ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.10 by root, Sat Mar 24 01:15:22 2007 UTC vs.
Revision 1.59 by root, Mon Aug 13 16:14:20 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	# 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
27		121
28	// structure used for encoding JSON	122	// structure used for encoding JSON
29	typedef struct	123	typedef struct
30	{	124	{
31	char *cur;	125	char *cur; // SvPVX (sv) + current output position
32	STRLEN len; // SvLEN (sv)
33	char *end; // SvEND (sv)	126	char *end; // SvEND (sv)
34	SV *sv;	127	SV *sv; // result scalar
35	UV flags;	128	JSON json;
36	int max_recurse;	129	U32 indent; // indentation level
37	int indent;	130	U32 maxdepth; // max. indentation/recursion level
38	} enc_t;	131	} enc_t;
39		132
40	// structure used for decoding JSON	133	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)	134	need (enc_t *enc, STRLEN len)
71	{	135	{
72	if (enc->cur + len >= enc->end)	136	if (expect_false (enc->cur + len >= enc->end))
73	{	137	{
74	STRLEN cur = enc->cur - SvPVX (enc->sv);	138	STRLEN cur = enc->cur - SvPVX (enc->sv);
75	SvGROW (enc->sv, cur + len + 1);	139	SvGROW (enc->sv, cur + len + 1);
76	enc->cur = SvPVX (enc->sv) + cur;	140	enc->cur = SvPVX (enc->sv) + cur;
77	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	141	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
78	}	142	}
79	}	143	}
80		144
81	static void	145	inline void
82	encode_ch (enc_t *enc, char ch)	146	encode_ch (enc_t *enc, char ch)
83	{	147	{
84	need (enc, 1);	148	need (enc, 1);
85	*enc->cur++ = ch;	149	*enc->cur++ = ch;
86	}	150	}
…		…
94		158
95	while (str < end)	159	while (str < end)
96	{	160	{
97	unsigned char ch = *(unsigned char *)str;	161	unsigned char ch = *(unsigned char *)str;
98		162
99	if (ch >= 0x20 && ch < 0x80) // most common case	163	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
100	{	164	{
101	if (ch == '"') // but with slow exceptions	165	if (expect_false (ch == '"')) // but with slow exceptions
102	{	166	{
103	need (enc, len += 1);	167	need (enc, len += 1);
104	*enc->cur++ = '\\';	168	*enc->cur++ = '\\';
105	*enc->cur++ = '"';	169	*enc->cur++ = '"';
106	}	170	}
107	else if (ch == '\\')	171	else if (expect_false (ch == '\\'))
108	{	172	{
109	need (enc, len += 1);	173	need (enc, len += 1);
110	*enc->cur++ = '\\';	174	*enc->cur++ = '\\';
111	*enc->cur++ = '\\';	175	*enc->cur++ = '\\';
112	}	176	}
…		…
130	STRLEN clen;	194	STRLEN clen;
131	UV uch;	195	UV uch;
132		196
133	if (is_utf8)	197	if (is_utf8)
134	{	198	{
135	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	199	uch = decode_utf8 (str, end - str, &clen);
136	if (clen == (STRLEN)-1)	200	if (clen == (STRLEN)-1)
137	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);
138	}	202	}
139	else	203	else
140	{	204	{
…		…
143	}	207	}
144		208
145	if (uch > 0x10FFFFUL)	209	if (uch > 0x10FFFFUL)
146	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);
147		211
148	if (uch < 0x80 || enc->flags & F_ASCII)	212	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
149	{	213	{
150	if (uch > 0xFFFFUL)	214	if (uch > 0xFFFFUL)
151	{	215	{
152	need (enc, len += 11);	216	need (enc, len += 11);
153	sprintf (enc->cur, "\\u%04x\\u%04x",	217	sprintf (enc->cur, "\\u%04x\\u%04x",
…		…
167	*enc->cur++ = hexdigit [(uch >> 0) & 15];	231	*enc->cur++ = hexdigit [(uch >> 0) & 15];
168	}	232	}
169		233
170	str += clen;	234	str += clen;
171	}	235	}
		236	else if (enc->json.flags & F_LATIN1)
		237	{
		238	*enc->cur++ = uch;
		239	str += clen;
		240	}
172	else if (is_utf8)	241	else if (is_utf8)
173	{	242	{
174	need (enc, len += clen);	243	need (enc, len += clen);
175	do	244	do
176	{	245	{
…		…
178	}	247	}
179	while (--clen);	248	while (--clen);
180	}	249	}
181	else	250	else
182	{	251	{
183	need (enc, len += 10); // never more than 11 bytes needed	252	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
184	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	253	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
185	++str;	254	++str;
186	}	255	}
187	}	256	}
188	}	257	}
…		…
190		259
191	--len;	260	--len;
192	}	261	}
193	}	262	}
194		263
195	#define INDENT SB \	264	inline void
		265	encode_indent (enc_t *enc)
		266	{
196	if (enc->flags & F_INDENT) \	267	if (enc->json.flags & F_INDENT)
197	{ \	268	{
198	int i_; \	269	int spaces = enc->indent * INDENT_STEP;
199	need (enc, enc->indent); \	270
200	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);
201	encode_ch (enc, ' '); \	290	encode_ch (enc, '\n');
202	} \	291	}
203	SE	292	}
204		293
205	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	294	inline void
206	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	295	encode_comma (enc_t *enc)
207	#define COMMA SB \	296	{
208	encode_ch (enc, ','); \	297	encode_ch (enc, ',');
		298
209	if (enc->flags & F_INDENT) \	299	if (enc->json.flags & F_INDENT)
210	NL; \	300	encode_nl (enc);
211	else if (enc->flags & F_SPACE_AFTER) \	301	else if (enc->json.flags & F_SPACE_AFTER)
212	SPACE; \	302	encode_space (enc);
213	SE	303	}
214		304
215	static void encode_sv (enc_t *enc, SV *sv);	305	static void encode_sv (enc_t *enc, SV *sv);
216		306
217	static void	307	static void
218	encode_av (enc_t *enc, AV *av)	308	encode_av (enc_t *enc, AV *av)
219	{	309	{
220	int i, len = av_len (av);	310	int i, len = av_len (av);
221		311
222	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);
223	++enc->indent;	316	++enc->indent;
224		317
225	for (i = 0; i <= len; ++i)	318	for (i = 0; i <= len; ++i)
226	{	319	{
227	INDENT;	320	SV **svp = av_fetch (av, i, 0);
228	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);
229		328
230	if (i < len)	329	if (i < len)
231	COMMA;	330	encode_comma (enc);
232	}	331	}
233		332
234	NL;	333	encode_nl (enc);
235		334
236	--enc->indent;	335	--enc->indent;
237	INDENT; encode_ch (enc, ']');	336	encode_indent (enc); encode_ch (enc, ']');
238	}	337	}
239		338
240	static void	339	static void
241	encode_he (enc_t *enc, HE *he)	340	encode_he (enc_t *enc, HE *he)
242	{	341	{
…		…
256	else	355	else
257	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	356	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
258		357
259	encode_ch (enc, '"');	358	encode_ch (enc, '"');
260		359
261	if (enc->flags & F_SPACE_BEFORE) SPACE;	360	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
262	encode_ch (enc, ':');	361	encode_ch (enc, ':');
263	if (enc->flags & F_SPACE_AFTER ) SPACE;	362	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
264	encode_sv (enc, HeVAL (he));	363	encode_sv (enc, HeVAL (he));
265	}	364	}
266		365
267	// compare hash entries, used when all keys are bytestrings	366	// compare hash entries, used when all keys are bytestrings
268	static int	367	static int
…		…
274	HE *b = *(HE **)b_;	373	HE *b = *(HE **)b_;
275		374
276	STRLEN la = HeKLEN (a);	375	STRLEN la = HeKLEN (a);
277	STRLEN lb = HeKLEN (b);	376	STRLEN lb = HeKLEN (b);
278		377
279	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	378	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))
280	cmp = la < lb ? -1 : la == lb ? 0 : 1;	379	cmp = la - lb;
281		380
282	return cmp;	381	return cmp;
283	}	382	}
284		383
285	// compare hash entries, used when some keys are sv's or utf-x	384	// compare hash entries, used when some keys are sv's or utf-x
…		…
292	static void	391	static void
293	encode_hv (enc_t *enc, HV *hv)	392	encode_hv (enc_t *enc, HV *hv)
294	{	393	{
295	int count, i;	394	int count, i;
296		395
		396	if (enc->indent >= enc->maxdepth)
		397	croak ("data structure too deep (hit recursion limit)");
		398
297	encode_ch (enc, '{'); NL; ++enc->indent;	399	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;
298		400
299	if ((count = hv_iterinit (hv)))	401	if ((count = hv_iterinit (hv)))
300	{	402	{
301	// for canonical output we have to sort by keys first	403	// for canonical output we have to sort by keys first
302	// actually, this is mostly due to the stupid so-called	404	// actually, this is mostly due to the stupid so-called
303	// security workaround added somewhere in 5.8.x.	405	// security workaround added somewhere in 5.8.x.
304	// that randomises hash orderings	406	// that randomises hash orderings
305	if (enc->flags & F_CANONICAL)	407	if (enc->json.flags & F_CANONICAL)
306	{	408	{
307	HE *he, *hes [count];
308	int fast = 1;	409	int fast = 1;
		410	HE *he;
		411	#if defined(__BORLANDC__) || defined(_MSC_VER)
		412	HE **hes = _alloca (count * sizeof (HE));
		413	#else
		414	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		415	#endif
309		416
310	i = 0;	417	i = 0;
311	while ((he = hv_iternext (hv)))	418	while ((he = hv_iternext (hv)))
312	{	419	{
313	hes [i++] = he;	420	hes [i++] = he;
…		…
337	LEAVE;	444	LEAVE;
338	}	445	}
339		446
340	for (i = 0; i < count; ++i)	447	for (i = 0; i < count; ++i)
341	{	448	{
342	INDENT;	449	encode_indent (enc);
343	encode_he (enc, hes [i]);	450	encode_he (enc, hes [i]);
344		451
345	if (i < count - 1)	452	if (i < count - 1)
346	COMMA;	453	encode_comma (enc);
347	}
348
349	NL;	454	}
		455
		456	encode_nl (enc);
350	}	457	}
351	else	458	else
352	{	459	{
353	SV *sv;
354	HE *he = hv_iternext (hv);	460	HE *he = hv_iternext (hv);
355		461
356	for (;;)	462	for (;;)
357	{	463	{
358	INDENT;	464	encode_indent (enc);
359	encode_he (enc, he);	465	encode_he (enc, he);
360		466
361	if (!(he = hv_iternext (hv)))	467	if (!(he = hv_iternext (hv)))
362	break;	468	break;
363		469
364	COMMA;	470	encode_comma (enc);
365	}
366
367	NL;	471	}
		472
		473	encode_nl (enc);
368	}	474	}
369	}	475	}
370		476
371	--enc->indent; INDENT; encode_ch (enc, '}');	477	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
		478	}
		479
		480	// encode objects, arrays and special \0=false and \1=true values.
		481	static void
		482	encode_rv (enc_t *enc, SV *sv)
		483	{
		484	svtype svt;
		485
		486	SvGETMAGIC (sv);
		487	svt = SvTYPE (sv);
		488
		489	if (expect_false (SvOBJECT (sv)))
		490	{
		491	HV *stash = !JSON_SLOW || json_boolean_stash
		492	? json_boolean_stash
		493	: gv_stashpv ("JSON::XS::Boolean", 1);
		494
		495	if (SvSTASH (sv) == stash)
		496	{
		497	if (SvIV (sv))
		498	encode_str (enc, "true", 4, 0);
		499	else
		500	encode_str (enc, "false", 5, 0);
		501	}
		502	else
		503	{
		504	#if 0
		505	if (0 && sv_derived_from (rv, "JSON::Literal"))
		506	{
		507	// not yet
		508	}
		509	#endif
		510	if (enc->json.flags & F_CONV_BLESSED)
		511	{
		512	// we re-bless the reference to get overload and other niceties right
		513	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		514
		515	if (to_json)
		516	{
		517	dSP;
		518
		519	ENTER; SAVETMPS; PUSHMARK (SP);
		520	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		521
		522	// calling with G_SCALAR ensures that we always get a 1 return value
		523	PUTBACK;
		524	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		525	SPAGAIN;
		526
		527	// catch this surprisingly common error
		528	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		529	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		530
		531	sv = POPs;
		532	PUTBACK;
		533
		534	encode_sv (enc, sv);
		535
		536	FREETMPS; LEAVE;
		537	}
		538	else if (enc->json.flags & F_ALLOW_BLESSED)
		539	encode_str (enc, "null", 4, 0);
		540	else
		541	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		542	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		543	}
		544	else if (enc->json.flags & F_ALLOW_BLESSED)
		545	encode_str (enc, "null", 4, 0);
		546	else
		547	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		548	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		549	}
		550	}
		551	else if (svt == SVt_PVHV)
		552	encode_hv (enc, (HV *)sv);
		553	else if (svt == SVt_PVAV)
		554	encode_av (enc, (AV *)sv);
		555	else if (svt < SVt_PVAV)
		556	{
		557	STRLEN len = 0;
		558	char *pv = svt ? SvPV (sv, len) : 0;
		559
		560	if (len == 1 && *pv == '1')
		561	encode_str (enc, "true", 4, 0);
		562	else if (len == 1 && *pv == '0')
		563	encode_str (enc, "false", 5, 0);
		564	else
		565	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		566	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		567	}
		568	else
		569	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		570	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
372	}	571	}
373		572
374	static void	573	static void
375	encode_sv (enc_t *enc, SV *sv)	574	encode_sv (enc_t *enc, SV *sv)
376	{	575	{
…		…
384	encode_str (enc, str, len, SvUTF8 (sv));	583	encode_str (enc, str, len, SvUTF8 (sv));
385	encode_ch (enc, '"');	584	encode_ch (enc, '"');
386	}	585	}
387	else if (SvNOKp (sv))	586	else if (SvNOKp (sv))
388	{	587	{
		588	// trust that perl will do the right thing w.r.t. JSON syntax.
389	need (enc, NV_DIG + 32);	589	need (enc, NV_DIG + 32);
390	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	590	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
391	enc->cur += strlen (enc->cur);	591	enc->cur += strlen (enc->cur);
392	}	592	}
393	else if (SvIOKp (sv))	593	else if (SvIOKp (sv))
394	{	594	{
395	need (enc, 64);	595	// we assume we can always read an IV as a UV
		596	if (SvUV (sv) & ~(UV)0x7fff)
		597	{
		598	// large integer, use the (rather slow) snprintf way.
		599	need (enc, sizeof (UV) * 3);
396	enc->cur +=	600	enc->cur +=
397	SvIsUV(sv)	601	SvIsUV(sv)
398	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	602	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
399	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	603	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		604	}
		605	else
		606	{
		607	// optimise the "small number case"
		608	// code will likely be branchless and use only a single multiplication
		609	I32 i = SvIV (sv);
		610	U32 u;
		611	char digit, nz = 0;
		612
		613	need (enc, 6);
		614
		615	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		616	u = i < 0 ? -i : i;
		617
		618	// convert to 4.28 fixed-point representation
		619	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		620
		621	// now output digit by digit, each time masking out the integer part
		622	// and multiplying by 5 while moving the decimal point one to the right,
		623	// resulting in a net multiplication by 10.
		624	// we always write the digit to memory but conditionally increment
		625	// the pointer, to ease the usage of conditional move instructions.
		626	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		627	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		628	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		629	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		630	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		631	}
400	}	632	}
401	else if (SvROK (sv))	633	else if (SvROK (sv))
402	{	634	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))	635	else if (!SvOK (sv))
419	encode_str (enc, "null", 4, 0);	636	encode_str (enc, "null", 4, 0);
420	else	637	else
421	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	638	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
422	SvPV_nolen (sv), SvFLAGS (sv));	639	SvPV_nolen (sv), SvFLAGS (sv));
423	}	640	}
424		641
425	static SV *	642	static SV *
426	encode_json (SV *scalar, UV flags)	643	encode_json (SV *scalar, JSON *json)
427	{	644	{
		645	enc_t enc;
		646
428	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	647	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
429	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	648	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
430		649
431	enc_t enc;	650	enc.json = *json;
432	enc.flags = flags;
433	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	651	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
434	enc.cur = SvPVX (enc.sv);	652	enc.cur = SvPVX (enc.sv);
435	enc.end = SvEND (enc.sv);	653	enc.end = SvEND (enc.sv);
436	enc.max_recurse = 0;
437	enc.indent = 0;	654	enc.indent = 0;
		655	enc.maxdepth = DEC_DEPTH (enc.json.flags);
438		656
439	SvPOK_only (enc.sv);	657	SvPOK_only (enc.sv);
440	encode_sv (&enc, scalar);	658	encode_sv (&enc, scalar);
441		659
		660	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		661	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		662
442	if (!(flags & (F_ASCII | F_UTF8)))	663	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
443	SvUTF8_on (enc.sv);	664	SvUTF8_on (enc.sv);
444		665
445	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
446
447	if (enc.flags & F_SHRINK)	666	if (enc.json.flags & F_SHRINK)
448	shrink (enc.sv);	667	shrink (enc.sv);
449		668
450	return enc.sv;	669	return enc.sv;
451	}	670	}
452		671
453	/////////////////////////////////////////////////////////////////////////////	672	/////////////////////////////////////////////////////////////////////////////
		673	// decoder
454		674
455	#define WS \	675	// structure used for decoding JSON
		676	typedef struct
		677	{
		678	char *cur; // current parser pointer
		679	char *end; // end of input string
		680	const char *err; // parse error, if != 0
		681	JSON json;
		682	U32 depth; // recursion depth
		683	U32 maxdepth; // recursion depth limit
		684	} dec_t;
		685
		686	inline void
		687	decode_ws (dec_t *dec)
		688	{
456	for (;;) \	689	for (;;)
457	{ \	690	{
458	char ch = *dec->cur; \	691	char ch = *dec->cur;
		692
459	if (ch > 0x20 \	693	if (ch > 0x20
460	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	694	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))
461	break; \	695	break;
		696
462	++dec->cur; \	697	++dec->cur;
463	}	698	}
		699	}
464		700
465	#define ERR(reason) SB dec->err = reason; goto fail; SE	701	#define ERR(reason) SB dec->err = reason; goto fail; SE
		702
466	#define EXPECT_CH(ch) SB \	703	#define EXPECT_CH(ch) SB \
467	if (*dec->cur != ch) \	704	if (*dec->cur != ch) \
468	ERR (# ch " expected"); \	705	ERR (# ch " expected"); \
469	++dec->cur; \	706	++dec->cur; \
470	SE	707	SE
471		708
		709	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		710	#define DEC_DEC_DEPTH --dec->depth
		711
472	static SV *decode_sv (dec_t *dec);	712	static SV *decode_sv (dec_t *dec);
473		713
474	static signed char decode_hexdigit[256];	714	static signed char decode_hexdigit[256];
475		715
476	static UV	716	static UV
477	decode_4hex (dec_t *dec)	717	decode_4hex (dec_t *dec)
478	{	718	{
479	signed char d1, d2, d3, d4;	719	signed char d1, d2, d3, d4;
		720	unsigned char *cur = (unsigned char *)dec->cur;
480		721
481	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	722	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
482	if (d1 < 0) ERR ("four hexadecimal digits expected");	723	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
483	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	724	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
484	if (d2 < 0) ERR ("four hexadecimal digits expected");	725	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
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		726
490	dec->cur += 4;	727	dec->cur += 4;
491		728
492	return ((UV)d1) << 12	729	return ((UV)d1) << 12
493	| ((UV)d2) << 8	730	| ((UV)d2) << 8
…		…
496		733
497	fail:	734	fail:
498	return (UV)-1;	735	return (UV)-1;
499	}	736	}
500		737
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 *	738	static SV *
517	decode_str (dec_t *dec)	739	decode_str (dec_t *dec)
518	{	740	{
519	SV *sv = NEWSV (0,2);	741	SV *sv = 0;
520	int utf8 = 0;	742	int utf8 = 0;
521	char *cur = SvPVX (sv);	743	char *dec_cur = dec->cur;
522	char *end = SvEND (sv);
523		744
524	for (;;)	745	do
525	{	746	{
526	unsigned char ch = *(unsigned char *)dec->cur;	747	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		748	char *cur = buf;
527		749
528	if (ch == '"')	750	do
529	break;
530	else if (ch == '\\')
531	{	751	{
532	switch (*++dec->cur)	752	unsigned char ch = *(unsigned char *)dec_cur++;
		753
		754	if (expect_false (ch == '"'))
		755	{
		756	--dec_cur;
		757	break;
533	{	758	}
534	case '\\':	759	else if (expect_false (ch == '\\'))
535	case '/':	760	{
536	case '"': APPEND_CH (*dec->cur++); break;	761	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	{	762	{
546	UV lo, hi;	763	case '\\':
547	++dec->cur;	764	case '/':
		765	case '"': *cur++ = *dec_cur++; break;
548		766
549	hi = decode_4hex (dec);	767	case 'b': ++dec_cur; *cur++ = '\010'; break;
550	if (hi == (UV)-1)	768	case 't': ++dec_cur; *cur++ = '\011'; break;
551	goto fail;	769	case 'n': ++dec_cur; *cur++ = '\012'; break;
		770	case 'f': ++dec_cur; *cur++ = '\014'; break;
		771	case 'r': ++dec_cur; *cur++ = '\015'; break;
552		772
553	// possibly a surrogate pair	773	case 'u':
554	if (hi >= 0xd800 && hi < 0xdc00)
555	{	774	{
556	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	775	UV lo, hi;
557	ERR ("missing low surrogate character in surrogate pair");	776	++dec_cur;
558		777
559	dec->cur += 2;	778	dec->cur = dec_cur;
560
561	lo = decode_4hex (dec);	779	hi = decode_4hex (dec);
		780	dec_cur = dec->cur;
562	if (lo == (UV)-1)	781	if (hi == (UV)-1)
563	goto fail;	782	goto fail;
564		783
		784	// possibly a surrogate pair
		785	if (hi >= 0xd800)
		786	if (hi < 0xdc00)
		787	{
		788	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		789	ERR ("missing low surrogate character in surrogate pair");
		790
		791	dec_cur += 2;
		792
		793	dec->cur = dec_cur;
		794	lo = decode_4hex (dec);
		795	dec_cur = dec->cur;
		796	if (lo == (UV)-1)
		797	goto fail;
		798
565	if (lo < 0xdc00 || lo >= 0xe000)	799	if (lo < 0xdc00 || lo >= 0xe000)
566	ERR ("surrogate pair expected");	800	ERR ("surrogate pair expected");
567		801
568	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	802	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		803	}
		804	else if (hi < 0xe000)
		805	ERR ("missing high surrogate character in surrogate pair");
		806
		807	if (hi >= 0x80)
		808	{
		809	utf8 = 1;
		810
		811	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		812	}
		813	else
		814	*cur++ = hi;
569	}	815	}
570	else if (hi >= 0xdc00 && hi < 0xe000)
571	ERR ("missing high surrogate character in surrogate pair");
572
573	if (hi >= 0x80)
574	{	816	break;
575	utf8 = 1;
576		817
577	APPEND_GROW (4); // at most 4 bytes for 21 bits
578	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
579	}
580	else	818	default:
581	APPEND_CH (hi);	819	--dec_cur;
		820	ERR ("illegal backslash escape sequence in string");
582	}	821	}
583	break;	822	}
		823	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		824	*cur++ = ch;
		825	else if (ch >= 0x80)
		826	{
		827	STRLEN clen;
		828	UV uch;
584		829
585	default:
586	--dec->cur;	830	--dec_cur;
587	ERR ("illegal backslash escape sequence in string");	831
		832	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		833	if (clen == (STRLEN)-1)
		834	ERR ("malformed UTF-8 character in JSON string");
		835
		836	do
		837	*cur++ = *dec_cur++;
		838	while (--clen);
		839
		840	utf8 = 1;
		841	}
		842	else
		843	{
		844	--dec_cur;
		845
		846	if (!ch)
		847	ERR ("unexpected end of string while parsing JSON string");
		848	else
		849	ERR ("invalid character encountered while parsing JSON string");
588	}	850	}
589	}	851	}
590	else if (ch >= 0x20 && ch <= 0x7f)	852	while (cur < buf + SHORT_STRING_LEN);
591	APPEND_CH (*dec->cur++);	853
592	else if (ch >= 0x80)
593	{	854	{
594	STRLEN clen;	855	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		856
599	APPEND_GROW (clen);	857	if (sv)
600	do
601	{	858	{
602	*cur++ = *dec->cur++;	859	SvGROW (sv, SvCUR (sv) + len + 1);
		860	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		861	SvCUR_set (sv, SvCUR (sv) + len);
603	}	862	}
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	863	else
611	ERR ("invalid character encountered");	864	sv = newSVpvn (buf, len);
612	}	865	}
		866	}
		867	while (*dec_cur != '"');
613		868
614	++dec->cur;	869	++dec_cur;
615		870
616	SvCUR_set (sv, cur - SvPVX (sv));	871	if (sv)
617		872	{
618	SvPOK_only (sv);	873	SvPOK_only (sv);
619	*SvEND (sv) = 0;	874	*SvEND (sv) = 0;
620		875
621	if (utf8)	876	if (utf8)
622	SvUTF8_on (sv);	877	SvUTF8_on (sv);
		878	}
		879	else
		880	sv = newSVpvn ("", 0);
623		881
624	if (dec->flags & F_SHRINK)	882	dec->cur = dec_cur;
625	shrink (sv);
626
627	return sv;	883	return sv;
628		884
629	fail:	885	fail:
630	SvREFCNT_dec (sv);	886	dec->cur = dec_cur;
631	return 0;	887	return 0;
632	}	888	}
633		889
634	static SV *	890	static SV *
635	decode_num (dec_t *dec)	891	decode_num (dec_t *dec)
…		…
693	is_nv = 1;	949	is_nv = 1;
694	}	950	}
695		951
696	if (!is_nv)	952	if (!is_nv)
697	{	953	{
698	UV uv;	954	int len = dec->cur - start;
699	int numtype = grok_number (start, dec->cur - start, &uv);	955
700	if (numtype & IS_NUMBER_IN_UV)	956	// special case the rather common 1..4-digit-int case, assumes 32 bit ints or so
701	if (numtype & IS_NUMBER_NEG)	957	if (*start == '-')
		958	switch (len)
702	{	959	{
703	if (uv < (UV)IV_MIN)	960	case 2: return newSViv (-( start [1] - '0' * 1));
704	return newSViv (-(IV)uv);	961	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		962	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		963	case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
705	}	964	}
		965	else
		966	switch (len)
		967	{
		968	case 1: return newSViv ( start [0] - '0' * 1);
		969	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		970	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		971	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		972	}
		973
		974	{
		975	UV uv;
		976	int numtype = grok_number (start, len, &uv);
		977	if (numtype & IS_NUMBER_IN_UV)
		978	if (numtype & IS_NUMBER_NEG)
		979	{
		980	if (uv < (UV)IV_MIN)
		981	return newSViv (-(IV)uv);
		982	}
706	else	983	else
707	return newSVuv (uv);	984	return newSVuv (uv);
708	}	985	}
709		986
		987	len -= *start == '-' ? 1 : 0;
		988
		989	// does not fit into IV or UV, try NV
		990	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		991	#if defined (LDBL_DIG)
		992	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		993	#endif
		994	)
		995	// fits into NV without loss of precision
		996	return newSVnv (Atof (start));
		997
		998	// everything else fails, convert it to a string
		999	return newSVpvn (start, dec->cur - start);
		1000	}
		1001
		1002	// loss of precision here
710	return newSVnv (Atof (start));	1003	return newSVnv (Atof (start));
711		1004
712	fail:	1005	fail:
713	return 0;	1006	return 0;
714	}	1007	}
…		…
716	static SV *	1009	static SV *
717	decode_av (dec_t *dec)	1010	decode_av (dec_t *dec)
718	{	1011	{
719	AV *av = newAV ();	1012	AV *av = newAV ();
720		1013
721	WS;	1014	DEC_INC_DEPTH;
		1015	decode_ws (dec);
		1016
722	if (*dec->cur == ']')	1017	if (*dec->cur == ']')
723	++dec->cur;	1018	++dec->cur;
724	else	1019	else
725	for (;;)	1020	for (;;)
726	{	1021	{
…		…
730	if (!value)	1025	if (!value)
731	goto fail;	1026	goto fail;
732		1027
733	av_push (av, value);	1028	av_push (av, value);
734		1029
735	WS;	1030	decode_ws (dec);
736		1031
737	if (*dec->cur == ']')	1032	if (*dec->cur == ']')
738	{	1033	{
739	++dec->cur;	1034	++dec->cur;
740	break;	1035	break;
…		…
744	ERR (", or ] expected while parsing array");	1039	ERR (", or ] expected while parsing array");
745		1040
746	++dec->cur;	1041	++dec->cur;
747	}	1042	}
748		1043
		1044	DEC_DEC_DEPTH;
749	return newRV_noinc ((SV *)av);	1045	return newRV_noinc ((SV *)av);
750		1046
751	fail:	1047	fail:
752	SvREFCNT_dec (av);	1048	SvREFCNT_dec (av);
		1049	DEC_DEC_DEPTH;
753	return 0;	1050	return 0;
754	}	1051	}
755		1052
756	static SV *	1053	static SV *
757	decode_hv (dec_t *dec)	1054	decode_hv (dec_t *dec)
758	{	1055	{
		1056	SV *sv;
759	HV *hv = newHV ();	1057	HV *hv = newHV ();
760		1058
761	WS;	1059	DEC_INC_DEPTH;
		1060	decode_ws (dec);
		1061
762	if (*dec->cur == '}')	1062	if (*dec->cur == '}')
763	++dec->cur;	1063	++dec->cur;
764	else	1064	else
765	for (;;)	1065	for (;;)
766	{	1066	{
		1067	decode_ws (dec); EXPECT_CH ('"');
		1068
		1069	// heuristic: assume that
		1070	// a) decode_str + hv_store_ent are abysmally slow.
		1071	// b) most hash keys are short, simple ascii text.
		1072	// => try to "fast-match" such strings to avoid
		1073	// the overhead of decode_str + hv_store_ent.
		1074	{
767	SV *key, *value;	1075	SV *value;
		1076	char *p = dec->cur;
		1077	char *e = p + 24; // only try up to 24 bytes
768		1078
769	WS; EXPECT_CH ('"');	1079	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	{	1080	{
		1081	// the >= 0x80 is true on most architectures
		1082	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1083	{
		1084	// slow path, back up and use decode_str
		1085	SV *key = decode_str (dec);
		1086	if (!key)
		1087	goto fail;
		1088
		1089	decode_ws (dec); EXPECT_CH (':');
		1090
		1091	value = decode_sv (dec);
		1092	if (!value)
		1093	{
		1094	SvREFCNT_dec (key);
		1095	goto fail;
		1096	}
		1097
		1098	hv_store_ent (hv, key, value, 0);
780	SvREFCNT_dec (key);	1099	SvREFCNT_dec (key);
		1100
		1101	break;
		1102	}
		1103	else if (*p == '"')
		1104	{
		1105	// fast path, got a simple key
		1106	char *key = dec->cur;
		1107	int len = p - key;
		1108	dec->cur = p + 1;
		1109
		1110	decode_ws (dec); EXPECT_CH (':');
		1111
		1112	value = decode_sv (dec);
		1113	if (!value)
781	goto fail;	1114	goto fail;
		1115
		1116	hv_store (hv, key, len, value, 0);
		1117
		1118	break;
		1119	}
		1120
		1121	++p;
782	}	1122	}
783
784	//TODO: optimise
785	hv_store_ent (hv, key, value, 0);
786
787	WS;	1123	}
		1124
		1125	decode_ws (dec);
788		1126
789	if (*dec->cur == '}')	1127	if (*dec->cur == '}')
790	{	1128	{
791	++dec->cur;	1129	++dec->cur;
792	break;	1130	break;
…		…
796	ERR (", or } expected while parsing object/hash");	1134	ERR (", or } expected while parsing object/hash");
797		1135
798	++dec->cur;	1136	++dec->cur;
799	}	1137	}
800		1138
		1139	DEC_DEC_DEPTH;
801	return newRV_noinc ((SV *)hv);	1140	sv = newRV_noinc ((SV *)hv);
		1141
		1142	// check filter callbacks
		1143	if (dec->json.flags & F_HOOK)
		1144	{
		1145	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1146	{
		1147	HE *cb, *he;
		1148
		1149	hv_iterinit (hv);
		1150	he = hv_iternext (hv);
		1151	hv_iterinit (hv);
		1152
		1153	// the next line creates a mortal sv each time its called.
		1154	// might want to optimise this for common cases.
		1155	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1156
		1157	if (cb)
		1158	{
		1159	dSP;
		1160	int count;
		1161
		1162	ENTER; SAVETMPS; PUSHMARK (SP);
		1163	XPUSHs (HeVAL (he));
		1164
		1165	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1166
		1167	if (count == 1)
		1168	{
		1169	sv = newSVsv (POPs);
		1170	FREETMPS; LEAVE;
		1171	return sv;
		1172	}
		1173
		1174	FREETMPS; LEAVE;
		1175	}
		1176	}
		1177
		1178	if (dec->json.cb_object)
		1179	{
		1180	dSP;
		1181	int count;
		1182
		1183	ENTER; SAVETMPS; PUSHMARK (SP);
		1184	XPUSHs (sv_2mortal (sv));
		1185
		1186	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1187
		1188	if (count == 1)
		1189	{
		1190	sv = newSVsv (POPs);
		1191	FREETMPS; LEAVE;
		1192	return sv;
		1193	}
		1194
		1195	SvREFCNT_inc (sv);
		1196	FREETMPS; LEAVE;
		1197	}
		1198	}
		1199
		1200	return sv;
802		1201
803	fail:	1202	fail:
804	SvREFCNT_dec (hv);	1203	SvREFCNT_dec (hv);
		1204	DEC_DEC_DEPTH;
805	return 0;	1205	return 0;
806	}	1206	}
807		1207
808	static SV *	1208	static SV *
809	decode_sv (dec_t *dec)	1209	decode_sv (dec_t *dec)
810	{	1210	{
811	WS;	1211	decode_ws (dec);
		1212
		1213	// the beauty of JSON: you need exactly one character lookahead
		1214	// to parse anything.
812	switch (*dec->cur)	1215	switch (*dec->cur)
813	{	1216	{
814	case '"': ++dec->cur; return decode_str (dec);	1217	case '"': ++dec->cur; return decode_str (dec);
815	case '[': ++dec->cur; return decode_av (dec);	1218	case '[': ++dec->cur; return decode_av (dec);
816	case '{': ++dec->cur; return decode_hv (dec);	1219	case '{': ++dec->cur; return decode_hv (dec);
…		…
822		1225
823	case 't':	1226	case 't':
824	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1227	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
825	{	1228	{
826	dec->cur += 4;	1229	dec->cur += 4;
827	return newSViv (1);	1230	return SvREFCNT_inc (json_true);
828	}	1231	}
829	else	1232	else
830	ERR ("'true' expected");	1233	ERR ("'true' expected");
831		1234
832	break;	1235	break;
833		1236
834	case 'f':	1237	case 'f':
835	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1238	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
836	{	1239	{
837	dec->cur += 5;	1240	dec->cur += 5;
838	return newSViv (0);	1241	return SvREFCNT_inc (json_false);
839	}	1242	}
840	else	1243	else
841	ERR ("'false' expected");	1244	ERR ("'false' expected");
842		1245
843	break;	1246	break;
…		…
852	ERR ("'null' expected");	1255	ERR ("'null' expected");
853		1256
854	break;	1257	break;
855		1258
856	default:	1259	default:
857	ERR ("malformed json string, neither array, object, number, string or atom");	1260	ERR ("malformed JSON string, neither array, object, number, string or atom");
858	break;	1261	break;
859	}	1262	}
860		1263
861	fail:	1264	fail:
862	return 0;	1265	return 0;
863	}	1266	}
864		1267
865	static SV *	1268	static SV *
866	decode_json (SV *string, UV flags)	1269	decode_json (SV *string, JSON *json, UV *offset_return)
867	{	1270	{
		1271	dec_t dec;
		1272	UV offset;
868	SV *sv;	1273	SV *sv;
869		1274
		1275	SvGETMAGIC (string);
		1276	SvUPGRADE (string, SVt_PV);
		1277
		1278	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1279	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1280	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1281
870	if (flags & F_UTF8)	1282	if (json->flags & F_UTF8)
871	sv_utf8_downgrade (string, 0);	1283	sv_utf8_downgrade (string, 0);
872	else	1284	else
873	sv_utf8_upgrade (string);	1285	sv_utf8_upgrade (string);
874		1286
875	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1287	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
876		1288
877	dec_t dec;	1289	dec.json = *json;
878	dec.flags = flags;
879	dec.cur = SvPVX (string);	1290	dec.cur = SvPVX (string);
880	dec.end = SvEND (string);	1291	dec.end = SvEND (string);
881	dec.err = 0;	1292	dec.err = 0;
		1293	dec.depth = 0;
		1294	dec.maxdepth = DEC_DEPTH (dec.json.flags);
882		1295
		1296	if (dec.json.cb_object || dec.json.cb_sk_object)
		1297	dec.json.flags |= F_HOOK;
		1298
		1299	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
883	sv = decode_sv (&dec);	1300	sv = decode_sv (&dec);
884		1301
		1302	if (!(offset_return || !sv))
		1303	{
		1304	// check for trailing garbage
		1305	decode_ws (&dec);
		1306
		1307	if (*dec.cur)
		1308	{
		1309	dec.err = "garbage after JSON object";
		1310	SvREFCNT_dec (sv);
		1311	sv = 0;
		1312	}
		1313	}
		1314
		1315	if (offset_return || !sv)
		1316	{
		1317	offset = dec.json.flags & F_UTF8
		1318	? dec.cur - SvPVX (string)
		1319	: utf8_distance (dec.cur, SvPVX (string));
		1320
		1321	if (offset_return)
		1322	*offset_return = offset;
		1323	}
		1324
885	if (!sv)	1325	if (!sv)
886	{	1326	{
887	IV offset = dec.flags & F_UTF8
888	? dec.cur - SvPVX (string)
889	: utf8_distance (dec.cur, SvPVX (string));
890	SV *uni = sv_newmortal ();	1327	SV *uni = sv_newmortal ();
891		1328
892	// horrible hack to silence warning inside pv_uni_display	1329	// horrible hack to silence warning inside pv_uni_display
893	COP cop = *PL_curcop;	1330	COP cop = *PL_curcop;
894	cop.cop_warnings = pWARN_NONE;	1331	cop.cop_warnings = pWARN_NONE;
…		…
896	SAVEVPTR (PL_curcop);	1333	SAVEVPTR (PL_curcop);
897	PL_curcop = &cop;	1334	PL_curcop = &cop;
898	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1335	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
899	LEAVE;	1336	LEAVE;
900		1337
901	croak ("%s, at character offset %d (%s)",	1338	croak ("%s, at character offset %d [\"%s\"]",
902	dec.err,	1339	dec.err,
903	(int)offset,	1340	(int)offset,
904	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1341	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
905	}	1342	}
906		1343
907	sv = sv_2mortal (sv);	1344	sv = sv_2mortal (sv);
908		1345
909	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1346	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)");	1347	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
911		1348
912	return sv;	1349	return sv;
913	}	1350	}
914		1351
		1352	/////////////////////////////////////////////////////////////////////////////
		1353	// XS interface functions
		1354
915	MODULE = JSON::XS PACKAGE = JSON::XS	1355	MODULE = JSON::XS PACKAGE = JSON::XS
916		1356
917	BOOT:	1357	BOOT:
918	{	1358	{
919	int i;	1359	int i;
920		1360
921	memset (decode_hexdigit, 0xff, 256);
922	for (i = 10; i--; )	1361	for (i = 0; i < 256; ++i)
923	decode_hexdigit ['0' + i] = i;	1362	decode_hexdigit [i] =
		1363	i >= '0' && i <= '9' ? i - '0'
		1364	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1365	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1366	: -1;
924		1367
925	for (i = 7; i--; )	1368	json_stash = gv_stashpv ("JSON::XS" , 1);
		1369	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1370
		1371	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1372	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1373	}
		1374
		1375	PROTOTYPES: DISABLE
		1376
		1377	void CLONE (...)
		1378	CODE:
		1379	json_stash = 0;
		1380	json_boolean_stash = 0;
		1381
		1382	void new (char *klass)
		1383	PPCODE:
		1384	{
		1385	SV *pv = NEWSV (0, sizeof (JSON));
		1386	SvPOK_only (pv);
		1387	Zero (SvPVX (pv), 1, JSON);
		1388	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1389	XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH)));
		1390	}
		1391
		1392	void ascii (JSON *self, int enable = 1)
		1393	ALIAS:
		1394	ascii = F_ASCII
		1395	latin1 = F_LATIN1
		1396	utf8 = F_UTF8
		1397	indent = F_INDENT
		1398	canonical = F_CANONICAL
		1399	space_before = F_SPACE_BEFORE
		1400	space_after = F_SPACE_AFTER
		1401	pretty = F_PRETTY
		1402	allow_nonref = F_ALLOW_NONREF
		1403	shrink = F_SHRINK
		1404	allow_blessed = F_ALLOW_BLESSED
		1405	convert_blessed = F_CONV_BLESSED
		1406	PPCODE:
		1407	{
		1408	if (enable)
		1409	self->flags |= ix;
		1410	else
		1411	self->flags &= ~ix;
		1412
		1413	XPUSHs (ST (0));
		1414	}
		1415
		1416	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1417	PPCODE:
		1418	{
		1419	UV log2 = 0;
		1420
		1421	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1422
		1423	while ((1UL << log2) < max_depth)
		1424	++log2;
		1425
		1426	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1427
		1428	XPUSHs (ST (0));
		1429	}
		1430
		1431	void max_size (JSON *self, UV max_size = 0)
		1432	PPCODE:
		1433	{
		1434	UV log2 = 0;
		1435
		1436	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
		1437	if (max_size == 1) max_size = 2;
		1438
		1439	while ((1UL << log2) < max_size)
		1440	++log2;
		1441
		1442	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1443
		1444	XPUSHs (ST (0));
		1445	}
		1446
		1447	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1448	PPCODE:
		1449	{
		1450	SvREFCNT_dec (self->cb_object);
		1451	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1452
		1453	XPUSHs (ST (0));
		1454	}
		1455
		1456	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1457	PPCODE:
		1458	{
		1459	if (!self->cb_sk_object)
		1460	self->cb_sk_object = newHV ();
		1461
		1462	if (SvOK (cb))
		1463	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1464	else
926	{	1465	{
927	decode_hexdigit ['a' + i] = 10 + i;	1466	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
928	decode_hexdigit ['A' + i] = 10 + i;	1467
		1468	if (!HvKEYS (self->cb_sk_object))
		1469	{
		1470	SvREFCNT_dec (self->cb_sk_object);
		1471	self->cb_sk_object = 0;
		1472	}
929	}	1473	}
930		1474
931	json_stash = gv_stashpv ("JSON::XS", 1);	1475	XPUSHs (ST (0));
932	}	1476	}
933		1477
934	PROTOTYPES: DISABLE	1478	void encode (JSON *self, SV *scalar)
		1479	PPCODE:
		1480	XPUSHs (encode_json (scalar, self));
935		1481
936	SV *new (char *dummy)	1482	void decode (JSON *self, SV *jsonstr)
		1483	PPCODE:
		1484	XPUSHs (decode_json (jsonstr, self, 0));
		1485
		1486	void decode_prefix (JSON *self, SV *jsonstr)
		1487	PPCODE:
		1488	{
		1489	UV offset;
		1490	EXTEND (SP, 2);
		1491	PUSHs (decode_json (jsonstr, self, &offset));
		1492	PUSHs (sv_2mortal (newSVuv (offset)));
		1493	}
		1494
		1495	void DESTROY (JSON *self)
937	CODE:	1496	CODE:
938	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1497	SvREFCNT_dec (self->cb_sk_object);
939	OUTPUT:	1498	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		1499
974	PROTOTYPES: ENABLE	1500	PROTOTYPES: ENABLE
975		1501
976	void to_json (SV *scalar)	1502	void to_json (SV *scalar)
977	PPCODE:	1503	PPCODE:
		1504	{
		1505	JSON json = { F_DEFAULT | F_UTF8 };
978	XPUSHs (encode_json (scalar, F_UTF8));	1506	XPUSHs (encode_json (scalar, &json));
		1507	}
979		1508
980	void from_json (SV *jsonstr)	1509	void from_json (SV *jsonstr)
981	PPCODE:	1510	PPCODE:
		1511	{
		1512	JSON json = { F_DEFAULT | F_UTF8 };
982	XPUSHs (decode_json (jsonstr, F_UTF8));	1513	XPUSHs (decode_json (jsonstr, &json, 0));
		1514	}
983		1515

Diff Legend

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