ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.1 by root, Thu Mar 22 16:40:16 2007 UTC vs.
Revision 1.65 by root, Sat Oct 13 01:55:31 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	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_RELAXED 0x00001000UL
16		33
17	#define F_DEFAULT 0	34	#define F_MAXDEPTH 0xf8000000UL
		35	#define S_MAXDEPTH 27
		36	#define F_MAXSIZE 0x01f00000UL
		37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		39
		40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
		42
		43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
		44	#define F_DEFAULT (9UL << S_MAXDEPTH)
18		45
19	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
		47	#define INDENT_STEP 3 // spaces per indentation level
		48
		49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
20		50
21	#define SB do {	51	#define SB do {
22	#define SE } while (0)	52	#define SE } while (0)
23		53
24	static HV *json_stash;	54	#if __GNUC__ >= 3
		55	# define expect(expr,value) __builtin_expect ((expr),(value))
		56	# define inline inline
		57	#else
		58	# define expect(expr,value) (expr)
		59	# define inline static
		60	#endif
		61
		62	#define expect_false(expr) expect ((expr) != 0, 0)
		63	#define expect_true(expr) expect ((expr) != 0, 1)
		64
		65	#ifdef USE_ITHREADS
		66	# define JSON_SLOW 1
		67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		68	#else
		69	# define JSON_SLOW 0
		70	# define JSON_STASH json_stash
		71	#endif
		72
		73	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		74	static SV *json_true, *json_false;
		75
		76	typedef struct {
		77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
		80	} JSON;
		81
		82	/////////////////////////////////////////////////////////////////////////////
		83	// utility functions
		84
		85	inline void
		86	shrink (SV *sv)
		87	{
		88	sv_utf8_downgrade (sv, 1);
		89	if (SvLEN (sv) > SvCUR (sv) + 1)
		90	{
		91	#ifdef SvPV_shrink_to_cur
		92	SvPV_shrink_to_cur (sv);
		93	#elif defined (SvPV_renew)
		94	SvPV_renew (sv, SvCUR (sv) + 1);
		95	#endif
		96	}
		97	}
		98
		99	// decode an utf-8 character and return it, or (UV)-1 in
		100	// case of an error.
		101	// we special-case "safe" characters from U+80 .. U+7FF,
		102	// but use the very good perl function to parse anything else.
		103	// note that we never call this function for a ascii codepoints
		104	inline UV
		105	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		106	{
		107	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		110	{
		111	*clen = 2;
		112	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		113	}
		114	else
		115	{
		116	*clen = (STRLEN)-1;
		117	return (UV)-1;
		118	}
		119	}
		120
		121	/////////////////////////////////////////////////////////////////////////////
		122	// encoder
25		123
26	// structure used for encoding JSON	124	// structure used for encoding JSON
27	typedef struct	125	typedef struct
28	{	126	{
29	char *cur;	127	char *cur; // SvPVX (sv) + current output position
30	STRLEN len; // SvLEN (sv)
31	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
32	SV *sv;	129	SV *sv; // result scalar
33	UV flags;	130	JSON json;
34	int max_recurse;	131	U32 indent; // indentation level
35	int indent;	132	U32 maxdepth; // max. indentation/recursion level
36	} enc_t;	133	} enc_t;
37		134
38	// structure used for decoding JSON	135	inline void
39	typedef struct
40	{
41	char *cur;
42	char *end;
43	char *err;
44	UV flags;
45	} dec_t;
46
47	static UV *
48	SvJSON (SV *sv)
49	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
51	croak ("object is not of type JSON::XS");
52
53	return &SvUVX (SvRV (sv));
54	}
55
56	/////////////////////////////////////////////////////////////////////////////
57
58	static void
59	need (enc_t *enc, STRLEN len)	136	need (enc_t *enc, STRLEN len)
60	{	137	{
61	if (enc->cur + len >= enc->end)	138	if (expect_false (enc->cur + len >= enc->end))
62	{	139	{
63	STRLEN cur = enc->cur - SvPVX (enc->sv);	140	STRLEN cur = enc->cur - SvPVX (enc->sv);
64	SvGROW (enc->sv, cur + len + 1);	141	SvGROW (enc->sv, cur + len + 1);
65	enc->cur = SvPVX (enc->sv) + cur;	142	enc->cur = SvPVX (enc->sv) + cur;
66	enc->end = SvEND (enc->sv);	143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
67	}	144	}
68	}	145	}
69		146
70	static void	147	inline void
71	encode_ch (enc_t *enc, char ch)	148	encode_ch (enc_t *enc, char ch)
72	{	149	{
73	need (enc, 1);	150	need (enc, 1);
74	*enc->cur++ = ch;	151	*enc->cur++ = ch;
75	}	152	}
…		…
77	static void	154	static void
78	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)	155	encode_str (enc_t *enc, char *str, STRLEN len, int is_utf8)
79	{	156	{
80	char *end = str + len;	157	char *end = str + len;
81		158
		159	need (enc, len);
		160
82	while (str < end)	161	while (str < end)
83	{	162	{
84	unsigned char ch = *(unsigned char *)str;	163	unsigned char ch = *(unsigned char *)str;
		164
85	if (ch >= 0x20 && ch < 0x80) // most common case	165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
		166	{
		167	if (expect_false (ch == '"')) // but with slow exceptions
86	{	168	{
		169	need (enc, len += 1);
		170	*enc->cur++ = '\\';
		171	*enc->cur++ = '"';
		172	}
		173	else if (expect_false (ch == '\\'))
		174	{
		175	need (enc, len += 1);
		176	*enc->cur++ = '\\';
		177	*enc->cur++ = '\\';
		178	}
		179	else
87	*enc->cur++ = ch;	180	*enc->cur++ = ch;
		181
88	str++;	182	++str;
89	}	183	}
90	else	184	else
91	{	185	{
92	STRLEN clen;	186	switch (ch)
93	UV uch;
94
95	if (is_utf8)
96	{
97	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
98	if (clen < 0)
99	croak ("malformed UTF-8 character in string, cannot convert to JSON");
100	}	187	{
101	else	188	case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
102	{	189	case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
103	uch = ch;	190	case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
104	clen = 1;	191	case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
105	}	192	case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
106		193
107	need (enc, len += 6);	194	default:
108
109	if (uch < 0xa0 || enc->flags & F_ASCII)
110	{
111	if (uch > 0xFFFFUL)
112	{	195	{
		196	STRLEN clen;
		197	UV uch;
		198
		199	if (is_utf8)
		200	{
		201	uch = decode_utf8 (str, end - str, &clen);
		202	if (clen == (STRLEN)-1)
		203	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
		204	}
		205	else
		206	{
		207	uch = ch;
113	len += 6;	208	clen = 1;
		209	}
		210
		211	if (uch > 0x10FFFFUL)
		212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		213
		214	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
		215	{
		216	if (uch > 0xFFFFUL)
		217	{
114	need (enc, len += 6);	218	need (enc, len += 11);
115	sprintf (enc->cur, "\\u%04x\\u%04x",	219	sprintf (enc->cur, "\\u%04x\\u%04x",
116	(uch - 0x10000) / 0x400 + 0xD800,	220	(int)((uch - 0x10000) / 0x400 + 0xD800),
117	(uch - 0x10000) % 0x400 + 0xDC00);	221	(int)((uch - 0x10000) % 0x400 + 0xDC00));
118	enc->cur += 12;	222	enc->cur += 12;
		223	}
		224	else
		225	{
		226	static char hexdigit [16] = "0123456789abcdef";
		227	need (enc, len += 5);
		228	*enc->cur++ = '\\';
		229	*enc->cur++ = 'u';
		230	*enc->cur++ = hexdigit [ uch >> 12 ];
		231	*enc->cur++ = hexdigit [(uch >> 8) & 15];
		232	*enc->cur++ = hexdigit [(uch >> 4) & 15];
		233	*enc->cur++ = hexdigit [(uch >> 0) & 15];
		234	}
		235
		236	str += clen;
		237	}
		238	else if (enc->json.flags & F_LATIN1)
		239	{
		240	*enc->cur++ = uch;
		241	str += clen;
		242	}
		243	else if (is_utf8)
		244	{
		245	need (enc, len += clen);
		246	do
		247	{
		248	*enc->cur++ = *str++;
		249	}
		250	while (--clen);
		251	}
		252	else
		253	{
		254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
		255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
		256	++str;
		257	}
119	}	258	}
120	else
121	{
122	sprintf (enc->cur, "\\u%04x", uch);
123	enc->cur += 6;
124	}
125	}
126	else if (is_utf8)
127	{	259	}
128	memcpy (enc->cur, str, clen);
129	enc->cur += clen;
130	}
131	else
132	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
133
134	str += clen;
135	}	260	}
136		261
137	--len;	262	--len;
138	}	263	}
139	}	264	}
140		265
141	#define INDENT SB \	266	inline void
		267	encode_indent (enc_t *enc)
		268	{
142	if (enc->flags & F_INDENT) \	269	if (enc->json.flags & F_INDENT)
143	{ \	270	{
144	int i_; \	271	int spaces = enc->indent * INDENT_STEP;
145	need (enc, enc->indent); \	272
146	for (i_ = enc->indent * 3; i_--; )\	273	need (enc, spaces);
		274	memset (enc->cur, ' ', spaces);
		275	enc->cur += spaces;
		276	}
		277	}
		278
		279	inline void
		280	encode_space (enc_t *enc)
		281	{
		282	need (enc, 1);
		283	encode_ch (enc, ' ');
		284	}
		285
		286	inline void
		287	encode_nl (enc_t *enc)
		288	{
		289	if (enc->json.flags & F_INDENT)
		290	{
		291	need (enc, 1);
147	encode_ch (enc, ' '); \	292	encode_ch (enc, '\n');
148	} \	293	}
149	SE	294	}
150		295
151	#define SPACE SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, ' '); } SE	296	inline void
152	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	297	encode_comma (enc_t *enc)
153	#define COMMA SB \	298	{
154	encode_ch (enc, ','); \	299	encode_ch (enc, ',');
		300
155	if (enc->flags & F_INDENT) \	301	if (enc->json.flags & F_INDENT)
156	NL; \	302	encode_nl (enc);
157	else if (enc->flags & F_SPACE_AFTER) \	303	else if (enc->json.flags & F_SPACE_AFTER)
158	SPACE; \	304	encode_space (enc);
159	SE	305	}
160		306
161	static void encode_sv (enc_t *enc, SV *sv);	307	static void encode_sv (enc_t *enc, SV *sv);
162		308
163	static void	309	static void
164	encode_av (enc_t *enc, AV *av)	310	encode_av (enc_t *enc, AV *av)
165	{	311	{
166	int i, len = av_len (av);	312	int i, len = av_len (av);
167		313
		314	if (enc->indent >= enc->maxdepth)
		315	croak ("data structure too deep (hit recursion limit)");
		316
168	encode_ch (enc, '['); NL;	317	encode_ch (enc, '[');
169	++enc->indent;	318
		319	if (len >= 0)
		320	{
		321	encode_nl (enc); ++enc->indent;
170		322
171	for (i = 0; i <= len; ++i)	323	for (i = 0; i <= len; ++i)
172	{	324	{
173	INDENT;	325	SV **svp = av_fetch (av, i, 0);
174	encode_sv (enc, *av_fetch (av, i, 0));
175		326
		327	encode_indent (enc);
		328
		329	if (svp)
		330	encode_sv (enc, *svp);
		331	else
		332	encode_str (enc, "null", 4, 0);
		333
176	if (i < len)	334	if (i < len)
177	COMMA;	335	encode_comma (enc);
178	}	336	}
179		337
180	NL;	338	encode_nl (enc); --enc->indent; encode_indent (enc);
181		339	}
182	--enc->indent;	340
183	INDENT; encode_ch (enc, ']');	341	encode_ch (enc, ']');
184	}	342	}
185		343
186	static void	344	static void
187	encode_he (enc_t *enc, HE *he)	345	encode_hk (enc_t *enc, HE *he)
188	{	346	{
189	encode_ch (enc, '"');	347	encode_ch (enc, '"');
190		348
191	if (HeKLEN (he) == HEf_SVKEY)	349	if (HeKLEN (he) == HEf_SVKEY)
192	{	350	{
193	SV *sv = HeSVKEY (he);	351	SV *sv = HeSVKEY (he);
194	STRLEN len;	352	STRLEN len;
		353	char *str;
		354
		355	SvGETMAGIC (sv);
195	char *str = SvPV (sv, len);	356	str = SvPV (sv, len);
196		357
197	encode_str (enc, str, len, SvUTF8 (sv));	358	encode_str (enc, str, len, SvUTF8 (sv));
198	}	359	}
199	else	360	else
200	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	361	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
201		362
202	encode_ch (enc, '"');	363	encode_ch (enc, '"');
203		364
204	if (enc->flags & F_SPACE_BEFORE) SPACE;	365	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
205	encode_ch (enc, ':');	366	encode_ch (enc, ':');
206	if (enc->flags & F_SPACE_AFTER ) SPACE;	367	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
207	encode_sv (enc, HeVAL (he));
208	}	368	}
209		369
210	// compare hash entries, used when all keys are bytestrings	370	// compare hash entries, used when all keys are bytestrings
211	static int	371	static int
212	he_cmp_fast (const void *a_, const void *b_)	372	he_cmp_fast (const void *a_, const void *b_)
…		…
217	HE *b = *(HE **)b_;	377	HE *b = *(HE **)b_;
218		378
219	STRLEN la = HeKLEN (a);	379	STRLEN la = HeKLEN (a);
220	STRLEN lb = HeKLEN (b);	380	STRLEN lb = HeKLEN (b);
221		381
222	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	382	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
223	cmp = la < lb ? -1 : la == lb ? 0 : 1;	383	cmp = lb - la;
224		384
225	return cmp;	385	return cmp;
226	}	386	}
227		387
228	// compare hash entries, used when some keys are sv's or utf-x	388	// compare hash entries, used when some keys are sv's or utf-x
229	static int	389	static int
230	he_cmp_slow (const void *a, const void *b)	390	he_cmp_slow (const void *a, const void *b)
231	{	391	{
232	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	392	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
233	}	393	}
234		394
235	static void	395	static void
236	encode_hv (enc_t *enc, HV *hv)	396	encode_hv (enc_t *enc, HV *hv)
237	{	397	{
		398	HE *he;
238	int count, i;	399	int count;
239		400
240	encode_ch (enc, '{'); NL; ++enc->indent;	401	if (enc->indent >= enc->maxdepth)
		402	croak ("data structure too deep (hit recursion limit)");
241		403
242	if ((count = hv_iterinit (hv)))	404	encode_ch (enc, '{');
243	{	405
244	// for canonical output we have to sort by keys first	406	// for canonical output we have to sort by keys first
245	// actually, this is mostly due to the stupid so-called	407	// actually, this is mostly due to the stupid so-called
246	// security workaround added somewhere in 5.8.x.	408	// security workaround added somewhere in 5.8.x.
247	// that randomises hash orderings	409	// that randomises hash orderings
248	if (enc->flags & F_CANONICAL)	410	if (enc->json.flags & F_CANONICAL)
		411	{
		412	int count = hv_iterinit (hv);
		413
		414	if (SvMAGICAL (hv))
		415	{
		416	// need to count by iterating. could improve by dynamically building the vector below
		417	// but I don't care for the speed of this special case.
		418	// note also that we will run into undefined behaviour when the two iterations
		419	// do not result in the same count, something I might care for in some later release.
		420
		421	count = 0;
		422	while (hv_iternext (hv))
		423	++count;
		424
		425	hv_iterinit (hv);
249	{	426	}
250	HE *he, *hes [count];	427
		428	if (count)
		429	{
251	int fast = 1;	430	int i, fast = 1;
		431	#if defined(__BORLANDC__) || defined(_MSC_VER)
		432	HE **hes = _alloca (count * sizeof (HE));
		433	#else
		434	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		435	#endif
252		436
253	i = 0;	437	i = 0;
254	while ((he = hv_iternext (hv)))	438	while ((he = hv_iternext (hv)))
255	{	439	{
256	hes [i++] = he;	440	hes [i++] = he;
…		…
262		446
263	if (fast)	447	if (fast)
264	qsort (hes, count, sizeof (HE *), he_cmp_fast);	448	qsort (hes, count, sizeof (HE *), he_cmp_fast);
265	else	449	else
266	{	450	{
267	// hack to disable "use bytes"	451	// hack to forcefully disable "use bytes"
268	COP *oldcop = PL_curcop, cop;	452	COP cop = *PL_curcop;
269	cop.op_private = 0;	453	cop.op_private = 0;
		454
		455	ENTER;
		456	SAVETMPS;
		457
		458	SAVEVPTR (PL_curcop);
270	PL_curcop = &cop;	459	PL_curcop = &cop;
271		460
272	SAVETMPS;
273	qsort (hes, count, sizeof (HE *), he_cmp_slow);	461	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		462
274	FREETMPS;	463	FREETMPS;
		464	LEAVE;
		465	}
275		466
276	PL_curcop = oldcop;	467	encode_nl (enc); ++enc->indent;
		468
		469	while (count--)
277	}	470	{
		471	encode_indent (enc);
		472	he = hes [count];
		473	encode_hk (enc, he);
		474	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
278		475
279	for (i = 0; i < count; ++i)	476	if (count)
		477	encode_comma (enc);
		478	}
		479
		480	encode_nl (enc); --enc->indent; encode_indent (enc);
		481	}
		482	}
		483	else
		484	{
		485	if (hv_iterinit (hv) || SvMAGICAL (hv))
		486	if ((he = hv_iternext (hv)))
		487	{
		488	encode_nl (enc); ++enc->indent;
		489
		490	for (;;)
280	{	491	{
281	INDENT;	492	encode_indent (enc);
282	encode_he (enc, hes [i]);	493	encode_hk (enc, he);
		494	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
283		495
284	if (i < count - 1)	496	if (!(he = hv_iternext (hv)))
285	COMMA;	497	break;
		498
		499	encode_comma (enc);
286	}	500	}
287		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
288	NL;	503	}
		504	}
		505
		506	encode_ch (enc, '}');
		507	}
		508
		509	// encode objects, arrays and special \0=false and \1=true values.
		510	static void
		511	encode_rv (enc_t *enc, SV *sv)
		512	{
		513	svtype svt;
		514
		515	SvGETMAGIC (sv);
		516	svt = SvTYPE (sv);
		517
		518	if (expect_false (SvOBJECT (sv)))
		519	{
		520	HV *stash = !JSON_SLOW || json_boolean_stash
		521	? json_boolean_stash
		522	: gv_stashpv ("JSON::XS::Boolean", 1);
		523
		524	if (SvSTASH (sv) == stash)
		525	{
		526	if (SvIV (sv))
		527	encode_str (enc, "true", 4, 0);
		528	else
		529	encode_str (enc, "false", 5, 0);
289	}	530	}
290	else	531	else
291	{	532	{
292	SV *sv;	533	#if 0
293	HE *he = hv_iternext (hv);	534	if (0 && sv_derived_from (rv, "JSON::Literal"))
294
295	for (;;)
296	{
297	INDENT;
298	encode_he (enc, he);
299
300	if (!(he = hv_iternext (hv)))
301	break;
302
303	COMMA;
304	}	535	{
305		536	// not yet
306	NL;
307	}	537	}
		538	#endif
		539	if (enc->json.flags & F_CONV_BLESSED)
		540	{
		541	// we re-bless the reference to get overload and other niceties right
		542	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		543
		544	if (to_json)
		545	{
		546	dSP;
		547
		548	ENTER; SAVETMPS; PUSHMARK (SP);
		549	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		550
		551	// calling with G_SCALAR ensures that we always get a 1 return value
		552	PUTBACK;
		553	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		554	SPAGAIN;
		555
		556	// catch this surprisingly common error
		557	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		558	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		559
		560	sv = POPs;
		561	PUTBACK;
		562
		563	encode_sv (enc, sv);
		564
		565	FREETMPS; LEAVE;
		566	}
		567	else if (enc->json.flags & F_ALLOW_BLESSED)
		568	encode_str (enc, "null", 4, 0);
		569	else
		570	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		571	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		572	}
		573	else if (enc->json.flags & F_ALLOW_BLESSED)
		574	encode_str (enc, "null", 4, 0);
		575	else
		576	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		577	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		578	}
		579	}
		580	else if (svt == SVt_PVHV)
		581	encode_hv (enc, (HV *)sv);
		582	else if (svt == SVt_PVAV)
		583	encode_av (enc, (AV *)sv);
		584	else if (svt < SVt_PVAV)
308	}	585	{
		586	STRLEN len = 0;
		587	char *pv = svt ? SvPV (sv, len) : 0;
309		588
310	--enc->indent; INDENT; encode_ch (enc, '}');	589	if (len == 1 && *pv == '1')
		590	encode_str (enc, "true", 4, 0);
		591	else if (len == 1 && *pv == '0')
		592	encode_str (enc, "false", 5, 0);
		593	else
		594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		596	}
		597	else
		598	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
311	}	600	}
312		601
313	static void	602	static void
314	encode_sv (enc_t *enc, SV *sv)	603	encode_sv (enc_t *enc, SV *sv)
315	{	604	{
		605	SvGETMAGIC (sv);
		606
316	if (SvPOKp (sv))	607	if (SvPOKp (sv))
317	{	608	{
318	STRLEN len;	609	STRLEN len;
319	char *str = SvPV (sv, len);	610	char *str = SvPV (sv, len);
320	encode_ch (enc, '"');	611	encode_ch (enc, '"');
321	encode_str (enc, str, len, SvUTF8 (sv));	612	encode_str (enc, str, len, SvUTF8 (sv));
322	encode_ch (enc, '"');	613	encode_ch (enc, '"');
323	}	614	}
324	else if (SvNOKp (sv))	615	else if (SvNOKp (sv))
325	{	616	{
		617	// trust that perl will do the right thing w.r.t. JSON syntax.
326	need (enc, NV_DIG + 32);	618	need (enc, NV_DIG + 32);
327	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
328	enc->cur += strlen (enc->cur);	620	enc->cur += strlen (enc->cur);
329	}	621	}
330	else if (SvIOKp (sv))	622	else if (SvIOKp (sv))
331	{	623	{
332	need (enc, 64);	624	// we assume we can always read an IV as a UV
		625	if (SvUV (sv) & ~(UV)0x7fff)
		626	{
		627	// large integer, use the (rather slow) snprintf way.
		628	need (enc, sizeof (UV) * 3);
333	enc->cur +=	629	enc->cur +=
334	SvIsUV(sv)	630	SvIsUV(sv)
335	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
336	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		633	}
		634	else
		635	{
		636	// optimise the "small number case"
		637	// code will likely be branchless and use only a single multiplication
		638	I32 i = SvIV (sv);
		639	U32 u;
		640	char digit, nz = 0;
		641
		642	need (enc, 6);
		643
		644	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		645	u = i < 0 ? -i : i;
		646
		647	// convert to 4.28 fixed-point representation
		648	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		649
		650	// now output digit by digit, each time masking out the integer part
		651	// and multiplying by 5 while moving the decimal point one to the right,
		652	// resulting in a net multiplication by 10.
		653	// we always write the digit to memory but conditionally increment
		654	// the pointer, to ease the usage of conditional move instructions.
		655	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		656	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		657	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		658	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		660	}
337	}	661	}
338	else if (SvROK (sv))	662	else if (SvROK (sv))
339	{	663	encode_rv (enc, SvRV (sv));
340	if (!--enc->max_recurse)
341	croak ("data structure too deep (hit recursion limit)");
342
343	sv = SvRV (sv);
344
345	switch (SvTYPE (sv))
346	{
347	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
348	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
349
350	default:
351	croak ("JSON can only represent references to arrays or hashes");
352	}
353	}
354	else if (!SvOK (sv))	664	else if (!SvOK (sv))
355	encode_str (enc, "null", 4, 0);	665	encode_str (enc, "null", 4, 0);
356	else	666	else
357	croak ("encountered perl type that JSON cannot handle");	667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		668	SvPV_nolen (sv), SvFLAGS (sv));
358	}	669	}
359		670
360	static SV *	671	static SV *
361	encode_json (SV *scalar, UV flags)	672	encode_json (SV *scalar, JSON *json)
362	{	673	{
363	enc_t enc;	674	enc_t enc;
364	enc.flags = flags;	675
		676	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		677	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		678
		679	enc.json = *json;
365	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
366	enc.cur = SvPVX (enc.sv);	681	enc.cur = SvPVX (enc.sv);
367	enc.end = SvEND (enc.sv);	682	enc.end = SvEND (enc.sv);
368	enc.max_recurse = 0;
369	enc.indent = 0;	683	enc.indent = 0;
		684	enc.maxdepth = DEC_DEPTH (enc.json.flags);
370		685
371	SvPOK_only (enc.sv);	686	SvPOK_only (enc.sv);
372	encode_sv (&enc, scalar);	687	encode_sv (&enc, scalar);
373		688
		689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		691
374	if (!(flags & (F_ASCII | F_UTF8)))	692	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
375	SvUTF8_on (enc.sv);	693	SvUTF8_on (enc.sv);
376		694
377	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	695	if (enc.json.flags & F_SHRINK)
		696	shrink (enc.sv);
		697
378	return enc.sv;	698	return enc.sv;
379	}	699	}
380		700
381	/////////////////////////////////////////////////////////////////////////////	701	/////////////////////////////////////////////////////////////////////////////
		702	// decoder
382		703
383	#define WS \	704	// structure used for decoding JSON
		705	typedef struct
		706	{
		707	char *cur; // current parser pointer
		708	char *end; // end of input string
		709	const char *err; // parse error, if != 0
		710	JSON json;
		711	U32 depth; // recursion depth
		712	U32 maxdepth; // recursion depth limit
		713	} dec_t;
		714
		715	inline void
		716	decode_comment (dec_t *dec)
		717	{
		718	// only '#'-style comments allowed a.t.m.
		719
		720	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		721	++dec->cur;
		722	}
		723
		724	inline void
		725	decode_ws (dec_t *dec)
		726	{
384	for (;;) \	727	for (;;)
385	{ \	728	{
386	char ch = *dec->cur; \	729	char ch = *dec->cur;
		730
387	if (ch > 0x20 \	731	if (ch > 0x20)
		732	{
		733	if (expect_false (ch == '#'))
		734	{
		735	if (dec->json.flags & F_RELAXED)
		736	decode_comment (dec);
		737	else
		738	break;
		739	}
		740	else
		741	break;
		742	}
388	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	743	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
389	break; \	744	break; // parse error, but let higher level handle it, gives better error messages
		745
390	++dec->cur; \	746	++dec->cur;
391	}	747	}
		748	}
392		749
393	#define ERR(reason) SB dec->err = reason; goto fail; SE	750	#define ERR(reason) SB dec->err = reason; goto fail; SE
		751
394	#define EXPECT_CH(ch) SB \	752	#define EXPECT_CH(ch) SB \
395	if (*dec->cur != ch) \	753	if (*dec->cur != ch) \
396	ERR (# ch " expected"); \	754	ERR (# ch " expected"); \
397	++dec->cur; \	755	++dec->cur; \
398	SE	756	SE
399		757
		758	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		759	#define DEC_DEC_DEPTH --dec->depth
		760
400	static SV *decode_sv (dec_t *dec);	761	static SV *decode_sv (dec_t *dec);
401
402	#define APPEND_CH(ch) SB \
403	SvGROW (sv, cur + 1 + 1); \
404	SvPVX (sv)[cur++] = (ch); \
405	SE
406		762
407	static signed char decode_hexdigit[256];	763	static signed char decode_hexdigit[256];
408		764
409	static UV	765	static UV
410	decode_4hex (dec_t *dec)	766	decode_4hex (dec_t *dec)
411	{	767	{
412	signed char d1, d2, d3, d4;	768	signed char d1, d2, d3, d4;
		769	unsigned char *cur = (unsigned char *)dec->cur;
413		770
414	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	771	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
415	if (d1 < 0) ERR ("four hexadecimal digits expected");	772	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
416	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	773	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
417	if (d2 < 0) ERR ("four hexadecimal digits expected");	774	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
418	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
419	if (d3 < 0) ERR ("four hexadecimal digits expected");
420	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
421	if (d4 < 0) ERR ("four hexadecimal digits expected");
422		775
423	dec->cur += 4;	776	dec->cur += 4;
424		777
425	return ((UV)d1) << 12	778	return ((UV)d1) << 12
426	| ((UV)d2) << 8	779	| ((UV)d2) << 8
…		…
432	}	785	}
433		786
434	static SV *	787	static SV *
435	decode_str (dec_t *dec)	788	decode_str (dec_t *dec)
436	{	789	{
437	SV *sv = NEWSV (0,2);	790	SV *sv = 0;
438	STRLEN cur = 0;
439	int utf8 = 0;	791	int utf8 = 0;
		792	char *dec_cur = dec->cur;
440		793
441	for (;;)	794	do
442	{	795	{
		796	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		797	char *cur = buf;
		798
		799	do
		800	{
443	unsigned char ch = *(unsigned char *)dec->cur;	801	unsigned char ch = *(unsigned char *)dec_cur++;
444		802
445	if (ch == '"')	803	if (expect_false (ch == '"'))
		804	{
		805	--dec_cur;
446	break;	806	break;
447	else if (ch == '\\')
448	{
449	switch (*++dec->cur)
450	{	807	}
451	case '\\':	808	else if (expect_false (ch == '\\'))
452	case '/':	809	{
453	case '"': APPEND_CH (*dec->cur++); break;	810	switch (*dec_cur)
454
455	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
456	case 't': APPEND_CH ('\011'); ++dec->cur; break;
457	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
458	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
459	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
460
461	case 'u':
462	{	811	{
463	UV lo, hi;	812	case '\\':
464	++dec->cur;	813	case '/':
		814	case '"': *cur++ = *dec_cur++; break;
465		815
466	hi = decode_4hex (dec);	816	case 'b': ++dec_cur; *cur++ = '\010'; break;
467	if (hi == (UV)-1)	817	case 't': ++dec_cur; *cur++ = '\011'; break;
468	goto fail;	818	case 'n': ++dec_cur; *cur++ = '\012'; break;
		819	case 'f': ++dec_cur; *cur++ = '\014'; break;
		820	case 'r': ++dec_cur; *cur++ = '\015'; break;
469		821
470	// possibly a surrogate pair	822	case 'u':
471	if (hi >= 0xd800 && hi < 0xdc00)
472	{	823	{
473	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	824	UV lo, hi;
474	ERR ("illegal surrogate character");	825	++dec_cur;
475		826
476	dec->cur += 2;	827	dec->cur = dec_cur;
477
478	lo = decode_4hex (dec);	828	hi = decode_4hex (dec);
		829	dec_cur = dec->cur;
479	if (lo == (UV)-1)	830	if (hi == (UV)-1)
480	goto fail;	831	goto fail;
481		832
		833	// possibly a surrogate pair
		834	if (hi >= 0xd800)
		835	if (hi < 0xdc00)
		836	{
		837	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		838	ERR ("missing low surrogate character in surrogate pair");
		839
		840	dec_cur += 2;
		841
		842	dec->cur = dec_cur;
		843	lo = decode_4hex (dec);
		844	dec_cur = dec->cur;
		845	if (lo == (UV)-1)
		846	goto fail;
		847
482	if (lo < 0xdc00 || lo >= 0xe000)	848	if (lo < 0xdc00 || lo >= 0xe000)
483	ERR ("surrogate pair expected");	849	ERR ("surrogate pair expected");
484		850
485	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	851	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		852	}
		853	else if (hi < 0xe000)
		854	ERR ("missing high surrogate character in surrogate pair");
		855
		856	if (hi >= 0x80)
		857	{
		858	utf8 = 1;
		859
		860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		861	}
		862	else
		863	*cur++ = hi;
486	}	864	}
487	else if (lo >= 0xdc00 && lo < 0xe000)
488	ERR ("illegal surrogate character");
489
490	if (hi >= 0x80)
491	{	865	break;
492	utf8 = 1;
493		866
494	SvGROW (sv, cur + 4 + 1); // at most 4 bytes for 21 bits
495	cur = (char *)uvuni_to_utf8_flags (SvPVX (sv) + cur, hi, 0) - SvPVX (sv);
496	}
497	else	867	default:
498	APPEND_CH (hi);	868	--dec_cur;
		869	ERR ("illegal backslash escape sequence in string");
499	}	870	}
500	break;
501	}
502	}	871	}
503	else if (ch >= 0x20 && ch <= 0x7f)	872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
504	APPEND_CH (*dec->cur++);	873	*cur++ = ch;
505	else if (ch >= 0x80)	874	else if (ch >= 0x80)
506	{	875	{
507	STRLEN clen;	876	STRLEN clen;
508	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	877	UV uch;
509	if (clen < 0)	878
		879	--dec_cur;
		880
		881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		882	if (clen == (STRLEN)-1)
510	ERR ("malformed UTF-8 character in string, cannot convert to JSON");	883	ERR ("malformed UTF-8 character in JSON string");
511		884
512	SvGROW (sv, cur + clen + 1); // at most 4 bytes for 21 bits	885	do
513	memcpy (SvPVX (sv) + cur, dec->cur, clen);	886	*cur++ = *dec_cur++;
514	dec->cur += clen;	887	while (--clen);
		888
		889	utf8 = 1;
		890	}
		891	else
		892	{
		893	--dec_cur;
		894
		895	if (!ch)
		896	ERR ("unexpected end of string while parsing JSON string");
		897	else
		898	ERR ("invalid character encountered while parsing JSON string");
		899	}
		900	}
		901	while (cur < buf + SHORT_STRING_LEN);
		902
		903	{
		904	STRLEN len = cur - buf;
		905
		906	if (sv)
		907	{
		908	SvGROW (sv, SvCUR (sv) + len + 1);
		909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		910	SvCUR_set (sv, SvCUR (sv) + len);
515	}	911	}
516	else	912	else
517	ERR ("invalid character encountered");	913	sv = newSVpvn (buf, len);
518	}	914	}
		915	}
		916	while (*dec_cur != '"');
519		917
520	++dec->cur;	918	++dec_cur;
521		919
		920	if (sv)
		921	{
522	SvPOK_only (sv);	922	SvPOK_only (sv);
523
524	SvCUR_set (sv, cur);
525	*SvEND (sv) = 0;	923	*SvEND (sv) = 0;
526		924
527	if (utf8)	925	if (utf8)
528	SvUTF8_on (sv);	926	SvUTF8_on (sv);
		927	}
		928	else
		929	sv = newSVpvn ("", 0);
529		930
		931	dec->cur = dec_cur;
530	return sv;	932	return sv;
531		933
532	fail:	934	fail:
533	SvREFCNT_dec (sv);	935	dec->cur = dec_cur;
534	return 0;	936	return 0;
535	}	937	}
536		938
537	static SV *	939	static SV *
538	decode_num (dec_t *dec)	940	decode_num (dec_t *dec)
…		…
548	{	950	{
549	++dec->cur;	951	++dec->cur;
550	if (*dec->cur >= '0' && *dec->cur <= '9')	952	if (*dec->cur >= '0' && *dec->cur <= '9')
551	ERR ("malformed number (leading zero must not be followed by another digit)");	953	ERR ("malformed number (leading zero must not be followed by another digit)");
552	}	954	}
553		955	else if (*dec->cur < '0' || *dec->cur > '9')
554	// int	956	ERR ("malformed number (no digits after initial minus)");
		957	else
		958	do
		959	{
		960	++dec->cur;
		961	}
555	while (*dec->cur >= '0' && *dec->cur <= '9')	962	while (*dec->cur >= '0' && *dec->cur <= '9');
556	++dec->cur;
557		963
558	// [frac]	964	// [frac]
559	if (*dec->cur == '.')	965	if (*dec->cur == '.')
560	{	966	{
		967	++dec->cur;
		968
		969	if (*dec->cur < '0' || *dec->cur > '9')
		970	ERR ("malformed number (no digits after decimal point)");
		971
		972	do
		973	{
		974	++dec->cur;
		975	}
		976	while (*dec->cur >= '0' && *dec->cur <= '9');
		977
561	is_nv = 1;	978	is_nv = 1;
562
563	do
564	{
565	++dec->cur;
566	}
567	while (*dec->cur >= '0' && *dec->cur <= '9');
568	}	979	}
569		980
570	// [exp]	981	// [exp]
571	if (*dec->cur == 'e' || *dec->cur == 'E')	982	if (*dec->cur == 'e' || *dec->cur == 'E')
572	{	983	{
573	is_nv = 1;
574
575	++dec->cur;	984	++dec->cur;
		985
576	if (*dec->cur == '-' || *dec->cur == '+')	986	if (*dec->cur == '-' || *dec->cur == '+')
577	++dec->cur;	987	++dec->cur;
578		988
		989	if (*dec->cur < '0' || *dec->cur > '9')
		990	ERR ("malformed number (no digits after exp sign)");
		991
		992	do
		993	{
		994	++dec->cur;
		995	}
579	while (*dec->cur >= '0' && *dec->cur <= '9')	996	while (*dec->cur >= '0' && *dec->cur <= '9');
580	++dec->cur;	997
		998	is_nv = 1;
581	}	999	}
582		1000
583	if (!is_nv)	1001	if (!is_nv)
584	{	1002	{
585	UV uv;	1003	int len = dec->cur - start;
586	int numtype = grok_number (start, dec->cur - start, &uv);	1004
587	if (numtype & IS_NUMBER_IN_UV)	1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
588	if (numtype & IS_NUMBER_NEG)	1006	if (*start == '-')
		1007	switch (len)
589	{	1008	{
590	if (uv < (UV)IV_MIN)	1009	case 2: return newSViv (-( start [1] - '0' * 1));
591	return newSViv (-(IV)uv);	1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1011	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1012	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
592	}	1013	}
		1014	else
		1015	switch (len)
		1016	{
		1017	case 1: return newSViv ( start [0] - '0' * 1);
		1018	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1019	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1020	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1021	}
		1022
		1023	{
		1024	UV uv;
		1025	int numtype = grok_number (start, len, &uv);
		1026	if (numtype & IS_NUMBER_IN_UV)
		1027	if (numtype & IS_NUMBER_NEG)
		1028	{
		1029	if (uv < (UV)IV_MIN)
		1030	return newSViv (-(IV)uv);
		1031	}
593	else	1032	else
594	return newSVuv (uv);	1033	return newSVuv (uv);
595	}	1034	}
596		1035
		1036	len -= *start == '-' ? 1 : 0;
		1037
		1038	// does not fit into IV or UV, try NV
		1039	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1040	#if defined (LDBL_DIG)
		1041	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1042	#endif
		1043	)
		1044	// fits into NV without loss of precision
		1045	return newSVnv (Atof (start));
		1046
		1047	// everything else fails, convert it to a string
		1048	return newSVpvn (start, dec->cur - start);
		1049	}
		1050
		1051	// loss of precision here
597	return newSVnv (Atof (start));	1052	return newSVnv (Atof (start));
598		1053
599	fail:	1054	fail:
600	return 0;	1055	return 0;
601	}	1056	}
…		…
603	static SV *	1058	static SV *
604	decode_av (dec_t *dec)	1059	decode_av (dec_t *dec)
605	{	1060	{
606	AV *av = newAV ();	1061	AV *av = newAV ();
607		1062
		1063	DEC_INC_DEPTH;
		1064	decode_ws (dec);
		1065
		1066	if (*dec->cur == ']')
		1067	++dec->cur;
		1068	else
608	for (;;)	1069	for (;;)
609	{	1070	{
610	SV *value;	1071	SV *value;
611		1072
612	value = decode_sv (dec);	1073	value = decode_sv (dec);
613	if (!value)	1074	if (!value)
614	goto fail;	1075	goto fail;
615		1076
616	av_push (av, value);	1077	av_push (av, value);
617		1078
618	WS;	1079	decode_ws (dec);
619		1080
620	if (*dec->cur == ']')	1081	if (*dec->cur == ']')
621	{	1082	{
622	++dec->cur;	1083	++dec->cur;
623	break;	1084	break;
		1085	}
624	}	1086
625
626	if (*dec->cur != ',')	1087	if (*dec->cur != ',')
627	ERR (", or ] expected while parsing array");	1088	ERR (", or ] expected while parsing array");
628		1089
629	++dec->cur;	1090	++dec->cur;
		1091
		1092	decode_ws (dec);
		1093
		1094	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1095	{
		1096	++dec->cur;
		1097	break;
		1098	}
630	}	1099	}
631		1100
		1101	DEC_DEC_DEPTH;
632	return newRV_noinc ((SV *)av);	1102	return newRV_noinc ((SV *)av);
633		1103
634	fail:	1104	fail:
635	SvREFCNT_dec (av);	1105	SvREFCNT_dec (av);
		1106	DEC_DEC_DEPTH;
636	return 0;	1107	return 0;
637	}	1108	}
638		1109
639	static SV *	1110	static SV *
640	decode_hv (dec_t *dec)	1111	decode_hv (dec_t *dec)
641	{	1112	{
		1113	SV *sv;
642	HV *hv = newHV ();	1114	HV *hv = newHV ();
643		1115
		1116	DEC_INC_DEPTH;
		1117	decode_ws (dec);
		1118
		1119	if (*dec->cur == '}')
		1120	++dec->cur;
		1121	else
644	for (;;)	1122	for (;;)
645	{	1123	{
646	SV *key, *value;
647
648	WS; EXPECT_CH ('"');	1124	EXPECT_CH ('"');
649		1125
650	key = decode_str (dec);	1126	// heuristic: assume that
651	if (!key)	1127	// a) decode_str + hv_store_ent are abysmally slow.
652	goto fail;	1128	// b) most hash keys are short, simple ascii text.
653		1129	// => try to "fast-match" such strings to avoid
654	WS; EXPECT_CH (':');	1130	// the overhead of decode_str + hv_store_ent.
655
656	value = decode_sv (dec);
657	if (!value)
658	{
659	SvREFCNT_dec (key);
660	goto fail;
661	}	1131	{
		1132	SV *value;
		1133	char *p = dec->cur;
		1134	char *e = p + 24; // only try up to 24 bytes
662		1135
663	//TODO: optimise	1136	for (;;)
		1137	{
		1138	// the >= 0x80 is true on most architectures
		1139	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1140	{
		1141	// slow path, back up and use decode_str
		1142	SV *key = decode_str (dec);
		1143	if (!key)
		1144	goto fail;
		1145
		1146	decode_ws (dec); EXPECT_CH (':');
		1147
		1148	decode_ws (dec);
		1149	value = decode_sv (dec);
		1150	if (!value)
		1151	{
		1152	SvREFCNT_dec (key);
		1153	goto fail;
		1154	}
		1155
664	hv_store_ent (hv, key, value, 0);	1156	hv_store_ent (hv, key, value, 0);
		1157	SvREFCNT_dec (key);
665		1158
666	WS;	1159	break;
		1160	}
		1161	else if (*p == '"')
		1162	{
		1163	// fast path, got a simple key
		1164	char *key = dec->cur;
		1165	int len = p - key;
		1166	dec->cur = p + 1;
667		1167
		1168	decode_ws (dec); EXPECT_CH (':');
		1169
		1170	decode_ws (dec);
		1171	value = decode_sv (dec);
		1172	if (!value)
		1173	goto fail;
		1174
		1175	hv_store (hv, key, len, value, 0);
		1176
		1177	break;
		1178	}
		1179
		1180	++p;
		1181	}
		1182	}
		1183
		1184	decode_ws (dec);
		1185
668	if (*dec->cur == '}')	1186	if (*dec->cur == '}')
669	{	1187	{
670	++dec->cur;	1188	++dec->cur;
671	break;	1189	break;
672	}	1190	}
673		1191
674	if (*dec->cur != ',')	1192	if (*dec->cur != ',')
675	ERR (", or } expected while parsing object/hash");	1193	ERR (", or } expected while parsing object/hash");
676		1194
677	++dec->cur;	1195	++dec->cur;
		1196
		1197	decode_ws (dec);
		1198
		1199	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1200	{
		1201	++dec->cur;
		1202	break;
		1203	}
678	}	1204	}
679		1205
		1206	DEC_DEC_DEPTH;
680	return newRV_noinc ((SV *)hv);	1207	sv = newRV_noinc ((SV *)hv);
		1208
		1209	// check filter callbacks
		1210	if (dec->json.flags & F_HOOK)
		1211	{
		1212	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1213	{
		1214	HE *cb, *he;
		1215
		1216	hv_iterinit (hv);
		1217	he = hv_iternext (hv);
		1218	hv_iterinit (hv);
		1219
		1220	// the next line creates a mortal sv each time its called.
		1221	// might want to optimise this for common cases.
		1222	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1223
		1224	if (cb)
		1225	{
		1226	dSP;
		1227	int count;
		1228
		1229	ENTER; SAVETMPS; PUSHMARK (SP);
		1230	XPUSHs (HeVAL (he));
		1231
		1232	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1233
		1234	if (count == 1)
		1235	{
		1236	sv = newSVsv (POPs);
		1237	FREETMPS; LEAVE;
		1238	return sv;
		1239	}
		1240
		1241	FREETMPS; LEAVE;
		1242	}
		1243	}
		1244
		1245	if (dec->json.cb_object)
		1246	{
		1247	dSP;
		1248	int count;
		1249
		1250	ENTER; SAVETMPS; PUSHMARK (SP);
		1251	XPUSHs (sv_2mortal (sv));
		1252
		1253	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1254
		1255	if (count == 1)
		1256	{
		1257	sv = newSVsv (POPs);
		1258	FREETMPS; LEAVE;
		1259	return sv;
		1260	}
		1261
		1262	SvREFCNT_inc (sv);
		1263	FREETMPS; LEAVE;
		1264	}
		1265	}
		1266
		1267	return sv;
681		1268
682	fail:	1269	fail:
683	SvREFCNT_dec (hv);	1270	SvREFCNT_dec (hv);
		1271	DEC_DEC_DEPTH;
684	return 0;	1272	return 0;
685	}	1273	}
686		1274
687	static SV *	1275	static SV *
688	decode_sv (dec_t *dec)	1276	decode_sv (dec_t *dec)
689	{	1277	{
690	WS;	1278	// the beauty of JSON: you need exactly one character lookahead
		1279	// to parse anything.
691	switch (*dec->cur)	1280	switch (*dec->cur)
692	{	1281	{
693	case '"': ++dec->cur; return decode_str (dec);	1282	case '"': ++dec->cur; return decode_str (dec);
694	case '[': ++dec->cur; return decode_av (dec);	1283	case '[': ++dec->cur; return decode_av (dec);
695	case '{': ++dec->cur; return decode_hv (dec);	1284	case '{': ++dec->cur; return decode_hv (dec);
…		…
701		1290
702	case 't':	1291	case 't':
703	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
704	{	1293	{
705	dec->cur += 4;	1294	dec->cur += 4;
706	return newSViv (1);	1295	#if JSON_SLOW
		1296	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1297	#endif
		1298	return SvREFCNT_inc (json_true);
707	}	1299	}
708	else	1300	else
709	ERR ("'true' expected");	1301	ERR ("'true' expected");
710		1302
711	break;	1303	break;
712		1304
713	case 'f':	1305	case 'f':
714	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
715	{	1307	{
716	dec->cur += 5;	1308	dec->cur += 5;
717	return newSViv (0);	1309	#if JSON_SLOW
		1310	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1311	#endif
		1312	return SvREFCNT_inc (json_false);
718	}	1313	}
719	else	1314	else
720	ERR ("'false' expected");	1315	ERR ("'false' expected");
721		1316
722	break;	1317	break;
723		1318
724	case 'n':	1319	case 'n':
725	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))	1320	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
726	{	1321	{
727	dec->cur += 4;	1322	dec->cur += 4;
728	return newSViv (1);	1323	return newSVsv (&PL_sv_undef);
729	}	1324	}
730	else	1325	else
731	ERR ("'null' expected");	1326	ERR ("'null' expected");
732		1327
733	break;	1328	break;
734		1329
735	default:	1330	default:
736	ERR ("malformed json string");	1331	ERR ("malformed JSON string, neither array, object, number, string or atom");
737	break;	1332	break;
738	}	1333	}
739		1334
740	fail:	1335	fail:
741	return 0;	1336	return 0;
742	}	1337	}
743		1338
744	static SV *	1339	static SV *
745	decode_json (SV *string, UV flags)	1340	decode_json (SV *string, JSON *json, UV *offset_return)
746	{	1341	{
		1342	dec_t dec;
		1343	UV offset;
747	SV *sv;	1344	SV *sv;
748		1345
		1346	SvGETMAGIC (string);
		1347	SvUPGRADE (string, SVt_PV);
		1348
		1349	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1350	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1351	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1352
749	if (!(flags & F_UTF8))	1353	if (json->flags & F_UTF8)
		1354	sv_utf8_downgrade (string, 0);
		1355	else
750	sv_utf8_upgrade (string);	1356	sv_utf8_upgrade (string);
751		1357
752	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
753		1359
754	dec_t dec;	1360	dec.json = *json;
755	dec.flags = flags;
756	dec.cur = SvPVX (string);	1361	dec.cur = SvPVX (string);
757	dec.end = SvEND (string);	1362	dec.end = SvEND (string);
758	dec.err = 0;	1363	dec.err = 0;
		1364	dec.depth = 0;
		1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
759		1366
760	*dec.end = 1; // invalid anywhere	1367	if (dec.json.cb_object || dec.json.cb_sk_object)
		1368	dec.json.flags |= F_HOOK;
		1369
		1370	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1371
		1372	decode_ws (&dec);
761	sv = decode_sv (&dec);	1373	sv = decode_sv (&dec);
762	*dec.end = 0;	1374
		1375	if (!(offset_return || !sv))
		1376	{
		1377	// check for trailing garbage
		1378	decode_ws (&dec);
		1379
		1380	if (*dec.cur)
		1381	{
		1382	dec.err = "garbage after JSON object";
		1383	SvREFCNT_dec (sv);
		1384	sv = 0;
		1385	}
		1386	}
		1387
		1388	if (offset_return || !sv)
		1389	{
		1390	offset = dec.json.flags & F_UTF8
		1391	? dec.cur - SvPVX (string)
		1392	: utf8_distance (dec.cur, SvPVX (string));
		1393
		1394	if (offset_return)
		1395	*offset_return = offset;
		1396	}
763		1397
764	if (!sv)	1398	if (!sv)
765	{	1399	{
766	IV offset = utf8_distance (dec.cur, SvPVX (string));
767	SV *uni = sv_newmortal ();	1400	SV *uni = sv_newmortal ();
768		1401
		1402	// horrible hack to silence warning inside pv_uni_display
		1403	COP cop = *PL_curcop;
		1404	cop.cop_warnings = pWARN_NONE;
		1405	ENTER;
		1406	SAVEVPTR (PL_curcop);
		1407	PL_curcop = &cop;
769	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1408	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1409	LEAVE;
		1410
770	croak ("%s, at character %d (%s)",	1411	croak ("%s, at character offset %d [\"%s\"]",
771	dec.err,	1412	dec.err,
772	(int)offset,	1413	(int)offset,
773	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
774	}	1415	}
775		1416
776	sv_dump (sv);//D
777	return sv_2mortal (sv);	1417	sv = sv_2mortal (sv);
		1418
		1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
		1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
		1421
		1422	return sv;
778	}	1423	}
		1424
		1425	/////////////////////////////////////////////////////////////////////////////
		1426	// XS interface functions
779		1427
780	MODULE = JSON::XS PACKAGE = JSON::XS	1428	MODULE = JSON::XS PACKAGE = JSON::XS
781		1429
782	BOOT:	1430	BOOT:
783	{	1431	{
784	int i;	1432	int i;
785		1433
786	memset (decode_hexdigit, 0xff, 256);
787	for (i = 10; i--; )	1434	for (i = 0; i < 256; ++i)
788	decode_hexdigit ['0' + i] = i;	1435	decode_hexdigit [i] =
		1436	i >= '0' && i <= '9' ? i - '0'
		1437	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1438	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1439	: -1;
789		1440
790	for (i = 6; --i; )	1441	json_stash = gv_stashpv ("JSON::XS" , 1);
		1442	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1443
		1444	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1445	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1446	}
		1447
		1448	PROTOTYPES: DISABLE
		1449
		1450	void CLONE (...)
		1451	CODE:
		1452	json_stash = 0;
		1453	json_boolean_stash = 0;
		1454
		1455	void new (char *klass)
		1456	PPCODE:
		1457	{
		1458	SV *pv = NEWSV (0, sizeof (JSON));
		1459	SvPOK_only (pv);
		1460	Zero (SvPVX (pv), 1, JSON);
		1461	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1462	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1463	}
		1464
		1465	void ascii (JSON *self, int enable = 1)
		1466	ALIAS:
		1467	ascii = F_ASCII
		1468	latin1 = F_LATIN1
		1469	utf8 = F_UTF8
		1470	indent = F_INDENT
		1471	canonical = F_CANONICAL
		1472	space_before = F_SPACE_BEFORE
		1473	space_after = F_SPACE_AFTER
		1474	pretty = F_PRETTY
		1475	allow_nonref = F_ALLOW_NONREF
		1476	shrink = F_SHRINK
		1477	allow_blessed = F_ALLOW_BLESSED
		1478	convert_blessed = F_CONV_BLESSED
		1479	relaxed = F_RELAXED
		1480	PPCODE:
		1481	{
		1482	if (enable)
		1483	self->flags |= ix;
		1484	else
		1485	self->flags &= ~ix;
		1486
		1487	XPUSHs (ST (0));
		1488	}
		1489
		1490	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1491	PPCODE:
		1492	{
		1493	UV log2 = 0;
		1494
		1495	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1496
		1497	while ((1UL << log2) < max_depth)
		1498	++log2;
		1499
		1500	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1501
		1502	XPUSHs (ST (0));
		1503	}
		1504
		1505	void max_size (JSON *self, UV max_size = 0)
		1506	PPCODE:
		1507	{
		1508	UV log2 = 0;
		1509
		1510	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1511	if (max_size == 1) max_size = 2;
		1512
		1513	while ((1UL << log2) < max_size)
		1514	++log2;
		1515
		1516	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1517
		1518	XPUSHs (ST (0));
		1519	}
		1520
		1521	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1522	PPCODE:
		1523	{
		1524	SvREFCNT_dec (self->cb_object);
		1525	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1526
		1527	XPUSHs (ST (0));
		1528	}
		1529
		1530	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1531	PPCODE:
		1532	{
		1533	if (!self->cb_sk_object)
		1534	self->cb_sk_object = newHV ();
		1535
		1536	if (SvOK (cb))
		1537	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1538	else
791	{	1539	{
792	decode_hexdigit ['a' + i] = 10 + i;	1540	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
793	decode_hexdigit ['A' + i] = 10 + i;	1541
		1542	if (!HvKEYS (self->cb_sk_object))
		1543	{
		1544	SvREFCNT_dec (self->cb_sk_object);
		1545	self->cb_sk_object = 0;
		1546	}
794	}	1547	}
795		1548
796	json_stash = gv_stashpv ("JSON::XS", 1);	1549	XPUSHs (ST (0));
797	}	1550	}
798		1551
799	SV *new (char *dummy)	1552	void encode (JSON *self, SV *scalar)
		1553	PPCODE:
		1554	XPUSHs (encode_json (scalar, self));
		1555
		1556	void decode (JSON *self, SV *jsonstr)
		1557	PPCODE:
		1558	XPUSHs (decode_json (jsonstr, self, 0));
		1559
		1560	void decode_prefix (JSON *self, SV *jsonstr)
		1561	PPCODE:
		1562	{
		1563	UV offset;
		1564	EXTEND (SP, 2);
		1565	PUSHs (decode_json (jsonstr, self, &offset));
		1566	PUSHs (sv_2mortal (newSVuv (offset)));
		1567	}
		1568
		1569	void DESTROY (JSON *self)
800	CODE:	1570	CODE:
801	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1571	SvREFCNT_dec (self->cb_sk_object);
802	OUTPUT:	1572	SvREFCNT_dec (self->cb_object);
803	RETVAL
804		1573
805	SV *ascii (SV *self, int enable)	1574	PROTOTYPES: ENABLE
806	ALIAS:
807	ascii = F_ASCII
808	utf8 = F_UTF8
809	indent = F_INDENT
810	canonical = F_CANONICAL
811	space_before = F_SPACE_BEFORE
812	space_after = F_SPACE_AFTER
813	json_rpc = F_JSON_RPC
814	CODE:
815	{
816	UV *uv = SvJSON (self);
817	if (enable)
818	*uv |= ix;
819	else
820	*uv &= ~ix;
821		1575
822	RETVAL = newSVsv (self);	1576	void to_json (SV *scalar)
823	}
824	OUTPUT:
825	RETVAL
826
827	void encode (SV *self, SV *scalar)
828	PPCODE:	1577	PPCODE:
		1578	{
		1579	JSON json = { F_DEFAULT | F_UTF8 };
829	XPUSHs (encode_json (scalar, *SvJSON (self)));	1580	XPUSHs (encode_json (scalar, &json));
		1581	}
830		1582
831	void decode (SV *self, SV *jsondata)	1583	void from_json (SV *jsonstr)
832	PPCODE:	1584	PPCODE:
		1585	{
		1586	JSON json = { F_DEFAULT | F_UTF8 };
833	XPUSHs (decode_json (jsondata, *SvJSON (self)));	1587	XPUSHs (decode_json (jsonstr, &json, 0));
		1588	}
834		1589

Diff Legend

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