[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 1.55 by root, Mon Jul 23 22:57:40 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_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		36
18	// F_SKIPINVALID?	37	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
19	// F_EXECCODEREF?	38	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
20	// F_SELFCONVERT? <=> { &__class__ => }
21		39
22	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	40	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
23	#define F_DEFAULT 0	41	#define F_DEFAULT (9UL << S_MAXDEPTH)
24		42
25	#define INIT_SIZE 32 // initial scalar size to be allocated	43	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	44	#define INDENT_STEP 3 // spaces per indentation level
27		45
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	46	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
30		47
31	#define SB do {	48	#define SB do {
32	#define SE } while (0)	49	#define SE } while (0)
33		50
		51	#if __GNUC__ >= 3
		52	# define expect(expr,value) __builtin_expect ((expr),(value))
		53	# define inline inline
		54	#else
		55	# define expect(expr,value) (expr)
		56	# define inline static
		57	#endif
		58
		59	#define expect_false(expr) expect ((expr) != 0, 0)
		60	#define expect_true(expr) expect ((expr) != 0, 1)
		61
34	static HV *json_stash; // JSON::XS::	62	static HV json_stash, json_boolean_stash; // JSON::XS::
		63	static SV json_true, json_false;
		64
		65	typedef struct {
		66	U32 flags;
		67	SV *cb_object;
		68	HV *cb_sk_object;
		69	} JSON;
35		70
36	/////////////////////////////////////////////////////////////////////////////	71	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	72	// utility functions
38		73
39	static UV *	74	inline void
40	SvJSON (SV *sv)
41	{
42	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
43	croak ("object is not of type JSON::XS");
44
45	return &SvUVX (SvRV (sv));
46	}
47
48	static void
49	shrink (SV *sv)	75	shrink (SV *sv)
50	{	76	{
51	sv_utf8_downgrade (sv, 1);	77	sv_utf8_downgrade (sv, 1);
52	if (SvLEN (sv) > SvCUR (sv) + 1)	78	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	79	{
…		…
62	// decode an utf-8 character and return it, or (UV)-1 in	88	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	89	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	90	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	91	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	92	// note that we never call this function for a ascii codepoints
67	static UV	93	inline UV
68	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	94	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
69	{	95	{
70	if (s[0] > 0xdf \|\| s[0] < 0xc2)	96	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	97	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	98	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
73	{	99	{
74	*clen = 2;	100	*clen = 2;
75	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	101	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
76	}	102	}
77	else	103	else
		104	{
		105	*clen = (STRLEN)-1;
78	return (UV)-1;	106	return (UV)-1;
		107	}
79	}	108	}
80		109
81	/////////////////////////////////////////////////////////////////////////////	110	/////////////////////////////////////////////////////////////////////////////
82	// encoder	111	// encoder
83		112
…		…
85	typedef struct	114	typedef struct
86	{	115	{
87	char *cur; // SvPVX (sv) + current output position	116	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	117	char *end; // SvEND (sv)
89	SV *sv; // result scalar	118	SV *sv; // result scalar
90	UV flags; // F_*	119	JSON json;
91	int indent; // indentation level	120	U32 indent; // indentation level
92	int max_depth; // max. recursion level	121	U32 maxdepth; // max. indentation/recursion level
93	} enc_t;	122	} enc_t;
94		123
95	static void	124	inline void
96	need (enc_t *enc, STRLEN len)	125	need (enc_t *enc, STRLEN len)
97	{	126	{
98	if (enc->cur + len >= enc->end)	127	if (expect_false (enc->cur + len >= enc->end))
99	{	128	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	129	STRLEN cur = enc->cur - SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	130	SvGROW (enc->sv, cur + len + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	131	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	132	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	133	}
105	}	134	}
106		135
107	static void	136	inline void
108	encode_ch (enc_t *enc, char ch)	137	encode_ch (enc_t *enc, char ch)
109	{	138	{
110	need (enc, 1);	139	need (enc, 1);
111	*enc->cur++ = ch;	140	*enc->cur++ = ch;
112	}	141	}
…		…
120		149
121	while (str < end)	150	while (str < end)
122	{	151	{
123	unsigned char ch = (unsigned char )str;	152	unsigned char ch = (unsigned char )str;
124		153
125	if (ch >= 0x20 && ch < 0x80) // most common case	154	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	155	{
127	if (ch == '"') // but with slow exceptions	156	if (expect_false (ch == '"')) // but with slow exceptions
128	{	157	{
129	need (enc, len += 1);	158	need (enc, len += 1);
130	*enc->cur++ = '\\';	159	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	160	*enc->cur++ = '"';
132	}	161	}
133	else if (ch == '\\')	162	else if (expect_false (ch == '\\'))
134	{	163	{
135	need (enc, len += 1);	164	need (enc, len += 1);
136	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
138	}	167	}
…		…
156	STRLEN clen;	185	STRLEN clen;
157	UV uch;	186	UV uch;
158		187
159	if (is_utf8)	188	if (is_utf8)
160	{	189	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	190	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	191	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	192	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	193	}
166	else	194	else
…		…
170	}	198	}
171		199
172	if (uch > 0x10FFFFUL)	200	if (uch > 0x10FFFFUL)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	201	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174		202
175	if (uch < 0x80 \|\| enc->flags & F_ASCII)	203	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
176	{	204	{
177	if (uch > 0xFFFFUL)	205	if (uch > 0xFFFFUL)
178	{	206	{
179	need (enc, len += 11);	207	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	208	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	222	*enc->cur++ = hexdigit [(uch >> 0) & 15];
195	}	223	}
196		224
197	str += clen;	225	str += clen;
198	}	226	}
		227	else if (enc->json.flags & F_LATIN1)
		228	{
		229	*enc->cur++ = uch;
		230	str += clen;
		231	}
199	else if (is_utf8)	232	else if (is_utf8)
200	{	233	{
201	need (enc, len += clen);	234	need (enc, len += clen);
202	do	235	do
203	{	236	{
…		…
205	}	238	}
206	while (--clen);	239	while (--clen);
207	}	240	}
208	else	241	else
209	{	242	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	243	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	244	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
212	++str;	245	++str;
213	}	246	}
214	}	247	}
215	}	248	}
…		…
217		250
218	--len;	251	--len;
219	}	252	}
220	}	253	}
221		254
222	static void	255	inline void
223	encode_indent (enc_t *enc)	256	encode_indent (enc_t *enc)
224	{	257	{
225	if (enc->flags & F_INDENT)	258	if (enc->json.flags & F_INDENT)
226	{	259	{
227	int spaces = enc->indent * INDENT_STEP;	260	int spaces = enc->indent * INDENT_STEP;
228		261
229	need (enc, spaces);	262	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	263	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	264	enc->cur += spaces;
232	}	265	}
233	}	266	}
234		267
235	static void	268	inline void
236	encode_space (enc_t *enc)	269	encode_space (enc_t *enc)
237	{	270	{
238	need (enc, 1);	271	need (enc, 1);
239	encode_ch (enc, ' ');	272	encode_ch (enc, ' ');
240	}	273	}
241		274
242	static void	275	inline void
243	encode_nl (enc_t *enc)	276	encode_nl (enc_t *enc)
244	{	277	{
245	if (enc->flags & F_INDENT)	278	if (enc->json.flags & F_INDENT)
246	{	279	{
247	need (enc, 1);	280	need (enc, 1);
248	encode_ch (enc, '\n');	281	encode_ch (enc, '\n');
249	}	282	}
250	}	283	}
251		284
252	static void	285	inline void
253	encode_comma (enc_t *enc)	286	encode_comma (enc_t *enc)
254	{	287	{
255	encode_ch (enc, ',');	288	encode_ch (enc, ',');
256		289
257	if (enc->flags & F_INDENT)	290	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	291	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	292	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	293	encode_space (enc);
261	}	294	}
262		295
263	static void encode_sv (enc_t enc, SV sv);	296	static void encode_sv (enc_t enc, SV sv);
264		297
265	static void	298	static void
266	encode_av (enc_t enc, AV av)	299	encode_av (enc_t enc, AV av)
267	{	300	{
268	int i, len = av_len (av);	301	int i, len = av_len (av);
269		302
		303	if (enc->indent >= enc->maxdepth)
		304	croak ("data structure too deep (hit recursion limit)");
		305
270	encode_ch (enc, '['); encode_nl (enc);	306	encode_ch (enc, '['); encode_nl (enc);
271	++enc->indent;	307	++enc->indent;
272		308
273	for (i = 0; i <= len; ++i)	309	for (i = 0; i <= len; ++i)
274	{	310	{
		311	SV **svp = av_fetch (av, i, 0);
		312
275	encode_indent (enc);	313	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));	314
		315	if (svp)
		316	encode_sv (enc, *svp);
		317	else
		318	encode_str (enc, "null", 4, 0);
277		319
278	if (i < len)	320	if (i < len)
279	encode_comma (enc);	321	encode_comma (enc);
280	}	322	}
281		323
…		…
304	else	346	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	347	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		348
307	encode_ch (enc, '"');	349	encode_ch (enc, '"');
308		350
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	352	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));	354	encode_sv (enc, HeVAL (he));
313	}	355	}
314		356
315	// compare hash entries, used when all keys are bytestrings	357	// compare hash entries, used when all keys are bytestrings
316	static int	358	static int
…		…
339		381
340	static void	382	static void
341	encode_hv (enc_t enc, HV hv)	383	encode_hv (enc_t enc, HV hv)
342	{	384	{
343	int count, i;	385	int count, i;
		386
		387	if (enc->indent >= enc->maxdepth)
		388	croak ("data structure too deep (hit recursion limit)");
344		389
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
346		391
347	if ((count = hv_iterinit (hv)))	392	if ((count = hv_iterinit (hv)))
348	{	393	{
349	// for canonical output we have to sort by keys first	394	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	395	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	396	// security workaround added somewhere in 5.8.x.
352	// that randomises hash orderings	397	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	398	if (enc->json.flags & F_CANONICAL)
354	{	399	{
355	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	400	int fast = 1;
		401	HE *he;
		402	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		403	HE *hes = _alloca (count sizeof (HE));
		404	#else
		405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		406	#endif
357		407
358	i = 0;	408	i = 0;
359	while ((he = hv_iternext (hv)))	409	while ((he = hv_iternext (hv)))
360	{	410	{
361	hes [i++] = he;	411	hes [i++] = he;
…		…
396		446
397	encode_nl (enc);	447	encode_nl (enc);
398	}	448	}
399	else	449	else
400	{	450	{
401	SV *sv;
402	HE *he = hv_iternext (hv);	451	HE *he = hv_iternext (hv);
403		452
404	for (;;)	453	for (;;)
405	{	454	{
406	encode_indent (enc);	455	encode_indent (enc);
…		…
415	encode_nl (enc);	464	encode_nl (enc);
416	}	465	}
417	}	466	}
418		467
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		469	}
		470
		471	// encode objects, arrays and special \0=false and \1=true values.
		472	static void
		473	encode_rv (enc_t enc, SV sv)
		474	{
		475	svtype svt;
		476
		477	SvGETMAGIC (sv);
		478	svt = SvTYPE (sv);
		479
		480	if (expect_false (SvOBJECT (sv)))
		481	{
		482	if (SvSTASH (sv) == json_boolean_stash)
		483	{
		484	if (SvIV (sv))
		485	encode_str (enc, "true", 4, 0);
		486	else
		487	encode_str (enc, "false", 5, 0);
		488	}
		489	else
		490	{
		491	#if 0
		492	if (0 && sv_derived_from (rv, "JSON::Literal"))
		493	{
		494	// not yet
		495	}
		496	#endif
		497	if (enc->json.flags & F_CONV_BLESSED)
		498	{
		499	// we re-bless the reference to get overload and other niceties right
		500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		501
		502	if (to_json)
		503	{
		504	int count;
		505	dSP;
		506
		507	ENTER; SAVETMPS; PUSHMARK (SP);
		508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		509
		510	// calling with G_SCALAR ensures that we always get a 1 return value
		511	PUTBACK;
		512	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		513	SPAGAIN;
		514
		515	// catch this surprisingly common error
		516	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		517	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		518
		519	sv = POPs;
		520	PUTBACK;
		521
		522	encode_sv (enc, sv);
		523
		524	FREETMPS; LEAVE;
		525	}
		526	else if (enc->json.flags & F_ALLOW_BLESSED)
		527	encode_str (enc, "null", 4, 0);
		528	else
		529	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		530	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		531	}
		532	else if (enc->json.flags & F_ALLOW_BLESSED)
		533	encode_str (enc, "null", 4, 0);
		534	else
		535	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		536	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		537	}
		538	}
		539	else if (svt == SVt_PVHV)
		540	encode_hv (enc, (HV *)sv);
		541	else if (svt == SVt_PVAV)
		542	encode_av (enc, (AV *)sv);
		543	else if (svt < SVt_PVAV)
		544	{
		545	STRLEN len = 0;
		546	char *pv = svt ? SvPV (sv, len) : 0;
		547
		548	if (len == 1 && *pv == '1')
		549	encode_str (enc, "true", 4, 0);
		550	else if (len == 1 && *pv == '0')
		551	encode_str (enc, "false", 5, 0);
		552	else
		553	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		554	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		555	}
		556	else
		557	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		558	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	559	}
421		560
422	static void	561	static void
423	encode_sv (enc_t enc, SV sv)	562	encode_sv (enc_t enc, SV sv)
424	{	563	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	571	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	572	encode_ch (enc, '"');
434	}	573	}
435	else if (SvNOKp (sv))	574	else if (SvNOKp (sv))
436	{	575	{
		576	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	577	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	578	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	579	enc->cur += strlen (enc->cur);
440	}	580	}
441	else if (SvIOKp (sv))	581	else if (SvIOKp (sv))
442	{	582	{
443	need (enc, 64);	583	// we assume we can always read an IV as a UV
		584	if (SvUV (sv) & ~(UV)0x7fff)
		585	{
		586	// large integer, use the (rather slow) snprintf way.
		587	need (enc, sizeof (UV) * 3);
444	enc->cur +=	588	enc->cur +=
445	SvIsUV(sv)	589	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	590	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	591	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		592	}
		593	else
		594	{
		595	// optimise the "small number case"
		596	// code will likely be branchless and use only a single multiplication
		597	I32 i = SvIV (sv);
		598	U32 u;
		599	char digit, nz = 0;
		600
		601	need (enc, 6);
		602
		603	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		604	u = i < 0 ? -i : i;
		605
		606	// convert to 4.28 fixed-point representation
		607	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		608
		609	// now output digit by digit, each time masking out the integer part
		610	// and multiplying by 5 while moving the decimal point one to the right,
		611	// resulting in a net multiplication by 10.
		612	// we always write the digit to memory but conditionally increment
		613	// the pointer, to ease the usage of conditional move instructions.
		614	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		615	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		616	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		617	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		618	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		619	}
448	}	620	}
449	else if (SvROK (sv))	621	else if (SvROK (sv))
450	{	622	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))	623	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	624	encode_str (enc, "null", 4, 0);
468	else	625	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	626	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	627	SvPV_nolen (sv), SvFLAGS (sv));
471	}	628	}
472		629
473	static SV *	630	static SV *
474	encode_json (SV *scalar, UV flags)	631	encode_json (SV scalar, JSON json)
475	{	632	{
		633	enc_t enc;
		634
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	635	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	636	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		637
479	enc_t enc;	638	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	639	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	640	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	641	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	642	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	643	enc.maxdepth = DEC_DEPTH (enc.json.flags);
486		644
487	SvPOK_only (enc.sv);	645	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	646	encode_sv (&enc, scalar);
489		647
		648	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		649	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		650
490	if (!(flags & (F_ASCII \| F_UTF8)))	651	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
491	SvUTF8_on (enc.sv);	652	SvUTF8_on (enc.sv);
492		653
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	654	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	655	shrink (enc.sv);
497		656
498	return enc.sv;	657	return enc.sv;
499	}	658	}
500		659
…		…
505	typedef struct	664	typedef struct
506	{	665	{
507	char *cur; // current parser pointer	666	char *cur; // current parser pointer
508	char *end; // end of input string	667	char *end; // end of input string
509	const char *err; // parse error, if != 0	668	const char *err; // parse error, if != 0
510	UV flags; // F_*	669	JSON json;
		670	U32 depth; // recursion depth
		671	U32 maxdepth; // recursion depth limit
511	} dec_t;	672	} dec_t;
512		673
513	static void	674	inline void
514	decode_ws (dec_t *dec)	675	decode_ws (dec_t *dec)
515	{	676	{
516	for (;;)	677	for (;;)
517	{	678	{
518	char ch = *dec->cur;	679	char ch = *dec->cur;
…		…
524	++dec->cur;	685	++dec->cur;
525	}	686	}
526	}	687	}
527		688
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	689	#define ERR(reason) SB dec->err = reason; goto fail; SE
		690
529	#define EXPECT_CH(ch) SB \	691	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	692	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	693	ERR (# ch " expected"); \
532	++dec->cur; \	694	++dec->cur; \
533	SE	695	SE
534		696
		697	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		698	#define DEC_DEC_DEPTH --dec->depth
		699
535	static SV decode_sv (dec_t dec);	700	static SV decode_sv (dec_t dec);
536		701
537	static signed char decode_hexdigit[256];	702	static signed char decode_hexdigit[256];
538		703
539	static UV	704	static UV
540	decode_4hex (dec_t *dec)	705	decode_4hex (dec_t *dec)
541	{	706	{
542	signed char d1, d2, d3, d4;	707	signed char d1, d2, d3, d4;
543	unsigned char cur = (unsigned char )dec->cur;	708	unsigned char cur = (unsigned char )dec->cur;
544		709
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	710	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");	711	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");	712	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");	713	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		714
550	dec->cur += 4;	715	dec->cur += 4;
551		716
552	return ((UV)d1) << 12	717	return ((UV)d1) << 12
553	\| ((UV)d2) << 8	718	\| ((UV)d2) << 8
…		…
561	static SV *	726	static SV *
562	decode_str (dec_t *dec)	727	decode_str (dec_t *dec)
563	{	728	{
564	SV *sv = 0;	729	SV *sv = 0;
565	int utf8 = 0;	730	int utf8 = 0;
		731	char *dec_cur = dec->cur;
566		732
567	do	733	do
568	{	734	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	735	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	736	char *cur = buf;
571		737
572	do	738	do
573	{	739	{
574	unsigned char ch = (unsigned char )dec->cur++;	740	unsigned char ch = (unsigned char )dec_cur++;
575		741
576	if (ch == '"')	742	if (expect_false (ch == '"'))
577	{	743	{
578	--dec->cur;	744	--dec_cur;
579	break;	745	break;
580	}	746	}
581	else if (ch == '\\')	747	else if (expect_false (ch == '\\'))
582	{	748	{
583	switch (*dec->cur)	749	switch (*dec_cur)
584	{	750	{
585	case '\\':	751	case '\\':
586	case '/':	752	case '/':
587	case '"': cur++ = dec->cur++; break;	753	case '"': cur++ = dec_cur++; break;
588		754
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	755	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	756	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	757	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	758	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	759	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		760
595	case 'u':	761	case 'u':
596	{	762	{
597	UV lo, hi;	763	UV lo, hi;
598	++dec->cur;	764	++dec_cur;
599		765
		766	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	767	hi = decode_4hex (dec);
		768	dec_cur = dec->cur;
601	if (hi == (UV)-1)	769	if (hi == (UV)-1)
602	goto fail;	770	goto fail;
603		771
604	// possibly a surrogate pair	772	// possibly a surrogate pair
605	if (hi >= 0xd800)	773	if (hi >= 0xd800)
606	if (hi < 0xdc00)	774	if (hi < 0xdc00)
607	{	775	{
608	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	776	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	777	ERR ("missing low surrogate character in surrogate pair");
610		778
611	dec->cur += 2;	779	dec_cur += 2;
612		780
		781	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	782	lo = decode_4hex (dec);
		783	dec_cur = dec->cur;
614	if (lo == (UV)-1)	784	if (lo == (UV)-1)
615	goto fail;	785	goto fail;
616		786
617	if (lo < 0xdc00 \|\| lo >= 0xe000)	787	if (lo < 0xdc00 \|\| lo >= 0xe000)
618	ERR ("surrogate pair expected");	788	ERR ("surrogate pair expected");
…		…
632	*cur++ = hi;	802	*cur++ = hi;
633	}	803	}
634	break;	804	break;
635		805
636	default:	806	default:
637	--dec->cur;	807	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	808	ERR ("illegal backslash escape sequence in string");
639	}	809	}
640	}	810	}
641	else if (ch >= 0x20 && ch <= 0x7f)	811	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
642	*cur++ = ch;	812	*cur++ = ch;
643	else if (ch >= 0x80)	813	else if (ch >= 0x80)
644	{	814	{
645	--dec->cur;
646
647	STRLEN clen;	815	STRLEN clen;
		816	UV uch;
		817
		818	--dec_cur;
		819
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	820	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	821	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	822	ERR ("malformed UTF-8 character in JSON string");
651		823
652	do	824	do
653	{
654	cur++ = dec->cur++;	825	cur++ = dec_cur++;
655	}
656	while (--clen);	826	while (--clen);
657		827
658	utf8 = 1;	828	utf8 = 1;
659	}	829	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	830	else
		831	{
		832	--dec_cur;
		833
		834	if (!ch)
		835	ERR ("unexpected end of string while parsing JSON string");
		836	else
663	ERR ("invalid character encountered");	837	ERR ("invalid character encountered while parsing JSON string");
664		838	}
665	}	839	}
666	while (cur < buf + SHORT_STRING_LEN);	840	while (cur < buf + SHORT_STRING_LEN);
667		841
		842	{
668	STRLEN len = cur - buf;	843	STRLEN len = cur - buf;
669		844
670	if (sv)	845	if (sv)
671	{	846	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	847	SvGROW (sv, SvCUR (sv) + len + 1);
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	848	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	849	SvCUR_set (sv, SvCUR (sv) + len);
675	}	850	}
676	else	851	else
677	sv = newSVpvn (buf, len);	852	sv = newSVpvn (buf, len);
678	}	853	}
		854	}
679	while (*dec->cur != '"');	855	while (*dec_cur != '"');
680		856
681	++dec->cur;	857	++dec_cur;
682		858
683	if (sv)	859	if (sv)
684	{	860	{
685	SvPOK_only (sv);	861	SvPOK_only (sv);
686	*SvEND (sv) = 0;	862	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	865	SvUTF8_on (sv);
690	}	866	}
691	else	867	else
692	sv = newSVpvn ("", 0);	868	sv = newSVpvn ("", 0);
693		869
		870	dec->cur = dec_cur;
694	return sv;	871	return sv;
695		872
696	fail:	873	fail:
		874	dec->cur = dec_cur;
697	return 0;	875	return 0;
698	}	876	}
699		877
700	static SV *	878	static SV *
701	decode_num (dec_t *dec)	879	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	937	is_nv = 1;
760	}	938	}
761		939
762	if (!is_nv)	940	if (!is_nv)
763	{	941	{
764	UV uv;	942	// 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);	943	if (*start == '-')
766	if (numtype & IS_NUMBER_IN_UV)	944	switch (dec->cur - start)
767	if (numtype & IS_NUMBER_NEG)
768	{	945	{
769	if (uv < (UV)IV_MIN)	946	case 2: return newSViv (-( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	947	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		948	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		949	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
771	}	950	}
		951	else
		952	switch (dec->cur - start)
		953	{
		954	case 1: return newSViv ( start [0] - '0' * 1);
		955	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		956	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		957	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		958	}
		959
		960	{
		961	UV uv;
		962	int numtype = grok_number (start, dec->cur - start, &uv);
		963	if (numtype & IS_NUMBER_IN_UV)
		964	if (numtype & IS_NUMBER_NEG)
		965	{
		966	if (uv < (UV)IV_MIN)
		967	return newSViv (-(IV)uv);
		968	}
772	else	969	else
773	return newSVuv (uv);	970	return newSVuv (uv);
		971
		972	// here would likely be the place for bigint support
774	}	973	}
		974	}
775		975
		976	// if we ever support bigint or bigfloat, this is the place for bigfloat
776	return newSVnv (Atof (start));	977	return newSVnv (Atof (start));
777		978
778	fail:	979	fail:
779	return 0;	980	return 0;
780	}	981	}
…		…
782	static SV *	983	static SV *
783	decode_av (dec_t *dec)	984	decode_av (dec_t *dec)
784	{	985	{
785	AV *av = newAV ();	986	AV *av = newAV ();
786		987
		988	DEC_INC_DEPTH;
787	decode_ws (dec);	989	decode_ws (dec);
		990
788	if (*dec->cur == ']')	991	if (*dec->cur == ']')
789	++dec->cur;	992	++dec->cur;
790	else	993	else
791	for (;;)	994	for (;;)
792	{	995	{
…		…
810	ERR (", or ] expected while parsing array");	1013	ERR (", or ] expected while parsing array");
811		1014
812	++dec->cur;	1015	++dec->cur;
813	}	1016	}
814		1017
		1018	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1019	return newRV_noinc ((SV *)av);
816		1020
817	fail:	1021	fail:
818	SvREFCNT_dec (av);	1022	SvREFCNT_dec (av);
		1023	DEC_DEC_DEPTH;
819	return 0;	1024	return 0;
820	}	1025	}
821		1026
822	static SV *	1027	static SV *
823	decode_hv (dec_t *dec)	1028	decode_hv (dec_t *dec)
824	{	1029	{
		1030	SV *sv;
825	HV *hv = newHV ();	1031	HV *hv = newHV ();
826		1032
		1033	DEC_INC_DEPTH;
827	decode_ws (dec);	1034	decode_ws (dec);
		1035
828	if (*dec->cur == '}')	1036	if (*dec->cur == '}')
829	++dec->cur;	1037	++dec->cur;
830	else	1038	else
831	for (;;)	1039	for (;;)
832	{	1040	{
833	SV key, value;
834
835	decode_ws (dec); EXPECT_CH ('"');	1041	decode_ws (dec); EXPECT_CH ('"');
836		1042
837	key = decode_str (dec);	1043	// heuristic: assume that
838	if (!key)	1044	// a) decode_str + hv_store_ent are abysmally slow.
839	goto fail;	1045	// b) most hash keys are short, simple ascii text.
		1046	// => try to "fast-match" such strings to avoid
		1047	// the overhead of decode_str + hv_store_ent.
		1048	{
		1049	SV *value;
		1050	char *p = dec->cur;
		1051	char *e = p + 24; // only try up to 24 bytes
840		1052
841	decode_ws (dec); EXPECT_CH (':');	1053	for (;;)
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1054	{
		1055	// the >= 0x80 is true on most architectures
		1056	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1057	{
		1058	// slow path, back up and use decode_str
		1059	SV *key = decode_str (dec);
		1060	if (!key)
		1061	goto fail;
		1062
		1063	decode_ws (dec); EXPECT_CH (':');
		1064
		1065	value = decode_sv (dec);
		1066	if (!value)
		1067	{
		1068	SvREFCNT_dec (key);
		1069	goto fail;
		1070	}
		1071
		1072	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1073	SvREFCNT_dec (key);
		1074
		1075	break;
		1076	}
		1077	else if (*p == '"')
		1078	{
		1079	// fast path, got a simple key
		1080	char *key = dec->cur;
		1081	int len = p - key;
		1082	dec->cur = p + 1;
		1083
		1084	decode_ws (dec); EXPECT_CH (':');
		1085
		1086	value = decode_sv (dec);
		1087	if (!value)
847	goto fail;	1088	goto fail;
		1089
		1090	hv_store (hv, key, len, value, 0);
		1091
		1092	break;
		1093	}
		1094
		1095	++p;
848	}	1096	}
849		1097	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1098
853	decode_ws (dec);	1099	decode_ws (dec);
854		1100
855	if (*dec->cur == '}')	1101	if (*dec->cur == '}')
856	{	1102	{
…		…
862	ERR (", or } expected while parsing object/hash");	1108	ERR (", or } expected while parsing object/hash");
863		1109
864	++dec->cur;	1110	++dec->cur;
865	}	1111	}
866		1112
		1113	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1114	sv = newRV_noinc ((SV *)hv);
		1115
		1116	// check filter callbacks
		1117	if (dec->json.flags & F_HOOK)
		1118	{
		1119	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1120	{
		1121	HE cb, he;
		1122
		1123	hv_iterinit (hv);
		1124	he = hv_iternext (hv);
		1125	hv_iterinit (hv);
		1126
		1127	// the next line creates a mortal sv each time its called.
		1128	// might want to optimise this for common cases.
		1129	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1130
		1131	if (cb)
		1132	{
		1133	dSP;
		1134	int count;
		1135
		1136	ENTER; SAVETMPS; PUSHMARK (SP);
		1137	XPUSHs (HeVAL (he));
		1138
		1139	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1140
		1141	if (count == 1)
		1142	{
		1143	sv = newSVsv (POPs);
		1144	FREETMPS; LEAVE;
		1145	return sv;
		1146	}
		1147
		1148	FREETMPS; LEAVE;
		1149	}
		1150	}
		1151
		1152	if (dec->json.cb_object)
		1153	{
		1154	dSP;
		1155	int count;
		1156
		1157	ENTER; SAVETMPS; PUSHMARK (SP);
		1158	XPUSHs (sv_2mortal (sv));
		1159
		1160	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1161
		1162	if (count == 1)
		1163	{
		1164	sv = newSVsv (POPs);
		1165	FREETMPS; LEAVE;
		1166	return sv;
		1167	}
		1168
		1169	SvREFCNT_inc (sv);
		1170	FREETMPS; LEAVE;
		1171	}
		1172	}
		1173
		1174	return sv;
868		1175
869	fail:	1176	fail:
870	SvREFCNT_dec (hv);	1177	SvREFCNT_dec (hv);
		1178	DEC_DEC_DEPTH;
871	return 0;	1179	return 0;
872	}	1180	}
873		1181
874	static SV *	1182	static SV *
875	decode_sv (dec_t *dec)	1183	decode_sv (dec_t *dec)
876	{	1184	{
877	decode_ws (dec);	1185	decode_ws (dec);
		1186
		1187	// the beauty of JSON: you need exactly one character lookahead
		1188	// to parse anything.
878	switch (*dec->cur)	1189	switch (*dec->cur)
879	{	1190	{
880	case '"': ++dec->cur; return decode_str (dec);	1191	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1192	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1193	case '{': ++dec->cur; return decode_hv (dec);
…		…
888		1199
889	case 't':	1200	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1201	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1202	{
892	dec->cur += 4;	1203	dec->cur += 4;
893	return newSViv (1);	1204	return SvREFCNT_inc (json_true);
894	}	1205	}
895	else	1206	else
896	ERR ("'true' expected");	1207	ERR ("'true' expected");
897		1208
898	break;	1209	break;
899		1210
900	case 'f':	1211	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1212	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1213	{
903	dec->cur += 5;	1214	dec->cur += 5;
904	return newSViv (0);	1215	return SvREFCNT_inc (json_false);
905	}	1216	}
906	else	1217	else
907	ERR ("'false' expected");	1218	ERR ("'false' expected");
908		1219
909	break;	1220	break;
…		…
918	ERR ("'null' expected");	1229	ERR ("'null' expected");
919		1230
920	break;	1231	break;
921		1232
922	default:	1233	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1234	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1235	break;
925	}	1236	}
926		1237
927	fail:	1238	fail:
928	return 0;	1239	return 0;
929	}	1240	}
930		1241
931	static SV *	1242	static SV *
932	decode_json (SV *string, UV flags)	1243	decode_json (SV string, JSON json, UV *offset_return)
933	{	1244	{
		1245	dec_t dec;
		1246	UV offset;
934	SV *sv;	1247	SV *sv;
935		1248
		1249	SvGETMAGIC (string);
		1250	SvUPGRADE (string, SVt_PV);
		1251
		1252	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1253	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1254	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1255
936	if (flags & F_UTF8)	1256	if (json->flags & F_UTF8)
937	sv_utf8_downgrade (string, 0);	1257	sv_utf8_downgrade (string, 0);
938	else	1258	else
939	sv_utf8_upgrade (string);	1259	sv_utf8_upgrade (string);
940		1260
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1261	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1262
943	dec_t dec;	1263	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1264	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1265	dec.end = SvEND (string);
947	dec.err = 0;	1266	dec.err = 0;
		1267	dec.depth = 0;
		1268	dec.maxdepth = DEC_DEPTH (dec.json.flags);
948		1269
		1270	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1271	dec.json.flags \|= F_HOOK;
		1272
		1273	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
949	sv = decode_sv (&dec);	1274	sv = decode_sv (&dec);
950		1275
		1276	if (!(offset_return \|\| !sv))
		1277	{
		1278	// check for trailing garbage
		1279	decode_ws (&dec);
		1280
		1281	if (*dec.cur)
		1282	{
		1283	dec.err = "garbage after JSON object";
		1284	SvREFCNT_dec (sv);
		1285	sv = 0;
		1286	}
		1287	}
		1288
		1289	if (offset_return \|\| !sv)
		1290	{
		1291	offset = dec.json.flags & F_UTF8
		1292	? dec.cur - SvPVX (string)
		1293	: utf8_distance (dec.cur, SvPVX (string));
		1294
		1295	if (offset_return)
		1296	*offset_return = offset;
		1297	}
		1298
951	if (!sv)	1299	if (!sv)
952	{	1300	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1301	SV *uni = sv_newmortal ();
957		1302
958	// horrible hack to silence warning inside pv_uni_display	1303	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1304	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1305	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1307	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1308	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1309	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1310	LEAVE;
966		1311
967	croak ("%s, at character offset %d (%s)",	1312	croak ("%s, at character offset %d [\"%s\"]",
968	dec.err,	1313	dec.err,
969	(int)offset,	1314	(int)offset,
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1315	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1316	}
972		1317
973	sv = sv_2mortal (sv);	1318	sv = sv_2mortal (sv);
974		1319
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1320	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
976	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1321	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1322
978	return sv;	1323	return sv;
979	}	1324	}
980		1325
…		…
985		1330
986	BOOT:	1331	BOOT:
987	{	1332	{
988	int i;	1333	int i;
989		1334
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1335	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1336	decode_hexdigit [i] =
		1337	i >= '0' && i <= '9' ? i - '0'
		1338	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1339	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1340	: -1;
993		1341
994	for (i = 7; i--; )	1342	json_stash = gv_stashpv ("JSON::XS" , 1);
		1343	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1344
		1345	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1346	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1347	}
		1348
		1349	PROTOTYPES: DISABLE
		1350
		1351	void new (char *klass)
		1352	PPCODE:
		1353	{
		1354	SV *pv = NEWSV (0, sizeof (JSON));
		1355	SvPOK_only (pv);
		1356	Zero (SvPVX (pv), 1, JSON);
		1357	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1358	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1359	}
		1360
		1361	void ascii (JSON *self, int enable = 1)
		1362	ALIAS:
		1363	ascii = F_ASCII
		1364	latin1 = F_LATIN1
		1365	utf8 = F_UTF8
		1366	indent = F_INDENT
		1367	canonical = F_CANONICAL
		1368	space_before = F_SPACE_BEFORE
		1369	space_after = F_SPACE_AFTER
		1370	pretty = F_PRETTY
		1371	allow_nonref = F_ALLOW_NONREF
		1372	shrink = F_SHRINK
		1373	allow_blessed = F_ALLOW_BLESSED
		1374	convert_blessed = F_CONV_BLESSED
		1375	PPCODE:
		1376	{
		1377	if (enable)
		1378	self->flags \|= ix;
		1379	else
		1380	self->flags &= ~ix;
		1381
		1382	XPUSHs (ST (0));
		1383	}
		1384
		1385	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1386	PPCODE:
		1387	{
		1388	UV log2 = 0;
		1389
		1390	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1391
		1392	while ((1UL << log2) < max_depth)
		1393	++log2;
		1394
		1395	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1396
		1397	XPUSHs (ST (0));
		1398	}
		1399
		1400	void max_size (JSON *self, UV max_size = 0)
		1401	PPCODE:
		1402	{
		1403	UV log2 = 0;
		1404
		1405	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1406	if (max_size == 1) max_size = 2;
		1407
		1408	while ((1UL << log2) < max_size)
		1409	++log2;
		1410
		1411	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1412
		1413	XPUSHs (ST (0));
		1414	}
		1415
		1416	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1417	PPCODE:
		1418	{
		1419	SvREFCNT_dec (self->cb_object);
		1420	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1421
		1422	XPUSHs (ST (0));
		1423	}
		1424
		1425	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1426	PPCODE:
		1427	{
		1428	if (!self->cb_sk_object)
		1429	self->cb_sk_object = newHV ();
		1430
		1431	if (SvOK (cb))
		1432	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1433	else
995	{	1434	{
996	decode_hexdigit ['a' + i] = 10 + i;	1435	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
997	decode_hexdigit ['A' + i] = 10 + i;	1436
		1437	if (!HvKEYS (self->cb_sk_object))
		1438	{
		1439	SvREFCNT_dec (self->cb_sk_object);
		1440	self->cb_sk_object = 0;
		1441	}
998	}	1442	}
999		1443
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1444	XPUSHs (ST (0));
1001	}	1445	}
1002		1446
1003	PROTOTYPES: DISABLE	1447	void encode (JSON self, SV scalar)
		1448	PPCODE:
		1449	XPUSHs (encode_json (scalar, self));
1004		1450
1005	SV new (char dummy)	1451	void decode (JSON self, SV jsonstr)
		1452	PPCODE:
		1453	XPUSHs (decode_json (jsonstr, self, 0));
		1454
		1455	void decode_prefix (JSON self, SV jsonstr)
		1456	PPCODE:
		1457	{
		1458	UV offset;
		1459	EXTEND (SP, 2);
		1460	PUSHs (decode_json (jsonstr, self, &offset));
		1461	PUSHs (sv_2mortal (newSVuv (offset)));
		1462	}
		1463
		1464	void DESTROY (JSON *self)
1006	CODE:	1465	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1466	SvREFCNT_dec (self->cb_sk_object);
1008	OUTPUT:	1467	SvREFCNT_dec (self->cb_object);
1009	RETVAL
1010		1468
1011	SV ascii (SV self, int enable = 1)	1469	PROTOTYPES: ENABLE
1012	ALIAS:
1013	ascii = F_ASCII
1014	utf8 = F_UTF8
1015	indent = F_INDENT
1016	canonical = F_CANONICAL
1017	space_before = F_SPACE_BEFORE
1018	space_after = F_SPACE_AFTER
1019	pretty = F_PRETTY
1020	allow_nonref = F_ALLOW_NONREF
1021	shrink = F_SHRINK
1022	CODE:
1023	{
1024	UV *uv = SvJSON (self);
1025	if (enable)
1026	*uv \|= ix;
1027	else
1028	*uv &= ~ix;
1029		1470
1030	RETVAL = newSVsv (self);	1471	void to_json (SV *scalar)
1031	}
1032	OUTPUT:
1033	RETVAL
1034
1035	void encode (SV self, SV scalar)
1036	PPCODE:	1472	PPCODE:
		1473	{
		1474	JSON json = { F_DEFAULT \| F_UTF8 };
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1475	XPUSHs (encode_json (scalar, &json));
		1476	}
1038		1477
1039	void decode (SV self, SV jsonstr)	1478	void from_json (SV *jsonstr)
1040	PPCODE:	1479	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1480	{
1042		1481	JSON json = { F_DEFAULT \| F_UTF8 };
1043	PROTOTYPES: ENABLE
1044
1045	void to_json (SV *scalar)
1046	ALIAS:
1047	objToJson = 0
1048	PPCODE:
1049	XPUSHs (encode_json (scalar, F_UTF8));
1050
1051	void from_json (SV *jsonstr)
1052	ALIAS:
1053	jsonToObj = 0
1054	PPCODE:
1055	XPUSHs (decode_json (jsonstr, F_UTF8));	1482	XPUSHs (decode_json (jsonstr, &json, 0));
		1483	}
1056		1484

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 1.55 by root, Mon Jul 23 22:57:40 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 1.55 by root, Mon Jul 23 22:57:40 2007 UTC