ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.12 by root, Sat Mar 24 22:10:08 2007 UTC vs.
Revision 1.75 by root, Wed Mar 19 15:37:54 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 <limits.h>
		10	#include <float.h>
8		11
		12	#if defined(__BORLANDC__) || defined(_MSC_VER)
		13	# define snprintf _snprintf // C compilers have this in stdio.h
		14	#endif
		15
		16	// some old perls do not have this, try to make it work, no
		17	// guarentees, though. if it breaks, you get to keep the pieces.
		18	#ifndef UTF8_MAXBYTES
		19	# define UTF8_MAXBYTES 13
		20	#endif
		21
		22	#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2)
		23
9	#define F_ASCII 0x00000001	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
10	#define F_UTF8 0x00000002	26	#define F_UTF8 0x00000004UL
11	#define F_INDENT 0x00000004	27	#define F_INDENT 0x00000008UL
12	#define F_CANONICAL 0x00000008	28	#define F_CANONICAL 0x00000010UL
13	#define F_SPACE_BEFORE 0x00000010	29	#define F_SPACE_BEFORE 0x00000020UL
14	#define F_SPACE_AFTER 0x00000020	30	#define F_SPACE_AFTER 0x00000040UL
15	#define F_ALLOW_NONREF 0x00000080	31	#define F_ALLOW_NONREF 0x00000100UL
16	#define F_SHRINK 0x00000100	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
		35	#define F_RELAXED 0x00001000UL
		36
		37	#define F_MAXDEPTH 0xf8000000UL
		38	#define S_MAXDEPTH 27
		39	#define F_MAXSIZE 0x01f00000UL
		40	#define S_MAXSIZE 20
		41	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		42
		43	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		44	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		45
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	46	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	47	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		48
21	#define INIT_SIZE 32 // initial scalar size to be allocated	49	#define INIT_SIZE 32 // initial scalar size to be allocated
22	#define INDENT_STEP 3 // spaces per indentation level	50	#define INDENT_STEP 3 // spaces per indentation level
23		51
24	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
25	#define SHORT_STRING_LEN 256 // special-case strings of up to this size	52	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
26		53
27	#define SB do {	54	#define SB do {
28	#define SE } while (0)	55	#define SE } while (0)
29		56
		57	#if __GNUC__ >= 3
		58	# define expect(expr,value) __builtin_expect ((expr), (value))
		59	# define INLINE static inline
		60	#else
		61	# define expect(expr,value) (expr)
		62	# define INLINE static
		63	#endif
		64
		65	#define expect_false(expr) expect ((expr) != 0, 0)
		66	#define expect_true(expr) expect ((expr) != 0, 1)
		67
		68	#define IN_RANGE_INC(type,val,beg,end) \
		69	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		70	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		71
		72	#ifdef USE_ITHREADS
		73	# define JSON_SLOW 1
		74	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		75	#else
		76	# define JSON_SLOW 0
		77	# define JSON_STASH json_stash
		78	#endif
		79
30	static HV *json_stash; // JSON::XS::	80	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		81	static SV *json_true, *json_false;
		82
		83	typedef struct {
		84	U32 flags;
		85	SV *cb_object;
		86	HV *cb_sk_object;
		87	} JSON;
31		88
32	/////////////////////////////////////////////////////////////////////////////	89	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	90	// utility functions
34		91
35	static UV *	92	INLINE void
36	SvJSON (SV *sv)
37	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))
39	croak ("object is not of type JSON::XS");
40
41	return &SvUVX (SvRV (sv));
42	}
43
44	static void
45	shrink (SV *sv)	93	shrink (SV *sv)
46	{	94	{
47	sv_utf8_downgrade (sv, 1);	95	sv_utf8_downgrade (sv, 1);
48	if (SvLEN (sv) > SvCUR (sv) + 1)	96	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	97	{
…		…
53	SvPV_renew (sv, SvCUR (sv) + 1);	101	SvPV_renew (sv, SvCUR (sv) + 1);
54	#endif	102	#endif
55	}	103	}
56	}	104	}
57		105
		106	// decode an utf-8 character and return it, or (UV)-1 in
		107	// case of an error.
		108	// we special-case "safe" characters from U+80 .. U+7FF,
		109	// but use the very good perl function to parse anything else.
		110	// note that we never call this function for a ascii codepoints
		111	INLINE UV
		112	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
		113	{
		114	if (expect_true (len >= 2
		115	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
		116	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
		117	{
		118	*clen = 2;
		119	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
		120	}
		121	else
		122	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		123	}
		124
		125	// likewise for encoding, also never called for ascii codepoints
		126	// this function takes advantage of this fact, although current gccs
		127	// seem to optimise the check for >= 0x80 away anyways
		128	INLINE unsigned char *
		129	encode_utf8 (unsigned char *s, UV ch)
		130	{
		131	if (expect_false (ch < 0x000080))
		132	*s++ = ch;
		133	else if (expect_true (ch < 0x000800))
		134	*s++ = 0xc0 | ( ch >> 6),
		135	*s++ = 0x80 | ( ch & 0x3f);
		136	else if ( ch < 0x010000)
		137	*s++ = 0xe0 | ( ch >> 12),
		138	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		139	*s++ = 0x80 | ( ch & 0x3f);
		140	else if ( ch < 0x110000)
		141	*s++ = 0xf0 | ( ch >> 18),
		142	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		143	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		144	*s++ = 0x80 | ( ch & 0x3f);
		145
		146	return s;
		147	}
		148
58	/////////////////////////////////////////////////////////////////////////////	149	/////////////////////////////////////////////////////////////////////////////
59	// encoder	150	// encoder
60		151
61	// structure used for encoding JSON	152	// structure used for encoding JSON
62	typedef struct	153	typedef struct
63	{	154	{
64	char *cur; // SvPVX (sv) + current output position	155	char *cur; // SvPVX (sv) + current output position
65	char *end; // SvEND (sv)	156	char *end; // SvEND (sv)
66	SV *sv; // result scalar	157	SV *sv; // result scalar
67	UV flags; // F_*	158	JSON json;
68	int indent; // indentation level	159	U32 indent; // indentation level
69	int max_depth; // max. recursion level	160	U32 maxdepth; // max. indentation/recursion level
		161	UV limit; // escape character values >= this value when encoding
70	} enc_t;	162	} enc_t;
71		163
72	static void	164	INLINE void
73	need (enc_t *enc, STRLEN len)	165	need (enc_t *enc, STRLEN len)
74	{	166	{
75	if (enc->cur + len >= enc->end)	167	if (expect_false (enc->cur + len >= enc->end))
76	{	168	{
77	STRLEN cur = enc->cur - SvPVX (enc->sv);	169	STRLEN cur = enc->cur - SvPVX (enc->sv);
78	SvGROW (enc->sv, cur + len + 1);	170	SvGROW (enc->sv, cur + len + 1);
79	enc->cur = SvPVX (enc->sv) + cur;	171	enc->cur = SvPVX (enc->sv) + cur;
80	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	172	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
81	}	173	}
82	}	174	}
83		175
84	static void	176	INLINE void
85	encode_ch (enc_t *enc, char ch)	177	encode_ch (enc_t *enc, char ch)
86	{	178	{
87	need (enc, 1);	179	need (enc, 1);
88	*enc->cur++ = ch;	180	*enc->cur++ = ch;
89	}	181	}
…		…
97		189
98	while (str < end)	190	while (str < end)
99	{	191	{
100	unsigned char ch = *(unsigned char *)str;	192	unsigned char ch = *(unsigned char *)str;
101		193
102	if (ch >= 0x20 && ch < 0x80) // most common case	194	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
103	{	195	{
104	if (ch == '"') // but with slow exceptions	196	if (expect_false (ch == '"')) // but with slow exceptions
105	{	197	{
106	need (enc, len += 1);	198	need (enc, len += 1);
107	*enc->cur++ = '\\';	199	*enc->cur++ = '\\';
108	*enc->cur++ = '"';	200	*enc->cur++ = '"';
109	}	201	}
110	else if (ch == '\\')	202	else if (expect_false (ch == '\\'))
111	{	203	{
112	need (enc, len += 1);	204	need (enc, len += 1);
113	*enc->cur++ = '\\';	205	*enc->cur++ = '\\';
114	*enc->cur++ = '\\';	206	*enc->cur++ = '\\';
115	}	207	}
…		…
133	STRLEN clen;	225	STRLEN clen;
134	UV uch;	226	UV uch;
135		227
136	if (is_utf8)	228	if (is_utf8)
137	{	229	{
138	uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);	230	uch = decode_utf8 (str, end - str, &clen);
139	if (clen == (STRLEN)-1)	231	if (clen == (STRLEN)-1)
140	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	232	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
141	}	233	}
142	else	234	else
143	{	235	{
144	uch = ch;	236	uch = ch;
145	clen = 1;	237	clen = 1;
146	}	238	}
147		239
148	if (uch > 0x10FFFFUL)	240	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
149	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
150
151	if (uch < 0x80 || enc->flags & F_ASCII)
152	{	241	{
153	if (uch > 0xFFFFUL)	242	if (uch >= 0x10000UL)
154	{	243	{
		244	if (uch >= 0x110000UL)
		245	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		246
155	need (enc, len += 11);	247	need (enc, len += 11);
156	sprintf (enc->cur, "\\u%04x\\u%04x",	248	sprintf (enc->cur, "\\u%04x\\u%04x",
157	(int)((uch - 0x10000) / 0x400 + 0xD800),	249	(int)((uch - 0x10000) / 0x400 + 0xD800),
158	(int)((uch - 0x10000) % 0x400 + 0xDC00));	250	(int)((uch - 0x10000) % 0x400 + 0xDC00));
159	enc->cur += 12;	251	enc->cur += 12;
…		…
170	*enc->cur++ = hexdigit [(uch >> 0) & 15];	262	*enc->cur++ = hexdigit [(uch >> 0) & 15];
171	}	263	}
172		264
173	str += clen;	265	str += clen;
174	}	266	}
		267	else if (enc->json.flags & F_LATIN1)
		268	{
		269	*enc->cur++ = uch;
		270	str += clen;
		271	}
175	else if (is_utf8)	272	else if (is_utf8)
176	{	273	{
177	need (enc, len += clen);	274	need (enc, len += clen);
178	do	275	do
179	{	276	{
…		…
181	}	278	}
182	while (--clen);	279	while (--clen);
183	}	280	}
184	else	281	else
185	{	282	{
186	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	283	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
187	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	284	enc->cur = encode_utf8 (enc->cur, uch);
188	++str;	285	++str;
189	}	286	}
190	}	287	}
191	}	288	}
192	}	289	}
193		290
194	--len;	291	--len;
195	}	292	}
196	}	293	}
197		294
198	static void	295	INLINE void
199	encode_indent (enc_t *enc)	296	encode_indent (enc_t *enc)
200	{	297	{
201	if (enc->flags & F_INDENT)	298	if (enc->json.flags & F_INDENT)
202	{	299	{
203	int spaces = enc->indent * INDENT_STEP;	300	int spaces = enc->indent * INDENT_STEP;
204		301
205	need (enc, spaces);	302	need (enc, spaces);
206	memset (enc->cur, ' ', spaces);	303	memset (enc->cur, ' ', spaces);
207	enc->cur += spaces;	304	enc->cur += spaces;
208	}	305	}
209	}	306	}
210		307
211	static void	308	INLINE void
212	encode_space (enc_t *enc)	309	encode_space (enc_t *enc)
213	{	310	{
214	need (enc, 1);	311	need (enc, 1);
215	encode_ch (enc, ' ');	312	encode_ch (enc, ' ');
216	}	313	}
217		314
218	static void	315	INLINE void
219	encode_nl (enc_t *enc)	316	encode_nl (enc_t *enc)
220	{	317	{
221	if (enc->flags & F_INDENT)	318	if (enc->json.flags & F_INDENT)
222	{	319	{
223	need (enc, 1);	320	need (enc, 1);
224	encode_ch (enc, '\n');	321	encode_ch (enc, '\n');
225	}	322	}
226	}	323	}
227		324
228	static void	325	INLINE void
229	encode_comma (enc_t *enc)	326	encode_comma (enc_t *enc)
230	{	327	{
231	encode_ch (enc, ',');	328	encode_ch (enc, ',');
232		329
233	if (enc->flags & F_INDENT)	330	if (enc->json.flags & F_INDENT)
234	encode_nl (enc);	331	encode_nl (enc);
235	else if (enc->flags & F_SPACE_AFTER)	332	else if (enc->json.flags & F_SPACE_AFTER)
236	encode_space (enc);	333	encode_space (enc);
237	}	334	}
238		335
239	static void encode_sv (enc_t *enc, SV *sv);	336	static void encode_sv (enc_t *enc, SV *sv);
240		337
241	static void	338	static void
242	encode_av (enc_t *enc, AV *av)	339	encode_av (enc_t *enc, AV *av)
243	{	340	{
244	int i, len = av_len (av);	341	int i, len = av_len (av);
245		342
246	encode_ch (enc, '['); encode_nl (enc);	343	if (enc->indent >= enc->maxdepth)
247	++enc->indent;	344	croak ("data structure too deep (hit recursion limit)");
248		345
		346	encode_ch (enc, '[');
		347
		348	if (len >= 0)
		349	{
		350	encode_nl (enc); ++enc->indent;
		351
249	for (i = 0; i <= len; ++i)	352	for (i = 0; i <= len; ++i)
250	{	353	{
		354	SV **svp = av_fetch (av, i, 0);
		355
251	encode_indent (enc);	356	encode_indent (enc);
252	encode_sv (enc, *av_fetch (av, i, 0));
253		357
		358	if (svp)
		359	encode_sv (enc, *svp);
		360	else
		361	encode_str (enc, "null", 4, 0);
		362
254	if (i < len)	363	if (i < len)
255	encode_comma (enc);	364	encode_comma (enc);
256	}	365	}
257		366
		367	encode_nl (enc); --enc->indent; encode_indent (enc);
		368	}
		369
258	encode_nl (enc);	370	encode_ch (enc, ']');
259
260	--enc->indent;
261	encode_indent (enc); encode_ch (enc, ']');
262	}	371	}
263		372
264	static void	373	static void
265	encode_he (enc_t *enc, HE *he)	374	encode_hk (enc_t *enc, HE *he)
266	{	375	{
267	encode_ch (enc, '"');	376	encode_ch (enc, '"');
268		377
269	if (HeKLEN (he) == HEf_SVKEY)	378	if (HeKLEN (he) == HEf_SVKEY)
270	{	379	{
…		…
280	else	389	else
281	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	390	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
282		391
283	encode_ch (enc, '"');	392	encode_ch (enc, '"');
284		393
285	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	394	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
286	encode_ch (enc, ':');	395	encode_ch (enc, ':');
287	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	396	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
288	encode_sv (enc, HeVAL (he));
289	}	397	}
290		398
291	// compare hash entries, used when all keys are bytestrings	399	// compare hash entries, used when all keys are bytestrings
292	static int	400	static int
293	he_cmp_fast (const void *a_, const void *b_)	401	he_cmp_fast (const void *a_, const void *b_)
…		…
298	HE *b = *(HE **)b_;	406	HE *b = *(HE **)b_;
299		407
300	STRLEN la = HeKLEN (a);	408	STRLEN la = HeKLEN (a);
301	STRLEN lb = HeKLEN (b);	409	STRLEN lb = HeKLEN (b);
302		410
303	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	411	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
304	cmp = la - lb;	412	cmp = lb - la;
305		413
306	return cmp;	414	return cmp;
307	}	415	}
308		416
309	// compare hash entries, used when some keys are sv's or utf-x	417	// compare hash entries, used when some keys are sv's or utf-x
310	static int	418	static int
311	he_cmp_slow (const void *a, const void *b)	419	he_cmp_slow (const void *a, const void *b)
312	{	420	{
313	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	421	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
314	}	422	}
315		423
316	static void	424	static void
317	encode_hv (enc_t *enc, HV *hv)	425	encode_hv (enc_t *enc, HV *hv)
318	{	426	{
		427	HE *he;
319	int count, i;	428	int count;
320		429
321	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	430	if (enc->indent >= enc->maxdepth)
		431	croak ("data structure too deep (hit recursion limit)");
322		432
323	if ((count = hv_iterinit (hv)))	433	encode_ch (enc, '{');
324	{	434
325	// for canonical output we have to sort by keys first	435	// for canonical output we have to sort by keys first
326	// actually, this is mostly due to the stupid so-called	436	// actually, this is mostly due to the stupid so-called
327	// security workaround added somewhere in 5.8.x.	437	// security workaround added somewhere in 5.8.x.
328	// that randomises hash orderings	438	// that randomises hash orderings
329	if (enc->flags & F_CANONICAL)	439	if (enc->json.flags & F_CANONICAL)
		440	{
		441	int count = hv_iterinit (hv);
		442
		443	if (SvMAGICAL (hv))
330	{	444	{
		445	// need to count by iterating. could improve by dynamically building the vector below
		446	// but I don't care for the speed of this special case.
		447	// note also that we will run into undefined behaviour when the two iterations
		448	// do not result in the same count, something I might care for in some later release.
		449
		450	count = 0;
		451	while (hv_iternext (hv))
		452	++count;
		453
		454	hv_iterinit (hv);
		455	}
		456
		457	if (count)
		458	{
		459	int i, fast = 1;
		460	#if defined(__BORLANDC__) || defined(_MSC_VER)
		461	HE **hes = _alloca (count * sizeof (HE));
		462	#else
331	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	463	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
332	int fast = 1;	464	#endif
333		465
334	i = 0;	466	i = 0;
335	while ((he = hv_iternext (hv)))	467	while ((he = hv_iternext (hv)))
336	{	468	{
337	hes [i++] = he;	469	hes [i++] = he;
…		…
359		491
360	FREETMPS;	492	FREETMPS;
361	LEAVE;	493	LEAVE;
362	}	494	}
363		495
364	for (i = 0; i < count; ++i)	496	encode_nl (enc); ++enc->indent;
		497
		498	while (count--)
365	{	499	{
366	encode_indent (enc);	500	encode_indent (enc);
		501	he = hes [count];
367	encode_he (enc, hes [i]);	502	encode_hk (enc, he);
		503	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
368		504
369	if (i < count - 1)	505	if (count)
370	encode_comma (enc);	506	encode_comma (enc);
371	}	507	}
372		508
		509	encode_nl (enc); --enc->indent; encode_indent (enc);
		510	}
		511	}
		512	else
		513	{
		514	if (hv_iterinit (hv) || SvMAGICAL (hv))
		515	if ((he = hv_iternext (hv)))
		516	{
		517	encode_nl (enc); ++enc->indent;
		518
		519	for (;;)
		520	{
373	encode_nl (enc);	521	encode_indent (enc);
		522	encode_hk (enc, he);
		523	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		524
		525	if (!(he = hv_iternext (hv)))
		526	break;
		527
		528	encode_comma (enc);
		529	}
		530
		531	encode_nl (enc); --enc->indent; encode_indent (enc);
		532	}
		533	}
		534
		535	encode_ch (enc, '}');
		536	}
		537
		538	// encode objects, arrays and special \0=false and \1=true values.
		539	static void
		540	encode_rv (enc_t *enc, SV *sv)
		541	{
		542	svtype svt;
		543
		544	SvGETMAGIC (sv);
		545	svt = SvTYPE (sv);
		546
		547	if (expect_false (SvOBJECT (sv)))
		548	{
		549	HV *stash = !JSON_SLOW || json_boolean_stash
		550	? json_boolean_stash
		551	: gv_stashpv ("JSON::XS::Boolean", 1);
		552
		553	if (SvSTASH (sv) == stash)
		554	{
		555	if (SvIV (sv))
		556	encode_str (enc, "true", 4, 0);
		557	else
		558	encode_str (enc, "false", 5, 0);
374	}	559	}
375	else	560	else
376	{	561	{
377	SV *sv;	562	#if 0
378	HE *he = hv_iternext (hv);	563	if (0 && sv_derived_from (rv, "JSON::Literal"))
379
380	for (;;)
381	{	564	{
382	encode_indent (enc);	565	// not yet
383	encode_he (enc, he);
384
385	if (!(he = hv_iternext (hv)))
386	break;
387
388	encode_comma (enc);
389	}	566	}
		567	#endif
		568	if (enc->json.flags & F_CONV_BLESSED)
		569	{
		570	// we re-bless the reference to get overload and other niceties right
		571	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
390		572
		573	if (to_json)
		574	{
		575	dSP;
		576
		577	ENTER; SAVETMPS; PUSHMARK (SP);
		578	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		579
		580	// calling with G_SCALAR ensures that we always get a 1 return value
		581	PUTBACK;
		582	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		583	SPAGAIN;
		584
		585	// catch this surprisingly common error
		586	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		587	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		588
		589	sv = POPs;
		590	PUTBACK;
		591
391	encode_nl (enc);	592	encode_sv (enc, sv);
		593
		594	FREETMPS; LEAVE;
		595	}
		596	else if (enc->json.flags & F_ALLOW_BLESSED)
		597	encode_str (enc, "null", 4, 0);
		598	else
		599	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		600	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		601	}
		602	else if (enc->json.flags & F_ALLOW_BLESSED)
		603	encode_str (enc, "null", 4, 0);
		604	else
		605	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		606	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
392	}	607	}
393	}	608	}
		609	else if (svt == SVt_PVHV)
		610	encode_hv (enc, (HV *)sv);
		611	else if (svt == SVt_PVAV)
		612	encode_av (enc, (AV *)sv);
		613	else if (svt < SVt_PVAV)
		614	{
		615	STRLEN len = 0;
		616	char *pv = svt ? SvPV (sv, len) : 0;
394		617
395	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	618	if (len == 1 && *pv == '1')
		619	encode_str (enc, "true", 4, 0);
		620	else if (len == 1 && *pv == '0')
		621	encode_str (enc, "false", 5, 0);
		622	else
		623	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		624	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		625	}
		626	else
		627	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		628	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
396	}	629	}
397		630
398	static void	631	static void
399	encode_sv (enc_t *enc, SV *sv)	632	encode_sv (enc_t *enc, SV *sv)
400	{	633	{
…		…
408	encode_str (enc, str, len, SvUTF8 (sv));	641	encode_str (enc, str, len, SvUTF8 (sv));
409	encode_ch (enc, '"');	642	encode_ch (enc, '"');
410	}	643	}
411	else if (SvNOKp (sv))	644	else if (SvNOKp (sv))
412	{	645	{
		646	// trust that perl will do the right thing w.r.t. JSON syntax.
413	need (enc, NV_DIG + 32);	647	need (enc, NV_DIG + 32);
414	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	648	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
415	enc->cur += strlen (enc->cur);	649	enc->cur += strlen (enc->cur);
416	}	650	}
417	else if (SvIOKp (sv))	651	else if (SvIOKp (sv))
418	{	652	{
		653	// we assume we can always read an IV as a UV and vice versa
		654	// we assume two's complement
		655	// we assume no aliasing issues in the union
		656	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		657	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		658	{
		659	// optimise the "small number case"
		660	// code will likely be branchless and use only a single multiplication
		661	// works for numbers up to 59074
		662	I32 i = SvIVX (sv);
		663	U32 u;
		664	char digit, nz = 0;
		665
419	need (enc, 64);	666	need (enc, 6);
		667
		668	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		669	u = i < 0 ? -i : i;
		670
		671	// convert to 4.28 fixed-point representation
		672	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		673
		674	// now output digit by digit, each time masking out the integer part
		675	// and multiplying by 5 while moving the decimal point one to the right,
		676	// resulting in a net multiplication by 10.
		677	// we always write the digit to memory but conditionally increment
		678	// the pointer, to enable the use of conditional move instructions.
		679	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		680	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		681	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		682	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		683	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		684	}
		685	else
		686	{
		687	// large integer, use the (rather slow) snprintf way.
		688	need (enc, IVUV_MAXCHARS);
420	enc->cur +=	689	enc->cur +=
421	SvIsUV(sv)	690	SvIsUV(sv)
422	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	691	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
423	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	692	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		693	}
424	}	694	}
425	else if (SvROK (sv))	695	else if (SvROK (sv))
426	{	696	encode_rv (enc, SvRV (sv));
427	SV *rv = SvRV (sv);
428
429	if (enc->indent >= enc->max_depth)
430	croak ("data structure too deep (hit recursion limit)");
431
432	switch (SvTYPE (rv))
433	{
434	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
435	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
436
437	default:
438	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
439	SvPV_nolen (sv));
440	}
441	}
442	else if (!SvOK (sv))	697	else if (!SvOK (sv))
443	encode_str (enc, "null", 4, 0);	698	encode_str (enc, "null", 4, 0);
444	else	699	else
445	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	700	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
446	SvPV_nolen (sv), SvFLAGS (sv));	701	SvPV_nolen (sv), SvFLAGS (sv));
447	}	702	}
448		703
449	static SV *	704	static SV *
450	encode_json (SV *scalar, UV flags)	705	encode_json (SV *scalar, JSON *json)
451	{	706	{
		707	enc_t enc;
		708
452	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	709	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
453	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	710	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
454		711
455	enc_t enc;	712	enc.json = *json;
456	enc.flags = flags;
457	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	713	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
458	enc.cur = SvPVX (enc.sv);	714	enc.cur = SvPVX (enc.sv);
459	enc.end = SvEND (enc.sv);	715	enc.end = SvEND (enc.sv);
460	enc.indent = 0;	716	enc.indent = 0;
461	enc.max_depth = 0x7fffffffUL;	717	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		718	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		719	: enc.json.flags & F_LATIN1 ? 0x000100UL
		720	: 0x110000UL;
462		721
463	SvPOK_only (enc.sv);	722	SvPOK_only (enc.sv);
464	encode_sv (&enc, scalar);	723	encode_sv (&enc, scalar);
465		724
		725	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		726	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		727
466	if (!(flags & (F_ASCII | F_UTF8)))	728	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
467	SvUTF8_on (enc.sv);	729	SvUTF8_on (enc.sv);
468		730
469	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
470
471	if (enc.flags & F_SHRINK)	731	if (enc.json.flags & F_SHRINK)
472	shrink (enc.sv);	732	shrink (enc.sv);
473		733
474	return enc.sv;	734	return enc.sv;
475	}	735	}
476		736
…		…
481	typedef struct	741	typedef struct
482	{	742	{
483	char *cur; // current parser pointer	743	char *cur; // current parser pointer
484	char *end; // end of input string	744	char *end; // end of input string
485	const char *err; // parse error, if != 0	745	const char *err; // parse error, if != 0
486	UV flags; // F_*	746	JSON json;
		747	U32 depth; // recursion depth
		748	U32 maxdepth; // recursion depth limit
487	} dec_t;	749	} dec_t;
488		750
489	static void	751	INLINE void
		752	decode_comment (dec_t *dec)
		753	{
		754	// only '#'-style comments allowed a.t.m.
		755
		756	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		757	++dec->cur;
		758	}
		759
		760	INLINE void
490	decode_ws (dec_t *dec)	761	decode_ws (dec_t *dec)
491	{	762	{
492	for (;;)	763	for (;;)
493	{	764	{
494	char ch = *dec->cur;	765	char ch = *dec->cur;
495		766
496	if (ch > 0x20	767	if (ch > 0x20)
		768	{
		769	if (expect_false (ch == '#'))
		770	{
		771	if (dec->json.flags & F_RELAXED)
		772	decode_comment (dec);
		773	else
		774	break;
		775	}
		776	else
		777	break;
		778	}
497	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	779	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
498	break;	780	break; // parse error, but let higher level handle it, gives better error messages
499		781
500	++dec->cur;	782	++dec->cur;
501	}	783	}
502	}	784	}
503		785
504	#define ERR(reason) SB dec->err = reason; goto fail; SE	786	#define ERR(reason) SB dec->err = reason; goto fail; SE
		787
505	#define EXPECT_CH(ch) SB \	788	#define EXPECT_CH(ch) SB \
506	if (*dec->cur != ch) \	789	if (*dec->cur != ch) \
507	ERR (# ch " expected"); \	790	ERR (# ch " expected"); \
508	++dec->cur; \	791	++dec->cur; \
509	SE	792	SE
510		793
		794	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		795	#define DEC_DEC_DEPTH --dec->depth
		796
511	static SV *decode_sv (dec_t *dec);	797	static SV *decode_sv (dec_t *dec);
512		798
513	static signed char decode_hexdigit[256];	799	static signed char decode_hexdigit[256];
514		800
515	static UV	801	static UV
516	decode_4hex (dec_t *dec)	802	decode_4hex (dec_t *dec)
517	{	803	{
518	signed char d1, d2, d3, d4;	804	signed char d1, d2, d3, d4;
519	unsigned char *cur = (unsigned char *)dec->cur;	805	unsigned char *cur = (unsigned char *)dec->cur;
520		806
521	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	807	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
522	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	808	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
523	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	809	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
524	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	810	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
525		811
526	dec->cur += 4;	812	dec->cur += 4;
527		813
528	return ((UV)d1) << 12	814	return ((UV)d1) << 12
529	| ((UV)d2) << 8	815	| ((UV)d2) << 8
…		…
537	static SV *	823	static SV *
538	decode_str (dec_t *dec)	824	decode_str (dec_t *dec)
539	{	825	{
540	SV *sv = 0;	826	SV *sv = 0;
541	int utf8 = 0;	827	int utf8 = 0;
		828	char *dec_cur = dec->cur;
542		829
543	do	830	do
544	{	831	{
545	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	832	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
546	char *cur = buf;	833	char *cur = buf;
547		834
548	do	835	do
549	{	836	{
550	unsigned char ch = *(unsigned char *)dec->cur++;	837	unsigned char ch = *(unsigned char *)dec_cur++;
551		838
552	if (ch == '"')	839	if (expect_false (ch == '"'))
553	{	840	{
554	--dec->cur;	841	--dec_cur;
555	break;	842	break;
556	}	843	}
557	else if (ch == '\\')	844	else if (expect_false (ch == '\\'))
558	{	845	{
559	switch (*dec->cur)	846	switch (*dec_cur)
560	{	847	{
561	case '\\':	848	case '\\':
562	case '/':	849	case '/':
563	case '"': *cur++ = *dec->cur++; break;	850	case '"': *cur++ = *dec_cur++; break;
564		851
565	case 'b': ++dec->cur; *cur++ = '\010'; break;	852	case 'b': ++dec_cur; *cur++ = '\010'; break;
566	case 't': ++dec->cur; *cur++ = '\011'; break;	853	case 't': ++dec_cur; *cur++ = '\011'; break;
567	case 'n': ++dec->cur; *cur++ = '\012'; break;	854	case 'n': ++dec_cur; *cur++ = '\012'; break;
568	case 'f': ++dec->cur; *cur++ = '\014'; break;	855	case 'f': ++dec_cur; *cur++ = '\014'; break;
569	case 'r': ++dec->cur; *cur++ = '\015'; break;	856	case 'r': ++dec_cur; *cur++ = '\015'; break;
570		857
571	case 'u':	858	case 'u':
572	{	859	{
573	UV lo, hi;	860	UV lo, hi;
574	++dec->cur;	861	++dec_cur;
575		862
		863	dec->cur = dec_cur;
576	hi = decode_4hex (dec);	864	hi = decode_4hex (dec);
		865	dec_cur = dec->cur;
577	if (hi == (UV)-1)	866	if (hi == (UV)-1)
578	goto fail;	867	goto fail;
579		868
580	// possibly a surrogate pair	869	// possibly a surrogate pair
581	if (hi >= 0xd800)	870	if (hi >= 0xd800)
582	if (hi < 0xdc00)	871	if (hi < 0xdc00)
583	{	872	{
584	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	873	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
585	ERR ("missing low surrogate character in surrogate pair");	874	ERR ("missing low surrogate character in surrogate pair");
586		875
587	dec->cur += 2;	876	dec_cur += 2;
588		877
		878	dec->cur = dec_cur;
589	lo = decode_4hex (dec);	879	lo = decode_4hex (dec);
		880	dec_cur = dec->cur;
590	if (lo == (UV)-1)	881	if (lo == (UV)-1)
591	goto fail;	882	goto fail;
592		883
593	if (lo < 0xdc00 || lo >= 0xe000)	884	if (lo < 0xdc00 || lo >= 0xe000)
594	ERR ("surrogate pair expected");	885	ERR ("surrogate pair expected");
…		…
600		891
601	if (hi >= 0x80)	892	if (hi >= 0x80)
602	{	893	{
603	utf8 = 1;	894	utf8 = 1;
604		895
605	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	896	cur = encode_utf8 (cur, hi);
606	}	897	}
607	else	898	else
608	*cur++ = hi;	899	*cur++ = hi;
609	}	900	}
610	break;	901	break;
611		902
612	default:	903	default:
613	--dec->cur;	904	--dec_cur;
614	ERR ("illegal backslash escape sequence in string");	905	ERR ("illegal backslash escape sequence in string");
615	}	906	}
616	}	907	}
617	else if (ch >= 0x20 && ch <= 0x7f)	908	else if (expect_true (ch >= 0x20 && ch < 0x80))
618	*cur++ = ch;	909	*cur++ = ch;
619	else if (ch >= 0x80)	910	else if (ch >= 0x80)
620	{	911	{
621	--dec->cur;
622
623	STRLEN clen;	912	STRLEN clen;
624	UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY);	913	UV uch;
		914
		915	--dec_cur;
		916
		917	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
625	if (clen == (STRLEN)-1)	918	if (clen == (STRLEN)-1)
626	ERR ("malformed UTF-8 character in JSON string");	919	ERR ("malformed UTF-8 character in JSON string");
627		920
628	do	921	do
629	{
630	*cur++ = *dec->cur++;	922	*cur++ = *dec_cur++;
631	}
632	while (--clen);	923	while (--clen);
633		924
634	utf8 = 1;	925	utf8 = 1;
635	}	926	}
636	else if (!ch)
637	ERR ("unexpected end of string while parsing json string");
638	else	927	else
		928	{
		929	--dec_cur;
		930
		931	if (!ch)
		932	ERR ("unexpected end of string while parsing JSON string");
		933	else
639	ERR ("invalid character encountered");	934	ERR ("invalid character encountered while parsing JSON string");
640		935	}
641	}	936	}
642	while (cur < buf + SHORT_STRING_LEN);	937	while (cur < buf + SHORT_STRING_LEN);
643		938
		939	{
644	STRLEN len = cur - buf;	940	STRLEN len = cur - buf;
645		941
646	if (sv)	942	if (sv)
647	{	943	{
648	SvGROW (sv, SvCUR (sv) + len + 1);	944	SvGROW (sv, SvCUR (sv) + len + 1);
649	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	945	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
650	SvCUR_set (sv, SvCUR (sv) + len);	946	SvCUR_set (sv, SvCUR (sv) + len);
651	}	947	}
652	else	948	else
653	sv = newSVpvn (buf, len);	949	sv = newSVpvn (buf, len);
654	}	950	}
		951	}
655	while (*dec->cur != '"');	952	while (*dec_cur != '"');
656		953
657	++dec->cur;	954	++dec_cur;
658		955
659	if (sv)	956	if (sv)
660	{	957	{
661	SvPOK_only (sv);	958	SvPOK_only (sv);
662	*SvEND (sv) = 0;	959	*SvEND (sv) = 0;
…		…
665	SvUTF8_on (sv);	962	SvUTF8_on (sv);
666	}	963	}
667	else	964	else
668	sv = newSVpvn ("", 0);	965	sv = newSVpvn ("", 0);
669		966
		967	dec->cur = dec_cur;
670	return sv;	968	return sv;
671		969
672	fail:	970	fail:
		971	dec->cur = dec_cur;
673	return 0;	972	return 0;
674	}	973	}
675		974
676	static SV *	975	static SV *
677	decode_num (dec_t *dec)	976	decode_num (dec_t *dec)
…		…
735	is_nv = 1;	1034	is_nv = 1;
736	}	1035	}
737		1036
738	if (!is_nv)	1037	if (!is_nv)
739	{	1038	{
740	UV uv;	1039	int len = dec->cur - start;
741	int numtype = grok_number (start, dec->cur - start, &uv);	1040
742	if (numtype & IS_NUMBER_IN_UV)	1041	// special case the rather common 1..5-digit-int case
743	if (numtype & IS_NUMBER_NEG)	1042	if (*start == '-')
		1043	switch (len)
744	{	1044	{
745	if (uv < (UV)IV_MIN)	1045	case 2: return newSViv (-( start [1] - '0' * 1));
746	return newSViv (-(IV)uv);	1046	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1047	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1048	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1049	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
747	}	1050	}
		1051	else
		1052	switch (len)
		1053	{
		1054	case 1: return newSViv ( start [0] - '0' * 1);
		1055	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1056	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1057	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1058	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1059	}
		1060
		1061	{
		1062	UV uv;
		1063	int numtype = grok_number (start, len, &uv);
		1064	if (numtype & IS_NUMBER_IN_UV)
		1065	if (numtype & IS_NUMBER_NEG)
		1066	{
		1067	if (uv < (UV)IV_MIN)
		1068	return newSViv (-(IV)uv);
		1069	}
748	else	1070	else
749	return newSVuv (uv);	1071	return newSVuv (uv);
750	}	1072	}
751		1073
		1074	len -= *start == '-' ? 1 : 0;
		1075
		1076	// does not fit into IV or UV, try NV
		1077	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1078	#if defined (LDBL_DIG)
		1079	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1080	#endif
		1081	)
		1082	// fits into NV without loss of precision
		1083	return newSVnv (Atof (start));
		1084
		1085	// everything else fails, convert it to a string
		1086	return newSVpvn (start, dec->cur - start);
		1087	}
		1088
		1089	// loss of precision here
752	return newSVnv (Atof (start));	1090	return newSVnv (Atof (start));
753		1091
754	fail:	1092	fail:
755	return 0;	1093	return 0;
756	}	1094	}
…		…
758	static SV *	1096	static SV *
759	decode_av (dec_t *dec)	1097	decode_av (dec_t *dec)
760	{	1098	{
761	AV *av = newAV ();	1099	AV *av = newAV ();
762		1100
		1101	DEC_INC_DEPTH;
763	decode_ws (dec);	1102	decode_ws (dec);
		1103
764	if (*dec->cur == ']')	1104	if (*dec->cur == ']')
765	++dec->cur;	1105	++dec->cur;
766	else	1106	else
767	for (;;)	1107	for (;;)
768	{	1108	{
…		…
784		1124
785	if (*dec->cur != ',')	1125	if (*dec->cur != ',')
786	ERR (", or ] expected while parsing array");	1126	ERR (", or ] expected while parsing array");
787		1127
788	++dec->cur;	1128	++dec->cur;
		1129
		1130	decode_ws (dec);
		1131
		1132	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1133	{
		1134	++dec->cur;
		1135	break;
		1136	}
789	}	1137	}
790		1138
		1139	DEC_DEC_DEPTH;
791	return newRV_noinc ((SV *)av);	1140	return newRV_noinc ((SV *)av);
792		1141
793	fail:	1142	fail:
794	SvREFCNT_dec (av);	1143	SvREFCNT_dec (av);
		1144	DEC_DEC_DEPTH;
795	return 0;	1145	return 0;
796	}	1146	}
797		1147
798	static SV *	1148	static SV *
799	decode_hv (dec_t *dec)	1149	decode_hv (dec_t *dec)
800	{	1150	{
		1151	SV *sv;
801	HV *hv = newHV ();	1152	HV *hv = newHV ();
802		1153
		1154	DEC_INC_DEPTH;
803	decode_ws (dec);	1155	decode_ws (dec);
		1156
804	if (*dec->cur == '}')	1157	if (*dec->cur == '}')
805	++dec->cur;	1158	++dec->cur;
806	else	1159	else
807	for (;;)	1160	for (;;)
808	{	1161	{
		1162	EXPECT_CH ('"');
		1163
		1164	// heuristic: assume that
		1165	// a) decode_str + hv_store_ent are abysmally slow.
		1166	// b) most hash keys are short, simple ascii text.
		1167	// => try to "fast-match" such strings to avoid
		1168	// the overhead of decode_str + hv_store_ent.
		1169	{
809	SV *key, *value;	1170	SV *value;
		1171	char *p = dec->cur;
		1172	char *e = p + 24; // only try up to 24 bytes
810		1173
811	decode_ws (dec); EXPECT_CH ('"');	1174	for (;;)
812
813	key = decode_str (dec);
814	if (!key)
815	goto fail;
816
817	decode_ws (dec); EXPECT_CH (':');
818
819	value = decode_sv (dec);
820	if (!value)
821	{	1175	{
		1176	// the >= 0x80 is true on most architectures
		1177	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1178	{
		1179	// slow path, back up and use decode_str
		1180	SV *key = decode_str (dec);
		1181	if (!key)
		1182	goto fail;
		1183
		1184	decode_ws (dec); EXPECT_CH (':');
		1185
		1186	decode_ws (dec);
		1187	value = decode_sv (dec);
		1188	if (!value)
		1189	{
		1190	SvREFCNT_dec (key);
		1191	goto fail;
		1192	}
		1193
		1194	hv_store_ent (hv, key, value, 0);
822	SvREFCNT_dec (key);	1195	SvREFCNT_dec (key);
		1196
		1197	break;
		1198	}
		1199	else if (*p == '"')
		1200	{
		1201	// fast path, got a simple key
		1202	char *key = dec->cur;
		1203	int len = p - key;
		1204	dec->cur = p + 1;
		1205
		1206	decode_ws (dec); EXPECT_CH (':');
		1207
		1208	decode_ws (dec);
		1209	value = decode_sv (dec);
		1210	if (!value)
823	goto fail;	1211	goto fail;
		1212
		1213	hv_store (hv, key, len, value, 0);
		1214
		1215	break;
		1216	}
		1217
		1218	++p;
824	}	1219	}
825		1220	}
826	//TODO: optimise
827	hv_store_ent (hv, key, value, 0);
828		1221
829	decode_ws (dec);	1222	decode_ws (dec);
830		1223
831	if (*dec->cur == '}')	1224	if (*dec->cur == '}')
832	{	1225	{
…		…
836		1229
837	if (*dec->cur != ',')	1230	if (*dec->cur != ',')
838	ERR (", or } expected while parsing object/hash");	1231	ERR (", or } expected while parsing object/hash");
839		1232
840	++dec->cur;	1233	++dec->cur;
		1234
		1235	decode_ws (dec);
		1236
		1237	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1238	{
		1239	++dec->cur;
		1240	break;
		1241	}
841	}	1242	}
842		1243
		1244	DEC_DEC_DEPTH;
843	return newRV_noinc ((SV *)hv);	1245	sv = newRV_noinc ((SV *)hv);
		1246
		1247	// check filter callbacks
		1248	if (dec->json.flags & F_HOOK)
		1249	{
		1250	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1251	{
		1252	HE *cb, *he;
		1253
		1254	hv_iterinit (hv);
		1255	he = hv_iternext (hv);
		1256	hv_iterinit (hv);
		1257
		1258	// the next line creates a mortal sv each time its called.
		1259	// might want to optimise this for common cases.
		1260	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1261
		1262	if (cb)
		1263	{
		1264	dSP;
		1265	int count;
		1266
		1267	ENTER; SAVETMPS; PUSHMARK (SP);
		1268	XPUSHs (HeVAL (he));
		1269
		1270	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1271
		1272	if (count == 1)
		1273	{
		1274	sv = newSVsv (POPs);
		1275	FREETMPS; LEAVE;
		1276	return sv;
		1277	}
		1278
		1279	FREETMPS; LEAVE;
		1280	}
		1281	}
		1282
		1283	if (dec->json.cb_object)
		1284	{
		1285	dSP;
		1286	int count;
		1287
		1288	ENTER; SAVETMPS; PUSHMARK (SP);
		1289	XPUSHs (sv_2mortal (sv));
		1290
		1291	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1292
		1293	if (count == 1)
		1294	{
		1295	sv = newSVsv (POPs);
		1296	FREETMPS; LEAVE;
		1297	return sv;
		1298	}
		1299
		1300	SvREFCNT_inc (sv);
		1301	FREETMPS; LEAVE;
		1302	}
		1303	}
		1304
		1305	return sv;
844		1306
845	fail:	1307	fail:
846	SvREFCNT_dec (hv);	1308	SvREFCNT_dec (hv);
		1309	DEC_DEC_DEPTH;
847	return 0;	1310	return 0;
848	}	1311	}
849		1312
850	static SV *	1313	static SV *
851	decode_sv (dec_t *dec)	1314	decode_sv (dec_t *dec)
852	{	1315	{
853	decode_ws (dec);	1316	// the beauty of JSON: you need exactly one character lookahead
		1317	// to parse everything.
854	switch (*dec->cur)	1318	switch (*dec->cur)
855	{	1319	{
856	case '"': ++dec->cur; return decode_str (dec);	1320	case '"': ++dec->cur; return decode_str (dec);
857	case '[': ++dec->cur; return decode_av (dec);	1321	case '[': ++dec->cur; return decode_av (dec);
858	case '{': ++dec->cur; return decode_hv (dec);	1322	case '{': ++dec->cur; return decode_hv (dec);
859		1323
860	case '-':	1324	case '-':
861	case '0': case '1': case '2': case '3': case '4':	1325	case '0': case '1': case '2': case '3': case '4':
862	case '5': case '6': case '7': case '8': case '9':	1326	case '5': case '6': case '7': case '8': case '9':
863	return decode_num (dec);	1327	return decode_num (dec);
864		1328
865	case 't':	1329	case 't':
866	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1330	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
867	{	1331	{
868	dec->cur += 4;	1332	dec->cur += 4;
869	return newSViv (1);	1333	#if JSON_SLOW
		1334	json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true);
		1335	#endif
		1336	return SvREFCNT_inc (json_true);
870	}	1337	}
871	else	1338	else
872	ERR ("'true' expected");	1339	ERR ("'true' expected");
873		1340
874	break;	1341	break;
875		1342
876	case 'f':	1343	case 'f':
877	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1344	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
878	{	1345	{
879	dec->cur += 5;	1346	dec->cur += 5;
880	return newSViv (0);	1347	#if JSON_SLOW
		1348	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
		1349	#endif
		1350	return SvREFCNT_inc (json_false);
881	}	1351	}
882	else	1352	else
883	ERR ("'false' expected");	1353	ERR ("'false' expected");
884		1354
885	break;	1355	break;
…		…
894	ERR ("'null' expected");	1364	ERR ("'null' expected");
895		1365
896	break;	1366	break;
897		1367
898	default:	1368	default:
899	ERR ("malformed json string, neither array, object, number, string or atom");	1369	ERR ("malformed JSON string, neither array, object, number, string or atom");
900	break;	1370	break;
901	}	1371	}
902		1372
903	fail:	1373	fail:
904	return 0;	1374	return 0;
905	}	1375	}
906		1376
907	static SV *	1377	static SV *
908	decode_json (SV *string, UV flags)	1378	decode_json (SV *string, JSON *json, UV *offset_return)
909	{	1379	{
		1380	dec_t dec;
		1381	UV offset;
910	SV *sv;	1382	SV *sv;
911		1383
		1384	SvGETMAGIC (string);
		1385	SvUPGRADE (string, SVt_PV);
		1386
		1387	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1388	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1389	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1390
912	if (flags & F_UTF8)	1391	if (json->flags & F_UTF8)
913	sv_utf8_downgrade (string, 0);	1392	sv_utf8_downgrade (string, 0);
914	else	1393	else
915	sv_utf8_upgrade (string);	1394	sv_utf8_upgrade (string);
916		1395
917	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1396	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
918		1397
919	dec_t dec;	1398	dec.json = *json;
920	dec.flags = flags;
921	dec.cur = SvPVX (string);	1399	dec.cur = SvPVX (string);
922	dec.end = SvEND (string);	1400	dec.end = SvEND (string);
923	dec.err = 0;	1401	dec.err = 0;
		1402	dec.depth = 0;
		1403	dec.maxdepth = DEC_DEPTH (dec.json.flags);
924		1404
		1405	if (dec.json.cb_object || dec.json.cb_sk_object)
		1406	dec.json.flags |= F_HOOK;
		1407
		1408	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1409
		1410	decode_ws (&dec);
925	sv = decode_sv (&dec);	1411	sv = decode_sv (&dec);
926		1412
		1413	if (!(offset_return || !sv))
		1414	{
		1415	// check for trailing garbage
		1416	decode_ws (&dec);
		1417
		1418	if (*dec.cur)
		1419	{
		1420	dec.err = "garbage after JSON object";
		1421	SvREFCNT_dec (sv);
		1422	sv = 0;
		1423	}
		1424	}
		1425
		1426	if (offset_return || !sv)
		1427	{
		1428	offset = dec.json.flags & F_UTF8
		1429	? dec.cur - SvPVX (string)
		1430	: utf8_distance (dec.cur, SvPVX (string));
		1431
		1432	if (offset_return)
		1433	*offset_return = offset;
		1434	}
		1435
927	if (!sv)	1436	if (!sv)
928	{	1437	{
929	IV offset = dec.flags & F_UTF8
930	? dec.cur - SvPVX (string)
931	: utf8_distance (dec.cur, SvPVX (string));
932	SV *uni = sv_newmortal ();	1438	SV *uni = sv_newmortal ();
933		1439
934	// horrible hack to silence warning inside pv_uni_display	1440	// horrible hack to silence warning inside pv_uni_display
935	COP cop = *PL_curcop;	1441	COP cop = *PL_curcop;
936	cop.cop_warnings = pWARN_NONE;	1442	cop.cop_warnings = pWARN_NONE;
…		…
938	SAVEVPTR (PL_curcop);	1444	SAVEVPTR (PL_curcop);
939	PL_curcop = &cop;	1445	PL_curcop = &cop;
940	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1446	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
941	LEAVE;	1447	LEAVE;
942		1448
943	croak ("%s, at character offset %d (%s)",	1449	croak ("%s, at character offset %d [\"%s\"]",
944	dec.err,	1450	dec.err,
945	(int)offset,	1451	(int)offset,
946	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1452	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
947	}	1453	}
948		1454
949	sv = sv_2mortal (sv);	1455	sv = sv_2mortal (sv);
950		1456
951	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1457	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
952	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1458	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
953		1459
954	return sv;	1460	return sv;
955	}	1461	}
956		1462
…		…
961		1467
962	BOOT:	1468	BOOT:
963	{	1469	{
964	int i;	1470	int i;
965		1471
966	memset (decode_hexdigit, 0xff, 256);
967	for (i = 10; i--; )	1472	for (i = 0; i < 256; ++i)
968	decode_hexdigit ['0' + i] = i;	1473	decode_hexdigit [i] =
		1474	i >= '0' && i <= '9' ? i - '0'
		1475	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1476	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1477	: -1;
969		1478
970	for (i = 7; i--; )
971	{
972	decode_hexdigit ['a' + i] = 10 + i;
973	decode_hexdigit ['A' + i] = 10 + i;
974	}
975
976	json_stash = gv_stashpv ("JSON::XS", 1);	1479	json_stash = gv_stashpv ("JSON::XS" , 1);
		1480	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1481
		1482	json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true );
		1483	json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false);
977	}	1484	}
978		1485
979	PROTOTYPES: DISABLE	1486	PROTOTYPES: DISABLE
980		1487
981	SV *new (char *dummy)	1488	void CLONE (...)
982	CODE:	1489	CODE:
983	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1490	json_stash = 0;
		1491	json_boolean_stash = 0;
		1492
		1493	void new (char *klass)
		1494	PPCODE:
		1495	{
		1496	SV *pv = NEWSV (0, sizeof (JSON));
		1497	SvPOK_only (pv);
		1498	Zero (SvPVX (pv), 1, JSON);
		1499	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1500	XPUSHs (sv_2mortal (sv_bless (
		1501	newRV_noinc (pv),
		1502	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1503	)));
		1504	}
		1505
		1506	void ascii (JSON *self, int enable = 1)
		1507	ALIAS:
		1508	ascii = F_ASCII
		1509	latin1 = F_LATIN1
		1510	utf8 = F_UTF8
		1511	indent = F_INDENT
		1512	canonical = F_CANONICAL
		1513	space_before = F_SPACE_BEFORE
		1514	space_after = F_SPACE_AFTER
		1515	pretty = F_PRETTY
		1516	allow_nonref = F_ALLOW_NONREF
		1517	shrink = F_SHRINK
		1518	allow_blessed = F_ALLOW_BLESSED
		1519	convert_blessed = F_CONV_BLESSED
		1520	relaxed = F_RELAXED
		1521	PPCODE:
		1522	{
		1523	if (enable)
		1524	self->flags |= ix;
		1525	else
		1526	self->flags &= ~ix;
		1527
		1528	XPUSHs (ST (0));
		1529	}
		1530
		1531	void get_ascii (JSON *self)
		1532	ALIAS:
		1533	get_ascii = F_ASCII
		1534	get_latin1 = F_LATIN1
		1535	get_utf8 = F_UTF8
		1536	get_indent = F_INDENT
		1537	get_canonical = F_CANONICAL
		1538	get_space_before = F_SPACE_BEFORE
		1539	get_space_after = F_SPACE_AFTER
		1540	get_allow_nonref = F_ALLOW_NONREF
		1541	get_shrink = F_SHRINK
		1542	get_allow_blessed = F_ALLOW_BLESSED
		1543	get_convert_blessed = F_CONV_BLESSED
		1544	get_relaxed = F_RELAXED
		1545	PPCODE:
		1546	XPUSHs (boolSV (self->flags & ix));
		1547
		1548	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1549	PPCODE:
		1550	{
		1551	UV log2 = 0;
		1552
		1553	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1554
		1555	while ((1UL << log2) < max_depth)
		1556	++log2;
		1557
		1558	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1559
		1560	XPUSHs (ST (0));
		1561	}
		1562
		1563	U32 get_max_depth (JSON *self)
		1564	CODE:
		1565	RETVAL = DEC_DEPTH (self->flags);
984	OUTPUT:	1566	OUTPUT:
985	RETVAL	1567	RETVAL
986		1568
987	SV *ascii (SV *self, int enable = 1)	1569	void max_size (JSON *self, UV max_size = 0)
988	ALIAS:	1570	PPCODE:
989	ascii = F_ASCII	1571	{
990	utf8 = F_UTF8	1572	UV log2 = 0;
991	indent = F_INDENT	1573
992	canonical = F_CANONICAL	1574	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
993	space_before = F_SPACE_BEFORE	1575	if (max_size == 1) max_size = 2;
994	space_after = F_SPACE_AFTER	1576
995	pretty = F_PRETTY	1577	while ((1UL << log2) < max_size)
996	allow_nonref = F_ALLOW_NONREF	1578	++log2;
997	shrink = F_SHRINK	1579
		1580	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1581
		1582	XPUSHs (ST (0));
		1583	}
		1584
		1585	int get_max_size (JSON *self)
998	CODE:	1586	CODE:
999	{	1587	RETVAL = DEC_SIZE (self->flags);
1000	UV *uv = SvJSON (self);
1001	if (enable)
1002	*uv |= ix;
1003	else
1004	*uv &= ~ix;
1005
1006	RETVAL = newSVsv (self);
1007	}
1008	OUTPUT:	1588	OUTPUT:
1009	RETVAL	1589	RETVAL
1010		1590
1011	void encode (SV *self, SV *scalar)	1591	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1012	PPCODE:	1592	PPCODE:
1013	XPUSHs (encode_json (scalar, *SvJSON (self)));	1593	{
		1594	SvREFCNT_dec (self->cb_object);
		1595	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1014		1596
1015	void decode (SV *self, SV *jsonstr)	1597	XPUSHs (ST (0));
		1598	}
		1599
		1600	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1016	PPCODE:	1601	PPCODE:
		1602	{
		1603	if (!self->cb_sk_object)
		1604	self->cb_sk_object = newHV ();
		1605
		1606	if (SvOK (cb))
		1607	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1608	else
		1609	{
		1610	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1611
		1612	if (!HvKEYS (self->cb_sk_object))
		1613	{
		1614	SvREFCNT_dec (self->cb_sk_object);
		1615	self->cb_sk_object = 0;
		1616	}
		1617	}
		1618
		1619	XPUSHs (ST (0));
		1620	}
		1621
		1622	void encode (JSON *self, SV *scalar)
		1623	PPCODE:
		1624	XPUSHs (encode_json (scalar, self));
		1625
		1626	void decode (JSON *self, SV *jsonstr)
		1627	PPCODE:
1017	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1628	XPUSHs (decode_json (jsonstr, self, 0));
		1629
		1630	void decode_prefix (JSON *self, SV *jsonstr)
		1631	PPCODE:
		1632	{
		1633	UV offset;
		1634	EXTEND (SP, 2);
		1635	PUSHs (decode_json (jsonstr, self, &offset));
		1636	PUSHs (sv_2mortal (newSVuv (offset)));
		1637	}
		1638
		1639	void DESTROY (JSON *self)
		1640	CODE:
		1641	SvREFCNT_dec (self->cb_sk_object);
		1642	SvREFCNT_dec (self->cb_object);
1018		1643
1019	PROTOTYPES: ENABLE	1644	PROTOTYPES: ENABLE
1020		1645
1021	void to_json (SV *scalar)	1646	void encode_json (SV *scalar)
1022	PPCODE:	1647	PPCODE:
		1648	{
		1649	JSON json = { F_DEFAULT | F_UTF8 };
1023	XPUSHs (encode_json (scalar, F_UTF8));	1650	XPUSHs (encode_json (scalar, &json));
		1651	}
1024		1652
1025	void from_json (SV *jsonstr)	1653	void decode_json (SV *jsonstr)
1026	PPCODE:	1654	PPCODE:
		1655	{
		1656	JSON json = { F_DEFAULT | F_UTF8 };
1027	XPUSHs (decode_json (jsonstr, F_UTF8));	1657	XPUSHs (decode_json (jsonstr, &json, 0));
		1658	}
1028		1659

Diff Legend

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