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

Comparing JSON-XS/XS.xs (file contents):
Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs.
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC

…		…
1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <float.h>
8		10
		11	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		12	# define snprintf _snprintf // C compilers have this in stdio.h
		13	#endif
		14
		15	// some old perls do not have this, try to make it work, no
		16	// guarentees, though. if it breaks, you get to keep the pieces.
		17	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
9	#define F_ASCII 0x00000001	21	#define F_ASCII 0x00000001UL
		22	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	23	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	24	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	25	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	26	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	27	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_MAXDEPTH 0xf8000000UL
		33	#define S_MAXDEPTH 27
		34	#define F_MAXSIZE 0x01f00000UL
		35	#define S_MAXSIZE 20
		36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
17		37
18	// F_SKIPINVALID?	38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
19	// F_EXECCODEREF?	39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
20	// F_SELFCONVERT?
21		40
22	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER	41	#define F_PRETTY F_INDENT \| F_SPACE_BEFORE \| F_SPACE_AFTER
23	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
24		43
25	#define INIT_SIZE 32 // initial scalar size to be allocated	44	#define INIT_SIZE 32 // initial scalar size to be allocated
26	#define INDENT_STEP 3 // spaces per indentation level	45	#define INDENT_STEP 3 // spaces per indentation level
27		46
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	47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
30		48
31	#define SB do {	49	#define SB do {
32	#define SE } while (0)	50	#define SE } while (0)
33		51
		52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr),(value))
		54	# define inline inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define inline static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#ifdef USE_ITHREADS
		64	# define JSON_SLOW 1
		65	#else
		66	# define JSON_SLOW 0
		67	#endif
		68
34	static HV *json_stash; // JSON::XS::	69	static HV json_stash, json_boolean_stash; // JSON::XS::
		70	static SV json_true, json_false;
		71
		72	typedef struct {
		73	U32 flags;
		74	SV *cb_object;
		75	HV *cb_sk_object;
		76	} JSON;
35		77
36	/////////////////////////////////////////////////////////////////////////////	78	/////////////////////////////////////////////////////////////////////////////
37	// utility functions	79	// utility functions
38		80
39	static UV *	81	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)	82	shrink (SV *sv)
50	{	83	{
51	sv_utf8_downgrade (sv, 1);	84	sv_utf8_downgrade (sv, 1);
52	if (SvLEN (sv) > SvCUR (sv) + 1)	85	if (SvLEN (sv) > SvCUR (sv) + 1)
53	{	86	{
62	// decode an utf-8 character and return it, or (UV)-1 in	95	// decode an utf-8 character and return it, or (UV)-1 in
63	// case of an error.	96	// case of an error.
64	// we special-case "safe" characters from U+80 .. U+7FF,	97	// we special-case "safe" characters from U+80 .. U+7FF,
65	// but use the very good perl function to parse anything else.	98	// but use the very good perl function to parse anything else.
66	// note that we never call this function for a ascii codepoints	99	// note that we never call this function for a ascii codepoints
67	static UV	100	inline UV
68	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)	101	decode_utf8 (unsigned char s, STRLEN len, STRLEN clen)
69	{	102	{
70	if (s[0] > 0xdf \|\| s[0] < 0xc2)	103	if (expect_false (s[0] > 0xdf \|\| s[0] < 0xc2))
71	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
72	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
73	{	106	{
74	*clen = 2;	107	*clen = 2;
75	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);	108	return ((s[0] & 0x1f) << 6) \| (s[1] & 0x3f);
76	}	109	}
77	else	110	else
		111	{
		112	*clen = (STRLEN)-1;
78	return (UV)-1;	113	return (UV)-1;
		114	}
79	}	115	}
80		116
81	/////////////////////////////////////////////////////////////////////////////	117	/////////////////////////////////////////////////////////////////////////////
82	// encoder	118	// encoder
83		119
…		…
85	typedef struct	121	typedef struct
86	{	122	{
87	char *cur; // SvPVX (sv) + current output position	123	char *cur; // SvPVX (sv) + current output position
88	char *end; // SvEND (sv)	124	char *end; // SvEND (sv)
89	SV *sv; // result scalar	125	SV *sv; // result scalar
90	UV flags; // F_*	126	JSON json;
91	int indent; // indentation level	127	U32 indent; // indentation level
92	int max_depth; // max. recursion level	128	U32 maxdepth; // max. indentation/recursion level
93	} enc_t;	129	} enc_t;
94		130
95	static void	131	inline void
96	need (enc_t *enc, STRLEN len)	132	need (enc_t *enc, STRLEN len)
97	{	133	{
98	if (enc->cur + len >= enc->end)	134	if (expect_false (enc->cur + len >= enc->end))
99	{	135	{
100	STRLEN cur = enc->cur - SvPVX (enc->sv);	136	STRLEN cur = enc->cur - SvPVX (enc->sv);
101	SvGROW (enc->sv, cur + len + 1);	137	SvGROW (enc->sv, cur + len + 1);
102	enc->cur = SvPVX (enc->sv) + cur;	138	enc->cur = SvPVX (enc->sv) + cur;
103	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
104	}	140	}
105	}	141	}
106		142
107	static void	143	inline void
108	encode_ch (enc_t *enc, char ch)	144	encode_ch (enc_t *enc, char ch)
109	{	145	{
110	need (enc, 1);	146	need (enc, 1);
111	*enc->cur++ = ch;	147	*enc->cur++ = ch;
112	}	148	}
…		…
120		156
121	while (str < end)	157	while (str < end)
122	{	158	{
123	unsigned char ch = (unsigned char )str;	159	unsigned char ch = (unsigned char )str;
124		160
125	if (ch >= 0x20 && ch < 0x80) // most common case	161	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
126	{	162	{
127	if (ch == '"') // but with slow exceptions	163	if (expect_false (ch == '"')) // but with slow exceptions
128	{	164	{
129	need (enc, len += 1);	165	need (enc, len += 1);
130	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
131	*enc->cur++ = '"';	167	*enc->cur++ = '"';
132	}	168	}
133	else if (ch == '\\')	169	else if (expect_false (ch == '\\'))
134	{	170	{
135	need (enc, len += 1);	171	need (enc, len += 1);
136	*enc->cur++ = '\\';	172	*enc->cur++ = '\\';
137	*enc->cur++ = '\\';	173	*enc->cur++ = '\\';
138	}	174	}
…		…
156	STRLEN clen;	192	STRLEN clen;
157	UV uch;	193	UV uch;
158		194
159	if (is_utf8)	195	if (is_utf8)
160	{	196	{
161	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
162	uch = decode_utf8 (str, end - str, &clen);	197	uch = decode_utf8 (str, end - str, &clen);
163	if (clen == (STRLEN)-1)	198	if (clen == (STRLEN)-1)
164	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	199	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
165	}	200	}
166	else	201	else
…		…
170	}	205	}
171		206
172	if (uch > 0x10FFFFUL)	207	if (uch > 0x10FFFFUL)
173	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
174		209
175	if (uch < 0x80 \|\| enc->flags & F_ASCII)	210	if (uch < 0x80 \|\| enc->json.flags & F_ASCII \|\| (enc->json.flags & F_LATIN1 && uch > 0xFF))
176	{	211	{
177	if (uch > 0xFFFFUL)	212	if (uch > 0xFFFFUL)
178	{	213	{
179	need (enc, len += 11);	214	need (enc, len += 11);
180	sprintf (enc->cur, "\\u%04x\\u%04x",	215	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
194	*enc->cur++ = hexdigit [(uch >> 0) & 15];	229	*enc->cur++ = hexdigit [(uch >> 0) & 15];
195	}	230	}
196		231
197	str += clen;	232	str += clen;
198	}	233	}
		234	else if (enc->json.flags & F_LATIN1)
		235	{
		236	*enc->cur++ = uch;
		237	str += clen;
		238	}
199	else if (is_utf8)	239	else if (is_utf8)
200	{	240	{
201	need (enc, len += clen);	241	need (enc, len += clen);
202	do	242	do
203	{	243	{
…		…
205	}	245	}
206	while (--clen);	246	while (--clen);
207	}	247	}
208	else	248	else
209	{	249	{
210	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
211	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
212	++str;	252	++str;
213	}	253	}
214	}	254	}
215	}	255	}
…		…
217		257
218	--len;	258	--len;
219	}	259	}
220	}	260	}
221		261
222	static void	262	inline void
223	encode_indent (enc_t *enc)	263	encode_indent (enc_t *enc)
224	{	264	{
225	if (enc->flags & F_INDENT)	265	if (enc->json.flags & F_INDENT)
226	{	266	{
227	int spaces = enc->indent * INDENT_STEP;	267	int spaces = enc->indent * INDENT_STEP;
228		268
229	need (enc, spaces);	269	need (enc, spaces);
230	memset (enc->cur, ' ', spaces);	270	memset (enc->cur, ' ', spaces);
231	enc->cur += spaces;	271	enc->cur += spaces;
232	}	272	}
233	}	273	}
234		274
235	static void	275	inline void
236	encode_space (enc_t *enc)	276	encode_space (enc_t *enc)
237	{	277	{
238	need (enc, 1);	278	need (enc, 1);
239	encode_ch (enc, ' ');	279	encode_ch (enc, ' ');
240	}	280	}
241		281
242	static void	282	inline void
243	encode_nl (enc_t *enc)	283	encode_nl (enc_t *enc)
244	{	284	{
245	if (enc->flags & F_INDENT)	285	if (enc->json.flags & F_INDENT)
246	{	286	{
247	need (enc, 1);	287	need (enc, 1);
248	encode_ch (enc, '\n');	288	encode_ch (enc, '\n');
249	}	289	}
250	}	290	}
251		291
252	static void	292	inline void
253	encode_comma (enc_t *enc)	293	encode_comma (enc_t *enc)
254	{	294	{
255	encode_ch (enc, ',');	295	encode_ch (enc, ',');
256		296
257	if (enc->flags & F_INDENT)	297	if (enc->json.flags & F_INDENT)
258	encode_nl (enc);	298	encode_nl (enc);
259	else if (enc->flags & F_SPACE_AFTER)	299	else if (enc->json.flags & F_SPACE_AFTER)
260	encode_space (enc);	300	encode_space (enc);
261	}	301	}
262		302
263	static void encode_sv (enc_t enc, SV sv);	303	static void encode_sv (enc_t enc, SV sv);
264		304
265	static void	305	static void
266	encode_av (enc_t enc, AV av)	306	encode_av (enc_t enc, AV av)
267	{	307	{
268	int i, len = av_len (av);	308	int i, len = av_len (av);
269		309
		310	if (enc->indent >= enc->maxdepth)
		311	croak ("data structure too deep (hit recursion limit)");
		312
270	encode_ch (enc, '['); encode_nl (enc);	313	encode_ch (enc, '['); encode_nl (enc);
271	++enc->indent;	314	++enc->indent;
272		315
273	for (i = 0; i <= len; ++i)	316	for (i = 0; i <= len; ++i)
274	{	317	{
		318	SV **svp = av_fetch (av, i, 0);
		319
275	encode_indent (enc);	320	encode_indent (enc);
276	encode_sv (enc, *av_fetch (av, i, 0));	321
		322	if (svp)
		323	encode_sv (enc, *svp);
		324	else
		325	encode_str (enc, "null", 4, 0);
277		326
278	if (i < len)	327	if (i < len)
279	encode_comma (enc);	328	encode_comma (enc);
280	}	329	}
281		330
…		…
304	else	353	else
305	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	354	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
306		355
307	encode_ch (enc, '"');	356	encode_ch (enc, '"');
308		357
309	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
310	encode_ch (enc, ':');	359	encode_ch (enc, ':');
311	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
312	encode_sv (enc, HeVAL (he));	361	encode_sv (enc, HeVAL (he));
313	}	362	}
314		363
315	// compare hash entries, used when all keys are bytestrings	364	// compare hash entries, used when all keys are bytestrings
316	static int	365	static int
…		…
339		388
340	static void	389	static void
341	encode_hv (enc_t enc, HV hv)	390	encode_hv (enc_t enc, HV hv)
342	{	391	{
343	int count, i;	392	int count, i;
		393
		394	if (enc->indent >= enc->maxdepth)
		395	croak ("data structure too deep (hit recursion limit)");
344		396
345	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
346		398
347	if ((count = hv_iterinit (hv)))	399	if ((count = hv_iterinit (hv)))
348	{	400	{
349	// for canonical output we have to sort by keys first	401	// for canonical output we have to sort by keys first
350	// actually, this is mostly due to the stupid so-called	402	// actually, this is mostly due to the stupid so-called
351	// security workaround added somewhere in 5.8.x.	403	// security workaround added somewhere in 5.8.x.
352	// that randomises hash orderings	404	// that randomises hash orderings
353	if (enc->flags & F_CANONICAL)	405	if (enc->json.flags & F_CANONICAL)
354	{	406	{
355	HE he, hes [count]; // if your compiler dies here, you need to enable C99 mode
356	int fast = 1;	407	int fast = 1;
		408	HE *he;
		409	#if defined(__BORLANDC__) \|\| defined(_MSC_VER)
		410	HE *hes = _alloca (count sizeof (HE));
		411	#else
		412	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		413	#endif
357		414
358	i = 0;	415	i = 0;
359	while ((he = hv_iternext (hv)))	416	while ((he = hv_iternext (hv)))
360	{	417	{
361	hes [i++] = he;	418	hes [i++] = he;
…		…
396		453
397	encode_nl (enc);	454	encode_nl (enc);
398	}	455	}
399	else	456	else
400	{	457	{
401	SV *sv;
402	HE *he = hv_iternext (hv);	458	HE *he = hv_iternext (hv);
403		459
404	for (;;)	460	for (;;)
405	{	461	{
406	encode_indent (enc);	462	encode_indent (enc);
…		…
415	encode_nl (enc);	471	encode_nl (enc);
416	}	472	}
417	}	473	}
418		474
419	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		476	}
		477
		478	// encode objects, arrays and special \0=false and \1=true values.
		479	static void
		480	encode_rv (enc_t enc, SV sv)
		481	{
		482	svtype svt;
		483
		484	SvGETMAGIC (sv);
		485	svt = SvTYPE (sv);
		486
		487	if (expect_false (SvOBJECT (sv)))
		488	{
		489	HV *stash = !JSON_SLOW \|\| json_boolean_stash
		490	? json_boolean_stash
		491	: gv_stashpv ("JSON::XS::Boolean", 1);
		492
		493	if (SvSTASH (sv) == stash)
		494	{
		495	if (SvIV (sv))
		496	encode_str (enc, "true", 4, 0);
		497	else
		498	encode_str (enc, "false", 5, 0);
		499	}
		500	else
		501	{
		502	#if 0
		503	if (0 && sv_derived_from (rv, "JSON::Literal"))
		504	{
		505	// not yet
		506	}
		507	#endif
		508	if (enc->json.flags & F_CONV_BLESSED)
		509	{
		510	// we re-bless the reference to get overload and other niceties right
		511	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		512
		513	if (to_json)
		514	{
		515	dSP;
		516
		517	ENTER; SAVETMPS; PUSHMARK (SP);
		518	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		519
		520	// calling with G_SCALAR ensures that we always get a 1 return value
		521	PUTBACK;
		522	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		523	SPAGAIN;
		524
		525	// catch this surprisingly common error
		526	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		527	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		528
		529	sv = POPs;
		530	PUTBACK;
		531
		532	encode_sv (enc, sv);
		533
		534	FREETMPS; LEAVE;
		535	}
		536	else if (enc->json.flags & F_ALLOW_BLESSED)
		537	encode_str (enc, "null", 4, 0);
		538	else
		539	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		540	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		541	}
		542	else if (enc->json.flags & F_ALLOW_BLESSED)
		543	encode_str (enc, "null", 4, 0);
		544	else
		545	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		547	}
		548	}
		549	else if (svt == SVt_PVHV)
		550	encode_hv (enc, (HV *)sv);
		551	else if (svt == SVt_PVAV)
		552	encode_av (enc, (AV *)sv);
		553	else if (svt < SVt_PVAV)
		554	{
		555	STRLEN len = 0;
		556	char *pv = svt ? SvPV (sv, len) : 0;
		557
		558	if (len == 1 && *pv == '1')
		559	encode_str (enc, "true", 4, 0);
		560	else if (len == 1 && *pv == '0')
		561	encode_str (enc, "false", 5, 0);
		562	else
		563	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		564	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		565	}
		566	else
		567	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		568	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
420	}	569	}
421		570
422	static void	571	static void
423	encode_sv (enc_t enc, SV sv)	572	encode_sv (enc_t enc, SV sv)
424	{	573	{
…		…
432	encode_str (enc, str, len, SvUTF8 (sv));	581	encode_str (enc, str, len, SvUTF8 (sv));
433	encode_ch (enc, '"');	582	encode_ch (enc, '"');
434	}	583	}
435	else if (SvNOKp (sv))	584	else if (SvNOKp (sv))
436	{	585	{
		586	// trust that perl will do the right thing w.r.t. JSON syntax.
437	need (enc, NV_DIG + 32);	587	need (enc, NV_DIG + 32);
438	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	588	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
439	enc->cur += strlen (enc->cur);	589	enc->cur += strlen (enc->cur);
440	}	590	}
441	else if (SvIOKp (sv))	591	else if (SvIOKp (sv))
442	{	592	{
443	need (enc, 64);	593	// we assume we can always read an IV as a UV
		594	if (SvUV (sv) & ~(UV)0x7fff)
		595	{
		596	// large integer, use the (rather slow) snprintf way.
		597	need (enc, sizeof (UV) * 3);
444	enc->cur +=	598	enc->cur +=
445	SvIsUV(sv)	599	SvIsUV(sv)
446	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	600	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
447	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	601	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		602	}
		603	else
		604	{
		605	// optimise the "small number case"
		606	// code will likely be branchless and use only a single multiplication
		607	I32 i = SvIV (sv);
		608	U32 u;
		609	char digit, nz = 0;
		610
		611	need (enc, 6);
		612
		613	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		614	u = i < 0 ? -i : i;
		615
		616	// convert to 4.28 fixed-point representation
		617	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		618
		619	// now output digit by digit, each time masking out the integer part
		620	// and multiplying by 5 while moving the decimal point one to the right,
		621	// resulting in a net multiplication by 10.
		622	// we always write the digit to memory but conditionally increment
		623	// the pointer, to ease the usage of conditional move instructions.
		624	digit = u >> 28; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0xfffffff) 5;
		625	digit = u >> 27; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x7ffffff) 5;
		626	digit = u >> 26; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x3ffffff) 5;
		627	digit = u >> 25; enc->cur = digit + '0'; enc->cur += (nz = nz \|\| digit); u = (u & 0x1ffffff) 5;
		628	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		629	}
448	}	630	}
449	else if (SvROK (sv))	631	else if (SvROK (sv))
450	{	632	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))	633	else if (!SvOK (sv))
467	encode_str (enc, "null", 4, 0);	634	encode_str (enc, "null", 4, 0);
468	else	635	else
469	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	636	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
470	SvPV_nolen (sv), SvFLAGS (sv));	637	SvPV_nolen (sv), SvFLAGS (sv));
471	}	638	}
472		639
473	static SV *	640	static SV *
474	encode_json (SV *scalar, UV flags)	641	encode_json (SV scalar, JSON json)
475	{	642	{
		643	enc_t enc;
		644
476	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	645	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
477	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	646	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
478		647
479	enc_t enc;	648	enc.json = *json;
480	enc.flags = flags;
481	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	649	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
482	enc.cur = SvPVX (enc.sv);	650	enc.cur = SvPVX (enc.sv);
483	enc.end = SvEND (enc.sv);	651	enc.end = SvEND (enc.sv);
484	enc.indent = 0;	652	enc.indent = 0;
485	enc.max_depth = 0x7fffffffUL;	653	enc.maxdepth = DEC_DEPTH (enc.json.flags);
486		654
487	SvPOK_only (enc.sv);	655	SvPOK_only (enc.sv);
488	encode_sv (&enc, scalar);	656	encode_sv (&enc, scalar);
489		657
		658	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		659	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		660
490	if (!(flags & (F_ASCII \| F_UTF8)))	661	if (!(enc.json.flags & (F_ASCII \| F_LATIN1 \| F_UTF8)))
491	SvUTF8_on (enc.sv);	662	SvUTF8_on (enc.sv);
492		663
493	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
494
495	if (enc.flags & F_SHRINK)	664	if (enc.json.flags & F_SHRINK)
496	shrink (enc.sv);	665	shrink (enc.sv);
497		666
498	return enc.sv;	667	return enc.sv;
499	}	668	}
500		669
…		…
505	typedef struct	674	typedef struct
506	{	675	{
507	char *cur; // current parser pointer	676	char *cur; // current parser pointer
508	char *end; // end of input string	677	char *end; // end of input string
509	const char *err; // parse error, if != 0	678	const char *err; // parse error, if != 0
510	UV flags; // F_*	679	JSON json;
		680	U32 depth; // recursion depth
		681	U32 maxdepth; // recursion depth limit
511	} dec_t;	682	} dec_t;
512		683
513	static void	684	inline void
514	decode_ws (dec_t *dec)	685	decode_ws (dec_t *dec)
515	{	686	{
516	for (;;)	687	for (;;)
517	{	688	{
518	char ch = *dec->cur;	689	char ch = *dec->cur;
…		…
524	++dec->cur;	695	++dec->cur;
525	}	696	}
526	}	697	}
527		698
528	#define ERR(reason) SB dec->err = reason; goto fail; SE	699	#define ERR(reason) SB dec->err = reason; goto fail; SE
		700
529	#define EXPECT_CH(ch) SB \	701	#define EXPECT_CH(ch) SB \
530	if (*dec->cur != ch) \	702	if (*dec->cur != ch) \
531	ERR (# ch " expected"); \	703	ERR (# ch " expected"); \
532	++dec->cur; \	704	++dec->cur; \
533	SE	705	SE
534		706
		707	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		708	#define DEC_DEC_DEPTH --dec->depth
		709
535	static SV decode_sv (dec_t dec);	710	static SV decode_sv (dec_t dec);
536		711
537	static signed char decode_hexdigit[256];	712	static signed char decode_hexdigit[256];
538		713
539	static UV	714	static UV
540	decode_4hex (dec_t *dec)	715	decode_4hex (dec_t *dec)
541	{	716	{
542	signed char d1, d2, d3, d4;	717	signed char d1, d2, d3, d4;
543	unsigned char cur = (unsigned char )dec->cur;	718	unsigned char cur = (unsigned char )dec->cur;
544		719
545	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	720	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");	721	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");	722	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");	723	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
549		724
550	dec->cur += 4;	725	dec->cur += 4;
551		726
552	return ((UV)d1) << 12	727	return ((UV)d1) << 12
553	\| ((UV)d2) << 8	728	\| ((UV)d2) << 8
…		…
561	static SV *	736	static SV *
562	decode_str (dec_t *dec)	737	decode_str (dec_t *dec)
563	{	738	{
564	SV *sv = 0;	739	SV *sv = 0;
565	int utf8 = 0;	740	int utf8 = 0;
		741	char *dec_cur = dec->cur;
566		742
567	do	743	do
568	{	744	{
569	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	745	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
570	char *cur = buf;	746	char *cur = buf;
571		747
572	do	748	do
573	{	749	{
574	unsigned char ch = (unsigned char )dec->cur++;	750	unsigned char ch = (unsigned char )dec_cur++;
575		751
576	if (ch == '"')	752	if (expect_false (ch == '"'))
577	{	753	{
578	--dec->cur;	754	--dec_cur;
579	break;	755	break;
580	}	756	}
581	else if (ch == '\\')	757	else if (expect_false (ch == '\\'))
582	{	758	{
583	switch (*dec->cur)	759	switch (*dec_cur)
584	{	760	{
585	case '\\':	761	case '\\':
586	case '/':	762	case '/':
587	case '"': cur++ = dec->cur++; break;	763	case '"': cur++ = dec_cur++; break;
588		764
589	case 'b': ++dec->cur; *cur++ = '\010'; break;	765	case 'b': ++dec_cur; *cur++ = '\010'; break;
590	case 't': ++dec->cur; *cur++ = '\011'; break;	766	case 't': ++dec_cur; *cur++ = '\011'; break;
591	case 'n': ++dec->cur; *cur++ = '\012'; break;	767	case 'n': ++dec_cur; *cur++ = '\012'; break;
592	case 'f': ++dec->cur; *cur++ = '\014'; break;	768	case 'f': ++dec_cur; *cur++ = '\014'; break;
593	case 'r': ++dec->cur; *cur++ = '\015'; break;	769	case 'r': ++dec_cur; *cur++ = '\015'; break;
594		770
595	case 'u':	771	case 'u':
596	{	772	{
597	UV lo, hi;	773	UV lo, hi;
598	++dec->cur;	774	++dec_cur;
599		775
		776	dec->cur = dec_cur;
600	hi = decode_4hex (dec);	777	hi = decode_4hex (dec);
		778	dec_cur = dec->cur;
601	if (hi == (UV)-1)	779	if (hi == (UV)-1)
602	goto fail;	780	goto fail;
603		781
604	// possibly a surrogate pair	782	// possibly a surrogate pair
605	if (hi >= 0xd800)	783	if (hi >= 0xd800)
606	if (hi < 0xdc00)	784	if (hi < 0xdc00)
607	{	785	{
608	if (dec->cur [0] != '\\' \|\| dec->cur [1] != 'u')	786	if (dec_cur [0] != '\\' \|\| dec_cur [1] != 'u')
609	ERR ("missing low surrogate character in surrogate pair");	787	ERR ("missing low surrogate character in surrogate pair");
610		788
611	dec->cur += 2;	789	dec_cur += 2;
612		790
		791	dec->cur = dec_cur;
613	lo = decode_4hex (dec);	792	lo = decode_4hex (dec);
		793	dec_cur = dec->cur;
614	if (lo == (UV)-1)	794	if (lo == (UV)-1)
615	goto fail;	795	goto fail;
616		796
617	if (lo < 0xdc00 \|\| lo >= 0xe000)	797	if (lo < 0xdc00 \|\| lo >= 0xe000)
618	ERR ("surrogate pair expected");	798	ERR ("surrogate pair expected");
…		…
632	*cur++ = hi;	812	*cur++ = hi;
633	}	813	}
634	break;	814	break;
635		815
636	default:	816	default:
637	--dec->cur;	817	--dec_cur;
638	ERR ("illegal backslash escape sequence in string");	818	ERR ("illegal backslash escape sequence in string");
639	}	819	}
640	}	820	}
641	else if (ch >= 0x20 && ch <= 0x7f)	821	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
642	*cur++ = ch;	822	*cur++ = ch;
643	else if (ch >= 0x80)	823	else if (ch >= 0x80)
644	{	824	{
645	--dec->cur;
646
647	STRLEN clen;	825	STRLEN clen;
		826	UV uch;
		827
		828	--dec_cur;
		829
648	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	830	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
649	if (clen == (STRLEN)-1)	831	if (clen == (STRLEN)-1)
650	ERR ("malformed UTF-8 character in JSON string");	832	ERR ("malformed UTF-8 character in JSON string");
651		833
652	do	834	do
653	{
654	cur++ = dec->cur++;	835	cur++ = dec_cur++;
655	}
656	while (--clen);	836	while (--clen);
657		837
658	utf8 = 1;	838	utf8 = 1;
659	}	839	}
660	else if (!ch)
661	ERR ("unexpected end of string while parsing json string");
662	else	840	else
		841	{
		842	--dec_cur;
		843
		844	if (!ch)
		845	ERR ("unexpected end of string while parsing JSON string");
		846	else
663	ERR ("invalid character encountered");	847	ERR ("invalid character encountered while parsing JSON string");
664		848	}
665	}	849	}
666	while (cur < buf + SHORT_STRING_LEN);	850	while (cur < buf + SHORT_STRING_LEN);
667		851
		852	{
668	STRLEN len = cur - buf;	853	STRLEN len = cur - buf;
669		854
670	if (sv)	855	if (sv)
671	{	856	{
672	SvGROW (sv, SvCUR (sv) + len + 1);	857	SvGROW (sv, SvCUR (sv) + len + 1);
673	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	858	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
674	SvCUR_set (sv, SvCUR (sv) + len);	859	SvCUR_set (sv, SvCUR (sv) + len);
675	}	860	}
676	else	861	else
677	sv = newSVpvn (buf, len);	862	sv = newSVpvn (buf, len);
678	}	863	}
		864	}
679	while (*dec->cur != '"');	865	while (*dec_cur != '"');
680		866
681	++dec->cur;	867	++dec_cur;
682		868
683	if (sv)	869	if (sv)
684	{	870	{
685	SvPOK_only (sv);	871	SvPOK_only (sv);
686	*SvEND (sv) = 0;	872	*SvEND (sv) = 0;
…		…
689	SvUTF8_on (sv);	875	SvUTF8_on (sv);
690	}	876	}
691	else	877	else
692	sv = newSVpvn ("", 0);	878	sv = newSVpvn ("", 0);
693		879
		880	dec->cur = dec_cur;
694	return sv;	881	return sv;
695		882
696	fail:	883	fail:
		884	dec->cur = dec_cur;
697	return 0;	885	return 0;
698	}	886	}
699		887
700	static SV *	888	static SV *
701	decode_num (dec_t *dec)	889	decode_num (dec_t *dec)
…		…
759	is_nv = 1;	947	is_nv = 1;
760	}	948	}
761		949
762	if (!is_nv)	950	if (!is_nv)
763	{	951	{
764	UV uv;	952	int len = dec->cur - start;
765	int numtype = grok_number (start, dec->cur - start, &uv);	953
766	if (numtype & IS_NUMBER_IN_UV)	954	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
767	if (numtype & IS_NUMBER_NEG)	955	if (*start == '-')
		956	switch (len)
768	{	957	{
769	if (uv < (UV)IV_MIN)	958	case 2: return newSViv (-( start [1] - '0' * 1));
770	return newSViv (-(IV)uv);	959	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		960	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		961	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
771	}	962	}
		963	else
		964	switch (len)
		965	{
		966	case 1: return newSViv ( start [0] - '0' * 1);
		967	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		968	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		969	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		970	}
		971
		972	{
		973	UV uv;
		974	int numtype = grok_number (start, len, &uv);
		975	if (numtype & IS_NUMBER_IN_UV)
		976	if (numtype & IS_NUMBER_NEG)
		977	{
		978	if (uv < (UV)IV_MIN)
		979	return newSViv (-(IV)uv);
		980	}
772	else	981	else
773	return newSVuv (uv);	982	return newSVuv (uv);
774	}	983	}
775		984
		985	len -= *start == '-' ? 1 : 0;
		986
		987	// does not fit into IV or UV, try NV
		988	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		989	#if defined (LDBL_DIG)
		990	\|\| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		991	#endif
		992	)
		993	// fits into NV without loss of precision
		994	return newSVnv (Atof (start));
		995
		996	// everything else fails, convert it to a string
		997	return newSVpvn (start, dec->cur - start);
		998	}
		999
		1000	// loss of precision here
776	return newSVnv (Atof (start));	1001	return newSVnv (Atof (start));
777		1002
778	fail:	1003	fail:
779	return 0;	1004	return 0;
780	}	1005	}
…		…
782	static SV *	1007	static SV *
783	decode_av (dec_t *dec)	1008	decode_av (dec_t *dec)
784	{	1009	{
785	AV *av = newAV ();	1010	AV *av = newAV ();
786		1011
		1012	DEC_INC_DEPTH;
787	decode_ws (dec);	1013	decode_ws (dec);
		1014
788	if (*dec->cur == ']')	1015	if (*dec->cur == ']')
789	++dec->cur;	1016	++dec->cur;
790	else	1017	else
791	for (;;)	1018	for (;;)
792	{	1019	{
…		…
810	ERR (", or ] expected while parsing array");	1037	ERR (", or ] expected while parsing array");
811		1038
812	++dec->cur;	1039	++dec->cur;
813	}	1040	}
814		1041
		1042	DEC_DEC_DEPTH;
815	return newRV_noinc ((SV *)av);	1043	return newRV_noinc ((SV *)av);
816		1044
817	fail:	1045	fail:
818	SvREFCNT_dec (av);	1046	SvREFCNT_dec (av);
		1047	DEC_DEC_DEPTH;
819	return 0;	1048	return 0;
820	}	1049	}
821		1050
822	static SV *	1051	static SV *
823	decode_hv (dec_t *dec)	1052	decode_hv (dec_t *dec)
824	{	1053	{
		1054	SV *sv;
825	HV *hv = newHV ();	1055	HV *hv = newHV ();
826		1056
		1057	DEC_INC_DEPTH;
827	decode_ws (dec);	1058	decode_ws (dec);
		1059
828	if (*dec->cur == '}')	1060	if (*dec->cur == '}')
829	++dec->cur;	1061	++dec->cur;
830	else	1062	else
831	for (;;)	1063	for (;;)
832	{	1064	{
833	SV key, value;
834
835	decode_ws (dec); EXPECT_CH ('"');	1065	decode_ws (dec); EXPECT_CH ('"');
836		1066
837	key = decode_str (dec);	1067	// heuristic: assume that
838	if (!key)	1068	// a) decode_str + hv_store_ent are abysmally slow.
839	goto fail;	1069	// b) most hash keys are short, simple ascii text.
		1070	// => try to "fast-match" such strings to avoid
		1071	// the overhead of decode_str + hv_store_ent.
		1072	{
		1073	SV *value;
		1074	char *p = dec->cur;
		1075	char *e = p + 24; // only try up to 24 bytes
840		1076
841	decode_ws (dec); EXPECT_CH (':');	1077	for (;;)
842
843	value = decode_sv (dec);
844	if (!value)
845	{	1078	{
		1079	// the >= 0x80 is true on most architectures
		1080	if (p == e \|\| p < 0x20 \|\| p >= 0x80 \|\| *p == '\\')
		1081	{
		1082	// slow path, back up and use decode_str
		1083	SV *key = decode_str (dec);
		1084	if (!key)
		1085	goto fail;
		1086
		1087	decode_ws (dec); EXPECT_CH (':');
		1088
		1089	value = decode_sv (dec);
		1090	if (!value)
		1091	{
		1092	SvREFCNT_dec (key);
		1093	goto fail;
		1094	}
		1095
		1096	hv_store_ent (hv, key, value, 0);
846	SvREFCNT_dec (key);	1097	SvREFCNT_dec (key);
		1098
		1099	break;
		1100	}
		1101	else if (*p == '"')
		1102	{
		1103	// fast path, got a simple key
		1104	char *key = dec->cur;
		1105	int len = p - key;
		1106	dec->cur = p + 1;
		1107
		1108	decode_ws (dec); EXPECT_CH (':');
		1109
		1110	value = decode_sv (dec);
		1111	if (!value)
847	goto fail;	1112	goto fail;
		1113
		1114	hv_store (hv, key, len, value, 0);
		1115
		1116	break;
		1117	}
		1118
		1119	++p;
848	}	1120	}
849		1121	}
850	//TODO: optimise
851	hv_store_ent (hv, key, value, 0);
852		1122
853	decode_ws (dec);	1123	decode_ws (dec);
854		1124
855	if (*dec->cur == '}')	1125	if (*dec->cur == '}')
856	{	1126	{
…		…
862	ERR (", or } expected while parsing object/hash");	1132	ERR (", or } expected while parsing object/hash");
863		1133
864	++dec->cur;	1134	++dec->cur;
865	}	1135	}
866		1136
		1137	DEC_DEC_DEPTH;
867	return newRV_noinc ((SV *)hv);	1138	sv = newRV_noinc ((SV *)hv);
		1139
		1140	// check filter callbacks
		1141	if (dec->json.flags & F_HOOK)
		1142	{
		1143	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1144	{
		1145	HE cb, he;
		1146
		1147	hv_iterinit (hv);
		1148	he = hv_iternext (hv);
		1149	hv_iterinit (hv);
		1150
		1151	// the next line creates a mortal sv each time its called.
		1152	// might want to optimise this for common cases.
		1153	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1154
		1155	if (cb)
		1156	{
		1157	dSP;
		1158	int count;
		1159
		1160	ENTER; SAVETMPS; PUSHMARK (SP);
		1161	XPUSHs (HeVAL (he));
		1162
		1163	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1164
		1165	if (count == 1)
		1166	{
		1167	sv = newSVsv (POPs);
		1168	FREETMPS; LEAVE;
		1169	return sv;
		1170	}
		1171
		1172	FREETMPS; LEAVE;
		1173	}
		1174	}
		1175
		1176	if (dec->json.cb_object)
		1177	{
		1178	dSP;
		1179	int count;
		1180
		1181	ENTER; SAVETMPS; PUSHMARK (SP);
		1182	XPUSHs (sv_2mortal (sv));
		1183
		1184	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1185
		1186	if (count == 1)
		1187	{
		1188	sv = newSVsv (POPs);
		1189	FREETMPS; LEAVE;
		1190	return sv;
		1191	}
		1192
		1193	SvREFCNT_inc (sv);
		1194	FREETMPS; LEAVE;
		1195	}
		1196	}
		1197
		1198	return sv;
868		1199
869	fail:	1200	fail:
870	SvREFCNT_dec (hv);	1201	SvREFCNT_dec (hv);
		1202	DEC_DEC_DEPTH;
871	return 0;	1203	return 0;
872	}	1204	}
873		1205
874	static SV *	1206	static SV *
875	decode_sv (dec_t *dec)	1207	decode_sv (dec_t *dec)
876	{	1208	{
877	decode_ws (dec);	1209	decode_ws (dec);
		1210
		1211	// the beauty of JSON: you need exactly one character lookahead
		1212	// to parse anything.
878	switch (*dec->cur)	1213	switch (*dec->cur)
879	{	1214	{
880	case '"': ++dec->cur; return decode_str (dec);	1215	case '"': ++dec->cur; return decode_str (dec);
881	case '[': ++dec->cur; return decode_av (dec);	1216	case '[': ++dec->cur; return decode_av (dec);
882	case '{': ++dec->cur; return decode_hv (dec);	1217	case '{': ++dec->cur; return decode_hv (dec);
…		…
888		1223
889	case 't':	1224	case 't':
890	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1225	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
891	{	1226	{
892	dec->cur += 4;	1227	dec->cur += 4;
893	return newSViv (1);	1228	return SvREFCNT_inc (json_true);
894	}	1229	}
895	else	1230	else
896	ERR ("'true' expected");	1231	ERR ("'true' expected");
897		1232
898	break;	1233	break;
899		1234
900	case 'f':	1235	case 'f':
901	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1236	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
902	{	1237	{
903	dec->cur += 5;	1238	dec->cur += 5;
904	return newSViv (0);	1239	return SvREFCNT_inc (json_false);
905	}	1240	}
906	else	1241	else
907	ERR ("'false' expected");	1242	ERR ("'false' expected");
908		1243
909	break;	1244	break;
…		…
918	ERR ("'null' expected");	1253	ERR ("'null' expected");
919		1254
920	break;	1255	break;
921		1256
922	default:	1257	default:
923	ERR ("malformed json string, neither array, object, number, string or atom");	1258	ERR ("malformed JSON string, neither array, object, number, string or atom");
924	break;	1259	break;
925	}	1260	}
926		1261
927	fail:	1262	fail:
928	return 0;	1263	return 0;
929	}	1264	}
930		1265
931	static SV *	1266	static SV *
932	decode_json (SV *string, UV flags)	1267	decode_json (SV string, JSON json, UV *offset_return)
933	{	1268	{
		1269	dec_t dec;
		1270	UV offset;
934	SV *sv;	1271	SV *sv;
935		1272
		1273	SvGETMAGIC (string);
		1274	SvUPGRADE (string, SVt_PV);
		1275
		1276	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1277	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1278	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1279
936	if (flags & F_UTF8)	1280	if (json->flags & F_UTF8)
937	sv_utf8_downgrade (string, 0);	1281	sv_utf8_downgrade (string, 0);
938	else	1282	else
939	sv_utf8_upgrade (string);	1283	sv_utf8_upgrade (string);
940		1284
941	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1285	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
942		1286
943	dec_t dec;	1287	dec.json = *json;
944	dec.flags = flags;
945	dec.cur = SvPVX (string);	1288	dec.cur = SvPVX (string);
946	dec.end = SvEND (string);	1289	dec.end = SvEND (string);
947	dec.err = 0;	1290	dec.err = 0;
		1291	dec.depth = 0;
		1292	dec.maxdepth = DEC_DEPTH (dec.json.flags);
948		1293
		1294	if (dec.json.cb_object \|\| dec.json.cb_sk_object)
		1295	dec.json.flags \|= F_HOOK;
		1296
		1297	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
949	sv = decode_sv (&dec);	1298	sv = decode_sv (&dec);
950		1299
		1300	if (!(offset_return \|\| !sv))
		1301	{
		1302	// check for trailing garbage
		1303	decode_ws (&dec);
		1304
		1305	if (*dec.cur)
		1306	{
		1307	dec.err = "garbage after JSON object";
		1308	SvREFCNT_dec (sv);
		1309	sv = 0;
		1310	}
		1311	}
		1312
		1313	if (offset_return \|\| !sv)
		1314	{
		1315	offset = dec.json.flags & F_UTF8
		1316	? dec.cur - SvPVX (string)
		1317	: utf8_distance (dec.cur, SvPVX (string));
		1318
		1319	if (offset_return)
		1320	*offset_return = offset;
		1321	}
		1322
951	if (!sv)	1323	if (!sv)
952	{	1324	{
953	IV offset = dec.flags & F_UTF8
954	? dec.cur - SvPVX (string)
955	: utf8_distance (dec.cur, SvPVX (string));
956	SV *uni = sv_newmortal ();	1325	SV *uni = sv_newmortal ();
957		1326
958	// horrible hack to silence warning inside pv_uni_display	1327	// horrible hack to silence warning inside pv_uni_display
959	COP cop = *PL_curcop;	1328	COP cop = *PL_curcop;
960	cop.cop_warnings = pWARN_NONE;	1329	cop.cop_warnings = pWARN_NONE;
…		…
962	SAVEVPTR (PL_curcop);	1331	SAVEVPTR (PL_curcop);
963	PL_curcop = &cop;	1332	PL_curcop = &cop;
964	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1333	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
965	LEAVE;	1334	LEAVE;
966		1335
967	croak ("%s, at character offset %d (%s)",	1336	croak ("%s, at character offset %d [\"%s\"]",
968	dec.err,	1337	dec.err,
969	(int)offset,	1338	(int)offset,
970	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1339	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
971	}	1340	}
972		1341
973	sv = sv_2mortal (sv);	1342	sv = sv_2mortal (sv);
974		1343
975	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1344	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)");	1345	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
977		1346
978	return sv;	1347	return sv;
979	}	1348	}
980		1349
…		…
985		1354
986	BOOT:	1355	BOOT:
987	{	1356	{
988	int i;	1357	int i;
989		1358
990	memset (decode_hexdigit, 0xff, 256);
991	for (i = 10; i--; )	1359	for (i = 0; i < 256; ++i)
992	decode_hexdigit ['0' + i] = i;	1360	decode_hexdigit [i] =
		1361	i >= '0' && i <= '9' ? i - '0'
		1362	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1363	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1364	: -1;
993		1365
994	for (i = 7; i--; )	1366	json_stash = gv_stashpv ("JSON::XS" , 1);
		1367	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1368
		1369	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1370	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1371	}
		1372
		1373	PROTOTYPES: DISABLE
		1374
		1375	void CLONE (...)
		1376	CODE:
		1377	json_stash = 0;
		1378	json_boolean_stash = 0;
		1379
		1380	void new (char *klass)
		1381	PPCODE:
		1382	{
		1383	HV *stash = !JSON_SLOW \|\| json_stash
		1384	? json_stash
		1385	: gv_stashpv ("JSON::XS", 1);
		1386	SV *pv = NEWSV (0, sizeof (JSON));
		1387	SvPOK_only (pv);
		1388	Zero (SvPVX (pv), 1, JSON);
		1389	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1390	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));
		1391	}
		1392
		1393	void ascii (JSON *self, int enable = 1)
		1394	ALIAS:
		1395	ascii = F_ASCII
		1396	latin1 = F_LATIN1
		1397	utf8 = F_UTF8
		1398	indent = F_INDENT
		1399	canonical = F_CANONICAL
		1400	space_before = F_SPACE_BEFORE
		1401	space_after = F_SPACE_AFTER
		1402	pretty = F_PRETTY
		1403	allow_nonref = F_ALLOW_NONREF
		1404	shrink = F_SHRINK
		1405	allow_blessed = F_ALLOW_BLESSED
		1406	convert_blessed = F_CONV_BLESSED
		1407	PPCODE:
		1408	{
		1409	if (enable)
		1410	self->flags \|= ix;
		1411	else
		1412	self->flags &= ~ix;
		1413
		1414	XPUSHs (ST (0));
		1415	}
		1416
		1417	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1418	PPCODE:
		1419	{
		1420	UV log2 = 0;
		1421
		1422	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1423
		1424	while ((1UL << log2) < max_depth)
		1425	++log2;
		1426
		1427	self->flags = self->flags & ~F_MAXDEPTH \| (log2 << S_MAXDEPTH);
		1428
		1429	XPUSHs (ST (0));
		1430	}
		1431
		1432	void max_size (JSON *self, UV max_size = 0)
		1433	PPCODE:
		1434	{
		1435	UV log2 = 0;
		1436
		1437	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1438	if (max_size == 1) max_size = 2;
		1439
		1440	while ((1UL << log2) < max_size)
		1441	++log2;
		1442
		1443	self->flags = self->flags & ~F_MAXSIZE \| (log2 << S_MAXSIZE);
		1444
		1445	XPUSHs (ST (0));
		1446	}
		1447
		1448	void filter_json_object (JSON self, SV cb = &PL_sv_undef)
		1449	PPCODE:
		1450	{
		1451	SvREFCNT_dec (self->cb_object);
		1452	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1453
		1454	XPUSHs (ST (0));
		1455	}
		1456
		1457	void filter_json_single_key_object (JSON self, SV key, SV *cb = &PL_sv_undef)
		1458	PPCODE:
		1459	{
		1460	if (!self->cb_sk_object)
		1461	self->cb_sk_object = newHV ();
		1462
		1463	if (SvOK (cb))
		1464	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1465	else
995	{	1466	{
996	decode_hexdigit ['a' + i] = 10 + i;	1467	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
997	decode_hexdigit ['A' + i] = 10 + i;	1468
		1469	if (!HvKEYS (self->cb_sk_object))
		1470	{
		1471	SvREFCNT_dec (self->cb_sk_object);
		1472	self->cb_sk_object = 0;
		1473	}
998	}	1474	}
999		1475
1000	json_stash = gv_stashpv ("JSON::XS", 1);	1476	XPUSHs (ST (0));
1001	}	1477	}
1002		1478
1003	PROTOTYPES: DISABLE	1479	void encode (JSON self, SV scalar)
		1480	PPCODE:
		1481	XPUSHs (encode_json (scalar, self));
1004		1482
1005	SV new (char dummy)	1483	void decode (JSON self, SV jsonstr)
		1484	PPCODE:
		1485	XPUSHs (decode_json (jsonstr, self, 0));
		1486
		1487	void decode_prefix (JSON self, SV jsonstr)
		1488	PPCODE:
		1489	{
		1490	UV offset;
		1491	EXTEND (SP, 2);
		1492	PUSHs (decode_json (jsonstr, self, &offset));
		1493	PUSHs (sv_2mortal (newSVuv (offset)));
		1494	}
		1495
		1496	void DESTROY (JSON *self)
1006	CODE:	1497	CODE:
1007	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1498	SvREFCNT_dec (self->cb_sk_object);
1008	OUTPUT:	1499	SvREFCNT_dec (self->cb_object);
1009	RETVAL
1010		1500
1011	SV ascii (SV self, int enable = 1)	1501	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		1502
1030	RETVAL = newSVsv (self);	1503	void to_json (SV *scalar)
1031	}
1032	OUTPUT:
1033	RETVAL
1034
1035	void encode (SV self, SV scalar)
1036	PPCODE:	1504	PPCODE:
		1505	{
		1506	JSON json = { F_DEFAULT \| F_UTF8 };
1037	XPUSHs (encode_json (scalar, *SvJSON (self)));	1507	XPUSHs (encode_json (scalar, &json));
		1508	}
1038		1509
1039	void decode (SV self, SV jsonstr)	1510	void from_json (SV *jsonstr)
1040	PPCODE:	1511	PPCODE:
1041	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1512	{
1042		1513	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));	1514	XPUSHs (decode_json (jsonstr, &json, 0));
		1515	}
1056		1516

Diff Legend

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

Comparing JSON-XS/XS.xs (file contents): Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs. Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC

Diff Legend

Comparing JSON-XS/XS.xs (file contents):
Revision 1.16 by root, Sun Mar 25 02:37:00 2007 UTC vs.
Revision 1.58 by root, Mon Aug 13 16:06:25 2007 UTC