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 1.61 by root, Sun Aug 26 21:56:47 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 or illegal unicode character in string [%.11s], cannot convert to JSON", str);	201	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	202	}
140	else	203	else
141	{	204	{
…		…
144	}	207	}
145		208
146	if (uch > 0x10FFFFUL)	209	if (uch > 0x10FFFFUL)
147	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	210	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
148		211
149	if (uch < 0x80 || enc->flags & F_ASCII)	212	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
150	{	213	{
151	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
152	{	215	{
153	need (enc, len += 11);	216	need (enc, len += 11);
154	sprintf (enc->cur, "\\u%04x\\u%04x",	217	sprintf (enc->cur, "\\u%04x\\u%04x",
155	(uch - 0x10000) / 0x400 + 0xD800,	218	(int)((uch - 0x10000) / 0x400 + 0xD800),
156	(uch - 0x10000) % 0x400 + 0xDC00);	219	(int)((uch - 0x10000) % 0x400 + 0xDC00));
157	enc->cur += 12;	220	enc->cur += 12;
158	}	221	}
159	else	222	else
160	{	223	{
161	static char hexdigit [16] = "0123456789abcdef";	224	static char hexdigit [16] = "0123456789abcdef";
…		…
168	*enc->cur++ = hexdigit [(uch >> 0) & 15];	231	*enc->cur++ = hexdigit [(uch >> 0) & 15];
169	}	232	}
170		233
171	str += clen;	234	str += clen;
172	}	235	}
		236	else if (enc->json.flags & F_LATIN1)
		237	{
		238	*enc->cur++ = uch;
		239	str += clen;
		240	}
173	else if (is_utf8)	241	else if (is_utf8)
174	{	242	{
175	need (enc, len += clen);	243	need (enc, len += clen);
176	do	244	do
177	{	245	{
…		…
179	}	247	}
180	while (--clen);	248	while (--clen);
181	}	249	}
182	else	250	else
183	{	251	{
184	need (enc, len += 10); // never more than 11 bytes needed	252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
185	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
186	++str;	254	++str;
187	}	255	}
188	}	256	}
189	}	257	}
…		…
191		259
192	--len;	260	--len;
193	}	261	}
194	}	262	}
195		263
196	#define INDENT SB \	264	inline void
		265	encode_indent (enc_t *enc)
		266	{
197	if (enc->flags & F_INDENT) \	267	if (enc->json.flags & F_INDENT)
198	{ \	268	{
199	int i_; \	269	int spaces = enc->indent * INDENT_STEP;
200	need (enc, enc->indent); \	270
201	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);
202	encode_ch (enc, ' '); \	290	encode_ch (enc, '\n');
203	} \	291	}
204	SE	292	}
205		293
206	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	294	inline void
207	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	295	encode_comma (enc_t *enc)
208	#define COMMA SB \	296	{
209	encode_ch (enc, ','); \	297	encode_ch (enc, ',');
		298
210	if (enc->flags & F_INDENT) \	299	if (enc->json.flags & F_INDENT)
211	NL; \	300	encode_nl (enc);
212	else if (enc->flags & F_SPACE_AFTER) \	301	else if (enc->json.flags & F_SPACE_AFTER)
213	SPACE; \	302	encode_space (enc);
214	SE	303	}
215		304
216	static void encode_sv (enc_t *enc, SV *sv);	305	static void encode_sv (enc_t *enc, SV *sv);
217		306
218	static void	307	static void
219	encode_av (enc_t *enc, AV *av)	308	encode_av (enc_t *enc, AV *av)
220	{	309	{
221	int i, len = av_len (av);	310	int i, len = av_len (av);
222		311
223	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);
224	++enc->indent;	316	++enc->indent;
225		317
226	for (i = 0; i <= len; ++i)	318	for (i = 0; i <= len; ++i)
227	{	319	{
228	INDENT;	320	SV **svp = av_fetch (av, i, 0);
229	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);
230		328
231	if (i < len)	329	if (i < len)
232	COMMA;	330	encode_comma (enc);
233	}	331	}
234		332
235	NL;	333	encode_nl (enc);
236		334
237	--enc->indent;	335	--enc->indent;
238	INDENT; encode_ch (enc, ']');	336	encode_indent (enc); encode_ch (enc, ']');
239	}	337	}
240		338
241	static void	339	static void
242	encode_he (enc_t *enc, HE *he)	340	encode_hk (enc_t *enc, HE *he)
243	{	341	{
244	encode_ch (enc, '"');	342	encode_ch (enc, '"');
245		343
246	if (HeKLEN (he) == HEf_SVKEY)	344	if (HeKLEN (he) == HEf_SVKEY)
247	{	345	{
…		…
257	else	355	else
258	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	356	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
259		357
260	encode_ch (enc, '"');	358	encode_ch (enc, '"');
261		359
262	if (enc->flags & F_SPACE_BEFORE) SPACE;	360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
263	encode_ch (enc, ':');	361	encode_ch (enc, ':');
264	if (enc->flags & F_SPACE_AFTER ) SPACE;	362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
265	encode_sv (enc, HeVAL (he));
266	}	363	}
267		364
268	// compare hash entries, used when all keys are bytestrings	365	// compare hash entries, used when all keys are bytestrings
269	static int	366	static int
270	he_cmp_fast (const void *a_, const void *b_)	367	he_cmp_fast (const void *a_, const void *b_)
…		…
275	HE *b = *(HE **)b_;	372	HE *b = *(HE **)b_;
276		373
277	STRLEN la = HeKLEN (a);	374	STRLEN la = HeKLEN (a);
278	STRLEN lb = HeKLEN (b);	375	STRLEN lb = HeKLEN (b);
279		376
280	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	377	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
281	cmp = la < lb ? -1 : la == lb ? 0 : 1;	378	cmp = lb - la;
282		379
283	return cmp;	380	return cmp;
284	}	381	}
285		382
286	// compare hash entries, used when some keys are sv's or utf-x	383	// compare hash entries, used when some keys are sv's or utf-x
287	static int	384	static int
288	he_cmp_slow (const void *a, const void *b)	385	he_cmp_slow (const void *a, const void *b)
289	{	386	{
290	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	387	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
291	}	388	}
292		389
293	static void	390	static void
294	encode_hv (enc_t *enc, HV *hv)	391	encode_hv (enc_t *enc, HV *hv)
295	{	392	{
		393	HE *he;
296	int count, i;	394	int count;
297		395
		396	if (enc->indent >= enc->maxdepth)
		397	croak ("data structure too deep (hit recursion limit)");
		398
298	encode_ch (enc, '{'); NL; ++enc->indent;	399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
299		400
300	if ((count = hv_iterinit (hv)))
301	{
302	// for canonical output we have to sort by keys first	401	// for canonical output we have to sort by keys first
303	// actually, this is mostly due to the stupid so-called	402	// actually, this is mostly due to the stupid so-called
304	// security workaround added somewhere in 5.8.x.	403	// security workaround added somewhere in 5.8.x.
305	// that randomises hash orderings	404	// that randomises hash orderings
306	if (enc->flags & F_CANONICAL)	405	if (enc->json.flags & F_CANONICAL)
		406	{
		407	int count = hv_iterinit (hv);
		408
		409	if (SvMAGICAL (hv))
307	{	410	{
308	HE *he, *hes [count];	411	// need to count by iterating. could improve by dynamically building the vector below
		412	// but I don't care for the speed of this special case.
		413	// note also that we will run into undefined behaviour when the two iterations
		414	// do not result in the same count, something I might care for in some later release.
		415
		416	count = 0;
		417	while (hv_iternext (hv))
		418	++count;
		419
		420	hv_iterinit (hv);
		421	}
		422
		423	if (count)
		424	{
309	int fast = 1;	425	int i, fast = 1;
		426	#if defined(__BORLANDC__) || defined(_MSC_VER)
		427	HE **hes = _alloca (count * sizeof (HE));
		428	#else
		429	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		430	#endif
310		431
311	i = 0;	432	i = 0;
312	while ((he = hv_iternext (hv)))	433	while ((he = hv_iternext (hv)))
313	{	434	{
314	hes [i++] = he;	435	hes [i++] = he;
…		…
336		457
337	FREETMPS;	458	FREETMPS;
338	LEAVE;	459	LEAVE;
339	}	460	}
340		461
341	for (i = 0; i < count; ++i)	462	while (count--)
342	{
343	INDENT;
344	encode_he (enc, hes [i]);
345
346	if (i < count - 1)
347	COMMA;
348	}	463	{
		464	encode_indent (enc);
		465	he = hes [count];
		466	encode_hk (enc, he);
		467	encode_sv (enc, SvMAGICAL (hv) ? hv_iterval (hv, he) : HeVAL (he));
349		468
		469	if (count)
		470	encode_comma (enc);
350	NL;	471	}
		472	}
		473	}
		474	else
		475	{
		476	if (hv_iterinit (hv) || SvMAGICAL (hv))
		477	if ((he = hv_iternext (hv)))
		478	for (;;)
		479	{
		480	encode_indent (enc);
		481	encode_hk (enc, he);
		482	encode_sv (enc, SvMAGICAL (hv) ? hv_iterval (hv, he) : HeVAL (he));
		483
		484	if (!(he = hv_iternext (hv)))
		485	break;
		486
		487	encode_comma (enc);
		488	}
		489	}
		490
		491	encode_nl (enc);
		492
		493	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		494	}
		495
		496	// encode objects, arrays and special \0=false and \1=true values.
		497	static void
		498	encode_rv (enc_t *enc, SV *sv)
		499	{
		500	svtype svt;
		501
		502	SvGETMAGIC (sv);
		503	svt = SvTYPE (sv);
		504
		505	if (expect_false (SvOBJECT (sv)))
		506	{
		507	HV *stash = !JSON_SLOW || json_boolean_stash
		508	? json_boolean_stash
		509	: gv_stashpv ("JSON::XS::Boolean", 1);
		510
		511	if (SvSTASH (sv) == stash)
		512	{
		513	if (SvIV (sv))
		514	encode_str (enc, "true", 4, 0);
		515	else
		516	encode_str (enc, "false", 5, 0);
351	}	517	}
352	else	518	else
353	{	519	{
354	SV *sv;	520	#if 0
355	HE *he = hv_iternext (hv);	521	if (0 && sv_derived_from (rv, "JSON::Literal"))
356
357	for (;;)
358	{
359	INDENT;
360	encode_he (enc, he);
361
362	if (!(he = hv_iternext (hv)))
363	break;
364
365	COMMA;
366	}	522	{
367		523	// not yet
368	NL;	524	}
		525	#endif
		526	if (enc->json.flags & F_CONV_BLESSED)
		527	{
		528	// we re-bless the reference to get overload and other niceties right
		529	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		530
		531	if (to_json)
		532	{
		533	dSP;
		534
		535	ENTER; SAVETMPS; PUSHMARK (SP);
		536	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		537
		538	// calling with G_SCALAR ensures that we always get a 1 return value
		539	PUTBACK;
		540	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		541	SPAGAIN;
		542
		543	// catch this surprisingly common error
		544	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		545	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		546
		547	sv = POPs;
		548	PUTBACK;
		549
		550	encode_sv (enc, sv);
		551
		552	FREETMPS; LEAVE;
		553	}
		554	else if (enc->json.flags & F_ALLOW_BLESSED)
		555	encode_str (enc, "null", 4, 0);
		556	else
		557	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		558	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		559	}
		560	else if (enc->json.flags & F_ALLOW_BLESSED)
		561	encode_str (enc, "null", 4, 0);
		562	else
		563	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		564	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
369	}	565	}
370	}	566	}
		567	else if (svt == SVt_PVHV)
		568	encode_hv (enc, (HV *)sv);
		569	else if (svt == SVt_PVAV)
		570	encode_av (enc, (AV *)sv);
		571	else if (svt < SVt_PVAV)
		572	{
		573	STRLEN len = 0;
		574	char *pv = svt ? SvPV (sv, len) : 0;
371		575
372	--enc->indent; INDENT; encode_ch (enc, '}');	576	if (len == 1 && *pv == '1')
		577	encode_str (enc, "true", 4, 0);
		578	else if (len == 1 && *pv == '0')
		579	encode_str (enc, "false", 5, 0);
		580	else
		581	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		582	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		583	}
		584	else
		585	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		586	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
373	}	587	}
374		588
375	static void	589	static void
376	encode_sv (enc_t *enc, SV *sv)	590	encode_sv (enc_t *enc, SV *sv)
377	{	591	{
…		…
385	encode_str (enc, str, len, SvUTF8 (sv));	599	encode_str (enc, str, len, SvUTF8 (sv));
386	encode_ch (enc, '"');	600	encode_ch (enc, '"');
387	}	601	}
388	else if (SvNOKp (sv))	602	else if (SvNOKp (sv))
389	{	603	{
		604	// trust that perl will do the right thing w.r.t. JSON syntax.
390	need (enc, NV_DIG + 32);	605	need (enc, NV_DIG + 32);
391	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	606	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
392	enc->cur += strlen (enc->cur);	607	enc->cur += strlen (enc->cur);
393	}	608	}
394	else if (SvIOKp (sv))	609	else if (SvIOKp (sv))
395	{	610	{
396	need (enc, 64);	611	// we assume we can always read an IV as a UV
		612	if (SvUV (sv) & ~(UV)0x7fff)
		613	{
		614	// large integer, use the (rather slow) snprintf way.
		615	need (enc, sizeof (UV) * 3);
397	enc->cur +=	616	enc->cur +=
398	SvIsUV(sv)	617	SvIsUV(sv)
399	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	618	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
400	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	619	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		620	}
		621	else
		622	{
		623	// optimise the "small number case"
		624	// code will likely be branchless and use only a single multiplication
		625	I32 i = SvIV (sv);
		626	U32 u;
		627	char digit, nz = 0;
		628
		629	need (enc, 6);
		630
		631	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		632	u = i < 0 ? -i : i;
		633
		634	// convert to 4.28 fixed-point representation
		635	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		636
		637	// now output digit by digit, each time masking out the integer part
		638	// and multiplying by 5 while moving the decimal point one to the right,
		639	// resulting in a net multiplication by 10.
		640	// we always write the digit to memory but conditionally increment
		641	// the pointer, to ease the usage of conditional move instructions.
		642	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		643	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		644	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		645	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		646	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		647	}
401	}	648	}
402	else if (SvROK (sv))	649	else if (SvROK (sv))
403	{	650	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))	651	else if (!SvOK (sv))
420	encode_str (enc, "null", 4, 0);	652	encode_str (enc, "null", 4, 0);
421	else	653	else
422	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	654	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
423	SvPV_nolen (sv), SvFLAGS (sv));	655	SvPV_nolen (sv), SvFLAGS (sv));
424	}	656	}
425		657
426	static SV *	658	static SV *
427	encode_json (SV *scalar, UV flags)	659	encode_json (SV *scalar, JSON *json)
428	{	660	{
		661	enc_t enc;
		662
429	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	663	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
430	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	664	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
431		665
432	enc_t enc;	666	enc.json = *json;
433	enc.flags = flags;
434	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	667	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
435	enc.cur = SvPVX (enc.sv);	668	enc.cur = SvPVX (enc.sv);
436	enc.end = SvEND (enc.sv);	669	enc.end = SvEND (enc.sv);
437	enc.max_recurse = 0;
438	enc.indent = 0;	670	enc.indent = 0;
		671	enc.maxdepth = DEC_DEPTH (enc.json.flags);
439		672
440	SvPOK_only (enc.sv);	673	SvPOK_only (enc.sv);
441	encode_sv (&enc, scalar);	674	encode_sv (&enc, scalar);
442		675
		676	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		677	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		678
443	if (!(flags & (F_ASCII | F_UTF8)))	679	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
444	SvUTF8_on (enc.sv);	680	SvUTF8_on (enc.sv);
445		681
446	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
447
448	if (enc.flags & F_SHRINK)	682	if (enc.json.flags & F_SHRINK)
449	shrink (enc.sv);	683	shrink (enc.sv);
450		684
451	return enc.sv;	685	return enc.sv;
452	}	686	}
453		687
454	/////////////////////////////////////////////////////////////////////////////	688	/////////////////////////////////////////////////////////////////////////////
		689	// decoder
455		690
456	#define WS \	691	// structure used for decoding JSON
		692	typedef struct
		693	{
		694	char *cur; // current parser pointer
		695	char *end; // end of input string
		696	const char *err; // parse error, if != 0
		697	JSON json;
		698	U32 depth; // recursion depth
		699	U32 maxdepth; // recursion depth limit
		700	} dec_t;
		701
		702	inline void
		703	decode_ws (dec_t *dec)
		704	{
457	for (;;) \	705	for (;;)
458	{ \	706	{
459	char ch = *dec->cur; \	707	char ch = *dec->cur;
		708
460	if (ch > 0x20 \	709	if (ch > 0x20
461	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	710	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
462	break; \	711	break;
		712
463	++dec->cur; \	713	++dec->cur;
464	}	714	}
		715	}
465		716
466	#define ERR(reason) SB dec->err = reason; goto fail; SE	717	#define ERR(reason) SB dec->err = reason; goto fail; SE
		718
467	#define EXPECT_CH(ch) SB \	719	#define EXPECT_CH(ch) SB \
468	if (*dec->cur != ch) \	720	if (*dec->cur != ch) \
469	ERR (# ch " expected"); \	721	ERR (# ch " expected"); \
470	++dec->cur; \	722	++dec->cur; \
471	SE	723	SE
472		724
		725	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		726	#define DEC_DEC_DEPTH --dec->depth
		727
473	static SV *decode_sv (dec_t *dec);	728	static SV *decode_sv (dec_t *dec);
474		729
475	static signed char decode_hexdigit[256];	730	static signed char decode_hexdigit[256];
476		731
477	static UV	732	static UV
478	decode_4hex (dec_t *dec)	733	decode_4hex (dec_t *dec)
479	{	734	{
480	signed char d1, d2, d3, d4;	735	signed char d1, d2, d3, d4;
		736	unsigned char *cur = (unsigned char *)dec->cur;
481		737
482	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	738	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");	739	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]];	740	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");	741	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		742
491	dec->cur += 4;	743	dec->cur += 4;
492		744
493	return ((UV)d1) << 12	745	return ((UV)d1) << 12
494	| ((UV)d2) << 8	746	| ((UV)d2) << 8
…		…
497		749
498	fail:	750	fail:
499	return (UV)-1;	751	return (UV)-1;
500	}	752	}
501		753
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 *	754	static SV *
518	decode_str (dec_t *dec)	755	decode_str (dec_t *dec)
519	{	756	{
520	SV *sv = NEWSV (0,2);	757	SV *sv = 0;
521	int utf8 = 0;	758	int utf8 = 0;
522	char *cur = SvPVX (sv);	759	char *dec_cur = dec->cur;
523	char *end = SvEND (sv);
524		760
525	for (;;)	761	do
526	{	762	{
527	unsigned char ch = *(unsigned char *)dec->cur;	763	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		764	char *cur = buf;
528		765
529	if (ch == '"')	766	do
530	break;
531	else if (ch == '\\')
532	{	767	{
533	switch (*++dec->cur)	768	unsigned char ch = *(unsigned char *)dec_cur++;
		769
		770	if (expect_false (ch == '"'))
		771	{
		772	--dec_cur;
		773	break;
534	{	774	}
535	case '\\':	775	else if (expect_false (ch == '\\'))
536	case '/':	776	{
537	case '"': APPEND_CH (*dec->cur++); break;	777	switch (*dec_cur)
538
539	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
540	case 't': APPEND_CH ('\011'); ++dec->cur; break;
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	{	778	{
547	UV lo, hi;	779	case '\\':
548	++dec->cur;	780	case '/':
		781	case '"': *cur++ = *dec_cur++; break;
549		782
550	hi = decode_4hex (dec);	783	case 'b': ++dec_cur; *cur++ = '\010'; break;
551	if (hi == (UV)-1)	784	case 't': ++dec_cur; *cur++ = '\011'; break;
552	goto fail;	785	case 'n': ++dec_cur; *cur++ = '\012'; break;
		786	case 'f': ++dec_cur; *cur++ = '\014'; break;
		787	case 'r': ++dec_cur; *cur++ = '\015'; break;
553		788
554	// possibly a surrogate pair	789	case 'u':
555	if (hi >= 0xd800 && hi < 0xdc00)
556	{	790	{
557	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	791	UV lo, hi;
558	ERR ("missing low surrogate character in surrogate pair");	792	++dec_cur;
559		793
560	dec->cur += 2;	794	dec->cur = dec_cur;
561
562	lo = decode_4hex (dec);	795	hi = decode_4hex (dec);
		796	dec_cur = dec->cur;
563	if (lo == (UV)-1)	797	if (hi == (UV)-1)
564	goto fail;	798	goto fail;
565		799
		800	// possibly a surrogate pair
		801	if (hi >= 0xd800)
		802	if (hi < 0xdc00)
		803	{
		804	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		805	ERR ("missing low surrogate character in surrogate pair");
		806
		807	dec_cur += 2;
		808
		809	dec->cur = dec_cur;
		810	lo = decode_4hex (dec);
		811	dec_cur = dec->cur;
		812	if (lo == (UV)-1)
		813	goto fail;
		814
566	if (lo < 0xdc00 || lo >= 0xe000)	815	if (lo < 0xdc00 || lo >= 0xe000)
567	ERR ("surrogate pair expected");	816	ERR ("surrogate pair expected");
568		817
569	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	818	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		819	}
		820	else if (hi < 0xe000)
		821	ERR ("missing high surrogate character in surrogate pair");
		822
		823	if (hi >= 0x80)
		824	{
		825	utf8 = 1;
		826
		827	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		828	}
		829	else
		830	*cur++ = hi;
570	}	831	}
571	else if (hi >= 0xdc00 && hi < 0xe000)
572	ERR ("missing high surrogate character in surrogate pair");
573
574	if (hi >= 0x80)
575	{	832	break;
576	utf8 = 1;
577		833
578	APPEND_GROW (4); // at most 4 bytes for 21 bits
579	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
580	}
581	else	834	default:
582	APPEND_CH (hi);	835	--dec_cur;
		836	ERR ("illegal backslash escape sequence in string");
583	}	837	}
584	break;	838	}
		839	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		840	*cur++ = ch;
		841	else if (ch >= 0x80)
		842	{
		843	STRLEN clen;
		844	UV uch;
585		845
586	default:
587	--dec->cur;	846	--dec_cur;
588	ERR ("illegal backslash escape sequence in string");	847
		848	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		849	if (clen == (STRLEN)-1)
		850	ERR ("malformed UTF-8 character in JSON string");
		851
		852	do
		853	*cur++ = *dec_cur++;
		854	while (--clen);
		855
		856	utf8 = 1;
		857	}
		858	else
		859	{
		860	--dec_cur;
		861
		862	if (!ch)
		863	ERR ("unexpected end of string while parsing JSON string");
		864	else
		865	ERR ("invalid character encountered while parsing JSON string");
589	}	866	}
590	}	867	}
591	else if (ch >= 0x20 && ch <= 0x7f)	868	while (cur < buf + SHORT_STRING_LEN);
592	APPEND_CH (*dec->cur++);	869
593	else if (ch >= 0x80)
594	{	870	{
595	STRLEN clen;	871	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		872
600	APPEND_GROW (clen);	873	if (sv)
601	do
602	{	874	{
603	*cur++ = *dec->cur++;	875	SvGROW (sv, SvCUR (sv) + len + 1);
		876	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		877	SvCUR_set (sv, SvCUR (sv) + len);
604	}	878	}
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	879	else
612	ERR ("invalid character encountered");	880	sv = newSVpvn (buf, len);
613	}	881	}
		882	}
		883	while (*dec_cur != '"');
614		884
615	++dec->cur;	885	++dec_cur;
616		886
617	SvCUR_set (sv, cur - SvPVX (sv));	887	if (sv)
618		888	{
619	SvPOK_only (sv);	889	SvPOK_only (sv);
620	*SvEND (sv) = 0;	890	*SvEND (sv) = 0;
621		891
622	if (utf8)	892	if (utf8)
623	SvUTF8_on (sv);	893	SvUTF8_on (sv);
		894	}
		895	else
		896	sv = newSVpvn ("", 0);
624		897
625	if (dec->flags & F_SHRINK)	898	dec->cur = dec_cur;
626	shrink (sv);
627
628	return sv;	899	return sv;
629		900
630	fail:	901	fail:
631	SvREFCNT_dec (sv);	902	dec->cur = dec_cur;
632	return 0;	903	return 0;
633	}	904	}
634		905
635	static SV *	906	static SV *
636	decode_num (dec_t *dec)	907	decode_num (dec_t *dec)
…		…
694	is_nv = 1;	965	is_nv = 1;
695	}	966	}
696		967
697	if (!is_nv)	968	if (!is_nv)
698	{	969	{
699	UV uv;	970	int len = dec->cur - start;
700	int numtype = grok_number (start, dec->cur - start, &uv);	971
701	if (numtype & IS_NUMBER_IN_UV)	972	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
702	if (numtype & IS_NUMBER_NEG)	973	if (*start == '-')
		974	switch (len)
703	{	975	{
704	if (uv < (UV)IV_MIN)	976	case 2: return newSViv (-( start [1] - '0' * 1));
705	return newSViv (-(IV)uv);	977	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		978	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		979	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
706	}	980	}
		981	else
		982	switch (len)
		983	{
		984	case 1: return newSViv ( start [0] - '0' * 1);
		985	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		986	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		987	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		988	}
		989
		990	{
		991	UV uv;
		992	int numtype = grok_number (start, len, &uv);
		993	if (numtype & IS_NUMBER_IN_UV)
		994	if (numtype & IS_NUMBER_NEG)
		995	{
		996	if (uv < (UV)IV_MIN)
		997	return newSViv (-(IV)uv);
		998	}
707	else	999	else
708	return newSVuv (uv);	1000	return newSVuv (uv);
709	}	1001	}
710		1002
		1003	len -= *start == '-' ? 1 : 0;
		1004
		1005	// does not fit into IV or UV, try NV
		1006	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1007	#if defined (LDBL_DIG)
		1008	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1009	#endif
		1010	)
		1011	// fits into NV without loss of precision
		1012	return newSVnv (Atof (start));
		1013
		1014	// everything else fails, convert it to a string
		1015	return newSVpvn (start, dec->cur - start);
		1016	}
		1017
		1018	// loss of precision here
711	return newSVnv (Atof (start));	1019	return newSVnv (Atof (start));
712		1020
713	fail:	1021	fail:
714	return 0;	1022	return 0;
715	}	1023	}
…		…
717	static SV *	1025	static SV *
718	decode_av (dec_t *dec)	1026	decode_av (dec_t *dec)
719	{	1027	{
720	AV *av = newAV ();	1028	AV *av = newAV ();
721		1029
722	WS;	1030	DEC_INC_DEPTH;
		1031	decode_ws (dec);
		1032
723	if (*dec->cur == ']')	1033	if (*dec->cur == ']')
724	++dec->cur;	1034	++dec->cur;
725	else	1035	else
726	for (;;)	1036	for (;;)
727	{	1037	{
…		…
731	if (!value)	1041	if (!value)
732	goto fail;	1042	goto fail;
733		1043
734	av_push (av, value);	1044	av_push (av, value);
735		1045
736	WS;	1046	decode_ws (dec);
737		1047
738	if (*dec->cur == ']')	1048	if (*dec->cur == ']')
739	{	1049	{
740	++dec->cur;	1050	++dec->cur;
741	break;	1051	break;
…		…
745	ERR (", or ] expected while parsing array");	1055	ERR (", or ] expected while parsing array");
746		1056
747	++dec->cur;	1057	++dec->cur;
748	}	1058	}
749		1059
		1060	DEC_DEC_DEPTH;
750	return newRV_noinc ((SV *)av);	1061	return newRV_noinc ((SV *)av);
751		1062
752	fail:	1063	fail:
753	SvREFCNT_dec (av);	1064	SvREFCNT_dec (av);
		1065	DEC_DEC_DEPTH;
754	return 0;	1066	return 0;
755	}	1067	}
756		1068
757	static SV *	1069	static SV *
758	decode_hv (dec_t *dec)	1070	decode_hv (dec_t *dec)
759	{	1071	{
		1072	SV *sv;
760	HV *hv = newHV ();	1073	HV *hv = newHV ();
761		1074
762	WS;	1075	DEC_INC_DEPTH;
		1076	decode_ws (dec);
		1077
763	if (*dec->cur == '}')	1078	if (*dec->cur == '}')
764	++dec->cur;	1079	++dec->cur;
765	else	1080	else
766	for (;;)	1081	for (;;)
767	{	1082	{
		1083	decode_ws (dec); EXPECT_CH ('"');
		1084
		1085	// heuristic: assume that
		1086	// a) decode_str + hv_store_ent are abysmally slow.
		1087	// b) most hash keys are short, simple ascii text.
		1088	// => try to "fast-match" such strings to avoid
		1089	// the overhead of decode_str + hv_store_ent.
		1090	{
768	SV *key, *value;	1091	SV *value;
		1092	char *p = dec->cur;
		1093	char *e = p + 24; // only try up to 24 bytes
769		1094
770	WS; EXPECT_CH ('"');	1095	for (;;)
771
772	key = decode_str (dec);
773	if (!key)
774	goto fail;
775
776	WS; EXPECT_CH (':');
777
778	value = decode_sv (dec);
779	if (!value)
780	{	1096	{
		1097	// the >= 0x80 is true on most architectures
		1098	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1099	{
		1100	// slow path, back up and use decode_str
		1101	SV *key = decode_str (dec);
		1102	if (!key)
		1103	goto fail;
		1104
		1105	decode_ws (dec); EXPECT_CH (':');
		1106
		1107	value = decode_sv (dec);
		1108	if (!value)
		1109	{
		1110	SvREFCNT_dec (key);
		1111	goto fail;
		1112	}
		1113
		1114	hv_store_ent (hv, key, value, 0);
781	SvREFCNT_dec (key);	1115	SvREFCNT_dec (key);
		1116
		1117	break;
		1118	}
		1119	else if (*p == '"')
		1120	{
		1121	// fast path, got a simple key
		1122	char *key = dec->cur;
		1123	int len = p - key;
		1124	dec->cur = p + 1;
		1125
		1126	decode_ws (dec); EXPECT_CH (':');
		1127
		1128	value = decode_sv (dec);
		1129	if (!value)
782	goto fail;	1130	goto fail;
		1131
		1132	hv_store (hv, key, len, value, 0);
		1133
		1134	break;
		1135	}
		1136
		1137	++p;
783	}	1138	}
784
785	//TODO: optimise
786	hv_store_ent (hv, key, value, 0);
787
788	WS;	1139	}
		1140
		1141	decode_ws (dec);
789		1142
790	if (*dec->cur == '}')	1143	if (*dec->cur == '}')
791	{	1144	{
792	++dec->cur;	1145	++dec->cur;
793	break;	1146	break;
…		…
797	ERR (", or } expected while parsing object/hash");	1150	ERR (", or } expected while parsing object/hash");
798		1151
799	++dec->cur;	1152	++dec->cur;
800	}	1153	}
801		1154
		1155	DEC_DEC_DEPTH;
802	return newRV_noinc ((SV *)hv);	1156	sv = newRV_noinc ((SV *)hv);
		1157
		1158	// check filter callbacks
		1159	if (dec->json.flags & F_HOOK)
		1160	{
		1161	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1162	{
		1163	HE *cb, *he;
		1164
		1165	hv_iterinit (hv);
		1166	he = hv_iternext (hv);
		1167	hv_iterinit (hv);
		1168
		1169	// the next line creates a mortal sv each time its called.
		1170	// might want to optimise this for common cases.
		1171	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1172
		1173	if (cb)
		1174	{
		1175	dSP;
		1176	int count;
		1177
		1178	ENTER; SAVETMPS; PUSHMARK (SP);
		1179	XPUSHs (HeVAL (he));
		1180
		1181	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1182
		1183	if (count == 1)
		1184	{
		1185	sv = newSVsv (POPs);
		1186	FREETMPS; LEAVE;
		1187	return sv;
		1188	}
		1189
		1190	FREETMPS; LEAVE;
		1191	}
		1192	}
		1193
		1194	if (dec->json.cb_object)
		1195	{
		1196	dSP;
		1197	int count;
		1198
		1199	ENTER; SAVETMPS; PUSHMARK (SP);
		1200	XPUSHs (sv_2mortal (sv));
		1201
		1202	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1203
		1204	if (count == 1)
		1205	{
		1206	sv = newSVsv (POPs);
		1207	FREETMPS; LEAVE;
		1208	return sv;
		1209	}
		1210
		1211	SvREFCNT_inc (sv);
		1212	FREETMPS; LEAVE;
		1213	}
		1214	}
		1215
		1216	return sv;
803		1217
804	fail:	1218	fail:
805	SvREFCNT_dec (hv);	1219	SvREFCNT_dec (hv);
		1220	DEC_DEC_DEPTH;
806	return 0;	1221	return 0;
807	}	1222	}
808		1223
809	static SV *	1224	static SV *
810	decode_sv (dec_t *dec)	1225	decode_sv (dec_t *dec)
811	{	1226	{
812	WS;	1227	decode_ws (dec);
		1228
		1229	// the beauty of JSON: you need exactly one character lookahead
		1230	// to parse anything.
813	switch (*dec->cur)	1231	switch (*dec->cur)
814	{	1232	{
815	case '"': ++dec->cur; return decode_str (dec);	1233	case '"': ++dec->cur; return decode_str (dec);
816	case '[': ++dec->cur; return decode_av (dec);	1234	case '[': ++dec->cur; return decode_av (dec);
817	case '{': ++dec->cur; return decode_hv (dec);	1235	case '{': ++dec->cur; return decode_hv (dec);
…		…
823		1241
824	case 't':	1242	case 't':
825	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1243	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
826	{	1244	{
827	dec->cur += 4;	1245	dec->cur += 4;
828	return newSViv (1);	1246	#if JSON_SLOW
		1247	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1248	#endif
		1249	return SvREFCNT_inc (json_true);
829	}	1250	}
830	else	1251	else
831	ERR ("'true' expected");	1252	ERR ("'true' expected");
832		1253
833	break;	1254	break;
834		1255
835	case 'f':	1256	case 'f':
836	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1257	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
837	{	1258	{
838	dec->cur += 5;	1259	dec->cur += 5;
839	return newSViv (0);	1260	#if JSON_SLOW
		1261	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1262	#endif
		1263	return SvREFCNT_inc (json_false);
840	}	1264	}
841	else	1265	else
842	ERR ("'false' expected");	1266	ERR ("'false' expected");
843		1267
844	break;	1268	break;
…		…
853	ERR ("'null' expected");	1277	ERR ("'null' expected");
854		1278
855	break;	1279	break;
856		1280
857	default:	1281	default:
858	ERR ("malformed json string, neither array, object, number, string or atom");	1282	ERR ("malformed JSON string, neither array, object, number, string or atom");
859	break;	1283	break;
860	}	1284	}
861		1285
862	fail:	1286	fail:
863	return 0;	1287	return 0;
864	}	1288	}
865		1289
866	static SV *	1290	static SV *
867	decode_json (SV *string, UV flags)	1291	decode_json (SV *string, JSON *json, UV *offset_return)
868	{	1292	{
		1293	dec_t dec;
		1294	UV offset;
869	SV *sv;	1295	SV *sv;
870		1296
		1297	SvGETMAGIC (string);
		1298	SvUPGRADE (string, SVt_PV);
		1299
		1300	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1301	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1302	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1303
871	if (flags & F_UTF8)	1304	if (json->flags & F_UTF8)
872	sv_utf8_downgrade (string, 0);	1305	sv_utf8_downgrade (string, 0);
873	else	1306	else
874	sv_utf8_upgrade (string);	1307	sv_utf8_upgrade (string);
875		1308
876	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1309	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
877		1310
878	dec_t dec;	1311	dec.json = *json;
879	dec.flags = flags;
880	dec.cur = SvPVX (string);	1312	dec.cur = SvPVX (string);
881	dec.end = SvEND (string);	1313	dec.end = SvEND (string);
882	dec.err = 0;	1314	dec.err = 0;
		1315	dec.depth = 0;
		1316	dec.maxdepth = DEC_DEPTH (dec.json.flags);
883		1317
		1318	if (dec.json.cb_object || dec.json.cb_sk_object)
		1319	dec.json.flags |= F_HOOK;
		1320
		1321	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
884	sv = decode_sv (&dec);	1322	sv = decode_sv (&dec);
885		1323
		1324	if (!(offset_return || !sv))
		1325	{
		1326	// check for trailing garbage
		1327	decode_ws (&dec);
		1328
		1329	if (*dec.cur)
		1330	{
		1331	dec.err = "garbage after JSON object";
		1332	SvREFCNT_dec (sv);
		1333	sv = 0;
		1334	}
		1335	}
		1336
		1337	if (offset_return || !sv)
		1338	{
		1339	offset = dec.json.flags & F_UTF8
		1340	? dec.cur - SvPVX (string)
		1341	: utf8_distance (dec.cur, SvPVX (string));
		1342
		1343	if (offset_return)
		1344	*offset_return = offset;
		1345	}
		1346
886	if (!sv)	1347	if (!sv)
887	{	1348	{
888	IV offset = dec.flags & F_UTF8
889	? dec.cur - SvPVX (string)
890	: utf8_distance (dec.cur, SvPVX (string));
891	SV *uni = sv_newmortal ();	1349	SV *uni = sv_newmortal ();
892		1350
893	// horrible hack to silence warning inside pv_uni_display	1351	// horrible hack to silence warning inside pv_uni_display
894	COP cop = *PL_curcop;	1352	COP cop = *PL_curcop;
895	cop.cop_warnings = pWARN_NONE;	1353	cop.cop_warnings = pWARN_NONE;
…		…
897	SAVEVPTR (PL_curcop);	1355	SAVEVPTR (PL_curcop);
898	PL_curcop = &cop;	1356	PL_curcop = &cop;
899	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1357	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
900	LEAVE;	1358	LEAVE;
901		1359
902	croak ("%s, at character offset %d (%s)",	1360	croak ("%s, at character offset %d [\"%s\"]",
903	dec.err,	1361	dec.err,
904	(int)offset,	1362	(int)offset,
905	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1363	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
906	}	1364	}
907		1365
908	sv = sv_2mortal (sv);	1366	sv = sv_2mortal (sv);
909		1367
910	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1368	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)");	1369	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
912		1370
913	return sv;	1371	return sv;
914	}	1372	}
915		1373
		1374	/////////////////////////////////////////////////////////////////////////////
		1375	// XS interface functions
		1376
916	MODULE = JSON::XS PACKAGE = JSON::XS	1377	MODULE = JSON::XS PACKAGE = JSON::XS
917		1378
918	BOOT:	1379	BOOT:
919	{	1380	{
920	int i;	1381	int i;
921		1382
922	memset (decode_hexdigit, 0xff, 256);
923	for (i = 10; i--; )	1383	for (i = 0; i < 256; ++i)
924	decode_hexdigit ['0' + i] = i;	1384	decode_hexdigit [i] =
		1385	i >= '0' && i <= '9' ? i - '0'
		1386	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1387	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1388	: -1;
925		1389
926	for (i = 7; i--; )	1390	json_stash = gv_stashpv ("JSON::XS" , 1);
		1391	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1392
		1393	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1394	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1395	}
		1396
		1397	PROTOTYPES: DISABLE
		1398
		1399	void CLONE (...)
		1400	CODE:
		1401	json_stash = 0;
		1402	json_boolean_stash = 0;
		1403
		1404	void new (char *klass)
		1405	PPCODE:
		1406	{
		1407	SV *pv = NEWSV (0, sizeof (JSON));
		1408	SvPOK_only (pv);
		1409	Zero (SvPVX (pv), 1, JSON);
		1410	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1411	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1412	}
		1413
		1414	void ascii (JSON *self, int enable = 1)
		1415	ALIAS:
		1416	ascii = F_ASCII
		1417	latin1 = F_LATIN1
		1418	utf8 = F_UTF8
		1419	indent = F_INDENT
		1420	canonical = F_CANONICAL
		1421	space_before = F_SPACE_BEFORE
		1422	space_after = F_SPACE_AFTER
		1423	pretty = F_PRETTY
		1424	allow_nonref = F_ALLOW_NONREF
		1425	shrink = F_SHRINK
		1426	allow_blessed = F_ALLOW_BLESSED
		1427	convert_blessed = F_CONV_BLESSED
		1428	PPCODE:
		1429	{
		1430	if (enable)
		1431	self->flags |= ix;
		1432	else
		1433	self->flags &= ~ix;
		1434
		1435	XPUSHs (ST (0));
		1436	}
		1437
		1438	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1439	PPCODE:
		1440	{
		1441	UV log2 = 0;
		1442
		1443	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1444
		1445	while ((1UL << log2) < max_depth)
		1446	++log2;
		1447
		1448	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1449
		1450	XPUSHs (ST (0));
		1451	}
		1452
		1453	void max_size (JSON *self, UV max_size = 0)
		1454	PPCODE:
		1455	{
		1456	UV log2 = 0;
		1457
		1458	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1459	if (max_size == 1) max_size = 2;
		1460
		1461	while ((1UL << log2) < max_size)
		1462	++log2;
		1463
		1464	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1465
		1466	XPUSHs (ST (0));
		1467	}
		1468
		1469	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1470	PPCODE:
		1471	{
		1472	SvREFCNT_dec (self->cb_object);
		1473	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1474
		1475	XPUSHs (ST (0));
		1476	}
		1477
		1478	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1479	PPCODE:
		1480	{
		1481	if (!self->cb_sk_object)
		1482	self->cb_sk_object = newHV ();
		1483
		1484	if (SvOK (cb))
		1485	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1486	else
927	{	1487	{
928	decode_hexdigit ['a' + i] = 10 + i;	1488	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
929	decode_hexdigit ['A' + i] = 10 + i;	1489
		1490	if (!HvKEYS (self->cb_sk_object))
		1491	{
		1492	SvREFCNT_dec (self->cb_sk_object);
		1493	self->cb_sk_object = 0;
		1494	}
930	}	1495	}
931		1496
932	json_stash = gv_stashpv ("JSON::XS", 1);	1497	XPUSHs (ST (0));
933	}	1498	}
934		1499
935	PROTOTYPES: DISABLE	1500	void encode (JSON *self, SV *scalar)
		1501	PPCODE:
		1502	XPUSHs (encode_json (scalar, self));
936		1503
937	SV *new (char *dummy)	1504	void decode (JSON *self, SV *jsonstr)
		1505	PPCODE:
		1506	XPUSHs (decode_json (jsonstr, self, 0));
		1507
		1508	void decode_prefix (JSON *self, SV *jsonstr)
		1509	PPCODE:
		1510	{
		1511	UV offset;
		1512	EXTEND (SP, 2);
		1513	PUSHs (decode_json (jsonstr, self, &offset));
		1514	PUSHs (sv_2mortal (newSVuv (offset)));
		1515	}
		1516
		1517	void DESTROY (JSON *self)
938	CODE:	1518	CODE:
939	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1519	SvREFCNT_dec (self->cb_sk_object);
940	OUTPUT:	1520	SvREFCNT_dec (self->cb_object);
941	RETVAL
942
943	SV *ascii (SV *self, int enable = 1)
944	ALIAS:
945	ascii = F_ASCII
946	utf8 = F_UTF8
947	indent = F_INDENT
948	canonical = F_CANONICAL
949	space_before = F_SPACE_BEFORE
950	space_after = F_SPACE_AFTER
951	json_rpc = F_JSON_RPC
952	pretty = F_PRETTY
953	allow_nonref = F_ALLOW_NONREF
954	shrink = F_SHRINK
955	CODE:
956	{
957	UV *uv = SvJSON (self);
958	if (enable)
959	*uv |= ix;
960	else
961	*uv &= ~ix;
962
963	RETVAL = newSVsv (self);
964	}
965	OUTPUT:
966	RETVAL
967
968	void encode (SV *self, SV *scalar)
969	PPCODE:
970	XPUSHs (encode_json (scalar, *SvJSON (self)));
971
972	void decode (SV *self, SV *jsonstr)
973	PPCODE:
974	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
975		1521
976	PROTOTYPES: ENABLE	1522	PROTOTYPES: ENABLE
977		1523
978	void to_json (SV *scalar)	1524	void to_json (SV *scalar)
979	PPCODE:	1525	PPCODE:
		1526	{
		1527	JSON json = { F_DEFAULT | F_UTF8 };
980	XPUSHs (encode_json (scalar, F_UTF8));	1528	XPUSHs (encode_json (scalar, &json));
		1529	}
981		1530
982	void from_json (SV *jsonstr)	1531	void from_json (SV *jsonstr)
983	PPCODE:	1532	PPCODE:
		1533	{
		1534	JSON json = { F_DEFAULT | F_UTF8 };
984	XPUSHs (decode_json (jsonstr, F_UTF8));	1535	XPUSHs (decode_json (jsonstr, &json, 0));
		1536	}
985		1537

Diff Legend

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