ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.8 by root, Fri Mar 23 16:13:59 2007 UTC vs.
Revision 1.57 by root, Mon Aug 13 16:05:42 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_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
17	#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
		37
		38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		40
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	41	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		43
22	#define INIT_SIZE 32 // initial scalar size to be allocated	44	#define INIT_SIZE 32 // initial scalar size to be allocated
		45	#define INDENT_STEP 3 // spaces per indentation level
		46
		47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		48
24	#define SB do {	49	#define SB do {
25	#define SE } while (0)	50	#define SE } while (0)
26		51
27	static HV *json_stash;	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 1
		67	#endif
		68
		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;
		77
		78	/////////////////////////////////////////////////////////////////////////////
		79	// utility functions
		80
		81	inline void
		82	shrink (SV *sv)
		83	{
		84	sv_utf8_downgrade (sv, 1);
		85	if (SvLEN (sv) > SvCUR (sv) + 1)
		86	{
		87	#ifdef SvPV_shrink_to_cur
		88	SvPV_shrink_to_cur (sv);
		89	#elif defined (SvPV_renew)
		90	SvPV_renew (sv, SvCUR (sv) + 1);
		91	#endif
		92	}
		93	}
		94
		95	// decode an utf-8 character and return it, or (UV)-1 in
		96	// case of an error.
		97	// we special-case "safe" characters from U+80 .. U+7FF,
		98	// but use the very good perl function to parse anything else.
		99	// note that we never call this function for a ascii codepoints
		100	inline UV
		101	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		102	{
		103	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		106	{
		107	*clen = 2;
		108	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		109	}
		110	else
		111	{
		112	*clen = (STRLEN)-1;
		113	return (UV)-1;
		114	}
		115	}
		116
		117	/////////////////////////////////////////////////////////////////////////////
		118	// encoder
28		119
29	// structure used for encoding JSON	120	// structure used for encoding JSON
30	typedef struct	121	typedef struct
31	{	122	{
32	char *cur;	123	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	124	char *end; // SvEND (sv)
35	SV *sv;	125	SV *sv; // result scalar
36	UV flags;	126	JSON json;
37	int max_recurse;	127	U32 indent; // indentation level
38	int indent;	128	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	129	} enc_t;
40		130
41	// structure used for decoding JSON	131	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	132	need (enc_t *enc, STRLEN len)
72	{	133	{
73	if (enc->cur + len >= enc->end)	134	if (expect_false (enc->cur + len >= enc->end))
74	{	135	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	136	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	137	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	138	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	140	}
80	}	141	}
81		142
82	static void	143	inline void
83	encode_ch (enc_t *enc, char ch)	144	encode_ch (enc_t *enc, char ch)
84	{	145	{
85	need (enc, 1);	146	need (enc, 1);
86	*enc->cur++ = ch;	147	*enc->cur++ = ch;
87	}	148	}
…		…
95		156
96	while (str < end)	157	while (str < end)
97	{	158	{
98	unsigned char ch = *(unsigned char *)str;	159	unsigned char ch = *(unsigned char *)str;
99		160
100	if (ch >= 0x20 && ch < 0x80) // most common case	161	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	162	{
102	if (ch == '"') // but with slow exceptions	163	if (expect_false (ch == '"')) // but with slow exceptions
103	{	164	{
104	need (enc, len += 1);	165	need (enc, len += 1);
105	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	167	*enc->cur++ = '"';
107	}	168	}
108	else if (ch == '\\')	169	else if (expect_false (ch == '\\'))
109	{	170	{
110	need (enc, len += 1);	171	need (enc, len += 1);
111	*enc->cur++ = '\\';	172	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	173	*enc->cur++ = '\\';
113	}	174	}
…		…
131	STRLEN clen;	192	STRLEN clen;
132	UV uch;	193	UV uch;
133		194
134	if (is_utf8)	195	if (is_utf8)
135	{	196	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	197	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	198	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	199	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	200	}
140	else	201	else
141	{	202	{
142	uch = ch;	203	uch = ch;
143	clen = 1;	204	clen = 1;
144	}	205	}
145		206
146	if (uch < 0x80 || enc->flags & F_ASCII)	207	if (uch > 0x10FFFFUL)
		208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		209
		210	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
147	{	211	{
148	if (uch > 0xFFFFUL)	212	if (uch > 0xFFFFUL)
149	{	213	{
150	need (enc, len += 11);	214	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	215	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	216	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	217	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	218	enc->cur += 12;
155	}	219	}
156	else	220	else
157	{	221	{
158	static char hexdigit [16] = "0123456789abcdef";	222	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	229	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	230	}
167		231
168	str += clen;	232	str += clen;
169	}	233	}
		234	else if (enc->json.flags & F_LATIN1)
		235	{
		236	*enc->cur++ = uch;
		237	str += clen;
		238	}
170	else if (is_utf8)	239	else if (is_utf8)
171	{	240	{
172	need (enc, len += clen);	241	need (enc, len += clen);
173	do	242	do
174	{	243	{
…		…
176	}	245	}
177	while (--clen);	246	while (--clen);
178	}	247	}
179	else	248	else
180	{	249	{
181	need (enc, len += 10); // never more than 11 bytes needed	250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	252	++str;
184	}	253	}
185	}	254	}
186	}	255	}
…		…
188		257
189	--len;	258	--len;
190	}	259	}
191	}	260	}
192		261
193	#define INDENT SB \	262	inline void
		263	encode_indent (enc_t *enc)
		264	{
194	if (enc->flags & F_INDENT) \	265	if (enc->json.flags & F_INDENT)
195	{ \	266	{
196	int i_; \	267	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	268
198	for (i_ = enc->indent * 3; i_--; )\	269	need (enc, spaces);
		270	memset (enc->cur, ' ', spaces);
		271	enc->cur += spaces;
		272	}
		273	}
		274
		275	inline void
		276	encode_space (enc_t *enc)
		277	{
		278	need (enc, 1);
		279	encode_ch (enc, ' ');
		280	}
		281
		282	inline void
		283	encode_nl (enc_t *enc)
		284	{
		285	if (enc->json.flags & F_INDENT)
		286	{
		287	need (enc, 1);
199	encode_ch (enc, ' '); \	288	encode_ch (enc, '\n');
200	} \	289	}
201	SE	290	}
202		291
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	292	inline void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	293	encode_comma (enc_t *enc)
205	#define COMMA SB \	294	{
206	encode_ch (enc, ','); \	295	encode_ch (enc, ',');
		296
207	if (enc->flags & F_INDENT) \	297	if (enc->json.flags & F_INDENT)
208	NL; \	298	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	299	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	300	encode_space (enc);
211	SE	301	}
212		302
213	static void encode_sv (enc_t *enc, SV *sv);	303	static void encode_sv (enc_t *enc, SV *sv);
214		304
215	static void	305	static void
216	encode_av (enc_t *enc, AV *av)	306	encode_av (enc_t *enc, AV *av)
217	{	307	{
218	int i, len = av_len (av);	308	int i, len = av_len (av);
219		309
220	encode_ch (enc, '['); NL;	310	if (enc->indent >= enc->maxdepth)
		311	croak ("data structure too deep (hit recursion limit)");
		312
		313	encode_ch (enc, '['); encode_nl (enc);
221	++enc->indent;	314	++enc->indent;
222		315
223	for (i = 0; i <= len; ++i)	316	for (i = 0; i <= len; ++i)
224	{	317	{
225	INDENT;	318	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));	319
		320	encode_indent (enc);
		321
		322	if (svp)
		323	encode_sv (enc, *svp);
		324	else
		325	encode_str (enc, "null", 4, 0);
227		326
228	if (i < len)	327	if (i < len)
229	COMMA;	328	encode_comma (enc);
230	}	329	}
231		330
232	NL;	331	encode_nl (enc);
233		332
234	--enc->indent;	333	--enc->indent;
235	INDENT; encode_ch (enc, ']');	334	encode_indent (enc); encode_ch (enc, ']');
236	}	335	}
237		336
238	static void	337	static void
239	encode_he (enc_t *enc, HE *he)	338	encode_he (enc_t *enc, HE *he)
240	{	339	{
…		…
254	else	353	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	354	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		355
257	encode_ch (enc, '"');	356	encode_ch (enc, '"');
258		357
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	359	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));	361	encode_sv (enc, HeVAL (he));
263	}	362	}
264		363
265	// compare hash entries, used when all keys are bytestrings	364	// compare hash entries, used when all keys are bytestrings
266	static int	365	static int
…		…
272	HE *b = *(HE **)b_;	371	HE *b = *(HE **)b_;
273		372
274	STRLEN la = HeKLEN (a);	373	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	374	STRLEN lb = HeKLEN (b);
276		375
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	376	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	377	cmp = la - lb;
279		378
280	return cmp;	379	return cmp;
281	}	380	}
282		381
283	// compare hash entries, used when some keys are sv's or utf-x	382	// compare hash entries, used when some keys are sv's or utf-x
…		…
290	static void	389	static void
291	encode_hv (enc_t *enc, HV *hv)	390	encode_hv (enc_t *enc, HV *hv)
292	{	391	{
293	int count, i;	392	int count, i;
294		393
		394	if (enc->indent >= enc->maxdepth)
		395	croak ("data structure too deep (hit recursion limit)");
		396
295	encode_ch (enc, '{'); NL; ++enc->indent;	397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
296		398
297	if ((count = hv_iterinit (hv)))	399	if ((count = hv_iterinit (hv)))
298	{	400	{
299	// for canonical output we have to sort by keys first	401	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	402	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	403	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	404	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	405	if (enc->json.flags & F_CANONICAL)
304	{	406	{
305	HE *he, *hes [count];
306	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
307		414
308	i = 0;	415	i = 0;
309	while ((he = hv_iternext (hv)))	416	while ((he = hv_iternext (hv)))
310	{	417	{
311	hes [i++] = he;	418	hes [i++] = he;
…		…
335	LEAVE;	442	LEAVE;
336	}	443	}
337		444
338	for (i = 0; i < count; ++i)	445	for (i = 0; i < count; ++i)
339	{	446	{
340	INDENT;	447	encode_indent (enc);
341	encode_he (enc, hes [i]);	448	encode_he (enc, hes [i]);
342		449
343	if (i < count - 1)	450	if (i < count - 1)
344	COMMA;	451	encode_comma (enc);
345	}
346
347	NL;	452	}
		453
		454	encode_nl (enc);
348	}	455	}
349	else	456	else
350	{	457	{
351	SV *sv;
352	HE *he = hv_iternext (hv);	458	HE *he = hv_iternext (hv);
353		459
354	for (;;)	460	for (;;)
355	{	461	{
356	INDENT;	462	encode_indent (enc);
357	encode_he (enc, he);	463	encode_he (enc, he);
358		464
359	if (!(he = hv_iternext (hv)))	465	if (!(he = hv_iternext (hv)))
360	break;	466	break;
361		467
362	COMMA;	468	encode_comma (enc);
363	}
364
365	NL;	469	}
		470
		471	encode_nl (enc);
366	}	472	}
367	}	473	}
368		474
369	--enc->indent; INDENT; 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	int count;
		516	dSP;
		517
		518	ENTER; SAVETMPS; PUSHMARK (SP);
		519	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		520
		521	// calling with G_SCALAR ensures that we always get a 1 return value
		522	PUTBACK;
		523	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		524	SPAGAIN;
		525
		526	// catch this surprisingly common error
		527	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		528	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		529
		530	sv = POPs;
		531	PUTBACK;
		532
		533	encode_sv (enc, sv);
		534
		535	FREETMPS; LEAVE;
		536	}
		537	else if (enc->json.flags & F_ALLOW_BLESSED)
		538	encode_str (enc, "null", 4, 0);
		539	else
		540	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		541	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		542	}
		543	else if (enc->json.flags & F_ALLOW_BLESSED)
		544	encode_str (enc, "null", 4, 0);
		545	else
		546	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		547	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		548	}
		549	}
		550	else if (svt == SVt_PVHV)
		551	encode_hv (enc, (HV *)sv);
		552	else if (svt == SVt_PVAV)
		553	encode_av (enc, (AV *)sv);
		554	else if (svt < SVt_PVAV)
		555	{
		556	STRLEN len = 0;
		557	char *pv = svt ? SvPV (sv, len) : 0;
		558
		559	if (len == 1 && *pv == '1')
		560	encode_str (enc, "true", 4, 0);
		561	else if (len == 1 && *pv == '0')
		562	encode_str (enc, "false", 5, 0);
		563	else
		564	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		565	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		566	}
		567	else
		568	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		569	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
370	}	570	}
371		571
372	static void	572	static void
373	encode_sv (enc_t *enc, SV *sv)	573	encode_sv (enc_t *enc, SV *sv)
374	{	574	{
…		…
382	encode_str (enc, str, len, SvUTF8 (sv));	582	encode_str (enc, str, len, SvUTF8 (sv));
383	encode_ch (enc, '"');	583	encode_ch (enc, '"');
384	}	584	}
385	else if (SvNOKp (sv))	585	else if (SvNOKp (sv))
386	{	586	{
		587	// trust that perl will do the right thing w.r.t. JSON syntax.
387	need (enc, NV_DIG + 32);	588	need (enc, NV_DIG + 32);
388	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	589	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
389	enc->cur += strlen (enc->cur);	590	enc->cur += strlen (enc->cur);
390	}	591	}
391	else if (SvIOKp (sv))	592	else if (SvIOKp (sv))
392	{	593	{
393	need (enc, 64);	594	// we assume we can always read an IV as a UV
		595	if (SvUV (sv) & ~(UV)0x7fff)
		596	{
		597	// large integer, use the (rather slow) snprintf way.
		598	need (enc, sizeof (UV) * 3);
394	enc->cur +=	599	enc->cur +=
395	SvIsUV(sv)	600	SvIsUV(sv)
396	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	601	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
397	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	602	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		603	}
		604	else
		605	{
		606	// optimise the "small number case"
		607	// code will likely be branchless and use only a single multiplication
		608	I32 i = SvIV (sv);
		609	U32 u;
		610	char digit, nz = 0;
		611
		612	need (enc, 6);
		613
		614	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		615	u = i < 0 ? -i : i;
		616
		617	// convert to 4.28 fixed-point representation
		618	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		619
		620	// now output digit by digit, each time masking out the integer part
		621	// and multiplying by 5 while moving the decimal point one to the right,
		622	// resulting in a net multiplication by 10.
		623	// we always write the digit to memory but conditionally increment
		624	// the pointer, to ease the usage of conditional move instructions.
		625	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		626	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		627	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		628	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		629	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		630	}
398	}	631	}
399	else if (SvROK (sv))	632	else if (SvROK (sv))
400	{	633	encode_rv (enc, SvRV (sv));
401	if (!--enc->max_recurse)
402	croak ("data structure too deep (hit recursion limit)");
403
404	sv = SvRV (sv);
405
406	switch (SvTYPE (sv))
407	{
408	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
409	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
410
411	default:
412	croak ("JSON can only represent references to arrays or hashes");
413	}
414	}
415	else if (!SvOK (sv))	634	else if (!SvOK (sv))
416	encode_str (enc, "null", 4, 0);	635	encode_str (enc, "null", 4, 0);
417	else	636	else
418	croak ("encountered perl type that JSON cannot handle");	637	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		638	SvPV_nolen (sv), SvFLAGS (sv));
419	}	639	}
420		640
421	static SV *	641	static SV *
422	encode_json (SV *scalar, UV flags)	642	encode_json (SV *scalar, JSON *json)
423	{	643	{
424	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
425	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
426
427	enc_t enc;	644	enc_t enc;
428	enc.flags = flags;	645
		646	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		647	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		648
		649	enc.json = *json;
429	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	650	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
430	enc.cur = SvPVX (enc.sv);	651	enc.cur = SvPVX (enc.sv);
431	enc.end = SvEND (enc.sv);	652	enc.end = SvEND (enc.sv);
432	enc.max_recurse = 0;
433	enc.indent = 0;	653	enc.indent = 0;
		654	enc.maxdepth = DEC_DEPTH (enc.json.flags);
434		655
435	SvPOK_only (enc.sv);	656	SvPOK_only (enc.sv);
436	encode_sv (&enc, scalar);	657	encode_sv (&enc, scalar);
437		658
		659	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		660	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		661
438	if (!(flags & (F_ASCII | F_UTF8)))	662	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
439	SvUTF8_on (enc.sv);	663	SvUTF8_on (enc.sv);
440		664
441	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
442
443	if (enc.flags & F_SHRINK)	665	if (enc.json.flags & F_SHRINK)
444	shrink (enc.sv);	666	shrink (enc.sv);
445		667
446	return enc.sv;	668	return enc.sv;
447	}	669	}
448		670
449	/////////////////////////////////////////////////////////////////////////////	671	/////////////////////////////////////////////////////////////////////////////
		672	// decoder
450		673
451	#define WS \	674	// structure used for decoding JSON
		675	typedef struct
		676	{
		677	char *cur; // current parser pointer
		678	char *end; // end of input string
		679	const char *err; // parse error, if != 0
		680	JSON json;
		681	U32 depth; // recursion depth
		682	U32 maxdepth; // recursion depth limit
		683	} dec_t;
		684
		685	inline void
		686	decode_ws (dec_t *dec)
		687	{
452	for (;;) \	688	for (;;)
453	{ \	689	{
454	char ch = *dec->cur; \	690	char ch = *dec->cur;
		691
455	if (ch > 0x20 \	692	if (ch > 0x20
456	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	693	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
457	break; \	694	break;
		695
458	++dec->cur; \	696	++dec->cur;
459	}	697	}
		698	}
460		699
461	#define ERR(reason) SB dec->err = reason; goto fail; SE	700	#define ERR(reason) SB dec->err = reason; goto fail; SE
		701
462	#define EXPECT_CH(ch) SB \	702	#define EXPECT_CH(ch) SB \
463	if (*dec->cur != ch) \	703	if (*dec->cur != ch) \
464	ERR (# ch " expected"); \	704	ERR (# ch " expected"); \
465	++dec->cur; \	705	++dec->cur; \
466	SE	706	SE
467		707
		708	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		709	#define DEC_DEC_DEPTH --dec->depth
		710
468	static SV *decode_sv (dec_t *dec);	711	static SV *decode_sv (dec_t *dec);
469		712
470	static signed char decode_hexdigit[256];	713	static signed char decode_hexdigit[256];
471		714
472	static UV	715	static UV
473	decode_4hex (dec_t *dec)	716	decode_4hex (dec_t *dec)
474	{	717	{
475	signed char d1, d2, d3, d4;	718	signed char d1, d2, d3, d4;
		719	unsigned char *cur = (unsigned char *)dec->cur;
476		720
477	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	721	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
478	if (d1 < 0) ERR ("four hexadecimal digits expected");	722	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
479	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	723	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
480	if (d2 < 0) ERR ("four hexadecimal digits expected");	724	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
481	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
482	if (d3 < 0) ERR ("four hexadecimal digits expected");
483	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
484	if (d4 < 0) ERR ("four hexadecimal digits expected");
485		725
486	dec->cur += 4;	726	dec->cur += 4;
487		727
488	return ((UV)d1) << 12	728	return ((UV)d1) << 12
489	| ((UV)d2) << 8	729	| ((UV)d2) << 8
…		…
492		732
493	fail:	733	fail:
494	return (UV)-1;	734	return (UV)-1;
495	}	735	}
496		736
497	#define APPEND_GROW(n) SB \
498	if (cur + (n) >= end) \
499	{ \
500	STRLEN ofs = cur - SvPVX (sv); \
501	SvGROW (sv, ofs + (n) + 1); \
502	cur = SvPVX (sv) + ofs; \
503	end = SvEND (sv); \
504	} \
505	SE
506
507	#define APPEND_CH(ch) SB \
508	APPEND_GROW (1); \
509	*cur++ = (ch); \
510	SE
511
512	static SV *	737	static SV *
513	decode_str (dec_t *dec)	738	decode_str (dec_t *dec)
514	{	739	{
515	SV *sv = NEWSV (0,2);	740	SV *sv = 0;
516	int utf8 = 0;	741	int utf8 = 0;
517	char *cur = SvPVX (sv);	742	char *dec_cur = dec->cur;
518	char *end = SvEND (sv);
519		743
520	for (;;)	744	do
521	{	745	{
522	unsigned char ch = *(unsigned char *)dec->cur;	746	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		747	char *cur = buf;
523		748
524	if (ch == '"')	749	do
525	break;
526	else if (ch == '\\')
527	{	750	{
528	switch (*++dec->cur)	751	unsigned char ch = *(unsigned char *)dec_cur++;
		752
		753	if (expect_false (ch == '"'))
		754	{
		755	--dec_cur;
		756	break;
529	{	757	}
530	case '\\':	758	else if (expect_false (ch == '\\'))
531	case '/':	759	{
532	case '"': APPEND_CH (*dec->cur++); break;	760	switch (*dec_cur)
533
534	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
535	case 't': APPEND_CH ('\011'); ++dec->cur; break;
536	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
537	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
538	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
539
540	case 'u':
541	{	761	{
542	UV lo, hi;	762	case '\\':
543	++dec->cur;	763	case '/':
		764	case '"': *cur++ = *dec_cur++; break;
544		765
545	hi = decode_4hex (dec);	766	case 'b': ++dec_cur; *cur++ = '\010'; break;
546	if (hi == (UV)-1)	767	case 't': ++dec_cur; *cur++ = '\011'; break;
547	goto fail;	768	case 'n': ++dec_cur; *cur++ = '\012'; break;
		769	case 'f': ++dec_cur; *cur++ = '\014'; break;
		770	case 'r': ++dec_cur; *cur++ = '\015'; break;
548		771
549	// possibly a surrogate pair	772	case 'u':
550	if (hi >= 0xd800 && hi < 0xdc00)
551	{	773	{
552	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	774	UV lo, hi;
553	ERR ("missing low surrogate character in surrogate pair");	775	++dec_cur;
554		776
555	dec->cur += 2;	777	dec->cur = dec_cur;
556
557	lo = decode_4hex (dec);	778	hi = decode_4hex (dec);
		779	dec_cur = dec->cur;
558	if (lo == (UV)-1)	780	if (hi == (UV)-1)
559	goto fail;	781	goto fail;
560		782
		783	// possibly a surrogate pair
		784	if (hi >= 0xd800)
		785	if (hi < 0xdc00)
		786	{
		787	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		788	ERR ("missing low surrogate character in surrogate pair");
		789
		790	dec_cur += 2;
		791
		792	dec->cur = dec_cur;
		793	lo = decode_4hex (dec);
		794	dec_cur = dec->cur;
		795	if (lo == (UV)-1)
		796	goto fail;
		797
561	if (lo < 0xdc00 || lo >= 0xe000)	798	if (lo < 0xdc00 || lo >= 0xe000)
562	ERR ("surrogate pair expected");	799	ERR ("surrogate pair expected");
563		800
564	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	801	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		802	}
		803	else if (hi < 0xe000)
		804	ERR ("missing high surrogate character in surrogate pair");
		805
		806	if (hi >= 0x80)
		807	{
		808	utf8 = 1;
		809
		810	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		811	}
		812	else
		813	*cur++ = hi;
565	}	814	}
566	else if (hi >= 0xdc00 && hi < 0xe000)
567	ERR ("missing high surrogate character in surrogate pair");
568
569	if (hi >= 0x80)
570	{	815	break;
571	utf8 = 1;
572		816
573	APPEND_GROW (4); // at most 4 bytes for 21 bits
574	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
575	}
576	else	817	default:
577	APPEND_CH (hi);	818	--dec_cur;
		819	ERR ("illegal backslash escape sequence in string");
578	}	820	}
579	break;	821	}
		822	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		823	*cur++ = ch;
		824	else if (ch >= 0x80)
		825	{
		826	STRLEN clen;
		827	UV uch;
580		828
581	default:
582	--dec->cur;	829	--dec_cur;
583	ERR ("illegal backslash escape sequence in string");	830
		831	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		832	if (clen == (STRLEN)-1)
		833	ERR ("malformed UTF-8 character in JSON string");
		834
		835	do
		836	*cur++ = *dec_cur++;
		837	while (--clen);
		838
		839	utf8 = 1;
		840	}
		841	else
		842	{
		843	--dec_cur;
		844
		845	if (!ch)
		846	ERR ("unexpected end of string while parsing JSON string");
		847	else
		848	ERR ("invalid character encountered while parsing JSON string");
584	}	849	}
585	}	850	}
586	else if (ch >= 0x20 && ch <= 0x7f)	851	while (cur < buf + SHORT_STRING_LEN);
587	APPEND_CH (*dec->cur++);	852
588	else if (ch >= 0x80)
589	{	853	{
590	STRLEN clen;	854	STRLEN len = cur - buf;
591	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
592	if (clen == (STRLEN)-1)
593	ERR ("malformed UTF-8 character in JSON string");
594		855
595	APPEND_GROW (clen);	856	if (sv)
596	do
597	{	857	{
598	*cur++ = *dec->cur++;	858	SvGROW (sv, SvCUR (sv) + len + 1);
		859	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		860	SvCUR_set (sv, SvCUR (sv) + len);
599	}	861	}
600	while (--clen);
601
602	utf8 = 1;
603	}
604	else if (dec->cur == dec->end)
605	ERR ("unexpected end of string while parsing json string");
606	else	862	else
607	ERR ("invalid character encountered");	863	sv = newSVpvn (buf, len);
608	}	864	}
		865	}
		866	while (*dec_cur != '"');
609		867
610	++dec->cur;	868	++dec_cur;
611		869
612	SvCUR_set (sv, cur - SvPVX (sv));	870	if (sv)
613		871	{
614	SvPOK_only (sv);	872	SvPOK_only (sv);
615	*SvEND (sv) = 0;	873	*SvEND (sv) = 0;
616		874
617	if (utf8)	875	if (utf8)
618	SvUTF8_on (sv);	876	SvUTF8_on (sv);
		877	}
		878	else
		879	sv = newSVpvn ("", 0);
619		880
620	if (dec->flags & F_SHRINK)	881	dec->cur = dec_cur;
621	shrink (sv);
622
623	return sv;	882	return sv;
624		883
625	fail:	884	fail:
626	SvREFCNT_dec (sv);	885	dec->cur = dec_cur;
627	return 0;	886	return 0;
628	}	887	}
629		888
630	static SV *	889	static SV *
631	decode_num (dec_t *dec)	890	decode_num (dec_t *dec)
…		…
689	is_nv = 1;	948	is_nv = 1;
690	}	949	}
691		950
692	if (!is_nv)	951	if (!is_nv)
693	{	952	{
694	UV uv;	953	int len = dec->cur - start;
695	int numtype = grok_number (start, dec->cur - start, &uv);	954
696	if (numtype & IS_NUMBER_IN_UV)	955	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
697	if (numtype & IS_NUMBER_NEG)	956	if (*start == '-')
		957	switch (len)
698	{	958	{
699	if (uv < (UV)IV_MIN)	959	case 2: return newSViv (-( start [1] - '0' * 1));
700	return newSViv (-(IV)uv);	960	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		961	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		962	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
701	}	963	}
		964	else
		965	switch (len)
		966	{
		967	case 1: return newSViv ( start [0] - '0' * 1);
		968	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		969	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		970	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		971	}
		972
		973	{
		974	UV uv;
		975	int numtype = grok_number (start, len, &uv);
		976	if (numtype & IS_NUMBER_IN_UV)
		977	if (numtype & IS_NUMBER_NEG)
		978	{
		979	if (uv < (UV)IV_MIN)
		980	return newSViv (-(IV)uv);
		981	}
702	else	982	else
703	return newSVuv (uv);	983	return newSVuv (uv);
704	}	984	}
705		985
		986	len -= *start == '-' ? 1 : 0;
		987
		988	// does not fit into IV or UV, try NV
		989	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		990	#if defined (LDBL_DIG)
		991	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		992	#endif
		993	)
		994	// fits into NV without loss of precision
		995	return newSVnv (Atof (start));
		996
		997	// everything else fails, convert it to a string
		998	return newSVpvn (start, dec->cur - start);
		999	}
		1000
		1001	// loss of precision here
706	return newSVnv (Atof (start));	1002	return newSVnv (Atof (start));
707		1003
708	fail:	1004	fail:
709	return 0;	1005	return 0;
710	}	1006	}
…		…
712	static SV *	1008	static SV *
713	decode_av (dec_t *dec)	1009	decode_av (dec_t *dec)
714	{	1010	{
715	AV *av = newAV ();	1011	AV *av = newAV ();
716		1012
717	WS;	1013	DEC_INC_DEPTH;
		1014	decode_ws (dec);
		1015
718	if (*dec->cur == ']')	1016	if (*dec->cur == ']')
719	++dec->cur;	1017	++dec->cur;
720	else	1018	else
721	for (;;)	1019	for (;;)
722	{	1020	{
…		…
726	if (!value)	1024	if (!value)
727	goto fail;	1025	goto fail;
728		1026
729	av_push (av, value);	1027	av_push (av, value);
730		1028
731	WS;	1029	decode_ws (dec);
732		1030
733	if (*dec->cur == ']')	1031	if (*dec->cur == ']')
734	{	1032	{
735	++dec->cur;	1033	++dec->cur;
736	break;	1034	break;
…		…
740	ERR (", or ] expected while parsing array");	1038	ERR (", or ] expected while parsing array");
741		1039
742	++dec->cur;	1040	++dec->cur;
743	}	1041	}
744		1042
		1043	DEC_DEC_DEPTH;
745	return newRV_noinc ((SV *)av);	1044	return newRV_noinc ((SV *)av);
746		1045
747	fail:	1046	fail:
748	SvREFCNT_dec (av);	1047	SvREFCNT_dec (av);
		1048	DEC_DEC_DEPTH;
749	return 0;	1049	return 0;
750	}	1050	}
751		1051
752	static SV *	1052	static SV *
753	decode_hv (dec_t *dec)	1053	decode_hv (dec_t *dec)
754	{	1054	{
		1055	SV *sv;
755	HV *hv = newHV ();	1056	HV *hv = newHV ();
756		1057
757	WS;	1058	DEC_INC_DEPTH;
		1059	decode_ws (dec);
		1060
758	if (*dec->cur == '}')	1061	if (*dec->cur == '}')
759	++dec->cur;	1062	++dec->cur;
760	else	1063	else
761	for (;;)	1064	for (;;)
762	{	1065	{
		1066	decode_ws (dec); EXPECT_CH ('"');
		1067
		1068	// heuristic: assume that
		1069	// a) decode_str + hv_store_ent are abysmally slow.
		1070	// b) most hash keys are short, simple ascii text.
		1071	// => try to "fast-match" such strings to avoid
		1072	// the overhead of decode_str + hv_store_ent.
		1073	{
763	SV *key, *value;	1074	SV *value;
		1075	char *p = dec->cur;
		1076	char *e = p + 24; // only try up to 24 bytes
764		1077
765	WS; EXPECT_CH ('"');	1078	for (;;)
766
767	key = decode_str (dec);
768	if (!key)
769	goto fail;
770
771	WS; EXPECT_CH (':');
772
773	value = decode_sv (dec);
774	if (!value)
775	{	1079	{
		1080	// the >= 0x80 is true on most architectures
		1081	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1082	{
		1083	// slow path, back up and use decode_str
		1084	SV *key = decode_str (dec);
		1085	if (!key)
		1086	goto fail;
		1087
		1088	decode_ws (dec); EXPECT_CH (':');
		1089
		1090	value = decode_sv (dec);
		1091	if (!value)
		1092	{
		1093	SvREFCNT_dec (key);
		1094	goto fail;
		1095	}
		1096
		1097	hv_store_ent (hv, key, value, 0);
776	SvREFCNT_dec (key);	1098	SvREFCNT_dec (key);
		1099
		1100	break;
		1101	}
		1102	else if (*p == '"')
		1103	{
		1104	// fast path, got a simple key
		1105	char *key = dec->cur;
		1106	int len = p - key;
		1107	dec->cur = p + 1;
		1108
		1109	decode_ws (dec); EXPECT_CH (':');
		1110
		1111	value = decode_sv (dec);
		1112	if (!value)
777	goto fail;	1113	goto fail;
		1114
		1115	hv_store (hv, key, len, value, 0);
		1116
		1117	break;
		1118	}
		1119
		1120	++p;
778	}	1121	}
779
780	//TODO: optimise
781	hv_store_ent (hv, key, value, 0);
782
783	WS;	1122	}
		1123
		1124	decode_ws (dec);
784		1125
785	if (*dec->cur == '}')	1126	if (*dec->cur == '}')
786	{	1127	{
787	++dec->cur;	1128	++dec->cur;
788	break;	1129	break;
…		…
792	ERR (", or } expected while parsing object/hash");	1133	ERR (", or } expected while parsing object/hash");
793		1134
794	++dec->cur;	1135	++dec->cur;
795	}	1136	}
796		1137
		1138	DEC_DEC_DEPTH;
797	return newRV_noinc ((SV *)hv);	1139	sv = newRV_noinc ((SV *)hv);
		1140
		1141	// check filter callbacks
		1142	if (dec->json.flags & F_HOOK)
		1143	{
		1144	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1145	{
		1146	HE *cb, *he;
		1147
		1148	hv_iterinit (hv);
		1149	he = hv_iternext (hv);
		1150	hv_iterinit (hv);
		1151
		1152	// the next line creates a mortal sv each time its called.
		1153	// might want to optimise this for common cases.
		1154	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1155
		1156	if (cb)
		1157	{
		1158	dSP;
		1159	int count;
		1160
		1161	ENTER; SAVETMPS; PUSHMARK (SP);
		1162	XPUSHs (HeVAL (he));
		1163
		1164	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1165
		1166	if (count == 1)
		1167	{
		1168	sv = newSVsv (POPs);
		1169	FREETMPS; LEAVE;
		1170	return sv;
		1171	}
		1172
		1173	FREETMPS; LEAVE;
		1174	}
		1175	}
		1176
		1177	if (dec->json.cb_object)
		1178	{
		1179	dSP;
		1180	int count;
		1181
		1182	ENTER; SAVETMPS; PUSHMARK (SP);
		1183	XPUSHs (sv_2mortal (sv));
		1184
		1185	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1186
		1187	if (count == 1)
		1188	{
		1189	sv = newSVsv (POPs);
		1190	FREETMPS; LEAVE;
		1191	return sv;
		1192	}
		1193
		1194	SvREFCNT_inc (sv);
		1195	FREETMPS; LEAVE;
		1196	}
		1197	}
		1198
		1199	return sv;
798		1200
799	fail:	1201	fail:
800	SvREFCNT_dec (hv);	1202	SvREFCNT_dec (hv);
		1203	DEC_DEC_DEPTH;
801	return 0;	1204	return 0;
802	}	1205	}
803		1206
804	static SV *	1207	static SV *
805	decode_sv (dec_t *dec)	1208	decode_sv (dec_t *dec)
806	{	1209	{
807	WS;	1210	decode_ws (dec);
		1211
		1212	// the beauty of JSON: you need exactly one character lookahead
		1213	// to parse anything.
808	switch (*dec->cur)	1214	switch (*dec->cur)
809	{	1215	{
810	case '"': ++dec->cur; return decode_str (dec);	1216	case '"': ++dec->cur; return decode_str (dec);
811	case '[': ++dec->cur; return decode_av (dec);	1217	case '[': ++dec->cur; return decode_av (dec);
812	case '{': ++dec->cur; return decode_hv (dec);	1218	case '{': ++dec->cur; return decode_hv (dec);
…		…
818		1224
819	case 't':	1225	case 't':
820	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1226	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
821	{	1227	{
822	dec->cur += 4;	1228	dec->cur += 4;
823	return newSViv (1);	1229	return SvREFCNT_inc (json_true);
824	}	1230	}
825	else	1231	else
826	ERR ("'true' expected");	1232	ERR ("'true' expected");
827		1233
828	break;	1234	break;
829		1235
830	case 'f':	1236	case 'f':
831	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1237	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
832	{	1238	{
833	dec->cur += 5;	1239	dec->cur += 5;
834	return newSViv (0);	1240	return SvREFCNT_inc (json_false);
835	}	1241	}
836	else	1242	else
837	ERR ("'false' expected");	1243	ERR ("'false' expected");
838		1244
839	break;	1245	break;
…		…
848	ERR ("'null' expected");	1254	ERR ("'null' expected");
849		1255
850	break;	1256	break;
851		1257
852	default:	1258	default:
853	ERR ("malformed json string, neither array, object, number, string or atom");	1259	ERR ("malformed JSON string, neither array, object, number, string or atom");
854	break;	1260	break;
855	}	1261	}
856		1262
857	fail:	1263	fail:
858	return 0;	1264	return 0;
859	}	1265	}
860		1266
861	static SV *	1267	static SV *
862	decode_json (SV *string, UV flags)	1268	decode_json (SV *string, JSON *json, UV *offset_return)
863	{	1269	{
		1270	dec_t dec;
		1271	UV offset;
864	SV *sv;	1272	SV *sv;
865		1273
		1274	SvGETMAGIC (string);
		1275	SvUPGRADE (string, SVt_PV);
		1276
		1277	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1278	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1279	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1280
866	if (flags & F_UTF8)	1281	if (json->flags & F_UTF8)
867	sv_utf8_downgrade (string, 0);	1282	sv_utf8_downgrade (string, 0);
868	else	1283	else
869	sv_utf8_upgrade (string);	1284	sv_utf8_upgrade (string);
870		1285
871	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1286	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
872		1287
873	dec_t dec;	1288	dec.json = *json;
874	dec.flags = flags;
875	dec.cur = SvPVX (string);	1289	dec.cur = SvPVX (string);
876	dec.end = SvEND (string);	1290	dec.end = SvEND (string);
877	dec.err = 0;	1291	dec.err = 0;
		1292	dec.depth = 0;
		1293	dec.maxdepth = DEC_DEPTH (dec.json.flags);
878		1294
		1295	if (dec.json.cb_object || dec.json.cb_sk_object)
		1296	dec.json.flags |= F_HOOK;
		1297
		1298	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
879	sv = decode_sv (&dec);	1299	sv = decode_sv (&dec);
880		1300
		1301	if (!(offset_return || !sv))
		1302	{
		1303	// check for trailing garbage
		1304	decode_ws (&dec);
		1305
		1306	if (*dec.cur)
		1307	{
		1308	dec.err = "garbage after JSON object";
		1309	SvREFCNT_dec (sv);
		1310	sv = 0;
		1311	}
		1312	}
		1313
		1314	if (offset_return || !sv)
		1315	{
		1316	offset = dec.json.flags & F_UTF8
		1317	? dec.cur - SvPVX (string)
		1318	: utf8_distance (dec.cur, SvPVX (string));
		1319
		1320	if (offset_return)
		1321	*offset_return = offset;
		1322	}
		1323
881	if (!sv)	1324	if (!sv)
882	{	1325	{
883	IV offset = dec.flags & F_UTF8
884	? dec.cur - SvPVX (string)
885	: utf8_distance (dec.cur, SvPVX (string));
886	SV *uni = sv_newmortal ();	1326	SV *uni = sv_newmortal ();
887		1327
888	// horrible hack to silence warning inside pv_uni_display	1328	// horrible hack to silence warning inside pv_uni_display
889	COP cop = *PL_curcop;	1329	COP cop = *PL_curcop;
890	cop.cop_warnings = pWARN_NONE;	1330	cop.cop_warnings = pWARN_NONE;
…		…
892	SAVEVPTR (PL_curcop);	1332	SAVEVPTR (PL_curcop);
893	PL_curcop = &cop;	1333	PL_curcop = &cop;
894	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1334	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
895	LEAVE;	1335	LEAVE;
896		1336
897	croak ("%s, at character offset %d (%s)",	1337	croak ("%s, at character offset %d [\"%s\"]",
898	dec.err,	1338	dec.err,
899	(int)offset,	1339	(int)offset,
900	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1340	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
901	}	1341	}
902		1342
903	sv = sv_2mortal (sv);	1343	sv = sv_2mortal (sv);
904		1344
905	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1345	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
906	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1346	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
907		1347
908	return sv;	1348	return sv;
909	}	1349	}
910		1350
		1351	/////////////////////////////////////////////////////////////////////////////
		1352	// XS interface functions
		1353
911	MODULE = JSON::XS PACKAGE = JSON::XS	1354	MODULE = JSON::XS PACKAGE = JSON::XS
912		1355
913	BOOT:	1356	BOOT:
914	{	1357	{
915	int i;	1358	int i;
916		1359
917	memset (decode_hexdigit, 0xff, 256);
918	for (i = 10; i--; )	1360	for (i = 0; i < 256; ++i)
919	decode_hexdigit ['0' + i] = i;	1361	decode_hexdigit [i] =
		1362	i >= '0' && i <= '9' ? i - '0'
		1363	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1364	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1365	: -1;
920		1366
921	for (i = 7; i--; )	1367	json_stash = gv_stashpv ("JSON::XS" , 1);
		1368	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1369
		1370	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1371	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1372	}
		1373
		1374	PROTOTYPES: DISABLE
		1375
		1376	void CLONE (...)
		1377	CODE:
		1378	json_stash = 0;
		1379	json_boolean_stash = 0;
		1380
		1381	void new (char *klass)
		1382	PPCODE:
		1383	{
		1384	HV *stash = !JSON_SLOW || json_stash
		1385	? json_stash
		1386	: gv_stashpv ("JSON::XS", 1);
		1387	SV *pv = NEWSV (0, sizeof (JSON));
		1388	SvPOK_only (pv);
		1389	Zero (SvPVX (pv), 1, JSON);
		1390	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1391	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));
		1392	}
		1393
		1394	void ascii (JSON *self, int enable = 1)
		1395	ALIAS:
		1396	ascii = F_ASCII
		1397	latin1 = F_LATIN1
		1398	utf8 = F_UTF8
		1399	indent = F_INDENT
		1400	canonical = F_CANONICAL
		1401	space_before = F_SPACE_BEFORE
		1402	space_after = F_SPACE_AFTER
		1403	pretty = F_PRETTY
		1404	allow_nonref = F_ALLOW_NONREF
		1405	shrink = F_SHRINK
		1406	allow_blessed = F_ALLOW_BLESSED
		1407	convert_blessed = F_CONV_BLESSED
		1408	PPCODE:
		1409	{
		1410	if (enable)
		1411	self->flags |= ix;
		1412	else
		1413	self->flags &= ~ix;
		1414
		1415	XPUSHs (ST (0));
		1416	}
		1417
		1418	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1419	PPCODE:
		1420	{
		1421	UV log2 = 0;
		1422
		1423	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1424
		1425	while ((1UL << log2) < max_depth)
		1426	++log2;
		1427
		1428	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1429
		1430	XPUSHs (ST (0));
		1431	}
		1432
		1433	void max_size (JSON *self, UV max_size = 0)
		1434	PPCODE:
		1435	{
		1436	UV log2 = 0;
		1437
		1438	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1439	if (max_size == 1) max_size = 2;
		1440
		1441	while ((1UL << log2) < max_size)
		1442	++log2;
		1443
		1444	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1445
		1446	XPUSHs (ST (0));
		1447	}
		1448
		1449	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1450	PPCODE:
		1451	{
		1452	SvREFCNT_dec (self->cb_object);
		1453	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1454
		1455	XPUSHs (ST (0));
		1456	}
		1457
		1458	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1459	PPCODE:
		1460	{
		1461	if (!self->cb_sk_object)
		1462	self->cb_sk_object = newHV ();
		1463
		1464	if (SvOK (cb))
		1465	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1466	else
922	{	1467	{
923	decode_hexdigit ['a' + i] = 10 + i;	1468	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
924	decode_hexdigit ['A' + i] = 10 + i;	1469
		1470	if (!HvKEYS (self->cb_sk_object))
		1471	{
		1472	SvREFCNT_dec (self->cb_sk_object);
		1473	self->cb_sk_object = 0;
		1474	}
925	}	1475	}
926		1476
927	json_stash = gv_stashpv ("JSON::XS", 1);	1477	XPUSHs (ST (0));
928	}	1478	}
929		1479
930	PROTOTYPES: DISABLE	1480	void encode (JSON *self, SV *scalar)
		1481	PPCODE:
		1482	XPUSHs (encode_json (scalar, self));
931		1483
932	SV *new (char *dummy)	1484	void decode (JSON *self, SV *jsonstr)
		1485	PPCODE:
		1486	XPUSHs (decode_json (jsonstr, self, 0));
		1487
		1488	void decode_prefix (JSON *self, SV *jsonstr)
		1489	PPCODE:
		1490	{
		1491	UV offset;
		1492	EXTEND (SP, 2);
		1493	PUSHs (decode_json (jsonstr, self, &offset));
		1494	PUSHs (sv_2mortal (newSVuv (offset)));
		1495	}
		1496
		1497	void DESTROY (JSON *self)
933	CODE:	1498	CODE:
934	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1499	SvREFCNT_dec (self->cb_sk_object);
935	OUTPUT:	1500	SvREFCNT_dec (self->cb_object);
936	RETVAL
937
938	SV *ascii (SV *self, int enable = 1)
939	ALIAS:
940	ascii = F_ASCII
941	utf8 = F_UTF8
942	indent = F_INDENT
943	canonical = F_CANONICAL
944	space_before = F_SPACE_BEFORE
945	space_after = F_SPACE_AFTER
946	json_rpc = F_JSON_RPC
947	pretty = F_PRETTY
948	allow_nonref = F_ALLOW_NONREF
949	shrink = F_SHRINK
950	CODE:
951	{
952	UV *uv = SvJSON (self);
953	if (enable)
954	*uv |= ix;
955	else
956	*uv &= ~ix;
957
958	RETVAL = newSVsv (self);
959	}
960	OUTPUT:
961	RETVAL
962
963	void encode (SV *self, SV *scalar)
964	PPCODE:
965	XPUSHs (encode_json (scalar, *SvJSON (self)));
966
967	void decode (SV *self, SV *jsonstr)
968	PPCODE:
969	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
970		1501
971	PROTOTYPES: ENABLE	1502	PROTOTYPES: ENABLE
972		1503
973	void to_json (SV *scalar)	1504	void to_json (SV *scalar)
974	PPCODE:	1505	PPCODE:
		1506	{
		1507	JSON json = { F_DEFAULT | F_UTF8 };
975	XPUSHs (encode_json (scalar, F_UTF8));	1508	XPUSHs (encode_json (scalar, &json));
		1509	}
976		1510
977	void from_json (SV *jsonstr)	1511	void from_json (SV *jsonstr)
978	PPCODE:	1512	PPCODE:
		1513	{
		1514	JSON json = { F_DEFAULT | F_UTF8 };
979	XPUSHs (decode_json (jsonstr, F_UTF8));	1515	XPUSHs (decode_json (jsonstr, &json, 0));
		1516	}
980		1517

Diff Legend

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