ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/JSON-XS/XS.xs

Comparing JSON-XS/XS.xs (file contents):
Revision 1.27 by root, Mon Apr 9 05:09:57 2007 UTC vs.
Revision 1.87 by root, Tue Jun 3 06:43:45 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"	8	#include <stdio.h>
		9	#include <limits.h>
		10	#include <float.h>
9		11
10	#if defined(__BORLANDC__) || defined(_MSC_VER)	12	#if defined(__BORLANDC__) || defined(_MSC_VER)
11	# define snprintf _snprintf // C compilers have this in stdio.h	13	# define snprintf _snprintf // C compilers have this in stdio.h
12	#endif	14	#endif
13		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
14	#define F_ASCII 0x00000001UL	24	#define F_ASCII 0x00000001UL
		25	#define F_LATIN1 0x00000002UL
15	#define F_UTF8 0x00000002UL	26	#define F_UTF8 0x00000004UL
16	#define F_INDENT 0x00000004UL	27	#define F_INDENT 0x00000008UL
17	#define F_CANONICAL 0x00000008UL	28	#define F_CANONICAL 0x00000010UL
18	#define F_SPACE_BEFORE 0x00000010UL	29	#define F_SPACE_BEFORE 0x00000020UL
19	#define F_SPACE_AFTER 0x00000020UL	30	#define F_SPACE_AFTER 0x00000040UL
20	#define F_ALLOW_NONREF 0x00000080UL	31	#define F_ALLOW_NONREF 0x00000100UL
21	#define F_SHRINK 0x00000100UL	32	#define F_SHRINK 0x00000200UL
		33	#define F_ALLOW_BLESSED 0x00000400UL
		34	#define F_CONV_BLESSED 0x00000800UL
22	#define F_MAXDEPTH 0xf8000000UL	35	#define F_RELAXED 0x00001000UL
23	#define S_MAXDEPTH 27	36	#define F_ALLOW_UNKNOWN 0x00002000UL
24		37	#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
25	#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
26
27	// F_SELFCONVERT? <=> to_json/toJson
28	// F_BLESSED? <=> { $__class__$ => }
29		38
30	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER	39	#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
31	#define F_DEFAULT (9UL << S_MAXDEPTH)
32		40
33	#define INIT_SIZE 32 // initial scalar size to be allocated	41	#define INIT_SIZE 32 // initial scalar size to be allocated
34	#define INDENT_STEP 3 // spaces per indentation level	42	#define INDENT_STEP 3 // spaces per indentation level
35		43
36	#define UTF8_MAX_LEN 11 // for perls UTF-X: max. number of octets per character
37	#define SHORT_STRING_LEN 512 // special-case strings of up to this size	44	#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
38		45
39	#define SB do {	46	#define SB do {
40	#define SE } while (0)	47	#define SE } while (0)
41		48
		49	#if __GNUC__ >= 3
		50	# define expect(expr,value) __builtin_expect ((expr), (value))
		51	# define INLINE static inline
		52	#else
		53	# define expect(expr,value) (expr)
		54	# define INLINE static
		55	#endif
		56
		57	#define expect_false(expr) expect ((expr) != 0, 0)
		58	#define expect_true(expr) expect ((expr) != 0, 1)
		59
		60	#define IN_RANGE_INC(type,val,beg,end) \
		61	((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
		62	<= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
		63
		64	#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
		65
		66	#ifdef USE_ITHREADS
		67	# define JSON_SLOW 1
		68	# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1))
		69	#else
		70	# define JSON_SLOW 0
		71	# define JSON_STASH json_stash
		72	#endif
		73
42	static HV *json_stash; // JSON::XS::	74	static HV *json_stash, *json_boolean_stash; // JSON::XS::
		75	static SV *json_true, *json_false;
		76
		77	enum {
		78	INCR_M_WS = 0, // initial whitespace skipping, must be 0
		79	INCR_M_STR, // inside string
		80	INCR_M_BS, // inside backslash
		81	INCR_M_JSON // outside anything, count nesting
		82	};
		83
		84	#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON)
		85
		86	typedef struct {
		87	U32 flags;
		88	U32 max_depth;
		89	STRLEN max_size;
		90
		91	SV *cb_object;
		92	HV *cb_sk_object;
		93
		94	// for the incremental parser
		95	SV *incr_text; // the source text so far
		96	STRLEN incr_pos; // the current offset into the text
		97	unsigned char incr_nest; // {[]}-nesting level
		98	unsigned char incr_mode;
		99	} JSON;
		100
		101	INLINE void
		102	json_init (JSON *json)
		103	{
		104	Zero (json, 1, JSON);
		105	json->max_depth = 512;
		106	}
43		107
44	/////////////////////////////////////////////////////////////////////////////	108	/////////////////////////////////////////////////////////////////////////////
45	// utility functions	109	// utility functions
46		110
47	static UV *	111	INLINE SV *
48	SvJSON (SV *sv)	112	get_bool (const char *name)
49	{	113	{
50	if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash))	114	SV *sv = get_sv (name, 1);
51	croak ("object is not of type JSON::XS");
52		115
53	return &SvUVX (SvRV (sv));	116	SvREADONLY_on (sv);
54	}	117	SvREADONLY_on (SvRV (sv));
55		118
56	static void	119	return sv;
		120	}
		121
		122	INLINE void
57	shrink (SV *sv)	123	shrink (SV *sv)
58	{	124	{
59	sv_utf8_downgrade (sv, 1);	125	sv_utf8_downgrade (sv, 1);
		126
60	if (SvLEN (sv) > SvCUR (sv) + 1)	127	if (SvLEN (sv) > SvCUR (sv) + 1)
61	{	128	{
62	#ifdef SvPV_shrink_to_cur	129	#ifdef SvPV_shrink_to_cur
63	SvPV_shrink_to_cur (sv);	130	SvPV_shrink_to_cur (sv);
64	#elif defined (SvPV_renew)	131	#elif defined (SvPV_renew)
…		…
70	// decode an utf-8 character and return it, or (UV)-1 in	137	// decode an utf-8 character and return it, or (UV)-1 in
71	// case of an error.	138	// case of an error.
72	// we special-case "safe" characters from U+80 .. U+7FF,	139	// we special-case "safe" characters from U+80 .. U+7FF,
73	// but use the very good perl function to parse anything else.	140	// but use the very good perl function to parse anything else.
74	// note that we never call this function for a ascii codepoints	141	// note that we never call this function for a ascii codepoints
75	static UV	142	INLINE UV
76	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)	143	decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen)
77	{	144	{
78	if (s[0] > 0xdf || s[0] < 0xc2)	145	if (expect_true (len >= 2
79	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);	146	&& IN_RANGE_INC (char, s[0], 0xc2, 0xdf)
80	else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf)	147	&& IN_RANGE_INC (char, s[1], 0x80, 0xbf)))
81	{	148	{
82	*clen = 2;	149	*clen = 2;
83	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);	150	return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f);
84	}	151	}
85	else	152	else
86	{	153	return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY);
87	*clen = (STRLEN)-1;	154	}
88	return (UV)-1;	155
89	}	156	// likewise for encoding, also never called for ascii codepoints
		157	// this function takes advantage of this fact, although current gccs
		158	// seem to optimise the check for >= 0x80 away anyways
		159	INLINE unsigned char *
		160	encode_utf8 (unsigned char *s, UV ch)
		161	{
		162	if (expect_false (ch < 0x000080))
		163	*s++ = ch;
		164	else if (expect_true (ch < 0x000800))
		165	*s++ = 0xc0 | ( ch >> 6),
		166	*s++ = 0x80 | ( ch & 0x3f);
		167	else if ( ch < 0x010000)
		168	*s++ = 0xe0 | ( ch >> 12),
		169	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		170	*s++ = 0x80 | ( ch & 0x3f);
		171	else if ( ch < 0x110000)
		172	*s++ = 0xf0 | ( ch >> 18),
		173	*s++ = 0x80 | ((ch >> 12) & 0x3f),
		174	*s++ = 0x80 | ((ch >> 6) & 0x3f),
		175	*s++ = 0x80 | ( ch & 0x3f);
		176
		177	return s;
90	}	178	}
91		179
92	/////////////////////////////////////////////////////////////////////////////	180	/////////////////////////////////////////////////////////////////////////////
93	// encoder	181	// encoder
94		182
…		…
96	typedef struct	184	typedef struct
97	{	185	{
98	char *cur; // SvPVX (sv) + current output position	186	char *cur; // SvPVX (sv) + current output position
99	char *end; // SvEND (sv)	187	char *end; // SvEND (sv)
100	SV *sv; // result scalar	188	SV *sv; // result scalar
101	U32 flags; // F_*	189	JSON json;
102	U32 indent; // indentation level	190	U32 indent; // indentation level
103	U32 maxdepth; // max. indentation/recursion level	191	UV limit; // escape character values >= this value when encoding
104	} enc_t;	192	} enc_t;
105		193
106	static void	194	INLINE void
107	need (enc_t *enc, STRLEN len)	195	need (enc_t *enc, STRLEN len)
108	{	196	{
109	if (enc->cur + len >= enc->end)	197	if (expect_false (enc->cur + len >= enc->end))
110	{	198	{
111	STRLEN cur = enc->cur - SvPVX (enc->sv);	199	STRLEN cur = enc->cur - SvPVX (enc->sv);
112	SvGROW (enc->sv, cur + len + 1);	200	SvGROW (enc->sv, cur + len + 1);
113	enc->cur = SvPVX (enc->sv) + cur;	201	enc->cur = SvPVX (enc->sv) + cur;
114	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;	202	enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1;
115	}	203	}
116	}	204	}
117		205
118	static void	206	INLINE void
119	encode_ch (enc_t *enc, char ch)	207	encode_ch (enc_t *enc, char ch)
120	{	208	{
121	need (enc, 1);	209	need (enc, 1);
122	*enc->cur++ = ch;	210	*enc->cur++ = ch;
123	}	211	}
…		…
131		219
132	while (str < end)	220	while (str < end)
133	{	221	{
134	unsigned char ch = *(unsigned char *)str;	222	unsigned char ch = *(unsigned char *)str;
135		223
136	if (ch >= 0x20 && ch < 0x80) // most common case	224	if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
137	{	225	{
138	if (ch == '"') // but with slow exceptions	226	if (expect_false (ch == '"')) // but with slow exceptions
139	{	227	{
140	need (enc, len += 1);	228	need (enc, len += 1);
141	*enc->cur++ = '\\';	229	*enc->cur++ = '\\';
142	*enc->cur++ = '"';	230	*enc->cur++ = '"';
143	}	231	}
144	else if (ch == '\\')	232	else if (expect_false (ch == '\\'))
145	{	233	{
146	need (enc, len += 1);	234	need (enc, len += 1);
147	*enc->cur++ = '\\';	235	*enc->cur++ = '\\';
148	*enc->cur++ = '\\';	236	*enc->cur++ = '\\';
149	}	237	}
…		…
167	STRLEN clen;	255	STRLEN clen;
168	UV uch;	256	UV uch;
169		257
170	if (is_utf8)	258	if (is_utf8)
171	{	259	{
172	//uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY);
173	uch = decode_utf8 (str, end - str, &clen);	260	uch = decode_utf8 (str, end - str, &clen);
174	if (clen == (STRLEN)-1)	261	if (clen == (STRLEN)-1)
175	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);	262	croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str);
176	}	263	}
177	else	264	else
178	{	265	{
179	uch = ch;	266	uch = ch;
180	clen = 1;	267	clen = 1;
181	}	268	}
182		269
183	if (uch > 0x10FFFFUL)	270	if (uch < 0x80/*0x20*/ || uch >= enc->limit)
184	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
185
186	if (uch < 0x80 || enc->flags & F_ASCII)
187	{	271	{
188	if (uch > 0xFFFFUL)	272	if (uch >= 0x10000UL)
189	{	273	{
		274	if (uch >= 0x110000UL)
		275	croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
		276
190	need (enc, len += 11);	277	need (enc, len += 11);
191	sprintf (enc->cur, "\\u%04x\\u%04x",	278	sprintf (enc->cur, "\\u%04x\\u%04x",
192	(int)((uch - 0x10000) / 0x400 + 0xD800),	279	(int)((uch - 0x10000) / 0x400 + 0xD800),
193	(int)((uch - 0x10000) % 0x400 + 0xDC00));	280	(int)((uch - 0x10000) % 0x400 + 0xDC00));
194	enc->cur += 12;	281	enc->cur += 12;
…		…
205	*enc->cur++ = hexdigit [(uch >> 0) & 15];	292	*enc->cur++ = hexdigit [(uch >> 0) & 15];
206	}	293	}
207		294
208	str += clen;	295	str += clen;
209	}	296	}
		297	else if (enc->json.flags & F_LATIN1)
		298	{
		299	*enc->cur++ = uch;
		300	str += clen;
		301	}
210	else if (is_utf8)	302	else if (is_utf8)
211	{	303	{
212	need (enc, len += clen);	304	need (enc, len += clen);
213	do	305	do
214	{	306	{
…		…
216	}	308	}
217	while (--clen);	309	while (--clen);
218	}	310	}
219	else	311	else
220	{	312	{
221	need (enc, len += UTF8_MAX_LEN - 1); // never more than 11 bytes needed	313	need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed
222	enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0);	314	enc->cur = encode_utf8 (enc->cur, uch);
223	++str;	315	++str;
224	}	316	}
225	}	317	}
226	}	318	}
227	}	319	}
228		320
229	--len;	321	--len;
230	}	322	}
231	}	323	}
232		324
233	static void	325	INLINE void
234	encode_indent (enc_t *enc)	326	encode_indent (enc_t *enc)
235	{	327	{
236	if (enc->flags & F_INDENT)	328	if (enc->json.flags & F_INDENT)
237	{	329	{
238	int spaces = enc->indent * INDENT_STEP;	330	int spaces = enc->indent * INDENT_STEP;
239		331
240	need (enc, spaces);	332	need (enc, spaces);
241	memset (enc->cur, ' ', spaces);	333	memset (enc->cur, ' ', spaces);
242	enc->cur += spaces;	334	enc->cur += spaces;
243	}	335	}
244	}	336	}
245		337
246	static void	338	INLINE void
247	encode_space (enc_t *enc)	339	encode_space (enc_t *enc)
248	{	340	{
249	need (enc, 1);	341	need (enc, 1);
250	encode_ch (enc, ' ');	342	encode_ch (enc, ' ');
251	}	343	}
252		344
253	static void	345	INLINE void
254	encode_nl (enc_t *enc)	346	encode_nl (enc_t *enc)
255	{	347	{
256	if (enc->flags & F_INDENT)	348	if (enc->json.flags & F_INDENT)
257	{	349	{
258	need (enc, 1);	350	need (enc, 1);
259	encode_ch (enc, '\n');	351	encode_ch (enc, '\n');
260	}	352	}
261	}	353	}
262		354
263	static void	355	INLINE void
264	encode_comma (enc_t *enc)	356	encode_comma (enc_t *enc)
265	{	357	{
266	encode_ch (enc, ',');	358	encode_ch (enc, ',');
267		359
268	if (enc->flags & F_INDENT)	360	if (enc->json.flags & F_INDENT)
269	encode_nl (enc);	361	encode_nl (enc);
270	else if (enc->flags & F_SPACE_AFTER)	362	else if (enc->json.flags & F_SPACE_AFTER)
271	encode_space (enc);	363	encode_space (enc);
272	}	364	}
273		365
274	static void encode_sv (enc_t *enc, SV *sv);	366	static void encode_sv (enc_t *enc, SV *sv);
275		367
276	static void	368	static void
277	encode_av (enc_t *enc, AV *av)	369	encode_av (enc_t *enc, AV *av)
278	{	370	{
279	int i, len = av_len (av);	371	int i, len = av_len (av);
280		372
281	if (enc->indent >= enc->maxdepth)	373	if (enc->indent >= enc->json.max_depth)
282	croak ("data structure too deep (hit recursion limit)");	374	croak (ERR_NESTING_EXCEEDED);
283		375
284	encode_ch (enc, '['); encode_nl (enc);	376	encode_ch (enc, '[');
285	++enc->indent;	377
		378	if (len >= 0)
		379	{
		380	encode_nl (enc); ++enc->indent;
286		381
287	for (i = 0; i <= len; ++i)	382	for (i = 0; i <= len; ++i)
288	{	383	{
		384	SV **svp = av_fetch (av, i, 0);
		385
289	encode_indent (enc);	386	encode_indent (enc);
290	encode_sv (enc, *av_fetch (av, i, 0));
291		387
		388	if (svp)
		389	encode_sv (enc, *svp);
		390	else
		391	encode_str (enc, "null", 4, 0);
		392
292	if (i < len)	393	if (i < len)
293	encode_comma (enc);	394	encode_comma (enc);
294	}	395	}
295		396
		397	encode_nl (enc); --enc->indent; encode_indent (enc);
		398	}
		399
296	encode_nl (enc);	400	encode_ch (enc, ']');
297
298	--enc->indent;
299	encode_indent (enc); encode_ch (enc, ']');
300	}	401	}
301		402
302	static void	403	static void
303	encode_he (enc_t *enc, HE *he)	404	encode_hk (enc_t *enc, HE *he)
304	{	405	{
305	encode_ch (enc, '"');	406	encode_ch (enc, '"');
306		407
307	if (HeKLEN (he) == HEf_SVKEY)	408	if (HeKLEN (he) == HEf_SVKEY)
308	{	409	{
…		…
318	else	419	else
319	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));	420	encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he));
320		421
321	encode_ch (enc, '"');	422	encode_ch (enc, '"');
322		423
323	if (enc->flags & F_SPACE_BEFORE) encode_space (enc);	424	if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc);
324	encode_ch (enc, ':');	425	encode_ch (enc, ':');
325	if (enc->flags & F_SPACE_AFTER ) encode_space (enc);	426	if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc);
326	encode_sv (enc, HeVAL (he));
327	}	427	}
328		428
329	// compare hash entries, used when all keys are bytestrings	429	// compare hash entries, used when all keys are bytestrings
330	static int	430	static int
331	he_cmp_fast (const void *a_, const void *b_)	431	he_cmp_fast (const void *a_, const void *b_)
…		…
336	HE *b = *(HE **)b_;	436	HE *b = *(HE **)b_;
337		437
338	STRLEN la = HeKLEN (a);	438	STRLEN la = HeKLEN (a);
339	STRLEN lb = HeKLEN (b);	439	STRLEN lb = HeKLEN (b);
340		440
341	if (!(cmp = memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb)))	441	if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la)))
342	cmp = la - lb;	442	cmp = lb - la;
343		443
344	return cmp;	444	return cmp;
345	}	445	}
346		446
347	// compare hash entries, used when some keys are sv's or utf-x	447	// compare hash entries, used when some keys are sv's or utf-x
348	static int	448	static int
349	he_cmp_slow (const void *a, const void *b)	449	he_cmp_slow (const void *a, const void *b)
350	{	450	{
351	return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b));	451	return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a));
352	}	452	}
353		453
354	static void	454	static void
355	encode_hv (enc_t *enc, HV *hv)	455	encode_hv (enc_t *enc, HV *hv)
356	{	456	{
357	int count, i;	457	HE *he;
358		458
359	if (enc->indent >= enc->maxdepth)	459	if (enc->indent >= enc->json.max_depth)
360	croak ("data structure too deep (hit recursion limit)");	460	croak (ERR_NESTING_EXCEEDED);
361		461
362	encode_ch (enc, '{'); encode_nl (enc); ++enc->indent;	462	encode_ch (enc, '{');
363		463
364	if ((count = hv_iterinit (hv)))
365	{
366	// for canonical output we have to sort by keys first	464	// for canonical output we have to sort by keys first
367	// actually, this is mostly due to the stupid so-called	465	// actually, this is mostly due to the stupid so-called
368	// security workaround added somewhere in 5.8.x.	466	// security workaround added somewhere in 5.8.x.
369	// that randomises hash orderings	467	// that randomises hash orderings
370	if (enc->flags & F_CANONICAL)	468	if (enc->json.flags & F_CANONICAL)
		469	{
		470	int count = hv_iterinit (hv);
		471
		472	if (SvMAGICAL (hv))
371	{	473	{
		474	// need to count by iterating. could improve by dynamically building the vector below
		475	// but I don't care for the speed of this special case.
		476	// note also that we will run into undefined behaviour when the two iterations
		477	// do not result in the same count, something I might care for in some later release.
		478
		479	count = 0;
		480	while (hv_iternext (hv))
		481	++count;
		482
		483	hv_iterinit (hv);
		484	}
		485
		486	if (count)
		487	{
		488	int i, fast = 1;
		489	#if defined(__BORLANDC__) || defined(_MSC_VER)
		490	HE **hes = _alloca (count * sizeof (HE));
		491	#else
372	HE *he, *hes [count]; // if your compiler dies here, you need to enable C99 mode	492	HE *hes [count]; // if your compiler dies here, you need to enable C99 mode
373	int fast = 1;	493	#endif
374		494
375	i = 0;	495	i = 0;
376	while ((he = hv_iternext (hv)))	496	while ((he = hv_iternext (hv)))
377	{	497	{
378	hes [i++] = he;	498	hes [i++] = he;
…		…
400		520
401	FREETMPS;	521	FREETMPS;
402	LEAVE;	522	LEAVE;
403	}	523	}
404		524
405	for (i = 0; i < count; ++i)	525	encode_nl (enc); ++enc->indent;
		526
		527	while (count--)
406	{	528	{
407	encode_indent (enc);	529	encode_indent (enc);
		530	he = hes [count];
408	encode_he (enc, hes [i]);	531	encode_hk (enc, he);
		532	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
409		533
410	if (i < count - 1)	534	if (count)
411	encode_comma (enc);	535	encode_comma (enc);
412	}	536	}
413		537
		538	encode_nl (enc); --enc->indent; encode_indent (enc);
		539	}
		540	}
		541	else
		542	{
		543	if (hv_iterinit (hv) || SvMAGICAL (hv))
		544	if ((he = hv_iternext (hv)))
		545	{
		546	encode_nl (enc); ++enc->indent;
		547
		548	for (;;)
		549	{
414	encode_nl (enc);	550	encode_indent (enc);
		551	encode_hk (enc, he);
		552	encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he));
		553
		554	if (!(he = hv_iternext (hv)))
		555	break;
		556
		557	encode_comma (enc);
		558	}
		559
		560	encode_nl (enc); --enc->indent; encode_indent (enc);
		561	}
		562	}
		563
		564	encode_ch (enc, '}');
		565	}
		566
		567	// encode objects, arrays and special \0=false and \1=true values.
		568	static void
		569	encode_rv (enc_t *enc, SV *sv)
		570	{
		571	svtype svt;
		572
		573	SvGETMAGIC (sv);
		574	svt = SvTYPE (sv);
		575
		576	if (expect_false (SvOBJECT (sv)))
		577	{
		578	HV *stash = !JSON_SLOW || json_boolean_stash
		579	? json_boolean_stash
		580	: gv_stashpv ("JSON::XS::Boolean", 1);
		581
		582	if (SvSTASH (sv) == stash)
		583	{
		584	if (SvIV (sv))
		585	encode_str (enc, "true", 4, 0);
		586	else
		587	encode_str (enc, "false", 5, 0);
415	}	588	}
416	else	589	else
417	{	590	{
418	HE *he = hv_iternext (hv);	591	#if 0
419		592	if (0 && sv_derived_from (rv, "JSON::Literal"))
420	for (;;)
421	{	593	{
422	encode_indent (enc);	594	// not yet
423	encode_he (enc, he);
424
425	if (!(he = hv_iternext (hv)))
426	break;
427
428	encode_comma (enc);
429	}	595	}
		596	#endif
		597	if (enc->json.flags & F_CONV_BLESSED)
		598	{
		599	// we re-bless the reference to get overload and other niceties right
		600	GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
430		601
		602	if (to_json)
		603	{
		604	dSP;
		605
		606	ENTER; SAVETMPS; PUSHMARK (SP);
		607	XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
		608
		609	// calling with G_SCALAR ensures that we always get a 1 return value
		610	PUTBACK;
		611	call_sv ((SV *)GvCV (to_json), G_SCALAR);
		612	SPAGAIN;
		613
		614	// catch this surprisingly common error
		615	if (SvROK (TOPs) && SvRV (TOPs) == sv)
		616	croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
		617
		618	sv = POPs;
		619	PUTBACK;
		620
431	encode_nl (enc);	621	encode_sv (enc, sv);
		622
		623	FREETMPS; LEAVE;
		624	}
		625	else if (enc->json.flags & F_ALLOW_BLESSED)
		626	encode_str (enc, "null", 4, 0);
		627	else
		628	croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
		629	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
		630	}
		631	else if (enc->json.flags & F_ALLOW_BLESSED)
		632	encode_str (enc, "null", 4, 0);
		633	else
		634	croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled",
		635	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
432	}	636	}
433	}	637	}
434
435	--enc->indent; encode_indent (enc); encode_ch (enc, '}');
436	}
437
438	// encode objects, arrays and special \0=false and \1=true values.
439	static void
440	encode_rv (enc_t *enc, SV *sv)
441	{
442	svtype svt;
443
444	SvGETMAGIC (sv);
445	svt = SvTYPE (sv);
446
447	if (svt == SVt_PVHV)	638	else if (svt == SVt_PVHV)
448	encode_hv (enc, (HV *)sv);	639	encode_hv (enc, (HV *)sv);
449	else if (svt == SVt_PVAV)	640	else if (svt == SVt_PVAV)
450	encode_av (enc, (AV *)sv);	641	encode_av (enc, (AV *)sv);
451	else if (svt < SVt_PVAV)	642	else if (svt < SVt_PVAV)
452	{	643	{
453	if (SvNIOK (sv) && SvIV (sv) == 0)	644	STRLEN len = 0;
		645	char *pv = svt ? SvPV (sv, len) : 0;
		646
		647	if (len == 1 && *pv == '1')
		648	encode_str (enc, "true", 4, 0);
		649	else if (len == 1 && *pv == '0')
454	encode_str (enc, "false", 5, 0);	650	encode_str (enc, "false", 5, 0);
455	else if (SvNIOK (sv) && SvIV (sv) == 1)	651	else if (enc->json.flags & F_ALLOW_UNKNOWN)
456	encode_str (enc, "true", 4, 0);	652	encode_str (enc, "null", 4, 0);
457	else	653	else
458	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",	654	croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
459	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	655	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
460	}	656	}
		657	else if (enc->json.flags & F_ALLOW_UNKNOWN)
		658	encode_str (enc, "null", 4, 0);
461	else	659	else
462	croak ("encountered %s, but JSON can only represent references to arrays or hashes",	660	croak ("encountered %s, but JSON can only represent references to arrays or hashes",
463	SvPV_nolen (sv_2mortal (newRV_inc (sv))));	661	SvPV_nolen (sv_2mortal (newRV_inc (sv))));
464	}	662	}
465		663
…		…
476	encode_str (enc, str, len, SvUTF8 (sv));	674	encode_str (enc, str, len, SvUTF8 (sv));
477	encode_ch (enc, '"');	675	encode_ch (enc, '"');
478	}	676	}
479	else if (SvNOKp (sv))	677	else if (SvNOKp (sv))
480	{	678	{
		679	// trust that perl will do the right thing w.r.t. JSON syntax.
481	need (enc, NV_DIG + 32);	680	need (enc, NV_DIG + 32);
482	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);	681	Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur);
483	enc->cur += strlen (enc->cur);	682	enc->cur += strlen (enc->cur);
484	}	683	}
485	else if (SvIOKp (sv))	684	else if (SvIOKp (sv))
486	{	685	{
		686	// we assume we can always read an IV as a UV and vice versa
		687	// we assume two's complement
		688	// we assume no aliasing issues in the union
		689	if (SvIsUV (sv) ? SvUVX (sv) <= 59000
		690	: SvIVX (sv) <= 59000 && SvIVX (sv) >= -59000)
		691	{
		692	// optimise the "small number case"
		693	// code will likely be branchless and use only a single multiplication
		694	// works for numbers up to 59074
		695	I32 i = SvIVX (sv);
		696	U32 u;
		697	char digit, nz = 0;
		698
487	need (enc, 64);	699	need (enc, 6);
		700
		701	*enc->cur = '-'; enc->cur += i < 0 ? 1 : 0;
		702	u = i < 0 ? -i : i;
		703
		704	// convert to 4.28 fixed-point representation
		705	u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits
		706
		707	// now output digit by digit, each time masking out the integer part
		708	// and multiplying by 5 while moving the decimal point one to the right,
		709	// resulting in a net multiplication by 10.
		710	// we always write the digit to memory but conditionally increment
		711	// the pointer, to enable the use of conditional move instructions.
		712	digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffffUL) * 5;
		713	digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffffUL) * 5;
		714	digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffffUL) * 5;
		715	digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffffUL) * 5;
		716	digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0'
		717	}
		718	else
		719	{
		720	// large integer, use the (rather slow) snprintf way.
		721	need (enc, IVUV_MAXCHARS);
488	enc->cur +=	722	enc->cur +=
489	SvIsUV(sv)	723	SvIsUV(sv)
490	? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv))	724	? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
491	: snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv));	725	: snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
		726	}
492	}	727	}
493	else if (SvROK (sv))	728	else if (SvROK (sv))
494	encode_rv (enc, SvRV (sv));	729	encode_rv (enc, SvRV (sv));
495	else if (!SvOK (sv))	730	else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
496	encode_str (enc, "null", 4, 0);	731	encode_str (enc, "null", 4, 0);
497	else	732	else
498	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",	733	croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this",
499	SvPV_nolen (sv), SvFLAGS (sv));	734	SvPV_nolen (sv), SvFLAGS (sv));
500	}	735	}
501		736
502	static SV *	737	static SV *
503	encode_json (SV *scalar, U32 flags)	738	encode_json (SV *scalar, JSON *json)
504	{	739	{
505	enc_t enc;	740	enc_t enc;
506		741
507	if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar))	742	if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar))
508	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");	743	croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
509		744
510	enc.flags = flags;	745	enc.json = *json;
511	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));	746	enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
512	enc.cur = SvPVX (enc.sv);	747	enc.cur = SvPVX (enc.sv);
513	enc.end = SvEND (enc.sv);	748	enc.end = SvEND (enc.sv);
514	enc.indent = 0;	749	enc.indent = 0;
515	enc.maxdepth = DEC_DEPTH (flags);	750	enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
		751	: enc.json.flags & F_LATIN1 ? 0x000100UL
		752	: 0x110000UL;
516		753
517	SvPOK_only (enc.sv);	754	SvPOK_only (enc.sv);
518	encode_sv (&enc, scalar);	755	encode_sv (&enc, scalar);
519		756
520	if (!(flags & (F_ASCII | F_UTF8)))
521	SvUTF8_on (enc.sv);
522
523	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));	757	SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
524	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings	758	*SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
525		759
		760	if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
		761	SvUTF8_on (enc.sv);
		762
526	if (enc.flags & F_SHRINK)	763	if (enc.json.flags & F_SHRINK)
527	shrink (enc.sv);	764	shrink (enc.sv);
528		765
529	return enc.sv;	766	return enc.sv;
530	}	767	}
531		768
…		…
536	typedef struct	773	typedef struct
537	{	774	{
538	char *cur; // current parser pointer	775	char *cur; // current parser pointer
539	char *end; // end of input string	776	char *end; // end of input string
540	const char *err; // parse error, if != 0	777	const char *err; // parse error, if != 0
541	U32 flags; // F_*	778	JSON json;
542	U32 depth; // recursion depth	779	U32 depth; // recursion depth
543	U32 maxdepth; // recursion depth limit	780	U32 maxdepth; // recursion depth limit
544	} dec_t;	781	} dec_t;
545		782
546	static void	783	INLINE void
		784	decode_comment (dec_t *dec)
		785	{
		786	// only '#'-style comments allowed a.t.m.
		787
		788	while (*dec->cur && *dec->cur != 0x0a && *dec->cur != 0x0d)
		789	++dec->cur;
		790	}
		791
		792	INLINE void
547	decode_ws (dec_t *dec)	793	decode_ws (dec_t *dec)
548	{	794	{
549	for (;;)	795	for (;;)
550	{	796	{
551	char ch = *dec->cur;	797	char ch = *dec->cur;
552		798
553	if (ch > 0x20	799	if (ch > 0x20)
		800	{
		801	if (expect_false (ch == '#'))
		802	{
		803	if (dec->json.flags & F_RELAXED)
		804	decode_comment (dec);
		805	else
		806	break;
		807	}
		808	else
		809	break;
		810	}
554	|| (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09))	811	else if (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)
555	break;	812	break; // parse error, but let higher level handle it, gives better error messages
556		813
557	++dec->cur;	814	++dec->cur;
558	}	815	}
559	}	816	}
560		817
…		…
564	if (*dec->cur != ch) \	821	if (*dec->cur != ch) \
565	ERR (# ch " expected"); \	822	ERR (# ch " expected"); \
566	++dec->cur; \	823	++dec->cur; \
567	SE	824	SE
568		825
569	#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)")	826	#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
570	#define DEC_DEC_DEPTH --dec->depth	827	#define DEC_DEC_DEPTH --dec->depth
571		828
572	static SV *decode_sv (dec_t *dec);	829	static SV *decode_sv (dec_t *dec);
573		830
574	static signed char decode_hexdigit[256];	831	static signed char decode_hexdigit[256];
…		…
577	decode_4hex (dec_t *dec)	834	decode_4hex (dec_t *dec)
578	{	835	{
579	signed char d1, d2, d3, d4;	836	signed char d1, d2, d3, d4;
580	unsigned char *cur = (unsigned char *)dec->cur;	837	unsigned char *cur = (unsigned char *)dec->cur;
581		838
582	d1 = decode_hexdigit [cur [0]]; if (d1 < 0) ERR ("four hexadecimal digits expected");	839	d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected");
583	d2 = decode_hexdigit [cur [1]]; if (d2 < 0) ERR ("four hexadecimal digits expected");	840	d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected");
584	d3 = decode_hexdigit [cur [2]]; if (d3 < 0) ERR ("four hexadecimal digits expected");	841	d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected");
585	d4 = decode_hexdigit [cur [3]]; if (d4 < 0) ERR ("four hexadecimal digits expected");	842	d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected");
586		843
587	dec->cur += 4;	844	dec->cur += 4;
588		845
589	return ((UV)d1) << 12	846	return ((UV)d1) << 12
590	| ((UV)d2) << 8	847	| ((UV)d2) << 8
…		…
598	static SV *	855	static SV *
599	decode_str (dec_t *dec)	856	decode_str (dec_t *dec)
600	{	857	{
601	SV *sv = 0;	858	SV *sv = 0;
602	int utf8 = 0;	859	int utf8 = 0;
		860	char *dec_cur = dec->cur;
603		861
604	do	862	do
605	{	863	{
606	char buf [SHORT_STRING_LEN + UTF8_MAX_LEN];	864	char buf [SHORT_STRING_LEN + UTF8_MAXBYTES];
607	char *cur = buf;	865	char *cur = buf;
608		866
609	do	867	do
610	{	868	{
611	unsigned char ch = *(unsigned char *)dec->cur++;	869	unsigned char ch = *(unsigned char *)dec_cur++;
612		870
613	if (ch == '"')	871	if (expect_false (ch == '"'))
614	{	872	{
615	--dec->cur;	873	--dec_cur;
616	break;	874	break;
617	}	875	}
618	else if (ch == '\\')	876	else if (expect_false (ch == '\\'))
619	{	877	{
620	switch (*dec->cur)	878	switch (*dec_cur)
621	{	879	{
622	case '\\':	880	case '\\':
623	case '/':	881	case '/':
624	case '"': *cur++ = *dec->cur++; break;	882	case '"': *cur++ = *dec_cur++; break;
625		883
626	case 'b': ++dec->cur; *cur++ = '\010'; break;	884	case 'b': ++dec_cur; *cur++ = '\010'; break;
627	case 't': ++dec->cur; *cur++ = '\011'; break;	885	case 't': ++dec_cur; *cur++ = '\011'; break;
628	case 'n': ++dec->cur; *cur++ = '\012'; break;	886	case 'n': ++dec_cur; *cur++ = '\012'; break;
629	case 'f': ++dec->cur; *cur++ = '\014'; break;	887	case 'f': ++dec_cur; *cur++ = '\014'; break;
630	case 'r': ++dec->cur; *cur++ = '\015'; break;	888	case 'r': ++dec_cur; *cur++ = '\015'; break;
631		889
632	case 'u':	890	case 'u':
633	{	891	{
634	UV lo, hi;	892	UV lo, hi;
635	++dec->cur;	893	++dec_cur;
636		894
		895	dec->cur = dec_cur;
637	hi = decode_4hex (dec);	896	hi = decode_4hex (dec);
		897	dec_cur = dec->cur;
638	if (hi == (UV)-1)	898	if (hi == (UV)-1)
639	goto fail;	899	goto fail;
640		900
641	// possibly a surrogate pair	901	// possibly a surrogate pair
642	if (hi >= 0xd800)	902	if (hi >= 0xd800)
643	if (hi < 0xdc00)	903	if (hi < 0xdc00)
644	{	904	{
645	if (dec->cur [0] != '\\' || dec->cur [1] != 'u')	905	if (dec_cur [0] != '\\' || dec_cur [1] != 'u')
646	ERR ("missing low surrogate character in surrogate pair");	906	ERR ("missing low surrogate character in surrogate pair");
647		907
648	dec->cur += 2;	908	dec_cur += 2;
649		909
		910	dec->cur = dec_cur;
650	lo = decode_4hex (dec);	911	lo = decode_4hex (dec);
		912	dec_cur = dec->cur;
651	if (lo == (UV)-1)	913	if (lo == (UV)-1)
652	goto fail;	914	goto fail;
653		915
654	if (lo < 0xdc00 || lo >= 0xe000)	916	if (lo < 0xdc00 || lo >= 0xe000)
655	ERR ("surrogate pair expected");	917	ERR ("surrogate pair expected");
…		…
661		923
662	if (hi >= 0x80)	924	if (hi >= 0x80)
663	{	925	{
664	utf8 = 1;	926	utf8 = 1;
665		927
666	cur = (char *)uvuni_to_utf8_flags (cur, hi, 0);	928	cur = encode_utf8 (cur, hi);
667	}	929	}
668	else	930	else
669	*cur++ = hi;	931	*cur++ = hi;
670	}	932	}
671	break;	933	break;
672		934
673	default:	935	default:
674	--dec->cur;	936	--dec_cur;
675	ERR ("illegal backslash escape sequence in string");	937	ERR ("illegal backslash escape sequence in string");
676	}	938	}
677	}	939	}
678	else if (ch >= 0x20 && ch <= 0x7f)	940	else if (expect_true (ch >= 0x20 && ch < 0x80))
679	*cur++ = ch;	941	*cur++ = ch;
680	else if (ch >= 0x80)	942	else if (ch >= 0x80)
681	{	943	{
682	STRLEN clen;	944	STRLEN clen;
683	UV uch;	945	UV uch;
684		946
685	--dec->cur;	947	--dec_cur;
686		948
687	uch = decode_utf8 (dec->cur, dec->end - dec->cur, &clen);	949	uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
688	if (clen == (STRLEN)-1)	950	if (clen == (STRLEN)-1)
689	ERR ("malformed UTF-8 character in JSON string");	951	ERR ("malformed UTF-8 character in JSON string");
690		952
691	do	953	do
692	*cur++ = *dec->cur++;	954	*cur++ = *dec_cur++;
693	while (--clen);	955	while (--clen);
694		956
695	utf8 = 1;	957	utf8 = 1;
696	}	958	}
697	else	959	else
698	{	960	{
699	--dec->cur;	961	--dec_cur;
700		962
701	if (!ch)	963	if (!ch)
702	ERR ("unexpected end of string while parsing JSON string");	964	ERR ("unexpected end of string while parsing JSON string");
703	else	965	else
704	ERR ("invalid character encountered while parsing JSON string");	966	ERR ("invalid character encountered while parsing JSON string");
…		…
717	}	979	}
718	else	980	else
719	sv = newSVpvn (buf, len);	981	sv = newSVpvn (buf, len);
720	}	982	}
721	}	983	}
722	while (*dec->cur != '"');	984	while (*dec_cur != '"');
723		985
724	++dec->cur;	986	++dec_cur;
725		987
726	if (sv)	988	if (sv)
727	{	989	{
728	SvPOK_only (sv);	990	SvPOK_only (sv);
729	*SvEND (sv) = 0;	991	*SvEND (sv) = 0;
…		…
732	SvUTF8_on (sv);	994	SvUTF8_on (sv);
733	}	995	}
734	else	996	else
735	sv = newSVpvn ("", 0);	997	sv = newSVpvn ("", 0);
736		998
		999	dec->cur = dec_cur;
737	return sv;	1000	return sv;
738		1001
739	fail:	1002	fail:
		1003	dec->cur = dec_cur;
740	return 0;	1004	return 0;
741	}	1005	}
742		1006
743	static SV *	1007	static SV *
744	decode_num (dec_t *dec)	1008	decode_num (dec_t *dec)
…		…
802	is_nv = 1;	1066	is_nv = 1;
803	}	1067	}
804		1068
805	if (!is_nv)	1069	if (!is_nv)
806	{	1070	{
807	UV uv;	1071	int len = dec->cur - start;
808	int numtype = grok_number (start, dec->cur - start, &uv);	1072
809	if (numtype & IS_NUMBER_IN_UV)	1073	// special case the rather common 1..5-digit-int case
810	if (numtype & IS_NUMBER_NEG)	1074	if (*start == '-')
		1075	switch (len)
811	{	1076	{
812	if (uv < (UV)IV_MIN)	1077	case 2: return newSViv (-( start [1] - '0' * 1));
813	return newSViv (-(IV)uv);	1078	case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11));
		1079	case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
		1080	case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
		1081	case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
814	}	1082	}
		1083	else
		1084	switch (len)
		1085	{
		1086	case 1: return newSViv ( start [0] - '0' * 1);
		1087	case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
		1088	case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
		1089	case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
		1090	case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
		1091	}
		1092
		1093	{
		1094	UV uv;
		1095	int numtype = grok_number (start, len, &uv);
		1096	if (numtype & IS_NUMBER_IN_UV)
		1097	if (numtype & IS_NUMBER_NEG)
		1098	{
		1099	if (uv < (UV)IV_MIN)
		1100	return newSViv (-(IV)uv);
		1101	}
815	else	1102	else
816	return newSVuv (uv);	1103	return newSVuv (uv);
817	}	1104	}
818		1105
		1106	len -= *start == '-' ? 1 : 0;
		1107
		1108	// does not fit into IV or UV, try NV
		1109	if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len)
		1110	#if defined (LDBL_DIG)
		1111	|| (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
		1112	#endif
		1113	)
		1114	// fits into NV without loss of precision
		1115	return newSVnv (Atof (start));
		1116
		1117	// everything else fails, convert it to a string
		1118	return newSVpvn (start, dec->cur - start);
		1119	}
		1120
		1121	// loss of precision here
819	return newSVnv (Atof (start));	1122	return newSVnv (Atof (start));
820		1123
821	fail:	1124	fail:
822	return 0;	1125	return 0;
823	}	1126	}
…		…
853		1156
854	if (*dec->cur != ',')	1157	if (*dec->cur != ',')
855	ERR (", or ] expected while parsing array");	1158	ERR (", or ] expected while parsing array");
856		1159
857	++dec->cur;	1160	++dec->cur;
		1161
		1162	decode_ws (dec);
		1163
		1164	if (*dec->cur == ']' && dec->json.flags & F_RELAXED)
		1165	{
		1166	++dec->cur;
		1167	break;
		1168	}
858	}	1169	}
859		1170
860	DEC_DEC_DEPTH;	1171	DEC_DEC_DEPTH;
861	return newRV_noinc ((SV *)av);	1172	return newRV_noinc ((SV *)av);
862		1173
…		…
867	}	1178	}
868		1179
869	static SV *	1180	static SV *
870	decode_hv (dec_t *dec)	1181	decode_hv (dec_t *dec)
871	{	1182	{
		1183	SV *sv;
872	HV *hv = newHV ();	1184	HV *hv = newHV ();
873		1185
874	DEC_INC_DEPTH;	1186	DEC_INC_DEPTH;
875	decode_ws (dec);	1187	decode_ws (dec);
876		1188
877	if (*dec->cur == '}')	1189	if (*dec->cur == '}')
878	++dec->cur;	1190	++dec->cur;
879	else	1191	else
880	for (;;)	1192	for (;;)
881	{	1193	{
		1194	EXPECT_CH ('"');
		1195
		1196	// heuristic: assume that
		1197	// a) decode_str + hv_store_ent are abysmally slow.
		1198	// b) most hash keys are short, simple ascii text.
		1199	// => try to "fast-match" such strings to avoid
		1200	// the overhead of decode_str + hv_store_ent.
		1201	{
882	SV *key, *value;	1202	SV *value;
		1203	char *p = dec->cur;
		1204	char *e = p + 24; // only try up to 24 bytes
883		1205
884	decode_ws (dec); EXPECT_CH ('"');	1206	for (;;)
885
886	key = decode_str (dec);
887	if (!key)
888	goto fail;
889
890	decode_ws (dec); EXPECT_CH (':');
891
892	value = decode_sv (dec);
893	if (!value)
894	{	1207	{
		1208	// the >= 0x80 is false on most architectures
		1209	if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\')
		1210	{
		1211	// slow path, back up and use decode_str
		1212	SV *key = decode_str (dec);
		1213	if (!key)
		1214	goto fail;
		1215
		1216	decode_ws (dec); EXPECT_CH (':');
		1217
		1218	decode_ws (dec);
		1219	value = decode_sv (dec);
		1220	if (!value)
		1221	{
		1222	SvREFCNT_dec (key);
		1223	goto fail;
		1224	}
		1225
		1226	hv_store_ent (hv, key, value, 0);
895	SvREFCNT_dec (key);	1227	SvREFCNT_dec (key);
		1228
		1229	break;
		1230	}
		1231	else if (*p == '"')
		1232	{
		1233	// fast path, got a simple key
		1234	char *key = dec->cur;
		1235	int len = p - key;
		1236	dec->cur = p + 1;
		1237
		1238	decode_ws (dec); EXPECT_CH (':');
		1239
		1240	decode_ws (dec);
		1241	value = decode_sv (dec);
		1242	if (!value)
896	goto fail;	1243	goto fail;
		1244
		1245	hv_store (hv, key, len, value, 0);
		1246
		1247	break;
		1248	}
		1249
		1250	++p;
897	}	1251	}
898		1252	}
899	hv_store_ent (hv, key, value, 0);
900	SvREFCNT_dec (key);
901		1253
902	decode_ws (dec);	1254	decode_ws (dec);
903		1255
904	if (*dec->cur == '}')	1256	if (*dec->cur == '}')
905	{	1257	{
…		…
909		1261
910	if (*dec->cur != ',')	1262	if (*dec->cur != ',')
911	ERR (", or } expected while parsing object/hash");	1263	ERR (", or } expected while parsing object/hash");
912		1264
913	++dec->cur;	1265	++dec->cur;
		1266
		1267	decode_ws (dec);
		1268
		1269	if (*dec->cur == '}' && dec->json.flags & F_RELAXED)
		1270	{
		1271	++dec->cur;
		1272	break;
		1273	}
914	}	1274	}
915		1275
916	DEC_DEC_DEPTH;	1276	DEC_DEC_DEPTH;
917	return newRV_noinc ((SV *)hv);	1277	sv = newRV_noinc ((SV *)hv);
		1278
		1279	// check filter callbacks
		1280	if (dec->json.flags & F_HOOK)
		1281	{
		1282	if (dec->json.cb_sk_object && HvKEYS (hv) == 1)
		1283	{
		1284	HE *cb, *he;
		1285
		1286	hv_iterinit (hv);
		1287	he = hv_iternext (hv);
		1288	hv_iterinit (hv);
		1289
		1290	// the next line creates a mortal sv each time its called.
		1291	// might want to optimise this for common cases.
		1292	cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
		1293
		1294	if (cb)
		1295	{
		1296	dSP;
		1297	int count;
		1298
		1299	ENTER; SAVETMPS; PUSHMARK (SP);
		1300	XPUSHs (HeVAL (he));
		1301
		1302	PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
		1303
		1304	if (count == 1)
		1305	{
		1306	sv = newSVsv (POPs);
		1307	FREETMPS; LEAVE;
		1308	return sv;
		1309	}
		1310
		1311	FREETMPS; LEAVE;
		1312	}
		1313	}
		1314
		1315	if (dec->json.cb_object)
		1316	{
		1317	dSP;
		1318	int count;
		1319
		1320	ENTER; SAVETMPS; PUSHMARK (SP);
		1321	XPUSHs (sv_2mortal (sv));
		1322
		1323	PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
		1324
		1325	if (count == 1)
		1326	{
		1327	sv = newSVsv (POPs);
		1328	FREETMPS; LEAVE;
		1329	return sv;
		1330	}
		1331
		1332	SvREFCNT_inc (sv);
		1333	FREETMPS; LEAVE;
		1334	}
		1335	}
		1336
		1337	return sv;
918		1338
919	fail:	1339	fail:
920	SvREFCNT_dec (hv);	1340	SvREFCNT_dec (hv);
921	DEC_DEC_DEPTH;	1341	DEC_DEC_DEPTH;
922	return 0;	1342	return 0;
923	}	1343	}
924		1344
925	static SV *	1345	static SV *
926	decode_sv (dec_t *dec)	1346	decode_sv (dec_t *dec)
927	{	1347	{
928	decode_ws (dec);	1348	// the beauty of JSON: you need exactly one character lookahead
		1349	// to parse everything.
929	switch (*dec->cur)	1350	switch (*dec->cur)
930	{	1351	{
931	case '"': ++dec->cur; return decode_str (dec);	1352	case '"': ++dec->cur; return decode_str (dec);
932	case '[': ++dec->cur; return decode_av (dec);	1353	case '[': ++dec->cur; return decode_av (dec);
933	case '{': ++dec->cur; return decode_hv (dec);	1354	case '{': ++dec->cur; return decode_hv (dec);
934		1355
935	case '-':	1356	case '-':
936	case '0': case '1': case '2': case '3': case '4':	1357	case '0': case '1': case '2': case '3': case '4':
937	case '5': case '6': case '7': case '8': case '9':	1358	case '5': case '6': case '7': case '8': case '9':
938	return decode_num (dec);	1359	return decode_num (dec);
939		1360
940	case 't':	1361	case 't':
941	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))	1362	if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
942	{	1363	{
943	dec->cur += 4;	1364	dec->cur += 4;
		1365	#if JSON_SLOW
		1366	json_true = get_bool ("JSON::XS::true");
		1367	#endif
944	return newSViv (1);	1368	return newSVsv (json_true);
945	}	1369	}
946	else	1370	else
947	ERR ("'true' expected");	1371	ERR ("'true' expected");
948		1372
949	break;	1373	break;
950		1374
951	case 'f':	1375	case 'f':
952	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))	1376	if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
953	{	1377	{
954	dec->cur += 5;	1378	dec->cur += 5;
		1379	#if JSON_SLOW
		1380	json_false = get_bool ("JSON::XS::false");
		1381	#endif
955	return newSViv (0);	1382	return newSVsv (json_false);
956	}	1383	}
957	else	1384	else
958	ERR ("'false' expected");	1385	ERR ("'false' expected");
959		1386
960	break;	1387	break;
…		…
978	fail:	1405	fail:
979	return 0;	1406	return 0;
980	}	1407	}
981		1408
982	static SV *	1409	static SV *
983	decode_json (SV *string, U32 flags)	1410	decode_json (SV *string, JSON *json, STRLEN *offset_return)
984	{	1411	{
985	dec_t dec;	1412	dec_t dec;
		1413	STRLEN offset;
986	SV *sv;	1414	SV *sv;
987		1415
		1416	SvGETMAGIC (string);
988	SvUPGRADE (string, SVt_PV);	1417	SvUPGRADE (string, SVt_PV);
989		1418
		1419	/* work around a bug in perl 5.10, which causes SvCUR to fail an
		1420	* assertion with -DDEBUGGING, although SvCUR is documented to
		1421	* return the xpv_cur field which certainly exists after upgrading.
		1422	* according to nicholas clark, calling SvPOK fixes this.
		1423	*/
		1424	SvPOK (string);
		1425
		1426	if (SvCUR (string) > json->max_size && json->max_size)
		1427	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1428	(unsigned long)SvCUR (string), (unsigned long)json->max_size);
		1429
990	if (flags & F_UTF8)	1430	if (json->flags & F_UTF8)
991	sv_utf8_downgrade (string, 0);	1431	sv_utf8_downgrade (string, 0);
992	else	1432	else
993	sv_utf8_upgrade (string);	1433	sv_utf8_upgrade (string);
994		1434
995	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP	1435	SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
996		1436
997	dec.flags = flags;	1437	dec.json = *json;
998	dec.cur = SvPVX (string);	1438	dec.cur = SvPVX (string);
999	dec.end = SvEND (string);	1439	dec.end = SvEND (string);
1000	dec.err = 0;	1440	dec.err = 0;
1001	dec.depth = 0;	1441	dec.depth = 0;
1002	dec.maxdepth = DEC_DEPTH (dec.flags);
1003		1442
		1443	if (dec.json.cb_object || dec.json.cb_sk_object)
		1444	dec.json.flags |= F_HOOK;
		1445
1004	*dec.end = 0; // this should basically be a nop, too, but make sure its there	1446	*dec.end = 0; // this should basically be a nop, too, but make sure it's there
		1447
		1448	decode_ws (&dec);
1005	sv = decode_sv (&dec);	1449	sv = decode_sv (&dec);
1006		1450
		1451	if (!(offset_return || !sv))
		1452	{
		1453	// check for trailing garbage
		1454	decode_ws (&dec);
		1455
		1456	if (*dec.cur)
		1457	{
		1458	dec.err = "garbage after JSON object";
		1459	SvREFCNT_dec (sv);
		1460	sv = 0;
		1461	}
		1462	}
		1463
		1464	if (offset_return || !sv)
		1465	{
		1466	offset = dec.json.flags & F_UTF8
		1467	? dec.cur - SvPVX (string)
		1468	: utf8_distance (dec.cur, SvPVX (string));
		1469
		1470	if (offset_return)
		1471	*offset_return = offset;
		1472	}
		1473
1007	if (!sv)	1474	if (!sv)
1008	{	1475	{
1009	IV offset = dec.flags & F_UTF8
1010	? dec.cur - SvPVX (string)
1011	: utf8_distance (dec.cur, SvPVX (string));
1012	SV *uni = sv_newmortal ();	1476	SV *uni = sv_newmortal ();
1013		1477
1014	// horrible hack to silence warning inside pv_uni_display	1478	// horrible hack to silence warning inside pv_uni_display
1015	COP cop = *PL_curcop;	1479	COP cop = *PL_curcop;
1016	cop.cop_warnings = pWARN_NONE;	1480	cop.cop_warnings = pWARN_NONE;
…		…
1026	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");	1490	dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1027	}	1491	}
1028		1492
1029	sv = sv_2mortal (sv);	1493	sv = sv_2mortal (sv);
1030		1494
1031	if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv))	1495	if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv))
1032	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");	1496	croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1033		1497
1034	return sv;	1498	return sv;
		1499	}
		1500
		1501	/////////////////////////////////////////////////////////////////////////////
		1502	// incremental parser
		1503
		1504	static void
		1505	incr_parse (JSON *self)
		1506	{
		1507	const char *p = SvPVX (self->incr_text) + self->incr_pos;
		1508
		1509	for (;;)
		1510	{
		1511	//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
		1512	switch (self->incr_mode)
		1513	{
		1514	// only used for intiial whitespace skipping
		1515	case INCR_M_WS:
		1516	for (;;)
		1517	{
		1518	if (*p > 0x20)
		1519	{
		1520	self->incr_mode = INCR_M_JSON;
		1521	goto incr_m_json;
		1522	}
		1523	else if (!*p)
		1524	goto interrupt;
		1525
		1526	++p;
		1527	}
		1528
		1529	// skip a single char inside a string (for \\-processing)
		1530	case INCR_M_BS:
		1531	if (!*p)
		1532	goto interrupt;
		1533
		1534	++p;
		1535	self->incr_mode = INCR_M_STR;
		1536	goto incr_m_str;
		1537
		1538	// inside a string
		1539	case INCR_M_STR:
		1540	incr_m_str:
		1541	for (;;)
		1542	{
		1543	if (*p == '"')
		1544	{
		1545	++p;
		1546	self->incr_mode = INCR_M_JSON;
		1547
		1548	if (!self->incr_nest)
		1549	goto interrupt;
		1550
		1551	goto incr_m_json;
		1552	}
		1553	else if (*p == '\\')
		1554	{
		1555	++p; // "virtually" consumes character after \
		1556
		1557	if (!*p) // if at end of string we have to switch modes
		1558	{
		1559	self->incr_mode = INCR_M_BS;
		1560	goto interrupt;
		1561	}
		1562	}
		1563	else if (!*p)
		1564	goto interrupt;
		1565
		1566	++p;
		1567	}
		1568
		1569	// after initial ws, outside string
		1570	case INCR_M_JSON:
		1571	incr_m_json:
		1572	for (;;)
		1573	{
		1574	switch (*p++)
		1575	{
		1576	case 0:
		1577	--p;
		1578	goto interrupt;
		1579
		1580	case 0x09:
		1581	case 0x0a:
		1582	case 0x0d:
		1583	case 0x20:
		1584	if (!self->incr_nest)
		1585	{
		1586	--p; // do not eat the whitespace, let the next round do it
		1587	goto interrupt;
		1588	}
		1589	break;
		1590
		1591	case '"':
		1592	self->incr_mode = INCR_M_STR;
		1593	goto incr_m_str;
		1594
		1595	case '[':
		1596	case '{':
		1597	if (++self->incr_nest > self->max_depth)
		1598	croak (ERR_NESTING_EXCEEDED);
		1599	break;
		1600
		1601	case ']':
		1602	case '}':
		1603	if (!--self->incr_nest)
		1604	goto interrupt;
		1605	}
		1606	}
		1607	}
		1608
		1609	modechange:
		1610	;
		1611	}
		1612
		1613	interrupt:
		1614	self->incr_pos = p - SvPVX (self->incr_text);
		1615	//printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1035	}	1616	}
1036		1617
1037	/////////////////////////////////////////////////////////////////////////////	1618	/////////////////////////////////////////////////////////////////////////////
1038	// XS interface functions	1619	// XS interface functions
1039		1620
1040	MODULE = JSON::XS PACKAGE = JSON::XS	1621	MODULE = JSON::XS PACKAGE = JSON::XS
1041		1622
1042	BOOT:	1623	BOOT:
1043	{	1624	{
1044	int i;	1625	int i;
1045
1046	memset (decode_hexdigit, 0xff, 256);
1047		1626
1048	for (i = 0; i < 256; ++i)	1627	for (i = 0; i < 256; ++i)
1049	decode_hexdigit [i] =	1628	decode_hexdigit [i] =
1050	i >= '0' && i <= '9' ? i - '0'	1629	i >= '0' && i <= '9' ? i - '0'
1051	: i >= 'a' && i <= 'f' ? i - 'a' + 10	1630	: i >= 'a' && i <= 'f' ? i - 'a' + 10
1052	: i >= 'A' && i <= 'F' ? i - 'A' + 10	1631	: i >= 'A' && i <= 'F' ? i - 'A' + 10
1053	: -1;	1632	: -1;
1054		1633
1055	json_stash = gv_stashpv ("JSON::XS", 1);	1634	json_stash = gv_stashpv ("JSON::XS" , 1);
		1635	json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1);
		1636
		1637	json_true = get_bool ("JSON::XS::true");
		1638	json_false = get_bool ("JSON::XS::false");
1056	}	1639	}
1057		1640
1058	PROTOTYPES: DISABLE	1641	PROTOTYPES: DISABLE
1059		1642
1060	SV *new (char *dummy)	1643	void CLONE (...)
1061	CODE:	1644	CODE:
1062	RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash);	1645	json_stash = 0;
		1646	json_boolean_stash = 0;
		1647
		1648	void new (char *klass)
		1649	PPCODE:
		1650	{
		1651	SV *pv = NEWSV (0, sizeof (JSON));
		1652	SvPOK_only (pv);
		1653	json_init ((JSON *)SvPVX (pv));
		1654	XPUSHs (sv_2mortal (sv_bless (
		1655	newRV_noinc (pv),
		1656	strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
		1657	)));
		1658	}
		1659
		1660	void ascii (JSON *self, int enable = 1)
		1661	ALIAS:
		1662	ascii = F_ASCII
		1663	latin1 = F_LATIN1
		1664	utf8 = F_UTF8
		1665	indent = F_INDENT
		1666	canonical = F_CANONICAL
		1667	space_before = F_SPACE_BEFORE
		1668	space_after = F_SPACE_AFTER
		1669	pretty = F_PRETTY
		1670	allow_nonref = F_ALLOW_NONREF
		1671	shrink = F_SHRINK
		1672	allow_blessed = F_ALLOW_BLESSED
		1673	convert_blessed = F_CONV_BLESSED
		1674	relaxed = F_RELAXED
		1675	allow_unknown = F_ALLOW_UNKNOWN
		1676	PPCODE:
		1677	{
		1678	if (enable)
		1679	self->flags |= ix;
		1680	else
		1681	self->flags &= ~ix;
		1682
		1683	XPUSHs (ST (0));
		1684	}
		1685
		1686	void get_ascii (JSON *self)
		1687	ALIAS:
		1688	get_ascii = F_ASCII
		1689	get_latin1 = F_LATIN1
		1690	get_utf8 = F_UTF8
		1691	get_indent = F_INDENT
		1692	get_canonical = F_CANONICAL
		1693	get_space_before = F_SPACE_BEFORE
		1694	get_space_after = F_SPACE_AFTER
		1695	get_allow_nonref = F_ALLOW_NONREF
		1696	get_shrink = F_SHRINK
		1697	get_allow_blessed = F_ALLOW_BLESSED
		1698	get_convert_blessed = F_CONV_BLESSED
		1699	get_relaxed = F_RELAXED
		1700	get_allow_unknown = F_ALLOW_UNKNOWN
		1701	PPCODE:
		1702	XPUSHs (boolSV (self->flags & ix));
		1703
		1704	void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
		1705	PPCODE:
		1706	self->max_depth = max_depth;
		1707	XPUSHs (ST (0));
		1708
		1709	U32 get_max_depth (JSON *self)
		1710	CODE:
		1711	RETVAL = self->max_depth;
1063	OUTPUT:	1712	OUTPUT:
1064	RETVAL	1713	RETVAL
1065		1714
1066	SV *ascii (SV *self, int enable = 1)	1715	void max_size (JSON *self, U32 max_size = 0)
1067	ALIAS:	1716	PPCODE:
1068	ascii = F_ASCII	1717	self->max_size = max_size;
1069	utf8 = F_UTF8	1718	XPUSHs (ST (0));
1070	indent = F_INDENT	1719
1071	canonical = F_CANONICAL	1720	int get_max_size (JSON *self)
1072	space_before = F_SPACE_BEFORE
1073	space_after = F_SPACE_AFTER
1074	pretty = F_PRETTY
1075	allow_nonref = F_ALLOW_NONREF
1076	shrink = F_SHRINK
1077	CODE:	1721	CODE:
1078	{	1722	RETVAL = self->max_size;
1079	UV *uv = SvJSON (self);
1080	if (enable)
1081	*uv |= ix;
1082	else
1083	*uv &= ~ix;
1084
1085	RETVAL = newSVsv (self);
1086	}
1087	OUTPUT:	1723	OUTPUT:
1088	RETVAL	1724	RETVAL
1089		1725
1090	SV *max_depth (SV *self, UV max_depth = 0x80000000UL)	1726	void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
		1727	PPCODE:
		1728	{
		1729	SvREFCNT_dec (self->cb_object);
		1730	self->cb_object = SvOK (cb) ? newSVsv (cb) : 0;
		1731
		1732	XPUSHs (ST (0));
		1733	}
		1734
		1735	void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
		1736	PPCODE:
		1737	{
		1738	if (!self->cb_sk_object)
		1739	self->cb_sk_object = newHV ();
		1740
		1741	if (SvOK (cb))
		1742	hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
		1743	else
		1744	{
		1745	hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0);
		1746
		1747	if (!HvKEYS (self->cb_sk_object))
		1748	{
		1749	SvREFCNT_dec (self->cb_sk_object);
		1750	self->cb_sk_object = 0;
		1751	}
		1752	}
		1753
		1754	XPUSHs (ST (0));
		1755	}
		1756
		1757	void encode (JSON *self, SV *scalar)
		1758	PPCODE:
		1759	XPUSHs (encode_json (scalar, self));
		1760
		1761	void decode (JSON *self, SV *jsonstr)
		1762	PPCODE:
		1763	XPUSHs (decode_json (jsonstr, self, 0));
		1764
		1765	void decode_prefix (JSON *self, SV *jsonstr)
		1766	PPCODE:
		1767	{
		1768	STRLEN offset;
		1769	EXTEND (SP, 2);
		1770	PUSHs (decode_json (jsonstr, self, &offset));
		1771	PUSHs (sv_2mortal (newSVuv (offset)));
		1772	}
		1773
		1774	void incr_parse (JSON *self, SV *jsonstr = 0)
		1775	PPCODE:
		1776	{
		1777	if (!self->incr_text)
		1778	self->incr_text = newSVpvn ("", 0);
		1779
		1780	// append data, if any
		1781	if (jsonstr)
		1782	{
		1783	if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text))
		1784	{
		1785	/* utf-8-ness differs, need to upgrade */
		1786	sv_utf8_upgrade (self->incr_text);
		1787
		1788	if (self->incr_pos)
		1789	self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
		1790	- (U8 *)SvPVX (self->incr_text);
		1791	}
		1792
		1793	{
		1794	STRLEN len;
		1795	const char *str = SvPV (jsonstr, len);
		1796	SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1);
		1797	Move (str, SvEND (self->incr_text), len, char);
		1798	SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
		1799	*SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
		1800	}
		1801	}
		1802
		1803	if (GIMME_V != G_VOID)
		1804	do
		1805	{
		1806	STRLEN offset;
		1807
		1808	if (!INCR_DONE (self))
		1809	{
		1810	incr_parse (self);
		1811
		1812	if (self->incr_pos > self->max_size && self->max_size)
		1813	croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
		1814	(unsigned long)self->incr_pos, (unsigned long)self->max_size);
		1815
		1816	if (!INCR_DONE (self))
		1817	break;
		1818	}
		1819
		1820	XPUSHs (decode_json (self->incr_text, self, &offset));
		1821
		1822	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
		1823	self->incr_pos -= offset;
		1824	self->incr_nest = 0;
		1825	self->incr_mode = 0;
		1826	}
		1827	while (GIMME_V == G_ARRAY);
		1828	}
		1829
		1830	SV *incr_text (JSON *self)
		1831	ATTRS: lvalue
1091	CODE:	1832	CODE:
1092	{	1833	{
1093	UV *uv = SvJSON (self);	1834	if (self->incr_pos)
1094	UV log2 = 0;	1835	croak ("incr_text can not be called when the incremental parser already started parsing");
1095		1836
1096	if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;	1837	RETVAL = self->incr_text ? SvREFCNT_inc (self->incr_text) : &PL_sv_undef;
1097
1098	while ((1UL << log2) < max_depth)
1099	++log2;
1100
1101	*uv = *uv & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1102
1103	RETVAL = newSVsv (self);
1104	}	1838	}
1105	OUTPUT:	1839	OUTPUT:
1106	RETVAL	1840	RETVAL
1107		1841
		1842	void incr_skip (JSON *self)
		1843	CODE:
		1844	{
		1845	if (self->incr_pos)
		1846	{
		1847	sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + self->incr_pos);
		1848	self->incr_pos = 0;
		1849	self->incr_nest = 0;
		1850	self->incr_mode = 0;
		1851	}
		1852	}
		1853
		1854	void incr_reset (JSON *self)
		1855	CODE:
		1856	{
		1857	SvREFCNT_dec (self->incr_text);
		1858	self->incr_text = 0;
		1859	self->incr_pos = 0;
		1860	self->incr_nest = 0;
		1861	self->incr_mode = 0;
		1862	}
		1863
		1864	void DESTROY (JSON *self)
		1865	CODE:
		1866	SvREFCNT_dec (self->cb_sk_object);
		1867	SvREFCNT_dec (self->cb_object);
		1868	SvREFCNT_dec (self->incr_text);
		1869
		1870	PROTOTYPES: ENABLE
		1871
1108	void encode (SV *self, SV *scalar)	1872	void encode_json (SV *scalar)
		1873	ALIAS:
		1874	to_json_ = 0
		1875	encode_json = F_UTF8
1109	PPCODE:	1876	PPCODE:
		1877	{
		1878	JSON json;
		1879	json_init (&json);
		1880	json.flags |= ix;
1110	XPUSHs (encode_json (scalar, *SvJSON (self)));	1881	XPUSHs (encode_json (scalar, &json));
		1882	}
1111		1883
1112	void decode (SV *self, SV *jsonstr)	1884	void decode_json (SV *jsonstr)
		1885	ALIAS:
		1886	from_json_ = 0
		1887	decode_json = F_UTF8
1113	PPCODE:	1888	PPCODE:
		1889	{
		1890	JSON json;
		1891	json_init (&json);
		1892	json.flags |= ix;
1114	XPUSHs (decode_json (jsonstr, *SvJSON (self)));	1893	XPUSHs (decode_json (jsonstr, &json, 0));
		1894	}
1115		1895
1116	PROTOTYPES: ENABLE
1117		1896
1118	void to_json (SV *scalar)
1119	ALIAS:
1120	objToJson = 0
1121	PPCODE:
1122	XPUSHs (encode_json (scalar, F_DEFAULT | F_UTF8));
1123
1124	void from_json (SV *jsonstr)
1125	ALIAS:
1126	jsonToObj = 0
1127	PPCODE:
1128	XPUSHs (decode_json (jsonstr, F_DEFAULT | F_UTF8));
1129

Diff Legend

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