ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.9 by root, Fri Mar 23 17:40:29 2007 UTC vs.
Revision 1.67 by root, Wed Nov 28 14:01:01 2007 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <float.h>
8		10
		11	#if defined(__BORLANDC__) || defined(_MSC_VER)
		12	# define snprintf _snprintf // C compilers have this in stdio.h
		13	#endif
		14
		15	// some old perls do not have this, try to make it work, no
		16	// guarentees, though. if it breaks, you get to keep the pieces.
		17	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
9	#define F_ASCII 0x00000001	21	#define F_ASCII 0x00000001UL
		22	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	23	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	24	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	25	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	26	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	27	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
		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 ))
18		42
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		45
22	#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
23		50
24	#define SB do {	51	#define SB do {
25	#define SE } while (0)	52	#define SE } while (0)
26		53
27	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
28		123
29	// structure used for encoding JSON	124	// structure used for encoding JSON
30	typedef struct	125	typedef struct
31	{	126	{
32	char *cur;	127	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
35	SV *sv;	129	SV *sv; // result scalar
36	UV flags;	130	JSON json;
37	int max_recurse;	131	U32 indent; // indentation level
38	int indent;	132	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	133	} enc_t;
40		134
41	// structure used for decoding JSON	135	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	136	need (enc_t *enc, STRLEN len)
72	{	137	{
73	if (enc->cur + len >= enc->end)	138	if (expect_false (enc->cur + len >= enc->end))
74	{	139	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	140	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	141	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	142	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	144	}
80	}	145	}
81		146
82	static void	147	inline void
83	encode_ch (enc_t *enc, char ch)	148	encode_ch (enc_t *enc, char ch)
84	{	149	{
85	need (enc, 1);	150	need (enc, 1);
86	*enc->cur++ = ch;	151	*enc->cur++ = ch;
87	}	152	}
…		…
95		160
96	while (str < end)	161	while (str < end)
97	{	162	{
98	unsigned char ch = *(unsigned char *)str;	163	unsigned char ch = *(unsigned char *)str;
99		164
100	if (ch >= 0x20 && ch < 0x80) // most common case	165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	166	{
102	if (ch == '"') // but with slow exceptions	167	if (expect_false (ch == '"')) // but with slow exceptions
103	{	168	{
104	need (enc, len += 1);	169	need (enc, len += 1);
105	*enc->cur++ = '\\';	170	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	171	*enc->cur++ = '"';
107	}	172	}
108	else if (ch == '\\')	173	else if (expect_false (ch == '\\'))
109	{	174	{
110	need (enc, len += 1);	175	need (enc, len += 1);
111	*enc->cur++ = '\\';	176	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	177	*enc->cur++ = '\\';
113	}	178	}
…		…
131	STRLEN clen;	196	STRLEN clen;
132	UV uch;	197	UV uch;
133		198
134	if (is_utf8)	199	if (is_utf8)
135	{	200	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	201	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	202	if (clen == (STRLEN)-1)
138	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	203	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	204	}
140	else	205	else
141	{	206	{
…		…
144	}	209	}
145		210
146	if (uch > 0x10FFFFUL)	211	if (uch > 0x10FFFFUL)
147	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
148		213
149	if (uch < 0x80 || enc->flags & F_ASCII)	214	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
150	{	215	{
151	if (uch > 0xFFFFUL)	216	if (uch > 0xFFFFUL)
152	{	217	{
153	need (enc, len += 11);	218	need (enc, len += 11);
154	sprintf (enc->cur, "\\u%04x\\u%04x",	219	sprintf (enc->cur, "\\u%04x\\u%04x",
155	(uch - 0x10000) / 0x400 + 0xD800,	220	(int)((uch - 0x10000) / 0x400 + 0xD800),
156	(uch - 0x10000) % 0x400 + 0xDC00);	221	(int)((uch - 0x10000) % 0x400 + 0xDC00));
157	enc->cur += 12;	222	enc->cur += 12;
158	}	223	}
159	else	224	else
160	{	225	{
161	static char hexdigit [16] = "0123456789abcdef";	226	static char hexdigit [16] = "0123456789abcdef";
…		…
168	*enc->cur++ = hexdigit [(uch >> 0) & 15];	233	*enc->cur++ = hexdigit [(uch >> 0) & 15];
169	}	234	}
170		235
171	str += clen;	236	str += clen;
172	}	237	}
		238	else if (enc->json.flags & F_LATIN1)
		239	{
		240	*enc->cur++ = uch;
		241	str += clen;
		242	}
173	else if (is_utf8)	243	else if (is_utf8)
174	{	244	{
175	need (enc, len += clen);	245	need (enc, len += clen);
176	do	246	do
177	{	247	{
…		…
179	}	249	}
180	while (--clen);	250	while (--clen);
181	}	251	}
182	else	252	else
183	{	253	{
184	need (enc, len += 10); // never more than 11 bytes needed	254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
185	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
186	++str;	256	++str;
187	}	257	}
188	}	258	}
189	}	259	}
…		…
191		261
192	--len;	262	--len;
193	}	263	}
194	}	264	}
195		265
196	#define INDENT SB \	266	inline void
		267	encode_indent (enc_t *enc)
		268	{
197	if (enc->flags & F_INDENT) \	269	if (enc->json.flags & F_INDENT)
198	{ \	270	{
199	int i_; \	271	int spaces = enc->indent * INDENT_STEP;
200	need (enc, enc->indent); \	272
201	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);
202	encode_ch (enc, ' '); \	292	encode_ch (enc, '\n');
203	} \	293	}
204	SE	294	}
205		295
206	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	296	inline void
207	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	297	encode_comma (enc_t *enc)
208	#define COMMA SB \	298	{
209	encode_ch (enc, ','); \	299	encode_ch (enc, ',');
		300
210	if (enc->flags & F_INDENT) \	301	if (enc->json.flags & F_INDENT)
211	NL; \	302	encode_nl (enc);
212	else if (enc->flags & F_SPACE_AFTER) \	303	else if (enc->json.flags & F_SPACE_AFTER)
213	SPACE; \	304	encode_space (enc);
214	SE	305	}
215		306
216	static void encode_sv (enc_t *enc, SV *sv);	307	static void encode_sv (enc_t *enc, SV *sv);
217		308
218	static void	309	static void
219	encode_av (enc_t *enc, AV *av)	310	encode_av (enc_t *enc, AV *av)
220	{	311	{
221	int i, len = av_len (av);	312	int i, len = av_len (av);
222		313
		314	if (enc->indent >= enc->maxdepth)
		315	croak ("data structure too deep (hit recursion limit)");
		316
223	encode_ch (enc, '['); NL;	317	encode_ch (enc, '[');
224	++enc->indent;	318
		319	if (len >= 0)
		320	{
		321	encode_nl (enc); ++enc->indent;
225		322
226	for (i = 0; i <= len; ++i)	323	for (i = 0; i <= len; ++i)
227	{	324	{
228	INDENT;	325	SV **svp = av_fetch (av, i, 0);
229	encode_sv (enc, *av_fetch (av, i, 0));
230		326
		327	encode_indent (enc);
		328
		329	if (svp)
		330	encode_sv (enc, *svp);
		331	else
		332	encode_str (enc, "null", 4, 0);
		333
231	if (i < len)	334	if (i < len)
232	COMMA;	335	encode_comma (enc);
233	}	336	}
234		337
235	NL;	338	encode_nl (enc); --enc->indent; encode_indent (enc);
236		339	}
237	--enc->indent;	340
238	INDENT; encode_ch (enc, ']');	341	encode_ch (enc, ']');
239	}	342	}
240		343
241	static void	344	static void
242	encode_he (enc_t *enc, HE *he)	345	encode_hk (enc_t *enc, HE *he)
243	{	346	{
244	encode_ch (enc, '"');	347	encode_ch (enc, '"');
245		348
246	if (HeKLEN (he) == HEf_SVKEY)	349	if (HeKLEN (he) == HEf_SVKEY)
247	{	350	{
…		…
257	else	360	else
258	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	361	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259		362
260	encode_ch (enc, '"');	363	encode_ch (enc, '"');
261		364
262	if (enc->flags & F_SPACE_BEFORE) SPACE;	365	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
263	encode_ch (enc, ':');	366	encode_ch (enc, ':');
264	if (enc->flags & F_SPACE_AFTER ) SPACE;	367	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
265	encode_sv (enc, HeVAL (he));
266	}	368	}
267		369
268	// compare hash entries, used when all keys are bytestrings	370	// compare hash entries, used when all keys are bytestrings
269	static int	371	static int
270	he_cmp_fast (const void *a_, const void *b_)	372	he_cmp_fast (const void *a_, const void *b_)
…		…
275	HE *b = *(HE **)b_;	377	HE *b = *(HE **)b_;
276		378
277	STRLEN la = HeKLEN (a);	379	STRLEN la = HeKLEN (a);
278	STRLEN lb = HeKLEN (b);	380	STRLEN lb = HeKLEN (b);
279		381
280	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	382	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
281	cmp = la < lb ? -1 : la == lb ? 0 : 1;	383	cmp = lb - la;
282		384
283	return cmp;	385	return cmp;
284	}	386	}
285		387
286	// 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
287	static int	389	static int
288	he_cmp_slow (const void *a, const void *b)	390	he_cmp_slow (const void *a, const void *b)
289	{	391	{
290	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	392	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
291	}	393	}
292		394
293	static void	395	static void
294	encode_hv (enc_t *enc, HV *hv)	396	encode_hv (enc_t *enc, HV *hv)
295	{	397	{
		398	HE *he;
296	int count, i;	399	int count;
297		400
298	encode_ch (enc, '{'); NL; ++enc->indent;	401	if (enc->indent >= enc->maxdepth)
		402	croak ("data structure too deep (hit recursion limit)");
299		403
300	if ((count = hv_iterinit (hv)))	404	encode_ch (enc, '{');
301	{	405
302	// for canonical output we have to sort by keys first	406	// for canonical output we have to sort by keys first
303	// actually, this is mostly due to the stupid so-called	407	// actually, this is mostly due to the stupid so-called
304	// security workaround added somewhere in 5.8.x.	408	// security workaround added somewhere in 5.8.x.
305	// that randomises hash orderings	409	// that randomises hash orderings
306	if (enc->flags & F_CANONICAL)	410	if (enc->json.flags & F_CANONICAL)
		411	{
		412	int count = hv_iterinit (hv);
		413
		414	if (SvMAGICAL (hv))
307	{	415	{
308	HE *he, *hes [count];	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);
		426	}
		427
		428	if (count)
		429	{
309	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
310		436
311	i = 0;	437	i = 0;
312	while ((he = hv_iternext (hv)))	438	while ((he = hv_iternext (hv)))
313	{	439	{
314	hes [i++] = he;	440	hes [i++] = he;
…		…
336		462
337	FREETMPS;	463	FREETMPS;
338	LEAVE;	464	LEAVE;
339	}	465	}
340		466
341	for (i = 0; i < count; ++i)	467	encode_nl (enc); ++enc->indent;
		468
		469	while (count--)
342	{	470	{
343	INDENT;	471	encode_indent (enc);
		472	he = hes [count];
344	encode_he (enc, hes [i]);	473	encode_hk (enc, he);
		474	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
345		475
346	if (i < count - 1)	476	if (count)
347	COMMA;	477	encode_comma (enc);
348	}	478	}
349		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)))
350	NL;	487	{
		488	encode_nl (enc); ++enc->indent;
		489
		490	for (;;)
		491	{
		492	encode_indent (enc);
		493	encode_hk (enc, he);
		494	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		495
		496	if (!(he = hv_iternext (hv)))
		497	break;
		498
		499	encode_comma (enc);
		500	}
		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		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);
351	}	530	}
352	else	531	else
353	{	532	{
354	SV *sv;	533	#if 0
355	HE *he = hv_iternext (hv);	534	if (0 && sv_derived_from (rv, "JSON::Literal"))
356
357	for (;;)
358	{	535	{
359	INDENT;	536	// not yet
360	encode_he (enc, he);
361
362	if (!(he = hv_iternext (hv)))
363	break;
364
365	COMMA;
366	}	537	}
367		538	#endif
		539	if (enc->json.flags & F_CONV_BLESSED)
368	NL;	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))));
369	}	578	}
370	}	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)
		585	{
		586	STRLEN len = 0;
		587	char *pv = svt ? SvPV (sv, len) : 0;
371		588
372	--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))));
373	}	600	}
374		601
375	static void	602	static void
376	encode_sv (enc_t *enc, SV *sv)	603	encode_sv (enc_t *enc, SV *sv)
377	{	604	{
…		…
385	encode_str (enc, str, len, SvUTF8 (sv));	612	encode_str (enc, str, len, SvUTF8 (sv));
386	encode_ch (enc, '"');	613	encode_ch (enc, '"');
387	}	614	}
388	else if (SvNOKp (sv))	615	else if (SvNOKp (sv))
389	{	616	{
		617	// trust that perl will do the right thing w.r.t. JSON syntax.
390	need (enc, NV_DIG + 32);	618	need (enc, NV_DIG + 32);
391	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392	enc->cur += strlen (enc->cur);	620	enc->cur += strlen (enc->cur);
393	}	621	}
394	else if (SvIOKp (sv))	622	else if (SvIOKp (sv))
395	{	623	{
396	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);
397	enc->cur +=	629	enc->cur +=
398	SvIsUV(sv)	630	SvIsUV(sv)
399	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
400	: 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	}
401	}	661	}
402	else if (SvROK (sv))	662	else if (SvROK (sv))
403	{	663	encode_rv (enc, SvRV (sv));
404	SV *rv = SvRV (sv);
405
406	if (!--enc->max_recurse)
407	croak ("data structure too deep (hit recursion limit)");
408
409	switch (SvTYPE (rv))
410	{
411	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
412	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
413
414	default:
415	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
416	SvPV_nolen (sv));
417	}
418	}
419	else if (!SvOK (sv))	664	else if (!SvOK (sv))
420	encode_str (enc, "null", 4, 0);	665	encode_str (enc, "null", 4, 0);
421	else	666	else
422	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
423	SvPV_nolen (sv), SvFLAGS (sv));	668	SvPV_nolen (sv), SvFLAGS (sv));
424	}	669	}
425		670
426	static SV *	671	static SV *
427	encode_json (SV *scalar, UV flags)	672	encode_json (SV *scalar, JSON *json)
428	{	673	{
		674	enc_t enc;
		675
429	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	676	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	677	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
431		678
432	enc_t enc;	679	enc.json = *json;
433	enc.flags = flags;
434	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435	enc.cur = SvPVX (enc.sv);	681	enc.cur = SvPVX (enc.sv);
436	enc.end = SvEND (enc.sv);	682	enc.end = SvEND (enc.sv);
437	enc.max_recurse = 0;
438	enc.indent = 0;	683	enc.indent = 0;
		684	enc.maxdepth = DEC_DEPTH (enc.json.flags);
439		685
440	SvPOK_only (enc.sv);	686	SvPOK_only (enc.sv);
441	encode_sv (&enc, scalar);	687	encode_sv (&enc, scalar);
442		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
443	if (!(flags & (F_ASCII | F_UTF8)))	692	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
444	SvUTF8_on (enc.sv);	693	SvUTF8_on (enc.sv);
445		694
446	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447
448	if (enc.flags & F_SHRINK)	695	if (enc.json.flags & F_SHRINK)
449	shrink (enc.sv);	696	shrink (enc.sv);
450		697
451	return enc.sv;	698	return enc.sv;
452	}	699	}
453		700
454	/////////////////////////////////////////////////////////////////////////////	701	/////////////////////////////////////////////////////////////////////////////
		702	// decoder
455		703
456	#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	{
457	for (;;) \	727	for (;;)
458	{ \	728	{
459	char ch = *dec->cur; \	729	char ch = *dec->cur;
		730
460	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	}
461	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	743	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
462	break; \	744	break; // parse error, but let higher level handle it, gives better error messages
		745
463	++dec->cur; \	746	++dec->cur;
464	}	747	}
		748	}
465		749
466	#define ERR(reason) SB dec->err = reason; goto fail; SE	750	#define ERR(reason) SB dec->err = reason; goto fail; SE
		751
467	#define EXPECT_CH(ch) SB \	752	#define EXPECT_CH(ch) SB \
468	if (*dec->cur != ch) \	753	if (*dec->cur != ch) \
469	ERR (# ch " expected"); \	754	ERR (# ch " expected"); \
470	++dec->cur; \	755	++dec->cur; \
471	SE	756	SE
472		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
473	static SV *decode_sv (dec_t *dec);	761	static SV *decode_sv (dec_t *dec);
474		762
475	static signed char decode_hexdigit[256];	763	static signed char decode_hexdigit[256];
476		764
477	static UV	765	static UV
478	decode_4hex (dec_t *dec)	766	decode_4hex (dec_t *dec)
479	{	767	{
480	signed char d1, d2, d3, d4;	768	signed char d1, d2, d3, d4;
		769	unsigned char *cur = (unsigned char *)dec->cur;
481		770
482	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");
483	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");
484	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");
485	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");
486	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
487	if (d3 < 0) ERR ("four hexadecimal digits expected");
488	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
489	if (d4 < 0) ERR ("four hexadecimal digits expected");
490		775
491	dec->cur += 4;	776	dec->cur += 4;
492		777
493	return ((UV)d1) << 12	778	return ((UV)d1) << 12
494	| ((UV)d2) << 8	779	| ((UV)d2) << 8
…		…
497		782
498	fail:	783	fail:
499	return (UV)-1;	784	return (UV)-1;
500	}	785	}
501		786
502	#define APPEND_GROW(n) SB \
503	if (cur + (n) >= end) \
504	{ \
505	STRLEN ofs = cur - SvPVX (sv); \
506	SvGROW (sv, ofs + (n) + 1); \
507	cur = SvPVX (sv) + ofs; \
508	end = SvEND (sv); \
509	} \
510	SE
511
512	#define APPEND_CH(ch) SB \
513	APPEND_GROW (1); \
514	*cur++ = (ch); \
515	SE
516
517	static SV *	787	static SV *
518	decode_str (dec_t *dec)	788	decode_str (dec_t *dec)
519	{	789	{
520	SV *sv = NEWSV (0,2);	790	SV *sv = 0;
521	int utf8 = 0;	791	int utf8 = 0;
522	char *cur = SvPVX (sv);	792	char *dec_cur = dec->cur;
523	char *end = SvEND (sv);
524		793
525	for (;;)	794	do
526	{	795	{
527	unsigned char ch = *(unsigned char *)dec->cur;	796	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		797	char *cur = buf;
528		798
529	if (ch == '"')	799	do
530	break;
531	else if (ch == '\\')
532	{	800	{
533	switch (*++dec->cur)	801	unsigned char ch = *(unsigned char *)dec_cur++;
		802
		803	if (expect_false (ch == '"'))
534	{	804	{
535	case '\\':	805	--dec_cur;
536	case '/':	806	break;
537	case '"': APPEND_CH (*dec->cur++); break;	807	}
538		808	else if (expect_false (ch == '\\'))
539	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	809	{
540	case 't': APPEND_CH ('\011'); ++dec->cur; break;	810	switch (*dec_cur)
541	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
542	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
543	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
544
545	case 'u':
546	{	811	{
547	UV lo, hi;	812	case '\\':
548	++dec->cur;	813	case '/':
		814	case '"': *cur++ = *dec_cur++; break;
549		815
550	hi = decode_4hex (dec);	816	case 'b': ++dec_cur; *cur++ = '\010'; break;
551	if (hi == (UV)-1)	817	case 't': ++dec_cur; *cur++ = '\011'; break;
552	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;
553		821
554	// possibly a surrogate pair	822	case 'u':
555	if (hi >= 0xd800 && hi < 0xdc00)
556	{	823	{
557	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	824	UV lo, hi;
558	ERR ("missing low surrogate character in surrogate pair");	825	++dec_cur;
559		826
560	dec->cur += 2;	827	dec->cur = dec_cur;
561
562	lo = decode_4hex (dec);	828	hi = decode_4hex (dec);
		829	dec_cur = dec->cur;
563	if (lo == (UV)-1)	830	if (hi == (UV)-1)
564	goto fail;	831	goto fail;
565		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
566	if (lo < 0xdc00 || lo >= 0xe000)	848	if (lo < 0xdc00 || lo >= 0xe000)
567	ERR ("surrogate pair expected");	849	ERR ("surrogate pair expected");
568		850
569	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;
570	}	864	}
571	else if (hi >= 0xdc00 && hi < 0xe000)
572	ERR ("missing high surrogate character in surrogate pair");
573
574	if (hi >= 0x80)
575	{	865	break;
576	utf8 = 1;
577		866
578	APPEND_GROW (4); // at most 4 bytes for 21 bits
579	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	}
581	else	867	default:
582	APPEND_CH (hi);	868	--dec_cur;
		869	ERR ("illegal backslash escape sequence in string");
583	}	870	}
584	break;	871	}
		872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		873	*cur++ = ch;
		874	else if (ch >= 0x80)
		875	{
		876	STRLEN clen;
		877	UV uch;
585		878
586	default:
587	--dec->cur;	879	--dec_cur;
588	ERR ("illegal backslash escape sequence in string");	880
		881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		882	if (clen == (STRLEN)-1)
		883	ERR ("malformed UTF-8 character in JSON string");
		884
		885	do
		886	*cur++ = *dec_cur++;
		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");
589	}	899	}
590	}	900	}
591	else if (ch >= 0x20 && ch <= 0x7f)	901	while (cur < buf + SHORT_STRING_LEN);
592	APPEND_CH (*dec->cur++);	902
593	else if (ch >= 0x80)
594	{	903	{
595	STRLEN clen;	904	STRLEN len = cur - buf;
596	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
597	if (clen == (STRLEN)-1)
598	ERR ("malformed UTF-8 character in JSON string");
599		905
600	APPEND_GROW (clen);	906	if (sv)
601	do
602	{	907	{
603	*cur++ = *dec->cur++;	908	SvGROW (sv, SvCUR (sv) + len + 1);
		909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		910	SvCUR_set (sv, SvCUR (sv) + len);
604	}	911	}
605	while (--clen);
606
607	utf8 = 1;
608	}
609	else if (dec->cur == dec->end)
610	ERR ("unexpected end of string while parsing json string");
611	else	912	else
612	ERR ("invalid character encountered");	913	sv = newSVpvn (buf, len);
613	}	914	}
		915	}
		916	while (*dec_cur != '"');
614		917
615	++dec->cur;	918	++dec_cur;
616		919
617	SvCUR_set (sv, cur - SvPVX (sv));	920	if (sv)
618		921	{
619	SvPOK_only (sv);	922	SvPOK_only (sv);
620	*SvEND (sv) = 0;	923	*SvEND (sv) = 0;
621		924
622	if (utf8)	925	if (utf8)
623	SvUTF8_on (sv);	926	SvUTF8_on (sv);
		927	}
		928	else
		929	sv = newSVpvn ("", 0);
624		930
625	if (dec->flags & F_SHRINK)	931	dec->cur = dec_cur;
626	shrink (sv);
627
628	return sv;	932	return sv;
629		933
630	fail:	934	fail:
631	SvREFCNT_dec (sv);	935	dec->cur = dec_cur;
632	return 0;	936	return 0;
633	}	937	}
634		938
635	static SV *	939	static SV *
636	decode_num (dec_t *dec)	940	decode_num (dec_t *dec)
…		…
694	is_nv = 1;	998	is_nv = 1;
695	}	999	}
696		1000
697	if (!is_nv)	1001	if (!is_nv)
698	{	1002	{
699	UV uv;	1003	int len = dec->cur - start;
700	int numtype = grok_number (start, dec->cur - start, &uv);	1004
701	if (numtype & IS_NUMBER_IN_UV)	1005	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
702	if (numtype & IS_NUMBER_NEG)	1006	if (*start == '-')
		1007	switch (len)
703	{	1008	{
704	if (uv < (UV)IV_MIN)	1009	case 2: return newSViv (-( start [1] - '0' * 1));
705	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));
706	}	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	}
707	else	1032	else
708	return newSVuv (uv);	1033	return newSVuv (uv);
709	}	1034	}
710		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
711	return newSVnv (Atof (start));	1052	return newSVnv (Atof (start));
712		1053
713	fail:	1054	fail:
714	return 0;	1055	return 0;
715	}	1056	}
…		…
717	static SV *	1058	static SV *
718	decode_av (dec_t *dec)	1059	decode_av (dec_t *dec)
719	{	1060	{
720	AV *av = newAV ();	1061	AV *av = newAV ();
721		1062
722	WS;	1063	DEC_INC_DEPTH;
		1064	decode_ws (dec);
		1065
723	if (*dec->cur == ']')	1066	if (*dec->cur == ']')
724	++dec->cur;	1067	++dec->cur;
725	else	1068	else
726	for (;;)	1069	for (;;)
727	{	1070	{
…		…
731	if (!value)	1074	if (!value)
732	goto fail;	1075	goto fail;
733		1076
734	av_push (av, value);	1077	av_push (av, value);
735		1078
736	WS;	1079	decode_ws (dec);
737		1080
738	if (*dec->cur == ']')	1081	if (*dec->cur == ']')
739	{	1082	{
740	++dec->cur;	1083	++dec->cur;
741	break;	1084	break;
…		…
743		1086
744	if (*dec->cur != ',')	1087	if (*dec->cur != ',')
745	ERR (", or ] expected while parsing array");	1088	ERR (", or ] expected while parsing array");
746		1089
747	++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	}
748	}	1099	}
749		1100
		1101	DEC_DEC_DEPTH;
750	return newRV_noinc ((SV *)av);	1102	return newRV_noinc ((SV *)av);
751		1103
752	fail:	1104	fail:
753	SvREFCNT_dec (av);	1105	SvREFCNT_dec (av);
		1106	DEC_DEC_DEPTH;
754	return 0;	1107	return 0;
755	}	1108	}
756		1109
757	static SV *	1110	static SV *
758	decode_hv (dec_t *dec)	1111	decode_hv (dec_t *dec)
759	{	1112	{
		1113	SV *sv;
760	HV *hv = newHV ();	1114	HV *hv = newHV ();
761		1115
762	WS;	1116	DEC_INC_DEPTH;
		1117	decode_ws (dec);
		1118
763	if (*dec->cur == '}')	1119	if (*dec->cur == '}')
764	++dec->cur;	1120	++dec->cur;
765	else	1121	else
766	for (;;)	1122	for (;;)
767	{	1123	{
768	SV *key, *value;
769
770	WS; EXPECT_CH ('"');	1124	EXPECT_CH ('"');
771		1125
772	key = decode_str (dec);	1126	// heuristic: assume that
773	if (!key)	1127	// a) decode_str + hv_store_ent are abysmally slow.
774	goto fail;	1128	// b) most hash keys are short, simple ascii text.
		1129	// => try to "fast-match" such strings to avoid
		1130	// the overhead of decode_str + hv_store_ent.
		1131	{
		1132	SV *value;
		1133	char *p = dec->cur;
		1134	char *e = p + 24; // only try up to 24 bytes
775		1135
776	WS; EXPECT_CH (':');	1136	for (;;)
777
778	value = decode_sv (dec);
779	if (!value)
780	{	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
		1156	hv_store_ent (hv, key, value, 0);
781	SvREFCNT_dec (key);	1157	SvREFCNT_dec (key);
		1158
		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;
		1167
		1168	decode_ws (dec); EXPECT_CH (':');
		1169
		1170	decode_ws (dec);
		1171	value = decode_sv (dec);
		1172	if (!value)
782	goto fail;	1173	goto fail;
		1174
		1175	hv_store (hv, key, len, value, 0);
		1176
		1177	break;
		1178	}
		1179
		1180	++p;
783	}	1181	}
784
785	//TODO: optimise
786	hv_store_ent (hv, key, value, 0);
787
788	WS;	1182	}
		1183
		1184	decode_ws (dec);
789		1185
790	if (*dec->cur == '}')	1186	if (*dec->cur == '}')
791	{	1187	{
792	++dec->cur;	1188	++dec->cur;
793	break;	1189	break;
…		…
795		1191
796	if (*dec->cur != ',')	1192	if (*dec->cur != ',')
797	ERR (", or } expected while parsing object/hash");	1193	ERR (", or } expected while parsing object/hash");
798		1194
799	++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	}
800	}	1204	}
801		1205
		1206	DEC_DEC_DEPTH;
802	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;
803		1268
804	fail:	1269	fail:
805	SvREFCNT_dec (hv);	1270	SvREFCNT_dec (hv);
		1271	DEC_DEC_DEPTH;
806	return 0;	1272	return 0;
807	}	1273	}
808		1274
809	static SV *	1275	static SV *
810	decode_sv (dec_t *dec)	1276	decode_sv (dec_t *dec)
811	{	1277	{
812	WS;	1278	// the beauty of JSON: you need exactly one character lookahead
		1279	// to parse anything.
813	switch (*dec->cur)	1280	switch (*dec->cur)
814	{	1281	{
815	case '"': ++dec->cur; return decode_str (dec);	1282	case '"': ++dec->cur; return decode_str (dec);
816	case '[': ++dec->cur; return decode_av (dec);	1283	case '[': ++dec->cur; return decode_av (dec);
817	case '{': ++dec->cur; return decode_hv (dec);	1284	case '{': ++dec->cur; return decode_hv (dec);
…		…
823		1290
824	case 't':	1291	case 't':
825	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1292	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826	{	1293	{
827	dec->cur += 4;	1294	dec->cur += 4;
828	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);
829	}	1299	}
830	else	1300	else
831	ERR ("'true' expected");	1301	ERR ("'true' expected");
832		1302
833	break;	1303	break;
834		1304
835	case 'f':	1305	case 'f':
836	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1306	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837	{	1307	{
838	dec->cur += 5;	1308	dec->cur += 5;
839	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);
840	}	1313	}
841	else	1314	else
842	ERR ("'false' expected");	1315	ERR ("'false' expected");
843		1316
844	break;	1317	break;
…		…
853	ERR ("'null' expected");	1326	ERR ("'null' expected");
854		1327
855	break;	1328	break;
856		1329
857	default:	1330	default:
858	ERR ("malformed json string, neither array, object, number, string or atom");	1331	ERR ("malformed JSON string, neither array, object, number, string or atom");
859	break;	1332	break;
860	}	1333	}
861		1334
862	fail:	1335	fail:
863	return 0;	1336	return 0;
864	}	1337	}
865		1338
866	static SV *	1339	static SV *
867	decode_json (SV *string, UV flags)	1340	decode_json (SV *string, JSON *json, UV *offset_return)
868	{	1341	{
		1342	dec_t dec;
		1343	UV offset;
869	SV *sv;	1344	SV *sv;
870		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
871	if (flags & F_UTF8)	1353	if (json->flags & F_UTF8)
872	sv_utf8_downgrade (string, 0);	1354	sv_utf8_downgrade (string, 0);
873	else	1355	else
874	sv_utf8_upgrade (string);	1356	sv_utf8_upgrade (string);
875		1357
876	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1358	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877		1359
878	dec_t dec;	1360	dec.json = *json;
879	dec.flags = flags;
880	dec.cur = SvPVX (string);	1361	dec.cur = SvPVX (string);
881	dec.end = SvEND (string);	1362	dec.end = SvEND (string);
882	dec.err = 0;	1363	dec.err = 0;
		1364	dec.depth = 0;
		1365	dec.maxdepth = DEC_DEPTH (dec.json.flags);
883		1366
		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);
884	sv = decode_sv (&dec);	1373	sv = decode_sv (&dec);
885		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	}
		1397
886	if (!sv)	1398	if (!sv)
887	{	1399	{
888	IV offset = dec.flags & F_UTF8
889	? dec.cur - SvPVX (string)
890	: utf8_distance (dec.cur, SvPVX (string));
891	SV *uni = sv_newmortal ();	1400	SV *uni = sv_newmortal ();
892		1401
893	// horrible hack to silence warning inside pv_uni_display	1402	// horrible hack to silence warning inside pv_uni_display
894	COP cop = *PL_curcop;	1403	COP cop = *PL_curcop;
895	cop.cop_warnings = pWARN_NONE;	1404	cop.cop_warnings = pWARN_NONE;
…		…
897	SAVEVPTR (PL_curcop);	1406	SAVEVPTR (PL_curcop);
898	PL_curcop = &cop;	1407	PL_curcop = &cop;
899	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);
900	LEAVE;	1409	LEAVE;
901		1410
902	croak ("%s, at character offset %d (%s)",	1411	croak ("%s, at character offset %d [\"%s\"]",
903	dec.err,	1412	dec.err,
904	(int)offset,	1413	(int)offset,
905	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1414	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906	}	1415	}
907		1416
908	sv = sv_2mortal (sv);	1417	sv = sv_2mortal (sv);
909		1418
910	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1419	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
911	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1420	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
912		1421
913	return sv;	1422	return sv;
914	}	1423	}
915		1424
		1425	/////////////////////////////////////////////////////////////////////////////
		1426	// XS interface functions
		1427
916	MODULE = JSON::XS PACKAGE = JSON::XS	1428	MODULE = JSON::XS PACKAGE = JSON::XS
917		1429
918	BOOT:	1430	BOOT:
919	{	1431	{
920	int i;	1432	int i;
921		1433
922	memset (decode_hexdigit, 0xff, 256);
923	for (i = 10; i--; )	1434	for (i = 0; i < 256; ++i)
924	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;
925		1440
926	for (i = 7; i--; )
927	{
928	decode_hexdigit ['a' + i] = 10 + i;
929	decode_hexdigit ['A' + i] = 10 + i;
930	}
931
932	json_stash = gv_stashpv ("JSON::XS", 1);	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);
933	}	1446	}
934		1447
935	PROTOTYPES: DISABLE	1448	PROTOTYPES: DISABLE
936		1449
937	SV *new (char *dummy)	1450	void CLONE (...)
938	CODE:	1451	CODE:
939	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	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 get_ascii (JSON *self)
		1491	ALIAS:
		1492	get_ascii = F_ASCII
		1493	get_latin1 = F_LATIN1
		1494	get_utf8 = F_UTF8
		1495	get_indent = F_INDENT
		1496	get_canonical = F_CANONICAL
		1497	get_space_before = F_SPACE_BEFORE
		1498	get_space_after = F_SPACE_AFTER
		1499	get_pretty = F_PRETTY
		1500	get_allow_nonref = F_ALLOW_NONREF
		1501	get_shrink = F_SHRINK
		1502	get_allow_blessed = F_ALLOW_BLESSED
		1503	get_convert_blessed = F_CONV_BLESSED
		1504	get_relaxed = F_RELAXED
		1505	PPCODE:
		1506	XPUSHs (boolSV (self->flags & ix));
		1507
		1508	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1509	PPCODE:
		1510	{
		1511	UV log2 = 0;
		1512
		1513	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1514
		1515	while ((1UL << log2) < max_depth)
		1516	++log2;
		1517
		1518	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1519
		1520	XPUSHs (ST (0));
		1521	}
		1522
		1523	U32 get_max_depth (JSON *self)
		1524	CODE:
		1525	RETVAL = DEC_DEPTH (self->flags);
940	OUTPUT:	1526	OUTPUT:
941	RETVAL	1527	RETVAL
942		1528
943	SV *ascii (SV *self, int enable = 1)	1529	void max_size (JSON *self, UV max_size = 0)
944	ALIAS:	1530	PPCODE:
945	ascii = F_ASCII	1531	{
946	utf8 = F_UTF8	1532	UV log2 = 0;
947	indent = F_INDENT	1533
948	canonical = F_CANONICAL	1534	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
949	space_before = F_SPACE_BEFORE	1535	if (max_size == 1) max_size = 2;
950	space_after = F_SPACE_AFTER	1536
951	json_rpc = F_JSON_RPC	1537	while ((1UL << log2) < max_size)
952	pretty = F_PRETTY	1538	++log2;
953	allow_nonref = F_ALLOW_NONREF	1539
954	shrink = F_SHRINK	1540	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1541
		1542	XPUSHs (ST (0));
		1543	}
		1544
		1545	int get_max_size (JSON *self)
955	CODE:	1546	CODE:
956	{	1547	RETVAL = DEC_SIZE (self->flags);
957	UV *uv = SvJSON (self);
958	if (enable)
959	*uv |= ix;
960	else
961	*uv &= ~ix;
962
963	RETVAL = newSVsv (self);
964	}
965	OUTPUT:	1548	OUTPUT:
966	RETVAL	1549	RETVAL
967		1550
968	void encode (SV *self, SV *scalar)	1551	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
969	PPCODE:	1552	PPCODE:
970	XPUSHs (encode_json (scalar, *SvJSON (self)));	1553	{
		1554	SvREFCNT_dec (self->cb_object);
		1555	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
971		1556
972	void decode (SV *self, SV *jsonstr)	1557	XPUSHs (ST (0));
		1558	}
		1559
		1560	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
973	PPCODE:	1561	PPCODE:
		1562	{
		1563	if (!self->cb_sk_object)
		1564	self->cb_sk_object = newHV ();
		1565
		1566	if (SvOK (cb))
		1567	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1568	else
		1569	{
		1570	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1571
		1572	if (!HvKEYS (self->cb_sk_object))
		1573	{
		1574	SvREFCNT_dec (self->cb_sk_object);
		1575	self->cb_sk_object = 0;
		1576	}
		1577	}
		1578
		1579	XPUSHs (ST (0));
		1580	}
		1581
		1582	void encode (JSON *self, SV *scalar)
		1583	PPCODE:
		1584	XPUSHs (encode_json (scalar, self));
		1585
		1586	void decode (JSON *self, SV *jsonstr)
		1587	PPCODE:
974	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1588	XPUSHs (decode_json (jsonstr, self, 0));
		1589
		1590	void decode_prefix (JSON *self, SV *jsonstr)
		1591	PPCODE:
		1592	{
		1593	UV offset;
		1594	EXTEND (SP, 2);
		1595	PUSHs (decode_json (jsonstr, self, &offset));
		1596	PUSHs (sv_2mortal (newSVuv (offset)));
		1597	}
		1598
		1599	void DESTROY (JSON *self)
		1600	CODE:
		1601	SvREFCNT_dec (self->cb_sk_object);
		1602	SvREFCNT_dec (self->cb_object);
975		1603
976	PROTOTYPES: ENABLE	1604	PROTOTYPES: ENABLE
977		1605
978	void to_json (SV *scalar)	1606	void to_json (SV *scalar)
979	PPCODE:	1607	PPCODE:
		1608	{
		1609	JSON json = { F_DEFAULT | F_UTF8 };
980	XPUSHs (encode_json (scalar, F_UTF8));	1610	XPUSHs (encode_json (scalar, &json));
		1611	}
981		1612
982	void from_json (SV *jsonstr)	1613	void from_json (SV *jsonstr)
983	PPCODE:	1614	PPCODE:
		1615	{
		1616	JSON json = { F_DEFAULT | F_UTF8 };
984	XPUSHs (decode_json (jsonstr, F_UTF8));	1617	XPUSHs (decode_json (jsonstr, &json, 0));
		1618	}
985		1619

Diff Legend

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