ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.7 by root, Fri Mar 23 15:57:18 2007 UTC vs.
Revision 1.47 by root, Sun Jul 1 14:08:03 2007 UTC

…		…
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include "assert.h"
6	#include "string.h"	6	#include "string.h"
7	#include "stdlib.h"	7	#include "stdlib.h"
		8	#include "stdio.h"
8		9
		10	#if defined(__BORLANDC__) || defined(_MSC_VER)
		11	# define snprintf _snprintf // C compilers have this in stdio.h
		12	#endif
		13
		14	// some old perls do not have this, try to make it work, no
		15	// guarentees, though. if it breaks, you get to keep the pieces.
		16	#ifndef UTF8_MAXBYTES
		17	# define UTF8_MAXBYTES 13
		18	#endif
		19
9	#define F_ASCII 0x00000001	20	#define F_ASCII 0x00000001UL
		21	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	22	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	23	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	24	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	25	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	26	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	27	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	28	#define F_SHRINK 0x00000200UL
		29	#define F_ALLOW_BLESSED 0x00000400UL
		30	#define F_CONV_BLESSED 0x00000800UL // NYI
		31	#define F_MAXDEPTH 0xf8000000UL
		32	#define S_MAXDEPTH 27
		33	#define F_MAXSIZE 0x01f00000UL
		34	#define S_MAXSIZE 20
		35
		36	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		37	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		38
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	40	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		41
22	#define INIT_SIZE 32 // initial scalar size to be allocated	42	#define INIT_SIZE 32 // initial scalar size to be allocated
		43	#define INDENT_STEP 3 // spaces per indentation level
		44
		45	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		46
24	#define SB do {	47	#define SB do {
25	#define SE } while (0)	48	#define SE } while (0)
26		49
27	static HV *json_stash;	50	#if __GNUC__ >= 3
		51	# define expect(expr,value) __builtin_expect ((expr),(value))
		52	# define inline inline
		53	#else
		54	# define expect(expr,value) (expr)
		55	# define inline static
		56	#endif
		57
		58	#define expect_false(expr) expect ((expr) != 0, 0)
		59	#define expect_true(expr) expect ((expr) != 0, 1)
		60
		61	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		62	static SV *json_true, *json_false;
		63
		64	typedef struct json {
		65	U32 flags;
		66	} JSON__XS;
		67
		68	/////////////////////////////////////////////////////////////////////////////
		69	// utility functions
		70
		71	static UV *
		72	SvJSON (SV *sv)
		73	{
		74	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
		75	croak ("object is not of type JSON::XS");
		76
		77	return &SvUVX (SvRV (sv));
		78	}
		79
		80	static void
		81	shrink (SV *sv)
		82	{
		83	sv_utf8_downgrade (sv, 1);
		84	if (SvLEN (sv) > SvCUR (sv) + 1)
		85	{
		86	#ifdef SvPV_shrink_to_cur
		87	SvPV_shrink_to_cur (sv);
		88	#elif defined (SvPV_renew)
		89	SvPV_renew (sv, SvCUR (sv) + 1);
		90	#endif
		91	}
		92	}
		93
		94	// decode an utf-8 character and return it, or (UV)-1 in
		95	// case of an error.
		96	// we special-case "safe" characters from U+80 .. U+7FF,
		97	// but use the very good perl function to parse anything else.
		98	// note that we never call this function for a ascii codepoints
		99	inline UV
		100	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		101	{
		102	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		103	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		104	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		105	{
		106	*clen = 2;
		107	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		108	}
		109	else
		110	{
		111	*clen = (STRLEN)-1;
		112	return (UV)-1;
		113	}
		114	}
		115
		116	/////////////////////////////////////////////////////////////////////////////
		117	// encoder
28		118
29	// structure used for encoding JSON	119	// structure used for encoding JSON
30	typedef struct	120	typedef struct
31	{	121	{
32	char *cur;	122	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	123	char *end; // SvEND (sv)
35	SV *sv;	124	SV *sv; // result scalar
36	UV flags;	125	struct json json;
37	int max_recurse;	126	U32 indent; // indentation level
38	int indent;	127	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	128	} enc_t;
40		129
41	// structure used for decoding JSON	130	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)	131	need (enc_t *enc, STRLEN len)
72	{	132	{
73	if (enc->cur + len >= enc->end)	133	if (expect_false (enc->cur + len >= enc->end))
74	{	134	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	135	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	136	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	137	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	138	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	139	}
80	}	140	}
81		141
82	static void	142	inline void
83	encode_ch (enc_t *enc, char ch)	143	encode_ch (enc_t *enc, char ch)
84	{	144	{
85	need (enc, 1);	145	need (enc, 1);
86	*enc->cur++ = ch;	146	*enc->cur++ = ch;
87	}	147	}
…		…
95		155
96	while (str < end)	156	while (str < end)
97	{	157	{
98	unsigned char ch = *(unsigned char *)str;	158	unsigned char ch = *(unsigned char *)str;
99		159
100	if (ch >= 0x20 && ch < 0x80) // most common case	160	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	161	{
102	if (ch == '"') // but with slow exceptions	162	if (expect_false (ch == '"')) // but with slow exceptions
103	{	163	{
104	need (enc, len += 1);	164	need (enc, len += 1);
105	*enc->cur++ = '\\';	165	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	166	*enc->cur++ = '"';
107	}	167	}
108	else if (ch == '\\')	168	else if (expect_false (ch == '\\'))
109	{	169	{
110	need (enc, len += 1);	170	need (enc, len += 1);
111	*enc->cur++ = '\\';	171	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	172	*enc->cur++ = '\\';
113	}	173	}
…		…
131	STRLEN clen;	191	STRLEN clen;
132	UV uch;	192	UV uch;
133		193
134	if (is_utf8)	194	if (is_utf8)
135	{	195	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	196	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	197	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	198	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	199	}
140	else	200	else
141	{	201	{
142	uch = ch;	202	uch = ch;
143	clen = 1;	203	clen = 1;
144	}	204	}
145		205
146	if (uch < 0x80 || enc->flags & F_ASCII)	206	if (uch > 0x10FFFFUL)
		207	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		208
		209	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
147	{	210	{
148	if (uch > 0xFFFFUL)	211	if (uch > 0xFFFFUL)
149	{	212	{
150	need (enc, len += 11);	213	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	214	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	215	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	216	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	217	enc->cur += 12;
155	}	218	}
156	else	219	else
157	{	220	{
158	static char hexdigit [16] = "0123456789abcdef";	221	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	228	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	229	}
167		230
168	str += clen;	231	str += clen;
169	}	232	}
		233	else if (enc->json.flags & F_LATIN1)
		234	{
		235	*enc->cur++ = uch;
		236	str += clen;
		237	}
170	else if (is_utf8)	238	else if (is_utf8)
171	{	239	{
172	need (enc, len += clen);	240	need (enc, len += clen);
173	do	241	do
174	{	242	{
…		…
176	}	244	}
177	while (--clen);	245	while (--clen);
178	}	246	}
179	else	247	else
180	{	248	{
181	need (enc, len += 10); // never more than 11 bytes needed	249	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	250	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	251	++str;
184	}	252	}
185	}	253	}
186	}	254	}
…		…
188		256
189	--len;	257	--len;
190	}	258	}
191	}	259	}
192		260
193	#define INDENT SB \	261	inline void
		262	encode_indent (enc_t *enc)
		263	{
194	if (enc->flags & F_INDENT) \	264	if (enc->json.flags & F_INDENT)
195	{ \	265	{
196	int i_; \	266	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	267
198	for (i_ = enc->indent * 3; i_--; )\	268	need (enc, spaces);
		269	memset (enc->cur, ' ', spaces);
		270	enc->cur += spaces;
		271	}
		272	}
		273
		274	inline void
		275	encode_space (enc_t *enc)
		276	{
		277	need (enc, 1);
		278	encode_ch (enc, ' ');
		279	}
		280
		281	inline void
		282	encode_nl (enc_t *enc)
		283	{
		284	if (enc->json.flags & F_INDENT)
		285	{
		286	need (enc, 1);
199	encode_ch (enc, ' '); \	287	encode_ch (enc, '\n');
200	} \	288	}
201	SE	289	}
202		290
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	291	inline void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	292	encode_comma (enc_t *enc)
205	#define COMMA SB \	293	{
206	encode_ch (enc, ','); \	294	encode_ch (enc, ',');
		295
207	if (enc->flags & F_INDENT) \	296	if (enc->json.flags & F_INDENT)
208	NL; \	297	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	298	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	299	encode_space (enc);
211	SE	300	}
212		301
213	static void encode_sv (enc_t *enc, SV *sv);	302	static void encode_sv (enc_t *enc, SV *sv);
214		303
215	static void	304	static void
216	encode_av (enc_t *enc, AV *av)	305	encode_av (enc_t *enc, AV *av)
217	{	306	{
218	int i, len = av_len (av);	307	int i, len = av_len (av);
219		308
220	encode_ch (enc, '['); NL;	309	if (enc->indent >= enc->maxdepth)
		310	croak ("data structure too deep (hit recursion limit)");
		311
		312	encode_ch (enc, '['); encode_nl (enc);
221	++enc->indent;	313	++enc->indent;
222		314
223	for (i = 0; i <= len; ++i)	315	for (i = 0; i <= len; ++i)
224	{	316	{
225	INDENT;	317	encode_indent (enc);
226	encode_sv (enc, *av_fetch (av, i, 0));	318	encode_sv (enc, *av_fetch (av, i, 0));
227		319
228	if (i < len)	320	if (i < len)
229	COMMA;	321	encode_comma (enc);
230	}	322	}
231		323
232	NL;	324	encode_nl (enc);
233		325
234	--enc->indent;	326	--enc->indent;
235	INDENT; encode_ch (enc, ']');	327	encode_indent (enc); encode_ch (enc, ']');
236	}	328	}
237		329
238	static void	330	static void
239	encode_he (enc_t *enc, HE *he)	331	encode_he (enc_t *enc, HE *he)
240	{	332	{
…		…
254	else	346	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	347	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		348
257	encode_ch (enc, '"');	349	encode_ch (enc, '"');
258		350
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	351	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	352	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	353	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));	354	encode_sv (enc, HeVAL (he));
263	}	355	}
264		356
265	// compare hash entries, used when all keys are bytestrings	357	// compare hash entries, used when all keys are bytestrings
266	static int	358	static int
…		…
272	HE *b = *(HE **)b_;	364	HE *b = *(HE **)b_;
273		365
274	STRLEN la = HeKLEN (a);	366	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	367	STRLEN lb = HeKLEN (b);
276		368
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	369	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	370	cmp = la - lb;
279		371
280	return cmp;	372	return cmp;
281	}	373	}
282		374
283	// compare hash entries, used when some keys are sv's or utf-x	375	// compare hash entries, used when some keys are sv's or utf-x
…		…
290	static void	382	static void
291	encode_hv (enc_t *enc, HV *hv)	383	encode_hv (enc_t *enc, HV *hv)
292	{	384	{
293	int count, i;	385	int count, i;
294		386
		387	if (enc->indent >= enc->maxdepth)
		388	croak ("data structure too deep (hit recursion limit)");
		389
295	encode_ch (enc, '{'); NL; ++enc->indent;	390	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
296		391
297	if ((count = hv_iterinit (hv)))	392	if ((count = hv_iterinit (hv)))
298	{	393	{
299	// for canonical output we have to sort by keys first	394	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	395	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	396	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	397	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	398	if (enc->json.flags & F_CANONICAL)
304	{	399	{
305	HE *he, *hes [count];
306	int fast = 1;	400	int fast = 1;
		401	HE *he;
		402	#if defined(__BORLANDC__) || defined(_MSC_VER)
		403	HE **hes = _alloca (count * sizeof (HE));
		404	#else
		405	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		406	#endif
307		407
308	i = 0;	408	i = 0;
309	while ((he = hv_iternext (hv)))	409	while ((he = hv_iternext (hv)))
310	{	410	{
311	hes [i++] = he;	411	hes [i++] = he;
…		…
317		417
318	if (fast)	418	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	419	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	420	else
321	{	421	{
322	// hack to disable "use bytes"	422	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	423	COP cop = *PL_curcop;
324	cop.op_private = 0;	424	cop.op_private = 0;
		425
		426	ENTER;
		427	SAVETMPS;
		428
		429	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	430	PL_curcop = &cop;
326		431
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	432	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		433
329	FREETMPS;	434	FREETMPS;
330		435	LEAVE;
331	PL_curcop = oldcop;
332	}	436	}
333		437
334	for (i = 0; i < count; ++i)	438	for (i = 0; i < count; ++i)
335	{	439	{
336	INDENT;	440	encode_indent (enc);
337	encode_he (enc, hes [i]);	441	encode_he (enc, hes [i]);
338		442
339	if (i < count - 1)	443	if (i < count - 1)
340	COMMA;	444	encode_comma (enc);
341	}
342
343	NL;	445	}
		446
		447	encode_nl (enc);
344	}	448	}
345	else	449	else
346	{	450	{
347	SV *sv;
348	HE *he = hv_iternext (hv);	451	HE *he = hv_iternext (hv);
349		452
350	for (;;)	453	for (;;)
351	{	454	{
352	INDENT;	455	encode_indent (enc);
353	encode_he (enc, he);	456	encode_he (enc, he);
354		457
355	if (!(he = hv_iternext (hv)))	458	if (!(he = hv_iternext (hv)))
356	break;	459	break;
357		460
358	COMMA;	461	encode_comma (enc);
359	}
360
361	NL;	462	}
		463
		464	encode_nl (enc);
362	}	465	}
363	}	466	}
364		467
365	--enc->indent; INDENT; encode_ch (enc, '}');	468	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		469	}
		470
		471	// encode objects, arrays and special \0=false and \1=true values.
		472	static void
		473	encode_rv (enc_t *enc, SV *sv)
		474	{
		475	svtype svt;
		476
		477	SvGETMAGIC (sv);
		478	svt = SvTYPE (sv);
		479
		480	if (expect_false (SvOBJECT (sv)))
		481	{
		482	if (SvSTASH (sv) == json_boolean_stash)
		483	{
		484	if (SvIV (sv) == 0)
		485	encode_str (enc, "false", 5, 0);
		486	else
		487	encode_str (enc, "true", 4, 0);
		488	}
		489	else
		490	{
		491	#if 0
		492	if (0 && sv_derived_from (rv, "JSON::Literal"))
		493	{
		494	// not yet
		495	}
		496	#endif
		497	if (enc->json.flags & F_CONV_BLESSED)
		498	{
		499	// we re-bless the reference to get overload and other niceties right
		500	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 1);
		501
		502	if (to_json)
		503	{
		504	dSP;
		505	ENTER;
		506	SAVETMPS;
		507	PUSHMARK (SP);
		508	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		509
		510	// calling with G_SCALAR ensures that we always get a 1 reutrn value
		511	// check anyways.
		512	PUTBACK;
		513	assert (1 == call_sv ((SV *)GvCV (to_json), G_SCALAR));
		514	SPAGAIN;
		515
		516	encode_sv (enc, POPs);
		517
		518	FREETMPS;
		519	LEAVE;
		520	}
		521	else if (enc->json.flags & F_ALLOW_BLESSED)
		522	encode_str (enc, "null", 4, 0);
		523	else
		524	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		525	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		526	}
		527	else if (enc->json.flags & F_ALLOW_BLESSED)
		528	encode_str (enc, "null", 4, 0);
		529	else
		530	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		531	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		532	}
		533	}
		534	else if (svt == SVt_PVHV)
		535	encode_hv (enc, (HV *)sv);
		536	else if (svt == SVt_PVAV)
		537	encode_av (enc, (AV *)sv);
		538	else if (svt < SVt_PVAV)
		539	{
		540	if (SvNIOK (sv) && SvIV (sv) == 0)
		541	encode_str (enc, "false", 5, 0);
		542	else if (SvNIOK (sv) && SvIV (sv) == 1)
		543	encode_str (enc, "true", 4, 0);
		544	else
		545	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		546	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		547	}
		548	else
		549	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		550	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	551	}
367		552
368	static void	553	static void
369	encode_sv (enc_t *enc, SV *sv)	554	encode_sv (enc_t *enc, SV *sv)
370	{	555	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	563	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	564	encode_ch (enc, '"');
380	}	565	}
381	else if (SvNOKp (sv))	566	else if (SvNOKp (sv))
382	{	567	{
		568	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	569	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	570	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	571	enc->cur += strlen (enc->cur);
386	}	572	}
387	else if (SvIOKp (sv))	573	else if (SvIOKp (sv))
388	{	574	{
389	need (enc, 64);	575	// we assume we can always read an IV as a UV
		576	if (SvUV (sv) & ~(UV)0x7fff)
		577	{
		578	// large integer, use the (rather slow) snprintf way.
		579	need (enc, sizeof (UV) * 3);
390	enc->cur +=	580	enc->cur +=
391	SvIsUV(sv)	581	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	582	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	583	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		584	}
		585	else
		586	{
		587	// optimise the "small number case"
		588	// code will likely be branchless and use only a single multiplication
		589	I32 i = SvIV (sv);
		590	U32 u;
		591	char digit, nz = 0;
		592
		593	need (enc, 6);
		594
		595	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		596	u = i < 0 ? -i : i;
		597
		598	// convert to 4.28 fixed-point representation
		599	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		600
		601	// now output digit by digit, each time masking out the integer part
		602	// and multiplying by 5 while moving the decimal point one to the right,
		603	// resulting in a net multiplication by 10.
		604	// we always write the digit to memory but conditionally increment
		605	// the pointer, to ease the usage of conditional move instructions.
		606	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		607	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		608	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		609	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		610	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		611	}
394	}	612	}
395	else if (SvROK (sv))	613	else if (SvROK (sv))
396	{	614	encode_rv (enc, SvRV (sv));
397	if (!--enc->max_recurse)
398	croak ("data structure too deep (hit recursion limit)");
399
400	sv = SvRV (sv);
401
402	switch (SvTYPE (sv))
403	{
404	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407	default:
408	croak ("JSON can only represent references to arrays or hashes");
409	}
410	}
411	else if (!SvOK (sv))	615	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	616	encode_str (enc, "null", 4, 0);
413	else	617	else
414	croak ("encountered perl type that JSON cannot handle");	618	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		619	SvPV_nolen (sv), SvFLAGS (sv));
415	}	620	}
416		621
417	static SV *	622	static SV *
418	encode_json (SV *scalar, UV flags)	623	encode_json (SV *scalar, struct json *json)
419	{	624	{
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	enc_t enc;	625	enc_t enc;
424	enc.flags = flags;	626
		627	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		628	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		629
		630	enc.json = *json;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	631	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	632	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	633	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	634	enc.indent = 0;
		635	enc.maxdepth = DEC_DEPTH (enc.json.flags);
430		636
431	SvPOK_only (enc.sv);	637	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	638	encode_sv (&enc, scalar);
433		639
		640	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		641	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		642
434	if (!(flags & (F_ASCII | F_UTF8)))	643	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	644	SvUTF8_on (enc.sv);
436		645
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438
439	if (enc.flags & F_SHRINK)	646	if (enc.json.flags & F_SHRINK)
440	shrink (enc.sv);	647	shrink (enc.sv);
441		648
442	return enc.sv;	649	return enc.sv;
443	}	650	}
444		651
445	/////////////////////////////////////////////////////////////////////////////	652	/////////////////////////////////////////////////////////////////////////////
		653	// decoder
446		654
447	#define WS \	655	// structure used for decoding JSON
		656	typedef struct
		657	{
		658	char *cur; // current parser pointer
		659	char *end; // end of input string
		660	const char *err; // parse error, if != 0
		661	struct json json;
		662	U32 depth; // recursion depth
		663	U32 maxdepth; // recursion depth limit
		664	} dec_t;
		665
		666	inline void
		667	decode_ws (dec_t *dec)
		668	{
448	for (;;) \	669	for (;;)
449	{ \	670	{
450	char ch = *dec->cur; \	671	char ch = *dec->cur;
		672
451	if (ch > 0x20 \	673	if (ch > 0x20
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	674	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
453	break; \	675	break;
		676
454	++dec->cur; \	677	++dec->cur;
455	}	678	}
		679	}
456		680
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	681	#define ERR(reason) SB dec->err = reason; goto fail; SE
		682
458	#define EXPECT_CH(ch) SB \	683	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	684	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	685	ERR (# ch " expected"); \
461	++dec->cur; \	686	++dec->cur; \
462	SE	687	SE
463		688
		689	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		690	#define DEC_DEC_DEPTH --dec->depth
		691
464	static SV *decode_sv (dec_t *dec);	692	static SV *decode_sv (dec_t *dec);
465		693
466	static signed char decode_hexdigit[256];	694	static signed char decode_hexdigit[256];
467		695
468	static UV	696	static UV
469	decode_4hex (dec_t *dec)	697	decode_4hex (dec_t *dec)
470	{	698	{
471	signed char d1, d2, d3, d4;	699	signed char d1, d2, d3, d4;
		700	unsigned char *cur = (unsigned char *)dec->cur;
472		701
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	702	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	703	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	704	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	705	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
477	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478	if (d3 < 0) ERR ("four hexadecimal digits expected");
479	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480	if (d4 < 0) ERR ("four hexadecimal digits expected");
481		706
482	dec->cur += 4;	707	dec->cur += 4;
483		708
484	return ((UV)d1) << 12	709	return ((UV)d1) << 12
485	| ((UV)d2) << 8	710	| ((UV)d2) << 8
…		…
488		713
489	fail:	714	fail:
490	return (UV)-1;	715	return (UV)-1;
491	}	716	}
492		717
493	#define APPEND_GROW(n) SB \
494	if (cur + (n) >= end) \
495	{ \
496	STRLEN ofs = cur - SvPVX (sv); \
497	SvGROW (sv, ofs + (n) + 1); \
498	cur = SvPVX (sv) + ofs; \
499	end = SvEND (sv); \
500	} \
501	SE
502
503	#define APPEND_CH(ch) SB \
504	APPEND_GROW (1); \
505	*cur++ = (ch); \
506	SE
507
508	static SV *	718	static SV *
509	decode_str (dec_t *dec)	719	decode_str (dec_t *dec)
510	{	720	{
511	SV *sv = NEWSV (0,2);	721	SV *sv = 0;
512	int utf8 = 0;	722	int utf8 = 0;
513	char *cur = SvPVX (sv);	723	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		724
516	for (;;)	725	do
517	{	726	{
518	unsigned char ch = *(unsigned char *)dec->cur;	727	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		728	char *cur = buf;
519		729
520	if (ch == '"')	730	do
521	break;
522	else if (ch == '\\')
523	{	731	{
524	switch (*++dec->cur)	732	unsigned char ch = *(unsigned char *)dec_cur++;
		733
		734	if (expect_false (ch == '"'))
		735	{
		736	--dec_cur;
		737	break;
525	{	738	}
526	case '\\':	739	else if (expect_false (ch == '\\'))
527	case '/':	740	{
528	case '"': APPEND_CH (*dec->cur++); break;	741	switch (*dec_cur)
529
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;
532	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536	case 'u':
537	{	742	{
538	UV lo, hi;	743	case '\\':
539	++dec->cur;	744	case '/':
		745	case '"': *cur++ = *dec_cur++; break;
540		746
541	hi = decode_4hex (dec);	747	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	748	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	749	case 'n': ++dec_cur; *cur++ = '\012'; break;
		750	case 'f': ++dec_cur; *cur++ = '\014'; break;
		751	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		752
545	// possibly a surrogate pair	753	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	754	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	755	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	756	++dec_cur;
550		757
551	dec->cur += 2;	758	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	759	hi = decode_4hex (dec);
		760	dec_cur = dec->cur;
554	if (lo == (UV)-1)	761	if (hi == (UV)-1)
555	goto fail;	762	goto fail;
556		763
		764	// possibly a surrogate pair
		765	if (hi >= 0xd800)
		766	if (hi < 0xdc00)
		767	{
		768	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		769	ERR ("missing low surrogate character in surrogate pair");
		770
		771	dec_cur += 2;
		772
		773	dec->cur = dec_cur;
		774	lo = decode_4hex (dec);
		775	dec_cur = dec->cur;
		776	if (lo == (UV)-1)
		777	goto fail;
		778
557	if (lo < 0xdc00 || lo >= 0xe000)	779	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	780	ERR ("surrogate pair expected");
559		781
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	782	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		783	}
		784	else if (hi < 0xe000)
		785	ERR ("missing high surrogate character in surrogate pair");
		786
		787	if (hi >= 0x80)
		788	{
		789	utf8 = 1;
		790
		791	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		792	}
		793	else
		794	*cur++ = hi;
561	}	795	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	796	break;
567	utf8 = 1;
568		797
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	798	default:
573	APPEND_CH (hi);	799	--dec_cur;
		800	ERR ("illegal backslash escape sequence in string");
574	}	801	}
575	break;	802	}
		803	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		804	*cur++ = ch;
		805	else if (ch >= 0x80)
		806	{
		807	STRLEN clen;
		808	UV uch;
576		809
577	default:
578	--dec->cur;	810	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	811
		812	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		813	if (clen == (STRLEN)-1)
		814	ERR ("malformed UTF-8 character in JSON string");
		815
		816	do
		817	*cur++ = *dec_cur++;
		818	while (--clen);
		819
		820	utf8 = 1;
		821	}
		822	else
		823	{
		824	--dec_cur;
		825
		826	if (!ch)
		827	ERR ("unexpected end of string while parsing JSON string");
		828	else
		829	ERR ("invalid character encountered while parsing JSON string");
580	}	830	}
581	}	831	}
582	else if (ch >= 0x20 && ch <= 0x7f)	832	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	833
584	else if (ch >= 0x80)
585	{	834	{
586	STRLEN clen;	835	STRLEN len = cur - buf;
587	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	if (clen == (STRLEN)-1)
589	ERR ("malformed UTF-8 character in JSON string");
590		836
591	APPEND_GROW (clen);	837	if (sv)
592	do
593	{	838	{
594	*cur++ = *dec->cur++;	839	SvGROW (sv, SvCUR (sv) + len + 1);
		840	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		841	SvCUR_set (sv, SvCUR (sv) + len);
595	}	842	}
596	while (--clen);
597
598	utf8 = 1;
599	}
600	else if (dec->cur == dec->end)
601	ERR ("unexpected end of string while parsing json string");
602	else	843	else
603	ERR ("invalid character encountered");	844	sv = newSVpvn (buf, len);
604	}	845	}
		846	}
		847	while (*dec_cur != '"');
605		848
606	++dec->cur;	849	++dec_cur;
607		850
608	SvCUR_set (sv, cur - SvPVX (sv));	851	if (sv)
609		852	{
610	SvPOK_only (sv);	853	SvPOK_only (sv);
611	*SvEND (sv) = 0;	854	*SvEND (sv) = 0;
612		855
613	if (utf8)	856	if (utf8)
614	SvUTF8_on (sv);	857	SvUTF8_on (sv);
		858	}
		859	else
		860	sv = newSVpvn ("", 0);
615		861
616	if (dec->flags & F_SHRINK)	862	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	863	return sv;
620		864
621	fail:	865	fail:
622	SvREFCNT_dec (sv);	866	dec->cur = dec_cur;
623	return 0;	867	return 0;
624	}	868	}
625		869
626	static SV *	870	static SV *
627	decode_num (dec_t *dec)	871	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	929	is_nv = 1;
686	}	930	}
687		931
688	if (!is_nv)	932	if (!is_nv)
689	{	933	{
690	UV uv;	934	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
691	int numtype = grok_number (start, dec->cur - start, &uv);	935	if (*start == '-')
692	if (numtype & IS_NUMBER_IN_UV)	936	switch (dec->cur - start)
693	if (numtype & IS_NUMBER_NEG)
694	{	937	{
695	if (uv < (UV)IV_MIN)	938	case 2: return newSViv (-( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	939	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		940	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		941	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
697	}	942	}
		943	else
		944	switch (dec->cur - start)
		945	{
		946	case 1: return newSViv ( start [0] - '0' * 1);
		947	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		948	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		949	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		950	}
		951
		952	{
		953	UV uv;
		954	int numtype = grok_number (start, dec->cur - start, &uv);
		955	if (numtype & IS_NUMBER_IN_UV)
		956	if (numtype & IS_NUMBER_NEG)
		957	{
		958	if (uv < (UV)IV_MIN)
		959	return newSViv (-(IV)uv);
		960	}
698	else	961	else
699	return newSVuv (uv);	962	return newSVuv (uv);
		963
		964	// here would likely be the place for bigint support
700	}	965	}
		966	}
701		967
		968	// if we ever support bigint or bigfloat, this is the place for bigfloat
702	return newSVnv (Atof (start));	969	return newSVnv (Atof (start));
703		970
704	fail:	971	fail:
705	return 0;	972	return 0;
706	}	973	}
…		…
708	static SV *	975	static SV *
709	decode_av (dec_t *dec)	976	decode_av (dec_t *dec)
710	{	977	{
711	AV *av = newAV ();	978	AV *av = newAV ();
712		979
713	WS;	980	DEC_INC_DEPTH;
		981	decode_ws (dec);
		982
714	if (*dec->cur == ']')	983	if (*dec->cur == ']')
715	++dec->cur;	984	++dec->cur;
716	else	985	else
717	for (;;)	986	for (;;)
718	{	987	{
…		…
722	if (!value)	991	if (!value)
723	goto fail;	992	goto fail;
724		993
725	av_push (av, value);	994	av_push (av, value);
726		995
727	WS;	996	decode_ws (dec);
728		997
729	if (*dec->cur == ']')	998	if (*dec->cur == ']')
730	{	999	{
731	++dec->cur;	1000	++dec->cur;
732	break;	1001	break;
…		…
736	ERR (", or ] expected while parsing array");	1005	ERR (", or ] expected while parsing array");
737		1006
738	++dec->cur;	1007	++dec->cur;
739	}	1008	}
740		1009
		1010	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1011	return newRV_noinc ((SV *)av);
742		1012
743	fail:	1013	fail:
744	SvREFCNT_dec (av);	1014	SvREFCNT_dec (av);
		1015	DEC_DEC_DEPTH;
745	return 0;	1016	return 0;
746	}	1017	}
747		1018
748	static SV *	1019	static SV *
749	decode_hv (dec_t *dec)	1020	decode_hv (dec_t *dec)
750	{	1021	{
751	HV *hv = newHV ();	1022	HV *hv = newHV ();
752		1023
753	WS;	1024	DEC_INC_DEPTH;
		1025	decode_ws (dec);
		1026
754	if (*dec->cur == '}')	1027	if (*dec->cur == '}')
755	++dec->cur;	1028	++dec->cur;
756	else	1029	else
757	for (;;)	1030	for (;;)
758	{	1031	{
		1032	decode_ws (dec); EXPECT_CH ('"');
		1033
		1034	// heuristic: assume that
		1035	// a) decode_str + hv_store_ent are abysmally slow.
		1036	// b) most hash keys are short, simple ascii text.
		1037	// => try to "fast-match" such strings to avoid
		1038	// the overhead of decode_str + hv_store_ent.
		1039	{
759	SV *key, *value;	1040	SV *value;
		1041	char *p = dec->cur;
		1042	char *e = p + 24; // only try up to 24 bytes
760		1043
761	WS; EXPECT_CH ('"');	1044	for (;;)
762
763	key = decode_str (dec);
764	if (!key)
765	goto fail;
766
767	WS; EXPECT_CH (':');
768
769	value = decode_sv (dec);
770	if (!value)
771	{	1045	{
		1046	// the >= 0x80 is true on most architectures
		1047	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1048	{
		1049	// slow path, back up and use decode_str
		1050	SV *key = decode_str (dec);
		1051	if (!key)
		1052	goto fail;
		1053
		1054	decode_ws (dec); EXPECT_CH (':');
		1055
		1056	value = decode_sv (dec);
		1057	if (!value)
		1058	{
		1059	SvREFCNT_dec (key);
		1060	goto fail;
		1061	}
		1062
		1063	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1064	SvREFCNT_dec (key);
		1065
		1066	break;
		1067	}
		1068	else if (*p == '"')
		1069	{
		1070	// fast path, got a simple key
		1071	char *key = dec->cur;
		1072	int len = p - key;
		1073	dec->cur = p + 1;
		1074
		1075	decode_ws (dec); EXPECT_CH (':');
		1076
		1077	value = decode_sv (dec);
		1078	if (!value)
773	goto fail;	1079	goto fail;
		1080
		1081	hv_store (hv, key, len, value, 0);
		1082
		1083	break;
		1084	}
		1085
		1086	++p;
774	}	1087	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1088	}
		1089
		1090	decode_ws (dec);
780		1091
781	if (*dec->cur == '}')	1092	if (*dec->cur == '}')
782	{	1093	{
783	++dec->cur;	1094	++dec->cur;
784	break;	1095	break;
…		…
788	ERR (", or } expected while parsing object/hash");	1099	ERR (", or } expected while parsing object/hash");
789		1100
790	++dec->cur;	1101	++dec->cur;
791	}	1102	}
792		1103
		1104	DEC_DEC_DEPTH;
793	return newRV_noinc ((SV *)hv);	1105	return newRV_noinc ((SV *)hv);
794		1106
795	fail:	1107	fail:
796	SvREFCNT_dec (hv);	1108	SvREFCNT_dec (hv);
		1109	DEC_DEC_DEPTH;
797	return 0;	1110	return 0;
798	}	1111	}
799		1112
800	static SV *	1113	static SV *
801	decode_sv (dec_t *dec)	1114	decode_sv (dec_t *dec)
802	{	1115	{
803	WS;	1116	decode_ws (dec);
		1117
		1118	// the beauty of JSON: you need exactly one character lookahead
		1119	// to parse anything.
804	switch (*dec->cur)	1120	switch (*dec->cur)
805	{	1121	{
806	case '"': ++dec->cur; return decode_str (dec);	1122	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1123	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1124	case '{': ++dec->cur; return decode_hv (dec);
…		…
814		1130
815	case 't':	1131	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1132	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1133	{
818	dec->cur += 4;	1134	dec->cur += 4;
819	return newSViv (1);	1135	return SvREFCNT_inc (json_true);
820	}	1136	}
821	else	1137	else
822	ERR ("'true' expected");	1138	ERR ("'true' expected");
823		1139
824	break;	1140	break;
825		1141
826	case 'f':	1142	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1143	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1144	{
829	dec->cur += 5;	1145	dec->cur += 5;
830	return newSViv (0);	1146	return SvREFCNT_inc (json_false);
831	}	1147	}
832	else	1148	else
833	ERR ("'false' expected");	1149	ERR ("'false' expected");
834		1150
835	break;	1151	break;
…		…
844	ERR ("'null' expected");	1160	ERR ("'null' expected");
845		1161
846	break;	1162	break;
847		1163
848	default:	1164	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1165	ERR ("malformed JSON string, neither array, object, number, string or atom");
850	break;	1166	break;
851	}	1167	}
852		1168
853	fail:	1169	fail:
854	return 0;	1170	return 0;
855	}	1171	}
856		1172
857	static SV *	1173	static SV *
858	decode_json (SV *string, UV flags)	1174	decode_json (SV *string, struct json *json, UV *offset_return)
859	{	1175	{
		1176	dec_t dec;
		1177	UV offset;
860	SV *sv;	1178	SV *sv;
861		1179
		1180	SvGETMAGIC (string);
		1181	SvUPGRADE (string, SVt_PV);
		1182
		1183	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1184	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1185	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1186
862	if (flags & F_UTF8)	1187	if (json->flags & F_UTF8)
863	sv_utf8_downgrade (string, 0);	1188	sv_utf8_downgrade (string, 0);
864	else	1189	else
865	sv_utf8_upgrade (string);	1190	sv_utf8_upgrade (string);
866		1191
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1192	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1193
869	dec_t dec;	1194	dec.json = *json;
870	dec.flags = flags;
871	dec.cur = SvPVX (string);	1195	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1196	dec.end = SvEND (string);
873	dec.err = 0;	1197	dec.err = 0;
		1198	dec.depth = 0;
		1199	dec.maxdepth = DEC_DEPTH (dec.json.flags);
874		1200
		1201	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
875	sv = decode_sv (&dec);	1202	sv = decode_sv (&dec);
876		1203
		1204	if (!(offset_return || !sv))
		1205	{
		1206	// check for trailing garbage
		1207	decode_ws (&dec);
		1208
		1209	if (*dec.cur)
		1210	{
		1211	dec.err = "garbage after JSON object";
		1212	SvREFCNT_dec (sv);
		1213	sv = 0;
		1214	}
		1215	}
		1216
		1217	if (offset_return || !sv)
		1218	{
		1219	offset = dec.json.flags & F_UTF8
		1220	? dec.cur - SvPVX (string)
		1221	: utf8_distance (dec.cur, SvPVX (string));
		1222
		1223	if (offset_return)
		1224	*offset_return = offset;
		1225	}
		1226
877	if (!sv)	1227	if (!sv)
878	{	1228	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1229	SV *uni = sv_newmortal ();
		1230
883	// horrible hack to silence warning inside pv_uni_display	1231	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1232	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1233	cop.cop_warnings = pWARN_NONE;
		1234	ENTER;
887	SAVEVPTR (PL_curcop);	1235	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1236	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1237	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1238	LEAVE;
		1239
891	croak ("%s, at character offset %d (%s)",	1240	croak ("%s, at character offset %d [\"%s\"]",
892	dec.err,	1241	dec.err,
893	(int)offset,	1242	(int)offset,
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1243	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1244	}
896		1245
897	sv = sv_2mortal (sv);	1246	sv = sv_2mortal (sv);
898		1247
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1248	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
900	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1249	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1250
902	return sv;	1251	return sv;
903	}	1252	}
904		1253
		1254	/////////////////////////////////////////////////////////////////////////////
		1255	// XS interface functions
		1256
905	MODULE = JSON::XS PACKAGE = JSON::XS	1257	MODULE = JSON::XS PACKAGE = JSON::XS
906		1258
907	BOOT:	1259	BOOT:
908	{	1260	{
909	int i;	1261	int i;
910		1262
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1263	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1264	decode_hexdigit [i] =
		1265	i >= '0' && i <= '9' ? i - '0'
		1266	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1267	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1268	: -1;
914		1269
915	for (i = 7; i--; )
916	{
917	decode_hexdigit ['a' + i] = 10 + i;
918	decode_hexdigit ['A' + i] = 10 + i;
919	}
920
921	json_stash = gv_stashpv ("JSON::XS", 1);	1270	json_stash = gv_stashpv ("JSON::XS" , 1);
		1271	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1272
		1273	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1274	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
922	}	1275	}
923		1276
924	PROTOTYPES: DISABLE	1277	PROTOTYPES: DISABLE
925		1278
926	SV *new (char *dummy)	1279	SV *new (char *dummy)
…		…
929	OUTPUT:	1282	OUTPUT:
930	RETVAL	1283	RETVAL
931		1284
932	SV *ascii (SV *self, int enable = 1)	1285	SV *ascii (SV *self, int enable = 1)
933	ALIAS:	1286	ALIAS:
934	ascii = F_ASCII	1287	ascii = F_ASCII
		1288	latin1 = F_LATIN1
935	utf8 = F_UTF8	1289	utf8 = F_UTF8
936	indent = F_INDENT	1290	indent = F_INDENT
937	canonical = F_CANONICAL	1291	canonical = F_CANONICAL
938	space_before = F_SPACE_BEFORE	1292	space_before = F_SPACE_BEFORE
939	space_after = F_SPACE_AFTER	1293	space_after = F_SPACE_AFTER
940	json_rpc = F_JSON_RPC
941	pretty = F_PRETTY	1294	pretty = F_PRETTY
942	allow_nonref = F_ALLOW_NONREF	1295	allow_nonref = F_ALLOW_NONREF
943	shrink = F_SHRINK	1296	shrink = F_SHRINK
		1297	allow_blessed = F_ALLOW_BLESSED
		1298	convert_blessed = F_CONV_BLESSED
944	CODE:	1299	CODE:
945	{	1300	{
946	UV *uv = SvJSON (self);	1301	UV *uv = SvJSON (self);
947	if (enable)	1302	if (enable)
948	*uv |= ix;	1303	*uv |= ix;
…		…
952	RETVAL = newSVsv (self);	1307	RETVAL = newSVsv (self);
953	}	1308	}
954	OUTPUT:	1309	OUTPUT:
955	RETVAL	1310	RETVAL
956		1311
		1312	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)
		1313	CODE:
		1314	{
		1315	UV *uv = SvJSON (self);
		1316	UV log2 = 0;
		1317
		1318	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1319
		1320	while ((1UL << log2) < max_depth)
		1321	++log2;
		1322
		1323	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1324
		1325	RETVAL = newSVsv (self);
		1326	}
		1327	OUTPUT:
		1328	RETVAL
		1329
		1330	SV *max_size (SV *self, UV max_size = 0)
		1331	CODE:
		1332	{
		1333	UV *uv = SvJSON (self);
		1334	UV log2 = 0;
		1335
		1336	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1337	if (max_size == 1) max_size = 2;
		1338
		1339	while ((1UL << log2) < max_size)
		1340	++log2;
		1341
		1342	*uv = *uv & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1343
		1344	RETVAL = newSVsv (self);
		1345	}
		1346	OUTPUT:
		1347	RETVAL
		1348
957	void encode (SV *self, SV *scalar)	1349	void encode (SV *self, SV *scalar)
958	PPCODE:	1350	PPCODE:
		1351	{
		1352	struct json json = { *SvJSON (self) };
959	XPUSHs (encode_json (scalar, *SvJSON (self)));	1353	XPUSHs (encode_json (scalar, &json));
		1354	}
960		1355
961	void decode (SV *self, SV *jsonstr)	1356	void decode (SV *self, SV *jsonstr)
962	PPCODE:	1357	PPCODE:
		1358	{
		1359	struct json json = { *SvJSON (self) };
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1360	XPUSHs (decode_json (jsonstr, &json, 0));
		1361	}
		1362
		1363	void decode_prefix (SV *self, SV *jsonstr)
		1364	PPCODE:
		1365	{
		1366	UV offset;
		1367	struct json json = { *SvJSON (self) };
		1368	EXTEND (SP, 2);
		1369	PUSHs (decode_json (jsonstr, &json, &offset));
		1370	PUSHs (sv_2mortal (newSVuv (offset)));
		1371	}
964		1372
965	PROTOTYPES: ENABLE	1373	PROTOTYPES: ENABLE
966		1374
967	void to_json (SV *scalar)	1375	void to_json (SV *scalar)
968	PPCODE:	1376	PPCODE:
		1377	{
		1378	struct json json = { F_DEFAULT | F_UTF8 };
969	XPUSHs (encode_json (scalar, F_UTF8));	1379	XPUSHs (encode_json (scalar, &json));
		1380	}
970		1381
971	void from_json (SV *jsonstr)	1382	void from_json (SV *jsonstr)
972	PPCODE:	1383	PPCODE:
		1384	{
		1385	struct json json = { F_DEFAULT | F_UTF8 };
973	XPUSHs (decode_json (jsonstr, F_UTF8));	1386	XPUSHs (decode_json (jsonstr, &json, 0));
		1387	}
974		1388

Diff Legend

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