec_t;	603	} dec_t;
512		604
513	static void	605	inline void
514	decode_ws (dec_t *dec)	606	decode_ws (dec_t *dec)
515	{	607	{
516	for (;;)	608	for (;;)
517	{	609	{
518	char ch = *dec->cur;	610	char ch = *dec->cur;
…		…
524	++dec->cur;	616	++dec->cur;
525	}	617	}
526	}	618	}
527		619
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	620	#define ERR(reason) SB dec->err = reason; goto fail; SE
		621
529	#define EXPECT_CH(ch) SB \	622	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	623	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	624	ERR (# ch " expected"); \
532	++dec->cur; \	625	++dec->cur; \
533	SE	626	SE
534		627
		628	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		629	#define DEC_DEC_DEPTH --dec->depth
		630
535	static SV decode_sv (dec_t dec);	631	static SV decode_sv (dec_t dec);
536		632
537	static signed char decode_hexdigit[256];	633	static signed char decode_hexdigit[256];
538		634
539	static UV	635	static UV
540	decode_4hex (dec_t *dec)	636	decode_4hex (dec_t *dec)
541	{	637	{
542	signed char d1, d2, d3, d4;	638	signed char d1, d2, d3, d4;
543	unsigned char cur = (unsigned char )dec->cur;	639	unsigned char cur = (unsigned char )dec->cur;
544		640
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	641	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
546	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	642	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
547	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	643	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
548	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	644	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		645
550	dec->cur += 4;	646	dec->cur += 4;
551		647
552	return ((UV)d1) << 12	648	return ((UV)d1) << 12
553	\| ((UV)d2) << 8	649	\| ((UV)d2) << 8
…		…
561	static SV *	657	static SV *
562	decode_str (dec_t *dec)	658	decode_str (dec_t *dec)
563	{	659	{
564	SV *sv = 0;	660	SV *sv = 0;
565	int utf8 = 0;	661	int utf8 = 0;
		662	char *dec_cur = dec->cur;
566		663
567	do	664	do
568	{	665	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	666	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	667	char *cur = buf;
571		668
572	do	669	do
573	{	670	{
574	unsigned char ch = (unsigned char )dec->cur++;	671	unsigned char ch = (unsigned char )dec_cur++;
575		672
576	if (ch == '"')	673	if (expect_false (ch == '"'))
577	{	674	{
578	--dec->cur;	675	--dec_cur;
579	break;	676	break;
580	}	677	}
581	else if (ch == '\\')	678	else if (expect_false (ch == '\\'))
582	{	679	{
583	switch (*dec->cur)	680	switch (*dec_cur)
584	{	681	{
585	case '\\':	682	case '\\':
586	case '/':	683	case '/':
587	case '"': cur++ = dec->cur++; break;	684	case '"': cur++ = dec_cur++; break;
588		685
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	686	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	687	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	688	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	689	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	690	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		691
595	case 'u':	692	case 'u':
596	{	693	{
597	UV lo, hi;	694	UV lo, hi;
598	++dec->cur;	695	++dec_cur;
599		696
		697	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	698	hi = decode_4hex (dec);
		699	dec_cur = dec->cur;
601	if (hi == (UV)-1)	700	if (hi == (UV)-1)
602	goto fail;	701	goto fail;
603		702
604	// possibly a surrogate pair	703	// possibly a surrogate pair
605	if (hi >= 0xd800)	704	if (hi >= 0xd800)
606	if (hi < 0xdc00)	705	if (hi < 0xdc00)
607	{	706	{
608	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	707	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	708	ERR ("missing low surrogate character in surrogate pair");
610		709
611	dec->cur += 2;	710	dec_cur += 2;
612		711
		712	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	713	lo = decode_4hex (dec);
		714	dec_cur = dec->cur;
614	if (lo == (UV)-1)	715	if (lo == (UV)-1)
615	goto fail;	716	goto fail;
616		717
617	if (lo < 0xdc00 \|\| lo >= 0xe000)	718	if (lo < 0xdc00 \|\| lo >= 0xe000)
618	ERR ("surrogate pair expected");	719	ERR ("surrogate pair expected");
…		…
632	*cur++ = hi;	733	*cur++ = hi;
633	}	734	}
634	break;	735	break;
635		736
636	default:	737	default:
637	--dec->cur;	738	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	739	ERR ("illegal backslash escape sequence in string");
639	}	740	}
640	}	741	}
641	else if (ch >= 0x20 && ch <= 0x7f)	742	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
642	*cur++ = ch;	743	*cur++ = ch;
643	else if (ch >= 0x80)	744	else if (ch >= 0x80)
644	{	745	{
645	--dec->cur;
646
647	STRLEN clen;	746	STRLEN clen;
		747	UV uch;
		748
		749	--dec_cur;
		750
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	751	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	752	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	753	ERR ("malformed UTF-8 character in JSON string");
651		754
652	do	755	do
653	{
654	cur++ = dec->cur++;	756	cur++ = dec_cur++;
655	}
656	while (--clen);	757	while (--clen);
657		758
658	utf8 = 1;	759	utf8 = 1;
659	}	760	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	761	else
		762	{
		763	--dec_cur;
		764
		765	if (!ch)
		766	ERR ("unexpected end of string while parsing JSON string");
		767	else
663	ERR ("invalid character encountered");	768	ERR ("invalid character encountered while parsing JSON string");
664		769	}
665	}	770	}
666	while (cur < buf + SHORT_STRING_LEN);	771	while (cur < buf + SHORT_STRING_LEN);
667		772
		773	{
668	STRLEN len = cur - buf;	774	STRLEN len = cur - buf;
669		775
670	if (sv)	776	if (sv)
671	{	777	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	778	SvGROW (sv, SvCUR (sv) + len + 1);
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	779	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	780	SvCUR_set (sv, SvCUR (sv) + len);
675	}	781	}
676	else	782	else
677	sv = newSVpvn (buf, len);	783	sv = newSVpvn (buf, len);
678	}	784	}
		785	}
679	while (*dec->cur != '"');	786	while (*dec_cur != '"');
680		787
681	++dec->cur;	788	++dec_cur;
682		789
683	if (sv)	790	if (sv)
684	{	791	{
685	SvPOK_only (sv);	792	SvPOK_only (sv);
686	*SvEND (sv) = 0;	793	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	796	SvUTF8_on (sv);
690	}	797	}
691	else	798	else
692	sv = newSVpvn ("", 0);	799	sv = newSVpvn ("", 0);
693		800
		801	dec->cur = dec_cur;
694	return sv;	802	return sv;
695		803
696	fail:	804	fail:
		805	dec->cur = dec_cur;
697	return 0;	806	return 0;
698	}	807	}
699		808
700	static SV *	809	static SV *
701	decode_num (dec_t *dec)	810	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	868	is_nv = 1;
760	}	869	}
761		870
762	if (!is_nv)	871	if (!is_nv)
763	{	872	{
764	UV uv;	873	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
765	int numtype = grok_number (start, dec->cur - start, &uv);	874	if (*start == '-')
766	if (numtype & IS_NUMBER_IN_UV)	875	switch (dec->cur - start)
767	if (numtype & IS_NUMBER_NEG)
768	{	876	{
769	if (uv < (UV)IV_MIN)	877	case 2: return newSViv (-( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	878	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		879	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		880	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
771	}	881	}
		882	else
		883	switch (dec->cur - start)
		884	{
		885	case 1: return newSViv ( start [0] - '0' * 1);
		886	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		887	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		888	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		889	}
		890
		891	{
		892	UV uv;
		893	int numtype = grok_number (start, dec->cur - start, &uv);
		894	if (numtype & IS_NUMBER_IN_UV)
		895	if (numtype & IS_NUMBER_NEG)
		896	{
		897	if (uv < (UV)IV_MIN)
		898	return newSViv (-(IV)uv);
		899	}
772	else	900	else
773	return newSVuv (uv);	901	return newSVuv (uv);
		902
		903	// here would likely be the place for bigint support
774	}	904	}
		905	}
775		906
		907	// if we ever support bigint or bigfloat, this is the place for bigfloat
776	return newSVnv (Atof (start));	908	return newSVnv (Atof (start));
777		909
778	fail:	910	fail:
779	return 0;	911	return 0;
780	}	912	}
…		…
782	static SV *	914	static SV *
783	decode_av (dec_t *dec)	915	decode_av (dec_t *dec)
784	{	916	{
785	AV *av = newAV ();	917	AV *av = newAV ();
786		918
		919	DEC_INC_DEPTH;
787	decode_ws (dec);	920	decode_ws (dec);
		921
788	if (*dec->cur == ']')	922	if (*dec->cur == ']')
789	++dec->cur;	923	++dec->cur;
790	else	924	else
791	for (;;)	925	for (;;)
792	{	926	{
…		…
810	ERR (", or ] expected while parsing array");	944	ERR (", or ] expected while parsing array");
811		945
812	++dec->cur;	946	++dec->cur;
813	}	947	}
814		948
		949	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	950	return newRV_noinc ((SV *)av);
816		951
817	fail:	952	fail:
818	SvREFCNT_dec (av);	953	SvREFCNT_dec (av);
		954	DEC_DEC_DEPTH;
819	return 0;	955	return 0;
820	}	956	}
821		957
822	static SV *	958	static SV *
823	decode_hv (dec_t *dec)	959	decode_hv (dec_t *dec)
824	{	960	{
825	HV *hv = newHV ();	961	HV *hv = newHV ();
826		962
		963	DEC_INC_DEPTH;
827	decode_ws (dec);	964	decode_ws (dec);
		965
828	if (*dec->cur == '}')	966	if (*dec->cur == '}')
829	++dec->cur;	967	++dec->cur;
830	else	968	else
831	for (;;)	969	for (;;)
832	{	970	{
…		…
845	{	983	{
846	SvREFCNT_dec (key);	984	SvREFCNT_dec (key);
847	goto fail;	985	goto fail;
848	}	986	}
849		987
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);	988	hv_store_ent (hv, key, value, 0);
		989	SvREFCNT_dec (key);
852		990
853	decode_ws (dec);	991	decode_ws (dec);
854		992
855	if (*dec->cur == '}')	993	if (*dec->cur == '}')
856	{	994	{
…		…
862	ERR (", or } expected while parsing object/hash");	1000	ERR (", or } expected while parsing object/hash");
863		1001
864	++dec->cur;	1002	++dec->cur;
865	}	1003	}
866		1004
		1005	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1006	return newRV_noinc ((SV *)hv);
868		1007
869	fail:	1008	fail:
870	SvREFCNT_dec (hv);	1009	SvREFCNT_dec (hv);
		1010	DEC_DEC_DEPTH;
871	return 0;	1011	return 0;
872	}	1012	}
873		1013
874	static SV *	1014	static SV *
875	decode_sv (dec_t *dec)	1015	decode_sv (dec_t *dec)
876	{	1016	{
877	decode_ws (dec);	1017	decode_ws (dec);
		1018
		1019	// the beauty of JSON: you need exactly one character lookahead
		1020	// to parse anything.
878	switch (*dec->cur)	1021	switch (*dec->cur)
879	{	1022	{
880	case '"': ++dec->cur; return decode_str (dec);	1023	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1024	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1025	case '{': ++dec->cur; return decode_hv (dec);
…		…
918	ERR ("'null' expected");	1061	ERR ("'null' expected");
919		1062
920	break;	1063	break;
921		1064
922	default:	1065	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1066	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1067	break;
925	}	1068	}
926		1069
927	fail:	1070	fail:
928	return 0;	1071	return 0;
929	}	1072	}
930		1073
931	static SV *	1074	static SV *
932	decode_json (SV *string, UV flags)	1075	decode_json (SV string, U32 flags, UV offset_return)
933	{	1076	{
		1077	dec_t dec;
		1078	UV offset;
934	SV *sv;	1079	SV *sv;
		1080
		1081	SvGETMAGIC (string);
		1082	SvUPGRADE (string, SVt_PV);
935		1083
936	if (flags & F_UTF8)	1084	if (flags & F_UTF8)
937	sv_utf8_downgrade (string, 0);	1085	sv_utf8_downgrade (string, 0);
938	else	1086	else
939	sv_utf8_upgrade (string);	1087	sv_utf8_upgrade (string);
940		1088
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1089	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1090
943	dec_t dec;
944	dec.flags = flags;	1091	dec.flags = flags;
945	dec.cur = SvPVX (string);	1092	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1093	dec.end = SvEND (string);
947	dec.err = 0;	1094	dec.err = 0;
		1095	dec.depth = 0;
		1096	dec.maxdepth = DEC_DEPTH (dec.flags);
948		1097
		1098	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
949	sv = decode_sv (&dec);	1099	sv = decode_sv (&dec);
950		1100
		1101	if (!(offset_return \|\| !sv))
		1102	{
		1103	// check for trailing garbage
		1104	decode_ws (&dec);
		1105
		1106	if (*dec.cur)
		1107	{
		1108	dec.err = "garbage after JSON object";
		1109	SvREFCNT_dec (sv);
		1110	sv = 0;
		1111	}
		1112	}
		1113
		1114	if (offset_return \|\| !sv)
		1115	{
		1116	offset = dec.flags & F_UTF8
		1117	? dec.cur - SvPVX (string)
		1118	: utf8_distance (dec.cur, SvPVX (string));
		1119
		1120	if (offset_return)
		1121	*offset_return = offset;
		1122	}
		1123
951	if (!sv)	1124	if (!sv)
952	{	1125	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1126	SV *uni = sv_newmortal ();
957		1127
958	// horrible hack to silence warning inside pv_uni_display	1128	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1129	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1130	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1132	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1133	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1134	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1135	LEAVE;
966		1136
967	croak ("%s, at character offset %d (%s)",	1137	croak ("%s, at character offset %d [\"%s\"]",
968	dec.err,	1138	dec.err,
969	(int)offset,	1139	(int)offset,
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1140	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1141	}
972		1142
…		…
985		1155
986	BOOT:	1156	BOOT:
987	{	1157	{
988	int i;	1158	int i;
989		1159
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1160	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1161	decode_hexdigit [i] =
993		1162	i >= '0' && i <= '9' ? i - '0'
994	for (i = 7; i--; )	1163	: i >= 'a' && i <= 'f' ? i - 'a' + 10
995	{	1164	: i >= 'A' && i <= 'F' ? i - 'A' + 10
996	decode_hexdigit ['a' + i] = 10 + i;	1165	: -1;
997	decode_hexdigit ['A' + i] = 10 + i;
998	}
999		1166
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1167	json_stash = gv_stashpv ("JSON::XS", 1);
1001	}	1168	}
1002		1169
1003	PROTOTYPES: DISABLE	1170	PROTOTYPES: DISABLE
…		…
1009	RETVAL	1176	RETVAL
1010		1177
1011	SV ascii (SV self, int enable = 1)	1178	SV ascii (SV self, int enable = 1)
1012	ALIAS:	1179	ALIAS:
1013	ascii = F_ASCII	1180	ascii = F_ASCII
		1181	latin1 = F_LATIN1
1014	utf8 = F_UTF8	1182	utf8 = F_UTF8
1015	indent = F_INDENT	1183	indent = F_INDENT
1016	canonical = F_CANONICAL	1184	canonical = F_CANONICAL
1017	space_before = F_SPACE_BEFORE	1185	space_before = F_SPACE_BEFORE
1018	space_after = F_SPACE_AFTER	1186	space_after = F_SPACE_AFTER
…		…
1030	RETVAL = newSVsv (self);	1198	RETVAL = newSVsv (self);
1031	}	1199	}
1032	OUTPUT:	1200	OUTPUT:
1033	RETVAL	1201	RETVAL
1034		1202
		1203	SV max_depth (SV self, UV max_depth = 0x80000000UL)
		1204	CODE:
		1205	{
		1206	UV *uv = SvJSON (self);
		1207	UV log2 = 0;
		1208
		1209	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1210
		1211	while ((1UL << log2) < max_depth)
		1212	++log2;
		1213
		1214	uv = uv & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1215
		1216	RETVAL = newSVsv (self);
		1217	}
		1218	OUTPUT:
		1219	RETVAL
		1220
1035	void encode (SV self, SV scalar)	1221	void encode (SV self, SV scalar)
1036	PPCODE:	1222	PPCODE:
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1223	XPUSHs (encode_json (scalar, *SvJSON (self)));
1038		1224
1039	void decode (SV self, SV jsonstr)	1225	void decode (SV self, SV jsonstr)
1040	PPCODE:	1226	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1227	XPUSHs (decode_json (jsonstr, *SvJSON (self), 0));
		1228
		1229	void decode_prefix (SV self, SV jsonstr)
		1230	PPCODE:
		1231	{
		1232	UV offset;
		1233	EXTEND (SP, 2);
		1234	PUSHs (decode_json (jsonstr, *SvJSON (self), &offset));
		1235	PUSHs (sv_2mortal (newSVuv (offset)));
		1236	}
1042		1237
1043	PROTOTYPES: ENABLE	1238	PROTOTYPES: ENABLE
1044		1239
1045	void to_json (SV *scalar)	1240	void to_json (SV *scalar)
1046	ALIAS:	1241	ALIAS:
1047	objToJson = 0	1242	objToJson = 0
1048	PPCODE:	1243	PPCODE:
1049	XPUSHs (encode_json (scalar, F_UTF8));	1244	XPUSHs (encode_json (scalar, F_DEFAULT \| F_UTF8));
1050		1245
1051	void from_json (SV *jsonstr)	1246	void from_json (SV *jsonstr)
1052	ALIAS:	1247	ALIAS:
1053	jsonToObj = 0	1248	jsonToObj = 0
1054	PPCODE:	1249	PPCODE:
1055	XPUSHs (decode_json (jsonstr, F_UTF8));	1250	XPUSHs (decode_json (jsonstr, F_DEFAULT \| F_UTF8, 0));
1056		1251

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.17 by root, Sun Mar 25 02:46:41 2007 UTC vs. Revision 1.42 by root, Sat Jun 23 22:56:08 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.17 by root, Sun Mar 25 02:46:41 2007 UTC vs.
Revision 1.42 by root, Sat Jun 23 22:56:08 2007 UTC