ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.7 by root, Fri Mar 23 15:57:18 2007 UTC vs.
Revision 1.60 by root, Mon Aug 13 16:19:13 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_MAXDEPTH 0xf8000000UL
		33	#define S_MAXDEPTH 27
		34	#define F_MAXSIZE 0x01f00000UL
		35	#define S_MAXSIZE 20
		36	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		37
		38	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		39	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		40
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	41	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		43
22	#define INIT_SIZE 32 // initial scalar size to be allocated	44	#define INIT_SIZE 32 // initial scalar size to be allocated
		45	#define INDENT_STEP 3 // spaces per indentation level
		46
		47	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		48
24	#define SB do {	49	#define SB do {
25	#define SE } while (0)	50	#define SE } while (0)
26		51
27	static HV *json_stash;	52	#if __GNUC__ >= 3
		53	# define expect(expr,value) __builtin_expect ((expr),(value))
		54	# define inline inline
		55	#else
		56	# define expect(expr,value) (expr)
		57	# define inline static
		58	#endif
		59
		60	#define expect_false(expr) expect ((expr) != 0, 0)
		61	#define expect_true(expr) expect ((expr) != 0, 1)
		62
		63	#ifdef USE_ITHREADS
		64	# define JSON_SLOW 1
		65	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		66	#else
		67	# define JSON_SLOW 0
		68	# define JSON_STASH json_stash
		69	#endif
		70
		71	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		72	static SV *json_true, *json_false;
		73
		74	typedef struct {
		75	U32 flags;
		76	SV *cb_object;
		77	HV *cb_sk_object;
		78	} JSON;
		79
		80	/////////////////////////////////////////////////////////////////////////////
		81	// utility functions
		82
		83	inline void
		84	shrink (SV *sv)
		85	{
		86	sv_utf8_downgrade (sv, 1);
		87	if (SvLEN (sv) > SvCUR (sv) + 1)
		88	{
		89	#ifdef SvPV_shrink_to_cur
		90	SvPV_shrink_to_cur (sv);
		91	#elif defined (SvPV_renew)
		92	SvPV_renew (sv, SvCUR (sv) + 1);
		93	#endif
		94	}
		95	}
		96
		97	// decode an utf-8 character and return it, or (UV)-1 in
		98	// case of an error.
		99	// we special-case "safe" characters from U+80 .. U+7FF,
		100	// but use the very good perl function to parse anything else.
		101	// note that we never call this function for a ascii codepoints
		102	inline UV
		103	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		104	{
		105	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		106	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		107	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		108	{
		109	*clen = 2;
		110	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		111	}
		112	else
		113	{
		114	*clen = (STRLEN)-1;
		115	return (UV)-1;
		116	}
		117	}
		118
		119	/////////////////////////////////////////////////////////////////////////////
		120	// encoder
28		121
29	// structure used for encoding JSON	122	// structure used for encoding JSON
30	typedef struct	123	typedef struct
31	{	124	{
32	char *cur;	125	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	126	char *end; // SvEND (sv)
35	SV *sv;	127	SV *sv; // result scalar
36	UV flags;	128	JSON json;
37	int max_recurse;	129	U32 indent; // indentation level
38	int indent;	130	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	131	} enc_t;
40		132
41	// structure used for decoding JSON	133	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)	134	need (enc_t *enc, STRLEN len)
72	{	135	{
73	if (enc->cur + len >= enc->end)	136	if (expect_false (enc->cur + len >= enc->end))
74	{	137	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	138	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	139	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	140	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	142	}
80	}	143	}
81		144
82	static void	145	inline void
83	encode_ch (enc_t *enc, char ch)	146	encode_ch (enc_t *enc, char ch)
84	{	147	{
85	need (enc, 1);	148	need (enc, 1);
86	*enc->cur++ = ch;	149	*enc->cur++ = ch;
87	}	150	}
…		…
95		158
96	while (str < end)	159	while (str < end)
97	{	160	{
98	unsigned char ch = *(unsigned char *)str;	161	unsigned char ch = *(unsigned char *)str;
99		162
100	if (ch >= 0x20 && ch < 0x80) // most common case	163	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	164	{
102	if (ch == '"') // but with slow exceptions	165	if (expect_false (ch == '"')) // but with slow exceptions
103	{	166	{
104	need (enc, len += 1);	167	need (enc, len += 1);
105	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	169	*enc->cur++ = '"';
107	}	170	}
108	else if (ch == '\\')	171	else if (expect_false (ch == '\\'))
109	{	172	{
110	need (enc, len += 1);	173	need (enc, len += 1);
111	*enc->cur++ = '\\';	174	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	175	*enc->cur++ = '\\';
113	}	176	}
…		…
131	STRLEN clen;	194	STRLEN clen;
132	UV uch;	195	UV uch;
133		196
134	if (is_utf8)	197	if (is_utf8)
135	{	198	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	199	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	200	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	201	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	202	}
140	else	203	else
141	{	204	{
142	uch = ch;	205	uch = ch;
143	clen = 1;	206	clen = 1;
144	}	207	}
145		208
146	if (uch < 0x80 || enc->flags & F_ASCII)	209	if (uch > 0x10FFFFUL)
		210	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		211
		212	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
147	{	213	{
148	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
149	{	215	{
150	need (enc, len += 11);	216	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	217	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	218	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	219	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	220	enc->cur += 12;
155	}	221	}
156	else	222	else
157	{	223	{
158	static char hexdigit [16] = "0123456789abcdef";	224	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	231	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	232	}
167		233
168	str += clen;	234	str += clen;
169	}	235	}
		236	else if (enc->json.flags & F_LATIN1)
		237	{
		238	*enc->cur++ = uch;
		239	str += clen;
		240	}
170	else if (is_utf8)	241	else if (is_utf8)
171	{	242	{
172	need (enc, len += clen);	243	need (enc, len += clen);
173	do	244	do
174	{	245	{
…		…
176	}	247	}
177	while (--clen);	248	while (--clen);
178	}	249	}
179	else	250	else
180	{	251	{
181	need (enc, len += 10); // never more than 11 bytes needed	252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	254	++str;
184	}	255	}
185	}	256	}
186	}	257	}
…		…
188		259
189	--len;	260	--len;
190	}	261	}
191	}	262	}
192		263
193	#define INDENT SB \	264	inline void
		265	encode_indent (enc_t *enc)
		266	{
194	if (enc->flags & F_INDENT) \	267	if (enc->json.flags & F_INDENT)
195	{ \	268	{
196	int i_; \	269	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	270
198	for (i_ = enc->indent * 3; i_--; )\	271	need (enc, spaces);
		272	memset (enc->cur, ' ', spaces);
		273	enc->cur += spaces;
		274	}
		275	}
		276
		277	inline void
		278	encode_space (enc_t *enc)
		279	{
		280	need (enc, 1);
		281	encode_ch (enc, ' ');
		282	}
		283
		284	inline void
		285	encode_nl (enc_t *enc)
		286	{
		287	if (enc->json.flags & F_INDENT)
		288	{
		289	need (enc, 1);
199	encode_ch (enc, ' '); \	290	encode_ch (enc, '\n');
200	} \	291	}
201	SE	292	}
202		293
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	294	inline void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	295	encode_comma (enc_t *enc)
205	#define COMMA SB \	296	{
206	encode_ch (enc, ','); \	297	encode_ch (enc, ',');
		298
207	if (enc->flags & F_INDENT) \	299	if (enc->json.flags & F_INDENT)
208	NL; \	300	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	301	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	302	encode_space (enc);
211	SE	303	}
212		304
213	static void encode_sv (enc_t *enc, SV *sv);	305	static void encode_sv (enc_t *enc, SV *sv);
214		306
215	static void	307	static void
216	encode_av (enc_t *enc, AV *av)	308	encode_av (enc_t *enc, AV *av)
217	{	309	{
218	int i, len = av_len (av);	310	int i, len = av_len (av);
219		311
220	encode_ch (enc, '['); NL;	312	if (enc->indent >= enc->maxdepth)
		313	croak ("data structure too deep (hit recursion limit)");
		314
		315	encode_ch (enc, '['); encode_nl (enc);
221	++enc->indent;	316	++enc->indent;
222		317
223	for (i = 0; i <= len; ++i)	318	for (i = 0; i <= len; ++i)
224	{	319	{
225	INDENT;	320	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));	321
		322	encode_indent (enc);
		323
		324	if (svp)
		325	encode_sv (enc, *svp);
		326	else
		327	encode_str (enc, "null", 4, 0);
227		328
228	if (i < len)	329	if (i < len)
229	COMMA;	330	encode_comma (enc);
230	}	331	}
231		332
232	NL;	333	encode_nl (enc);
233		334
234	--enc->indent;	335	--enc->indent;
235	INDENT; encode_ch (enc, ']');	336	encode_indent (enc); encode_ch (enc, ']');
236	}	337	}
237		338
238	static void	339	static void
239	encode_he (enc_t *enc, HE *he)	340	encode_he (enc_t *enc, HE *he)
240	{	341	{
…		…
254	else	355	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	356	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		357
257	encode_ch (enc, '"');	358	encode_ch (enc, '"');
258		359
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	361	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));	363	encode_sv (enc, HeVAL (he));
263	}	364	}
264		365
265	// compare hash entries, used when all keys are bytestrings	366	// compare hash entries, used when all keys are bytestrings
266	static int	367	static int
…		…
272	HE *b = *(HE **)b_;	373	HE *b = *(HE **)b_;
273		374
274	STRLEN la = HeKLEN (a);	375	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	376	STRLEN lb = HeKLEN (b);
276		377
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	378	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	379	cmp = la - lb;
279		380
280	return cmp;	381	return cmp;
281	}	382	}
282		383
283	// compare hash entries, used when some keys are sv's or utf-x	384	// compare hash entries, used when some keys are sv's or utf-x
…		…
290	static void	391	static void
291	encode_hv (enc_t *enc, HV *hv)	392	encode_hv (enc_t *enc, HV *hv)
292	{	393	{
293	int count, i;	394	int count, i;
294		395
		396	if (enc->indent >= enc->maxdepth)
		397	croak ("data structure too deep (hit recursion limit)");
		398
295	encode_ch (enc, '{'); NL; ++enc->indent;	399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
296		400
297	if ((count = hv_iterinit (hv)))	401	if ((count = hv_iterinit (hv)))
298	{	402	{
299	// for canonical output we have to sort by keys first	403	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	404	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	405	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	406	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	407	if (enc->json.flags & F_CANONICAL)
304	{	408	{
305	HE *he, *hes [count];
306	int fast = 1;	409	int fast = 1;
		410	HE *he;
		411	#if defined(__BORLANDC__) || defined(_MSC_VER)
		412	HE **hes = _alloca (count * sizeof (HE));
		413	#else
		414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		415	#endif
307		416
308	i = 0;	417	i = 0;
309	while ((he = hv_iternext (hv)))	418	while ((he = hv_iternext (hv)))
310	{	419	{
311	hes [i++] = he;	420	hes [i++] = he;
…		…
317		426
318	if (fast)	427	if (fast)
319	qsort (hes, count, sizeof (HE *), he_cmp_fast);	428	qsort (hes, count, sizeof (HE *), he_cmp_fast);
320	else	429	else
321	{	430	{
322	// hack to disable "use bytes"	431	// hack to forcefully disable "use bytes"
323	COP *oldcop = PL_curcop, cop;	432	COP cop = *PL_curcop;
324	cop.op_private = 0;	433	cop.op_private = 0;
		434
		435	ENTER;
		436	SAVETMPS;
		437
		438	SAVEVPTR (PL_curcop);
325	PL_curcop = &cop;	439	PL_curcop = &cop;
326		440
327	SAVETMPS;
328	qsort (hes, count, sizeof (HE *), he_cmp_slow);	441	qsort (hes, count, sizeof (HE *), he_cmp_slow);
		442
329	FREETMPS;	443	FREETMPS;
330		444	LEAVE;
331	PL_curcop = oldcop;
332	}	445	}
333		446
334	for (i = 0; i < count; ++i)	447	for (i = 0; i < count; ++i)
335	{	448	{
336	INDENT;	449	encode_indent (enc);
337	encode_he (enc, hes [i]);	450	encode_he (enc, hes [i]);
338		451
339	if (i < count - 1)	452	if (i < count - 1)
340	COMMA;	453	encode_comma (enc);
341	}
342
343	NL;	454	}
		455
		456	encode_nl (enc);
344	}	457	}
345	else	458	else
346	{	459	{
347	SV *sv;
348	HE *he = hv_iternext (hv);	460	HE *he = hv_iternext (hv);
349		461
350	for (;;)	462	for (;;)
351	{	463	{
352	INDENT;	464	encode_indent (enc);
353	encode_he (enc, he);	465	encode_he (enc, he);
354		466
355	if (!(he = hv_iternext (hv)))	467	if (!(he = hv_iternext (hv)))
356	break;	468	break;
357		469
358	COMMA;	470	encode_comma (enc);
359	}
360
361	NL;	471	}
		472
		473	encode_nl (enc);
362	}	474	}
363	}	475	}
364		476
365	--enc->indent; INDENT; encode_ch (enc, '}');	477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		478	}
		479
		480	// encode objects, arrays and special \0=false and \1=true values.
		481	static void
		482	encode_rv (enc_t *enc, SV *sv)
		483	{
		484	svtype svt;
		485
		486	SvGETMAGIC (sv);
		487	svt = SvTYPE (sv);
		488
		489	if (expect_false (SvOBJECT (sv)))
		490	{
		491	HV *stash = !JSON_SLOW || json_boolean_stash
		492	? json_boolean_stash
		493	: gv_stashpv ("JSON::XS::Boolean", 1);
		494
		495	if (SvSTASH (sv) == stash)
		496	{
		497	if (SvIV (sv))
		498	encode_str (enc, "true", 4, 0);
		499	else
		500	encode_str (enc, "false", 5, 0);
		501	}
		502	else
		503	{
		504	#if 0
		505	if (0 && sv_derived_from (rv, "JSON::Literal"))
		506	{
		507	// not yet
		508	}
		509	#endif
		510	if (enc->json.flags & F_CONV_BLESSED)
		511	{
		512	// we re-bless the reference to get overload and other niceties right
		513	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		514
		515	if (to_json)
		516	{
		517	dSP;
		518
		519	ENTER; SAVETMPS; PUSHMARK (SP);
		520	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		521
		522	// calling with G_SCALAR ensures that we always get a 1 return value
		523	PUTBACK;
		524	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		525	SPAGAIN;
		526
		527	// catch this surprisingly common error
		528	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		529	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		530
		531	sv = POPs;
		532	PUTBACK;
		533
		534	encode_sv (enc, sv);
		535
		536	FREETMPS; LEAVE;
		537	}
		538	else if (enc->json.flags & F_ALLOW_BLESSED)
		539	encode_str (enc, "null", 4, 0);
		540	else
		541	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else if (enc->json.flags & F_ALLOW_BLESSED)
		545	encode_str (enc, "null", 4, 0);
		546	else
		547	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		548	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		549	}
		550	}
		551	else if (svt == SVt_PVHV)
		552	encode_hv (enc, (HV *)sv);
		553	else if (svt == SVt_PVAV)
		554	encode_av (enc, (AV *)sv);
		555	else if (svt < SVt_PVAV)
		556	{
		557	STRLEN len = 0;
		558	char *pv = svt ? SvPV (sv, len) : 0;
		559
		560	if (len == 1 && *pv == '1')
		561	encode_str (enc, "true", 4, 0);
		562	else if (len == 1 && *pv == '0')
		563	encode_str (enc, "false", 5, 0);
		564	else
		565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		567	}
		568	else
		569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	571	}
367		572
368	static void	573	static void
369	encode_sv (enc_t *enc, SV *sv)	574	encode_sv (enc_t *enc, SV *sv)
370	{	575	{
…		…
378	encode_str (enc, str, len, SvUTF8 (sv));	583	encode_str (enc, str, len, SvUTF8 (sv));
379	encode_ch (enc, '"');	584	encode_ch (enc, '"');
380	}	585	}
381	else if (SvNOKp (sv))	586	else if (SvNOKp (sv))
382	{	587	{
		588	// trust that perl will do the right thing w.r.t. JSON syntax.
383	need (enc, NV_DIG + 32);	589	need (enc, NV_DIG + 32);
384	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
385	enc->cur += strlen (enc->cur);	591	enc->cur += strlen (enc->cur);
386	}	592	}
387	else if (SvIOKp (sv))	593	else if (SvIOKp (sv))
388	{	594	{
389	need (enc, 64);	595	// we assume we can always read an IV as a UV
		596	if (SvUV (sv) & ~(UV)0x7fff)
		597	{
		598	// large integer, use the (rather slow) snprintf way.
		599	need (enc, sizeof (UV) * 3);
390	enc->cur +=	600	enc->cur +=
391	SvIsUV(sv)	601	SvIsUV(sv)
392	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	602	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
393	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	603	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		604	}
		605	else
		606	{
		607	// optimise the "small number case"
		608	// code will likely be branchless and use only a single multiplication
		609	I32 i = SvIV (sv);
		610	U32 u;
		611	char digit, nz = 0;
		612
		613	need (enc, 6);
		614
		615	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		616	u = i < 0 ? -i : i;
		617
		618	// convert to 4.28 fixed-point representation
		619	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		620
		621	// now output digit by digit, each time masking out the integer part
		622	// and multiplying by 5 while moving the decimal point one to the right,
		623	// resulting in a net multiplication by 10.
		624	// we always write the digit to memory but conditionally increment
		625	// the pointer, to ease the usage of conditional move instructions.
		626	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		627	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		628	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		629	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		630	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		631	}
394	}	632	}
395	else if (SvROK (sv))	633	else if (SvROK (sv))
396	{	634	encode_rv (enc, SvRV (sv));
397	if (!--enc->max_recurse)
398	croak ("data structure too deep (hit recursion limit)");
399
400	sv = SvRV (sv);
401
402	switch (SvTYPE (sv))
403	{
404	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
405	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
406
407	default:
408	croak ("JSON can only represent references to arrays or hashes");
409	}
410	}
411	else if (!SvOK (sv))	635	else if (!SvOK (sv))
412	encode_str (enc, "null", 4, 0);	636	encode_str (enc, "null", 4, 0);
413	else	637	else
414	croak ("encountered perl type that JSON cannot handle");	638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		639	SvPV_nolen (sv), SvFLAGS (sv));
415	}	640	}
416		641
417	static SV *	642	static SV *
418	encode_json (SV *scalar, UV flags)	643	encode_json (SV *scalar, JSON *json)
419	{	644	{
420	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
421	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
422
423	enc_t enc;	645	enc_t enc;
424	enc.flags = flags;	646
		647	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		648	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		649
		650	enc.json = *json;
425	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
426	enc.cur = SvPVX (enc.sv);	652	enc.cur = SvPVX (enc.sv);
427	enc.end = SvEND (enc.sv);	653	enc.end = SvEND (enc.sv);
428	enc.max_recurse = 0;
429	enc.indent = 0;	654	enc.indent = 0;
		655	enc.maxdepth = DEC_DEPTH (enc.json.flags);
430		656
431	SvPOK_only (enc.sv);	657	SvPOK_only (enc.sv);
432	encode_sv (&enc, scalar);	658	encode_sv (&enc, scalar);
433		659
		660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		661	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		662
434	if (!(flags & (F_ASCII | F_UTF8)))	663	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
435	SvUTF8_on (enc.sv);	664	SvUTF8_on (enc.sv);
436		665
437	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
438
439	if (enc.flags & F_SHRINK)	666	if (enc.json.flags & F_SHRINK)
440	shrink (enc.sv);	667	shrink (enc.sv);
441		668
442	return enc.sv;	669	return enc.sv;
443	}	670	}
444		671
445	/////////////////////////////////////////////////////////////////////////////	672	/////////////////////////////////////////////////////////////////////////////
		673	// decoder
446		674
447	#define WS \	675	// structure used for decoding JSON
		676	typedef struct
		677	{
		678	char *cur; // current parser pointer
		679	char *end; // end of input string
		680	const char *err; // parse error, if != 0
		681	JSON json;
		682	U32 depth; // recursion depth
		683	U32 maxdepth; // recursion depth limit
		684	} dec_t;
		685
		686	inline void
		687	decode_ws (dec_t *dec)
		688	{
448	for (;;) \	689	for (;;)
449	{ \	690	{
450	char ch = *dec->cur; \	691	char ch = *dec->cur;
		692
451	if (ch > 0x20 \	693	if (ch > 0x20
452	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	694	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
453	break; \	695	break;
		696
454	++dec->cur; \	697	++dec->cur;
455	}	698	}
		699	}
456		700
457	#define ERR(reason) SB dec->err = reason; goto fail; SE	701	#define ERR(reason) SB dec->err = reason; goto fail; SE
		702
458	#define EXPECT_CH(ch) SB \	703	#define EXPECT_CH(ch) SB \
459	if (*dec->cur != ch) \	704	if (*dec->cur != ch) \
460	ERR (# ch " expected"); \	705	ERR (# ch " expected"); \
461	++dec->cur; \	706	++dec->cur; \
462	SE	707	SE
463		708
		709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		710	#define DEC_DEC_DEPTH --dec->depth
		711
464	static SV *decode_sv (dec_t *dec);	712	static SV *decode_sv (dec_t *dec);
465		713
466	static signed char decode_hexdigit[256];	714	static signed char decode_hexdigit[256];
467		715
468	static UV	716	static UV
469	decode_4hex (dec_t *dec)	717	decode_4hex (dec_t *dec)
470	{	718	{
471	signed char d1, d2, d3, d4;	719	signed char d1, d2, d3, d4;
		720	unsigned char *cur = (unsigned char *)dec->cur;
472		721
473	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	722	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
474	if (d1 < 0) ERR ("four hexadecimal digits expected");	723	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
475	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	724	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
476	if (d2 < 0) ERR ("four hexadecimal digits expected");	725	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
477	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
478	if (d3 < 0) ERR ("four hexadecimal digits expected");
479	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
480	if (d4 < 0) ERR ("four hexadecimal digits expected");
481		726
482	dec->cur += 4;	727	dec->cur += 4;
483		728
484	return ((UV)d1) << 12	729	return ((UV)d1) << 12
485	| ((UV)d2) << 8	730	| ((UV)d2) << 8
…		…
488		733
489	fail:	734	fail:
490	return (UV)-1;	735	return (UV)-1;
491	}	736	}
492		737
493	#define APPEND_GROW(n) SB \
494	if (cur + (n) >= end) \
495	{ \
496	STRLEN ofs = cur - SvPVX (sv); \
497	SvGROW (sv, ofs + (n) + 1); \
498	cur = SvPVX (sv) + ofs; \
499	end = SvEND (sv); \
500	} \
501	SE
502
503	#define APPEND_CH(ch) SB \
504	APPEND_GROW (1); \
505	*cur++ = (ch); \
506	SE
507
508	static SV *	738	static SV *
509	decode_str (dec_t *dec)	739	decode_str (dec_t *dec)
510	{	740	{
511	SV *sv = NEWSV (0,2);	741	SV *sv = 0;
512	int utf8 = 0;	742	int utf8 = 0;
513	char *cur = SvPVX (sv);	743	char *dec_cur = dec->cur;
514	char *end = SvEND (sv);
515		744
516	for (;;)	745	do
517	{	746	{
518	unsigned char ch = *(unsigned char *)dec->cur;	747	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		748	char *cur = buf;
519		749
520	if (ch == '"')	750	do
521	break;
522	else if (ch == '\\')
523	{	751	{
524	switch (*++dec->cur)	752	unsigned char ch = *(unsigned char *)dec_cur++;
		753
		754	if (expect_false (ch == '"'))
		755	{
		756	--dec_cur;
		757	break;
525	{	758	}
526	case '\\':	759	else if (expect_false (ch == '\\'))
527	case '/':	760	{
528	case '"': APPEND_CH (*dec->cur++); break;	761	switch (*dec_cur)
529
530	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
531	case 't': APPEND_CH ('\011'); ++dec->cur; break;
532	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
533	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
534	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
535
536	case 'u':
537	{	762	{
538	UV lo, hi;	763	case '\\':
539	++dec->cur;	764	case '/':
		765	case '"': *cur++ = *dec_cur++; break;
540		766
541	hi = decode_4hex (dec);	767	case 'b': ++dec_cur; *cur++ = '\010'; break;
542	if (hi == (UV)-1)	768	case 't': ++dec_cur; *cur++ = '\011'; break;
543	goto fail;	769	case 'n': ++dec_cur; *cur++ = '\012'; break;
		770	case 'f': ++dec_cur; *cur++ = '\014'; break;
		771	case 'r': ++dec_cur; *cur++ = '\015'; break;
544		772
545	// possibly a surrogate pair	773	case 'u':
546	if (hi >= 0xd800 && hi < 0xdc00)
547	{	774	{
548	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	775	UV lo, hi;
549	ERR ("missing low surrogate character in surrogate pair");	776	++dec_cur;
550		777
551	dec->cur += 2;	778	dec->cur = dec_cur;
552
553	lo = decode_4hex (dec);	779	hi = decode_4hex (dec);
		780	dec_cur = dec->cur;
554	if (lo == (UV)-1)	781	if (hi == (UV)-1)
555	goto fail;	782	goto fail;
556		783
		784	// possibly a surrogate pair
		785	if (hi >= 0xd800)
		786	if (hi < 0xdc00)
		787	{
		788	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		789	ERR ("missing low surrogate character in surrogate pair");
		790
		791	dec_cur += 2;
		792
		793	dec->cur = dec_cur;
		794	lo = decode_4hex (dec);
		795	dec_cur = dec->cur;
		796	if (lo == (UV)-1)
		797	goto fail;
		798
557	if (lo < 0xdc00 || lo >= 0xe000)	799	if (lo < 0xdc00 || lo >= 0xe000)
558	ERR ("surrogate pair expected");	800	ERR ("surrogate pair expected");
559		801
560	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	802	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		803	}
		804	else if (hi < 0xe000)
		805	ERR ("missing high surrogate character in surrogate pair");
		806
		807	if (hi >= 0x80)
		808	{
		809	utf8 = 1;
		810
		811	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		812	}
		813	else
		814	*cur++ = hi;
561	}	815	}
562	else if (hi >= 0xdc00 && hi < 0xe000)
563	ERR ("missing high surrogate character in surrogate pair");
564
565	if (hi >= 0x80)
566	{	816	break;
567	utf8 = 1;
568		817
569	APPEND_GROW (4); // at most 4 bytes for 21 bits
570	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
571	}
572	else	818	default:
573	APPEND_CH (hi);	819	--dec_cur;
		820	ERR ("illegal backslash escape sequence in string");
574	}	821	}
575	break;	822	}
		823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		824	*cur++ = ch;
		825	else if (ch >= 0x80)
		826	{
		827	STRLEN clen;
		828	UV uch;
576		829
577	default:
578	--dec->cur;	830	--dec_cur;
579	ERR ("illegal backslash escape sequence in string");	831
		832	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		833	if (clen == (STRLEN)-1)
		834	ERR ("malformed UTF-8 character in JSON string");
		835
		836	do
		837	*cur++ = *dec_cur++;
		838	while (--clen);
		839
		840	utf8 = 1;
		841	}
		842	else
		843	{
		844	--dec_cur;
		845
		846	if (!ch)
		847	ERR ("unexpected end of string while parsing JSON string");
		848	else
		849	ERR ("invalid character encountered while parsing JSON string");
580	}	850	}
581	}	851	}
582	else if (ch >= 0x20 && ch <= 0x7f)	852	while (cur < buf + SHORT_STRING_LEN);
583	APPEND_CH (*dec->cur++);	853
584	else if (ch >= 0x80)
585	{	854	{
586	STRLEN clen;	855	STRLEN len = cur - buf;
587	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
588	if (clen == (STRLEN)-1)
589	ERR ("malformed UTF-8 character in JSON string");
590		856
591	APPEND_GROW (clen);	857	if (sv)
592	do
593	{	858	{
594	*cur++ = *dec->cur++;	859	SvGROW (sv, SvCUR (sv) + len + 1);
		860	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		861	SvCUR_set (sv, SvCUR (sv) + len);
595	}	862	}
596	while (--clen);
597
598	utf8 = 1;
599	}
600	else if (dec->cur == dec->end)
601	ERR ("unexpected end of string while parsing json string");
602	else	863	else
603	ERR ("invalid character encountered");	864	sv = newSVpvn (buf, len);
604	}	865	}
		866	}
		867	while (*dec_cur != '"');
605		868
606	++dec->cur;	869	++dec_cur;
607		870
608	SvCUR_set (sv, cur - SvPVX (sv));	871	if (sv)
609		872	{
610	SvPOK_only (sv);	873	SvPOK_only (sv);
611	*SvEND (sv) = 0;	874	*SvEND (sv) = 0;
612		875
613	if (utf8)	876	if (utf8)
614	SvUTF8_on (sv);	877	SvUTF8_on (sv);
		878	}
		879	else
		880	sv = newSVpvn ("", 0);
615		881
616	if (dec->flags & F_SHRINK)	882	dec->cur = dec_cur;
617	shrink (sv);
618
619	return sv;	883	return sv;
620		884
621	fail:	885	fail:
622	SvREFCNT_dec (sv);	886	dec->cur = dec_cur;
623	return 0;	887	return 0;
624	}	888	}
625		889
626	static SV *	890	static SV *
627	decode_num (dec_t *dec)	891	decode_num (dec_t *dec)
…		…
685	is_nv = 1;	949	is_nv = 1;
686	}	950	}
687		951
688	if (!is_nv)	952	if (!is_nv)
689	{	953	{
690	UV uv;	954	int len = dec->cur - start;
691	int numtype = grok_number (start, dec->cur - start, &uv);	955
692	if (numtype & IS_NUMBER_IN_UV)	956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
693	if (numtype & IS_NUMBER_NEG)	957	if (*start == '-')
		958	switch (len)
694	{	959	{
695	if (uv < (UV)IV_MIN)	960	case 2: return newSViv (-( start [1] - '0' * 1));
696	return newSViv (-(IV)uv);	961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		962	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		963	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
697	}	964	}
		965	else
		966	switch (len)
		967	{
		968	case 1: return newSViv ( start [0] - '0' * 1);
		969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		970	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		971	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		972	}
		973
		974	{
		975	UV uv;
		976	int numtype = grok_number (start, len, &uv);
		977	if (numtype & IS_NUMBER_IN_UV)
		978	if (numtype & IS_NUMBER_NEG)
		979	{
		980	if (uv < (UV)IV_MIN)
		981	return newSViv (-(IV)uv);
		982	}
698	else	983	else
699	return newSVuv (uv);	984	return newSVuv (uv);
700	}	985	}
701		986
		987	len -= *start == '-' ? 1 : 0;
		988
		989	// does not fit into IV or UV, try NV
		990	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		991	#if defined (LDBL_DIG)
		992	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		993	#endif
		994	)
		995	// fits into NV without loss of precision
		996	return newSVnv (Atof (start));
		997
		998	// everything else fails, convert it to a string
		999	return newSVpvn (start, dec->cur - start);
		1000	}
		1001
		1002	// loss of precision here
702	return newSVnv (Atof (start));	1003	return newSVnv (Atof (start));
703		1004
704	fail:	1005	fail:
705	return 0;	1006	return 0;
706	}	1007	}
…		…
708	static SV *	1009	static SV *
709	decode_av (dec_t *dec)	1010	decode_av (dec_t *dec)
710	{	1011	{
711	AV *av = newAV ();	1012	AV *av = newAV ();
712		1013
713	WS;	1014	DEC_INC_DEPTH;
		1015	decode_ws (dec);
		1016
714	if (*dec->cur == ']')	1017	if (*dec->cur == ']')
715	++dec->cur;	1018	++dec->cur;
716	else	1019	else
717	for (;;)	1020	for (;;)
718	{	1021	{
…		…
722	if (!value)	1025	if (!value)
723	goto fail;	1026	goto fail;
724		1027
725	av_push (av, value);	1028	av_push (av, value);
726		1029
727	WS;	1030	decode_ws (dec);
728		1031
729	if (*dec->cur == ']')	1032	if (*dec->cur == ']')
730	{	1033	{
731	++dec->cur;	1034	++dec->cur;
732	break;	1035	break;
…		…
736	ERR (", or ] expected while parsing array");	1039	ERR (", or ] expected while parsing array");
737		1040
738	++dec->cur;	1041	++dec->cur;
739	}	1042	}
740		1043
		1044	DEC_DEC_DEPTH;
741	return newRV_noinc ((SV *)av);	1045	return newRV_noinc ((SV *)av);
742		1046
743	fail:	1047	fail:
744	SvREFCNT_dec (av);	1048	SvREFCNT_dec (av);
		1049	DEC_DEC_DEPTH;
745	return 0;	1050	return 0;
746	}	1051	}
747		1052
748	static SV *	1053	static SV *
749	decode_hv (dec_t *dec)	1054	decode_hv (dec_t *dec)
750	{	1055	{
		1056	SV *sv;
751	HV *hv = newHV ();	1057	HV *hv = newHV ();
752		1058
753	WS;	1059	DEC_INC_DEPTH;
		1060	decode_ws (dec);
		1061
754	if (*dec->cur == '}')	1062	if (*dec->cur == '}')
755	++dec->cur;	1063	++dec->cur;
756	else	1064	else
757	for (;;)	1065	for (;;)
758	{	1066	{
		1067	decode_ws (dec); EXPECT_CH ('"');
		1068
		1069	// heuristic: assume that
		1070	// a) decode_str + hv_store_ent are abysmally slow.
		1071	// b) most hash keys are short, simple ascii text.
		1072	// => try to "fast-match" such strings to avoid
		1073	// the overhead of decode_str + hv_store_ent.
		1074	{
759	SV *key, *value;	1075	SV *value;
		1076	char *p = dec->cur;
		1077	char *e = p + 24; // only try up to 24 bytes
760		1078
761	WS; EXPECT_CH ('"');	1079	for (;;)
762
763	key = decode_str (dec);
764	if (!key)
765	goto fail;
766
767	WS; EXPECT_CH (':');
768
769	value = decode_sv (dec);
770	if (!value)
771	{	1080	{
		1081	// the >= 0x80 is true on most architectures
		1082	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1083	{
		1084	// slow path, back up and use decode_str
		1085	SV *key = decode_str (dec);
		1086	if (!key)
		1087	goto fail;
		1088
		1089	decode_ws (dec); EXPECT_CH (':');
		1090
		1091	value = decode_sv (dec);
		1092	if (!value)
		1093	{
		1094	SvREFCNT_dec (key);
		1095	goto fail;
		1096	}
		1097
		1098	hv_store_ent (hv, key, value, 0);
772	SvREFCNT_dec (key);	1099	SvREFCNT_dec (key);
		1100
		1101	break;
		1102	}
		1103	else if (*p == '"')
		1104	{
		1105	// fast path, got a simple key
		1106	char *key = dec->cur;
		1107	int len = p - key;
		1108	dec->cur = p + 1;
		1109
		1110	decode_ws (dec); EXPECT_CH (':');
		1111
		1112	value = decode_sv (dec);
		1113	if (!value)
773	goto fail;	1114	goto fail;
		1115
		1116	hv_store (hv, key, len, value, 0);
		1117
		1118	break;
		1119	}
		1120
		1121	++p;
774	}	1122	}
775
776	//TODO: optimise
777	hv_store_ent (hv, key, value, 0);
778
779	WS;	1123	}
		1124
		1125	decode_ws (dec);
780		1126
781	if (*dec->cur == '}')	1127	if (*dec->cur == '}')
782	{	1128	{
783	++dec->cur;	1129	++dec->cur;
784	break;	1130	break;
…		…
788	ERR (", or } expected while parsing object/hash");	1134	ERR (", or } expected while parsing object/hash");
789		1135
790	++dec->cur;	1136	++dec->cur;
791	}	1137	}
792		1138
		1139	DEC_DEC_DEPTH;
793	return newRV_noinc ((SV *)hv);	1140	sv = newRV_noinc ((SV *)hv);
		1141
		1142	// check filter callbacks
		1143	if (dec->json.flags & F_HOOK)
		1144	{
		1145	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1146	{
		1147	HE *cb, *he;
		1148
		1149	hv_iterinit (hv);
		1150	he = hv_iternext (hv);
		1151	hv_iterinit (hv);
		1152
		1153	// the next line creates a mortal sv each time its called.
		1154	// might want to optimise this for common cases.
		1155	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1156
		1157	if (cb)
		1158	{
		1159	dSP;
		1160	int count;
		1161
		1162	ENTER; SAVETMPS; PUSHMARK (SP);
		1163	XPUSHs (HeVAL (he));
		1164
		1165	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1166
		1167	if (count == 1)
		1168	{
		1169	sv = newSVsv (POPs);
		1170	FREETMPS; LEAVE;
		1171	return sv;
		1172	}
		1173
		1174	FREETMPS; LEAVE;
		1175	}
		1176	}
		1177
		1178	if (dec->json.cb_object)
		1179	{
		1180	dSP;
		1181	int count;
		1182
		1183	ENTER; SAVETMPS; PUSHMARK (SP);
		1184	XPUSHs (sv_2mortal (sv));
		1185
		1186	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1187
		1188	if (count == 1)
		1189	{
		1190	sv = newSVsv (POPs);
		1191	FREETMPS; LEAVE;
		1192	return sv;
		1193	}
		1194
		1195	SvREFCNT_inc (sv);
		1196	FREETMPS; LEAVE;
		1197	}
		1198	}
		1199
		1200	return sv;
794		1201
795	fail:	1202	fail:
796	SvREFCNT_dec (hv);	1203	SvREFCNT_dec (hv);
		1204	DEC_DEC_DEPTH;
797	return 0;	1205	return 0;
798	}	1206	}
799		1207
800	static SV *	1208	static SV *
801	decode_sv (dec_t *dec)	1209	decode_sv (dec_t *dec)
802	{	1210	{
803	WS;	1211	decode_ws (dec);
		1212
		1213	// the beauty of JSON: you need exactly one character lookahead
		1214	// to parse anything.
804	switch (*dec->cur)	1215	switch (*dec->cur)
805	{	1216	{
806	case '"': ++dec->cur; return decode_str (dec);	1217	case '"': ++dec->cur; return decode_str (dec);
807	case '[': ++dec->cur; return decode_av (dec);	1218	case '[': ++dec->cur; return decode_av (dec);
808	case '{': ++dec->cur; return decode_hv (dec);	1219	case '{': ++dec->cur; return decode_hv (dec);
…		…
814		1225
815	case 't':	1226	case 't':
816	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
817	{	1228	{
818	dec->cur += 4;	1229	dec->cur += 4;
819	return newSViv (1);	1230	#if JSON_SLOW
		1231	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1232	#endif
		1233	return SvREFCNT_inc (json_true);
820	}	1234	}
821	else	1235	else
822	ERR ("'true' expected");	1236	ERR ("'true' expected");
823		1237
824	break;	1238	break;
825		1239
826	case 'f':	1240	case 'f':
827	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1241	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
828	{	1242	{
829	dec->cur += 5;	1243	dec->cur += 5;
830	return newSViv (0);	1244	#if JSON_SLOW
		1245	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1246	#endif
		1247	return SvREFCNT_inc (json_false);
831	}	1248	}
832	else	1249	else
833	ERR ("'false' expected");	1250	ERR ("'false' expected");
834		1251
835	break;	1252	break;
…		…
844	ERR ("'null' expected");	1261	ERR ("'null' expected");
845		1262
846	break;	1263	break;
847		1264
848	default:	1265	default:
849	ERR ("malformed json string, neither array, object, number, string or atom");	1266	ERR ("malformed JSON string, neither array, object, number, string or atom");
850	break;	1267	break;
851	}	1268	}
852		1269
853	fail:	1270	fail:
854	return 0;	1271	return 0;
855	}	1272	}
856		1273
857	static SV *	1274	static SV *
858	decode_json (SV *string, UV flags)	1275	decode_json (SV *string, JSON *json, UV *offset_return)
859	{	1276	{
		1277	dec_t dec;
		1278	UV offset;
860	SV *sv;	1279	SV *sv;
861		1280
		1281	SvGETMAGIC (string);
		1282	SvUPGRADE (string, SVt_PV);
		1283
		1284	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1285	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1286	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1287
862	if (flags & F_UTF8)	1288	if (json->flags & F_UTF8)
863	sv_utf8_downgrade (string, 0);	1289	sv_utf8_downgrade (string, 0);
864	else	1290	else
865	sv_utf8_upgrade (string);	1291	sv_utf8_upgrade (string);
866		1292
867	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1293	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
868		1294
869	dec_t dec;	1295	dec.json = *json;
870	dec.flags = flags;
871	dec.cur = SvPVX (string);	1296	dec.cur = SvPVX (string);
872	dec.end = SvEND (string);	1297	dec.end = SvEND (string);
873	dec.err = 0;	1298	dec.err = 0;
		1299	dec.depth = 0;
		1300	dec.maxdepth = DEC_DEPTH (dec.json.flags);
874		1301
		1302	if (dec.json.cb_object || dec.json.cb_sk_object)
		1303	dec.json.flags |= F_HOOK;
		1304
		1305	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
875	sv = decode_sv (&dec);	1306	sv = decode_sv (&dec);
876		1307
		1308	if (!(offset_return || !sv))
		1309	{
		1310	// check for trailing garbage
		1311	decode_ws (&dec);
		1312
		1313	if (*dec.cur)
		1314	{
		1315	dec.err = "garbage after JSON object";
		1316	SvREFCNT_dec (sv);
		1317	sv = 0;
		1318	}
		1319	}
		1320
		1321	if (offset_return || !sv)
		1322	{
		1323	offset = dec.json.flags & F_UTF8
		1324	? dec.cur - SvPVX (string)
		1325	: utf8_distance (dec.cur, SvPVX (string));
		1326
		1327	if (offset_return)
		1328	*offset_return = offset;
		1329	}
		1330
877	if (!sv)	1331	if (!sv)
878	{	1332	{
879	IV offset = dec.flags & F_UTF8
880	? dec.cur - SvPVX (string)
881	: utf8_distance (dec.cur, SvPVX (string));
882	SV *uni = sv_newmortal ();	1333	SV *uni = sv_newmortal ();
		1334
883	// horrible hack to silence warning inside pv_uni_display	1335	// horrible hack to silence warning inside pv_uni_display
884	COP cop;	1336	COP cop = *PL_curcop;
885	memset (&cop, 0, sizeof (cop));
886	cop.cop_warnings = pWARN_NONE;	1337	cop.cop_warnings = pWARN_NONE;
		1338	ENTER;
887	SAVEVPTR (PL_curcop);	1339	SAVEVPTR (PL_curcop);
888	PL_curcop = &cop;	1340	PL_curcop = &cop;
889
890	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1341	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
		1342	LEAVE;
		1343
891	croak ("%s, at character offset %d (%s)",	1344	croak ("%s, at character offset %d [\"%s\"]",
892	dec.err,	1345	dec.err,
893	(int)offset,	1346	(int)offset,
894	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1347	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
895	}	1348	}
896		1349
897	sv = sv_2mortal (sv);	1350	sv = sv_2mortal (sv);
898		1351
899	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1352	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
900	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1353	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
901		1354
902	return sv;	1355	return sv;
903	}	1356	}
904		1357
		1358	/////////////////////////////////////////////////////////////////////////////
		1359	// XS interface functions
		1360
905	MODULE = JSON::XS PACKAGE = JSON::XS	1361	MODULE = JSON::XS PACKAGE = JSON::XS
906		1362
907	BOOT:	1363	BOOT:
908	{	1364	{
909	int i;	1365	int i;
910		1366
911	memset (decode_hexdigit, 0xff, 256);
912	for (i = 10; i--; )	1367	for (i = 0; i < 256; ++i)
913	decode_hexdigit ['0' + i] = i;	1368	decode_hexdigit [i] =
		1369	i >= '0' && i <= '9' ? i - '0'
		1370	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1371	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1372	: -1;
914		1373
915	for (i = 7; i--; )	1374	json_stash = gv_stashpv ("JSON::XS" , 1);
		1375	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1376
		1377	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1378	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1379	}
		1380
		1381	PROTOTYPES: DISABLE
		1382
		1383	void CLONE (...)
		1384	CODE:
		1385	json_stash = 0;
		1386	json_boolean_stash = 0;
		1387
		1388	void new (char *klass)
		1389	PPCODE:
		1390	{
		1391	SV *pv = NEWSV (0, sizeof (JSON));
		1392	SvPOK_only (pv);
		1393	Zero (SvPVX (pv), 1, JSON);
		1394	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1395	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1396	}
		1397
		1398	void ascii (JSON *self, int enable = 1)
		1399	ALIAS:
		1400	ascii = F_ASCII
		1401	latin1 = F_LATIN1
		1402	utf8 = F_UTF8
		1403	indent = F_INDENT
		1404	canonical = F_CANONICAL
		1405	space_before = F_SPACE_BEFORE
		1406	space_after = F_SPACE_AFTER
		1407	pretty = F_PRETTY
		1408	allow_nonref = F_ALLOW_NONREF
		1409	shrink = F_SHRINK
		1410	allow_blessed = F_ALLOW_BLESSED
		1411	convert_blessed = F_CONV_BLESSED
		1412	PPCODE:
		1413	{
		1414	if (enable)
		1415	self->flags |= ix;
		1416	else
		1417	self->flags &= ~ix;
		1418
		1419	XPUSHs (ST (0));
		1420	}
		1421
		1422	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1423	PPCODE:
		1424	{
		1425	UV log2 = 0;
		1426
		1427	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1428
		1429	while ((1UL << log2) < max_depth)
		1430	++log2;
		1431
		1432	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1433
		1434	XPUSHs (ST (0));
		1435	}
		1436
		1437	void max_size (JSON *self, UV max_size = 0)
		1438	PPCODE:
		1439	{
		1440	UV log2 = 0;
		1441
		1442	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1443	if (max_size == 1) max_size = 2;
		1444
		1445	while ((1UL << log2) < max_size)
		1446	++log2;
		1447
		1448	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1449
		1450	XPUSHs (ST (0));
		1451	}
		1452
		1453	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1454	PPCODE:
		1455	{
		1456	SvREFCNT_dec (self->cb_object);
		1457	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1458
		1459	XPUSHs (ST (0));
		1460	}
		1461
		1462	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1463	PPCODE:
		1464	{
		1465	if (!self->cb_sk_object)
		1466	self->cb_sk_object = newHV ();
		1467
		1468	if (SvOK (cb))
		1469	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1470	else
916	{	1471	{
917	decode_hexdigit ['a' + i] = 10 + i;	1472	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
918	decode_hexdigit ['A' + i] = 10 + i;	1473
		1474	if (!HvKEYS (self->cb_sk_object))
		1475	{
		1476	SvREFCNT_dec (self->cb_sk_object);
		1477	self->cb_sk_object = 0;
		1478	}
919	}	1479	}
920		1480
921	json_stash = gv_stashpv ("JSON::XS", 1);	1481	XPUSHs (ST (0));
922	}	1482	}
923		1483
924	PROTOTYPES: DISABLE	1484	void encode (JSON *self, SV *scalar)
		1485	PPCODE:
		1486	XPUSHs (encode_json (scalar, self));
925		1487
926	SV *new (char *dummy)	1488	void decode (JSON *self, SV *jsonstr)
		1489	PPCODE:
		1490	XPUSHs (decode_json (jsonstr, self, 0));
		1491
		1492	void decode_prefix (JSON *self, SV *jsonstr)
		1493	PPCODE:
		1494	{
		1495	UV offset;
		1496	EXTEND (SP, 2);
		1497	PUSHs (decode_json (jsonstr, self, &offset));
		1498	PUSHs (sv_2mortal (newSVuv (offset)));
		1499	}
		1500
		1501	void DESTROY (JSON *self)
927	CODE:	1502	CODE:
928	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1503	SvREFCNT_dec (self->cb_sk_object);
929	OUTPUT:	1504	SvREFCNT_dec (self->cb_object);
930	RETVAL
931
932	SV *ascii (SV *self, int enable = 1)
933	ALIAS:
934	ascii = F_ASCII
935	utf8 = F_UTF8
936	indent = F_INDENT
937	canonical = F_CANONICAL
938	space_before = F_SPACE_BEFORE
939	space_after = F_SPACE_AFTER
940	json_rpc = F_JSON_RPC
941	pretty = F_PRETTY
942	allow_nonref = F_ALLOW_NONREF
943	shrink = F_SHRINK
944	CODE:
945	{
946	UV *uv = SvJSON (self);
947	if (enable)
948	*uv |= ix;
949	else
950	*uv &= ~ix;
951
952	RETVAL = newSVsv (self);
953	}
954	OUTPUT:
955	RETVAL
956
957	void encode (SV *self, SV *scalar)
958	PPCODE:
959	XPUSHs (encode_json (scalar, *SvJSON (self)));
960
961	void decode (SV *self, SV *jsonstr)
962	PPCODE:
963	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
964		1505
965	PROTOTYPES: ENABLE	1506	PROTOTYPES: ENABLE
966		1507
967	void to_json (SV *scalar)	1508	void to_json (SV *scalar)
968	PPCODE:	1509	PPCODE:
		1510	{
		1511	JSON json = { F_DEFAULT | F_UTF8 };
969	XPUSHs (encode_json (scalar, F_UTF8));	1512	XPUSHs (encode_json (scalar, &json));
		1513	}
970		1514
971	void from_json (SV *jsonstr)	1515	void from_json (SV *jsonstr)
972	PPCODE:	1516	PPCODE:
		1517	{
		1518	JSON json = { F_DEFAULT | F_UTF8 };
973	XPUSHs (decode_json (jsonstr, F_UTF8));	1519	XPUSHs (decode_json (jsonstr, &json, 0));
		1520	}
974		1521

Diff Legend

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