ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.15 by root, Sun Mar 25 02:32:40 2007 UTC vs.
Revision 1.84 by root, Thu Mar 27 06:37:35 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	#define F_ALLOW_UNKNOWN 0x00002000UL
		37
		38	#define F_MAXDEPTH 0xf8000000UL
		39	#define S_MAXDEPTH 27
		40	#define F_MAXSIZE 0x01f00000UL
		41	#define S_MAXSIZE 20
		42	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
		43
		44	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
		45	#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
17		46
18	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	47	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
19	#define F_DEFAULT 0	48	#define F_DEFAULT (9UL << S_MAXDEPTH)
20		49
21	#define INIT_SIZE 32 // initial scalar size to be allocated	50	#define INIT_SIZE 32 // initial scalar size to be allocated
22	#define INDENT_STEP 3 // spaces per indentation level	51	#define INDENT_STEP 3 // spaces per indentation level
23		52
24	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
25	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	53	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
26		54
27	#define SB do {	55	#define SB do {
28	#define SE } while (0)	56	#define SE } while (0)
29		57
		58	#if __GNUC__ >= 3
		59	# define expect(expr,value) __builtin_expect ((expr), (value))
		60	# define INLINE static inline
		61	#else
		62	# define expect(expr,value) (expr)
		63	# define INLINE static
		64	#endif
		65
		66	#define expect_false(expr) expect ((expr) != 0, 0)
		67	#define expect_true(expr) expect ((expr) != 0, 1)
		68
		69	#define IN_RANGE_INC(type,val,beg,end) \
		70	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		71	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		72
		73	#ifdef USE_ITHREADS
		74	# define JSON_SLOW 1
		75	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		76	#else
		77	# define JSON_SLOW 0
		78	# define JSON_STASH json_stash
		79	#endif
		80
30	static HV *json_stash; // JSON::XS::	81	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		82	static SV *json_true, *json_false;
		83
		84	enum {
		85	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		86	INCR_M_STR, // inside string
		87	INCR_M_BS, // inside backslash
		88	INCR_M_JSON // outside anything, count nesting
		89	};
		90
		91	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		92
		93	typedef struct {
		94	U32 flags;
		95	SV *cb_object;
		96	HV *cb_sk_object;
		97
		98	// for the incremental parser
		99	SV *incr_text; // the source text so far
		100	STRLEN incr_pos; // the current offset into the text
		101	int incr_nest; // {[]}-nesting level
		102	int incr_mode;
		103	} JSON;
31		104
32	/////////////////////////////////////////////////////////////////////////////	105	/////////////////////////////////////////////////////////////////////////////
33	// utility functions	106	// utility functions
34		107
35	static UV *	108	INLINE SV *
36	SvJSON (SV *sv)	109	get_bool (const char *name)
37	{	110	{
38	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	111	SV *sv = get_sv (name, 1);
39	croak ("object is not of type JSON::XS");
40		112
41	return &SvUVX (SvRV (sv));	113	SvREADONLY_on (sv);
42	}	114	SvREADONLY_on (SvRV (sv));
43		115
44	static void	116	return sv;
		117	}
		118
		119	INLINE void
45	shrink (SV *sv)	120	shrink (SV *sv)
46	{	121	{
47	sv_utf8_downgrade (sv, 1);	122	sv_utf8_downgrade (sv, 1);
48	if (SvLEN (sv) > SvCUR (sv) + 1)	123	if (SvLEN (sv) > SvCUR (sv) + 1)
49	{	124	{
…		…
58	// decode an utf-8 character and return it, or (UV)-1 in	133	// decode an utf-8 character and return it, or (UV)-1 in
59	// case of an error.	134	// case of an error.
60	// we special-case "safe" characters from U+80 .. U+7FF,	135	// we special-case "safe" characters from U+80 .. U+7FF,
61	// but use the very good perl function to parse anything else.	136	// but use the very good perl function to parse anything else.
62	// note that we never call this function for a ascii codepoints	137	// note that we never call this function for a ascii codepoints
63	static UV	138	INLINE UV
64	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	139	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
65	{	140	{
66	if (s[0] > 0xdf || s[0] < 0xc2)	141	if (expect_true (len >= 2
67	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	142	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
68	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	143	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
69	{	144	{
70	*clen = 2;	145	*clen = 2;
71	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	146	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
72	}	147	}
73	else	148	else
74	return (UV)-1;	149	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
		150	}
		151
		152	// likewise for encoding, also never called for ascii codepoints
		153	// this function takes advantage of this fact, although current gccs
		154	// seem to optimise the check for >= 0x80 away anyways
		155	INLINE unsigned char *
		156	encode_utf8 (unsigned char *s, UV ch)
		157	{
		158	if (expect_false (ch < 0x000080))
		159	*s++ = ch;
		160	else if (expect_true (ch < 0x000800))
		161	*s++ = 0xc0 | ( ch >> 6),
		162	*s++ = 0x80 | ( ch & 0x3f);
		163	else if ( ch < 0x010000)
		164	*s++ = 0xe0 | ( ch >> 12),
		165	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		166	*s++ = 0x80 | ( ch & 0x3f);
		167	else if ( ch < 0x110000)
		168	*s++ = 0xf0 | ( ch >> 18),
		169	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		170	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		171	*s++ = 0x80 | ( ch & 0x3f);
		172
		173	return s;
75	}	174	}
76		175
77	/////////////////////////////////////////////////////////////////////////////	176	/////////////////////////////////////////////////////////////////////////////
78	// encoder	177	// encoder
79		178
…		…
81	typedef struct	180	typedef struct
82	{	181	{
83	char *cur; // SvPVX (sv) + current output position	182	char *cur; // SvPVX (sv) + current output position
84	char *end; // SvEND (sv)	183	char *end; // SvEND (sv)
85	SV *sv; // result scalar	184	SV *sv; // result scalar
86	UV flags; // F_*	185	JSON json;
87	int indent; // indentation level	186	U32 indent; // indentation level
88	int max_depth; // max. recursion level	187	U32 maxdepth; // max. indentation/recursion level
		188	UV limit; // escape character values >= this value when encoding
89	} enc_t;	189	} enc_t;
90		190
91	static void	191	INLINE void
92	need (enc_t *enc, STRLEN len)	192	need (enc_t *enc, STRLEN len)
93	{	193	{
94	if (enc->cur + len >= enc->end)	194	if (expect_false (enc->cur + len >= enc->end))
95	{	195	{
96	STRLEN cur = enc->cur - SvPVX (enc->sv);	196	STRLEN cur = enc->cur - SvPVX (enc->sv);
97	SvGROW (enc->sv, cur + len + 1);	197	SvGROW (enc->sv, cur + len + 1);
98	enc->cur = SvPVX (enc->sv) + cur;	198	enc->cur = SvPVX (enc->sv) + cur;
99	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv);	199	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
100	}	200	}
101	}	201	}
102		202
103	static void	203	INLINE void
104	encode_ch (enc_t *enc, char ch)	204	encode_ch (enc_t *enc, char ch)
105	{	205	{
106	need (enc, 1);	206	need (enc, 1);
107	*enc->cur++ = ch;	207	*enc->cur++ = ch;
108	}	208	}
…		…
116		216
117	while (str < end)	217	while (str < end)
118	{	218	{
119	unsigned char ch = *(unsigned char *)str;	219	unsigned char ch = *(unsigned char *)str;
120		220
121	if (ch >= 0x20 && ch < 0x80) // most common case	221	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
122	{	222	{
123	if (ch == '"') // but with slow exceptions	223	if (expect_false (ch == '"')) // but with slow exceptions
124	{	224	{
125	need (enc, len += 1);	225	need (enc, len += 1);
126	*enc->cur++ = '\\';	226	*enc->cur++ = '\\';
127	*enc->cur++ = '"';	227	*enc->cur++ = '"';
128	}	228	}
129	else if (ch == '\\')	229	else if (expect_false (ch == '\\'))
130	{	230	{
131	need (enc, len += 1);	231	need (enc, len += 1);
132	*enc->cur++ = '\\';	232	*enc->cur++ = '\\';
133	*enc->cur++ = '\\';	233	*enc->cur++ = '\\';
134	}	234	}
…		…
152	STRLEN clen;	252	STRLEN clen;
153	UV uch;	253	UV uch;
154		254
155	if (is_utf8)	255	if (is_utf8)
156	{	256	{
157	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
158	uch = decode_utf8 (str, end - str, &clen);	257	uch = decode_utf8 (str, end - str, &clen);
159	if (clen == (STRLEN)-1)	258	if (clen == (STRLEN)-1)
160	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	259	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
161	}	260	}
162	else	261	else
163	{	262	{
164	uch = ch;	263	uch = ch;
165	clen = 1;	264	clen = 1;
166	}	265	}
167		266
168	if (uch > 0x10FFFFUL)	267	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
169	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
170
171	if (uch < 0x80 || enc->flags & F_ASCII)
172	{	268	{
173	if (uch > 0xFFFFUL)	269	if (uch >= 0x10000UL)
174	{	270	{
		271	if (uch >= 0x110000UL)
		272	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		273
175	need (enc, len += 11);	274	need (enc, len += 11);
176	sprintf (enc->cur, "\\u%04x\\u%04x",	275	sprintf (enc->cur, "\\u%04x\\u%04x",
177	(int)((uch - 0x10000) / 0x400 + 0xD800),	276	(int)((uch - 0x10000) / 0x400 + 0xD800),
178	(int)((uch - 0x10000) % 0x400 + 0xDC00));	277	(int)((uch - 0x10000) % 0x400 + 0xDC00));
179	enc->cur += 12;	278	enc->cur += 12;
…		…
190	*enc->cur++ = hexdigit [(uch >> 0) & 15];	289	*enc->cur++ = hexdigit [(uch >> 0) & 15];
191	}	290	}
192		291
193	str += clen;	292	str += clen;
194	}	293	}
		294	else if (enc->json.flags & F_LATIN1)
		295	{
		296	*enc->cur++ = uch;
		297	str += clen;
		298	}
195	else if (is_utf8)	299	else if (is_utf8)
196	{	300	{
197	need (enc, len += clen);	301	need (enc, len += clen);
198	do	302	do
199	{	303	{
…		…
201	}	305	}
202	while (--clen);	306	while (--clen);
203	}	307	}
204	else	308	else
205	{	309	{
206	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	310	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
207	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	311	enc->cur = encode_utf8 (enc->cur, uch);
208	++str;	312	++str;
209	}	313	}
210	}	314	}
211	}	315	}
212	}	316	}
213		317
214	--len;	318	--len;
215	}	319	}
216	}	320	}
217		321
218	static void	322	INLINE void
219	encode_indent (enc_t *enc)	323	encode_indent (enc_t *enc)
220	{	324	{
221	if (enc->flags & F_INDENT)	325	if (enc->json.flags & F_INDENT)
222	{	326	{
223	int spaces = enc->indent * INDENT_STEP;	327	int spaces = enc->indent * INDENT_STEP;
224		328
225	need (enc, spaces);	329	need (enc, spaces);
226	memset (enc->cur, ' ', spaces);	330	memset (enc->cur, ' ', spaces);
227	enc->cur += spaces;	331	enc->cur += spaces;
228	}	332	}
229	}	333	}
230		334
231	static void	335	INLINE void
232	encode_space (enc_t *enc)	336	encode_space (enc_t *enc)
233	{	337	{
234	need (enc, 1);	338	need (enc, 1);
235	encode_ch (enc, ' ');	339	encode_ch (enc, ' ');
236	}	340	}
237		341
238	static void	342	INLINE void
239	encode_nl (enc_t *enc)	343	encode_nl (enc_t *enc)
240	{	344	{
241	if (enc->flags & F_INDENT)	345	if (enc->json.flags & F_INDENT)
242	{	346	{
243	need (enc, 1);	347	need (enc, 1);
244	encode_ch (enc, '\n');	348	encode_ch (enc, '\n');
245	}	349	}
246	}	350	}
247		351
248	static void	352	INLINE void
249	encode_comma (enc_t *enc)	353	encode_comma (enc_t *enc)
250	{	354	{
251	encode_ch (enc, ',');	355	encode_ch (enc, ',');
252		356
253	if (enc->flags & F_INDENT)	357	if (enc->json.flags & F_INDENT)
254	encode_nl (enc);	358	encode_nl (enc);
255	else if (enc->flags & F_SPACE_AFTER)	359	else if (enc->json.flags & F_SPACE_AFTER)
256	encode_space (enc);	360	encode_space (enc);
257	}	361	}
258		362
259	static void encode_sv (enc_t *enc, SV *sv);	363	static void encode_sv (enc_t *enc, SV *sv);
260		364
261	static void	365	static void
262	encode_av (enc_t *enc, AV *av)	366	encode_av (enc_t *enc, AV *av)
263	{	367	{
264	int i, len = av_len (av);	368	int i, len = av_len (av);
265		369
266	encode_ch (enc, '['); encode_nl (enc);	370	if (enc->indent >= enc->maxdepth)
267	++enc->indent;	371	croak ("data structure too deep (hit recursion limit)");
268		372
		373	encode_ch (enc, '[');
		374
		375	if (len >= 0)
		376	{
		377	encode_nl (enc); ++enc->indent;
		378
269	for (i = 0; i <= len; ++i)	379	for (i = 0; i <= len; ++i)
270	{	380	{
		381	SV **svp = av_fetch (av, i, 0);
		382
271	encode_indent (enc);	383	encode_indent (enc);
272	encode_sv (enc, *av_fetch (av, i, 0));
273		384
		385	if (svp)
		386	encode_sv (enc, *svp);
		387	else
		388	encode_str (enc, "null", 4, 0);
		389
274	if (i < len)	390	if (i < len)
275	encode_comma (enc);	391	encode_comma (enc);
276	}	392	}
277		393
		394	encode_nl (enc); --enc->indent; encode_indent (enc);
		395	}
		396
278	encode_nl (enc);	397	encode_ch (enc, ']');
279
280	--enc->indent;
281	encode_indent (enc); encode_ch (enc, ']');
282	}	398	}
283		399
284	static void	400	static void
285	encode_he (enc_t *enc, HE *he)	401	encode_hk (enc_t *enc, HE *he)
286	{	402	{
287	encode_ch (enc, '"');	403	encode_ch (enc, '"');
288		404
289	if (HeKLEN (he) == HEf_SVKEY)	405	if (HeKLEN (he) == HEf_SVKEY)
290	{	406	{
…		…
300	else	416	else
301	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	417	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
302		418
303	encode_ch (enc, '"');	419	encode_ch (enc, '"');
304		420
305	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	421	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
306	encode_ch (enc, ':');	422	encode_ch (enc, ':');
307	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	423	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
308	encode_sv (enc, HeVAL (he));
309	}	424	}
310		425
311	// compare hash entries, used when all keys are bytestrings	426	// compare hash entries, used when all keys are bytestrings
312	static int	427	static int
313	he_cmp_fast (const void *a_, const void *b_)	428	he_cmp_fast (const void *a_, const void *b_)
…		…
318	HE *b = *(HE **)b_;	433	HE *b = *(HE **)b_;
319		434
320	STRLEN la = HeKLEN (a);	435	STRLEN la = HeKLEN (a);
321	STRLEN lb = HeKLEN (b);	436	STRLEN lb = HeKLEN (b);
322		437
323	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	438	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
324	cmp = la - lb;	439	cmp = lb - la;
325		440
326	return cmp;	441	return cmp;
327	}	442	}
328		443
329	// compare hash entries, used when some keys are sv's or utf-x	444	// compare hash entries, used when some keys are sv's or utf-x
330	static int	445	static int
331	he_cmp_slow (const void *a, const void *b)	446	he_cmp_slow (const void *a, const void *b)
332	{	447	{
333	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	448	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
334	}	449	}
335		450
336	static void	451	static void
337	encode_hv (enc_t *enc, HV *hv)	452	encode_hv (enc_t *enc, HV *hv)
338	{	453	{
339	int count, i;	454	HE *he;
340		455
341	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	456	if (enc->indent >= enc->maxdepth)
		457	croak ("data structure too deep (hit recursion limit)");
342		458
343	if ((count = hv_iterinit (hv)))	459	encode_ch (enc, '{');
344	{	460
345	// for canonical output we have to sort by keys first	461	// for canonical output we have to sort by keys first
346	// actually, this is mostly due to the stupid so-called	462	// actually, this is mostly due to the stupid so-called
347	// security workaround added somewhere in 5.8.x.	463	// security workaround added somewhere in 5.8.x.
348	// that randomises hash orderings	464	// that randomises hash orderings
349	if (enc->flags & F_CANONICAL)	465	if (enc->json.flags & F_CANONICAL)
		466	{
		467	int count = hv_iterinit (hv);
		468
		469	if (SvMAGICAL (hv))
350	{	470	{
		471	// need to count by iterating. could improve by dynamically building the vector below
		472	// but I don't care for the speed of this special case.
		473	// note also that we will run into undefined behaviour when the two iterations
		474	// do not result in the same count, something I might care for in some later release.
		475
		476	count = 0;
		477	while (hv_iternext (hv))
		478	++count;
		479
		480	hv_iterinit (hv);
		481	}
		482
		483	if (count)
		484	{
		485	int i, fast = 1;
		486	#if defined(__BORLANDC__) || defined(_MSC_VER)
		487	HE **hes = _alloca (count * sizeof (HE));
		488	#else
351	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	489	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
352	int fast = 1;	490	#endif
353		491
354	i = 0;	492	i = 0;
355	while ((he = hv_iternext (hv)))	493	while ((he = hv_iternext (hv)))
356	{	494	{
357	hes [i++] = he;	495	hes [i++] = he;
…		…
379		517
380	FREETMPS;	518	FREETMPS;
381	LEAVE;	519	LEAVE;
382	}	520	}
383		521
384	for (i = 0; i < count; ++i)	522	encode_nl (enc); ++enc->indent;
		523
		524	while (count--)
385	{	525	{
386	encode_indent (enc);	526	encode_indent (enc);
		527	he = hes [count];
387	encode_he (enc, hes [i]);	528	encode_hk (enc, he);
		529	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
388		530
389	if (i < count - 1)	531	if (count)
390	encode_comma (enc);	532	encode_comma (enc);
391	}	533	}
392		534
		535	encode_nl (enc); --enc->indent; encode_indent (enc);
		536	}
		537	}
		538	else
		539	{
		540	if (hv_iterinit (hv) || SvMAGICAL (hv))
		541	if ((he = hv_iternext (hv)))
		542	{
		543	encode_nl (enc); ++enc->indent;
		544
		545	for (;;)
		546	{
393	encode_nl (enc);	547	encode_indent (enc);
		548	encode_hk (enc, he);
		549	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		550
		551	if (!(he = hv_iternext (hv)))
		552	break;
		553
		554	encode_comma (enc);
		555	}
		556
		557	encode_nl (enc); --enc->indent; encode_indent (enc);
		558	}
		559	}
		560
		561	encode_ch (enc, '}');
		562	}
		563
		564	// encode objects, arrays and special \0=false and \1=true values.
		565	static void
		566	encode_rv (enc_t *enc, SV *sv)
		567	{
		568	svtype svt;
		569
		570	SvGETMAGIC (sv);
		571	svt = SvTYPE (sv);
		572
		573	if (expect_false (SvOBJECT (sv)))
		574	{
		575	HV *stash = !JSON_SLOW || json_boolean_stash
		576	? json_boolean_stash
		577	: gv_stashpv ("JSON::XS::Boolean", 1);
		578
		579	if (SvSTASH (sv) == stash)
		580	{
		581	if (SvIV (sv))
		582	encode_str (enc, "true", 4, 0);
		583	else
		584	encode_str (enc, "false", 5, 0);
394	}	585	}
395	else	586	else
396	{	587	{
397	SV *sv;	588	#if 0
398	HE *he = hv_iternext (hv);	589	if (0 && sv_derived_from (rv, "JSON::Literal"))
399
400	for (;;)
401	{	590	{
402	encode_indent (enc);	591	// not yet
403	encode_he (enc, he);
404
405	if (!(he = hv_iternext (hv)))
406	break;
407
408	encode_comma (enc);
409	}	592	}
		593	#endif
		594	if (enc->json.flags & F_CONV_BLESSED)
		595	{
		596	// we re-bless the reference to get overload and other niceties right
		597	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
410		598
		599	if (to_json)
		600	{
		601	dSP;
		602
		603	ENTER; SAVETMPS; PUSHMARK (SP);
		604	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		605
		606	// calling with G_SCALAR ensures that we always get a 1 return value
		607	PUTBACK;
		608	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		609	SPAGAIN;
		610
		611	// catch this surprisingly common error
		612	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		613	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		614
		615	sv = POPs;
		616	PUTBACK;
		617
411	encode_nl (enc);	618	encode_sv (enc, sv);
		619
		620	FREETMPS; LEAVE;
		621	}
		622	else if (enc->json.flags & F_ALLOW_BLESSED)
		623	encode_str (enc, "null", 4, 0);
		624	else
		625	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		626	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		627	}
		628	else if (enc->json.flags & F_ALLOW_BLESSED)
		629	encode_str (enc, "null", 4, 0);
		630	else
		631	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		632	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
412	}	633	}
413	}	634	}
		635	else if (svt == SVt_PVHV)
		636	encode_hv (enc, (HV *)sv);
		637	else if (svt == SVt_PVAV)
		638	encode_av (enc, (AV *)sv);
		639	else if (svt < SVt_PVAV)
		640	{
		641	STRLEN len = 0;
		642	char *pv = svt ? SvPV (sv, len) : 0;
414		643
415	--enc->indent; encode_indent (enc); encode_ch (enc, '}');	644	if (len == 1 && *pv == '1')
		645	encode_str (enc, "true", 4, 0);
		646	else if (len == 1 && *pv == '0')
		647	encode_str (enc, "false", 5, 0);
		648	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		649	encode_str (enc, "null", 4, 0);
		650	else
		651	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
		652	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		653	}
		654	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		655	encode_str (enc, "null", 4, 0);
		656	else
		657	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
		658	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
416	}	659	}
417		660
418	static void	661	static void
419	encode_sv (enc_t *enc, SV *sv)	662	encode_sv (enc_t *enc, SV *sv)
420	{	663	{
…		…
428	encode_str (enc, str, len, SvUTF8 (sv));	671	encode_str (enc, str, len, SvUTF8 (sv));
429	encode_ch (enc, '"');	672	encode_ch (enc, '"');
430	}	673	}
431	else if (SvNOKp (sv))	674	else if (SvNOKp (sv))
432	{	675	{
		676	// trust that perl will do the right thing w.r.t. JSON syntax.
433	need (enc, NV_DIG + 32);	677	need (enc, NV_DIG + 32);
434	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	678	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
435	enc->cur += strlen (enc->cur);	679	enc->cur += strlen (enc->cur);
436	}	680	}
437	else if (SvIOKp (sv))	681	else if (SvIOKp (sv))
438	{	682	{
		683	// we assume we can always read an IV as a UV and vice versa
		684	// we assume two's complement
		685	// we assume no aliasing issues in the union
		686	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		687	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		688	{
		689	// optimise the "small number case"
		690	// code will likely be branchless and use only a single multiplication
		691	// works for numbers up to 59074
		692	I32 i = SvIVX (sv);
		693	U32 u;
		694	char digit, nz = 0;
		695
439	need (enc, 64);	696	need (enc, 6);
		697
		698	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		699	u = i < 0 ? -i : i;
		700
		701	// convert to 4.28 fixed-point representation
		702	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		703
		704	// now output digit by digit, each time masking out the integer part
		705	// and multiplying by 5 while moving the decimal point one to the right,
		706	// resulting in a net multiplication by 10.
		707	// we always write the digit to memory but conditionally increment
		708	// the pointer, to enable the use of conditional move instructions.
		709	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		710	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		711	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		712	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		713	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		714	}
		715	else
		716	{
		717	// large integer, use the (rather slow) snprintf way.
		718	need (enc, IVUV_MAXCHARS);
440	enc->cur +=	719	enc->cur +=
441	SvIsUV(sv)	720	SvIsUV(sv)
442	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	721	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
443	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	722	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		723	}
444	}	724	}
445	else if (SvROK (sv))	725	else if (SvROK (sv))
446	{	726	encode_rv (enc, SvRV (sv));
447	SV *rv = SvRV (sv);	727	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
448
449	if (enc->indent >= enc->max_depth)
450	croak ("data structure too deep (hit recursion limit)");
451
452	switch (SvTYPE (rv))
453	{
454	case SVt_PVAV: encode_av (enc, (AV *)rv); break;
455	case SVt_PVHV: encode_hv (enc, (HV *)rv); break;
456
457	default:
458	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
459	SvPV_nolen (sv));
460	}
461	}
462	else if (!SvOK (sv))
463	encode_str (enc, "null", 4, 0);	728	encode_str (enc, "null", 4, 0);
464	else	729	else
465	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	730	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
466	SvPV_nolen (sv), SvFLAGS (sv));	731	SvPV_nolen (sv), SvFLAGS (sv));
467	}	732	}
468		733
469	static SV *	734	static SV *
470	encode_json (SV *scalar, UV flags)	735	encode_json (SV *scalar, JSON *json)
471	{	736	{
		737	enc_t enc;
		738
472	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	739	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
473	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	740	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
474		741
475	enc_t enc;	742	enc.json = *json;
476	enc.flags = flags;
477	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	743	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
478	enc.cur = SvPVX (enc.sv);	744	enc.cur = SvPVX (enc.sv);
479	enc.end = SvEND (enc.sv);	745	enc.end = SvEND (enc.sv);
480	enc.indent = 0;	746	enc.indent = 0;
481	enc.max_depth = 0x7fffffffUL;	747	enc.maxdepth = DEC_DEPTH (enc.json.flags);
		748	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		749	: enc.json.flags & F_LATIN1 ? 0x000100UL
		750	: 0x110000UL;
482		751
483	SvPOK_only (enc.sv);	752	SvPOK_only (enc.sv);
484	encode_sv (&enc, scalar);	753	encode_sv (&enc, scalar);
485		754
		755	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
		756	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
		757
486	if (!(flags & (F_ASCII | F_UTF8)))	758	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
487	SvUTF8_on (enc.sv);	759	SvUTF8_on (enc.sv);
488		760
489	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
490
491	if (enc.flags & F_SHRINK)	761	if (enc.json.flags & F_SHRINK)
492	shrink (enc.sv);	762	shrink (enc.sv);
493		763
494	return enc.sv;	764	return enc.sv;
495	}	765	}
496		766
…		…
501	typedef struct	771	typedef struct
502	{	772	{
503	char *cur; // current parser pointer	773	char *cur; // current parser pointer
504	char *end; // end of input string	774	char *end; // end of input string
505	const char *err; // parse error, if != 0	775	const char *err; // parse error, if != 0
506	UV flags; // F_*	776	JSON json;
		777	U32 depth; // recursion depth
		778	U32 maxdepth; // recursion depth limit
507	} dec_t;	779	} dec_t;
508		780
509	static void	781	INLINE void
		782	decode_comment (dec_t *dec)
		783	{
		784	// only '#'-style comments allowed a.t.m.
		785
		786	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		787	++dec->cur;
		788	}
		789
		790	INLINE void
510	decode_ws (dec_t *dec)	791	decode_ws (dec_t *dec)
511	{	792	{
512	for (;;)	793	for (;;)
513	{	794	{
514	char ch = *dec->cur;	795	char ch = *dec->cur;
515		796
516	if (ch > 0x20	797	if (ch > 0x20)
		798	{
		799	if (expect_false (ch == '#'))
		800	{
		801	if (dec->json.flags & F_RELAXED)
		802	decode_comment (dec);
		803	else
		804	break;
		805	}
		806	else
		807	break;
		808	}
517	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	809	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
518	break;	810	break; // parse error, but let higher level handle it, gives better error messages
519		811
520	++dec->cur;	812	++dec->cur;
521	}	813	}
522	}	814	}
523		815
524	#define ERR(reason) SB dec->err = reason; goto fail; SE	816	#define ERR(reason) SB dec->err = reason; goto fail; SE
		817
525	#define EXPECT_CH(ch) SB \	818	#define EXPECT_CH(ch) SB \
526	if (*dec->cur != ch) \	819	if (*dec->cur != ch) \
527	ERR (# ch " expected"); \	820	ERR (# ch " expected"); \
528	++dec->cur; \	821	++dec->cur; \
529	SE	822	SE
530		823
		824	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")
		825	#define DEC_DEC_DEPTH --dec->depth
		826
531	static SV *decode_sv (dec_t *dec);	827	static SV *decode_sv (dec_t *dec);
532		828
533	static signed char decode_hexdigit[256];	829	static signed char decode_hexdigit[256];
534		830
535	static UV	831	static UV
536	decode_4hex (dec_t *dec)	832	decode_4hex (dec_t *dec)
537	{	833	{
538	signed char d1, d2, d3, d4;	834	signed char d1, d2, d3, d4;
539	unsigned char *cur = (unsigned char *)dec->cur;	835	unsigned char *cur = (unsigned char *)dec->cur;
540		836
541	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	837	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
542	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	838	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
543	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	839	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
544	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	840	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
545		841
546	dec->cur += 4;	842	dec->cur += 4;
547		843
548	return ((UV)d1) << 12	844	return ((UV)d1) << 12
549	| ((UV)d2) << 8	845	| ((UV)d2) << 8
…		…
557	static SV *	853	static SV *
558	decode_str (dec_t *dec)	854	decode_str (dec_t *dec)
559	{	855	{
560	SV *sv = 0;	856	SV *sv = 0;
561	int utf8 = 0;	857	int utf8 = 0;
		858	char *dec_cur = dec->cur;
562		859
563	do	860	do
564	{	861	{
565	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	862	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
566	char *cur = buf;	863	char *cur = buf;
567		864
568	do	865	do
569	{	866	{
570	unsigned char ch = *(unsigned char *)dec->cur++;	867	unsigned char ch = *(unsigned char *)dec_cur++;
571		868
572	if (ch == '"')	869	if (expect_false (ch == '"'))
573	{	870	{
574	--dec->cur;	871	--dec_cur;
575	break;	872	break;
576	}	873	}
577	else if (ch == '\\')	874	else if (expect_false (ch == '\\'))
578	{	875	{
579	switch (*dec->cur)	876	switch (*dec_cur)
580	{	877	{
581	case '\\':	878	case '\\':
582	case '/':	879	case '/':
583	case '"': *cur++ = *dec->cur++; break;	880	case '"': *cur++ = *dec_cur++; break;
584		881
585	case 'b': ++dec->cur; *cur++ = '\010'; break;	882	case 'b': ++dec_cur; *cur++ = '\010'; break;
586	case 't': ++dec->cur; *cur++ = '\011'; break;	883	case 't': ++dec_cur; *cur++ = '\011'; break;
587	case 'n': ++dec->cur; *cur++ = '\012'; break;	884	case 'n': ++dec_cur; *cur++ = '\012'; break;
588	case 'f': ++dec->cur; *cur++ = '\014'; break;	885	case 'f': ++dec_cur; *cur++ = '\014'; break;
589	case 'r': ++dec->cur; *cur++ = '\015'; break;	886	case 'r': ++dec_cur; *cur++ = '\015'; break;
590		887
591	case 'u':	888	case 'u':
592	{	889	{
593	UV lo, hi;	890	UV lo, hi;
594	++dec->cur;	891	++dec_cur;
595		892
		893	dec->cur = dec_cur;
596	hi = decode_4hex (dec);	894	hi = decode_4hex (dec);
		895	dec_cur = dec->cur;
597	if (hi == (UV)-1)	896	if (hi == (UV)-1)
598	goto fail;	897	goto fail;
599		898
600	// possibly a surrogate pair	899	// possibly a surrogate pair
601	if (hi >= 0xd800)	900	if (hi >= 0xd800)
602	if (hi < 0xdc00)	901	if (hi < 0xdc00)
603	{	902	{
604	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	903	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
605	ERR ("missing low surrogate character in surrogate pair");	904	ERR ("missing low surrogate character in surrogate pair");
606		905
607	dec->cur += 2;	906	dec_cur += 2;
608		907
		908	dec->cur = dec_cur;
609	lo = decode_4hex (dec);	909	lo = decode_4hex (dec);
		910	dec_cur = dec->cur;
610	if (lo == (UV)-1)	911	if (lo == (UV)-1)
611	goto fail;	912	goto fail;
612		913
613	if (lo < 0xdc00 || lo >= 0xe000)	914	if (lo < 0xdc00 || lo >= 0xe000)
614	ERR ("surrogate pair expected");	915	ERR ("surrogate pair expected");
…		…
620		921
621	if (hi >= 0x80)	922	if (hi >= 0x80)
622	{	923	{
623	utf8 = 1;	924	utf8 = 1;
624		925
625	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	926	cur = encode_utf8 (cur, hi);
626	}	927	}
627	else	928	else
628	*cur++ = hi;	929	*cur++ = hi;
629	}	930	}
630	break;	931	break;
631		932
632	default:	933	default:
633	--dec->cur;	934	--dec_cur;
634	ERR ("illegal backslash escape sequence in string");	935	ERR ("illegal backslash escape sequence in string");
635	}	936	}
636	}	937	}
637	else if (ch >= 0x20 && ch <= 0x7f)	938	else if (expect_true (ch >= 0x20 && ch < 0x80))
638	*cur++ = ch;	939	*cur++ = ch;
639	else if (ch >= 0x80)	940	else if (ch >= 0x80)
640	{	941	{
641	--dec->cur;
642
643	STRLEN clen;	942	STRLEN clen;
		943	UV uch;
		944
		945	--dec_cur;
		946
644	UV uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	947	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
645	if (clen == (STRLEN)-1)	948	if (clen == (STRLEN)-1)
646	ERR ("malformed UTF-8 character in JSON string");	949	ERR ("malformed UTF-8 character in JSON string");
647		950
648	do	951	do
649	{
650	*cur++ = *dec->cur++;	952	*cur++ = *dec_cur++;
651	}
652	while (--clen);	953	while (--clen);
653		954
654	utf8 = 1;	955	utf8 = 1;
655	}	956	}
656	else if (!ch)
657	ERR ("unexpected end of string while parsing json string");
658	else	957	else
		958	{
		959	--dec_cur;
		960
		961	if (!ch)
		962	ERR ("unexpected end of string while parsing JSON string");
		963	else
659	ERR ("invalid character encountered");	964	ERR ("invalid character encountered while parsing JSON string");
660		965	}
661	}	966	}
662	while (cur < buf + SHORT_STRING_LEN);	967	while (cur < buf + SHORT_STRING_LEN);
663		968
		969	{
664	STRLEN len = cur - buf;	970	STRLEN len = cur - buf;
665		971
666	if (sv)	972	if (sv)
667	{	973	{
668	SvGROW (sv, SvCUR (sv) + len + 1);	974	SvGROW (sv, SvCUR (sv) + len + 1);
669	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);	975	memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
670	SvCUR_set (sv, SvCUR (sv) + len);	976	SvCUR_set (sv, SvCUR (sv) + len);
671	}	977	}
672	else	978	else
673	sv = newSVpvn (buf, len);	979	sv = newSVpvn (buf, len);
674	}	980	}
		981	}
675	while (*dec->cur != '"');	982	while (*dec_cur != '"');
676		983
677	++dec->cur;	984	++dec_cur;
678		985
679	if (sv)	986	if (sv)
680	{	987	{
681	SvPOK_only (sv);	988	SvPOK_only (sv);
682	*SvEND (sv) = 0;	989	*SvEND (sv) = 0;
…		…
685	SvUTF8_on (sv);	992	SvUTF8_on (sv);
686	}	993	}
687	else	994	else
688	sv = newSVpvn ("", 0);	995	sv = newSVpvn ("", 0);
689		996
		997	dec->cur = dec_cur;
690	return sv;	998	return sv;
691		999
692	fail:	1000	fail:
		1001	dec->cur = dec_cur;
693	return 0;	1002	return 0;
694	}	1003	}
695		1004
696	static SV *	1005	static SV *
697	decode_num (dec_t *dec)	1006	decode_num (dec_t *dec)
…		…
755	is_nv = 1;	1064	is_nv = 1;
756	}	1065	}
757		1066
758	if (!is_nv)	1067	if (!is_nv)
759	{	1068	{
760	UV uv;	1069	int len = dec->cur - start;
761	int numtype = grok_number (start, dec->cur - start, &uv);	1070
762	if (numtype & IS_NUMBER_IN_UV)	1071	// special case the rather common 1..5-digit-int case
763	if (numtype & IS_NUMBER_NEG)	1072	if (*start == '-')
		1073	switch (len)
764	{	1074	{
765	if (uv < (UV)IV_MIN)	1075	case 2: return newSViv (-( start [1] - '0' * 1));
766	return newSViv (-(IV)uv);	1076	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1077	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1078	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1079	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
767	}	1080	}
		1081	else
		1082	switch (len)
		1083	{
		1084	case 1: return newSViv ( start [0] - '0' * 1);
		1085	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1086	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1087	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1088	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1089	}
		1090
		1091	{
		1092	UV uv;
		1093	int numtype = grok_number (start, len, &uv);
		1094	if (numtype & IS_NUMBER_IN_UV)
		1095	if (numtype & IS_NUMBER_NEG)
		1096	{
		1097	if (uv < (UV)IV_MIN)
		1098	return newSViv (-(IV)uv);
		1099	}
768	else	1100	else
769	return newSVuv (uv);	1101	return newSVuv (uv);
770	}	1102	}
771		1103
		1104	len -= *start == '-' ? 1 : 0;
		1105
		1106	// does not fit into IV or UV, try NV
		1107	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1108	#if defined (LDBL_DIG)
		1109	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1110	#endif
		1111	)
		1112	// fits into NV without loss of precision
		1113	return newSVnv (Atof (start));
		1114
		1115	// everything else fails, convert it to a string
		1116	return newSVpvn (start, dec->cur - start);
		1117	}
		1118
		1119	// loss of precision here
772	return newSVnv (Atof (start));	1120	return newSVnv (Atof (start));
773		1121
774	fail:	1122	fail:
775	return 0;	1123	return 0;
776	}	1124	}
…		…
778	static SV *	1126	static SV *
779	decode_av (dec_t *dec)	1127	decode_av (dec_t *dec)
780	{	1128	{
781	AV *av = newAV ();	1129	AV *av = newAV ();
782		1130
		1131	DEC_INC_DEPTH;
783	decode_ws (dec);	1132	decode_ws (dec);
		1133
784	if (*dec->cur == ']')	1134	if (*dec->cur == ']')
785	++dec->cur;	1135	++dec->cur;
786	else	1136	else
787	for (;;)	1137	for (;;)
788	{	1138	{
…		…
804		1154
805	if (*dec->cur != ',')	1155	if (*dec->cur != ',')
806	ERR (", or ] expected while parsing array");	1156	ERR (", or ] expected while parsing array");
807		1157
808	++dec->cur;	1158	++dec->cur;
		1159
		1160	decode_ws (dec);
		1161
		1162	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1163	{
		1164	++dec->cur;
		1165	break;
		1166	}
809	}	1167	}
810		1168
		1169	DEC_DEC_DEPTH;
811	return newRV_noinc ((SV *)av);	1170	return newRV_noinc ((SV *)av);
812		1171
813	fail:	1172	fail:
814	SvREFCNT_dec (av);	1173	SvREFCNT_dec (av);
		1174	DEC_DEC_DEPTH;
815	return 0;	1175	return 0;
816	}	1176	}
817		1177
818	static SV *	1178	static SV *
819	decode_hv (dec_t *dec)	1179	decode_hv (dec_t *dec)
820	{	1180	{
		1181	SV *sv;
821	HV *hv = newHV ();	1182	HV *hv = newHV ();
822		1183
		1184	DEC_INC_DEPTH;
823	decode_ws (dec);	1185	decode_ws (dec);
		1186
824	if (*dec->cur == '}')	1187	if (*dec->cur == '}')
825	++dec->cur;	1188	++dec->cur;
826	else	1189	else
827	for (;;)	1190	for (;;)
828	{	1191	{
		1192	EXPECT_CH ('"');
		1193
		1194	// heuristic: assume that
		1195	// a) decode_str + hv_store_ent are abysmally slow.
		1196	// b) most hash keys are short, simple ascii text.
		1197	// => try to "fast-match" such strings to avoid
		1198	// the overhead of decode_str + hv_store_ent.
		1199	{
829	SV *key, *value;	1200	SV *value;
		1201	char *p = dec->cur;
		1202	char *e = p + 24; // only try up to 24 bytes
830		1203
831	decode_ws (dec); EXPECT_CH ('"');	1204	for (;;)
832
833	key = decode_str (dec);
834	if (!key)
835	goto fail;
836
837	decode_ws (dec); EXPECT_CH (':');
838
839	value = decode_sv (dec);
840	if (!value)
841	{	1205	{
		1206	// the >= 0x80 is false on most architectures
		1207	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1208	{
		1209	// slow path, back up and use decode_str
		1210	SV *key = decode_str (dec);
		1211	if (!key)
		1212	goto fail;
		1213
		1214	decode_ws (dec); EXPECT_CH (':');
		1215
		1216	decode_ws (dec);
		1217	value = decode_sv (dec);
		1218	if (!value)
		1219	{
		1220	SvREFCNT_dec (key);
		1221	goto fail;
		1222	}
		1223
		1224	hv_store_ent (hv, key, value, 0);
842	SvREFCNT_dec (key);	1225	SvREFCNT_dec (key);
		1226
		1227	break;
		1228	}
		1229	else if (*p == '"')
		1230	{
		1231	// fast path, got a simple key
		1232	char *key = dec->cur;
		1233	int len = p - key;
		1234	dec->cur = p + 1;
		1235
		1236	decode_ws (dec); EXPECT_CH (':');
		1237
		1238	decode_ws (dec);
		1239	value = decode_sv (dec);
		1240	if (!value)
843	goto fail;	1241	goto fail;
		1242
		1243	hv_store (hv, key, len, value, 0);
		1244
		1245	break;
		1246	}
		1247
		1248	++p;
844	}	1249	}
845		1250	}
846	//TODO: optimise
847	hv_store_ent (hv, key, value, 0);
848		1251
849	decode_ws (dec);	1252	decode_ws (dec);
850		1253
851	if (*dec->cur == '}')	1254	if (*dec->cur == '}')
852	{	1255	{
…		…
856		1259
857	if (*dec->cur != ',')	1260	if (*dec->cur != ',')
858	ERR (", or } expected while parsing object/hash");	1261	ERR (", or } expected while parsing object/hash");
859		1262
860	++dec->cur;	1263	++dec->cur;
		1264
		1265	decode_ws (dec);
		1266
		1267	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1268	{
		1269	++dec->cur;
		1270	break;
		1271	}
861	}	1272	}
862		1273
		1274	DEC_DEC_DEPTH;
863	return newRV_noinc ((SV *)hv);	1275	sv = newRV_noinc ((SV *)hv);
		1276
		1277	// check filter callbacks
		1278	if (dec->json.flags & F_HOOK)
		1279	{
		1280	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1281	{
		1282	HE *cb, *he;
		1283
		1284	hv_iterinit (hv);
		1285	he = hv_iternext (hv);
		1286	hv_iterinit (hv);
		1287
		1288	// the next line creates a mortal sv each time its called.
		1289	// might want to optimise this for common cases.
		1290	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1291
		1292	if (cb)
		1293	{
		1294	dSP;
		1295	int count;
		1296
		1297	ENTER; SAVETMPS; PUSHMARK (SP);
		1298	XPUSHs (HeVAL (he));
		1299
		1300	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1301
		1302	if (count == 1)
		1303	{
		1304	sv = newSVsv (POPs);
		1305	FREETMPS; LEAVE;
		1306	return sv;
		1307	}
		1308
		1309	FREETMPS; LEAVE;
		1310	}
		1311	}
		1312
		1313	if (dec->json.cb_object)
		1314	{
		1315	dSP;
		1316	int count;
		1317
		1318	ENTER; SAVETMPS; PUSHMARK (SP);
		1319	XPUSHs (sv_2mortal (sv));
		1320
		1321	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1322
		1323	if (count == 1)
		1324	{
		1325	sv = newSVsv (POPs);
		1326	FREETMPS; LEAVE;
		1327	return sv;
		1328	}
		1329
		1330	SvREFCNT_inc (sv);
		1331	FREETMPS; LEAVE;
		1332	}
		1333	}
		1334
		1335	return sv;
864		1336
865	fail:	1337	fail:
866	SvREFCNT_dec (hv);	1338	SvREFCNT_dec (hv);
		1339	DEC_DEC_DEPTH;
867	return 0;	1340	return 0;
868	}	1341	}
869		1342
870	static SV *	1343	static SV *
871	decode_sv (dec_t *dec)	1344	decode_sv (dec_t *dec)
872	{	1345	{
873	decode_ws (dec);	1346	// the beauty of JSON: you need exactly one character lookahead
		1347	// to parse everything.
874	switch (*dec->cur)	1348	switch (*dec->cur)
875	{	1349	{
876	case '"': ++dec->cur; return decode_str (dec);	1350	case '"': ++dec->cur; return decode_str (dec);
877	case '[': ++dec->cur; return decode_av (dec);	1351	case '[': ++dec->cur; return decode_av (dec);
878	case '{': ++dec->cur; return decode_hv (dec);	1352	case '{': ++dec->cur; return decode_hv (dec);
879		1353
880	case '-':	1354	case '-':
881	case '0': case '1': case '2': case '3': case '4':	1355	case '0': case '1': case '2': case '3': case '4':
882	case '5': case '6': case '7': case '8': case '9':	1356	case '5': case '6': case '7': case '8': case '9':
883	return decode_num (dec);	1357	return decode_num (dec);
884		1358
885	case 't':	1359	case 't':
886	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1360	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
887	{	1361	{
888	dec->cur += 4;	1362	dec->cur += 4;
		1363	#if JSON_SLOW
		1364	json_true = get_bool ("JSON::XS::true");
		1365	#endif
889	return newSViv (1);	1366	return newSVsv (json_true);
890	}	1367	}
891	else	1368	else
892	ERR ("'true' expected");	1369	ERR ("'true' expected");
893		1370
894	break;	1371	break;
895		1372
896	case 'f':	1373	case 'f':
897	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1374	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
898	{	1375	{
899	dec->cur += 5;	1376	dec->cur += 5;
		1377	#if JSON_SLOW
		1378	json_false = get_bool ("JSON::XS::false");
		1379	#endif
900	return newSViv (0);	1380	return newSVsv (json_false);
901	}	1381	}
902	else	1382	else
903	ERR ("'false' expected");	1383	ERR ("'false' expected");
904		1384
905	break;	1385	break;
…		…
914	ERR ("'null' expected");	1394	ERR ("'null' expected");
915		1395
916	break;	1396	break;
917		1397
918	default:	1398	default:
919	ERR ("malformed json string, neither array, object, number, string or atom");	1399	ERR ("malformed JSON string, neither array, object, number, string or atom");
920	break;	1400	break;
921	}	1401	}
922		1402
923	fail:	1403	fail:
924	return 0;	1404	return 0;
925	}	1405	}
926		1406
927	static SV *	1407	static SV *
928	decode_json (SV *string, UV flags)	1408	decode_json (SV *string, JSON *json, STRLEN *offset_return)
929	{	1409	{
		1410	dec_t dec;
		1411	STRLEN offset;
930	SV *sv;	1412	SV *sv;
931		1413
		1414	SvGETMAGIC (string);
		1415	SvUPGRADE (string, SVt_PV);
		1416
		1417	if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags))
		1418	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1419	(unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags));
		1420
932	if (flags & F_UTF8)	1421	if (json->flags & F_UTF8)
933	sv_utf8_downgrade (string, 0);	1422	sv_utf8_downgrade (string, 0);
934	else	1423	else
935	sv_utf8_upgrade (string);	1424	sv_utf8_upgrade (string);
936		1425
937	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1426	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
938		1427
939	dec_t dec;	1428	dec.json = *json;
940	dec.flags = flags;
941	dec.cur = SvPVX (string);	1429	dec.cur = SvPVX (string);
942	dec.end = SvEND (string);	1430	dec.end = SvEND (string);
943	dec.err = 0;	1431	dec.err = 0;
		1432	dec.depth = 0;
		1433	dec.maxdepth = DEC_DEPTH (dec.json.flags);
944		1434
		1435	if (dec.json.cb_object || dec.json.cb_sk_object)
		1436	dec.json.flags |= F_HOOK;
		1437
		1438	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1439
		1440	decode_ws (&dec);
945	sv = decode_sv (&dec);	1441	sv = decode_sv (&dec);
946		1442
		1443	if (!(offset_return || !sv))
		1444	{
		1445	// check for trailing garbage
		1446	decode_ws (&dec);
		1447
		1448	if (*dec.cur)
		1449	{
		1450	dec.err = "garbage after JSON object";
		1451	SvREFCNT_dec (sv);
		1452	sv = 0;
		1453	}
		1454	}
		1455
		1456	if (offset_return || !sv)
		1457	{
		1458	offset = dec.json.flags & F_UTF8
		1459	? dec.cur - SvPVX (string)
		1460	: utf8_distance (dec.cur, SvPVX (string));
		1461
		1462	if (offset_return)
		1463	*offset_return = offset;
		1464	}
		1465
947	if (!sv)	1466	if (!sv)
948	{	1467	{
949	IV offset = dec.flags & F_UTF8
950	? dec.cur - SvPVX (string)
951	: utf8_distance (dec.cur, SvPVX (string));
952	SV *uni = sv_newmortal ();	1468	SV *uni = sv_newmortal ();
953		1469
954	// horrible hack to silence warning inside pv_uni_display	1470	// horrible hack to silence warning inside pv_uni_display
955	COP cop = *PL_curcop;	1471	COP cop = *PL_curcop;
956	cop.cop_warnings = pWARN_NONE;	1472	cop.cop_warnings = pWARN_NONE;
…		…
958	SAVEVPTR (PL_curcop);	1474	SAVEVPTR (PL_curcop);
959	PL_curcop = &cop;	1475	PL_curcop = &cop;
960	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);	1476	pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
961	LEAVE;	1477	LEAVE;
962		1478
963	croak ("%s, at character offset %d (%s)",	1479	croak ("%s, at character offset %d [\"%s\"]",
964	dec.err,	1480	dec.err,
965	(int)offset,	1481	(int)offset,
966	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1482	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
967	}	1483	}
968		1484
969	sv = sv_2mortal (sv);	1485	sv = sv_2mortal (sv);
970		1486
971	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1487	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
972	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1488	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
973		1489
974	return sv;	1490	return sv;
		1491	}
		1492
		1493	/////////////////////////////////////////////////////////////////////////////
		1494	// incremental parser
		1495
		1496	static void
		1497	incr_parse (JSON *self)
		1498	{
		1499	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1500
		1501	for (;;)
		1502	{
		1503	//printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
		1504	switch (self->incr_mode)
		1505	{
		1506	// only used for intiial whitespace skipping
		1507	case INCR_M_WS:
		1508	for (;;)
		1509	{
		1510	if (*p > 0x20)
		1511	{
		1512	self->incr_mode = INCR_M_JSON;
		1513	goto incr_m_json;
		1514	}
		1515	else if (!*p)
		1516	goto interrupt;
		1517
		1518	++p;
		1519	}
		1520
		1521	// skip a single char inside a string (for \\-processing)
		1522	case INCR_M_BS:
		1523	if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	self->incr_mode = INCR_M_STR;
		1528	goto incr_m_str;
		1529
		1530	// inside a string
		1531	case INCR_M_STR:
		1532	incr_m_str:
		1533	for (;;)
		1534	{
		1535	if (*p == '"')
		1536	{
		1537	++p;
		1538	self->incr_mode = INCR_M_JSON;
		1539
		1540	if (!self->incr_nest)
		1541	goto interrupt;
		1542
		1543	goto incr_m_json;
		1544	}
		1545	else if (*p == '\\')
		1546	{
		1547	++p; // "virtually" consumes character after \
		1548
		1549	if (!*p) // if at end of string we have to switch modes
		1550	{
		1551	self->incr_mode = INCR_M_BS;
		1552	goto interrupt;
		1553	}
		1554	}
		1555	else if (!*p)
		1556	goto interrupt;
		1557
		1558	++p;
		1559	}
		1560
		1561	// after initial ws, outside string
		1562	case INCR_M_JSON:
		1563	incr_m_json:
		1564	for (;;)
		1565	{
		1566	switch (*p++)
		1567	{
		1568	case 0:
		1569	--p;
		1570	goto interrupt;
		1571
		1572	case 0x09:
		1573	case 0x0a:
		1574	case 0x0d:
		1575	case 0x20:
		1576	if (!self->incr_nest)
		1577	{
		1578	--p; // do not eat the whitespace, let the next round do it
		1579	goto interrupt;
		1580	}
		1581	break;
		1582
		1583	case '"':
		1584	self->incr_mode = INCR_M_STR;
		1585	goto incr_m_str;
		1586
		1587	case '[':
		1588	case '{':
		1589	++self->incr_nest;
		1590	break;
		1591
		1592	case ']':
		1593	case '}':
		1594	if (!--self->incr_nest)
		1595	goto interrupt;
		1596	}
		1597	}
		1598	}
		1599
		1600	modechange:
		1601	;
		1602	}
		1603
		1604	interrupt:
		1605	self->incr_pos = p - SvPVX (self->incr_text);
		1606	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
975	}	1607	}
976		1608
977	/////////////////////////////////////////////////////////////////////////////	1609	/////////////////////////////////////////////////////////////////////////////
978	// XS interface functions	1610	// XS interface functions
979		1611
…		…
981		1613
982	BOOT:	1614	BOOT:
983	{	1615	{
984	int i;	1616	int i;
985		1617
986	memset (decode_hexdigit, 0xff, 256);
987	for (i = 10; i--; )	1618	for (i = 0; i < 256; ++i)
988	decode_hexdigit ['0' + i] = i;	1619	decode_hexdigit [i] =
		1620	i >= '0' && i <= '9' ? i - '0'
		1621	: i >= 'a' && i <= 'f' ? i - 'a' + 10
		1622	: i >= 'A' && i <= 'F' ? i - 'A' + 10
		1623	: -1;
989		1624
990	for (i = 7; i--; )
991	{
992	decode_hexdigit ['a' + i] = 10 + i;
993	decode_hexdigit ['A' + i] = 10 + i;
994	}
995
996	json_stash = gv_stashpv ("JSON::XS", 1);	1625	json_stash = gv_stashpv ("JSON::XS" , 1);
		1626	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1627
		1628	json_true = get_bool ("JSON::XS::true");
		1629	json_false = get_bool ("JSON::XS::false");
997	}	1630	}
998		1631
999	PROTOTYPES: DISABLE	1632	PROTOTYPES: DISABLE
1000		1633
1001	SV *new (char *dummy)	1634	void CLONE (...)
1002	CODE:	1635	CODE:
1003	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1636	json_stash = 0;
		1637	json_boolean_stash = 0;
		1638
		1639	void new (char *klass)
		1640	PPCODE:
		1641	{
		1642	SV *pv = NEWSV (0, sizeof (JSON));
		1643	SvPOK_only (pv);
		1644	Zero (SvPVX (pv), 1, JSON);
		1645	((JSON *)SvPVX (pv))->flags = F_DEFAULT;
		1646	XPUSHs (sv_2mortal (sv_bless (
		1647	newRV_noinc (pv),
		1648	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1649	)));
		1650	}
		1651
		1652	void ascii (JSON *self, int enable = 1)
		1653	ALIAS:
		1654	ascii = F_ASCII
		1655	latin1 = F_LATIN1
		1656	utf8 = F_UTF8
		1657	indent = F_INDENT
		1658	canonical = F_CANONICAL
		1659	space_before = F_SPACE_BEFORE
		1660	space_after = F_SPACE_AFTER
		1661	pretty = F_PRETTY
		1662	allow_nonref = F_ALLOW_NONREF
		1663	shrink = F_SHRINK
		1664	allow_blessed = F_ALLOW_BLESSED
		1665	convert_blessed = F_CONV_BLESSED
		1666	relaxed = F_RELAXED
		1667	allow_unknown = F_ALLOW_UNKNOWN
		1668	PPCODE:
		1669	{
		1670	if (enable)
		1671	self->flags |= ix;
		1672	else
		1673	self->flags &= ~ix;
		1674
		1675	XPUSHs (ST (0));
		1676	}
		1677
		1678	void get_ascii (JSON *self)
		1679	ALIAS:
		1680	get_ascii = F_ASCII
		1681	get_latin1 = F_LATIN1
		1682	get_utf8 = F_UTF8
		1683	get_indent = F_INDENT
		1684	get_canonical = F_CANONICAL
		1685	get_space_before = F_SPACE_BEFORE
		1686	get_space_after = F_SPACE_AFTER
		1687	get_allow_nonref = F_ALLOW_NONREF
		1688	get_shrink = F_SHRINK
		1689	get_allow_blessed = F_ALLOW_BLESSED
		1690	get_convert_blessed = F_CONV_BLESSED
		1691	get_relaxed = F_RELAXED
		1692	get_allow_unknown = F_ALLOW_UNKNOWN
		1693	PPCODE:
		1694	XPUSHs (boolSV (self->flags & ix));
		1695
		1696	void max_depth (JSON *self, UV max_depth = 0x80000000UL)
		1697	PPCODE:
		1698	{
		1699	UV log2 = 0;
		1700
		1701	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
		1702
		1703	while ((1UL << log2) < max_depth)
		1704	++log2;
		1705
		1706	self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
		1707
		1708	XPUSHs (ST (0));
		1709	}
		1710
		1711	U32 get_max_depth (JSON *self)
		1712	CODE:
		1713	RETVAL = DEC_DEPTH (self->flags);
1004	OUTPUT:	1714	OUTPUT:
1005	RETVAL	1715	RETVAL
1006		1716
1007	SV *ascii (SV *self, int enable = 1)	1717	void max_size (JSON *self, UV max_size = 0)
1008	ALIAS:	1718	PPCODE:
1009	ascii = F_ASCII	1719	{
1010	utf8 = F_UTF8	1720	UV log2 = 0;
1011	indent = F_INDENT	1721
1012	canonical = F_CANONICAL	1722	if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1013	space_before = F_SPACE_BEFORE	1723	if (max_size == 1) max_size = 2;
1014	space_after = F_SPACE_AFTER	1724
1015	pretty = F_PRETTY	1725	while ((1UL << log2) < max_size)
1016	allow_nonref = F_ALLOW_NONREF	1726	++log2;
1017	shrink = F_SHRINK	1727
		1728	self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
		1729
		1730	XPUSHs (ST (0));
		1731	}
		1732
		1733	int get_max_size (JSON *self)
1018	CODE:	1734	CODE:
1019	{	1735	RETVAL = DEC_SIZE (self->flags);
1020	UV *uv = SvJSON (self);
1021	if (enable)
1022	*uv |= ix;
1023	else
1024	*uv &= ~ix;
1025
1026	RETVAL = newSVsv (self);
1027	}
1028	OUTPUT:	1736	OUTPUT:
1029	RETVAL	1737	RETVAL
1030		1738
1031	void encode (SV *self, SV *scalar)	1739	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1032	PPCODE:	1740	PPCODE:
1033	XPUSHs (encode_json (scalar, *SvJSON (self)));	1741	{
		1742	SvREFCNT_dec (self->cb_object);
		1743	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
1034		1744
1035	void decode (SV *self, SV *jsonstr)	1745	XPUSHs (ST (0));
		1746	}
		1747
		1748	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1036	PPCODE:	1749	PPCODE:
		1750	{
		1751	if (!self->cb_sk_object)
		1752	self->cb_sk_object = newHV ();
		1753
		1754	if (SvOK (cb))
		1755	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1756	else
		1757	{
		1758	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1759
		1760	if (!HvKEYS (self->cb_sk_object))
		1761	{
		1762	SvREFCNT_dec (self->cb_sk_object);
		1763	self->cb_sk_object = 0;
		1764	}
		1765	}
		1766
		1767	XPUSHs (ST (0));
		1768	}
		1769
		1770	void encode (JSON *self, SV *scalar)
		1771	PPCODE:
		1772	XPUSHs (encode_json (scalar, self));
		1773
		1774	void decode (JSON *self, SV *jsonstr)
		1775	PPCODE:
1037	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1776	XPUSHs (decode_json (jsonstr, self, 0));
		1777
		1778	void decode_prefix (JSON *self, SV *jsonstr)
		1779	PPCODE:
		1780	{
		1781	STRLEN offset;
		1782	EXTEND (SP, 2);
		1783	PUSHs (decode_json (jsonstr, self, &offset));
		1784	PUSHs (sv_2mortal (newSVuv (offset)));
		1785	}
		1786
		1787	void incr_parse (JSON *self, SV *jsonstr = 0)
		1788	PPCODE:
		1789	{
		1790	if (!self->incr_text)
		1791	self->incr_text = newSVpvn ("", 0);
		1792
		1793	// append data, if any
		1794	if (jsonstr)
		1795	{
		1796	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1797	{
		1798	/* utf-8-ness differs, need to upgrade */
		1799	sv_utf8_upgrade (self->incr_text);
		1800
		1801	if (self->incr_pos)
		1802	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1803	- (U8 *)SvPVX (self->incr_text);
		1804	}
		1805
		1806	{
		1807	STRLEN len;
		1808	const char *str = SvPV (jsonstr, len);
		1809	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1810	Move (str, SvEND (self->incr_text), len, char);
		1811	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1812	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1813	}
		1814	}
		1815
		1816	if (GIMME_V != G_VOID)
		1817	do
		1818	{
		1819	STRLEN offset;
		1820
		1821	if (!INCR_DONE (self))
		1822	{
		1823	incr_parse (self);
		1824	if (!INCR_DONE (self))
		1825	break;
		1826	}
		1827
		1828	XPUSHs (decode_json (self->incr_text, self, &offset));
		1829
		1830	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1831	self->incr_pos -= offset;
		1832	self->incr_nest = 0;
		1833	self->incr_mode = 0;
		1834	}
		1835	while (GIMME_V == G_ARRAY);
		1836	}
		1837
		1838	SV *incr_text (JSON *self)
		1839	ATTRS: lvalue
		1840	CODE:
		1841	{
		1842	if (self->incr_pos)
		1843	croak ("incr_text can not be called when the incremental parser already started parsing");
		1844
		1845	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
		1846	}
		1847	OUTPUT:
		1848	RETVAL
		1849
		1850	void incr_skip (JSON *self)
		1851	CODE:
		1852	{
		1853	if (self->incr_pos)
		1854	{
		1855	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1856	self->incr_pos = 0;
		1857	self->incr_nest = 0;
		1858	self->incr_mode = 0;
		1859	}
		1860	}
		1861
		1862	void DESTROY (JSON *self)
		1863	CODE:
		1864	SvREFCNT_dec (self->cb_sk_object);
		1865	SvREFCNT_dec (self->cb_object);
		1866	SvREFCNT_dec (self->incr_text);
1038		1867
1039	PROTOTYPES: ENABLE	1868	PROTOTYPES: ENABLE
1040		1869
1041	void to_json (SV *scalar)	1870	void encode_json (SV *scalar)
1042	ALIAS:	1871	ALIAS:
1043	objToJson = 0	1872	to_json_ = 0
		1873	encode_json = F_UTF8
1044	PPCODE:	1874	PPCODE:
		1875	{
		1876	JSON json = { F_DEFAULT | ix };
1045	XPUSHs (encode_json (scalar, F_UTF8));	1877	XPUSHs (encode_json (scalar, &json));
		1878	}
1046		1879
1047	void from_json (SV *jsonstr)	1880	void decode_json (SV *jsonstr)
1048	ALIAS:	1881	ALIAS:
1049	jsonToObj = 0	1882	from_json_ = 0
		1883	decode_json = F_UTF8
1050	PPCODE:	1884	PPCODE:
		1885	{
		1886	JSON json = { F_DEFAULT | ix };
1051	XPUSHs (decode_json (jsonstr, F_UTF8));	1887	XPUSHs (decode_json (jsonstr, &json, 0));
		1888	}
1052		1889
		1890

Diff Legend

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