ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.8 by root, Fri Mar 23 16:13:59 2007 UTC vs.
Revision 1.70 by root, Wed Mar 19 00:44:54 2008 UTC

1	#include "EXTERN.h"	1	#include "EXTERN.h"
2	#include "perl.h"	2	#include "perl.h"
3	#include "XSUB.h"	3	#include "XSUB.h"
4		4
5	#include "assert.h"	5	#include <assert.h>
6	#include "string.h"	6	#include <string.h>
7	#include "stdlib.h"	7	#include <stdlib.h>
		8	#include <stdio.h>
		9	#include <float.h>
8		10
		11	#if defined(__BORLANDC__) || defined(_MSC_VER)
		12	# define snprintf _snprintf // C compilers have this in stdio.h
		13	#endif
		14
		15	// some old perls do not have this, try to make it work, no
		16	// guarentees, though. if it breaks, you get to keep the pieces.
		17	#ifndef UTF8_MAXBYTES
		18	# define UTF8_MAXBYTES 13
		19	#endif
		20
9	#define F_ASCII 0x00000001	21	#define F_ASCII 0x00000001UL
		22	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	23	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	24	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	25	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	26	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	27	#define F_SPACE_AFTER 0x00000040UL
15	#define F_JSON_RPC 0x00000040
16	#define F_ALLOW_NONREF 0x00000080	28	#define F_ALLOW_NONREF 0x00000100UL
17	#define F_SHRINK 0x00000100	29	#define F_SHRINK 0x00000200UL
		30	#define F_ALLOW_BLESSED 0x00000400UL
		31	#define F_CONV_BLESSED 0x00000800UL
		32	#define F_RELAXED 0x00001000UL
		33
		34	#define F_MAXDEPTH 0xf8000000UL
		35	#define S_MAXDEPTH 27
		36	#define F_MAXSIZE 0x01f00000UL
		37	#define S_MAXSIZE 20
		38	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		39
		40	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		41	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
18		42
19	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	43	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
20	#define F_DEFAULT 0	44	#define F_DEFAULT (9UL << S_MAXDEPTH)
21		45
22	#define INIT_SIZE 32 // initial scalar size to be allocated	46	#define INIT_SIZE 32 // initial scalar size to be allocated
		47	#define INDENT_STEP 3 // spaces per indentation level
		48
		49	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
23		50
24	#define SB do {	51	#define SB do {
25	#define SE } while (0)	52	#define SE } while (0)
26		53
27	static HV *json_stash;	54	#if __GNUC__ >= 3
		55	# define expect(expr,value) __builtin_expect ((expr),(value))
		56	# define inline inline
		57	#else
		58	# define expect(expr,value) (expr)
		59	# define inline static
		60	#endif
		61
		62	#define expect_false(expr) expect ((expr) != 0, 0)
		63	#define expect_true(expr) expect ((expr) != 0, 1)
		64
		65	#ifdef USE_ITHREADS
		66	# define JSON_SLOW 1
		67	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		68	#else
		69	# define JSON_SLOW 0
		70	# define JSON_STASH json_stash
		71	#endif
		72
		73	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		74	static SV *json_true, *json_false;
		75
		76	typedef struct {
		77	U32 flags;
		78	SV *cb_object;
		79	HV *cb_sk_object;
		80	} JSON;
		81
		82	/////////////////////////////////////////////////////////////////////////////
		83	// utility functions
		84
		85	inline void
		86	shrink (SV *sv)
		87	{
		88	sv_utf8_downgrade (sv, 1);
		89	if (SvLEN (sv) > SvCUR (sv) + 1)
		90	{
		91	#ifdef SvPV_shrink_to_cur
		92	SvPV_shrink_to_cur (sv);
		93	#elif defined (SvPV_renew)
		94	SvPV_renew (sv, SvCUR (sv) + 1);
		95	#endif
		96	}
		97	}
		98
		99	// decode an utf-8 character and return it, or (UV)-1 in
		100	// case of an error.
		101	// we special-case "safe" characters from U+80 .. U+7FF,
		102	// but use the very good perl function to parse anything else.
		103	// note that we never call this function for a ascii codepoints
		104	inline UV
		105	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		106	{
		107	if (expect_false (s[0] > 0xdf || s[0] < 0xc2))
		108	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		109	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)
		110	{
		111	*clen = 2;
		112	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		113	}
		114	else
		115	{
		116	*clen = (STRLEN)-1;
		117	return (UV)-1;
		118	}
		119	}
		120
		121	/////////////////////////////////////////////////////////////////////////////
		122	// encoder
28		123
29	// structure used for encoding JSON	124	// structure used for encoding JSON
30	typedef struct	125	typedef struct
31	{	126	{
32	char *cur;	127	char *cur; // SvPVX (sv) + current output position
33	STRLEN len; // SvLEN (sv)
34	char *end; // SvEND (sv)	128	char *end; // SvEND (sv)
35	SV *sv;	129	SV *sv; // result scalar
36	UV flags;	130	JSON json;
37	int max_recurse;	131	U32 indent; // indentation level
38	int indent;	132	U32 maxdepth; // max. indentation/recursion level
39	} enc_t;	133	} enc_t;
40		134
41	// structure used for decoding JSON	135	inline void
42	typedef struct
43	{
44	char *cur;
45	char *end;
46	const char *err;
47	UV flags;
48	} dec_t;
49
50	static UV *
51	SvJSON (SV *sv)
52	{
53	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
54	croak ("object is not of type JSON::XS");
55
56	return &SvUVX (SvRV (sv));
57	}
58
59	static void
60	shrink (SV *sv)
61	{
62	sv_utf8_downgrade (sv, 1);
63	#ifdef SvPV_shrink_to_cur
64	SvPV_shrink_to_cur (sv);
65	#endif
66	}
67
68	/////////////////////////////////////////////////////////////////////////////
69
70	static void
71	need (enc_t *enc, STRLEN len)	136	need (enc_t *enc, STRLEN len)
72	{	137	{
73	if (enc->cur + len >= enc->end)	138	if (expect_false (enc->cur + len >= enc->end))
74	{	139	{
75	STRLEN cur = enc->cur - SvPVX (enc->sv);	140	STRLEN cur = enc->cur - SvPVX (enc->sv);
76	SvGROW (enc->sv, cur + len + 1);	141	SvGROW (enc->sv, cur + len + 1);
77	enc->cur = SvPVX (enc->sv) + cur;	142	enc->cur = SvPVX (enc->sv) + cur;
78	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	143	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
79	}	144	}
80	}	145	}
81		146
82	static void	147	inline void
83	encode_ch (enc_t *enc, char ch)	148	encode_ch (enc_t *enc, char ch)
84	{	149	{
85	need (enc, 1);	150	need (enc, 1);
86	*enc->cur++ = ch;	151	*enc->cur++ = ch;
87	}	152	}
…		…
95		160
96	while (str < end)	161	while (str < end)
97	{	162	{
98	unsigned char ch = *(unsigned char *)str;	163	unsigned char ch = *(unsigned char *)str;
99		164
100	if (ch >= 0x20 && ch < 0x80) // most common case	165	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
101	{	166	{
102	if (ch == '"') // but with slow exceptions	167	if (expect_false (ch == '"')) // but with slow exceptions
103	{	168	{
104	need (enc, len += 1);	169	need (enc, len += 1);
105	*enc->cur++ = '\\';	170	*enc->cur++ = '\\';
106	*enc->cur++ = '"';	171	*enc->cur++ = '"';
107	}	172	}
108	else if (ch == '\\')	173	else if (expect_false (ch == '\\'))
109	{	174	{
110	need (enc, len += 1);	175	need (enc, len += 1);
111	*enc->cur++ = '\\';	176	*enc->cur++ = '\\';
112	*enc->cur++ = '\\';	177	*enc->cur++ = '\\';
113	}	178	}
…		…
131	STRLEN clen;	196	STRLEN clen;
132	UV uch;	197	UV uch;
133		198
134	if (is_utf8)	199	if (is_utf8)
135	{	200	{
136	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	201	uch = decode_utf8 (str, end - str, &clen);
137	if (clen == (STRLEN)-1)	202	if (clen == (STRLEN)-1)
138	croak ("malformed UTF-8 character in string, cannot convert to JSON");	203	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
139	}	204	}
140	else	205	else
141	{	206	{
142	uch = ch;	207	uch = ch;
143	clen = 1;	208	clen = 1;
144	}	209	}
145		210
146	if (uch < 0x80 || enc->flags & F_ASCII)	211	if (uch > 0x10FFFFUL)
		212	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		213
		214	if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF))
147	{	215	{
148	if (uch > 0xFFFFUL)	216	if (uch > 0xFFFFUL)
149	{	217	{
150	need (enc, len += 11);	218	need (enc, len += 11);
151	sprintf (enc->cur, "\\u%04x\\u%04x",	219	sprintf (enc->cur, "\\u%04x\\u%04x",
152	(uch - 0x10000) / 0x400 + 0xD800,	220	(int)((uch - 0x10000) / 0x400 + 0xD800),
153	(uch - 0x10000) % 0x400 + 0xDC00);	221	(int)((uch - 0x10000) % 0x400 + 0xDC00));
154	enc->cur += 12;	222	enc->cur += 12;
155	}	223	}
156	else	224	else
157	{	225	{
158	static char hexdigit [16] = "0123456789abcdef";	226	static char hexdigit [16] = "0123456789abcdef";
…		…
165	*enc->cur++ = hexdigit [(uch >> 0) & 15];	233	*enc->cur++ = hexdigit [(uch >> 0) & 15];
166	}	234	}
167		235
168	str += clen;	236	str += clen;
169	}	237	}
		238	else if (enc->json.flags & F_LATIN1)
		239	{
		240	*enc->cur++ = uch;
		241	str += clen;
		242	}
170	else if (is_utf8)	243	else if (is_utf8)
171	{	244	{
172	need (enc, len += clen);	245	need (enc, len += clen);
173	do	246	do
174	{	247	{
…		…
176	}	249	}
177	while (--clen);	250	while (--clen);
178	}	251	}
179	else	252	else
180	{	253	{
181	need (enc, len += 10); // never more than 11 bytes needed	254	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
182	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	255	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);
183	++str;	256	++str;
184	}	257	}
185	}	258	}
186	}	259	}
…		…
188		261
189	--len;	262	--len;
190	}	263	}
191	}	264	}
192		265
193	#define INDENT SB \	266	inline void
		267	encode_indent (enc_t *enc)
		268	{
194	if (enc->flags & F_INDENT) \	269	if (enc->json.flags & F_INDENT)
195	{ \	270	{
196	int i_; \	271	int spaces = enc->indent * INDENT_STEP;
197	need (enc, enc->indent); \	272
198	for (i_ = enc->indent * 3; i_--; )\	273	need (enc, spaces);
		274	memset (enc->cur, ' ', spaces);
		275	enc->cur += spaces;
		276	}
		277	}
		278
		279	inline void
		280	encode_space (enc_t *enc)
		281	{
		282	need (enc, 1);
		283	encode_ch (enc, ' ');
		284	}
		285
		286	inline void
		287	encode_nl (enc_t *enc)
		288	{
		289	if (enc->json.flags & F_INDENT)
		290	{
		291	need (enc, 1);
199	encode_ch (enc, ' '); \	292	encode_ch (enc, '\n');
200	} \	293	}
201	SE	294	}
202		295
203	#define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE	296	inline void
204	#define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE	297	encode_comma (enc_t *enc)
205	#define COMMA SB \	298	{
206	encode_ch (enc, ','); \	299	encode_ch (enc, ',');
		300
207	if (enc->flags & F_INDENT) \	301	if (enc->json.flags & F_INDENT)
208	NL; \	302	encode_nl (enc);
209	else if (enc->flags & F_SPACE_AFTER) \	303	else if (enc->json.flags & F_SPACE_AFTER)
210	SPACE; \	304	encode_space (enc);
211	SE	305	}
212		306
213	static void encode_sv (enc_t *enc, SV *sv);	307	static void encode_sv (enc_t *enc, SV *sv);
214		308
215	static void	309	static void
216	encode_av (enc_t *enc, AV *av)	310	encode_av (enc_t *enc, AV *av)
217	{	311	{
218	int i, len = av_len (av);	312	int i, len = av_len (av);
219		313
		314	if (enc->indent >= enc->maxdepth)
		315	croak ("data structure too deep (hit recursion limit)");
		316
220	encode_ch (enc, '['); NL;	317	encode_ch (enc, '[');
221	++enc->indent;	318
		319	if (len >= 0)
		320	{
		321	encode_nl (enc); ++enc->indent;
222		322
223	for (i = 0; i <= len; ++i)	323	for (i = 0; i <= len; ++i)
224	{	324	{
225	INDENT;	325	SV **svp = av_fetch (av, i, 0);
226	encode_sv (enc, *av_fetch (av, i, 0));
227		326
		327	encode_indent (enc);
		328
		329	if (svp)
		330	encode_sv (enc, *svp);
		331	else
		332	encode_str (enc, "null", 4, 0);
		333
228	if (i < len)	334	if (i < len)
229	COMMA;	335	encode_comma (enc);
230	}	336	}
231		337
232	NL;	338	encode_nl (enc); --enc->indent; encode_indent (enc);
233		339	}
234	--enc->indent;	340
235	INDENT; encode_ch (enc, ']');	341	encode_ch (enc, ']');
236	}	342	}
237		343
238	static void	344	static void
239	encode_he (enc_t *enc, HE *he)	345	encode_hk (enc_t *enc, HE *he)
240	{	346	{
241	encode_ch (enc, '"');	347	encode_ch (enc, '"');
242		348
243	if (HeKLEN (he) == HEf_SVKEY)	349	if (HeKLEN (he) == HEf_SVKEY)
244	{	350	{
…		…
254	else	360	else
255	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	361	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
256		362
257	encode_ch (enc, '"');	363	encode_ch (enc, '"');
258		364
259	if (enc->flags & F_SPACE_BEFORE) SPACE;	365	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
260	encode_ch (enc, ':');	366	encode_ch (enc, ':');
261	if (enc->flags & F_SPACE_AFTER ) SPACE;	367	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
262	encode_sv (enc, HeVAL (he));
263	}	368	}
264		369
265	// compare hash entries, used when all keys are bytestrings	370	// compare hash entries, used when all keys are bytestrings
266	static int	371	static int
267	he_cmp_fast (const void *a_, const void *b_)	372	he_cmp_fast (const void *a_, const void *b_)
…		…
272	HE *b = *(HE **)b_;	377	HE *b = *(HE **)b_;
273		378
274	STRLEN la = HeKLEN (a);	379	STRLEN la = HeKLEN (a);
275	STRLEN lb = HeKLEN (b);	380	STRLEN lb = HeKLEN (b);
276		381
277	if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	382	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
278	cmp = la < lb ? -1 : la == lb ? 0 : 1;	383	cmp = lb - la;
279		384
280	return cmp;	385	return cmp;
281	}	386	}
282		387
283	// compare hash entries, used when some keys are sv's or utf-x	388	// compare hash entries, used when some keys are sv's or utf-x
284	static int	389	static int
285	he_cmp_slow (const void *a, const void *b)	390	he_cmp_slow (const void *a, const void *b)
286	{	391	{
287	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	392	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
288	}	393	}
289		394
290	static void	395	static void
291	encode_hv (enc_t *enc, HV *hv)	396	encode_hv (enc_t *enc, HV *hv)
292	{	397	{
		398	HE *he;
293	int count, i;	399	int count;
294		400
295	encode_ch (enc, '{'); NL; ++enc->indent;	401	if (enc->indent >= enc->maxdepth)
		402	croak ("data structure too deep (hit recursion limit)");
296		403
297	if ((count = hv_iterinit (hv)))	404	encode_ch (enc, '{');
298	{	405
299	// for canonical output we have to sort by keys first	406	// for canonical output we have to sort by keys first
300	// actually, this is mostly due to the stupid so-called	407	// actually, this is mostly due to the stupid so-called
301	// security workaround added somewhere in 5.8.x.	408	// security workaround added somewhere in 5.8.x.
302	// that randomises hash orderings	409	// that randomises hash orderings
303	if (enc->flags & F_CANONICAL)	410	if (enc->json.flags & F_CANONICAL)
		411	{
		412	int count = hv_iterinit (hv);
		413
		414	if (SvMAGICAL (hv))
304	{	415	{
305	HE *he, *hes [count];	416	// need to count by iterating. could improve by dynamically building the vector below
		417	// but I don't care for the speed of this special case.
		418	// note also that we will run into undefined behaviour when the two iterations
		419	// do not result in the same count, something I might care for in some later release.
		420
		421	count = 0;
		422	while (hv_iternext (hv))
		423	++count;
		424
		425	hv_iterinit (hv);
		426	}
		427
		428	if (count)
		429	{
306	int fast = 1;	430	int i, fast = 1;
		431	#if defined(__BORLANDC__) || defined(_MSC_VER)
		432	HE **hes = _alloca (count * sizeof (HE));
		433	#else
		434	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
		435	#endif
307		436
308	i = 0;	437	i = 0;
309	while ((he = hv_iternext (hv)))	438	while ((he = hv_iternext (hv)))
310	{	439	{
311	hes [i++] = he;	440	hes [i++] = he;
…		…
333		462
334	FREETMPS;	463	FREETMPS;
335	LEAVE;	464	LEAVE;
336	}	465	}
337		466
338	for (i = 0; i < count; ++i)	467	encode_nl (enc); ++enc->indent;
		468
		469	while (count--)
339	{	470	{
340	INDENT;	471	encode_indent (enc);
		472	he = hes [count];
341	encode_he (enc, hes [i]);	473	encode_hk (enc, he);
		474	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
342		475
343	if (i < count - 1)	476	if (count)
344	COMMA;	477	encode_comma (enc);
345	}	478	}
346		479
		480	encode_nl (enc); --enc->indent; encode_indent (enc);
		481	}
		482	}
		483	else
		484	{
		485	if (hv_iterinit (hv) || SvMAGICAL (hv))
		486	if ((he = hv_iternext (hv)))
347	NL;	487	{
		488	encode_nl (enc); ++enc->indent;
		489
		490	for (;;)
		491	{
		492	encode_indent (enc);
		493	encode_hk (enc, he);
		494	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		495
		496	if (!(he = hv_iternext (hv)))
		497	break;
		498
		499	encode_comma (enc);
		500	}
		501
		502	encode_nl (enc); --enc->indent; encode_indent (enc);
		503	}
		504	}
		505
		506	encode_ch (enc, '}');
		507	}
		508
		509	// encode objects, arrays and special \0=false and \1=true values.
		510	static void
		511	encode_rv (enc_t *enc, SV *sv)
		512	{
		513	svtype svt;
		514
		515	SvGETMAGIC (sv);
		516	svt = SvTYPE (sv);
		517
		518	if (expect_false (SvOBJECT (sv)))
		519	{
		520	HV *stash = !JSON_SLOW || json_boolean_stash
		521	? json_boolean_stash
		522	: gv_stashpv ("JSON::XS::Boolean", 1);
		523
		524	if (SvSTASH (sv) == stash)
		525	{
		526	if (SvIV (sv))
		527	encode_str (enc, "true", 4, 0);
		528	else
		529	encode_str (enc, "false", 5, 0);
348	}	530	}
349	else	531	else
350	{	532	{
351	SV *sv;	533	#if 0
352	HE *he = hv_iternext (hv);	534	if (0 && sv_derived_from (rv, "JSON::Literal"))
353
354	for (;;)
355	{	535	{
356	INDENT;	536	// not yet
357	encode_he (enc, he);
358
359	if (!(he = hv_iternext (hv)))
360	break;
361
362	COMMA;
363	}	537	}
364		538	#endif
		539	if (enc->json.flags & F_CONV_BLESSED)
365	NL;	540	{
		541	// we re-bless the reference to get overload and other niceties right
		542	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
		543
		544	if (to_json)
		545	{
		546	dSP;
		547
		548	ENTER; SAVETMPS; PUSHMARK (SP);
		549	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		550
		551	// calling with G_SCALAR ensures that we always get a 1 return value
		552	PUTBACK;
		553	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		554	SPAGAIN;
		555
		556	// catch this surprisingly common error
		557	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		558	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		559
		560	sv = POPs;
		561	PUTBACK;
		562
		563	encode_sv (enc, sv);
		564
		565	FREETMPS; LEAVE;
		566	}
		567	else if (enc->json.flags & F_ALLOW_BLESSED)
		568	encode_str (enc, "null", 4, 0);
		569	else
		570	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		571	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		572	}
		573	else if (enc->json.flags & F_ALLOW_BLESSED)
		574	encode_str (enc, "null", 4, 0);
		575	else
		576	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		577	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
366	}	578	}
367	}	579	}
		580	else if (svt == SVt_PVHV)
		581	encode_hv (enc, (HV *)sv);
		582	else if (svt == SVt_PVAV)
		583	encode_av (enc, (AV *)sv);
		584	else if (svt < SVt_PVAV)
		585	{
		586	STRLEN len = 0;
		587	char *pv = svt ? SvPV (sv, len) : 0;
368		588
369	--enc->indent; INDENT; encode_ch (enc, '}');	589	if (len == 1 && *pv == '1')
		590	encode_str (enc, "true", 4, 0);
		591	else if (len == 1 && *pv == '0')
		592	encode_str (enc, "false", 5, 0);
		593	else
		594	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		595	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		596	}
		597	else
		598	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		599	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
370	}	600	}
371		601
372	static void	602	static void
373	encode_sv (enc_t *enc, SV *sv)	603	encode_sv (enc_t *enc, SV *sv)
374	{	604	{
…		…
382	encode_str (enc, str, len, SvUTF8 (sv));	612	encode_str (enc, str, len, SvUTF8 (sv));
383	encode_ch (enc, '"');	613	encode_ch (enc, '"');
384	}	614	}
385	else if (SvNOKp (sv))	615	else if (SvNOKp (sv))
386	{	616	{
		617	// trust that perl will do the right thing w.r.t. JSON syntax.
387	need (enc, NV_DIG + 32);	618	need (enc, NV_DIG + 32);
388	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	619	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
389	enc->cur += strlen (enc->cur);	620	enc->cur += strlen (enc->cur);
390	}	621	}
391	else if (SvIOKp (sv))	622	else if (SvIOKp (sv))
392	{	623	{
393	need (enc, 64);	624	// we assume we can always read an IV as a UV
		625	if (SvUV (sv) & ~(UV)0x7fff)
		626	{
		627	// large integer, use the (rather slow) snprintf way.
		628	need (enc, sizeof (UV) * 3);
394	enc->cur +=	629	enc->cur +=
395	SvIsUV(sv)	630	SvIsUV(sv)
396	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	631	? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv))
397	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	632	: snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv));
		633	}
		634	else
		635	{
		636	// optimise the "small number case"
		637	// code will likely be branchless and use only a single multiplication
		638	I32 i = SvIV (sv);
		639	U32 u;
		640	char digit, nz = 0;
		641
		642	need (enc, 6);
		643
		644	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		645	u = i < 0 ? -i : i;
		646
		647	// convert to 4.28 fixed-point representation
		648	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		649
		650	// now output digit by digit, each time masking out the integer part
		651	// and multiplying by 5 while moving the decimal point one to the right,
		652	// resulting in a net multiplication by 10.
		653	// we always write the digit to memory but conditionally increment
		654	// the pointer, to ease the usage of conditional move instructions.
		655	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5;
		656	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5;
		657	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5;
		658	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5;
		659	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		660	}
398	}	661	}
399	else if (SvROK (sv))	662	else if (SvROK (sv))
400	{	663	encode_rv (enc, SvRV (sv));
401	if (!--enc->max_recurse)
402	croak ("data structure too deep (hit recursion limit)");
403
404	sv = SvRV (sv);
405
406	switch (SvTYPE (sv))
407	{
408	case SVt_PVAV: encode_av (enc, (AV *)sv); break;
409	case SVt_PVHV: encode_hv (enc, (HV *)sv); break;
410
411	default:
412	croak ("JSON can only represent references to arrays or hashes");
413	}
414	}
415	else if (!SvOK (sv))	664	else if (!SvOK (sv))
416	encode_str (enc, "null", 4, 0);	665	encode_str (enc, "null", 4, 0);
417	else	666	else
418	croak ("encountered perl type that JSON cannot handle");	667	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
		668	SvPV_nolen (sv), SvFLAGS (sv));
419	}	669	}
420		670
421	static SV *	671	static SV *
422	encode_json (SV *scalar, UV flags)	672	encode_json (SV *scalar, JSON *json)
423	{	673	{
424	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))
425	croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)");
426
427	enc_t enc;	674	enc_t enc;
428	enc.flags = flags;	675
		676	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
		677	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
		678
		679	enc.json = *json;
429	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	680	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
430	enc.cur = SvPVX (enc.sv);	681	enc.cur = SvPVX (enc.sv);
431	enc.end = SvEND (enc.sv);	682	enc.end = SvEND (enc.sv);
432	enc.max_recurse = 0;
433	enc.indent = 0;	683	enc.indent = 0;
		684	enc.maxdepth = DEC_DEPTH (enc.json.flags);
434		685
435	SvPOK_only (enc.sv);	686	SvPOK_only (enc.sv);
436	encode_sv (&enc, scalar);	687	encode_sv (&enc, scalar);
437		688
		689	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		690	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		691
438	if (!(flags & (F_ASCII | F_UTF8)))	692	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
439	SvUTF8_on (enc.sv);	693	SvUTF8_on (enc.sv);
440		694
441	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
442
443	if (enc.flags & F_SHRINK)	695	if (enc.json.flags & F_SHRINK)
444	shrink (enc.sv);	696	shrink (enc.sv);
445		697
446	return enc.sv;	698	return enc.sv;
447	}	699	}
448		700
449	/////////////////////////////////////////////////////////////////////////////	701	/////////////////////////////////////////////////////////////////////////////
		702	// decoder
450		703
451	#define WS \	704	// structure used for decoding JSON
		705	typedef struct
		706	{
		707	char *cur; // current parser pointer
		708	char *end; // end of input string
		709	const char *err; // parse error, if != 0
		710	JSON json;
		711	U32 depth; // recursion depth
		712	U32 maxdepth; // recursion depth limit
		713	} dec_t;
		714
		715	inline void
		716	decode_comment (dec_t *dec)
		717	{
		718	// only '#'-style comments allowed a.t.m.
		719
		720	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		721	++dec->cur;
		722	}
		723
		724	inline void
		725	decode_ws (dec_t *dec)
		726	{
452	for (;;) \	727	for (;;)
453	{ \	728	{
454	char ch = *dec->cur; \	729	char ch = *dec->cur;
		730
455	if (ch > 0x20 \	731	if (ch > 0x20)
		732	{
		733	if (expect_false (ch == '#'))
		734	{
		735	if (dec->json.flags & F_RELAXED)
		736	decode_comment (dec);
		737	else
		738	break;
		739	}
		740	else
		741	break;
		742	}
456	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \	743	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
457	break; \	744	break; // parse error, but let higher level handle it, gives better error messages
		745
458	++dec->cur; \	746	++dec->cur;
459	}	747	}
		748	}
460		749
461	#define ERR(reason) SB dec->err = reason; goto fail; SE	750	#define ERR(reason) SB dec->err = reason; goto fail; SE
		751
462	#define EXPECT_CH(ch) SB \	752	#define EXPECT_CH(ch) SB \
463	if (*dec->cur != ch) \	753	if (*dec->cur != ch) \
464	ERR (# ch " expected"); \	754	ERR (# ch " expected"); \
465	++dec->cur; \	755	++dec->cur; \
466	SE	756	SE
467		757
		758	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		759	#define DEC_DEC_DEPTH --dec->depth
		760
468	static SV *decode_sv (dec_t *dec);	761	static SV *decode_sv (dec_t *dec);
469		762
470	static signed char decode_hexdigit[256];	763	static signed char decode_hexdigit[256];
471		764
472	static UV	765	static UV
473	decode_4hex (dec_t *dec)	766	decode_4hex (dec_t *dec)
474	{	767	{
475	signed char d1, d2, d3, d4;	768	signed char d1, d2, d3, d4;
		769	unsigned char *cur = (unsigned char *)dec->cur;
476		770
477	d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]];	771	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
478	if (d1 < 0) ERR ("four hexadecimal digits expected");	772	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
479	d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]];	773	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
480	if (d2 < 0) ERR ("four hexadecimal digits expected");	774	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
481	d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]];
482	if (d3 < 0) ERR ("four hexadecimal digits expected");
483	d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]];
484	if (d4 < 0) ERR ("four hexadecimal digits expected");
485		775
486	dec->cur += 4;	776	dec->cur += 4;
487		777
488	return ((UV)d1) << 12	778	return ((UV)d1) << 12
489	| ((UV)d2) << 8	779	| ((UV)d2) << 8
…		…
492		782
493	fail:	783	fail:
494	return (UV)-1;	784	return (UV)-1;
495	}	785	}
496		786
497	#define APPEND_GROW(n) SB \
498	if (cur + (n) >= end) \
499	{ \
500	STRLEN ofs = cur - SvPVX (sv); \
501	SvGROW (sv, ofs + (n) + 1); \
502	cur = SvPVX (sv) + ofs; \
503	end = SvEND (sv); \
504	} \
505	SE
506
507	#define APPEND_CH(ch) SB \
508	APPEND_GROW (1); \
509	*cur++ = (ch); \
510	SE
511
512	static SV *	787	static SV *
513	decode_str (dec_t *dec)	788	decode_str (dec_t *dec)
514	{	789	{
515	SV *sv = NEWSV (0,2);	790	SV *sv = 0;
516	int utf8 = 0;	791	int utf8 = 0;
517	char *cur = SvPVX (sv);	792	char *dec_cur = dec->cur;
518	char *end = SvEND (sv);
519		793
520	for (;;)	794	do
521	{	795	{
522	unsigned char ch = *(unsigned char *)dec->cur;	796	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
		797	char *cur = buf;
523		798
524	if (ch == '"')	799	do
525	break;
526	else if (ch == '\\')
527	{	800	{
528	switch (*++dec->cur)	801	unsigned char ch = *(unsigned char *)dec_cur++;
		802
		803	if (expect_false (ch == '"'))
529	{	804	{
530	case '\\':	805	--dec_cur;
531	case '/':	806	break;
532	case '"': APPEND_CH (*dec->cur++); break;	807	}
533		808	else if (expect_false (ch == '\\'))
534	case 'b': APPEND_CH ('\010'); ++dec->cur; break;	809	{
535	case 't': APPEND_CH ('\011'); ++dec->cur; break;	810	switch (*dec_cur)
536	case 'n': APPEND_CH ('\012'); ++dec->cur; break;
537	case 'f': APPEND_CH ('\014'); ++dec->cur; break;
538	case 'r': APPEND_CH ('\015'); ++dec->cur; break;
539
540	case 'u':
541	{	811	{
542	UV lo, hi;	812	case '\\':
543	++dec->cur;	813	case '/':
		814	case '"': *cur++ = *dec_cur++; break;
544		815
545	hi = decode_4hex (dec);	816	case 'b': ++dec_cur; *cur++ = '\010'; break;
546	if (hi == (UV)-1)	817	case 't': ++dec_cur; *cur++ = '\011'; break;
547	goto fail;	818	case 'n': ++dec_cur; *cur++ = '\012'; break;
		819	case 'f': ++dec_cur; *cur++ = '\014'; break;
		820	case 'r': ++dec_cur; *cur++ = '\015'; break;
548		821
549	// possibly a surrogate pair	822	case 'u':
550	if (hi >= 0xd800 && hi < 0xdc00)
551	{	823	{
552	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	824	UV lo, hi;
553	ERR ("missing low surrogate character in surrogate pair");	825	++dec_cur;
554		826
555	dec->cur += 2;	827	dec->cur = dec_cur;
556
557	lo = decode_4hex (dec);	828	hi = decode_4hex (dec);
		829	dec_cur = dec->cur;
558	if (lo == (UV)-1)	830	if (hi == (UV)-1)
559	goto fail;	831	goto fail;
560		832
		833	// possibly a surrogate pair
		834	if (hi >= 0xd800)
		835	if (hi < 0xdc00)
		836	{
		837	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
		838	ERR ("missing low surrogate character in surrogate pair");
		839
		840	dec_cur += 2;
		841
		842	dec->cur = dec_cur;
		843	lo = decode_4hex (dec);
		844	dec_cur = dec->cur;
		845	if (lo == (UV)-1)
		846	goto fail;
		847
561	if (lo < 0xdc00 || lo >= 0xe000)	848	if (lo < 0xdc00 || lo >= 0xe000)
562	ERR ("surrogate pair expected");	849	ERR ("surrogate pair expected");
563		850
564	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;	851	hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
		852	}
		853	else if (hi < 0xe000)
		854	ERR ("missing high surrogate character in surrogate pair");
		855
		856	if (hi >= 0x80)
		857	{
		858	utf8 = 1;
		859
		860	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
		861	}
		862	else
		863	*cur++ = hi;
565	}	864	}
566	else if (hi >= 0xdc00 && hi < 0xe000)
567	ERR ("missing high surrogate character in surrogate pair");
568
569	if (hi >= 0x80)
570	{	865	break;
571	utf8 = 1;
572		866
573	APPEND_GROW (4); // at most 4 bytes for 21 bits
574	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);
575	}
576	else	867	default:
577	APPEND_CH (hi);	868	--dec_cur;
		869	ERR ("illegal backslash escape sequence in string");
578	}	870	}
579	break;	871	}
		872	else if (expect_true (ch >= 0x20 && ch <= 0x7f))
		873	*cur++ = ch;
		874	else if (ch >= 0x80)
		875	{
		876	STRLEN clen;
		877	UV uch;
580		878
581	default:
582	--dec->cur;	879	--dec_cur;
583	ERR ("illegal backslash escape sequence in string");	880
		881	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
		882	if (clen == (STRLEN)-1)
		883	ERR ("malformed UTF-8 character in JSON string");
		884
		885	do
		886	*cur++ = *dec_cur++;
		887	while (--clen);
		888
		889	utf8 = 1;
		890	}
		891	else
		892	{
		893	--dec_cur;
		894
		895	if (!ch)
		896	ERR ("unexpected end of string while parsing JSON string");
		897	else
		898	ERR ("invalid character encountered while parsing JSON string");
584	}	899	}
585	}	900	}
586	else if (ch >= 0x20 && ch <= 0x7f)	901	while (cur < buf + SHORT_STRING_LEN);
587	APPEND_CH (*dec->cur++);	902
588	else if (ch >= 0x80)
589	{	903	{
590	STRLEN clen;	904	STRLEN len = cur - buf;
591	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);
592	if (clen == (STRLEN)-1)
593	ERR ("malformed UTF-8 character in JSON string");
594		905
595	APPEND_GROW (clen);	906	if (sv)
596	do
597	{	907	{
598	*cur++ = *dec->cur++;	908	SvGROW (sv, SvCUR (sv) + len + 1);
		909	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
		910	SvCUR_set (sv, SvCUR (sv) + len);
599	}	911	}
600	while (--clen);
601
602	utf8 = 1;
603	}
604	else if (dec->cur == dec->end)
605	ERR ("unexpected end of string while parsing json string");
606	else	912	else
607	ERR ("invalid character encountered");	913	sv = newSVpvn (buf, len);
608	}	914	}
		915	}
		916	while (*dec_cur != '"');
609		917
610	++dec->cur;	918	++dec_cur;
611		919
612	SvCUR_set (sv, cur - SvPVX (sv));	920	if (sv)
613		921	{
614	SvPOK_only (sv);	922	SvPOK_only (sv);
615	*SvEND (sv) = 0;	923	*SvEND (sv) = 0;
616		924
617	if (utf8)	925	if (utf8)
618	SvUTF8_on (sv);	926	SvUTF8_on (sv);
		927	}
		928	else
		929	sv = newSVpvn ("", 0);
619		930
620	if (dec->flags & F_SHRINK)	931	dec->cur = dec_cur;
621	shrink (sv);
622
623	return sv;	932	return sv;
624		933
625	fail:	934	fail:
626	SvREFCNT_dec (sv);	935	dec->cur = dec_cur;
627	return 0;	936	return 0;
628	}	937	}
629		938
630	static SV *	939	static SV *
631	decode_num (dec_t *dec)	940	decode_num (dec_t *dec)
…		…
689	is_nv = 1;	998	is_nv = 1;
690	}	999	}
691		1000
692	if (!is_nv)	1001	if (!is_nv)
693	{	1002	{
694	UV uv;	1003	int len = dec->cur - start;
695	int numtype = grok_number (start, dec->cur - start, &uv);	1004
696	if (numtype & IS_NUMBER_IN_UV)	1005	// special case the rather common 1..5-digit-int case
697	if (numtype & IS_NUMBER_NEG)	1006	if (*start == '-')
		1007	switch (len)
698	{	1008	{
699	if (uv < (UV)IV_MIN)	1009	case 2: return newSViv (-( start [1] - '0' * 1));
700	return newSViv (-(IV)uv);	1010	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1011	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1012	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1013	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
701	}	1014	}
		1015	else
		1016	switch (len)
		1017	{
		1018	case 1: return newSViv ( start [0] - '0' * 1);
		1019	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1020	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1021	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1022	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1023	}
		1024
		1025	{
		1026	UV uv;
		1027	int numtype = grok_number (start, len, &uv);
		1028	if (numtype & IS_NUMBER_IN_UV)
		1029	if (numtype & IS_NUMBER_NEG)
		1030	{
		1031	if (uv < (UV)IV_MIN)
		1032	return newSViv (-(IV)uv);
		1033	}
702	else	1034	else
703	return newSVuv (uv);	1035	return newSVuv (uv);
704	}	1036	}
705		1037
		1038	len -= *start == '-' ? 1 : 0;
		1039
		1040	// does not fit into IV or UV, try NV
		1041	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1042	#if defined (LDBL_DIG)
		1043	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1044	#endif
		1045	)
		1046	// fits into NV without loss of precision
		1047	return newSVnv (Atof (start));
		1048
		1049	// everything else fails, convert it to a string
		1050	return newSVpvn (start, dec->cur - start);
		1051	}
		1052
		1053	// loss of precision here
706	return newSVnv (Atof (start));	1054	return newSVnv (Atof (start));
707		1055
708	fail:	1056	fail:
709	return 0;	1057	return 0;
710	}	1058	}
…		…
712	static SV *	1060	static SV *
713	decode_av (dec_t *dec)	1061	decode_av (dec_t *dec)
714	{	1062	{
715	AV *av = newAV ();	1063	AV *av = newAV ();
716		1064
717	WS;	1065	DEC_INC_DEPTH;
		1066	decode_ws (dec);
		1067
718	if (*dec->cur == ']')	1068	if (*dec->cur == ']')
719	++dec->cur;	1069	++dec->cur;
720	else	1070	else
721	for (;;)	1071	for (;;)
722	{	1072	{
…		…
726	if (!value)	1076	if (!value)
727	goto fail;	1077	goto fail;
728		1078
729	av_push (av, value);	1079	av_push (av, value);
730		1080
731	WS;	1081	decode_ws (dec);
732		1082
733	if (*dec->cur == ']')	1083	if (*dec->cur == ']')
734	{	1084	{
735	++dec->cur;	1085	++dec->cur;
736	break;	1086	break;
…		…
738		1088
739	if (*dec->cur != ',')	1089	if (*dec->cur != ',')
740	ERR (", or ] expected while parsing array");	1090	ERR (", or ] expected while parsing array");
741		1091
742	++dec->cur;	1092	++dec->cur;
		1093
		1094	decode_ws (dec);
		1095
		1096	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1097	{
		1098	++dec->cur;
		1099	break;
		1100	}
743	}	1101	}
744		1102
		1103	DEC_DEC_DEPTH;
745	return newRV_noinc ((SV *)av);	1104	return newRV_noinc ((SV *)av);
746		1105
747	fail:	1106	fail:
748	SvREFCNT_dec (av);	1107	SvREFCNT_dec (av);
		1108	DEC_DEC_DEPTH;
749	return 0;	1109	return 0;
750	}	1110	}
751		1111
752	static SV *	1112	static SV *
753	decode_hv (dec_t *dec)	1113	decode_hv (dec_t *dec)
754	{	1114	{
		1115	SV *sv;
755	HV *hv = newHV ();	1116	HV *hv = newHV ();
756		1117
757	WS;	1118	DEC_INC_DEPTH;
		1119	decode_ws (dec);
		1120
758	if (*dec->cur == '}')	1121	if (*dec->cur == '}')
759	++dec->cur;	1122	++dec->cur;
760	else	1123	else
761	for (;;)	1124	for (;;)
762	{	1125	{
763	SV *key, *value;
764
765	WS; EXPECT_CH ('"');	1126	EXPECT_CH ('"');
766		1127
767	key = decode_str (dec);	1128	// heuristic: assume that
768	if (!key)	1129	// a) decode_str + hv_store_ent are abysmally slow.
769	goto fail;	1130	// b) most hash keys are short, simple ascii text.
		1131	// => try to "fast-match" such strings to avoid
		1132	// the overhead of decode_str + hv_store_ent.
		1133	{
		1134	SV *value;
		1135	char *p = dec->cur;
		1136	char *e = p + 24; // only try up to 24 bytes
770		1137
771	WS; EXPECT_CH (':');	1138	for (;;)
772
773	value = decode_sv (dec);
774	if (!value)
775	{	1139	{
		1140	// the >= 0x80 is true on most architectures
		1141	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1142	{
		1143	// slow path, back up and use decode_str
		1144	SV *key = decode_str (dec);
		1145	if (!key)
		1146	goto fail;
		1147
		1148	decode_ws (dec); EXPECT_CH (':');
		1149
		1150	decode_ws (dec);
		1151	value = decode_sv (dec);
		1152	if (!value)
		1153	{
		1154	SvREFCNT_dec (key);
		1155	goto fail;
		1156	}
		1157
		1158	hv_store_ent (hv, key, value, 0);
776	SvREFCNT_dec (key);	1159	SvREFCNT_dec (key);
		1160
		1161	break;
		1162	}
		1163	else if (*p == '"')
		1164	{
		1165	// fast path, got a simple key
		1166	char *key = dec->cur;
		1167	int len = p - key;
		1168	dec->cur = p + 1;
		1169
		1170	decode_ws (dec); EXPECT_CH (':');
		1171
		1172	decode_ws (dec);
		1173	value = decode_sv (dec);
		1174	if (!value)
777	goto fail;	1175	goto fail;
		1176
		1177	hv_store (hv, key, len, value, 0);
		1178
		1179	break;
		1180	}
		1181
		1182	++p;
778	}	1183	}
779
780	//TODO: optimise
781	hv_store_ent (hv, key, value, 0);
782
783	WS;	1184	}
		1185
		1186	decode_ws (dec);
784		1187
785	if (*dec->cur == '}')	1188	if (*dec->cur == '}')
786	{	1189	{
787	++dec->cur;	1190	++dec->cur;
788	break;	1191	break;
…		…
790		1193
791	if (*dec->cur != ',')	1194	if (*dec->cur != ',')
792	ERR (", or } expected while parsing object/hash");	1195	ERR (", or } expected while parsing object/hash");
793		1196
794	++dec->cur;	1197	++dec->cur;
		1198
		1199	decode_ws (dec);
		1200
		1201	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1202	{
		1203	++dec->cur;
		1204	break;
		1205	}
795	}	1206	}
796		1207
		1208	DEC_DEC_DEPTH;
797	return newRV_noinc ((SV *)hv);	1209	sv = newRV_noinc ((SV *)hv);
		1210
		1211	// check filter callbacks
		1212	if (dec->json.flags & F_HOOK)
		1213	{
		1214	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1215	{
		1216	HE *cb, *he;
		1217
		1218	hv_iterinit (hv);
		1219	he = hv_iternext (hv);
		1220	hv_iterinit (hv);
		1221
		1222	// the next line creates a mortal sv each time its called.
		1223	// might want to optimise this for common cases.
		1224	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1225
		1226	if (cb)
		1227	{
		1228	dSP;
		1229	int count;
		1230
		1231	ENTER; SAVETMPS; PUSHMARK (SP);
		1232	XPUSHs (HeVAL (he));
		1233
		1234	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1235
		1236	if (count == 1)
		1237	{
		1238	sv = newSVsv (POPs);
		1239	FREETMPS; LEAVE;
		1240	return sv;
		1241	}
		1242
		1243	FREETMPS; LEAVE;
		1244	}
		1245	}
		1246
		1247	if (dec->json.cb_object)
		1248	{
		1249	dSP;
		1250	int count;
		1251
		1252	ENTER; SAVETMPS; PUSHMARK (SP);
		1253	XPUSHs (sv_2mortal (sv));
		1254
		1255	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1256
		1257	if (count == 1)
		1258	{
		1259	sv = newSVsv (POPs);
		1260	FREETMPS; LEAVE;
		1261	return sv;
		1262	}
		1263
		1264	SvREFCNT_inc (sv);
		1265	FREETMPS; LEAVE;
		1266	}
		1267	}
		1268
		1269	return sv;
798		1270
799	fail:	1271	fail:
800	SvREFCNT_dec (hv);	1272	SvREFCNT_dec (hv);
		1273	DEC_DEC_DEPTH;
801	return 0;	1274	return 0;
802	}	1275	}
803		1276
804	static SV *	1277	static SV *
805	decode_sv (dec_t *dec)	1278	decode_sv (dec_t *dec)
806	{	1279	{
807	WS;	1280	// the beauty of JSON: you need exactly one character lookahead
		1281	// to parse anything.
808	switch (*dec->cur)	1282	switch (*dec->cur)
809	{	1283	{
810	case '"': ++dec->cur; return decode_str (dec);	1284	case '"': ++dec->cur; return decode_str (dec);
811	case '[': ++dec->cur; return decode_av (dec);	1285	case '[': ++dec->cur; return decode_av (dec);
812	case '{': ++dec->cur; return decode_hv (dec);	1286	case '{': ++dec->cur; return decode_hv (dec);
…		…
818		1292
819	case 't':	1293	case 't':
820	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1294	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
821	{	1295	{
822	dec->cur += 4;	1296	dec->cur += 4;
823	return newSViv (1);	1297	#if JSON_SLOW
		1298	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1299	#endif
		1300	return SvREFCNT_inc (json_true);
824	}	1301	}
825	else	1302	else
826	ERR ("'true' expected");	1303	ERR ("'true' expected");
827		1304
828	break;	1305	break;
829		1306
830	case 'f':	1307	case 'f':
831	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1308	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
832	{	1309	{
833	dec->cur += 5;	1310	dec->cur += 5;
834	return newSViv (0);	1311	#if JSON_SLOW
		1312	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1313	#endif
		1314	return SvREFCNT_inc (json_false);
835	}	1315	}
836	else	1316	else
837	ERR ("'false' expected");	1317	ERR ("'false' expected");
838		1318
839	break;	1319	break;
…		…
848	ERR ("'null' expected");	1328	ERR ("'null' expected");
849		1329
850	break;	1330	break;
851		1331
852	default:	1332	default:
853	ERR ("malformed json string, neither array, object, number, string or atom");	1333	ERR ("malformed JSON string, neither array, object, number, string or atom");
854	break;	1334	break;
855	}	1335	}
856		1336
857	fail:	1337	fail:
858	return 0;	1338	return 0;
859	}	1339	}
860		1340
861	static SV *	1341	static SV *
862	decode_json (SV *string, UV flags)	1342	decode_json (SV *string, JSON *json, UV *offset_return)
863	{	1343	{
		1344	dec_t dec;
		1345	UV offset;
864	SV *sv;	1346	SV *sv;
865		1347
		1348	SvGETMAGIC (string);
		1349	SvUPGRADE (string, SVt_PV);
		1350
		1351	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1352	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1353	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1354
866	if (flags & F_UTF8)	1355	if (json->flags & F_UTF8)
867	sv_utf8_downgrade (string, 0);	1356	sv_utf8_downgrade (string, 0);
868	else	1357	else
869	sv_utf8_upgrade (string);	1358	sv_utf8_upgrade (string);
870		1359
871	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1360	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
872		1361
873	dec_t dec;	1362	dec.json = *json;
874	dec.flags = flags;
875	dec.cur = SvPVX (string);	1363	dec.cur = SvPVX (string);
876	dec.end = SvEND (string);	1364	dec.end = SvEND (string);
877	dec.err = 0;	1365	dec.err = 0;
		1366	dec.depth = 0;
		1367	dec.maxdepth = DEC_DEPTH (dec.json.flags);
878		1368
		1369	if (dec.json.cb_object || dec.json.cb_sk_object)
		1370	dec.json.flags |= F_HOOK;
		1371
		1372	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1373
		1374	decode_ws (&dec);
879	sv = decode_sv (&dec);	1375	sv = decode_sv (&dec);
880		1376
		1377	if (!(offset_return || !sv))
		1378	{
		1379	// check for trailing garbage
		1380	decode_ws (&dec);
		1381
		1382	if (*dec.cur)
		1383	{
		1384	dec.err = "garbage after JSON object";
		1385	SvREFCNT_dec (sv);
		1386	sv = 0;
		1387	}
		1388	}
		1389
		1390	if (offset_return || !sv)
		1391	{
		1392	offset = dec.json.flags & F_UTF8
		1393	? dec.cur - SvPVX (string)
		1394	: utf8_distance (dec.cur, SvPVX (string));
		1395
		1396	if (offset_return)
		1397	*offset_return = offset;
		1398	}
		1399
881	if (!sv)	1400	if (!sv)
882	{	1401	{
883	IV offset = dec.flags & F_UTF8
884	? dec.cur - SvPVX (string)
885	: utf8_distance (dec.cur, SvPVX (string));
886	SV *uni = sv_newmortal ();	1402	SV *uni = sv_newmortal ();
887		1403
888	// horrible hack to silence warning inside pv_uni_display	1404	// horrible hack to silence warning inside pv_uni_display
889	COP cop = *PL_curcop;	1405	COP cop = *PL_curcop;
890	cop.cop_warnings = pWARN_NONE;	1406	cop.cop_warnings = pWARN_NONE;
…		…
892	SAVEVPTR (PL_curcop);	1408	SAVEVPTR (PL_curcop);
893	PL_curcop = &cop;	1409	PL_curcop = &cop;
894	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1410	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
895	LEAVE;	1411	LEAVE;
896		1412
897	croak ("%s, at character offset %d (%s)",	1413	croak ("%s, at character offset %d [\"%s\"]",
898	dec.err,	1414	dec.err,
899	(int)offset,	1415	(int)offset,
900	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1416	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
901	}	1417	}
902		1418
903	sv = sv_2mortal (sv);	1419	sv = sv_2mortal (sv);
904		1420
905	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1421	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
906	croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)");	1422	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
907		1423
908	return sv;	1424	return sv;
909	}	1425	}
910		1426
		1427	/////////////////////////////////////////////////////////////////////////////
		1428	// XS interface functions
		1429
911	MODULE = JSON::XS PACKAGE = JSON::XS	1430	MODULE = JSON::XS PACKAGE = JSON::XS
912		1431
913	BOOT:	1432	BOOT:
914	{	1433	{
915	int i;	1434	int i;
916		1435
917	memset (decode_hexdigit, 0xff, 256);
918	for (i = 10; i--; )	1436	for (i = 0; i < 256; ++i)
919	decode_hexdigit ['0' + i] = i;	1437	decode_hexdigit [i] =
		1438	i >= '0' && i <= '9' ? i - '0'
		1439	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1440	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1441	: -1;
920		1442
921	for (i = 7; i--; )
922	{
923	decode_hexdigit ['a' + i] = 10 + i;
924	decode_hexdigit ['A' + i] = 10 + i;
925	}
926
927	json_stash = gv_stashpv ("JSON::XS", 1);	1443	json_stash = gv_stashpv ("JSON::XS" , 1);
		1444	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1445
		1446	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1447	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
928	}	1448	}
929		1449
930	PROTOTYPES: DISABLE	1450	PROTOTYPES: DISABLE
931		1451
932	SV *new (char *dummy)	1452	void CLONE (...)
933	CODE:	1453	CODE:
934	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1454	json_stash = 0;
		1455	json_boolean_stash = 0;
		1456
		1457	void new (char *klass)
		1458	PPCODE:
		1459	{
		1460	SV *pv = NEWSV (0, sizeof (JSON));
		1461	SvPOK_only (pv);
		1462	Zero (SvPVX (pv), 1, JSON);
		1463	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1464	XPUSHs (sv_2mortal (sv_bless (
		1465	newRV_noinc (pv),
		1466	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1467	)));
		1468	}
		1469
		1470	void ascii (JSON *self, int enable = 1)
		1471	ALIAS:
		1472	ascii = F_ASCII
		1473	latin1 = F_LATIN1
		1474	utf8 = F_UTF8
		1475	indent = F_INDENT
		1476	canonical = F_CANONICAL
		1477	space_before = F_SPACE_BEFORE
		1478	space_after = F_SPACE_AFTER
		1479	pretty = F_PRETTY
		1480	allow_nonref = F_ALLOW_NONREF
		1481	shrink = F_SHRINK
		1482	allow_blessed = F_ALLOW_BLESSED
		1483	convert_blessed = F_CONV_BLESSED
		1484	relaxed = F_RELAXED
		1485	PPCODE:
		1486	{
		1487	if (enable)
		1488	self->flags |= ix;
		1489	else
		1490	self->flags &= ~ix;
		1491
		1492	XPUSHs (ST (0));
		1493	}
		1494
		1495	void get_ascii (JSON *self)
		1496	ALIAS:
		1497	get_ascii = F_ASCII
		1498	get_latin1 = F_LATIN1
		1499	get_utf8 = F_UTF8
		1500	get_indent = F_INDENT
		1501	get_canonical = F_CANONICAL
		1502	get_space_before = F_SPACE_BEFORE
		1503	get_space_after = F_SPACE_AFTER
		1504	get_allow_nonref = F_ALLOW_NONREF
		1505	get_shrink = F_SHRINK
		1506	get_allow_blessed = F_ALLOW_BLESSED
		1507	get_convert_blessed = F_CONV_BLESSED
		1508	get_relaxed = F_RELAXED
		1509	PPCODE:
		1510	XPUSHs (boolSV (self->flags & ix));
		1511
		1512	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1513	PPCODE:
		1514	{
		1515	UV log2 = 0;
		1516
		1517	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1518
		1519	while ((1UL << log2) < max_depth)
		1520	++log2;
		1521
		1522	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1523
		1524	XPUSHs (ST (0));
		1525	}
		1526
		1527	U32 get_max_depth (JSON *self)
		1528	CODE:
		1529	RETVAL = DEC_DEPTH (self->flags);
935	OUTPUT:	1530	OUTPUT:
936	RETVAL	1531	RETVAL
937		1532
938	SV *ascii (SV *self, int enable = 1)	1533	void max_size (JSON *self, UV max_size = 0)
939	ALIAS:	1534	PPCODE:
940	ascii = F_ASCII	1535	{
941	utf8 = F_UTF8	1536	UV log2 = 0;
942	indent = F_INDENT	1537
943	canonical = F_CANONICAL	1538	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
944	space_before = F_SPACE_BEFORE	1539	if (max_size == 1) max_size = 2;
945	space_after = F_SPACE_AFTER	1540
946	json_rpc = F_JSON_RPC	1541	while ((1UL << log2) < max_size)
947	pretty = F_PRETTY	1542	++log2;
948	allow_nonref = F_ALLOW_NONREF	1543
949	shrink = F_SHRINK	1544	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1545
		1546	XPUSHs (ST (0));
		1547	}
		1548
		1549	int get_max_size (JSON *self)
950	CODE:	1550	CODE:
951	{	1551	RETVAL = DEC_SIZE (self->flags);
952	UV *uv = SvJSON (self);
953	if (enable)
954	*uv |= ix;
955	else
956	*uv &= ~ix;
957
958	RETVAL = newSVsv (self);
959	}
960	OUTPUT:	1552	OUTPUT:
961	RETVAL	1553	RETVAL
962		1554
963	void encode (SV *self, SV *scalar)	1555	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
964	PPCODE:	1556	PPCODE:
965	XPUSHs (encode_json (scalar, *SvJSON (self)));	1557	{
		1558	SvREFCNT_dec (self->cb_object);
		1559	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
966		1560
967	void decode (SV *self, SV *jsonstr)	1561	XPUSHs (ST (0));
		1562	}
		1563
		1564	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
968	PPCODE:	1565	PPCODE:
		1566	{
		1567	if (!self->cb_sk_object)
		1568	self->cb_sk_object = newHV ();
		1569
		1570	if (SvOK (cb))
		1571	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1572	else
		1573	{
		1574	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1575
		1576	if (!HvKEYS (self->cb_sk_object))
		1577	{
		1578	SvREFCNT_dec (self->cb_sk_object);
		1579	self->cb_sk_object = 0;
		1580	}
		1581	}
		1582
		1583	XPUSHs (ST (0));
		1584	}
		1585
		1586	void encode (JSON *self, SV *scalar)
		1587	PPCODE:
		1588	XPUSHs (encode_json (scalar, self));
		1589
		1590	void decode (JSON *self, SV *jsonstr)
		1591	PPCODE:
969	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1592	XPUSHs (decode_json (jsonstr, self, 0));
		1593
		1594	void decode_prefix (JSON *self, SV *jsonstr)
		1595	PPCODE:
		1596	{
		1597	UV offset;
		1598	EXTEND (SP, 2);
		1599	PUSHs (decode_json (jsonstr, self, &offset));
		1600	PUSHs (sv_2mortal (newSVuv (offset)));
		1601	}
		1602
		1603	void DESTROY (JSON *self)
		1604	CODE:
		1605	SvREFCNT_dec (self->cb_sk_object);
		1606	SvREFCNT_dec (self->cb_object);
970		1607
971	PROTOTYPES: ENABLE	1608	PROTOTYPES: ENABLE
972		1609
973	void to_json (SV *scalar)	1610	void encode_json (SV *scalar)
974	PPCODE:	1611	PPCODE:
		1612	{
		1613	JSON json = { F_DEFAULT | F_UTF8 };
975	XPUSHs (encode_json (scalar, F_UTF8));	1614	XPUSHs (encode_json (scalar, &json));
		1615	}
976		1616
977	void from_json (SV *jsonstr)	1617	void decode_json (SV *jsonstr)
978	PPCODE:	1618	PPCODE:
		1619	{
		1620	JSON json = { F_DEFAULT | F_UTF8 };
979	XPUSHs (decode_json (jsonstr, F_UTF8));	1621	XPUSHs (decode_json (jsonstr, &json, 0));
		1622	}
980		1623

Diff Legend

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