ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.4 by root, Thu Mar 22 21:13:58 2007 UTC vs.
Revision 1.56 by root, Thu Jul 26 11:33:35 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
		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 ))
17		40
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	41	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		43
21	#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
22		48
23	#define SB do {	49	#define SB do {
24	#define SE } while (0)	50	#define SE } while (0)
25		51
26	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	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		64	static SV *json_true, *json_false;
		65
		66	typedef struct {
		67	U32 flags;
		68	SV *cb_object;
		69	HV *cb_sk_object;
		70	} JSON;
		71
		72	/////////////////////////////////////////////////////////////////////////////
		73	// utility functions
		74
		75	inline void
		76	shrink (SV *sv)
		77	{
		78	sv_utf8_downgrade (sv, 1);
		79	if (SvLEN (sv) > SvCUR (sv) + 1)
		80	{
		81	#ifdef SvPV_shrink_to_cur
		82	SvPV_shrink_to_cur (sv);
		83	#elif defined (SvPV_renew)
		84	SvPV_renew (sv, SvCUR (sv) + 1);
		85	#endif
		86	}
		87	}
		88
		89	// decode an utf-8 character and return it, or (UV)-1 in
		90	// case of an error.
		91	// we special-case "safe" characters from U+80 .. U+7FF,
		92	// but use the very good perl function to parse anything else.
		93	// note that we never call this function for a ascii codepoints
		94	inline UV
		95	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		96	{
		97	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		98	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		99	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		100	{
		101	*clen = 2;
		102	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		103	}
		104	else
		105	{
		106	*clen = (STRLEN)-1;
		107	return (UV)-1;
		108	}
		109	}
		110
		111	/////////////////////////////////////////////////////////////////////////////
		112	// encoder
27		113
28	// structure used for encoding JSON	114	// structure used for encoding JSON
29	typedef struct	115	typedef struct
30	{	116	{
31	char *cur;	117	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	118	char *end; // SvEND (sv)
34	SV *sv;	119	SV *sv; // result scalar
35	UV flags;	120	JSON json;
36	int max_recurse;	121	U32 indent; // indentation level
37	int indent;	122	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	123	} enc_t;
39		124
40	// structure used for decoding JSON	125	inline void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	/////////////////////////////////////////////////////////////////////////////
59
60	static void
61	need (enc_t *enc, STRLEN len)	126	need (enc_t *enc, STRLEN len)
62	{	127	{
63	if (enc->cur + len >= enc->end)	128	if (expect_false (enc->cur + len >= enc->end))
64	{	129	{
65	STRLEN cur = enc->cur - SvPVX (enc->sv);	130	STRLEN cur = enc->cur - SvPVX (enc->sv);
66	SvGROW (enc->sv, cur + len + 1);	131	SvGROW (enc->sv, cur + len + 1);
67	enc->cur = SvPVX (enc->sv) + cur;	132	enc->cur = SvPVX (enc->sv) + cur;
68	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	133	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
69	}	134	}
70	}	135	}
71		136
72	static void	137	inline void
73	encode_ch (enc_t *enc, char ch)	138	encode_ch (enc_t *enc, char ch)
74	{	139	{
75	need (enc, 1);	140	need (enc, 1);
76	*enc->cur++ = ch;	141	*enc->cur++ = ch;
77	}	142	}
…		…
85		150
86	while (str < end)	151	while (str < end)
87	{	152	{
88	unsigned char ch = *(unsigned char *)str;	153	unsigned char ch = *(unsigned char *)str;
89		154
90	if (ch == '"')	155	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
91	{	156	{
		157	if (expect_false (ch == '"')) // but with slow exceptions
		158	{
92	need (enc, len += 1);	159	need (enc, len += 1);
93	*enc->cur++ = '\\';	160	*enc->cur++ = '\\';
94	*enc->cur++ = '"';	161	*enc->cur++ = '"';
95	++str;
96	}	162	}
97	else if (ch == '\\')	163	else if (expect_false (ch == '\\'))
98	{	164	{
99	need (enc, len += 1);	165	need (enc, len += 1);
100	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
101	*enc->cur++ = '\\';	167	*enc->cur++ = '\\';
102	++str;
103	}	168	}
104	else if (ch >= 0x20 && ch < 0x80) // most common case	169	else
105	{
106	*enc->cur++ = ch;	170	*enc->cur++ = ch;
107	++str;	171
108	}
109	else if (ch == '\015')
110	{
111	need (enc, len += 1);
112	*enc->cur++ = '\\';
113	*enc->cur++ = 'r';
114	++str;
115	}
116	else if (ch == '\012')
117	{
118	need (enc, len += 1);
119	*enc->cur++ = '\\';
120	*enc->cur++ = 'n';
121	++str;	172	++str;
122	}	173	}
123	else	174	else
124	{	175	{
125	STRLEN clen;	176	switch (ch)
126	UV uch;
127
128	if (is_utf8)
129	{
130	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
131	if (clen < 0)
132	croak ("malformed UTF-8 character in string, cannot convert to JSON");
133	}	177	{
134	else	178	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
135	{	179	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
136	uch = ch;	180	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
137	clen = 1;	181	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
138	}	182	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
139		183
140	if (uch < 0x80 || enc->flags & F_ASCII)	184	default:
141	{
142	if (uch > 0xFFFFUL)
143	{	185	{
		186	STRLEN clen;
		187	UV uch;
		188
		189	if (is_utf8)
		190	{
		191	uch = decode_utf8 (str, end - str, &clen);
		192	if (clen == (STRLEN)-1)
		193	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		194	}
		195	else
		196	{
		197	uch = ch;
		198	clen = 1;
		199	}
		200
		201	if (uch > 0x10FFFFUL)
		202	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		203
		204	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		205	{
		206	if (uch > 0xFFFFUL)
		207	{
144	need (enc, len += 11);	208	need (enc, len += 11);
145	sprintf (enc->cur, "\\u%04x\\u%04x",	209	sprintf (enc->cur, "\\u%04x\\u%04x",
146	(uch - 0x10000) / 0x400 + 0xD800,	210	(int)((uch - 0x10000) / 0x400 + 0xD800),
147	(uch - 0x10000) % 0x400 + 0xDC00);	211	(int)((uch - 0x10000) % 0x400 + 0xDC00));
148	enc->cur += 12;	212	enc->cur += 12;
		213	}
		214	else
		215	{
		216	static char hexdigit [16] = "0123456789abcdef";
		217	need (enc, len += 5);
		218	*enc->cur++ = '\\';
		219	*enc->cur++ = 'u';
		220	*enc->cur++ = hexdigit [ uch >> 12 ];
		221	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		222	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		223	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		224	}
		225
		226	str += clen;
		227	}
		228	else if (enc->json.flags & F_LATIN1)
		229	{
		230	*enc->cur++ = uch;
		231	str += clen;
		232	}
		233	else if (is_utf8)
		234	{
		235	need (enc, len += clen);
		236	do
		237	{
		238	*enc->cur++ = *str++;
		239	}
		240	while (--clen);
		241	}
		242	else
		243	{
		244	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		245	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		246	++str;
		247	}
149	}	248	}
150	else
151	{
152	static char hexdigit [16] = "0123456789abcdef";
153	need (enc, len += 5);
154	*enc->cur++ = '\\';
155	*enc->cur++ = 'u';
156	*enc->cur++ = hexdigit [ uch >> 12 ];
157	*enc->cur++ = hexdigit [(uch >> 8) & 15];
158	*enc->cur++ = hexdigit [(uch >> 4) & 15];
159	*enc->cur++ = hexdigit [(uch >> 0) & 15];
160	}
161
162	str += clen;
163	}
164	else if (is_utf8)
165	{
166	need (enc, len += clen);
167	while (clen--)
168	*enc->cur++ = *str++;
169	}
170	else
171	{
172	need (enc, 10); // never more than 11 bytes needed
173	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
174	++str;
175	}	249	}
176	}	250	}
177		251
178	--len;	252	--len;
179	}	253	}
180	}	254	}
181		255
182	#define INDENT SB \	256	inline void
		257	encode_indent (enc_t *enc)
		258	{
183	if (enc->flags & F_INDENT) \	259	if (enc->json.flags & F_INDENT)
184	{ \	260	{
185	int i_; \	261	int spaces = enc->indent * INDENT_STEP;
186	need (enc, enc->indent); \	262
187	for (i_ = enc->indent * 3; i_--; )\	263	need (enc, spaces);
		264	memset (enc->cur, ' ', spaces);
		265	enc->cur += spaces;
		266	}
		267	}
		268
		269	inline void
		270	encode_space (enc_t *enc)
		271	{
		272	need (enc, 1);
		273	encode_ch (enc, ' ');
		274	}
		275
		276	inline void
		277	encode_nl (enc_t *enc)
		278	{
		279	if (enc->json.flags & F_INDENT)
		280	{
		281	need (enc, 1);
188	encode_ch (enc, ' '); \	282	encode_ch (enc, '\n');
189	} \	283	}
190	SE	284	}
191		285
192	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	286	inline void
193	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	287	encode_comma (enc_t *enc)
194	#define COMMA SB \	288	{
195	encode_ch (enc, ','); \	289	encode_ch (enc, ',');
		290
196	if (enc->flags & F_INDENT) \	291	if (enc->json.flags & F_INDENT)
197	NL; \	292	encode_nl (enc);
198	else if (enc->flags & F_SPACE_AFTER) \	293	else if (enc->json.flags & F_SPACE_AFTER)
199	SPACE; \	294	encode_space (enc);
200	SE	295	}
201		296
202	static void encode_sv (enc_t *enc, SV *sv);	297	static void encode_sv (enc_t *enc, SV *sv);
203		298
204	static void	299	static void
205	encode_av (enc_t *enc, AV *av)	300	encode_av (enc_t *enc, AV *av)
206	{	301	{
207	int i, len = av_len (av);	302	int i, len = av_len (av);
208		303
209	encode_ch (enc, '['); NL;	304	if (enc->indent >= enc->maxdepth)
		305	croak ("data structure too deep (hit recursion limit)");
		306
		307	encode_ch (enc, '['); encode_nl (enc);
210	++enc->indent;	308	++enc->indent;
211		309
212	for (i = 0; i <= len; ++i)	310	for (i = 0; i <= len; ++i)
213	{	311	{
214	INDENT;	312	SV **svp = av_fetch (av, i, 0);
215	encode_sv (enc, *av_fetch (av, i, 0));	313
		314	encode_indent (enc);
		315
		316	if (svp)
		317	encode_sv (enc, *svp);
		318	else
		319	encode_str (enc, "null", 4, 0);
216		320
217	if (i < len)	321	if (i < len)
218	COMMA;	322	encode_comma (enc);
219	}	323	}
220		324
221	NL;	325	encode_nl (enc);
222		326
223	--enc->indent;	327	--enc->indent;
224	INDENT; encode_ch (enc, ']');	328	encode_indent (enc); encode_ch (enc, ']');
225	}	329	}
226		330
227	static void	331	static void
228	encode_he (enc_t *enc, HE *he)	332	encode_he (enc_t *enc, HE *he)
229	{	333	{
…		…
243	else	347	else
244	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	348	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
245		349
246	encode_ch (enc, '"');	350	encode_ch (enc, '"');
247		351
248	if (enc->flags & F_SPACE_BEFORE) SPACE;	352	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
249	encode_ch (enc, ':');	353	encode_ch (enc, ':');
250	if (enc->flags & F_SPACE_AFTER ) SPACE;	354	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
251	encode_sv (enc, HeVAL (he));	355	encode_sv (enc, HeVAL (he));
252	}	356	}
253		357
254	// compare hash entries, used when all keys are bytestrings	358	// compare hash entries, used when all keys are bytestrings
255	static int	359	static int
…		…
261	HE *b = *(HE **)b_;	365	HE *b = *(HE **)b_;
262		366
263	STRLEN la = HeKLEN (a);	367	STRLEN la = HeKLEN (a);
264	STRLEN lb = HeKLEN (b);	368	STRLEN lb = HeKLEN (b);
265		369
266	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	370	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
267	cmp = la < lb ? -1 : la == lb ? 0 : 1;	371	cmp = la - lb;
268		372
269	return cmp;	373	return cmp;
270	}	374	}
271		375
272	// compare hash entries, used when some keys are sv's or utf-x	376	// compare hash entries, used when some keys are sv's or utf-x
…		…
279	static void	383	static void
280	encode_hv (enc_t *enc, HV *hv)	384	encode_hv (enc_t *enc, HV *hv)
281	{	385	{
282	int count, i;	386	int count, i;
283		387
		388	if (enc->indent >= enc->maxdepth)
		389	croak ("data structure too deep (hit recursion limit)");
		390
284	encode_ch (enc, '{'); NL; ++enc->indent;	391	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
285		392
286	if ((count = hv_iterinit (hv)))	393	if ((count = hv_iterinit (hv)))
287	{	394	{
288	// for canonical output we have to sort by keys first	395	// for canonical output we have to sort by keys first
289	// actually, this is mostly due to the stupid so-called	396	// actually, this is mostly due to the stupid so-called
290	// security workaround added somewhere in 5.8.x.	397	// security workaround added somewhere in 5.8.x.
291	// that randomises hash orderings	398	// that randomises hash orderings
292	if (enc->flags & F_CANONICAL)	399	if (enc->json.flags & F_CANONICAL)
293	{	400	{
294	HE *he, *hes [count];
295	int fast = 1;	401	int fast = 1;
		402	HE *he;
		403	#if defined(__BORLANDC__) || defined(_MSC_VER)
		404	HE **hes = _alloca (count * sizeof (HE));
		405	#else
		406	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		407	#endif
296		408
297	i = 0;	409	i = 0;
298	while ((he = hv_iternext (hv)))	410	while ((he = hv_iternext (hv)))
299	{	411	{
300	hes [i++] = he;	412	hes [i++] = he;
…		…
306		418
307	if (fast)	419	if (fast)
308	qsort (hes, count, sizeof (HE *), he_cmp_fast);	420	qsort (hes, count, sizeof (HE *), he_cmp_fast);
309	else	421	else
310	{	422	{
311	// hack to disable "use bytes"	423	// hack to forcefully disable "use bytes"
312	COP *oldcop = PL_curcop, cop;	424	COP cop = *PL_curcop;
313	cop.op_private = 0;	425	cop.op_private = 0;
		426
		427	ENTER;
		428	SAVETMPS;
		429
		430	SAVEVPTR (PL_curcop);
314	PL_curcop = &cop;	431	PL_curcop = &cop;
315		432
316	SAVETMPS;
317	qsort (hes, count, sizeof (HE *), he_cmp_slow);	433	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		434
318	FREETMPS;	435	FREETMPS;
319		436	LEAVE;
320	PL_curcop = oldcop;
321	}	437	}
322		438
323	for (i = 0; i < count; ++i)	439	for (i = 0; i < count; ++i)
324	{	440	{
325	INDENT;	441	encode_indent (enc);
326	encode_he (enc, hes [i]);	442	encode_he (enc, hes [i]);
327		443
328	if (i < count - 1)	444	if (i < count - 1)
329	COMMA;	445	encode_comma (enc);
330	}
331
332	NL;	446	}
		447
		448	encode_nl (enc);
333	}	449	}
334	else	450	else
335	{	451	{
336	SV *sv;
337	HE *he = hv_iternext (hv);	452	HE *he = hv_iternext (hv);
338		453
339	for (;;)	454	for (;;)
340	{	455	{
341	INDENT;	456	encode_indent (enc);
342	encode_he (enc, he);	457	encode_he (enc, he);
343		458
344	if (!(he = hv_iternext (hv)))	459	if (!(he = hv_iternext (hv)))
345	break;	460	break;
346		461
347	COMMA;	462	encode_comma (enc);
348	}
349
350	NL;	463	}
		464
		465	encode_nl (enc);
351	}	466	}
352	}	467	}
353		468
354	--enc->indent; INDENT; encode_ch (enc, '}');	469	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		470	}
		471
		472	// encode objects, arrays and special \0=false and \1=true values.
		473	static void
		474	encode_rv (enc_t *enc, SV *sv)
		475	{
		476	svtype svt;
		477
		478	SvGETMAGIC (sv);
		479	svt = SvTYPE (sv);
		480
		481	if (expect_false (SvOBJECT (sv)))
		482	{
		483	if (SvSTASH (sv) == json_boolean_stash)
		484	{
		485	if (SvIV (sv))
		486	encode_str (enc, "true", 4, 0);
		487	else
		488	encode_str (enc, "false", 5, 0);
		489	}
		490	else
		491	{
		492	#if 0
		493	if (0 && sv_derived_from (rv, "JSON::Literal"))
		494	{
		495	// not yet
		496	}
		497	#endif
		498	if (enc->json.flags & F_CONV_BLESSED)
		499	{
		500	// we re-bless the reference to get overload and other niceties right
		501	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		502
		503	if (to_json)
		504	{
		505	int count;
		506	dSP;
		507
		508	ENTER; SAVETMPS; PUSHMARK (SP);
		509	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		510
		511	// calling with G_SCALAR ensures that we always get a 1 return value
		512	PUTBACK;
		513	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		514	SPAGAIN;
		515
		516	// catch this surprisingly common error
		517	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		518	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		519
		520	sv = POPs;
		521	PUTBACK;
		522
		523	encode_sv (enc, sv);
		524
		525	FREETMPS; LEAVE;
		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 enabled nor TO_JSON method available on it",
		531	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		532	}
		533	else if (enc->json.flags & F_ALLOW_BLESSED)
		534	encode_str (enc, "null", 4, 0);
		535	else
		536	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		537	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		538	}
		539	}
		540	else if (svt == SVt_PVHV)
		541	encode_hv (enc, (HV *)sv);
		542	else if (svt == SVt_PVAV)
		543	encode_av (enc, (AV *)sv);
		544	else if (svt < SVt_PVAV)
		545	{
		546	STRLEN len = 0;
		547	char *pv = svt ? SvPV (sv, len) : 0;
		548
		549	if (len == 1 && *pv == '1')
		550	encode_str (enc, "true", 4, 0);
		551	else if (len == 1 && *pv == '0')
		552	encode_str (enc, "false", 5, 0);
		553	else
		554	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		555	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		556	}
		557	else
		558	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		559	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
355	}	560	}
356		561
357	static void	562	static void
358	encode_sv (enc_t *enc, SV *sv)	563	encode_sv (enc_t *enc, SV *sv)
359	{	564	{
…		…
367	encode_str (enc, str, len, SvUTF8 (sv));	572	encode_str (enc, str, len, SvUTF8 (sv));
368	encode_ch (enc, '"');	573	encode_ch (enc, '"');
369	}	574	}
370	else if (SvNOKp (sv))	575	else if (SvNOKp (sv))
371	{	576	{
		577	// trust that perl will do the right thing w.r.t. JSON syntax.
372	need (enc, NV_DIG + 32);	578	need (enc, NV_DIG + 32);
373	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	579	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
374	enc->cur += strlen (enc->cur);	580	enc->cur += strlen (enc->cur);
375	}	581	}
376	else if (SvIOKp (sv))	582	else if (SvIOKp (sv))
377	{	583	{
378	need (enc, 64);	584	// we assume we can always read an IV as a UV
		585	if (SvUV (sv) & ~(UV)0x7fff)
		586	{
		587	// large integer, use the (rather slow) snprintf way.
		588	need (enc, sizeof (UV) * 3);
379	enc->cur +=	589	enc->cur +=
380	SvIsUV(sv)	590	SvIsUV(sv)
381	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	591	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
382	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	592	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		593	}
		594	else
		595	{
		596	// optimise the "small number case"
		597	// code will likely be branchless and use only a single multiplication
		598	I32 i = SvIV (sv);
		599	U32 u;
		600	char digit, nz = 0;
		601
		602	need (enc, 6);
		603
		604	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		605	u = i < 0 ? -i : i;
		606
		607	// convert to 4.28 fixed-point representation
		608	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		609
		610	// now output digit by digit, each time masking out the integer part
		611	// and multiplying by 5 while moving the decimal point one to the right,
		612	// resulting in a net multiplication by 10.
		613	// we always write the digit to memory but conditionally increment
		614	// the pointer, to ease the usage of conditional move instructions.
		615	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		616	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		617	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		618	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		619	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		620	}
383	}	621	}
384	else if (SvROK (sv))	622	else if (SvROK (sv))
385	{	623	encode_rv (enc, SvRV (sv));
386	if (!--enc->max_recurse)
387	croak ("data structure too deep (hit recursion limit)");
388
389	sv = SvRV (sv);
390
391	switch (SvTYPE (sv))
392	{
393	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
394	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
395
396	default:
397	croak ("JSON can only represent references to arrays or hashes");
398	}
399	}
400	else if (!SvOK (sv))	624	else if (!SvOK (sv))
401	encode_str (enc, "null", 4, 0);	625	encode_str (enc, "null", 4, 0);
402	else	626	else
403	croak ("encountered perl type that JSON cannot handle");	627	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		628	SvPV_nolen (sv), SvFLAGS (sv));
404	}	629	}
405		630
406	static SV *	631	static SV *
407	encode_json (SV *scalar, UV flags)	632	encode_json (SV *scalar, JSON *json)
408	{	633	{
409	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
410	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
411
412	enc_t enc;	634	enc_t enc;
413	enc.flags = flags;	635
		636	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		637	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		638
		639	enc.json = *json;
414	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	640	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
415	enc.cur = SvPVX (enc.sv);	641	enc.cur = SvPVX (enc.sv);
416	enc.end = SvEND (enc.sv);	642	enc.end = SvEND (enc.sv);
417	enc.max_recurse = 0;
418	enc.indent = 0;	643	enc.indent = 0;
		644	enc.maxdepth = DEC_DEPTH (enc.json.flags);
419		645
420	SvPOK_only (enc.sv);	646	SvPOK_only (enc.sv);
421	encode_sv (&enc, scalar);	647	encode_sv (&enc, scalar);
422		648
		649	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		650	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		651
423	if (!(flags & (F_ASCII | F_UTF8)))	652	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
424	SvUTF8_on (enc.sv);	653	SvUTF8_on (enc.sv);
425		654
426	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	655	if (enc.json.flags & F_SHRINK)
		656	shrink (enc.sv);
		657
427	return enc.sv;	658	return enc.sv;
428	}	659	}
429		660
430	/////////////////////////////////////////////////////////////////////////////	661	/////////////////////////////////////////////////////////////////////////////
		662	// decoder
431		663
432	#define WS \	664	// structure used for decoding JSON
		665	typedef struct
		666	{
		667	char *cur; // current parser pointer
		668	char *end; // end of input string
		669	const char *err; // parse error, if != 0
		670	JSON json;
		671	U32 depth; // recursion depth
		672	U32 maxdepth; // recursion depth limit
		673	} dec_t;
		674
		675	inline void
		676	decode_ws (dec_t *dec)
		677	{
433	for (;;) \	678	for (;;)
434	{ \	679	{
435	char ch = *dec->cur; \	680	char ch = *dec->cur;
		681
436	if (ch > 0x20 \	682	if (ch > 0x20
437	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	683	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
438	break; \	684	break;
		685
439	++dec->cur; \	686	++dec->cur;
440	}	687	}
		688	}
441		689
442	#define ERR(reason) SB dec->err = reason; goto fail; SE	690	#define ERR(reason) SB dec->err = reason; goto fail; SE
		691
443	#define EXPECT_CH(ch) SB \	692	#define EXPECT_CH(ch) SB \
444	if (*dec->cur != ch) \	693	if (*dec->cur != ch) \
445	ERR (# ch " expected"); \	694	ERR (# ch " expected"); \
446	++dec->cur; \	695	++dec->cur; \
447	SE	696	SE
448		697
		698	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		699	#define DEC_DEC_DEPTH --dec->depth
		700
449	static SV *decode_sv (dec_t *dec);	701	static SV *decode_sv (dec_t *dec);
450		702
451	static signed char decode_hexdigit[256];	703	static signed char decode_hexdigit[256];
452		704
453	static UV	705	static UV
454	decode_4hex (dec_t *dec)	706	decode_4hex (dec_t *dec)
455	{	707	{
456	signed char d1, d2, d3, d4;	708	signed char d1, d2, d3, d4;
		709	unsigned char *cur = (unsigned char *)dec->cur;
457		710
458	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	711	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
459	if (d1 < 0) ERR ("four hexadecimal digits expected");	712	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
460	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	713	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
461	if (d2 < 0) ERR ("four hexadecimal digits expected");	714	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
462	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
463	if (d3 < 0) ERR ("four hexadecimal digits expected");
464	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
465	if (d4 < 0) ERR ("four hexadecimal digits expected");
466		715
467	dec->cur += 4;	716	dec->cur += 4;
468		717
469	return ((UV)d1) << 12	718	return ((UV)d1) << 12
470	| ((UV)d2) << 8	719	| ((UV)d2) << 8
…		…
473		722
474	fail:	723	fail:
475	return (UV)-1;	724	return (UV)-1;
476	}	725	}
477		726
478	#define APPEND_GROW(n) SB \
479	if (cur + (n) >= end) \
480	{ \
481	STRLEN ofs = cur - SvPVX (sv); \
482	SvGROW (sv, ofs + (n) + 1); \
483	cur = SvPVX (sv) + ofs; \
484	end = SvEND (sv); \
485	} \
486	SE
487
488	#define APPEND_CH(ch) SB \
489	APPEND_GROW (1); \
490	*cur++ = (ch); \
491	SE
492
493	static SV *	727	static SV *
494	decode_str (dec_t *dec)	728	decode_str (dec_t *dec)
495	{	729	{
496	SV *sv = NEWSV (0,2);	730	SV *sv = 0;
497	int utf8 = 0;	731	int utf8 = 0;
498	char *cur = SvPVX (sv);	732	char *dec_cur = dec->cur;
499	char *end = SvEND (sv);
500		733
501	for (;;)	734	do
502	{	735	{
503	unsigned char ch = *(unsigned char *)dec->cur;	736	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		737	char *cur = buf;
504		738
505	if (ch == '"')	739	do
506	break;
507	else if (ch == '\\')
508	{	740	{
509	switch (*++dec->cur)	741	unsigned char ch = *(unsigned char *)dec_cur++;
		742
		743	if (expect_false (ch == '"'))
		744	{
		745	--dec_cur;
		746	break;
510	{	747	}
511	case '\\':	748	else if (expect_false (ch == '\\'))
512	case '/':	749	{
513	case '"': APPEND_CH (*dec->cur++); break;	750	switch (*dec_cur)
514
515	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
516	case 't': APPEND_CH ('\011'); ++dec->cur; break;
517	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
518	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
519	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
520
521	case 'u':
522	{	751	{
523	UV lo, hi;	752	case '\\':
524	++dec->cur;	753	case '/':
		754	case '"': *cur++ = *dec_cur++; break;
525		755
526	hi = decode_4hex (dec);	756	case 'b': ++dec_cur; *cur++ = '\010'; break;
527	if (hi == (UV)-1)	757	case 't': ++dec_cur; *cur++ = '\011'; break;
528	goto fail;	758	case 'n': ++dec_cur; *cur++ = '\012'; break;
		759	case 'f': ++dec_cur; *cur++ = '\014'; break;
		760	case 'r': ++dec_cur; *cur++ = '\015'; break;
529		761
530	// possibly a surrogate pair	762	case 'u':
531	if (hi >= 0xd800 && hi < 0xdc00)
532	{	763	{
533	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	764	UV lo, hi;
534	ERR ("illegal surrogate character");	765	++dec_cur;
535		766
536	dec->cur += 2;	767	dec->cur = dec_cur;
537
538	lo = decode_4hex (dec);	768	hi = decode_4hex (dec);
		769	dec_cur = dec->cur;
539	if (lo == (UV)-1)	770	if (hi == (UV)-1)
540	goto fail;	771	goto fail;
541		772
		773	// possibly a surrogate pair
		774	if (hi >= 0xd800)
		775	if (hi < 0xdc00)
		776	{
		777	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		778	ERR ("missing low surrogate character in surrogate pair");
		779
		780	dec_cur += 2;
		781
		782	dec->cur = dec_cur;
		783	lo = decode_4hex (dec);
		784	dec_cur = dec->cur;
		785	if (lo == (UV)-1)
		786	goto fail;
		787
542	if (lo < 0xdc00 || lo >= 0xe000)	788	if (lo < 0xdc00 || lo >= 0xe000)
543	ERR ("surrogate pair expected");	789	ERR ("surrogate pair expected");
544		790
545	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	791	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		792	}
		793	else if (hi < 0xe000)
		794	ERR ("missing high surrogate character in surrogate pair");
		795
		796	if (hi >= 0x80)
		797	{
		798	utf8 = 1;
		799
		800	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		801	}
		802	else
		803	*cur++ = hi;
546	}	804	}
547	else if (lo >= 0xdc00 && lo < 0xe000)
548	ERR ("illegal surrogate character");
549
550	if (hi >= 0x80)
551	{	805	break;
552	utf8 = 1;
553		806
554	APPEND_GROW (4); // at most 4 bytes for 21 bits
555	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
556	}
557	else	807	default:
558	APPEND_CH (hi);	808	--dec_cur;
		809	ERR ("illegal backslash escape sequence in string");
559	}	810	}
		811	}
		812	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		813	*cur++ = ch;
		814	else if (ch >= 0x80)
		815	{
		816	STRLEN clen;
		817	UV uch;
		818
		819	--dec_cur;
		820
		821	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		822	if (clen == (STRLEN)-1)
		823	ERR ("malformed UTF-8 character in JSON string");
		824
		825	do
		826	*cur++ = *dec_cur++;
		827	while (--clen);
		828
		829	utf8 = 1;
		830	}
		831	else
		832	{
		833	--dec_cur;
		834
		835	if (!ch)
		836	ERR ("unexpected end of string while parsing JSON string");
560	break;	837	else
		838	ERR ("invalid character encountered while parsing JSON string");
561	}	839	}
562	}	840	}
563	else if (ch >= 0x20 && ch <= 0x7f)	841	while (cur < buf + SHORT_STRING_LEN);
564	APPEND_CH (*dec->cur++);	842
565	else if (ch >= 0x80)	843	{
		844	STRLEN len = cur - buf;
		845
		846	if (sv)
566	{	847	{
567	STRLEN clen;	848	SvGROW (sv, SvCUR (sv) + len + 1);
568	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	849	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
569	if (clen < 0)	850	SvCUR_set (sv, SvCUR (sv) + len);
570	ERR ("malformed UTF-8 character in string, cannot convert to JSON");
571
572	APPEND_GROW (clen);
573	memcpy (cur, dec->cur, clen);
574	cur += clen;
575	dec->cur += clen;
576	}	851	}
577	else	852	else
578	ERR ("invalid character encountered");	853	sv = newSVpvn (buf, len);
579	}	854	}
		855	}
		856	while (*dec_cur != '"');
580		857
581	++dec->cur;	858	++dec_cur;
582		859
583	SvCUR_set (sv, cur - SvPVX (sv));	860	if (sv)
584		861	{
585	SvPOK_only (sv);	862	SvPOK_only (sv);
586	*SvEND (sv) = 0;	863	*SvEND (sv) = 0;
587		864
588	if (utf8)	865	if (utf8)
589	SvUTF8_on (sv);	866	SvUTF8_on (sv);
		867	}
		868	else
		869	sv = newSVpvn ("", 0);
590		870
		871	dec->cur = dec_cur;
591	return sv;	872	return sv;
592		873
593	fail:	874	fail:
594	SvREFCNT_dec (sv);	875	dec->cur = dec_cur;
595	return 0;	876	return 0;
596	}	877	}
597		878
598	static SV *	879	static SV *
599	decode_num (dec_t *dec)	880	decode_num (dec_t *dec)
…		…
609	{	890	{
610	++dec->cur;	891	++dec->cur;
611	if (*dec->cur >= '0' && *dec->cur <= '9')	892	if (*dec->cur >= '0' && *dec->cur <= '9')
612	ERR ("malformed number (leading zero must not be followed by another digit)");	893	ERR ("malformed number (leading zero must not be followed by another digit)");
613	}	894	}
614		895	else if (*dec->cur < '0' || *dec->cur > '9')
615	// int	896	ERR ("malformed number (no digits after initial minus)");
		897	else
		898	do
		899	{
		900	++dec->cur;
		901	}
616	while (*dec->cur >= '0' && *dec->cur <= '9')	902	while (*dec->cur >= '0' && *dec->cur <= '9');
617	++dec->cur;
618		903
619	// [frac]	904	// [frac]
620	if (*dec->cur == '.')	905	if (*dec->cur == '.')
621	{	906	{
622	is_nv = 1;	907	++dec->cur;
		908
		909	if (*dec->cur < '0' || *dec->cur > '9')
		910	ERR ("malformed number (no digits after decimal point)");
623		911
624	do	912	do
625	{	913	{
626	++dec->cur;	914	++dec->cur;
627	}	915	}
628	while (*dec->cur >= '0' && *dec->cur <= '9');	916	while (*dec->cur >= '0' && *dec->cur <= '9');
		917
		918	is_nv = 1;
629	}	919	}
630		920
631	// [exp]	921	// [exp]
632	if (*dec->cur == 'e' || *dec->cur == 'E')	922	if (*dec->cur == 'e' || *dec->cur == 'E')
633	{	923	{
634	is_nv = 1;
635
636	++dec->cur;	924	++dec->cur;
		925
637	if (*dec->cur == '-' || *dec->cur == '+')	926	if (*dec->cur == '-' || *dec->cur == '+')
638	++dec->cur;	927	++dec->cur;
639		928
		929	if (*dec->cur < '0' || *dec->cur > '9')
		930	ERR ("malformed number (no digits after exp sign)");
		931
		932	do
		933	{
		934	++dec->cur;
		935	}
640	while (*dec->cur >= '0' && *dec->cur <= '9')	936	while (*dec->cur >= '0' && *dec->cur <= '9');
641	++dec->cur;	937
		938	is_nv = 1;
642	}	939	}
643		940
644	if (!is_nv)	941	if (!is_nv)
645	{	942	{
646	UV uv;	943	int len = dec->cur - start;
647	int numtype = grok_number (start, dec->cur - start, &uv);	944
648	if (numtype & IS_NUMBER_IN_UV)	945	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
649	if (numtype & IS_NUMBER_NEG)	946	if (*start == '-')
		947	switch (len)
650	{	948	{
651	if (uv < (UV)IV_MIN)	949	case 2: return newSViv (-( start [1] - '0' * 1));
652	return newSViv (-(IV)uv);	950	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		951	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		952	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
653	}	953	}
		954	else
		955	switch (len)
		956	{
		957	case 1: return newSViv ( start [0] - '0' * 1);
		958	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		959	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		960	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		961	}
		962
		963	{
		964	UV uv;
		965	int numtype = grok_number (start, len, &uv);
		966	if (numtype & IS_NUMBER_IN_UV)
		967	if (numtype & IS_NUMBER_NEG)
		968	{
		969	if (uv < (UV)IV_MIN)
		970	return newSViv (-(IV)uv);
		971	}
654	else	972	else
655	return newSVuv (uv);	973	return newSVuv (uv);
656	}	974	}
657		975
		976	len -= *start == '-' ? 1 : 0;
		977
		978	// does not fit into IV or UV, try NV
		979	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		980	#if defined (LDBL_DIG)
		981	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		982	#endif
		983	)
		984	// fits into NV without loss of precision
		985	return newSVnv (Atof (start));
		986
		987	// everything else fails, convert it to a string
		988	return newSVpvn (start, dec->cur - start);
		989	}
		990
		991	// loss of precision here
658	return newSVnv (Atof (start));	992	return newSVnv (Atof (start));
659		993
660	fail:	994	fail:
661	return 0;	995	return 0;
662	}	996	}
…		…
664	static SV *	998	static SV *
665	decode_av (dec_t *dec)	999	decode_av (dec_t *dec)
666	{	1000	{
667	AV *av = newAV ();	1001	AV *av = newAV ();
668		1002
		1003	DEC_INC_DEPTH;
		1004	decode_ws (dec);
		1005
		1006	if (*dec->cur == ']')
		1007	++dec->cur;
		1008	else
669	for (;;)	1009	for (;;)
670	{	1010	{
671	SV *value;	1011	SV *value;
672		1012
673	value = decode_sv (dec);	1013	value = decode_sv (dec);
674	if (!value)	1014	if (!value)
675	goto fail;	1015	goto fail;
676		1016
677	av_push (av, value);	1017	av_push (av, value);
678		1018
679	WS;	1019	decode_ws (dec);
680		1020
681	if (*dec->cur == ']')	1021	if (*dec->cur == ']')
682	{	1022	{
683	++dec->cur;	1023	++dec->cur;
684	break;	1024	break;
		1025	}
685	}	1026
686
687	if (*dec->cur != ',')	1027	if (*dec->cur != ',')
688	ERR (", or ] expected while parsing array");	1028	ERR (", or ] expected while parsing array");
689		1029
690	++dec->cur;	1030	++dec->cur;
691	}	1031	}
692		1032
		1033	DEC_DEC_DEPTH;
693	return newRV_noinc ((SV *)av);	1034	return newRV_noinc ((SV *)av);
694		1035
695	fail:	1036	fail:
696	SvREFCNT_dec (av);	1037	SvREFCNT_dec (av);
		1038	DEC_DEC_DEPTH;
697	return 0;	1039	return 0;
698	}	1040	}
699		1041
700	static SV *	1042	static SV *
701	decode_hv (dec_t *dec)	1043	decode_hv (dec_t *dec)
702	{	1044	{
		1045	SV *sv;
703	HV *hv = newHV ();	1046	HV *hv = newHV ();
704		1047
		1048	DEC_INC_DEPTH;
		1049	decode_ws (dec);
		1050
		1051	if (*dec->cur == '}')
		1052	++dec->cur;
		1053	else
705	for (;;)	1054	for (;;)
706	{	1055	{
707	SV *key, *value;	1056	decode_ws (dec); EXPECT_CH ('"');
708		1057
709	WS; EXPECT_CH ('"');	1058	// heuristic: assume that
710		1059	// a) decode_str + hv_store_ent are abysmally slow.
711	key = decode_str (dec);	1060	// b) most hash keys are short, simple ascii text.
712	if (!key)	1061	// => try to "fast-match" such strings to avoid
713	goto fail;	1062	// the overhead of decode_str + hv_store_ent.
714
715	WS; EXPECT_CH (':');
716
717	value = decode_sv (dec);
718	if (!value)
719	{	1063	{
		1064	SV *value;
		1065	char *p = dec->cur;
		1066	char *e = p + 24; // only try up to 24 bytes
		1067
		1068	for (;;)
		1069	{
		1070	// the >= 0x80 is true on most architectures
		1071	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1072	{
		1073	// slow path, back up and use decode_str
		1074	SV *key = decode_str (dec);
		1075	if (!key)
		1076	goto fail;
		1077
		1078	decode_ws (dec); EXPECT_CH (':');
		1079
		1080	value = decode_sv (dec);
		1081	if (!value)
		1082	{
		1083	SvREFCNT_dec (key);
		1084	goto fail;
		1085	}
		1086
		1087	hv_store_ent (hv, key, value, 0);
720	SvREFCNT_dec (key);	1088	SvREFCNT_dec (key);
		1089
		1090	break;
		1091	}
		1092	else if (*p == '"')
		1093	{
		1094	// fast path, got a simple key
		1095	char *key = dec->cur;
		1096	int len = p - key;
		1097	dec->cur = p + 1;
		1098
		1099	decode_ws (dec); EXPECT_CH (':');
		1100
		1101	value = decode_sv (dec);
		1102	if (!value)
721	goto fail;	1103	goto fail;
		1104
		1105	hv_store (hv, key, len, value, 0);
		1106
		1107	break;
		1108	}
		1109
		1110	++p;
		1111	}
722	}	1112	}
723		1113
724	//TODO: optimise	1114	decode_ws (dec);
725	hv_store_ent (hv, key, value, 0);
726		1115
727	WS;
728
729	if (*dec->cur == '}')	1116	if (*dec->cur == '}')
		1117	{
		1118	++dec->cur;
		1119	break;
		1120	}
		1121
		1122	if (*dec->cur != ',')
		1123	ERR (", or } expected while parsing object/hash");
		1124
		1125	++dec->cur;
		1126	}
		1127
		1128	DEC_DEC_DEPTH;
		1129	sv = newRV_noinc ((SV *)hv);
		1130
		1131	// check filter callbacks
		1132	if (dec->json.flags & F_HOOK)
		1133	{
		1134	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
730	{	1135	{
731	++dec->cur;	1136	HE *cb, *he;
732	break;	1137
		1138	hv_iterinit (hv);
		1139	he = hv_iternext (hv);
		1140	hv_iterinit (hv);
		1141
		1142	// the next line creates a mortal sv each time its called.
		1143	// might want to optimise this for common cases.
		1144	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1145
		1146	if (cb)
		1147	{
		1148	dSP;
		1149	int count;
		1150
		1151	ENTER; SAVETMPS; PUSHMARK (SP);
		1152	XPUSHs (HeVAL (he));
		1153
		1154	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1155
		1156	if (count == 1)
		1157	{
		1158	sv = newSVsv (POPs);
		1159	FREETMPS; LEAVE;
		1160	return sv;
		1161	}
		1162
		1163	FREETMPS; LEAVE;
		1164	}
733	}	1165	}
734		1166
735	if (*dec->cur != ',')	1167	if (dec->json.cb_object)
736	ERR (", or } expected while parsing object/hash");	1168	{
		1169	dSP;
		1170	int count;
737		1171
738	++dec->cur;	1172	ENTER; SAVETMPS; PUSHMARK (SP);
739	}	1173	XPUSHs (sv_2mortal (sv));
740		1174
741	return newRV_noinc ((SV *)hv);	1175	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1176
		1177	if (count == 1)
		1178	{
		1179	sv = newSVsv (POPs);
		1180	FREETMPS; LEAVE;
		1181	return sv;
		1182	}
		1183
		1184	SvREFCNT_inc (sv);
		1185	FREETMPS; LEAVE;
		1186	}
		1187	}
		1188
		1189	return sv;
742		1190
743	fail:	1191	fail:
744	SvREFCNT_dec (hv);	1192	SvREFCNT_dec (hv);
		1193	DEC_DEC_DEPTH;
745	return 0;	1194	return 0;
746	}	1195	}
747		1196
748	static SV *	1197	static SV *
749	decode_sv (dec_t *dec)	1198	decode_sv (dec_t *dec)
750	{	1199	{
751	WS;	1200	decode_ws (dec);
		1201
		1202	// the beauty of JSON: you need exactly one character lookahead
		1203	// to parse anything.
752	switch (*dec->cur)	1204	switch (*dec->cur)
753	{	1205	{
754	case '"': ++dec->cur; return decode_str (dec);	1206	case '"': ++dec->cur; return decode_str (dec);
755	case '[': ++dec->cur; return decode_av (dec);	1207	case '[': ++dec->cur; return decode_av (dec);
756	case '{': ++dec->cur; return decode_hv (dec);	1208	case '{': ++dec->cur; return decode_hv (dec);
…		…
762		1214
763	case 't':	1215	case 't':
764	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1216	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
765	{	1217	{
766	dec->cur += 4;	1218	dec->cur += 4;
767	return newSViv (1);	1219	return SvREFCNT_inc (json_true);
768	}	1220	}
769	else	1221	else
770	ERR ("'true' expected");	1222	ERR ("'true' expected");
771		1223
772	break;	1224	break;
773		1225
774	case 'f':	1226	case 'f':
775	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1227	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
776	{	1228	{
777	dec->cur += 5;	1229	dec->cur += 5;
778	return newSViv (0);	1230	return SvREFCNT_inc (json_false);
779	}	1231	}
780	else	1232	else
781	ERR ("'false' expected");	1233	ERR ("'false' expected");
782		1234
783	break;	1235	break;
784		1236
785	case 'n':	1237	case 'n':
786	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1238	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
787	{	1239	{
788	dec->cur += 4;	1240	dec->cur += 4;
789	return newSViv (1);	1241	return newSVsv (&PL_sv_undef);
790	}	1242	}
791	else	1243	else
792	ERR ("'null' expected");	1244	ERR ("'null' expected");
793		1245
794	break;	1246	break;
795		1247
796	default:	1248	default:
797	ERR ("malformed json string");	1249	ERR ("malformed JSON string, neither array, object, number, string or atom");
798	break;	1250	break;
799	}	1251	}
800		1252
801	fail:	1253	fail:
802	return 0;	1254	return 0;
803	}	1255	}
804		1256
805	static SV *	1257	static SV *
806	decode_json (SV *string, UV flags)	1258	decode_json (SV *string, JSON *json, UV *offset_return)
807	{	1259	{
		1260	dec_t dec;
		1261	UV offset;
808	SV *sv;	1262	SV *sv;
809		1263
		1264	SvGETMAGIC (string);
		1265	SvUPGRADE (string, SVt_PV);
		1266
		1267	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1268	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1269	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1270
810	if (!(flags & F_UTF8))	1271	if (json->flags & F_UTF8)
		1272	sv_utf8_downgrade (string, 0);
		1273	else
811	sv_utf8_upgrade (string);	1274	sv_utf8_upgrade (string);
812		1275
813	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1276	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
814		1277
815	dec_t dec;	1278	dec.json = *json;
816	dec.flags = flags;
817	dec.cur = SvPVX (string);	1279	dec.cur = SvPVX (string);
818	dec.end = SvEND (string);	1280	dec.end = SvEND (string);
819	dec.err = 0;	1281	dec.err = 0;
		1282	dec.depth = 0;
		1283	dec.maxdepth = DEC_DEPTH (dec.json.flags);
820		1284
821	*dec.end = 1; // invalid anywhere	1285	if (dec.json.cb_object || dec.json.cb_sk_object)
		1286	dec.json.flags |= F_HOOK;
		1287
		1288	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
822	sv = decode_sv (&dec);	1289	sv = decode_sv (&dec);
823	*dec.end = 0;	1290
		1291	if (!(offset_return || !sv))
		1292	{
		1293	// check for trailing garbage
		1294	decode_ws (&dec);
		1295
		1296	if (*dec.cur)
		1297	{
		1298	dec.err = "garbage after JSON object";
		1299	SvREFCNT_dec (sv);
		1300	sv = 0;
		1301	}
		1302	}
		1303
		1304	if (offset_return || !sv)
		1305	{
		1306	offset = dec.json.flags & F_UTF8
		1307	? dec.cur - SvPVX (string)
		1308	: utf8_distance (dec.cur, SvPVX (string));
		1309
		1310	if (offset_return)
		1311	*offset_return = offset;
		1312	}
824		1313
825	if (!sv)	1314	if (!sv)
826	{	1315	{
827	IV offset = utf8_distance (dec.cur, SvPVX (string));
828	SV *uni = sv_newmortal ();	1316	SV *uni = sv_newmortal ();
829		1317
		1318	// horrible hack to silence warning inside pv_uni_display
		1319	COP cop = *PL_curcop;
		1320	cop.cop_warnings = pWARN_NONE;
		1321	ENTER;
		1322	SAVEVPTR (PL_curcop);
		1323	PL_curcop = &cop;
830	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1324	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1325	LEAVE;
		1326
831	croak ("%s, at character %d (%s)",	1327	croak ("%s, at character offset %d [\"%s\"]",
832	dec.err,	1328	dec.err,
833	(int)offset,	1329	(int)offset,
834	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1330	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
835	}	1331	}
836		1332
837	sv = sv_2mortal (sv);	1333	sv = sv_2mortal (sv);
838		1334
839	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1335	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
840	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1336	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
841		1337
842	return sv;	1338	return sv;
843	}	1339	}
844		1340
		1341	/////////////////////////////////////////////////////////////////////////////
		1342	// XS interface functions
		1343
845	MODULE = JSON::XS PACKAGE = JSON::XS	1344	MODULE = JSON::XS PACKAGE = JSON::XS
846		1345
847	BOOT:	1346	BOOT:
848	{	1347	{
849	int i;	1348	int i;
850		1349
851	memset (decode_hexdigit, 0xff, 256);
852	for (i = 10; i--; )	1350	for (i = 0; i < 256; ++i)
853	decode_hexdigit ['0' + i] = i;	1351	decode_hexdigit [i] =
		1352	i >= '0' && i <= '9' ? i - '0'
		1353	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1354	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1355	: -1;
854		1356
855	for (i = 7; i--; )	1357	json_stash = gv_stashpv ("JSON::XS" , 1);
		1358	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1359
		1360	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1361	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1362	}
		1363
		1364	PROTOTYPES: DISABLE
		1365
		1366	void new (char *klass)
		1367	PPCODE:
		1368	{
		1369	SV *pv = NEWSV (0, sizeof (JSON));
		1370	SvPOK_only (pv);
		1371	Zero (SvPVX (pv), 1, JSON);
		1372	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1373	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), json_stash)));
		1374	}
		1375
		1376	void ascii (JSON *self, int enable = 1)
		1377	ALIAS:
		1378	ascii = F_ASCII
		1379	latin1 = F_LATIN1
		1380	utf8 = F_UTF8
		1381	indent = F_INDENT
		1382	canonical = F_CANONICAL
		1383	space_before = F_SPACE_BEFORE
		1384	space_after = F_SPACE_AFTER
		1385	pretty = F_PRETTY
		1386	allow_nonref = F_ALLOW_NONREF
		1387	shrink = F_SHRINK
		1388	allow_blessed = F_ALLOW_BLESSED
		1389	convert_blessed = F_CONV_BLESSED
		1390	PPCODE:
		1391	{
		1392	if (enable)
		1393	self->flags |= ix;
		1394	else
		1395	self->flags &= ~ix;
		1396
		1397	XPUSHs (ST (0));
		1398	}
		1399
		1400	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1401	PPCODE:
		1402	{
		1403	UV log2 = 0;
		1404
		1405	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1406
		1407	while ((1UL << log2) < max_depth)
		1408	++log2;
		1409
		1410	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1411
		1412	XPUSHs (ST (0));
		1413	}
		1414
		1415	void max_size (JSON *self, UV max_size = 0)
		1416	PPCODE:
		1417	{
		1418	UV log2 = 0;
		1419
		1420	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1421	if (max_size == 1) max_size = 2;
		1422
		1423	while ((1UL << log2) < max_size)
		1424	++log2;
		1425
		1426	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1427
		1428	XPUSHs (ST (0));
		1429	}
		1430
		1431	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1432	PPCODE:
		1433	{
		1434	SvREFCNT_dec (self->cb_object);
		1435	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1436
		1437	XPUSHs (ST (0));
		1438	}
		1439
		1440	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1441	PPCODE:
		1442	{
		1443	if (!self->cb_sk_object)
		1444	self->cb_sk_object = newHV ();
		1445
		1446	if (SvOK (cb))
		1447	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1448	else
856	{	1449	{
857	decode_hexdigit ['a' + i] = 10 + i;	1450	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
858	decode_hexdigit ['A' + i] = 10 + i;	1451
		1452	if (!HvKEYS (self->cb_sk_object))
		1453	{
		1454	SvREFCNT_dec (self->cb_sk_object);
		1455	self->cb_sk_object = 0;
		1456	}
859	}	1457	}
860		1458
861	json_stash = gv_stashpv ("JSON::XS", 1);	1459	XPUSHs (ST (0));
862	}	1460	}
863		1461
864	PROTOTYPES: DISABLE	1462	void encode (JSON *self, SV *scalar)
		1463	PPCODE:
		1464	XPUSHs (encode_json (scalar, self));
865		1465
866	SV *new (char *dummy)	1466	void decode (JSON *self, SV *jsonstr)
		1467	PPCODE:
		1468	XPUSHs (decode_json (jsonstr, self, 0));
		1469
		1470	void decode_prefix (JSON *self, SV *jsonstr)
		1471	PPCODE:
		1472	{
		1473	UV offset;
		1474	EXTEND (SP, 2);
		1475	PUSHs (decode_json (jsonstr, self, &offset));
		1476	PUSHs (sv_2mortal (newSVuv (offset)));
		1477	}
		1478
		1479	void DESTROY (JSON *self)
867	CODE:	1480	CODE:
868	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1481	SvREFCNT_dec (self->cb_sk_object);
869	OUTPUT:	1482	SvREFCNT_dec (self->cb_object);
870	RETVAL
871
872	SV *ascii (SV *self, int enable)
873	ALIAS:
874	ascii = F_ASCII
875	utf8 = F_UTF8
876	indent = F_INDENT
877	canonical = F_CANONICAL
878	space_before = F_SPACE_BEFORE
879	space_after = F_SPACE_AFTER
880	json_rpc = F_JSON_RPC
881	pretty = F_PRETTY
882	allow_nonref = F_ALLOW_NONREF
883	CODE:
884	{
885	UV *uv = SvJSON (self);
886	if (enable)
887	*uv |= ix;
888	else
889	*uv &= ~ix;
890
891	RETVAL = newSVsv (self);
892	}
893	OUTPUT:
894	RETVAL
895
896	void encode (SV *self, SV *scalar)
897	PPCODE:
898	XPUSHs (encode_json (scalar, *SvJSON (self)));
899
900	void decode (SV *self, SV *jsonstr)
901	PPCODE:
902	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
903		1483
904	PROTOTYPES: ENABLE	1484	PROTOTYPES: ENABLE
905		1485
906	void to_json (SV *scalar)	1486	void to_json (SV *scalar)
907	PPCODE:	1487	PPCODE:
		1488	{
		1489	JSON json = { F_DEFAULT | F_UTF8 };
908	XPUSHs (encode_json (scalar, F_UTF8));	1490	XPUSHs (encode_json (scalar, &json));
		1491	}
909		1492
910	void from_json (SV *jsonstr)	1493	void from_json (SV *jsonstr)
911	PPCODE:	1494	PPCODE:
		1495	{
		1496	JSON json = { F_DEFAULT | F_UTF8 };
912	XPUSHs (decode_json (jsonstr, F_UTF8));	1497	XPUSHs (decode_json (jsonstr, &json, 0));
		1498	}
913		1499

Diff Legend

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