ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.11 by root, Sat Mar 24 19:42:14 2007 UTC vs.
Revision 1.57 by root, Mon Aug 13 16:05:42 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_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
16	#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 ))
17		40
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	41	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	42	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		43
21	#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
22		48
23	#define SB do {	49	#define SB do {
24	#define SE } while (0)	50	#define SE } while (0)
25		51
26	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	#else
		66	# define JSON_SLOW 1
		67	#endif
		68
		69	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		70	static SV *json_true, *json_false;
		71
		72	typedef struct {
		73	U32 flags;
		74	SV *cb_object;
		75	HV *cb_sk_object;
		76	} JSON;
		77
		78	/////////////////////////////////////////////////////////////////////////////
		79	// utility functions
		80
		81	inline void
		82	shrink (SV *sv)
		83	{
		84	sv_utf8_downgrade (sv, 1);
		85	if (SvLEN (sv) > SvCUR (sv) + 1)
		86	{
		87	#ifdef SvPV_shrink_to_cur
		88	SvPV_shrink_to_cur (sv);
		89	#elif defined (SvPV_renew)
		90	SvPV_renew (sv, SvCUR (sv) + 1);
		91	#endif
		92	}
		93	}
		94
		95	// decode an utf-8 character and return it, or (UV)-1 in
		96	// case of an error.
		97	// we special-case "safe" characters from U+80 .. U+7FF,
		98	// but use the very good perl function to parse anything else.
		99	// note that we never call this function for a ascii codepoints
		100	inline UV
		101	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		102	{
		103	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		104	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		105	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		106	{
		107	*clen = 2;
		108	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		109	}
		110	else
		111	{
		112	*clen = (STRLEN)-1;
		113	return (UV)-1;
		114	}
		115	}
		116
		117	/////////////////////////////////////////////////////////////////////////////
		118	// encoder
27		119
28	// structure used for encoding JSON	120	// structure used for encoding JSON
29	typedef struct	121	typedef struct
30	{	122	{
31	char *cur;	123	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	124	char *end; // SvEND (sv)
34	SV *sv;	125	SV *sv; // result scalar
35	UV flags;	126	JSON json;
36	int max_recurse;	127	U32 indent; // indentation level
37	int indent;	128	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	129	} enc_t;
39		130
40	// structure used for decoding JSON	131	inline void
41	typedef struct
42	{
43	char *cur;
44	char *end;
45	const char *err;
46	UV flags;
47	} dec_t;
48
49	static UV *
50	SvJSON (SV *sv)
51	{
52	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
53	croak ("object is not of type JSON::XS");
54
55	return &SvUVX (SvRV (sv));
56	}
57
58	static void
59	shrink (SV *sv)
60	{
61	sv_utf8_downgrade (sv, 1);
62	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);
64	#endif
65	}
66
67	/////////////////////////////////////////////////////////////////////////////
68
69	static void
70	need (enc_t *enc, STRLEN len)	132	need (enc_t *enc, STRLEN len)
71	{	133	{
72	if (enc->cur + len >= enc->end)	134	if (expect_false (enc->cur + len >= enc->end))
73	{	135	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	136	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	137	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	138	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	139	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	140	}
79	}	141	}
80		142
81	static void	143	inline void
82	encode_ch (enc_t *enc, char ch)	144	encode_ch (enc_t *enc, char ch)
83	{	145	{
84	need (enc, 1);	146	need (enc, 1);
85	*enc->cur++ = ch;	147	*enc->cur++ = ch;
86	}	148	}
…		…
94		156
95	while (str < end)	157	while (str < end)
96	{	158	{
97	unsigned char ch = *(unsigned char *)str;	159	unsigned char ch = *(unsigned char *)str;
98		160
99	if (ch >= 0x20 && ch < 0x80) // most common case	161	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	162	{
101	if (ch == '"') // but with slow exceptions	163	if (expect_false (ch == '"')) // but with slow exceptions
102	{	164	{
103	need (enc, len += 1);	165	need (enc, len += 1);
104	*enc->cur++ = '\\';	166	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	167	*enc->cur++ = '"';
106	}	168	}
107	else if (ch == '\\')	169	else if (expect_false (ch == '\\'))
108	{	170	{
109	need (enc, len += 1);	171	need (enc, len += 1);
110	*enc->cur++ = '\\';	172	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	173	*enc->cur++ = '\\';
112	}	174	}
…		…
130	STRLEN clen;	192	STRLEN clen;
131	UV uch;	193	UV uch;
132		194
133	if (is_utf8)	195	if (is_utf8)
134	{	196	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	197	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	198	if (clen == (STRLEN)-1)
137	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	199	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
138	}	200	}
139	else	201	else
140	{	202	{
…		…
143	}	205	}
144		206
145	if (uch > 0x10FFFFUL)	207	if (uch > 0x10FFFFUL)
146	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);	208	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
147		209
148	if (uch < 0x80 || enc->flags & F_ASCII)	210	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149	{	211	{
150	if (uch > 0xFFFFUL)	212	if (uch > 0xFFFFUL)
151	{	213	{
152	need (enc, len += 11);	214	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	215	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	229	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	230	}
169		231
170	str += clen;	232	str += clen;
171	}	233	}
		234	else if (enc->json.flags & F_LATIN1)
		235	{
		236	*enc->cur++ = uch;
		237	str += clen;
		238	}
172	else if (is_utf8)	239	else if (is_utf8)
173	{	240	{
174	need (enc, len += clen);	241	need (enc, len += clen);
175	do	242	do
176	{	243	{
…		…
178	}	245	}
179	while (--clen);	246	while (--clen);
180	}	247	}
181	else	248	else
182	{	249	{
183	need (enc, len += 10); // never more than 11 bytes needed	250	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	251	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185	++str;	252	++str;
186	}	253	}
187	}	254	}
188	}	255	}
…		…
190		257
191	--len;	258	--len;
192	}	259	}
193	}	260	}
194		261
195	#define INDENT SB \	262	inline void
		263	encode_indent (enc_t *enc)
		264	{
196	if (enc->flags & F_INDENT) \	265	if (enc->json.flags & F_INDENT)
197	{ \	266	{
198	int i_; \	267	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	268
200	for (i_ = enc->indent * 3; i_--; )\	269	need (enc, spaces);
		270	memset (enc->cur, ' ', spaces);
		271	enc->cur += spaces;
		272	}
		273	}
		274
		275	inline void
		276	encode_space (enc_t *enc)
		277	{
		278	need (enc, 1);
		279	encode_ch (enc, ' ');
		280	}
		281
		282	inline void
		283	encode_nl (enc_t *enc)
		284	{
		285	if (enc->json.flags & F_INDENT)
		286	{
		287	need (enc, 1);
201	encode_ch (enc, ' '); \	288	encode_ch (enc, '\n');
202	} \	289	}
203	SE	290	}
204		291
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	292	inline void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	293	encode_comma (enc_t *enc)
207	#define COMMA SB \	294	{
208	encode_ch (enc, ','); \	295	encode_ch (enc, ',');
		296
209	if (enc->flags & F_INDENT) \	297	if (enc->json.flags & F_INDENT)
210	NL; \	298	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	299	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	300	encode_space (enc);
213	SE	301	}
214		302
215	static void encode_sv (enc_t *enc, SV *sv);	303	static void encode_sv (enc_t *enc, SV *sv);
216		304
217	static void	305	static void
218	encode_av (enc_t *enc, AV *av)	306	encode_av (enc_t *enc, AV *av)
219	{	307	{
220	int i, len = av_len (av);	308	int i, len = av_len (av);
221		309
222	encode_ch (enc, '['); NL;	310	if (enc->indent >= enc->maxdepth)
		311	croak ("data structure too deep (hit recursion limit)");
		312
		313	encode_ch (enc, '['); encode_nl (enc);
223	++enc->indent;	314	++enc->indent;
224		315
225	for (i = 0; i <= len; ++i)	316	for (i = 0; i <= len; ++i)
226	{	317	{
227	INDENT;	318	SV **svp = av_fetch (av, i, 0);
228	encode_sv (enc, *av_fetch (av, i, 0));	319
		320	encode_indent (enc);
		321
		322	if (svp)
		323	encode_sv (enc, *svp);
		324	else
		325	encode_str (enc, "null", 4, 0);
229		326
230	if (i < len)	327	if (i < len)
231	COMMA;	328	encode_comma (enc);
232	}	329	}
233		330
234	NL;	331	encode_nl (enc);
235		332
236	--enc->indent;	333	--enc->indent;
237	INDENT; encode_ch (enc, ']');	334	encode_indent (enc); encode_ch (enc, ']');
238	}	335	}
239		336
240	static void	337	static void
241	encode_he (enc_t *enc, HE *he)	338	encode_he (enc_t *enc, HE *he)
242	{	339	{
…		…
256	else	353	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	354	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		355
259	encode_ch (enc, '"');	356	encode_ch (enc, '"');
260		357
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	358	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	359	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	360	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));	361	encode_sv (enc, HeVAL (he));
265	}	362	}
266		363
267	// compare hash entries, used when all keys are bytestrings	364	// compare hash entries, used when all keys are bytestrings
268	static int	365	static int
…		…
292	static void	389	static void
293	encode_hv (enc_t *enc, HV *hv)	390	encode_hv (enc_t *enc, HV *hv)
294	{	391	{
295	int count, i;	392	int count, i;
296		393
		394	if (enc->indent >= enc->maxdepth)
		395	croak ("data structure too deep (hit recursion limit)");
		396
297	encode_ch (enc, '{'); NL; ++enc->indent;	397	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298		398
299	if ((count = hv_iterinit (hv)))	399	if ((count = hv_iterinit (hv)))
300	{	400	{
301	// for canonical output we have to sort by keys first	401	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	402	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	403	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	404	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	405	if (enc->json.flags & F_CANONICAL)
306	{	406	{
307	HE *he, *hes [count];
308	int fast = 1;	407	int fast = 1;
		408	HE *he;
		409	#if defined(__BORLANDC__) || defined(_MSC_VER)
		410	HE **hes = _alloca (count * sizeof (HE));
		411	#else
		412	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		413	#endif
309		414
310	i = 0;	415	i = 0;
311	while ((he = hv_iternext (hv)))	416	while ((he = hv_iternext (hv)))
312	{	417	{
313	hes [i++] = he;	418	hes [i++] = he;
…		…
337	LEAVE;	442	LEAVE;
338	}	443	}
339		444
340	for (i = 0; i < count; ++i)	445	for (i = 0; i < count; ++i)
341	{	446	{
342	INDENT;	447	encode_indent (enc);
343	encode_he (enc, hes [i]);	448	encode_he (enc, hes [i]);
344		449
345	if (i < count - 1)	450	if (i < count - 1)
346	COMMA;	451	encode_comma (enc);
347	}
348
349	NL;	452	}
		453
		454	encode_nl (enc);
350	}	455	}
351	else	456	else
352	{	457	{
353	SV *sv;
354	HE *he = hv_iternext (hv);	458	HE *he = hv_iternext (hv);
355		459
356	for (;;)	460	for (;;)
357	{	461	{
358	INDENT;	462	encode_indent (enc);
359	encode_he (enc, he);	463	encode_he (enc, he);
360		464
361	if (!(he = hv_iternext (hv)))	465	if (!(he = hv_iternext (hv)))
362	break;	466	break;
363		467
364	COMMA;	468	encode_comma (enc);
365	}
366
367	NL;	469	}
		470
		471	encode_nl (enc);
368	}	472	}
369	}	473	}
370		474
371	--enc->indent; INDENT; encode_ch (enc, '}');	475	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		476	}
		477
		478	// encode objects, arrays and special \0=false and \1=true values.
		479	static void
		480	encode_rv (enc_t *enc, SV *sv)
		481	{
		482	svtype svt;
		483
		484	SvGETMAGIC (sv);
		485	svt = SvTYPE (sv);
		486
		487	if (expect_false (SvOBJECT (sv)))
		488	{
		489	HV *stash = !JSON_SLOW || json_boolean_stash
		490	? json_boolean_stash
		491	: gv_stashpv ("JSON::XS::Boolean", 1);
		492
		493	if (SvSTASH (sv) == stash)
		494	{
		495	if (SvIV (sv))
		496	encode_str (enc, "true", 4, 0);
		497	else
		498	encode_str (enc, "false", 5, 0);
		499	}
		500	else
		501	{
		502	#if 0
		503	if (0 && sv_derived_from (rv, "JSON::Literal"))
		504	{
		505	// not yet
		506	}
		507	#endif
		508	if (enc->json.flags & F_CONV_BLESSED)
		509	{
		510	// we re-bless the reference to get overload and other niceties right
		511	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		512
		513	if (to_json)
		514	{
		515	int count;
		516	dSP;
		517
		518	ENTER; SAVETMPS; PUSHMARK (SP);
		519	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		520
		521	// calling with G_SCALAR ensures that we always get a 1 return value
		522	PUTBACK;
		523	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		524	SPAGAIN;
		525
		526	// catch this surprisingly common error
		527	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		528	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		529
		530	sv = POPs;
		531	PUTBACK;
		532
		533	encode_sv (enc, sv);
		534
		535	FREETMPS; LEAVE;
		536	}
		537	else if (enc->json.flags & F_ALLOW_BLESSED)
		538	encode_str (enc, "null", 4, 0);
		539	else
		540	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		541	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		542	}
		543	else if (enc->json.flags & F_ALLOW_BLESSED)
		544	encode_str (enc, "null", 4, 0);
		545	else
		546	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		547	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		548	}
		549	}
		550	else if (svt == SVt_PVHV)
		551	encode_hv (enc, (HV *)sv);
		552	else if (svt == SVt_PVAV)
		553	encode_av (enc, (AV *)sv);
		554	else if (svt < SVt_PVAV)
		555	{
		556	STRLEN len = 0;
		557	char *pv = svt ? SvPV (sv, len) : 0;
		558
		559	if (len == 1 && *pv == '1')
		560	encode_str (enc, "true", 4, 0);
		561	else if (len == 1 && *pv == '0')
		562	encode_str (enc, "false", 5, 0);
		563	else
		564	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		565	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		566	}
		567	else
		568	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		569	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	570	}
373		571
374	static void	572	static void
375	encode_sv (enc_t *enc, SV *sv)	573	encode_sv (enc_t *enc, SV *sv)
376	{	574	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	582	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	583	encode_ch (enc, '"');
386	}	584	}
387	else if (SvNOKp (sv))	585	else if (SvNOKp (sv))
388	{	586	{
		587	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	588	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	589	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	590	enc->cur += strlen (enc->cur);
392	}	591	}
393	else if (SvIOKp (sv))	592	else if (SvIOKp (sv))
394	{	593	{
395	need (enc, 64);	594	// we assume we can always read an IV as a UV
		595	if (SvUV (sv) & ~(UV)0x7fff)
		596	{
		597	// large integer, use the (rather slow) snprintf way.
		598	need (enc, sizeof (UV) * 3);
396	enc->cur +=	599	enc->cur +=
397	SvIsUV(sv)	600	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	601	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	602	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		603	}
		604	else
		605	{
		606	// optimise the "small number case"
		607	// code will likely be branchless and use only a single multiplication
		608	I32 i = SvIV (sv);
		609	U32 u;
		610	char digit, nz = 0;
		611
		612	need (enc, 6);
		613
		614	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		615	u = i < 0 ? -i : i;
		616
		617	// convert to 4.28 fixed-point representation
		618	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		619
		620	// now output digit by digit, each time masking out the integer part
		621	// and multiplying by 5 while moving the decimal point one to the right,
		622	// resulting in a net multiplication by 10.
		623	// we always write the digit to memory but conditionally increment
		624	// the pointer, to ease the usage of conditional move instructions.
		625	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		626	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		627	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		628	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		629	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		630	}
400	}	631	}
401	else if (SvROK (sv))	632	else if (SvROK (sv))
402	{	633	encode_rv (enc, SvRV (sv));
403	SV *rv = SvRV (sv);
404
405	if (!--enc->max_recurse)
406	croak ("data structure too deep (hit recursion limit)");
407
408	switch (SvTYPE (rv))
409	{
410	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
411	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
412
413	default:
414	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
415	SvPV_nolen (sv));
416	}
417	}
418	else if (!SvOK (sv))	634	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	635	encode_str (enc, "null", 4, 0);
420	else	636	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	637	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	638	SvPV_nolen (sv), SvFLAGS (sv));
423	}	639	}
424		640
425	static SV *	641	static SV *
426	encode_json (SV *scalar, UV flags)	642	encode_json (SV *scalar, JSON *json)
427	{	643	{
		644	enc_t enc;
		645
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	646	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	647	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		648
431	enc_t enc;	649	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	650	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	651	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	652	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	653	enc.indent = 0;
		654	enc.maxdepth = DEC_DEPTH (enc.json.flags);
438		655
439	SvPOK_only (enc.sv);	656	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	657	encode_sv (&enc, scalar);
441		658
		659	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		660	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		661
442	if (!(flags & (F_ASCII | F_UTF8)))	662	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	663	SvUTF8_on (enc.sv);
444		664
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	665	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	666	shrink (enc.sv);
449		667
450	return enc.sv;	668	return enc.sv;
451	}	669	}
452		670
453	/////////////////////////////////////////////////////////////////////////////	671	/////////////////////////////////////////////////////////////////////////////
		672	// decoder
454		673
455	#define WS \	674	// structure used for decoding JSON
		675	typedef struct
		676	{
		677	char *cur; // current parser pointer
		678	char *end; // end of input string
		679	const char *err; // parse error, if != 0
		680	JSON json;
		681	U32 depth; // recursion depth
		682	U32 maxdepth; // recursion depth limit
		683	} dec_t;
		684
		685	inline void
		686	decode_ws (dec_t *dec)
		687	{
456	for (;;) \	688	for (;;)
457	{ \	689	{
458	char ch = *dec->cur; \	690	char ch = *dec->cur;
		691
459	if (ch > 0x20 \	692	if (ch > 0x20
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	693	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461	break; \	694	break;
		695
462	++dec->cur; \	696	++dec->cur;
463	}	697	}
		698	}
464		699
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	700	#define ERR(reason) SB dec->err = reason; goto fail; SE
		701
466	#define EXPECT_CH(ch) SB \	702	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	703	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	704	ERR (# ch " expected"); \
469	++dec->cur; \	705	++dec->cur; \
470	SE	706	SE
471		707
		708	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		709	#define DEC_DEC_DEPTH --dec->depth
		710
472	static SV *decode_sv (dec_t *dec);	711	static SV *decode_sv (dec_t *dec);
473		712
474	static signed char decode_hexdigit[256];	713	static signed char decode_hexdigit[256];
475		714
476	static UV	715	static UV
477	decode_4hex (dec_t *dec)	716	decode_4hex (dec_t *dec)
478	{	717	{
479	signed char d1, d2, d3, d4;	718	signed char d1, d2, d3, d4;
		719	unsigned char *cur = (unsigned char *)dec->cur;
480		720
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	721	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	722	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	723	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	724	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
485	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
486	if (d3 < 0) ERR ("four hexadecimal digits expected");
487	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
488	if (d4 < 0) ERR ("four hexadecimal digits expected");
489		725
490	dec->cur += 4;	726	dec->cur += 4;
491		727
492	return ((UV)d1) << 12	728	return ((UV)d1) << 12
493	| ((UV)d2) << 8	729	| ((UV)d2) << 8
…		…
496		732
497	fail:	733	fail:
498	return (UV)-1;	734	return (UV)-1;
499	}	735	}
500		736
501	#define APPEND_GROW(n) SB \
502	if (cur + (n) >= end) \
503	{ \
504	STRLEN ofs = cur - SvPVX (sv); \
505	SvGROW (sv, ofs + (n) + 1); \
506	cur = SvPVX (sv) + ofs; \
507	end = SvEND (sv); \
508	} \
509	SE
510
511	#define APPEND_CH(ch) SB \
512	APPEND_GROW (1); \
513	*cur++ = (ch); \
514	SE
515
516	static SV *	737	static SV *
517	decode_str (dec_t *dec)	738	decode_str (dec_t *dec)
518	{	739	{
519	SV *sv = NEWSV (0,2);	740	SV *sv = 0;
520	int utf8 = 0;	741	int utf8 = 0;
521	char *cur = SvPVX (sv);	742	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		743
524	for (;;)	744	do
525	{	745	{
526	unsigned char ch = *(unsigned char *)dec->cur;	746	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		747	char *cur = buf;
527		748
528	if (ch == '"')	749	do
529	break;
530	else if (ch == '\\')
531	{	750	{
532	switch (*++dec->cur)	751	unsigned char ch = *(unsigned char *)dec_cur++;
		752
		753	if (expect_false (ch == '"'))
		754	{
		755	--dec_cur;
		756	break;
533	{	757	}
534	case '\\':	758	else if (expect_false (ch == '\\'))
535	case '/':	759	{
536	case '"': APPEND_CH (*dec->cur++); break;	760	switch (*dec_cur)
537
538	case 'b': APPEND_CH ('\010'); ++dec->cur; break;
539	case 't': APPEND_CH ('\011'); ++dec->cur; break;
540	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
541	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
542	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
543
544	case 'u':
545	{	761	{
546	UV lo, hi;	762	case '\\':
547	++dec->cur;	763	case '/':
		764	case '"': *cur++ = *dec_cur++; break;
548		765
549	hi = decode_4hex (dec);	766	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	767	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	768	case 'n': ++dec_cur; *cur++ = '\012'; break;
		769	case 'f': ++dec_cur; *cur++ = '\014'; break;
		770	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		771
553	// possibly a surrogate pair	772	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	773	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	774	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	775	++dec_cur;
558		776
559	dec->cur += 2;	777	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	778	hi = decode_4hex (dec);
		779	dec_cur = dec->cur;
562	if (lo == (UV)-1)	780	if (hi == (UV)-1)
563	goto fail;	781	goto fail;
564		782
		783	// possibly a surrogate pair
		784	if (hi >= 0xd800)
		785	if (hi < 0xdc00)
		786	{
		787	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		788	ERR ("missing low surrogate character in surrogate pair");
		789
		790	dec_cur += 2;
		791
		792	dec->cur = dec_cur;
		793	lo = decode_4hex (dec);
		794	dec_cur = dec->cur;
		795	if (lo == (UV)-1)
		796	goto fail;
		797
565	if (lo < 0xdc00 || lo >= 0xe000)	798	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	799	ERR ("surrogate pair expected");
567		800
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	801	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		802	}
		803	else if (hi < 0xe000)
		804	ERR ("missing high surrogate character in surrogate pair");
		805
		806	if (hi >= 0x80)
		807	{
		808	utf8 = 1;
		809
		810	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		811	}
		812	else
		813	*cur++ = hi;
569	}	814	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	815	break;
575	utf8 = 1;
576		816
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	817	default:
581	APPEND_CH (hi);	818	--dec_cur;
		819	ERR ("illegal backslash escape sequence in string");
582	}	820	}
583	break;	821	}
		822	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		823	*cur++ = ch;
		824	else if (ch >= 0x80)
		825	{
		826	STRLEN clen;
		827	UV uch;
584		828
585	default:
586	--dec->cur;	829	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	830
		831	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		832	if (clen == (STRLEN)-1)
		833	ERR ("malformed UTF-8 character in JSON string");
		834
		835	do
		836	*cur++ = *dec_cur++;
		837	while (--clen);
		838
		839	utf8 = 1;
		840	}
		841	else
		842	{
		843	--dec_cur;
		844
		845	if (!ch)
		846	ERR ("unexpected end of string while parsing JSON string");
		847	else
		848	ERR ("invalid character encountered while parsing JSON string");
588	}	849	}
589	}	850	}
590	else if (ch >= 0x20 && ch <= 0x7f)	851	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	852
592	else if (ch >= 0x80)
593	{	853	{
594	STRLEN clen;	854	STRLEN len = cur - buf;
595	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
596	if (clen == (STRLEN)-1)
597	ERR ("malformed UTF-8 character in JSON string");
598		855
599	APPEND_GROW (clen);	856	if (sv)
600	do
601	{	857	{
602	*cur++ = *dec->cur++;	858	SvGROW (sv, SvCUR (sv) + len + 1);
		859	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		860	SvCUR_set (sv, SvCUR (sv) + len);
603	}	861	}
604	while (--clen);
605
606	utf8 = 1;
607	}
608	else if (dec->cur == dec->end)
609	ERR ("unexpected end of string while parsing json string");
610	else	862	else
611	ERR ("invalid character encountered");	863	sv = newSVpvn (buf, len);
612	}	864	}
		865	}
		866	while (*dec_cur != '"');
613		867
614	++dec->cur;	868	++dec_cur;
615		869
616	SvCUR_set (sv, cur - SvPVX (sv));	870	if (sv)
617		871	{
618	SvPOK_only (sv);	872	SvPOK_only (sv);
619	*SvEND (sv) = 0;	873	*SvEND (sv) = 0;
620		874
621	if (utf8)	875	if (utf8)
622	SvUTF8_on (sv);	876	SvUTF8_on (sv);
		877	}
		878	else
		879	sv = newSVpvn ("", 0);
623		880
624	if (dec->flags & F_SHRINK)	881	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	882	return sv;
628		883
629	fail:	884	fail:
630	SvREFCNT_dec (sv);	885	dec->cur = dec_cur;
631	return 0;	886	return 0;
632	}	887	}
633		888
634	static SV *	889	static SV *
635	decode_num (dec_t *dec)	890	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	948	is_nv = 1;
694	}	949	}
695		950
696	if (!is_nv)	951	if (!is_nv)
697	{	952	{
698	UV uv;	953	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	954
700	if (numtype & IS_NUMBER_IN_UV)	955	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
701	if (numtype & IS_NUMBER_NEG)	956	if (*start == '-')
		957	switch (len)
702	{	958	{
703	if (uv < (UV)IV_MIN)	959	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	960	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		961	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		962	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705	}	963	}
		964	else
		965	switch (len)
		966	{
		967	case 1: return newSViv ( start [0] - '0' * 1);
		968	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		969	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		970	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		971	}
		972
		973	{
		974	UV uv;
		975	int numtype = grok_number (start, len, &uv);
		976	if (numtype & IS_NUMBER_IN_UV)
		977	if (numtype & IS_NUMBER_NEG)
		978	{
		979	if (uv < (UV)IV_MIN)
		980	return newSViv (-(IV)uv);
		981	}
706	else	982	else
707	return newSVuv (uv);	983	return newSVuv (uv);
708	}	984	}
709		985
		986	len -= *start == '-' ? 1 : 0;
		987
		988	// does not fit into IV or UV, try NV
		989	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		990	#if defined (LDBL_DIG)
		991	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		992	#endif
		993	)
		994	// fits into NV without loss of precision
		995	return newSVnv (Atof (start));
		996
		997	// everything else fails, convert it to a string
		998	return newSVpvn (start, dec->cur - start);
		999	}
		1000
		1001	// loss of precision here
710	return newSVnv (Atof (start));	1002	return newSVnv (Atof (start));
711		1003
712	fail:	1004	fail:
713	return 0;	1005	return 0;
714	}	1006	}
…		…
716	static SV *	1008	static SV *
717	decode_av (dec_t *dec)	1009	decode_av (dec_t *dec)
718	{	1010	{
719	AV *av = newAV ();	1011	AV *av = newAV ();
720		1012
721	WS;	1013	DEC_INC_DEPTH;
		1014	decode_ws (dec);
		1015
722	if (*dec->cur == ']')	1016	if (*dec->cur == ']')
723	++dec->cur;	1017	++dec->cur;
724	else	1018	else
725	for (;;)	1019	for (;;)
726	{	1020	{
…		…
730	if (!value)	1024	if (!value)
731	goto fail;	1025	goto fail;
732		1026
733	av_push (av, value);	1027	av_push (av, value);
734		1028
735	WS;	1029	decode_ws (dec);
736		1030
737	if (*dec->cur == ']')	1031	if (*dec->cur == ']')
738	{	1032	{
739	++dec->cur;	1033	++dec->cur;
740	break;	1034	break;
…		…
744	ERR (", or ] expected while parsing array");	1038	ERR (", or ] expected while parsing array");
745		1039
746	++dec->cur;	1040	++dec->cur;
747	}	1041	}
748		1042
		1043	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1044	return newRV_noinc ((SV *)av);
750		1045
751	fail:	1046	fail:
752	SvREFCNT_dec (av);	1047	SvREFCNT_dec (av);
		1048	DEC_DEC_DEPTH;
753	return 0;	1049	return 0;
754	}	1050	}
755		1051
756	static SV *	1052	static SV *
757	decode_hv (dec_t *dec)	1053	decode_hv (dec_t *dec)
758	{	1054	{
		1055	SV *sv;
759	HV *hv = newHV ();	1056	HV *hv = newHV ();
760		1057
761	WS;	1058	DEC_INC_DEPTH;
		1059	decode_ws (dec);
		1060
762	if (*dec->cur == '}')	1061	if (*dec->cur == '}')
763	++dec->cur;	1062	++dec->cur;
764	else	1063	else
765	for (;;)	1064	for (;;)
766	{	1065	{
		1066	decode_ws (dec); EXPECT_CH ('"');
		1067
		1068	// heuristic: assume that
		1069	// a) decode_str + hv_store_ent are abysmally slow.
		1070	// b) most hash keys are short, simple ascii text.
		1071	// => try to "fast-match" such strings to avoid
		1072	// the overhead of decode_str + hv_store_ent.
		1073	{
767	SV *key, *value;	1074	SV *value;
		1075	char *p = dec->cur;
		1076	char *e = p + 24; // only try up to 24 bytes
768		1077
769	WS; EXPECT_CH ('"');	1078	for (;;)
770
771	key = decode_str (dec);
772	if (!key)
773	goto fail;
774
775	WS; EXPECT_CH (':');
776
777	value = decode_sv (dec);
778	if (!value)
779	{	1079	{
		1080	// the >= 0x80 is true on most architectures
		1081	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1082	{
		1083	// slow path, back up and use decode_str
		1084	SV *key = decode_str (dec);
		1085	if (!key)
		1086	goto fail;
		1087
		1088	decode_ws (dec); EXPECT_CH (':');
		1089
		1090	value = decode_sv (dec);
		1091	if (!value)
		1092	{
		1093	SvREFCNT_dec (key);
		1094	goto fail;
		1095	}
		1096
		1097	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1098	SvREFCNT_dec (key);
		1099
		1100	break;
		1101	}
		1102	else if (*p == '"')
		1103	{
		1104	// fast path, got a simple key
		1105	char *key = dec->cur;
		1106	int len = p - key;
		1107	dec->cur = p + 1;
		1108
		1109	decode_ws (dec); EXPECT_CH (':');
		1110
		1111	value = decode_sv (dec);
		1112	if (!value)
781	goto fail;	1113	goto fail;
		1114
		1115	hv_store (hv, key, len, value, 0);
		1116
		1117	break;
		1118	}
		1119
		1120	++p;
782	}	1121	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1122	}
		1123
		1124	decode_ws (dec);
788		1125
789	if (*dec->cur == '}')	1126	if (*dec->cur == '}')
790	{	1127	{
791	++dec->cur;	1128	++dec->cur;
792	break;	1129	break;
…		…
796	ERR (", or } expected while parsing object/hash");	1133	ERR (", or } expected while parsing object/hash");
797		1134
798	++dec->cur;	1135	++dec->cur;
799	}	1136	}
800		1137
		1138	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1139	sv = newRV_noinc ((SV *)hv);
		1140
		1141	// check filter callbacks
		1142	if (dec->json.flags & F_HOOK)
		1143	{
		1144	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1145	{
		1146	HE *cb, *he;
		1147
		1148	hv_iterinit (hv);
		1149	he = hv_iternext (hv);
		1150	hv_iterinit (hv);
		1151
		1152	// the next line creates a mortal sv each time its called.
		1153	// might want to optimise this for common cases.
		1154	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1155
		1156	if (cb)
		1157	{
		1158	dSP;
		1159	int count;
		1160
		1161	ENTER; SAVETMPS; PUSHMARK (SP);
		1162	XPUSHs (HeVAL (he));
		1163
		1164	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1165
		1166	if (count == 1)
		1167	{
		1168	sv = newSVsv (POPs);
		1169	FREETMPS; LEAVE;
		1170	return sv;
		1171	}
		1172
		1173	FREETMPS; LEAVE;
		1174	}
		1175	}
		1176
		1177	if (dec->json.cb_object)
		1178	{
		1179	dSP;
		1180	int count;
		1181
		1182	ENTER; SAVETMPS; PUSHMARK (SP);
		1183	XPUSHs (sv_2mortal (sv));
		1184
		1185	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1186
		1187	if (count == 1)
		1188	{
		1189	sv = newSVsv (POPs);
		1190	FREETMPS; LEAVE;
		1191	return sv;
		1192	}
		1193
		1194	SvREFCNT_inc (sv);
		1195	FREETMPS; LEAVE;
		1196	}
		1197	}
		1198
		1199	return sv;
802		1200
803	fail:	1201	fail:
804	SvREFCNT_dec (hv);	1202	SvREFCNT_dec (hv);
		1203	DEC_DEC_DEPTH;
805	return 0;	1204	return 0;
806	}	1205	}
807		1206
808	static SV *	1207	static SV *
809	decode_sv (dec_t *dec)	1208	decode_sv (dec_t *dec)
810	{	1209	{
811	WS;	1210	decode_ws (dec);
		1211
		1212	// the beauty of JSON: you need exactly one character lookahead
		1213	// to parse anything.
812	switch (*dec->cur)	1214	switch (*dec->cur)
813	{	1215	{
814	case '"': ++dec->cur; return decode_str (dec);	1216	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1217	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1218	case '{': ++dec->cur; return decode_hv (dec);
…		…
822		1224
823	case 't':	1225	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1226	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1227	{
826	dec->cur += 4;	1228	dec->cur += 4;
827	return newSViv (1);	1229	return SvREFCNT_inc (json_true);
828	}	1230	}
829	else	1231	else
830	ERR ("'true' expected");	1232	ERR ("'true' expected");
831		1233
832	break;	1234	break;
833		1235
834	case 'f':	1236	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1237	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1238	{
837	dec->cur += 5;	1239	dec->cur += 5;
838	return newSViv (0);	1240	return SvREFCNT_inc (json_false);
839	}	1241	}
840	else	1242	else
841	ERR ("'false' expected");	1243	ERR ("'false' expected");
842		1244
843	break;	1245	break;
…		…
852	ERR ("'null' expected");	1254	ERR ("'null' expected");
853		1255
854	break;	1256	break;
855		1257
856	default:	1258	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1259	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1260	break;
859	}	1261	}
860		1262
861	fail:	1263	fail:
862	return 0;	1264	return 0;
863	}	1265	}
864		1266
865	static SV *	1267	static SV *
866	decode_json (SV *string, UV flags)	1268	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1269	{
		1270	dec_t dec;
		1271	UV offset;
868	SV *sv;	1272	SV *sv;
869		1273
		1274	SvGETMAGIC (string);
		1275	SvUPGRADE (string, SVt_PV);
		1276
		1277	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1278	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1279	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1280
870	if (flags & F_UTF8)	1281	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1282	sv_utf8_downgrade (string, 0);
872	else	1283	else
873	sv_utf8_upgrade (string);	1284	sv_utf8_upgrade (string);
874		1285
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1286	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1287
877	dec_t dec;	1288	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1289	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1290	dec.end = SvEND (string);
881	dec.err = 0;	1291	dec.err = 0;
		1292	dec.depth = 0;
		1293	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1294
		1295	if (dec.json.cb_object || dec.json.cb_sk_object)
		1296	dec.json.flags |= F_HOOK;
		1297
		1298	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
883	sv = decode_sv (&dec);	1299	sv = decode_sv (&dec);
884		1300
		1301	if (!(offset_return || !sv))
		1302	{
		1303	// check for trailing garbage
		1304	decode_ws (&dec);
		1305
		1306	if (*dec.cur)
		1307	{
		1308	dec.err = "garbage after JSON object";
		1309	SvREFCNT_dec (sv);
		1310	sv = 0;
		1311	}
		1312	}
		1313
		1314	if (offset_return || !sv)
		1315	{
		1316	offset = dec.json.flags & F_UTF8
		1317	? dec.cur - SvPVX (string)
		1318	: utf8_distance (dec.cur, SvPVX (string));
		1319
		1320	if (offset_return)
		1321	*offset_return = offset;
		1322	}
		1323
885	if (!sv)	1324	if (!sv)
886	{	1325	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1326	SV *uni = sv_newmortal ();
891		1327
892	// horrible hack to silence warning inside pv_uni_display	1328	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1329	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1330	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1332	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1333	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1334	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1335	LEAVE;
900		1336
901	croak ("%s, at character offset %d (%s)",	1337	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1338	dec.err,
903	(int)offset,	1339	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1340	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1341	}
906		1342
907	sv = sv_2mortal (sv);	1343	sv = sv_2mortal (sv);
908		1344
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1345	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
910	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1346	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1347
912	return sv;	1348	return sv;
913	}	1349	}
914		1350
		1351	/////////////////////////////////////////////////////////////////////////////
		1352	// XS interface functions
		1353
915	MODULE = JSON::XS PACKAGE = JSON::XS	1354	MODULE = JSON::XS PACKAGE = JSON::XS
916		1355
917	BOOT:	1356	BOOT:
918	{	1357	{
919	int i;	1358	int i;
920		1359
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1360	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1361	decode_hexdigit [i] =
		1362	i >= '0' && i <= '9' ? i - '0'
		1363	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1364	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1365	: -1;
924		1366
925	for (i = 7; i--; )	1367	json_stash = gv_stashpv ("JSON::XS" , 1);
		1368	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1369
		1370	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1371	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1372	}
		1373
		1374	PROTOTYPES: DISABLE
		1375
		1376	void CLONE (...)
		1377	CODE:
		1378	json_stash = 0;
		1379	json_boolean_stash = 0;
		1380
		1381	void new (char *klass)
		1382	PPCODE:
		1383	{
		1384	HV *stash = !JSON_SLOW || json_stash
		1385	? json_stash
		1386	: gv_stashpv ("JSON::XS", 1);
		1387	SV *pv = NEWSV (0, sizeof (JSON));
		1388	SvPOK_only (pv);
		1389	Zero (SvPVX (pv), 1, JSON);
		1390	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1391	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), stash)));
		1392	}
		1393
		1394	void ascii (JSON *self, int enable = 1)
		1395	ALIAS:
		1396	ascii = F_ASCII
		1397	latin1 = F_LATIN1
		1398	utf8 = F_UTF8
		1399	indent = F_INDENT
		1400	canonical = F_CANONICAL
		1401	space_before = F_SPACE_BEFORE
		1402	space_after = F_SPACE_AFTER
		1403	pretty = F_PRETTY
		1404	allow_nonref = F_ALLOW_NONREF
		1405	shrink = F_SHRINK
		1406	allow_blessed = F_ALLOW_BLESSED
		1407	convert_blessed = F_CONV_BLESSED
		1408	PPCODE:
		1409	{
		1410	if (enable)
		1411	self->flags |= ix;
		1412	else
		1413	self->flags &= ~ix;
		1414
		1415	XPUSHs (ST (0));
		1416	}
		1417
		1418	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1419	PPCODE:
		1420	{
		1421	UV log2 = 0;
		1422
		1423	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1424
		1425	while ((1UL << log2) < max_depth)
		1426	++log2;
		1427
		1428	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1429
		1430	XPUSHs (ST (0));
		1431	}
		1432
		1433	void max_size (JSON *self, UV max_size = 0)
		1434	PPCODE:
		1435	{
		1436	UV log2 = 0;
		1437
		1438	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1439	if (max_size == 1) max_size = 2;
		1440
		1441	while ((1UL << log2) < max_size)
		1442	++log2;
		1443
		1444	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1445
		1446	XPUSHs (ST (0));
		1447	}
		1448
		1449	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1450	PPCODE:
		1451	{
		1452	SvREFCNT_dec (self->cb_object);
		1453	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1454
		1455	XPUSHs (ST (0));
		1456	}
		1457
		1458	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1459	PPCODE:
		1460	{
		1461	if (!self->cb_sk_object)
		1462	self->cb_sk_object = newHV ();
		1463
		1464	if (SvOK (cb))
		1465	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1466	else
926	{	1467	{
927	decode_hexdigit ['a' + i] = 10 + i;	1468	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
928	decode_hexdigit ['A' + i] = 10 + i;	1469
		1470	if (!HvKEYS (self->cb_sk_object))
		1471	{
		1472	SvREFCNT_dec (self->cb_sk_object);
		1473	self->cb_sk_object = 0;
		1474	}
929	}	1475	}
930		1476
931	json_stash = gv_stashpv ("JSON::XS", 1);	1477	XPUSHs (ST (0));
932	}	1478	}
933		1479
934	PROTOTYPES: DISABLE	1480	void encode (JSON *self, SV *scalar)
		1481	PPCODE:
		1482	XPUSHs (encode_json (scalar, self));
935		1483
936	SV *new (char *dummy)	1484	void decode (JSON *self, SV *jsonstr)
		1485	PPCODE:
		1486	XPUSHs (decode_json (jsonstr, self, 0));
		1487
		1488	void decode_prefix (JSON *self, SV *jsonstr)
		1489	PPCODE:
		1490	{
		1491	UV offset;
		1492	EXTEND (SP, 2);
		1493	PUSHs (decode_json (jsonstr, self, &offset));
		1494	PUSHs (sv_2mortal (newSVuv (offset)));
		1495	}
		1496
		1497	void DESTROY (JSON *self)
937	CODE:	1498	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1499	SvREFCNT_dec (self->cb_sk_object);
939	OUTPUT:	1500	SvREFCNT_dec (self->cb_object);
940	RETVAL
941
942	SV *ascii (SV *self, int enable = 1)
943	ALIAS:
944	ascii = F_ASCII
945	utf8 = F_UTF8
946	indent = F_INDENT
947	canonical = F_CANONICAL
948	space_before = F_SPACE_BEFORE
949	space_after = F_SPACE_AFTER
950	pretty = F_PRETTY
951	allow_nonref = F_ALLOW_NONREF
952	shrink = F_SHRINK
953	CODE:
954	{
955	UV *uv = SvJSON (self);
956	if (enable)
957	*uv |= ix;
958	else
959	*uv &= ~ix;
960
961	RETVAL = newSVsv (self);
962	}
963	OUTPUT:
964	RETVAL
965
966	void encode (SV *self, SV *scalar)
967	PPCODE:
968	XPUSHs (encode_json (scalar, *SvJSON (self)));
969
970	void decode (SV *self, SV *jsonstr)
971	PPCODE:
972	XPUSHs (decode_json (jsonstr, *SvJSON (self)));
973		1501
974	PROTOTYPES: ENABLE	1502	PROTOTYPES: ENABLE
975		1503
976	void to_json (SV *scalar)	1504	void to_json (SV *scalar)
977	PPCODE:	1505	PPCODE:
		1506	{
		1507	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1508	XPUSHs (encode_json (scalar, &json));
		1509	}
979		1510
980	void from_json (SV *jsonstr)	1511	void from_json (SV *jsonstr)
981	PPCODE:	1512	PPCODE:
		1513	{
		1514	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1515	XPUSHs (decode_json (jsonstr, &json, 0));
		1516	}
983		1517

Diff Legend

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