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 1.72 by root, Wed Mar 19 04:08:22 2008 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 static 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	#define IN_RANGE_INC(type,val,beg,end) \
		66	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		67	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		68
		69	#ifdef USE_ITHREADS
		70	# define JSON_SLOW 1
		71	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		72	#else
		73	# define JSON_SLOW 0
		74	# define JSON_STASH json_stash
		75	#endif
		76
		77	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		78	static SV *json_true, *json_false;
		79
		80	typedef struct {
		81	U32 flags;
		82	SV *cb_object;
		83	HV *cb_sk_object;
		84	} JSON;
		85
		86	/////////////////////////////////////////////////////////////////////////////
		87	// utility functions
		88
		89	INLINE void
		90	shrink (SV *sv)
		91	{
		92	sv_utf8_downgrade (sv, 1);
		93	if (SvLEN (sv) > SvCUR (sv) + 1)
		94	{
		95	#ifdef SvPV_shrink_to_cur
		96	SvPV_shrink_to_cur (sv);
		97	#elif defined (SvPV_renew)
		98	SvPV_renew (sv, SvCUR (sv) + 1);
		99	#endif
		100	}
		101	}
		102
		103	// decode an utf-8 character and return it, or (UV)-1 in
		104	// case of an error.
		105	// we special-case "safe" characters from U+80 .. U+7FF,
		106	// but use the very good perl function to parse anything else.
		107	// note that we never call this function for a ascii codepoints
		108	INLINE UV
		109	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		110	{
		111	if (expect_true (len >= 2
		112	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		113	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		114	{
		115	*clen = 2;
		116	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		117	}
		118	else
		119	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		120	}
		121
		122	// likewise for encoding, also never called for ascii codepoints
		123	// this function takes advantage of this fact, although current gccs
		124	// seem to optimise the check for >= 0x80 away anyways
		125	INLINE unsigned char *
		126	encode_utf8 (unsigned char *s, UV ch)
		127	{
		128	if (ch <= 0x7FF)
		129	{
		130	*s++ = (ch >> 6) | 0xc0;
		131	*s++ = (ch & 0x3f) | 0x80;
		132	}
		133	else
		134	s = uvuni_to_utf8_flags (s, ch, 0);
		135
		136	return s;
		137	}
		138
		139	/////////////////////////////////////////////////////////////////////////////
		140	// encoder
28		141
29	// structure used for encoding JSON	142	// structure used for encoding JSON
30	typedef struct	143	typedef struct
31	{	144	{
32	char *cur;	145	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	146	char *end; // SvEND (sv)
35	SV *sv;	147	SV *sv; // result scalar
36	UV flags;	148	JSON json;
37	int max_recurse;	149	U32 indent; // indentation level
38	int indent;	150	U32 maxdepth; // max. indentation/recursion level
		151	UV limit; // escape character values >= this value when encoding
39	} enc_t;	152	} enc_t;
40		153
41	// structure used for decoding JSON	154	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)	155	need (enc_t *enc, STRLEN len)
72	{	156	{
73	if (enc->cur + len >= enc->end)	157	if (expect_false (enc->cur + len >= enc->end))
74	{	158	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	159	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	160	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	161	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	162	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	163	}
80	}	164	}
81		165
82	static void	166	INLINE void
83	encode_ch (enc_t *enc, char ch)	167	encode_ch (enc_t *enc, char ch)
84	{	168	{
85	need (enc, 1);	169	need (enc, 1);
86	*enc->cur++ = ch;	170	*enc->cur++ = ch;
87	}	171	}
…		…
95		179
96	while (str < end)	180	while (str < end)
97	{	181	{
98	unsigned char ch = *(unsigned char *)str;	182	unsigned char ch = *(unsigned char *)str;
99		183
100	if (ch >= 0x20 && ch < 0x80) // most common case	184	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	185	{
102	if (ch == '"') // but with slow exceptions	186	if (expect_false (ch == '"')) // but with slow exceptions
103	{	187	{
104	need (enc, len += 1);	188	need (enc, len += 1);
105	*enc->cur++ = '\\';	189	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	190	*enc->cur++ = '"';
107	}	191	}
108	else if (ch == '\\')	192	else if (expect_false (ch == '\\'))
109	{	193	{
110	need (enc, len += 1);	194	need (enc, len += 1);
111	*enc->cur++ = '\\';	195	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	196	*enc->cur++ = '\\';
113	}	197	}
…		…
131	STRLEN clen;	215	STRLEN clen;
132	UV uch;	216	UV uch;
133		217
134	if (is_utf8)	218	if (is_utf8)
135	{	219	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	220	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	221	if (clen == (STRLEN)-1)
138	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	222	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	223	}
140	else	224	else
141	{	225	{
142	uch = ch;	226	uch = ch;
143	clen = 1;	227	clen = 1;
144	}	228	}
145		229
146	if (uch > 0x10FFFFUL)	230	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
147	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
148
149	if (uch < 0x80 || enc->flags & F_ASCII)
150	{	231	{
151	if (uch > 0xFFFFUL)	232	if (uch > 0xFFFFUL)
152	{	233	{
		234	if (uch > 0x10FFFFUL)
		235	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		236
153	need (enc, len += 11);	237	need (enc, len += 11);
154	sprintf (enc->cur, "\\u%04x\\u%04x",	238	sprintf (enc->cur, "\\u%04x\\u%04x",
155	(uch - 0x10000) / 0x400 + 0xD800,	239	(int)((uch - 0x10000) / 0x400 + 0xD800),
156	(uch - 0x10000) % 0x400 + 0xDC00);	240	(int)((uch - 0x10000) % 0x400 + 0xDC00));
157	enc->cur += 12;	241	enc->cur += 12;
158	}	242	}
159	else	243	else
160	{	244	{
161	static char hexdigit [16] = "0123456789abcdef";	245	static char hexdigit [16] = "0123456789abcdef";
…		…
168	*enc->cur++ = hexdigit [(uch >> 0) & 15];	252	*enc->cur++ = hexdigit [(uch >> 0) & 15];
169	}	253	}
170		254
171	str += clen;	255	str += clen;
172	}	256	}
		257	else if (enc->json.flags & F_LATIN1)
		258	{
		259	*enc->cur++ = uch;
		260	str += clen;
		261	}
173	else if (is_utf8)	262	else if (is_utf8)
174	{	263	{
175	need (enc, len += clen);	264	need (enc, len += clen);
176	do	265	do
177	{	266	{
…		…
179	}	268	}
180	while (--clen);	269	while (--clen);
181	}	270	}
182	else	271	else
183	{	272	{
184	need (enc, len += 10); // never more than 11 bytes needed	273	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
185	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	274	enc->cur = encode_utf8 (enc->cur, uch);
186	++str;	275	++str;
187	}	276	}
188	}	277	}
189	}	278	}
190	}	279	}
191		280
192	--len;	281	--len;
193	}	282	}
194	}	283	}
195		284
196	#define INDENT SB \	285	INLINE void
		286	encode_indent (enc_t *enc)
		287	{
197	if (enc->flags & F_INDENT) \	288	if (enc->json.flags & F_INDENT)
198	{ \	289	{
199	int i_; \	290	int spaces = enc->indent * INDENT_STEP;
200	need (enc, enc->indent); \	291
201	for (i_ = enc->indent * 3; i_--; )\	292	need (enc, spaces);
		293	memset (enc->cur, ' ', spaces);
		294	enc->cur += spaces;
		295	}
		296	}
		297
		298	INLINE void
		299	encode_space (enc_t *enc)
		300	{
		301	need (enc, 1);
		302	encode_ch (enc, ' ');
		303	}
		304
		305	INLINE void
		306	encode_nl (enc_t *enc)
		307	{
		308	if (enc->json.flags & F_INDENT)
		309	{
		310	need (enc, 1);
202	encode_ch (enc, ' '); \	311	encode_ch (enc, '\n');
203	} \	312	}
204	SE	313	}
205		314
206	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	315	INLINE void
207	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	316	encode_comma (enc_t *enc)
208	#define COMMA SB \	317	{
209	encode_ch (enc, ','); \	318	encode_ch (enc, ',');
		319
210	if (enc->flags & F_INDENT) \	320	if (enc->json.flags & F_INDENT)
211	NL; \	321	encode_nl (enc);
212	else if (enc->flags & F_SPACE_AFTER) \	322	else if (enc->json.flags & F_SPACE_AFTER)
213	SPACE; \	323	encode_space (enc);
214	SE	324	}
215		325
216	static void encode_sv (enc_t *enc, SV *sv);	326	static void encode_sv (enc_t *enc, SV *sv);
217		327
218	static void	328	static void
219	encode_av (enc_t *enc, AV *av)	329	encode_av (enc_t *enc, AV *av)
220	{	330	{
221	int i, len = av_len (av);	331	int i, len = av_len (av);
222		332
		333	if (enc->indent >= enc->maxdepth)
		334	croak ("data structure too deep (hit recursion limit)");
		335
223	encode_ch (enc, '['); NL;	336	encode_ch (enc, '[');
224	++enc->indent;	337
		338	if (len >= 0)
		339	{
		340	encode_nl (enc); ++enc->indent;
225		341
226	for (i = 0; i <= len; ++i)	342	for (i = 0; i <= len; ++i)
227	{	343	{
228	INDENT;	344	SV **svp = av_fetch (av, i, 0);
229	encode_sv (enc, *av_fetch (av, i, 0));
230		345
		346	encode_indent (enc);
		347
		348	if (svp)
		349	encode_sv (enc, *svp);
		350	else
		351	encode_str (enc, "null", 4, 0);
		352
231	if (i < len)	353	if (i < len)
232	COMMA;	354	encode_comma (enc);
233	}	355	}
234		356
235	NL;	357	encode_nl (enc); --enc->indent; encode_indent (enc);
236		358	}
237	--enc->indent;	359
238	INDENT; encode_ch (enc, ']');	360	encode_ch (enc, ']');
239	}	361	}
240		362
241	static void	363	static void
242	encode_he (enc_t *enc, HE *he)	364	encode_hk (enc_t *enc, HE *he)
243	{	365	{
244	encode_ch (enc, '"');	366	encode_ch (enc, '"');
245		367
246	if (HeKLEN (he) == HEf_SVKEY)	368	if (HeKLEN (he) == HEf_SVKEY)
247	{	369	{
…		…
257	else	379	else
258	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	380	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259		381
260	encode_ch (enc, '"');	382	encode_ch (enc, '"');
261		383
262	if (enc->flags & F_SPACE_BEFORE) SPACE;	384	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
263	encode_ch (enc, ':');	385	encode_ch (enc, ':');
264	if (enc->flags & F_SPACE_AFTER ) SPACE;	386	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
265	encode_sv (enc, HeVAL (he));
266	}	387	}
267		388
268	// compare hash entries, used when all keys are bytestrings	389	// compare hash entries, used when all keys are bytestrings
269	static int	390	static int
270	he_cmp_fast (const void *a_, const void *b_)	391	he_cmp_fast (const void *a_, const void *b_)
…		…
275	HE *b = *(HE **)b_;	396	HE *b = *(HE **)b_;
276		397
277	STRLEN la = HeKLEN (a);	398	STRLEN la = HeKLEN (a);
278	STRLEN lb = HeKLEN (b);	399	STRLEN lb = HeKLEN (b);
279		400
280	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	401	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
281	cmp = la < lb ? -1 : la == lb ? 0 : 1;	402	cmp = lb - la;
282		403
283	return cmp;	404	return cmp;
284	}	405	}
285		406
286	// compare hash entries, used when some keys are sv's or utf-x	407	// compare hash entries, used when some keys are sv's or utf-x
287	static int	408	static int
288	he_cmp_slow (const void *a, const void *b)	409	he_cmp_slow (const void *a, const void *b)
289	{	410	{
290	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	411	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
291	}	412	}
292		413
293	static void	414	static void
294	encode_hv (enc_t *enc, HV *hv)	415	encode_hv (enc_t *enc, HV *hv)
295	{	416	{
		417	HE *he;
296	int count, i;	418	int count;
297		419
298	encode_ch (enc, '{'); NL; ++enc->indent;	420	if (enc->indent >= enc->maxdepth)
		421	croak ("data structure too deep (hit recursion limit)");
299		422
300	if ((count = hv_iterinit (hv)))	423	encode_ch (enc, '{');
301	{	424
302	// for canonical output we have to sort by keys first	425	// for canonical output we have to sort by keys first
303	// actually, this is mostly due to the stupid so-called	426	// actually, this is mostly due to the stupid so-called
304	// security workaround added somewhere in 5.8.x.	427	// security workaround added somewhere in 5.8.x.
305	// that randomises hash orderings	428	// that randomises hash orderings
306	if (enc->flags & F_CANONICAL)	429	if (enc->json.flags & F_CANONICAL)
		430	{
		431	int count = hv_iterinit (hv);
		432
		433	if (SvMAGICAL (hv))
307	{	434	{
308	HE *he, *hes [count];	435	// need to count by iterating. could improve by dynamically building the vector below
		436	// but I don't care for the speed of this special case.
		437	// note also that we will run into undefined behaviour when the two iterations
		438	// do not result in the same count, something I might care for in some later release.
		439
		440	count = 0;
		441	while (hv_iternext (hv))
		442	++count;
		443
		444	hv_iterinit (hv);
		445	}
		446
		447	if (count)
		448	{
309	int fast = 1;	449	int i, fast = 1;
		450	#if defined(__BORLANDC__) || defined(_MSC_VER)
		451	HE **hes = _alloca (count * sizeof (HE));
		452	#else
		453	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		454	#endif
310		455
311	i = 0;	456	i = 0;
312	while ((he = hv_iternext (hv)))	457	while ((he = hv_iternext (hv)))
313	{	458	{
314	hes [i++] = he;	459	hes [i++] = he;
…		…
336		481
337	FREETMPS;	482	FREETMPS;
338	LEAVE;	483	LEAVE;
339	}	484	}
340		485
341	for (i = 0; i < count; ++i)	486	encode_nl (enc); ++enc->indent;
		487
		488	while (count--)
342	{	489	{
343	INDENT;	490	encode_indent (enc);
		491	he = hes [count];
344	encode_he (enc, hes [i]);	492	encode_hk (enc, he);
		493	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
345		494
346	if (i < count - 1)	495	if (count)
347	COMMA;	496	encode_comma (enc);
348	}	497	}
349		498
		499	encode_nl (enc); --enc->indent; encode_indent (enc);
		500	}
		501	}
		502	else
		503	{
		504	if (hv_iterinit (hv) || SvMAGICAL (hv))
		505	if ((he = hv_iternext (hv)))
350	NL;	506	{
		507	encode_nl (enc); ++enc->indent;
		508
		509	for (;;)
		510	{
		511	encode_indent (enc);
		512	encode_hk (enc, he);
		513	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		514
		515	if (!(he = hv_iternext (hv)))
		516	break;
		517
		518	encode_comma (enc);
		519	}
		520
		521	encode_nl (enc); --enc->indent; encode_indent (enc);
		522	}
		523	}
		524
		525	encode_ch (enc, '}');
		526	}
		527
		528	// encode objects, arrays and special \0=false and \1=true values.
		529	static void
		530	encode_rv (enc_t *enc, SV *sv)
		531	{
		532	svtype svt;
		533
		534	SvGETMAGIC (sv);
		535	svt = SvTYPE (sv);
		536
		537	if (expect_false (SvOBJECT (sv)))
		538	{
		539	HV *stash = !JSON_SLOW || json_boolean_stash
		540	? json_boolean_stash
		541	: gv_stashpv ("JSON::XS::Boolean", 1);
		542
		543	if (SvSTASH (sv) == stash)
		544	{
		545	if (SvIV (sv))
		546	encode_str (enc, "true", 4, 0);
		547	else
		548	encode_str (enc, "false", 5, 0);
351	}	549	}
352	else	550	else
353	{	551	{
354	SV *sv;	552	#if 0
355	HE *he = hv_iternext (hv);	553	if (0 && sv_derived_from (rv, "JSON::Literal"))
356
357	for (;;)
358	{	554	{
359	INDENT;	555	// not yet
360	encode_he (enc, he);
361
362	if (!(he = hv_iternext (hv)))
363	break;
364
365	COMMA;
366	}	556	}
367		557	#endif
		558	if (enc->json.flags & F_CONV_BLESSED)
368	NL;	559	{
		560	// we re-bless the reference to get overload and other niceties right
		561	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		562
		563	if (to_json)
		564	{
		565	dSP;
		566
		567	ENTER; SAVETMPS; PUSHMARK (SP);
		568	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		569
		570	// calling with G_SCALAR ensures that we always get a 1 return value
		571	PUTBACK;
		572	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		573	SPAGAIN;
		574
		575	// catch this surprisingly common error
		576	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		577	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		578
		579	sv = POPs;
		580	PUTBACK;
		581
		582	encode_sv (enc, sv);
		583
		584	FREETMPS; LEAVE;
		585	}
		586	else if (enc->json.flags & F_ALLOW_BLESSED)
		587	encode_str (enc, "null", 4, 0);
		588	else
		589	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		590	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		591	}
		592	else if (enc->json.flags & F_ALLOW_BLESSED)
		593	encode_str (enc, "null", 4, 0);
		594	else
		595	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		596	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
369	}	597	}
370	}	598	}
		599	else if (svt == SVt_PVHV)
		600	encode_hv (enc, (HV *)sv);
		601	else if (svt == SVt_PVAV)
		602	encode_av (enc, (AV *)sv);
		603	else if (svt < SVt_PVAV)
		604	{
		605	STRLEN len = 0;
		606	char *pv = svt ? SvPV (sv, len) : 0;
371		607
372	--enc->indent; INDENT; encode_ch (enc, '}');	608	if (len == 1 && *pv == '1')
		609	encode_str (enc, "true", 4, 0);
		610	else if (len == 1 && *pv == '0')
		611	encode_str (enc, "false", 5, 0);
		612	else
		613	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		614	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		615	}
		616	else
		617	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		618	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
373	}	619	}
374		620
375	static void	621	static void
376	encode_sv (enc_t *enc, SV *sv)	622	encode_sv (enc_t *enc, SV *sv)
377	{	623	{
…		…
385	encode_str (enc, str, len, SvUTF8 (sv));	631	encode_str (enc, str, len, SvUTF8 (sv));
386	encode_ch (enc, '"');	632	encode_ch (enc, '"');
387	}	633	}
388	else if (SvNOKp (sv))	634	else if (SvNOKp (sv))
389	{	635	{
		636	// trust that perl will do the right thing w.r.t. JSON syntax.
390	need (enc, NV_DIG + 32);	637	need (enc, NV_DIG + 32);
391	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	638	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392	enc->cur += strlen (enc->cur);	639	enc->cur += strlen (enc->cur);
393	}	640	}
394	else if (SvIOKp (sv))	641	else if (SvIOKp (sv))
395	{	642	{
396	need (enc, 64);	643	// we assume we can always read an IV as a UV
		644	if (SvUV (sv) & ~(UV)0x7fff)
		645	{
		646	// large integer, use the (rather slow) snprintf way.
		647	need (enc, sizeof (UV) * 3);
397	enc->cur +=	648	enc->cur +=
398	SvIsUV(sv)	649	SvIsUV(sv)
399	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	650	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
400	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	651	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		652	}
		653	else
		654	{
		655	// optimise the "small number case"
		656	// code will likely be branchless and use only a single multiplication
		657	I32 i = SvIV (sv);
		658	U32 u;
		659	char digit, nz = 0;
		660
		661	need (enc, 6);
		662
		663	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		664	u = i < 0 ? -i : i;
		665
		666	// convert to 4.28 fixed-point representation
		667	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		668
		669	// now output digit by digit, each time masking out the integer part
		670	// and multiplying by 5 while moving the decimal point one to the right,
		671	// resulting in a net multiplication by 10.
		672	// we always write the digit to memory but conditionally increment
		673	// the pointer, to ease the usage of conditional move instructions.
		674	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		675	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		676	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		677	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		678	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		679	}
401	}	680	}
402	else if (SvROK (sv))	681	else if (SvROK (sv))
403	{	682	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))	683	else if (!SvOK (sv))
420	encode_str (enc, "null", 4, 0);	684	encode_str (enc, "null", 4, 0);
421	else	685	else
422	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	686	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
423	SvPV_nolen (sv), SvFLAGS (sv));	687	SvPV_nolen (sv), SvFLAGS (sv));
424	}	688	}
425		689
426	static SV *	690	static SV *
427	encode_json (SV *scalar, UV flags)	691	encode_json (SV *scalar, JSON *json)
428	{	692	{
		693	enc_t enc;
		694
429	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	695	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	696	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
431		697
432	enc_t enc;	698	enc.json = *json;
433	enc.flags = flags;
434	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	699	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435	enc.cur = SvPVX (enc.sv);	700	enc.cur = SvPVX (enc.sv);
436	enc.end = SvEND (enc.sv);	701	enc.end = SvEND (enc.sv);
437	enc.max_recurse = 0;
438	enc.indent = 0;	702	enc.indent = 0;
		703	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		704	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		705	: enc.json.flags & F_LATIN1 ? 0x000100UL
		706	: 0x10FFFFUL;
439		707
440	SvPOK_only (enc.sv);	708	SvPOK_only (enc.sv);
441	encode_sv (&enc, scalar);	709	encode_sv (&enc, scalar);
442		710
		711	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		712	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		713
443	if (!(flags & (F_ASCII | F_UTF8)))	714	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
444	SvUTF8_on (enc.sv);	715	SvUTF8_on (enc.sv);
445		716
446	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447
448	if (enc.flags & F_SHRINK)	717	if (enc.json.flags & F_SHRINK)
449	shrink (enc.sv);	718	shrink (enc.sv);
450		719
451	return enc.sv;	720	return enc.sv;
452	}	721	}
453		722
454	/////////////////////////////////////////////////////////////////////////////	723	/////////////////////////////////////////////////////////////////////////////
		724	// decoder
455		725
456	#define WS \	726	// structure used for decoding JSON
		727	typedef struct
		728	{
		729	char *cur; // current parser pointer
		730	char *end; // end of input string
		731	const char *err; // parse error, if != 0
		732	JSON json;
		733	U32 depth; // recursion depth
		734	U32 maxdepth; // recursion depth limit
		735	} dec_t;
		736
		737	INLINE void
		738	decode_comment (dec_t *dec)
		739	{
		740	// only '#'-style comments allowed a.t.m.
		741
		742	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		743	++dec->cur;
		744	}
		745
		746	INLINE void
		747	decode_ws (dec_t *dec)
		748	{
457	for (;;) \	749	for (;;)
458	{ \	750	{
459	char ch = *dec->cur; \	751	char ch = *dec->cur;
		752
460	if (ch > 0x20 \	753	if (ch > 0x20)
		754	{
		755	if (expect_false (ch == '#'))
		756	{
		757	if (dec->json.flags & F_RELAXED)
		758	decode_comment (dec);
		759	else
		760	break;
		761	}
		762	else
		763	break;
		764	}
461	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	765	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
462	break; \	766	break; // parse error, but let higher level handle it, gives better error messages
		767
463	++dec->cur; \	768	++dec->cur;
464	}	769	}
		770	}
465		771
466	#define ERR(reason) SB dec->err = reason; goto fail; SE	772	#define ERR(reason) SB dec->err = reason; goto fail; SE
		773
467	#define EXPECT_CH(ch) SB \	774	#define EXPECT_CH(ch) SB \
468	if (*dec->cur != ch) \	775	if (*dec->cur != ch) \
469	ERR (# ch " expected"); \	776	ERR (# ch " expected"); \
470	++dec->cur; \	777	++dec->cur; \
471	SE	778	SE
472		779
		780	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		781	#define DEC_DEC_DEPTH --dec->depth
		782
473	static SV *decode_sv (dec_t *dec);	783	static SV *decode_sv (dec_t *dec);
474		784
475	static signed char decode_hexdigit[256];	785	static signed char decode_hexdigit[256];
476		786
477	static UV	787	static UV
478	decode_4hex (dec_t *dec)	788	decode_4hex (dec_t *dec)
479	{	789	{
480	signed char d1, d2, d3, d4;	790	signed char d1, d2, d3, d4;
		791	unsigned char *cur = (unsigned char *)dec->cur;
481		792
482	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	793	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");	794	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]];	795	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");	796	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		797
491	dec->cur += 4;	798	dec->cur += 4;
492		799
493	return ((UV)d1) << 12	800	return ((UV)d1) << 12
494	| ((UV)d2) << 8	801	| ((UV)d2) << 8
…		…
497		804
498	fail:	805	fail:
499	return (UV)-1;	806	return (UV)-1;
500	}	807	}
501		808
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 *	809	static SV *
518	decode_str (dec_t *dec)	810	decode_str (dec_t *dec)
519	{	811	{
520	SV *sv = NEWSV (0,2);	812	SV *sv = 0;
521	int utf8 = 0;	813	int utf8 = 0;
522	char *cur = SvPVX (sv);	814	char *dec_cur = dec->cur;
523	char *end = SvEND (sv);
524		815
525	for (;;)	816	do
526	{	817	{
527	unsigned char ch = *(unsigned char *)dec->cur;	818	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		819	char *cur = buf;
528		820
529	if (ch == '"')	821	do
530	break;
531	else if (ch == '\\')
532	{	822	{
533	switch (*++dec->cur)	823	unsigned char ch = *(unsigned char *)dec_cur++;
		824
		825	if (expect_false (ch == '"'))
534	{	826	{
535	case '\\':	827	--dec_cur;
536	case '/':	828	break;
537	case '"': APPEND_CH (*dec->cur++); break;	829	}
538		830	else if (expect_false (ch == '\\'))
539	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	831	{
540	case 't': APPEND_CH ('\011'); ++dec->cur; break;	832	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	{	833	{
547	UV lo, hi;	834	case '\\':
548	++dec->cur;	835	case '/':
		836	case '"': *cur++ = *dec_cur++; break;
549		837
550	hi = decode_4hex (dec);	838	case 'b': ++dec_cur; *cur++ = '\010'; break;
551	if (hi == (UV)-1)	839	case 't': ++dec_cur; *cur++ = '\011'; break;
552	goto fail;	840	case 'n': ++dec_cur; *cur++ = '\012'; break;
		841	case 'f': ++dec_cur; *cur++ = '\014'; break;
		842	case 'r': ++dec_cur; *cur++ = '\015'; break;
553		843
554	// possibly a surrogate pair	844	case 'u':
555	if (hi >= 0xd800 && hi < 0xdc00)
556	{	845	{
557	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	846	UV lo, hi;
558	ERR ("missing low surrogate character in surrogate pair");	847	++dec_cur;
559		848
560	dec->cur += 2;	849	dec->cur = dec_cur;
561
562	lo = decode_4hex (dec);	850	hi = decode_4hex (dec);
		851	dec_cur = dec->cur;
563	if (lo == (UV)-1)	852	if (hi == (UV)-1)
564	goto fail;	853	goto fail;
565		854
		855	// possibly a surrogate pair
		856	if (hi >= 0xd800)
		857	if (hi < 0xdc00)
		858	{
		859	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		860	ERR ("missing low surrogate character in surrogate pair");
		861
		862	dec_cur += 2;
		863
		864	dec->cur = dec_cur;
		865	lo = decode_4hex (dec);
		866	dec_cur = dec->cur;
		867	if (lo == (UV)-1)
		868	goto fail;
		869
566	if (lo < 0xdc00 || lo >= 0xe000)	870	if (lo < 0xdc00 || lo >= 0xe000)
567	ERR ("surrogate pair expected");	871	ERR ("surrogate pair expected");
568		872
569	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	873	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		874	}
		875	else if (hi < 0xe000)
		876	ERR ("missing high surrogate character in surrogate pair");
		877
		878	if (hi >= 0x80)
		879	{
		880	utf8 = 1;
		881
		882	cur = encode_utf8 (cur, hi);
		883	}
		884	else
		885	*cur++ = hi;
570	}	886	}
571	else if (hi >= 0xdc00 && hi < 0xe000)
572	ERR ("missing high surrogate character in surrogate pair");
573
574	if (hi >= 0x80)
575	{	887	break;
576	utf8 = 1;
577		888
578	APPEND_GROW (4); // at most 4 bytes for 21 bits
579	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	}
581	else	889	default:
582	APPEND_CH (hi);	890	--dec_cur;
		891	ERR ("illegal backslash escape sequence in string");
583	}	892	}
584	break;	893	}
		894	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		895	*cur++ = ch;
		896	else if (ch >= 0x80)
		897	{
		898	STRLEN clen;
		899	UV uch;
585		900
586	default:
587	--dec->cur;	901	--dec_cur;
588	ERR ("illegal backslash escape sequence in string");	902
		903	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		904	if (clen == (STRLEN)-1)
		905	ERR ("malformed UTF-8 character in JSON string");
		906
		907	do
		908	*cur++ = *dec_cur++;
		909	while (--clen);
		910
		911	utf8 = 1;
		912	}
		913	else
		914	{
		915	--dec_cur;
		916
		917	if (!ch)
		918	ERR ("unexpected end of string while parsing JSON string");
		919	else
		920	ERR ("invalid character encountered while parsing JSON string");
589	}	921	}
590	}	922	}
591	else if (ch >= 0x20 && ch <= 0x7f)	923	while (cur < buf + SHORT_STRING_LEN);
592	APPEND_CH (*dec->cur++);	924
593	else if (ch >= 0x80)
594	{	925	{
595	STRLEN clen;	926	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		927
600	APPEND_GROW (clen);	928	if (sv)
601	do
602	{	929	{
603	*cur++ = *dec->cur++;	930	SvGROW (sv, SvCUR (sv) + len + 1);
		931	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		932	SvCUR_set (sv, SvCUR (sv) + len);
604	}	933	}
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	934	else
612	ERR ("invalid character encountered");	935	sv = newSVpvn (buf, len);
613	}	936	}
		937	}
		938	while (*dec_cur != '"');
614		939
615	++dec->cur;	940	++dec_cur;
616		941
617	SvCUR_set (sv, cur - SvPVX (sv));	942	if (sv)
618		943	{
619	SvPOK_only (sv);	944	SvPOK_only (sv);
620	*SvEND (sv) = 0;	945	*SvEND (sv) = 0;
621		946
622	if (utf8)	947	if (utf8)
623	SvUTF8_on (sv);	948	SvUTF8_on (sv);
		949	}
		950	else
		951	sv = newSVpvn ("", 0);
624		952
625	if (dec->flags & F_SHRINK)	953	dec->cur = dec_cur;
626	shrink (sv);
627
628	return sv;	954	return sv;
629		955
630	fail:	956	fail:
631	SvREFCNT_dec (sv);	957	dec->cur = dec_cur;
632	return 0;	958	return 0;
633	}	959	}
634		960
635	static SV *	961	static SV *
636	decode_num (dec_t *dec)	962	decode_num (dec_t *dec)
…		…
694	is_nv = 1;	1020	is_nv = 1;
695	}	1021	}
696		1022
697	if (!is_nv)	1023	if (!is_nv)
698	{	1024	{
699	UV uv;	1025	int len = dec->cur - start;
700	int numtype = grok_number (start, dec->cur - start, &uv);	1026
701	if (numtype & IS_NUMBER_IN_UV)	1027	// special case the rather common 1..5-digit-int case
702	if (numtype & IS_NUMBER_NEG)	1028	if (*start == '-')
		1029	switch (len)
703	{	1030	{
704	if (uv < (UV)IV_MIN)	1031	case 2: return newSViv (-( start [1] - '0' * 1));
705	return newSViv (-(IV)uv);	1032	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1033	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1034	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1035	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
706	}	1036	}
		1037	else
		1038	switch (len)
		1039	{
		1040	case 1: return newSViv ( start [0] - '0' * 1);
		1041	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1042	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1043	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1044	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1045	}
		1046
		1047	{
		1048	UV uv;
		1049	int numtype = grok_number (start, len, &uv);
		1050	if (numtype & IS_NUMBER_IN_UV)
		1051	if (numtype & IS_NUMBER_NEG)
		1052	{
		1053	if (uv < (UV)IV_MIN)
		1054	return newSViv (-(IV)uv);
		1055	}
707	else	1056	else
708	return newSVuv (uv);	1057	return newSVuv (uv);
709	}	1058	}
710		1059
		1060	len -= *start == '-' ? 1 : 0;
		1061
		1062	// does not fit into IV or UV, try NV
		1063	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1064	#if defined (LDBL_DIG)
		1065	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1066	#endif
		1067	)
		1068	// fits into NV without loss of precision
		1069	return newSVnv (Atof (start));
		1070
		1071	// everything else fails, convert it to a string
		1072	return newSVpvn (start, dec->cur - start);
		1073	}
		1074
		1075	// loss of precision here
711	return newSVnv (Atof (start));	1076	return newSVnv (Atof (start));
712		1077
713	fail:	1078	fail:
714	return 0;	1079	return 0;
715	}	1080	}
…		…
717	static SV *	1082	static SV *
718	decode_av (dec_t *dec)	1083	decode_av (dec_t *dec)
719	{	1084	{
720	AV *av = newAV ();	1085	AV *av = newAV ();
721		1086
722	WS;	1087	DEC_INC_DEPTH;
		1088	decode_ws (dec);
		1089
723	if (*dec->cur == ']')	1090	if (*dec->cur == ']')
724	++dec->cur;	1091	++dec->cur;
725	else	1092	else
726	for (;;)	1093	for (;;)
727	{	1094	{
…		…
731	if (!value)	1098	if (!value)
732	goto fail;	1099	goto fail;
733		1100
734	av_push (av, value);	1101	av_push (av, value);
735		1102
736	WS;	1103	decode_ws (dec);
737		1104
738	if (*dec->cur == ']')	1105	if (*dec->cur == ']')
739	{	1106	{
740	++dec->cur;	1107	++dec->cur;
741	break;	1108	break;
…		…
743		1110
744	if (*dec->cur != ',')	1111	if (*dec->cur != ',')
745	ERR (", or ] expected while parsing array");	1112	ERR (", or ] expected while parsing array");
746		1113
747	++dec->cur;	1114	++dec->cur;
		1115
		1116	decode_ws (dec);
		1117
		1118	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1119	{
		1120	++dec->cur;
		1121	break;
		1122	}
748	}	1123	}
749		1124
		1125	DEC_DEC_DEPTH;
750	return newRV_noinc ((SV *)av);	1126	return newRV_noinc ((SV *)av);
751		1127
752	fail:	1128	fail:
753	SvREFCNT_dec (av);	1129	SvREFCNT_dec (av);
		1130	DEC_DEC_DEPTH;
754	return 0;	1131	return 0;
755	}	1132	}
756		1133
757	static SV *	1134	static SV *
758	decode_hv (dec_t *dec)	1135	decode_hv (dec_t *dec)
759	{	1136	{
		1137	SV *sv;
760	HV *hv = newHV ();	1138	HV *hv = newHV ();
761		1139
762	WS;	1140	DEC_INC_DEPTH;
		1141	decode_ws (dec);
		1142
763	if (*dec->cur == '}')	1143	if (*dec->cur == '}')
764	++dec->cur;	1144	++dec->cur;
765	else	1145	else
766	for (;;)	1146	for (;;)
767	{	1147	{
768	SV *key, *value;
769
770	WS; EXPECT_CH ('"');	1148	EXPECT_CH ('"');
771		1149
772	key = decode_str (dec);	1150	// heuristic: assume that
773	if (!key)	1151	// a) decode_str + hv_store_ent are abysmally slow.
774	goto fail;	1152	// b) most hash keys are short, simple ascii text.
		1153	// => try to "fast-match" such strings to avoid
		1154	// the overhead of decode_str + hv_store_ent.
		1155	{
		1156	SV *value;
		1157	char *p = dec->cur;
		1158	char *e = p + 24; // only try up to 24 bytes
775		1159
776	WS; EXPECT_CH (':');	1160	for (;;)
777
778	value = decode_sv (dec);
779	if (!value)
780	{	1161	{
		1162	// the >= 0x80 is true on most architectures
		1163	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1164	{
		1165	// slow path, back up and use decode_str
		1166	SV *key = decode_str (dec);
		1167	if (!key)
		1168	goto fail;
		1169
		1170	decode_ws (dec); EXPECT_CH (':');
		1171
		1172	decode_ws (dec);
		1173	value = decode_sv (dec);
		1174	if (!value)
		1175	{
		1176	SvREFCNT_dec (key);
		1177	goto fail;
		1178	}
		1179
		1180	hv_store_ent (hv, key, value, 0);
781	SvREFCNT_dec (key);	1181	SvREFCNT_dec (key);
		1182
		1183	break;
		1184	}
		1185	else if (*p == '"')
		1186	{
		1187	// fast path, got a simple key
		1188	char *key = dec->cur;
		1189	int len = p - key;
		1190	dec->cur = p + 1;
		1191
		1192	decode_ws (dec); EXPECT_CH (':');
		1193
		1194	decode_ws (dec);
		1195	value = decode_sv (dec);
		1196	if (!value)
782	goto fail;	1197	goto fail;
		1198
		1199	hv_store (hv, key, len, value, 0);
		1200
		1201	break;
		1202	}
		1203
		1204	++p;
783	}	1205	}
784
785	//TODO: optimise
786	hv_store_ent (hv, key, value, 0);
787
788	WS;	1206	}
		1207
		1208	decode_ws (dec);
789		1209
790	if (*dec->cur == '}')	1210	if (*dec->cur == '}')
791	{	1211	{
792	++dec->cur;	1212	++dec->cur;
793	break;	1213	break;
…		…
795		1215
796	if (*dec->cur != ',')	1216	if (*dec->cur != ',')
797	ERR (", or } expected while parsing object/hash");	1217	ERR (", or } expected while parsing object/hash");
798		1218
799	++dec->cur;	1219	++dec->cur;
		1220
		1221	decode_ws (dec);
		1222
		1223	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1224	{
		1225	++dec->cur;
		1226	break;
		1227	}
800	}	1228	}
801		1229
		1230	DEC_DEC_DEPTH;
802	return newRV_noinc ((SV *)hv);	1231	sv = newRV_noinc ((SV *)hv);
		1232
		1233	// check filter callbacks
		1234	if (dec->json.flags & F_HOOK)
		1235	{
		1236	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1237	{
		1238	HE *cb, *he;
		1239
		1240	hv_iterinit (hv);
		1241	he = hv_iternext (hv);
		1242	hv_iterinit (hv);
		1243
		1244	// the next line creates a mortal sv each time its called.
		1245	// might want to optimise this for common cases.
		1246	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1247
		1248	if (cb)
		1249	{
		1250	dSP;
		1251	int count;
		1252
		1253	ENTER; SAVETMPS; PUSHMARK (SP);
		1254	XPUSHs (HeVAL (he));
		1255
		1256	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1257
		1258	if (count == 1)
		1259	{
		1260	sv = newSVsv (POPs);
		1261	FREETMPS; LEAVE;
		1262	return sv;
		1263	}
		1264
		1265	FREETMPS; LEAVE;
		1266	}
		1267	}
		1268
		1269	if (dec->json.cb_object)
		1270	{
		1271	dSP;
		1272	int count;
		1273
		1274	ENTER; SAVETMPS; PUSHMARK (SP);
		1275	XPUSHs (sv_2mortal (sv));
		1276
		1277	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1278
		1279	if (count == 1)
		1280	{
		1281	sv = newSVsv (POPs);
		1282	FREETMPS; LEAVE;
		1283	return sv;
		1284	}
		1285
		1286	SvREFCNT_inc (sv);
		1287	FREETMPS; LEAVE;
		1288	}
		1289	}
		1290
		1291	return sv;
803		1292
804	fail:	1293	fail:
805	SvREFCNT_dec (hv);	1294	SvREFCNT_dec (hv);
		1295	DEC_DEC_DEPTH;
806	return 0;	1296	return 0;
807	}	1297	}
808		1298
809	static SV *	1299	static SV *
810	decode_sv (dec_t *dec)	1300	decode_sv (dec_t *dec)
811	{	1301	{
812	WS;	1302	// the beauty of JSON: you need exactly one character lookahead
		1303	// to parse anything.
813	switch (*dec->cur)	1304	switch (*dec->cur)
814	{	1305	{
815	case '"': ++dec->cur; return decode_str (dec);	1306	case '"': ++dec->cur; return decode_str (dec);
816	case '[': ++dec->cur; return decode_av (dec);	1307	case '[': ++dec->cur; return decode_av (dec);
817	case '{': ++dec->cur; return decode_hv (dec);	1308	case '{': ++dec->cur; return decode_hv (dec);
…		…
823		1314
824	case 't':	1315	case 't':
825	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1316	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826	{	1317	{
827	dec->cur += 4;	1318	dec->cur += 4;
828	return newSViv (1);	1319	#if JSON_SLOW
		1320	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1321	#endif
		1322	return SvREFCNT_inc (json_true);
829	}	1323	}
830	else	1324	else
831	ERR ("'true' expected");	1325	ERR ("'true' expected");
832		1326
833	break;	1327	break;
834		1328
835	case 'f':	1329	case 'f':
836	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1330	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837	{	1331	{
838	dec->cur += 5;	1332	dec->cur += 5;
839	return newSViv (0);	1333	#if JSON_SLOW
		1334	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1335	#endif
		1336	return SvREFCNT_inc (json_false);
840	}	1337	}
841	else	1338	else
842	ERR ("'false' expected");	1339	ERR ("'false' expected");
843		1340
844	break;	1341	break;
…		…
853	ERR ("'null' expected");	1350	ERR ("'null' expected");
854		1351
855	break;	1352	break;
856		1353
857	default:	1354	default:
858	ERR ("malformed json string, neither array, object, number, string or atom");	1355	ERR ("malformed JSON string, neither array, object, number, string or atom");
859	break;	1356	break;
860	}	1357	}
861		1358
862	fail:	1359	fail:
863	return 0;	1360	return 0;
864	}	1361	}
865		1362
866	static SV *	1363	static SV *
867	decode_json (SV *string, UV flags)	1364	decode_json (SV *string, JSON *json, UV *offset_return)
868	{	1365	{
		1366	dec_t dec;
		1367	UV offset;
869	SV *sv;	1368	SV *sv;
870		1369
		1370	SvGETMAGIC (string);
		1371	SvUPGRADE (string, SVt_PV);
		1372
		1373	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1374	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1375	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1376
871	if (flags & F_UTF8)	1377	if (json->flags & F_UTF8)
872	sv_utf8_downgrade (string, 0);	1378	sv_utf8_downgrade (string, 0);
873	else	1379	else
874	sv_utf8_upgrade (string);	1380	sv_utf8_upgrade (string);
875		1381
876	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1382	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877		1383
878	dec_t dec;	1384	dec.json = *json;
879	dec.flags = flags;
880	dec.cur = SvPVX (string);	1385	dec.cur = SvPVX (string);
881	dec.end = SvEND (string);	1386	dec.end = SvEND (string);
882	dec.err = 0;	1387	dec.err = 0;
		1388	dec.depth = 0;
		1389	dec.maxdepth = DEC_DEPTH (dec.json.flags);
883		1390
		1391	if (dec.json.cb_object || dec.json.cb_sk_object)
		1392	dec.json.flags |= F_HOOK;
		1393
		1394	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1395
		1396	decode_ws (&dec);
884	sv = decode_sv (&dec);	1397	sv = decode_sv (&dec);
885		1398
		1399	if (!(offset_return || !sv))
		1400	{
		1401	// check for trailing garbage
		1402	decode_ws (&dec);
		1403
		1404	if (*dec.cur)
		1405	{
		1406	dec.err = "garbage after JSON object";
		1407	SvREFCNT_dec (sv);
		1408	sv = 0;
		1409	}
		1410	}
		1411
		1412	if (offset_return || !sv)
		1413	{
		1414	offset = dec.json.flags & F_UTF8
		1415	? dec.cur - SvPVX (string)
		1416	: utf8_distance (dec.cur, SvPVX (string));
		1417
		1418	if (offset_return)
		1419	*offset_return = offset;
		1420	}
		1421
886	if (!sv)	1422	if (!sv)
887	{	1423	{
888	IV offset = dec.flags & F_UTF8
889	? dec.cur - SvPVX (string)
890	: utf8_distance (dec.cur, SvPVX (string));
891	SV *uni = sv_newmortal ();	1424	SV *uni = sv_newmortal ();
892		1425
893	// horrible hack to silence warning inside pv_uni_display	1426	// horrible hack to silence warning inside pv_uni_display
894	COP cop = *PL_curcop;	1427	COP cop = *PL_curcop;
895	cop.cop_warnings = pWARN_NONE;	1428	cop.cop_warnings = pWARN_NONE;
…		…
897	SAVEVPTR (PL_curcop);	1430	SAVEVPTR (PL_curcop);
898	PL_curcop = &cop;	1431	PL_curcop = &cop;
899	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1432	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
900	LEAVE;	1433	LEAVE;
901		1434
902	croak ("%s, at character offset %d (%s)",	1435	croak ("%s, at character offset %d [\"%s\"]",
903	dec.err,	1436	dec.err,
904	(int)offset,	1437	(int)offset,
905	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1438	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906	}	1439	}
907		1440
908	sv = sv_2mortal (sv);	1441	sv = sv_2mortal (sv);
909		1442
910	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1443	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)");	1444	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
912		1445
913	return sv;	1446	return sv;
914	}	1447	}
915		1448
		1449	/////////////////////////////////////////////////////////////////////////////
		1450	// XS interface functions
		1451
916	MODULE = JSON::XS PACKAGE = JSON::XS	1452	MODULE = JSON::XS PACKAGE = JSON::XS
917		1453
918	BOOT:	1454	BOOT:
919	{	1455	{
920	int i;	1456	int i;
921		1457
922	memset (decode_hexdigit, 0xff, 256);
923	for (i = 10; i--; )	1458	for (i = 0; i < 256; ++i)
924	decode_hexdigit ['0' + i] = i;	1459	decode_hexdigit [i] =
		1460	i >= '0' && i <= '9' ? i - '0'
		1461	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1462	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1463	: -1;
925		1464
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);	1465	json_stash = gv_stashpv ("JSON::XS" , 1);
		1466	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1467
		1468	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1469	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
933	}	1470	}
934		1471
935	PROTOTYPES: DISABLE	1472	PROTOTYPES: DISABLE
936		1473
937	SV *new (char *dummy)	1474	void CLONE (...)
938	CODE:	1475	CODE:
939	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1476	json_stash = 0;
		1477	json_boolean_stash = 0;
		1478
		1479	void new (char *klass)
		1480	PPCODE:
		1481	{
		1482	SV *pv = NEWSV (0, sizeof (JSON));
		1483	SvPOK_only (pv);
		1484	Zero (SvPVX (pv), 1, JSON);
		1485	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1486	XPUSHs (sv_2mortal (sv_bless (
		1487	newRV_noinc (pv),
		1488	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1489	)));
		1490	}
		1491
		1492	void ascii (JSON *self, int enable = 1)
		1493	ALIAS:
		1494	ascii = F_ASCII
		1495	latin1 = F_LATIN1
		1496	utf8 = F_UTF8
		1497	indent = F_INDENT
		1498	canonical = F_CANONICAL
		1499	space_before = F_SPACE_BEFORE
		1500	space_after = F_SPACE_AFTER
		1501	pretty = F_PRETTY
		1502	allow_nonref = F_ALLOW_NONREF
		1503	shrink = F_SHRINK
		1504	allow_blessed = F_ALLOW_BLESSED
		1505	convert_blessed = F_CONV_BLESSED
		1506	relaxed = F_RELAXED
		1507	PPCODE:
		1508	{
		1509	if (enable)
		1510	self->flags |= ix;
		1511	else
		1512	self->flags &= ~ix;
		1513
		1514	XPUSHs (ST (0));
		1515	}
		1516
		1517	void get_ascii (JSON *self)
		1518	ALIAS:
		1519	get_ascii = F_ASCII
		1520	get_latin1 = F_LATIN1
		1521	get_utf8 = F_UTF8
		1522	get_indent = F_INDENT
		1523	get_canonical = F_CANONICAL
		1524	get_space_before = F_SPACE_BEFORE
		1525	get_space_after = F_SPACE_AFTER
		1526	get_allow_nonref = F_ALLOW_NONREF
		1527	get_shrink = F_SHRINK
		1528	get_allow_blessed = F_ALLOW_BLESSED
		1529	get_convert_blessed = F_CONV_BLESSED
		1530	get_relaxed = F_RELAXED
		1531	PPCODE:
		1532	XPUSHs (boolSV (self->flags & ix));
		1533
		1534	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1535	PPCODE:
		1536	{
		1537	UV log2 = 0;
		1538
		1539	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1540
		1541	while ((1UL << log2) < max_depth)
		1542	++log2;
		1543
		1544	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1545
		1546	XPUSHs (ST (0));
		1547	}
		1548
		1549	U32 get_max_depth (JSON *self)
		1550	CODE:
		1551	RETVAL = DEC_DEPTH (self->flags);
940	OUTPUT:	1552	OUTPUT:
941	RETVAL	1553	RETVAL
942		1554
943	SV *ascii (SV *self, int enable = 1)	1555	void max_size (JSON *self, UV max_size = 0)
944	ALIAS:	1556	PPCODE:
945	ascii = F_ASCII	1557	{
946	utf8 = F_UTF8	1558	UV log2 = 0;
947	indent = F_INDENT	1559
948	canonical = F_CANONICAL	1560	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
949	space_before = F_SPACE_BEFORE	1561	if (max_size == 1) max_size = 2;
950	space_after = F_SPACE_AFTER	1562
951	json_rpc = F_JSON_RPC	1563	while ((1UL << log2) < max_size)
952	pretty = F_PRETTY	1564	++log2;
953	allow_nonref = F_ALLOW_NONREF	1565
954	shrink = F_SHRINK	1566	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1567
		1568	XPUSHs (ST (0));
		1569	}
		1570
		1571	int get_max_size (JSON *self)
955	CODE:	1572	CODE:
956	{	1573	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:	1574	OUTPUT:
966	RETVAL	1575	RETVAL
967		1576
968	void encode (SV *self, SV *scalar)	1577	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
969	PPCODE:	1578	PPCODE:
970	XPUSHs (encode_json (scalar, *SvJSON (self)));	1579	{
		1580	SvREFCNT_dec (self->cb_object);
		1581	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
971		1582
972	void decode (SV *self, SV *jsonstr)	1583	XPUSHs (ST (0));
		1584	}
		1585
		1586	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
973	PPCODE:	1587	PPCODE:
		1588	{
		1589	if (!self->cb_sk_object)
		1590	self->cb_sk_object = newHV ();
		1591
		1592	if (SvOK (cb))
		1593	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1594	else
		1595	{
		1596	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1597
		1598	if (!HvKEYS (self->cb_sk_object))
		1599	{
		1600	SvREFCNT_dec (self->cb_sk_object);
		1601	self->cb_sk_object = 0;
		1602	}
		1603	}
		1604
		1605	XPUSHs (ST (0));
		1606	}
		1607
		1608	void encode (JSON *self, SV *scalar)
		1609	PPCODE:
		1610	XPUSHs (encode_json (scalar, self));
		1611
		1612	void decode (JSON *self, SV *jsonstr)
		1613	PPCODE:
974	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1614	XPUSHs (decode_json (jsonstr, self, 0));
		1615
		1616	void decode_prefix (JSON *self, SV *jsonstr)
		1617	PPCODE:
		1618	{
		1619	UV offset;
		1620	EXTEND (SP, 2);
		1621	PUSHs (decode_json (jsonstr, self, &offset));
		1622	PUSHs (sv_2mortal (newSVuv (offset)));
		1623	}
		1624
		1625	void DESTROY (JSON *self)
		1626	CODE:
		1627	SvREFCNT_dec (self->cb_sk_object);
		1628	SvREFCNT_dec (self->cb_object);
975		1629
976	PROTOTYPES: ENABLE	1630	PROTOTYPES: ENABLE
977		1631
978	void to_json (SV *scalar)	1632	void encode_json (SV *scalar)
979	PPCODE:	1633	PPCODE:
		1634	{
		1635	JSON json = { F_DEFAULT | F_UTF8 };
980	XPUSHs (encode_json (scalar, F_UTF8));	1636	XPUSHs (encode_json (scalar, &json));
		1637	}
981		1638
982	void from_json (SV *jsonstr)	1639	void decode_json (SV *jsonstr)
983	PPCODE:	1640	PPCODE:
		1641	{
		1642	JSON json = { F_DEFAULT | F_UTF8 };
984	XPUSHs (decode_json (jsonstr, F_UTF8));	1643	XPUSHs (decode_json (jsonstr, &json, 0));
		1644	}
985		1645

Diff Legend

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